<%BANNER%>

Complex Host-Microbe Interactions of the Oral Cavity Revealed by Epithelial Transcriptomics

Permanent Link: http://ufdc.ufl.edu/UFE0019817/00001

Material Information

Title: Complex Host-Microbe Interactions of the Oral Cavity Revealed by Epithelial Transcriptomics
Physical Description: 1 online resource (226 p.)
Language: english
Creator: Mans, Jeffrey J
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: actinomycetemcomitans, aggregatibacter, genome, gingivalis, host, infectome, microarray, ontology, pathogen, periodontal, porphyromonas, transcriptome, virulence
Immunology and Microbiology (IDP) -- Dissertations, Academic -- UF
Genre: Medical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Periodontal diseases result from bacterial infection by several pathogens combined with a destructive host immune response. Host-pathogen interactions are inherently dynamic and complex and the unique environment of the oral cavity further complicates the relationship between hosts and microbes. It is estimated that more than 700 bacterial species can inhabit the oral cavity, and a single individual may support 30 to 80 bacterial species at a given time. These bacteria interact with each other as well as with the host, and the impact of additional factors such as smoking or diabetes mellitus further influences the state of health or disease. Although there are many risk factors associated with periodontal disease, true prognostic indicators are lacking and require development. To achieve this goal, gaining an understanding of the host pathogen interactions in the oral cavity, and determining key events that shift the balance from health to illness are crucial aspects of effectively treating periodontal disease. This study utilized transcriptional profiling to investigate the interactions between human immortalized gingival keratinocytes (HIGKs) and two oral pathogens, Aggregatibacter actinomycetemcomitans and Porphyromonas gingivalis. A model using HIGK cells was developed, and the most-impacted host pathways were further characterized phenotypically. Insights gained were related to tropism of oral bacteria to oral epithelia, a specific epithelial response to different species of bacteria, and the contribution of specific bacterial components to the bacterial-host interactions. The impacts of P. gingivalis YPF1, A. actinomycetemcomitans CDT, ORF859, and Aae upon the host transcriptome were investigated. In collaboration with other researchers, the impact of P. gingivalis SerB was also studied. Also in collaboration, the interactions of epithelial cells to commensal S. gordonii and the opportunistic commensal F. nucleatum have been initiated. This work focuses on the host-pathogen interplay occurring in the oral epithelium. The model system established, and the insights provided herein based on host cell global responses will lead to a greater understanding of host pathogen interactions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jeffrey J Mans.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Handfield, Martin.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2008-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0019817:00001

Permanent Link: http://ufdc.ufl.edu/UFE0019817/00001

Material Information

Title: Complex Host-Microbe Interactions of the Oral Cavity Revealed by Epithelial Transcriptomics
Physical Description: 1 online resource (226 p.)
Language: english
Creator: Mans, Jeffrey J
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: actinomycetemcomitans, aggregatibacter, genome, gingivalis, host, infectome, microarray, ontology, pathogen, periodontal, porphyromonas, transcriptome, virulence
Immunology and Microbiology (IDP) -- Dissertations, Academic -- UF
Genre: Medical Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Periodontal diseases result from bacterial infection by several pathogens combined with a destructive host immune response. Host-pathogen interactions are inherently dynamic and complex and the unique environment of the oral cavity further complicates the relationship between hosts and microbes. It is estimated that more than 700 bacterial species can inhabit the oral cavity, and a single individual may support 30 to 80 bacterial species at a given time. These bacteria interact with each other as well as with the host, and the impact of additional factors such as smoking or diabetes mellitus further influences the state of health or disease. Although there are many risk factors associated with periodontal disease, true prognostic indicators are lacking and require development. To achieve this goal, gaining an understanding of the host pathogen interactions in the oral cavity, and determining key events that shift the balance from health to illness are crucial aspects of effectively treating periodontal disease. This study utilized transcriptional profiling to investigate the interactions between human immortalized gingival keratinocytes (HIGKs) and two oral pathogens, Aggregatibacter actinomycetemcomitans and Porphyromonas gingivalis. A model using HIGK cells was developed, and the most-impacted host pathways were further characterized phenotypically. Insights gained were related to tropism of oral bacteria to oral epithelia, a specific epithelial response to different species of bacteria, and the contribution of specific bacterial components to the bacterial-host interactions. The impacts of P. gingivalis YPF1, A. actinomycetemcomitans CDT, ORF859, and Aae upon the host transcriptome were investigated. In collaboration with other researchers, the impact of P. gingivalis SerB was also studied. Also in collaboration, the interactions of epithelial cells to commensal S. gordonii and the opportunistic commensal F. nucleatum have been initiated. This work focuses on the host-pathogen interplay occurring in the oral epithelium. The model system established, and the insights provided herein based on host cell global responses will lead to a greater understanding of host pathogen interactions.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Jeffrey J Mans.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Handfield, Martin.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2008-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0019817:00001


This item has the following downloads:


Full Text

PAGE 1

1 COMPLEX HOST-MICROBE INTERACTIONS OF THE ORAL CAVITY REVEALED BY EPITHELIAL TRANSCRIPTOMICS By JEFFREY JAY MANS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Jeffrey Jay Mans

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS First and foremost, I wish to thank my wonderf ul wife Lori, who has shared this journey with me and has made numerous sacrifices and cont ributions to help me achieve this goal. I look forward to meeting countless more of Life’s challenges and triumphs with her by my side. I thank my parents, Glenn and Judy, for the upbringing they provided, and the encouragement they provided for my athletic, hobby, and educational interests. I thank my brother Rob for the adventures we shared grow ing up, such as collecti ng crawdads and sharks teeth in our neighborhood creek. I thank my extended family—t he Weavers, Hubers, Irbys, Mans (C___, North), Bramells, and McMillans—as well as my grandparents Rusty and Beulah, and Marilyn for providing a family environment th at valued education and learning, but was also fun, loving, safe, secure, and free from want. I al so thank the Kings and their family for their love and support now and always. The amazing st ability and support of such an environment is an advantage that makes any goal achievable ; the foundation I am blessed to stand upon undoubtedly contributed to the completion of this academic endeavor. Traveling along the educational path that led me to pursu e the degree of Doctor of Philosophy, I was taught by many dedicat ed and wonderful educators at all levels. In particular, several of these teachers provided a learning environment that was stimulating and memorable. I especially thank the following individuals for their positive social and educational impact: Ms. Margaret McMeekin, Ms. Pat Ridaught, Ms. Li ssa McGann, Ms. Virginia Christensen, Mrs. Nancy Smith, Mr. John Wilson, Mrs. Beverly Gris eck, Dr. Marnie Jones, Dr. William Slaughter, and Dr. Kerry Clark. I especially thank my committee chairman, Dr Martin Handfield for the opportunity to study under his guidance. I appreciate the tec hnical, academic, and professional training; for providing “enough rope to hang myself with,” for the encouragement to publish, and for the

PAGE 5

5 opportunities to present papers at conferences—especially in Bris bane, Australia. I wish to thank my committee members, Dr. Richard Lam ont, Dr. Henry Baker, Dr. Ann Progulske-Fox, and Dr. Keith Mintz, for their challenges, input guidance and valuable time invested in my education. I also wish to recognize Joyce Conners for her invaluable work to ensure the aspects of registration, enrollment, and documentation were completed, while simultaneously providing support and encouragement above th e call of duty. Thank you as we ll to the entire Department of Oral Biology, chaired by Dr. Robert Burne, Dr .William McArthur for his administration of the T32 Training Grant funding this work, and all th e lab members with whom I have interacted. Your knowledge and collaboration has aided in the progression of th is project, and your friendships and support have made this a memorable and enjoyable journey. In closing, all the family, friends, and teacher s who I have known are a wonderful asset, and have a special place in my heart. GO GATORS!

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ........10 LIST OF FIGURES................................................................................................................ .......11 LIST OF ABBREVIATIONS........................................................................................................12 ABSTRACT....................................................................................................................... ............17 Chapter 1 INTRODUCTION................................................................................................................. .19 Periodontal Disease Is a Si gnificant Health Concern.............................................................19 The Oral Cavity................................................................................................................ ......19 Multifactorial Disease......................................................................................................... ....23 Risk Factors................................................................................................................... ..23 Damage by Bacteria and Destructive Immune Response................................................24 Aggregatibacter actinomycetemcomitans .......................................................................27 General characteristics.............................................................................................27 Virulence factors......................................................................................................29 Porphyromonas gingivalis ...............................................................................................34 General characteristics.............................................................................................34 Virulence factors......................................................................................................36 Probing the Complex Interplay between Ho st Responses and Bacterial Virulence Factors........................................................................................................................ .........39 Specific Aims.................................................................................................................. ........40 2 MICROARRAY ANALYSIS OF HUMAN EPITHELIAL CELL RESPONSES TO BACTERIAL INTERACTION..............................................................................................42 Introduction................................................................................................................... ..........42 Class Discovery, Class Comparison, and Class Prediction Paradigms..................................43 Microarrays and the Study of Host-Pathogen Interactions.....................................................47 Epithelial Responses to Pathogenic and Commensal Microorganisms..................................51 Bacterial Mutant Analysis via Host Transcriptomics.............................................................55 Epithelial Cells Interacting with Purifi ed Bacterial Products and Components.....................60 The Environmental Contribution to Host-Pathogen Interactions...........................................62 Revised Role of Epithelial Cells in Host-Pat hogen Interactions at the Mucosal Surface......65 Current Limitations of Microarray s and Gene Ontology Annotations...................................68 Future Directions/Perspective.................................................................................................7 0

PAGE 7

7 3 DISTINCTIVE CHARACTERISTICS OF TRANSCRIPTIONAL PROFILES FROM TWO EPITHELIAL CELL LINES UPON INTERACTION WITH Aggretatibacter actinomycetemcomitans ..........................................................................................................76 Introduction................................................................................................................... ..........76 Specific Host and Pathogen Interact ions and the Host Transcriptome...................................77 Materials and Methods.......................................................................................................... .79 Bacteria and Cell Lines...................................................................................................79 Microarray Analysis........................................................................................................80 Ontology Analysis...........................................................................................................81 Results and Discussion......................................................................................................... ..82 4 DISTINCT TRANSCRIPTIONAL PROFILES CHARACTERIZE ORAL EPITHELIUM-MICROBIOTA INTERACTIONS................................................................96 Introduction................................................................................................................... ..........96 Background..................................................................................................................... ........97 Results and Discussion......................................................................................................... 100 General Considerations.................................................................................................100 Association of A. actinomycetemcomitans and P. gingivalis With Epithelial Cells.....101 Gene Expression in Gingival Epithelial Cells Regulated by A. actinomycetemcomitans and P. gingivalis .................................................................102 Ontology Analysis.........................................................................................................103 Apoptosis in Gingival Epith elial Cells Modulated by A. actinomycetemcomitans or P. gingivalis ...............................................................................................................107 Gene Expression in Response to Isogenic Mutants.......................................................108 Conclusions.................................................................................................................... .......110 Methods........................................................................................................................ ........111 Bacterial Strains.............................................................................................................1 11 Eukaryotic Cell Lines....................................................................................................111 Microbial/Host Cell Co-Culture....................................................................................112 RNA Isolation, cRNA Synthesis and Chip Hybridization............................................112 Microarray Data Analysis and Expression Filter..........................................................113 Variation Filter, Normalizat ion, and Cluster Analysis..................................................113 Supervised Learning, Discrimination Analysis, and Cross Validation.........................113 Ontology Analysis.........................................................................................................114 Assessment of HIGK Cell Apoptosis............................................................................114 5 IMPACT OF Aggregatibacter actinomycetemcomitans ADHERENCE ON GINGIVAL EPITHELIAL CELL TRANSCRIPTOME..........................................................................125 Introduction...........................................................................................................................125 Background..................................................................................................................... ......127 Materials and Methods.........................................................................................................1 31 Bacteria and Cell Lines.................................................................................................131 Microarray Analysis......................................................................................................133 Functional Categorization by Gene Ontol ogy (GO) and Bioinformatics Analyses......134

PAGE 8

8 Confocal Fluorescent Microscopy.................................................................................135 Quantification of A. actinomycetemcomitans Adherence.....................................................137 Results........................................................................................................................ ...........137 A. actinomycetemcomitans Viable Counts....................................................................137 Transcriptional Profiling...............................................................................................138 Ontology Analysis.........................................................................................................139 Preliminary Phenotypic Confirmation...........................................................................144 Adhesion.................................................................................................................144 Rearrangement of Actin Cytoskeleton..........................................................................145 Discussion..................................................................................................................... ........146 Conclusions.................................................................................................................... .......152 6 DISCUSSION AND PERSPECTIVES................................................................................166 Additional Collaborative Work............................................................................................167 Lessons Learned................................................................................................................ ...168 Specific Aim 1: Establish the Epithelia l Transcriptome Experimental Model.............168 Choosing the microarray and data analysis platform.............................................168 Determining standard infection conditions............................................................169 Effect of host cell lineage on transcriptional response...........................................170 Specific Aim 2: Establish the Baseli ne Oral Epithelial Transcriptome Upon Infection.....................................................................................................................1 72 P. gingivalis and A. actinomycetemtomitans -specific responses...........................172 Baseline host response to S. gordonii and F. nucleatum ........................................172 Specific Aim 3: Investigate the Impact of Individual Bacterial Components on the Host Cell Transcriptome............................................................................................174 Impacts of P. gingivalis fimbriae and A. actinomycetemcomitans ORF859..........174 A. actinomycetemcomitans CDT and Aae impact on hos t pathogen interactions..174 Advantages and Limitations of the Current Epithelial System............................................176 Advantages....................................................................................................................1 76 Limitations.................................................................................................................... .177 Future Endeavors............................................................................................................... ...178 Revisit the Database Periodically..................................................................................178 Uncharacterized Bacterial IVIAT Genes.......................................................................179 Complex Flora...............................................................................................................179 Time course of Infection and Parall el Host and Pathogen Array Analysis...................180 Improvements to the Transcript ome Reporter System Model......................................181 Summary........................................................................................................................ .......183 7 GENERAL CONCLUSIONS...............................................................................................184 Appendix A GINGIVAL EPITHELIAL CELL TR ANSCRIPTIONAL RESPONSES TO COMMENSAL AND OPPORTUNISTIC ORAL MICROBIAL SPECIES.......................186

PAGE 9

9 B INTERNATIONAL ASSOCIATION FOR DENTAL RESEARCH (IADR) ABSTRACTS RESULTING FROM TRANSCRIPTOMICS PROJECT............................194 C TEMPORAL VARIATION OF THE TRAN SCRIPTOME OF GINGIVAL CELLS INTERACTING WITH Aggregatibacter actinomycetemcomitans .....................................196 LIST OF REFERENCES............................................................................................................. 198 BIOGRAPHICAL SKETCH.......................................................................................................226

PAGE 10

10 LIST OF TABLES Table page 2-1. Transcriptional profiling of human epithelial cells to study host pathogen interactions...73 3-1. Transcriptional regulation of common probe sets to A. actinomycetemcomitans infected HeLa (KB) and HIGK epithelial cells..................................................................89 4-1 Microbial-epithelial cell interaction char acteristics of human primary (GEC) and transformed (HIGK) gingival cells..................................................................................115 4-2. Pathways common to P. gingivalis and A. actinomycetemcomitans -infected HIGK cells. ........................................................................................................................ ....115 4-3. Pathways specific to P. gingivalis -infected HIGK cells..................................................116 4-4. Pathways specific to A. actinomycetemcomitans -infected HIGK cells...........................116 4-5. Gene ontology analysis for HIGK cells infected with A. actinomycetemcomitans mutant strain JMS04........................................................................................................117 4-6. Gene ontology analysis for HIGK cells infected with P. gingivalis mutant strain YPF1. ......................................................................................................................... ...117 5-1. Gene ontology analysis (P<0.05) of the most impacted pathways caused by a mutation of aae (VT1565) in the parent strain A. actinomycetemcomitans (SUNY465)...................................................................................................................... 155

PAGE 11

11 LIST OF FIGURES Figure page 3-1. Divergence of HeLa and HIGK ce ll transcriptional profiles.............................................90 3-2. Different patterns of gene expression by HeLa and HIGK cells upon co-culture with A. actinomycetemcomitans .................................................................................................92 3-3. Processes associated with stress response in HeLa and HIGK cells that are impacted upon A. actinomycetemcomitans interaction......................................................................94 4-2. Differential modulation of the P53-mediated apoptosis pathway by oral bacteria..........120 4-3. Apoptotic responses of HIGK cells to A. actinomycetemcomitans or P. gingivalis by ELISA of cytoplasmic histone-associated DNA fragments.............................................122 5-1. HIGK transcriptome upon VT1565-, SUNY465-, or mock-infection.............................156 5-2. Impact of A. actinomycetemcomitans SUNY465 interaction with HIGK cells upon the Regulation of Actin Cytoskeleton pathway...............................................................158 5-3. Impact of A. actinomycetemcomitans SUNY1565..........................................................160 5-4. Specific contribution of Aae to the HI GK cell Regulation of Actin Cytoskeleton pathway........................................................................................................................ ....162 5-5. HIGK cells interacting with Aggregatibacter actinomycetemcomitans ..........................164

PAGE 12

12 LIST OF ABBREVIATIONS 1KNN 1st nearest neighbor 3KNN 3rd nearest neighbors A absorbance units Aa Aggregatibacter actinomycetemcomitans AFLP amplified (restriction) fragment length polymorphism AIDS Acquired Immunodeficiency Virus ATCC American Type Culture Collection BEC(s) buccal epithelial cell(s) BoP bleeding on probing BRB Biometric Research Branch BSA bovine serum albumin CA California Ca2+ Calcium CAL clinical attachment loss CAM camptothecin C celsius cDNA complementary deoxyribonucleic acid CFU colony forming units CMV cytomegalovirus CO2 carbon dioxide COOH carboxyl CP chronic periodontitis CPS capsular polysaccharide

PAGE 13

13 CTRL control DAG directed acyclic graphs DMEM Dulbecco’s modified eagle’s medium DNA deoxyribonucleic acid dPBS Dulbecco’s phosphate buffered saline ELISA enzyme-linked immunosorbant assay EPEC enteropathogenic E. coli EST expressed sequence tag F-actin filamentous actin G0 quiescence G1 first gap phase G2 second gap phase GCOS GeneChip operating software GEC(s) gingival epithelial cell(s) GI gastrointestinal GO ID gene ontology identity GO gene ontology HACEK Haemophilus spp. (except H. influenzae ), Aggregatibacter actinomycetemcomitans, Cardiobacteri um hominis, Eikenella corrodens and Kingella kingae complex HG human genome h hour(s) HIGK(s) human immortalized gingival keratinocyte(s) HIV human immunodeficiency virus

PAGE 14

14 HPV human papillomavirus IADR International Association for Dental Research IL interleukin iNOS inducible nitric oxide synthase IVIAT in vivo -induced antigen technology k kilo kDa kiloDaltons KEGG Kyoto Encyclopedia of Genes and Genomes KGP lysine specific gingipain KSFM keratinocyte serum free medium LAP localized aggressive periodontitis LOOCV leave one out cross validation LPS lipopolysaccharide mg milligram mL milliliter M mitosis mM millimolar MN Minnesota MOI multiplicity of infection MPSS massively parallel signature sequencing mRNA messenger ribonucleic acid NalR nalidixic acid resistant NFAT nuclear factor of activated T-cells

PAGE 15

15 NH2 amino NIDCR National Institute of Dental and Craniofacial Research nm nanometers NY New York OD optical density ORF open reading frame P probability value due to chance PAI pathogenicity island PAMP(s) pathogen-associated molecular pattern(s) PBS phosphate buffered saline PCR polymerase chain reaction PD pocket depth PEDANT protein extraction, desc ription, and analysis tool pH potential of Hydrogen PHUEC(s) primary human urethral epithelial cell(s) rcf relative centrifugal force RGP arginine specific gingipain RifR rifampicin resistant RNA ribonucleic acid rRNA ribosomal ribonucleic acid RT room temperature RT-PCR reverse transcription polymerase chain reaction RTX repeats in toxin

PAGE 16

16 SAGE serial analysis of gene expression STS staurosporine SUNY State University of New York T4SS type four secretion system TLR TOLL-like receptor TNF tumor necrosis factor TSBYE trypticase soy broth yeast extract UEC(s) urethral epithelial cell(s) UFCD University of Florida College of Dentistry USA United States of America U units VT Vermont YoP Yersinia outer protein g microgram

PAGE 17

17 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COMPLEX HOST-MICROBE INTERACTIONS OF THE ORAL CAVITY REVEALED BY EPITHELIAL TRANSCRIPTOMICS By Jeffrey Jay Mans August, 2007 Chair: Martin Handfield Major: Medical Sciences –Immunology and Microbiology Periodontal diseases result from bacterial inf ection by several pathogens combined with a destructive host immune respons e. Host-pathogen interactio ns are inherently dynamic and complex and the unique environment of the oral cavity further complicates the relationship between hosts and microbes. It is estimated that more than 700 bacterial species can inhabit the oral cavity, and a single individual may support 30 to 80 bacter ial species at a given time. These bacteria interact with each other as well as with the host, and the impact of additional factors such as smoking or diabetes mellitus further influe nces the state of health or disease. Although there are many risk factors associated with peri odontal disease, true pr ognostic indicators are lacking and require development. To achieve this goal, gaining an understanding of the host pathogen interactions in the oral cavity, and determining key events that shift the balance from health to illness are crucia l aspects of effectively tr eating periodontal disease. This study utilized transcripti onal profiling to investigate th e interactions between human immortalized gingival keratinocytes (HIGKs) and two oral pathogens, Aggregatibacter actinomycetemcomitans and Porphyromonas gingivalis A model using HIGK cells was developed, and the most-impacted host pathways were further charac terized phenotypically. Insights gained were related to tr opism of oral bacteria to oral epithelia, a specific epithelial

PAGE 18

18 response to different species of bacteria, and the contribution of specific bacteria l components to the bacterial-host intera ctions. The impacts of P. gingivalis YPF1, A. actinomycetemcomitans CDT, ORF859, and Aae upon the host transcriptome were investigated. In collaboration with other researchers, the impact of P. gingivalis SerB was also studied. Also in collaboration, the interactions of epithelial cells to commensal S. gordonii and the opportunistic commensal F. nucleatum have been initiated. This work focuses on the host-pathogen interplay occurring in the oral epithelium. The model system established, and the insights provi ded herein based on host cell global responses will lead to a greater understanding of host pathogen interactions.

PAGE 19

19 CHAPTER 1 INTRODUCTION Periodontal Disease Is a Si gnificant Health Concern According to the World Health Organization, Oral health means being free of chronic mouth and facial pain, oral and thro at cancer, oral sores, birth def ects such as cleft lip and palate, periodontal disease, tooth decay and tooth loss, a nd other diseases and disorders that affect the mouth and oral cavity (World Health Organization, 2007). Unfortunately, both internationally and in the United States, society is far from atta ining complete oral health among its citizens. Worldwide, the most common oral diseases are de ntal cavities and periodontal disease. In each country, 60-90% of school children worldwide have dent al cavities and 5-20% of middle-aged adults have severe peri odontal disease, with rates varying acr oss geographical regions. Incidence of oral cancer, birth defects such as cleft lip and pala te, and occurrences of bacterial or fungal oral infections among immune-compromised individua ls is also a significant health concern. In many high income countries, oral health accounts for 5-10% of public health expenditures. In 2006, Americans made approximately 500 million visits to dentists, and an estimated $94 billion was spent on dental services (Office of the U.S. Surgeon General, 2007). Additionally, there is a strong association between periodon tal disease in pregnant women and the increased birthrate of pre-term low birth weight infants (Offenbacher et al., 1996; Loesche, 1997; Dasanayake, 1998; Li et al., 2000; Yeo et al., 2005; Alves and Ribeiro, 2006), as well as an association between periodontal disease and cardiovascular di sease (Demmer and Desvarieux, 2006; Ellis et al., 2007; Tonetti et al., 2007). The Oral Cavity The oral cavity, or cavum oris, is divided in to the vestibule and the mouth cavity proper. The outer boundary of the vestibule is the ar ea comprising the cheeks a nd lips, and the inner

PAGE 20

20 boundary is formed by the gums and teeth. The m outh cavity proper consists of the area within the aleveolar arches and teeth, ha rd and soft palate, tongue and mucous membrane to the isthmus faucium just before the pharynx (Lewis and Gra y, 1918). The tissues that support the teeth are known as the periodontium and consist of the cemen tum, alveolar processes of the maxillae and mandible, periodontal ligaments, and gingiva (SweetHaven Publishing Services, 2006). The oral cavity is a unique and complex e nvironment. As such, a high diversity of bacterial species exists in this microecologi cal niche. Cloning and sequencing of 16s rRNA originally isolated from the s ubgingival plaque of both healthy and diseased volunteers revealed the presence of 347 species or phylotypes (Paster et al., 2001). Predictive statistical methods revealed that 68 species of bact eria were not sampled, but woul d be found in the population at large upon 10,000 additional clones being analyz ed—increasing the pot ential total to 415 species. The authors estimated that by includin g other oral surfaces—such as the cheek, tongue, and teeth—the number of bacterial inhabitants in the oral cavity was very likely to be 500 to 600 species (Paster et al., 2001). Identical methods were used to assess the microbial diversity that is found colonizing the tongue dorsa of healthy individuals, and those with halitosis confirming this prediction. In the above study, 630 total bacterial species were id entified from the mouths of volunteers, and estimated the tota l number to surpass 700 (Kazor et al., 2003). An investigation of the bacterial flora found in h ealthy volunteers showed that a given individual is colonized by 30 to 80 of the possible 700 species at any given time (Aas et al., 2005). A study performed prior to the detection of these 700 species estimated that approxi mately 25 species of the oral bacteria are pathogenic, or at least have association w ith periodontal disease (Moore et al., 1982). A more recent study, which included a scr een for recently discovered bacteria, has implicated several new bacteria asso ciated with periodontal disease (Kumar et al., 2003). The

PAGE 21

21 estimated number of oral pathogens is very likely to increase, as new t ools allowing the detection of uncultivable bacteria are applie d to periodontal disease research. The anatomy of the oral cavity has several features that promote microbial growth and survival and simultaneously present a challenge to clinicians who wish to treat microbial infections (Haffajee and Socransky, 1994). The teeth are a non-shedding foundation for bacterial colonization, which allows for the accumulation ove r time of bacterial biofilms in the form of plaque. A recent review summari zed the general features of or al biofilms and their impact on periodontal health and dis ease (ten Cate, 2006). The structure and physiology of oral biofilm s resembles those found in other anatomical sites or in the external environment. Pillars of co-aggregated bacteria have been described as mushroom like, interspersed with bacteria-f ree channels that are probably filled with extracellular polysaccharide (EPS). These channels are a conduit for the flow of nutrients, waste products, and salivary components of the host. According to the ecological plaque hypothesis (Marsh, 1994), the composition of an oral biofilm significantly influences the development of dental caries, periodontal disease or can promote dental health. The sequence of oral plaque biofilm formation generally begins with gram positive early colonizers from the Streptococcus genus. Fusobacterium nucleatum is considered a bridge organism, which joins the plaque biofilm at an intermediate stage and then allows late colonizers to eventually attach (ten Cate, 2006). The members of the resulting bact erial community communicate through quorum sensing, and change their gene expression based on population densities. Both antagonistic and synergistic interactions occur simultaneously, and the outcome is a diverse microenvironment within the biofilm that displays variable redox potential, salivar y flow, nutrient availability, and pH depending on the location. Individual bacteria l species can make the environment more or

PAGE 22

22 less favorable for other species through severa l actions, such as bact eriocin production, acid production, or secretion of EPS. One strategy proposed in the ecological plaque hypothesis is to infl uence the formation of the plaque biofilm to make the conditions favorable to health -related bacteria, and simultaneously prevent biofilm communities favo rable to pathogens (Marsh, 1994). Cariogenic biofilms are promoted by high-sucrose di ets and other factors which lead to mutans Streptococci proliferation. The biofilm eventually becomes acidic, and leads to the predomination of acid tolerant bacteria that produce lactic acid and damage tooth en amel. In the periodontal disease model, the availability of oxygen decreases as the oral biofilm increas es in density. Gram negative anaerobes begin to dominate and thrive, including the late colonizers P. gingivalis and A. actinomycetemcomitans The formation of such a biofilm places the gingival and underlying tissues in danger, and the prolonged assault by bacteria upon these tissu es can occur from a relatively stable “home base” which the plaq ue provides (Haffajee and Socransky, 1994). The difficulty of eradicating bacteria in a bi ofilm is also greater than the sum of treating each strain individually. The homeo static balance of the biofilm of ten induces bacteria to slow their metabolism, rendering many antib iotics that target actively growing bacteria ineffective. The EPS secreted by bacteria often binds pharm aceutical agents before they can target the bacterial cells. As a result, the required concentr ation of antibiotic to kill bacteria in a biofilm may be 1000 fold greater than the dose required to kill an equal number of planktonic bacterial cells. Insufficient dose leads to the developmen t of persisting bacteria, which increases the likelihood of resistant strains evolving. Th e measurement Minimum Biofilm Eradication Concentration (MBEC) has replaced the Minimu m Inhibitory Concentration (MIC) in many

PAGE 23

23 cases, which reflects the challenge of treating bacter ial infections occurring as a biofilm, such as dental plaque (ten Cate, 2006). Multifactorial Disease Destruction of the periodont al ligament and resorption of the alveolar bone leading to tooth loss is the hallmark of periodontal disease. Periodontal disease has recently been described as “a heterogeneous group of pathoses character ized by a predominance of specific infectious agents in the face of inadequa te local host defenses” (Slots, 2000); a characterization which reflects the complexity of periodontal disease. Many risk factors for periodontal disease are well documented, and events that may contribute to periodontal disease progre ssion are characterized. Descriptive trends have been re alized for actual clinical outc omes based on these factors. However, predictable outcomes based on known vi rulence factors are unreliable, and the exact mechanisms of the disease process are unknow n (Slots, 2000). Indi viduals may remain periodontially healthy for prolonged periods in th e presence of these risk factors until an unknown triggering event shifts the bala nce towards periodontal disease. Risk Factors The body of evidence in periodontal disease re search has provided strong support for the existence of several risk factors which increa se an individual’s lik elihood of developing periodontal disease. A recent epidemiological re view of periodontitis ca ses cited several major risk factors, such as cigarette smoking, poorly controlled diabetes mellitus, and infection by specific bacteria at high levels Obesity, osteopenia/osteoporos is, socioeconomic status, and HIV infection are additional factors which tend to coincide with periodontal disease (Borrell and Papapanou, 2005). Genetic predisposition of certai n individuals to periodontal disease (Kinane et al., 2005; Loos et al., 2005; Shapira et al., 2005; James et al., 2007) and the potential contribution of herpesviruses to disease (Slots, 2005; Wu et al., 2006; Teughels et al., 2007) are

PAGE 24

24 additional factors that deserve further investig ation. Therefore, a multitude of factors can influence the development and progression of pe riodontal disease. One goal of periodontal disease research is to determine consistent pr ognostic indicators from th ese factors to improve treatment of this comple x and challenging disease. Damage by Bacteria and Dest ructive Immune Response The risk factor most implicated in peri odontal disease is the infection by specific bacterial pathogens at high levels combined w ith a destructive host im mune response (Sakamoto et al., 2005; Teng, 2006a). Subgingival plaque accumu lation consists of a multitude of oral bacteria allowing pathogenic ba cteria to colonize the periodontiu m. The actions of bacterial virulence factors directly—or i ndirectly through the activation of the i mmune system—cause swelling, inflammation, and pocket formation. Eventu ally, this can lead to detachment of the periodontal ligament and tooth loss through reso rption of supporting alveolar bone. The balance between protective and destructive immune re sponses is a key determinant of disease progression. This balance is strongly infl uenced by the host res ponse to challenge by subgingival bacteria (Teng, 2006b). Recently, two competing models of disease progression, the Random Burst theory (Socransky et al., 1984) and the Linear Pr ogression theory were purportedly reconciled as essentia lly the same phenomenon (Gilthorpe et al., 2003). Linear progression held that destructive peri odontal disease is a con tinuous phenomenon that occurs at a relatively constant pace and only varies between population groups (Loe et al., 1978; Gilthorpe et al., 2003). The Random Burst theory was proposed to explain findings from longitudinal studies which contradi cted the linear progression mode l. Clinical attachment loss (CAL) and loss of alveolar bone were found to occur at rates both faster and slower than the rate predicted under the linear progre ssion hypothesis. Periodic bursts of destruction followed by periods of remission were proposed to explain the discrepancies. To reconcile these two

PAGE 25

25 theories, multilevel modeling examined the patte rns of disease progression while considering several covariates in the context of a hierarchical structure (Gilthorpe et al., 2003). Patterns consistent with both linear and random burst theo ries were uncovered, depending on the variable studied. The frequency and randomness of sampling from diseased or healthy sites was found to influence the interpretation of clinical data towa rds either the linear or random burst pattern of progression. The underlying phenomenon that led to both patterns previously described were cycles of damage and repair of the periodontiu m. These observations were found to represent real changes in the health or disease state of individual sites, and not simply a result of measurement error. Thus, the temporal com ponent of periodontal di sease progression or regression further emphasizes the dy namic nature of this disease. Studies have been undertaken to identify the causative organisms that participate in this cycle of destruction and repair. In one study, the community associatio ns of oral bacteria correlating to health or disease were described. Subgingival plaque samples were obtained from healthy and diseased individuals for analysis. Several statistical models were utilized to calculate relatedness coefficients, which demonstr ate how frequently specific bacterial species are found complexed together in bot h healthy and diseased sites. The result of this study is the formulation of the Red, Orange, Yellow, Gree n, and Violet complexe s of oral bacteria (Socransky et al., 1998). The Red complex, which included Porphyromonas gingivalis Treponema denticola and Tannerella forsythia, strongly correlates to Chronic Periodontal disease. Aggregatibacter actinomycetemcomitans complexed with other bacterial species and was not strongly correlated with CP, consistent with its association with Localized Aggressive Periodontitis. This work implicates several bacterial species for th eir putative roles in periodontal health or disease. Another st udy utilized checkerboard DNA-DNA hybridization to

PAGE 26

26 assess the predominant organisms in subgingi val plaque among patients from different geographical regions. The dist ribution of bacterial species with putative involvement in periodontal disease was investigated in plaque sa mples from patients in the USA, Sweden, Chile, and Brazil. These patients were also assessed cl inically for severity of disease. Significant differences between the bacterial floras isolated from the participants that correlated to their country of origin. Thus, a geographical component also appears to play a role in periodontal disease (Haffajee et al., 2004). A similar finding was made in a study that inves tigated the serum antibody levels to several strains of A. actinomycetemcomitans and P. gingivalis found in patients from the USA and Turkey. The findi ngs demonstrated the potential for antigenic diversity among A. actinomycetemcomitans and P. gingivalis strains from different geographical regions (Celenligil a nd Ebersole, 1998). The 1996 World Workshop in Periodontics Cons ensus Report recognized three species as causative organisms of periodontal disease (Bor rell and Papapanou, 2005). These organisms are: Tannerella forsythia [(Maiden et al., 2003), formerly Tannerella forsythensis (Sakamoto et al., 2002) and Bacteroides forsythus ], Porphorymonas gingivalis and Aggregatibacter [(NorskovLauritsen and Kilian, 2006), formerly Actinobacillus actinomycetemcomitans ]. However, in light of continuing efforts to identify key periodont al pathogens, and th e identification of uncultivable bacteria from the oral cavity, the number of fo rmally recognized periodontal pathogens is likely to increase (Kumar et al., 2003). The following will briefly introduce A. actinomycetemcomitans and P. gingivalis the recognized etiologica l agents of Localized Aggressive Periodontitis (LAP) and Chronic Pe riodontitis, (CP), respec tively, and the model organisms for this study.

PAGE 27

27 Aggregatibacter actinomycetemcomitans General characteristics Aggregatibacter actinomycetemcomitans was originally named Bacterium Actinomycetem comitans by Klinger in 1912, as a consequence of its initial isolation from oral actinomycotic lesions in combination with Actinomyces israelii (Kachlany et al., 2001). A. actinomycetemcomitans has been demonstrated to support the growth of Actinomyces (Hoffler et al., 1980), and complexes with other bacter ia in subgingival plaque (Socransky et al., 1998). Actinomycetemcomitans literally means “accompanying an actinomycete” (Quercia et al., 2006). This bacterium has also b een briefly referred to as Haemophilus actinomycetemcomitans (Potts et al., 1985; Ohta et al., 1986; Henderson et al., 2003). The creation of a new genus to classify A. actinomycetemcomitans and three closely related species and subsequent name change to Aggregatibacter actinomycetemcomitans satisfies categorization i ssues with this bacterium (Norskov-Lauritsen and Kilian, 2006). Th e 2105 kb genome of strain HK1651 has been sequenced and is publicly available (Roe et al., 2003). A. actinomycetemcomitans is a gram negative coccob acillus that is non motile, saccharolytic, and capnophilic. Isolation of A. actinomycetemcomitans from within buccal epithelial cells obtained fr om periodontal disease patients confir ms this bacterium is an invasive organism in vivo (Rudney et al., 2001; Rudney et al., 2005). The ability of this bacterium to invade several cell types in vitro has also been demonstrated to varying extents in several cell lines (Brissette and Five s-Taylor, 1999; Schenkein et al., 2000; Handfield et al., 2005). The mechanism of invasion is variously described as involving both host cell actin and microtubule rearrangement that resembles invasion st rategies of enteric bacteria (Meyer et al., 1996; Meyer et al., 1997; Brissette and Fi ves-Taylor, 1999; Meyer et al., 1999). A. actinomycetemcomitans also belongs to the HACEK group ( Haemophilus spp. (except H. influenzae), Aggregatibacter

PAGE 28

28 actinomycetemcomitans, Cardiobacteri um hominis, Eikenella corrodens and Kingella kingae ) of pathogens responsible for infect ive endocarditis in 2%-5% of culture-positive cases (Paturel et al., 2004; Petti et al., 2006). References associating A. actinomycetemcomitans and bacterial endocarditis can be found as early as 1966 (Overholt, 1966). There currently are 6 recognized serotypes describing A. actinomycetemcomitans isolates based on the O-polysaccharide structure of the lipopolysaccharide, designated a-f (Zambon et al., 1983; Saarela et al., 1992; Gmur et al., 1993; Kaplan et al., 2001; Henderson et al., 2003). Some serotypes are more prevalent in specific geographical areas or among certain ethnic groups (Henderson et al., 2003). Additionally, the majority of i ndividuals tend to be stably infected with A. actinomycetmcomitans from a single serotype rather th an population of mixed clonality. A change in the serotype infecti ng an individual is a rare event, even after year s of treatment directed towards eradication of A. actinomycetemcomitans infection (Saarela et al., 1992; Saarela et al., 1999). Generally, serotype b st rains are most frequently isolat ed from localized aggressive periodontal disease patients, and serotype c stra ins are most often asso ciated with healthy subjects (Asikainen et al., 1991; Henderson et al., 2003; Yang et al., 2005). However, an association of serotype c strains among Japanese chronic and generalized aggressive periodontal patients has recently emerged (Wang et al., 2005a; Thiha et al., 2007). A. actinomycetemcomitans forms rough phenotype colonies with characteristic crossed cigar morphology visible under low power magnifi cation upon initial clin ical isolation and growth on agar. This characteristic is the result of fibril formation, which are composed of Flp-1 pili and aid in the bacterial adhesion to various surfaces, including plastic and saliva coated hydroxyapatite—a condition that approximates the tooth surface. In broth culture, the rough phenotype A. actinomycetemcomitans autoaggregates and forms a f ilm on exposed culture flask

PAGE 29

29 surfaces. Upon subsequent subcultures, a smooth phenotype A. actinomycetemcomitans emerges, which grows as plankton ic bacteria in liquid culture and eventually predominate over rough phenotype A. actinomycetemcomitans On agar, smooth phenotype A. actinomycetemcomitans form colonies that are opaque and lack the crossed cigar pattern. The transition from the rough to smooth phenotype is partially based on spontaneous mutations in the flp operon promoter region and rend ers the clones deficient in fimb riae, although this is not the only mechanism of the rough to smooth conversion (Wang et al., 2005b). Mutations engineered into the tad locus, which encodes a molecular transpor t system dedicated to the production of Flp pili, have been demonstrated to abrogate the formation of fibrils in vitro Conceivably, similar mutations occurring spontaneously could also result in the loss of this non specific tight adherence phenotype (Schreiner et al., 2003; Perez et al., 2006). Virulence factors A. actinomycetemcomitans possesses several adhesins which allow colonization of the oral cavity in addition to Flp pili. Three nonfimbrial adhesins are currently recognized: Aae, Omp100, and EmaA. The autotransporter adhe sin, Aae, confers specific adherence of A. actinomycetemcomitans to epithelial cells of humans and old world primates (Rose et al., 2003; Fine et al., 2005). The 130 kDa Aae proteins are su rface expressed and possess homology to the autotransporter proteins Hap and Hia of Haemophilus influenzae (Rose et al., 2003). Different strains of A. actinomycetemcomitans have one to four copies of aae The number of copies can affect the binding of host defense factors, such as lactoferrin and milk wh ey to these strains. Strains with one or two copies of aae seem to escape interaction with these factors and therefore polymorphisms of aae may affect the pathoge nic potential of some A. actinomycetemcomitans strains (Rose et al., 2003). Omp100, or ApiA, also is important for A. actinomycetemcomitans binding to human epithelial cells (Yue et al., 2007). This protein binds buccal epithelial cells

PAGE 30

30 with lower affinity than Aae, yet demonstrates the same specificity for human epithelial cells described previously (Fine et al., 2005; Yue et al., 2007). The binding ability of an AaeOmp100 double mutant was completely abrogated in a recent report (Yue et al., 2007). Omp100 is a 100 kDa outer membrane protein with a role in host cell invasion in addition to adhesion (Li et al., 2004). The protein also is able to induc e pro-inflammatory cyt okine expression in epithelial cells (Komatsuzawa et al., 2002; Asakawa et al., 2003). EmaA, also known as ApiB is an orthologue of Yersinia enterocolitica YadA, and is important in the binding of A. actinomycetemcomitans to collagen (Mintz, 2004; Ruiz et al., 2006). EmaA is predicted to be 201 kDa, but runs aberrantly by gel electrophoresis. This protein consists of a membrane anchor region, a stalk, and a head domain (Ruiz et al., 2006). ApiC is also belongs to the oligomeric coiled-coil adhesins (Oca) represented by Y. enterocolitica YadA and participates in collagen binding of A. actinomycetemcomitans Finally, the matrix polysaccharaide PGA and glycoside hydrolase Dispersin B (DspB) play a role in the dynamics of A. actinomycetemcomitans biofilm formation and release, respectively (Kaplan et al., 2004; Ramasubbu et al., 2005). A. actinomycetemcomitans also possesses several components which directly act upon various eukaryotic cell types and cause toxicit y. The LtxA leukotoxin kills polymorphonuclear leukocytes and monocytes of humans, old wo rld primates, and great apes (Taichman et al., 1980; Diaz et al., 2006) and is a member of the RTX (repeats in toxin) exotoxin family. The generic RTX toxin operon consists of four genes transcribed in the order rtxCABD RtxA is generally the secreted active toxin, and the RtxB protei n is generally involved in post-translational modifications which are required for the toxin to be active biologically. The C and D products appear to be involved in the tran sport of RtxA. RTX toxins cont ains a characteristic consensus amino acid sequence of GGXGXDX[L/I/V/W/Y/F ]X, where X is any amino acid. This

PAGE 31

31 sequence is tandemly repeated between six and 40 times, depending on the toxin (Lally et al., 1999). RTX toxins generally are believed to affect ta rget cells in two stages The first stage is a non-specific adsorption to the cell surface. The second phase involves membrane insertion of the toxin. According to one model of RTX toxi city, a hydrophobic region of the RtxA toxin is believed to form pores in the target cells, leading to cell death. In high concentrations, cell death is characteristic of necrosis. Low doses, however result in apoptotic cell death. A second model proposes that the toxin incompletely inserts into the target cell membranes, and remains in the outer leaflet. By displacing lipids in the outer leaflet, pressure placed on the monolayer causes the collapse of the membrane (Kachlany et al., 2000). Leukotoxin is unique from othe r RTX toxins, in that the ma jority of LtxA produced in rough phenotype strains remains attached to th e bacterial surface, although small amounts of leukotoxin are secreted by selected rough strains. This implies that direct bacteria-cell contact is necessary before the leukotoxi n killing activity oc curs. In contrast, the smooth phenotype A. actinomycetemcomitans isolates secrete large quantities of leukotoxin during early growth stages, but the amount of leukotoxin secreted decreas es as cultures age. After prolonged growth, leukotoxin is apparently degraded by an unknown mechanism (Kachlany et al., 2000). The expression of leukotoxin is re gulated by oxygen tension, iron concentration (Spitznagel et al., 1995) and pH (Kachlany et al 2000), which suggest s adaptation to the envi ronment of the oral cavity. The specificity of A. actinomycetemcomitans leukotoxin to leukocytes of humans, old world primates and great apes (Taichman et al., 1987), combined with the finding that these target cells uniquely produce Lymphocyte Functi on Associated Antigen-1 (LFA-1), led to the

PAGE 32

32 discovery that LFA-1 is the specific target of leukotoxi n. In 2006, Fong and colleagues demonstrated the requirement of lipid rafts for this proc ess to occur (Fong et al., 2006). Leukotoxin treatment was observed to cause elevated levels of cytosolic Ca2+ and the accumulation of LFA-1 clustering in lipid rafts. Dispersal of lipid rafts was found to prevent the cytotoxic effects of LtxA, which could be restored by reconstituting the lipid rafts. Consistent with the two-stage hypothesis of RTX toxin activ ity, the authors showed that LtxA initially signals targeted cells to release Ca2+, which activated the intracellu lar protease calpain. Calpain cleaves talin, which anchors LFA1 to the cytoskeleton. This cl eavage event allows LFA-1 to cluster on the cell surface in the association with lipid rafts. LtxA then binds LFA-1, and the toxin moves into the lipid microdomain according to the general RTX two-stage mechanism. Recently, CD18 was identified as the specific receptor for LtxA (Dileepan et al., 2007). CD18 is the 2 integrin subunit that combines with CD11a to form the LFA-1 complex. Using chimeric human (LtxA susceptible) and bovine (L txA resistant) LFA-1 co mposed of different CD18 and CD11a fragments, the f unctional receptor for LtxA was mapped to specific integrinepidermal growth factor like domains (I-EGF) 2, 3, and 4 of CD18. One advantage arising from the effects of A. actinomycetemcomitans leukotoxin is to enable th e bacterium to escape local immune defenses and establ ish an infection (Narayanan et al., 2002). The JP2 clone is hypervirulent and is associated with severe forms of LAP. It possess es a 530 bp promoter deletion that confers higher leukotoxi n activity, corroborating the importance of this virulence factor (Brogan et al., 1994; Haubek et al., 2007). Associated with leukotoxin tr ansport, the protein TdeA is implicated as a toxin and antibiotic exporte r (Crosby and Kachlany, 2007). A. actinomycetemcomitans also produces a cytolethal distending toxin, Cdt, which has been implicated in causing G0/G1 or G2/M cell cycle

PAGE 33

33 arrest in some oral cell types l eading to apoptotic cell death (Sugai et al., 1998; Mayer et al., 1999; Kang et al., 2005; Kanno et al., 2005). Cdt holotoxin is formed by CdtA, CdtB, and CdtC subunits. CdtA is able to bind target ce ll membranes though lipid binding activity (Ueno et al., 2006). CdtC works in concert with CdtA to de liver the B subunit into target cells. Once internalized, CdtB enters the nucleus and causes double-strand br eaks through DNAse I-like activity. CDT contributes to pathogenesis by ki lling immune response ce lls, epithelial cells, and fibroblasts (Smith and Bayles, 2006). In epithelial cells, a CagE homologue induces morphological changes consistent with a pro-a poptotic phenotype (Teng and Zhang, 2005). This CagE homologue is located in the bacterial cy toplasm and is associated with a Type Four Secretion System (T4SS). Even A. actinomycetemcomitans LPS has a demonstrated ability to induce apoptosis in the presence of cycloheximide in human macrophages (Suzuki et al., 2004; Rogers et al., 2007). This is a brief listing of the most prominent A. actinomycetemcomitans virulence factors that are characterized for their probable roles as direct players in the causation of tissue destruction or the ability to establish an infection. Othe r gene products of importance include A. actinomycetemcomitans luxS which encodes an autoinducer 2 (AI-2) like molecule. AI-2 plays a role in quorum sensing and can regulate th e gene expression for ir on acquisition in this bacterium (Fong et al., 2001; Fong et al., 2003). Reviewed elsewhere, the roles of phosphorylcholine incorporation in LPS, the hgpA iron transport system, and the role of Type IV PilA pili in bacterial transformation are aspects of the A. actinomycetemcomitans lifestyle that deserve consideration for their roles in the pa thogenic potential of this bacterium (Fine et al., 2006).

PAGE 34

34 Porphyromonas gingivalis General characteristics Porphyromonas gingivalis is an anaerobic, gram negative, black pigmented, asaccharolytic coccobacillus, and is an etiologica l agent of Chronic Periodontitis (CP). This bacterium is also implicated as a possible risk factor for the developmen t of certain types of cardiovascular disease (Brodala et al., 2005; Renvert et al., 2006) and pregnancy complications (Dasanayake et al., 2001; Contreras et al., 2006). P. gingivalis was first recognized as a distinct species from Bacteroides asaccharolyticus in 1980 on the basis of percentage of total guanine and cytosine content, DNA-DNA hybridization assays, differences between oral and non-oral habitation, and the production of phe nylacetic acid, when it was named Bacteroides gingivalis (Kaczmarek and Coykendall, 1980) The current name was assi gned in 1988, when a new genus was proposed for several Bacteroides species and P. gingivalis became recognized as the accepted name for this oral pathogen (Shah, H.N. and Collins, M.D., 1988). The sequenced genome of P. gingivalis strain W83 is approximately 2343 kb (Nelson et al., 2003). In vitro growth of P. gingivalis is routinely performed on blood agar plates supplemented with hemin and vitamin K1 or K3. The colony morphology of clinical isolates has been described as rough, semi-rough, or smooth (Dahlen et al., 2007). Early colonies of P. gingivalis can appear white or tan in color, but afte r several days of growth, the co lonies become strongly blackpigmented. This characteristic results from th e storage of acquired heme on the bacterial cell surface (Shah et al., 1979; Genco, 1995; Lamont and Jenkinson, 1998) and subsequent aggregation of iron (Olczak et al., 2005). P. gingivalis obligately requires iron for growth, and hemin satisfies this dependen ce (Lamont and Jenkinson, 1998). P. gingivalis relies on small peptide molecules for nutrition because of its inability to effectively utilize sugars as an energy source (Shah and Gharbia, 1989; Lamont and Jenki nson, 1998). The strict nutrient requirement

PAGE 35

35 is suggested to place selective pressure on P. gingivalis to maintain several proteases, which hydrolyze proteins into small peptides and am ino acids, but can simultaneously damage host tissues. P. gingivalis has been categorized into six reacti ve serogroups, designated K1-K6 based on the antigenicity of its capsular carbohydrates, (Laine et al., 1996). There is also a K(-) group for non reactive types (Aduse-Opoku et al., 2006). Additionally, the LPS of P. gingivalis has given rise to three serotypes, O1 to O3. Alternat e criteria have also yielded a third classification system, designating A and B serotype P. gingivalis (Dahlen et al., 2007). As is the case with A. actinomycetemcomitans individuals tend to be colonized by a singl e clonal type of P. gingivalis and thus it follows that isolates from periodonta l disease patients are us ually from a single K and O serotype group Bacteria isolated from disease site s display a wide diversity of K and O antigen combinations, and CP does not appe ar to result from any single clone of P. gingivalis exclusively (Sims et al., 2001; Yoshino et al., 2007). Increased prevalence of some K/O serotype combinations among peri odontal disease patients has been reported, which may also be specific to geographical region and ethnicity of the populatio ns sampled (Van Winkelhoff et al., 1999; Sims et al., 2001; Yoshino et al., 2007). P. gingivalis can be isolated from dis eased periodontal tissue (Thiha et al., 2007) and has the ability to invade ging ival epithelial cells (Lamont et al., 1995; Tribble et al., 2006) as well as vascular and cardiac endothelial cells (Deshpande et al., 1998; Rodrigues and Progulske-Fox, 2005). The invasion process is not fully understood, but involves cytoskelet al rearrangement and the formation of membrane invaginations at the location of P. gingivalis adherence, resembling the bacterium-directed phagocytosis observe d during cellular invasi on by enteric pathogens (Sansonetti, 1993; Lamont and Jenkinson, 1998). The newly described haloacid dehalogenase

PAGE 36

36 family phosphatase, SerB, has been demonstrated to play a direct role in the invasion process (Tribble et al., 2006). This bacterium also has the ability to override host cell apoptotic programs (Nakhjiri et al., 2001; Handfield et al., 2005), which presumably maintains a favorable intracellular environment for bacterial survival. Th e diversity of strategies displayed by invasive bacteria, including P. gingivalis is an interesting area of i nvestigation for host-pathogen interaction research. Virulence factors The ability of a pathogen to colonize its host is considered to be a virulence factor. P. gingivalis produces major fimbriae which are composed of FimA fimbrillin subunits and are associated with the ability of this bacterium to adhere to various surfaces. FimA is a 67 kDa protein encoded by the fimA gene (Dickinson et al., 1988). Six types of fimbriae have been described, designated Type I to Type V and Type Ib, based on the variable nucleotide sequence of the FimA gene from different strains of P. gingivalis (Amano et al., 2000; Nakagawa et al., 2002). P. gingivalis with Type II fimbriae are most often isolated from patients with CP, and Type I fimbriae have been associat ed with periodontal health (Amano et al., 2000; Yoshino et al., 2007). Different regions of P gingivalis fimbria have been found to interact with various substrates, such as lactoferrin, fibronectin and erythrocytes, de monstrating the dynamic range of capabilities afforded by this bacterial structure (Amano et al., 1996; Lamont and Jenkinson, 1998). FimA isogenic mutant strains have impaired adherence, as well as a deficiency to invade host cells as compared with the wild type strain (Weinberg et al., 1997). The FimA(-) phenotype allowed for the discovery of shorter fimbrial structures, which have been termed minor fimbriae (Hamada et al., 1996). The minor fimbriae are distinct from major fimbriae and are 67 kDa proteins encoded by the mfa 1 gene (Hamada et al., 1996). The ability of P. gingivalis fim Aand mfa 1singleand double-mutant strains to form biofilms in

PAGE 37

37 vitro was assessed and compared to the wild type strain. The fim Asingle mutant strain was attenuated in the ability to form a mono-species biofilm on saliva-coated plates. By comparison, the mfa 1strains were completely devoid of the ability to form biofilms (Lin et al., 2006). These bacterial strains were also assessed for autoaggr egation in broth culture. Approximately 50% of the wild type bacterial culture autoaggr egated and fell out of suspension. The mfa 1strains did not autoaggregate to any detectab le level. In contrast, the fim Astrain demonstrated an increased level of autoaggretation to 90% of th e total culture. In the wild type P. gingivalis the larger major fimbriae interfere with autoaggregation by hindering close cell-cell contact. When the major fimbriae are not expressed, the minor fimbri ae are able to come into closer contact, explaining the increased autoaggregation observed in the fim Astrain. That deletion of mfa 1 completely abrogates autoaggreg ation is additional evidence that minor fimbriae are responsible for cell-cell interactions in P. gingivalis (Lin et al., 2006). Additionally, minor fimbriae have a demonstrated interaction with streptococcal species (Lamont et al., 2002; Park et al., 2005). From these experiments, it is evident th at minor fimbriae complement the adhesive characteristics afforded by P. gingivalis FimA fimbriae. Impacted processes are biofilm formation, autoaggregation, and microcolony formation of P. gingivalis The ability of this bacterium to withstand shearing forces in the or al cavity and avoid elimination by salivary flow and swallowing is important for colonization. Therefore, both major and minor fimbriae are important P. gingivalis virulence factors. As previously mentioned, th e nutritional requirements of P. gingivalis provide selective pressure to maintain genes involved w ith iron and small peptide acquisition. P. gingivalis produces several proteinases, such as PrtT, pe riodontain, PrtC collagenase, Trp peptidase, and prolyl tripeptidyl peptidase (Lamont and Je nkinson, 2000). A recent review of proteolytic

PAGE 38

38 activity by selected oral pa thogens listed thirteen know n proteinases produced by P. gingivalis (Potempa et al., 2000). In addition to aidi ng this bacterium with nutri tional requirements, these enzymes can play a role in pr ocessing surface components of P. gingivalis or cause the destruction of host tissue (L amont and Jenkinson, 2000). The most infamous proteinases, are the Arg-X (Chen et al., 1992; Okamoto et al., 1995; Pavloff et al., 1995)and Lys-X (Okamoto et al., 1996; Pavloff et al., 1997; Potempa et al., 1997) proteases, or gingipains The term gingipain wa s proposed in 1992 to desc ribe the first Arg-X gingipain characterized, Gingipain -1 (now RgpA), in light of its specificity for arginine substrates and to reflect its source, class of prot einase, and similarity to other well characterized enzymes such as papain, cl ostripain, and calpain (Chen et al., 1992). It is estimated that 85% of the proteolytic activity of P. gingivalis is attributable to gingipains (Potempa et al., 1997; Imamura, 2003). Porphyromonas gingivalis research conducted over th e last 25 years has often focused on these proteinases, and has shed cons iderable light on their remarkable pathogenic potential. Among other functions, the co mbined activities th e gingipains give P. gingivalis the capacity to affect gingival crevicular flui d production by inducing vascular permeability, increase inflammation through activating the blood coagula tion pathway, and prevent blood clotting through degradation of fibrinogen (Imamura, 2003). The binding activity of gingipains to proteins of the extracellular ma trix is also thought to promote in vivo colonization by P. gingivalis and upon proteolysis, to periodontal pock et formation and eventual tooth loss (Potempa et al., 2000). Furthermore, these enzymes prev ent activation of leukocytes through the degradation of macrophage CD 14, a co-receptor involved in r ecognizing and binding bacterial LPS. Additionally, gingipains bind and lyse er ythrocytes for heme acquisition (Imamura, 2003).

PAGE 39

39 Porphyromonas gingivalis possesses several hemagglutinins and hemolysins. These molecules also contribute to iron acquisition through specific binding of erythrocytes and eventual lysis. Most notably, th e HagA-E proteins (Progulske-Fox et al., 1989; Progulske-Fox et al., 1995; Han et al., 1996; Lepine and Progulske-Fox, 1996; Lepine et al., 1996) demonstrate this binding capability, and addi tional hemagglutinins have been identified in the sequenced genome (Nelson et al., 2003). Iron acquisition appears to be associated with quorum sensing systems in P. gingivalis (James et al., 2006). There are at least 51 components of P. gingivalis with a putative role in iron or heme utilization (Olczak et al., 2005), emphasizing the importance of iron to this bacterium. Further studie s dedicated to the iron uptake and acquisition components of P. gingivalis are likely to reveal further insigh ts to the pathogeni c potential of this organism. Probing the Complex Interplay between Host Responses and Bacterial Virulence Factors Clearly, periodontal disease is a complex cha llenge for researchers and clinicians, with a myriad of microecological, microbial, and host factors interacting to determine a healthy or pathological state. The pa thogenic potential alone of Aggregatibacter actinomycetemcomitans and Porphyromonas gingivalis is daunting to understand. Th e interrelated nature of these bacterial systems—gene regulation by iron or ot her environmental cues, stress responses, or production of toxins to overtly damage host ti ssue—is magnified and complicated when these bacteria are forced to adapt to the host res ponse. The situation beco mes inreasingly complex when the enormous numbers of possible inter-bact erial interactions between the oral cavity’s 700 species are also considered. Several methods such as In Vivo -Induced Antigen Technology (IVIAT) (Handfield et al., 2000) have been utilized to overcome the challe nge of understanding host-pathogen interactions in the oral cavity. Many intere sting bacterial products with a putative role in periodontal disease

PAGE 40

40 have been identified using IVIAT for both A. actinomycetemcomitans and P. gingivalis (Song et al., 2002). However, this only considers the infe ction from the bacterial perspective. As lamented at the time this di ssertation project was undertaken, little is known concerning the interplay of the host responses that regulate these pathogens (Kinane et al., 1999). Investigating the host perspective provides a more comple te understanding of disease pathogenesis. The oral epithelium is now appreciated as more than a passive barrier to infection. Oral epithelial cells respond actively to microbial encounters through an timicrobial peptide production and cytokine expres sion in response to microbial stimulation (Dale, 2002). Paradoxically, this antimicrobial host response is not necessarily beneficial. Destructive host immune responses are largely credited for peri odontal disease initiati on and progression (Teng, 2006a; Teng, 2006b). In this situatio n, the host is perhaps just as cu lpable as specific bacteria for disease pathogenesis. Understanding the benefici al and harmful aspects of the host response is therefore equally important as uncovering bacteria l virulence factors. Specific Aims In light of the complex questions at hand, the goal of this study was to gain insights into host pathogen interactions that may occur in pe riodontal disease. This was performed with global host transcriptional profili ng of oral epithelial cells responding to bact erial challenge. As reviewed in Chapter 2, microarrays have been us ed successfully to gain insights into epithelial cell responses to various bacteria in sites other than the oral cav ity. The hypothesis driving this project was that, in contrast to the previous paradigm of a passive role of the epithelium, oral gingival epithelial cells actively and specifically re spond to oral pathogens. Three Specific Aims were completed in the course of this study: SA 1: Establish an experimental model usi ng transcriptiona l profiling to study host pathogen interactions in epithelial cells. The appr opriate cell line, inf ection conditions, and

PAGE 41

41 presence of a specific respons e to bacterial challenge by oral epithelial cells were optimized. SA 2: Establish the epithelial host cell baseli ne transcriptional profile. The uninfected transcriptome was compared to the transcri ptome of epithelial cells encountering wild type pathogens to establish this point of reference. SA 3: Assess the impact of individual bact erial factors upon the host cell transcriptome and confirm predicted phenotypes dire ctly, such as antior proapoptotic host cell responses assessed with DNA fragmentation assays. Mutant analysis co mbined with tr anscriptional profiling of epithelial cells and subsequent phenotypic c onfirmations addressed the impact of specific bacterial components upon the interaction with host epithelial cells. This work investigated the dynami c host-pathogen interplay that occu rs in the oral epithelium. The insights provided herein ar e anticipated to further our un derstanding of pathogenesis by periodontal pathogens.

PAGE 42

42 CHAPTER 21 MICROARRAY ANALYSIS OF HUMAN EPITHELIAL CELL RESPONSES TO BACTERIAL INTERACTION2 Introduction Host-pathogen interactions ar e inherently complex and dynamic. The recent use of human microarrays has been invaluable to monitor the effects of various bacter ial and viral pathogens upon host cell gene expression programs. Th is methodology has allowed the host response transcriptome of several cell lines to be studied on a global scal e. To this point, the great majority of reports have focused on the res ponse of immune cells, in cluding macrophages and dendritic cells. These studies revealed that the immune response to microbial pathogens is tailored to different microbial challenges. Conversely, the pa radigm for epithelial cells has— until recently—held that the epithelium mostly served as a relatively pass ive physical barrier to infection. It is now generally accepted that the epithelial barrie r contributes more actively to signaling events in the immune response. The purpose of the work reported in this chap ter is to address the feasibility of using epithelial cell transcripti onal profiles as a reporter system and to establish the groundwork for completing Specific Aim 1: establish a model system Reviewed here, the strategies utilized by other researchers to gain insight s into host-pathogen interactions serve as a framework for the development of a microarray based model for oral ep ithelial cells. The dive rsity of experimental conditions that can be tested usi ng microarrays is explored. Differe nces in host responses to live versus dead bacteria, bacteria of different strains, mutant vers us wild-type comparisons, purified bacterial components and pharmacol ogical treatments are discussed. The case for ep ithelial cells 1 The following manuscript is reprin ted with permission from Bentham Scie nce Publisher Ltd. To access the definitive version, please refer to: Mans, J.J., Lamont, R.J. and Handfield, M. (2 006) Microarray analysis of human epithelial cell responses to bacterial interaction. Infect Disord Drug Targets 6 : 299-309. 2 Supported by NIH/NIDCR T32 Grant DE07200 (JJM), RO1 DE11111 (RJL) and RO1 DE16715 (MH).

PAGE 43

43 eliciting specific pathogen-direct ed responses that are not simp ly extensions of the immune system is bolstered. The challenges, advantages, and limitations of such a study are explored. Experiments that defined both a common core response, as well as pathogen-specific host re sponses are discussed. This review also summarizes the contributions that tran scriptional profiling an alysis has made to our understanding of bacterial physio -pathogensis of infection. Th is includes a discussion of how host transcriptional responses can be used to infer the func tion of virulence determinants from bacterial pathogens inter acting with epithelial mucosa. In particular, we expand upon the lessons that have been learne d from gastro-intestinal and or al pathogens, as well as from members of the commensal flora. Class Discovery, Class Comparison, and Class Prediction Paradigms Global gene expression data presents a unique opportunity to explore the innermost workings of cells that are not revealed by ot her methods. The snapshot of gene regulation afforded by microarray transcriptional profiling has been used to gain insights into cancer genetics and host pathogen inte ractions, among other fields. Th ere are three primary goals of global transcriptional pr ofiling experiments: (1) Class pred iction, (2) Class comparison, and (3) Class Discovery (Simon et al., 2003). Determining which of these approaches is most applicable to the questions at hand, and utilizing valid analytical methods are critical to conducting successful transcriptional profiling experiments. The situation where th e least is known about experimental samples is the class discovery analysis. Class discovery analys is is an unsupervised method th at delineates the relationships between experimental samples, and often focuses on the differences between these samples. Hierarchical clustering (Eisen et al., 1998) and self organizing maps (Tamayo et al., 1999) are two algorithms that are useful in class discovery. Both of thes e algorithms employ a distance

PAGE 44

44 metric, such as Euclidian distance or Pears on’s correlation coefficien t, to describe the relationships between samples or genes (McShane et al 2002). It is important to note that clusters or classes will always emerge in a cla ss discovery analysis, even from random data. For this reason, one should never use cl asses derived from class discove ry to drive a class prediction or class comparison analysis on the same dataset. This practice will lead to over-fitting the dataset and inaccurate classifiers. However, cl asses discovered with unsupervised methods can be further studied using class prediction methods on subsequent independent datasets (McShane et al 2002). Class comparison analysis is used to unc over genes that are differentially regulated between two or more predefined classes. Cla ss prediction is similar to class comparison, with the added goal of deriving classifi ers that are predictive of the id entities of future samples based on the expression profiles. Just as several methods exist for cl ass discovery purposes, there are several algorithms designed to accomplish the supe rvised analyses of class comparison and class prediction. Diagonal linear di scriminant, nearest neighbor, weighted voting method, support vector machines, and compound covariate predicti on are all examples of methods designed to identify classifiers in the cl ass prediction paradigm (Simon et al., 2003; Radmacher et al., 2002). The specifics of each of these methods are beyond the scope of the current work. Suffice is to say that each approach utilizes a formula th at is designed to account for the variance among experimental samples, the mean expression levels for samples in the same class, and in some cases the weighted ability of a specific classifi er to distinguish between classes. The class prediction paradigm is the approach undertaken in the current study, following the framework set forth by Radmacher, McShane, and Simon in 2002. The general framework of a class prediction study is to (1) determine the approp riateness for class pred iction with a given da taset, (2) select a

PAGE 45

45 predictive method, (3) perform cross-validated class prediction, and (4) determine the significance of the class predicti on by permutation testing. The typical outcome of a microarray experime nt is the differential expression levels of thousands of genes for a handful of samples. When the number of covariates exceeds the number of classes, a situation referred to as the multiple hypothesis tests problem arises (Cheng and Pounds, 2007). A significance level of P<0.05 which is suitable for a small number of variables would reveal 1500 genes by chance alone in a 30,000 gene array experiment. Therefore a more stringent leve l of significance, such as P<0. 001 is often used to reduce the number of false positives identified, also known as a Type I error. Most researchers want to identify genes that are truly pred ictors of a specific condition, a nd desire a target genelist with high probability of being confirmed downstream with multiple methods. As the Type I error rate declines, however, the Type II erro r rate increases. A Type II erro r is erroneously leaving a true positive out of the list of regulated genes. Conve rsely, decreasing the Type II error rate includes more false positives, also known as the False Discovery Rate (FDR). There are five currently recognized methods of addressing the multiple hypothesis tests problem. These paradigms are (1) Control of the FDR, (2) estimation of the FDR, (3) significance threshold criteria, (4) control of fa mily-wise error rate (FWER), and (5) empirical Bayesian approaches. The details of these met hods are beyond the scope of this study, yet the diversity of methods available that address th e multiple hypothesis tests problem demonstrates the importance of obtaining reliable ge nelists for further characterization. One of the pitfalls of a class prediction anal ysis is over-fitting the da ta. As demonstrated by Radmacher and colleagues (2002), it is alwa ys possible to derive classifiers from transcriptional data, even when that data is ra ndomly generated from the same distribution. This

PAGE 46

46 occurs when the same dataset is used to build a nd then test the predicti on model that is designed to identify classifiers. As a result of the re latively few samples compared to the number of covariates, and the power of the algorithms chosen, it is always possible for a prediction model to discriminate between two cl asses based on variance noise in the dataset. To avoid this situation, ideally the datase t is split evenly into a training set and a test set. The prediction model is applied to the training set and the classifiers are then applied to the test set to determine their predictive ability. Since experimental samples are often precious, an alternative to this setup is to use leave-one-out cr oss-validation (LOOCV). Leave one out cross-validation approximates the situation of havi ng a training set and a test set. With LOOCV, one expe rimental sample is left out of the dataset, and the rest of the samples are used to determine classifiers based on the prediction model c hosen. The classifiers are then all tested for their ability to correctly pr edict the identity of the sample that has been left out of the training set, which is basically a one-samp le test set. Each sample is left out in turn while the prediction model is built, and th en cross-validated. The cross-validated misclassification rate is an indica tion of the predictive ability of the classifiers. This method avoids over-fitting the dataset because the test sample was not used to build the prediction model and is thus independent of the training set (Radmacher et al., 2002). LOOCV is not the final stage of a sound statis tical analysis of transcriptional data. An important component of the class prediction paradi gm is to test the significance of the crossvalidated misclassification rate. Permutation te sting is a method to compare the ability of the classifiers to correctly predict the identity of the test sample to that expected by chance alone. With permutation testing, the class labels are randomly permuted and the LOOCV analysis is repeated on the scrambled dataset. The cross-va lidated misclassification rate of the permuted

PAGE 47

47 dataset should be higher than the misclassificatio n rate of the true clas sifiers. Because every possible permutation is impossible based on curre nt computing power, a Monte Carlo method is used to perform a user specified number of permut ations on the dataset. A typical analysis will involve 2000 random permutations of the data set. The proportion of these 2000 permuted LOOCV analyses with cross validated miscla ssification rates equal or lower than the misclassification rates for the clas sifiers tested is determined. When this proportion is less than 0.05 (or 100 per 2000 permutations), the nul l hypothesis is rejected (Radmacher et al., 2002). Statistical analysis of micr oarray experiments is a superior method to filtering candidate genes based on the simple mean fold-change of give n transcript level. Simple mean calculations or visual examination of clusters do not account for the variance in a dataset, and will produce unreliable classifiers (Simon et al., 2003). These methods are acceptable for class discovery projects, but not fo r class prediction. Several software programs are available which apply statistical algorithms to transcriptional data. One such software program is BRB Array Tools. Rather than choosing a single statistical algorithm, BR B Array Tools is designed to implement several different hypothesis testing methods that perform well. Li near discriminant analysis, nearest neighbor, and support vector machines are examples of fo rmulae that attempt to identify genes that are truly differentially regulated. Additionally, BRB Array Tools softare allows implementation of the cross-validation and permuta tion testing aspects of class prediction that are critical to obtaining classifiers with a known level of pr edictive ability. As bioinformatic methods improve, the reliability of microarray experiments will follow suit. Microarrays and the Study of Host-Pathogen Interactions The balance between health and disease is an evolutionary arms race involving complex host-pathogen intera ctions (Bergelson et al., 2001; Woolhouse et al., 2002). In recent years,

PAGE 48

48 transcriptional profiling of host-pathogen interac tions has increasingly been used to understand the dynamics of this race in a concerted effort to fight infectious disease. Analysis of the transcriptome changes that o ccur during the cross talk between host and pathogen can be accomplished through several methods, such as microarra ys, or serial analysis of gene expression (SAGE). Similar techniques have been used to the same endeavor and include amplified restriction fragment length polymorphism (AFLP )-derived technique for RNA fingerprinting (cDNA-AFLP), random EST sequencing and massive ly parallel signature sequencing (MPSS) (Matsumura et al., 2005). Host and microbe microarrays offer several advantages to other methods of transcriptional prof iling, including high throughput, pa rallelism, miniaturization, and automation (Miller et al., 2002; Bryant et al., 2004; Dharmadi and Gonzalez, 2004; Kirmizis and Farnham, 2004). These advantages have been in strumental to analyze the genetic polymorphisms of pathogens that are resistant to antimicrobials, explore the di stribution of genes among isolates from the same species, or to investigate the e volutionary relationships between closely related species (Kato-Maeda et al., 2001). Several versions of targeted virulence factor arrays have been useful to detect the presence of pathogenic microorganisms in environmental samples (Bekal et al., 2003). Transcriptional profiling has been us ed to evaluate the simultaneous and genomewide response of host and pathogen cells to drug treatments or antimicrobials in order to predict the safety and efficacy of new treatments (Yowe et al., 2001). The study of host-pathogen interactions has also benefited from microa rray technology where the host and pathogen gene regulation patterns have been evaluate d. The field of infectomics (Huang et al., 2002; McManus et al., 2002; Jong and Huang, 2005) or infectogenomi cs (Kellam and Weiss, 2006) has emerged from these seminal experiments.

PAGE 49

49 Since global differential gene e xpression was first monitored in Arabidopsis thaliana (Schena et al., 1995), microarray technology has advanced significantly. The pot ential of arrays is demonstrated by the comparison of differential gene expression th at occurs between root tissue and leaf tissue with a microarray consisting of onl y 45 genes. In this initial report, genes that were known to be involved with photosynthesis were over-expresse d in leaf tissue relative to root tissue. In addition, both di rect and downstream effects of di fferential gene regulation were observed. This was performed through the comparison of wild-type Arabidopsis to a transgenic plant known to over-express the transcription factor HAT4 at the protein level. The developmental regulator, HAT4, wa s previously discovered in A. thaliana and characterized as a homeo domain-leucine zipper protein (Schena et al., 1993; Schena and Davis, 1994). The upregulation of HAT4 was observed at the transcript level, and th e subsequent cascade of gene regulation changes caused by HAT4 was noted. This pioneering report also attempted to address many challenges still relevant today, including is sues with the standard ization of replicates, problems with the selection of criteria to determin e what gene changes qualify as significant, and how best to organize and interp ret the overwhelming amount of data that can be obtained with a single experiment. Despite thes e challenges, global transcriptional profiling offers an unprecedented potential to further understand the f undamental basis of infectious diseases. This has been suggested to represent a milestone th at will ultimately lead to the development of innovative diagnostics, prophylactics and ther apeutics (Cummings and Relman, 2000; Yowe et al., 2001; Huang et al., 2002; Bryant et al., 2004; Aldridge et al., 2005; Kellam and Weiss, 2006). Cytomegalovirus (CMV) is a virus transmitted eas ily at mucosal surfaces and is a critical concern for solid organ transplant and haematopoiet ic stem cell transplant patients. This virus

PAGE 50

50 can also cause devastating neurodevelopmental se quelae in newborns (Schleiss, 2005). In 1998, the first transcriptional profiling of human ce lls to probe host-CMV in teractions utilizing microarrays was performed. To i nvestigate the impact of infecti on with this virus on host cells, the expression level of 6600 human transcript s from CMV-infected primary human foreskin fibroblasts was compared to uninfected cells. Th e consistency amongst biological replicates was evaluated and assessed the amount of variation that resulted fr om background differences. With a threshold of variation set at three-fold over background, 258 genes were found to be significant, and were further investigated (Zhu et al., 1998). In 2000, microarrays were first used to profile the host response to a bacterial pathogen, Salmonella typhimurium (Rosenberger et al., 2000). Gastrointestinal S. typhimurium infection is usually non-systemic and self-limiting, but can ca use debilitating enteroco litis lasting between two and three weeks. A milder gastroenteriti s can last between three and five days, with symptoms for both manifestations consisting of nausea, vomiting, fever, abdominal pain, and bloody or mucous diarrhea (Hapfe lmeier and Hardt, 2005). At the time of this microarray investigation, specific bact erial effectors were known to modify host cell responses at the protein level, as exemplified by Type III secretion systems (Hueck, 1998). This microarray study demonstrated that the transcriptome of the host c ould also be impacted by bacterial factors. In addition to the host’s response to live Salmonella the effect of a single bacterial component, the lipopolysaccharide (LPS), was also determin ed. Although LPS was known to have immune effects on host cells, the extent of these effects was not well understood at the transcriptional level. In an attempt to mimic an in vivo infection by using an in vitro model, the authors stimulated the host cells with interferon gamma (IFN) and repeated their initial experiment to investigate the effect of this cytokine on th e macrophage transcripti onal responses. This

PAGE 51

51 reconstructed some of the complex interactions that are thought to occur between host and pathogenic factors. The combina tion of isogenic mutant analysis in pathogens with host arrays was also proposed to investigate additional bacterial effectors whos e influence on the host response may be masked by LPS-induced responses. Following the classification of Jenner (J enner and Young, 2005), four categories of experimental strategy have been used to dissect host-pathogen interactio ns. These include (1) host cells infected with a wild type pathogen, (2) host cells infected with a pathogen containing an isogenic mutation, (3) host cell s exposed to purified bacteria l components (such as LPS or pili), and (4) host cells exposed to a physico-chemical treatment (pH, temperature, antibiotic treatment, or pre-exposure to host-derived comp onents). Many other exhaustive reviews have been presented in recent years that cover the transc riptional response of immune cells to infection (Cummings and Relman, 2 000; Kellam, 2000; Kellam, 2001; Yowe et al., 2001; Dharmadi and Gonzalez, 2004; Aldridge et al., 2005; Jenner and Young, 2005). The following is a summary of the progress made in understanding the nature of the interaction between epithelial cells and commensal or pathogenic bacter ia that co-exist as part of the human microflora. Since no new drugs, vacc ines, or diagnostics have yet emerged from targets found with host arrays, we will close with a speculative view of how reverse vaccinology (Rappuoli, 2001) may be applied to the results of host transcriptional profiling to develop novel diagnostics, therapeutics, or prophylactics. Epithelial Responses to Pathogeni c and Commensal Microorganisms. Initial microarray experiments have consiste ntly compared uninfect ed host cells with cells exposed to one or several pathogens to de termine the baseline epith elial host response to a bacterial challenge. Table 2-1 reveals numerous re ports that have exploite d this strategy. In a great majority of cases, investig ators have focused on host gene re gulation related to a specific

PAGE 52

52 class of genes—usually cytokines and chemoki nes—that are known to play a role during infection in vivo This approach sheds some light on the alarm signal activities of epithelial cells, and has identified several novel f actors involved with recruiting i mmune cells to the site of infection. Custom arrays, or commercially availa ble chemokine arrays have been used in several studies to specifically investigate the host gene regulation of these factor s. For example, the novel involvement of seve ral chemokines with host response to pathogens was determined by array analysis of 277 cDNA elements in the human colorectal epithelial cell line HT-29 interacting with Salmonella dublin (Eckmann et al., 2000). S. dublin naturally infects cattle, but can occasionally be spread to human s where it is highly invasive. Th e direct route of infection is unknown, although it is frequently isolated from bl ood, and can cause metastatic infections in virtually all organs. Presum ably, unpasteurized cheese a nd milk are responsible for S. dublin transmission from cattle to humans (Lester et al., 1995). To investigate th e host response to this highly invasive pathogen, the i nvestigators focused on genes m odulated at time points that demonstrated the highest degree of up-regul ation in HT-29 cells upon interaction with Salmonella Two new targets of NFB regulation were identified by this strategy. In addition, the interaction of Salmonella with AGS gastric adenocarcinoma cells confirmed the role of NFB as a regulator of pro-inflammatory cyt okines. Similarly, th e activation of NFB has significance in SV40 transformed human colon carcinoma cells interacting with the causative agent of whooping cough: Bordetella pertussis (Belcher et al., 2000). Besides NFB activation, eight of the 33 up-re gulated genes in these epithelia l cells were pro-inflammatory cytokines. The chemo-attractant activity of these factor s is consistent with the infiltration of monocytes, neutrophils, and lympho cytes which has been observed in other models of infection (Belcher et al., 2000). Likewise, clues into the mechanis ms behind the clinical observations of

PAGE 53

53 acute inflammation occurring in human infections by Salmonella typhimurium (Zeng et al., 2003), Chlamydia trachomatis (Dessus-Babus et al., 2000; Hess et al., 2001), and Neisseria gonorrhoeae (Binnicker et al., 2003) have been provided by global transcriptional profiling, revealing the specific regulation characteristics of vari ous immuno-stimulatory factors. A second group of host responses that has b een addressed with microarrays is the perturbation of host cell apoptotic mechanisms by invasive bacteria. A study performed using a 12,626-element human array profiled primary human urethral epithelial cells (PHUECs) interacting with Neisseria gonorrhoeae and demonstrated the upre gulation of anti-apoptotic regulators, bfl-1, cox-2, and cIAP-2 (Binnicker et al., 2003). The upregulation of antiapoptotic regulators was also shown to depend on the cell type, as RT-PCR analysis of N. gonorrhoeae infected Chang/HeLa epitheli al cells (American Type Cultu re Collection (ATCC); CCL-20.2) did not show the same up regulation of these anti-apoptotic genes. It is interesting to speculate if this dichotomy of host cell respons e also occurs with regard to pro-inflammatory cytokines, and if the different clinical manifestations in men and women are attributable to this host-cell based difference (Binnicker et al., 2003). Further, N. gonorrhoeae -infected urethral epithelial cells were protected from staurosporine (STS)-induced apoptosis. This pathogen’s strategy of immune system avoidance and establishment of an intracellular environment conducive to replication has been demonstr ated in other bacteria. In either epithelial cells or immune cells, Chlamydia trachomatis (Fan et al., 1998), Chlamydia pneumoniae (Geng et al., 2000; Fischer et al., 2001), Helicobacter pylori (Shirin et al., 2000), and Porphyromonas gingivalis (Hiroi et al., 1998; Nakhjiri et al., 2001; Handfield et al., 2005a) have demonstrated anti-apoptotic activity that complements an intracellular lifestyle. Transcriptional profiling has revealed how apoptotic pathways are perturbed by P. gingivalis in

PAGE 54

54 Human Immortalized Gingival Epithelial cells (HIGK). P. gingivalis which establishes a chronic periodontal infection (L oesche and Grossman, 2001), modul ates several factors mapping to the P53 branch of the apoptotic pathway. Similar to the effect of N. gonorrhoeae on urethral epithelial cells, and consistent with earlier work (Nakhjiri et al., 2001), phenotypic confirmation showed that P. gingivalis infection was protective against camptothecin-induced apoptosis in these oral epithelial cells (Handfield et al., 2005a). This silencing of the apoptotic pathways in epithelial cells requires living bacteria, as RT-P CR analysis of primary gingival epithelial cells exposed to heat-killed P. gingivalis induced apoptosis, in contrast to the inhibition observed with live bacteria (Brozovic et al., 2006). The transcriptional profil es of cells encountering killed P. gingivalis are consistent with the transcriptome of cells infected by other pro-inflammatory bacteria; killed P. gingivalis or P. gingivalis extracts induce NFB and up-regulate the innate immune response. Together, these examples furthe r illustrate that the sp ecificity of bacteriaepithelium interactions can be dissected with transcriptional profiling of host cells. The strategy highlighted above is not restricted to the study of the modulation of the innate immune defenses upon bacterial challe nge. Although the dissection of apoptosis and immuno-regulatory pathways has been a significant focal point, many additional cellular processes are involved during epithelium-microbe in teractions. Numerous transcription factors and cell-cell communication fact ors, for example, have been consistently found to be differentially regulated in infected host cells compared to uninfected controls (Handfield et al., 2005a; Jenner and Young, 2005). Studies of the in teraction of HIGK cells with two commensal bacterial species of the oral cavity, Fusobacterium nucleatum ATCC 25586 and Streptococcus gordonii DL-1 Challis, elicit transcrip tional profiles that are highly similar to each other, and far less disruptive to the baseline HIGK cell gene regulation when when compared to pathogen-

PAGE 55

55 infected and uninfected cells (Hasegawa et al., 2007, Appendix A). The significance of this profile is currently under investigation. Clear ly, the complexity involved in predicting and confirming a testable phenotype currently hampers the downstream functional characterization of many of these genes. The ultimate validation of other factors modulated upon infection and the testing of their biologica l relevance in relation to disease in itiation and progression compared to health remains a significant and widespread challenge. Bacterial Mutant Analysis via Host Transcriptomics. Combining a classical bacterial isogenic mu tant analysis with host profiling has been performed with a number of human pathogens, including Pseudomonas aeruginosa (Ichikawa et al., 2000), Helicobacter pylori (Chiou et al., 2001; Cox et al., 2001; Maeda et al., 2001a; Maeda et al., 2001b; Bach et al., 2002; Guillemin et al., 2002), Yersinia enterocolitica (Bohn et al., 2004), A. actinomycetemcomitans and P. gingivalis (Handfield et al., 2005a). The first report describing th e epithelial host cell response to Pseudomonas aeruginosa interaction was performed using microarr ay profiling of A549 human lung carcinoma pneumocytes (Ichikawa et al., 2000). P. aeruginosa is a Gram-negative opportunistic pathogen associated, amongst other types of infection, with deteri orating lung function in cystic fibrosis patients. A microarray consisting of 1506 pneum ocyte genes first esta blished a baseline response to P. aeruginosa interaction at 0h and 3h post infect ion. The contribution of Type IV Pili-dependent bacterial adhesion to the host re sponse was investigated using a non-piliated mutant derivative of strain PAK in parallel with th e parent strain to infect epithelial cells. The baseline response to infection revealed a differen tial regulation of 24 genes. The regulation of 16 genes was directly attributed to Type IV pili, including the up-regulation of Interferon Regulatory Factor-1 (IRF-1). RT-PCR further ch aracterized the factors contributing to IRF-1 activation. It was determined that pili play a critical role in this host response. Subsequent

PAGE 56

56 downstream effects are also associated with a dhesion and included the targeting of inducible nitric oxide synthase and certain inte rferon-induced immune effects (Ichikawa et al., 2000). This finding determined that the cellular adhesion th rough these pili was directly responsible for triggering the host response, in addition to dete rmining a physical association between the host and microbe. Gastric epithelial cell response to Helicobacter pylori is frequently studied with microarray global profiling. This microorganism is the causative agent of active chronic gastritis and is associated with gastri c cancer, in addition to compli cations resulting from pathogeninduced inflammatory damage. In particular the Cag pathogenicity island (CagPAI) of H. pylori has been extensively investigated for its contribution to the NFB and IL-8 induction upon infection (Cox et al., 2001), as well as for the inhibition of apoptosis and cell cycle interference that may lead to gastric cancer. For example, two parallel time-course studies were performed comparing gastric adenocarcinoma (AGS) epithelia l cells interacting with wild type, mockinfected, and five mutant strains of H. pylori. These mutant strains lack ed either the Vacuolating cytotoxin (VacA), or one of severa l structural (CagE) or effector (CagA) proteins from a type IV secretion system encoded by the CagPAI. A microarray of 22,571 elements was used and detected the differential regulation of hundreds of genes under several infection conditions. This ultimately distinguished between CagA, CagE, a nd CagPAI-dependent host responses. In particular, a CagA mutant strain of H. pylori was found to induce fewer cytoskeletal genes than the wild type strain, thus suppor ting the hypothesis that CagA is th e bacterial component directly involved in triggering this host re sponse. Strikingly, the deletion of the entire PAI results in a host response that closely resembles the mock in fection profile, suggestin g the PAI is the major contributor to the AGS cell response to Helicobacter pylori (Guillemin et al., 2002).

PAGE 57

57 Similar strategies have b een utilized to study the cont ribution of specific putative virulence determinant of Y. enterocolitica, A. actinomycetemcomitans, and P. gingivalis For Y. enterocolitica an enteropathogen that causes self-lim iting infections (Pujol and Bliska, 2005), mutant strains of the Yersinia virulence plasmid pYV ( inv +/pYV-), the Yersinia outer protein YopP (pYV+/ yop P-), and the double mutant for Yersinia virulence plasmid and invasin protein ( inv -/pYV-) were used to infect He La cervical epithelial cells (Bohn et al., 2004). The experimental design used uninfected HeLa cells as the baseline gene expression level, and investigated four mutant stra ins over three time points (1, 3 a nd 7 hours). A differential host response was determined for each strain us ing the Affymetrix Hu man Genome HG-U133A GeneChip, which consists of 22,283 probesets The differential responses support the contribution of each virulence factor and establis hed the temporal expression of host factors in response to Yersinia challenge.. After data filtering, two-fold changes in gene expression revealed that 193 probesets are differentially re gulated between experimental conditions. The functional categories impacted included tran scription factors, receptors and signaling components, growth factors, chemokines and cyt okines, and cytoskeletal reorganization factors (Bohn et al., 2004). Interestingly, IL-8 was highly indu ced in HeLa cells upon interaction with the mutants lacking YopP, yet still containing the invasin protein (I nv+/pYV+/YopP-) compared to all other conditions. This finding is consiste nt with other studies and supported the concept that Inv triggers IL -8 production (Schulte et al., 1996), while YopP ma y counteract this induction (Denecker et al., 2002). A comparison of the transc ription profile upon infection with Yersinia lacking Inv to Yersinia lacking both Inv and the vi rulence plasmid identified 117 probesets. Moreover, 24 genes we re coordinately induced in HeLa cells compared to uninfected

PAGE 58

58 cells under these conditions. This suggest ed that other chromosomally encoded Yersinia factors were responsible for the m odulation of these 24 genes. The same strategy has been used to study th e contribution of speci fic putative virulence determinants and extended to make inferences on the role of factor s with unknown function in A. actinomycetemcomitans, and P. gingivalis These oral pathogens are both implicated in periodontal infections (Loesc he and Grossman, 2001; Sanz et al., 2004). In P. gingivalis the major fimbrial protein FimA is important for bacterial adhesion and necessary for effective invasion of host cells (Hamada et al., 1994; Malek et al., 1994; Njoroge et al., 1997). An isogenic mutant strain for FimA (YPF-1) was used to profile Human Immortalized Gingival Keratinocyte (HIGK) (Oda et al., 1996) cells upon a two-hour co -culture using Affymetrix HGU133A human GeneChips (Handfield et al., 2005a). The patterns of gene regulation between HIGK cells encountering w ild type and mutant P. gingivalis revealed many genes involved with host cytoskeleton and membrane receptor activity, consistent with the observed adhesion and invasion deficiencies of this mutant. This conf irmed the critical role of FimA in the adhesion and subsequent invasion even ts of epithelial cells by P. gingivalis Further investigations are underway to determine if the observed effect s are FimA dependent, or adhesion dependent, which can be accomplished by treating HIGK cells w ith purified FimA protein and observing the host response. The regulation of the individual genes involved in these processes can shed light on the details of the bacterial-epithelial cell intera ctions that result in th e observed phenotypes of the host and pathogen. Previously, an in vivo induced antigen technology (IVIAT) screen of human patients with localized aggressive peri odontitis (LAP) (Handfield et al., 2000; Cao et al., 2004a) revealed the in vivo induction of A. actinomycetemcomitans orf859 (PEDANT database). The bioinformatic

PAGE 59

59 analysis of this gene produc t did not predict any significant homologies in the databases, although the gene product was high ly conserved across genera. The orf859 gene product of was further shown to be induced in plaque from infected patients (Handfield et al., 2000) and in various human cell lines in vitro (Richardson et al., 2005). Thus orf859 is an attractive potential marker for active disease in LAP patients (Cao et al., 2004b). Although expresse d specifically in human LAP in vivo the precise function of this bact erial gene was uncertain based on bioinformatics alone (Cao et al., 2004b). In an attempt to elucidate the role of ORF 859 in host-pathogen inte ractions, microarray profiling and ontology analysis were perf ormed for HIGK cells interacting with A. actinomycetemcomitans wild-type strain VT1169, and the is ogenic mutant strain JMS04 ( orf859-). Specific genes were differentially regulate d between HIGK cells infected by wild type and mutant strains of A. actinomycetemcomitans For instance, HIGK infected with JMS04 failed to induce IL-27, which is closely related to IL -12 and has been shown to promote Th1 cellmediated immune responses (Hunter et al., 2004). Furthermore, genes involved with intermediate metabolism, signal transduction, and cytokine activity were differentially expressed in HIGK cells compared to host ce lls interacting with wild type A. actinomycetemcomitans This suggests that the function of ORF859 is likel y to be related to intracellular survival of A. actinomycetemcomitans and probably contributes to imm une recognition by epithelial cells (Handfield et al., 2005a). The ontology analysis furthe r revealed that the most significant variations are associated with intermediate metabolism functions, stress response, and signal transduction (Handfield et al., 2005a). This is consistent with the phenotype of the mutant strain as characterized by total inter action, invasion, and competition assa ys with the wild-type strain (Cao et al., 2004b). The product of this gene is uni nvolved with adhesion, but conferred a

PAGE 60

60 significant persistence defect within epithelial cells in the isogenic mutant strain. Collectively, this gene product is specifically induced intrace llularly during infection a nd confers a selectable advantage to the microorganism in the intracellular environment. T hus, the role of a protein with unknown function has been investig ated through the use of host transcriptional profiling and mutant analysis, in addition to investig ating the impact of known bacterial genes. Epithelial Cells Interacting with Purifi ed Bacterial Products and Components. In certain instances, directly investigati ng the effects of purifie d virulence factors on target cells has been an option (Yokoyama et al., 2005). This strategy has been used to confirm a putative cause-effect function revealed by isog enic mutant analysis. As described in the previous section, H. pylori up-regulates several virulence factors upon interaction with gastrointestinal epithelial host ce lls. In particular, CagA and VacA, have substantial phenotypic effects on these cells. The global transcriptional response of AG S human gastric epithelial cells has been investigated upon direct e xposure to CagA (Yokoyama et al., 2005). CagA was transfected into AGS cells with an expression vector contai ning either hemagglutinin-tagged CagA, or a phosphorylation-deficient CagA. CagA is normally injected into host cells through a type IV secretion system (Censini et al., 1996; Akopyants et al., 1998), which justifie d the use of the transfection method. One of the advantages of this system was the re duction of the possible noise introduced by live bacteria-c ell contact. Four time points were chosen to investigate both direct and downstream events attributable to Ca gA. Overall, 8500 genes we re analyzed using the Affymetrix Genome Focus Array. This reveal ed 339 up-regulated genes and 145 down-regulated genes at a late time point (24 hour s). In contrast, at an earlie r time point (12H), 71 genes were up regulated in gastric epithelia l cells relative to AGS cells not exposed to CagA. The later inductions (24 hours) were consistent with the downstream effects of a cascade initiated at 12

PAGE 61

61 hours post-infection. Of particular interest, th e putative –1000 to –1 promoter regions of the activated genes were analyzed to identify cis elem ents from a panel of tr anscription factors. Binding sites for the nuclear factor of activated T cells (NFAT) family of transcription factors were over-represented among the 71 genes as compar ed to a pool of genes selected at random. Further investigation of these NFAT transcriptional factors determined that Helicobacter CagA initiates a cascade by directly stimulating PLC to activate calcineurin. Calcineurin is a calcium/calmodulin-dependent phosphatase, which acts upon a conserved N-terminal domain of NFAT family transc ription factors (Rao et al., 1997). Activated calc ineurin dephosphorylates cytoplasmic NFAT, which is then translocat ed to the host cell nucleus and activates transcription. When secreted VacA, was allowed to interact with the experimental system described above, it prevented NFATc3 dephosphorylation by calci neurin, inhibiting its translocation to the nucleus and antagonizing CagA. This balance of cellular state linked to specific bacterial virulence factors supports th e clinically observed divers e degrees of severity of H. pylori infection. This is also consistent with the clonality of Helicobacter strains and their variable degrees of CagA and VacA expression. Human intestinal epithelial cell line HT29 and murine RAW 264.7 macrophages have also been subjected to the Aeromonas hydrophila cytotoxic enterotoxin, Act. This bacterium is a water-borne pathogen of fish, am phibians, reptiles, and other co ld-blooded animals. It is generally considered an opportuni stic pathogen of humans, although it has been reported as a primary pathogen of healthy i ndividuals in rare cases (Kao et al., 2003). To study possible effects of A. hydrophila upon human cells, a sub-lethal dos e of purified Act (12ng/mL) was applied to polarized and non-polarized human epith elial HT29 cells at various timepoints (0, 2,

PAGE 62

62 and 12 hours). Affymetrix HU133 array GeneChips were used to profile host cells, and four different software programs were used to impl ement statistical algorithms designed to identify significantly regulated genes (GCOS, SAM, Spotfire 7.3, and ANOVA). All four methods consistently identified 34 gene s that were induced between the 0 and 2-hour timepoints, which were investigated further using ELISA, cytokine profiles, western blot analysis, and phosphorprotein screens. Host responses related to apoptosis, cellula r reorganization, cell growth/cycle/differentiation, cy tokines, signal transduction, a nd transcription factors were identified. Calcium mobilization in human intestinal epithelial cells was found to be important, and assays confirmed this process was essent ial for the maximal production of IL-8 in Acttreated cells. In contrast, a different set of host response genes was up regulated in the murine RAW 264.7 macrophages upon interaction with the same toxin. This was consistent with the specialized roles that diffe rent cell types play in th e encounter with microbes. The Environmental Contribution to Host-Pathogen Interactions. Multiple artificial treatments can be us ed to subject a host-pathogen system to perturbations that are thought to occur during a normal infec tion process. These treatments encompass perturbations such as the addition of purified host prot eins, or the stimulation by IFN, for instance (Rosenberger et al., 2000). Another example of such treatment includes the treatment with pharmaceutical agents. In this case, the therapeutic efficacy can be assessed with the host transcriptome to confirm that the intera ctions between bacteria and host cell are being affected according to the drug’s design. Simila rly, the knockdown of host factors can also be achieved to assess their contribution to th e host-pathogen interac tion. The following are examples of studies that have ut ilized this environmental impact approach to dissect bacterialepithelium interactions.

PAGE 63

63 The experimental strategy mentioned earlier for the HT-29 human intestinal cell line and Salmonella dublin was modified through perturbation of the baseline system (Eckmann et al., 2000). This was performed in order to dissect the specific contributi on of nuclear factorB (NFB) to the induction of several genes previo usly revealed through th e use of Gene Filter GF211 (Research Genetics Inc, Huntsville, AL) and Atlas Human Cytokine/Receptor cDNA arrays (CLONTECH, Palo Alto, CA). NFB is a critical transcri ption factor that helps coordinate the innate and adaptive immune response. Activation of NFB occurs through the phosphorylation of its in hibitors, such as I B family members (Zingare lli, 2005). The authors hypothesized that NFB is also involved in the transcriptional regulation of the cytokine induction revealed in their microarray studies An adenovirus-based system was used to transfect cells with a construct th at expressed a mutant form of I B This defective I B could not be phosphorylated at two positions and thus acted as a super-repressor of NFB activation. Infections by S. dublin of the super-repressor HT-29 cells were performed under conditions identical to the original array experiment, and the regulation of seven genes were characterized with RT-PCR. Although not entirely abolished, th ese previously induced genes were noticeably repressed. These results sugge st a direct role of NFB in the observed HT-29 transcriptional response to Salmonella infection. Additionally, this approa ch led to the discovery of the NFB regulation of two host genes, EBI3 and BGT-1, as well at the confirmation of several other genes being targets of NFB (Eckmann et al., 2000). Another study investiga ting the role of NFB-dependent host responses to infection utilized chemical inhibitors to alter the infection system of Bordetella pertussis with BEAS-2B bronchial epithelial cells. The baseline micr oarray experiment utilizing Affymetrix HU6800 GeneChip arrays confirmed a domin ant pro-inflammatory phenotype upon B. pertussis infection,

PAGE 64

64 consistent with the clinical manifestations and in vivo models of whooping cough (Belcher et al., 2000). The effects of NFB repression were characterized by an anti-inflammatory treatment of host cells with Dexamethasone and sodium salicylate prior to infection with B. pertussis (Belcher et al., 2000). Dexamethasone is a glucocorticoid that has been suggested for therapeutic treatment of pertussis and seemed to augment I B synthesis. Sodium salicylate inhibits NFB activation by preventing I B phosphorylation. Interestingly, pre-treatment of BEAS-2B cells with these compounds prior to B. pertussis infection eliminated the induction of GRO-2, GRO-1, IL-8, IL-6, IL-1 MCP-1, MIP-1 GRO-3, CIITA, HLA-DQ, API-2 /IAP and many other genes that previously were induced in the original mi croarray study. In addition, dexamethasone failed to block the induction of TNFAIP3 while sodium salicylate blocked i nduction. Dexamethasone also up regulated (4.8-fold ) inducible nitric oxide synthase (iNOS), which was not observed with untreated or sodium salicylate-pr etreated-infected epithelial cells. In light of this latter finding, the authors urged caution regard ing the use of dexamethasone as a possible treatment for whooping cough. Potentially, the clinical conditi on could be exacerbated by the induction of iNOS, as collateral damage of neighboring, uninf ected cells could occur through iNOS activity (Belcher et al., 2000). Similarly, dissection of pro-inflammatory gene activation wa s undertaken in T84 intestinal epithelia l cells by exposure to purifie d Tumor Necrosis Factor(TNF). Parallel infection of T84 cells with various S. typhimurium isolates, Escherichia coli Salmonella typhi and treatment with TNFwas performed and the host respons e at 4 hours was assessed with a custom array of >600 genes. Hierar chical cluster anal ysis grouped the S. typhimurium -infected and TNFtreated cells together, and distinctly from the E. coli and S. typhi infected cells. This indirectly implicated TNFin the transcriptional profiles observed under S. typhimurium -

PAGE 65

65 infected conditions, and led to further investigat ion of possible mechanisms that could lead to direct TNFinduction by S. typhimurium (Zeng et al., 2003). Revised Role of Epithelial Cells in Host-P athogen Interactions at the Mucosal Surface. The epithelium is a critical barrier to preven t pathogen entry to the human host. As such, this first line of defense to systemic infecti on has been an obstacle to pathogens that provoked evolutionary changes for bacterial entr y and access to underlying tissue (Eckmann et al., 2000; Aldridge et al., 2005). This process of pathogen entry a nd tissue invasion is an active process that involves signaling pathwa ys from the pathogen and host, which differ between microorganisms (Finlay and Cossart, 1997). The role of epithelial cells has been proposed as an early warning system or se nsor for infection (Eckmann et al., 2000; Aldridge et al., 2005). Many studies have determined that the gene regu lation of host cells, in response to interaction with pathogens, is to specifically regulate pr o-inflammatory and chemoattractant signals which subsequently recruit immune eff ectors to the area of infection. As a result, the focus of much work has been on identifying chemokines and cy tokines produced by epithelial cells. To this point, most microarray-based studies of host-pat hogen interactions have focused on the primary role of the immune system in the host-pathoge n encounter using dendritic cells, macrophages, and other immune cells. However, epithelial cells remain the firs t cell type that most mucosal pathogens and commensal organisms encounter and thus may contribute to host defense in ways not previously recognized. Recently, the role of the epithe lial layer as a sensor was emphasized with the discovery that host epithelia preemptively search for pathogenic bacteria th rough the monitoring for pathogen-associated molecula r patterns (PAMPs) (Ayabe et al., 2004). These microbe-host interactions encompass both pat hogenic and commensal organisms in scope, as host epithelial

PAGE 66

66 cells use TOLL-like receptors to screen for PA MPs of lipopolysaccharide (LPS), flagellins, and bacterial DNA. A detected commensal will not trigger the same host defense response that occurs when a pathogen is encountered (Akira and Hemmi, 2003). For example, the lung epithelium is markedly intolerant to bacter ial colonization, whereas the gastrointestinal epithelium harbors many commensal species (Aldridge et al., 2005). This situation requires the gastrointestinal host cells to de-sensitize the response to LPS, which is accomplished though the down-regulation of the LPS-sensi ng TOLL-like receptor (TLR)-4 (Abreu et al., 2001). Additionally, TLR-5 mediated fl agellin recognition is also do wn-regulated on the apical cell surface where commensal bacteria are known to colonize. However, TLR-5 remains active on the basolateral side where undesi rable tissue invasion past the ep ithelial barrier would still be detected (Gewirtz et al., 2001). Alternatively, in duction of the mucosal inflammatory response through NFB activation can be accomplished by the cy tosolic recognition factor Nod1. This alternative to TLR-based recogni tion of infection by th e host is thus tolerant of commensal inhabitants, while able to trigger the i mmune response to invasive bacteria (Kim et al., 2004). These consistent findings seem to point to a default response of epithelial cells—almost always involving NFB and IL-8—to recruit immune effectors to the site of infection. The observed induction of apoptotic mechanisms as an innate response to infection has also been noted, although bacteria are able to prevent this process in some instances to preserve an intracellular niche for replication and growth. These consiste nt findings have been broadly emphasized, but it remains unclear if this constitutes the entire scop e of the epithelium’s role during interaction with bacteria. In an attempt to define a conserved tran scriptional response by human host cells to infection, Jenner and Young described host responses to pathogens that have been investigated

PAGE 67

67 by transcriptional profiling of various cell lines (Jenner and Young, 2005). The existence of a conserved epithelial cell response to interaction with invasive pathogens was strongly suggested by early reports (Eckmann et al., 2000). Included in this progra m were several recurring genes, such as iNOS and cyclooxygenase-2, as well as pro-inflammatory chemokines and cytokines (Finlay and Cossart, 1997; Kaufmann and Hess, 1999). This putative core response was further investigated using a wide range of cell types, which also allowe d the elucidation of cell-type specific and pathogen-specific responses. To de fine the conserved response, a common pattern of regulation by cluster algorithms was used to overcome the differences in cell lines, time points, and pathogens used in these studies. The rationale was that the hierarchy for gene regulation would group affected genes together, a nd in discernable pattern s, regardless of the experimental design. For all cell types, the au thors determined that the core response to pathogens was the activation of 511 genes, ba sed on their co-regulation during host-pathogen interactions (Jenner and Young, 2005). Genes th at are preferentially induced by various cell types were also identified, such as immune re sponse genes and genes for neutrophil cytosolic factor, important for skin immune responses. Indeed, comparing host transcriptional profil es across all known arrays in search of response signatures was a tantaliz ing prospect. Cross-comparison has yielded useful information (Jenner and Young, 2005), but can be misleading if uniformity in the array process is lacking (Kellam, 2001). Potential pitfalls include the absence of given genes from experimental arrays, or the filtering out of these genes during the anal ysis being interpreted as these genes not being regulated (Kellam, 2001). These potential pitfalls appear to be in play in this instance, and suggest the description of “the core response” w ould more accurately be represented as “a core response.” This is more than a question of semantics, as the in consistencies between

PAGE 68

68 experimental conditions potential ly introduce significant variabil ity and make the existence of other conserved host responses likely. Table 21 illustrates the studies that have utilized transcriptional profiling of epithelial cells to stud y hostbacterial interactions, which have almost exclusively relied upon different a rrays to provide a core response. Further, many studies used custom arrays that were specifically targeted to study the cytokine and chemokine profiles elicited in host cells upon bacter ial interaction. This strategy biases the final outcome of any attempt to define host response programs towa rds processes that involve innate immune defenses, such as apoptosis, cytokine/chemokine expression, and recruitment of immune cells. As these genes are reinforced for their roles in the host response, the simultaneous omission of cell-line specific and pathogen specific response s occurs. These specific responses may reveal highly diverse host strategies that are currently under-appreciated. For instance, the ability of epithelial cells to metabolize specific bacter ial signaling molecules and virulence factors (Aldridge et al., 2005), such as the Pseudomonas quorum sensing molecule 3OC12-HSL (Chun et al., 2004), supports a more active role of the epith elium than simply recruiting immune cells. The preferential habitation of the mucosal layer by Pseudomonas in patients may arise from this metabolic activity by the epith elial layer (Doring and Wor litzsch, 2000). Additionally, the inhibition of apoptosis by invasi ve bacteria already been discussed is another example of an active epithelial response to bacter ia. Further investigation of this process is likely to reveal a great degree of pathogen-specific responses. Although the relationship between epithelial cells and inflammation is undeniable, a dditional work is necessary to reveal epithelial cell functions above and beyond the current paradigm. Current Limitations of Microarrays and Gene Ontology Annotations A common thread to the majority of global host transcriptional prof iling experiments is the challenge of making biologi cal sense of the hundreds of di fferentially regulated genes

PAGE 69

69 amongst different experimental conditions. Due to data mining limitations, further confirmation steps have been largely restricted to genes that are up-regulated in infected cells as compared to the baseline uninfected controls (Eckmann et al., 2000). This strategy is focused on the discovery of inducers of host response and will uncover general trends in the host response. However, down-regulated genes cannot completely be discounted in hostpathogen interactions. For example, the down-regulation of a given repressor may result in the induction of a gene/pathway. Discounting the genes that are do wn regulated may result in the omission of important checkpoints in response cascades. Likewise, downstream phenotypic confirmations would be limited to a small number of gene products and undermine the potential of global transcriptional profiling. In addition to improving the accuracy of re gulated genelists, methods to predict a biologically relevant outcome have emerged. One developing strategy to mine complex transcriptional datasets entails the use of ge ne ontology tools. Bio-ontologies have been described as formal representa tions of knowledge areas in whic h essential terms are combined with structuring rules that describe the relatio nships between the terms. This structured knowledge can then be linked to molecular data bases (Bard and Rhee, 2004). Thus, the products of genes that are differentially regulated can be classified according to their biological process, the molecular function that they are associated with, or their cellula r localization. Those individual ontology terms that are overly impacted in the output ge ne lists resulting from a given experimental condition are candidates for future investigation. Just as microarrays can only detect the genes that are included on the ch ip/membrane, gene ontology will only return functional information for genes that have been annotated. A ffymetrix HG-U133A chips, for instance, have 22283 total probesets, but only 16 264 of these have been annotated with a

PAGE 70

70 biological process as of March 2006. Thus, approximately a third of the probesets of interest may still not be classified by ontology analysis. An additional source of complexity derives from the multifunctional nature of many genes. Hence, the final number of biological annotations may be significantly greater than th e total number of probesets fed to an ontology analysis. The current genomic annotations bei ng greatly driven by current research, a positive feedback loop remains where, in the case of ho st pathogen interactions, gene annotations weighted by immune responses will beget the disc overy of pathways also involved with immune responses. Another potential area of bias lies in the in vitro growth of the microorganism of interest for host-pathogen interaction studies. In almost all cases, bacterial strains are grown under conditions that poorly approximate the natural envi ronment encountered at the site of infection. It is generally accepted that microorganisms modul ate their gene expressi on profile in response to their immediate environment. Thus, critical bacterial determinants may be absent from array analysis, and consequently ma y not elicit all of the respons es that would normally occur in vivo (Handfield et al., 2005b; Handfield and Hillman, 2006). Future Directions/Perspective As predicted by Cummings et al., global transcriptional pr ofiling can potentially be applied to the paradigm of dia gnosis of infectious diseases (Cummings and Relman, 2000). The hypothesis is that a pathogen expresses unique viru lence factors, which w ill trigger a signature response in the host. This signature respons e will be amplified through a cascade of unique events, thus leaving a trail fo r investigators to trace back to the perpetrating organism (Cummings and Relman, 2000). Employing such a strategy, however, requires the characterization of the host response to hundreds of pathogens, which is no small task (Cummings and Relman, 2000).

PAGE 71

71 Characterization of host responses at an actual infection site is a potentially informative exercise, although cellular heterogeneity is a potential complication (Cummings and Relman, 2000). This challenge has supported the use of cel l culture models in cell lines likely to encounter a particular pat hogen (Cummings and Relman, 20 00). The rapid progress of microarray technology may already be on the verge of solving some of these complications. As Ichimura and colleagues demonstrated, the potential exists to combinine transcriptional profiling with histology and laser microdissection to address several co mplications of transcriptional profiling of patient samples. Once the cell type of interest has been isolated and excised, the mRNA expressed within cells can be amplifie d with sensitivity and fidelity acceptable for microarray analysis (Ichimura et al., 2006). A timecourse analysis of host response regul ation performed in pa rallel with pathogen gene regulation may demonstrate temporal inte rplay during host-pathogen interactions. The simultaneous monitoring of the host and pathogen global transcriptomes in timecourse, or the “interaction transcriptome,” woul d be the preferred method to demonstrate cause and effect relationships between host and pa thogen gene expression (Matsumura et al., 2005). Experimental conditions that incl ude dead or inert bacteria in addition to live wild type and mutant pathogens are likely to demonstrate importa nt differences in host cell responses to active manipulation by pathogens. A core response to dead bacteria of many types might consist of the innate immune functions consis tently studied up to this point while the live pathogen and cell type specific differences are li kely to be apparent upon active a lteration of the host transcriptome by bacteria, and subsequent interplay. The probl ems that may arise are likely to not lie in the technology or ability to perform the experiment, but again in the familiar problem of sorting through an enormous amount of data on both the host and pathogen side. Clearly, annotation of

PAGE 72

72 genes and the degree of sequencing for the host cells and pathogens of interest are limiting factors that can contribute to this problem. Additionally, the familiar conundrum of choosing an appropriate experimental model is always pres ent; the balance between accurately reproducing the conditions of actual real world infections—trending toward s complexity—is paradoxically balanced by the need for simplicity in a model in order to yield understandable data. A large amount of data from inconsistent platforms, ti mepoints, host cell lines and other experimental conditions, is likely to lose im portant regulatory information and dilute the potential and resolution of arrays rather than maxi mizing their capacity (Jenner and Young, 2005). Despite the current challenges associated with transcriptional profiling to study hostpathogen interactions, the bioinf ormatics era is anticipated to resolve biologically relevant questions. Improvements in data mining, completi on of gene annotation, and a focused study of epithelial cells in teracting with bacteria are expected to result in novel therapies. There is some evidence that host arrays may provide interes ting candidates for vaccine development, but no new drugs, vaccines or diagnostics have yet emer ged from targets found with these arrays. Attempts with “reverse vaccinology” have consis ted of identifying genes of pathogens that are regulated upon interaction with a host cell. These in vivo -induced genes are potentially protective antigens that may be us ed in vaccine design. However, the specific interactions that are occurring between host and pa thogen are potentially far more informative. Clearly, a more thorough investigation of the differential host res ponses is necessary, especially in light of differences observed between comm ensal and pathogenic bacteria. In this endeavor, microarray analysis will continue to be an essential tool for studying hos t-pathogen interactions of the epithelium.

PAGE 73

73 Table 2-1. Transcriptional profiling of hu man epithelial cells to study host pathogen interactions 3 Epithelial Cell Line Bacterial species Strain (mutant gene) Array Type # Genes References Epithelial cells interacting with wild type pathogens: T84 Salmonella typhimurium SL1344, SL3201, KK1004, TML83, SR11 Glass 650 (Zeng et al., 2003) E schericia coli Strain 4 Glass 650 (Zeng et al., 2003) Salmonella typhi Clinical Isolate Glass 650 (Zeng et al., 2003) AGS H elicobacter pylori TN2, G27 Microarray 2304; 22630 (Guillemin et al.,2002; Maeda et al., 2001a) P12, K8 Membrane 588 (Chiou et al., 2001; Bach et al., 2002) HeLa Chlamydia trachomatis D/UW-3/Cx Membrane 1176 (Hess et al., 2001) LGV L2/434/Bu Glass, Membrane 15000; 268 (Dessus-Babus et al., 2000) Yersinia enterocolitica Serotype O8 p YV+ Affymetrix HGU133A 22283 (Bohn et al., 2004) A ctinobacillus actinomycetemcomitans VT 1169 Affymetrix HGU133A 22283 (Mans et al., 2006) Kato3 H pylori N CTC 11637 Membrane 136; 588; 10752; 46302 (Cox et al., 2001) IHGK P orphyromonas g ingivalis 33277 Affymetrix HGU133A 22283 (Handfield et al., 2005) A actinomycetemcomitans VT 1169 Affymetrix HGU133A 22283 (Handfield et al., 2005; Mans et al., 2006) F usobacterium nucleatum ATCC 25586 Affymetrix HGU133A 22283 (Hasegawa et al., 2007) Streptococcus gordonii DL1-Challis Affymetrix HGU133A 22283 (Hasegawa et al., 2007) HT-29 Salmonella dublin Lane Membrane 277 (Eckmann et al., 2000) A549 P seudomonas aeruginosa PAK Glass 1506 (Ichikawa et al., 2000) BEAS-2B B ordetella pertussis BP536 Affymetrix Hu6800 7070 (Belcher et al., 2000) MKN45 H pylori TN2 Membrane 268 (Maeda et al., 2001b) Primary UECs N eisseria gonorrhoeae Strain 1291 Affymetrix HGU95A 12626 (Binnicker et al., 2003) 3 Table adapted from: (Jenner and Young, 2005):

PAGE 74

74 Table 2-1. Continued. Epithelial Cell Line Bacterial species Strain (mutant gene) Array Type # Genes References T84 S. typhimurium SL1344 (spi1-), SL2301 (flhD-, fliB-, fliC-, fliBfliC-), KK1004 (fliD-), SR11 (invA-), TML-83 (invG-) Glass 650 (Zeng et al., 2003) E coli Strain 7 Glass 650 (Zeng et al., 2003) AGS H pylori TN2 (cagE-) Microarray 2304 (Maeda et al., 2001a) (cagA-, vacA-, cagE-, cagN-, PAI-) Microarray 22630 (Guillemin et al., 2002) P17 (cagA-) Membrane 588 (Bach et al., 2002) HeLa Y. enterocolitica p YV-, pYV-(inv), pYV+ (yopP-) Affymetrix HGU133A 22283 (Bohn et al., 2004) Kato3 H pylori G50 (PAI-) Membrane 136; 588; 10752; 46302 (Cox et al., 2001) IHGK P gingivalis YPF-1 (fimA-) Affymetrix HGU133A 22283 (Handfield et al., 2005) A actinomycetemcomitans JMS04 (orf1402-) Affymetrix HGU133A 22283 (Handfield et al., 2005) A549 P aeruginosa PAK, PAK-NP Glass 1506 (Ichikawa et al., 2000) BEAS-2B B pertussis BP356, BP9K/129G Affymetrix Hu6800 7070 (Belcher et al., 2000) MKN45 H pylori TN2 (cagE-) Membrane 268 (Maeda et al., 2001b) Epithelial cells interacting with bacterial components: AGS H pylori ABCCC-type CagA of NCTC 11637 Affymetrix Genome Focus 8500 8500 (Yokoyama et al., 2005) HT29 A eromonas hydrophila Cytotoxic enterotoxin Act Affymetrix HGU133A 22283 (Galindo et al., 2005) T84 S. typhimurium Flagellin of SL2301 Glass 650 (Zeng et al., 2003) BEAS-2B B pertussis Pertussis toxin of BP536 Affymetrix Hu6800 7070 (Belcher et al., 2000)

PAGE 75

75 Table 2-1. Continued. Epithelial Cell Line Bacterial species Strain (mutant gene) Array Type # Genes References Environmental treatments perturbing host-bacterial interactions: T84 N /A TNF Glass 650 (Zeng et al., 2003) AGS H pylori cagA transfection Affymetrix HG Focus 8500 (Yokoyama et al., 2005) BEAS-2B B pertussis BP536 Sodium salicylate, BP536 Dexamethasone Affymetrix Hu6800 7070 (Belcher et al., 2000)

PAGE 76

76 CHAPTER 34 DISTINCTIVE CHARACTERISTICS OF TR ANSCRIPTIONAL PROFILES FROM TWO EPITHELIAL CELL LINES UPON INTERACTION WITH Aggretatibacter actinomycetemcomitans5 Introduction Utilizing several strategies introduced in Ch apter 2 herein constitu tes the completion of Specific Aim 1. A reporter system using epithelial cell transcriptional profiles is established, and several experimental conditions for this system ar e optimized. One purpose of this chapter is to demonstrate the importance of choosing a biologica lly relevant model system. To this point, transcriptional profiling and ge ne ontology analyses were perfor med to investigate the unique properties of two different ep ithelial cell lines to an Aggregatibacter actinomycetemcomitans challenge. A total of 2867 genes are differentially regul ated among all experime ntal conditions. The analysis of these 2867 genes reveal s that the predominant specific response to infection in HeLa cells is associated with the regulation of enzyme activity, RNA metabolism, nucleoside and nucleic acid transport and prot ein modification. The predomin ant specific response in HIGK cells is associated with the regulation of a ngiogenesis, chemotaxis, transmembrane receptor protein tyrosine kinase signali ng, cell differentiation, apoptosis and response to stress. Of particular interest, stress response genes are si gnificantly, yet differently, affected in both cell lines. In HeLa cells, only three genes impact the response to stress, and the response to unfolded protein was the only term passing th e ontology filters. This strikingl y contrasts with the profiles 4 This work was supported in part by NIDCR grants DE13523 (MH), DE11111 and DE14955 (RJL) and T32 training grant DE07200 (JM). Analyses were performe d using BRB Array Tools developed by Dr. Richard Simon and Amy Peng Lam, National Cancer Institute 5 The following manuscript is reprin ted with permission from Blackwell Sy nergy To access the definitive version, please refer to: Mans, J.J., Baker, H.V., Oda, D., Lamont, R.J. and Handfield, M. (2006) Distinctive characteristics of transcriptional profiles from two epithelial cell lines upon interaction with Actinobacillus actinomycetemcomitans. Oral Microbiol Immunol 21 : 261-267.

PAGE 77

77 obtained for HIGK cells, where 61 re gulated genes impact the res ponse to stress, and constitute an extensive network of cell responses to A. actinomycetemcomitans interaction (response to pathogens, oxidative stress, unfolded proteins DNA damage, starvation and wounding). Hence, the vast majority of genes and ontology terms th at are currently associ ated with host-pathogen interactions are not common to HeLa and HIGK cells. Based on published reports of specific epithelial responses, the c hoice of an appropriate hos t cell line is important in light of bacterial tropism. Another condition optimized is the MOI n ecessary to elicit a representative host response. Assessing the ratios of bacteria interacting pe r host cell is a useful criterion to increase the likelihood that every host cell encounters at least a single bacter ium. A ratio of at least one bacterium per host cell should allow a homoge neous, representative mRNA response sampled for each treatment class. Furt her discussion of this hypothesi s by microscopy may be found in Chapter 5. Additionally, the baseline transcriptome of HIGK cells is established for uninfected cells, and for cells interacting with A. actinomycetemcomitans for two hours, addressing the second Specific Aim. Specific Host and Pathogen Interact ions and the Host Transcriptome Bacteria that colonize mucosal surfaces enga ge host epithelial cells in multifaceted and intimate interactions (Handfield et al 2005). For example, bacterial inhabitants of the urogenital, gastro-intestinal (GI) and respirat ory tracts can manipulate epithelial cell signal transduction pathways, often to direct their inte rnalization within these otherwise non-phagocytic host cells (Cossart and Sansonetti, 2004). Subsequentl y, epithelial cells infected with bacteria can exhibit major changes in the expresse d proteome and transcriptome (Handfield et al 2005; Hardwidge et al 2004). As a model system to study epithelial cell responses to bacterial challenge, the HeLa cell line and its derivatives have often been utiliz ed. These cells—derived

PAGE 78

78 from a cervical carcinoma—have generated much information concerning the pathogenic properties of organisms such as Shigella, Salmonella, Yersinia, EPEC, Helicobacter and many others (Aldridge et al 2005; Garbom et al 2004; Lucchini et al 2005). Similarly, various cell lines including KB (ATCC, CCL-17) and HEp-2 (A TCC, CCL-23) are often used in the study of oral periodontal pathogens such as P. gingivalis and A. actinomycetemcomitans (Cao et al 2004; Richardson et al 2005; Sandros et al 1996). Once thought to be oral in origin, both KB and HEp-2 cells are now known to be HeLa derivatives that contaminated the original cultures (American Type Culture Collection). Recently, Kang et al. demonstrated that A. actinomycetemcomitans cytolethal distending toxin (CDT) inhibits epithelial ce ll proliferation, but doe s not affect fibroblasts, when these cells are grown together in culture (Kang et al 2005). Additionally, Fine et al. have shown that the A. actinomycetemcomitans autotransporter adhesin Aae, is the adhesin res ponsible for the binding to Buccal Epithelial Cells (BEC) isolated from humans and old world primates, but not for BEC derived from new world primates and several other mammalian species (Fine et al 2005a). These examples demonstr ate that interactions between A. actinomycetemcomitans and host cells appear to exhibit specificity and tropism. However, it is unclear if this tropism extends beyond the initial attachment of A. actinomycetemcomitans to oral cells and impacts the host-cell transcriptome. This question is particularly relevant in light of the increasing recognition of oral pathogens for their role in nonperiodontal conditions—such as coronary artery disease (Beck et al 2005; Genco et al 2002) and birth of preterm-low birth weight infants (Scannapieco, 2005)—and subsequent interactions with different tissue types. Undirected methods, such as DNA microarra ys, can be used to survey the global transcriptional profiles of host cells in response to many differen t conditions. This approach is

PAGE 79

79 highly useful for uncovering new pr ocesses involved with bacteria l interaction beyond the effects of well-characterized adhesins a nd toxins. The value of this a pproach is especially apparent when host cell phenotypic changes are outwardly subtle and not easily observable using other methods. In this study, we have compared th e transcriptional responses of two cell lines commonly used to study host-pathogen inte ractions—Human Immortalized Gingival Keratinocytes (HIGK) and HeLa cells—following A. actinomycetemcomitans co-culture. Materials and Methods Bacteria and Cell Lines HeLa (KB cells; CCL-17, American Type Culture Collection) and HIGK cells (Oda et al 1996) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) and Keratinocyte Serum-Free Medium (KSFM), respectively, as a monol ayer to 95% confluence in an atmosphere of 5% CO2 at 37o C (Cao et al 2004; Oda et al 1996). Both cell culture media were supplemented with 50 U penicillin/streptomycin mL-1 (Gibco). A. actinomycetemcomitans smooth strain VT1169 (SUNY465 NalR RifR) (Mintz and Fives-Tayl or, 2000) was grown in liquid culture at 37o C and 10% CO2 to mid-logarithmic phase, and prepared for host cell coculture according to standard methods (Richardson et al 2005). Briefly, ep ithelial cells were washed three times with 1X Dulbecco’s Phos phate-Buffered Saline (Cambrex, Walkersville, MD) to remove residual antibioti cs and waste products. In biol ogical replicates of four per condition, epithelial cells were sham-infected with cell culture media or co-cultured with A. actinomycetemcomitans resuspended in culture media, resulting in a multiplicity of infection (MOI) of 1000:1. Previous studies in our lab (data unpublished) determined that 1000 was the lowest MOI ensuring every host ce ll encountered at least a singl e bacterium, resulting in a homogeneous population of infected host cells, and thus a represen tative mRNA sample of the infected state. Two hours co-culture was the ti me point previously determined to display a

PAGE 80

80 phenotype that can be characterized in terms of host cell monolayer inte grity, and chosen in order to maintain consistency with previous work in our lab (Handfield et al 2005). Epithelial cells were lysed with Trizol (Invitrogen Life Technologies, Carlsbad, CA) and RNA was prepared for GeneChip hybridization as recently described (Handfield et al 2005). Microarray Analysis Assessment of the host cellular responses to bacterial challenge was accomplished by transcriptional profiling us ing Affymetrix Human Genome HG U133A DNA microarrays (Feezor et al 2003). Infected and uninfected HeLa and HIGK cells were tested in four independent replicates. Subsequent array anal ysis was performed as recently presented (Handfield et al 2005). In brief, expression filters were applied to remove Affymetrix control oligos and probesets whose signal was undetected acr oss all samples. The signal intensity values of the resulting dataset were variance-normalized mean-centered, and ranked by their coefficient of variation. Normalization was performed in orde r to give equal weight to all probesets in the analysis, regardless of the ra w signal intensity order of magni tude. To reduce the confounding effect of background signal variation on the analys is, the half of the dataset demonstrating the most variation across samples was used to perf orm unsupervised hierarch ical cluster analysis using Cluster software (Eisen et al 1998). The resulting heat-map and Cluster dendrograms were visualized with Treeview (Eisen et al 1998) to reveal the extent of characteristic host-cell responses to each infection state, defined as identical treatments clustering together. Following initial assessment of the host cell response to each condition, supervised analysis was performed to investigate differe nces in gene regulation among experimental conditions. For this analysis, th e raw signal intensities were l og-transformed for all probesets passing initial expression filters and correlate d using BRB Array Tools (Simon and Peng-Lam). In each supervised analysis, biological replicates were grouped into classes according to host cell

PAGE 81

81 type and infection state during co -culture experiments. Several methods of class prediction were utilized (compound covariate predic tor, nearest neighbor predicto r, and support vector machine predictor) in order to generate lists of classi fiers whose expression stat e changes between classes at P <0.001 and P <0.01 levels of significance. Leave-one-out crossvalidation (LOOCV) was performed to test the accuracy of each classifier and compared to the probability of a correct class prediction by chance alone, based on the Pvalue and total number of genes analyzed. In order to visualize the differentially-regulated genes, Microsoft Access database queries were used to match the subset of significantly regul ated genes with their associated variancenormalized, mean-centered signal values calculated previously. Cluster and Treeview were used to visualize the correlatio ns among genes and samples. Ontology Analysis The biological significance of the transcriptional profiles wa s investigated using gene ontology tools available online thro ugh NetAffx Analysis Center (A ffymetrix Inc., Santa Clara, CA). Cross-validated probesets from the HG-U133A Gene Chip Arrays that were differentially expressed between classes at the P <0.001 leve l of significance were annotated with their associated biological process ontology terms. Biological processes impa cted by two or more regulated genes were visually examined via dire cted acyclic graphs (DAG) to gain insight of epithelial cell responses to A. actinomycetemcomitans co-culture. The total number of genes regulated per biological proce ss, percentage of total genes impacted per term, and P-values calculated by NetAffx, were the criteria used to prioritize biological pr ocesses. Consistency between parent and child ontology terms was a prer equisite for additional characterization of the predicted biological response of HeLa and HIGK cells upon A. actinomycetemcomitans interaction.

PAGE 82

82 Results and Discussion Previous work suggested that a number of hos t factors are differentially expressed in response to challenges by oral pathogens. In particular, A. actinomycetemcomitans has been shown to display tissue tropism (Fine et al 2005b) and its well-characterized toxins have drastically different effects on different cell types (Kang et al 2005; Kanno et al 2005). Although these are clinically rele vant examples of host gene m odulation responding to bacterial challenge, the extent to which the transcript ome is impacted in a tissue specific manner upon direct cellular interaction with A. actinomycetemcomitans remains unclear. Hence, extensive transcriptional profiling and gene ontology analysis was performed to investigate the similarities and differences in the transcriptional response by two different lineages of epithelial cells to an A. actinomycetemcomitans challenge. Initially, all samples from uninfected and in fected HeLa and HIGK cells were used to determine the overall similarity of the transcript ional profiles of these two epithelial cell lines. Signal intensity data for the 14,171 probesets that pa ssed initial expression filters were used to perform unsupervised cluster anal ysis and supervised class pr ediction as described in the Materials and Methods section. Unsupervised hierarchical cluster analysis revealed a characteristic host cell transcriptional profile, as biological replicates clus tered together (data not shown). Class prediction at the P <0.001 leve l of stringency reveal ed that 2867 genes were differentially regulated among all experimental conditions; by chance alone and with a normal distribution, one would expect that fourteen genes be identified as false positives. In addition, linear discriminant analysis and 1-nearest ne ighbor classifications were 100% accurate by LOOCV for 2000 random permutations, while n earest centroid and 3-nearest neighbors classifications were 80% accurate. Both rates of 80% and 100% are significantly more accurate than 25% correct classification rate expected by chance alone for class prediction using four

PAGE 83

83 classes. This analysis conferred a high degree of confidence that these 2867 genes were indeed differentially regulated among all classes tested. As 2867 genes repr esents approxim ately 20% of the total genes analyzed, 80% of the transcri ptome is neither signifi cantly changed between HeLa and HIGK cells, nor is it impacted significantly upon an A. actinomycetemcomitans 2-hour co-culture. This level of similarity is consistent with the fact that both cell lines are epithelial in nature. Processes that are universally important to cell homeostasis would be predicted to be unaffected by bacterial challenges and thus constit utively expressed by epit helial cells regardless of their lineage. Indeed, a partial survey of the constitutively expressed biological processes revealed cellular functions rela ted to RNA synthesis, metabolism, protein synthesis, and other generalized cellular processes (d ata not shown). Limitations in the algorithms used herein restricted the analysis to 1500 pr obesets per query, preventing an exhaustive ontology analysis for all 11,304 genes that were detected with the arrays but not differen tially modulated between conditions or cell lines. Treeview visualization of the 2867 probesets differentially expressed among all four classes (Figure 3-1) revealed interesting charac teristics of HeLa and HIGK cells. The measured distance required to connect samples along the scaled dendrogram path reflects how closely related the transcriptional profiles of each sample are, based on Pearson’s correlation coefficient. Thus, if a difference exists between two classes of tr eatments (infected versus uninfected, for example) replicates of a treatment will be more closely related to each other than to all other samples, as the intra -class distances required to connect these samples is shorter than the inter class distances. This analysis also represen ted an indirect measure of the degree of noise introduced in that experimental system. In Figure 3-1, the major node of separation occurred between HeLa and HIGK cells, regardless of inf ection state. Less pronoun ced, yet significant,

PAGE 84

84 nodes of separation could also be detected betw een uninfected and infected cells of the same lineage. As the differences between cell lines overshadowed the observable differences between infected and uninfected states, this dendrogram suggested a significant, lineage-based difference between the global transcriptional responses of these two epithelial cell types despite their high degree of similarity in housekeeping functions and intermediate metabolism. However, it cannot be ruled out that some differences discovered be tween the transcriptiona l profiles of HeLa and HIGK cells resulted from the different cell culture media used, differences in the growth rates of these cell lines, or differences that exist be tween HPV Type 16 (HIGK) and HPV Type 18 (HeLa) immortalization (Aguilar-Lemarroy et al 2001). To eliminate these variables, and further investigate the cell line-specific transcri ptional profiles uncovered by our initial analysis, a comparison of the A. actinomycetemcomitans -infected state with th e corresponding, baselineuninfected state was performed for both cell line s independently. For these analyses, signal intensities were re-normalized across all samples, and both unsupervised and supervised analyses were repeated as presented above. In HeLa cells 10,921 genes passed the initial expression filters, while 13,176 genes were analyzed in HIGK cells. Class prediction for HeLa cells revealed th at only 67 genes were differentially expressed upon A. actinomycetemcomitans infection at the significance level of P <0.001 (Figure 3-2A). In contrast, this anal ysis performed on HIGK cells yielded 625 significantly modulated genes (F igure 3-2B). LOOCV analysis for 2000 random permutations confirmed these predictors at 100% correct classi fication rate, using a number of analyses such as the compound covariate predictor, the diagon al linear discriminant, the 1 and 3-nearest neighbor, the nearest centroid, and th e support vector machines analysis.

PAGE 85

85 A directed effort was undertaken to inve stigate the extent of the common core transcriptional response by these two cell lines upon A. actinomycetemcomitans infection. Using Microsoft Access database queries, the 625 signifi cantly-modulated genes in HIGK cells, and the 67 genes significantly-regulated in HeLa cells were correlated. Upon A. actinomycetemcomitans interaction, the common transcrip tional response of HeLa and HIGK cells consisted of eight probe sets. Although statistically si gnificant (P=1.4E -05), this cons titutes only 1.17% of the total 684 genes regulated in both cell lines combined. In other words, a common core response to infection was found, but it was minimal and its biological significance remains uncertain. The expression patterns for seven of these eight genes showed a consistent pattern of regulation for both cell lines, six of which were up-regulat ed and one was down regulated (Table 1). To investigate the possible effect of sampling error on this outcome, we repeated the class prediction analysis at lower stringency of P <0.01. At that level of significance, HeLa cells modulated 404 genes and HIGK cells modulated 2011 genes (c ompared to 109 and 132, respectively, which would be expected by chance alone at this confidence level). Mi crosoft Access queries of this dataset revealed 84 genes regulated by both cel l lines, representing 3.6% of the total genes modulated. This is of the sa me order of magnitude as the 1.17% of genes found to be in common at P <0.001 significance. Thus, this sup ports our contention that the low number of genes found to be modulated in both cell lines at P <0.001 was not the resu lt of statis tical error caused by the low sample number from HeLa cells. The biological significan ce of this core transc riptional response to A. actinomycetemcomitans interaction was further investigated using the gene ontology tools as described in the Materials and Methods. The annot ations were available for five of the eight genes presented in Table 1. The resulting output was 11 biological processes organized into four

PAGE 86

86 main branches and associated with de velopment (P=0.08484), morphogenesis (P=0.02327), primary metabolism (P=0.80176) and signal tran sduction (P=0.21298). Consistent with the stringent analysis presented above, the ontology analysis repeated for the 84 common genes identified at the P <0.01 threshold also impli cated generalized cellula r processes as being impacted in both cell lines. This corroborates our initial finding that ap proximately 80% of the transcriptome is modulated simila rly between the two cell lines upon A. actinomycetemcomitans interaction for processes important to general ho meostasis and not specifically related to hostpathogen interactions. The biological processes that were differe ntially impacted in the two cell lines upon A. actinomycetemcomitans infection were analyzed using th e same gene ontology algorithms described above. The 625 genes of HIGK cells that were found to be differentially regulated at a level of significance of P <0.001 were annotated and visualized. Using the filters described above, seven high-priority groups of host respon ses were identified. Th e biological processes identified included the regulation of angiogenesi s, chemotaxis, transmembrane receptor protein tyrosine kinase signaling pathwa y, cell differentiation, and response to stress. Similarly, HeLa cells revealed a predominant specific response as sociated with the regulation of enzyme activity, RNA metabolism, nucleoside and nucleic acid tran sport, and protein modification. Consistent with previous reports, ontology terms related to cell death and apoptosis were uncovered in both cell lines (Handfield et al 2005). Of immediate interest in the context of host-pathogen interac tions were genes associated with stress-response. This biol ogical process was significantly im pacted in both HIGK and HeLa cells. However, a side-by-side comparison of th e child ontology terms for the response to stress in HIGK (Figure 3-3A) and HeLa cells (Figure 3-3B) revealed significan t differences in the

PAGE 87

87 extent and nature of the transcriptional res ponse uncovered. In HeLa cells, only three genes impacted the response to stress, and the respon se to unfolded protein was the only child term present that passed the ontology filters. This strikingly cont rasted with the DAG obtained for HIGK cells, where 61 regulated genes impacted the response to stress, and constituted an extensive network of cell responses to A. actinomycetemcomitans interaction. For example, six child terms, including the res ponse to wounding and the response to DNA damage stimulus were uncovered. These are consistent with the effect of the cytolethal distending toxin which arrests cell growth at the G2/M phase through DNA dama ge in immune cells and other cell types (Belibasakis et al 2005). In addition, 24 genes impacted the ontology terms associated with the response to pest, pathogen or para site in HIGK cells. Included in this list were genes involved in the inflammatory response such as IL-1 beta an d IL-6. IL-1 beta up-regul ation in HIGK cells is consistent with reports documenting increased ex pression of this pro-inflammatory cytokine in primary GEC (Sfakianakis et al 2001) and has previously been reported (Handfield et al 2005). IL-6 stimulation in gingival fibroblasts by A. actinomycetemcomitans has been demonstrated in connection to CDT (Belibasakis et al 2005). This representative example illustrated that biological processes impacted in both HeLa and HIGK cells may still be regulated differently Consequently, solely identifying a list of differentially regulated genes between two conditi ons is not sufficient to predict a biologically significant outcome. Hence, the adjuncti on of a thorough ontology analysis favorably complements the transcriptional profile analysis and is invaluable in the context of a complex host-pathogen interaction. The extensive transcriptional profiling and ge ne ontology analysis de scribed herein did uncover a large number of common biological pr ocesses shared between both epithelial cell

PAGE 88

88 lines. However, the vast majority of genes and on tology terms that are currently associated with host-pathogen interac tions were not common to HeLa and HIGK cells. The relatively high number of differentially regulated genes found in oral HIGK cells (625) as compared to HeLa (67) is consistent with the tissue tr opism displayed by the Aae adhesin of A. actinomycetemcomitans (Fine et al 2005a). The data presented he re further suggested that the host transcriptional response to A. actinomycetemcomitans challenge is substantial. Further, in contrast to the current paradigm, the response of or al epithelial cells in ho st defense to infection appeared to be tailored, and have ramifications extending be yond specific toxicity and tissue tropism. To our knowledge, this study presen ts the first report of the intr insic differences that exist at the global host cellular level for two different epithelial cell lines in co-culture with the same pathogenic oral organism. By extrapolation, this study also em phasizes that caution should be exercised in the choice of epithel ial cell lines or animal models of infection, regardless if the specific model behaves similarly in te rms of adhesion and cytotoxicity. Finally, the current study has evolutionary implications for the investigation of bacterial adaptation to association with host cells. Gene regulation in a dhering or invading bacteria may depend not only on the presence of specific ad hesins, but also to some extent, on the physiological status of the host ce lls. Thus, this report exemplifie s that host-pat hogen interaction may be more relevant if performed in the cont ext of host cells derived from the tissue and the host of interest. The deta iled analysis presented here supports the use of transcri ptional profiling as a powerful tool to establis h the basis of intrinsic simila rities and discrepancies amongst different models of infection. This may be particul arly useful to substantiate some contradictory reports in the literature pertaining to a variety of oral and other microorganisms.

PAGE 89

89 Table 3-1. Transcriptional regula tion of common probe sets to A. actinomycetemcomitans infected HeLa (KB) and HIGK epithelial cells Probe Set ID HeLa HIGK Gene Title Gene Symbol 202028_s_at UP UP ----202499_s_at UP UP solute carrier family 2 SLC2A3 206323_x_at UP UP oligophrenin 1 OPHN1 210095_s_at DOWN DOWN insulin-like growth factor binding protein 3 IGFBP3 212368_at UP UP zinc finger protein 292 ZNF292 216609_at UP UP ----221943_x_at UP UP ----222155_s_at DOWN UP G protein-coupled receptor 172A GPR172A

PAGE 90

90 Figure 3-1. Divergence of HeLa and HIGK cell tr anscriptional profiles. This heat map and dendrogram were constructed from 2867 probe sets differentially expressed between the four experimental classes at the significance level of P <0.001. Probe set signal intensities were variance-no rmalized, mean-centered across samples, and subjected to hierarchi cal cluster analysis. Av erage linkage clustering by uncentered correlation was performed for genes and samples. The degree of similarity between the transcriptional profiles of each sample is expressed by Pearson’s correlation coefficient distance me tric, according to the adjacent scale. The expression state of each data point is represented as standard deviations from the mean expression level for that gene in all samples. Red indicates a relative increase, green indicates a relative decr ease, and black indi cates no relative change of mRNA transcripts for a given gene. Labels. Uninfected HIGK Cells, IHGK CTRL 01-04; A. actinomycetemcomitans -infected HIGK Cells, IHGK Aa 01-04; Uninfected HeLa Cells, HeLa CTRL 01-04; A. actinomycetemcomitans infected HeLa Cells, HeLa Aa 01-04.

PAGE 91

91

PAGE 92

92 Figure 3-2. Different patterns of gene expr ession by HeLa and HIGK cells upon co-culture with A. actinomycetemcomitans RNA was isolated and purified after 2h coculture with A. actinomycetemcomitans and compared to uninfected cells, for both cell lines independently. Probe set signa l intensities were variance-normalized, mean-centered across samples, and subjected to hierarchical cluster analysis. Average linkage clustering by uncentered correlation was performed for genes and samples. Heat maps and dendrograms were constructed from 67 probe sets for HeLa cells (A), and 625 probe sets in HIGK cells (B), diffe rentially expressed between uninfected and A. actinomycetemcomitans -infected treatments. The level of significance was P <0.001. The de gree of similarity between the transcriptional profiles of each sample is expressed by Pearson’s correlation coefficient distance metric, according to th e adjacent scale. The expression state of each data point is represented as sta ndard deviations from the mean expression level for that gene in all samples. Red indicates a relative increase, green indicates a relative decr ease, and black indicates no relative change of mRNA transcripts for a given gene. Labels Uninfected HIGK Cells, IHGK CTRL 0104; A. actinomycetemcomitans -infected HIGK Cells, IHGK Aa 01-04; Uninfected HeLa Cells, HeLa CTRL 01-04; A. actinomycetemcomitans -infected HeLa Cells, HeLa Aa 01-04.

PAGE 93

93

PAGE 94

94 Figure 3-3. Processes associated with stress response in HeLa and HIGK cells that are impacted upon A. actinomycetemcomitans interaction. Differentially regulated probe sets were annotated, and their associated gene ontology terms were visualized with NetAffx. Biological pro cesses were organized by directed acyclic graphs (DAG), consisting of parent and child terms progressing from left to right. The degree of impact upon the ontology ne twork caused by each transcriptional profile was expressed in te rms of the percentage of total probe sets on the HGU133A array. This DAG is one representa tive example of the total biological response to A. actinomycetemcomitans interaction by HIGK cells (A) and the corresponding analysis in HeLa cells (B ). Individual nodes are color-coded on a spectrum of blue to red, w ith the latter indicating ontology terms most impacted. Biological processes missing in HeLa cells relative to HIGK cells are shown in gray.

PAGE 95

95 A response to unfolded protein (4) response to pest, pathogen or parasite (24) response to stress (61) response to wounding (19) response to oxidative stress (3) response to DNA damage stimulus (17) response to starvation (2) response to reactive oxygen species (2) cellular defense response (5) wound healing (5) acute-phase response (2) DNA repair (13) humoral immune response (4) humoral defense mechanism (sensu Vertebrata) (3) blood coagulation (5) nucleotide-excision repair (3) inflammatory response (8) mismatch repair (4) response to virus (5) response to pathogen (2) cellular response to starvation (2) response to pest, pathogen or parasite response to wounding response to oxidative stress response to DNA damage stimulus response to starvation response to reactive oxygen species cellular defense response wound healing acute-phase response DNA repair humoral immune response humoral defense mechanism (sensu Vertebrata) blood coagulation nucleotide-excision repair inflammatory response mismatch repair response to virus response to pathogen cellular response to starvation response to stress (3) response to unfolded protein (2)B response to unfolded protein (4) response to pest, pathogen or parasite (24) response to stress (61) response to wounding (19) response to oxidative stress (3) response to DNA damage stimulus (17) response to starvation (2) response to reactive oxygen species (2) cellular defense response (5) wound healing (5) acute-phase response (2) DNA repair (13) humoral immune response (4) humoral defense mechanism (sensu Vertebrata) (3) blood coagulation (5) nucleotide-excision repair (3) inflammatory response (8) mismatch repair (4) response to virus (5) response to pathogen (2) cellular response to starvation (2) response to pest, pathogen or parasite response to wounding response to oxidative stress response to DNA damage stimulus response to starvation response to reactive oxygen species cellular defense response wound healing acute-phase response DNA repair humoral immune response humoral defense mechanism (sensu Vertebrata) blood coagulation nucleotide-excision repair inflammatory response mismatch repair response to virus response to pathogen cellular response to starvation response to stress (3) response to unfolded protein (2)B response to unfolded protein (4) response to pest, pathogen or parasite (24) response to stress (61) response to wounding (19) response to oxidative stress (3) response to DNA damage stimulus (17) response to starvation (2) response to reactive oxygen species (2) cellular defense response (5) wound healing (5) acute-phase response (2) DNA repair (13) humoral immune response (4) humoral defense mechanism (sensu Vertebrata) (3) blood coagulation (5) nucleotide-excision repair (3) inflammatory response (8) mismatch repair (4) response to virus (5) response to pathogen (2) cellular response to starvation (2) response to unfolded protein (4) response to pest, pathogen or parasite (24) response to stress (61) response to wounding (19) response to oxidative stress (3) response to DNA damage stimulus (17) response to starvation (2) response to reactive oxygen species (2) cellular defense response (5) wound healing (5) acute-phase response (2) DNA repair (13) humoral immune response (4) humoral defense mechanism (sensu Vertebrata) (3) blood coagulation (5) nucleotide-excision repair (3) inflammatory response (8) mismatch repair (4) response to virus (5) response to pathogen (2) cellular response to starvation (2) response to pest, pathogen or parasite response to wounding response to oxidative stress response to DNA damage stimulus response to starvation response to reactive oxygen species cellular defense response wound healing acute-phase response DNA repair humoral immune response humoral defense mechanism (sensu Vertebrata) blood coagulation nucleotide-excision repair inflammatory response mismatch repair response to virus response to pathogen cellular response to starvation response to stress (3) response to unfolded protein (2)B response to pest, pathogen or parasite response to wounding response to oxidative stress response to DNA damage stimulus response to starvation response to reactive oxygen species cellular defense response wound healing acute-phase response DNA repair humoral immune response humoral defense mechanism (sensu Vertebrata) blood coagulation nucleotide-excision repair inflammatory response mismatch repair response to virus response to pathogen cellular response to starvation response to stress (3) response to unfolded protein (2)B

PAGE 96

96 CHAPTER 46 DISTINCT TRANSCRIPTIONAL PROFILES CHARACTERIZE ORAL EPITHELIUMMICROBIOTA INTERACTIONS7 Introduction Transcriptional profiling, bioinformatics, statistical, and ontology tools were used to uncover and dissect genes and pathways of huma n gingival epithelial cell s that are modulated upon interaction with th e periodontal pathogens A. actinomycetemcomitans and P. gingivalis The baseline transcriptome of HIGK cells is cons istent with findings from the previous chapter for uninfected HIGK cells and cells interacting with Aggregatibacter actinomycetemcomitans for two hours at the MOI of 1000:1. Additionally, the baseline response of HIGK cells to Porphyromonas gingivalis interaction is determined for tw o hours co-culture at the MOI of 100:1. These experiments constitute the completion of Specific Aim 2. Furthermore, the studies reported in this chapter move beyond simply establishing a model using transcriptional profiles of oral epithelial cells. Some of the anticipated insights into host-pathogen intera ctions are revealed. The differential regulation of pathways in epithelial cells upon bacterial inter action, such as apoptosis, is demons trated. Consistent with their biological and clinical differences, the common core transcriptional response of epithelial cells to both organisms is limited, and organism-specifi c responses predominate. A large number of differentially regulated genes linked to the P53 apoptotic network were found with both 6 This work was supported in part by NIDCR grants DE13523 (MH), DE11111 and DE14955 (RJL), T32 training grant DE07200 (JM) and funds from the Center for Molecular Microbiology (APF). The research presented here has complied with all relevant federal guidelines and institu tional policies regarding the use of human subjects. We thank Dr. Dolphine Oda (University of Washington) for kindly providing HIGK cells; and Drs. K. Mintz and P. Fives-Taylor for kindly providing strain VT1169 and pl asmid pVT1461. Analyses were performed using BRB Array Tools developed by Dr. Richard Simon and Amy Peng Lam. We thank Renata Salas Collazo and Paolo Rodrigues for their technical assistance. 7 The following manuscript is reprin ted with permission from Blackwell Sy nergy To access the definitive version, please refer to: Handfield, M., Mans, J.J., Zheng, G., Lopez, M.C., Mao, S., Progulske-Fox, A., et al (2005) Distinct transcriptiona l profiles characterize oral epit helium-microbiota interactions. Cell Microbiol 7 : 811823.

PAGE 97

97 organisms, which is consistent with th e pro-apoptotic pheno type observed with A. actinomycetemcomitans and anti-apoptotic phenotype of P. gingivalis. Furthermore, with A. actinomycetemcomitans the induction of apoptosis does not appear to be Fasor TNF mediated. The limits of the reporter system are pre ssed by combining mutant analysis with transcriptional profiling. Linka ge of specific bacterial comp onents to host pathways and networks provides additional insigh t into the pathogenic process. In so doing, this chapter also addresses the third and final specific aim, the c ontribution of specific ba cterial components to host pathogen interactions. Comparison of the transcriptional responses of epithelial cells challenged with parental P. gingivalis or with a mutant of P. gingivalis deficient in production of major fimbriae, which are required for optimal invasion, showed major expression differences that reverberated thr oughout the host cell tran scriptome. In contrast, gene ORF859 in A. actinomycetemcomitans, which may play a role in intracellular homeostasis, had a more subtle effect on the transcriptome. These studies begin to unravel the complex and dynamic interactions between host epitheli al cells and endogenous bacteria that can cause opportunistic infections. Background The human microbiota comprises a comp lex ecosystem characterized by the simultaneous presence of a large number of nor mal colonizers, associated with health and thriving in a dynamic environment. Since health is the most common state of a host, it has been speculated that the autochthonous flora has co-evolved with its host to inter act in a balanced state that is beneficial to both the host and th e microbiota (Galan and Zhou, 2000). There are an appreciable number of benefits to the host that the indigenous microbiota is thought to provide, including the synthesis of vitamins (B complex and K), the prevention of infection by pathogens

PAGE 98

98 (by direct competition for niches or by immune cross-reactivity), and impacting the normal development of the immune system (Hooper a nd Gordon, 2001). Furthermore, there is an increasing realization that co mplex societies of indigenous microbes can influence human physiology and development. For example, in the GI tract the Gram-negative anaerobe Bacteroides thetaiotaomicron can modulate expression of ileal epithelial cell genes involved in nutrient adsorption, mucosal barrier fortifica tion, xenobiotic metabolism, angiogenesis, and maturation (Hooper et. al., 2001). Since host and microbiota interactions are dynamic, disease may arise at the mucosal surface of a susceptible hos t when a perturbation occurs in the epithelial environment, for example, when the host becomes immunocompromised, or as a result of the unintended (in an evolutionary se nse) consequences of bacterial activity (Galan and Zhou, 2000). In the oral cavity, periodontal infections that affect and ultimately destroy the tissues supporting the teeth are among the most common diseases of hu mans. According to the 2000 Surgeons General’s Report on Oral Health (NID CR, 2000), these conditions afflict 14% of adults aged 45-54 and 23% of those aged 65-74 years. Fu rthermore, an epidemiological association is emerging between periodontal infect ions and serious systemic conditi ons such as coronary artery disease and preterm delivery of low birth weight infants (S cannapieco and Genco, 1999). The etiology of oral infectious dis eases is complex and involves cons ortia of bacteria thriving in biofilms and exploiting immunological susceptibilit ies in the host. De spite the multifactorial nature of these diseases, there is a consistent relationship between the Gram-negative capnophile Aggregatibacter actinomycetemcomitans and localized aggressive pe riodontal disease (Slots and Genco, 1984; Zambon, 1985; Haffajee and Socr ansky, 1999; Offenbacher, 1996; Meyer and Fives-Taylor, 1997), and between the Gram-negative anaerobe Porphyromonas gingivalis and

PAGE 99

99 severe, chronic manifestati ons of the disease (Slots et al., 1986, Slots and Genco, 1984; Haffajee and Socransky, 1999). The initial interface between the host and the potentially periodontopathic organisms, such as P. gingivalis and A. actinomycetemcomitans is the epithelial layer that lines the subgingival crevice. Epithelial ce lls are both a physical barrier to infection and a component of a network that efficiently signals microbial intrus ion to the immune cells to insure effective mobilization of the innate and specific defe nse mechanisms (Kagnoff and Eckmann, 1997). Both A. actinomycetemcomitans and P. gingivalis are capable of inva ding gingival epithelial cells and can remain viable intracellularly. Fu rthermore, epithelial ce lls maintain viability following intracellular penetration by either P. gingivalis or A. actinomycetemcomitans (Nakhjiri et al., 2001; Takayama et al., 2003; Kato et al., 2000). However, the entry mechanisms employed by these invasive organisms are distinct. A. actinomycetemcomitans enters epithelial cells through a dynamic multistep process whereupon the organisms are first constrained in an intracellular vacuole from which they subsequently escape and spread cell-to -cell with the aid of microtubules (Meyer et al., 1996, 1999). Within epithelial cells A. actinomycetemcomitans upregulates a distinct set of genes that facilitate adaptation to the intracellular environment (Cao et al., 2004, Richardson et al., 2004). Among these genes is ORF859 encoding a conserved protein of unknown function. In the case of P. gingivalis, the major fimbriae (comprised of the FimA protein) bind to integrins on the surfaces of gingival epitheli al cells and stimulate integrindependent signaling to effect i nvasion through both microfilame nt and microtubule remodeling (Yilmaz et al., 2002, 2003). P. gingivalis also impacts the MAP-kinase pathway and causes transient increases in intracellular Ca2+ concentrations (Watanabe et al., 2001; Belton et al., 2004). Both of these signal transduction pathways can converge on nuclear transcription factors

PAGE 100

100 and modulate gene expression. Indeed, P. gingivalis has been shown to affect expression of individual genes in epithelial cells includi ng those encoding IL-8 and Bcl-2 (Darveau et al., 1998; Nakhjiri et al., 2001). Transcriptional profiling using microarrays provides a means to monitor epithelial cell responses to invading microorganisms on a gl obal scale (Kagnoff and Eckmann, 2001; Yowe et al., 2001; Kellam, 2000, 2001; Kato-Maeda et al., 2001; Manger and Re lman, 2000; Cummings and Relman, 2000; Lory and Ichikawa, 2002; Ichikawa et al., 2000; Sepulveda et al., 2002). Results from such studies suggest that the en counter between host and microbiota may involve a finely tuned set of interactions whereby both cell type s adapt and co-exist with each other. Consequently, the regulation of normal processes such as cell di vision or apoptosis may be key to maintaining a balanced longstanding intracel lular state whereby both cell types inflict a minimal degree of harm on each other. In support of this concept, epithelial cells recovered from the oral cavity show high levels of intracellular P. gingivalis and A. actinomycetemcomitans (Rudney et al., 2001; Christersson et al., 1993, 1987a, b). Hence an intr acellular location may be a natural component of th e lifestyle of these oral organisms. In this study, we have utilized human microarrays to determine the transcriptiona l response of gingival epithelial cells to coculture with P. gingivalis or A. actinomycetemcomitans Moreover, we have extended these studies to investigate the transcriptional response s of epithelial cells that are manipulated by the major fimbriae (FimA) of P. gingivalis and the intracellulary upregulated ORF859 of A. actinomycetemcomitans Results and Discussion General Considerations To investigate early even ts in oral infection by P. gingivalis and A. actinomycetemcomitans we analysed differential gene e xpression in human gingival epithelial

PAGE 101

101 cells using the Affymetrix HG U 133-A oligonucleotide arrays th at contain over 22,000 different probe sets. In addition, isogenic mutant strains were utilized to assess the roles of specific bacterial proteins in modulation of the host cell transcriptome. Host cell apoptosis, the major pathway impacted by P. gingivalis and A. actinomycetemcomitans was validated by phenotypic assays. Association of A. actinomycetemcomitans and P. gingivalis With Epithelial Cells P. gingivalis and A. actinomycetemcomitans demonstrate differing e fficiencies of binding to, and internalization within, human immortalized gingival keratinocytes (HIGK). In order to compare epithelial cell transcrip tional profiles in response to an equivalent challenge of the two organisms, we first compared adhesion and invasi on at MOIs predicted to result in the same number of epithelial cell -associated bacteria for each species. As shown in Table 4-1, at a MOI of 100:1 for P. gingivalis and 3,000:1 for A. actinomycetemcomitans the numbers of bacteria associated with the epithelial cells were of the same order of magnitude. These MOIs were then used in subsequent experiments. In contrast, the levels of invasion we re significantly different; P. gingivalis being a considerably more efficient invasive microorganism as compared to A. actinomycetemcomitans Notably, HIGK cells behaved simila rly to primary gingival epithelial cells (GEC) in co-cultures with A. actinomycetemcomitans and P. gingivalis with regard to both adhesion and invasion. Further confirmation of the relevance of the HIGK cell model was provided by the finding that the gene for IL-1 beta was upregulated in HIGK cells in co-cultures with both periopathogens (not shown). This is consistent with repor ts documenting increased expression of this pro-inflammatory cy tokine in primary GEC (Sfakianakis et al., 2001; Sandros et al., 2000).

PAGE 102

102 Gene Expression in Gingival Epithelial Cells Regulated by A. actinomycetemcomitans and P. gingivalis To characterize epithe lial cell responses to A. actinomycetemcomitans and P. gingivalis, and to assess the extent to which host res ponses may depend on the challenging organism, we used human cDNA microarrays to monitor relati ve abundance of HIGK cell transcripts following co-culture with A. actinomycetemcomitans or P. gingivalis Array-to-array comparisons were carried out using unsupervised and supervised me thods to assess the rela tedness of the specimens (arrays) under investigat ion using the Cluster and TreeView Software (Eisen et al., 1998). The significance level used in identifying genes th at were differentially expressed was P < 0.001. Hierarchical clustering was first used to perfor m an unsupervised analysis. Visual representation of the unsupervised cluster analysis of P. gingivalis -infected, A. actinomycetemcomitans infected, and uninfected cells wa s performed using Treeview software. The resulting dendrogram revealed that the array chips from each infection state clustered together ( not shown). Thus, each infection state elicited a specific and distinct transcriptome in HIGK cells. This was also an indication of the quality and consiste ncy of the hybridization procedure. Supervised analyses were next performed to identify gene expression differences between the P. gingivalis-infected or A. actinomycetemcomitans -infected as compared to uninfected HIGK cells, at a significance level of P < 0.001. To test the predictive validity of the probe sets identified at this level of significance, a leav e-one-out cross-validati on (LOOCV) was performed with four different prediction models (linear discriminant, 1K NN, 3KNN and nearest centroid). This validation step addressed the ability of pr obe sets to distinguish between the different classes (i.e. infection states). Br iefly, this analysis determined if the classifier performed better than one would expect by chance alone. In the present study there were th ree classes; on average one would expect to correctly classify the ar rays by chance alone 33% of the time. Using the

PAGE 103

103 gene expression classifier, the ar rays were correctly classified 92% of the time. Thus, the gene expression differences significant at P < 0.001 can be used to dis tinguish between the strains and their miss-classification rate of 8% is much lower than the miss-classification rate of 67% expected by chance. Figure 4-1 shows the K-mean s clustering patterns of probe sets where the expression patterns were significantly different between the treatment classes. Several interesting clusters of genes are highlighted by blocks one through six in Figure 4-1. For instance, block 1 (182 probe sets) represents gene s that are down-regulated in A. actinomycetemcomitans -infected cells, but not modulated in P. gingivalis -infected or control uninfect ed cells. Conversely, block 5 (252 probe sets) represents ge nes that are up-regulated in A. actinomycetemcomitans -infected cells, but not modulated in P. gingivalis -infected or control uninfected cells. These two clusters of genes may be characteristic of cellular interactions sp ecifically associated with A. actinomycetemcomitans Similarly, P. gingivalis elicited a transcriptiona l response in HIGK cells that is specific to this organism (block 2, up -regulated; and block 3, dow n-regulated). Overall, transcriptional response appear ed to be diametrically opposed between the two organisms with only a small number of genes (41 probe sets) upor down-regulated by both species (blocks 3 and 6). These common genes may be involved in a general host cell response to infection that may be universal for oral Gram-negative organism s or possibly even for b acterial stimulation in general. However, possibly more importantly, th e data also suggest th at individual organisms may have evolved to modulate a finite number of pa thways that are characte ristics of the genus. Moreover, host cells appear to be able to di stinguish between infecti ng organisms and tailor transcriptional responses accordingly. Ontology Analysis In order to mine the array data for biological ly relevant information, an ontology analysis based on relatedness to known metabolic pathways was performed. The ontology analysis was

PAGE 104

104 performed at P <0.005 against the 354 different bi ological processes that have been identified thus far in the human GO syntax ontology database ( http://obo.sourceforge.net/ ). 16 gene ontology pathways, including molecular functions, cellular components, a nd biological processes were identified as representi ng the canonical response to bot h organisms. Moreover, 21 additional gene ontology pathwa ys were specifically found am ong the genes that responded to exposure to P. gingivalis Similarly, a specific response for A. actinomycetemcomitans -infected HIGK cells resulted in modula tion of 49 pathways. Those pathways with relevance to documented host-pathogen inte ractions are presented in Tables 4-2, 4-3, and 4-4. Of particular interest to host-pathogen in teractions in the oral cavity, one of the overrepresented pathways was the apoptosis pathway. P. gingivalis and A. actinomycetemcomitans have been shown to alter cytoki ne expression and modulate apoptosis in various cell types. The induction of apoptosis in immune cells of the oral cavity is thought to have a significant immunomodulatory (immunosuppressive) effect and contribute to the pathogenesis of A. actinomycetemcomitans in periodontal diseases (Lally et al., 1989ab; Ebersole et al., 1990; Spitznagel et al., 1995; Korostoff et al., 1998, Demuth et al., 2003; Kato et al., 2002). Furthermore, A. actinomycetemcomitans can also induce apoptosis in a leukotoxin-independent manner in oral epithelial cel ls, periodontal ligament cells, and gingival fibroblasts (Kato et al., 2000; Belibasakis et al., 2002; Teng and Hu, 2003). A recent re port suggests that the effector molecule associated with A. actinomycetemcomitans apoptosis in human gingival epithelial cells is a CagE homologue, which encodes a component of a putative type IV secretion system (Teng and Hu, 2003). P. gingivalis by contrast, suppresses apoptosis in primary cultures of gingival epithelial cells. P. gingivalis -induced suppression of apoptosis is correlated with activation of

PAGE 105

105 Bcl-2 at the transcriptional le vel and inhibition of cytochrome c release from the mitochondria (Nakhjiri, et al., 2001; Yilmaz et al., 2004). The ontology analysis presented in Tables 42, 4-3 and 4-4 revealed that a total of fiftyfive distinct apoptosis-assoc iated genes were modulated upon A. actinomycetemcomitans or P. gingivalis co-cultures of HIGK cells at a signifi cance of P <0.005. Of these, eight were modulated in both organisms, thirty-one were modulated only in A. actinomycetemcomitans and seventeen were differentially transcribed in P. gingivalis only. Interestingly, a large number of differentially regulated genes linked to the P53 network were found in both organisms. The P53 protein is a tumor suppressor gene that is positioned at a major node of a network that is involved in cell division and apoptosis. There are three majo r types of stress that modulate P53: aberrant growth signals, DNA damage and physico-chemi cal stress. The apoptosis induction by P53 can be mediated either by stimulation of Bax and Fa s antigen expression, or by repression of Bcl-2 expression. A summary of the ma jor apoptotic effector molecules impacted by the organisms is presented in Figure 4-2. A. actinomycetemcomitans activated the proapoptotic molecules BBC3, GADD45A, E2F1 and ATM and repressed cMYC. P. gingivalis activated cMYC and SGK both of which are anti-apoptotic and play a role in cel l survival and prolifera tion. cMYC can repress transcription of the proapoptotic GADD45A while SGK phosphorylat es and negatively regulates the transcription factor FOXO3A that can participate in apoptosis, in part through the GADD45a protein (Barsyte-Lovejoy et al., 2004; Brunet et al., 2001; Tran et al., 2002). SGKs are related to Akt, a serine/threoni ne kinase that plays a crucial role in promoting cell survival and has been shown to be activated by P. gingivalis in primary GEC (Yilmaz et al, 2004). Most of the activity of P. gingivalis however, revolved around the mitochondrial pathway, with upregulation of Bcl-2 and Bfl1. Bcl-2 inhibits release of cytochrome c from the mitochondria

PAGE 106

106 and can inhibit P53 (Cory and Adams, 2002). Bfl-1 can also inhibit cytoch rome c release and, in addition, suppresses Bid, which is an activator of the proapoptotic medi ators Bax and Bak (Cory and Adams 2002). Thus, the transcriptional prof iles are consistent with a pro-apoptotic phenotype of A. actinomycetemcomitans and an anti-apoptotic phenotype of P. gingivalis. Besides the P53 pathway, A. actinomycetemcomitans and P. gingivalis modulated the apoptosis pathway via a number of other effectors. With A. actinomycetemcomitans proapoptotic activity was restricted to the P 53 pathway and multiple genes were found to be regulated in a pattern that is consistent with the repr ession of the Fasand TNF mediated pathways. For instance, the Tumor necrosis fact or receptor superfamily member 6B precursor (TNFRSF6B or decoy receptor 3) was found to be up-regulated. This factor is a soluble receptor that binds to the Fas ligand and is thought to pl ay a regulatory role in suppression of FasLand LIGHT-mediated cell death (Yu et al., 1999). Similarly, CFLAR (CASP8 and FADD-like apoptosis regulator precursor, aka FLIP) was f ound to be up-regulated. This factor is a well known inhibitor of Fas and all othe r known human death receptors (Irmler et al., 1997). FOXO3a, which is a factor that can promote apoptosis through, among other pathways, FLIP down-regulation (Skurk et al., 2004), was transcrip tionally repressed by A. actinomycetemcomitans This is consistent with the observed up-regulatio n of FLIP. The programmed cell death protein 6 (PDCD6, alias AL G-2) was found to be repressed. This factor is thought to mediate calcium-re gulated signals along the death pa thway, and is required for Fasinduced cell death (Vito et al., 1996). Similarly, RIPK1 (TNFRSFinteracting serine-threonine kinase 1) was down-regulated. This protein inte racts with the death dom ain of FAS and TRADD and initiates apoptosis (Kreuz et al., 2004). In addition, TNFAIP3 th at is known to inhibit TNFinduced NF-kappa-Bdependent gene expression by interfering with a RIPor TRAF2mediated

PAGE 107

107 transactivation signal (Ferran et al., 1998), was induced by A. actinomycetemcomitans Also induced was NFKB1A that inhibits NF-kappa -B by complexing with and trapping it in the cytoplasm (Haskill et al., 1991). The induction of NFKB1A has also been observed with pathogenic Pseudomonas aeruginosa (Perez et al., 2004). IER3 (alias IEX-1 or immediate early response gene X-1), upregulated by A. actinomycetemcomitans, is controlled by multiple transcription factors among which P53, NF-kappaB/ rel, Sp1 and c-Myc play central roles. Overexpression of IER3 has been known to re nder some cells sensitive to apoptosis (Wu et al., 2001). Receptor interacting protei n-2 (RIP2) that mediates the recruitment of caspase death proteases (McCarthy et al., 1998) was over expressed in response to A. actinomycetemcomitans which is consistent with the activation of cell death. Collectively, th ese observations are consistent with the repression of the Fas and TNF signaling pathways. Hence, in the case of A. actinomycetemcomitans the transcriptional profile argues that the observed apoptotic phenotype is not Fasor TNF -mediated, but P53dependant. In common with A. actinomycetemcomitans P. gingivalis upregulated the anti-apoptotic molecules TNFAIP3 and CFLAR. Additionally, P. gingivalis downregulated CDC2L2, a serinetheronine kinase that may play multiple roles in apoptosis. Moreover, CDC2L2 is deleted/translocated in neuroblastomas with MY CN gene amplification, a subset of malignant melanomas (Gururajan et al., 1998.). Taken together, these data show that P. gingivalis can prevent the induction of apoptosis in HIGK cells at multiple levels. The long term consequences of this activity for normal physiologic function of epithelial cells remain to be established. Apoptosis in Gingival Epithelial Cells Modulated by A. actinomycetemcomitans or P. gingivalis The apoptotic responses of HI GK cells at the transcripti onal level revealed by array analyses were verified by a phenotypic a ssay for apoptosis. As shown in Figure 4-3, A.

PAGE 108

108 actinomycetemcomitans induced apoptosis in HIGK cells whereas P. gingivalis did not stimulate apoptotic activity. Furthermore, P. gingivalis cells were capable of inhibiting camptothecininduced apoptosis in HIGK cells. These results both corroborate other re ports in the literature with different epit helial cells (Kato et al., 2000; Teng and Hu, 2003; Nakhjiri, et al., 2001; Yilmaz et al., 2004) and show that the mRNA expre ssion levels correlate with phenotypic properties, at least with regard to some genes involved with apoptosis. Gene Expression in Response to Isogenic Mutants A bacterial mutant analysis was combined with host transcriptional profiling to assess the role of specific bacterial products or phenotypes on the epit helial cell gene expression programs. A. actinomycetemcomitans ORF859 (PEDANT database) was initially found to be induced in vivo in infected humans us ing IVIAT (Handfield et al., 2000). The product of this gene was further shown to be induced in pla que from infected patients (Handfield, 2000, 2002) and in various cell lines, including HIGK cells (R ichardson, 2004), and is a potential marker for active disease in LAP patients (Cao et al., 2004). A bioinformatic analys is of this gene product did not reveal a predicted func tion, although the gene product is hi ghly conserved across genera (Cao et al., 2004). As shown in Figure 4-4a, a supervised hierarchical clustering analysis showed that several genes were differentially regul ated by the ORF859 mutant strain JMS04 in comparison to the parental strain. The ontology an alysis presented in Table 4-5 further revealed that the most significant and numerous variations (P<0.001) were associated with intermediate metabolism functions, signal transduction and cy tokine activity. Interestingly, IL-27 was found to be induced by wild-type A. actinomycetemcomitans, but not by the JMS04 mutant (not shown) IL-27 is closely related to IL-12 in both sequence and structure (Artis et al., 2004), and has been shown to promote Th1 cellmediated immune responses (Hunter et al., 2004). Together, this suggests that the product of ORF859 may be related to the intrace llular adaptation and

PAGE 109

109 homeostasis of A. actinomycetemcomitans a process that does not impact large numbers of host cell pathways. However, upregulation of IL-27 will stimulate host cell mediated immunity and hence the ORF859 gene product may contribute to th e inflammatory properties of the organism. The P. gingivalis mutant (YPF1) tested is deficient in production of the major fimbrial protein, FimA, a multifunctional adhesin (Lam ont and Jenkinson, 1998). FimA mediates attachment of P. gingivalis to gingival epithelial cells through engaging an integrin receptor on the host cell surface (Yilmaz et al., 2002). Fimbrial-integrin intera ction results in assembly of integrin focal adhesion complexes, and the initiation of signaling pathways that induce remodeling of cytoskeletal architecture th at allows entry of the organism (Yilmaz et al., 2003). The YPF1 mutant is thus significantly impair ed in invasion and in cytoskeletal remodeling activity (Yilmaz et al., 2002, 2003). In contrast to the profiles obtained with the A. actinomycetemcomitans mutant, the P. gingivalis mutant strain YPF1 had a transcriptional pattern strikingly divergent from the parental st rain. As shown in Figure 4-4b and Table 4-6, and consistent with the phenotypic pr operties of the mutant, a large pr oportion of genes related to the cytoskeleton and to membrane a nd receptor activity were underrepresented in the transcriptional profile of YPF1-infected cells. For example, YPF1 failed to upregul ate actin binding LIM protein 1 which may play a general role in bridgi ng the actin-based cytoskeleton with an array of potential LIM protein-binding pa rtners, filamin B beta (act in binding protein 278) which connects cell membrane constituen ts to the actin cytoskeleton, and coronin 2A another actin binding protein (Roof et al., 1997.; Feng and Walsh, 2004; de Hostos, 1999). Additionally, YPF1 did not upregulate beta 3, 4 and 6 integri n, along with alpha V, 3 and 4 integrin, and CD47 an, integrin-associated signal transducer. YPF1 also demonstrated a significant inability to impinge on cell cycle and cell prol iferation pathways indicating th at a successful invasion event

PAGE 110

110 may be necessary for P. gingivalis to manipulate these pathwa ys. The apoptosis ontology pathway was not differentially influenced by YPF1, indicating that the fimbriae deficient mutant strain should capable of inhibiting apoptosis in HIGK cells to the same extent as the parental strain. This was confirmed by the phenotypic apoptosis assay (Figure 4-3) that showed YPF1 could antagonize chemically induc ed apoptosis to the same exte nt as the parental strain. Conclusions The transcriptional profiling pr esented herein begins to pr ovide insights into both the intricate biological phenomena occurring during host-pathogen in teractions and the distinct pathophysiology of A. actinomycetemcomitans and P. gingivalis A characteristic clinical outcome is associated with inf ection with either organism. A. actinomycetemcomitans -associated disease involves acute tissue destruction in the absence of overt inflammation, whereas P. gingivalis -associated disease is chroni c and involves inflammatory tissue destructio n. Moreover, the mechanism of intracellular i nvasion of both organisms is dis tinct. Consistent with these biological and clinical differences, the common core transcriptional response of epithelial cells to these organisms is very limited, and organism-speci fic responses predominate. Interestingly, this contrasts with disease models in other cell types. For example, infecti on of dendritic cells with Escherichia coli Candida albicans or the influenza virus resulted in a substantial shared core response along with a pathogen-specific pattern of gene expression (Huang et al., 2001). Thus oral epithelial cells, that encount er an array of microbes with va rying degrees of pathogenicity, may direct a measured response that is tailored to the pathogenic pot ential of the infecting organism. These responses can then influence disease progression. For example induction of apoptosis in epithelial cells by A. actinomycetemcomitans could contribute to immunologically silent tissue destruction. Inhibi tion of apoptosis by intracellular P. gingivalis in contrast, could contribute to bacterial pe rsistence and chronic, slowly progre ssing tissue destruction. Linkage of

PAGE 111

111 specific bacterial components to host pathways a nd networks provides additional insight into the pathogenic process. The loss of fimbriae form P. gingivalis retards adherence and invasion and the consequences of this reverberate throughout the transcriptome. Genes such as ORF859 in A. actinomycetemcomitans that appear to be involved in in tracellular homeostasis have a more subtle effect on the transcriptome. Such patter ns of gene expression di fferences in response to isogenic mutants may provide a means to evalua te the biological function of as yet undefined bacteria products. Methods Bacterial Strains A. actinomycetemcomitans strain VT1169 is a nalidixic-ac id and rifampin-resistant clone derived from the clinical strain SUNY465 (Mintz et al., 2000). JMS04 is an isogenic mutant for ORF859 constructed in VT1169, and obtained by in sertional inactivation with a spectinomycin cassette (Cao et al, 2004). A. actinomycetemcomitans strains were grown in Trypticase Soy Broth supplemented with 0.6% yeast extrac t (TSB-YE) in a humidified, 10% CO2 atmosphere, at 37C. P. gingivalis strains ATCC 33277 and its fimbriae deficient mutant YPF1 (Yilmaz et al., 2002), were cultured anaerobically for 24 h at 37C in trypticase soy broth supplemented with yeast extract (1 mg mL –1), hemin (5 g mL –1), and menadione (1 g mL –1). Eukaryotic Cell Lines HIGK cells (human HPV-immortal ized gingival keratinocyte) were originally generated by transfection of primary gingival epithe lial cells with E6/E7 from HPV (Oda et al., 1996;). HIGK cells are capable of normal kera tin synthesis and exhibit degree of differentiation similar to parent normal cells (Oda et al. 1996). HIGK cells were cultured under 5% CO2 in keratinocyte serum-free medium (K-SFM, Gibco/Invitroge n, Carlsbad, CA) suppl emented with: 0.05 mM calcium chloride, 200 mM L-glutamine (Gibco/Inv itrogen, Carlsbad, CA). Primary cultures of

PAGE 112

112 gingival epithelial cells (GEC ) were generated as described previously (Lamont et al., 1992; Oda and Watson, 1990). Briefly, healthy gingival tiss ue was collected from patients undergoing surgery for removal of impacted third mola rs and following Institutional Review Board Guidelines. Basal epithelial cells were separa ted and cultured in kera tinocyte growth medium (KGM; Cambrix, East Rutherford, N.J.), at 37C in 5% CO2. GEC were used at passage four. Microbial/Host Cell Co-Culture Bacteria were harvested and washed by centr ifugation, and resuspende d in antibiotic-free K-SFM media. HIGK cells (105) were washed three times with phosphate-buffered saline (PBS) and incubated with bacteria at a multi plicity of infecti on (MOI) of 100:1 for P. gingivalis and 3000:1 for A. actinomycetemcomitans After 2 hours at 370C in 5% CO2, the cells were washed three times with PBS and lysed with Trizol (Invitrogen, Carlsba d, CA) prior to RNA extraction. In parallel, total numbers of bact eria associated with the HIGK cells, both external and internal, after 2 h incubation and washing, were dete rmined by lysis and pl ate counts (Meyer et al., 1996). In addition, levels of A. actinomycetemcomitans and P. gingivalis invasion were measured by antibiotic protection assays as previously described (Lamont et al., 1995; Meyer et al., 1996). Co-cultures were carried out in quadruplicate. RNA Isolation, cRNA Synthesis and Chip Hybridization Total RNA was extracted, DNAse-treated, purifi ed and quantified according to standard methods (Qiagen and Affymetrix). cRNA synthesis was performed with 5-8 g of total cellular RNA, based on the Affymetrix protocol. Doubl e-stranded cDNA was synthesized according to a standardized protocol (SuperS cript double-stranded cDNA synt hesis kit; Invitrogen Corp., Carlsbad, Calif.). cRNA was transcribed in vitro incorporating biotinylat ed nucleotides by using a BioArray high-yield RNA transcri pt labeling kit (T7) (Enzo Life Sciences, Inc., Farmingdale, N.Y.), and hybridized onto the human HG U133A oligonucleotide arrays (Affymetrix). Each

PAGE 113

113 sample was studied in parallel, and the sa mples were not pooled. The microarrays were hybridized for 16 h at 45C, stained with phycoe rythrin-conjugated streptavidin and washed according to the Affymetrix protocol (EukGE-WS2v4) using an Affymetrix fluidics station, and scanned with an Affymetrix scanner. Microarray Data Analysis and Expression Filter Probe sets that were flagge d as absent on all arrays analyzed in this study by the Affymetrix GCOS software (with default settings) were removed from the datasets and were not included in the analyses. The signal intensity m easurements as detected reflect the degree of hybridization of synthesized cRNA to the probe sets on the microarray chip. These probe sets represent genes or DNA sequences within genes. Some genes are represented by more than one probe set on a given microarray, and hence probe sets are not uniquely correlated to genes. However, for ease of discussion, we use the te rms “probe sets” and “g enes” interchangeably (Freezor et al., 2003). Variation Filter, Normalization, and Cluster Analysis The signal intensities of the probe sets remain ing after applying the expression filter were ranked according to the coeffici ent of variation, and 50% of th e data set with the greatest coefficient of variation was then normalized to a mean of 0 and a standard deviation of 1. Kmeans clustering and hierarchical cluster analyses were performed with the variance-normalized data set and viewed with the algorithms in the software packages Cluster and TreeView developed by Eisen et al. (1998; Freezor et al., 2003). Supervised Learning, Discrimination Analysis, and Cross Validation The hybridization signal intensities of the ge nes passing the initial expression filter were analyzed (in part) with BRB Array Tools 3.01 (developed by Dr. Richar d Simon and Amy Peng Lam, National Cancer Institute, Bethesda, MD) to identify genes differentially expressed among

PAGE 114

114 the treatment classes: uninfected cells, cells infected with A. actinomycetemcomitans or mutant strain JMS04, or cel ls infected with P. gingivalis or mutant strain YPF1 (P < 0.001). The ability of gene identification to predict treatment class was assesse d by a leave-one-out cross-validation using each of four methods of class prediction: nearest-neighbor prediction, three-nearestneighbors prediction, linear disc riminant analysis, and neares t-centroid analysis (Feezor et al., 2003). Ontology Analysis The procedure delineated in Zheng et al. (i n press) was followed to perform the ontology analysis. Briefly, sets of genes differentially expressed under experiment al conditions were fed into the GoMiner software and P-values were co mputed for each GO term using the Fisher exact test (Zeeberg et al., 2003). The Gene Ontology (GO) database organizes genes into hierarchical categories based on biological process, molecu lar function and subcel lular location. GoMiner helps to identify all the GO-terms or categories that have been particularly enriched or depleted in the set of significantly differentiated genes (Zeeberg et al., 2003). Assessment of HIGK Cell Apoptosis To detect fragmentation of DNA in apoptotic epithelial cells, histone associated DNA fragments were examined in a cell death detect ion ELISA kit (Roche, Indianapolis, IN). HIGK cytoplasmic extracts were added to wells of ELISA plates coated with monoclonal antibodies against histones. The presence of histone-associ ated DNA fragments was then detected in a sandwich ELISA using anti-DNA peroxi dase-conjugated antibodies, with 2,2 -azino-di-[3ethylbenzthiazoline-sulfonate] substrate. Abso rbance was measured at 405 nm and background at 490 nm. As a positive control for apoptosis, HI GK cells were incubated with camptothecin (2 g mL 1) for 4 h.

PAGE 115

115 Table 4-1 Microbial-epithelial cell interacti on characteristics of human primary (GEC) and transformed (HIGK) gingival cells. Epithelial Cells Microorganism MOIaTotal Interactionb (CFU/cell) % of Interacting Bacteria that Invadec GEC A. actinomycetemcomitans 3000:1404<0.02 HIGK A. actinomycetemcomitans 3000:13525<0.05 GEC P. gingivalis 100:114825.00 HIGK P. gingivalis 100:118424.00 a Multiplicity of infection (bacteria:epithelial cell). b Total count of adhering and invading organisms after co-culture and cell lysis at 60 min. Data are re ported as mean value from two independent assays in triplicat e the standard deviation. c Calculated from intracellular CFU counts after antibio tic treatment. Table 4-2. Pathways common to P. gingivalis and A. actinomycetemcomitans -infected HIGK cells. GO ID Term P. gingivalisinfected (Total change)a P-Value A. actinomycetemcomitans infected (Total change) a P-Value 45449 regulation of transcription 490.0017118 0 16021 integral to membrane 390.002596 0 5215 transporter activity 170.003455 0.0036 12501 programmed cell death 25039 0.0031 17017 MAP kinase phosphatase activity 40.00016 0 Only pathways with documented relevance to host-pathogen interact ions are presented. aThe total change represents the to tal number of underand over-repres ented genes in a particular pathway.

PAGE 116

116 Table 4-3. Pathways specific to P. gingivalis -infected HIGK cells. GO ID Term Changea P-Value 7275 development 64 0.0031 9653 morphogenesis 47 0.0044 8283 cell proliferation 42 0.0009 9888 histogenesis 10 0.0004 8544 epidermal differentiation 7 0.0004 16265 death 25 0.0001 74 regulation of cell cycle 22 0.0009 5125 cytokine activity 12 0.0029 45073 regulation of chemokine biosynthesis 2 0.0034 8138 protein tyrosine/serine/thre onine phosphatase activity 4 0.0048 5149 interleukin-1 receptor binding 3 0.0010 Only pathways with documented relevance to host-pathogen interact ions are presented. aThe change represents the total number of underand ove r-represented genes in a particular pathway. Table 4-4. Pathways specific to A. actinomycetemcomitans -infected HIGK cells. GO ID Term Changea P-Value 8152 metabolism 330 0.0005 30528 transcription regulator activity 93 0 4888 transmembrane receptor activity 26 0.0040 9581 detection of external stimulus 6 0.0006 4930 G-protein coupled receptor activity 5 0 15268 alpha-type channel activity 5 0.0002 5216 ion channel activity 4 0.0002 5261 cation channel activity 3 0.0025 43066 negative regulation of apoptosis 14 0.0018 3773 heat shock protein activity 8 0.0015 16337 cell-cell adhesion 3 0.0036 3786 actin lateral binding 2 0.0036 Only pathways with documented relevance to host-pathogen interact ions are presented. aThe total change represents the to tal number of underand over-repres ented genes in a particular pathway.

PAGE 117

117 Table 4-5. Gene ontology analysis for HIGK cells infected with A. actinomycetemcomitans mutant strain JMS04 GO ID Under Over ChangeP-ValueTerm 6082 14 1 150.0001organic acid metabolism 9451 7 0 70RNA modification 5625 9 2 110.0001soluble fraction 5125 5 4 90.0014cytokine activity 6950 15 7 220.0022response to stress 3754 7 0 70.0025chaperone activity 7165 8 8 160.0090signal transduction 6983 2 0 20.0002response to ER-overload Only pathways with documented relevance to ho st-pathogen interactions are presented. Baseline for comparison is cells infect ed with parental strain. Table 4-6. Gene ontology analysis for HIGK cells infected with P. gingivalis mutant strain YPF1. GO ID Under Over Change P-Value Term 8283 112 26 138 0 cell proliferation 7049 94 16 110 0 cell cycle 166 87 29 116 0 nucleotide binding 5856 64 13 77 0 cytoskeleton 4872 31 21 52 0.0006 receptor activity 6811 6 10 16 0.0002 ion transport 16020 104 76 180 0 membrane 6974 24 6 30 0.0001 response to DNA damage stimulus 7186 15 10 25 0.0087 G-protein coupled receptor protein signaling pathway 19207 12 1 13 0.0014 kinase regulator activity Only pathways with documented relevance to ho st-pathogen interactions are presented. Baseline for comparison is cells infect ed with parental strain.

PAGE 118

118 Figure 4-1. Different patterns of gene expr ession of oral epithelial HIGK cells upon coculture with A. actinomycetemcomitans or P. gingivalis Hierarchical clustering of variance-normalized gene expression data from uninfected human HIGK cells and from cells in co-culture with either organism for 2 h prior to RNA isolation and purification. Expression and variation filt ers were applied to the data set prior to clustering. Probe sets giving hybridi zation signal intensity at or below background levels on all arrays tested were eliminated from further analysis. The resulting data set was culled by ranki ng on the coefficient of variation and eliminating the bottom half of the data se t to remove probe sets whose expression did not vary between the treatment regimens. The gene expression observations were variance normalized to a mean of 0 and a standard deviation of 1, and this normalized data set was subjected to hier archical cluster analysis with average linkage clustering of the nodes. The varia tion in gene expression for a given gene is expressed as distance from the mean obs ervation for that gene. Each expression data point represents the ratio of the fluorescence intensity of the cRNA from A. actinomycetemcomitans -infected (columns Aa R1-R4) or P. gingivalis -infected HIGK cells (columns Pg R1-R4) to the fl uorescence intensity of the cDNA from mock-infected HIGK cells (columns CTRL R1-R5). The scale adjacent to the dendrogram relates to Pears on’s correlation coefficient. Highlighted blocks are described in the text.

PAGE 119

119

PAGE 120

120 Figure 4-2. Differential modulati on of the P53-mediated apoptos is pathway by oral bacteria. HIGK cell response to (A) A. actinomycetemcomitans and (B) P. gingivalis Red terms are transcriptionally induced, while green terms are repressed. See text for description of individual molecules.

PAGE 121

121

PAGE 122

122 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8Cytoplasmic Histone-Associated DNA Fragments (OD 405nm)C CAM Aa Pg YPF1 Pg YPF1 + CAM* Figure 4-3. Apoptotic responses of HIGK cells to A. actinomycetemcomitans or P. gingivalis by ELISA of cytoplasmic histone-asso ciated DNA fragments. Control (C) represents HIGK under normal culture conditions. A. actinomycetemcomitans (Aa) was incubated with HIGK ce lls at a MOI of 3000:1 for 4h. P. gingivalis parental (Pg) or mutant (YPF1) strains were incubated with HIGK at a MOI of 100:1 for 20 h Camptothecin (CAM) was incubated with HIGK for 4 h. For inhibition of camptothecin-induced apopt osis, HIGK cells were incubated with P. gingivalis strains for 16 h followed by camptoth ecin for 4 h. Error bars represent standard deviation, n=3. denotes statistically diffe rent from control P <0.005

PAGE 123

123 Figure 4-4. Impact of specific bacterial co mponents upon HIGK cell transcriptomes. (A) A. actinomycetemcomitans or (B) P. gingival is bacterial strains with an isogenic mutation were allowed to interact with HIGK cells for 2 hours. The transcriptional profiles of HIGK cells encountering mutant strains are compared to those from HIGK cells interacting with w ild type parental strains or uninfected. This heat map and dendrogram were cons tructed from probe sets differentially expressed between experimental classe s at the significance level of P <0.001. Probe set signal intensities were variance-normalized, mean-centered across samples, and subjected to hierarchical cluster analysis. Average linkage clustering by uncentered correlation was performed for genes and samples. The degree of similarity between the transc riptional profiles of each sample is expressed by Pearson’s correlation coeffi cient distance metric, according to the adjacent scale. The expression state of each data point is represented as standard deviations from the mean expression leve l for that gene in all samples. Red indicates a relative increase, green in dicates a relative decrease, and black indicates no relative change of mRNA transcripts for a given gene. Labels. (A) CTRL R1-R6; uninfected HIGK cells Aa VT1169 R1-R4; Wild type A actinomycetemcomitans Aa JMS04 R1-R4; orf859strain of A. actinomycetemcomitans (B) CTRL R1-R5; uninfected HIGK cells, Pg 33277 R1-R4; Wild type P. gingivalis Pg YPF-1 R1-R4; FimAstrain of P.gingivalis

PAGE 124

124

PAGE 125

125 CHAPTER 5 IMPACT OF Aggregatibacter actinomycetemcomitans ADHERENCE ON GINGIVAL EPITHELIAL CELL TRANSCRIPTOME8 Introduction The ability of A. actinomycetemcomitans to colonize the oral cavity is an important component of the disease process. Biofilm fo rmation at disease sites can directly cause inflammation and tissue destruc tion, and allow pathogens to pe rsist during the course of periodontal disease. Aae is the first adhesin of A. actinomycetemcomitans implicated in the specific and direct adhesion of this ba cterium to oral epithelial cells (Rose et al., 2003; Fine et al., 2005). Although some phenotypic charact erization has been performed (Rose et al., 2003; Fine et al., 2005), the impact on the host transcriptom e by attachment of A. actinomycetemcomitans has not been documented. Furthermor e, little is known about this newly discovered A. actinomycetemcomitans virulence factor. For inst ance, it is not known which of the documented HIGK cell responses to A. actinomycetemcomitans are contact-dependent and involve signaling events attributable to direct Aae binding. Conversely, it is possible that other virulence factors are responsible for the observed host responses and the binding of A. actinomycetemcomitans merely serve to increase the loca l concentration of these secreted factors. For example, CDT may reach a higher local concentration at the hose cell surface when A. actinomycetemcomitans adhere, compared to the levels a ttained by CDT secretion into the culture media by planktonic bacteria. In accordance with Specific Aim 3, th e goal of this chapter is to assess the contribution of a specific bacteria l component, Aae, to hostpathogen interactions. 8 This work was supported by an NIH/NIDCR T32 Grant DE07200 (JJM) and RO1 DE16715 (MH). We thank Dr. Scott Grieshaber, University of Florida, Department of Oral Biology, for the use of his confocal fluorescent microscope.

PAGE 126

126 Transcriptional profiling of HIGK oral epithe lial cells combined with mutant analysis was performed to gain insights into th e contribution of acherence to the global A. actinomycetemcomitans -HIGK interaction. A differential host response was uncovered between the experimental conditions. Among several pathways impacted by Aae, Focal Adhesion, and the Regulation of Actin Cytoskeleton were highly modulated. Although Focal Adhesion was more significantly populated, the av ailability of convenient methods to assess the Regulation of Actin Cytoskeleton pathway justified further inve stigation of this host function. The initial stages of phenotypic co nfirmation have been initiated usi ng confocal fluorescence microscopy. In addition, adhesion to gingi val epithelial cells by the A. actinomycetemcomitans aaeisogenic mutant strain VT1565 was compared to that of the parental strain SUNY465. A standard assay to enumerate viable bacteria in teracting with host cells (adhesi on and invasion) demonstrated a 65% reduction in aaebacteria compared with the parent al strain. Conversely, microscopic visualization and enumeration of bacteria inte racting with HIGK cells failed to reveal a difference in the adherence of the aaestrain. Confirmation of this preliminary data is ongoing. If it holds, the data herein would sugge st a novel function of Aae pertaining to A. actinomycetemcomitans viability upon host cell interaction. This first attempt to investigate the Regulat ion of Actin Cytoskeleton upon infection with aaeor wild type bacteria failed to reveal definitive remodeling of F-actin architecture under the conditions tested. The transcript ional profile of HIGK cells sugge sts that other factors in the Regulation of Actin Cytoskeleton pathway may be i nvestigated for their role in the regulation of this pathway upon Aae-mediated adherence by A. actinomycetemcomitans In the absence of definitive and conclusive evidence, the confocal microscopy will be repe ated using additional

PAGE 127

127 markers for cytoskeletal rearrangement and Fo cal Adhesion amongst others and to assess actin architecture rearrangements for earlier time points not studied in the initial experiment. Herein, the assessment of the impact by a single bacterial component, Aae, was inconclusively attempted in co mpletion of Specific Aim 3. Upon confirmation of this initial study, a possibly novel function for Aae may be desc ribed. The transcriptional study described here will provide the basis for future e xperiments investigating the contribution of A. actinomycetemcomitans adhesion on the pathogenic pers onality of this microorganism. Background The colonization of the periodontium by A. actinomycetemcomitans is mediated by several adhesins. Some of the ear liest pili studied incl ude the Flp-1 pili. Fl p-1 pilin subunits are 6.5 kDa polypeptides which are secreted through a Tad secretion system and assembled into long and functional pili (Kachlany et al., 2001). These pili bundle into thick fibrils, and play a significant role in autoaggregation of A. actinomycetemcomitans in vitro as well as nonspecific adherence to glass, plastic and saliva-coated hyd roxyapatite, a surface that mimics the tooth. An animal model has also demonstrated the re quirement of Flp-1 pili for colonization and pathogenesis of A. actinomycetemcomitans based on mutant analysis studies of flp-1and tadAstrains (Schreiner et al., 2003). A. actinomycetemcomitans adhesion to the extracellular matrix of connective tissue prompted th e discovery of EmaA, an adhesi n that attaches to collagen independently of fibrils (Mintz, 2004; Ruiz et al., 2006). EmaA expression in A. actinomycetemcomitans results in the formation of antenn ae-like protrusions described as long stalks tipped by an ellips oidal head region (Ruiz et al., 2006). These non-fimbrial adhesins are believed to contribute to co lonization and persistence of A. actinomycetemcomitans to the extracellular matrix in vivo (Tang et al., 2007).

PAGE 128

128 The first adhesin implicated in the direct attachment of A. actinomycetemcomitans to oral epithelial cells was Aae, an autotransporter adhesi n with specificity to buccal epithelial cells of humans and old-world primates (Rose et al., 2003; Fine et al., 2005). Aae was discovered by computational biology and found to possess homology to the autotransporter adhesins Hap and Hia found in Haemophilus influenzae (Rose et al., 2003). The adhesive ability of Aae, first suggest ed by its homology to Hap and Hia, has been investigated in two separate studies. The or iginal characterization was performed with KB (HeLa CCL-17) cervical ep ithelial cells grown in a monolayer and infected by wild type or aaeA. actinomycetemcomitans at a MOI of 100:1. The authors di d not report the number of adherent bacteria per epithelial cell, but viability assays demonstrated a 70% re duction of binding by the aaestrain VT1565 compared to the wild type SUNY465 (smooth, serotype b). There was a similar percentage reduction in binding by a s econd mutant strain, VT1568, when compared to the parental strain ATCC29523 (sm ooth, serotype a). However, when the adhesion data of all four strains is normalized to a common denominat or, SUNY465 binds epithelia l cells in three to four-fold greater numbers than strain ATCC29523. Futhermore, binding of A. actinomycetemcomitans to epithelial cells was not comp letely abolished upon mutation of aae This suggests an additional adhesin, and does not preclude the differentia l expression of this adhesin and aae among different strains. Additionall y, lactoferrin from human milk, which cleaves Hap from H. influenzae is able to cleave Aae from the ATCC 29523 strain but not from SUNY465. Taken together, these observations s uggest the expression and substrate variability of this adhesin from strain to strain and fu rther exemplifies the diversity of interaction characteristics displayed by di fferent clinical strains of A. actinomycetemcomitans. Further

PAGE 129

129 studies are necessary to determine the influe nce of post translational modifications, length polymorphism, or copy number upon th e binding characteristics of different bacterial strains. A second study further characterized Aae-medi ated adhesion to host cells using a rough, serotype d strain of A. actinomycetemcomitans transformed with a mutated aae gene originally isolated from the rough, serotype f strain CU1000 (Fine et al., 2005). Buccal epithelial cells (BECs) from humans, old world primates, new world primates, and non-primate mammals were exposed to A. actinomycetemcomitans at MOI ranging from 1000:1 to 10,000:1 for 90 minutes while suspended in liquid and under constant rotation. It is unclear if the BECs remained viable throughout the isolation and testi ng procedure, with implications for the host response to adherence, and subsequent pathoge n-directed endocytosis events. In this assay, 210 +/14 wild type bacteria bound per BEC co mpared to14 +/4 for the aaestrain. This is approximately a 93% reduction in binding efficiency of human an d old world primate buccal epithelial cells. No difference was observed between the binding of mutant and wild-type A. actinomycetemcomitans to rat and cow BECs. No bindi ng was observed by wild type or aaestrains incubated with BECs from eleven other mammalian sources, in cluding mouse, pig, and new world primates. Although A. actinomycetemcomitans have been isolated from BECs in vivo the relevance of BECs to periodontal disease beyond a reservoir for bacteria is unclear. The binding efficiency of wild type or aaebacteria to Ginvival epithelial cells (GECs) was not reported. Escherichia coli transformed to express Aae on their outer membranes were also able to bind BECs in high numbers, with 10 to 100 tr ansformed bacteria adhering per BEC. Interestingly, the only other epitheli al cell type bound by Aae-transformed E. coli were primary gingival epithelial cells (GEC), and at the lower ra tio of 1 to 3 bacteria per GEC. These GECs were also isolated using a scra ping method, and the cells may not have been viable. Less than

PAGE 130

130 one Aae-transformed E.coli bound per 100 cells during assays te sting human alveolar, bronchial, palatal, tongue and cervical epithelial cells These studies conf irm the tropism of A. actinomycetemcomitans to BECs and to a lesser extent, GECs (Fine et al., 2005). Although the extent of Aae-sp ecific binding appears to be lower in GECs compared to BECs, Aae-mediated binding to GECs has been demonstrated with aae -transformed E. coli (Fine et al., 2005). Furthermore, gingival epithelial cells are amongst the first cell types to interact with A. actinomycetemcomitans in the oral cavity, and the outcome of this interaction is critical in periodontal disease. The in itial study reporting the discover y of Aae by Rose and colleagues used cervical epithelial cells to demons trate a 70% reduction in the binding of aaestrains compared to the parental strains. However, considering the experime ntal model and methods, neither study has definitively addr essed the Aae-mediated binding of A. actinomycetemcomitans to GECs. Furthermore, beyond the phenotypic ch aracterization of the Aae binding function for A. actinomycetemcomitans little is known about the e ffects of Aae upon host-pathogen interactions. The specificity of the Aae-mediated binding by A. actinomycetemcomitans to BECs and GECs prompted this study to investigate the role of Aae during adhesion to gingival epithelial cells. Potential alternative functions for Aae besi des adherence remain possible. In other bacteria, adhesion triggers a speci fic host response, such as an upt ake signal, which can lead to pathogen-directed endocytosis. This interaction has been observed for Shigella spp. and Salmonella typhimurium (Steele-Mortimer et al., 2000), and remains a likely outcome for A. actinomycetemcomitans Furthermore, the role of specifi c adhesion is not fully understood as it relates to the delivery of bacterial effectors that are presumed to rely on contact with host cells prior to their injection through a secr etion system, such as CagE (Teng et al., 2003). Conversely,

PAGE 131

131 it is not understood if secr eted factors, such as CDT, contin uously stimulate ep ithelial cells and override any specific response elicited by adhesi on. As an initial st ep to address these considerations, the effect of Aae-mediated bindi ng to human immortalized gingival keratinocytes (HIGKs) on the host transcriptome is thus examined. Materials and Methods Bacteria and Cell Lines HIGK cells originally generated by transfection of primary gingival epithelial cells with E6/E7 from HPV (Oda et al., 1996) were grown in Keratinoc yte Serum-Free Medium (KSFM) as a monolayer to 95% confluen ce in an atmosphere of 5% CO2 at 37oC (Oda et al., 1996; Cao et al., 2004). Cell culture media contained L-glutamine and was supplemented with 50 U penicillin-streptomycin per mL (Gibco) and calcium chloride to 0.05 mM final concentration. Bacterial strains were obtained from Dr Fives-Taylor, University of Vermont. A. actinomycetemcomitans SUNY465 was maintained on 1.5% ag ar plates consisting of 3% trypticase soy broth and 0.6% y east extract (TSBYE). VT15 65 was grown on TSBYE plates supplemented with 50g/mL kanamycin (Rose et al., 2003). Both strains were inoculated into liquid culture and grown at 37o C and 10% CO2 to mid-logarithmic phase, and prepared for host cell co-culture according to standard methods (Richardson et al., 2005). Briefly, epithelial cells were washed three times with 1X Dulbecco’s Phosphate-Buffered Saline (dPBS) (Cambrex, Walkersville, MD) to remove resi dual antibiotics, cellular waste products, and secreted toxins. In biological replicates of f our per condition, epithelial cells were sham-infected with KSFM alone or co-cultured with A. actinomycetemcomitans resuspended in culture media, resulting in a multiplicity of infection (MOI) of 2500:1, which is the same order of magnitude compared to previous experiments (1000:1) and yielded consistent total interac tion ratios for the strains being studied. This slightly increased MOI allo wed the detection of a baseline level for aaestrain

PAGE 132

132 interaction per host cell and subse quent calculations for th e adhesion defect of the mutant strain. Previous studies in our lab (Handfield, Mans et al., 2005) determined that 1000:1 was an MOI ensuring a that every host cell was likely to enco unter at least a single bacterium—typically 25 to 30—, resulting in a homogeneous population of inf ected host cells, and thus a representative mRNA sample of the infected state. Two hours co-culture was the time point previously determined to display a phenotype that can be characterized in terms of host cell monolayer integrity, and chosen in order to maintain consis tency with previous work in our lab (Handfield et al., 2005). After two hours co-culture, epithelial ce lls were lysed with Tr izol (Invitrogen Life Technologies, Carlsbad, CA) and RNA was prepar ed for GeneChip hybridization as recently described (Handfield et al., 2005). Total RNA was extracted fr om Trizol-lysed cells, treated with DNase I, purified, and qua ntified according to standard methods (Qiagen, Valencia, CA; and Affymetrix, Santa Clara, CA). Comple mentary DNA (cDNA) synt hesis was performed according to the Affymetrix protocol (Supe rScript double-stranded cDNA synthesis kit; Invitrogen,) with 5-8 g of total cellular RNA used as a template to amplify mRNA species for detection. Double-stranded c DNA was purified, and used as a template for labeled complementary RNA (cRNA) synthesis. In vitro transcription was perf ormed using a BioArray high-yield RNA transcript labeling kit (T7) (Enzo Life Science, Farmingdale, NY), to incorporate biotinylated nucleotides. cRNA wa s subsequently fragment ed and hybridized onto Human Genome (HG) U133 Plus 2.0 GeneChip DNA microarrays (Affymetrix) with appropriate controls. Each sample was studied in parallel, and the samples were not pooled. The microarrays were hybridized for 16 h at 45C, stained with phycoerythrin-conjuga ted streptavidin and washed according to the Affymetrix protocol (EukGE-WS2v4) using an Affymetrix fluidics station, and scanned with an A ffymetrix GeneChip 3000 scanner.

PAGE 133

133 Microarray Analysis Assessment of the host cellular responses to bacterial challenge was accomplished by transcriptional profiling using Human Genome (HG) U133 Plus 2.0 GeneChip DNA microarrays (Affymetrix). This generation of GeneChips al lows the experimental assessment of over 47,000 human transcripts. Infected and uninfected HIGK cells were tested in four independent replicates. Subsequent array analysis was performed as recently presented (Handfield et al., 2005). In brief, expression filters were applie d to remove Affymetrix control oligos and probesets whose signal was undetected across all sa mples. The signal intensity values of the resulting dataset were variance-normalized, mean -centered, and ranked by their coefficients of variation. Normalization was performed in order to give equal weight to all probesets in the analysis, regardless of the ra w signal intensity order of magni tude. To reduce the confounding effect of background signal variation on the analys is, the half of the dataset demonstrating the most variation across samples was used to perf orm unsupervised hierarch ical cluster analysis using Cluster software (Eisen et al., 1998). The resulting heat-map and Cluster dendrograms were visualized with Treeview (Eisen et al., 1998) to reveal the extent of characteristic host-cell responses to each infection state, defined as identical treatments clustering together. Following initial assessment of the host cell response to each condition, supervised analysis was performed to investigate differe nces in gene regulation among experimental conditions. For this analysis, th e raw signal intensities were l og-transformed for all probesets passing initial expression filters and correlate d using BRB Array Tools (Simon and Peng-Lam). In each supervised analysis, biological replic ates were grouped into classes according to infection state during co-culture experiments. Several methods of class prediction were utilized (compound covariate predictor, nearest neighbor predictor, and support vector machine predictor) in order to generate lists of cl ass predictors whose expression state changes

PAGE 134

134 significantly between classes. Leave-one-out cross-validation (LOOCV) was performed to test the accuracy of each class predicto r and compared to the probability of a correct class prediction by chance alone assuming normal distribution, base d on the P-value and total number of genes analyzed. In order to visualize the differentia lly-regulated genes, Microsoft Access database queries were used to match the subset of signi ficantly regulated genes with their associated variance-normalized, mean-centered signal values calculated previously. Cluster and Treeview software were used to visualize the co rrelations among genes and samples (Eisen et al., 1998). Functional Categorization by Gene Onto logy (GO) and Bioinformatics Analyses. GO tables and KEGG pathways were popul ated using Pathway Express (Khatri et al., 2005) available at http://vortex.cs.wayne.edu/projects.htm This software makes pairwise comparisons of experimental conditions and populates KEGG pathways based on fold induction or decrease for each impacted gene. Therefore, the supervised analysis was repeated for twoclass comparison of the HIGK response to infecti on at P <0.05. This threshold was chosen to increase the number of genes populated for a given pathway in order to allow a phenotypic prediction based on the gene regulation observed. Three analyses were performed: (1) HIGK cocultured with wild type SUNY465 compared to baseline uninfected controls, (2) HIGK cocultured with aaestrain VT1565 compared to baseline uninfected controls, and (3) HIGK cocultured with aaeVT1565 compared to wild type SUNY 465-infected baseline. An arbitrary minimum threshold filter of 2-fold expression change was also applied. This additional condition increased the likelihood that expression level differe nces for a given gene were detectable using downstream phenotypic confir matory methods. Gene annotations were obtained through the Genecards on line database available at http://www.genecards.org/index.shtml

PAGE 135

135 Confocal Fluorescent Microscopy HIGK cells were resuspended in KSFM without antibiotic/an timycotic and the concentration of cells was determined with a pa rticle counter (Beckman Coulter, Inc., Miami, FL). The Lab Tek II four chamber slide system was used for microscopy experiments (Nalge Nunc International, Rochester, NY). A total of 1.25x105 HIGK cells were seeded per chamber, resulting in a 70-90% confluent mo nolayer of cells for the duration of the infection experiment. Starting with a 95% confluent m onolayer would have led to cell overgrowth for later timepoints, such as 24 H post inoculation. A. actinomycetemcomitans SUNY465 and VT1565 ( aae-) overnight cultures were diluted to an optical de nsity (OD) of approximately 0.2 A at 495 nm. CellTracker Green BODIPY live dye (Invitrogen Corporation, Ca rlsbad, California) was added at 10 M final concentration to VT1565 ( aae-) and SUNY465. Bacteria we re incubated for 3 to 4 hours and until cultures reached mid-logarithmi c growth stage (OD 495 nm, 0.6 A), in order for the bacteria to incorporate the fluorescent dye. Optimization experiments (data not shown) revealed no adverse effects on ba cterial growth or viability in vitro due to live dye incorporation. Bacteria were collected by centrifugation at 3000 rcf for 10 minutes, and resuspended in 1x dPBS to wash away residual dye that coul d be incorporated by HI GK host cells during the infection. Centrifugation was repeated, and bact eria were resuspended in antibiotic-free KSFM at a concentration to result in an MOI of 2500:1 upon infection. Pr ior to co-culture, host cells were washed twice with 1x dPBS to remove wast es and antibiotics, and then inoculated with fluorescent dye-label ed SUNY465, VT1565 ( aae-), or mock-infected with KSFM. One mL of inoculate was dispensed per slide chamber. D ilution plating was performed in parallel to confirm the inoculate MOI for each strain. Post -inoculation time points investigated were 20 minutes, 2 hours, 6 hours, and 24 hours.

PAGE 136

136 At the 20 minute and 2 hour timepoints, HI GK cells were washed three times with 1x dPBS to remove non-tightly adherent bacteria and fixed with 4% para formaldehyde in 1x dPBS for 30 minutes at room temperature. For 6 hour and 24 hour timepoi nts, HIGK cells were washed twice with 1x dPBS at 2 hours, and cells were incubated in KSFM without antibiotics for the remaining time. At 6 and 24 hours post inoc ulation, HIGK cells were washed once with 1x dPBS and fixed with 4% paraformaldehyde in 1x dPBS. Parallel adhesion assays were conducted at 2 hours post inoculation to maintain consistency with microarray experiments and assess the initial total interaction levels of the bacteria. HIGK cells were processed and stained accord ing to the commercial protocol for staining with Texas Red phalloidin (Invitrogen Corporati on, Carlsbad, California). Briefly, cells were washed with 1x dPBS once, then made perm eable with 0.1% Triton x-100 in 1x dPBS, PBSwashed again, and blocked prior to staining with 1% BSA in 1x dPBS. Cells were stained with 2 Units of Texas Red phalloidin in 1% BSA-dPBS for 30 minutes per chamber. Cells were washed again, the slide chamber walls were removed, and slides were air dried and hard mounted with Vectashield HardSet antifade and mounting medium (Vector Laboratories, Burlingame, CA). Slides were dried ove rnight at RT in the dark, and sealed. The Factin cytoskeleton of HIGK cells wa s visualized with a spinning disk confocal microscope under oil immersion at x600 magni fication. VoxCell software (VisiTech International, Sunderland, United Kingdom) and live camera images taken through an x1.5 conversion lens were used to captu re a series of fluorescent optical x-y sections to create digitally reconstructed images ( z -stacks) of HIGK cells interacting with A. actinomycetemcomitans A collection of slices were captured for the entire height ( z -dimenstion) of the tallest HIGK cell in the visual field at 0.2 micron step s. Bacteria labeled with Cell Tracker Green were excited with

PAGE 137

137 a 491 nm laser and the emission spectrum was captured using a 525 nm narrow-pass filter. HIGK cells stained with Texas-Re d phalloidin (F-actin directed) at 561 nm and the emission spectrum was captured with a 595 nm narrow pass filter. Green and Red channel images were merged using Imaris imaging software (Bitplane Inc, Saint Paul, MN), and three-dimensional reconstructions of the cells and bacteria were an alyzed to measure total levels of interaction (adhesion and invasion) by A. actinomycetemcomitans and to visualize HIGK cell actin cytoskeletal rearrangements. Quantification of A. actinomycetemcomitans Adherence Five to ten optical fields were select ed per condition for image capture based on representative actin morphology of HIGK cells visualized by Texa s Red phalloidin. In general, selected cells were not rounded and were judged to be tightly adhering to th e slide surface. More than 50 total fields were analyzed to examine HIGK-bacterial interactions for both strains during four timepoints. Bacteria were not visualized until after the fields had been captured to avoid biasing the selected fields based on the numbers of bacteria associated per cell. The numbers of bacteria per cell were counted fo r whole cells only. Ce lls that were truncated from the visual field were not considered and their associated ba cteria were not counted. The mean and standard deviations of bacteria observed in three di mensional Z-stacks was calculated for SUNY465 and VT1565. A two-tailed, unpaired st udent’s t-test was performe d with Microsoft Excel to determine if the mean numbers of aaeor wild type bacteria per cell were significantly different. Results A. actinomycetemcomitans Viable Counts Viable counts of the inoculums and total inte racting bacteria were measured in parallel with spectrophotometry readings to corroborate the predicted MOI and subsequent interaction ratios for a given experiment. The mean MO I for SUNY465 calculate d by viable counts was

PAGE 138

138 (2.8+/-0.4) x103. The mean MOI of VT1565 was (2.6+/-0.2) x103. The total interaction counts were used to confirm the apparent adhesion deficiency of the aaestrain VT1565 with HIGK epithelial cells. The mean tota l interaction count of SUNY465 was 22+/-7. The mean total interaction count for VT1565 was 8+/-3. The probability these two means are equal as determined by a two tailed, unpaired student’s t-test was P=0.07. The adhesion deficiency of this strain was approximately 65% reduction from that of the parental strain by dilution plating. This is consistent with the 70% reduc tion observed during the initial ch aracterization of this strain interacting with cervical epitheli al cells using a slightly high er MOI and comparable methods (Rose et al., 2003). Transcriptional Profiling Three-class supervised analysis was perf ormed for the 30,395 probesets which passed the initial expression filt ers. BRB Array Tools revealed 4138 and 161 probe sets that were differentially expressed among the treatment c onditions at the P <0.05 and P <0.001 levels of significance, respectively. With normal dist ribution, one would expect 1520 (P<0.05) and 30 (P<0.001) modulated genes by chance alone. The results of Leave-One-Out Cross-Validation (LOOCV) ranged from 67% nearest centroid correct to 75% nearest neighbor and diagonal linear discriminant correct prediction for the P <0.001 da taset. For the P <0.05 dataset, the nearest neighbor prediction and nearest centroid were also 75% and 67 % correct. The diagonal linear discriminant analysis was correct 67% of the time. This is more accurate than 33% correct prediction by chance alone for LOOCV for three cl asses and normal distribu tion. The statistical significance of the cross-validated misclassification rate base d on 2000 random permutations of the dataset was P=0.02 for all prediction models. Treeview visualization of significant genes by supervised analysis were organized with Clus ter and revealed several distinct nodes which differentiate the infection c onditions. The P <0.05 significant gene expression patterns are

PAGE 139

139 shown in Figure 5-1. HIGK cells in co-culture with VT1565 ( aae-) clustered more closely with the uninfected control HIGK cells than wild type SUNY465-infected cells during all three-class analyses as determined by Pears on’s correlation coefficient. Two-class supervised analysis was performed to generate genelists for ontology analysis. At the P <0.05 level of significance, 2762 probese ts were regulated in SUNY465-infected cells compared to uninfected controls. Also compar ed to uninfected controls, HIGK cells infected with VT1565 ( aae-) differentially regulated 3494 probese ts. The comparison of HIGK cells infected with aaestrain of A. actinomycetemcomitans regulated 5147 probesets when compared to HIGK cells co-cultured with the parental strain SUNY465. LOOCV for the two-class supervised analyses ranged from 75% to 100% accurate, which is better than 50% correct classification expected by chance al one for LOOCV with two classes. Ontology Analysis The Pathway Express algorithm displayed many pathways of HIGK cells that were impacted upon interaction w ith SUNY465 and VT1565. Changi ng the applied level of stringency between P<0.05, P<0.01, or P<0.001 sub tly changed the ranking order of implicated pathways. Yet, the overall composition of impacted pathway lists was identical regardless of the level of significance (dat a not shown). P <0.05 was chosen sin ce that level of st ringency allowed implicated pathways to be populated to a greate r extent, which aided pheno typic predictions. At this threshold and without additional filters, the two class comparision of VT1565 to SUNY465 infected cells yielded a geneli st that populated over 50% of the Rearrangement of Actin Cytoskeleton pathway. To clarify the pattern of regulation for this pathway under the three experimental conditions, a second fi lter was applied to reduce the number of genes considered in the phenotypic prediction (Figures 5-2 to 5-4). Genes whose expression level changed by at least two-fold and were also significantly modulat ed at P<0.05 were considered for additional

PAGE 140

140 analysis. This filter reduced the number of total genes to 15 in the Regulation of Actin Cytoskeleton pathway (Table 5-1). This fold change filter removed statistically significant genes with expression levels that were thought to fall below the sensitivity of downstream confirmatory methods. As shown in Table 5-1, multiple HIGK pa thways were impacted by mutation in aae The Regulation of Actin Cytoskeleton was consistent ly a top 5-impacted pathway for supervised analysis performed at P <0.001, P <0.01, and P < 0.05. When the entire P<0.05 genelists were used for ontology analysis, the pathway P-values for Regulation of Actin Cytoskeleton were 4.926 x10-3, 4.660 x10-2, and 1.391 x10-4 for the wild type /Ctrl, Aae-/Ctrl, and Aae-/ wild type comparisons, respectively. Increasing the stringe ncy with the additional filter applied to the P<0.05 dataset of at least two-fold expression change artificially increased the P-values 10 to 100 fold. This occurs because the extra filter lowe rs the total number of im pacted genes populating each pathway and used in the pathway significance tests. Therefore, Regulation of Actin Cytoskeleton is one of the most significantl y impacted pathways under the experimental conditions tested. This is consistent with ear ly reports of a putati ve role for host actin rearrangement in the invasion of cultur ed HeLa cervical epithelial cells by A. actinomycetemcomitans SUNY465 (Fives-Taylor et al., 1995). Therefore, this pathway was chosen for further characterization, taking in to account the availability of phenotypical confirmatory assays. Figure 5-2 shows the baseline impact of A. actinomycetemcomitans adhesion upon the Regulation of Actin Cytoskeleton pathway in HIGK cells. Of the genes passing filtering thresholds, ten genes were up-regulated and none were down-regulated in this pathway upon epithelial cellA. actinomycetemcomitans SUNY465 interaction. Func tional annotations were

PAGE 141

141 obtained from the GeneCards database, accessible at http://www.genecards.org/index.shtml The four genes PIR121, Nap125, ERM, and MyosinII we re impacted exclusively in the SUNY465infected HIGK cells. APC is involved with cell adhesion and active cell mi gration. This product also participates in signaling ev ents. NWASP binds se veral proteins, including the regulatory GTPase, Rac. This gene product binds the Arp2/3 complex and is an effect or protein that links Arp2/3 to the Rho-type GTPases. It has been reported that NWASP is important for efficient actin polymerization and regulates the structure and dynamics of the actin cytoskeleton. PIR121 (CYFIP1) also interacts with active GTP-bound Rac. PIR121 binds to F-actin and is involved in the formation of membrane ruffles and lamelli podia protrusions. Nap125 (NCKAP1) forms a lamellipodial complex with PIR121. Nap125 is also involved in Rac-dependent actin remodeling. ITG (ITGA11) is a cell surface adhe sion receptor that mediates cell adhesion to extra cellular matrix and to other cells. GF (EGF) stimulates the growth of epithelial tissue. Its receptor, RTK, was also up-regulated. RT K binding to GF leads to dimerization and internalization of the GF-recept or complex. Potentially, GF-RTK mimicry by bacteria could be explored as a mechanism for pat hogen-directed endocytosis. ERK is involved with the initiation and regulation of mitosis and meio sis. Bacterial stimulation c ould conceivably affect the host cell growth rate and explain the up-regulation of this gene. CDT is a known factor that affects the cell cycle, and this may be another avenue for A. actinomycetemcomitans to modulate host cells. ERM is believed to participate in connect ions of major cytoskeletal structures to the plasma membrane. ERM is a filopodial protei n involved in cell recogn ition and morphological changes. MyosinII plays a role in cytokinesis and cell shape. This contractile motor protein moves towards the plus ends of actin filaments. Taken together, the up-regulation of the above

PAGE 142

142 factors suggests an increase in lamellipodi a and filopodia formation, and the induction of membrane ruffling upon infection with A. actinomycetemcomitans Figure 5-3 illustrates the Regulation of Actin Cytoskeleton pathway impacted by A. actinomycetemcomitans strain SUNY1565 ( aae-) compared to uninfected HIGK cells. Of the genes passing the selection crit eria, twelve total genes were up-regulated and none were downregulated. Six of the genes up-regulated upon SUNY465 interaction were consistently upregulated in HIGK cells encountering SUNY1565 ( aae-) compared to the baseline uninfected transcript levels. These six genes are APC, NWASP, ITG, GF, RTK and ERK, and their functions are listed in the preceding paragraph. Cdc42 is a GTPase which cycles between act ive and inactive states. When active, Cdc42 binds other effector proteins to regulate various cell res ponses, such as epithelial cell polarization. Another function of Cdc42 is to cause the formation of filopodia. As previously mentioned, Rac is also a GTPase which regula tes many proteins in the actin rearrangement pathway. PI3K is a regulatory kinase. RhoG EF regulates RhoA GTPase. MLCP and MLCK are a phosphatase and kinase re spectively that oppositely regu late MLC. The downstream function is to mediate binding to myosin and re gulate myosin phosphatase activity. Since both of these opposing regulators are induced, more info rmation is required to predict the downstream outcome of the transcriptional regulation of thes e genes. As was the case with wild type A. actinomycetemtomitans the ontology for HIGK cells encountering the aaestrain SUNY1565 also predicts for the increas ed activity of filopodia, lamellipodia, and membrane ruffling compared to the uninfected cells. Thus, these may be adhesion-dependent responses (since no difference was observed in adhesion) but not related to Aae signaling.

PAGE 143

143 Figure 5-4 illustrates the tr anscript levels in SUNY1565 ( aae-)-infected HIGK cells compared to the wild type-infected baseline leve l. This ontology analysis specifically addressed the impact of Aae on the cytoskeleton. This an alysis is more sensitive than the separate comparisons of mutant and wild t ype infected cells to the baseline uninfected state because it is a direct comparison of the two infected conditions. However, the uninfected state is a necessary point of reference to place gene -specific differences in context. To accomplish this for genes regulated in all conditions, it is possible to ex trapolate the pairwise data into a three-class comparison. For instance, ITG is up-regulat ed upon interaction with both SUNY465 and SUNY1565. The Aae-/ wild type comparison reveals that while ITG is up-regulated upon wild type infection, it is up-r egulated to an even greater exte nt in SUNY1565-infected HIGK cells. Eleven genes were regulated in mutant-infected ce lls compared to wild-typ e infected. Of these eleven, five genes were consis tently regulated when infected HIGKs were compared to uninfected cells. The remaining six genes that we re not differentially re gulated when infected HIGK cells were compared to uninfecte d, but were signifi cantly different in aaestrain infected HIGKs compared to wild type, are PIX, PAK, Ra s, FN1, Mena and mDia. Only Mena and mDia were down-regulated. All other ge nes were up-regulated upon SUNY1565 ( aae-) infection of HIGK cells compared to infection with SUNY46 5. The gene product PIX interacts with PAK kinases and acts as a RAC1 guani ne nucleotide exchange factor. As a result, PIX can induce membrane ruffling. PAK may act to stabilize ac tin filaments through the inhibition of cofilin activity. PAK plays a role in th e reorganization of the actin cyto skeleton and in the formation of filopodia. Ras has GTPase activity and may tran sduce growth inhibitory signals. FN1 binds various cell surfaces including collagen and fi brin. FN1 is involved in cell adhesion, cell motility and maintenance of cell shape. Mena is a scaffold protein that stabilizes microtubules

PAGE 144

144 and promotes cell migration. Through binding the barbed end of the actin filament, Mena slows down actin polymerization and depolymerization. Th e factor mDia is required for actin cables and stress fibers. This product also slows the ra te of actin polymerization and depolymerization by binding to the barbed end of ac tin filaments. Activity by mDia also stabilizes microtubules. This is consistent with the finding that invasion by A. actinomycetemcomitans apprears to involve both actin a nd microtubules (Meyer et al., 1996, Meyer et al., 1999). The predicted down-regulation of in the mutant of actin-stabilizing genes Mena and mDia would result in the net increase of rearrangement activity. This is al so in agreement with th e up-regulation of genes that induce lamellipodia and filopodia formation. These genes were not impacted in SUNY465infected cells compared to uninfected HIGK cells. Thus, the comparison of mutant-infected cells to wild type infected cells revealed that A ae may impact the Actin Cytoskeleton Rearrangement pathway through Mena and mDia, which was not re vealed by comparing in fected cells to the uninfected state. Taken together, the ontology analysis predicts that in the absence of Aae, more actin rearrangement will occur in HIGK cells, a lthough both strains are capable of inducing actin rearrangments above uninfected levels. The pr edicted phenotypes would be increased numbers and/or size of filopodia, lamellipodia, and memb rane ruffling in HIGK interacting with the aaestrain. Preliminary Phenotypic Confirmation Adhesion In contrast to the data obtained from en umeration of viable co lony forming units (CFU) interacting per host cell, confocal fluorescen ce microscopy performed on a limited number of observations and a single experiment failed to reveal a difference between the ability of the A. actinomycetemcomitans aaeto adhere to HIGK cells compared to the parental strain (Figure 55). More than 50 optical z -stacks consisting of merged x-y fields were analyzed. A total of 116

PAGE 145

145 HIGK cells were observed independently, of which 56 were exposed to SUNY1565 ( aae-) and 60 were infected with SUNY465. The z-stacks a llowed the entire cell to be visualized for interaction with A. actinomycetemcomitans Thus, counts were not based on single z -sections individually, but rather the en tire volume of the HIGK cell. The first analysis examined the binding and invasion ability of the aaestrain compared to the wild type strain for each timepoint sepa rately during the course of the infection. No statistically significant differences between th e means were observed for any timepoint. The mean of adherent and invasive ba cteria per cell for the entire timecourse covered a range with the lower limit 3.8+/-4.9 and the upper limit 5.6 +/4.6. The result of an unpaired, two-tailed student’s t-test performed to compare the mean adhering bacteria of each strain at 20 minutes, two hours, six hours and 24 hours ranged from P=0.55 to P=0.75. These results suggest no difference in the total binding capability or the rate of bindi ng over time for the aaestrain. To increase the number of samples per conditi on, the total mean numbers of adherent and invading bacteria were considered for each st rain separately, pooling the results from each timepoint. A. actinomycetemcomitans strain SUNY1565 ( aae-) interacted with HIGK cells at ratio of 4.7+/-4.3 (n=56) bacteria per cell. 5.0 +/-3.9 (n=60) SUNY465 interacted per epithelial cell. Unpaired, two-taile d student’s t-test revealed the proba bility of the two means being equal was P=0.66. Rearrangement of Actin Cytoskeleton The rearrangement of the actin cytoskel eton was not observed at any timepoint or condition attempted thus far (Figure 5-5). Based on the ontology analysis illustrated in Figures 5-2 to 5-4, both wild type and aaeA. actinomycetemcomitans were expected to induce membrane ruffling and increased lamellipodial protrusions through the up-regulation of PIR121 (CYFIP1) and Nap125. The anticipated actin foci beneath attached A. actinomycetemcomitans

PAGE 146

146 previously reported in the literatu re were not obser ved (Fives-Taylor et al., 1995). No obvious shortening or lengthening of filopodia was detect ed, and the intensity of actin staining which would indicate an increase in the amount of actin present was not observed. No changes in the localization of concentrated structur es of F-actin were visible. Stru ctures that are believed to be stress fibers were observed in all conditions, but the number or characteristics of these structures were not changed at any time point. Discussion Previous reports (Rose et al., 2003; Fine et al., 2005) and our own results using viable counts have consistently supported that aaestrains of A. actinomycetemcomitans are impaired in their ability to adhere to GECs, BECs, and HIGKs. A binding reduction ranging from 65% to 90% has been documented by this method. The curre nt study attempted to visualize and confirm the aaeand wild type interaction and differential adhesion to epithelial cells using confocal fluorescence microscopy. Based on a limited number of observations and a single experiement, no difference in the ability of these strains to inte ract with oral epithelial cells was observed. To confirm the validity of these results, a larger sample will be taken consisting of 100 randomlyselected fields for each infection condition. Po sitive controls for invasion and cytoskeleton rearrangement will also be included in the analys is, such as parallel invasion of HIGK cells by P. gingivalis a prototype invader that induces observable F-actin changes (Hasegawa et al., 2007, submitted). Additional markers for the cell membra ne of epithelial cells will be used, such as fluorescent pan-cadherin antibody (Abcam, Camb ridge, UK) to more clearly delineate internalized bacteria from extern ally attached bacteria. Clearl y the discrepancy between direct microscopic observation and plate counting requires resolution. We cannot rule out at this point that this experiment requires further optimization and calibration.

PAGE 147

147 If these results are duplicated and definitively confirmed, one possible explanation for this discrepancy is a viability issue for aaebacteria exposed to epithe lial cells. Colony plating and spectrophotometric analysis of the i noculums have eliminated a higher MOI for aaestrains as a possible source of the unexpectedly high ad hesion observed with microscopy. A situation where aaemutant bacteria were presen t but non-viable would explai n the differences observed. A fluorescent stain that is able to differentiate live versus dead bacteria would aid in addressing this issue, and will be explored. A product such as SYTOX Green stain (Molecular Probes Inc., Eugene, Oreg.) will be considered to enumer ate dead and living bacteria under infection conditions with HIGK cells. Ontology analysis led to predictions of in creased actin rearrangement manifested in filopodia formation, lamellipodia formation, a nd membrane ruffling in HIGK upon interaction with A. actinomycetemcomitans These events are not predic ted to be Aae dependent, although absence of Aae seems to indica te an increased actin rearrang ement activity above the levels observed in HIGK cells interacting with wild type bacteria. The current experiment was not able to validate these changes phenotypically using F-actin staining and fluorescence microscopy. An explanation for the inability of the conf ocal fluorescence micros copy to corroborate the predicted actin cytoskeleton re arrangements is that the time poi nts chosen may not have been able to detect a dynamic event. Previous studies using cervical ep ithelial cells have demonstrated A. actinomycetemcomitans interaction occurs within ten minutes of inoculation, and bacterial entry and exit of host cells can occur in less than 30 minutes (Fives-Taylor et al., 1995). Although the 20 minute time point was chosen to account for these observations, it is possible that the rearrangement events of actin in HIGK cells occurred before the first time point was visualized microscopically. Ac tin foci beneath attached SUNY465 A.

PAGE 148

148 actinomycetemcomitans have been previously de monstrated (Fives-Taylor et al., 1995). Our transcriptional profiling furthe r supports the observation that actin rearrangement does occur upon microbe-host interaction. The discrepancies between ontology pr edictions, live counts and microscopic determination are under investiga tion. These experiments will include shorter time points and the combination of directed methods to qua ntify the size and numb ers of filopodia and lamellipodia. ImageJ (Rasband W; NIH, Bethesda MD) and Imaris (Bitplane Inc, Saint Paul, MN) software are capable of quantifying the size a nd intensity of specified features that appear in digitized images. Additional controls will al so be included to assure the technical soundness of our assays, including P. gingivalis invasion, which can induce actin cytoskeleton rearrangements (Hasegawa et al., 2007, submitted). Additionally, in light of the down regulation of the microtubule regulators, mDia and Mena, tubulin rearrangement may be a confirmable phenotype that is dependent upon Aae. Previous work has demonstrated the involvement of actin rearrangement of host cells during A. actinomycetemcomitans invasion which resembles pathogen-directed endocytosis of other organisms (Fives-Taylor et al., 1995; Meyer et al., 1996). Microtubules have been shown to play a role in the cellto-cell spread of A. actinomycetemcomitans as well as the egress of bacteria in to the extracellular medium (Meyer et al., 1999). The role of Aae in microtubule a ssociation and rearrangement thus warrants investigation in concert with the experiments examining actin cytoskeletal rearrangement. Chemical inhibitors of actin and tubulin assemb ly are a treatment which may help elucidate the effects of Aae predicted by ontology analysis. The inhibitor of ac tin assembly cytochalasin D, the inhibitor of microtubule polymerization colchi cin, and the microtubule st abilizing agent taxol

PAGE 149

149 are examples of such chemical agents th at have been used previously to study A. actinomycetemcomitans invasion (Meyer et al., 1999). To our knowledge, this study is the first direct examination of the properties of A. actinomycetemcomitans aaestrain adhesion to gingival epithelial cells compared to wild type strains. This is also the first report of the transcriptional response by HIGKs to Aae. Previous studies have relied upon viable counts to quantify the adhesion deficiency attributed to Aae disruption in A. actinomycetemcomitans mutant strains (Rose et al., 2003; Fine et al., 2005). The only previous microscopic visual examination of Aae-mediated adhesion was performed using E. coli transformed to express Aae, and assessed their binding to BECs. Interestingly, Aaetransformed E. coli robustly adhered to BECs at the rati o of ten to 100 bacteria per BEC at the saturating MOI of 1000:1 to 10,000:1. Yet, the same assay performed using GECs found only three Aae-transformed E. coli adhering per cell (Fine et al., 2005). Further, the adhesion assays of Aae-transformed E. coli and assays comparing wild type to mutant A. actinomycetemcomitans were performed on epithelial cells suspended in culture and under constant rotation. These conditions could create host cell binding charact eristics much different from cells already adhering to a substrate and forming a monolayer, not the least of which is a greater surface area available for bacterial binding in the suspended cells The authors’ observa tions that BECs seem to have more binding sites for Aae than GECs also has implications in our model. If we assume each GEC has only 3 receptors for Aae-mediated binding, the microscopic observation that equal numbers of mutant and wild type A. actinomycetemcomitans bound HIGK may have resulted from another receptor-adhesin inte raction, such as ApiA. This ma y have been enhanced with our infection conditions (MOI 2500:1). The original authors who discovered Aae in A.

PAGE 150

150 actinomycetemcomitans used an MOI of only 100:1, which ma y have been optimal conditions to avoid non-specific interactions (Rose et al., 2003). Although Aae has been clearly demonstrat ed to act as an adhesin under certain conditions, it is not unreasonabl e to speculate this protein may have additional unforeseen functions. Aae is homologous to Hap and Hia, a nd subsequent characterization of this protein revealed that Aae was most closely related to the IgA1 proteases of Neisseria and Haemophilus species (Rose et al., 2003). Based on this homology of Aae to IgA1 proteases and the adhesin/protease Hap of Neisseria and Haemophilus this protein could conceivably possess proteolytic activity in A. actinomycetemcomitans in addition to its role as an adhesin. Reexamination of the microarray data may rev eal a potential proteolytic effect of Aae transcriptionally. Haemophilus influenzae Hap plays a role in epithelial cell binding, binding to collagen IV, fibronectin and laminin, and also plays a role in autoaggregation of H. influenzae (Fink et al., 2003). Characteristic of autotransporter adhesins Hap consists of a passenger domain flanked by an NH2-terminal signal sequence and a COOH-termi nal translocator domain (Henderson and Nataro, 2001). Interstingly, in H. influenzae, the passenger domain po ssesses all the adherence capability, as well as the ability to release itself into the extra cellular environment through autoproteolysis (Fink et al., 2001). The proteolytic activity of the released adhesin/protease Hap has been proposed to aid initial colonization of th is bacterium through proteolysis of host immune factors or epithelia l components ahead of bacterial arriva l, although a host substrate for this potential activity has not been identified for Hap (Fink et al., 2001). Conversely, proteolysis occurring while Hap is stably adherent to an in vivo surface would result in the release of H. influenzae from its attachment site. While predictive modeling for Aae fails to predict a

PAGE 151

151 proteolytic domain, assays for prot eolytic activity by this protein in A. actinomycetemcomitans have not been conducted (Rose et al., 2003) and cannot be ruled out. The Hia protein of Haemophilus influenzae lacks the proteolytic activity of Hap, and has been proposed to better reflect the role of Aae in A. actinomycetemcomitans pathogenesis (Rose et al., 2003) according to current characterizations. The lack of proteolytic activity by this protein causes Hia to remain associ ated with the outer membrane of H. influenzae in contrast to Hap (St Geme and Cutter, 2000). Hia has been demonstrated to po ssess two binding domains that recognize the same receptor of human conjunc tival epithelial Chang cells, and distinct from the receptor of Hap. A lthough the two domains individually di ffered in their binding affinity by 10-20 fold, both domains were demo nstrated as essential for Hia binding to Chang cells to occur to the fullest extent (Laarmann et al., 2002). Further differentiating Hia from Hap, the COOHterminal domain of Hia is arranged as a trimeric translocator. This char acteristic constitutes a subclass of autotransporter molecules, which a ppear to specialize in high affinity adhesive interactions with host surfaces (Cotter et al., 2005). The IgA1 protease of Neisseria species is named for its ability to cleave the hinge regi on of IgA1 secreted by mucosal epithelium, presumably aiding bacterial colonization of this environment. Another possible role for this protein is based upon the IgA1 preotease-medi ated proteolysis of lysosome/late endosome associated membrane protein, LAMP 1. The intracellular lifestyle of N. gonorrhea has been studied, and demonstrates the survival of this bacterium within intracellular vesicles, the escape from said vescicles, and the association of N. gonorrhea with LAMP-1 proteins of endosomes (Hauck and Meyer, 1997). Additionally, an IgA1 protease-deficient mutant of N. gonorrhoeae has been reported to suffer a pronounced gr owth defect within epithelial cells (Lin et al., 1997; Henderson and Nataro, 2001; Ayala et al., 2002). Furthermore, LAMP1 proteolysis has been

PAGE 152

152 proposed as a mechanism for trafficking of N. gonorrhea within epithelial cells. Neisseria gonorrhea mutants deficient in the IgA1 protease have a demonstrated impairment in their ability to traverse polarized epithelial monolayer s of the T84 human colorectal epithelial cell line (Hopper et al., 2000). A. actinomycetemcomitans has also been demonstrated to reside within a membranous vesicle following invasion of cultured epithelial cells (Meyer et al., 1991). Additionally, the ontology an alysis of the unfiltered dataset re vealed the up regul ation of PI4P5K (PIP5K3) (data not shown), whic h is involved in endosome relate d membrane trafficking. These shared characteristics between A. actinomycetemcomitans and N. gonorrhea as well as the homology between Aae and the IgA1 protease, gr ant further investigation for a proteolytic activity of Aae, which may play a role in intracellular bacterial survival. Conclusions The deletion of A. actinomycetemcomitans Aae significantly impacted the transcriptome of HIGK oral epithelial cells, and revealed many pathways th at appear to be involved in hostpathogen interactions. The re gulation of actin cytoskeleton was one pathway highly impacted though transcriptional profiling and ontology anal ysis. Depending on the pairwise analysis performed, the pathway P-values ranged from 4.6 x102 to 1.4 x104. Ontology analysis predicted increased lamellipodia, filopodia and membrane ruffling in infected cells compared to uninfected. The ontology for aaeinfected HIGK predicts for a further increase of actin cytoskeleton rearrangements. Confocal fluoresce nce microscopy was used to assess the role of Aae on HIGK cell actin rearrangement Under the experimental cond itions used in the current study, microscopy failed to capture cl ear evidence of actin rearrangement. Therefore, the direct role of Aae in actin rearrangement is inconclusive In the absence of additional controls and in a single experiment, no evidence was obtained that duplicate the co-localization of A. actinomycetemcomitans and host actin foci reporte d previously (Fives-Taylor et al., 1995).

PAGE 153

153 The microscopic visualization of bacteria a nd host cells performed did not correlate with the data obtained with viable counts, nor support primary role of Aae as an adhesin of GECs. Direct visualization methods failed to confirm the adhesion deficiency of aaestrains previously reported through methods relying upon viable count s. Future experiments will be conducted to reconcile these discrepanc ies. Additionally, actin rearrangement events that may be occurring earlier than the first time point used in this st udy will be further investigated. The spinning disk confocal microscope utilized for this study is ca pable of recording time-lapse images of bacterial and host co-cultures, and will be utilized to record the interactions of A. actinomycetemcomitans with HIGK cells over an extended period of time. This will potentially capture actin rearrangments that occur very quickly, a nd capture the adherence and invasion of A. actinomycetemcomitans in real time, eliminating the comp lication of choosing static timepoints to capture rapid and dynamic events. Although Aae has defined capability to mediate A. actinomycetemcomitans adherence to the KB cervical epithelial cell line (Rose et al., 2003) and to primary buccal epithelial cells (Fine et al., 2005), it is reasonable to speculate that Aae possesses additional activ ities in oral gingival keratinocytes. As reviewed in Chapter 1, P. gingivalis gingipains and FimA pili are both capable of multiple interactions and functions in various models. Thus, in addition to exploring the role of Aae-mediated adhesion to gingival epithe lial cells, host transcriptional profiling may yet reveal unanticipated functions for Aae. The homology of Aae to IgA1 proteases, the ontology that predicts membrane ruffling, endocytosis, and endosomal traffick ing are potential clues to the function of Aae in host-pathogen interactions. Although Aae lack s a defined proteolytic motif, perhaps a proteolytic function of this component may allow the escape of A. actinomycetemcomitans from intracellular vacuoles following invasion, and the mutation of this

PAGE 154

154 gene decreases viability of bacteria interacting with epithelia l cells. The role of Hia, Hap, and IgA1 proteases are areas of interest due to their roles in pathogenesis of various bacteria and can provide clues into the functions of Aae (Henderson and Nataro, 2001). We have previously demonstrated that the specific host response of oral epithelial cells mediated by individual bacterial component s can provide insights into host pathogen interactions. Predicting phenotypi c outcomes of specific pathways from transcriptional data has remains daunting, as many pathways interact in co mplex networks. Caution in the interpretation of transcriptomes is thus critical, and th e importance of phenotypic characterizations to complement microarray studies remains invaluab le. Although the particular method chosen to validate the transcriptional profiles obtained from this study was not definitive, the regulation of gingival epithelial cell cytoskeleton rearrangements will be investigated further. Experimental conditions will be optimized, including proper controls and methods of quantifying actin rearrangement definitively. The potential to identify an unknown function for Aae also has arisen from analysis of the host cell transcript ome interacting with oral pathogens and will be pursued.

PAGE 155

155 Table 5-1. Gene ontology analysis (P<0.05) of the most impacted pathways caused by a mutation of aae (VT1565) in the parent strain A. actinomycetemcomitans (SUNY465). Impact Factora Pathway Nameb Modulated (total) genes in pathwayc 5.675 Regulation of actin cytoskeleton 15 (206) 5.143 Focal adhesion 18 (194) 4.376 ECM-receptor interaction 9 (87) 3.931 Jak-STAT signaling pathway 13 (153) 3.856 Adipocytokine signaling pathway 7 (69) 3.185 Leukocyte transendothelial migration 10 (117) 3.136 Cell adhesion molecules (CAMs) 10 (132) 2.856 Tight junction 10 (119) 2.581 MAPK signaling pathway 17 (273) 2.374 Cell cycle 9 (112) 2.254 Calcium signaling pathway 12 (176) 1.935 Apoptosis 4 (84) 1.769 Toll-like receptor signaling pathway 5 (91) 1.436 Cytokine-cytokine receptor interaction 14 (256) 1.247 Epithelial cell signaling in He licobacter pylori infection 2 (46) The epithelial cell pathways were determined by Pathway Express aThe impact factor measures the pathways most affected by the changes in ge ne expression by considering the proportion of differentially regulated genes, the perturbation factors of a ll the pathway genes, and the propagation of these perturba tions throughout the pathway. bAccording to the Kyoto Encyclopedia of Genes and Genomes pathways ( http://www.genome.jp/kegg/ ). cNumber of regulated genes in a pathway /total number of genes curren tly mapped to this pathway.

PAGE 156

156 Figure 5-1. HIGK transcriptome upon VT1565, SUNY465-, or mock-infection. RNA was isolated and purified af ter 2h co-culture with A. actinomycetemcomitans SUNY465 or VT1565 and compared to uninfected cells. This heat map and dendrogram were constructed from 4138 probe sets differentially expressed between the three experimental classes at the significance level of P <0.05. Probe set signal intensities were variance-no rmalized, mean-centered across samples, and subjected to hierarchi cal cluster analysis. Av erage linkage clustering by uncentered correlation was performed for genes and samples. The degree of similarity between the transcriptional profiles of each sample is expressed by Pearson’s correlation coefficient distance me tric, according to the adjacent scale. The expression state of each data point is represented as standard deviations from the mean expression level for that gene in all samples. Red indicates a relative increase, green indicates a relative decr ease, and black indi cates no relative change of mRNA transcripts for a given gene. Labels. Uninfected HIGK Cells, CTRL R1-R4; A. actinomycetemcomitans SUNY465-infected HIGK Cells, SUNYR1-R4; A. actinomycetemcomitans VT1565-infected HIGK Cells, ^Aae R1-R4

PAGE 157

157

PAGE 158

158 Figure 5-2. Impact of A. actinomycetemcomitans SUNY465 interaction with HIGK cells upon the Regulation of Actin Cytoskeleton pathway. Differential modulation of pathway genes is represented as expre ssion levels from Human Immortalized Gingival Keratinocytes upon co-culture with A. actinomycetemcomitans parent strain SUNY465 compared to transcript levels obtained from Uninfected HIGK cells. Red terms are transcriptionally up-regulated, while blue terms are downregulated compared to baseline conditions. Terms in green were not significantly modulated. Modulated gene s were significant at the P <0.05 threshold and were changed by at least two-fold magnitude.

PAGE 159

159

PAGE 160

160 Figure 5-3. Impact of A. actinomycetemcomitans SUNY1565( aae-) interaction with HIGK cells upon the Regulation of Actin Cytoskeleton pathway. Differential modulation of pathway genes is represen ted as expression levels from Human Immortalized Gingival Keratin ocytes upon co-culture with A. actinomycetemcomitans isogenic mutant strain SUNY1565( aae-) compared to transcript levels obtained from Un infected HIGK cells. Red terms are transcriptionally up-regulated, while blue terms are down-regulated compared to baseline conditions. Terms in green were not significantly modulated. Modulated genes were significant at the P <0.05 threshold and were changed by at least two-fold magnitude.

PAGE 161

161

PAGE 162

162 Figure 5-4. Specific contribu tion of Aae to the HIGK cell Re gulation of Actin Cytoskeleton pathway. Differential modulation of path way genes is represented as expression levels from Human Immortalized Gingiva l Keratinocytes upon co-culture with A. actinomycetemcomitans isogenic mutant strain SUNY1565( aae-) compared to transcript levels obtained from HIGK cells interacting with A. actinomycetemcomitans parental strain SUNY465. Red terms are transcriptionally up-regulated, while blue terms are down-regulated compared to baseline conditions. Terms in green were not significantly modulated. Modulated genes were significant at the P <0.05 threshold and were changed by at least two-fold magnitude.

PAGE 163

163

PAGE 164

164 Figure 5-5. HIGK cells interacting with Aggregatibacter actinomycetemcomitans Human Immortalized Gingival Keratinocytes we re visualized by Texas Red phalloidin staining of host cell actin cytoskeleton (red). A. actinomycetemcomitans (green or turquoise dots) were stained with Cell Tracker Green BODIPY live dye prior to co-culture experiments and allowed to in teract for the indicated times. Imaris software was used to digita lly recolor signal channels. A. actinomycetemcomitans VT1565 ( aae-) are shown in turquoise and SUNY465 (wild type) are shown in green. Columns shown from left to right are the three experimental conditions, Uninfected HIGK cells, VT1565-infected HIGK cells, and SUNY465-infected HIGK cells. Rows from top to bottom are the four timepoints chosen to investigate, 20 minutes, Two Hours, Six Hours, and 24 Hours post-innoculation.

PAGE 165

165

PAGE 166

166 CHAPTER 6 DISCUSSION AND PERSPECTIVES The fundamental questions that arise in this work are not only “W hat are pathogens doing to cause disease?” and “How can we promote hea lth?” but “What is health besides the absence of disease?” From the microecological standpoint, is health simply a microbial imbalance with constant turnover and the inability for a single bacterium to set up residence and cause tissue damage? Is health the opposite situation, the ability of truly commensal b acteria to stably colonize the oral cavity and prevent pathogens from ever occupying the niche that the “good” flora occupies? The intestinal flora has been shown to positively benefit the host in several instances, such as maintenance of homeostasis an d prevention of injury in the gut through TLR signaling (Rakoff-Nahoum et al., 2004), nutrient absorption (Hooper et al., 2001; Backhed et al., 2004), mucosal barrier fortific ation, and angiogenesis (Hooper et al., 2001). While this study cannot answer all these questions—however—it does lay the groundwork fo r investigating these complicated issues at a future time. In light of the complex questions at hand, the goal of this study was to gain insights into host-pathogen interactio ns that may occur in periodontal disease. This was performed with global host transcriptional profili ng of oral epithelial cells. Th e hypothesis driving this project was that oral epithelial cells ac tively and specifically respond to oral pathogens, and can thus serve as a reporter system to shed insight on bacterial pathog enesis. To accomplish this endeavor, three specific ai ms were completed: SA-I Establish a periodontal disease experimental model of epithelial cells and transcriptional profiling SA-2 Establish the epithelial host cell baseline gl obal transcriptional prof iles for uninfected and infected conditions under investigation

PAGE 167

167 SA-3 Assess the impact of indi vidual bacterial fact ors upon the host cell transcriptome and confirm biologically relevant predicted outcomes phenotypically Additional Collaborative Work In addition to the work presented as the focus of this study, the strategies developed herein have been applied to address additiona l questions addressing host-pathogen interactions likely to occur in the oral cavity that did not pertain directly to our specific aims. Through collaborations between the laboratory of Dr. Martin Handfield, as well as the lab of Dr. Richard Lamont, a wealth of transcriptional data has b een collected. As these tools continue to be applied to various organisms or bacterial compon ents, a broader perspective is to be gained. Afterall, with P. gingivalis and A. actinomycetemcomitans studies well underway, at least 698 other oral microbes await further study! In collaboration with Dr. Yoshiaki Hasegawa (Lamont Lab), the baseline transcriptional profiling for the commensal bacterium Streptococcus gordonii and the opportuni stic commensal Fusobacterium nucleatum was performed. The host epitheli al response to these organisms was compared to the baseline previously determined for P. gingivalis and A. actinomycetemcomitans (Chapter 4). One finding was a common response shared by all four of these microbes, as presented in Appendix A (Hasegawa et al., 2007). Also uncovered, were various epithelial responses which were completely unique to each organism. These unique responses are under investigation for possible im plications in actual human oral infections. The impact of specific bacterial compone nts was extended to study the effect of A. actinomycetemcomitans CDT on epithelial host cell transcript omes in collaboration with Dr. Mounia Alaoui El Azher (Handfield La b). As described for studies of A. actinomycetemcomitans ORF859 and P. gingivalis FimA, mutant analysis was combined with transcriptional profiling to reveal the CDT eff ects on HIGK cells. Consistent with the putative

PAGE 168

168 activity of CDT, the Cell Cycle and Apoptosis pa thways were both implicated. The dissection of these pathways was enabled through microarray analysis, and the involv ement of specific host genes in the epithelial response was further confirmed by Wester n analysis (Appendix B, AlaouEl-Azher). The effects of interf ering with the host response were studied with RNA interference, and the results have implications for the specifi c activity and targets of CDT (Alaoui El Azher et al., unpublished). Mutant analysis of another P. gingivalis component, SerB, has also been conducted (Hasegawa et al., submitted). Although the specific findings are beyond the scope of this project, these examples illustrate how th e epithelial host transcriptome has broadly been applied to study oral microbes. Lessons Learned Specific Aim 1: Establish the Epithelia l Transcriptome Experimental Model As reviewed in Chapter 2, pr ior to this study, global tran scriptional profiling with DNA microarrays had been utilized successfully to gain insights of host-pa thogen interactions, primarily between gut epithelium and enteri c pathogens. In part icular, the study of Helicobacter pylori which incorporated mutant analysis and assessment of the host transcriptome (Guillemin et al., 2002) provided an encouraging ba sis to attempt a model of ep ithelial cell interactions for the oral cavity. Prior to executing a similar mo del in the oral epithelium, however, several factors had to be addressed. Choosing the microarray and data analysis platform In contrast to other works at the time, th e current project was de signed to reveal genes whose modulations were statisti cally different between classes —not simply those that met a threshold for fold induction or decrease. Usi ng statistical methods to find genes that are differentially regulated provides a measurable probability that uncovered genes are truly predictors of a given experimental condition. Add itionally, as replicates are added to a statistical

PAGE 169

169 analysis, variance is reduced and the differences between mean values become more clearly defined. This strategy allowed genes to be unc overed during the analysis which would normally be missed due to the lack of power by other fi ltering methods, such as those based on foldchange or visual interest in clus ters of like-regulated genes. Fu rther, the variability of a given gene’s expression within samples of the same class was considered in statistical methods, which reduced false positives with large fold-expression changes but even larger standard deviations between samples. The advantage of the significa nce threshold is a reliab le method to anticipate the quality of the dataset under study. In addition to utilizing sta tistical analysis software, the use of commercially available, annotated microarrays were an important feature of this study. The a nnotations for impacted genes were readily available onlin e (Affymetrix), and ontology tool s were available to organize impacted genes into biologically relevant pathwa ys. The use of these tools has been an often underappreciated aspect of the current study, bu t a critical component of interpreting the transcriptional data. The advent of Pathway Express has been a welcome improvement to the bioinformatic aspect of the process, which automatically generates populated KEGG pathway maps that previously had to be populated manually. These considerations di d not directly impact the biological aspects of our experiments, yet their importance to successful array experiments cannot be overlooked. Determining standard infection conditions The actual benchwork that comprises a transc riptional profiling experiment was refined through trial and error until the experimental cond itions that came to define a standard array experiment were developed. The growth stage of the infecting bacteria to use for co-culture experiments, the extent of monolayer confluence, and even the appropriate cell line to use had to be optimized. The first microarray experime nt was conducted using the MOI of 100:1 for A.

PAGE 170

170 actinomycetemcomitan s and P. gingivalis in parallel with uninfected cells. An interesting finding was the close clustering of A. actinomycetemcomitans -infected epithelial cells with uninfected cells under these condi tions. Review of the phenotypic data, which showed the ratio of adherent A. actinomycetemcomitans to host cells was less than one per cell, suggested a majority of the cells were profiling as uninfected. This led to the hypothesis that the MOI had to be high enough for every host cell to encounter a single bacterium in order for a homogenous mRNA pool reflective of infection to result. The next round of experiments incorporated this change, and a robust host response to A. actinomycetemcomitans was observed. The importance of including proper controls to monitor the actual co-culture interactions was reinforced by this experience in addition to establishing an MOI of 1000:1 to 2500:1 as the target for A. actinomycetemcomitans The ratio of total interacting bact eria thus became the benchmark for a successful co-culture experiment, ra ther than the initial MOI. More aggressive bacteria, such as P. gingivalis thus required fewer initial bacteria to obt ain the same level of interaction with host cells as compared to A. actinomycetemcomitans Interstingly, the latest confocal microscopy experiments demonstrate that not every host cell in fact encounters at least a single bacterium. This raises the possibility that a dose response between the MOI and host transcrip tional profiles exists, possible dependent on secreted factors. Effect of host cell lineage on transcriptional response Reliance upon the KB (HeLa CCL-17) cell li ne was widespread upon the onset of this study, because KB cells were believed to descend from an oral carcinoma, and in vitro invasion assays between these cells and oral pathogens could successfully re plicate the clinical observations of an invasive b acterial phenotype. The agreemen t of phenotypic characteristics during these assays between KB cells and primary oral epithelial cells also supported their use. The discovery that KB cells were actually HeLa cervical epithelial cells that contaminated the

PAGE 171

171 original cultures combined with the realization that epithelial cel ls are able to respond actively and specifically to infection raised questions regarding the relevance of using KB cells in periodontal research. Thus, in what is believed to be the only study of its kind, two different epithelial cell lines—KB and HIGK— were co-cultured with A. actinomycetemcomitans in parallel, and the transcriptomes were analyzed. Not surprising in relation to the hypothesis that epithelial cells mount an active a specific response to infect ion (Dale, 2002), significant differences were found between the responses of th ese two cell lines to the same bacteria. As detailed in Chapter 3, a large degree of simila rity was shared for many genes that involve homeostasis and other functions not within the host-p athogen interaction realm. This was not surprising since both cell lines were epithelial ce lls and shared some major characteristics. However, the differences in Response to Pest, Pa thogen, and Parasite pathways and others with documented relevance to host-pathog en interactions revealed specifi c interactions that could be attributed to bacterial tropism. Establishing th is cell-line based specifi city under the optimized infection conditions already discus sed was an important aspect of completing Specific Aim 1, as well as establishing the transcriptional model for probing hostpathogen interactions of the oral epithelium. This lesson also cautions that alt hough a high degree of similarity has been shown between HIGK cells and primary GECs in terms of their behavior durin g invasion assays and other phenotypic responses to oral bacteria, differences in the tr anscriptional responses of these cells will also likely exist. One potential source of differences is the HPV immortalization of HIGK cells (Oda et al., 1996). Over the course of 350 passages, several chromosomal abnormalities were detected, including a mutation in P53 (Oda et al., 1996b). Although the impact on cellular regulatory pathways, such as cell cycle and apoptosis, were not severe enough to cause these cells to become tumorigenic, the overall effects of HPV immortalization are

PAGE 172

172 unknown as related to HIGK responses to bacteria l interaction. Theref ore, this may be a limitation of the current model. Specific Aim 2: Establish the Baseline Oral Epithelial Transcriptome Upon Infection P. gingivalis and A. actinomycetemtomitans -specific responses Now established, the epithelial transcripti onal model system began to provide insights into the effects of A. actinomycetemcomitans and P. gingivalis upon the epithelium. The experiments described in Chapter 4 which compared co-culture of wild type bacteria and HIGK cells to uninfected HIGK cells revealed both common and exclusive pathways for a given experimental treatment. Many pathways await furt her investigation into their potential role in health or disease of the oral epitheliu m. Notably, the different impact of P. gingivalis and A. actinomycetemcomitans upon HIGK apoptotic pathways was revealed using transcriptional profiling. The subsequent ontology analysis an d phenotypic confirmations demonstrated the ability of P. gingivalis to actively prevent a poptosis in these cells, and the pro-apoptotic phenotype initiated by A. actinomycetemcomitans These findings were presented to the 83rd General Session of The IADR in a symposiu m (Appendix B, Mans). The different host apoptotic response to these two pathogens could help explain w hy LAP progression is rapid, and CP is a chronic infection, as the host tissue is prevented from undergoing cell suicide in one case, and promoted in the other situat ion. Further dissection of the A. actinomycetemcomitans impacted apoptotic pathway was performed as a dire ct result of the transcriptional findings. The induction of several apoptosis prot eins were studied using Western blot analysis, and were also presented to the 83rd General Session and Exhibition of the IADR (Appendix B, Naselsker). Baseline host response to S. gordonii and F. nucleatum Although not a specific component of the current study, the methods developed herein have been used in collaboration with others (Appendix A). As a result, the baseline

PAGE 173

173 transcriptome was also established for HIGK cells interacting with the commensal Streptococcus gordonii and the opportunistic commensal Fusobacterium nucleatum When compared to the transcriptional profiles and ontology established for A. actinomycetemcomitans and P. gingivalis interaction, core pathways were found for all four infection c onditions. The specific manner in which each bacterium impacts these common pathways is currently under investigation. Also in support of an active response to in fection by oral epithe lial cells were the pa thways that were distinct for each bacterial species. These path ways will be examined for their potential to promote health or disease, and may also serve as markers of active bacterial infection compared to generalized colonization or mere presen ce of bacteria in the oral cavity. A more detailed analysis of the HIGK response to S. gordonii and F. nucleatum have been completed, and the findings were presented to the 84th General Session and Exhibition of the IADR (Appendix B, Hasegawa) and later pub lished in the journal In fection and Immunity (Appendix A). The baseline HIGK response to inf ection yielded useful insights which allowed species-specific outcomes to be studied, as pres ented, such as cytokine regulation. Of great interest, although F. nucleatum increased the expression of th e pro-inflammatory cytokines Interleukin-6 (IL-6) and IL-8, S. gordonii actually repressed these cytokines below levels observed in uninfected HIGK cells. This finding suggests that S. gordonii may function to counteract a destructive immune response caused by other bacteria and has implications in periodontal disease that will be explored futh er. Additionally, the baseline response to interaction with wild type b acteria became the foundation of completing the third and final specific aim, the contribution of specific bacterial components.

PAGE 174

174 Specific Aim 3: Investigate th e Impact of Individual Bacteri al Components on the Host Cell Transcriptome. Impacts of P. gingivalis fimbriae and A. actinomycetemcomitans ORF859 The combination of host transcriptional profiling with a mutant analysis strategy allowed the completion of Specific Aim 3: assess the c ontribution of individual bacterial components on host-pathogen interac tions. As first performed in Chapter 4, the host response to P. gingivalis YPF-1, a strain deficient in major fimbriae, and A. actinomycetemcomitans JMS04, as strain with a mutation in the ORF859 gene implicated by IVIAT as an in vivo induced gene in actual human LAP infections, were compared to the baseline host responses of HIGK cells to the wild type parental strains. Strain JMS04 was characterized prior to this array st udy, and mutation of the ORF859 gene was found to confer a def ect in intracellular survival by A. actinomycetemcomitans encountering epith elial cells (Cao et al., 2004). Interestingly, the imp act of this bacterial product upon the HIGK transcriptome was found to be minima l, and may reflect a lack of interaction between this bacterial pr otein and the host cell. This is c onsistent with a role for ORF859 in bacterial homeostasis and that this component is not a toxin. In contrast, the mutation in strain YPF-1, which abrogated the expression of func tional major fimbriae, was shown to highly impact the HIGK transcriptome. Previous work had shown a diversity of functions for major fimbriae, as reviewed in Chapter 1, and the ex tent of interaction between this bacterial component and host cells was reflected by the tr anscriptional profile. This work was also presented to the 83rd General Session and Exhibition of the IADR. A. actinomycetemcomitans CDT and Aae impact on host pathogen interactions The contributions by two additional A. actinomycetemcomitans components to hostpathogen interactions have been investigated. As introduced in Chapter 1, the Cytolethal Distending Toxin (CDT) and autotransporter adhe sin Aae are important virulence factors of A.

PAGE 175

175 actinomycetemcomitans CDT is known to cause cell cycle arre st leading to apoptosis in several cell types, and Aae is currently credited for spec ific adhesion to oral ep ithelial cells. To probe the transcriptional impact upon the host cells of these functions, the array analysis, ontology, and phenotypic assays were performed in collaborati on with Dr. Alaoui-El-Hazer. Not surprisingly, CDT impacted the Apoptosis and Cell Cyle path ways significantly in HIGK cells. Importantly, the transcriptional profiling not only confirmed the reported func tion of this exotoxin, it also revealed the manner in which cell cycle arrest was perturbed at the level of individual genes of the cell cycle pathway. Although not a specific aim of the cu rrent project, development of Specific Aims 1 and 2 directly led to progress in the study of CDT. The results of this collaboration were presented to the 84th General Session and Exhibi tion of the IADR during a symposium (Appendix B, Mans). Additional wo rk to dissect the cell cyle pathway was performed using RNAi, and was also presented to the IADR (Appendix B, Alaoui-El-Azher). As discussed in Chapter 5, mutation of the aae gene conferred an adherence deficiency by viable counts, as previously de termined by two other groups (Rose et al., 2003; Fine et al., 2005). In addition, transcriptional profiling reve aled a significant impact on HIGK regulation of actin cytoskeleton among the most significantly impacted pathways. This transcriptional data was presented to the 85th General Session and Exhibition of th e IADR (Appendix B, Comerford). Confocal fluorescence microscopy failed to c onclusively and definitively reveal actin rearrangement at several timepoints and after repeated co-culture e xperiments. This situation has several possible explanations. The phenotypic pred iction could have been incorrect, as the result of an interpretation error of the resultant eff ect on the pathway based on the transcriptional profiles obtained. Alternatively, the timepoints chosen to confir m actin rearrangement may have been inadequate to capture a fleeting event. Th e possibility also exists that this pathway is

PAGE 176

176 simply a false positive, as the stringency for this particular experiment would predict 5% of the total genes are called significant by chance alone. To investigate these qu estions, other proteins in the pathway will be chosen to confirm the predicted phenotype arising from the transcriptional profile and ontology analysis. Microscopic visualization of a single experime nt suggested that an equal number of wild type and mutant bacteria adhe red to and invaded host cells. Further confirmation experiments will be performed with additional controls to conc lusively investigate these initial findings. If confirmed, the data reported he re would suggest a novel function for Aae that may have been discovered through the use of transcriptional profiling. A possible explanation based on the homology of Aae to other bacterial component s is a defect in intracellular viability by A. actinomycetemcomitans Aae mutants. Neisseria gonorrhea deficient in the close homolog of Aae, IgA1 protease, is less viable than the parent al strain in trans-endothe lial migration assays. The discrepancy found in the mutant bact eria VT1565 between the adherence defects demonstrated by indirect colony counts compared to direct visualization requires resolution. If nothing else, this situation emphasizes the crit ical importance of phenotypic confirmations of transcriptional profile data. Advantages and Limitations of the Current Epithelial System Advantages The obvious advantage and primary reason for conducting global transc riptional profiling is the enormous amount of data that can be collected from a single experiment. Microarray experiments are time consuming to optimize, but once the experimental system is established, throughput of samples becomes highly efficient. Additionally, the use of a microarray for all known human genes avoids biasing the lists of significant genes to reflect known paradigms of host-pathogen interactio ns, such as concentrating on immune effectors and signaling molecules

PAGE 177

177 as reviewed in Chapter 2. The global screen ing afforded by this te chnology facilitates the discovery of new host responses, as is potential ly the case for Aae. Once biological pathways have been implicated through transcriptome and ontology analysis, more traditional assays can be used to validate the findings phe notypically. This strategy quick ly focuses investigations to a manageable experimental question that is highly likely to provide meaningful data. A single array experiment can provide a workable datase t for confirmation that will drive laboratory research for many years after the initial experiment As the case with the current model, only the most relevant impacted pathways are typica lly confirmed immediatel y following a successful array experiment. Although lower pr iority at first glance, the nume rous pathways that are first passed over for phenotypic confirmation still yield ve ry valuable information. Datasets can be revisited as ontology tools improve, and new discoveries can come from old data in this manner. Limitations The most glaring limitation of the current strate gy is the exorbitant costs associated with transcriptional profili ng of human host cells. Roughly $800 to $900 dollars must be spent per sample, which limits the possible experiments that can be reasonably run. The total cost of the project described herein, not including colla borations, was estimated at $45,000. Since the current strategy relies on biologi cal replicates of four per co ndition, financial costs limit the number to between two and four experiment al conditions per microarray experiment. Another limitation lies in the infection mode l used for host profiling. Quite clearly, cells grown in tissue culture are different from cells in the mouth of a human ho st. The intercellular interactions that occur in differentiated tissue are different from the interactions between cell monolayers and plastic. Additiona lly, the grown media provides different nutrients than what are found in the oral cavity. As IVIAT work previously demonstrated (Song et al., 2002; Handfield et al., 2000), bacteria grown in liquid culture al so express different genes compared to

PAGE 178

178 those found in an actual infection. The MOI used to elicit a representative response may also be different than normal in vivo infection conditions in order for the transcriptional profiling to reveal a consistent pattern of gene expressi on. Thus, any findings made using the current in vitro model have to be accepted for their limitati ons as a model of peri odontal disease and not periodontal disease proper. C onfirming the hypotheses from in vitro models in a clinical setting is of course a necessary exercise to fully advance the understanding of periodontal disease pathogenesis. Future Endeavors Revisit the Database Periodically. Including recent collaborations, 123 GeneChips for the Human Genome have been assayed for ten separate transcriptomic studies directed towa rds host-microbe interacti ons of the oral cavity. The approximate costs of these studies was $105,000. This wealth of information reveals the global status of oral epithelial ce lls in the presence of 4 bacteria l species, three isogenic mutant strains of A. actinomycetemcomitans and two mutant strains in P. gingivalis. As gene ontology and pathway mapping tools improve, this informa tion will be invaluable to unraveling the complex interplay between host and pathogen. Th e current study presents a first level of the interpretation and validation of the transcri ptional profiles alrea dy obtained. Phenotypic confirmations for a handful of predicted pheno types have been performed, but potentially many more experiments can be based on the data alr eady on hand. It is becoming increasingly clear that the interrelatedness of the pathways implicat ed tell an important story. Transcriptome wide reverberations attributable to single bacterial components have provided a glimpse of how host cell pathways are all intertwined at some leve l. As pathways become better characterized, identifying the connections will become more manageable.

PAGE 179

179 Uncharacterized Bacterial IVIAT Genes A major motivating factor to developing the current experimental system of host transcriptional profiling was the prospect to assi gn probable functions to bacterial genes with no known homology or function in current databases. In Vivo Induced Antigen Technology (Handfield et al., 2000; Rollins et al., 2005) has identified 116 A. actinomycetemcomitans and 144 P. gingivalis sequences specifically induced in human periodontal infections (Song et al., 2002). Many of these sequences have no known homology or biological function. Proof of principle was established when the A. actinomycetemcomitans strain JMS04 isogenic mutant for ORF859 was studied using the current transcript ome model. A biological impact upon the host cell was uncovered, although the to tal numbers of impacted genes was lower than what was observed for known virulence determinants. Although expensive to ch aracterize over 200 IVIAT genes in this manner, using the host transc riptome as a reporter system seems to be the most efficient method available to assign functi ons to these uncharacterized genes. Thus, mutants for IVIAT genes with unknown functions should be inve stigated usi ng the current strategy. A phenotypic screen prior to the array experiment to identify prospective clones with an observable impact upon the host cells would help identify bacterial gene s that are likely to yield a definitive result and formal assignment of function in pathogenesis. Complex Flora As reviewed prior to this project being st arted, one strate gy to prevent periodontal disease may be to control key members of the oral cavit y that disrupt the microbi al ecological balance from a symbiotic and healthy host-pathogen re lationship towards pathogenesis and disease (Kinane et al., 1999). Studies attempting to understand th e outcome of inter-bacterial encounters and their subsequent effects on host cells have al ready been undertaken in another collaboration study performed during the course of this st udy (Dr. Kate Von Lackum, Hanfield Lab).

PAGE 180

180 Transcriptome analysis presented to the 85th General Session of the In ternational Association for Dental Research (IADR) (Appendix B, Von Lackum ) investigated the differences in epithelial cell responses to a mixed S. gordonii P. gingivalis culture compared to HIGK cells interacting with each bacteria alone or uni nfected. This four-class anal ysis revealed transcriptional differences between classes, and gene ontology work is underway to predict the biological outcome of these interactions. It will be interesting to see if S. gordonii is able to confer a protective effect to gingival epithelial cells during mixed co-culture experiments, when P. gingivalis infection alone modulates host cells in a manner that is presumed to directly cause a chronic infection. Path ways already demonstated to be affected by P. gingivalis such as apoptosis as presented in Chapter 4, will be especially interesting under mixed culture conditions. Another antagonistic relationship that may exist in vivo was discovered from the transcriptome of HIGK cells interacting with P. gingivalis and A. actinomycetemtomitans and mentioned previously. The anti -apoptotic phenotype promoted by P. gingivalis and proapoptotic phenotype caused by A. actinomycetemcomitans infection would presumably counteract in epithelial cells en countering both bacteria simultane ously. The outcome of a mixed species co-culture to investigat e the resultant apoptotic state of the cell is one example of additional host-microbe inte ractions worth future study. Time course of Infection and Paralle l Host and Pathogen Array Analysis A tantalizing prospect introduced in Chap ter 2 is to study the host transcriptome and pathogen transcriptome simultaneously. A timec ourse experiment mon itoring the dynamics of gene expression by both participants in the hos t-pathogen interplay woul d truly reveal pointcounterpoint modulations that occu r as both entities try to cope w ith the other. The progression of the responses conceivably would reveal causeand effect relationships between the actions of host and pathogen.

PAGE 181

181 As presented during the 2007 UFCD Research Day (Appendix C) a timecourse analysis of the host transcriptome is challenging and more complex than using static “snapshots” of an interaction. The wavelike modulati on of genes is difficult to reso lve with basic statistics, and complicated models must be applied to interpre t the data. Based on this initial experiment, biological replicates for each timepoint under investigation w ould help resolve expression patterns, while unfortunately increasing the fina ncial cost of an experiment. However, the statistical tools are available to undertake such an experiment and a para llel profiling of host and pathogen transcriptomes over time potentially could yield highly in formative data that can not be obtained any other way. Improvements to the Transcriptome Reporter System Model Adapting the current in vitro model to more accurately reflect the oral cavity is a goal that co-culture assays using mixed bact erial infections could bring closer to realit y. Another option is to use a more complicated host tissue model to probe the host response more deeply. Although the specific responses of epithe lial cells in a monolayer have been studied successfully, how a more complex model that more closely mirrors multi layered host tissue would react is unknown. Previously, such an endeavor was limited by the inability to separate homogenous populations of cells from mixed models. As mentioned in Chapter 2, laser microdissection (Schutze et al., 2007) and improvements in the amplifica tion of minute quantities of mRNA (Viale et al., 2007) make complicated host model systems feasible. Of course, an alternative to complex in vitro models, whether cell lines are derived from primary cells, immortalized gingival tissue, or differentiated tissue models, is to study the transcriptome of diseased tissue sites from actua l patients with LAP or CP. An initial concern with this strategy is the patient-to-patient geneti c variability that could overshadow differences between healthy and diseased tissue. Additi onally, the complexity of periodontal disease

PAGE 182

182 discussed in Chapter 1 introduces numerous variab les into a transcriptomic analysis that would add noise to the analysis. The number of matc hed samples that would need to come from healthy and diseased sites from the same patient is likely to incur a significant cost of materials. This type of experiment would cost more than the in vitro studies described herein, and would present several technical challenges, such as obtaining a homogenous cell type from the tissue samples in order to obtain repr esentative transcripti onal profiles. A pilo t study to investigate transcriptional profile signatures that differen tiate CP and LAP patients was recently performed using 35 arrays covering 14 total patients (Papapanou et al., 2004). No significant differences were found between transcripti onal profiles from CP and LAP patients. However, a new grouping of patients was uncovered by microarray analysis of patie nt-isolated tissue. Although all patients had similar clinical presentation, the profiles of these patien ts separated into two groups along the lines of antibody titers to severa l periodontal pathogens. This study indicates a larger sample size of patients is necessary to uncover additional gene expression differences in LAP or CP patients. Two other recent studie s that involved clini cal patient samples under microarray analysis utilized 312 (Buness et al., 2007) and 1422 indivi dual samples (Brodsky et al., 2006)! The cost of materials alone would cu rrently preclude a 100 sample microarray experiment with our current level of understand ing the host response to specific bacteria and specific bacterial components. A st udy of this nature would also address different questions that are generally more descriptive a nd predictive of periodontal disease outcome, in contrast to the current study which investigated specific host-pa thogen interactions. Until the current model is no longer yielding results, and the targeted inves tigations of specific virulence factors is no longer worth the cost, the in vitro system developed during this pr oject will be highly beneficial.

PAGE 183

183 However, as the questions being asked increase in complexity, and the utility of epithelial cell monlayers wanes, the necessity to conduct transcriptional pr ofiling experiments from actual patients will become pressing. At some point, these experiments will simply have to be conducted, despite the financial cost. Perhaps th e cost of materials will decrease as array technologies improve to aid in this transition. Summary Altogether, the current project has succe ssfully developed a model that used the transcriptome of oral epithelial cells as a reporter system to di ssect host-pathogen interactions. In addition to establishing a platform technology th at will provide useful data for years to come, the development process itself has yielded insights into the interplay between oral epithelial cells and several oral microbes. A host response that is specific to bacter ial species was uncovered, and the degree of specificity also allowed the impact of individual bacterial products to be assessed. Improvements to the established mode l will allow more complex questions to be addressed, shedding light on the amazing array of inte ractions that occur in the oral cavity and result in a healthy or diseased outcome. Indeed the paradox is true, “the more we learn, the less we know.” However the insights gained herein se t the stage for exciting discoveries in the future of periodontal disease research.

PAGE 184

184 CHAPTER 7 GENERAL CONCLUSIONS The work described herein has demonstrated the successful development and application of a powerful strategy to probe th e interactions between gingival ep ithelial host cells of the oral cavity and oral microbes. The specificity of the host response to bacter ial challenge has proven useful, and subsequently the cells themselves have served as a reporter system to reveal previously unknown genome-wide re verberations caused by whole bact eria and specific bacterial components. This reporter system has been co mpletely conceived and developed within the scope of the current research pr oject, and the collaborations ma king this work possible represent the only instances of host transcri ptional profiling curren tly used in oral biology to study disease pathogenesis. In parallel with the developm ent of the current model to st udy host-pathogen interactions, insights into the pathogenes is of species such as Aggregatibacter actinomycetemcomitans and Porphyromonas gingivalis have been gained. The host regula tion of many biological processes, such as Apoptosis and Regulation of Cell Cycl e have been distinct upon exposure to these pathogens, and many more candidate processe s await phenotypic confirmation. A specific response has also been demonstrated by epithe lial cells encountering an isogenic mutant for specific bacterial components when compared to the baseline respons e to a parental strain. The specificity of the host response has been demons trated as dependant upon the origins of the epithelial cell r eacting to bacterial challenge, as the transcriptome from cervical and oral epithelial cells were distinct. Many lessons have been learned and the voi d in understanding the complex interplay between oral bacteria and host cells has begun to fill, if only slightly. This project has established a solid foundation for future investigations of the f actors contributing to LAP and CP

PAGE 185

185 progression, from both the standpoint of a reliable model for host-pathogen interactions, as well as a wealth of transcriptional data that will help unravel the comp lications of periodontal disease.

PAGE 186

186 APPENDIX A GINGIVAL EPITHELIAL CELL TRANSCRIPT IONAL RESPONSES TO COMMENSAL AND OPPORTUNISTIC ORAL MICROBIAL SPECIES9 9 The following manuscript is reprinted with permission from American Society for Mi crobiology. To access the definitive version, please refer to: Hasegawa, Y., Mans, J.J., et al (2007) Gingival epithelial cell transcriptional responses to commensal and opportu nistic oral microbial species. Infect Immun 75 : 2540-2547.

PAGE 187

187

PAGE 188

188

PAGE 189

189

PAGE 190

190

PAGE 191

191

PAGE 192

192

PAGE 193

193

PAGE 194

194 APPENDIX B INTERNATIONAL ASSOCIATION FOR DENTAL RESEARCH (IADR) ABSTRACTS RESULTING FROM TRANSCRIPTOMICS PROJECT10 83rd General Session and Exhibiti on of the IADR, Baltimore 2005 Seq#66, Thursday 10 March 2005, 10:45 AM 12:45 PM Oral, Room 330 Microbiology/Immunology and Infecti on Control Proteomics/Genomics 0328 12:15 Transcriptional Profiling of Epithelia l Cells Interacting wi th Oral Pathogens J J. Mans*, H.V. Baker, G. Narasimhan, A. Progulske-Fox, R.J. Lamont, M. Handfield Seq# 246, Friday 11 March 2005, 2;00 PM 4:00 PM, Poster, Exhibit hall E-F Microbiology/Immunology and Infection Cont rolMicrobiology/Im munology of Periodontal Diseases 2244 Modulation of Apoptosis-Associated Proteins of Gingival Cells by Actinobacillus actinomycetemcomitans D Naselsker*, M. Nolin, J.J. Mans, M Handfield 84th General Session and Exhibiti on of the IADR, Brisbane 2006 Seq #113, Thursday 29 June 2006, 3:30 PM 4:30 PM, Poster, Exhibit Hall 1 Microbiology/Immunology and Infection ControlClinical Oral Microbiology 1374 Insight into Bacteria-Epithelium Inte ractions from Transcriptional Profiling. Y. Hasegawa*, J.J. Mans, H.V. Baker, R.J. Lamont, M Handfield Seq# 163, Friday 30 June 2006, 1:30 PM 3:30 PM, Oral, Room M4 Microbiology/Immunology and Infection C ontrol Oral Microb iology and Immunology 1988 ATM Pathway Involvement in Actinobacillus acti nomycetemcomitans Induced-Apoptosis of IHGK Cells M. Alaoui-El-Azher*, N. Come rford, A. Progulske-Fox, R.J. Lamont, and M. Handfield Seq# 163, Friday 30 June 2006, 1:30 PM 3:30 PM, Oral, Room M4 Microbiology/Immunology and Infection C ontrol Oral Microb iology and Immunology 1990 Dissecting Specific Actinobacillus acti nomycetemcomitans -Epithelial Cell Interactions with Transcriptional Profiling. J.J. Mans*, W. Chen, C. Chen, H.V. Baker, R.J. Lamont, M. Handfield 85th General Session and Exhibition of the IADR, New Orleans 2007 Seq #268 Saturday, 24 March 2007, 10:45 AM-12:00 PM Poster, Ernest N. Morial Convention Center Exhibit Hall I2-J AADR/Pfizer Hatton Awards 2574 Characterization of Epithelial Respons es to a Complex Microbiota Challenge K Von Lackum*, J.J. Mans, J. Van Puymbrouck, R.J. Lamont, and M. Handfield 10 *Presenting author

PAGE 195

195 Seq #89 Thursday, 22 March 2007, 2:00 PM-3:15 PM Poster, Ernest N. Morial Convention Center Exhibit Hall I2-J Microbiology / Immunology and Infection ControlAggregatibacter actinomycetemcomitans II 0614 Effect of Aggregatibacter actinomycetemcomitans Adhesion on the Epithelial Transcriptome N.P. Comerford*, J.J. Mans, P. Fives-Ta ylor, R.J. Lamont, and M. Handfield

PAGE 196

196 APPENDIX C TEMPORAL VARIATION OF THE TRANSC RIPTOME OF GINGIVAL CELLS INTERACTING WITH Aggregatibacter actinomycetemcomitans11 J.J. Mans*, H.V. Baker, R.J. Lamont and M. Handfield Fifth annual UFCD Research Day. Friday, April 13, 2007. 1:30-3:30, Founders Gallery PhD/Post-doc Division, Abstract 30 *Presenting author (poster) We have previously described how Aggregatibacter actinomycetemcomitans (A.a.) can modulate the transcriptome of epith elial cells upon interaction. Further, this method has been useful in determining impacted biological pa thways, and in predicting downstream phenotypic changes resulting from the host-pathogen encounter. Objective : The purpose of this study was to expand the predictive power of transcriptional profiling by a dding a temporal component to our previously established model of host-pathogen interaction. Methods : Human immortalized gingival keratinocytes (H IGK) were grown in vitro following standard procedures and cocultured with A.a. VT1169, a nalid ixic acid/rifampici n resistant spontaneous mutant derived from a serotype b clinical isolate, A.a. SU NY465 (smooth phenotype). Host cell to tal RNA was extracted for seven timepoints at 20-minute in tervals for two hours. The RNA was purified, quantified, and reverse transcrip tion was performed to generate complementary-RNA. This cRNA and relevant controls were labeled and used to probe the Affy metrix HG-U133A human microarrays according to the manufacturer's recommendations. Expression patterns were analyzed with bio-informatic, st atistical and gene ontology tools. Antibiotic protection assays were performed in parallel to measure total levels of bacterial interaction. Results : At 20 minutes post infection, A.a. interacted with HI GK cells at a ratio of 19+/-3 A.a. per gingival 11 This work was supported by NIH/NIDCR T32 Grant DE07200 (JJM), DE11111 (RJL) and R01 DE16715 (MH).

PAGE 197

197 epithelial cell. Th e interaction peaked at 40 minutes post infection, with 36+/-10 A.a. per HIGK cell at this timepoint, and gr adually declined to 18+/-2 pe r HIGK cell at two hours. The unsupervised microarray analysis by Cluster re vealed two distinct clusters of upand downregulated genes over the course of infection. Performance of s upervised analysis with Edge revealed genes that were significantly modulated over ti me in a non-linear manner, demonstrating the dynamics and complexities of the HIGK transcriptional profile as it impacts biological response pathways. Pairwise ontology analysis of each timepoint relative to time 0 revealed biological functions most highly imp acted in HIGK cells upon infection, and how they were modulated temporally. The 20 biological f unctions with the highest impact factors ranged from phosphatidylinisitol signaling (10.02) and cell cycle (4.5), to actin cytoskeleton rearrangement (2.5). Predicted phenotype s are currently under investigation. Conclusion : Combining transcriptomic, ontology analysis and temporal methods constitute a predictive tool to study host pathogen interactions. Phenotypic confirmations of the transcriptomic data generated here will provide further insights a nd definitive evidence of the contribution of novel host pathogen interactions in the oral cavity.

PAGE 198

198 LIST OF REFERENCES Aas, J.A., Paster, B.J., Stokes, L.N., Olsen, I. and Dewhirst, F.E. (2005) Defining the normal bacterial flora of the oral cavity J Clin Microbiol 43: 5721-5732. Abreu, M.T., Vora, P., Faure, E., Thomas, L.S ., Arnold, E.T. and Arditi M. (2001) Decreased expression of Toll-like receptor-4 and MD-2 correlates with intestinal epithelial cell protection against dysregulated proinflammatory gene expression in response to bacterial lipopolysaccharide. J Immunol 167 : 1609-1616. Aduse-Opoku, J., Slaney, J.M., Hashim, A., Galla gher, A., Gallagher, R.P., Rangarajan, M., et al (2006) Identification and ch aracterization of the capsula r polysaccharide (K-antigen) locus of Porphyromonas gingivalis. Infect Immun 74 : 449-460. Aguilar-Lemarroy A., Kirchhoff S., Whitaker N., Gariglio P., zur Hausen H., Krammer P. H. and Rosl F. (2001) Differential sensitivity of huma n papillomavirus type 16(+) and type 18(+) cervical carcinoma cells to CD95-mediated apoptosis. Int J Cancer 93 :823-831. Akira, S. and Hemmi, H. (2003) Recognition of pathogen-associated molecular patterns by TLR family. Immunol Lett 85 : 85-95. Akopyants, N.S., Clifton, S.W., Kersulyte, D., Crabtree, J.E., Youree, B.E., Reece, C.A., et al (1998) Analyses of the ca g pathogenicity island of Helicobacter pylori Mol Microbiol 28 : 37-53. Aldridge, P.D., Gray, M.A., Hirst, B.H. a nd Khan, C.M. (2005) Who's talking to whom? Epithelial-bacterial pathogen interactions Mol Microbiol 55 : 655-663. Alves, R.T. and Ribeiro, R.A. (2006) Relations hip between maternal periodontal disease and birth of preterm low weight babies. Pesqui Odontol Bras 20 : 318-323. Amano, A., Kuboniwa, M., Nakagawa, I., Akiyama, S., Morisaki, I. and Hamada, S. (2000) Prevalence of specific genotypes of Porphyromonas gingivalis fimA and periodontal health status. J Dent Res 79 : 1664-1668. Amano, A., Sharma, A., Lee, J.Y., Sojar, H.T., Raj, P.A. and Genco, R.J. (1996) Structural domains of Porphyromonas gingivalis recombinant fimbrillin that mediate binding to salivary proline-rich protein and statherin. Infect Immun 64 : 1631-1637. Armitage, G.C. (2002) Classifying periodont al diseases--a long -standing dilemma. Periodontol 2000 30 : 9-23. Artis, D., Villarino, A., Silverman, M., He, W ., Thornton, E.M., Mu, S., Summer, S., Covey, T.M., Huang, E., Yoshida, H., Koretzky, G., Goldschmidt, M., Wu, G.D., de Sauvage, F., Miller, H.R., Saris, C.J., Scott, P., and H unter, C.A. (2004) The IL-27 receptor (WSX-1) is an inhibitor of innate and adap tive elements of type 2 immunity. J Immunol 173 : 562634.

PAGE 199

199 Asakawa, R., Komatsuzawa, H., Kawai, T., Yamada, S., Goncalves, R.B., Izumi, S., et al (2003) Outer membrane protein 100, a ve rsatile virulence factor of Actinobacillus actinomycetemcomitans. Mol Microbiol 50 : 1125-1139. Asikainen, S., Lai, C.H., Alaluusua, S. and Slots, J. (1991) Distribution of Actinobacillus actinomycetemcomitans serotypes in periodontal health and disease. Oral Microbiol Immunol 6 : 115-118. Ayabe, T., Ashida, T., Kohgo, Y. and Kono, T. (2004) The role of Paneth cells and their antimicrobial peptides in innate host defense. Trends Microbiol 12 : 394-398. Ayala, P., Vasquez, B., Wetzler, L. and So, M. (2002) Neisseria gonorrhoeae porin P1.B induces endosome exocytosis and a redistributi on of Lamp1 to the plasma membrane. Infect Immun 70 : 5965-5971. Bach, S., Makristathis, A., Rotter, M. and Hirs chl, A.M. (2002) Gene expression profiling in AGS cells stimulated with Helicobacter pylori isogenic strains (cagA positive or cagA negative). Infect Immun 70 : 988-992. Backhed, F., Ding, H., Wang, T., Hooper, L.V., Koh, G.Y., Nagy, A., et al (2004) The gut microbiota as an environmental fact or that regulates fat storage. Proc Natl Acad Sci U S A 101 : 15718-15723. Bard, J.B.L. and Rhee, S.Y. (2004) Ontologies in biology: Design, applications and future challenges. Nature Reviews Genetics 5 : 213-222. Barsyte-Lovejoy, D., Mao, D.Y., and Penn, L.Z. (2004) c-Myc represses the proximal promoters of GADD45a and GADD153 by a post-RNA polymerase II recruitment mechanism. Oncogene 23 : 3481-3486. Beck J. D., Eke P., Heiss G., Madianos P., Coupe r D., Lin D., Moss K., Elter J. and Offenbacher S. (2005) Periodontal disease a nd coronary heart disease: a reappraisal of the exposure. Circulation 112 :19-24. Bekal, S., Brousseau, R., Masson, L., Prefontaine, G ., Fairbrother, J. and Harel, J. (2003) Rapid identification of Escherichia coli pathotypes by virulence gene detection with DNA microarrays. J Clin Microbiol 41 : 2113-2125. Belcher, C.E., Drenkow, J., Kehoe, B., Gingeras, T.R., McNamara, N., Lemjabbar, H., et al (2000) The transcriptional responses of respiratory ep ithelial cells to Bordetella pertussis reveal host defensive and pathoge n counter-defensive strategies. Proc Natl Acad Sci U S A 97 : 13847-13852.

PAGE 200

200 Belibasakis G. N., Johansson A., Wang Y., Chen C ., Lagergard T., Kalfas S. and Lerner U. H. (2005) Cytokine responses of human gingival fibroblasts to Actinobacillus actinomycetemcomitans cytolethal distending toxin. Cytokine 30 :56-63. Belibasakis, G., Johansson, A., Wang, Y., Claess on, R., Chen, C., Asikainen, S., and Kalfas, S. (2002) Inhibited proliferation of human periodontal ligament cells and gingival fibroblasts by Actinobacillus actinomycetemcomitans : involvement of the cytolethal distending toxin. Eur J Oral Sci 110 : 366-73. Belton, C.M., Goodwin, P.C., Fatherazi, S., Schu bert, M.M., Lamont, R.J., and Izutsu, K.T. (2004) Calcium oscillations in gingival epithelial cells infected with Porphyromonas gingivalis Microbes Infect 6 : 440-447. Bergelson, J., Kreitman, M., Stahl, E.A. and Ti an, D. (2001) Evolutionary dynamics of plant Rgenes. Science 292 : 2281-2285. Binnicker, M.J., Williams, R.D. and Apicella, M.A. (2003) Infection of human urethral epithelium with Neisseria gonorrhoeae elicits an upregulation of host anti-apoptotic factors and protects cells from staurosporine-induced apoptosis. Cell Microbiol 5 : 549560. Bohn, E., Muller, S., Lauber, J., Geffers, R., Speer, N., Spieth, C., et al (2004) Gene expression patterns of epithelial cells modulated by pat hogenicity factors of Yersinia enterocolitica. Cell Microbiol 6 : 129-141. Borrell, L.N. and Papapanou, P.N. (2005) An alytical epidemiology of periodontitis. J Clin Periodontol 32 Suppl 6 : 132-158. Brissette, C.A. and Fi ves-Taylor, P.M. (1999) Actinobacillus acti nomycetemcomitans may utilize either actin-dependent or actin-i ndependent mechanisms of invasion. Oral Microbiol Immunol 14 : 137-142. Brodala, N., Merricks, E.P., Bellinger, D.A ., Damrongsri, D., Offenbacher, S., Beck, J., et al (2005) Porphyromonas gingivalis bacteremia induces coronary and aortic atherosclerosis in normocholesterolemic and hypercholesterolemic pigs. Arterioscler Thromb Vasc Biol 25 : 1446-1451. Brodsky, L.I., Wahed, A.S., Li, J., Tavis, J.E., Tsukahara, T. and Taylor M.W. (2006) A Novel Unsupervised Method to Identify Genes Important in the Anti-viral Response: Application to Interferon/Ribav irin in Hepatitis C Patients. PLoS ONE 1 : e584. Brogan, J.M., Lally, E.T., Poulsen, K., Kilian, M. and Demuth, D.R. (1994) Regulation of Actinobacillus actinomycetemcomitans leukotoxin expression: analysis of the promoter regions of leukotoxic and mi nimally leukotoxic strains Infect Immun 62: 501-508.

PAGE 201

201 Brozovic, S., Sahoo, R., Barve, S., Shiba, H., Uriarte, S., Blumberg, R.S., et al (2006) Porphyromonas gingivalis enhances FasL expression vi a up-regulation of NF{kappa}Bmediated gene transcription and induces apopt otic cell death in human gingival epithelial cells. Microbiology 152 : 797-806. Brunet, A., Park, J., Tran, H., Hu, L.S., Hemm ings, B.A., and Greenberg, M.E. (2001) Protein kinase SGK mediates survival signals by phosphorylating the forkhead transcription factor FKHRL1 (FOXO3a). Mol Cell Biol 21 : 952-965. Bryant, P.A., Venter, D., Robins-Browne, R. a nd Curtis, N. (2004) Chips with everything: DNA microarrays in infectious diseases. Lancet Infect Dis 4 : 100-111. Buness, A., Kuner, R., Ruschhaupt, M., Poustk a, A., Sultmann, H. and Tresch, A. (2007) Identification of aberrant chromosomal regi ons from gene expression microarray studies applied to human breast cancer. Bioinformatics Cao, S.L., Progulske-Fox, A., Hillman, J.D. and Handfield, M. (2004) In vivo induced antigenic determinants of Actinobacillus actinomycetemcomitans. FEMS Microbiol Lett 237 : 97103. Celenligil, H. and Ebersole, J.L. (1998) Analysis of serum antibody responses to periodontopathogens in early-onset periodontit is patients from different geographical locations. J Clin Periodontol 25 : 994-1002. Censini, S., Lange, C., Xiang, Z., Crab tree, J.E., Ghiara, P., Borodovsky, M., et al (1996) cag, a pathogenicity island of Helicobacter pylori encodes type I-specific and diseaseassociated virulence factors. Proc Natl Acad Sci U S A 93 : 14648-14653. Chen, Z., Potempa, J., Polanowski, A., Wikstrom M. and Travis, J. (1992) Purification and characterization of a 50-kDa cystei ne proteinase (gingipain) from Porphyromonas gingivalis J Biol Chem 267 : 18896-18901. Cheng, C. and Pounds, S. (2007) False discovery rate paradigms for statistical analyses of microarray gene expression data Bioinformation 1: 436-446. Chiou, C.C., Chan, C.C., Sheu, D.L., Che n, K.T., Li, Y.S. and Chan, E.C. (2001) Helicobacter pylori infection induced alteration of gene expression in human gastric cells. Gut 48 : 598604. Christersson, L.A.(1993) Actinobacillus acti nomycetemcomitans and localized juvenile periodontitis. Clinical, microbiologic and hi stologic studies. Swed Dent J Suppl 90 : 1-46. Christersson, L.A., Albini, B., Zambon, J.J., Wikesjo, U.M., and Genco R.J. (1987) Tissue localization of Actinobacillus actinomycetemcomitans in human periodontitis. I. Light, immunofluorescence and elec tron microscopic studies. J Periodontol 58 : 529-39.

PAGE 202

202 Christersson, L.A., Wikesjo, U.M., Albini, B ., Zambon, J.J., and Genco, R.J. (1987) Tissue localization of Actinobacillus actinomycetemcomitans in human periodontitis. II. Correlation between immunofluores cence and culture techniques. J Periodontol 58 : 5405. Chun, C.K., Ozer, E.A., Welsh, M.J., Zabner, J. and Greenberg, E.P. (2004) Inactivation of a Pseudomonas aeruginosa quorum-sensing signal by human airway epithelia. Proc Natl Acad Sci U S A 101 : 3587-3590. Contreras, A., Herrera, J.A., Soto, J.E., Arce, R.M., Jaramillo, A. and Botero, J.E. (2006) Periodontitis is associated with preeclampsia in pregnant women. J Periodontol 77 : 182188. Cory, S. and Adams, J.M. (2002) The Bcl2 family: regulators of th e cellular life-or-death switch. Nat Rev Cancer 2 : 647-656. Cossart P. and Sansonetti P. J. (2004) Bacteria l invasion: the paradigms of enteroinvasive pathogens. Science 304 :242-248. Cotter, S.E., Surana, N.K. and St Geme, J.W.,3r d. (2005) Trimeric autotr ansporters: a distinct subfamily of autotransporter proteins. Trends Microbiol 13 : 199-205. Cox, J.M., Clayton, C.L., Tomita, T., Wallace, D. M., Robinson, P.A. and Crabtree, J.E. (2001) cDNA array analysis of cag pa thogenicity island-associated Helicobacter pylori epithelial cell response genes. Infect Immun 69 : 6970-6980. Crosby, J.A. and Kachlany, S.C. (2007) TdeA, a TolC-like protein requi red for toxin and drug export in Aggregatibacter (Actinobacillus) actinomycetemcomitans Gene 388 : 83-92. Cummings, C.A. and Relman, D.A. (2000) Us ing DNA microarrays to study host-microbe interactions. Emerg Infect Dis 6 : 513-525. Dahlen, G., Gmur, R. and Yoshino, T. (2007) Phe notypes, serotypes and an tibiotic susceptibility of Swedish Porphyromonas gingivalis isolates from periodontitis and periodontal abscesses. Oral Microbiol Immunol 22 : 80-86. Dale, B.A. (2002) Periodontal epithelium: a ne wly recognized role in health and disease. Periodontol 2000 30 : 70-78. Darveau,R., Belton,C.M., Reife,R., and Lamont,R.J (1998) Local chemokine paralysis: a novel pathogenic mechanism for Porphyromonas gingivalis Infect Immun 66 : 1660 1665. Dasanayake, A.P. (1998) Poor peri odontal health of the pregnant woman as a risk factor for low birth weight. Ann Periodontol 3 : 206-212.

PAGE 203

203 Dasanayake, A.P., Boyd, D., Madianos, P.N ., Offenbacher, S. and Hills, E. (2001) The association between Porphyromonas gingivalis -specific maternal serum IgG and low birth weight. J Periodontol 72 : 1491-1497. de Hostos, E.L. (1999) The coronin family of actin-associated proteins. Trends Cell Biol 9 : 345350. Demmer, R.T. and Desvarieux, M. (2006) Periodontal infections a nd cardiovascular disease: the heart of the matter. J Am Dent Assoc 137 Suppl: 14S-20S; quiz 38S. Demuth, D.R., James, D., Kowashi, Y ., and Kato, S. (2003) Interaction of Actinobacillus actinomycetemcomitans outer membrane vesicles with HL60 cells does not require leukotoxin. Cell Microbiol 5 : 111-21. Denecker, G., Totemeyer, S., Mota, L.J., Troi sfontaines, P., Lambermont, I., Youta, C., et al (2002) Effect of lowand high-virulence Yersinia enterocolitica strains on the inflammatory response of human um bilical vein endothelial cells. Infect Immun 70 : 35103520. Deshpande, R.G., Khan, M.B. and Genco, C.A. ( 1998) Invasion of aortic and heart endothelial cells by Porphyromonas gingivalis Infect Immun 66 : 5337-5343. Dessus-Babus, S., Knight, S.T. and Wyrick, P.B. (2000) Chlamydial infection of polarized HeLa cells induces PMN chemotaxis but the cytoki ne profile varies be tween disseminating and non-disseminating strains. Cell Microbiol 2 : 317-327. Dharmadi, Y. and Gonzalez, R. (2004) DNA microarr ays: experimental issu es, data analysis, and application to bacterial systems. Biotechnol Prog 20 : 1309-1324. Diaz, R., Ghofaily, L.A., Patel, J., Ba lashova, N.V., Freitas, A.C., Labib, I., et al (2006) Characterization of leukotoxin from a clinical strain of Actinobacillus actinomycetemcomitans Microb Pathog 40 : 48-55. Dickinson, D.P., Kubiniec, M.A., Yoshimura, F. and Genco, R.J. (1988) Molecular cloning and sequencing of the gene encoding the fimbrial subunit protein of Bacteroides gingivalis J Bacteriol 170: 1658-1665. Dileepan, T., Kachlany, S.C., Balashova, N.V., Pa tel, J. and Maheswar an, S.K. (2007) Human CD18 is the functional receptor for Aggregat ibacter actinomycetemcomitans leukotoxin Infect Immun. Doring, G. and Worlitzsch, D. (2000) Inflammati on in cystic fibrosis and its management. Paediatr Respir Rev 1 : 101-106.

PAGE 204

204 Ebersole, J.L., Kraig, E., Bauman, G., Spitznage l, J.K., and Kolodrubetz, D. (1990) Molecular approaches to leukotoxin as a virulence component in Actinobacillus actinomycetemcomitans Arch Oral Biol 35 : 69S-78S. Eckmann, L., Smith, J.R., Housley, M.P., Dwinell, M.B. and Kagnoff, M.F. (2000) Analysis by high density cDNA arrays of a ltered gene expression in human intestinal ep ithelial cells in response to infection with the invasive enteric bacteria Salmonella. J Biol Chem 275 : 14084-14094. Eisen M. B., Spellman P. T., Brown P. O. and Bo tstein D. (1998) Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95 :14863-14868. Ellis, J.S., Averley, P.A., Preshaw, P.M., Steele, J.G., Seymour, R.A. and Thomason, J.M. (2007) Change in cardiovascular risk status after dental clearance. Br Dent J 202 : 543544. Fan, T., Lu, H., Hu, H., Shi, L., McClarty, G.A., Nance, D.M., et al (1998) Inhibition of apoptosis in chlamydia -infected cells: blockade of m itochondrial cytochrome c release and caspase activation. J Exp Med 187 : 487-496. Feezor R. J., Oberholzer C., Baker H. V., Novick D., Rubinstein M., Moldawer L. L., Pribble J., Souza S., Dinarello C. A., Ertel W. and Ob erholzer A. (2003) Molecular characterization of the acute inflammatory response to infec tions with gram-negative versus gram-positive bacteria. Infect Immun 71 :5803-5813. Feng, Y., and Walsh, C.A. (2004) The many faces of filamin: a versatile mo lecular scaffold for cell motility and signalling. Nat Cell Biol 6 : 1034-1038. Ferran C, Stroka DM, Badrichani AZ, Cooper JT Wrighton CJ, Soares M, Grey ST, and Bach FH. (1998) A20 inhibits NF-ka ppaB activation in endothelial cells without sensitizing to tumor necrosis factor-mediated apoptosis. Blood 91 : 2249-58 Fine, D.H., Kaplan, J.B., Kachlany, S.C. and Schr einer, H.C. (2006) How we got attached to Actinobacillus acti nomycetemcomitans : A model for infectious diseases. Periodontol 2000 42 : 114-157. Fine, D.H., Velliyagounder, K., Furgang, D. and Kaplan, J.B. (2005) The Actinobacillus actinomycetemcomitans autotransporter adhesin Aae e xhibits specificity for buccal epithelial cells from humans and old world primates. Infect Immun 73 : 1947-1953. Fink, D.L., Buscher, A.Z., Green, B., Fernst en, P. and St Geme, J.W.,3rd. (2003) The Haemophilus influenzae Hap autotransporter mediat es microcolony formation and adherence to epithelial cells and extracellular matrix via binding regions in the C-terminal end of the passenger domain. Cell Microbiol 5 : 175-186.

PAGE 205

205 Fink, D.L., Cope, L.D., Hansen, E. J. and Geme, J.W.,3rd. (2001) The Haemophilus influenzae Hap autotransporter is a chymotrypsin clan serine protease and unde rgoes autoproteolysis via an intermolecular mechanism. J Biol Chem 276 : 39492-39500. Finlay, B.B. and Cossart, P. (1997) Exploitati on of mammalian host cell functions by bacterial pathogens. Science 276 : 718-725. Fischer, S.F., Schwarz, C., Vier, J. and Hack er, G. (2001) Characterization of antiapoptotic activities of Chlamydia pneumoniae in human cells. Infect Immun 69 : 7121-7129. Fives-Taylor, P., Meyer, D. and Mi ntz, K. (1995) Characteristics of Actinobacillus actinomycetemcomitans invasion of and adhesion to cultured epithelial cells. Adv Dent Res 9 : 55-62. Fong, K.P., Chung, W.O., Lamont, R.J. and Demu th, D.R. (2001) Intraand interspecies regulation of gene expression by Actinobacillus actinomycetemcomitans LuxS. Infect Immun 69 : 7625-7634. Fong, K.P., Pacheco, C.M., Otis, L.L., Baranwal, S., Kieba, I.R., Harrison, G., et al. (2006) Actinobacillus actinomycetemcomitans leukotox in requires lipid microdomains for target cell cytotoxicity Cell Microbiol 8: 1753-1767. Fong, K.P., Gao, L. and Demuth, D.R. (2003) lu xS and arcB control aerobic growth of Actinobacillus acti nomycetemcomitans under iron limitation. Infect Immun 71 : 298-308. Galan, J.E., and Zhou, D. (2000) Striking a bala nce: modulation of th e actin cytoskeleton by Salmonella Proc Natl Acad Sci U S A 97 : 8754-8761. Galindo, C.L., Fadl, A.A., Sha, J., Pillai, L., Gutie rrez, C.,Jr and Chopra, A.K. (2005) Microarray and proteomics analyses of human intestinal epithelial cells treated with the Aeromonas hydrophila cytotoxic enterotoxin Infect Immun 73: 2628-2643. Garbom S., Forsberg A., Wolf-Watz H. and Ki hlberg B. M. (2004) Identification of novel virulence-associated genes via geno me analysis of hypothetical genes. Infect Immun 72 :1333-1340. Genco R., Offenbacher S. and Beck J. (2002) Pe riodontal disease and card iovascular disease: epidemiology and possible mechanisms. J Am Dent Assoc 133 Suppl:14S-22S. Genco, C.A. (1995) Regulation of hemin and iron transport in Porphyromonas gingivalis Adv Dent Res 9 : 41-47. Geng, Y., Shane, R.B., Berencsi, K., Gon czol, E., Zaki, M.H., Margolis, D.J., et al (2000) Chlamydia pneumoniae inhibits apoptosis in human peripheral blood mononuclear cells through induction of IL-10. J Immunol 164 : 5522-5529.

PAGE 206

206 Gewirtz, A.T., Navas, T.A., Lyons, S., Godowski, P.J. and Madara, J.L. (2001) Cutting edge: bacterial flagellin activates basolaterally expressed TLR5 to induce epithelial proinflammatory gene expression. J Immunol 167 : 1882-1885. Gilthorpe, M.S., Zamzuri, A.T., Griffiths, G. S., Maddick, I.H., Eaton, K.A. and Johnson, N.W. (2003) Unification of the "burst and "linear" theories of pe riodontal disease progression: a multilevel manifestation of the same phenomenon. J Dent Res 82 : 200-205. Gmur, R., McNabb, H., van Steenbergen, T.J., Bae hni, P., Mombelli, A., van Winkelhoff, A.J., et al (1993) Seroclassification of hitherto nontypeable Actinobacillus actinomycetemcomitans strains: evidence for a new serotype e. Oral Microbiol Immunol 8 : 116-120. Guillemin, K., Salama, N.R., Tompkins, L.S. and Falkow, S. (2002) Cag pathogenicity islandspecific responses of gastric epithelial cells to Helicobacter pylori infection. Proc Natl Acad Sci U S A 99 : 15136-15141. Gururajan, R., Lahti, J.M., Grenet J., Easton, J., Gruber, I., Ambr os, P.F., and Kidd, V.J. (1998) Duplication of a genomic region contai ning the Cdc2L1-2 and MMP21-22 genes on human chromosome 1p36.3 and their linkage to D1Z2. Genome Res 8 : 929-939. Haffajee, A.D. and Socransky, S.S. (1994) Mi crobial etiological ag ents of destructive periodontal diseases. Periodontol 2000 5 : 78-111. Haffajee, A.D., Bogren, A., Hasturk, H., Feres, M., Lopez, N.J. and Socransky, S.S. (2004) Subgingival microbiota of ch ronic periodontitis subjects from different geographic locations. J Clin Periodontol 31 : 996-1002. Hamada, N., Sojar, H.T., Cho, M.I. and Genco, R. J. (1996) Isolation and characterization of a minor fimbria from Porphyromonas gingivalis. Infect Immun 64 : 4788-4794. Hamada, N., Watanabe, K., Sasakawa, C., Yosh ikawa, M., Yoshimura, F. and Umemoto, T. (1994) Construction and characteri zation of a fimA mutant of Porphyromonas gingivalis Infect Immun 62 : 1696-1704. Han, N., Whitlock, J. and Progulske-Fox, A. (1996) The hemagglutinin gene A (hagA) of Porphyromonas gingivalis 381 contains four large, contiguous, direct repeats. Infect Immun 64 : 4000-4007. Handfield, M., Brady, L.J., Progulske-Fox, A. a nd Hillman, J.D. (2000) IVIAT: a novel method to identify microbial gene s expressed specifically during human infections. Trends Microbiol 8 : 336-339. Handfield, M. and Hillman, J.D. (2006) In vivo induced antigen technology (IVIAT) and change mediated antigen technology (CMAT) Infect Disord Drug Targets 6: 327-334.

PAGE 207

207 Handfield, M., Mans, J.J., Zheng, G., Lopez, M.C., Mao, S., Progulske-Fox, A., et al (2005) Distinct transcriptional profiles characterize oral epithelium-microbiota interactions. Cell Microbiol 7 : 811-823. Handfield, M., Progulske-Fox, A. and Hillman, J.D. (2005b) In vivo induced genes in human diseases. Periodontol 2000 38 : 123-134. Handfield, M., Seifert, T ., and Hillman, J.D. (2002) In vivo expression of bact erial genes during human infections. Methods Mol Med 71 : 225-42. Hapfelmeier, S. and Hardt, W.D. (2005) A mouse model for S. typhimurium -induced enterocolitis. Trends Microbiol 13 : 497-503. Hardwidge P. R., Rodriguez-Escudero I., G oode D., Donohoe S., Eng J., Goodlett D. R., Aebersold R. and Finlay B. B. (2004) Proteomic analysis of the intestinal epithelial cell response to enteropathogenic Escherichia coli J Biol Chem 279 :20127-20136. Hasegawa, Y., Mans, J.J., Mao, S., Lopez, M.C., Baker, H.V., Handfield, M., et al (2007) Gingival epithelial cell transcriptional re sponses to commensal and opportunistic oral microbial species. Infect Immun 75 : 2540-2547. Haskill, S., Beg, A.A., Tompkins, S.M., Morris J.S., Yurochko, A.D., Sampson-Johannes, A., Mondal, K., Ralph, P., and Baldwin, A.S. Jr. (1991) Characterization of an immediateearly gene induced in adherent monocyt es that encodes I kappa B-like activity. Cell 65 : 1281-9. Haubek, D., Poulsen, K. and Kilian, M. (2007) Micr oevolution and Patterns of Dissemination of the JP2 Clone of Aggregatibacter (Actinobacil lus) actinomycetemcomitans. Infect Immun 75 : 3080-3088. Hauck, C.R. and Meyer, T.F. (1997) The lysosoma l/phagosomal membrane protein h-lamp-1 is a target of the IgA1 protease of Neisseria gonorrhoeae FEBS Lett 405 : 86-90. Hedges, S.R., Mayo, M.S., Kallman, L., Mesteck y, J., Hook, E.W.,3rd and Russell, M.W. (1998) Evaluation of immunoglobulin A1 (IgA1) protea se and IgA1 proteaseinhibitory activity in human female genital infection with Neisseria gonorrhoeae Infect Immun 66 : 58265832. Henderson, B., Nair, S.P., Ward, J.M. and Wilson, M. (2003) Molecular pat hogenicity of the oral opportunistic pathogen Actinobacillus actinomycetemco mitans. Annu Rev Microbiol 57 : 29-55. Henderson, I.R. and Nataro, J.P. (2001) Virule nce functions of auto transporter proteins. Infect Immun 69 : 1231-1243.

PAGE 208

208 Hess, S., Rheinheimer, C., Tidow, F., Bartling, G., Kaps, C., Lauber, J., et al (2001) The reprogrammed host: Chlamydia trachomatis -induced up-regulation of glycoprotein 130 cytokines, transcription fact ors, and antiapoptotic genes. Arthritis Rheum 44 : 2392-2401. Hiroi, M., Shimojima, T., Kashimata, M., Miyata, T., Takano, H., Takahama, M., et al (1998) Inhibition by Porphyromonas gingivalis LPS of apoptosis inducti on in human peripheral blood polymorphonuclear leukocytes. Anticancer Res 18 : 3475-3479. Hoffler, U., Niederau, W. and Pulverer, G. (1980) Susceptibility of Bacterium actinomycetem comitans to 45 antibiotics. Antimicrob Agents Chemother 17 : 943-946. Hooper, L.V. and Gordon, J.I. (2001) Commensa l Host-Bacterial Relationships in the Gut. Science 292 : 1115-1118. Hooper, L.V., Wong, M.H., Thelin, A., Hanss on, L., Falk, P.G. and Gordon, J.I. (2001) Molecular analysis of commensal host-micr obial relationships in the intestine. Science 291 : 881-884. Hopper, S., Vasquez, B., Merz, A., Clary, S., W ilbur, J.S. and So, M. (2000) Effects of the immunoglobulin A1 protease on Neisseria gonorrhoeae trafficking across polarized T84 epithelial monolayers. Infect Immun 68 : 906-911. Huang, Q., Liu, D., Majewski, P., Schulte, L.C. Korn, J.M., Young, R.A., Lander, E.S., and Hacohen, N. (2001) The plasticity of dendri tic cell responses to pathogens and their components. Science 294 : 870-5. Huang, S.H., Triche, T. and J ong, A.Y. (2002) Infectomics: genomics and proteomics of microbial infections. Funct Integr Genomics 1 : 331-344. Hueck, C.J. (1998) Type III protein secretion sy stems in bacterial pathogens of animals and plants. Microbiol Mol Biol Rev 62 : 379-433. Hunter, C.A., Villarino, A., Artis, D. and Scott, P. (2004) The role of IL-27 in the development of T-cell responses durin g parasitic infections. Immunol Rev 202 : 106-114. Hunter, C.A., Villarino, A., Artis, D ., and Scott, P. (2004) The role of IL-27 in the development of T-cell responses durin g parasitic infections. Immunol Rev 202 : 106-14. Ichikawa, J.K., Norris, A., Bangera, M.G., Geiss, G.K., van 't Wout, A.B., Bumgarner, R.E., et al (2000) Interaction of Pseudomonas aeruginosa with epithelial cells: identification of differentially regulated genes by expressi on microarray analysis of human cDNAs. Proc Natl Acad Sci U S A 97 : 9659-9664. Ichimura, M., Hiratsuka, K., Ogura, N., Utsunomiya, T., Sakamaki, H., Kondoh, T., et al (2006) Expression profile of chemokines and chemokine receptors in epithelial cell layers of oral lichen planus. J Oral Pathol Med 35 : 167-174.

PAGE 209

209 Imamura, T. (2003) The role of gingipains in the pathogenesis of periodontal disease. J Periodontol 74 : 111-118. Irmler, M., Thome, M., Hahne, M., Schneider, P., Hofmann, K., Steiner, V., Bodmer, J.L., Schroter, M., Burns, K., Mattmann, C., Rimold i, D., French, L.E., and Tschopp, J. (1997) Inhibition of death receptor signals by cellular FLIP. Nature 388 : 190-5. James, C.E., Hasegawa, Y., Park, Y., Yeung, V., Tribble, G.D., Kuboniwa, M., et al (2006) LuxS involvement in the regulation of genes coding for hemin and iron acquisition systems in Porphyromonas gingivalis Infect Immun 74 : 3834-3844. James, J.A., Poulton, K.V., Haworth, S.E ., Payne, D., McKay, I.J., Clarke, F.M., et al (2007) Polymorphisms of TLR4 but not CD14 are associ ated with a decreased risk of aggressive periodontitis. J Clin Periodontol 34 : 111-117. Jenner, R.G. and Young, R.A. (2005) Insights into host responses against pathogens from transcriptional profiling. Nat Rev Microbiol 3 : 281-294. Johannsen, D.B., Johnston, D.M., Koymen, H. O., Cohen, M.S. and Cannon, J.G. (1999) A Neisseria gonorrhoeae immunoglobulin A1 protease mutant is infectious in the human challenge model of urethral infection. Infect Immun 67 : 3009-3013. Jong, A. and Huang, S.H. (2005) Blood-brain barr ier drug discovery for central nervous system infections. Curr Drug Targets Infect Disord 5 : 65-72. Kachlany, S.C., Fine, D.H. and Figurski, D. H. (2000) Secretion of RTX leukotoxin by Actinobacillus actinomycetemcomitans Infect Immun 68: 6094-6100. Kachlany, S.C., Planet, P.J., DeSalle, R., Fine, D. H. and Figurski, D.H. (2001) Genes for tight adherence of Actinobacillus actinomycetemcomitans : from plaque to plague to pond scum. Trends Microbiol 9 : 429-437. Kaczmarek, F.S. and Coykendall, A.L. (1980) Production of phenylacetic acid by strains of Bacteroides asaccharolyticus and Bacteroides gingivalis (sp. nov.). J Clin Microbiol 12 : 288-290. Kagnoff, M.F., and Eckmann, L. (1997) Epithelia l cells as sensors for microbial infection. J Clin Invest 100: 6-10. Kang, P., Korostoff, J., Volgina, A., Grzesik, W. a nd DiRienzo, J.M. (2005) Differential effect of the cytolethal distending toxin of Actinobacillus acti nomycetemcomitans on co-cultures of human oral cells. J Med Microbiol 54 : 785-794.

PAGE 210

210 Kanno, F., Korostoff, J., Volgina, A. and Dirien zo, J.M. (2005) Resistance of Human Periodontal Ligament Fibroblasts to the Cy tolethal Distending Toxin of Actinobacillus actinomycetemcomitans J Periodontol 76 : 1189-1201. Kao, H.T., Huang, Y.C. and Lin, T.Y. (2003) Fatal bacteremic pneumonia caused by Aeromonas hydrophila in a previously healthy child. J Microbiol Immunol Infect 36 : 209-211. Kaplan, J.B., Perry, M.B., MacLean, L.L., Furg ang, D., Wilson, M.E. and Fine, D.H. (2001) Structural and genetic analyses of O polysaccharide from Actinobacillus actinomycetemcomitans serotype f. Infect Immun 69 : 5375-5384. Kaplan, J.B., Velliyagounder, K., Ragunath, C., Rohde, H., Mack, D., Knobloch, J.K., et al (2004) Genes involved in the synthesis and degradation of matrix polysaccharide in Actinobacillus acti nomycetemcomitans and Actinobacillus pleuropneumoniae biofilms. J Bacteriol 186 : 8213-8220. Kato, S., Nakashima, K., Inoue, M., Tomioka, J., Nonaka, K., Nishihara, T., and Kowashi, Y. (2000) Human epithelial cell death caused by Actinobacillus actinomycetemcomitans infection. J Med Microbiol 49 : 739-45. Kato-Maeda, M., Gao, Q. and Small, P.M. (2001) Microarray analysis of pathogens and their interaction with hosts. Cell Microbiol 3 : 713-719. Kaufmann, S.H. and Hess, J. (1999) Impact of in tracellular loca tion of and antigen display by intracellular bacteria: implications for vaccine development. Immunol Lett 65 : 81-84. Kazor, C.E., Mitchell, P.M., Lee, A.M., Stokes, L.N., Loesche, W.J., Dewhirst, F.E., et al. (2003) Diversity of bacterial populations on th e tongue dorsa of patie nts with halitosis and healthy patients J Clin Microbiol 41: 558-563. Kellam, P. (2000) Host-pathogen stud ies in the post-genomic era. Genome Biol 1 : REVIEWS1009. Kellam, P. (2001) Post-genomic virology: th e impact of bioinformatics, microarrays and proteomics on investigating hos t and pathogen interactions. Rev Med Virol 11 : 313-329. Kellam, P. and Weiss, R.A. (2006) Infectoge nomics: insights from the host genome into infectious diseases. Cell 124 : 695-697. Khatri, P., Sellamuthu, S., Malhotra, P., Amin, K., Done, A. and Draghici, S. (2005) Recent additions and improvements to the Onto-Tools. Nucleic Acids Res 33 : W762-5. Kim, J.G., Lee, S.J. and Kagnoff, M.F. (2004) Nod1 is an essential signal tr ansducer in intestinal epithelial cells infected with bacteria that avoid recogni tion by toll-like receptors. Infect Immun 72 : 1487-1495.

PAGE 211

211 Kinane, D.F., Mooney, J. and Ebersole, J. L. (1999) Humoral immune response to Actinobacillus actinomycetemcomitans and Porphyromonas gingivalis in periodontal disease. Periodontol 2000 20 : 289-340. Kinane, D.F., Shiba, H. and Hart, T.C. (2005) The genetic basis of periodontitis. Periodontol 2000 39 : 91-117. Kirmizis, A. and Farnham, P.J. (2004) Genomic approaches that aid in the identification of transcription factor target genes. Exp Biol Med (Maywood) 229 : 705-721. Komatsuzawa, H., Asakawa, R., Kawai, T ., Ochiai, K., Fujiwara, T., Taubman, M.A., et al (2002) Identification of six majo r outer membrane proteins from Actinobacillus actinomycetemcomitans. Gene 288 : 195-201. Korostoff, J., Wang, J.F., Kieba, I., Miller M., Shenker, B.J., and Lally, E.T. (1998) Actinobacillus acti nomycetemcomitans leukotoxin induces apoptosis in HL-60 cells. Infect Immun 66 : 4474-83. Kreuz, S., Siegmund, D., Rumpf, J.J., Samel, D ., Leverkus, M., Janssen, O., Hacker, G., DittrichBreiholz, O., Kracht, M., Scheurich, P., and Wajant, H. (2004) NFkappaB activation by Fas is mediated through FADD, caspase8, and RIP and is inhibited by FLIP. J Cell Biol 166 : 369-80. Kumar, P.S., Griffen, A.L., Bart on, J.A., Paster, B.J., Moeschberg er, M.L. and Leys, E.J. (2003) New bacterial species associat ed with chronic periodontitis J Dent Res 82: 338-344. Laarmann, S., Cutter, D., Juehne, T., Barenka mp, S.J. and St Geme, J.W. (2002) The Haemophilus influenzae Hia autotransporter ha rbours two adhesive pockets that reside in the passenger domain and recognize the same host cell receptor. Mol Microbiol 46 : 731743. Laine, M.L., Appelmelk, B.J. and van Winkel hoff, A.J. (1996) Novel polysaccharide capsular serotypes in Porphyromonas gingivalis J Periodontal Res 31 : 278-284. Lally, E.T., Golub, E.E., Kieba, I.R., Taichm an, N.S., Rosenbloom, J., Rosenbloom, J.C., Gibson, C.W., and Demuth, D.R. (1989) Analysis of the Actinobacillus actinomycetemcomitans leukotoxin gene. Delineation of unique features and comparison to homologous toxins. J Biol Chem 264 : 15451-6. Lally, E.T., Hill, R.B., Kieba, I.R. and Korosto ff, J. (1999) The interaction between RTX toxins and target cells Trends Microbiol 7: 356-361. Lally, E.T., Kieba, I.R., Demuth, D.R., Rose nbloom, J., Golub, E.E ., Taichman, N.S., and Gibson, C.W. (1989) Identification and expression of the Actinobacillus actinomycetemcomitans leukotoxin gene. Biochem Biophys Res Commun 159 : 256-62.

PAGE 212

212 Lamont, R.J. and Jenkinson, H.F. (1998) Life be low the gum line: pathogenic mechanisms of Porphyromonas gingivalis. Microbiol Mol Biol Rev 62 : 1244-1263. Lamont, R.J. and Jenkinson, H.F. (2000) Subgingival colonization by Porphyromonas gingivalis. Oral Microbiol Immunol 15 : 341-349. Lamont, R.J., Chan, A., Belton, C.M., Izutsu, K.T., Vasel, D. and Weinberg, A. (1995) Porphyromonas gingivalis invasion of gingival epithelial cells. Infect Immun 63 : 38783885. Lamont, R.J., El-Sabaeny, A., Park, Y., Cook, G. S., Costerton, J.W. and Demuth, D.R. (2002) Role of the Streptococcus gordonii SspB protein in the development of Porphyromonas gingivalis biofilms on streptococcal substrates. Microbiology 148 : 1627-1636. Lepine, G. and Progulske-Fox, A. (1996) D uplication and differential expression of hemagglutinin genes in Porphyromonas gingivalis. Oral Microbiol Immunol 11 : 65-78. Lepine, G., Ellen, R.P. and Progulske-Fox, A. (1996) Construction and preliminary characterization of three hemagglutinin mutants of Porphyromonas gingivalis. Infect Immun 64 : 1467-1472. Lester, A., Bruun, B.G., Husum, P., Kolmos H.J., Nielsen, B.B., Scheibel, J.H., et al (1995) Salmonella Dublin Ugeskr Laeger 157 : 20-24. Lewis, W.H. and Gray, H. (1918) Anatomy of the Human Body 20 : 1396. Li, L., Matevski, D., Aspiras, M., Ellen, R.P. a nd Lepine, G. (2004) Two epithelial cell invasionrelated loci of the oral pathogen Actinobacillus actinomycetemc omitans. Oral Microbiol Immunol 19 : 16-25. Li, X., Kolltveit, K.M., Tronstad, L. and Olsen, I. (2000) Systemic diseases caused by oral infection. Clin Microbiol Rev 13 : 547-558. Lin, L., Ayala, P., Larson, J., Mulks, M., Fukuda, M., Carlsson, S.R., et al (1997) The Neisseria type 2 IgA1 protease cleaves LAMP1 and promot es survival of bacteria within epithelial cells. Mol Microbiol 24 : 1083-1094. Lin, X., Wu, J. and Xie, H. (2006) Porphyromonas gingivalis minor fimbriae are required for cell-cell interactions. Infect Immun 74 : 6011-6015. Loe, H., Anerud, A., Boysen, H. and Smith, M. ( 1978) The natural history of periodontal disease in man. The rate of periodontal de struction before 40 years of age J Periodontol 49: 607620

PAGE 213

213 Loesche, W.J. (1997) Association of the or al flora with important medical diseases. Curr Opin Periodontol 4 : 21-28. Loesche, W.J. and Grossman, N.S. (2001) Period ontal disease as a specific, albeit chronic, infection: diagnosis and treatment. Clin Microbiol Rev 14 : 727-52, table of contents. Loos, B.G., John, R.P. and Laine, M.L. (2005) Identification of gene tic risk factors for periodontitis and possible mechanisms of action. J Clin Periodontol 32 Suppl 6: 159-179. Lory, S., and Ichikawa, J.K. (2002) Pseudomonas -epithelial cell interactions dissected with DNA microarrays. Chest 121 : 36S-39S. Lucchini S., Liu H., Jin Q., Hinton J. C. and Yu J. (2005). Transcri ptional adaptation of Shigella flexneri during infection of macropha ges and epithelial cells: insi ghts into the strategies of a cytosolic bacterial pathogen. Infect Immun 73 :88-102. Maeda, S., Akanuma, M., Mitsuno, Y., Hirata, Y., Ogura, K., Yoshida, H., et al (2001a) Distinct mechanism of Helicobacter pylori -mediated NF-kappa B activation between gastric cancer cells and monocytic cells. J Biol Chem 276 : 44856-44864. Maeda, S., Otsuka, M., Hirata, Y., M itsuno, Y., Yoshida, H., Shiratori, Y., et al (2001b) cDNA microarray analysis of Helicobacter pylori -mediated alteration of gene expression in gastric cancer cells. Biochem Biophys Res Commun 284 : 443-449. Maiden, M.F., Cohee, P. and Tanner, A.C. (2003) Pr oposal to conserve the ad jectival form of the specific epithet in th e reclassification of Bacteroides forsythus Tanner et al 1986 to the genus Tannerella Sakamoto et al 2002 as Tannerella forsythia corrig., gen. nov., comb. nov. Request for an Opinion. Int J Syst Evol Microbiol 53 : 2111-2112. Malek, R., Fisher, J.G., Caleca, A., S tinson, M., van Oss, C.J., Lee, J.Y., et al (1994) Inactivation of the Porphyromonas gingivalis fimA gene blocks periodontal damage in gnotobiotic rats. J Bacteriol 176: 1052-1059. Manger, I.D. and Relman, D.A. (2000) How the host 'sees' pathogens: global gene expression responses to infection. Curr Opin Immunol 12 : 215-8. Mans, J.J., Baker, H.V., Oda, D., Lamont, R.J. and Handfield, M. (2006) Distinctive characteristics of transcriptional profiles fr om two epithelial cell lines upon interaction with Actinobacillus actinomycetemcomitans. Oral Microbiol Immunol 21 : 261-267. Mans, J.J., Lamont, R.J. and Handfield, M. (2006 ) Microarray analysis of human epithelial cell responses to bact erial interaction. Infect Disord Drug Targets 6 : 299-309. Marsh, P.D. (1994) Microbial ecolog y of dental plaque and its signi ficance in health and disease Adv Dent Res 8: 263-271.

PAGE 214

214 Matsumura, H., Ito, A., Saitoh, H., Winter, P., Kahl, G., Reuter, M., et al (2005) SuperSAGE. Cell Microbiol 7 : 11-18. Mayer, M.P., Bueno, L.C., Hansen, E.J. and DiRien zo, J.M. (1999) Identifi cation of a cytolethal distending toxin gene locus and features of a virulence-associated region in Actinobacillus actinomycetem comitans. Infect Immun 67 : 1227-1237. McCarthy, J.V., Ni, J., and Dixit, V.M. (1998) RIP2 is a novel NF-kappaB-activating and cell death-inducing kinase. J Biol Chem 273 : 16968-75. McManus, B.M., Yanagawa, B., Rezai, N., Luo, H., Taylor, L., Zhang, M., et al (2002) Genetic determinants of coxsackievirus B3 pathogenesis. Ann N Y Acad Sci 975 : 169-179. McShane, L.M., Radmacher, M.D., Freidlin, B., Y u, R., Li, M.C. and Simon, R. (2002) Methods for assessing reproducibility of clustering pa tterns observed in analyses of microarray data Bioinformatics 18: 1462-1469. Meyer, D.H., and Fives-Taylor, P.M. (1997) The role of Actinobacillus acti nomycetemcomitans in the pathogenesis of periodontal disease. Trends Microbiol 5 : 224-8. Meyer, D.H., Lippmann, J.E. and Fives-Taylor P.M. (1996) Invasion of epithelial cells by Actinobacillus acti nomycetemcomitans : a dynamic, multistep process. Infect Immun 64 : 2988-2997. Meyer, D.H., Mintz, K.P. and Fives-Taylor, P. M. (1997) Models of invasion of enteric and periodontal pathogens into epitheli al cells: a comparative analysis. Crit Rev Oral Biol Med 8 : 389-409. Meyer, D.H., Rose, J.E., Lippmann, J.E. and Fives-Taylor, P.M. ( 1999) Microtubules are associated with intracellular movement and spread of the periodontopathogen Actinobacillus acti nomycetemcomitans Infect Immun 67 : 6518-6525. Meyer, D.H., Sreenivasan, P.K. and Fives-Taylor P.M. (1991) Evidence for invasion of a human oral cell line by Actinobacillus actinomycetemcomitans. Infect Immun 59 : 2719-2726. Miller, L.D., Long, P.M., Wong, L., Mukherjee, S., McShane, L.M. and Liu, E.T. (2002) Optimal gene expression analysis by microarrays. Cancer Cell 2 : 353-361. Mintz K. P. and Fives-Taylor P. M. (2000) impA a gene coding for an inner membrane protein, influences colonial morphology of Actinobacillus acti nomycetemcomitans Infect Immun 68 :6580-6586. Mintz, K.P. (2004) Identification of an extr acellular matrix protein adhesin, EmaA, which mediates the adhesion of Actinobacillus acti nomycetemcomitans to collagen. Microbiology 150 : 2677-2688.

PAGE 215

215 Mintz, K.P., Brissette, C., and Fives-Taylor, P.M. (2002) A reco mbinase A-deficient strain of Actinobacillus acti nomycetemcomitans constructed by insertional mutagenesis using a mobilizable plasmid. FEMS Microbiol Lett 206 : 87-92. Moore, W.E., Holdeman, L.V., Smibert, R.M., Hash, D.E., Burmeister, J.A. and Ranney, R.R. (1982) Bacteriology of severe peri odontitis in young adult humans. Infect Immun 38 : 1137-1148. Nakagawa, I., Amano, A., Ohara-Nemoto, Y., Endoh, N., Morisaki, I., Kimura, S., et al (2002) Identification of a new va riant of fimA gene of Porphyromonas gingivalis and its distribution in adults and disabl ed populations with periodontitis. J Periodontal Res 37 : 425-432. Nakhjiri, S.F., Park, Y., Yilmaz, O., Chung, W.O., Watanabe, K., El-Sabaeny, A., et al (2001) Inhibition of epithe lial cell apoptosis by Porphyromonas gingivalis. FEMS Microbiol Lett 200 : 145-149. Narayanan, S.K., Nagaraja, T.G., Chengappa, M.M. and Stewart, G.C. (2002) Leukotoxins of gram-negative bacteria. Vet Microbiol 84 : 337-356. Nelson, K.E., Fleischmann, R.D., DeBoy, R.T., Paulsen, I.T., Fouts, D.E., Eisen, J.A., et al (2003) Complete genome sequence of the oral pathogenic Bacterium porphyromonas gingivalis strain W83. J Bacteriol 185 : 5591-5601. Njoroge, T., Genco, R.J., Sojar, H.T., Hamada, N. and Genco, C.A. (1997) A role for fimbriae in Porphyromonas gingivalis invasion of oral epithelial cells. Infect Immun 65 : 1980-1984. Norskov-Lauritsen, N. and Kilian, M. (2006) Reclassification of Actinobacillus actinomycetemcomitans, Haemophilus aphrophilus, Haemophilus paraphrophilus and Haemophilus segnis as Aggregatibacter actinomycetemcomitans gen. nov., comb. nov., Aggregatibacter aphrophilus comb. nov. and Aggregatibacter segnis comb. nov., and emended description of Aggregatibacter aphrophilus to include V factor-dependent and V factor-independent isolates. Int J Syst Evol Microbiol 56 : 2135-2146. Oda, D., and Watson, E. (1990) Human oral ep ithelial cell culture. Improved conditions for reproducible culture in serum-free medium. In Vitro Cell Dev Biol 26 : 589-595. Oda, D., Bigler, L., Lee, P. and Blanton, R. (1996) HPV immortalization of human oral epithelial cells: a model for carcinogenesis. Exp Cell Res 226 : 164-169. Oda, D., Bigler, L., Mao, E.J. and Disteche, C.M. (1996b) Chromosomal abnormalities in HPV16-immortalized oral epithelial cells Carcinogenesis 17: 2003-2008. Offenbacher S. (1996) Periodon tal diseases: pathogenesis. Ann Periodontol (1996) 1 : 821-78.

PAGE 216

216 Offenbacher, S., Katz, V., Fertik, G., Collins, J., Boyd, D., Maynor, G., et al (1996) Periodontal infection as a possible risk fact or for preterm low birth weight. J Periodontol 67 : 11031113. Office of the U.S. Surgeon General. (2007) Oral Health: Preventing Cavities, Gum Disease, and Tooth Loss. 2007 : 4. Ohta, H., Kokeguchi, S., Fukui, K. and Kato, K. (1986) Actinobacillus (Haemophilus) actinomycetemcomitans in periodontal disease. Microbiol Immunol 30 : 629-643. Okamoto, K., Kadowaki, T., Nakayama, K. a nd Yamamoto, K. (1996) Cloning and sequencing of the gene encoding a novel lysine-specifi c cysteine proteinase (Lys-gingipain) in Porphyromonas gingivalis : structural relationship with the arginine-specific cysteine proteinase (Arg-gingipain). J Biochem (Tokyo) 120 : 398-406. Okamoto, K., Misumi, Y., Kadowaki, T., Yoneda M., Yamamoto, K. and Ikehara, Y. (1995) Structural characterization of argingipain, a novel arginine-sp ecific cysteine proteinase as a major periodontal pathogenic factor from Porphyromonas gingivalis. Arch Biochem Biophys 316 : 917-925. Olczak, T., Simpson, W., Liu, X. and Genco, C. A. (2005) Iron and heme utilization in Porphyromonas gingivalis. FEMS Microbiol Rev 29 : 119-144. Overholt, B.F. (1966) Actinobacillus actinomycetemcomitans endocarditis. Arch Intern Med 117 : 99-102. Papapanou, P.N., Abron, A., Verbitsky, M., Picolo s, D., Yang, J., Qin, J., et al. (2004) Gene expression signatures in chronic and a ggressive periodontit is: a pilot study Eur J Oral Sci 112: 216-223. Park, Y., Simionato, M.R., Sekiya, K., Murakami, Y., James, D., Chen, W., et al (2005) Short fimbriae of Porphyromonas gingivalis and their role in coadhesion with Streptococcus gordonii Infect Immun 73 : 3983-3989. Paster, B.J., Boches, S.K., Galvin, J.L ., Ericson, R.E., Lau, C.N., Levanos, V.A., et al (2001) Bacterial diversity in human subgingival plaque. J Bacteriol 183 : 3770-3783. Paturel, L., Casalta, J.P., Habib, G ., Nezri, M. and Raoult, D. (2004) Actinobacillus actinomycetemcomitans endocarditis. Clin Microbiol Infect 10 : 98-118. Pavloff, N., Pemberton, P.A., Potempa, J ., Chen, W.C., Pike, R.N., Prochazka, V., et al (1997) Molecular cloning and characterization of Porphyromonas gingivalis lysine-specific gingipain. A new member of an emerging fa mily of pathogenic bacterial cysteine proteinases. J Biol Chem 272 : 1595-1600.

PAGE 217

217 Pavloff, N., Potempa, J., Pike, R.N., Pr ochazka, V., Kiefer, M.C., Travis, J., et al (1995) Molecular cloning and structural characteriza tion of the Arg-gingipa in proteinase of Porphyromonas gingivalis Biosynthesis as a protei nase-adhesin polyprotein. J Biol Chem 270 : 1007-1010. Perez, A., and Davis, P.B. ( 2004) Gene profile changes after Pseudomonas aeruginosa exposure in immortalized airw ay epithelial cells. J Struct Funct Genomics 5 : 179-94. Perez, B.A., Planet, P.J., Kachlany, S.C., Tomi ch, M., Fine, D.H. and Figurski, D.H. (2006) Genetic analysis of the requirement for flp-2, tadV, and rcpB in Actinobacillus actinomycetemcomitans biofilm formation. J Bacteriol 188 : 6361-6375. Petti, C.A., Bhally, H.S., Weinstein, M.P ., Joho, K., Wakefield, T., Reller, L.B., et al (2006) Utility of extended blood culture incubation for isolation of Haemophilus, Actinobacillus Cardiobacterium, Eikenella, and Kingella organisms: a retrospective multicenter evaluation. J Clin Microbiol 44 : 257-259. Potempa, J., Banbula, A. and Travis, J. (2000) Role of bacterial proteinases in matrix destruction and modulation of host responses. Periodontol 2000 24 : 153-192. Potempa, J., Pike, R. and Travis, J. (1997) Titra tion and mapping of the active site of cysteine proteinases from Porphyromonas gingivalis (gingipains) using peptidyl chloromethanes. Biol Chem 378 : 223-230. Potts, T.V., Zambon, J.J. and Genco, R.J. (1985) Reassignment of Actinobacillus actinomycetemcomitans to the genus Haemophilus as Haemophilus actinomycetemcomitans comb. nov. Int J Syst Bacteriol 35 : 337-341. Progulske-Fox, A., Tumwasorn, S. and Holt, S. C. (1989) The expression and function of a Bacteroides gingivalis hemagglutinin gene in Escherichia coli. Oral Microbiol Immunol 4 : 121-131. Progulske-Fox, A., Tumwasorn, S., Lepine, G., Whitlock, J., Savett, D., Ferretti, J.J., et al (1995) The cloning, expression and sequence analysis of a second Porphyromonas gingivalis gene that codes for a protein involved in hemagglutination. Oral Microbiol Immunol 10 : 311-318. Pujol, C. and Bliska, J.B. (2005) Turning Yersinia pathogenesis outside in: subversion of macrophage function by intracellular yersiniae Clin Immunol 114 : 216-226. Quercia, R., Bani Sadr, F., Cortez, A., Ar let, G. and Pialoux, G. (2006) Genital tract actinomycosis caused by Actimyces israelii Med Mal Infect 36 : 393-395. Radmacher, M.D., McShane, L.M. and Simon, R. (2002) A paradigm for class prediction using gene expression profiles J Comput Biol 9: 505-511.

PAGE 218

218 Rakoff-Nahoum, S., Paglino, J., Eslami-Varzane h, F., Edberg, S. and Medzhitov, R. (2004) Recognition of commensal microflora by tolllike receptors is required for intestinal homeostasis. Cell 118 : 229-241. Ramasubbu, N., Thomas, L.M., Ragunath, C. and Ka plan, J.B. (2005) Structural analysis of dispersin B, a biofilm-releasing glycos ide hydrolase from the periodontopathogen Actinobacillus actinomycet emcomitans. J Mol Biol 349 : 475-486. Rao, A., Luo, C. and Hogan, P.G. (1997) Transcri ption factors of the NFAT family: regulation and function. Annu Rev Immunol 15 : 707-747. Rappuoli, R. (2001) Reverse vaccinology, a genome -based approach to vaccine development. Vaccine 19 : 2688-2691. Renvert, S., Pettersson, T., Ohlss on, O. and Persson, G.R. (2006) Bacterial profile and burden of periodontal infection in subj ects with a diagnosis of acute coronary syndrome. J Periodontol 77 : 1110-1119. Richardson, J., Craighead, J.C., Cao, S.L. and Handfield, M. (2005) Concurrence between the gene expression pattern of Actinobacillus actinomycetemcomitans in localized aggressive periodontitis and in human epithelial cells. J Med Microbiol 54 : 497-504. Rodrigues, P.H. and Progulske-Fox, A. (2005) Gene expression profile analysis of Porphyromonas gingivalis during invasion of human coronary artery endothelial cells. Infect Immun 73 : 6169-6173. Roe, B.A., Najar, F.Z., Gillaspy, A., Clifton, S., Ducey, T., Lewis, L., et al (2003) Actinobacillus Genome Sequencing Project. 2007 : 1. Rogers, J.E., Li, F., Coatney, D.D., Rossa, C., Bronson, P., Krieder, J.M., et al (2007) Actinobacillus acti nomycetemcomitans lipopolysaccharide-mediated experimental bone loss model for aggressive periodontitis. J Periodontol 78 : 550-558. Rollins, S.M., Peppercorn, A., Hang, L., Hillm an, J.D., Calderwood, S.B., Handfield, M., et al (2005) In vivo induced antigen technology (IVIAT). Cell Microbiol 7 : 1-9. Roof, D.J., Hayes, A., Adamian, M., Chishti, A. H., and Li, T. (1997) Molecular characterization of abLIM, a novel actin-binding and double zinc finger protein. J Cell Biol 138 : 575-588. Rose, J.E., Meyer, D.H. and Fives-Taylor, P.M. (2003) Aae, an autotransporter involved in adhesion of Actinobacillus acti nomycetemcomitans to epithelial cells. Infect Immun 71 : 2384-2393. Rosenberger, C.M., Scott, M.G., Gold, M.R ., Hancock, R.E. and Finlay, B.B. (2000) Salmonella typhimurium infection and lipopolysaccharide stimulation induce similar changes in macrophage gene expression. J Immunol 164 : 5894-5904.

PAGE 219

219 Rudney, J.D., Chen, R. and Sedgewick, G.J. (2001) Intracellular Actinobacillus actinomycetemcomitans and Porphyromonas gingivalis in buccal epithelial cells collected from human subjects. Infect Immun 69 : 2700-2707. Rudney, J.D., Chen, R. and Sedgewick, G.J. (2005) Actinobacillus actinomycetemcomitans, Porphyromonas gingivalis, and Tannerella forsythensis are components of a polymicrobial intracellular flora within human buccal cells. J Dent Res 84 : 59-63. Ruiz, T., Lenox, C., Radermacher, M. and Mintz, K.P. (2006) Novel surface structures are associated with the adhesion of Actinobacillus acti nomycetemcomitans to collagen. Infect Immun 74 : 6163-6170. Saarela, M., Asikainen, S., Alaluusua, S., Pyhala, L., Lai, C.H. and Jousimies-Somer, H. (1992) Frequency and stability of monoor poly-infection by Actinobacillus actinomycetemcomitans serotypes a, b, c, d or e. Oral Microbiol Immunol 7 : 277-279. Saarela, M.H., Dogan, B., Alaluusua, S. and Asikainen, S. (1999) Persistence of oral colonization by the same Actinobacillus acti nomycetemcomitans strain(s). J Periodontol 70 : 504-509. Sakamoto, M., Suzuki, M., Umeda, M., Ishikawa I. and Benno, Y. (2002) Reclassification of Bacteroides forsythus (Tanner et al 1986) as Tannerella forsythensis corrig., gen. nov., comb. nov. Int J Syst Evol Microbiol 52 : 841-849. Sakamoto, M., Umeda, M. and Benno, Y. (2005) Mo lecular analysis of hu man oral microbiota. J Periodontal Res 40 : 277-285. Sandros J., Madianos P. N. and Papapanou P. N. (1996) Cellular events concurrent with Porphyromonas gingivalis invasion of oral epithelium in vitro. Eur J Oral Sci 104 :363371. Sandros, J., Karlsson, C., Lappin, D.F., Madian os, P.N., Kinane, D.F. and Papapanou, P.N. (2000) Cytokine responses of oral epithe lial cells to Porphyromonas gingivalis infection. J Dent Res 79 : 1808-14. Sansonetti, P.J. (1993) Bacterial pathogens, from adherence to i nvasion: comparative strategies. Med Microbiol Immunol 182 : 223-232. Sanz, M., Lau, L., Herrera, D., Morillo, J.M. and Silva, A. (2004) Methods of detection of Actinobacillus actinomycetemcomita ns, Porphyromonas gingivalis and Tannerella forsythensis in periodontal microbiology, with speci al emphasis on advanced molecular techniques: a review. J Clin Periodontol 31 : 1034-1047. Scannapieco F. A. (2005) Systemic effects of periodontal diseases. Dent Clin North Am 49 :53350, vi.

PAGE 220

220 Scannapieco, F.A., and Genco, R.J. (1999) A ssociation of periodontal infections with atherosclerotic and pulmonary diseases. J Periodont Res 34 : 340-345. Schena, M. and Davis, R.W. (1994) Structure of homeobox-leuc ine zipper genes suggests a model for the evolution of gene families. Proc Natl Acad Sci U S A 91 : 8393-8397. Schena, M., Lloyd, A.M. and Davis, R.W. (1993) The HAT4 gene of Arabidopsis encodes a developmental regulator. Genes Dev 7 : 367-379. Schena, M., Shalon, D., Davis, R.W. and Brown, P.O. (1995) Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270 : 467-470. Schenkein, H.A., Barbour, S.E., Berry, C.R., Kipps B. and Tew, J.G. (2000) Invasion of human vascular endothelial cells by Actinobacillus actinomycetemcomitans via the receptor for platelet-activating factor. Infect Immun 68 : 5416-5419. Schilling, D., Pittelkow, M.R., and Kumar, R. (2001) IEX-1, an immediate early gene, increases the rate of apoptosis in keratinocytes. Oncogene 20 : 7992-7. Schleiss, M. (2005) Progress in cyto megalovirus vaccine development. Herpes 12 : 66-75. Schreiner, H.C., Sinatra, K., Kaplan, J.B ., Furgang, D., Kachlany, S.C., Planet, P.J., et al (2003) Tight-adherence genes of Actinobacillus actinomycetemcomitans are required for virulence in a rat model. Proc Natl Acad Sci U S A 100 : 7295-7300. Schulte, R., Wattiau, P., Hartland, E.L., Robins -Browne, R.M. and Cornelis, G.R. (1996) Differential secretion of interleukin-8 by human epithelial cell lines upon entry of virulent or nonvirulent Yersinia enterocolitica Infect Immun 64 : 2106-2113. Schutze, K., Niyaz, Y., Stich, M. and Buchstalle r, A. (2007) Noncontact laser microdissection and catapulting for pure sample capture. Methods Cell Biol 82 : 647-673. Sepulveda, A.R., Tao, H., Carloni E., Sepulveda, J., Graham, D. Y., and Peterson, L.E. (2002) Screening of gene expression profiles in gastric epithelial cells induced by Helicobacter pylori using microarray analysis. Aliment Pharmacol Ther 16 : 145-57. Sfakianakis A., Barr C. E. and Kreutzer D. L. (2001) Actinobacillus actinomycetemcomitans induced expression of IL-1alpha and IL-1beta in human gingival epithe lial cells: role in IL-8 expression. Eur J Oral Sci 109 :393-401. Shah, H.N. and Collins, M.D. (1988) Proposal for reclassification of Bacteroides asaccharolyticus, Bacteroides gingivalis, and Bacteroides endodontalis in a new genus, Porphyromonas. Int J Syst Bacteriol 38 : 128-131.

PAGE 221

221 Shah, H.N. and Gharbia, S.E. (1989) Ecologica l events in subgingival dental plaque with reference to Bacteroides and Fusobacterium species. Infection 17 : 264-268. Shah, H.N., Bonnett, R., Mateen, B. and Williams R.A. (1979) The porphyrin pigmentation of subspecies of Bacteroides melaninogenicus. Biochem J 180 : 45-50. Shapira, L., Wilensky, A. and Kinane, D.F. (2005) Effect of genetic variability on the inflammatory response to periodontal infection. J Clin Periodontol 32 Suppl 6 : 72-86. Shirin, H., Sordillo, E.M., Kolevska, T. K., Hibshoosh, H., Kawabata, Y., Oh, S.H., et al (2000) Chronic Helicobacter pylori infection induces an a poptosis-resistant phenotype associated with decreased expression of p27(kip1). Infect Immun 68 : 5321-5328. Simon, R., Radmacher, M.D., Dobbin, K. and McShan e, L.M. (2003) Pitfalls in the use of DNA microarray data for diagnostic and prognostic classification J Natl Cancer Inst 95: 14-18. Sims, T.J., Schifferle, R.E., Ali, R.W., Ska ug, N. and Page, R.C. (2001) Immunoglobulin G response of periodontitis patients to Porphyromonas gingivalis capsular carbohydrate and lipopolysaccharide antigens. Oral Microbiol Immunol 16 : 193-201. Skurk, C., Maatz, H., Kim, H.S., Yang, J., Abi d, M.R., Aird, W.C., and Walsh, K. (2004) The Akt-regulated forkhead transcription factor FOXO3a controls endothelial cell viability through modulation of the caspase-8 inhibitor FLIP. J Biol Chem 279 : 1513-25. Slots, J. (2005) Herpesviru ses in periodontal diseases. Periodontol 2000 38 : 33-62. Slots, J., and Genco, R.J. (1984) Black-pigmented Bacteroides species, Capnocytophaga species, and Actinobacillus actinomycetemcomitans in human periodontal disease: virulence factors in colonization, surv ival, and tissue destruction. J Dent Res 63 : 412-21. Slots, J., Bragd, L., Wikstrom, M., a nd Dahlen, G. (1986) The occurrence of Actinobacillus actinomycetemcomitans, Bacteroides gingivalis and Bacteroides intermedius in destructive periodontal disease in adults. J Clin Periodontol 13 : 570-7. Smith, J.L. and Bayles, D.O. (2006) The contributi on of cytolethal disten ding toxin to bacterial pathogenesis Crit Rev Microbiol 32: 227-248. Socransky, S.S., Haffajee, A.D., Cugini, M.A., Smith, C. and Kent, R.L.,Jr. (1998) Microbial complexes in subgingival plaque. J Clin Periodontol 25 : 134-144. Socransky, S.S., Haffajee, A.D., Goodson, J. M. and Lindhe, J. (1984) New concepts of destructive periodontal disease. J Clin Periodontol 11 : 21-32. Song, Y.-., Kozarov, E.V., Walters, S.M., Cao, S.L., Handfield, M., Hillman, J.D., et al (2002) Genes of Periodontopathogens E xpressed During Human Disease. Ann Periodontol 7 : 38-42.

PAGE 222

222 Spitznagel, J. Jr, Kraig, E., and Kolodrubetz, D. (1995) The regulation of leukotoxin production in Actinobacillus actinomycetemcomitans strain JP2. Adv Dent Res 9 : 48-54. Steele-Mortimer, O., Kn odler, L.A. and Finlay, B.B. (2000) Poisons, ruffles and rockets: bacterial pathogens and th e host cell cytoskeleton Traffic 1: 107-118. St Geme, J.W.,3rd and Cutter, D. (2000) The Haemophilus influenzae Hia adhesin is an autotransporter protein that remains uncleaved at the C terminus and fully cell associated. J Bacteriol 182 : 6005-6013. Sugai, M., Kawamoto, T., Peres, S.Y., Ueno, Y., Komatsuzawa, H., Fujiwara, T., et al (1998) The cell cycle-specific growth -inhibitory factor produced by Actinobacillus actinomycetemcomitans is a cytolethal distending toxin. Infect Immun 66 : 5008-5019. Suzuki, T., Kobayashi, M., Isatsu, K., Nishihara, T., Aiuchi, T., Nakaya, K., et al (2004) Mechanisms involved in apoptosis of hu man macrophages induced by lipopolysaccharide from Actinobacillus actinomycetemcomitans in the presence of cycloheximide. Infect Immun 72 : 1856-1865. SweetHaven Publishing Services. (2006) Fundamentals of Dental Technology 2007 : 1. Taichman, N.S., Dean, R.T. and Sanderson, C.J. (1980) Biochemi cal and morphological characterization of the killing of human monocytes by a leukotoxin derived from Actinobacillus actinomycetem comitans. Infect Immun 28 : 258-268. Takayama, A., Satoh, A., Ngai, T., Nishimura, T., Ikawa, K., Matsuyama, T., Shimauchi, H., Takada, H., and Sugawara, S. (2003) Augmentation of Actinobacillus actinomycetemcomitans invasion of human or al epithelial cells and up-regulation of interleukin-8 production by saliva CD14. Infect Immun 71 : 5598-604. Tamayo, P., Slonim, D., Mesir ov, J., Zhu, Q., Kitareewan, S. Dmitrovsky, E., et al. (1999) Interpreting patterns of gene expression with sel f-organizing maps: methods and application to hematopoietic differentiation Proc Natl Acad Sci U S A 96: 2907-2912. Tang, G., Ruiz, T., Barrantes-Reynolds, R. and Mi ntz, K.P. (2007) Molecular heterogeneity of EmaA, an oligomeric autotransporter adhe sin of Aggregatibac ter (Actinobacillus) actinomycetemcomitans Microbiology 153: 2447-2457. ten Cate, J.M. (2006) Biofilms, a new appro ach to the microbiology of dental plaque Odontology 94: 1-9. Teng, Y.T. (2006a) Protective and destructive immun ity in the periodontium: Part 1--innate and humoral immunity and the periodontium. J Dent Res 85 : 198-208. Teng, Y.T. (2006b) Protective and destructive imm unity in the periodontium: Part 2--T-cellmediated immunity in the periodontium. J Dent Res 85 : 209-219.

PAGE 223

223 Teng, Y.T. and Zhang, X. (2005) Apoptotic activ ity and sub-cellular lo calization of a T4SSassociated CagE-homologue in Actinobacillus actinomycetemcomitans. Microb Pathog 38 : 125-132. Teng, Y.T., and Hu, W. (2003) Expression cloni ng of a periodontitis-a ssociated apoptotic effector, cagE homologue, in Actinobacillus actinomycetemcomitans. Biochem Biophys Res Commun 303 : 1086-94. Teughels, W., Sliepen, I., Quirynen, M., Haake, S.K., Van Eldere, J., Fives-Taylor, P., et al (2007) Human cytomegalovirus enhances A. actinomycetemcomitans adherence to cells. J Dent Res 86 : 175-180. Thiha, K., Takeuchi, Y., Umeda, M., Huang, Y., Ohnishi, M. and Ishikawa, I. (2007) Identification of periodontopathic bacteria in gingival tissu e of Japanese periodontitis patients. Oral Microbiol Immunol 22 : 201-207. Tonetti, M.S., D'Aiuto, F., Nibali, L., Donald, A., Storry, C., Parkar, M., et al (2007) Treatment of periodontitis and endothelial function. N Engl J Med 356 : 911-920. Tran, H., Brunet, A., Grenier, J.M., Datta, S.R ., Fornace, A.J. Jr, DiStefano, P.S., Chiang, L.W., and Greenberg, M.E. (2002) DNA repair pathway stimulated by the forkhead transcription factor FOXO3a through the Gadd45 protein. Science 296 : 530-534. Tribble, G.D., Mao, S., James, C.E. and Lamont, R.J. (2006) A Porphyromonas gingivalis haloacid dehalogenase family phosphatase interacts with human phosphoproteins and is important for invasion. Proc Natl Acad Sci U S A 103 : 11027-11032. Ueno, Y., Ohara, M., Kawamoto, T., Fujiwara, T. Komatsuzawa, H., Oswald, E., et al. (2006) Biogenesis of the Actinobacillus actinomycet emcomitans cytolethal distending toxin holotoxin Infect Immun 74: 3480-3487. Van Winkelhoff, A.J., Laine, M.L., Timmerman M.F., Van der Weijden, G.A., Abbas, F., Winkel, E.G., et al (1999) Prevalence and serotyping of Porphyromonas gingivalis in an Indonesian population. J Clin Periodontol 26 : 301-305. Viale, A., Li, J., Tiesman, J., He ster, S., Massimi, A., Griffin, C., et al (2007) Big results from small samples: evaluation of amplificati on protocols for gene expression profiling. J Biomol Tech 18: 150-161. Vito, P., Lacana, E., and D'Adamio, L. (1996) Interfering with a poptosis: Ca(2+)-binding protein ALG-2 and Alzheimer's disease gene ALG-3. Science 271 : 521-5. Wang, D., Kawashima, Y., Nagasawa, T., Takeuchi, Y., Kojima, T., Umeda, M., et al (2005a) Elevated serum IgG titer and avidity to Actinobacillus actinomycetemcomitans serotype c in Japanese periodontitis patients. Oral Microbiol Immunol 20 : 172-179.

PAGE 224

224 Wang, Y., Liu, A. and Chen, C. (2005b) Genetic basis for conversion of rough-to-smooth colony morphology in Actinobacillus actinomycetemcomitans. Infect Immun 73 : 3749-3753. Watanabe,K., Yilmaz,O., Nakhjiri,S.F., Belton,C .M., and Lamont,R.J. (2001) Association of mitogen activated protein kinase pathways with gingival epithelia l cell responses to Porphyromonas gingivalis. Infect Immun 69 : 6731 6737. Weinberg, A., Belton, C.M., Park, Y. and La mont, R.J. (1997) Role of fimbriae in Porphyromonas gingivalis invasion of gingival epithelial cells. Infect Immun 65 : 313316. Woolhouse, M.E., Webster, J.P., Domingo, E., Charlesworth, B. and Levin, B.R. (2002) Biological and biomedical imp lications of the co-evolution of pathogens and their hosts. Nat Genet 32 : 569-577. World Health Organization. (2007) Fact sheet N318 2007 : 1. Wu, M.X. (2003) Roles of the stress-induced gene IEX-1 in regulation of cell death and oncogenesis. Apoptosis 8 : 11-8. Wu, Y.M., Yan, J., Chen, L.L., Sun, W.L. and G u, Z.Y. (2006) Infection frequency of EpsteinBarr virus in subgingival samples from patient s with different peri odontal status and its correlation with clinical parameters. J Zhejiang Univ Sci B 7 : 876-883. Yang, H.W., Huang, Y.F., Chan, Y. a nd Chou, M.Y. (2005) Relationship of Actinobacillus actinomycetemcomitans serotypes to periodontal condition: prevalence and proportions in subgingival plaque. Eur J Oral Sci 113 : 28-33. Yeo, B.K., Lim, L.P., Paquette, D.W. and W illiams, R.C. (2005) Periodontal disease -the emergence of a risk for systemic conditions: pre-term low birth weight. Ann Acad Med Singapore 34 : 111-116. Yilmaz, O., Watanabe, K., and Lamont, R.J. (2 002) Involvement of integrins in fimbriaemediated binding and invasion by Porphyromonas gingivalis Cell Microbiol 4 : 305-14. Yilmaz, O., Young, P.A., Lamont, R.J., and Ke nny, G.E. (2003) Gingival epithelial cell signalling and cytoskeletal responses to Porphyromonas gingivalis invasion. Microbiol 149 : 2417-2426. Yokoyama, K., Higashi, H., Ishikawa S., Fujii, Y., Kondo, S., Kato, H., et al (2005) Functional antagonism between Helicobacter pylori CagA and vacuolating t oxin VacA in control of the NFAT signaling pathway in gastric epithelial cells. Proc Natl Acad Sci U S A 102 : 9661-9666.

PAGE 225

225 Yoshino, T., Laine, M.L., van Winkelhoff, A.J. and Dahlen, G. (2007) Genotype variation and capsular serotypes of Porphyromonas gingivalis from chronic periodontitis and periodontal abscesses. FEMS Microbiol Lett 270 : 75-81. Yowe, D., Cook, W.J. and Gutierrez-Ramos, J. C. (2001) Microarrays for studying the host transcriptional response to microbial inf ection and for the identification of host drug targets. Microbes Infect 3 : 813-821. Yu, K.Y., Kwon, B., Ni, J., Zhai, Y., Ebner, R ., and Kwon, B.S. (1999) A newly identified member of tumor necrosis factor rece ptor superfamily (TR6) suppresses LIGHTmediated apoptosis. J Biol Chem 274 : 13733-6. Yue, G., Kaplan, J.B., Furgang, D., Mansfi eld, K.G. and Fine, D.H. (2007) A Second Aggregatibacter actinomycetemcomitans Au totransporter Adhesin that Exhibits Specificity for Buccal Epithelial Cells of Humans and Old World Primates Infect Immun Zambon, J.J. (1985) Actinobacillus actinomycetemcomitans in human periodontal disease. J Clin Periodontol 12 : 1-20. Zambon, J.J., Slots, J. and Genc o, R.J. (1983) Serology of oral Actinobacillus actinomycetemcomitans and serotype distribution in human periodontal disease. Infect Immun 41 : 19-27. Zeeberg, B.R., Feng, W., Wang, G ., Wang, M.D., Fojo, A.T., Suns hine, M., Narasimhan, S., Kane, D.W., Reinhold, W.C., Lababidi, S., Buss ey, K.J., Riss, J., Barrett, J.C. and J.N. Weinstein. (2003) GoMiner: A Resource for Biological Interpreta tion of Genomic and Proteomic Data. Genome Biol 4 : R28. Zeng, H., Carlson, A.Q., Guo, Y., Yu, Y ., Collier-Hyams, L.S., Madara, J.L., et al (2003) Flagellin is the major proinflammatory determinant of enteropathogenic Salmonella J Immunol 171 : 3668-3674. Zheng, G., George, E.O., Narasimhan, G. (2004) Ne ural Network Classifier s and Gene Selection Methods for Microarray Data on Human Lung Adenocarcinoma, In Press: Methods of Microarray Data Analysis IV, Springer. Zhu, H., Cong, J.P., Mamtora, G., Gingeras, T. and Shenk, T. (1998) Cellular gene expression altered by human cytomegalovirus: global monitoring with oligonucleotide arrays Proc Natl Acad Sci U S A 95 : 14470-14475. Zingarelli, B. (2005) Nuclear factor-kappaB. Crit Care Med 33 : S414-6.

PAGE 226

226 BIOGRAPHICAL SKETCH Jeffrey Jay Mans was born January 22, 1979 to Glenn and Judy Mans of Gainesville, FL. He is the grandson of Dr. Rusty and Beulah Mans of Gainesville, FL and Donald and Marilyn McMillan of Nashville, TN. A lifelong Floridian, Jeff attended Gainesville High School (GHS) where he competed in junior varsity soccer and varsity baseball for GHS and American Legion baseball for Gainesville Post 16. Advanced Placement biology taught by Mrs. Nancy Smith was the most influential class in Jeff’s high school education, and impact ed Jeff’s decision to pursue a biology degree after his graduation from GHS in 1997. Jeff was awarded the Presidential Scholarship to attend the University of North Florida (UNF). While a student at UNF Jeff was actively involved in the Honors Program and earned High Pass distinction for his Honors portfoli o, as well as University Honors for his undergraduate research, “Prevalen ce of Lyme Disease Bacteria in Northeast Florida” conducted under the guidance of Dr. Kerry Clark. The Honor s Program and mentoring of Dr. Clark also allowed Jeff to complete summer coursework ab road at the University of Cambridge in Cambridge, UK. He graduated from UNF Cum La ude with his Bachelor of Science degree for biology in 2001. After graduation, Jeff worked as a Nurse T ech I in Labor, Delivery, Recovery and Postpartum at North Florida Regional Medical Center under guidance of many wonderful nurses and Mrs. Beverly Griseck. Jeff later worked in the laboratory of Dr. Bill Castleman at the University of Florida, College of Veterinary Medicine, Department of Pathobiology before his acceptance into the University of Florida Interdisciplinary Program (IDP). Jeff joined the lab of Dr. Martin Handfield in 2002 to co mplete his Doctoral research. Jeff is 4 years happily married to Lori K. Mans of Jacksonville, FL and UNF sweetheart.