Stochasticity in Quorum Sensing


Material Information

Stochasticity in Quorum Sensing lux Expression in Vibrio fischeri at the Single Cell Level
Physical Description:
1 online resource (132 p.)
Perez,Pablo D
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Committee Chair:
Hagen, Stephen J
Committee Members:
Reitze, David H
Roitberg, Adrian E
Tanner, David B
Purich, Daniel L


Subjects / Keywords:
autoinducer -- bacteria -- cell -- expression -- fischeri -- gene -- heterogeneity -- homoserine -- individual -- lactone -- luminescence -- lux -- microfluidics -- noise -- quorum -- regulation -- sensing -- single -- stochasticity -- vibrio
Physics -- Dissertations, Academic -- UF
Physics thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation


Like many bacterial species, Vibrio fischeri can detect its own population density through a quorum sensing (QS) mechanism. The bacterium releases small signal molecules (AI, autoinducers), which accumulate at high population density and trigger a genetic switch. In V.fischeri this leads to bioluminescence. Little is known about how stochastic gene expression affects QS at the level of single cells. We imaged the luminescence of individual V.fischeri cells in a flow chamber, and the fluorescence of a gfp reporter using microfluidics. We directly measured the intercellular variability in lux activation. We found a wide heterogeneous response to the external 3OC6HSL AI signal among cells and over time: some cells may be strongly luminescent while others are virtually dark. We also studied how the addition of C8HSL AI signal modulated the heterogeneous response induced by 3OC6HSL. Our results confirmed at the single cell level the C8HSL inhibition effect described in previous studies. Furthermore, they showed that the broad distribution of lux responses depends only on the average activation when different 3OC6HSL and C8HSL concentrations are provided. The analysis of noise in the individual cell response can eventually lead to a better understanding of how cells use QS to gather information about their environment.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
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 Pablo D Perez.
Thesis (Ph.D.)--University of Florida, 2011.
Adviser: Hagen, Stephen J.
Electronic Access:

Record Information

Source Institution:
Rights Management:
Applicable rights reserved.
lcc - LD1780 2011
System ID:

This item is only available as the following downloads:

Full Text




2 2011 Pablo Delfino Prez


3 To my mother and my late father, who always encouraged and supported me in my studies, and to my fiance, whose love and company made po ssible completing this arduous and rewarding journey


4 ACKNOWLEDGMENTS Completing a PhD dissertation in experimenta l biophysics was a long, difficult but very rewarding journey. There were many times when I could only make it through thanks to the people who provided me invaluable assistance. I woul d like to first thank my mother Silvia and my late father Delfino. Without their help, care and encouragement it would not have been possible for me to achieve what I have. They both unconditionally h elped me al ong all these years. During my first years of graduate school, I met my fiance Sohyun. Her love and patience kept me moving forward when facing extremely difficult personal and academic challenges. I could not have done it without her. I would like to tha nk my advisor Dr Stephen J. Hagen for his guidance and the sharing of his knowledge. He provided me a great place to develop my scientific skills. Dr Hagen consistently provided suggestions and smart insights into all my research work. He continually pushe d me forward to see the big picture and to be attentive to the small details at the same time. My thanks also go to my talented supervisory committee members Dr Adrian Roitberg, Dr David Reitze, Dr David Tanner and Dr Daniel Purich. Their suggestions have been invaluable in completing this research work. I would like to express my gratitude to oth er members of my research group: Minjun Son Joel Weiss and Jonathan Young. They developed a microfluidic device making technique and instructed me on how to make my own devices. I want to th ank Joel Weiss for providing the technical details of the wafer and device making process as it is described in this dissertation. Joel also provided very significant help in several microfluid ic and bulk experiments described i n Chapters 4 and 5. I am also grateful to Minjun Son who made the wafer template that we repeatedly used later to make the


5 three channel PDMS device s I also want to thank Elaine Johnson, Leslie Pelakh and Rachel West as t hey made important contributions in the first stages of the research. I would like to express my gratitude to other researchers for their generous contributions. Dr Edward Ruby and Dr Mark Mandel provided us with the V.fischeri MJ11 strain and Dr Eric Stabb provided the V.fischeri JB10 st rain. My thanks also go to my friends and former lab members Ranjani Narayanan and Omjoy (OJ) Ganesh. Rese arch is not easy by definition and having more experienced and encouraging lab mates as they were was invaluable. Minjun Son, G abriel Dilanji and Inh ae Kwak, the newest lab additions, have been great lab mates. I wish them the best of successes. Working in experimental physics is about having good ideas for experiments, but it is also about making things work. Several staff members at the UF Physics De partment provided priceless help in developing the experimental hardware. I thank t he machine shop staff (Marc Link, Ed Storch, Bill Malphurs, Raymond Frommeyer John Van Leer and Mike Herlevich ), the electronics shop staff Jay Horton and Tim Noland. I wo uld like to acknowledge the National Science Foundation for funding support ( MCB 0347124 ). L ast but not least, I want to thank the University of Florida in Gainesville I was able to interact with very talented faculty and researchers, made many friends an d enjoyed the warm and sunny weather.


6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Quorum Sensing ................................ ................................ ................................ ..... 13 Quorum Sensing in Vibrio fischeri ................................ ................................ .... 15 Bioluminescence reaction in Vibrio fischeri ................................ ................ 17 3OC6HSL lux regulation ................................ ................................ ............ 19 C8HSL and AI 2 lux regulation ................................ ................................ ... 20 Noise in Gene Expression ................................ ................................ ...................... 22 Objectives of this Dissertation ................................ ................................ ................. 25 2 LUX EXPRESSION IN RESPON SE TO 3OC6HSL ................................ ................ 28 Measuring Single Cell Vibrio fischeri Bioluminescence ................................ .......... 28 Bioluminescence Bulk Measurements with MJ11 V.fischeri Strain ......................... 29 Using Well trays to Measure Bulk Cultures ................................ ...................... 31 MJ11 Bulk Well tray Measurement Results ................................ ...................... 32 Single cell Bioluminescence of V.fischeri MJ11 in Response to 3OC6HSL ............ 35 Acquiring Single cell V.fischeri Bioluminescence ................................ ............. 35 Single cell Bioluminescence Result s ................................ ................................ 37 ....................... 43 Autocorrelating the Light Output of Single Cells and C ontrol Microspheres ..... 46 Additional Complementary Experiments ................................ ................................ 49 Testing the Heterogeneity in Single Colony Cultures ................................ ....... 49 Substrate Deficiency ................................ ................................ ........................ 50 Comparing the Heterogeneity of the Native Bioluminescence as a Reporter with that of JB10 V.fischeri Using Fluores cence as a Reporter ..................... 51 Rich Medium Effect ................................ ................................ ................................ 53 Discussion ................................ ................................ ................................ .............. 56 Biolu minescence Heterogeneities in Single Cells ................................ ............. 57 Conclusions and Additional Thoughts ................................ .............................. 61 3 LUX EXPRESSION IN RESPON SE TO 3OC6HSL: MA TERIALS AND METHODS ................................ ................................ ................................ .............. 64


7 Vibrio fischeri MJ11, JB10 and ATCC7744 Strains ................................ ................. 64 Vibrio fischeri Cultures ................................ ................................ ............................ 64 Growing the Cultures ................................ ................................ ........................ 64 Culture Media: Defined and Rich Photobacterium Media ................................ 65 Making Sin gle Colony Cultures to Obtain Monoclonal Cells ............................ 67 Bulk Measurements with Varying AI Concentrations ................................ .............. 68 Using Well Plates to Measu re Many Samples in Parallel ................................ 68 Well Plate Readers and Bulk Measurements ................................ ................... 69 Samples for Bulk Measurements ................................ ................................ ...... 70 Measuring Single Cell V.fischeri Luminescence in a Perfusion Chamber over Time ................................ ................................ ................................ ..................... 71 Perfusion Chamber and Optical System Configuration ................................ .... 72 Perfusing and Adhering Single V.fischeri Cells to a Glass Surface .................. 74 Physical Stability of the Optical System ................................ ........................... 76 Orientation of the Cells on the Glass ................................ ................................ 77 Single Cell Data Analysis ................................ ................................ ........................ 78 Programming the Data Analysis Code ................................ ............................. 78 Characterization of the Intensified CCD (iCCD) Camera ................................ ........ 81 Noise Sources in the iCCD Camera ................................ ................................ 81 Calculating the iCCD Camera Gain ................................ ................................ .. 82 4 LUX EXPRESSION IN RESPONSE TO 3OC6HSL AND C8HSL ........................... 85 Introduction ................................ ................................ ................................ ............. 85 Measuring V.fischeri lux Operon Expression in Response to Different 3OC6HSL and C8 Concentration Combinations ................................ ................................ ... 86 V.fischeri JB10 Bulk Measurements ................................ ................................ ....... 87 V.fischeri JB10 Bulk Measurements Using Well trays ................................ ...... 87 V.fischeri JB10 Bulk Measurem ents Results ................................ .................... 88 Studying V.fischeri JB10 Quorum Sensing Induction at the Single Cell Level ........ 92 Measuring Single Cell lux Expressi on in a Microfluidic Device in Response to Different 3OC6HSL and C8HSL Combinations ................................ ......... 92 Single Cell lux Activation Measurements Results ................................ ............. 94 Discussion ................................ ................................ ................................ ............ 102 5 LUX EX PRESSION IN RESPONSE TO 3OC6HSL AND C8HSL: M ATERIALS AND METHODS ................................ ................................ ................................ ... 106 Vibrio fischeri JB10 Strai n and Cultures ................................ ................................ 106 Defined Growth Medium ................................ ................................ ................. 108 Bulk Measurements ................................ ................................ ........................ 108 Cell Sample Preparation for Single Cell Experiments ................................ .... 109 Experimental Configuration and Procedures for Single cell Measurements ......... 110 Optical Equipment and Image Acquisition ................................ ...................... 112 Microfluidic Devices ................................ ................................ ........................ 113 Adhering Single Cells to the Observation Surface ................................ .......... 116 Keeping Adhered Single cells under a Constant External Medium ................ 116


8 Single cell Fluorescence Data Analysis ................................ ................................ 117 Preliminary Trials with a Microfluidic Device Using a Concentration Gradient ...... 119 6 CONCLUSIONS AND FUTU RE DIRECTIONS ................................ .................... 120 LIST OF REFERENCES ................................ ................................ ............................. 125 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 132


9 LIST OF FIGURES Figure page 1 1 Schematic re presentation of gene expression ................................ .................... 14 1 2 V.fischeri quorum sensing regulatory network wi th three autoinducers .............. 16 1 3 V.fischeri luciferase bi oluminescence emission reaction ................................ .... 17 1 4 V.fischeri bi oluminescence emission spectrum ................................ .................. 18 1 5 Chemical structure of HSL autoinducers in V.fischeri ................................ ......... 20 1 6 Noise in gene expression ................................ ................................ ................... 24 1 7 Individual V.fisch eri imaged in dark field and bioluminescence .......................... 27 2 1 Cell density vs. optical density for V.fischeri ................................ ....................... 30 2 2 Bioluminescence emiss ion for wild type V.fischeri MJ1 1 normalized to optical density ................................ ................................ ................................ ................ 33 2 3 Luminescence of single cells following addition of 3OC6HSL ( AI ) ...................... 38 2 4 Sequential dark field and luminescence images for a single V.fischeri cell ........ 39 2 5 Single cell luminescence distribution of activated V.fischeri cells ....................... 40 2 6 Lu minescence activation over time ................................ ................................ ..... 41 2 7 Onset times at high 3OC6HSL ( AI ) concentration ................................ .............. 42 2 8 Variability in signal levels for V.fischeri cells and for c ontrol fluorescent microspheres ................................ ................................ ................................ ...... 44 2 9 Temporal autocorrelation of individual cell luminescence ................................ ... 47 2 10 V.fischeri JB10 lux regulation ................................ ................................ ............. 51 2 11 Heterogeneity of native luminescence versus GF P fluorescence as a QS reporter ................................ ................................ ................................ ............... 53 2 12 Inhibition of V.fischeri biolumines cence by rich (complete) medium ................... 55 3 1 Rich fresh medium suppresses luminescence ................................ .................... 67 3 2 Perfusion cham ber optical configuration ................................ ............................. 72


10 3 3 iCCD camera gain G ................................ ................................ .......................... 84 4 1 V.fischeri JB10 bu lk luminescence and fluorescence at constant OD ................ 89 4 2 Fluorescence vs. bioluminescence of a b ulk JB10 culture at constant OD ......... 91 4 3 GFP fluorescence of V.fischeri JB10 strain as derived from bulk measurements in a plate reader ................................ ................................ ......... 93 4 4 Single cell V.fischeri JB10 GFP fluorescence at 1000 nM 3OC6HSL (Experiment #1) ................................ ................................ ................................ .. 95 4 5 Single cell V.fischeri JB10 GFP fluorescence at 100 nM 3OC6HSL (Experiment #2) ................................ ................................ ................................ .. 96 4 6 Single cell V.fischeri JB10 GFP fluoresc ence at different combinations of 3O C6HSL and C8HSL (Experiment #3) ................................ ............................. 97 4 7 Single cell V.fischeri JB10 GFP fluorescence at different combinations of 3O C6HSL and C8HSL (Experiment #4) ................................ ............................. 98 4 8 Graphical representation of a t test on single cell JB 10 lux expression distributions ................................ ................................ ................................ ...... 100 4 9 Coefficients of variation ( CV F ) of s ingle cell V.fischeri JB10 GFP fluorescence for four different sample se ts using a microfluidics system ............................... 101 5 1 Microfluidics experimental configurat ion picture ................................ ............... 107 5 2 Microf luidic three way device pattern ................................ ............................... 111 5 3 Single cell fluorescence data analysis ................................ .............................. 115


11 LIST OF ABBREVIATION S 3OC6HS L N (3 o xo hexanoyl) L homoserine lactone ( autoinducer ) AHL Acyl homoserine lactone AI Autoinducer C8HSL N o ctanoyl L homoserine lactone ( autoinducer ) CCD Charge coupled device DNA Deoxyribonucleic acid GFP Green fluorescent protein iCCD Intensified charge coupled device OD Optical density PDMS Poly(dimethylsiloxane) QS Quorum sensing RNA Ribonucleic acid


12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy STOCHASTICITY IN QUORUM SENSING : LUX EXPRESSION IN VIBRIO FISCHERI AT THE SINGLE CELL LEVEL By Pablo Delfino Prez August 2011 Chair: Stephen James Hagen Major: Physics Like many bacterial species, Vibrio fischeri can det ect its own population density through a quorum sensing (QS) mech anism. The bacterium releases small signal molecule s ( AI autoinducer s ) which accumulate at high population density and trigger a genetic switch. In V.fischeri this leads to bioluminescence. Little is known about how stochastic gene expression affects QS at th e level of single cells. We imaged the luminescence of individual V.fischeri cells in a flo w chamber and the fluorescence of a gfp reporter using microfluidics. We directly measured the intercell ular variability in lux activation We found a wide heterogeneous response to the external 3OC6HSL AI signal among cells and over time: s ome cells may be strongly luminescent while others are virtually dark. We also studied how the addition of C8 HSL AI signal modulated the heterogeneous response induced by 3OC6HSL. Our results confirmed at the single cell level the C8HSL inhibition effect described in previous studies. Furthermore, they showed that the broad distribution of lux responses depends o nly on the average activation when different 3OC6HSL and C8HSL concentrations are provided. The analysis of noise in the individual cell response can eventually lead to a better understanding of how cells use QS to gather information about their environmen t.


13 CHAPTER 1 INTRODUCTION Quorum Sensing Gene expression in the living cell is carefully regulated to meet the physiological needs of the cell. The expression of protein from a gene is regulated by genetic networks in response to conditions of cell cycle, growth, response to signals, adaptation to the environment, etc. Figure 1 1 sketches gene expression in a prokaryote. Quorum sensing (QS) is one mechanism used by numerous species of bacteria to synchronize and regulate gene expression in accordance with the local bacterial density (1) The QS mechanism employs small diffusible molecules that are synthesized by the bacteria and accumulate in the local environment (2) Higher densit y of these signaling molecules usually indicates a high density of cells. QS was first studied in the bioluminescence regulation of marine bacteria Vibrio harveyi and Vibrio fischeri (3, 4) makin g them par adigms in quorum sensing among G ram negative 1 bacteria. This dissertation focuses on Vibrio fischeri a bacterium that colonizes the light o rgans of sepiolida squids and monocentrid fish like the Hawaiian bobtail squid Euprymna scolopes (5) and the Japanese pinecone fish Monocentris japonica (6 8) The rich environment provided by the host light organ allows the bacteria to grow and re ach enough density to induce the expressio n of its bioluminescence genes. The squid profits from this relationship as the light emitted can mask the shadow a potential predator might see from below. V.fischeri also benefits from this symbiotic relationship because it gets a rich environment that allows it to grow to very high concentrations (10 10 cells per 1 Gram staining: It is a widespread empirical way of separating bacteria into two large groups, positive and negative, based on the properties of their cell walls, which affect the way they react to the Gram stain.


14 ml of light organ fluid) compared to the low populations levels when found in free seawater (<100 per ml) (2) Fi gure 1 1: Schematic representation of gene expression. The DNA molecule encodes the amino acid sequences necessary for the synthesis of proteins. The double stranded DNA molecule is first used as a template to synthesize an mRNA single strand molecule in a process called transcription. In a second step called translation, the ribosome and tRNA molecules use the mRNA as a template to synthesize the corresponding protein. In prokaryotes, as in Vibrio fischeri the cell does not have a nucleus membrane, and th e mRNA does not need to mature and can be immediately translated into proteins. There are a few broad groups of QS mechanisms described in the literature. Acyl homoserine lactones ( AHL ) are used as signaling molecules by Gram negative bacteria (1, 9) as in the LuxR/LuxI system of V.fischeri Gram positive bacteria use small peptides as QS signals (1, 9) AI 2 is a group of homolog molecules used by hundreds of bacterial species, including V.fischeri as well as other Gram positive and negative bacteria (10, 11) The LuxS/AI 2 system has been proposed to function as an interspecies communication mechanism (9)


15 QS regulates both light production and host colonization behavior in V.fischeri (12) However, other bacteria use a QS mechanism to coordinate very different activities su ch as the formation of biofilms, competence for DNA uptake, antibiotic production, sporulation, etc. (9, 13, 14) There are even bacteria that coordinate the production of toxins to attack a host tissue only when the bacterial population has reached high enough levels (15 18) tim e fighting back an infection that has just been detected. Efforts directed to a better understanding of the QS mechanisms in general can thus provide novel ways to fight pathogens by disrupting their communication systems, instead of directly targeting the bacteria in the manner of conventional antibiotics (19 22) The QS mechanisms, including those of V.fischeri have been under intense study for the last two or three decades but there is much that is still unknown. It has been proposed that QS not only serves as a cell communication system, but it can be extended as a way for bacteria to probe its environment (23 25) For example, the bacteria would release their signaling molecules to test how far the signals diffuse sen approaches (26) Quorum Sensing in Vibrio fischeri V.fischeri has been a model system for the study of QS for several decades (27) The V.fischeri ES114 genome the model strain symbiont of the squid Euprymna scolopes has recently been fully sequenced (28) revealing a total of 4.2 million base pairs between its two chromosomes. Much successful work has been done towards revealing the QS biochemical pathways and their corresponding key chemical


16 molecules. However, little is known about how the cell actually integrates different autoinducer signal channels to generate a response. Figure 1 2 : V.fischeri quorum sen sing regulatory network with the three autoinducer inputs. Figure adapted from Visick (63) The V.fischeri QS system is regulated by three autoinducers: 3OC6HSL, C8HSL and AI 2. The two parallel lines in the upper part represent the outer (top) and inner (bottom) cell membranes. The autoinducers are synthesized in the cytoplasm but can diffuse out of the cell and back in. 3OC6HSL forms a complex with LuxR that str ongly activates luxICDABE transcription producing luminescence. However, C8HSL can also bind to LuxR and weakly activate luxICDABE transcription. C8HSL induces the synthesis of LuxR receptor by dephosphorylating LuxO and preventing the synthesis of small R NAs (sRNA), which in turn repress litR AI 2 has minimal effect on luminescence regulation. The cooperative host colonization is species specific and metabolically expensive to the bacteria. It is cooperative because both the host and bacteria benefit from the


17 relationship; and species specific because each particular V.fischeri strain colonizes a particular marine animal. The symbiosis is tightly regulated by the host and V.fischeri On one hand, it was observed that t he squid host developed mechanisms to prevent dark mutants (non luminescent V.fischeri ) from benefiting from the symbiosis in the light organ (5, 29, 30) On the other hand, V.fischeri produces the metabolically expens ive luminescence only at high colonization densities (31 33) The QS system regulates the lux operon 2 which gene rates the luminescent emission. The QS also regulates behaviors associated with host colonization, including cell motility and biofilm formation Figure 1 2 shows the detail of QS regulation in V.fischeri by three known autoinducers. Figure 1 3: V.fischeri luciferase bioluminescence emission reaction. The genes for s ynthesis of the luciferase enzyme are contained within the lux operon and regulated by the LuxI /LuxR quorum sensing mechanism. Luciferase uses flavin mononucleotide (FMNH 2 ), a long chain fatty acid and oxygen as substrates for the reaction. Bioluminescence reaction in Vibrio fischeri The cytoplasmic luciferase enzyme catalyzes the luminescence emission reaction in V.fischeri and is synthesized when the lux system is activated. It uses oxygen, a 2 Operon : Group of genes transcribed as a unit in response to the activation of its promoter.


18 long chain fatty aldehyde and a reduced flavin mononucleotide (FMNH 2 ) as substrates (34) Figure 1 3 depicts the bioluminescence emission reaction. In the reaction the oxygen is reduced to H 2 O; FMNH 2 and the long chain fatty aldehyde are oxidized to FMN and a long chain fatty ac id; and photons are emitted with a maximum wavelength at ~ 493 nm Figure 1 4 shows the luminescence emission spectrum of V.fischeri which appears to the eye as a blue green light Figure 1 4: V.fischeri bioluminescence emission spectrum. V. fischeri cel ls produce their own light when the quorum sensing mechanism is activated. The curve shows the spectrum of the luminescence emission intensity which peaks near 493 nm wavelength. The emission has a blue green color when a bright culture is observed by eye. The spectrum was measured in a bulk culture with a Jasco FP 750 Spectrofluorometer. The bioluminescence is energy intensive for the cells. It was estimated that in very bright induced cells growing in a rich medium, luciferase accounts for 5% of the total cell proteins and the bioluminescence can use up to 20% of the oxygen consumption (4) The luciferase reaction emits between 0.1 and 1 photon per cycle, and 6 to 60 ATP


19 molecules are used per emitted photon. The biol uminescence emission from one V.fischeri cell is on the order of 0.01 1000 photon/s depending on the induction level (34) 3OC6HSL lux regulation The lux genes controlling bioluminescence in V.fischeri comprise the l ux operon (or luciferase luxICDABEG operon) (1, 9) The expression of the lux operon for the production of light is controlled by two proteins, LuxI and LuxR. LuxI is the autoinducer synthase 3 which p roduces the N (3 oxohexanoyl) L homoserine lactone (3OC6HSL) autoinducer (35 37) Figure 1 5 shows 3OC6HSL chemical structure. The LuxR cytoplasmic 4 receptor is activat ed by binding to 3OC6HSL, and then consequently binds to the DNA to promote the lux operon for the production of light (35, 38) When the LuxR rece ptor is free and unbound to the autoinducer ( AI ) signal, it is unstable and quickly degraded (39) The AI s can freely diffuse in and out of the cell (2) and their external concentration generally depends on the concentratio n of bacteria. More AI in the environment indicates a higher concentration of cells. 3OC6HSL binds to the LuxR receptor, forming a dimeric complex that has two funct ions in activating the luxICDABEG operon (40) On o ne hand, it activates the transcription of luxAB and luxCDEG which encode the two subunits of the bacterial luciferase enzyme ( Lux A and Lux B) and the fatty acid reductase complex ( Lux C, Lux D and Lux E) for the metabolism of lu minescence precursors respecti vely (34) On the other hand, the LuxR 3OC6HSL complex also 3 Synthase: Enzyme that catalyzes a synthesis process. 4 Cytoplasm: Aqueous f luid that fills the cell, and the place where most of cellular activities are performed


20 induces the expression of the 3OC6HSL synthase LuxI. The production of LuxI generates more 3OC6HSL in the environment, setting up a positive feedback loop ( F igure 1 2 ) This regulatory system causes the entire population of cells to start producing light when a critical concentration of 3OC6HSL (or effective density of V.fischeri producing 3OC6HSL) is reached. Thus, the QS mechanism allows the cell to switch b etween low and high luminescence phenotypes depending on the local 3OC6HSL concentration. Figure 1 5: Chemical structure of HSL autoinducers in V.fischeri The N (3 oxohexanoyl) homoserine lactone (3OC6HSL) and N octanoyl L homoserine lactone (C8HSL) are shown. 3OC6HSL regulates the LuxI/LuxR system and C8HSL regulates the AinS/AinR system. C8HSL and AI 2 lux regulation The host colonization and luminescence emissi on is carefully regulated in multiple stages by the QS system (12, 29, 41) Besides 3OC6HSL autoinducer, QS in V.fischeri is regulated by two other known autoinducer pathways: N o ctanoyl L homoserine lactone (C8HSL) autoinducer synthesized by the AinS synthase (42) ( Figure 1 5 ); and AI 2 autoinducer, a furanosyl borate diester produced by the LuxS synthase (10, 32)


21 The 3OC6HSL, C8HSL and AI 2 autoinducers are simultaneously pr esent in the cell cytoplasm, although their relative and absolute concentrations vary. The mechanism by which an individual cell integrates several signals to generate its responses is under active study. In this dissertation we studied how V.fischeri int egrates the 3OC6HSL and C8HSL signal s into the lux operon expression. Figure 1 2 depicts the three autoinducer pathways that regulate bioluminescence and colonization phenotypes in V.fischeri The second QS signal C8HSL is synthesized by the AinS enzyme, w hich belongs to the LuxM group of acyl HSL synthases (42) At very low cell densities, phosphorylated LuxO regulator represses the luminescence by indirectly repressing the litR transcriptional regulator via small RNAs (sRNA) (29, 43, 44) At higher intermediate cell densities, C8HSL concentration increases. It inactivates (by dephosphorylation) LuxO regulator which represses the tra nscription of litR regulator. As a second effect, it increases the transcription of litR (43) litR encodes the transcriptional regulator LitR that positively promotes the transcription of luxR C8HSL participates i n the regulation of the lux system through the AinS/AinR channel, and is involved in the regulation of host colonization cell behaviors (12, 29, 41) Furthermore, C8HSL is capable of cross talking with the lux system acting directly on LuxR. C8HSL autoinducer can bind to LuxR regulator and form a complex that directly promotes the expression of the lux box, but with a lower affinity than the 3OC6HSL LuxR complex. C8HSL can induce l uminescence at very low cell density when little or no 3OC6HSL is present. However, it acts as a luminescence inhibitor at higher cell density by competing with 3OC6HSL in binding to LuxR (45)


22 The third autoinducer in V. fischeri is AI 2, a furanosyl borate diester. The LuxS synthase that produces AI 2 was found in several Gram positiv e and negative bacteria, and has been proposed to have a role in bacterial interspecies communication (11, 46) In V.fischeri the effect of AI 2 signal has been shown to have a role in lux operon and colonization regulation. AI 2 is detected by the periplasmic LuxP and LuxQ receptors, and it feeds the same channel as the AinS/AinR sy stem. The effect on lux expression, however, is quantitatively smaller than the one generated by 3OC6HSL and C8HSL signals. In a previous study, a luxS mutant V.fischeri produced 70% of the luminescence of a V.fischeri wild type (47) Furthermore, strain mutants lacking the HSL synthases did not produce detectable luminescence (47) In all our single cell experiments described in this dissertation, the exogenous AI 2 concentration was eff ectively set a t zero. The description of the QS genetic network we present here is only the core showing the dominant interactions that regulate luminescence. In fact, a recent study discovered 18 additional genes in V.fischeri directly regulated by 3OC6HS L autoinducer, and most of them directly by the LuxR transcriptional regulator (48) Another study (12) has shown that the AinS pathway controls (positively and negative ly) at least 30 other genes including cell motility genes essential for successful colonization (49, 50) Noise in Gene Expression The biochemical processes occurring inside cells (as well as outsid e them) are fund amentally stochastic in nature. The molecules participating in the reaction move randomly in the medium and have a certain chance of reacting when they interact. If the number of molecules in the system is very large, then it is reasonable and useful to use


23 an averaged description of the reactions. Nevertheless, the number of copies of regulatory molecules in cells, including DNA, is often very small (~100 500 or less ) (51, 52) S tochasticity and fluctuations can then play a very important role in the generation of phenotypic differences (noise) between cells in a population with the same genetic composition (53 55) when the molecule numbers are low (51, 56, 57) are cl onal, or genetically identical (55) Noise in gene expression can be categorized as extrinsic or intrinsic in its origin (51, 58) As was mentioned before, th e production of proteins from the transcription 5 of genes is controlled by regulatory molecules in the cytoplasm. The concentrations and distributions of these molecules can vary from one cell to another, making possible fluctuations in the output prod ucti on of genes in two isogenic 6 cells. This kind of noise is define d as extrinsic or global noise. Variations in ribosomes, RNA polymerase, transcription factors and other cellular machinery shared by different genes contribute to extrinsic noise (51) On the other hand, even two identical cells with the same number of copies of regulatory molecules and the s ame concentrations of nutrients would still be subject to fluctuations or variations in the small numbers of molec ules involve d in key biochemical reactions. This would lead to differences in the expression of the same gene in two otherwise ident ical cells (in the same state). This variation is known as intrinsic noise. From another point of view, if a single cell has two identical copies of the same gene, 5 T ranscription: Process by which messenger RNA is synthesized from a DNA gene, resulting in a transfer of information that will lead to the generation of proteins. 6 Isogenic: Containing the same genetic information.


24 there will be small differences in their expression due to intrinsic noise. Different population levels of mRNA, transcription, small numbers of regulatory proteins, mRNA decay and other discrete biochemical events c ontribute to intrinsic noise (51, 59) Figure 1 6 depicts the intrinsic and extrinsic noise in gene expression. Figure 1 6: Noise in gene expression. Figure adapted from Elowitz et al. (51) Schematic of intrinsic and extrinsic noise. Two genes identically regulated are inserted in a genome, indicated by red and green genes. In any genetic pathway, both intrinsic and extrinsic noise are often prese nt. (A B) Extrinsic noise is produced by cytoplasmic fluctuations in the genetic machinery that controls protein synthesis from both genes (ribosome numbers, levels of RNA polymerase, transcription factors, etc). (A) In a single cell both red and green are equally expressed; however (B) there are differences among cells. (C D) Intrinsic noise is produced by discrete stochastic variations of biochemical origin. (C) In a single cell red and green expression are different. (D) Extrinsic noise produces differen t overall levels among cells (as in B); however, the addition of intrinsic noise produces different relative green/red expressions in different cells. The sources of genetic noise are still under st udy but progress has been made. When studying systems at t he single cell level, the timescale (or characteristic time) at which fluctuations occur can suggest whether the noise is intrinsic or extrinsic. Recent


25 studies by Rosenfeld et al. (52) in Escherichia coli strains fo und characteristic times of less than 10 minutes for intrinsic noise and ~40 minutes for extrinsic noise. It was suggested that the longer timescales of extrinsic noise were more relevant in generating phenotypic differences. In our experiments described i n Chapters 2 and 3 of this dissertation, we observed small fluctuations while measuring single cell bioluminescence signals. We associated with intrinsic noise the occasional short term (~10 20 min) uncorrelated (among cells) fluctuations in the signal of an individual cell over time. On the other hand, the long term (~1 h our) variations among cells possibly originate in extrinsic noise It is not completely understood to what extent noise affects different genetic networks. Furthermore, it is of interest t o investigate how genetic systems incorporate noise. There are known cases where bacteria take adv antage and rely on it (60, 61) Single cell measurements are necessary to fully understand the se mechanisms. Objectives of this Dissertation Vibrio fischeri is a model G ram negative organism that has a quorum sensing (QS) system with many structural similarities to tho se of other bacteria. Numerous G ram negative bacteria, including pathogens, emplo y an acyl homoserine lactone ( AHL ) signaling system (62) and many of them synthesize more than one AHL The AHL QS systems are usually of the LuxI/LuxR type as the 3OC6HSL channel in V.fischeri However, some AHLs are linked to a LuxM type system like the AinS/AinR channel in V.fischeri LuxM systems are not associated with a dedicated cognate transcriptional receptor that forms a complex with the autoinducer and directly regulates transcription. I nstead LuxM sy s te ms us e a phosphorelay cascade that involves the transference of phosphoryl groups in a sequence of steps that end with the regulation of


26 the target genes Furthermore, V.fischeri possesses an AI 2 autoinducer (that uses the same channel as C8HSL) also foun d in many Gram positive and negative species. V.fischeri is a desirable choice for experimental studies, as it is non pathogenic, its QS activation produces a directly measurable response (bioluminescence), and it shares similarities with other species as explained above. The bioluminescence is a direct product of the QS activation and, however weak, it is measurable at the single cell level in a wild type strain. In the first set of experiments described in Chapters 2 (Results and discussion) and 3 (Method ology) we used a wild type strain and studied the QS system under different concentrations of the 3OC6HSL QS signal by measuring the reporting luminescence. We aimed at quantitatively measuring how the population was activated at the single cell level and how the single cell emission fluctuated over time. We studied how accurately the single cell lux activation responded to (or processed) a cons tant and well defined 3OC6HSL signal. To achieve this we first acquired the bioluminescence signal of bulk bacteri al cultures, essentially characterizing the average behavior. In a second stage, we used a high sensitivity microscopy perfusion system. We imaged and measured the single cell luminescence emission under different well controlled concentrations of exogenou s 3OC6HSL ( Figure 1 7 ) This configuration allowed us to define the external medium of the individual bacterium while quenching the LuxI/LuxR autoinduction feedback. Our second set of experiments is described in Chapters 4 (Results and discussion) and 5 (M ethodology). Adding a second variable to the system, we exposed V.fischeri populations to different 3OC6HSL and C8HSL autoinducer concentration combinations. For these experiments we used a V.fischeri JB10 strain that synthesizes


27 green fluorescent protein as a lux reporter 7 becoming fluorescent when the QS system is induced. We first characterized the bulk population behavior. We designed a microfluidic microscopy system to perform the needed higher number of measurements, acquiring data as in the perfusio n chamber but with a higher throughput. We aimed at observing and measuring how the addition of C8HSL modulated the single cell activation across the cell population. Our objective was to measure how the cell integrated the two inputs provided by 3OC6HSL a nd C8HSL into a single lux response. Figure 1 7: Individual V.fischeri imaged in dark field and bioluminescence. (A) Dark field (externally illuminated) and (B) bioluminescence (light emission) images of V.fischeri cells adhered to the glass window o f th e perfusion chamber at 24 C in the presence of 500 nM exogenous 3OC6HSL autoinducer. The cells appear as rods (~3 5 m long) in the dark field image. The bioluminescence image shows in false color the luminescent emission de tected in a 16 minute exposure. Images we re collected in an inverted microscopy configuration with an intensified CCD camera and a 100 oil immersion objective. Reprinted from (64) 7 Reporter: A gene that encodes an easily observable product and is inserted under the same promoter as the genes of interest. By measuring the product of the reporter, it is possible to measure when the genes of interest are activated.


28 CHAPTER 2 LUX EXPRESSION IN RESPONSE TO 3OC6HSL Measuring Single Cell Vibrio fischer i Biol uminescence Single V.fischeri cells emit a very weak luminescence that is much dimmer than the light produced by the fluorescence of a cell expressing E GFP 8 A single cell produces an estimated bioluminescence of 10 2 to 10 3 photons/s d epending on it s metabolic state, induction level and the bacterial strain (34, 65) This emission is weaker than that of a cell expressing a fluorescent protein as a reporter. However, the weak bioluminesce nce emission of single V.fischeri cells can be detected with a photomultiplier (66) or an intensified CCD camera (67 69) During our single cell luminescence experiments (Chapters 2 and 3), V.fischeri had an estimated luminescence emission of 30 photons/s per cell from which we were able to detect 1.7 photons/s. On the other hand, in Chap ters 4 and 5 we describe single cell fluorescence measur ements where the estimated fluorescence emission was 2500 photons/s per cell from which we were able to detect 100 photons/s. We needed to take many experimental measures to ensure that the collected bio luminescence output and its fluctuations were appropr iately recorded. The experimental configuration we used, described in detail in Chapter 3, needed to have enough stability to allow us to record over long periods of time small intensity light fluctuations in the order of 50 photons per minute per cell We also had to take into account that b acteria grow and repli cate on very short time scales. They constantly modify their environment, move vigorously and can mutate to adapt to new conditions. 8 EGFP: Enhanced green fluorescent protein. It is a modified version of the wild type GFP found in the Pacific jellyfish Aequoria Victoria EGFP is mutated to make it more suitable for laboratory use. Its excitation peak is shifted from 396 nm to 488 nm wavelength, its fluorescence emission is brighter and it under goes faster maturation (91, 100)


29 In this cha pter we describe and discuss our bulk and single cell bio luminescence measurements. We implemented a perfusion flow in the single cell experiments that fixed the extracellular concentrations. Save for some preliminary experiments, we used MJ11 V.fischeri wild type 9 strain for these measurements (Chapters 2 a nd 3) The MJ11 strain was extracted from the light organ of the Japanese pinecone fish ( Monocentris japonica ) and was provided to us by Dr M.J. Mandel and Dr E.G. Ruby 10 Biol uminescence Bulk Measurements with MJ11 V.fischeri Strain We performed bulk measu rements in order to characterize the average bioluminescence prior to designing and interpreting the single cell experiments. Bulk measurements are the more usual approach when studying bacteria. In these experiments, a large population of bacteria is stud ied ( ~ 200 l of liquid culture at ~ 10 5 10 8 cells/cm 3 ). Bulk experiments are generally done in a relatively closed system : d epending on th e system configuration there can be only some very mild medium evaporation and gas exchanges between the liquid medium and the ro om air. The cells are generally placed in a small volume of liquid culture (100 500 l) where they can grow and perform their metabolic functions. The cells will indeed modify the environment changing the chemical concentrations when consuming nutrients an d producing excretions. They will also replicate until the nutrients are severely diminished and will eventually die T he latter can take a few days depending on th e strain and growing conditions 9 Wild type: A wild type strain of a bacterial species is one that does not have any mutation in its genome introduced by a laboratory procedure. 10 Mandel M.J., S tabb E.V., Ruby E.G., Ferriera S., Johnson J., Kravitz S., Beeson K., Sutton G., Rogers Y. H., Friedman R., Frazier M., Venter J.C. ; "Complete sequence of Vibrio fischeri strain MJ11."; Submitted (AUG 2008) to the EMBL/GenBank/DDBJ databases.


30 Figure 2 1: Cell density vs. optical density for V.fis cheri The optical density (OD) was measured in a Shimadzu spectrophotometer at 600 nm wavelength and 1 cm path length. The OD and cell density are proportional for OD values ~ 0.1 1. Around 100 cells were counted in a hemocytometer to measure the cell dens ity from an initial sample with 3.13x10 9 cell/ml. Serial dilutions were later performed and the OD was measured to calculate the cell density of the diluted samples. During our bulk measurements we observed ~ 10 5 10 7 cells at the same time confined in a few hundred microliters according to the optical density vs. cell density calibration described in the following section These measurements showed us how the


31 average V.fischeri cell behaves when put under some typical 3OC6HSL AI concentrations ( ranging from 0 to 5000 nM ) B ulk measurements are qualitatively different than single cel l ones. As mentioned above, bulk measurements provide a lower level of control on the bacterial environment than perfusion single cell experiment s that fix the external concentrat ions. However, they provide information on characteristic autoinducer concentrations response times and response dynamics to different autoinducers. Obtaining this information is important to know what to expec t from single cell data and how to interpret them Using Well trays to Measure Bulk Cultures Well trays are widely used in the biological science community. These small plastic trays come in a variety of formats and the y are a few tens of centimeters wide and long. They are made of transparent polyst yrene with clear lid and bottom for optical measurements. More technical details about the sample preparation and we ll plate experiments can be found in Chapter 3 of this dissertation. Our first preliminary experiments were done in a custom made well tray reader using a photomultiplier that measured bioluminescence emission Jonathan Young, Elaine Johnson and Leslie Pelakh performed some of these bulk experiments. We later started using a Biotek automated well tray reader (Biotek Synergy 2, Biotek Instrumen ts) that als o allowed us to measure turbidity luminescence and fluorescence (see Chapters 4 and 5 for the latter ) The plate is placed in a dark light sealed compartment (temperature ~ 25 C) and the bio luminescence measurements are taken from the bottom or top of each well. The plate is shaken in between measurements for keeping the liquid sample homoge nized and oxygenated. Wells typically contained 150 200


32 (detailed ingredient list in Chapter 3) and a small drop (10 culture. The culture was washed in fres h medium to remove any autoinducers the cell might have produced while growing. We added different concentrations of exogenous 3OC6HSL autoinducer in each well (ranging from 0 to 5000 nM) (N ( ketocaproyl) L homoserine lactone, Sigma Aldrich CAS: 143537 62 6 ) The cells were left to grow in the plate for several hours (6 12 hours) while performing measurements on the wells. The Biotek programmable reader also al lowed us to monitor the turbidity of each well during the experiment. The turbidity wa s measured at 600 nm wavelength The turbidity per 1 cm path length is the specific optical density (OD) and for low values ( up to 0.1 1 ) the OD is proportional to the concentration of cells in the sample (70) Knowing the OD of the sample is very important as it allowed us to know when the cells reached exponential pha se. We observed lag ph ases of ~3 5 hours (depending on the i nitial cell density) follow ed by an exponential phase of ~4 hours. Using a Shimadzu spectrophotometer, we calibrated OD to cell density values. Figure 2 1 shows the linear relationship between OD and cell density. MJ11 Bulk Well tray Measurement Results Cu ltures were left to grow in the well plate for several hours under different initial 3OC6HSL autoinducer concentrations. When the bio luminescence of a well plate sample is measured, the obtained number is proportional to the absolute bio lumin escence by an unknown constant. However, knowledge of the value of this constant is not necessary as long as the same constant applies to different wells and over time in a given experiment. The absolute measure of the biolumine scence would be the total count of photons emitted by the observed sample in all spatial directions. By contrast, our measured value for the bio luminescence depends on the fraction of


33 collected light given by the optics geometry of the well tray reader, its photomultiplier gain (kept constant duri ng all the experiment) and the ADC converter. The Figure 2 2 shows the bio luminescence of V.fischeri in a 48 well tray growing in defined medium normalized to OD, making it proportional to the bio luminescence emission per cell. We can see here how the cell s are starting to satu rate their bio luminescence at 3OC6HSL AI concentration of ~ 1000 nM This saturation concentration was also observed in the average behavior of preliminary single cell experiments. The figure shows that adding increasing 3OC6HSL to the samples increases the bio luminescence emission, with a steep increase between 100 nM and 500 nM. We can also observe how the bio luminescence activation in response to the exogenous 3OC6HSL takes ~ 100 minutes to reach a steady state. Figure 2 2: Biolumin escence emission for wild type V.fischeri MJ11 normalized to optical density. The measurement was taken in a 48 well plate in defined photobacterium medium. Exogenous 3OC6HSL AI was added at time t = 0. At t = 70 minutes the response is still developing, a nd for the curves t > 130 minutes it has already reached a steady state. The dotted curve corresponds to the fit to a cooperative binding model, with equilibrium constant and Hill coefficient Reprinted from (64)


34 We can fit this data using a simplified cooperative binding model (71) The 3OC6HSL AI binds directly to the LuxR receptor forming a complex C [2 1] This model predicts a complex C concentration proportional to the concentration of AI We assume a number n of complexes C bind directly to the lux promoter region in the DNA ( Plux ) forming an active complex ( n C : Plux ) [2 2] The activated lux promoter complex ( nC: Plux ) would be directly responsible for the light production. The fraction of activated lux promoter Y would be proportional to the light intensity output. In the equilibrium state, the fraction of activated l ux promoter would be Y. [2 3] T he brackets indicate molar concentrations. Since C will be proportional to AI we can absorb the proportionality constant in and make it K which is the dissociation constant. [2 4] The fraction Y of activated lux promoter is proportional to the bio luminescence output. We fitted our bulk measurement data to this cooperative binding model using a


35 least squares method (dotted curves in Figure 2 2 ). We obtained the values and This information about the average behavior does not contain single cell information, and it hides the high heterogeneity we observed in the single cell experiments we describe next. However, it provided i nformation about the activat ion time and 3OC6HSL induction concentrations for our V.fischeri strain; as well as a reference for the future single cell experiments. Single cell Biol uminescence of V.fischeri MJ11 in Response to 3OC6HSL Bulk measurements allo w a good control of sample conditions. Nevertheless V.fischeri can st ill modify the environment. Bacteria are constantly consuming nutrients and producing secretions. In the closed environment of the bulk sample the growth medium is not renewed, and eventu ally the nutrients deplete and the secretions accumulate. Using a flow perfusion chamber allowe d us to maintain the cells in a controlled environment where the extracellular concentrations are fixed and known. Any AIs produced by the bacteria are washed aw ay. Thus, t he lux autoexcitation (by 3OC6HSL AI accumulation) is negated and the cells are left exposed only to the exogenous 3OC6HSL AI provided More importantly, single cell studies in the chamber provided us with not only the average behavior but they also allowed us to see a big heterogeneity in the single cell response. Acquiring Single cell s cheri Biol uminescence We used a custom microscopy configuration we built in our laboratory to perform these measurements. T he schematics of our home built ap paratus and more technical details can be found in Chapter 3. The bacteria were placed inside a small cylindrical perfusion chamber of 1.5 cm 3 capacity. The chamber had two transparent glass


36 coverslip windows spaced by 5 mm on its to p and bottom circular b ases. The cells were immobilized inside the chamber on the upper surface of the bottom window. An upward directed 100x microscope objective (P lan oil immersion, NA 1.25, in finite conjugate ) collected the emitted light from below. W e used an intensified CCD cooled camera (512x512 pixel, I MAX 512 T operating at 35 C, Princeton Instruments, Princeton NJ) to acquire all our images. The perfusion chamber was carefully filled and perfused with a low flow rate (0.2 ml/h) of fresh defined growth medium. The mediu m contained nutrients, oxygen and a known concentration of exogenous 3OC6HSL autoinducer. Short chain acyl homoserine lactones such as 3OC6HSL can freely diffuse in and out of the cells (2) effectively setting the cyt oplasmic concentration equal to that of the perfused medium. Our experiments were usually 6 8 hours long and the data were collected during approximately 4 hours. During thi s time we alternatively collected dark field and bio l uminescence images. We illumin ated the cells with external light to take dark field images. Each of these images was taken in less than a minute and they gave us the location of the cells. Using the dark field images we could tell whether any cells moved or somehow changed in between b iol uminescence images. Immediately after eac h dark field image we acquired its corresponding biol uminescence image. Using the iCCD camera intensification and shutting off all external light, we acquired a 10 minute long exposure collecting single cells bio luminescence emission. Figure 1 7 shows the imaging of single cells in a pair of corresponding dark field and bio luminescence images.


37 We performed several tests to ens ure the stability of the system and to be sure the small single cell fluctuations we obs erved had biological and not artifactual origin. We checked that the cells remained stationary and in focus during the 4 6 hours of measurements, and we verified the s tability of the physical system and the data acquisition When selecting bacteria in the images, we took care to select only isolated cells that were not in contact with others. The analyzed cells were no less than ~ 10 m from their closest neighbors. The very few bacterial clusters on the observation surface were typically composed of 2 4 cells. These were left aside and were not part of the data analysis. Thus, the perfusion flow m aintained the cells isolated and negat ed crosstalk among observed cells. Single cell Biol uminescence Results In this chapter, we present measurements of single cell V.fischeri MJ11 bioluminescence over time for four different 3OC6HSL concentrations: 50, 100, 200 and 1000 nM ( Figure 2 3 ). We sa w no activation of bio luminescence when no exogenous 3OC6HSL autoinducer ( AI ) was provided, showing that the chamber flow removed any significant natural AI the cells produced. The luminescence reached a saturating level at 3OC6HSL concentrations of 1000 n M and above. This is similar to the behavior of bulk samples ( Figure 2 2 ). U sing the described optical configuration we could track in time the bioluminescence of single cells. As in the bulk measurements, we observed that the aver age single cell took aro und 100 150 minutes to achieve stable bioluminescence output. Figure 2 4 shows the bioluminescence output of a single cell. The image also shows for different points in time the corresp onding bioluminescence and dark field images for that particular V.fisc heri cell.


38 Figure 2 3: Luminescence of single cells f ollowing addition of 3OC6HSL ( AI ). At each AI concentration, roughly 25 40 MJ11 cells were imaged repeatedly over a period of ~4 h following introduction (at t = 0) of exogenous AI at the indicated con centration. The light emission from each cell was quantified thr ough analysis of a series of 10 minute camera expo sures The state of induction of the initial cell culture determines the luminescence of the cells at t = 0. However, once adhered in the flow chamber and exposed to the flow of medium (containing exogenous AI ), the cells respond by adjus ting their luminescent output. This leads to a transient increase or decrease i n the emission over the next ~1 2 h. After ~3 h the cells have a dapted to the app lied AI level. The control shows an experimental verification of the stability and sensitivity of microscopy and data analysis. For this control green fluorescent latex spheres were illuminated with a n attenuated light source and then imaged with the same experimental parameters and data analysis, as used for the V.fischeri measurements. The time dependence of all emission vs time curves in this figure has been smoothed by a Gaussian filter with width = 10 minutes. Reprinted from (64)


39 Figure 2 4: Sequential dark field and lumin escence images for a single V.fischeri cell. (A) Dark field and (B) bioluminescence images of an individual cell adhered to the window of the perfusion chamber, and (C) luminescence levels extracted from the se images. (The luminescence trajectory ha s not been Gaussian filtered.) Images were collected at the numbered time points indicated in (C). Reprinted from (64) Each individual curve i n Figure 2 3 represents the bio luminescence detected from a single cell. The first half of each graph ( 0 100 minutes approximately) is a transient phase before the cells stabilize to the exogenous AI concentration. In some cases, as in the insets correspon ding to 50 and 100 nM, the cells are actually losing some transient initial activation they had acquired in the bulk culture in which they were prepared However, this transient is useful to show that the cells lose previous activation o n a similar time sc ale as they gain it ( ~ 120 minutes) and that the signal we observe when the cells stabilize (after ~100 minutes) is due to the exogenous 3OC6HSL For a given 3O C6HSL concentration, the single cell signals present heterogeneity in different respects : in the degree of activation of the cells in the time scale of activation and in the level of activation of a single cell over time. We discuss these heterogeneities in the following paragraphs.


40 Figure 2 5: Single cell luminescence distribution of activated V .fischeri cells. The 3D histogram shows the detected luminescence distribution of four different populations under different 3OC6HSL concentrations. All the shown luminescence corresponds to a late time ( ~ 200 minutes) when the cell population is fully acti vated (as if it were a snapshot in time). The luminescence distribution broadens the more the cells activate (200 and 1000 nM in yellow and red). The inset shows the same information in logarithmic scale. The variation in the bio luminescence among activate d cells is possibly the most visible in Figure 2 3 When the V.fischeri population is fully activated the single cell bio luminescence covers a very wide range of values. This can be easily seen for [3OC 6HSL] of 200 and 1000 nM. In the later times of these insets some cells are still very dark, some have a moderate activation and a few are very bright. Taking a snapshot in time when the cells are fully activated (~200 minutes from exogenous AI introduction ), we can compare in Figure 2 5 the bio luminescence distribution of the four populations at different AI concentrations. When a population activates, it is not just the


41 mean bio luminescence that becomes higher as we could observe in the bulk measurements. The activation also br ings a very wide distribution, especially in the populations at 200 and 1000 nM. Figure 2 6: Luminescence activation over time. A) V.fischeri MJ11 luminescence histograms for the indicated times after introduction of 1000 nM of 3OC6HSL at t = 0. B) Median cell brightness and 25% and 75% percentiles of brightness. As the cells become activated due to introduced exogenous AI the distribution broadens. Even at 4 hours some cells still emit at detection threshold ( ~ 10 20 photons/minute/cell). Reprinted from (64)


42 The heterogeneity can also be observed in the activation time. Figure 2 3 shows that some cells respond very quickly after the 3OC6HSL AI introduction ( t = 0) and some others take several hours to start showing a response Figure 2 6 shows how the population bioluminesc ence distribution spreads over time when they are exposed to 1000 nM of 3OC6HSL. After ~120 minutes the population is activated and there is a wide dispersion in the single cell bioluminesc ence when adding a high concent ration of 3OC6HSL (usually 200 nM or more). Nevertheless, even when the population is activated, a significant fraction of the population still remain s dark as can be seen in the low value of the 25% percentile in Figure 2 6 and in Figure 2 3 Figure 2 7: Onset times at high 3OC6HSL ( AI ) concentration. The figure shows the distribution of luminescence onset times t 1/2 for single cells at 3OC6HSL concentrations of 200 nM and 1000 nM. We define the onset time t 1/2 of an individual cell as the time at which the luminescence output I ( t ) of that single cell is halfway between its initial value I ( t = 0) and its final value I ( t 250), when AI was introduced at t=0. Reprinted from (64) We show this activation time heterogeneity by defining the onset time t 1/2 of a single cell. We define the onset time t 1/2 of an individual cell as th e time at which the


43 bioluminesc ence output I(t) of that single cell is halfway between its initial value I(t = 0) and its final value I(t 250) when AI was introduced at t = 0. The distribution of onset times t 1/2 is shown in Figure 2 7 We observed a ve ry flat and broad distribution of onset times t 1/2 The figure shows how this distribution remains broad as the 3OC6HSL concentration increases from 200 nM to 1000 nM. We could also observe temporal fluctuation in the emission from individual cells at late r times ( t > 2 hours) As can be observed in Figure 2 3 the individual signals suggest that the bioluminesc ence of some cells occasionally fluctuate by ~20 40% of their light output o n time scales of ~30 minutes. Testing the data acquisition with very sta The observed variability in bioluminesc ence output motivated us to test the stability of our detection, using a stable light source as a control. Our objective was to quantify the uncertainty in the light detection and verify whethe r the apparent emission variations were actually originating in the cells. The premise was using weak microscopic li ght sources of similar size and brightness as the cells but with uniform and temporally stable light output. A good fit for this need were m icron sized polystyrene microspheres (FluoSpheres, Duke Scientific type G700 ) with a green fluorescent emission (508 nm) similar to the V.fischeri luminescence emission (493 nm) The microspheres had a 0.7 m diameter and were dyed with Firefli fluorescen t green (468/508 nm) incorporated in the polymer matrix. We adhered the microspheres to the lower glass of the perfusion chamber in a scattered pattern the same way the cells were during measurements. We used a blue LED to excite the sphere fluorescence an d tuned down its intensity until the fluorescence per microsphere average was approximat ely 100 photons/minute/particle,


44 similar to the detected average from V.fischeri bioluminesc ence The microspheres were imaged with identical parameters as the cells: e xposure time, magnification, camera gain, etc. We analyzed the single microsphere signal data exactly the same way as the cell bioluminesc ence data. Figure 2 8: Variability in signal levels for V.fischeri cell s and for control fluorescent microspheres. A ) Histograms comparing the luminescent emission from individual V.fischeri cells under two different AI concentrations. B) F luorescent emission under weak excitation of a control sample of individual m icron sized latex spheres. Each histogram shows the num ber of individual emission measurements falling into the indicated brightness bin, over a ~30 minute period comprising thr ee 10 minute camera exposures. A and B insets ha ve the same horizontal scales: All images for both cells and fluorospheres were collec ted in ten minute exposures using identical camera and microscop e settings and image analysis. For the fluorospheres, we used a highly attenuated blue LED as excitation source and inserted a Schott longpass filter G G485 into the detection path. The coeffic ient of variation for the fluorospheres is 0.12, while the coefficient of variation for the V.fischeri cells i s 1.3 (200 nM AI ) and 1.0 (1000 nM AI ). Reprinted from (64)


45 The fluorescence emission from differen t spheres is more tightly packed than the bioluminescence emission from different cells ( Figure 2 8 ). The coefficient of variation (CV = standard deviation / mean) of different fluorescent microspheres is 0.12, while the CV of different V.fischeri induced cells is CV 1. The bioluminescence emission across cells after the population reached a steady state among cells is very wide (above the 10 photons/min/particle). Some very dark cells have detected emission ~ 5 photons/min/cell, while other very bright on es (in 200 nM and 1000 nM 3OC6HSL) have an emission ~ 200 photons/min/cell. Figure 2 3 shows that the microsphere output is very stable over several hours, showing that the system wa s stable during the experiment. Taking a single sphere, its signal fluctuat ions over time were 10 12 photons/minute/particle ( ~ 10% of the signal) which would be the uncertainty in our measurement This uncertainty includes the shot noise from the photon emission the iCCD dark noise and the process of extracting the data from the iCCD images (see Chapter 3) While developing our code, we observed tha t fluctuations were introduced and properly t uning the fitting parameters contributed to reducing the uncertainty introduced by the software analysis We occasionally observed fluctuat ions over time in single cell bioluminesc ence in the order of ~ 2 0 40 photons/min/cell. This is above the 10 photo ns/min/particle measurement uncertainty, and suggests that some cells do have biochemical fluctuations over time in bioluminesc ence. The biolum inesc ence of single cells was collected with 10 minute exposures. The average cell had a detected emission of ~ 100 photons/min/cell. This amounts to 1000 detected photons per cell for a single bioluminesc ence image acquisition. The signal to noise ratio SN R of the shot noise is given by the number of photons N


46 [2 5] The SNR of the shot noise is then ~30 (per image) so that the noise level is ~ 3% and below our estimated measurement uncertainty of 10%. Overall, the main contributio ns to the uncertainty ( ~ 10%) in the single cell luminescence acquisition were the shot noise ( ~ 3%), the iCCD dark noise ( ~ 3%) and the introduced in the image software processing ( ~ 4%). Autocorrelating the Light Output of Single Cells and Control Microspher es The emission of single cells can be autocorrelated in time, comparing the bioluminesc ence signal I(t) of an individual cell with its signal at a later time I(t+ ) However, the single cell signals have an overall trend to increase. Furthermore, the cells have a very wide onset times t 1/2 distribution equally spanning the whole experiment ( Figure 2 7 ). It is then difficult performing an autocorrelation in the sing le cell signal looking for a periodicity, when also considering that the experiments have only a total of ~ 15 points (images) in time. However, we calculated an autocorrelation to measure by how much the cells diverged from the average trend as time progre ssed. In Figure 2 9 we show the result of these autocorrelations for different cell populations induced with different concentrations of 3OC6HSL. The first four panels (A D) in Figure 2 9 show in detail the autocorrelation for the same sample perfused with 1000 nM. For the correlation we took as starting point in time an instant over 10 0 minutes when the average cell is fully induced. The coefficient represents the time increment. We can see how as the increment increases, the single cell autocorrelati on moves away from the best fit line as expected. Nevertheless,


47 for small to medium increments = 10 20 minutes, the autocorrelation is close to the best fit (the distance d showed in panel A is small). This shows how the signal of an individual cell at a given time is a good predictor of its brigh tness at a later time. This result i s expected since the cell brightness remains mostly stable over short time s once the cell reached equilibrium with the external AI concentrations. Figure 2 9: Temporal au tocorrelation of individual cell luminescence. The emission level I ( t ) of a cell at time t is compared to its emission at a later time I ( t + ). Data represent individual cell emission levels measured at least 10 0 minutes aft er introduction of 1000 nM AI : (A) (D) For small values of the data are close to p redictor o f its intensity at time t + However as approaches 40 60 minutes, the scatter around the average line increases, indicating that the brightness of the cell at later times (relative to the average or best fit trend) is poorly predicted by its earlier brightness or by the averag e behavior of the other cells. The vertical distance d of each point from the trend line becomes larger at large Panel (E) shows d (the standard de viation of d) as a function of At high AI concentrations the stand ard deviation continues to grow for many minutes, indicating tha t the brightness of the cells continues to diverge both from its initial value and from the average growth trend. The d of the control (fluorescence spheres) is essentially flat as e xpected, except for a dip near = 10 minutes (due to Gaussian filtering of the trajectories) Reprinted from (64)


48 In panel E of Figure 2 9 we plot the standard deviation d of the distance d not only for [3OC6HSL] = 1 000 nM as in panels A D but also for other samples at different exogenous 3OC6HSL AI We can see in panel E how at higher time increments d increases for high AI concentration; showing that as time progresses the cells are more susceptible to change th eir brightness and move away from the average behavior. For lower AI concentrations AI = 50 100 nM the cells do not change their brightness so much and are more predictable. The control microspheres show a stable d as expected. Over time, the cells tend t o become brighter, but the individual cell brightness also moves away from the average trend. This divergence (characterized by d ) acquires values at least double the uncertainty (~10 photons/min/cell) at 40 60 minutes for the well induced samples wit h 200 nM and 1000 nM [3OC6HSL]. The long time scale of this divergence suggests extrinsic noise by overall changes in the metabolic state of the cell (variation in regulatory molecules, nutrients, RNA polymerase etc) ; and not intrinsic noise given by the inherent stochasticity in discrete biochemical reactions If we extrapolate the curves d vs to the limit = 0, we can see that they roughly result in d 5 15 photons/min/cell which is the same order of noise estimation we obtained from the fluctuation of single fluorescent microspheres ( 10 12 photons/minute/cell). However, the experim ental noise is smaller than the observed light output of induced single cells ( ~ 100 photons/cell/min) and somewhat smaller than the occasionally measured single cell fluctuations (20 40 photons/cell/min) at short time scales ( ~ 30 minutes).


49 Additional Com plementary Experiments We performed a series of exp eriments to explore some possibilities for the source of the high level noise we observed. It was possible the bioluminesc ence difference among cells was being originated by differences in the genome among cell s. For example, the apparition of some dark mutants (non luminescent V.fischeri ) might have produced the observed heterogeneity. Therefore w e tested this hypothesis with a single colony experiment. In another experiment, we added a key substrate to kn ow whether some cells had some substrate deficiency for synthesizing luciferase. Finally, we used the fluorescent V.fischeri JB10 reporter strain to further investigate the noise in the bioluminesc ence. Testing the Heterogeneity in Single Colony Cultures B acteria are capable of accumulating mutations in their genetic information during replications. Our V.fischeri bacterium has a replication rate of ~ 1 hour. The individual bacteria we used for acquiring single cell data was obtained from up to 2 or 3 recolo nizations away from our original strain. We generally did not start our cultures from a single colony culture (monoclonal), which is derived from a single cell, because we wished to avoid the possibility of basing an entire experimental analysis on the des cende nts of a single atypical mutant. However, we wanted to verify that the heterogeneity in V.fischeri bioluminesc ence output tha t we observed was not due to mutations in the DNA. Using a single colony V.fischeri MJ11 strain culture we repeated a biolumin esc ence experiment using 1000 nM of 3OC6HSL to induce the cells. Under these conditions, the cells used are all derived from a single cell and only a few generations away from our original strain.


50 Details on how we obtained the monoclonal strain are found in Chapter 3 (Methodology). Using the monoclonal culture we obtained a similar very wide luminescence brightness distribution (CV 1) when it was fully induced at 1000 nM 3OC6HSL concentration. T he observed cells had similar size and similar locations in the field of view (and not in contact with each other) but they had heterogeneous brightness even when derived from a single parent cell This indicates that the heterogeneity probably cannot be attributed t Substrate Defici ency We wanted to know whether the wide variability in the single cell light production might be caused by a lack of substrate for the luciferase reaction. The light production in V.fischeri is due to the luciferase reaction, which involves the oxidation o f tetradecanal (a C14 aldehyde) to produce the fatty acid (tetradecanoic acid). The fatty acid is then recycled back to aldehyde (72, 73) If the cells have a deficiency in the production of aldehyd e substrate, adding exogenous tetradecanal or tetradecanoic acid would restore the C14 to the reaction and greatly increase the bioluminesc ence output. We repeated the experiment with 1000 nM concentration of 3OC6HSL in the same conditions previously descr ibed, except that this time we also added 1 M of tetradecanoic acid to the perfused medium. After over four hours of observation (more than enough time to stabilize the cells) we found no difference in the bioluminesc ence output: the average single cell i ntensity, the onset time t 1/2 distribution and the cell variability were all very similar to the original experiment results without added exogenous tetradecanoic acid.


51 We concluded that the high variability in the single cell bioluminesc ence does not orig inate from C14 substrate deficiency. Figure 2 10: V.fischeri JB10 lux regulation. The V.fischeri JB10 reporter strain contains a gfp chromosomal insertion in the lux box, between luxI and luxC (74) Green fluore scent protein (GFP) is synthesized along with luxICDABEG genes when the lux operon is promoted. Measuring GFP fluorescence allows the quantification of lux expression. The C8HSL and AI 2 autoinducer channel is also present in JB10 but not represented in th is figure. V.fischeri JB10 strain was provided by Dr. Eric Stabb. Comparing t he Heterogeneity of the Native Biol uminescence as a Reporter with that of JB10 V.fischeri Using Fluorescence as a Reporter We wanted to test if another gene regulated under the sa me promoter as the luciferase would also present a wide heterogeneous distribution in its expression. We used the Vibrio fischeri JB10 strain for the additional experiment discussed in this section and the experiments detailed in Chapters 4 and 5. Vibrio f ischeri JB10 strain


52 was provided to us by Dr Eric Stabb. This strain is derived from the V.fischeri wild type strain ES114 which was extracted from the Hawaiian bobtail squid ( Euprymna scolopes ). V.fischeri strain JB10 contains a chromosomal gfp insertion between luxI and luxC ( luxI gfp luxCDABEG ) in the lux operon (74) Having the gfp insertion in this location means that when the lux operon is activated, it not only synthesizes luciferase but also GFP (green fluo rescent protein). Figure 2 10 shows a diagram of the gfp insertion. The insertion in the chromosome makes the mutation stable and kept at a low copy number so that it is not overexpressed (75) In practical terms, the cells will behave the same as the wild type strain but will also produce GFP molecules when activated by autoinducer (76) GFP fluorescence can be measured to quantify the level of expression of the set of genes under the control of the sam e promoter. The comparison between single cell results of JB10 V.fischeri fluorescence and bioluminesc ence of the wild type V.fischeri can be observed in Figure 2 1 1 We compare the distribution of the two different populations, both induced with 1000 nM 3OC6HSL. The two populations show a broad heterogeneous response when fully induced. The fluorescence of JB10 has a coefficient of variation CV F 0.8, and the bioluminesc ence of the wild type shows a CV L 1. In Chapter 4, we discuss experiments that combine bioluminescence and fluorescence, and show that the noise in expression of the lux operon manifests a relationship. We concluded that the wide distribution (among cells) in the single cell bioluminesc ence i s not exclusive to the luciferase as a reporter, but related to the expression of the lux operon.


53 Figure 2 11: Heterogeneity of native luminescence versus GFP fluorescence as a QS reporter. A) Histogram of luminescence emission levels from 47 individual wild type MJ11 V.fischeri cells, following induction by 1000 nM of 3OC6HSL. The luminescence levels are normalized to the median value. B) Histogram of fluorescence levels for 127 individual V.fischeri cells of mutant JB10, also induced by 1000 nM of 3OC6H SL. The fluorescence values are normalized to the median. The JB10 mutant contains a chromosomal gfp insertion between luxI and luxC in the LuxI/LuxR system. Both luminescence and fluorescence reporters for the QS system show heterogeneous response at full induction. The fluorescence shows a CV 0.8 and the luminescence a CV 1. Reprinted from (64) Rich Medium Effect Single cell experiments in a fixed controlled environment provide more information than bulk experiments. However, single cell experiments are more complicated to perform experimentally, and sometimes make evident some effects that are hidden in bulk experiments. The rich medium effect we encountered and characterized for our V.fischeri MJ11 strai n is an important example. Commercial photobacterium medium is commonly used in bulk cultures for culturing bioluminesc ent bacteria as it allows rapid growth We started using one of


54 them in our first V.fischeri experiments (Carolina photobacterium powder No. 786230, Carolina Biological, Burlington NC). This is a rich medium based on artificial seawater sa lts, yeast extract, tryptone and glycerol (see Chapter 3 for a detailed component list). The rich medium effect had been previously described but we were not aware of it in our first single cell experiments (2, 77) A similar inhibition of the homoserine lactone dependent gene expression has also been observed in Pseudomonas aeruginosa QS (78) and has been attributed to a small molecule present in rich medium. V.fischeri cells readily bioluminesce in bulk experiments However, we found that in the perfusion chamber the cells would not bioluminesc e at all with fresh rich medium perfusion even when supplied with high concentrations of 3OC6HSL autoinducer (2000 nM) Further experimentation, led us to conclude that the weak bioluminesc ence in the flow chamber was caused by our use of a rich growth medium During bu lk experiments, the cells were in a confined space without medium renewal. After a transient time, the cell density increased and the cells removed the inhibitor present in the fresh rich medium. On the other hand, in the case of single cell experiments in a perfusion chamber, the cells are in a fixed external environment. The medium is continually renewed and the cells do not have a chance to metabolize the inhibitor. Even cells already induced and bioluminesc ent, turned off in a few minutes when exposed t o fresh rich medium flow. perfusion chamber they did bioluminesc e, unlike when they were under fresh rich medium (77) Conditioned mediu m is a growth medium that has been pretreated through the growth of cell culture. The main problem in using conditioned medium for


55 our experiments was that the cells produced an unknown amount of autoinducers during the conditioning. This prevented us from accurately setting the external AI concentration and from using concentrations lower than the amount of AI produced during conditioning. Figure 2 12 : Inhibition of V.fischeri bio luminescence by rich (complete) medi um. Light emission from individual cell s in the perfusion chamber was tracked over time as the flowing medium was switched from an initial (100% defined medium) to a final (70% defined medium, 30 % rich medium) composition. 3OC6HSL concentration remained 1 000 nM at all times. Image times repr ese nt the starting time of a 16 m inute bioluminescence exposure. The histograms, showing the fraction of observed cells emitting at the indicated level, collapse rapidly as complete medium is introduced. Reprinted from (64)


56 We started using defined photobacterium medium since it would not produce this inhibiting effect (77) The detailed component list of our defined medium can be found in Chapter 3 of this dissertation (6) Figure 2 1 2 shows that rich medium quenches the bioluminesc ence of the cells. We initially introduced the cells in the perfusion chamber under 100% fresh defined medium with [3OC6HSL] = 1000 nM. The cells were soon read ily bioluminesc ent and presented a wide bioluminesc ence distribution, where some cells were very bright and some other s relatively dim. Around 80 minutes later, we changed the perfusion to a mix of 70% fresh defined medium and 30% fresh rich medium. This n e w perfusion also contained 1000 nM 3OC6HSL. The new perfusion made the brightness histogram collapse very quickly (~20 minutes) to lower values, even when both media contained the same 1000 nM 3OC6HSL. We found this effect to generally occur in a faster t imescale than the time the cells take to adapt to a new 3OC6HSL concentration (~1 2 hours). The fact that the bioluminesc ence collapse is produced by only adding 20% of rich medium strongly suggests the presence of an inhibitory molecule in this rich mediu m. If instead the rich medium lacked a required molecule, this would be provided by the 70% of defined medium of the mix. Discussion V.fischeri regulates its bioluminescence through a quorum sensing (QS) system that controls symbiotic colonization factors in addition to bioluminesc ence. When the cell concentration surpasses a certain threshold (indicating host colonization), the bioluminesc ence is acti vated. The bacterial population detects this threshold through the accumulation of small diffusible molecul es, or autoinducers ( AI s ) secreted by the cells. 3OC6HSL and C8HSL signals modulate the expression of the lux genes. Both AI s


57 can bind to the transcriptional regulator LuxR forming a complex that binds to the lux promoter and induces the transcription of the lux genes (12) The LuxR 3OC6HSL has the highest capability to induce lux operon (45) The mai n question we have asked in this chapter is how tightly the single cell bioluminesc ence is reg u lated by a well defined 3OC6HSL concentration We found that although the bulk population response is predictable given an external concentration of 3OC6HSL the response of individual V.fischeri cells is very heterogeneous in activation level, onse t time and in short term brightness fluctuations. Biol uminescence Heterogeneities in Single Cells The bioluminesc ence emission of V.fischeri cells is below detectable levels when the external AI concentrations are fixed to zero. When an exo genous concentra tion of 3OC6HSL is added the cell population becomes significantly brighter on average after an activation period of around 150 250 minutes. The average activation of the population was consistent with our bulk experiments. However, when comparing single cells we observed that the brightness activation level differences among them spanned an order of magnitude ( Figure 2 5 ). The presence of 3OC6HSL not only increases the average cell bioluminesc ence, but also greatly increases the difference among cells. We confirmed that our single cell results were consistent with the ones obtained through bulk experiments in similar growth conditions. Using a least squares fit with a simple cooperative binding model, we compared our single cell data with our bulk data F o r the single cell experiments we obtained an equilibrium constant and a Hill coefficient When fitting the bulk cultures, we obtained and Thus the single cell and bulk culture results are


58 close agreement in the value of the Hill coefficient and are consistent in the order of magnitude of the equilibrium constant. T he single cell and bulk experiments are actually qual itatively different and the first ones have a very wide dispersion Despite the low cell density in the bulk experiments, the bacteria might still consume nutrients and produce excretions (including AI s) that accumulate in the medium over time. The single cells in the perfusion chamber are maintained in a constant known environment that is constantly renewed by the liquid medium flow. Furthermore, in the well tray the cells are growing exponentially and have not been adhered to a surface for a few hours as in the perfusion chamber We found quite rema rkable the high variation in emission between individual cells in a homogeneous and well defined environment. Even after several hours of exposure, the luminescence brightness variability between cells generated a very high coefficient of variation (CV = s tandard deviation / mean ~ 1). This variation is higher than that expected from stochastic simulations of the LuxI/LuxR system. Cox et al. predicted a very small variation in the activation of luxI respect to AI concentrations (79) showing LuxI variations for low AI concentrations (< 50 nM). However, for high inducing AI concentrations the simulations predicted a lower CV 0.1. The origin of this high variabilit y is not likely to be caused by heterogeneity of intracellular AI concentrations. The AI molecules diffuse readily both across the cell membrane (2) and the external AI concent ration is well maintained by the perfusion flow. Little is known about how cells process the information carried by the AI s or how they integrate several input AI signals into one or more output behaviors. It is also under investigation how genetic noise a ffects this processing and its corresponding results.


59 Single cell experiments provide a unique view not available with bulk experiments. To this effect, Anetzberger et al. recently performed a study with individual V.harveyi bacteria where a significant ce ll to cell QS regulated bioluminesc ence variability was also found (80) In this study, V.harveyi cells were not in a controlled environment. They grew in an enclosed medium (without perfusion) until the accumulate d natural AI triggered the QS. The lux system that V.harveyi possesses is different than that of V.fischeri but they also observed a big fraction of cells (25%) remaining dark when the cell population as a whole was induced. H owever, another recent study by Long et al. in V.harveyi with mutants using fluorescent reporter s for the lux system, found a smaller variability (CV ~ 0.2 0.4) in gfp expression (81) These GFP reporter strains lacked autoinducer synthases eliminat ing any possible feedback loop mediated by AI which might have explained the more homogeneous behavior observed in this case as Anetzberger et al. suggested. However, our experiment fixed the AI concentrations eliminating the chance of a posit ive feedback mechanism to exist, reducing the chance of direct AI feedback to explain the high variability. Much is still unknown so a feedback mechanism is a possible explanation for the observed heterogeneity. Williams et al. recently studied LuxI/LuxR in an E.coli mutant strain lux01 where the LuxR regulatory protein is activated by 3OC6HSL to induce the expression of gfp but the autoinducer synthase LuxI is absent (82) Using flow cytometry, they found a bimodal respon se in GFP production. They suggested that external concentrations of AI feed into an autoregulatory loop for LuxR expression. In this scenario, the cells would present some hysteresis and this might help explain the


60 high variability. The individual initial levels of LuxR in each cell might be amplified to different levels by a feedback mechanism, resulting in significantly different levels of GFP fluorescence. A hysteresis mechanism might also expla in the wide distribution of onset activation times we obser ved ( Figure 2 7 ). Another possible explanation for the light output variability we found might be heterogeneity in the energy resources among cells. However, we observed that even when diff erent cells activated to different levels, each of them maintained roughly constant bioluminesc ence output over long periods of time. These time periods were on a similar t imescale as the replication (~1 2 hours), and energy depletion cycles mig ht be shorter. Also, addition of the C14 substrate for the luciferase reaction did not cause significant difference when compared to our experiments without added substrate. Furthermore, we observed a high heterogeneity in the fluorescence of the V.fischeri JB10 mutant that expresses GFP when activated. All these findings point away from energy resource heterogeneity as an explanation for high cell to cell variability. We also observed an occasional ~ 20 40% fluctuation over time in the bioluminesc ence output when a single induced cell is considered An early study showed no significa nt oscillations in single cells in a high frequency range of 0.01 10 Hz (66) but did not explore the lower frequencies of ~ 10 3 Hz studied in this dissertation Finding whether the bioluminesc ence output of single cells really is fluctuating, intermittent or bursting in any way is a topic that requires more study. It might be possible that the short term fluctuations we observed over time in single cells were due to intrinsic noise (biochemical noise) (52, 55) However, the characteristic times for


61 intrinsic noise ( ~ 10 min utes ) really are at or below our time resolution for single cell experiments ( ~ 15 min utes). On the other hand, extrinsic noise time scales are l onger ( ~ 40 minutes) (52, 55) They are similar to our observed onset time variability (spread over a ~ 150 minutes interval, Figure 2 7 ) and our wide distribution in bioluminescence among cells (br ight cells tend to remain bright). This similarity between timescales is suggestive of extrinsic noise originated in differences in global variables (within a single cell) among different bacteria such as concentrations of key regulatory proteins, ribosome s, different growth stages, etc. (56) Many systems under positive feedback regulation present oscillations. Our single cell experiments did not show oscillations in the LuxI/LuxR system activation. The perfusion flow maintained the cells in a constant environment. Any extra natural autoinducers were washed away by the flow, negating the feedback mechanism and preventing oscillations. Conclusions and Additional Thoughts In V.fischeri QS regulates bioluminescence emissi on as well as colonization behaviors. For example, it would be very interesting to investigate if the heterogeneities we observed are also present in the in vivo symbiotic environment. When colonizing their host, V.fischeri cells can reach very high biolum inesc ence outputs of up to ~1000 photons/second/cell (34) The host environment can produce biological events to which the cell has to adapt, possibly making it a different environment than the laboratory culture medi um (63) According to our results, the individual V.fischeri cell response does not seem to contain accurate information about its local microenvironment (23, 26, 83) The


62 individual cell bioluminesc ence is a poor indicator of the local AI con centration levels, in the sense that the AI concentration cannot reliably predict the response of a given individual cell. Our group of cell s generated very different responses in a well defined, stable and homogeneous environment ; so the QS system is not a reliable sensor of the local concentrations or at least it generates a very unreliable response. In a natural environment, the medium wou ld likely be more temporal and spatially heterogeneous so the variability among cells might increase (63) V.fischeri bioluminesce in the course of their symbiot ic relationship with their host. The latter provides the bacteria with a medium where they can grow to higher densities than in the open seawater The natural question s to ask are whether the noisy response in the bioluminesc ence benefits the bacteria and affects the symbiosis. There are many known scenario s where a heterogeneous phenotypic response is beneficial to bacteria (60, 61) In the case at hand, V.fischeri cells are under a very strong selective pressure by the host to maintain their b ioluminesc ence output. The E.scolopes squid for example, does not tolerate colonization by a dark mutant V.fischeri (5, 84) The host can select a strain for its average bioluminesc ence output, bu t is unlikely to be able to detect temporal fluctuations or brightness variations at the single cell level. In other words, the host possibly can detect the average bioluminesc ence but not its variance. A possible explanation is that the brightness level i s not coordinated across the bacterial population simply because they are not under selective pressure to do so. The host cannot (or does not need to) require the bacteria to have a well defined respon se down to the single cell when the overall brightness is already satisfactory.


63 Light production is a very taxing activity for the bacteria. A fully induced cell may use more than ~ 10 4 ATP molecules per second, and up to 20% of the oxygen consumption when fully induced (34 ) We observed during the course of our bulk experiments how cultures grew slower when fully bioluminesc ent. It is possible that when bacteria are fully bioluminesc ent they have to reduce or shut down other important metabolic activities. If the populatio n bio luminesces in a heterogeneous fashion, even when all cells are genetically identical, the dark cells escape the burden of light production and can pursue other required activities like growth, colonization behaviors, etc. In a sense, it would be as if the cells divided their tasks for an overall increased efficiency. As light production is energetically costly, it is likely for the cells to have evolved to avoid bioluminesc ing if they do not have to. This requires the host to exert a strong selection p ressure to weed out dark mutants. We confirmed that in an overall bright population the dark cells were not mutant cheaters. Using a monoclonal strain (originated from a single cell) we observed the same wide variability in the population as in the polyclo nal cultures. Thus, the appearance of cheaters as an explanation seems unlikel y: t he cells have a strong selection pressure in the host proba bly reducing diversity through mutants, and typically kin selection favors cooperation in closely related populatio ns (15)


64 CHAPTER 3 LUX EXPRESSION IN RESPONSE TO 3OC6HSL: MATERIALS AND METHOD S Vibrio fischeri MJ11, JB10 and ATCC7744 Strains For studying the lux expression of V.fischeri in response to varying 3OC6HSL we u sed the strains MJ11 and ATCC 7744 The MJ11 strain (NCBI Taxonomy ID: 388396) was provided by Dr. Mark Mandel and Dr. Edward G. Ruby. This strain is a wild type strain extracted from the light organ of the pinecone fish ( Monocentris japonica ) (85) The JB10 strain was provided by Dr. Eric Stabb and is a fluorescent reporter strain derived from the ES114 wild type extracted from the Hawaiian bobtail squid ( Euprymna scolopes ). In the JB10 strain a chromosomal gfp rep orter is placed under the control of the LuxI/LuxR system by insert ion between luxI and luxC ( luxI gfp luxCDABEG ) so as to express GFP when the LuxI/LuxR system is activated by 3OC6HS L (74) The ATCC 7744 strain w as obtained from the American Type Culture Collection (ATCC) resource center. ATCC 7744 is a wild type strain that colonizes the Hawaiian bobtail squid ( Euprymna scolopes ) We used this strain for initial tests with Vibrio fischeri We stopped using this s train since its luminescence output is over an order of magnitude lower than that of MJ11 strain. ATCC 7744 also has a metabolic deficiency 11 such that it grows poorly in defined medium preventing us from performing certain experiments with the strain Vib rio fischeri C ultures Growing the C ultures Bacterial stock cultures were stored in liquid nitrogen vapor phase ( 195 C) The frozen stocks were prepared by growing V.fischeri to stationary phase in rich medium 11 The sugges tion of a possible metabolic deficiency was provided by Dr. EG Ruby (private communication).


65 cultures for ~ 16 hours G lycerol was added at 15% to prevent cell damage when freezing (86) The cells used were grown directly from a frozen stock or from an agar plate culture (grown from frozen stock). The lat t er method was usually preferred to avoid repe atedly thawing and freezing the stocks. The plated culture s from stock were periodically controlled for contaminations. The culture samples used in the experiments were grown overnight (16 to 1 8 hours) at room temperature ( 22C) on an orbital shaker to k eep them oxygenated. Culture M edia: Defined and Rich P hotobacterium M edi a We used a defined photobacterium broth recipe provided by Dr. Edward G. Ruby laboratory (6) This defined broth consists of an artificial seawate r mixture Tris (organic buffer), phosphate, ammonium (nitrogen sourc e) and glycerol (carbon source) at the concentrations listed below. 50 mM Tris 50 mM MgSO 4 7 mM CaCl 2 300 mM NaCl 10 mM KCl 0.333 mM K 2 PO 4 18.5 mM NH 4 Cl 32.6 mM Glycerol Trace minerals 12 We also used a commercial photobacterium growth medium (No. 786230, Carolina Biological, Burlington NC). We used this rich medium extensively in preliminary single cell experiments. The cells gro 12 Deionized water is around 50% of the liquid volume, the rest (around 50%) is completed with tap water which contains trace minerals necessary for cell developmen t.


66 preve nts its use for single cell experiments in the perfusion chamber. The Carolina Biological rich photobacterium ingredient list is detailed below. 5 g Difco TM Bacto TM tryptone 2.5 g Yeast extract 0.3 g Ammonium chloride 0.3 g Magnesium sulfate 0.01 g Fe rric chloride 1 g Calcium carbonate 3 g Monobasic potassium phosphate 23.5 g Sodium glycerol phosphate 30 g Sodium chloride 1 L Distilled water It is possible to condition rich medium before use to avoid the rich medium effect Conditioning medium in experimental biology means growing a culture in the medium until a low concentration of cells is reached and then removing the cells When we started becoming conditioned media in preliminary studies. Figure 3 1 shows the luminescence suppression effect from rich medium in single cell images. The main drawba ck is that during the conditioning the cells produce an unknown amount of autoinducer. This would not allow us to take accurate measurements of gene expressio n response at low concentrations of exogenous autoinducer. In some experiments, it was ne cessary to remove the natural autoinducers produced when growing the culture wit hout significantly damaging the cells The objective was to remove the liquid medium (d efined or rich) and replace it with fresh one while keeping the cells. We achieved this with three successive washes by centrifugation In each 1 0 minute centrifugation cycle, the used supernatant was disposed of and replaced with fresh defined broth, and the cells resuspended. This


67 procedure dilutes the AI by over 3000 times. Even if the culture d cells had produced an AI concentration of over 1000 nM, this method reduced it to undetectable levels of less than 1 nM. Figure 3 1: Rich fresh medium suppresse s luminescence. V.fischeri were immobilized in the perfusion chamber. The chamber was initially filled with conditioned rich medium (Carolina photobacterium powder) that also contained natural autoinducers. At ~ 120 min the chamber was flushed and filled wi th a mix of rich medium (75% conditioned / 25% fresh). The histograms show single cell luminescence distributions for 10 minute long exposure images. Making Single Colony Cultures to Obtain Monoclonal Cells Making single colony cultures requires some extra steps when compared to the standard working cultures. We made these single colony cultures as explain ed in the


68 following text. Starting from a frozen stock we made a culture in liquid fresh rich photobacterium medium. The culture was left overnight (aroun d 15 hours) to grow on a n orbital shaker at room temperature. The culture was then spread on agar photobacterium plates forming a thin film on the surface. We used several dilutions of culture in fresh medium to find a suitable cell concentration that prod uced many monoclonal colonies The plates were left to grow at room temperature overnight. After 24 48 hours, we could observe single isolated colonies in the plates spread with a suitable dilution. Each colony can be assumed to have originated from a sing le cell. We took a little culture material from a single bright colony and made another culture in liquid fresh rich medium. We left this culture to grow overnight and used i t to perform the described control luminescence experiment. Bulk Measurements with V arying AI C oncentrations The single cell work we present in this dissertation was compa red with bulk measurements of macroscopic cell culture s. Bulk measurements were also important when deciding which sing le cell experiments to perform as they provide i nformation about conditions required for the single cell experiments. Using Well Plates to Measure Many Samples in Parallel Bulk experiments were performed in very small volumes of culture of around 0.2 2 ml of sample. We generally wanted to measure the ce ll response under a wide array of AI concentrations, so it was necessary to test many samples in parallel. The samples were placed in well tray sample holders. The plates are plastic trays with small cylindrical wells laid out in an array. They have a lid that is not air tight but that prevents evaporation from the samples. The lid and the bottom of the wells are transparent. Well plates are commercially standardized in different sizes. We used 48 well (6x8) and 96


69 (8x12) well plates (Falcon Multiwell polys tyrene sterile 48 well plate, Becton Dickinson Labware; Costar polystyrene black wall 96 well plate, Corning Inc.). We performed some preliminary experiments with 24 well plates (Falcon Multiwell polystyrene sterile 24 well plate, Becton Dickinson Labware) Well Plate Readers and Bulk Measurements The objective of the bulk experiments was to measure the bioluminescence response t o defined concentrations of AI. We collected our first bulk measurements of bioluminescence using a home made well plate reader ba sed on a photomultiplier (PMT Photo M ultiplier Tube ). The well plate with the samples was positioned on a stage so that only the light with the well being measured reached the PMT. This configuration provided us with valuable preliminary information. Howe ver, its lack of automation did not allow us to acquire the high density data we later required from overnight measurements. We moved from the PMT experimental configuration to using a commercial well plate reader (Biotek Synergy 2, Biotek Instruments). Th is automated well plate reader allows the user to read the optical density (OD), fluorescence and luminescence of a sample in a programmable sequence that was generally repeated every 5 10 minutes for long periods of many hours. Completing a sequence takes the read er around 5 minutes depending on the sequ ence steps and the number of wells in the tray. The reader was set to softly shake the sample in between measurements when the reader was idle. Shaking the sample serves to oxygenate the liquid culture medi um and to keep the liquid samples homogenized during the experiment. The luminescence is acquired by an optical reader head that comes very close to the measured well (from the top or the bottom). The luminescence light is very weak


70 and is generated in all directions from the well contents. Unlike the well plates made in their entirety with transparent plastic, plates with opaque black walls give some better results for luminescence measurements. The opaque black walls of each individual well prevent signal interference between we lls, and were used for the V.fischeri JB10 bulk experiment described in Chapters 4 and 5. The optical density was measured at 600 nm, a standard wavelength in cell densitometry. Samples for Bulk Measurements When preparing V.fischer i for bulk measurements, a well plate was prepared introducing in each well around 200 l of fresh growth medium with the desired exogenous 3OC6HSL AI The wells were then filled with a liquid layer of a few millimeters. The plates we used had 48 or 96 wells so we could test several AI concentrations with a few controls at the same time. Whe n using 96 well plates we reduced the volumes accordingly to fit samples in the smaller wells. The bacterial cells were previously grown in fresh medium for a few hours (5 8 hours) on an orbital shaker at room temperature. Under these conditions the sample reached a good concentration (typically 0.1 0.3 of OD at 600 nm and 1 cm path length ) and the cells were in exponential phase. The cells were washed to remove any AI the cells had already produced. We diluted the wash 50 to 100 times (depending on the ini tial sample density) to a very low concentration. We introduced in each well of the plate around 5 l of the diluted washed culture. The cells were diluted to an optical density below the detection level. We added to the wells the exogenous 3OC6HSL AI conc entrations to be tested. The sample was then set in the well plate reader and the measurement sequence started. The period


71 following the exponential phase do not give very useful information as the secreted AIs raise the AI concentrations above the exogen ous levels and the nutrients are deplete d Measuring Single Cell V.fischeri Luminescence in a Perfusion Chamber over Time Vibrio fischeri cells generate a very we ak light output of up to ~ 10 3 photons/s when colonizing the host (34) The emission intensity can significantly vary depending on their metabolic state. A sufficiently sensitive and stable experimental configuration is required to measure this signal and its fluctuations over a period of over 5 hours. A single layer of s pread immobilized V.fischeri cells was adhered to the inside surface of a perfusion chamber. We filled the chamber with fresh defined medium with oxygen and nutrients and with the desired 3OC6HSL AI concentration A slow perfusion flow was start ed. This flow was slow enough not to disturb the cells but sufficiently high to remove any secretions and AI produced by the cells. Under these conditions the cells were in a controlled and constant environment during the experiment. Dark field and lumines cence images w ere alternatively taken every 15 25 minutes. The dark field images provided the physical location of the cells in the observed area of the glass surface. The measurements could not be extended much more than 5 hours since the doubling time of cells was around 2 3 hours. After 5 hours most single cells immobilized on the glass had become clusters of cells preventing us from acquiring single cell bioluminescence of neighboring cells. The cell replication set a hard time limit for the experiment. A t approximately 2.5 hours most cells stabilized to the exogenous AI concentration as can be seen in ( Figure 2 7 ). Appropriately timing and coordinating image acquisition, cell adherence and growth, and AI response time was necessary for the experiment su ccess.


72 Figure 3 2: Perfusion chamber optical configuration. The optical apparatus shown in the figure was placed inside a dark light proof box to prevent external light interfering with luminescence measurements (except the body of the iCCD to properly c ool the camera). The green LED was turned on during for dark field images and turned off during luminescence ones. Both types of images were acquired with the iCCD. The color CCD camera was only used for live video feed to monitor the sample. V.fischeri we re attached to the lower glass in scattered layer of single cells and their light collected with the microscope objective. The perfusion medium was pumped into the chamber with a programmable syringe pump at very low rate (0.2 ml/h). Perfusion Chamber and Optical System Configuration The single cell measurements were performed on individual cells in a perfusion chamber. The perfusion c hamber is a cylinder of approximately 25 mm diameter and 5 mm high. The bases of this c ylinder a re two parallel circul ar gla ss coverslips and its


73 volume was approximately 1.5 cm The chamber has an inlet and an outlet on opposite sides to provide fresh liquid medium perfusion with a programmable syringe pump. The Figure 3 2 shows a detail of the perfusion chamber. The chamber was mounted on an aluminum stage attached to a vertical frame. We used a 100x microscope objective (P lan oil immersion, in finite conjugate NA 1.25) to collect the light. This objective was placed under the chamber focusing on the upper surface of the lowe r glass window where the cells were located. The collected light was then diverted with a mirror and focused onto an intensified CCD camera (512x512 pixel, I MAX 512 T operating at 35 C, Princeton Instruments, Princeton NJ) with an achromatic doublet lens (f = 15 cm) The final image on the iCCD sensor had a scale of 0.278 m per pixel which we measured with a scaled micrometric slide The system magnification was 87x resulting from the combination of the 100x objective and the 15 cm tube length given by the achromatic lens. In order to help the iCCD camera have its sensor at 35 C, we refrigerated the camera with running water at 15 C with a programmable water bath. The chilled running water helped in cooling the hot side of the thermoelectric plate inbuilt in the camera; and allowed the cold side to maintain the iCCD sens or at 35 C. Figure 3 2 shows a schematic of the optical experimental configuration We used the optical system with the chamber, the objective and the iCCD camera when taking dark field or luminescence images. We collected pairs of images, each consisting of a dark field immediately followed by a luminescence image. When acquiring dark field images, we turned on a light excitation system that directed light downwards through the chamber glass windows. The excitation light was collimated


74 light from an ultra bright green LED. We preferred using green light when focusing the cells to have a wavelength similar to the luminescence output light. We collected the scattered excitation light from the cells to form an image with the cell locations. No intensification of the iCCD sensor was necessary for dark field images. The LED was then turned off and unplugged during luminescence image acquisition. A dark box prevented external light from reaching the system and the camera intensifier was set to maximum gain (which generated 60 counts per detected photon). The only light collected in luminescence image acquisition was the bioluminescence emission from the cells. The collect ed light could be diverted from the iCCD camera with a flip mirror to be focused on a color CCD camera connected to a television screen. T his color camera provided us with a real time video in dark field imaging to find appropriate spots in the sample to acquire data, and to check the focusing and sample status T hese checks were very frequently per formed in betwe en images during the experiment (typically 15 20 times during the ~5 hours of data acquisition). Perfusing and Adhering Single V.fischeri Cells to a Glass Surface Having the bacterial single cells adhere to a surface and remain stationary fo r 5 to 6 hours is a challenge as the cells are normally flagellated and can swim at startling speeds. Flagellated wild type V.fischeri were observed to have swimming speeds of ~70 (50) This means that in 1 sec ond a V.fischeri cell can cover a distance around 30 times its body length. We tried s everal methods for promoting adhesion of the cells to the glass window. Some bacterial species stick to glass surfaces without the need of any special treatment. This was not the case for V.fischeri The lower coverslip was covered with a


75 dried thin layer of poly L lysine ( ~ 40 l drop of 0.5 mg/ml solution) Using poly L lysine for cell adhesion is a standard procedure in b iological sciences microscopy that works with many cellular types (87) This surface treatment works very well with big and malleable eukaryoti c cells, and it also works (to much lesser degree) with prokaryotic cells like V.fischeri Poly L lysine molecules are positive ly charged and interact with the negative charges of the glass surface and the cell exterior wall promoting adhesion. The chamber had to be prepared careful ly to have an appropriate scattered layer of cells for the experiment. A small drop of approximately 15 l of culture in exponential phase was put on the lower glass window covered with poly L lysine. We put the chamber on the optical stage and observed the sample with the color camera CCD in real time. The time it took for the cells to firmly stick to t he glass was variable (between 15 and 60 minutes). On one hand, it was necessary to wait long enough to have a sufficient number of scattered cells in the field of view to perform statistics. On the other hand if we waited too long too many cells would sti ck, and since the cells eventually repli cate the sample would become crowded too soon. Once the culture drop provided a suitable cell distribution, the chamber top glass was locked in place and the chamber was very slowly filled with fresh defined medium w ith the desired concentrations of AI I f we filled the chamber very slowly and carefully most cells would remain sticking to t he lower glass. We also had to take care to remove any air bubbles as these would scatter light and prevent image collection (esp ecially dark field images) Just by filling the chamber the culture drop is diluted around 100 times. Immediately afterwards, an amount of fresh medium equal to around two times the volume of the chamber was flowed through in 15 20 minutes. This wash essen tially


76 set the cells at the fresh medium exogenous AI concentration and removed any extra natural AI s the cells may have produced. The flow rate of this wash could not be increased too much or the cells would detach from the glass. The flow was finally red uced to 0.2 ml/h which was slow enough to not distu rb the cells and set them to the desired exogenous AI concentration. A numerical integration of the diffusion equation predicts an AI concentration of less t han 50 pM at the glass surface (well below 1 10 nM detectable levels) assuming a rate of AI synthesis equal to 10 21 g/s/cell and diffusion at 100 m 2 /s (88) Also, perfusing cells with fresh medium without AI did not generate any detectable luminescence. Going a step beyond, previously induced cells shut down their luminescence when we perfused them with fresh medium without 3OC6HSL AI Physical Stability of the Optical System We set the optical configuration on an optical table to reduce vibrations. The whole system, except the body of the iCCD camera, was enclosed in a light proof black b ox and the room was kept dark as a precaution. We performed several tests to ensure that any ambient light that could reach the iCCD was negligible We mounted t he microscope objective, the dark field excitation system and a reflecting mirror on a vertical custom made platform. This platform was cut and drilled at the machine shop of the Physics Department. We added extra reinforcements on the vertical platform to reduce microm etric vibrations that might affect the measurements. The acquisition of t he dark field image took only 15 seconds approximately, while the luminescence i mage took 10 minutes. Having a d ark field image taken right before each luminescence one allowed us to correct small drifts in the focus and the cell locations. We observed some very small occasional m ovements of the stage of 1 or 2 m. These shifts, which we presume were due to thermal expansions, only occurred


77 once or twice during the 5 to 7 hours of data acquisition They were easily corrected during the data analysis. When looking at Figure 1 7 carefully, it can be observed that the luminescence image contour of the cells very slightly exceeds the boundaries of the dark field image contours. We obtained the dark field images by illuminating the cells from above with collimate d light This light produced some mild diffrac tion patterns around the cells that prevented a perfect focusing. Nevertheless, this slight defocusing did not affect the extraction of l uminescence data from the images. The focusing of the objective on the perfusion chamber was checked before each image pair when taking the dark field image. If the focusing drifted, it did so in the timescale of an hour which was significantly longer than the 10 12 minutes necessary to acquire an image pair. These drifts were likely due to thermal expansion and pressure changes in the chamber when the liquid medium perfusion rate was changed. The sample of cells adhered to the chamber glass also had change s during the several hours of experiments. Some cells occasionally detached and were washed away. S ometimes a new cell adhered righ t next to a cell being observed; or sometimes an individual cell replicated very fast leaving a cluster at the end where ther e was a single cell at the beginning of the image sequence. We performed the data analysis with cells that did not change or were not externally affected. The dark field images provided us with a constant check of the cell locations. Orientation of the Cel ls on the Glass The sample cells observed during the measurements were adhered to the lower glass of the perfusion chamber. The V.fischeri cells have a rod like shape of around 2 3 m long and less than a m wide The bacteria usually stuck to the glass ly ing flat


78 against the surface to increase the contact area. The orientations of their long axes were found in many directions on the glass plane. The perfusion flow did not seem to affect the axes distribution on the glass A few cells stuck to the glass st anding upright with their long axis perpendicular to the glass surface. In this case the cell tip farther from th e glass was slightly defocused when compared to the tip closer to the glass The cell orientation on the glass, even for standing cells, was no t a problem when acquiring the cell luminescence. While collecting the luminescence of a single cell, we analyzed a small image area that completely included the cell as well as some background area All the collected light above the local background was r ecorded as signal and not only the signal per pixel. Even in the case of standing cells, no significant light was lost even for the slightly defocu sed tip farther from the glass. Most of the cells selected for analysis ( less than 95%) were l ying flat on th e glass. We did not perform any specific analysis to test if there was a significant difference in luminescence intensity between cells laying flat and the ones adhered standing upright. Even when taking all the described precautions, the cells selected fo r analysis were the ones with a clear and focused shape in the dark field image. The cells that detached, moved or w ere somehow affected during the experiment were discarded for the data analysis. Single Cell Data Analysis Programming the Data Analysis Cod e We analyzed the single cell luminescence data acquired with the perfusion chamber and the iCCD camera using a custom Matlab code that we developed. The luminescence images we obtained show clear shapes w h ere the brightest cells are located. Figure 1 7 sh ows a pair of dark field and luminescence image views for a fully


79 induced sample. Nevertheless, we had to be careful when analyzing the data to introduce as little noise as possible with the data processing itself. The dark field images were used to genera te a map with the cell locations. We drew an area around each cell of 30 to 40 pixels side T he iCCD camera was used with minimal gain for the dark field images The cell location selections in each dark field image carried over in the sequence of images o ver the 5 7 hours We performed several trials to automate the cell selection using custom code we developed and open access code already developed by other researchers. In all cases using automated cell selection, the errors introduced by false positives and missed detections were much higher than the fluctuations we expected to measure. Even when difficult and laborious, manual cell selection in the dark field images was the technique that produced the best results. After the cell selection we took care of the rare slight micrometric shifts of thermal origin of the stage holding the chamber in the X Y plane This was easily achieved by carefully shifting the selections in each image if necessary. The very occasional shifts in the Z direction were constant ly checked for and corrected during acquisition. The luminesc ence images were taken with the intensifier of the iCCD at its maximum setting (yielding around 6 0 counts per detected photon). The first step when dealing with the luminescence images was always subtracting a dark image (acquired with the iCCD completely blocked from entering light) Once the dark image was subtracted, each luminescence image was analyzed separately taking one by one the rectangular pixel ranges surrounding each cell Save for th e very dark cells, the luminescence shape was typically identified inside the cell


80 ranges selected in the d ark field image ( Figure 1 7 ) For each cell, we binned by 2x2 the region to improve the signal to noise ratio. We then generated an intensity histogr am for the pixels in the cell region (after binning). The shape of this distribution was typically a big curve at low intensity owing to the background pixels and a small peak overlapping the high end tail of the big curve due to the bright cell pixels W e fit only the lower background curve with a Gaussian curve and subtracted it to the whole signal. This fit was first subtracted to the whole signal histogram. The high signal tail containing the bright pixels was added up after the subtraction. This value represented the signal count for that particular cell range and image. We further divided this value by the exposure time (10 minut es) and the iCCD camera gain (6 0 counts per detected photon) to finally obtain the detected photon count per minute per cell This method provided a satis factory way to count the single cell detected luminescence emission. We confirmed that the data processing was not dependent on small changes of the sel ected parameters (intensity binning, background cutoff, etc.) W e also sel ected many empty boxes in the images and processed them the same way as the ones containing cells. These empty boxes gave an output signal of ~ 20 photons/minute defining a baseline background noise level in the data processing. By fitting the local inten sity background leve l in each box surrounding a bacterium image we were able to take care of small background variations in the field of view By taking not a global but a local background level, we could consider and include these s mall variations in the analysis. It is interesting to note that this data processing technique does not depend on the cell shape. Before developing this analysis method we tried several other s that


81 depended on the cell shape and location within the selected box. These methods did not provide results as robust a s the one previously described. Characterization of the Intensified CCD (iCCD) Camera As we performed all the luminescence measurements with the iCCD camera we needed to appropriately characterize its signal response Noi se Sources in the iCCD Camera In general iCCD cameras have several sources of noise: pixel non uniformity, dark current, read out noise, electronic interference, intensifier and shot noise (89, 90) Variations of 1% to 2% among pixels from the average signal are not uncom mon. Our data analysis required us to take several pixels ( ~ 1200) to calculate the signal of a single cell, which tends to ave rage out the se variations. The pixels of th e CCD accumulate electrons that will later be counted by the ADC Electrons can be accumulated without any incoming signal by thermal excitation (dark current ) Lowering the sensor temperature reduces this current and consequently reduces the dark noise H owever it does not eliminate it completely. We cooled the thermoelectric stage with running chilled water and set the sensor temperature at 35 C, the minimum stable temperature possible in our configuration. The running chilled water was set at 15 C to h elp cool the hot side of the iCCD thermoelectric plate If the temperature was set lower than 15 C, we ran the risk of condensation. The longer the exposure time and the higher the camera gain, the more dark current is accumulated in the sensor. We subtrac ted a dark image to all our data images. The read out noise is usually a small value noise introduced when the camera reads the signal values of each pixel. We characterized this noise in our camera to be in the order 1.2 counts per pixel. This noise can b e neglected if we consider that for


82 acquiring the signal of one V.fischeri cell we typically took 30 50 signal pixels (of the total ~ 1200 pixels in the region) each yielding hundreds of counts and the observed biological cell fluctuations where in the orde r of 30% of the signal. When the iCCD was operated at high gain, it was common observing a higher background signal in one corner of the field of view due to electronic interference. This electronic background signal was subtracted from the images. The sho t noise is not caused by the iCCD camera but by nature itself. It is related to the detection of light quanta at very low intensities Photons are detected according to Poisson statistics, so that the signal to noise ratio ( SNR ) is given by the average num ber of photons N Increasing the measuring time increases the SNR. [3 1] Characterizing each noise source can be quite difficult and we only needed the overall noise introduced by the camera and data processing Subtracting an up dated dark image compen sa tes for the bias, the dark current and the electronic background We estimated the overall noise in the measurement for individual cells with the fluorescent spheres control. Calculating the iCCD Camera Gain It was necessary to det ermine the gain G of the iCCD, which relates th e number of detected photons N to the number of recorded S ADU analog di gital units in the final image for our luminescence measurement microchannel plate (MCP) setting with [3 2] We estimated this gain with the following procedure (90) We used an LED to evenly illuminate the fi eld of view. We fixed the MCP gain setting to a single value (in its


83 scale 0 255), varied the LED intensity and acquir ed images using four different LED brightness levels. Taking these four images, we plotted the pixel signal average S ADU vs. pixel variance 2 ADU and fitted these points to a straight line. The slope of this line was the gain g ( photoelectrons per ADU count ) for the single selected MCP setting (0 255) [3 3] [3 4] [3 5] 2 ADU is the total pixel variance and 2 1(A DU ) is the noise contribution independe nt from the signal intensity (dark noise and readout noise) The desired gain G (ADU counts per photoelectron) is the inverse of g Performing this fitting with an MCP setting 255 was technically enough for our purposes, as this was the only setting we use d. However, w e repeated this procedure several times to obtain more points and check the consistency of the method Figure 3 3 shows the fit for the maximum MCP setting 255, as well as the fit that combines all the measured MCP settings. All the luminescen ce measurements were performed with the maximum 255 gain setting, which corresponded to an estimated gain G of 60 counts per photoelectron.


84 Figure 3 3: iCCD camera gain G A) Linear fit of the 2 vs. mean intensity for a region of pixels in a set of 4 images. All images (points) have MCP setting 255 but different brightness level. B) Repeating the procedure with different MCP settings 0 255 the gain G calibration curve is obtained. The gain error bars in the figure have negligible size ( 0.15 ADU/photoe lectron; ADU: analog digital units) and are not included. All luminescence experiments were performed with MCP setting 255, which corresponds to an iCCD camera gain of ~ 60 counts per photoelectron (last point in inset B). When MCP setting is 0, the gain G is 1 (one ADU count per photoelectron).


85 CHAPTER 4 LUX EXPRESSION IN RESPONSE TO 3OC6HSL AND C8HSL Introduction The quorum sensing (QS) system that regulates luminescence in V.fischeri is known to respond to three distinct autoinducer species ( AI ): 3OC6HSL C8 HSL and AI 2 (44, 63) The experiments described in Chapters 3 and 4 focused on the variability in the QS response to 3OC6HSL concentration. In Chapters 4 (Results and discussion) an d Chapter 5 (Methodology) of this dissertation we describe a study of the lux activity in response to exogenous 3OC6HSL and C8HSL. We wanted to measure h ow the response and wide single cell heterogeneity previously described was now affected or modulat ed b y the presence of C8HSL. In other words, we studied how the individual cell processed the two signals ( 3OC6HSL and C8HSL ) into a single output ( lux activation). These experiments show ed that, while C8HSL su p p ressed the average lux activity of the populatio n, it did not reduce the heterogeneity in lux expression. When testing samples with different 3OC6HSL and C8HSL concentration combinations that produced the same average lux activity; we obtained results that showed very similar lux expression distribution among these samples. In the experiments described in Chapters 2 and 3, we measured the bioluminescent emission from a V.fischeri wild type strain. F or the experiments desc ribed here in Chapters 4 and 5 w e used a V.fischeri JB10 mutant (74) provided to us by Dr Eric Stabb V.fischeri JB10 has a reporter gfp insertion in the lux operon I n this chapter we discuss the difference between the bio luminescence and fluorescence reporter signals.


86 We acqu ired images of thre e independently perfused samples with independent exposures. Using gfp as a reporter allowed us to acquire images with lower exposure times, which gave us more flexibility to design the experiments. This was possible because the single cell fluorescence em ission is several orders of magnitude higher than the single cell bioluminescence A single V.fischeri cell produces 0.01 to 1000 photon/sec (34) while a single gfp m olecule fluorescence emission can reach ~ 500 phot on/sec (91) depending on the excitation It has been estimated that the number of cytoplasmic fluorescent proteins in an expressing cell is ~ 1000 10,000 (5 2, 92) A single GFP molecule has a fluorescence emission of the same order of magnitude as the bioluminescence emission of a single cell. The SNR of the photon emission is improved in fluorescent images when compared to luminescence for equal exposure ti mes. In these Chapters 4 and 5 experiments, V.fischeri JB10 single cells had a fluorescence emission of ~ 2500 p hoton/sec in all directions, and we detected ~ 100 p hoton/sec This fluorescence emission is significantly higher than the luminescence emission o f ~ 30 photon/sec ( ~ 1.7 photon/sec detected) described in Chapters 2 and 3. Measuring V.fischeri l ux Operon Expression in Response to Different 3OC6HSL and C8 Concentration Combinations Measuring the response of single V.fischeri cells to different combinat ions of two different autoinducers requires more data acquisition than when varying only one autoinducer (as was the case in Chapters 2 and 3). The perfusion chamber configuration, while very useful for the data acquisition described in the previous chapte rs is limited in terms of testing a high number of AI concentration combinations.


87 Using the perfusion chamber the experimenter can study only one AI combination per experimental session. Thus, we developed a mic rofluidic device that allowed us to test and measure three different samples at the same time in one experimental session instead of one This microfluidic system was equipped with three independ ent channels that held and perfused three populations of V.fischeri JB10 strain cells. Each of the three samples was perfused with its own corresponding exogenous c oncentrations of 3OC6HSL and C8 HSL autoinducers We measured the single cell GFP fluorescence reporter signal of the lux activation of the isolated V.fischeri JB10 cells adhered to the interior mic roflui dic channels. However, we first performed bulk experiments to characterize the average response of the cells. In this Chapter, we describe those experiments and results The detailed methodology for the single cell and bulk experiments can be found i n the following Chapter 5. Joel Weiss performed many bulk experiments and preliminary single cell experiments. Minjun Son made the silicon wafer that we later used to make the PDMS devices. V.fischeri JB10 Bulk Measurements V.fischeri JB10 Bulk Measurement s Using W ell trays To study the average behavior of V.fischeri JB10, we performed bulk measurements using well trays in a very similar manner as we described in Chapters 2 and 3. We used a programmable Biotek well plate reader to observe and measure 96 bul k samples (including 8 blank controls). During these exp eriments we set 88 diluted washed culture sample s ( ~ 0.001 0.005 optical density at 600 nm and 1 cm path length) in defined photobacterium medium with different added concentrations of 3OC6HSL


88 and C8HS L autoindu cers. We acquired the turbidity luminescence and fluorescence independently from all the sa mples while the cultures grew for a period of ~ 20 hours. The purpose of measuring the response of the cells under different AI combinations was to charact erize the concentration range at which C8HSL affected the action of 3OC6HSL. This bulk data provided an idea of what to expect when doing single cell measurements, but more importantly we would know which AI combinations would be relevant to test. We test e d 3OC6HSL from 0 nM to 1000 nM (0, 5, 20, 50, 100, 200, 500 and 1000 nM), and C8HSL from 0 nM to 2000 nM (0, 5, 10, 20, 30, 50, 100, 200, 500, 1000 and 2000 nM). We used 88 of the 96 wells for culture samples, while the other 8 were used for blank controls V.fischeri JB10 Bulk Measurements Results Each of the 88 wells we monitored provided data over time for several hours ( ~ 20 hours). The objective was to record data while the cells were in exponential phase. There are usually diffe rences among wells in th e number of cells introduced, making the different wells slightly out of phase with each other ( up to one hour in turbidity development) We took an absolute turbidity of 0.1 as a reference for further data processing or OD = 0.16 for 1 cm path length. Th e Figure 4 1 depicts the luminescence and fluorescence of the well cultures in the form of a surface in the plane of exogenous added concentrations ([3OC6HSL], [C8HSL]). The fluorescence emission surface gave us a necessary characterization for choosing co mbinations of the two autoinducers to explore in the single cell experiments using microfluidics. The fluorescence and luminescence si gnals from the same c ulture are correlated, as shown in Figure 4 1 However there is an offset in the fluorescence signal. Plotting the se luminescence and fluorescence signals against each other we obtain Figure 4 2


8 9 Empirically we find that the fluorescence scales approximately as the square root of the bioluminescence. This linear relationship may originate from the fact th at the luciferase enzyme is a heterodimer. That is, it forms from the association of two different peptide subunits encoded by the genes luxA and luxB (9, 93) By contrast the GFP protein is a monome r (94) Figure 4 1: V.fischeri JB10 bulk luminescence and fluorescence at constant OD. The four surfaces show the luminescence (A B) and the fluorescence (C D) of bulk cultures of V.fischeri JB10 at constant OD 0.16 (per cm path length) and varying concentrations of autoinducers (in nM). Panels A B show the same luminescence surface from a different point of view (same as C D for fluorescence). 3OC6HSL has an increasingly inducing effect, and C8HSL has an inhibit ing effect. Luminescence and fluorescence were measured with different gains. The luciferase molecule is a dimer formed by the combination of two monomers. [4 1]


90 A and B are the luciferase subunits concentrations and D is the dim er concentration. At equilibrium, the concentration of the dimer D is proportional to that of the square of the concentration of the monomers LuxA and LuxB If we take A = B equal, we obtain Equations 4 2 and 4 3. [4 2] [4 3] We take the luminescence output L proportional to the luciferase concentration D and the fluorescence output F proportional to the GFP concentration G [4 4] [4 5] T he monomer concentration A is equal to the GFP concentration G W e finally obtain Equation 4 6 grouping all the constants in C [4 6] This is the relationship that we empirically observe in Figure 4 2 after subtracting the fluo rescence of fset. This offset is likely contributed by a fluorescence background from the cells and not related to lux activation (95) Given the relationship in a population of single cells we obtain Equati on 4 7, where F and L are the standard deviations of the fluorescence F and luminescence L respectively. [4 7]


91 We can then obtain a relationship for the coefficients of variation of the fluorescence CV F and luminescence CV L [4 8] Figure 4 2: Fluorescence vs. bioluminescence of a bulk JB10 culture at constant OD. Data acquired at OD 0.16 per cm path length. The fit shows an empirical linear relationship between the GFP fluorescence and the square root of the l uciferase bioluminescence of individual V.fischeri JB10 bulk cultures. Each point represents a culture in one of 88 wells of the well tray. The different wells contain different combinations of autoinducers, making them induced at different levels. The flu orescence offset is likely given by autofluorescence from the cells and not due to lux activity.


92 I n Chapters 2 and 3 of this diss ertation, we obtained CV L values ~1 in the bioluminescence response In the following section of this Chapter 4 we show single cell fluorescence measurements using microfluidics where we obtained CV F values ~0.5 which is consistent with the equation above and explains the difference Studying V.fischeri JB10 Quorum Sensing Induction at the Single C ell L evel We explain in this sec tion our studies of the QS induction at the single cell level by measuring single cell fluorescence using microfluidic devices. The technical details of the experiments are given in Chapter 5. Measuring Single Cell l ux Expression in a Microfluidic Device i n Response to Different 3OC6HSL and C8HSL Combinations We adhered V.fischeri cells to the interior of the microfluidic channels of the de vice in its observation region The cells were perfused with fresh defined photobacterium medium with the three differe nt selected combination s of exogenous 3OC6HSL and C8HSL autoinducers one for each observation channel T he medium flow maintained the cells under the constant perfused exogenous concentrations, while also providing them with nutrients and oxygen. Once the exogenous concentration s stabilized in the measurement channel, we started acquiring images. We collected pairs of images repeatedly every ~ 15 minutes in each of the three independent observation channels. Each pair consisted of a phase contrast image (0. 15 second exposure) immediately followed by a fluorescence image (10 second exposure). The first image provided us with the location coordinates of the cells, while the second image recorded the GFP fluorescence signal. The microscope focusing, microflu idi c device stage, cell culture general condition s etc. were manually monitored every ~ 15 minutes before each image pair acquisition to ensure the system stability.


93 Figure 4 3: GFP fluorescence of V.fischeri J B10 strain as derived from bulk me asurements in a plate reader. Data are shown for OD 0.1 6 per cm path length in defined photobacterium medium at 25 C. The map is generated from data collected at 88 signal combinations. The four groups of three connected dots indicate the (3OC6HSL;C8HSL) combinations c hosen in the microfluidics single cell experiments: low 3OC6HSL at 100 nM, high 3OC6HSL at 1000 nM, constant fluorescence at 37% of maximum emission, and constant fluorescence at 67% of maximum emission. Since the microfluidic pattern we used was equipped with three independent observation channels, we could observe and measure three samples per experimental session. The three chosen sampl es in a session belonged to a characteristic curve in the fluorescence surface shown in Figure 4 3 This dissertation sh ows dat a of four different experiments that follow four curves in the surface: high 3OC6HSL (1000 nM), low 3OC6HSL (100 nM), constant fluorescence signal at 37% of maximum output, and co nstant fluorescence signal at 67 % of maximum output. The maximum measu red fluorescence (referenced as 100%) in the surface corresponds to the well with 1000 nM


94 3OC6HSL and no C8HSL. These percentage values are a n approximate guideline obtained from the interpolation of the well tray data that generated Figure 4 3 The curves we followed in each of the data sets are shown in Figure 4 3 Single Cell l ux Activation Measurements Results We show in this subsection the details of the single cell results obtained for the four different sample sets represented and connected by dotted lines in the bulk data graph in Figure 4 3 In each set, we introduced the culture in the three microchannels and perfused them independently under well defined combinations of 3OC6HSL and C8HSL concentrations. We later discuss a t test to evaluate the st atistical significance of the differences between distributions collected under different conditions. We acquired the data from a pair of images (phase contrast and fluorescence) for each experiment. The images w ere taken at 3 4 hours from exogenous AI int roduction for all the experiments We had observed in our previous experiments with 3OC6HSL that at 3 hours after AI introduction the average single cell response had already stabilized ( Figure 2 7 and Figure 2 6 ). Ad ditional analysis of the 67% maximum ou tput set gave more evidence in favor of the cells stabiliz ing their response at ~ 3.5 hours. Comparing images taken at 3 and 3.6 hours gave differences of only 5% in the median and standard deviation of the total single cell fluorescence for the three chann els We performed Experiments #1 and #2 to investigate how the addition of increasing concentrations of C8HSL modulated the single cell lux activity of a sample activated by exogenous 3OC6HSL. The Figure 4 4 represents three populations of cells under the same concentration 1000 nM of 3OC6HSL autoinducer and different C8HSL concentration s (Experiment #1) The histograms show the detected single cell GFP fluorescence distributions T he sample with no C8HSL has higher median and standard


95 deviation values foll owed by the one with 500 nM C8HSL; while the sample with 2000 nM C8HSL has its distribution the most collapsed to the low intensity end C 8HSL suppressed the single cell average lux activity as was observed in the bulk measurements ( Figure 4 1 ). However, t he addition of C8HSL did not simply shift the single cell distribution but it reduced the standard deviation This feature was hidden in the bulk measurements since the only data available were average values. Figure 4 4: Single cell V.fischeri JB10 GFP fluorescence at 1000 nM 3OC6HSL (Experiment #1). The detected GFP fluorescence distributions for three independent populations of V.fischeri JB10 are shown. The red lines indicate the median and standard deviation. The three samples contain 1000 nM 3OC6HSL and different C8HSL concentrations. The addition of C8HSL collapses the distribution to the low intensity end. Each of the three samples consists of 200 250 individual cells. When testing the set with three samples at a 3OC6HSL concentration of 100 nM and different C8HSL (Experiment #2) we obtained the distributions shown in


96 Figur e 4 5 Again, C8HSL suppressed the GFP fluorescence signal while narrowing the distributions (smaller standard deviation) Figure 4 5: Single cell V.fischeri JB10 GFP fluoresce nce at 100 nM 3OC6HSL (Experiment #2). The detected GFP fluorescence distributions for three independent populations of V.fischeri JB10 are shown. The red lines indicate the median and standard deviation. The three samples contain 100 nM 3OC6HSL and differ ent C8HSL concentrations. The addition of C8HSL clearly collapses the distribution for the samples with 500 and 1000 nM. Each of the three samples consists of ~ 200 individual cells. We then asked the question of how the distributions would change if we cho se the samples in a set so that the three of them would have the same approximate average lux activity reported by GFP fluorescence. The objective was measuring if V.fischeri processed these different input values ( AI combinations) any differently to produ ce the final gfp expression. To this effect we chose t wo sets, one of them with three samples at 67% average fluorescence intensity (Experiment #3) and another with samples at 37% average intensity (Experiment #4) We took as the 100% reference the intens ity of the bulk sample with 1000 nM of 3OC6HSL and no C8HSL. Figure 4 6 and Figure 4 7


97 show the single cell fluorescence distributions for these two sets. The figures show that the median intensities of the samples in a set remain c lose within 15% of each other. Furthermore the widths of the distributions (standard deviations) in a set also remain similar (again within ~ 15% of each other) and do not show the different widths presented by experiments #1 and #2 ( Figure 4 4 and Figure 4 5 ). Figure 4 6: Sing le cell V.fischeri JB10 GFP fluorescence at different combinations of 3OC6HSL and C8HSL (Experiment #3). The detected GFP fluorescence distributions for three independent populations of V.fischeri JB10 are shown. The red lines indicate the median and stand ard deviation. The three samples contain three combinations of 3OC6HSL and C8HSL concentrations that lie on the 67% activation contour. The populations show little difference ( ~ 15%) in the median fluorescence and coefficient of variation (CV) compared to o ther sample groups with constant 3OC6HSL. Each of the three samples consists of 180 250 individual cells. According to the sample sets shown here, C8HSL has the inhibiting effect observed in the bulk experiment C8HSL also collapses the single cell intensi ty distribution as can be seen in the set with constant [3OC6HSL] = 1000 nM ( Figure 4 4 ) H ow ever, the lux activity distribution seem s to be insensitive to different combinations of


98 3OC6HSL and C8HSL concentrations that produc e the same average lux activit y. These last Experiments #3 and #4 produced very similar distributions (average and standard deviation). Figure 4 7: Single cell V.fischeri JB10 GFP fluorescence at different combinations of 3OC6HSL and C8HSL (Experiment #4). The detected GFP fluorescen ce distributions for three independent populations of V.fischeri JB10 are shown. The red lines indicate the median and standard deviation. The three samples contain three combinations of 3OC6HSL and C8HSL concentrations that lie in the on the 37% activatio n contour. The populations show a smaller difference ( ~ 20%) in the median fluorescence and coefficient of variation (CV) when compared to other sample groups with constant 3OC6HSL. Each of the three samples consists of 145 230 individual cells. We performe d t tests (at the 5% significance) to statistically test these observations and compare with each other the samples within the experiment sets. Table 4 1 shows the numerical results and Figure 4 8 shows a graphical representation of the test When comparin g the samples of Experiment #1 (high 1000 nM 3OC6HSL) with each other, the test found statistical difference among all the samples.


99 Table 4 1 V.fischeri JB10 single cell fluorescence distributions comparison. Pairs of samples were compared in each set usi ng an unpaired t test at the 5% significance. The table shows if the t test finds similarity between the two distributions and the corresponding p value. Experiment Number Sample pair ([3OC6HSL], [C8HSL]) (nM) Are distributions statistically different? p v alue 1 (1000, 2000) (1000, 500) Yes < 0.001 1 (1000, 2000) (1000, 0) Yes < 0.001 1 (1000, 500) (1000, 0) Yes < 0.001 2 (100, 0) (100, 500) Yes < 0.001 2 (100, 0) (100, 1000) Yes < 0.001 2 (100, 500) (100, 1000) No 0.88 3 (800, 100) (200, 0) Yes 0.04 5 3 (800, 100) (500, 20) Yes 0.0013 3 (200, 0) (500, 20) No 0.42 4 (500, 300) (1000, 1000) Yes 0.013 4 (500, 300) (50, 0) No 0.91 4 (1000, 1000) (50, 0) Yes 0.0027 Experiment #2 (low 100 nM 3OC6HSL) found difference between all pairs except (100 nM 3OC6HSL; 500 nM C8HSL) and (100 nM 3OC6HSL; 1000 nM C8HSL). The bulk results indicate that there is little difference in the average response between these

PAGE 100

100 two AI combinations (within 10%). Furthermore, these two samples combinations are less than 10% abov e the lowest signal samples (0 nM 3OC6HSL). It is possible the suppression produced by 500 nM and 1000 nM C8HSL is not very different, making the samples relatively similar. Figure 4 8: Graphical representation of a t test on single cell JB10 lux express ion distributions. Pairs of samples were compared in each set using an unpaired t test at 5% significance. Each point represents the comparison of a pair of distributions with its corresponding p value label. The closer the median values are to each other (closer to diagonal), the more similar the distributions (higher p value). Evaluating e xperiment #3 samples (67% emission) with the t test, we found that (200 nM 3OC6HSL; 0 nM C8HSL) and (500 nM 3OC6HSL; 20 nM C8HSL) samples are statistically similar. Howe ver, the first sample (800 nM 3OC6HSL; 100 nM C8HSL) was found to be statistically different than the two others. Experiment #4 (37% emission) presented a similar situation, sample (1000 nM 3OC6HSL; 1000 nM C8HSL) was found statistically different to the t wo others. In experiments #3 and #4, i t is possible that the

PAGE 101

101 samples found dissimilar to the two others in the set actually did not have the right 3OC6HSL/C8HSL combination. Small movements in the (3OC6HSL; C8HSL) plane represented in Figure 4 3 may produc e big intensity variations. Taking the data of each population independently, we can calculate t he coefficients of variation ( CV F ) of single cell GFP fluorescence Figure 4 9 shows the CV F values obtained for these sets. As shown in the different figure pa nels, t he CV F for single cell GFP fluorescence as a reporter for the lux system is approximately ~ 0.5 ( CV F ~ 0.4 0.6 ) Figure 4 9: Coefficients of variation ( CV F ) of single cell V.fischeri JB10 GFP fluorescence for four different sample sets using a micro fluidics system. Each panel shows the CV F (standard deviation over median) of each of the three independent populations measured. The corresponding 3OC6HSL and C8HSL concentrations used for each sample are indicated in each panel. We observed that the CV F values remained approximately 0.4 0.6.

PAGE 102

102 Discussion We used the V.fischeri JB10 mutant for investigating how the bacterium integrates the two simultaneous quorum sensing (QS) signals given by 3OC6HSL and C8HSL autoinducers. The bacterial strain contains a gf p chromosomal insertion in the lux system under the control of the lux promoter The bacterium synthesizes green fluorescent protein (GFP) along with luciferase when the lux system is a ctivated via QS We first performed bulk experiments using a programmab le well tray reader The bulk measurements allowed us to characterize both the luciferase bioluminescence and the GFP fluorescence for cultures in defined photobacterium medium under different concentrations of 3O C6HSL and C8HSL autoinducers. W e observed t hat C8HSL has an inhibitory effect at all tested positive 3OC6HSL concentrations, with a sharper step in the range ~ 200 800 nM C8HSL in fully activated cultures ( Figure 4 1 ). In a second phase, w e tested different single cell populations under different co mbinations of 3OC6HSL and C8HSL autoinducers. We designed and used a microfluidics system to streamline our single cell measurements and increase the data acquisition throughput. By measuring GFP fluorescence instead of luminescence, we were able to acquir e more images much faster: 10 second exposures against the 10 minute exposures required for luminescence. We used this experimental system to simultaneously measure three independent samples. We perfused them with an external liquid medium flow that fixed the extracellular concentrations of 3OC6HSL and C8HSL to the desired value s, and negated the QS autoinduction by washing away any secreted AI In our studied single cell samples, we found C8HSL both inhibited the average GFP fluorescence emission and signi ficantly narrowed the intensity distribution (up to a

PAGE 103

103 third in a well induced population, Figure 4 1 ) when comparing induced populations under the same 3OC6HSL concentration. The average inhibiting effect of C8HSL had been observed in previous studies (45) and we confirmed it with our bulk measurements, so it made sense also finding it in our single cell experiments. In addition, we studied different 3OC6HSL/C8HSL concentration combinations that produced the same ave rage GF P fluorescence emission. W e o btained results showing that the width of the distribution of V.fischeri JB10 activation levels is independent from the 3OC6HSL and C8HSL combination and seems to depend only on the average activation level ( Figure 4 6 and Figu re 4 7 ) The addition of C8HSL did not improve the V.fischeri lux response precision. When correlating luminescence and fluorescence of individual bulk cultures, we observed that the GFP fluorescence was linearly correlated with the square root of the luci ferase luminescence ( Figure 4 2 ). We hypothesized that this relationship might be due to the luciferase being a dimer and the GFP a monomer. The CV F fluorescence values we later found in our single cell experiments were CV F ~ 0.4 0.6. This CV F combined with the luminescence CV L ~ 1 we found in the perfusion chamber experiments described in Chapters 2 and 3, point in the direction of the relationship we would expect given the square root correlation we found in the bulk measurements. If this is the case, GFP fluorescence is a more direct indicator of the noise in lux expression. L uciferase bioluminescence has a higher noise (twice the CV ) than fluorescence emission due to being a dimer. The CV F ~ 0.4 0.6 values we obtained were larger than the reported values in other recent works that studied QS in Vibrio harveyi another model Gram negative bacterium

PAGE 104

104 Anetzberger et al. (80) studied the bioluminescence of V.harveyi single cells and reported a CV L ~ 0.25 lower than the CV L ~ 1 we found in our measurements. Another study by Long et al. (81) in V.harveyi using GFP as a reporter of QS activation measured a CV F ~ 0.2 0.4 somehow lower than the CV F ~ 0.4 0.6 we described fo r V.fischeri JB10 in this chapte r. These differences are perhaps explained by the differences in V.fischeri and V.harveyi QS networks and the experimental conditions V.harveyi uses three AinS/AinR type autoinducers in an additive way, while 3OC6HSL and C 8HSL in V.fischeri interact competitively in binding to LuxR. Furthermore, in Anetzberger et al. experiment the exogenous concentra tions were not controlled. Also, i n Long et al. experiment they did not measure the lux response but the reporter of an inter mediary in the activation network. C8HSL competes with 3OC6HSL by binding to LuxR (45) effectively reducing the fraction of LuxR receptor that 3OC6HSL can bind to. Fewer available LuxR receptors for 3OC6HSL among cells wo uld lower the average lux activity, as well as the chance of having cells with very high output. This would explain the reduction of the average and the width of the activation in the single cell distribution. C8HSL autoinducer regulates the synthesis of t he 3OC6HSL LuxR receptor (63) is capable of activating luminescence by binding to LuxR (29) and also participates in host colonization (12) At lo w cell densities, when 3OC6HS L concentration is not high enough, C8HSL binds to LuxR to stabilize it, prevent its rapid degradation and induce a weak luminescence. The LuxR C8HSL complex is less active in promoting over the lux operon than is the LuxR 3OC6 HSL complex (44) This causes C8HSL to suppress the average and width of the luminescence distribution when 3OC6HSL is present.

PAGE 105

105 Furthermore, different combinations of 3OC6HSL and C8HSL with the same average also have the same distribution width. The suppression of lux activation by C8HSL could be an adaptation mechanism for the cells to prevent luminescence when it is not necessary at low cell densities (45) By suppressing t he average and variance of the luminescence distribution, the population ensures at early colonization stages that most cells will be dedicated to other possibly more important functions. We used a t test to compare single cell distributions since they wer e reasonably similar to Gaussian functions. However, to perform a deeper analysis another function would be more suitable for fitting the histograms. In a previous study Cai et al. (2006) proposed a gamma distribution as a basis for modeling this distribu tion (96) Proteins are expressed by the formation of mRNA molecules by transcription. Each mRNA Cai et al. showed that a function using the gamma function appropriately fits this phenomenon. [4 9] p(x) is the steady state probability distribu tion of protein number per cell The and parameters are only dependent on th e mean m and standard deviation of the distribution.

PAGE 106

106 CHAPTER 5 LUX EXPRESSION IN RESPONSE TO 3 OC6HSL AND C8HSL: MATERIALS AND METHOD S We studied the response of the V.fischeri l ux operon under well defined external concentrations of aut oinducers. In this part of our research we varied the exogenous concentrations of 3OC6HSL and C8HSL autoinducers. The concentration of the third known autoinducer (AI 2), has little effect on l ux activation A previous study found that a luxS mutant, the A I 2 synthase, had 70% the bioluminescence emission than the wild type (47) In our experiment, the external AI 2 concentration was effectively set as zero by the perfusion device We used microfluidic devices and observed the fluorescence of a GFP reporter for the lux genes Single cells were immobilized inside the microfluidic channels and perfusion flow maintained the cells under a constant concentration of autoinducers. We exposed the cells under different combinations of concentrations of exogenous 3 OC6HSL and C8HSL autoinducers. Using the microfluidic device instead of the perfusion chamber allowed us to measure three samples in one experimental session instead of one. In the following sections we describe the details of these experiments. There are a few common points between this set of experiments and the one described in Chapters 2 and 3. We will occasionally refer the re ader to these chapters to avoid repetition. Figure 5 1 shows a panoramic view of our microscopy configuration apparatus. Vibrio fischeri JB10 Strain and Cultures We used the Vibrio fischeri JB10 strain for all the experiments detailed in Chapters 4 and 5 as well as for an additional experiment discussed in Chapters 2 and 3 Vibrio

PAGE 107

107 fischeri JB10 stra in was provided by Dr Eric Stabb. This strain is derived from the V.fischeri wild type strain ES114 and produces fluorescent GFP when lux is induced Figure 5 1: Microfluidics experimental configuration picture. The photograph shows the experimental appa ratus we used for measuring the lux activity of V.fischeri cells using microfluidics. The Nikon microscope can be seen at the center. The Teflon base with microfluidic device is fastened at the center of the microscope stage. Teflon tubing is attached to each inlet and perfused with independent syringe pumps. Our original JB10 strain was stored in the vap or phase of liquid nitrogen (~ 196C) in medium containing 15% glycerol. To prepare a culture for experimentation, we removed a s mall macroscopic amount of culture from the frozen stock to make an overnight culture in fresh growth medium. The cells were grown at room temperature ( ~ 22 C) on an orbital shaker to keep the culture oxygenated. This overnight was used to start a culture i n a photobacterium agar plate (agar

PAGE 108

108 photobacterium medium, No. 786230 Carolina Biological, Burlington NC). The agar plates were stored in a laboratory refrigerator at ~ 5 C. An agar plate culture is usually useful for up to 7 10 days. After this time, it is necessary to make a ne w agar plate culture. The plates were regularly checked for contaminations. Defined Growth Medium In the experimental procedures described in these Chapters 4 and 5 we used a defined photobacterium medium, except for the pl ate cultures where we used photob acterium medium (Carolina photobacterium powder) The defined medium is based on an artificial seawater extract and contains Tris (organic buffer), phosphate, ammonium (nitrogen source) and glycerol (carbon source). See Chapter 3 for the detailed ingredien t list of the growth media We found that V.fischeri JB10 strain had trouble growing in the defined medium we used for V.fischeri MJ11 in the perfusion chamber experiments (Chapters 2 and 3) To better support the V.fischeri JB10 cells, the medium was furt her supplemented with 0.3 g/ml of casamino acid extract (97) Bulk Measurements The procedures we followed for performing the bulk measurements were very similar to those described in Chapter 3. We will not re peat them again here but only mention the differences. In this second project we studied the response of V.fischeri cells under variations of both 3OC6HSL and C8HSL, unlike in Chapters 2 and 3 where we only varied 3OC6HSL. In the present case we needed to test many more samples as we were studying points on an AI concentration plane and not on an AI concentration line. We

PAGE 109

109 exclusively used 96 well plates (Costar polystyrene black wall 96 well plate, Corning Inc.) with black walls to reduce cross talk among w ells. During thes e measurements all 96 wells w ere occupie d with samples and a few (6 to 8 ) with blank controls. Each well contained a different combination of 3OC6HSL and C8 HSL. Given that filling all the wells with the right concentrations took a signific ant amount of time (~1 hour), we were careful to leave the introduction of the diluted V.fischeri culture for the end so that all wells started growing at a similar time. The loaded 96 well plate was then placed inside the Biotek well plate reader overnigh t for measuring in an approximately 20 hour long experimental run at ~ 26 C. The reader was programmed with a sequence that acquired fluorescence, luminescence and turbidity from each well at regular intervals ( ~ 5 minutes). The well plate was covered all th e time with a non airtight anti evaporation lid and softly shaken in between measurements. Cell Sample Preparation for Single Cell Experiments When preparing for our single cell experiments, we scraped a small amount of solid culture from the agar plate su rface and used it to start a culture ( ~ 12 hours) in defined liquid medium on an orbital shaker. We then took a small liquid amount ( ~ 50 l) from the recent culture and introduced it in 7 ml of fresh defined medium. This culture was left to grow on the orbital shaker for 2 5 hours until it reached an optical density (OD) near 0.05 at 600 nm wavelength and 1 cm path length. The OD was measur ed with reference to fresh defined medium (which has the same OD as water at that wavelength). We performed the OD measurements with a Shimadzu UV 1601 spectrophotometer. We use the optical density (OD) as a measure of cell culture density. The 0.05 OD cul ture was further diluted to an OD of 0.015 0.03, which we

PAGE 110

110 found was a suitable V.fischeri density to perform the experiment as explained below. This last diluted culture is introduced in the microfluidic device. There were some differences between cell pre paration for the microfluidic experiments and cell preparation for bulk experiments described earlier. We performed several preliminary experiments to establish a suitable protocol for preparing cells for microfluidic experiments. One difference between bu lk experiments and microfluidic experiments is that the cells we introduced in the microfluidic device were mostly the same cells that adhered to the surface and we later observed. By contrast, in the bulk experiments w e introduced a diluted culture with O D at the detection level of our well plate reader and let it grow while measuring. The V.fischeri density increased a couple of orders of magnitude or more. Therefore the original cells contributed to a small fraction of the bulk measurement and most of th e detected signal is produced by their daughters. Furthermore in single cell protocols it is preferable to maintain the cells at low density (OD < ~ 0.1) to reduce exposure to AI s, as it has been suggested that the lux system may present hysteresis (82) The cells that are introduced in the microfluidic device need ed to be in exponential phase, which means they must have been growing for around 2 5 hours in fresh defined medium at low concentrations (OD < 0. 0 5) We found that a V.fischeri culture in exponential phase at OD ~ 0.015 0.03 satisfied our need. A lower cell density would not give us enough single cells (> 100) on the observed surface. A higher density would produce too many cell clusters by replication b efore the experiment was completed. Experimental Configuration and Procedures for Single cell Measurements The V.fischeri cells were observed in a micro fluidic device that consisted of a poly(dimethylsiloxane) (PDMS) elastomer film sealed by a glass covers lip. The culture

PAGE 111

111 was loaded into the microchannels and V.fischeri adhered to the inside of the device window during the course of the experiment. The data was collected using a microfluidic device with three independent channels, which allowed us to simult aneously observe the prepared culture at three different exogenous 3OC6HSL and C8HSL combinations. Figure 5 2 shows the microfluidic pattern. Microchannels ( ~10 x 400 2 cross section) were embedded in the device by soft lithography The PDMS film is opti cally transparent down to 230 nm wavelength. It is also non toxic to cells, impermeable to water and permeable to gases (98) Figure 5 2: Microfluidic three way device pattern. The figure shows in detail the pattern e tched (carved) into the PDMS film. This etched side is bonded to a glass coverslip to complete the channel structure. The whole device is around 6 cm long and 3 cm wide. The observation area is shown where the V.fischeri cells are attached in the channel i nteriors. We could only observe and measure 1/10 of this observation area. The three channels are independent and perfused by the three indicated inlets. The channel distance from the three inlets to the three observation channels is the same to have simil ar hydrodynamic resistance. The channels combine only at the outlet to drain the perfused media that already washed over the cells. Each independent channel is 10 mm long, 400 m wide and 10 15 m deep.

PAGE 112

112 Optical Equipment and Image Acquisition The microfluidic device was fastened to a Teflon base with four inlet/outlets that sealed to the PDMS film openings by light pressure O rings. The base with the device was placed on the st age of a Nikon Eclipse TE2000 U inverted microscope. The sample light was collected with a 20x objective (NA 0.5). The Teflon base was attached to the microscope stage to reduce small micrometric displacements during the several hours ( ~ 5 hours) the devic e needed to be under observation. The images were acquired in a dark room to reduce external light. The microscope was equipped with two separate sets of lenses and filters for acquiring green fluorescence emission (GFP) and red fluorescence emission (red dye fluorescent for control ; Sulforhodamine, Acros Organics, CAS: 60311 02 6 ) We sequentially measured the samples of the three samples in the independent microchannels. Each microchannel had an observation area of 400 x 32 0 cells adhered to the glass window (and occasionally to the PDMS surface). We usually aimed to fill the channels so as to have ~200 250 single isolated cells in the observation area during the measurement. We acquired images for 4 5 hours with a cooled CCD camera (Photometr ics Coolsnap HQ 2 ) with the sensor kept at 35 C to reduce dark current. The CCD ch ip had 1392x1040 pixels (6.45x 6.45 ) and an imaging area of 8.98x 6.71 mm. The microscope provided x21 magnifica tion (x20 objective, x 0.7 relay lens and x 1.5 lens switch) which made the pixels correspond to 0.3 x0.3 field of view Therefore V.fischeri cells with a rod shape (typically ~3 imaged on a rectangle ~10 pixels long ( Figure 5 3 ). Each pair of images consisted of a phase contrast image immediately followed by a fluorescence image. The sample was externally illuminated with white light to acquire

PAGE 113

113 th e 15 millisecond exposure phase contrast image, which gave us the locations of the cells. The fluorescence image (10 seconds exposure) was taken immediately afterwards to image the fluorescence emission of the GFP contained inside the induced cells using a high intensity lamp (Fluorescent lamp, Nikon Intensilight C HGFI). This repetitive and frequent image pairing gave us an almost real time control of the correspondence between the phase contrast and the fluorescence images in terms of focusing drifts, eve ntual micrometric stage movements, cell detachments, etc. We acquired a pair of images every 15 20 minutes for 4 5 hours. In between acquisition of image pairs, we manually checked the culture sample and experimental configura tion Microfluidic Devices Fig ure 5 2 shows the device we used for acquiring the data. At the observation area, each channel is 10 mm long, 400 m wide and 10 15 m high. Each channel was independently perfused by its own syringe using Teflon tubing and a programmable syringe pump (KD Scientific). Microfluidic devices were fabricated as follows. The microfluidic pattern w as first designed using A utocad This pattern design was then transferred to make a transparency mask in a 1:1 scale Minjun Son performed all the lithography work to make the wafer that we later used. The silicon wafer was made from the transparency mask by deep reactive ion etch ing. In this process a blank 4 inch silicon wafer surface was sequentially washed with acetone, isopropanol and deionized water. The wafer was briefly heat dried on a plate and treated with HMDS to improve the photoresist adhesion. The wafer was then coate d with S1813 photoresist on a Suss Delta 80 spin coater at 4000 rpm for 50 seconds, bringing the photoresist to a thickness of 1.125 m. The wafer was soft baked for 80 seconds on a 110 C hotplate. The photoresist was

PAGE 114

114 exposed to 395 nm UV light through the transparency mask on a Karl Suss MA6 mask aligner for 15 seconds. Later, the treated wafer was developed in 300MIF diluted in DI water in a 1:1 ratio. The wafer was swirled in the solution for 2 to 3 minutes until the pattern was resolved. The pattern was then placed in a STS Deep Reactive Ion Etcher, bringing the depth of the raised pattern to 10 15 m. The etched wafer was finally exposed to oxygen plasma to remove the photoresist and coated with C4F8 to prevent adhesion of PDMS to the silicon wafer. The finished wafer with the etched corresponding pattern was reusable and worked as a mold for making the PDMS layer of the microfluidic device. This wafer was placed in an aluminum cast ing dish with a close fitting lucite lid. The lid had several op enings to inject PDMS through one of them and let the air escape through the others while covering the etch ed wafer. For the PDMS mix we used a 10:1 ratio of Sylgard 184 silicon elastomer base and curing agent. The PDMS filled aluminum base was baked at 110C for around 15 minutes or until the PDMS acquired a spongy con sistency. We let the baked PDMS cool overn ight in the dish The PDMS film with the pattern etched was carefully peeled away from the wafer. Using a scalpel, the etched pattern was cut from the wafer All this process, as well as the following bonding, had to be performed under a laminar flow hood so as not to let small particles (or tiny cloth fibers) adhere to the PDMS film on the etched side. If particles end up between the PDMS film and the glass coverslip, they can easily disrupt the channel pattern making the device unusable. We exposed the PD MS and glass coverslip surfaces to oxygen plasma for around 2 minutes with an Electro Technic high voltage generator to bind them. Both surfaces

PAGE 115

115 were gently pressed together and the finished microfluidic device was left to sit overnight. Figure 5 3: Sing le cell fluorescence data analysis. The four panels correspond to the same single V.fischeri cell The axes show the pixel scale for this particular selection. A and B show the box selection in the phase contrast and fluorescence acquired images, respectiv ely. C shows the painted pixels from A that we used to estimate the cell size, 36 pixels in this case. D shows the painted pixels from B we used as signal pixels. The signal pixel intensities are added and subtracted the local background to obtain the tota l fluorescence for this V.fischeri cell. The red crosses at the center of the cell shapes are the centers automatically selected by the code. Most of the time, the centers in A and B would match but in the case of shifts there may be small micrometric diff erences (1 3 pixels).

PAGE 116

116 Adhering Single Cells to the Observation Surface We observed that V.fischeri cells had a very mild tendency to attach to PDMS and glass surfaces However, only a minority of cells attached to the surfaces and most that did attach wou ld easily detach after a short period of time from 5 to 20 minutes. We utilized a poly L lysine treatment to increase cell adhesion. The positive charges of poly L lysine contribute to the adhesive electrostatic interaction of the negatively charged glass surface and outer cell membrane (99) The protocol we followed was introducing a concentrated solution 1 mg/ml of poly L lysine (Sigma Aldrich, poly L lysine hydrobromide, 300,000 MW, CAS: 25988 63 0) in the chan nels of the finished microfluidic device. The device was filled with the solution and incubated overnight in a refrigerator at 5C. Then the cell culture with an appropriate concentration was directly introduced into the device. Enough flow rate and time w as given to completely wash away the bulk of the poly L lysine solution and leave a poly L lysine coating on the glass window. We performed a well plate experiment to test whether the poly L lysine affected the behavior of V.fischeri cells in our experimen tal conditions. Several V.fischeri JB10 culture samples with different poly L lysine concentrations were left to grow and compared with control V.fischeri JB10 samples We did not detect difference in luminescence or fluorescence intensity between the test ed and control samples. Keeping Adhered S ingle cells under a Constant External M edium We introduced the prepared sample culture in the microfluidic device by back flowing it through the outlet (up to 1 2 ml/h divided in the three channels) for ~ 10 15 minut es. Once the device was properly filled with the sample, we slowed down

PAGE 117

117 the flow to let the bacteria adhere to the glass surface. The syringe pump with the sample, simultaneously feeding the three channels, was set to 0.02 ml/h We waited until ~ 200 cells were adhered in the 400 x 320 Obtaining this cell distribution could take a variable time from 15 minutes to up to 1 hour. Having the cells adhering fast prevented the early formation of cells clusters during measurements. We stopped t he culture flow when we had obtained the desired distribution of adhered cells. We connected the three fresh defined media to the three channel inlet s Each perfusion medium contained a different combination of exogenous 3OC6HSL (N ( ketocaproyl) L homose rine lactone, Sigma Aldrich, CAS: 143537 62 6) and C8HSL ( N octanoyl L homoserine lactone, Cayman Chemical company, CAS: 147852 84 4). We used concentrations ranging 0 to 2000 nM. We added 10 1 5 ng/ml of red fluorescent dye (sulforhodamine) to the perfusio n media. The red fluorescence would indicate us when the perfusion had properly reached the microchannels and the bacteria were under stable concentrations of AI s. The three perfusion media flows were finally set at 0.02 ml/h for image acquisition. Single cell Fluorescence Data Analysis We performed the data analysis of the acquired images with a custom code we programmed in Matlab. The images were acquired using the software Winview 32 which controlled the CCD camera. The analysis technique we applied took into account the contour of the cells. The single cell fluorescence images are brighter and the shapes have better definition than the luminescence images w e analy zed in Chapters 2 and 3. We estimate d the GFP

PAGE 118

118 fluorescence emission to be ~ 100 times the bio luminescence. Therefore the SNR of the photon emission is ~ 1 0 times higher for the fluorescent cell. We applied a G aussian filter (sigma width of 1 pixel ) to the fluorescence images to reduc e spatial high frequency noise We subtracted corresponding backgr ound images to both phase contrast and fluorescence images. A flat field correction was then applied to the fluorescence image. After this correction, the pixel intensities in the fluorescence image are representative of the local GFP fluorescence and are independent of the local excitation illumination (save for a global constant). The next step in the analysis was to identify individual cells in the 10 40x1392 phase contrast image. We defined a rect angular region approximately 25 35 pixels per side around each V.fischeri cell. Given the cell distribution and the f ield of view area, we could typically locate ~ 200 cells per image. The code independently processed each rectangular region. It fitted the intensity histograms of both phase contrast and fluorescen ce to determine cutoff values from the local background. Using the cutoff values it looked for connected pixels above these values to determine the signal pixels. In the fluorescence pixel range, adding the pixel values and subtracting a local background y ielded the total cell fluorescence signal. The code used the cell position in the phase contrast and fluorescence pixel ranges to account for rare micrometric shifts in between both images. Figure 5 3 shows four panels corresponding to the same V.fischeri cell. T he single cell box selection is shown in the phase contrast and the fluorescence images, along with the corresponding painted pixels for each.

PAGE 119

119 This code method was developed and tuned to avoid it being confused by rare micrometric shifts in individu al cells, any extraneous cell appearing in the outer border of the box selection or any other not so frequent anomaly. The box selection process of the cells across the 1040x1392 phase contrast images was performed manually. Most of the rest of the procedu re was automated. Preliminary Trials with a Microfluidic D evice Using a Concentration G radient We tested different microfluidic patterns when developing the right techniques and protocols for the experimental project described in C hapters 4 and 5. One of t hem was particularly interesting, and we worked the most with it before switching. In this case, the device had a region where a gradient was formed after two input perfusion media (with 3OC6HSL and C8HSL each) at different concentrations were mixed at dif ferent degrees. We found the gradient had fluctuations over time while building and took too long to stabilize. By the time we could start the measurements, too many cell clusters had formed. Furthermore, the gradient fluctuations during the build up expos ed the cells to the wrong AI concentrations. Anothe r difficulty was that the number of cells per gradient band ( AI concentration combination) was too small to perform statistics ( ~ 10 cells or less). Using microfluidics opens many possibilities to investiga te single cell samples. For the more sophisticated studies, g reat part of the work is designing and making work the right device. It only takes looking at the expert groups in the community to see some amazing examples of what can be achieved with these te chniques.

PAGE 120

120 CHAPTER 6 CONCLUSIONS AND FUTU RE DIRECTIONS A considerable amount of investigation has been done on quorum sensing (QS) networks in bulk po pulations. However, studies of QS regulat ion at the single cell level have just begun in the past few year s. V ery little work has been done on single cells in controlled environment s. Like all gene regulatory networks, QS networks are subject to genetic noise. This noise plays an important role in many systems to the point that networks have been identified w here cells have incorporated noise and rely on it for their physiological function. Traditional bulk experiments only show the average population features Therefore, a deep understanding of QS, as well as the role that noise plays in its regulation, canno t be achieved without looking at individual cell responses. In this dissertation we first studied the impact of genetic noise in the output of the luxI/luxR system of the model QS marine bacterium Vibrio fischeri To this end we built a high sensitivity mi croscopy experimental system with a perfusion chamber. Utilizing this measurement equipment, we maintained single cells under a perfusion flow in a well defined external environment with known concentrations of 3OC6HSL autoinducer. We measured the low inte nsity bioluminescence of single cells, emitted through activation of the lux operon, and compared it against the behavior of bulk cultures. The average behavior correlated to that found in bulk cultures, but w e found a high degree of heterogeneity ( CV L ~ 1) in luminescence emission among cells exposed for several hours to the same medium. Our results also revealed heterogeneity in the time scale of activation, and fluctuations above the detection level of single cell bioluminescence over time. The high hetero geneity among cells in activation levels and

PAGE 121

121 in the activation times is likely extrinsic noise originated. On the other hand, the small single cell fluctuations over time suggest an intrinsic noise origin given their shorter ~ 20 30 minutes time scales. V.f ischeri also detects two other autoinducers in its lux regulatory network besides 3OC6HSL. The other autoinducer with most impact in bioluminescence regulation is C8HSL. We asked the question of how much the addition of C8HSL modulated the previously obser ved high heterogeneity in luminescence. We utilized for these measurements a V.fischeri fluorescent lux reporter ( V.fischeri JB10) that has a gfp insertion in its lux operon. W hen the QS network activates lux the bacterium also produces green fluorescent protein that we detected with fluorescence mi croscopy. We employed a microfluidic device that allowed us to measure three samples at the same time. We perfused each of the three samples with different combinations of exogenous 3OC6HSL and C8HSL autoinducer s while measuring single cell fluorescence. W e observed that C8HSL suppresses the average fluorescence corroborating our bulk culture characterization. Previous studies have shown that C8HSL competes with 3OC6HSL in binding to LuxR receptor. We found that C8HSL along with the suppression of the average fluorescence also suppresses the variance among cells when constant 3OC6HSL is provided We also found evidence suggesting that lux activity is insensitive to different combinations of 3OC6HSL and C8HSL th at produce the same average lux activation. The latter result suggests that the precision of the lux response at the single cell level is not improved by the addition of C8HSL. On the other hand, the C8HSL suppression possibly has a

PAGE 122

122 physiological role at l ow cell densities to avoid unnecessary light emission by very bright bacteria. As a futur e investigation avenue it would be very interesting to study in dep th the time depende n ce of the lux activity under 3OC6HSL and C8HSL. The use of microfluidics is an experimental technique that offers immense possibilities. We used basic microfluidic patterns in our research, but the door is open to use some of the more sophisticated patterns in use in the community. During our single cell measurements, V.fischeri cell s were adhered and immobilized to a glass surface treated with poly L lysine for several hours. It would be interesting to find out if the results we obtained were affected in any way by this fact. Was the experiment biased by only looking at a subpopulati on of immobilized cells that have not divided for 3 4 hours and in contact with a poly L lysine treated surface? It is unlikely but further tests are require d. Using mutant strains lacking flagella could be a viable alternative. Another possibility could b e having an experimental configuration without adhering the cells, but observing them in a flow cytometry system. Of course, there would still be side effects requiring attention. The lux activity has been connected to the cell motility, which is something to consider if using non flagellum mutants. Finally, flow cytometry would require having the cells in suspension at some point which could make difficult fixing the external conditions. However, a strong point is made if similar results are obtained with different methods. We performed some preliminary experiments shearing the bacterial flagella. This was performed by manually pumping the liquid culture back and forth between two

PAGE 123

123 syringes connected by very thin tubing ( ~ 500 m diameter). We found this procedure did not improve our cell adhesion. Furthermore, shearing the flagella puts the cells in a very high stress situation which could trigger additional responses or modify existing ones. Cell replication was an important l imiting factor during our measurements. We estimated a ~ 1.1 hours average replication time in a bulk culture. However, our experiments required observing immobilized isolated cells for ~ 3 4 hours. Furthermore, the lux response time is ~ 2 hours (albeit high ly heterogeneous). A future investigator might want to know if a significant fraction of the cells we considered had already replicated their chromosomes. A DNA dye combined with higher resolution microscopy and appropriate data analysis wou ld help in prov iding an answer. The cells replicating in the same time scale as the phenomena observed was a strong constraint. Timing all the experiment phases was critical, and having to repeat long experiments was very common due to the majority of cells dividing befo re we could acquire data. A mechanism to slow cell replication could significantly help in adding flexibility to future experiments. Some possibilities to consider are using a slow replicating mutant, an agent that inhibits DNA replication, or a modified l iquid growth medium. We observed a high heterogeneity in lux response with two reporters, luciferase bioluminescence and GFP fluorescence. However, we did not accurately identify the step with the highest contribution to noise. It is very possible that Lux R receptor is playing an important role here. An interesting experiment would be finding a correlation between LuxR copy numbers and lux activation level. A second fluorescent reporter

PAGE 124

124 gene like mCherry could be inserted under the same promoter along with the luxR gene. Measuring red mCherry fluorescence would provide an indicator of LuxR numbers, while green GFP fluorescence would report lux expression. It has been observed by other researchers that lux might be susceptible to hysteresis (82) The levels of LuxR can be different depending on the cell past history due to a LuxR autoregulation mechanism not studied in this dissertation. It would be very interesting to measure if there is a correlation between pres ent and past activations in individual cells. In such experiment, single cell lux response would be measured while cells are activated, shut down and reactivated. Performing such experiment would most likely require slowing down cell replication, as both a ctivation and deactivation are in the same time scale as cell replication. In our research we extensively utilized a wild type and a fluorescent lux reporter strain. However, the use of a wider variety of custom strains would provide a significant edge in isolating and testing other properties of the network as explained above Further studies would be needed to find out if other functions also regulated by QS are subject to the high heterogeneity we observed in lux activation. Even more, it would be very i mportant to answer whether the bacteria utilize or rely on the genetic noise for QS regulation.

PAGE 125

125 LIST OF REFERENCES 1. Waters CM & Bassler BL (2005) Quorum sensing: Cell to cell communication in bacteria Annu Rev Cell Dev Biol 21: 319 346. 2. Kaplan HB & Greenberg EP (1985) Diffusion of autoinducer is involved in regulation of the vibrio fischeri luminescence system J Bacteriol 163: 1210 1214. 3. Nealson KH, Platt T & Hastings JW (1970) Cellular control of synthesis and activity of bacter ial luminescent system J Bacteriol 104: 313 &. 4. Nealson KH & Hastings JW (1979) Bacterial bioluminescence its control and ecological significance Microbiol Rev 43: 496 518. 5. Visick KL, Foster J, Doino J, McFall Ngai M & Ruby EG (2000) Vibrio fisc heri lux genes play an important role in colonization and development of the host light organ J Bacteriol 182: 4578 4586. 6. Ruby EG & Nealson KH (1976) Symbiotic association of photobacterium fischeri with marine luminous fish monocentris japonica mod el of symbiosis based on bacterial studies Biol Bull 151: 574 586. 7. Ruby EG & Nealson KH (1978) Seasonal changes in species composition of luminous bacteria in nearshore seawater Limnol Oceanogr 23: 530 533. 8. Ruby EG, Greenberg EP & Hastings JW (19 80) Planktonic marine luminous bacteria species distribution in the water column Appl Environ Microbiol 39: 302 306. 9. Ng WL & Bassler BL (2009) Bacterial quorum sensing network architectures Annu Rev Genet 43: 197 222. 10. Chen X et al (2002) Stru ctural identification of a bacterial quorum sensing signal containing boron Nature 415: 545 549. 11. Xavier KB & Bassler BL (2003) LuxS quorum sensing: More than just a numbers game Curr Opin Microbiol 6: 191 197. 12. Lupp C & Ruby EG (2005) Vibrio fis cheri uses two quorum sensing systems for the regulation of early and late colonization factors J Bacteriol 187: 3620 3629. 13. Kolter R & Greenberg EP (2006) Microbial sciences the superficial life of microbes Nature 441: 300 302. 14. Parsek MR & Gr eenberg EP (2005) Sociomicrobiology: The connections between quorum sensing and biofilms Trends Microbiol 13: 27 33.

PAGE 126

126 15. Diggle SP, Griffin AS, Campbell GS & West SA (2007) Cooperation and conflict in quorum sensing bacterial populations Nature 450: 411 U7. 16. Dong YH et al (2001) Quenching quorum sensing dependent bacterial infection by an N acyl homoserine lactonase Nature 411: 813 817. 17. Manefield M & Turner SL (2002) Quorum sensing in context: Out of molecular biology and into microbial ecolog y Microbiology (UK) 148: 3762 3764. 18. Tarighi S & Taheri P (2011) Different aspects of bacterial communication signals World J Microbiol Biotechnol 27: 1267 1280. 19. Sifri CD (2008) Quorum sensing: Bacteria talk sense Clin Infect Dis 47: 1070 1076. 20. Atkinson S & Williams P (2009) Quorum sensing and social networking in the microbial world J R Soc Interface 6: 959 978. 21. de Kievit TR & Iglewski BH (2000) Bacterial quorum sensing in pathogenic relationships Infect Immun 68: 4839 4849. 22. An tunes LCM & Ferreira RBR (2009) Intercellular communication in bacteria Crit Rev Microbiol 35: 69 80. 23. Redfield RJ (2002) Is quorum sensing a side effect of diffusion sensing? Trends Microbiol 10: 365 370. 24. Boyer M & Wisniewski Dye F (2009) Cell cell signalling in bacteria: Not simply a matter of quorum FEMS Microbiol Ecol 70: 1 19. 25. Dunn AK & Stabb EV (2007) Beyond quorum sensing: The complexities of prokaryotic parliamentary procedures Anal Bioanal Chem 387: 391 398. 26. Hense BA et al ( 2007) Opinion does efficiency sensing unify diffusion and quorum sensing? Nat Rev Microbiol 5: 230 239. 27. Ruby EG (2008) Symbiotic conversations are revealed under genetic interrogation Nat Rev Microbiol 6: 752 762. 28. Ruby EG et al (2005) Comple te genome sequence of vibrio fischeri: A symbiotic bacterium with pathogenic congeners Proc Natl Acad Sci U S A 102: 3004 3009. 29. Lupp C, Urbanowski M, Greenberg EP & Ruby EG (2003) The vibrio fischeri quorum sensing systems ain and lux sequentially in duce luminescence gene expression and are important for persistence in the squid host Mol Microbiol 50: 319 331.

PAGE 127

127 30. Stabb EV (2005) Shedding light on the bioluminescence "paradox" ASM News 71: 223 +. 31. Fuqua C, Parsek MR & Greenberg EP (2001) Regula tion of gene expression by cell to cell communication: Acyl homoserine lactone quorum sensing Annu Rev Genet 35: 439 468. 32. Whitehead NA, Barnard AML, Slater H, Simpson NJL & Salmond GPC (2001) Quorum sensing in gram negative bacteria FEMS Microbiol R ev 25: 365 404. 33. Stabb EV, Butler MS & Adin DM (2004) Correlation between osmolarity and luminescence of symbiotic vibrio fischeri strain ES114 J Bacteriol 186: 2906 2908. 34. Dunlap PV & Greenberg EP (1991) Role of Intercellular Chemical Communicati on in the Vibrio Fischeri Monocentrid Fish Symbiosis, pp 254. 35. Engebrecht J & Silverman M (1984) Identification of genes and gene products necessary for bacterial bioluminescence Proceedings of the National Academy of Sciences of the United States of America Biological Sciences 81: 4154 4158. 36. Schaefer AL, Val DL, Hanzelka BL, Cronan JE & Greenberg EP (1996) Generation of cell to cell signals in quorum sensing: Acyl homoserine lactone synthase activity of a purified vibrio fescheri LuxI protein Pr oc Natl Acad Sci U S A 93: 9505 9509. 37. Dunlap PV (1999) Quorum regulation of luminescence in vibrio fischeri J Mol Microbiol Biotechnol 1: 5 12. 38. Engebrecht J, Nealson K & Silverman M (1983) Bacterial bioluminescence isolation and genetic analys is of functions from vibrio fischeri Cell 32: 773 781. 39. Reading NC & Sperandio V (2006) Quorum sensing: The many languages of bacteria FEMS Microbiol Lett 254: 1 11. 40. Stevens AM, Dolan KM & Greenberg EP (1994) Synergistic binding of the vibrio fi scheri luxr transcriptional activator domain and rna polymerase to the lux promoter region Proc Natl Acad Sci U S A 91: 12619 12623. 41. Visick KL & Ruby EG (2006) Vibrio fischeri and its host: It takes two to tango Curr Opin Microbiol 9: 632 638. 42. Gilson L, Kuo A & Dunlap PV (1995) Ains and a new family of autoinducer synthesis proteins J Bacteriol 177: 6946 6951. 43. Miyamoto CM, Dunlap PV, Ruby EG & Meighen EA (2003) LuxO controls luxR expression in vibrio harveyi: Evidence for a common regulato ry mechanism in vibrio Mol Microbiol 48: 537 548.

PAGE 128

128 44. von Bodman SB, Willey JM & Diggle SP (2008) Cell cell communication in bacteria: United we stand J Bacteriol 190: 4377 4391. 45. Kuo A, Callahan SM & Dunlap PV (1996) Modulation of luminescence oper on expression by N octanoyl L homoserine lactone in ainS mutants of vibrio fischeri J Bacteriol 178: 971 976. 46. Miller MB & Bassler BL (2001) Quorum sensing in bacteria Annu Rev Microbiol 55: 165 199. 47. Lupp C & Ruby EG (2004) Vibrio fischeri LuxS and AinS: Comparative study of two signal synthases J Bacteriol 186: 3873 3881. 48. Antunes LCM et al (2007) Transcriptome analysis of the vibrio fischeri LuxR LuxI regulon J Bacteriol 189: 8387 8391. 49. Graf J, Dunlap PV & Ruby EG (1994) Effect of t ransposon induced motility mutations on colonization of the host light organ by vibrio fischeri J Bacteriol 176: 6986 6991. 50. Millikan DS & Ruby EG (2002) Alterations in vibrio fischeri motility correlate with a delay in symbiosis initiation and are as sociated with additional symbiotic colonization defects Appl Environ Microbiol 68: 2519 2528. 51. Elowitz MB, Levine AJ, Siggia ED & Swain PS (2002) Stochastic gene expression in a single cell Science 297: 1183 1186. 52. Rosenfeld N, Young JW, Alon U, Swain PS & Elowitz MB (2005) Gene regulation at the single cell level Science 307: 1962 1965. 53. Spudich JL & Koshland DE (1976) Non genetic individuality chance in single cell Nature 262: 467 471. 54. McAdams HH & Arkin A (1997) Stochastic mechanis ms in gene expression Proc Natl Acad Sci U S A 94: 814 819. 55. Raser JM & O'Shea EK (2005) Noise in gene expression: Origins, consequences, and control Science 309: 2010 2013. 56. Swain PS, Elowitz MB & Siggia ED (2002) Intrinsic and extrinsic contrib utions to stochasticity in gene expression Proc Natl Acad Sci U S A 99: 12795 12800. 57. Ozbudak EM, Thattai M, Kurtser I, Grossman AD & van Oudenaarden A (2002) Regulation of noise in the expression of a single gene Nat Genet 31: 69 73.

PAGE 129

129 58. Smits WK, Kuipers OP & Veening JW (2006) Phenotypic variation in bacteria: The role of feedback regulation Nat Rev Microbiol 4: 259 271. 59. Simpson ML et al (2009) Noise in biological circuits Wiley Interdiscip Rev Nanomed Nanobiotechnol 1: 214 225. 60. Balab an NQ, Merrin J, Chait R, Kowalik L & Leibler S (2004) Bacterial persistence as a phenotypic switch Science 305: 1622 1625. 61. Maamar H, Raj A & Dubnau D (2007) Noise in gene expression determines cell fate in bacillus subtilis Science 317: 526 529. 6 2. Williams P, Winzer K, Chan WC & Camara M (2007) Look who's talking: Communication and quorum sensing in the bacterial world Philos Trans R Soc B Biol Sci 362: 1119 1134. 63. Visick KL (2005) Layers of signaling in a bacterium host association J Bacte riol 187: 3603 3606. 64. Perez PD & Hagen SJ (2010) Heterogeneous response to a quorum sensing signal in the luminescence of individual vibrio fischeri PLoS One 5: e15473. 65. Bose JL, Rosenberg CS & Stabb EV (2008) Effects of luxCDABEG induction in vib rio fischeri: Enhancement of symbiotic colonization and conditional attenuation of growth in culture Arch Microbiol 190: 169 183. 66. Haas E (1980) Bioluminescence from single bacterial cells exhibits no oscillation Biophys J 31: 301 312. 67. Sternberg C, Eberl L, Poulsen LK & Molin S (1997) Detection of bioluminescence from individual bacterial cells: A comparison of two different low light imaging systems J Biolumin Chemilumin 12: 7 13. 68. Phiefer CB, Palmer RJ & White DC (1999) Comparison of relat ive photon flux from single cells of the bioluminescent marine bacteria vibrio fischeri and vibrio harveyi using photon counting microscopy Luminescence 14: 147 +. 69. Greer LF & Szalay AA (2002) Imaging of light emission from the expression of luciferas es in living cells and organisms: A review Luminescence 17: 43 74. 70. Koch AL (1970) Turbidity measurements of bacterial cultures in some available commercial instruments Anal Biochem 38: 252 &. 71. Goutelle S et al (2008) The hill equation: A review of its capabilities in pharmacological modelling Fundam Clin Pharmacol 22: 633 648.

PAGE 130

130 72. Ulitzur S & Hastings JW (1978) Myristic acid stimulation of bacterial bioluminescence in aldehyde mutants Proc Natl Acad Sci U S A 75: 266 269. 73. Ulitzur S & Has tings JW (1979) Evidence for tetradecanal as the natural aldehyde in bacterial bioluminescence Proc Natl Acad Sci U S A 76: 265 267. 74. Bose JL et al (2007) Bioluminescence in vibrio fischeri is controlled by the redox responsive regulator ArcA Mol Mi crobiol 65: 538 553. 75. Herrero M, Delorenzo V & Timmis KN (1990) Transposon vectors containing non antibiotic resistance selection markers for cloning and stable chromosomal insertion of foreign genes in gram negative bacteria J Bacteriol 172: 6557 656 7. 76. Chalfie M, Tu Y, Euskirchen G, Ward WW & Prasher DC (1994) Green fluorescent protein as a marker for gene expression Science 263: 802 805. 77. Eberhard A (1972) Inhibition and activation of bacterial luciferase synthesis J Bacteriol 109: 1101 &. 78. Yarwood JM, Volper EM & Greenberg EP (2005) Delays in pseudomonas aeruginosa quorum controlled gene expression are conditional Proc Natl Acad Sci U S A 102: 9008 9013. 79. Cox CD et al (2003) Analysis of noise in quorum sensing. OMICS A Journal of Integrative Biology 7: 317 334. 80. Anetzberger C, Pirch T & Jung K (2009) Heterogeneity in quorum sensing regulated bioluminescence of vibrio harveyi Mol Microbiol 73: 267 277. 81. Long T et al (2009) Quantifying the integration of quorum sensing sig nals with single cell resolution PLoS Biol 7: 640 649. 82. Williams JW, Cui X, Levchenko A & Stevens AM (2008) Robust and sensitive control of a quorum sensing circuit by two interlocked feedback loops Mol Syst Biol 4: 234. 83. Hagen SJ, Son M, Weiss J T & Young JH (2010) Bacterium in a box: Sensing of quorum and environment by the LuxI/LuxR gene regulatory circuit J Biol Phys 36: 317 327. 84. Tong D et al (2009) Evidence for light perception in a bioluminescent organ Proc Natl Acad Sci U S A 106: 98 36 9841. 85. Mandel MJ, Wollenberg MS, Stabb EV, Visick KL & Ruby EG (2009) A single regulatory gene is sufficient to alter bacterial host range Nature 458: 215 U7.

PAGE 131

131 86. Goldman E & Green LH (2009) Practical handbook of microbiology, (CRC Press, Boca Rat on), pp 853. 87. Mazia D, Schatten G & Sale W (1975) Adhesion of cells to surfaces coated with polylysine applications to electron microscopy J Cell Biol 66: 198 200. 88. Parent ME, Snyder CE, Kopp ND & Velegol D (2008) Localized quorum sensing in vib rio fischeri Colloid Surf B Biointerfaces 62: 180 187. 89. Martinez P & Klotz A (1998) A Practical Guide to CCD Astronomy, (Cambridge University Press, 90. Buil C (1991) CCD astronomy : construction and use of an astronomical CCD camera, (Willmann Bell, Richmond, Va.), pp 321. 91. Shaner NC, Steinbach PA & Tsien RY (2005) A guide to choosing fluorescent proteins Nat Methods 2: 905 909. 92. Bagh S et al (2008) Plasmid borne prokaryotic gene expression: Sources of variability and quantitative system ch aracterization Phys Rev E 77: 021919. 93. Meighen EA (1991) Molecular biology of bacterial bioluminescence Microbiol Rev 55: 123 142. 94. Schwille P, Haupts U, Maiti S & Webb WW (1999) Molecular dynamics in living cells observed by fluorescence correla tion spectroscopy with one and two photon excitation Biophys J 77: 2251 2265. 95. Bao N, Jagadeesan B, Bhunia AK, Yao Y & Lu C (2008) Quantification of bacterial cells based on autofluorescence on a microfluidic platform J Chromatogr A 1181: 153 158. 96. Cai L, Friedman N & Xie XS (2006) Stochastic protein expression in individual cells at the single molecule level Nature 440: 358 362. 97. Hussa EA, Darnell CL & Visick KL (2008) RscS functions upstream of SypG to control the syp locus and biofilm for mation in vibrio fischeri J Bacteriol 190: 4576 4583. 98. Sia SK & Whitesides GM (2003) Microfluidic devices fabricated in poly(dimethylsiloxane) for biological studies Electrophoresis 24: 3563 3576. 99. Colville K, Tompkins N, Rutenberg AD & Jericho M H (2010) Effects of poly(L lysine) substrates on attached escherichia coli bacteria Langmuir 26: 2639 2644. 100. Tsien RY (1998) The green fluorescent protein Annu Rev Biochem 67: 509 544.

PAGE 132

132 BIOGRAPHICAL SKETCH Pablo was born and raised in Buenos Aires Argentina, a huge, lively and chaotic city on the shores of the Rio de la Plata (River of the Silver). In his early youth, he attended elementary and high school in the Dmaso Centeno Institute. He later studied biomedical sciences at the University of B uenos Aires (UBA). Although he was very interested in the courses he was taking, doing very well and gaining an extremely solid foundation in biological sciences, his enthusiasm for research was not being satisfied. In a pivotal decision, he left the Schoo l of Medicine and started studying physics in the School of Exact and Natural Sciences in the UBA. After several years of hard study, he gained h is Licenciado degree in physics He completed his Licenciado thesis in fMRI (functional magnetic resonance imag ing) with Dr Susana Blanco and Dr Mirta Villarreal as co advisors. In 2005 he arrived to the University of Florida to work on a Physics PhD. He worked under the supervision of Dr Stephen J. Hagen, studying noise in genetic networks. He completed his resear ch and obtained his PhD in 2011.