<%BANNER%>

Theoretical Studies on Structure-Activity Relationships of FLP Neuropeptides

Permanent Link: http://ufdc.ufl.edu/UFE0022548/00001

Material Information

Title: Theoretical Studies on Structure-Activity Relationships of FLP Neuropeptides
Physical Description: 1 online resource (141 p.)
Language: english
Creator: Leonis, Georgios
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The animal nervous system employs chemical neurotransmitters to relay messages from one neuron to the next. We focused our interest on a family of neurotransmitters called FMRFamide-like (or FLP) neuropeptides. FLPs are particularly interesting because they are involved in a wide range of neural functions, including learning, feeding processes, and pain modulation. In the first part of our study, the different potencies of two FLPs against neuropeptide receptor 1 (NPR-1) in the nematode Caenorhabditis elegans were evaluated. DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2 exhibit different activities against NPR-1, with the longer peptide presenting a lower potency. Replica-Exchange Molecular Dynamics (REMD) simulations suggested that DFDGAMPGVLRF-NH2 adopts mainly two conformations, forming either a C-terminal loop, or a bicyclic structure involving N-terminal with C-terminal-loop interactions. Conversely, EMPGVLRF-NH2 presents one dominant conformation, with the PGVLRF-NH2 region resembling the corresponding region of DFDGAMPGVLRF-NH2. Furthermore, it was concluded that when the N-terminal-associated interactions are eliminated, the activity of the peptide is enhanced, whereas the PGVLRF-NH2 motif is a unique conformation. These findings suggest that the N-terminal region behaves independently and is determinant for the differences in binding affinity. In the second part of our study, a mammalian neuropeptide involved in pain modulation was considered. REMD methodology was applied to identify structural characteristics associated with the high binding affinity of NPFF (FLFQPQRF-NH2) for a specific receptor. Backbone conformations dictated by interactions among Gln4 and the C-terminal Arg and Phe residues, are necessary for high affinity. However, these interactions alone are not sufficient for binding to the receptor; we observed that the C-terminal?Gln4 interactions enable appropriate positioning for the side chains of Phe and Arg. A free and solvent-exposed aromatic ring, along with an oriented (and positively charged) arginine side chain, are possibly the main structural units that recognize receptor binding sites. These studies provide further insight concerning structural characteristics of FLPs that are critical for binding to receptors in the animal nervous system. Although the structure-activity relationship of FLPs remains elusive, the outcomes of this investigation shed new light into this area, and will hopefully be a stepping stone towards the understanding of the pharmacological effects of FLPs.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Georgios Leonis.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Roitberg, Adrian E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022548:00001

Permanent Link: http://ufdc.ufl.edu/UFE0022548/00001

Material Information

Title: Theoretical Studies on Structure-Activity Relationships of FLP Neuropeptides
Physical Description: 1 online resource (141 p.)
Language: english
Creator: Leonis, Georgios
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

Subjects

Subjects / Keywords: Chemistry -- Dissertations, Academic -- UF
Genre: Chemistry thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The animal nervous system employs chemical neurotransmitters to relay messages from one neuron to the next. We focused our interest on a family of neurotransmitters called FMRFamide-like (or FLP) neuropeptides. FLPs are particularly interesting because they are involved in a wide range of neural functions, including learning, feeding processes, and pain modulation. In the first part of our study, the different potencies of two FLPs against neuropeptide receptor 1 (NPR-1) in the nematode Caenorhabditis elegans were evaluated. DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2 exhibit different activities against NPR-1, with the longer peptide presenting a lower potency. Replica-Exchange Molecular Dynamics (REMD) simulations suggested that DFDGAMPGVLRF-NH2 adopts mainly two conformations, forming either a C-terminal loop, or a bicyclic structure involving N-terminal with C-terminal-loop interactions. Conversely, EMPGVLRF-NH2 presents one dominant conformation, with the PGVLRF-NH2 region resembling the corresponding region of DFDGAMPGVLRF-NH2. Furthermore, it was concluded that when the N-terminal-associated interactions are eliminated, the activity of the peptide is enhanced, whereas the PGVLRF-NH2 motif is a unique conformation. These findings suggest that the N-terminal region behaves independently and is determinant for the differences in binding affinity. In the second part of our study, a mammalian neuropeptide involved in pain modulation was considered. REMD methodology was applied to identify structural characteristics associated with the high binding affinity of NPFF (FLFQPQRF-NH2) for a specific receptor. Backbone conformations dictated by interactions among Gln4 and the C-terminal Arg and Phe residues, are necessary for high affinity. However, these interactions alone are not sufficient for binding to the receptor; we observed that the C-terminal?Gln4 interactions enable appropriate positioning for the side chains of Phe and Arg. A free and solvent-exposed aromatic ring, along with an oriented (and positively charged) arginine side chain, are possibly the main structural units that recognize receptor binding sites. These studies provide further insight concerning structural characteristics of FLPs that are critical for binding to receptors in the animal nervous system. Although the structure-activity relationship of FLPs remains elusive, the outcomes of this investigation shed new light into this area, and will hopefully be a stepping stone towards the understanding of the pharmacological effects of FLPs.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Georgios Leonis.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.
Local: Adviser: Roitberg, Adrian E.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2009-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0022548:00001


This item has the following downloads:


Full Text

PAGE 1

THEORETICAL STUDIES ON STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES By GEORGIOS LEONIS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008 1

PAGE 2

2008 Georgios Leonis 2

PAGE 3

To my parents, Constantine and Stavroula 3

PAGE 4

ACKNOWLEDGMENTS This work would not have been possible without the support and guidance of several important people. I gratefully thank my mentor, Dr. Adrian Roitberg, for his continuous guidance, support and understanding. I also thank my committee members, Dr. Rodney Bartlett, Dr. Erik Deumens, Dr. Arthur Edison, and Dr. Thomas Lyons for being such insightful and helpful committee. Special thanks to Dr. Edison for his valuable advice and constructive discussions. I thank Remigio Cabbrera-Trujillo for all his help during my first years in the Quantum Theory Project (QTP), and Seonah Kim for her continuous help and for being such a great friend. Mehrnoosh Arrar and Aaron Dossey have also made a significant contribution to my project. During my time in QTP, I have made good friends all of whom I have shared great time. First, I would like to thank Andrew Taube and Julio Palma for their sincere friendship and fun times. I also give special thanks to Josh McClellan, Christina Crecca, Tom Hughes, Ozlem Demir, Dan Sindhikara, Yilin Meng, Joey Nicely, Gustavo Seabra, Dominika Zgid, and Martin Peters. I have shared many memorable moments with my good friends from Chemistry Department: I primarily thank Jorge Chavez, Daniel Kuroda, Fedra Leonik and Josh Smith. I would also like to thank Lilly Alexaki and my sisters, Joanna and Vassileia, along with my grandmother Jenny, for their love and support throughout all these years. Finally and above all, I thank Asimina Hiona for her enormous support, understanding, andmore importantlyfor her true, pure love. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES...........................................................................................................................7 LIST OF FIGURES.........................................................................................................................9 ABSTRACT...................................................................................................................................12 CHAPTER 1 INTRODUCTION..................................................................................................................14 1.1 Introduction to Protein Structure......................................................................................14 1.2 Basic Neuroscience...........................................................................................................19 1.2.1 Synaptic Transmission............................................................................................19 1.2.2 Neurotransmitters...................................................................................................22 1.2.3 FLP Neuropeptides.................................................................................................23 1.2.4 Receptors................................................................................................................29 1.3 Experimental Methods......................................................................................................37 1.3.1 Structure Determination.........................................................................................37 1.3.1.1 Nuclear magnetic resonance spectroscopy...................................................37 1.3.1.2 X-ray crystallography...................................................................................40 1.3.2 Evaluation of Binding Affinity and Biological Activity........................................42 1.4 Motivation.........................................................................................................................42 2 THEORY AND METHODOLOGY......................................................................................43 2.1 Theoretical Considerations...............................................................................................43 2.2 Mathematical Model of Replica-Exchange Molecular Dynamics Method......................43 2.3 Force Field........................................................................................................................47 2.4 Generalized Born Solvation Model..................................................................................51 2.5 Computational Schemes...................................................................................................53 2.5.1 Clustering Methodology.........................................................................................53 2.5.2 NMR Calculation....................................................................................................55 2.5.3 Solvent-Accessible Surface Area Calculations......................................................56 3 STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES: C.ELEGANS NEUROPEPTIDES.........................................................................................58 3.1 Introduction.......................................................................................................................58 3.2 Computational Methods and Systems..............................................................................62 3.2.1 Systems...................................................................................................................62 3.2.2 Computational Methods.........................................................................................63 3.3 Results and Discussion.....................................................................................................67 5

PAGE 6

3.3.1 Hydrogen Bonding and Clustering Analyses.........................................................67 3.3.2 Different Protonation States...................................................................................75 3.3.3 Study of Mutations.................................................................................................79 3.3.4 The C-terminal PGVLRF-NH2 Region...................................................................82 3.3.5 The N-terminal DFDGAM Region........................................................................85 3.4 Conclusions.......................................................................................................................89 4 STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES: NPFF NEUROPEPTIDE...................................................................................................................91 4.1 Introduction.......................................................................................................................91 4.2 Computational Methods and Systems..............................................................................93 4.2.1 Systems...................................................................................................................93 4.2.2 Computational Methods.........................................................................................94 4.3 Results and Discussion.....................................................................................................97 4.3.1 NPFF Neuropeptide................................................................................................97 4.3.2 C-terminal Substitutions: The Role of Phenylalanine..........................................102 4.3.3 C-terminal Substitutions: The Role of Arginine..................................................113 4.3.4 C-terminal Substitutions: The Role of Glutamine................................................116 4.3.5 N-terminal Substitutions.......................................................................................119 4.4 Conclusions.....................................................................................................................120 5 AN OVERALL VIEW.........................................................................................................122 APPENDIX A SUPPLEMENTAL INFORMATION ON PROJECT I.......................................................124 B SUPPLEMENTAL INFORMATION ON PROJECT II......................................................125 LIST OF REFERENCES.............................................................................................................129 BIOGRAPHICAL SKETCH.......................................................................................................141 6

PAGE 7

LIST OF TABLES Table page 3-1 Peptides considered in this study.......................................................................................63 3-2 Principal HB interactions in DFDGAMPGVLRF-NH2 (full REMD trajectory)...............68 3-3 Principal HB interactions in DFDGAMPGVLRF-NH2 (Cluster 1)..................................70 3-4 Principal HB interactions in DFDGAMPGVLRF-NH2 (Cluster 2)..................................70 3-5 Coexistence of principal HB interactions in DFDGAMPGVLRF-NH2............................72 3-6 Principal HB interactions in EMPGVLRF-NH2................................................................73 3-7 Main HB interactions in the fully protonated form and in the deprotonated form of DFDGAMPGVLRF-NH2..................................................................................................76 3-8 Principal HB interactions for DFD(H)GAMPGVLRF-NH2, the partially protonated form of peptide III..............................................................................................................77 3-9 Comparison of principal HB interactions in DFDGEMPGVLRF-NH2 and its protonated form, DFD(H)GE(H)MPGVLRF-NH2............................................................79 3-10 HB populations for DFDGAMPGVLRF-NH2 and its mutant, SGSGAMPGVLRF-NH2....................................................................................................................................80 3-11 HB patterns for DFDGAMPGVLRF-NH2 and DFDGEMPGVLRF-NH2........................82 3-12 Summary of the principal HB interactions (peptides I, III, IV, VII and X), in decreasing intensity............................................................................................................83 3-13 Amide and alpha proton chemical shift calculations for DFDGAMPGVLRF-NH2 and DFDGAM-NH2...........................................................................................................86 3-14 Significant HB interactions of the 2-6 amide protons within the DFDGAM region of A: DFDGAM-NH2 peptide and B: DFDGAMPGVLRF-NH2 peptide..............................87 4-1 Peptides considered in this study.......................................................................................94 4-2 Principal HB interactions in NPFF (full REMD trajectory)..............................................97 4-3 Comparison of HB patterns obtained from implicit-solvent and hybrid-REMD simulations for NPFF.........................................................................................................99 4-4 Principal HB interactions in NPFF (Cluster 2)................................................................101 4-5 Principal HB interactions in NPFF (Cluster 3)................................................................101 7

PAGE 8

4-6 Principal HB interactions in peptide II (full trajectory)...................................................103 4-7 Principal HB interactions in peptide II (Clusters 1, 2, and 3)..........................................105 4-8 Principal HB interactions in peptide III (full trajectory).................................................113 4-9 Principal HB interactions in peptide IV (full trajectory).................................................117 4-10 Principal HB interactions in peptide V (full trajectory)...................................................118 4-11 Principal HB interactions in peptide VI (full trajectory).................................................119 4-12 Principal HB interactions in peptide VII (full trajectory)................................................120 A-1 Principal HB interactions in D(H)FDGAMPGVLRF-NH2, the partially protonated form of peptide III............................................................................................................124 B-1 Principle HB interactions in NPFF (Cluster 1)................................................................125 B-2 Principal HB interactions in peptide IV (full trajectory).................................................126 B-3 Principle HB interactions in peptide VI (Cluster 1).........................................................127 B-4 Principle HB interactions in peptide VI (Cluster 2).........................................................127 B-5 Principle HB interactions in peptide VI (Cluster 3).........................................................127 B-6 Principle HB interactions in peptide VII (Cluster 1).......................................................128 B-7 Principle HB interactions in peptide VI (Cluster 2).........................................................128 8

PAGE 9

LIST OF FIGURES Figure page 1-1 The general form of an amino acid....................................................................................14 1-2 Ionization states of an amino acid are pH dependent........................................................15 1-3 The two configurations for the optical isomers L and D in amino acids...........................15 1-4 The twenty amino acids in proteins...................................................................................16 1-5 A tripeptide unit in a peptide chain....................................................................................17 1-6 Levels of structure in protein architecture.........................................................................18 1-7 The structure of a neuron...................................................................................................19 1-8 A typical chemical synapse................................................................................................21 1-9 Neuropeptide processing in the flp-16 gene of C. elegans................................................26 1-10 Putative FLP neuropeptides in C. elegans.........................................................................27 1-10 Continued...........................................................................................................................28 1-11 A G protein-coupled receptor............................................................................................30 1-12 Ligand-receptor binding curves for ligands A and B, with kdA
PAGE 10

3-4 The two prevalent configurations for DFDGAMPGVLRF-NH2 peptide: representative structures for A) Cluster 1 and B) Cluster 2...............................................70 3-5 Backbone dihedral angles for each residue in EMPGVLRF-NH2 as a function of the simulation time...................................................................................................................73 3-6 Ramachandran plots for each residue in EMPGVLRF-NH2.............................................74 3-7 Principal HB interactions for EMPGVLRF-NH2 peptide..................................................74 3-8 Principal HB interactions in A) DFDGAMPGVLRF-NH2 (peptide III) and B) its mutant, SGSGAMPGVLRF-NH2 (peptide VI).................................................................80 3-9 Ramachandran plots for the common MPGVLRF C-terminal region in DFDGAMPGVLRF-NH2 (red) and EMPGVLRF-NH2 (black)........................................84 3-10 Representative conformations for A) DFDGAMPGVLRF-NH2 and B) EMPGVLRF-NH2....................................................................................................................................85 3-11 Amide proton NMR chemical shift as a function of residue number for DFDGAM-NH2 (blue curve) and DFDGAMPGVLRF-NH2 (purple curve)........................................86 3-12 Alpha proton NMR chemical shift as a function of residue number for DFDGAM-NH2 (blue curve) and DFDGAMPGVLRF-NH2 (purple curve).......................................87 3-13 Residue 2 amide proton interactions in DFDGAM-NH2...............................................88 3-14 Residue 2 amide proton interactions in DFDGAMPGVLRF-NH2................................88 4-1 Backbone dihedral angles for each residue in NPFF as a function of the simulation time....................................................................................................................................98 4-2 Ramachandran plots for each residue in NPFF..................................................................98 4-3 Ramachandran plots about Gln-4 for the three NPFF clusters........................................100 4-4 The three prevalent configurations for NPFF: representative structures for A) Cluster 1, B) Cluster 2 and C) Cluster 3.......................................................................................102 4-5 Backbone dihedral angles for each residue in peptide II as a function of the simulation time.................................................................................................................103 4-6 Ramachandran plots for each residue in peptide II..........................................................104 4-7 The three prevalent configurations for peptide II: representative structures for A) Cluster 1, B) Cluster 2 and C) Cluster 3..........................................................................106 4-8 Arginine side chain interactions in A) NPFF and B) peptide II......................................107 10

PAGE 11

4-9 SASA histograms for the arginine residue in NPFF (red) and peptide II (black)............108 4-10 The aromatic ring A) in NPFF and B) in peptide II.........................................................109 4-11 Relative-SASA calculations for residues 4-8 in NPFF....................................................110 4-12 SASA histograms for the phenylalanine residue in NPFF (black) and for the tyrosine residue in peptide II (red).................................................................................................112 4-13 SASA histograms for the phenylalanine residue in Cluster 1 and Cluster 3 of A) NPFF and B) peptide II....................................................................................................112 4-14 Backbone dihedral angles for each residue in peptide III as a function of the simulation time.................................................................................................................114 4-15 Ramachandran plots for each residue in peptide III........................................................114 4-16 Amino acids arginine and lysine......................................................................................116 11

PAGE 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 THEORETICAL STUDIES ON STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES By Georgios Leonis December 2008 Chair: Adrian E. Roitberg Major: Chemistry The animal nervous system employs chemical neurotransmitters to relay messages from one neuron to the next. We focused our interest on a family of neurotransmitters called FMRFamide-like (or FLP) neuropeptides. FLPs are particularly interesting because they are involved in a wide range of neural functions, including learning, feeding processes, and pain modulation. In the first part of our study, the different potencies of two FLPs against neuropeptide receptor 1 (NPR-1) in the nematode Caenorhabditis elegans were evaluated. DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2 exhibit different activities against NPR-1, with the longer peptide presenting a lower potency. Replica-Exchange Molecular Dynamics (REMD) simulations suggested that DFDGAMPGVLRF-NH2 adopts mainly two conformations, forming either a C-terminal loop, or a bicyclic structure involving N-terminal with C-terminal-loop interactions. Conversely, EMPGVLRF-NH2 presents one dominant conformation, with the PGVLRF-NH2 region resembling the corresponding region of DFDGAMPGVLRF-NH2. Furthermore, it was concluded that when the N-terminal-associated interactions are eliminated, the activity of the peptide is enhanced, whereas the PGVLRF-NH2 motif is a unique 12

PAGE 13

conformation. These findings suggest that the N-terminal region behaves independently and is determinant for the differences in binding affinity. In the second part of our study, a mammalian neuropeptide involved in pain modulation was considered. REMD methodology was applied to identify structural characteristics associated with the high binding affinity of NPFF (FLFQPQRF-NH2) for a specific receptor. Backbone conformations dictated by interactions among Gln4 and the C-terminal Arg and Phe residues, are necessary for high affinity. However, these interactions alone are not sufficient for binding to the receptor; we observed that the C-terminalGln4 interactions enable appropriate positioning for the side chains of Phe and Arg. A free and solvent-exposed aromatic ring, along with an oriented (and positively charged) arginine side chain, are possibly the main structural units that recognize receptor binding sites. These studies provide further insight concerning structural characteristics of FLPs that are critical for binding to receptors in the animal nervous system. Although the structure-activity relationship of FLPs remains elusive, the outcomes of this investigation shed new light into this area, and will hopefully be a stepping stone towards the understanding of the pharmacological effects of FLPs. 13

PAGE 14

CHAPTER 1 INTRODUCTION 1.1 Introduction to Protein Structure Proteins participate in almost all biological processes with a great variety of functions: enzymatic catalysis, carriage and storage of several small molecules, mechanical support, combined movement, creation of neural pulses and immune protection are some examples that demonstrate the importance of proteins. The building blocks of proteins are called amino acids. A naturally occurring -amino acid consists of one amino group, one carboxyl group, one hydrogen atom and a distinctive R group (side chain) attached to the -carbon atom (Figure 1-1). igure 1-1. The general form of an amino acid. The central carbon atom is attached to a hydrogen alpha atom, a side chain, an amino group and a carboxyl group. t physiological pH the amino acids are mainly zwitterions with the amino group being protonated (NH3+), whereas the carboxyl group is negatively charged (COO-). The degree of ionization varies with pH (Figure 1-2). AAmmiinnoo ggrroouupp NNHH22 F A CCaa HH CCaarrbboonn aallpphhaa aattoomm ((CCaa)) SSiiddee cchhaaiinn RR CCOOOOHH CCaarrbbooxxyyll ggrroouupp 14

PAGE 15

igure 1-2. Ionis the /aa-neutral-ph.jpg F zation states of an amino acid are pH dependent. In acidic solutioncarboxylate is not charged, whereas the amino group is protonated. The opposite happens in basic solutions. [Adapted from: http://dbs.umt.edu/courses/fall2006/bioc380/lectures/008/images ibutes optical actid D isomers) are displayed in Figure 1-3.t only the L configuration and they are all consist of various combinations of only 20 amino acids. Amino acids are classified by the chemical propel s. Figure 1-3. The two configurations for the optical isomers L and D in amino acids. In proteins, only the L configuration exists. Last accessed September, 2008]. The tetrahedral structure that four different groups acquire around an -carbon attrvity to the amino acids. The two possibilities (L an It is interesting that proteins adop rties their side chains possess (Figure 1-4). The variety in shape, size, charge and chemicareactivity of the side chains, is the main reason for the great functional variety of the protein CCaa HH NNHH++ 33 CCOOOO-RR LL--iissoommeerr RR CCaaHH NNHH++ 33 CCOOOO-DD--iissoommeerr 15

PAGE 16

Nonpolaramino acids +H3NC H COO CH3 Ala,Alanine(A)H3CCH CH3 C +H3N COO H Val,Valine(V)H3C CH CH 3 CH 2 C +H3N COOH Leu,Leucine(L)CH2C H3 HC CH 3 C +H3N COOH Ile,Isoleucine(I)HCH2CHNCH2 2CH Pro,Proline(P)C COOH +H3N CH2 C +H 3N COOH Phe,PhenylalanineCH3S (F) CH2 CH2 C COOH3C H Met,Meth ionine(M) HN CH2 C COOH +H3N Trp,Tryptophan(W) SHCH2 C H +HN 3 COOCys,+H3NC Cysteine(C) COO H H Gly,Glycine(G)ChargedpolaraminoacidsOC OCH2 C COOH +H3N Asp,Asparticacid(D)OC OCH2 CH2 C +H3N COOH Glu,Glutamicacid(E)NH3+CH2 CH2 CH2 CH2 C +H3N COOH Lys,Lysine(K)NH2C NH CH2 CH2 CH2 C NH2+ COO+H3N H Arg,Arginine(R)NH+HN CH2 C COOH +H3N His,Histidine(H)UnchargedpolaraminoacidsOC NH2 CH2 C COO+H3N H Asn,Asparagine(N)CH2C COO+H3N H CH2 C NH2 O Gln,Glutamine(Q)OHCH2 C COO+H3N H Ser,Serine(S)CCOOH +H3N CH HO CH3 Thr,Threonine(T) CH2 C COOH +H3N Tyr,Tyrosine(Y)OH Figure 1-4. The twenty amino acids in proteins. [Adapted from Kim, S. 2007. Simulating temperature jumps for protein folding studies. Ph.D. dissertation (Page 15, Figure 1-2). University of Florida, Gainesville, Florida]. 16

PAGE 17

17 In proteins, the -carboxyl group of an amino acid is bonded to the -amino group of another amino acid to form a peptide bond (or amide bond). Several amino acids can be joined together with amide bonds to form peptide chains of different lengths (Figure 1-5). Small peptides (less than 20 amino acids) are called oligopeptides and longer chains could be considered as polypeptides. Proteins usually are long amino acid chains. An amino acid group in a peptide is called a residue. Figure 1-5. A tripeptide unit in a peptide chain. The unit starts at the N-terminal and ends at the C-terminal. [Adapted from Kim, S. 2007. Simulating temperature jumps for protein folding studies. Ph.D. dissertation (Page 16, Figure 1-4). University of Florida, Gainesville, Florida]. Four structural levels are related to a proteins architecture. The primary structure is the amino acid sequence and the position of disulfide bonds, if any. Thus, the primary structure is a escription of the covalent bonds of the protein (Figure 1-6A). The secondary structure is the The most common secondary structural elemlix and the-sheet (Figure 1-6B). NN--tteerrmmiinnaall CC--tteerrmmiinnaall PPeeppttiiddee bboonndd d regularly repeating configurations of the amino acids, generally stabilized by hydrogen bonds. ents are the -he

PAGE 18

The tn re. move disulfide bonds. Quaternary structure is the result of the interaction among polypeptide chains (Figure 1-6D). Figure 1-6. Levels of structure in protein architecture. A) Primary structure. B) Secondary ertiary structure describes the relationship among regions with different secondary structure (Figure 1-6C). Thus, the tertiary structure provides the overall shape of a single proteimolecule. Additionally, many nonlocal interactions are involved to stabilize such a structuFor example, in order to minimize their exposure to water, hydrophobic residues tend totowards the proteins core. Other important factors that stabilize the structure include salt bridges, hydrogen bonds, and TTyyrr GGllnn AArrggPPrroo AA aa--hheelliixx --sshheeeett BB DD CC structure. C) Tertiary structure. D) Quaternary structure. 18

PAGE 19

1.2 Basic Neuroscience In order to understand the biological basis of consciousness and the mental processewhich we behave, perceive, learn and remember, the ultimate challenge would be the unification of the study of behaviorscience of the mindand neural science, the science of the brain. Such a unified scientific approach is particularly difficult to be realized due to the fact that our behavior is determined by the complex interactions of more than 100 billion individual nercells in the human brain.1 Nevertheless, one step towards that direction would be the s by ve investigation of how neurons are organized into signaling pathways in the brain and how they he rele coity ofn behavior is a result of the function of nerve cells and the interactionm. A typical neuron is showe Figure 1-7. The structure of a neuron. [Adapted from: http://www.morphonix.com/software/education/science/brain/game/specimens/image communicate with each other, by means of synaptic transmission. 1.2.1 Synaptic Transmission The nervous system employs two types of cells: glial cells (support cells) and nerve cells (neuro ns). T markab mplex huma s among the n in Figur 1-7. s/neuron_parts.gif Last accessed September, 2008]. 19

PAGE 20

It consists of four regions: (1) The cbolic center of the neuron and is the pl n ms-1. re of events. The most on that was revealed after investigating the role of action poten unicate information about vision. Another notable feature is that action potentials within a neuron flow only in one direction, namely from the cell body to the end of the axon (principle of dynamic polarization). (4) The last part of the cell is the presynaptic terminal, which transmits the signal to the dendrite (postsynaptic terminal) of another neuron. The site where electrical or chemical transmission occurs is called synapse (Figure 1-8). Two important characteristics concerning synapses are that: a) the presynaptic terminal is not in contact with the postsynaptic terminal but they are separated by a space, the synaptic cleft, and b) each presynaptic terminal makes specific in postsynaptic terminals to form definite networks (principle of ell body serves as the meta ace where proteins are formed. (2) Several short dendritesat the exterior of the cell bodyare the sites where incoming signals arrive from other neurons. (3) The axon which, also extends away from the cell body, carries electrical signals (action potentials) to other neuronsalong a great range of distances, from 0.1mm to 3m. The amplitude of action potentials is 100 mV and they last approximately 1 ms.1 After an action potential is initiated at the trigger regio(axon hillock) in the origin of the axon, it travels down the axon with a speed of 1-100Action potentials are the signals by which the brain draws and analyzes information and they ahighly stereotyped, even though they are initiated by a great variety important aspect of brain functi tials, is that the key factor for the information conveyed is the pathway the signal follows inthe brain, and not the form of the signalwhich is always the same.1 Thus for example, the signals that communicate information about pain are identical to those that comm connections with certa connectional specificity). 20

PAGE 21

by the synaptic cleft. [Adapted from: http://webspace.ship.edu/cgboer/synapse Figure 1-8. A typical chemical synapse. The presynaptic and postsynaptic terminals are separated .gif Last accessed September, 2008]. An average nerve cell forms about 1000 synaptic connections and receives up to ten times more inputs. Synapses can be either electrical (where instantaneous signal transmission is provided), or chemical. Figure 1-8 shows a chemical synapse. In this case, when an action potential arrives at the presynaptic terminal of the neuron, it stimulates the release of specific chemical transmitters from the cell. These chemical transmitters are stored in organelles called synaptic vesicles, which are found in specific release sites (active zones) of the presynaptic terminal. The transmitter is then released from the presynaptic terminal and diffuses across the synaptic cleft to binding receptor sites in the postsynaptic terminal. The binding of the transmitter to the receptor results in the generation of a new potential on the next neuron. The importance of transmitters and receptors becomes evident, since they are the key elements for neuronal communication. 21

PAGE 22

Transmitters can be either small-molecule substances, such as acetylcholine, biogenic amine transmitters, amino acid transmitters, ATP and adenosine, or neuroactive peptides (neuropeptides).1,2 Neuropeptides are short polymers of amino acids and along with their receptors are the main focus of this study. 1.2.2 Neurotransmitters A neurotransmitter can be defined as a substance that is released in a synapse by one neuron and that interacts with a postsynaptic cell (either an effector organ or neuron), in a specific manner.1 However, since it is difficult to demonstrate a transmitter function (for a ar synapse, we probably cannot arrive at a comprehensive and precise ical substances must fulfill the following focriteria to be considered transme synthesized in the neuron, 2) They are found in the presynaptic terminal and are released in amounts sufficient to produce a defined action on a postsynaptic cell, 3) When exogenously administered, they mimic the action of the endogenously released transmitter exactly, and 4) They are removed from the synaptic cleft via a specific mechanism. A plethora of short neuropeptides are pharmacologically active in appropriate target neurons by causing excitation, inhibition, or both. Some of these peptides act as neurotransmitters, whereas others have been identified as hormones, or as neuroendocrine-secretion products (oxytocin). Neuroactive peptides serving as neurotransmitters are derived from secretory proteins, which are formed in the cell body of the neuron. The study of neuropeptides is particularly important because some are involved in regulating emotions and sensory perception. For example, some peptides (NPFF neuropeptide) are located in specific regions of the central nervous system and they are implicated in pain modulation;7 others substance) at a particul definition of a neurotransmitter. In general, chemur itters: 1) They ar 1,4,531,2,6 22

PAGE 23

mo polyprotein. One single precursor protein produces ecific proteolytic cleavages, sometimes amplifying the production of one po se re vesicle makes this process extraordinarily complex.1 ilies uropeptide nd y t-Arg-Phe-NH2), because most of them share similar amino acid sequence with it (and they all present dulate the feeding behavior of nematodes (DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2, in C. elegans).8 Usually, the production of neuropeptides occurs after a single continuous mRNA is translated to one large precursor protein, or several neuropeptides by sp articular peptide (yielding more than one copy). The processing of precursor proteins tneuropeptides occurs in vesicles. An important aspect on the processing of precursor proteins is that if a particular geneencodes a polyprotein in different neurons, the resulting neuropeptides may be different becauof the different ways each neuron processes the polyprotein.1 Neuropeptides do not transport anyunique information (as DNA does, for example), they rather convey the information to the receptors by binding and causing them to reshape. Once this happens, the receptor produces an electrical signal. A single synaptic action involves the simultaneous release of several neuropeptides towards specific receptors. The fact that about 5000 neuropeptide molecules astored in each Although the diversity of neuropeptides is prodigious, they could be classified in famwith their members sharing similar amino acid sequences. Among many different nefamilies that have been defined thus far, we are particularly interested in the opioid family amore specifically in a category of peptides called FMRFamide-like neuropeptides. 1.2.3 FLP Neuropeptides FMRFamide-like (or FLPs)9-16 neuropeptides belong to the opioid peptide family and theare short polymers of amino acids derived from secretory proteins that are formed in the cell body of the neurons.3 They are named after their parent peptide FMRFamide17 (Phe-Me 23

PAGE 24

the RF-NH2 sequence at the C-terminus). FMRFamide was discovered in 1977 as a cardioactive peptide extracted from the clam Macrocallista nimbosa.17 tide genomes has provided valuable information abouting id ilies:31 r large non polar groups (valin, t position -rse correlation between the populations of reverse turns for a series of FL FLPs are present in every organism studied to-date10 (they form the largest neuropepfamily in invertebrates) and they have been implicated in a variety of neural activities such as cardioregulation, muscle control, learning and feeding behavior in nematodes,17-26 as well as sensory modulation in mammals.27-30 The DNA sequence determination of different the diversity of numerous neuropeptides present in the animal kingdom. A very interestfeature of FLPs is that they present similar conserved C-terminal amino acid sequences, with a decreased amino acid conservation pattern towards the N-terminal. Since all FLPs share the RF-NH2 C-terminal sequence, they can be classified in subfamilies based on the third amino acfrom the C-terminal (position -3). Thus, high frequencies of appearance have been observed for four amino acids at the position before the arginine, yielding the following four subfamLRF-NH2 MRF-NH2 IRF-NH2 and GRF-NH2. Further moving towards the N-terminal, position -4 is mainly occupied by aromatic (mostly phenylalanine) o e, leucine); glutamic acid also occurs frequently. At position -5, even though the amino acid diversity becomes greater, we observe a preference towards polar (charged or uncharged) residues such as Asn, Asp, Pro, Gly, Glu, Ser, and Gln. Edison et al., and Wilmot and Thorntonhave recognized the importance of reverse turns due to the presence of Pro and Gly a 5.32,33 For example, the inve Ps and their receptor binding affinities, indicates the role of N-terminal amino acid diversity in FLPs:32 a multitude of conformations for neuropeptides in the unbound state may result in a variety of different binding affinities to the receptors. 24

PAGE 25

The fascinating amino acid diversity in FLPs has motivated us to investigate the functional role of these peptides. For instance, the soil nematode Caen orhabditis elegans expresses relativf t d or they are identical to neuropeptides of other parasitic nema ely simple patterns of behavior, even though it contains a very large and diverse number oneuropeptide genes. To date, 109 neuropeptide genes have been identified in C. elegans, with aleast 30 being flp genes.34 A very important observation is that neuropeptide sequences isolatefrom C. elegans present high similarity todes.15,35 This may indicate that these neuropeptides participate in particular processes, by functioning similarly across some (or all) organisms. As previously discussed, large precursor proteins are cleaved to yield specific neuropeptides. Proper processing of the precursors to active peptides is ensured via the secretorypathway.3,36 As an example, in Figure 1-9 we describe the neuropeptide processing in the C. elegans gene flp-16. The processing sites in C. elegans consist of dibasic (KR, RK, RR, KK), basic (R or K), or tribasic residues.37-42 The basic site cleavages are realized by kex2/subtilisin-like proprotein convertases (PC).34 The precursor proteins are further processed by carboxypeptidases E (CPEs), which remove the basic residues from the peptide sequence.43 Finally, a glycine at the C-terminal of a peptide results in amidation, by donating an amino group. It has been suggested that amidation happens to generate an active form of the peptide, and/or to protect against degradation.43,44 Even though the C.elegans amidation involves unknown enzymes, homology studies have shown that C. elegans contains at least one PAM-likemolecule.45 25

PAGE 26

Gene flp-16 MNFSGFEFSSIVAFFLLILQLSTAAVLPADYAYGVADEMSALPDSGSLFAEQRPSKRAQTFVRFGKRAQTFVRFGKRGQTFVRFGRSAPFEQ 2x AQTFVRFGKRGQTFVRFGR 2x AQTFVRFGGQTFVRFG2x AQTFVRF-NH2 GQTFVRF-NH2 FLP NeuropeptidesPCCPE PAM Figurpeptides are produced from a large precursor protein (two copies are obtained for AQTFVRF-e e resulted in re in 1o-dimensional nanoscale liquid chromatography in tandem with mass pectrometry.38 Since the truncated peptide was not predicted based on the precursor processing scheme we followed, we can assume that the peptide is either processed via a new cleavage site, or it is a degradation product.34 In any case, the indication that Figure 1-10 may not be complete, is a strong possibility. e 1-9. Neuropeptide processing in the flp-16 gene of C. elegans. Two FLP neuroNH2 peptide). To date, at least 30 flp genes have been identified in C. elegans,15,46 and in Figure 1-10 wpresent putative peptide sequences for these genes. Predicted peptides are identified based on thcharacteristics of neuropeptide precursor processing as described in Figure 1-9. This85 different FLP neuropeptides in C. elegans. Peptides that are not experimentally isolated aparentheses. It is important to mention that the number of predicted neuropeptides in Figure10 is most likely an underestimate of the total number of FLPs in C. elegans. This conclusion issupported by the following example: Husson et al. have isolated the truncated peptide GAMPGVLRF-NH2 (along with the longer predicted peptide, DFDGAMPGVLRF-NH2), performing tw s 26

PAGE 27

AGSDPNFLRFaSQPNFLRFaASGDPNFLRFaSDPNFLRFaAAADPNFLRFaSADPNFLRFaPNFLRFaLRGEPIRFa(GEPIRFa)SPREPIRFaSPLGTMRFaTPLGTMRFaSAEPFGTMRFaNPENDTPFGTMRFaASEDALFGTMRFaEAEEPLGTMRFaSADDSAPFGTMRFaNPLGTMRFaPTFIRFaASPSFIRFa APKPKFIRFaAGAKFIRFaGAKFIRFa KSAYMRFa flp-2flp-1flp-3fl flpgenesFLP neuropeptides flp-4 p-6flp-5 TPMQRSSMVRFa (SSMVRFa)SPMQRSSMVRFaSPMERSAMVRFaKNEFIRFaKPSFVRFa(PSFVRFa)(SGYIRFa) (ARSGYIRFa)AMRNALVRFa(NALVRFa)ASGGMRNALVRFaRNKFEFIRFaSDRPTRAMDSPLIAMDSPLIRFaAPEASPFIRFaASPSAPLIRFaSPSAVPLIRFaASSAPLIRFaflp-8flp-7flp-10flp-9flp-13flp-12flp-11 SPMDRSKMVRFaQPKARSGYIRFaNGAPQPFVRFa(FEFIRFa)RFaAADGAPLIRFaSAAAPLIRFa KHEYLRFa flp-14entally isolated. (HEYLRFa) Figure 1-10. Putative FLP neuropeptides in C. elegans. The different lengths of FLP precursor proteins are normalized to the same length. Sections with same color in the precursor denotes multiple copies of the same peptide, and shorter (truncated) peptides are color matched to the corresponding longer peptides. Predicted peptides in parentheses have not been experim 27

PAGE 28

GG PQGPLRFaAFVRFa)QYIRFaDGAMPGVLRFANQVRFa RGPSGPLRFa(GPSGPLRFa)AQTFVRFaGQTFVRFaKSAFVRFa(SKS(SQYIRFa)DFaGAMPGVLRFaEMPGVLRFaSVEIPGVLRFaSEVPGVLRFaDVPGVLRFaSYVLRFaWASWASSVRFaAMMRFaGLGPRPLRFaSPSAKWMRFaVVGQQDFLRFaflp-23flp-22flp-21flp-20flp-19flp-18flp-17flp-16flp-15 PGVLRFa FDEKKSVPG EFNADDLTLRFaFNAD DLTLRFaGEPLAFSPDMLSL GGARFa VPSAGDMMVRFaaaaVLMRFaFa(PRFa)flp-33flp-28flp-27flp-26flp-25flp-24 DYDFVRFASYDYIRFGKGLGGRMRFAMRNSLVR(NSLVRFa) flp-32 Figure 1-10. Continued. diversity eand mammals are vC. The intriguing part of FLP neuropeptides is that they present an enormous amino acid xpressed by the utilization of similar motifs repeatedly. Furthermore, other nematodes ery likely to share the neuropeptide diversity (and function) present in 28

PAGE 29

elegans. To date, the challenging task of determining the functional roles of individual FLPs is far from complete. Future research on important aspects such as the matching of receptors with specific peptide ligands, the enigma of the amino acid patterns in FLPs, and their roles in the nervous system, can lead to a profound understanding of neural functions. In this study, we investigate the structural properties of DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2 in the worm C. elegans, as well as of NPFF, a mammalian neuropeptide participating in pain regulation. 1.2.4 Receptors The two main functions of synaptic receptors include the recognitionific neurotransmitters and the activation of effectors by altering the membrane potential of the cell. Two major types of synaptic receptors can be identified: (1) Ionotropic rece47ichdirectly open ion channels, and (2) Metabotropic receptors,48 where ion channels are gated indirectly only. The latter receptor type can be divided into two families: the G protein-coupled receptors (GPCRs) and the receptor tyrosine kinases.49 The G protein-coupled receptors50-55 are coupled to an effector via a guanine nucleotide-binding protein (G protein). After being activated by the action of neuropeptides, the GPCR shifts conformation, causing activation and detachment of the G protein. At that point the receptor either returns to its previous inactive state or activates another G prhe G protein is responsible for further effects such as, activation of second-messenger (for example cyclic alteration of ion channel activity. These second-messenger actions last from seconds to minutes and are implicated in emotional states, memory and learning.1 Members of thein of specptors, whotein. T AMP) cascades or direct GPCR family include and adrenergic receptors,56 the GABAB receptors,57 serotonreceptors,58 receptors for neuropeptides, such as NPR-1 receptor,8 as well as NPFF1 and NPFF2 receptors.7 29

PAGE 30

A typical G protein-coupled receptor is shown in Figure 1-11. GPCRs are integral membrane proteins with seven characteristic membrane-spanning regions. The C-terminal regionof the protein is found at the cytoplasmic region of the cell, whereas the N-terminal is at the extracellular area. The receptor structure is stabilized by disulfide bonds form ed in the extrac007, cture (2-adrenergic receptor) was solved,60 followed by a much 61,62 ellular loops (by two cysteine residues). In 2000, the crystal structure for the bovine rhodopsin was solved,59 and this was the only mammalian GPCR structure analyzed before 2where the first human GPCR stru higher resolution. Figure 1-11. A G protein-coupled receptor. NN CC CCyyttooppllaassmm EExxttrraacceelllluullaarr TTrraannssmm eemmbbrraannee DDoommaaiinn SSppaaccee 30

PAGE 31

In addition to the aforementioned opioid receptors, a novel opioid-like orphan recepbeen discovered. Its ligand is a 17 amino acid neuropeptide called nociceptin63 (or orphanin FQ). The nociceptin receptor is found extensively in the nervous system, and the neuropeptide appto participate in many behavioral functions, as well as in the regulation of nociception (response to tissue damage). It has been known that the brain contains specific receptors for opiates.64 The three major classes are and An interesting observation is that there is a high correlation between the potency of an analgesi tor has ears the c substance (such as morphine) and its affinity to the receptor.1 Endogenous opioid peptides that interact with these receptors have been identified and include enkephalins, -endorphin and dynorphins. Mammalian NPFF is a morphine modulating (1-1) where R is the free (unbound) receptor and LR is the bound ligand-receptor complex. The dissociation constant Kd = koff/kon determines the fraction of receptors in each state, at equilibrium. Although Kd is temperature dependent, it can be considered constant for any particular receptor-ligand pair, since threlatively constant. The relationship between bound and unbound receptor is given by Eq. 1-2 neuropeptide (acts as an endogenous anti-opioid peptide in rodents), which also induces opioid-like effects. In Chapter 4 we investigate the structural characteristics of NPFF that are responsible for high affinity to a specific receptor. The interaction between a neuropeptide (ligand, L) and a receptor R can be described by Eq. 1-1 LRRLonoffkk e body temperature is ][]][[LRRLKd (1-2) 31

PAGE 32

whereuropeptide (ligand) effect can result from an increase in the conce [L] is the free ligand concentration, [R] is the unbound receptor concentration and [LR] is the concentration of the ligand-receptor complex. Since Kd is constant, we observe that the bound receptor concentration increases with an increase in the ligand concentration. Moreover, as free receptor concentration is increased, bound receptor concentration increases as well. Therefore, an increase in a ne ntration of either the receptor or the ligand. If we assume that the total receptor concentration R0 is constant, we can write dKRLRLRRR]][[][][][][ Solving Eq. 1-3 for [R] and substituting to Eq. 1-2, we obtain 0 (1-3) dKL LRLR][][][][ (1-4 0) The binding of two different ligands to the same receptor (as described by the semilogarithmic form of Eq. 1-4) is shown in Figure 1-12. The sigmoidal plots are known as ligand-receptor binding curves. The two ligand-receptor interactions are described by different Kd values, with KdA
PAGE 33

0 K[L] K 1.0 Ligand B Ligand A 0.5 [LR]/R0 dA dB Figure 1-12. Ligand-receptor binding curves for ligands A and B, with kdA
PAGE 34

produced by the l igand. Efficacy can be considered as the state where additional ligand will not produce additional response. E is a measurable reample, a decrease in blood pressure).Thus, in Figure 1-13, ligand A is more potent than B, even though A and B have the same efficacy. This indicates that a high potency is not necessarily accompanied by a high efficacy for a given ligand (drug). Figure 1-13. Dose-response curves for ligands A and B. The efft a ligd as aunctioconcentration is demonstrated. to sponse to a drug (for ex 0 1.0 E/Emax [L] Ligand B Ligand A 0.5 EC50[A] EC50[B] ecofan fn of its It is usually assumed that a receptor is in a conformational equilibrium between an active and an inactive state. Based on the ability to control the existence of the receptor in each conformation, ligands can be classified as agonists or antagonists. A ligand that, after bindingthe receptor stabilizes one specific conformation (usually the active) is called an agonist, 34

PAGE 35

whereas a ligand acts as an antagonist if it hinders activation of the receptor by preventing the agonist binding to the receptor. An extended form of Eq. 1-1 is a useful approximation which describes the relationship between agonist binding and activation of the receptor (Eq. 1-6) *DRDRRLkkon (1 k koff-6) here LR* is the activated state of the receptor, k is the rate constant for the activation of the receptor, k is the rate constant which corresponds to the receptor deactivation. Eq. 1-6 provides a relationship between potency (Kd=koff/kon) and agonist binding (L+RLR), as well as efficacy (k/k) and the conformational changes associatedtor (LRLR*). The relationship between the affinity of a peptide for its receptor and the efficacy to produce an effect is still unclear.65 The term efficacy was first introduced by Stephenson to describe the ability of an agonist to produce a response.66 Another empirical method which described the relative ability of agonists to cause a response for a particular receptor occupancy (intrinsic efficacy) was later proposed by Furchgott.67 Receptor models based on linkage theory provide a more explicit description of efficacy by linking the protein molecules through energy conversions that are equal with respect to the ach a specific final point (based on the first law of thermodynamics).68,69 In such a case, the energy required to get from the couple to Gof the G protein subunits, which in turn activate 70-72 The ability of ligands to produce such an effect is a representation of efficacy. w with the activation of the recep path follo wed to re reactants to the products is independent of the path followed. As previously mentioned, GPCRs proteins and induce separation effectors. Thus, efficacy can be considered as a function of the entire GPCR system, which includes the receptor and its cellular environment. Technological advances have enabled us to observe a broad array of GPCR behaviors, which include pleiotropic coupling to different G proteins, oligomerization, desensitization, 35

PAGE 36

internalization, and interaction with other proteins.65,73 Research on these behaviors has shown that some peptides can induce some of these effects but not others, and that a molecule can havefficacy by altering the behavior of the receptor. e Peptides change receptor behavior via selective affinity for the various cothe receptor. The ligand binds most strongly to the receptor state for which it has the highest 22resulting in receptoDD)cholecystokinin-26-32]-phenethyl ester does not result in t s r nformations of affinity, and produces a bias towards an active receptor state and a subsequent physiological response. This demonstrates that affinity is not simply the ligand-receptor binding, but a dynamic process that results in a change in receptor conformation. Such conformational changes have been observed in the binding of -adrenoceptor antagonists to -adrenoceptors, without r activation.74 This suggests that the relationship between affinity and efficacy is very compelling, since efficacy is the result of the ligands action to the receptor. As already mentioned, it has also been observed that some ligands can directly produce several secondary effects (dimerization, internalization, desensitization), without producing a response (as required by Stephensons definition for efficacy). On the other hand, there are ligands which produce a response, but not other secondary effects. For instance, even though GPCR internalization is related to receptor activation, the cholecystokinin (CCK)-receptor antagonist -Tyr-Gly-[(Ile28,31,-Trp30receptor activation, but it produces receptor internalization.75 The chemokine peptide RANTES produces receptor activation as well as internalization,76 while a RANTES analogue does noproduce receptor stimulation, but induces a fast receptor internalization.77,78 All these peptidepresent no efficacy concerning the production of a response, but they have efficacy for receptointernalization. 36

PAGE 37

The historical separation of efficacy from affinity has led to the misconception that affinitis simply the binding of a ligand to the receptor. Since now affinity has been related to other GPCR effects, it would be very interesting to test y ligands (with known affinities) for specific activion Thse lecule and UV spectroscopy provides information about the functional groups, NMR spectroscopy is a complementary technique, which reveals a map of the carbon skeleton along with the hydrogen atoms, in an organic molecule. NMR spectroscopy is extensively used ties. Another important example that connects the efficacy with the affinity is the phenomenof inverse agonism.65 In systems where receptors are spontaneously present in their active state,there are peptides that preferentially change the state of the receptor into the inactive. Thus, the constitutive activity of the system is reversed and efficacy is not observed even though it happens. 1.3 Experimental Methods 1.3.1 Structure Determination ere is a vast array of physical methods for investigating the structures of molecules, with each technique having its strengths and weaknesses. Mass spectrometry, Ultraviolet-Visible (UV/Vis) spectroscopy, Infrared (IR) and Circular Dichroism (CD) spectroscopies, Nuclear Magnetic Resonance (NMR) and X-ray crystallography are only some typical examples of themethods. In the present study, we briefly describe the last two, because of their extensive applicability in the study of proteins, their relevance to our systems and their historical importance. 1.3.1.1 Nuclear magnetic resonance spectroscopy Nuclear Magnetic Resonance spectroscopy (NMR) is the most useful spectroscopic technique for structure determination. While mass spectroscopy helps identify the molecular formula of a mo 37

PAGE 38

to stuhe molecular structure of solids and to investigate molecular dynamics in solutions, respeolecule is placed in a magnetic field, its electrons create their own small magnetic fields by circulatin. These local magnetic ed field, so as the real (effective) magnetic field B at the nu dy chemical structure using simple one-dimensional techniques, whereas two-dimensional methodologies are employed to determine chemical structures of more complex molecules.79 In addition, solid state and time domain NMR spectroscopic techniques have been developed to determine t ctively.80,81 Nuclear magnetic resonance occurs when the magnetic moment of a nucleus is aligned to a static magnetic field and then exposed to a second oscillating magnetic field. The most commonly used nuclei are 1H and 13C, even though isotopes of other atoms (14N, 15N, 19F among others) could also be measured. When a m g around the direction of the applied field fields oppose the externally appli cleus is slightly less than the applied field B0 by a fraction )1(0 BB (Thus, we could say that the nuclei are shielded from the complete effect of the applied field dto the surrounding electrons. On the other hand, there are cases, such as the benzene molecule, where the circulation of the electrons in the aromatic orbitals creates a magnetic field, wenhances the applied field, resulting in the deshielding of the nucleus. The che 1-7) ue hich mical shift phenomenon is based on the fact that the effective field at each eld would be different in each case due to the different electron density around0 nucleus in a molecule will vary, since the opposing fi each nucleus. Two different types of nuclei may have different resonance frequencies, depending on the strength of the applied magnetic field. The greater the value of B, the greater the frequency difference. Thus, NMR spectra obtained by different field strengths, would be very difficult to compare. The term chemical shift was 38

PAGE 39

introduced to overcome this problem. The chemical shift of a nucleus is defined as the difference between the resonance frequency of the nucleus and a reference frequency, relative tothe reference. 610)(REFREF (1-8) The reference frequency is usually tetramethylsilane [TMS, Si(CH3)4], and the chemical shift fTMS is defined to be zero. Chemical shift is reported in parts per million (ppm). One interesting aspect of NMR sp or ectroscopy is that the spins of nuclei generate small magnns ure ions are observable if the distance between non-equivalent nuclei is less than82close to each other, but they do not have to belong to neighboring residues in the protein sequence. The intensity of the NOE is proportional to the inverse sixth power of the distance small distances of approximately 2-5 the following procedure: (1) Sample preparation (the protein solution), (2) Obtain the NMR signals by measuring the electromagnetic radiation emitted after perturbing the equilibrium spin atoms in the moleculeknowledge of protein sequence is required, (4) Consider possible NOEs, chemical shifts, or J-coupling effects to obtain conformational information, (5) Construct the etic fields, which in turn interact with neighboring nuclei. These spin-spin interactiobetween two nuclei can be realized either through-space or through-bond. The former interactions formulate the basis for the Nuclear Overhauser Effect (NOE), which helps measthe distance between hydrogen nuclei, whereas the latter interact or equal to three bond lengths and they are called spin-spin coupling or J-coupling. Similarly, the nuclei involved in NOE observation are between the two interacting nuclei, something that restricts the NOE observation to relatively NMR experiments for structure determination can be performed in solution according to states of the nuclei, (3) Analyze the NMR results by assigning the signals to the corresponding 39

PAGE 40

three-dimensional structure of the molecule. It is important to note that only average strucare obtained since the NMR measurements provide a range of possible values. Thus, an ensemble of str tures uctures is generated instead of one unique structure assignment. 1.3.1.2 X-ray crystallography The understanding of protein structure has been enriched with the applications of X-ray act that the tomic m f the tain ow X-ray beam impinges the crystal. Part of the beam t crystallography, a technique which reveals the detailed position of most of the atomsand therefore the bond lengths and anglesin a protein. This method is based on the f aspacings in crystals are of the same order of magnitude as the wavelength of X-rays, namely of the order of 1. The crystal acts as a three-dimensional diffraction grating to a beaof X-rays, and the resulting diffraction pattern can be analyzed to obtain the positions oatoms in the crystal with a high precision (a few tens of pm). The first step towards the structure determination with X-ray crystallography is to obcrystals of the protein in interest. Crystals are usually formed by adding ammonium sulfate or other salts into a highly concentrated solution of the protein, thus reducing its solubility. Crystallization is often the most difficult part of the procedure and it is approached with insistence and patience. The three components of an X-ray crystallographic analysis are: an X-ray source, the crystallized protein and a detector. A narr is scattered by the atoms (primarily through their electrons) of the crystal. The scattered beam can be detected on a photographic film, or by an electronic detector. The technique is based in the following principles: (1) The electrons scatter the X-rays (2) The scattered waves add constructively in few directions, even though they cancel each other ouin most directions (destructive interference). The few directions where constructive interference occurs are determined by Braggs law (Eq. 1-9 and Figure 1-14). 40

PAGE 41

nd sin2 (1-9where n can be any integer and is the wavelength of the X-ray. ) Figure 1-14. Bragg reflection from a set of crystal planes with a spacing d. [Adapted from: http://upload.wikimedia.org/wikipedia/commons/0/0a/Bragg_diffraction.png Last accessed September, 2008]. (3) The way scattered waves are combined depends only on the arrangement of the atoms in the crystal. The protein crystal is placed in a capillary tube with a particular orientation with respect to the X-ray beam and the film. The rotation of the crystal results in an X-ray picture (set of diffraction patterns) consisting of a regular array of spots (reflections). The intensities of the spots on the diffraction pattern are the basic experimental data of an X-ray crystallographic analysis. 41

PAGE 42

1.3.2 uropeptide activity is to nthesize the peptide. Standard solid-phase methods are commonly used for that purpose.7 ext, the peptide is purified, usually by high-performance liquid chromatography (HPLC) and verified by electro-spray mass spectrometry7 or by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry.8 To investigate the binding to the receptor, the peptide is usually iodinated with 125I. Then, autoradiography can be used to assess binding of radioactive ligands on tissue sections. Another technique for binding assays is rapid filtration. The final step would be the evaluation of the degree of activation of the receptor; in the case of GPCR, activation can be determined by measuring intracellural radioactive [3H]cAMP, usually by means of chromatography. 1.4 Motivation present study, the relationship between the structure of FLPs and their activagainst specific receptors in the brain is explored. We employ computational methodologies to identify important FLP configurations that are responsible for particular neuropeptide-receptor FLPs is the main goal of our research that will hopefully contribute to Evaluation of Binding Affinity and Biological Activity The first step towards the experimental evaluation of the ne sy N In theity interactions in nematode and mammalian brains. The characterization of specific structure-activity relationships of the understanding of many biological processes. 42

PAGE 43

CHAPTER 2 THEORY AND METHODOLOGY 2.1 Theoretical Considerations Sampling the conformational space of complex systems such as proteins and peptides is a difficult problem in theoretical chemistry; at low temperatures, regular molecular dynamics (MD) simulations tend to get trapped in local-minimum-energy states, separated by high energy barriers. Thus, the probability of finding the global minimum greatly diminishes. One way to overcome the so-called quasi-ergodicity problem is to perform the simulation in a generalized ensemble. Such an ensemble is based on a non-Boltzmann probability distribution, so that a random walk in potential energy space is realized. A multitude of generalized-ensemble methodologies has been developed: the multicanonical algorithm (MUCA, also referred as adaptive umbrella sampling86 or entropic sampling), the simulated tempering87,88 (ST, or expanded ensemble), the 1/k-sampling,89 a generalized-ensemble algorithm based on the Tsallis weight factor, and replica-exchange method (REM) are some examples of well-known generalized-ensemble techniques. One of the most reliable computational methodologies to study complex systems such as proteins and peptides is Replica-Exchange Molecular Dynamics (REMD). REMD has been primarily used for biological systems, but its usefulness covers a broad range of applications, such as crystal structures identification, polymeric systems,95 spin glasses,96-98 quantum level systems,99,100 and general optimization problems101-103. 2.2 Mathematical Model of Replica-Exchange Molecular Dynamics Method In Replica-Exchange Method104,105 (REM, also called parallel tempering104 or multiple Markov chain method106), a number of non-interacting copies (replicas) are simulated independently and simultaneously at different temperatures by the conventional molecular 8384,85 9091-9394, 43

PAGE 44

dynamics (MD) or Monte-Carlo (MC) menged between different replicas every few steps with a specified transition probability that is defined by a usual Metropo107protein systems104,105,108-112 and to clustering studies in quantum chemistry.99 In this work, we replica exchange molecular dynamics105ting copies (or replicas) is considered at M different temperatures, T thods. Conformations are excha lis criterion. The weight factor is obtained as the product of Boltzmann factors. This exchange of replicas realizes a random walk in temperature space, which in turn leads to a random walk in potential energy space. As a consequence, REM has been widely applied to employ the REM algorithm, modified to be combined with molecular dynamics, the so-called (REMD). In REM, an artificial system of M non-interac m (m=1, 2,, M). The state of the generalized ensemble is defined as )()1( iii ip iq ,...,Mixx with Mii mqpx, where represent momenta and coordinates, respectively, for replica i at temperature m. Since the replicas are non-interacting, the weighfactor (W) for the state X is obtained as the product of Boltzmann factors for each replica or temperature, (Eq. 2-1): t (2-1) MiiiimREMqpHXW1][][)(),(exp)( where TkB with kB the Boltzmann constant and the ),(][][iiqpH the sum of kinand potential energy. If we attempt to exchange tem 1etic peratures between the ith and jth replicas at temperatures Tm and Tn, r]nmnimxxXxxX (2-2) espectively, the new state of the system becomes: ,...),...,(...,,...),...,(..., ''][]['][[ijjThe detailed balance condition needed to converge to an equilibrium ensemble is: 44

PAGE 45

)()()()('''XXwXWXXwXWREMREM (2-3) where )'(XXw is the transition probability from state X to X and )(XWREM is the wfactor of the state X. From Eq. 2-1, 2-2, and 2-3, the exchange probability (P) is obtained: eight eeeeeXWXXwPjijijjiijiijjjiiEEEEEEEEEEREMREM))(('')()( (2where ))((jijiEE eeXWXXwjiij')()(-4) and E is the pote ntial energy for each replica. Thus, we obtain: the acceptance probability of replica exchange, P(accept), by using a Metropolis criterion eeacceptPjiEEji,1min,1min)())(( (2-EM sime is reathe following steps (Figure 2-1): (1) Each replica (corresponding to a fixed temperature) is simulated simultaneously and independently for a certain number of MD steps. (2) Periodically, replicas with neighboring temperatures are exchan 5) The general Rulation schemlized in ged with acceptance probability, P(accept) from Eq. 2-5. (3) The process is repeated. Exchanges are allowed only between adjacent replicas (or temperatures), since the acceptio of thefactors increases. In Figure 2-1 we describe an REMD simulation, by illustrating the replica exchange between different tem tance ra exchange decreases exponentially as the difference between the two peratures. 45

PAGE 46

he generalized-ensemble algorithms, the determination of the weight factors is a timee al-utational limitation: as the system size (N) increases, the required number of replicas increases according to the square root of the number of degrees of freedom.113-115 Thus, for large proteins or for systems that represent solvent molecules explicitly, a lot of computer power is demanded. The computational cost can be reduced significantly if the explicit solvent molecules are neglected by using continuum solvent models that provide an estimation of the free energy of solvation. The application of continuum solvent TT22 TT33 TT44 Figure 2-1. Replica-Exchange Molecular Dynamics. In most of t -consuming procedure. REM surpasses these methods since the weight factor can bobtained beforehand (Eq.2-1). This greatly decreases the possibility of getting trapped in a locminimum energy state. However, REM presents also a comp TT11 TT TTiimmee RReepplliiccaa 44 RReepplliiccaa 33 RReepplliiccaa 22 RReepplliiccaa 11 EExxcchhaannggee aacccceepptteedd EExxcchhaannggee rreejjeecctteed d 46

PAGE 47

models is very efficient for large systems, keeping thuired relatively small. To incorporate explicit solvation effects into implicit solvent models, several himplicit-explicit models have been developed.116-120 In hybrid solvent REMD, each replica is sim explicit solvent, but the exchange probabilities are calculated in a differpredetermined closest-solvent shell is kept, while the remaining solvent molecules are trarily represented by a continuum solvent description. After the exchange probbeen calculated using the hybrid model, the original solvent molecules are restored and the sation proceeds in a fully explicit solvation manner. Thus, the exchange probabicalculated based on a drastically reduced system size, and the number of replicas needed decreases significantly. In hybrid-REMD121 the hybrid solvent model is used only to calculate the exchange proly solvated systems are used for the replica exchange 2.3 Force Field terand intramolecular forces) within a system. An important feature of thene of which was also used in the present study. AMBER uses more specific atom types than other e number of replicas req ybrid ulated in ent way: a empo ability has imul lities are bability, whereas explicit simulation. Molecular mechanics methodologies have been developed to perform simulations on systems with large number of atoms by calculating the energy as a function of the nuclear positions only (electronic motions are ignored). The force field is based on a simple model of the molecular interactions (in force field is that it uses a set of fitting parameters tested on a small number of cases and successfully applied to a much greater number of systems. A variety of different empirical force fields has been developed over the years, with AMBER,122 GROMACS,123 GROMOS,124 ENZYMIX,125 OPLS,126,127 and CHARMM128 being the most representative examples. Othe most popular force fields for exploring biomolecular systems is the AMBER force field, 47

PAGE 48

force fields which were developed for calculations on small molecules. For example, the carbon atom in a benzene ring is different from the carbon atom assigned in a five-membered ring. When p roperly parameterized, AMBER has also many applications in small molecules and polymcids in f side /mol and 3.55 kJ/mol, respectively. Another related study obtained an 11.3 kJ/mol rms difference employing the GROMOS et al. have reported calculated solvad model e with ers. It produces relatively accurate results for solvation free energies, vibrational frequencies, and conformational energies. Gu et al. have performed molecular dynamics simulations in explicit solvent (with the AMBER99 force field) using multiconfiguration thermodynamic integration129 to calculate the solvation free energies of neutral amino awater and in chloroform.130 Solvation free-energy changes and transfer free energies for transferring the 20 amino acids from chloroform (hydrophobic solution) to water were in agreement with experimental data, with root-mean-square (rms) differences that do not exceed5.1 kJ/mol and 6.4 kJ/mol, respectively. Previous studies for the hydration free energies ochain analogues by Shirts et al. produced similar results, with 5.65 kJ/mol rms difference compared to experimental values;131 the corresponding differences for CHARMM and OPLS were 5.48 kJ 96 force field.132 Jayaram tion free energies for a multitude of molecules, chosen as prototypes of nucleic acid anprotein constituents.133 Using a parametrized generalized Born (see next section) solvation under AMBER, they obtained a mean unsigned error of less than 3 kJ/mol. Even though the error is larger on individual basis, the solvation energies were in remarkable agreement with the experimental values. Li et al. used the AMBER99 force field to calculate the normal modes and the very far IR absorption spectra of the double stranded RNA chain poly(rG)-poly(rC) and the DNA chain poly(dA)-poly(dT).134 Their absorption results for the RNA homopolymer agrethe measured absorption spectra; moreover, the majority of the experimentally obtained 48

PAGE 49

resonance peaks have been reproduced by the modeling of the DNA homopolymer. This example illustrates the suitability of AMBER force field in the prediction and analysis of the light absorption signatures of biological molecules. Like CHARMM which is also widely usmodeling proteins and nucleic acids, AMBER was parameterized by experimental data. Evethough CHARMM performs well over a range of applications, it describes hydrogen bonmainly through van der Waals and electrostatic terms, whereas AMBER includes hydrogebonding as a separate term. In the AMBER model, the potential energy function U(R) includes the stretching of bondthe bending of angles, rotations about single bonds and non-bonded interactions (van derand electrostatic interactions), along with the corresponding fitting parameters for each ter2-6) ed in n ding n s, Waals m (Eq. ticelectrostaqqWaalsdervanRBRAVanglesKbondsrrKRUatomsjiatomsjiijijdihedralsneqeqr612222)()()( The first two terms of Eq. 2-6 account for the deviation of bonds and angles away from their equilibrium states req and eq respectively. Kr and K are the force constants. The third considers rotations about single bonds, with Vn indicating the relative barriers to rotation, n beingthe multiplicity, is the phase factor and is the torsion angle. The last two terms include the van der Waals (Lennard-Jones function) interactions and the electrostatic potential. Aij is a Rsdihedralnjiijijijanglesbonds])cos[1( (2-6) term 49

PAGE 50

parameter associated with the collision diameter and Bij with the well depth. More sophisforce fields may include additional terms such as polarization models or replacing the LenJones 6-12 potential with another expression which accounts for hydrogen bonding interactionThe force fields used in molecular modeling should be considered as single entities, consisting of two parts: the functional form and the empirical parameters. Two force fields may have the same functional forms but different parameters. The assignment of appropriate parameters to successfully describe a variety of systems is a complicated and time-consumintask, yet a crucial one. Gener ticated nard-s. g ally, parameters are obtained by fitting to experimental data or to high lstatic parameters are considered using a restrained electrostatic potential fit model (RESP). Molecular dynamics simulations for biological systems have been performed with these force fields for a long time and the results are usually in agreement with experimental observations. In an attempt to decide which force field should be used for protein structure predictions (without using information from a database), the research group of Okamoto applied replica-exchange methodology to an -helical peptide and a -hairpin peptide, in explicit water or six different force fields).135 They showed that for the -helical peptide, AMBER99 and CHA evel quantum mechanical calculations. As an example, Kr, K, and Vn are obtained after the combination of experimental data with ab initio calculations; the Lennard-Jones parameters Aij and Bij are determined from thermodynamic properties of various pure liquids and the electro (f RMM22 produced the most accurate results, whereas AMBER94 and GROMOS96 were not consistent with experiments. In the case of -hairpin peptide, the best results were obtained by OPLS-AA/L and GROMOS96; on the other hand, AMBER99 and CHARMM22 appeared to disagree. These findings suggest that the choice of a force field for protein folding simulations 50

PAGE 51

(starting from a random initial conformation) should be made based on the forms of secondstructure (, etc.) of each target protein. Recently, the conformational evolution and structural properties of insulin have been studied in an effort to compare AMBER03, CHARMM27, OPLS-AA, and GROMOS 43A1 force fields.136 It was concluded that different force fields favor ary differsulin nd l rrent ticated ill be developed, with the increase in computer performance. accuracy rst ent structural motifs, with CHARMM and GROMOS representing the dynamics of inmore successfully. Both force fields produced similar structural trends, such as the formation of some biologically crucial states. AMBER satisfied most of the NMR distance restraints, aproduced well conserved helical regions. However, this over-stabilization of -helix induced a loss in flexibility of the system, which resulted in an inadequate sampling of the conformationaspace. This is a well-known limitation of older AMBER-type force fields. In order to overcome these deficiencies, new parameters (obtained from high-level ab initio calculations) were introduced.137 Finally, it is important to mention that due to computational limitations, all cuforce fields sacrifice accuracy for computational efficiency. It is expected that more sophismodels w 2.4 Generalized Born Solvation Model The use of individual solvent molecules in MD simulations provides high-levelbut is of great computational cost. Alternatively, implicit solvation (continuum solvation) methods that represent the solvent as a continuous medium have been introduced. Such a representation of the solvent increases the computational speed and provides a more accurate statistical average, since the sampling of solvent conformations is omitted. Historically, the fiimplicit solvation models were based on solvent accessible surface areas (SASA) or on the Poisson-Boltzmann (PB) equation, which describes the electrostatic environment of a solute in aionic solvent environment. 51

PAGE 52

The generalized Born equation represents the electrostatic contribution to the free energof solvation. The total electrostatic free energy of a system consisting of N particles with chargeqi and radii ai is considered as the sum of the Coulomb energy and the Born free energy of solvation: y s NiiiNijijjNiaqrqqG12111121 (2-7) where is the relative permittivity of the medium. If we express the first term in Eq. 2-7 as the sum of a Coulomb interaction in vacuo and in (1-1/), then we have: ielec NiiiNijijjiNiNijijjiNielecqqqqqG12111111111 (2-The difference (Gelec) between Gelec and the Coulomb energy in vacuo is the generalized Born equation: arr28) NiNjiNelecqqqG11111 (2-9) iiijijiar12112The generalized Born equation is commonly used in molecular dynamics, as well as in semi-empirical quantum mechanical calculations. In these cases, Eq. 2-9 is usually expressed in a single term form: jielecarfG,1121 (2-10) NjijijNiqq11Where f(rij,aij) is a function of the distance between particles rij and the Born radii ai. The functional form of f can be expressed as: )4/(,)(,)(,222222ijijjiijDijijijijrDwhereerarf ( 112-11) 52

PAGE 53

The qffective Born radius i is usually calculated by a pairwise approximation as: uantity ai is known as the effective Born radius and it describes the degree of burial of each atom. It has dimensions of length and it can be considered as the distance between the center of an atom and the molecular surface. The e ijjijiiirrg),,,(11 (2-12) where i is an intrinsic radius for atom i, and g is a positive function which depends on the tions is a GB model which incorporates the hydrophobic SASA term and it is called generalized Borntsurface area (GB/SA). Although this approach has been very successful in identifying the native in larger systems have been observed. 2.5 Computational Schemes In this chapter, the basic methodologies used in the present study are described. A generic introduction on the principles of each techniquealong with the related computational detailsare provided. 2.5.1 Clustering Methodology Clustering is a data analysis method which, when apidentifies homogeneous subsets as defined by a particular measure of similarity.138 Molecular dynamics simulations yield trajectories which contain tremendous amount of data. In that case, clustering becomes very useful because we can group all molecular configurations into subsets, thus useful information is gathered, and at the same time the amount of dat be analyzed drastically decreases. positions and radii of the atoms. One of the most widely used continuum solvent model combina /solven accessible states of small proteins and oligopeptides, some limitations such as -helix and salt bridges over-stabilization plied to a set of heterogeneous objects, a tofurther 53

PAGE 54

The general clustering procedure involves the following steps: (1) Identify specific descriptors for each object (molecule, for example) in the data set. (2) Select a model or measureof similarity. (3) Apply an appropriat e technique to cluster the data set. (4) Analyze the results. similar conformations together; the conformations belonging to oneach ingle clustering algorithm which successfully groups every molec among others) methods. A typical example of hierarchical clustering is show Clustering the molecular conformations from a molecular dynamics trajectory groups e cluster are more similar to other than to conformations from other clusters. One important feature of clustering methodologies is that there is no s ular configuration set. Namely, the appropriate algorithm for clustering a particular set of data, depends on the data. The most commonly used clustering algorithms for chemical applications are divided in hierarchical methods (agglomerative or divisive) and nonhierarchical (single-pass, nearest neighbor, relocation, mixture model and topographic, n in Figure (2-2). Figure 2-2. Demonstration of typical hierarchical clustering. [Adapted from: http://images.wikia.com/psychology/images/a/ab/Hierarchical_clustering_diagram.pn g Last accessed September, 2008]. A set of six objects (a-f) are to be assigned in clusters based on the distances between them. Initially, each object is assigned to its own cluster. We then define a cutoff of 0.5 mm as the 54

PAGE 55

criters of cluste e mple, a hole, it is often useful to select an object as a representative of the cluste ion for two objects to belong to the same cluster. The criterion is satisfied for the pairb,c and d,e. Thus, we reduce the number of clusters to four. If we relax our constraint, say a 0.5cm cutoff, then the number of clusters reduces to three, since object f has merged in the (d,e) r, satisfying the broader condition. By increasing the cutoff even more, we end up withonly one cluster that includes every object. The selection of the appropriate cutoff depends on thspecific problem we address, or the specific metric of similarity we want to apply. For exaif we want to group several protein configurations based on their backbone structure similarityC-RMSD cutoff of 1.5 would be a good choice. If we want to have a general idea about the properties of a cluster as a w r. For a cluster of s compounds, the vector of the cluster centroid, x(c), is defined as metic mean of the cluster members vectors, and it is used to represent the cluster as a whole. Another useful concept is the square-error, e2, defined as the sum of squared Euclidian distances to the centroid (2-14) 2.5.2 NMR Calculation R chemical shifts proeful information about the molecular les. Calculating the chemicalterpretation of structural information on macromolecules. Empirical methods,139-142 semi-empirical models143-146 and ab initio quantum srrxscx1)(1)( (2-13) Where x(r) is the rth vector. The centroid is the arith 212)()(srcxrxe As stated before, NMvide us structure because they are sensitive to local variations in bond lengths and ang shift is, therefore, important for in approaches have been used to compute chemical shifts. 55

PAGE 56

An equation for the proton chemical shift is generally described in terms of various contributions: miscsideeHBringtorrctotal (2-15)where rc is the random coil chemical shift value of a residue, tor is the backbone torsional contribution, ring the ring current contribution, HB the hydrogen bond contribution, e the electric field contribution, side the side chain torsional contribution, and misc other chemical shiftcontributions such as solvent, temperature, and covalent bond geometries. This empirical model was developed and parameterized to experimental chemical shift data through literature search. However, it simply provides an empirical knowledge of chemical shift propensities, rather than a unique and quantitative description of proton chemical shifts (H). 1Quantum chemical shift calculations were performed to increase the accuracy of the previous emrrent and electrostatic The area over which contact between a biological molecule and the solvent can occur is called Solvent-Accessible Surface Area (SASA). SASA was first described in 1971 by Lee and antifying hydrophobic burial and it is calculated by employing the rollian e ius ere hich is approximately the value for the radius of a water molecule. It is important to mention that the SASA is different than the pirical models, by including ring cueffects, structural dependence of magnetic anisotropy, and close contact contributions. As a result, a new empiricalmodel was developed by combining the empirical parameters with quantum calculations. 2.5.3 Solvent-Accessible Surface Area Calculations Richards147 as a way of qu ng ball algorithm.148 This method uses a sphere with a particular radius to roll over the vder Waals surface of the protein (Figure 2-3A). The SASA is determined by the set of points thcenter of the probe sphere covers. The sphere represents the solvent molecule and its radaffects the observed surface area, since using a larger radius the sphfeels less surface details. A typical radius for SASA calculations is 1.4, w 56

PAGE 57

PPrroobbee sspphheerree vvaann ddeerr WWaaaallss vvaann ddeerr WWaaaallss AAccccee((ttrraaccee ssuurrffaaccee PPrroobbee sspphheerree ssuurrffaaccee molece ity in bulk solvent, called solvent-excluded surface. Figur: ular surface (Figure 2-3B). SASA is also related to its inverse effect, namely a surfacdescribing a cav ssssiibbllee ssuurrffaaccee ooff pprroobbee cceenntteerr)) MMoolleeccuullaarr ssuurrffaaccee e 2-3. Rolling ball algorithm for calculation of the A) solvent-accessible surface area (SASA) which should not be confused with B) the molecular surface. [Adapted fromhttp://www.ccp4.ac.uk/newsletters/newsletter38/03_surfarea.html Last accessed September, 2008]. AA BB 57

PAGE 58

58 STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES: C.ELEGANS 3.1 Introduction Section 1.2, we described that chemical transmission at synapses proceeds via four main steps:149 (1) The synthesis of a transmitter molecule, (2) the storage and release of the r, (3) the transmitters inostsynaptic terminal, and (4s removal from the synaptic cleft. The majority of substances classified as transmitters belongs to the class of neuroactive peptides, even though a few small-molecule chemical messengers, such as acetylcholine, ATP and adeso serve as transmitters.1,2 A plethora of short neuropeptides is pharmacologically active in nerve cells. These peptides cause excitation or en applied to appropriate target neurons.4,5 The study of neuroactive eptides is of great interest because they are involved in regulating sensory perception and motions.1,2,6 Although the diversity of neuroactive peptides is immense, as a class these hemical messengers share common cell biology. A startling generality is that neuroactive ilies whose members have similar sequences of amino acid residues. FMRterminal Rsequence and are present in every animal studied thus far. Their functional roles include learning, pain modulation and involvement in feeding behavior, among others (Section 1.2.3). A very important aspect is that they usually have highly mobile structures and this renders them as a very interesting system to be tested theoretically, since experimental-only studies can provide average properties, over time and over many molecules. In order to elaborate into the hidden diversity of form these systems present, computational approaches become pivotal.31,32,150 CHAPTER 3 NEUROPEPTIDES In 1, transm itte teractio n in the p ) the tran smitter nosine, can al i nhibition, or both, wh p e c peptides are grouped in fam Famide-like (or FLPs)9-16 neuropeptides are opioid peptides that share a common C-F-NH2 amino acid

PAGE 59

DFDGAMPGVLRF-NH2 (Asp-Phe-Asp-Gly-Ala-Met-Pro-Gly-Val-Leu-Arg-Phe-NH2) and EMPGVLRF-NH2 (Glu-Met-Pro--NH2) come from Caenorhabditis elegans,11,34,151,152 a soil nematode, abouhich lives in a temperate soil enviro2 e 1-PGsent Gly-Val-Leu-Arg-Phe t 1mm in length, w nment (Figure 3-1). Approximately 80% of all animals are nematodes, with 20,000 species described so far.153,154 C.elegans has a very simple nervous system composed of only 30neurons. It consists of less than 1000 cells and its entire cell lineage and anatomy are known.155 However, information about its behavior or chemical control is far from complete. The nervous system of C. elegans contains at least 30 FLP genes, which encode at least 85 possible FLPs, characterized by several bioinformatics and mass spectrometry approaches.15,34,38,156 In C. elegans, the flp-18 gene encodes one large protein precursor or polyprotein, flp-18 (see Figur10); several FLPs peptides are produced by the single flp-18 protein, by specific proteolytic cleavages that are catalyzed by proteases.157,158 Two of the produced FLPs are DFDGAMVLRF-NH2 and EMPGVLRF-NH2, which will be the focal point of our preresearch work. Figure 3-1. Caenorhabditis elegans. [Adapted from: http://www.space.gc.ca/asc/img/celegans_esa.jpg and http://i43.photobucket.com/albums/e353/lizng20/celgansgfp.g if Last accessed September, 2008]. 59

PAGE 60

In recent years, a shift in interest focus from FLP ligands towards their receptors biologyhas been observed, with several attempts to match receptors with their cognate ligands (deorphanization process).151 FLPs are associated with two main types of receptors: 1) protein-coupled receptors (GPCRs)50-55 are the principal class of receptors related to FLPs, with eleven C. elegans GPCRs identified thus far,14 and 2) an FMRF-NH2 gated sodium channe161 Neuropeptide receptor 1 (NPR-1), the first-discovered GPCR that modulates feeding behavior in C. elegans, exists in two forms that differ by a single amino acid at position 215;55the NPR-1 form with valine at position 215, impels the C. elegans worms to present a solifeeding behavior, whereas if a phenylalanine is present at position 215, the worms tend to aggregate during feeding (social feeding behavior).162 In C. elegans, NPR-1 is activated byfamilies of FLPs: a) the AF9 neuropeptide (GLGPRPLRF-NH2), encoded by thb) a set of PGVLRF-NH2 C-terminus ending neuropeptides, encoded by the flp-18 gene.51,55 AlFLP-18 neurope G l.159-,162 tary two e flp-21 gene and l ptides are derived from one single precursor protein, and there is a strong indication that they are released simultaneously.11 Previous studies have shown that the most potent (the one that activates the NPR-1 receptor the most) peptide is EMPGVLRF-NH2, whereas DFDGAMPGVLRF-NH2 activates NPR-1 significantly less.8,55 The structural properties of these two peptides have been experimentally investigated by the research groups of Edison and Evans using NMR chemical shift analyses.8 The results suggested signie two terminals, with the structured, whereas the DFDG N-terminal region bein interactions. Also, two important interactions appear to exist, that may give rise to differences in receptor binding affinities: a Nto C-terminal ficant differences between th C-terminal PGVLRF-NH2 region being practically ung stabilized by hydrogen bonding 60

PAGE 61

loop form a N-terminal region influence the receptor binding affinity by drastically changing the popul,105 their tes and we have also applied REMD methoF-NH2 e estigate ormed when one or more acidic N-terminal residues interact with the side chain of the arginine, and a combination of this loop with the previously mentioned DFDG loop to fbicyclic structure. Additionally, previous NMR experiments on a series of short FLPs (5-7 residues, all sharing the PFLRF-NH2 C-terminal) has demonstrated that specific amino acid substitutions in the ations of reverse turns in solution (see also Section 1.2.3).32 The difference in NPR-1 activation, along with the existing hypothesis that N-terminal hydrogen bonding could influence activity, will be the main focus of our research. In the present work, we employ Replica-Exchange Molecular Dynamics, (REMD)104techniques to unravel the most dominant conformations of EMPGVLRF-NH2 and DFDGAMPGVLRF-NH2. Our interest is focused on the structural characteristics that may give rise to the difference in potencies for the two peptides. In order to better understand the biological functions of DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2, we have exploredconformational diversity in different protonation sta dology to a series of different peptide variants, with most of them ending in PGVLR(PGVLRF-NH2, DFDGAM-NH2, SGSGAMPGVLRF-NH2 and DFDGEMPGVLRF-NH2). Wperformed hydrogen bonding (HB), NMR chemical shift and clustering analyses to invthe structural properties that account for the difference in NPR-1 activation efficiency of DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2. 61

PAGE 62

3.2 Computational Methods and Systems 3.2.1 Systems The following series of FLPs have been studied in this work: EMPGVLRF-NH2, DFDGH2 F-NH2], ain; similar notation is used for pepossess natura et al.8 and ssing tify AMPGVLRF-NH2, SGSGAMPGVLRF-NH2, DFDGEMPGVLRF-NH2, PGVLRF-Nand DFDGAM-NH2, in different protonation states (Table 3-1). The protonated form of peptide I (EMPGVLRF-NH2) is peptide II [E(H)MPGVLRwhere a hydrogen atom has been added to the glutamic acid side ch ptides IV, V, VIII, and XI. It is important to mention that among these neuropeptides, the only ones that lly occurring amino acid sequences are DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2, whereas the four remaining are synthesized.8 Activity data were taken from Dosseythey are presented as a percentage of the response to (EMPGVLRF-NH2) Standard Error of theMean (SEM) (all 10-6M peptides were subjected to 2 m pulses to Xenopus oocytes expreNPR-1 215V). Since DFDGAMPGVLRF-NH2 is the least active native FLP-18 peptide and EMPGVLRF-NH2 the most active, all other peptides were chosen such that we could idenuseful relationships between structure and activity for the two peptides of interest; for example, SGSGAMPGVLRF-NH2 has the same sequence length and only three amino acids different thanDFDGAMPGVLRF-NH2 but its NPR-1 activation ability is much higher (Table 3-1). 62

PAGE 63

Table 3-1. Peptides consid Amino acid sequence Activitya I 100 II E(H)MPGVLRF-NH VI VII bDFDGAM-NH2 0 (3) D(H)GAM-NH2 Note:-6 M control pulse of iers. In order to overcome this problem, replica-exchange molecular dynamics (REMD) simulations have been carried out for the aforementioned neuropeptides with the SANDER MD module from the AMBER 9.0 simulation package.164 The peptides were represented by using the modified AMBER ff99SB force field,137 which resulted in a total of 183 atoms for DFDGAMPGVLRF-NH2, 137 atoms for EMPGVLRF-NH2, and 168, 105, 83 and 188 atoms for SGSGAMPGVLRF-NH2, PGVLRF-NH2, DFDGAM-NH2 and DFDGEMPGVLRF-NH2, respectively. The ered in this study. Peptide symbol EMPGVLRF-NH2 2III DFDGAMPGVLRF-NH2 29.1 5.7 (16) IV D(H)FD(H)GAMPGVLRF-NH2 V DFD(H)GAMPGVLRF-NH2 SGSGAMPGVLRF-NH2 118.7 11.0 (4) DFDGEMPGVLRF-NH2 19.0 2.6 (8) VIII DFD(H)GE(H)MPGVLRF-NH2 IX PGVLRF-NH2 43.0 5.5 (14) X XI D(H)F a) Activity is expressed as the magnitude of the potassium current produced by a 10-6M pulse of the peptide as a % of the response of the same oocyte to a 10EMPGVLRF-NH2.8 b) 40 ns REMD simulation. The highly conserved PGVLRF-NH2 region is in green. Hydrogens in parentheses denote a protonated side chain. If we could understand the reason why this happensin terms of differences in the configurations of the two peptidesit may provide us with a more complete insight about the different activities of DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2. 3.2.2 Computational Methods Sampling the conformational space of complex systems, such as proteins is a standing problem in theoretical chemistry.163 At low temperatures, molecular dynamics simulations get trapped in one of the many local-minimum-energy states separated by high energy barr 63

PAGE 64

generalized Born/solvent accessible surface a rea (GB/SA) implicit solvent model was used (in all y i an Case167 (the 0.8 algorithm168 was applied to erature constant. 00 as the initial conformation for every subsequent replica-exchange simulation. over a om walk in potential ong simulation time was applied in order to ensure that the confo simulations) to model the effects of solvation: in particular, the Hawkins, Cramer, Truhlar pairwise generalized Born model (GBHCT) was employed,165,166 with the radii parameters used b Tsudradius of hydrogen bonded to oxygen is; hydrogen bonded to carbon is 1.3 and hydrogen bonded to nitrogen is 1.3). The SHAKE constrain all bond lengths involving hydrogen to their equilibrium distance, and a 2 fs time step was used. The weak-coupling algorithm (Berendsen thermostat)169 was used to keep the temp Our peptides were initially built in extended conformations before being subjected to 2,0steps of steepest descent minimization; each final state obtained after the minimization was used In this study, replicas for all six peptides (Table 3-1) were simulated for 400ns each,range of exponentially distributed temperatures (Appendix A),105 with a 13% target exchange ratio. The exponential distribution of temperatures guarantees that a rand energy space is realized. Such a l rmational space is adequately sampled. As already mentioned, we have also considered each peptide in its different (partial and total) protonation states. For instance, the side chain of either (or both) the aspartic acids in DFDGAMPGVLRF-NH2 can carry a proton, resulting in the following derivatives: D(H)FDGAMPGVLRF-NH2, DFD(H)GAMPGVLRF-NH2 and D(H)FD(H)GAMPGVLRF-NH2. These three peptides, along with E(H)MPGVLRF-NH2, DFDGE(H)MPGVLRF-NH2 andD(H)FD(H)GAM-NH2, have also been included in our research scheme. In water solution, the pKa value for an aspartic acid side chain is 3.86, while for a glutamic acid side chain is 4.07; a 64

PAGE 65

10ns constant-pH REMD test calculation on DFDGEMPGVLRF-NH2 at pH=3.5 has shown that the (Asp3/Glu5) protonated form of the peptide is ~70% populated, with the other (deprotoor partially protonated) forms being significantly less populated. This is the reason we chose DFD(H)GE(H)MPGVLRF-NH2 to be our sole protonated form. The number of replicas for each protonated peptide is identical to the number of replicas for the corresponding non-protonatpeptides, since the total number of atoms in the protonated forms of the peptides is basically the same. Thus, exchange attempts were made after every 400ps and conformations were savthe simulation of each nated ed ed from replica every 2ps in the case of peptides III, IX, X and XI, and every 0.2ps and 1ile the number of exchae I, l peptides, the first 40 ns of the simulation were discar, ature s 0ps, respectively, for the remaining peptides (Table 3-1). During the course of this research, our group established that for optimal REMD performance and for a specific total simulation time, the number of MD steps between exchange attempts should be small, wh nges must be kept large.170 This change can be shown to increase the rate of convergence of the properties and has no bearing on the equilibrium of the system. This is the reason why thtimes reported above are not the same for all peptides. This yielded a total of 200,000 conformations for peptides I, III, VIII, IX, X and XI, and 40,000 conformations for peptides IIV, V, VI and VII, at each temperature. For al ded and only the remaining 360 ns were used for further calculations. Hydrogen bondingstructural clustering and NMR chemical shift analyses were performed on the lowest tempertrajectories obtained from each simulation. The REMD trajectories were analyzed for hydrogen bonding patterns with the ptraj tool171 under AMBER. We used 3.5 as a distance cutoff, along with an angle cutoff of 120, for all distances and angles respectively. The backbone-backbone hydrogen bonding interactionwere considered based on the distance between the carbonyl oxygen and the amide nitrogen, as 65

PAGE 66

well as on the angle that the carbonyl oxygen, amide hydrogen and amide nitrogen form. The side chain-backbone or side chain-side chain hydrogen bonding interactions follow similarly, since they involve analogous patterns (carboxylate oxygens and N-H systems on the side chai A very efficient way to manipulate the tremendous amount of data obtained from our MD simulations is to group all molecular configurations into subsets using the so-called clustering methodologies.138 The representative conformational (structural) clusters for each peptide wereobtained after performing an all-residue C-RMSD based hierarchical138 cluster analysis with MOIL-View Version 10.0, written by Carlos Simmerling.172 A cutoff of 1.5 was used for classifying 200,000 (or 40,000) conformations. Hydrogen bonding calculations were then performed on the resulting clustered trajectories. Additionally, NMR chemical shift analysis provide useful structural information: we performed chemical shift calculations on the trajectories obtained from the REMD simulation aswell as for each set of configurations obtained from the clustering analysis. The results reveal important information with regard to conformational diversity of the FLPs by helping ithe most-dominant configurations sampled. It has been known for some time that the structural chemical shifts (the differences between the resonance positions in a protein and in a random coil peptide carry useful information about the structure.173,174 We employ SHIFTS 4.1 program145,175 for estimating amide proton, as well as H chemical shifts for DF ns). dentify DGAMPGVLRF-NH2 and DFDGAM-NH2 and their protorom nated analogues. SHIFTS uses a protein structure and computes proton chemical shifts fempirical equations that represent magnetic anisotropy and electrostatic interactions.145 A consistent, physically reasonable set of empirical parameters has been established, which describes both the simplest, isolated shifts, as well as shifts in more complicated systems.146 66

PAGE 67

3.3 Results and Discussion 3.3.1 Hydrogen Bonding and Clustering Analyses DFDGAMPGVLRF-NH2 adopts two major conformations: we begin our investigation with the longe n at ilar phrasing will be used during our analysis. Morefor d et al., further supporting the possible interaction betwetoms Val9 4, peptide III is presented in Figure 3-3. st and least potent of the two peptides, DFDGAMPGVLRF-NH2. The hydrogen bonding (HB) analysison the full trajectory obtained from REMDshows that almost throughout thesimulation (98% of the time), a structural loop, formed by HB interactions exists between the side chains of aspartic acid either at position one or three (Asp1 or Asp3) and the arginine sidechain (Arg11,SC); in that particular case, side chain interactions indicate interactions betweeleast one of the carboxylate ions on the side chain of Asp1 or Asp3, with at least one of thehydrogen atoms on the Arg11 side chain, and sim specifically, we observed that the Asp1 side chain interacts with the Arg11 side chain 73% of the time, while the Asp3 side chain with Arg11 side chain interaction appears for 58%. Furthermore, the loop is strongly stabilized for 31% of the time, namely, during that time periowhen the Asp1-Arg11 and Asp3-Arg11 interactions coexist. These findings are in agreement with the NMR experimental results of Dossey en the two carboxylate residues and the arginine side chain in peptide III.8 Other major HB patterns include backbone interactions between carbonyl oxygen a(C=O) and amide protons (N-H), such as the principal interaction between the proline at positionseven and the leucine at position 10 (Pro7,O-Leu10,H), which occurs for 91% of the simulation time. The dominant HB interactions are summarized in Table 3-2, and a representation of the dynamics for peptide III is shown in Figure 3-2. We notice that residues such as Met6 and appear relatively rigid with no major fluctuations, whereas others (for example, Phe2, GlyAla5, Gly8) present significant flexibility. The and angles distribution for each residue in 67

PAGE 68

Table 3-2. Principal HB interactio2 (full REMD trajectory) Percentage Asp1 SC with Arg11 SC 72.6% ns in DFDGAMPGVLRF-NH Interaction (Asp1 or Asp3) SC with Arg11 SC 98.8% Asp3 SC with Arg11 SC 57.7% (Asp1 and Asp3) SC with Arg11 SC 31.4% Pro7,O with Leu10,H 91.1% Pro7,O with Arg11,H 64.3% Phe2,O with Ala5,H 61.7% Met6,O with Arg11 SC ~60% Note: Backbone interactions are shown in red and SC denotes a side chain. Amino 180 acposionD3,F2,G4,P7,V9,G8,L10,D3,F2,G4,P7,G8,V9,180 id itA5,-180Time (ns) M6,R11,F12,A5,M6,L10,R11,F12,16028040040 Time (ns)-180 Figure 3-2. Backbone dihedral angles for each residue in DFDGAMPGVLRF-NH2 as a function 40160280400of the simulation time. 68

PAGE 69

180 -180-180 180Phe2Phe12Arg11Leu10Val9Gly8PrAsp3 Ala5Gly4 o7Met6 Figure 3-3. Ramachandran plots for each residue in DFDGAMPGVLRF-NH2. Clustering analysis on the trajectory revealed two major clusters: Cluster 1 and Cluster 2 consist of 37,504 and 30,605 members, respectively. A minor Cluster 3 with 12,888 members could also be considered; nevertheless, the HB patterns in Cluster 3 resemble very much the ones observed for Cluster 1 hence, we could safely join those two clusters into one without loss of useful information. This provides us with the two most dominant configurations for the peptide (the representative structures for each cluster are shown in Figure 3-4, each one having the following HB patterns (Tables 3-3 and 3-4): 69

PAGE 70

Table 3-3. Principal HB interactions in DFDGAMPGVLRF-NH2 (Cluster 1) Interaction Cluster 1 Cluster 2 Asp1 SC with Arg11 SC (except H) ~100% Pro7,O with Leu10,H 92.2% 21.6% Met6,O with Arg11 SC >80% 0 Pro7,O with Arg11,H 70.3% 0 Gly8,O with Phe12,H 64.6% 0 Note: Main HB interactions in peptide III: Cluster 1; the corresponding values for Cluster 2 are contrasted. The Asp1-Arg11 interaction involves all hydrogen atoms of the Arg11 side chain, except the epsilon hydrogen atom. Table 3-4. Principal HB interactions in DFDGAMPGVLRF-NH2 (Cluster 2) Interaction Cluster 2 Cluster 1 (Asp1 or Asp 3) SC with Arg11 SC ~100% Phe2,O with Ala5,H 70.3% 43.7% Gly4,O with Phe12,H 69.8% 0 Asp1,O with Gly4,H 68.1% 0 Asp1,O with Met6,H 59.8% 0 or cluster 1 are contrasted. Figure 3-4. The two prevalent configurations for DFDGAMPGVLRF-NH2 peptide: representative structures for A) Cluster 1 and B) Cluster 2. Note: Main HB interactions in peptide III: Cluster 2; the corresponding values f AAsspp11 AArrgg1111 PPrroo77 LLeeuu1100 AAsspp11 AArrgg1111 AAllaa55 AA PPhhee22 BB 70

PAGE 71

Cluster 1 appears to form a strong loop towards the C-terminal (Pro7-Leu10 and Met6-, also combined with the stable Asp1-Arg11 loop. On the other hand, Cluster 2 shows an -termal struh interacts with a second (Asp1 or Asp3 to Arg11) loop, to form In order to gain any useful information that may be hidden in the complexity of data, we spect to the second-half (residues 7-12) of the peptide: all resulting clusters seem almost entical to the two main clusters described above. This further supports the existence of the two ed conformations for DFDGAMPGVLRF-NH2. Te HB red so far would be more valuable if it can be further developed: We considered the six major HB patterns for DFDGAMPGVLRF-NH2 (Table 3-2) and we em in the context of the time that two of them coexist (Table 3-5). Considering these cases where HB interactions show up in pairs, we conclude that the combination of a C-terminus loop with the (Asp1 or Asp3)-Arg11 side chain interaction, has a strong likelihood to occur. It appears that the formation of a bicyclic structure is further supported: a) the C-terminal loop is stabiliultaneous binding of Pro7,O with Leu10,H and Arg11,H and b) the major end-to-end loop is formed by the eous coexistence of the (Asp1 and Asp3) side chains with the Arg11 side chain. More imly, the results presented in Table 3-5 suggest that a) and b) occur at the same time, for most of the time. Arg11) Ninctural loop, whic a bicyclic structure. Notice that, with the exception of the principal interaction Asp1 to Arg11, which is present in both clusters, almost all other major HB patterns are different. performed additional clustering with respect to the first-half (residues 1-6), as well as with re id propos hinformation gathe The important question of which HB interactions are present simultaneously, could be elucidated by analyzing the existing HB data. organized th zed by the sim simultan portant 71

PAGE 72

Coexistence of principal HB interactions in DFDGAMPGVL Table 3-5. RF-NH2 CInteraction Leu10,H Arg11,H Ala5,H Arg11 SC (H) Arg11 SC Arg11 SC (1) Pro7,O91.1%68.8%/ 59.5%/ 54.0%/ 71.8%/ 55.2%/ 60.3%/ Ala5,H (4) Met6,O97.1% 83.6% 51.3% 50.7% 65.2%/ 44.7%/ / Arg11 SC 31.4% Note: HB coexistence for principal interactions in peptide III: a) % of time a single HB exists, b) for thare both present. nt conformation: The main HB interactions for EMPGnd the Ramager ptide I es 3(1) Pro7,O(2) Pro7,O(3) Phe2,O(4) Met6,O(5) Asp1 SC(6) Asp3 S a bLeu10,H c62.6% 54.2% 49.2% 65.4% 50.3% (2) Pro7,O97.4% 64.3% 52.1%/ 65.8%/ 67.7%/ 46.2%/ Arg11,H 33.5% 42.3% 43.5% 29.7% (3) Phe2,O87.9% 54.4% 61.7% 42.1%/ 74.1%/ 26.0% 45.7% 37.2% Arg11 SC (H) 33.0% 22.7%(5) Asp1 SC90.0% 59.9% 62.9% 45.5% 72.6% 43.3% (6) Asp3 SC87.2% 51.6% 64.5% 39.3% 54.5% 57.7% Arg11 SC e % of time that 1 is present, 1 and 2 are both present and c) Overall % of time that 1 and 2 EMPGVLRF-NH2 adopts one domina VLRF-NH2 are presented in Table 3-6 and a representative configuration is shown in Figure 3-7. Figures 3-5 and 3-6 provide the backbone dihedral angle evolution a chandran ( and ) plots respectively, for each residue in peptide I. Similarly to the lonpeptide III, we observe essentially rigid regions (Met2, Val5 and Leu6), while others appear to be more flexible (Pro3, Gly4 and Arg7). However, the overall degree of flexibility for peis smaller than peptide III, since the fluctuations for all residues are reduced (compare Figur2 and 3-5). 72

PAGE 73

Table 3-6. Principal HB interactions in EMPGVLRF-NH2 Interaction Percentage G lu1 SC with Ag7 SC 94.5% r Pro3,O with Leu6,H 92.3% Pro3,O with Arg7,H 76.2% Met2,O with Arg7 SC 75.5% (mostly H, 44.6%) Gly4,O with Phe8,H 64.6%lu1 SC with Met2,H ~30% G AminoF8,G4,F8,R7,V5,G4,4040160160280280400400 aci,L6,V5,0Time (ns)Time (ns) M2,1 18080 M2,-18018 L6, d positionR7,P3, P3 Figure 3-5. Backbone dihedral angles for each residue in EMPGVLRF-NH2 as a function of the ulation time. sim 73

PAGE 74

0Met2Pro3 180Phe8Arg7Leu6VaGly4 18 -180 -180 l5 LLeeuu66 A A r rgg77 A A r rgg77 PPrroo33 GGlluu11 Figure 3-6. Ramachandran plots for each residue in EMPGVLRF-NH2. Figure 3-7. Principal HB interactions for EMPGVLRF-NH2 peptide. 74

PAGE 75

75 Similarly to DFDGAMPGVLRF-NH2, we notice that analogous HB trends exist in the case of EMPGVLRF-NH2: the end-to-end terminal interaction between the glutamic acid and the arginine side chain is present, along with the already observed double HB between the proline and both the leucine and the arginine. This suggests that the common PGVLRF-NH2 region acquires similar configurations, in both peptides. Later on, we examine the PGVLRF-NH2 region in more depth. The dominant conformation proposed for EMPGVLRF-NH2 is further supported by the clustering analysis: we obtained only one major cluster with the same HB features as in the unclustered trajectory. 3.3.2 Different Protonation States w 6 to prevent amide roton exchange.176,177 When the pH is decreased from 5.5 to 2, the protonation state of the aspartic acid side chain changes from .8 Thus, HB nalysis on the fully protonated form of the peptide (present at pH < 2) shows that the (Asp1 or Asp3)-Arg11 side chain interaction does not form and that the backbone hydrogen bonds rearrange when compared with the deprotonated form. The most dominant HB patterns are shown in Table 3-7. Amide proton NMR experiments are usually performed at pH belo p the deprotonated to the protonated form a

PAGE 76

Table 3-7. Main HB interactions in the fully protonated form and in the deprotonated form of (mostly with H, 64.0%) 76.3% 0 Gly4,O with Leu10,H 44.3% 0 Phe2,O with Val9,H 32.3% 0 11,H 29.9% 0 Pro7,O with Leu10,H 23.9% 91.1% Pro7,ave been replaced by others; consequently the peptide adopts a substantially different conformation Notice how the Pro7 with Leu10 and Arg11 (the dominating backbone interactions in the deprotonated formped to 24% and 19%, respectively. In the low pH form of the peptide, the most significant HB interaction is the Gly8,O with Arg11 side chain, whereas in high pH (deprotonated form) this interaction is not observed. Thus, the configurations that the peptide adopts are pH dependent, something that may reflect in its biological function as well. Clustering analysis revealed only one cluster (1st and also 2nd part of clustering performed), further supporting the HB interactions already proposed for the fully protonated form of the peptide. Previous experimental studies have concluded that the pKa values for the aspartic acids in positions one and three are around 3 and 4, respectively.8 Based on these pKa values, as the pH decreases, we infer that protonation occurs first on the aspartic acid side chain in position three. In addition, by the time the protonated Asp3 side chain is significantly populated, the Asp1 side chain remains mainly negatively charged (deprotonated). Thus, we assume that between the fully DFDGAMPGVLRF-NH2 Interaction D(H)FD(H)GAMPGVLRF-NH2 DFDGAMPGVLRF-NH2 Gly8, O with Arg11 SC Arg11,O with Met6,H 52.7% 0Pro7,O with Phe2,H 34.5% 0 Gly8,O with Arg O with Arg11,H 18.9% 64.3% Note: As opposed to the deprotonated form, the dominant Pro7-(Leu10/Arg11) interactions hat low pH. ) percentages drop 76

PAGE 77

p rotonated form of the peptide [D(H)FD(H)GAMPGVLRF-NH2] and the deprotonated form (DFDGAMFD(H)GAMPGVLRF-NH2 intermediate forms. The main teractions for DFDRF-NH2, the partially protonated ide III Perceage 92% 62% et6,O with Arg11 SC ~65% (H 53.9%) de ions with DFDGix PGVLRF-NH2), only the D interactions for the partially protonated form of the peptide, are shown in Table 3-8: Table 3-8. Principal HB in(H)GAMPGVL form of pept Interaction nt Asp1 SC with Arg11 SC 7. Pro7,O with Leu10,H 87.8% Pro7,O with Arg11,H 6. M Asp3 SC with Ala5,H 41.0% Asp3 SC with Met6,H 38.1% By observing the HB patterns for the intermediate pH form of the peptide, we concluthat DFD(H)GAMPGVLRF-NH2 shares most of its principal HB interact AMPGVLRF-NH2, except the Asp3 side chain interactions (lost because of the proton addition) and the Phe2-Ala5 backbone interaction (present only in the deprotonated form). Additionally, we noted that the other partially protonated form (DHD) acquires an average configuration, which appears to be an intermediate or transition configuration between the deprotonated (DD) and fully protonated (DHDH) form of the peptide (data shown in AppendA, Table A-1). For EMPGVLRF-NH2, with varying the pH, the dominant Glu1-Arg7 interaction is not observed any more, since the proton added to the carboxylate on the glutamic acid side chain hasresulted in the impediment of such a bonding, as expected. Otherwise, we notice that the other major HB patterns remain the same, with frequencies very close to the ones observed for the deprotonated form. That leads us to the conclusion that, contrary to the DFDGAMPGVLRF-NH2 77

PAGE 78

case, changes in pH influence only the HB ability of the particular residue associated with theaddition of the proton, without affecting the remaining interactions. Thus, even though we recognize that the structure of EMPGVLRF-NH2 is dependent on pH, we conclude th at the pH ariance does not produce a combined effect, as in the case of DFDGAMPGVLRF-NH2. DFD can be protonated in three different positions, namely, the s can cept protono formompound. ed (see ), we aim thic acid and on ion the. He report our analysis only for this particular rotonation state: the protonated and deprotonated forms of the peptide present different HB trends, demonstrating that DFDGEMPGVLRF-NH2 HB interactions are pH dependent. Contrary to the case of EMPGVLRF-NH2, we noticed that protonation of the glutamic and aspartic acids results in the replacement of specific hydrogen bonds; namely, two strong HB interactions between Asp3 (side chain) and (Glu5,H 80%/Met6,H 80%) appear (less than 2% occurrence in the deprotonated form), while the significant Phe2,O with (Glu5,H 89% and Met6,H 80%) interaction in the deprotonated form dropped to less than 5% upon protonation. In the partially protonated form of the peptide, the Glu5 and Met6 backbone hydrogens remain the key HB participants, but with different partners: Asp3 (side chain) instead of Phe2,O (Table 3-9). Other HB patterns such as Pro7,O-Leu10,H (~90%) remain the same in both deprotonated and protonated forms. v GEMPGVLRF-NH2 glutamic acid and the two aspartic acidaca t a lower pH c Based on the relative pKa values and the constant-pH REMD calculation we perform Computational Methodsclat protonation occurs mainly on the glutam the aspartic acid at positrere, we p 78

PAGE 79

protonated form, DFD(H)GE(H)MPGVLRF-NH2 Interaction DFDGEMPGVLRF-NH2 DFD(H)GE(H)MPGVLRF-NH2Asp1 SC Table 3-9. Comparison of principal HB interactions in DFDGEMPGVLRF-NH2 and its with Arg11 SC ~70% ~70% Pro7,.0% Pro7,O with Arg11,H 56.1% 72.1% % Phe2,O with Glu5,H 88.6% 4.6% Asp3 SC with Met6,H <2% 79.9% ent only in the protonated form are in red. iffer icant e nine, s are y more. Additionally, the replacement of a phenylalanine with the much smaller glycine adds flexibility to the side chain. The HB analysis for peptides III and VI is summarized in Table 3-10. O with Leu10,H 90.8% 92Met6,O with Arg11 SC ~95% (H, 76.1%) ~60Phe2,O with Met6,H 79.8% 3.4%Asp3 SC with Glu5,H <2% 79.8% Note: Interactions present only in the deprotonated form are in green; interactions pres 3.3.3 Study of Mutations SGSGAMPGVLRF-NH2 versus DFDGAMPGVLRF-NH2: SGSGAMPGVLRF-NH2 (peptide VI) and DFDGAMPGVLRF-NH2 (peptide III) have the same number of residues and they donly in the first three amino acids; nevertheless, peptide VI activates the NPR-1 receptor much more than peptide III (Table 3-1). Even though both peptides have similar sequences, signifdifferences in conformation are anticipated, to account for this difference in potency. As we havalready noted, peptide III presents strong interactions between the aspartic acids and the argito form highly stabilized Nto C-terminal loops. By replacing an aspartic acid with a serine, the N-terminal side chain HB interactions are expected to diminish, since the carboxylate ionnot participating an 79

PAGE 80

80 Table 3-10. HB populations for DFDGAMPGVLRF-NH2 (in the fully protonated and unprotonated forms, peptides IV and III respectivel Asp1AArrgg1111 LLeeuu1100 AA BB SSeerr11 LLeeuu1100 Asp1PPrroo77 PPrroo77 y) and its mutant, SGSGAMPGVLRF-NH2 (peptide VI). SGSGAMPGVLRF-NH2 DFDGAMPGVLRF-NH2 D(H)FD(H)GAMPGVLRF-NH2 0 98.8% 0 98.1% 91.1% 64.3% ~95% (H, 89.4%) ~60% 4% 61.7% 6.4% ~25% rg11,O with Gly/Phe2,H 55.2% <1% <1% s, aspartic acid side chain interactions are eliminated, with peptide VI having almost all other HB interactions very similar to peptide III. The main structural differences are illustrated in Figure 3-8. In general, we could state that the two HB analyses present similar trends, except the Asp1 (and less Asp3) to Arg11 Interaction (Asp1 or Asp3) SC with Arg11 SC Pro7,O with Leu10,H 23.9% Pro7,O with Arg11,H 95.3% 18.9% Met6,O with Arg11 SC <10% Gly/Phe2,O with Ala5,H 9.3% Ser/Asp3,O with Met6,H 8 5.5% Gly8,O with Phe12,H 59.7% 0 <1% A Note: In both IV and VI peptide Figure 3-8. Principal HB interactions in A) DFDGAMPGVLRF-NH2 (peptide III) and B) its mutant, SGSGAMPGVLRF-NH2 (peptide VI).

PAGE 81

teraction, which is present only in DFDGAMPGVLRF-NH2, and may support the hypothesis that the parity: MPGVLRF-NH2 is significantly more active than DFDGAMPGVLRF-NH2, possibly hat it does not present any stable cyclic configurations associated with N-wathat g Ninveral HB interactions in common. The importance of the amino acid substitutions is further D(H)FD(H)GAMPGVLRF-NH2). The fact that we simply minimized the HB interactions associated with the aspartic acid side chains (by protonating them), does not guarantee that the rest of the interactions will remain the same as before. Also, in the case of SGSGAMPGVLRF-NH2, elimination of the side chain HB interaction alone, was not sufficient for lowering the potency. Specific mutations gave rise to a cooperative effect that resulted in reduced activity for the peptide. DFDF-NH2 versus DFDGAMPGVLRF-NH2: A single amino acid substitution at position 5 (glutamic acid instead of alanine) resulted in decreasing effectiveness for the long eptide (Tle 3-1). This observation suggests that the pree of a glutad residue in a for the effectivnt HB interactions remain the same, except the Gly8-Phe12 backbone interaction, which is not present in DFDGAMPGVLRF-NH2 and the Phe2-Met6 interaction, which 2 in ticipation of the N-terminal in HB formation can change the FLP activ SGSGA due to the fact t terminalC-terminal HB interactions. Furthermore, it seems that specific amino acid substitutions can substantially alter the peptide conformation, in such a y reat differences in activity may occur. Specifically, a three amino acid substitution in the-termal gives rise to a very big difference in NPR-1 activation between the two peptides, even though they have se emphasized if we consider the fully protonated form of the peptide (Table 3-10, GEMPGVLR p ab senc mic aci position corresponding to the first residue position of EMPGVLRF-NH2 is not a determinant eness. After the Glu-Ala mutation, the most domina occurs more often in DFDGEMPGVLRF-NH. 81

PAGE 82

We also observed that the side chain of the glutamic acid has almost no interaction with the arginine side chain and more importantly, it presents only minor interactions with aresidue (principal interactions: E (side chain)-D1 (H) 13%, and E (side chain)-F2 (H) 12%). Thmain structural differences between DFDGEMPGVLRF-NH2 (peptide VII) and DFDGAMPGVLRF-NH2 (peptide III) are presented in Table 3-11. (Asp1 or Asp3) SC with Arg11 SC ~70% (Asp1 SC only) 98.8%Pro7,O with Arg11,H 56.1% Phe2,O with Glu/Ala5,H 88.6% 61Gly8,O with Phe12,H 55.5% 0 Asp3 SC with Gly8,H ~50% 4% Note: The replacement of the alanine with a glutamic acid did not result in significant change 3.3.4 The Cny other e Table 3-11. HB patterns for DFDGAMPGVLRF-NH2 and DFDGEMPGVLRF-NH2. Interaction DFDGEMPGVLRF-NH2 DFDGAMPGVLRF-NH2 Pro7,O with Leu10,H 90.8% 91.1% 64.3% Met6,O with Arg11 SC ~95% (H, 76.1%) ~60% .7% Asp3,O with Met6,H 9.4% ~25% Phe2,O with Met6,H 79.8% ~30% in structure. terminal PGVLRF-NH2 Region Fg to the receptor. In order to test this suggestion, we must examine the structure of the PGVLRF-NH2 After investigating the role of N-terminal on the configuration and potency of the peptideswe focused our interest on the C-terminal region. More specifically, we examined the PGVLRNH2 end, which is present in both peptides I and III; in order to have a more complete understanding of how the PGVLRF-NH2 end influences the structure and activity of the peptides,the HB analysis has been performed on the clustered trajectories of I and III, as well as on the clusters of II, IV, VI, VII and IX peptides. It has been indicated8 that the PGVLRF-NH2 terminal adopts random conformations in solution, and that C-terminal flexibility is crucial for bindin 82

PAGE 83

sequeTable 3-12. Summary of the principal HB interactions (peptides I, III, IV, VII and X), in IV Peptide IX lu1 (SC)-Arg7 (SC) Asp1,3 (SC)-Arg11 (SC) Gly-Arg Gly-Arg ro-Leu e Pro-Arg Gly4-Leu10 Val3-Phe6 ) G Met-Arg (SC) Phe2-Val9 Gly-Phe are colour matched to their corresponding een oitted e to t te peptse hierchieswe nohe principV, Gly2rg5 aepdes prII) pees share principal interactions between: i) the proline and the leucine and ii) the proline PGVLRF region has a well defined HB pattern assocNH2 (full nce. Table 3-12 shows the principal HB interactions (in decreasing intensity) present for each case. decreasing intensity. Peptide I Peptides III and VI Peptide G Pro-Arg Pro-Leu Arg11-Met6 P Gly-Ph Met-Arg (SC ly-Phe Pro7-Phe2 Gly2-Arg5 Note: Peptide II presents the same HB pattern as peptide I, except the Glu1-Arg7 bonding. Interactions in commonpeptides; for these cases, the numbering scheme has bmduhe different length ofhides. Non-numbered residues correspond to the (MPGVLRF-NH2) region, common in all peptides. If we compare thera, te for instance that, tal interactions for I Gly8-Arg11 and for IX,-Are the same. Namely, both ptiesent a dominant interaction between the glycine and the arginine. Similarly, we observe that other (III, VI, I, and ptid and the arginine. This strongly suggests that the iated with a unique structure, which is in contrast to the claim that the conserved C-terminal PGVLRF-NH2 is largely unstructured in solution. The and angle distributions of the common C-terminal MPGVLRF region in DFDGAMPGVLRF-NH2 and EMPGVLRF-trajectories) further support the structural similarity between the C-terminal regions of these peptides (Figure 3-9). In conclusion, it is very likely that the C-terminal flexibility is not important for binding to the receptor. In Figure 3-10, two representative structures of DFDGAMPGVLRF-NH2 and EMPGVLRF-NH2 peptides are shown. The remarkable similarity 83

PAGE 84

between the two PGVLRF C-terminal regions suggests that the structural loop formed by the Nterminal of -DFDGAMPGVLRF-NH2 may influence the peptides binding affinity for the ceptor. Since both peptides present almost identical C-terminal structures, relative NPR-1 inhibition by the (DFDGAM)C-terminal bicyclic configuration is a strong possibility. re Met2/6Gly4/8Pro3/7 180 180-180 -180 Phe8/12Arg7/11Leu6/10Val5/9 Figure 3-9. Ramachandran plots for the common MPGVLRF C-terminal region in DFDGAMPGVLRF-NH (red) and EMPGVLRF-NH (black). The similarity between the (/) distributions suggests that the C-terminal region is associated with a unique structure. 22 84

PAGE 85

A B N N --tteerrmmiinnaall N N --tteerrmmiinnaall CC--tteerrmmiinnaall CC--tmmiinnaall teerr Figure 3-10. Representative conformations for A) DFDGAMPGVLRF-NH2 and B) EMPGVLRF-NH2. The common PGVLRF region is shown as a tube. DFDGAMPGVLRF-NH2 activates NPR-1 less effectively, possibly because of the N-terminal structural loop formed. 3.3.5 The N-terminal DFDGAM Region Chemical shifts analysis: In order to further investigate whether the N-terminal extensions behave independently or not, we performed NMR chemical shift calculations on DFDGAM-NH2 and DFDGAMPGVLRF-NH2 peptides. By comparing amide proton (NH) and alpha proton (H) chemical shifts for the first six amino acids and for both peptides, we observe a similarity that able 3-13, Figures 3-1DFDGAMPGVLRF-NH2 peptide behaves in a particular way that reduces its potency to NPR-1 receptor; it is apparent that it does not just acquire a random conformation. Edisons research group performed NMR resonance experiments for DFDGAM-NH2 and DFDGAMPGVLRF-NH2 using standard two-dimensional 1H-based methods177 at 288K.8 Comparison of experimental chemical shifts to random-coil178-181 values revealed regions with significant populations of allows us to conclude that the N-terminal extensions behave as independent units (T1 and 3-12). This may be an additional indication that the N-terminal of the 85

PAGE 86

secondary structure. In particular, amide and chemical shifts for the N-terminal (DFDG) at pH ~2.3 showed significant deviation from random-coil values, further supporting that this is an area with significant structure. In agreement with our results (Table 3-13), the greatest deviation was also experimentally observed for the Gly4 amide proton. Additionally, 1D NMR pH-titration experiments by Edisons lab showed that the DFDGAM-NH2 spectra are pH dependent and resemble the spectrar the correspGAM region in DFDGAMPGdemonstrating that the N-terminal units do not adopt random conformations. Table 3-13. Amide and alpha proton chemical shift calculations for DFDGAMPGVLRF-NH2 and DFDGAM-NH Residue ical shift D1 8.13 8.15 4.04 4.02 2 8.14 8.07 4.57 4.58 D3 7.59 7.71 4.35 4.36 7.54 3.77 3.84 4.41 VLRF-NH2, fo onding DFD 2 Amide proton chemical shift Alpha proton chem DFDGAM-NH2 DFDGAMPGVLRF-NH2 DFDGAM-NH2 DFDGAMPGVLRF-NH2 F G4 7.47 A5 7.50 7.57 4.26 4.06 M6 7.85 7.79 4.31 Note: The greatest deviations from random-coil values correspond to Gly4 hydrogens (in red). Figure 3-11. Amide proton NMR chemical shift as a function of residue number for DFDGAM-NH2 (blue curve) and DFDGAMPGVLRF-NH2 (purple curve). 7.47.688.2Residue 7.801234567hcaft DFDGAM DFDGAMPGVLRF 86

PAGE 87

87 3.5 3.74.14.501234567Residue 3.94.34.7m sh DFDGAM DFDGAMPGVLRF DFDGAMPGVLRF-NH2 (purple curve). Ife fur A B A B A B A B A B 2H 9% 7% 3H 6% 13% 4% 4H 5% 8% 11% 16% 1% 5% 5% 5H 5% 3% 42% 46% 10% 2% 4% 7% 6H 11% 23% 21% 1% 2% 50% 19% Note: Interactions present for more than 10% of the time are in red, while those being present for less than 0.5% are not shown. For DFDGAM-NH2, it appears that hydrogen bonding proton-participation obeys the following trend: 6H>5H>4H>2H>3H, with the methionine amide hydrogen (6H) being involved Figure 3-12. Alpha proton NMR chemical shift as a function of residue number for DFDGAMNH2 (blue curve) and wther compare the HB patterns that the first 2 amide protons present we have Table 3-14 and Figures 3-13 and 3-14: Table 3-14. Significant HB interactions of the 2-6 amide protons within the DFDGAM region of A: DFDGAM-NH2 peptide and B: DFDGAMPGVLRF-NH2 peptide. 1OD1 or 1OD2 1O 2O 3OD1 or 3OD2 3O in HB for a longer time than any other amide hydrogen. This does not contradict

PAGE 88

DFDGAMPGVLRF-NH2, but for the longer peptide we could also attribute: 5H>6H>4H>3H>2H. Maybe this is an insignificant detail, but it could also be an indication that the N-terminal extensions do not behave absolutely independently. Nevertheless, N-terminal interactions seem to be realized mostly because of the N-terminal extensions, without being significantly dictated by other parts. 0 1020 304050HB percentage 2H3H4H5H6HAmide Proton 1OD1 1O 2O 3OD1 3O Figure 3-13. Residue 2 amide proton in teractions in DFDGAM-NH2. 50 40 1OD1 1O 30 HB percentage 20 2O 10 3OD1 3O 0 2H3H4H5H6H Amide Proton Figure 3-14. Residue 2 amide proton interactions in DFDGAMPGVLRF-NH2. 88

PAGE 89

3.4 Conclusions The conformational characteristics of two unbound neuropeptides were investigated in order to understand the difference in their potencies on a specific C. elegans receptor. Our analysis is based on the fact that EMPGVLRF-NH2 activates the NPR-1 receptor more efficiently than DFDGAMPGVLRF-N H2. REMD calculations were performed to both peptides, as well as to a series of several test peptides. DFDGAMPGVLRF-NH2 adopts two main conformations, forming either a strong loop towards the C-terminal, or a bicyclic structure which combines N-terminal with C-terminal loop interactions. On the other hand, EMPGVLRF-NH2 presents one dominant conformation with the PGVLRF-NH2 region resembling the corresponding region of DFDGAMPGVLRF-NH2. This structural similarity suggests that the bicyclic region on DFDGAMPGVLRF-NH2 may be responsible for its lower potency on NPR-1. MPGVLRF-NH2 change, whereas in the case of EMPGVLRF-NH2 the pH variation does not affect the existing interactions significantly. The role of the N-terminal region was further examined by comparing the hydrogen bond patterns of DFDGAMPGVLRF-NH2 to those of two mutants, SGSGAMPGVLRF-NH2 and DFDGEMPGVLRF-NH2. An interesting observation is that when the N-terminal-associated hydrogen bonding interactions are not significant (SGSGAMPGVLRF-NH2) the activity of the peptide is enhanced, further suggesting that the bicyclic structure observed in DFDGAMPGVLRF-NH2 diminishes the potency. Additionally, erved that specific amino acid substitutions (DFDGEMPGVLRF-NH2) alter the activity al shift analysis on DFDGAMPGVLRF-NH2 and DFDGAM-NH2 peptides suggests that the N-terminal region (DFDGAM) behaves as an independent structural unit. The C-terminal region was next considered; for a series of neuropeptides sharing the same C-terminal amino acid sequence, we With varying pH, the dominant hydrogen bonding interactions in DFDGA we obs of the peptides via significant conformational changes. Furthermore, NMR chemic 89

PAGE 90

observed that the PGVLRF-NH2 motif iation, with the same structural featured t the s a unique conform es for all peptides. This observation contradicts previous studies arguing that the conservC-terminal PGVLRF-NH2 region is mostly unstructured in solution, and rather proposes thaN-terminal region is determinant for the differences in binding affinity. 90

PAGE 91

CHAPTER 4 NEUROPEPTIDE 4.1 Introduction Pain is a complex perception; it depends on emotional experience, STRUCTURE-ACTIVITY RELATIONSHIPS OF FLP NEUROPEPTIDES: NPFF environmental factors, and therefore varies from person to person. Pain can be defined as a sensory and emotional state related to actual or potential tissue damage, and because of its highly subjective nature, it is difficult to be treated clinically.1 Nevertheless, many advances have contributed in understanding e physiology of pain mechanisms, such as the discovery that stimulation of specific receptors the brain produces analgesia, and that opioid peptides (acting as neurotransmitters) participate the endogenous pain control system.1,182-189 Neuropeptide NPFF (FF, Phe-Leu-Phe-Gln-Pro-Gln-Arg-Phe-NH2) is a mammalian eurotransmitter, which interacts with specific receptors present in the central nervous stem.190-196 It is involved in pain transmission, as well as in a variety of other physiological nctions, including cardiovascular and memory regulation, opioid tolerance, and feeding rocesses.6,30,194,195,197-201 NPFF also belongs to the opioid family of FLP neuropeptides.9-16 FLPs, s described in previous sections, are considered to be present in all animals and they have milar amino acid sequences to the molluscan cardioexcitatory neuropeptide FMRFamide (Phe-et-Arg-Phe-NH2), with all of them sharing the RF-NH2 C-terminus.17 NPFF, similarly to all LPs, is encoded by precursor proteins that are processed into mature peptides by specific roteolytic cleavages, catalyzed by proteases. Two polyproteins (pro-NPFFA and pro-PFFB)30,202,203 participate in the NPFF precursor system, with pro-NPFFA202 producing NPFF, mong other peptides. NPFF was initially isolated from bovine brain, and depending to its route f administration it induces both pro-opioid and anti-opioid effects. The mechanisms underlying these opposing physiological functions of NPFF are not completely understood, but there is th in in n sy fu p a si M F p N a o 91

PAGE 92

strong evidence that NPFF acts so variousteracts with more than one receptor subtype. Two G protein-coupled (GPCR) rated with the NPFF peptide:204,205 NPFFnal d neureptides, respectively. Therefore, NPFF2 receptor is the of peptide ly because it in eceptors are associ 1 and NPFF2 receptors, which are related to the neuropeptide Y (NPY)206-208 and orexin receptors209 (30-35% homology), have been cloned and characterized in human and rat. In rodents, NPFF2 is found in the brain and spinal cord, whereas NPFF1 is detected in supraspiregions. They present distinct functional differences, even though they are structurally 50% related. Although NPFF1 and NPFF2 receptors are not very discriminative towards the peptides from both pro-NPFFA and pro-NPFFB precursors, they have been observed to interact mostly with pro-NPFFB and pro-NPFFA deriveop primary target for NPFF, even though binding and activation of NPFF1 by NPFF is important in biological functions. In order to relate the structural characteristics of NPFF with its high-affinity binding to NPFF2 receptor, a series of analogues with modifications on the C-terminal were computationally investigated. Previous experimental studies have shown that specific NPFF mutations alter significantly the binding affinity and in turn, the pharmacological propertiesNPFF. In the present work, we applied replica-exchange molecular dynamics (REMD) methodologies to study the conformational properties of selected NPFF analogues. By correlating the structural characteristics of each analogue with its binding affinity for the receptor, important relationships between the structure and the activity of NPFF neurocould be established. 92

PAGE 93

4.2 Computational Methods and Systems 4.2.1 Syst ems et r lation, ethylation of the peptide bond, had a minor effect on the Ki value were kept changes e Several NPFF analogues with different binding affinities to NPFF receptor have been considered in this study (Table 4-1). Peptides were synthesized such to identify specific structural characteristics that may give rise to varying NPFF receptor activation. Gouardresal. have reported experimental quantitative autoradiography studies for evaluating NPFF receptoaffinities with respect to peptides I-VII. Thus, in Table 4-1, affinity Ki is represented as the ability of NPFF (and its corresponding mutants) to inhibit the binding of [125I]1DMe (D.Tyr-Leu-(NMe)Phe-Gln-Pro-Gln-Arg-Phe-NH2) to rat spinal cord membrane preparations. Previous experiments have demonstrated that N-terminal modifications on NPFF do not significantly alter the binding affinity: changes such as blocking of the N-terminal by acetyinsertion of a bulky residue, or N-m s. Similarly, shortening of the peptide sequence by deleting residues from the N-terminalside produced a moderate decrease in affinity, until the last four NPFF amino acids(peptide VII). In that case, peptide VII presents a 450-fold loss in affinity to the receptor. In contrast, C-terminal modificationsespecially in positions seven and eighthave produced a substantial loss in affinity. For peptides II-VI, a single C-terminal amino acidsubstitution decreases greatly the affinity for the receptor. The possible conformational induced by the specific amino acid substitutions, along with the relationship between thesstructural changes and the affinities to the NPFF receptor will be the main focus of this investigation. 93

PAGE 94

Table 4-1. Peptides consid bol Amino-acid sequence Mean SD I II Phe-Leu-Phe-Gln-Pro-Gln-Arg-Tyr-NH 34.0 10.2 0 aautoradiography as the displacement of [125I]1DMe binding sites of the rat spinal cord. 4.2.2 on of 156 aI and rface odel: similarly to the previous project, the Hawkins, Cramer, Truhlar pairwise neralized Born model (GBHCT) was employed, with the radii parameters used by Tsui and Case (the radius of hydrogen bonded to oxygen is 0.8; hydrogen bonded to carbon is 1.3 and hydrogen bonded to nitrogen is 1.3). The SHAKE algorithm was used to constrain all hydrogenered in this study Peptide sym (NPFF) Phe-Leu-Phe-Gln-Pro-Gln-Arg-Phe-NH2 0.34 0.07 2 III Phe-Leu-Phe-Gln-Pro-Gln-Lys-Phe-NH2 245 9 IV Phe-Leu-Phe-Gln-Pro-Glu-Arg-Phe-NH2 307 87 V Phe-Leu-Phe-Gln-Pro-Asn -Arg-Phe-NH2 7.32 1.67 VI Phe-Leu-Phe-Glu-Pro-Gln-Arg-Phe-NH2 1.15 0.26 VII Pro-Gln-Arg-Phe-NH2 15.5 2.3 Note: Neuropeptide NPFF (I) and its analogues (II-VII) considered in this study. Amino acid substitutions for the mutants are in red. Affinities (Ki) are calculated by quntitative Computational Methods In order to minimize the possibility that our simulations get trapped in a local low-energy state without ensuring an adequate sampling of the conformational space, replica-exchange molecular dynamics (REMD) simulations were performed for NPFF neuropeptide and its analogues. All calculations have been carried out with the AMBER 9.0 simulation package by employing the modified AMBER ff99SB molecular mechanics force field for the simulati toms for NPFF, and 157, 154, 154, 153, 154 and 80 atoms for peptides II, III, IV, V, VVII, respectively. Each peptide was initially built in a fully extended chain and was energyminimized for 2,000 steps of steepest descent minimization to relax any possible strains introduced. The solvent was treated implicitly with the generalized Born/solvent accessible suarea (GB/SA) m ge 94

PAGE 95

involved bond distances, with an integration t ime step of 2fs. The weak-coupling algorithm e f 00,000 conformations for each of the peptides, at each temperature. In order not to misdirect ormations, the first 20ns (10%) of each simulation were discarded and only the remaining 180ns were used nding, clustering, and SASA calculations were applied to the lowesplicit r icas ging from 282.8K-540.6K (Appendix B). A total simulation time of 50ns was divide (Berendsen thermostat) was used to keep the temperature constant. In this study, replicas for all seven peptides (Table 4-1) were simulated for 200ns each over a range of exponentially distributed temperatures, with a 13% target exchange ratio. The temperatures chosen for the replicas are presented in Appendix B. We attempted to exchang replicas after every 0.1ps and conformations were saved every 2ps. This resulted in a total o 1 our analysis by taking into account the initial events primarily related to random conf for further analysis. Hydrogen bo t temperature trajectories from each simulation. Additionally, a hybrid-solvent replica exchange molecular dynamics (hybrid-REMD) simulation was performed for NPFF (peptide I). The actual simulation was performed in exsolvent with truncated octahedral periodic boundary conditions. The TIP3P water model was used.210 The number of water molecules retained in the hybrid model was 76, a sufficient population even for relative extended conformations. After minimization and equilibration, the explicit solvent simulation was performed on a system consisting of 3933 atoms (1259 watemolecules added to the 156-atom peptide). To calculate the exchange probability, 12 replwere used ran d in 500,000 exchange attempts every 100ps. The Berendsen thermostat and the SHAKE algorithm were also used to keep the temperature constant and to restrain the bonds involving hydrogens, respectively. A total of 10,000 frames were generated and the last 9,000 snapshots were considered for the HB calculations. 95

PAGE 96

A distance cutoff of 3.5 and an angle cutoff of 120 were used to perform the hydrogbonding analyses with the ptraj tool under AMBER. The backbone-ba en ckbone, side chain-side chain e ing techniques in the case of small peptidibed D es nt r e simulation was discarded and from the remaining and side chain-backbone interactions are defined as in the previous project, involving interactions among amide hydrogens, carboxylate oxygens, carbonyl oxygens, hydroxides and N-H systems on the side chains. Analysis of the large amount of data contained in the REMD trajectories (90,000 framesper each trajectory) provides an adequate average description of the system, but at the same timuseful information may be hidden in the complexity of that average. If all molecular configurations were grouped into subsets according to a measure of similarity (clustering), a more realistic description of the system is obtained, by identifying the most probable conformational states present. The usefulness of cluster es becomes apparent, for such very mobile systems are unlikely to be adequately descrby a single conformation. The representative conformational clusters for each peptide are obtained after a C-RMSbased hierarchical cluster analysis with the MOIL-View program (version 10.0). All residuwere included, while the first ten percent of each REMD trajectory was discarded, with the remaining used for the clustering. A cutoff of 1.5 was used for classifying the 90,000 differeconformations. The resulting clustered trajectories were used for further hydrogen bonding andSASA calculations. Solvent accessible surface area (SASA) calculations were based on the rolling ball algorithm described in Section 2.5.3. A value of 1.4 was used as a probe radius for the water molecule. SASA values were obtained for the five C-terminal residues of peptides I-VI. Similato the HB calculations, the initial 10% of th 96

PAGE 97

90,00 the identification of dominant NPFF configurations. Hydrogen bondi ydrogen en onyl rg7 ure 4-1, and the corresponding Ramachandran plots are presented in Figur Gln4,OE1(SC)-Arg7, H (SC) 28.80% 27.83% (SC) 27.14% Note: The two m 0 conformations, the 9,000 were used for estimating the SASA of each residue. SASA calculations for individual clusters considered every member of the cluster. 4.3 Results and Discussion 4.3.1 NPFF Neuropeptide We begin our analysis with ng (HB) calculations on the complete REMD trajectory suggest that interactions involvingthe arginine at position seven (Arg7), the phenylalanine at position eight (Phe8) and the glutamine at position four (Gln4) are prevalent (Table 4-2). In particular, the amide hatoms on Arg7 and Phe8 interact frequently with the carbonyl oxygen on Gln4 to form hydrogbonds (Gln4,O-Arg7,H and Gln4,O-Phe8,H). Side chain interactions between the carboxygen on the side chain of Gln4 and the epsilon hydrogen (H) atom on the side chain of Aare also important (Gln4,OE1-Arg7, H). The and values for each NPFF residue during the dynamics are shown in Fig e 4-2. Table 4-2. Principal HB interactions in NPFF (full REMD trajectory) Interactions Percentage Gln4, O-Arg7, H 67.26% Gln4, O-Phe8, H 46.71% Phe3, O-Gln6, H Leu2, O-Arg7, H ain HB interactions in NPFF are in green. 97

PAGE 98

98 180L2,Q6,-180 180incsitionF3,L2,R7,Q6,Q4,-180 Amo aid poF8,R7,P5,Q4,F8,F3,P5,Time (ns)Time (ns)20208080140140200200 e. Figure 4-1. Backbone dihedral angles for each residue in NPFF as a function of the simulation tim Leu2Phe3Gln4 180 Pro5Gln6A rg7Phe880 1-180 -180 igure 4-2. Ramachandran plots for each residue in NPFF. F

PAGE 99

The validity of the results was confirmed by performing a hybrid replica-exchange molecular dynamics simulation (Hybrid-REMD). In agreement with the implicit-solvent results, the dominant HB patterns involve similar interactions (Table 4-3). Table 4-3. Comparison of HB patterns obtained from implicit-solvent and hybrid-REMD simulations for NPFF Interactions Percentage Implicit-solvent REMD Hybrid-REMD Gln4, O-Arg7, H 67.26% 27.89% Gln4, O-Phe8, H 46.71% 13.65% Gln4, OE1 (SC)-Arg7, H (SC) 28.80% Phe3, O-Gln6, H 27.83% Leu2, O-Arg7, H (SC) 27.14% Pro5, O-Phe8, H 10.49% Pro5, O-Arg7, H 5.77% Phe8, O-Gln4, H 5.16% Note: interactions occurring less than 20% (implicit-REMD) and less than 5% (hybrid-REMD) are not shown. Additionally, it is suggested that some other HB interactions may be hidden in the great amount of data the implicit trajectory yields (90,000 frames): for example, backbone interactions between the proline at position 5 (Pro5) and the phenylalanine at position 8 (Phe8) are less than 20% populated when implicit-solvent REMD is performed, whereas for hybrid-REMD it is the third most frequent interaction. This indicates that some specific configurations appear for an efficient amount of time but they pass unnoticed under the grand average of the one conformation obtained. In that case, clustering analysis can be a very useful tool. The agreement between the more accurateyet very computationally expensive hybrid-REMD methodology and the implicit-solvent REMD calculations allows us to perform any subsequent calculations using implicit REMD with confidence. 99

PAGE 100

100 Gln-4 Gln-4 PPII Cluster 1: 35.9% Cluster 2: 27.6% Cluster 3: 13.4% Gln-4 T ition four (Gln-4) for the Cluster 3 a polyproline II helical (PPII, poly-Pro II) conformation about Gln4. The main HB interactions for Cluster 1 are almost identical with the hus, clustering the implicit-solvent REMD trajectory revealed three major clusters:Cluster 1, Cluster 2 and Cluster 3 comprise the 35.9%, 27.6% and 13.4% of the full trajectory, respectively. The and angles distribution about glutamine at pos three clusters are shown in Ramachandran plots (Figure 4-3). Figure 4-3. Ramachandran plots about Gln-4 for the three NPFF clusters. Each member of the group belongs to a well defined cluster, having a very distinct secondary structure associated with the glutamine. Specifically, Cluster 1 adopts a -sheet configuration, Cluster 2 an -helix, and

PAGE 101

101 full tr 53.56% 33.24% Gln4, O-Arg7, H 53.10% 71.14% Note: Main HB interactions in NPFF for Cluster 2; the corresponding values for Cluster 3 are contrasted. The two dominant interactions obtained by the full trajectory analysis occur less frequently in Cluster 2 (in green). Table 4-5. Principal HB interactions in NPFF (Cluster 3) Interactions Percentage Cluster 3 Cluster 2 Gln4, O-Arg7, H 71.14% 53.10% Pro5, O-Phe8, H 33.24% 10.13% Gln 33.24% 53.56% Leu2, O-Gln4, H 32.07% 0 Leu2, O-Arg7, H (SC) 21.57% 73.53% Note: Main HB interactions in NPFF for Cluster 3; the corresponding values for Cluster 2 are contrasted. The three clusters represent three different configurations, with Cluster 1 mainly resembling the average conformation of the full trajectory (Figure 4-4A). On the contrary, Cluster 2 is dominated by the Phe3-Gln6 backbone interaction and by the Leu2 (carbonyl oxygen)-Arg7 (side chain) interaction (Figure 4-4B); the most frequent HB interaction in Cluster 3 is the backbone interaction between Gln4 and Arg7 (Figure 4-4C). ajectory, and the populations for each HB are cited in Appendix B (Table B-1). The principal HB interactions for Clusters 2 and 3 are shown in Tables 4-4 and 4-5. Table 4-4. Principal HB interactions in NPFF (Cluster 2) Interactions Percentage Cluster 2 Cluster 3 Phe3, O-Gln6, H 87.76% 0 Leu2, O-Arg7, H (SC) 73.53% 21.57% Phe3, O-Arg7, H 60.10% 0 G ln 4, O-Phe8, H 4, O-Phe8, H

PAGE 102

102 So far, it is suggested that the high affinity of the peptide for the NPFF2 receptor is primrelated to the interactions among residues Gln4, Arg7 and Phe8 with an additiona Gln-4 Phe-8 Arg-7 Gln-4 Phe-8 Arg-7 Arg7 Gln6 Leu2 Phe3 AA BB CC arily l indication that me other secondary interactions may be involved. However, various aspects may be also ositioning and orientation of ffinity for the ter .36. Figure 4-5 provides a so crucial for high affinity, such as the charge, polarity, length, p individual residues. To associate the nature of each residue with the peptides a recep tor, specific amino acid substitutions along with their induced changes in affinity are investigated. F igure 4-4. The three prevalent configurations for NPFF: representative structures for A) Clus 1, B) Cluster 2 and C) Cluster 3. 4 .2 C-terminal Substitutions: The Role of Phenylalanine Substitution of the phenylalanine at position eight with a tyrosine, leads to a 100-fold loss in binding affinity (Table 4-1). To question this significant change, REMD calculations have been performed for the NPFF analogue, Phe-Leu-Phe-Gln-Pro-Gln-Arg-Tyr-NH2 (peptide II). The main HB interactions for the full trajectory are shown in Table 4-

PAGE 103

103 rep Pro5, O,-Tyr8, H 37.86% Tyr8, OH-Arg7, H (SC) 27.76% Tyr8, H 23.53% resentation of the NPFF dynamics, and Figure 4-6 shows the structural summary ( and plots) of the peptide. Table 4-6. Principal HB interactions in peptide II (full trajectory) Interactions Percentage Gln4, O-Arg7, H 59.86% Leu2, O-Arg7, HH (SC) 35.42% Gln4,O N ote: The main HB interactions in NPFF are in green. The additional Tyr-Arg interaction occurring in peptide II is shown in red. Am inQ4,R7, o acid position180180L2,R7,F3,Q6,L2,F3,Q6,Q4,-180-180Time (ns)Time (ns) Y8,P5, Y8,P5,20208080140140200200 simulation time. Figure 4-5. Backbone dihedral angles for each residue in peptide II as a function of the

PAGE 104

Leu2Phe3Gln4 Pro5 Gln6Arg7Tyr8180-180 180-180 Figure 4-6. Ramachandran plots for each residue in peptide II. At a first look, the average picture for peptide II resembles the hydrogen bonding pattern of NPFF, with the backbone Gln4-Arg7 interaction remaining the predominant, and the Pro5-Tyr8 gaining ground over the Gln4-Phe8 interaction. It seems that the replacement of the phenylalanine with the tyrosine drives the amide hydrogen on residue eight to interact more preferably with the Pro5 rather than with the Gln4 (Table 4-6). None the less, a noticeable difference from the NPFF HB pattern is a new interactionwhich appears for almost 1/3 of the simulation timebetween the hydroxyl on the tyrosine ring and the side chain of the arginine. Thus, it is reasonable to assume that the conformational changes induced by the OH group on the tyrosine ring may be responsible for the lowering in affinity. The identification of principal hydrogen bonds that coexist with the OH interaction would be a step closer to this assumption. Three distinct groups are obtained after clustering the trajectory of peptide II (Table 4-7 and 13.1% correspond to Cluster 2 and Cluster 3, respectively. and Figure 4-7): 31.2% of the REMD trajectory corresponds to Cluster 1, whereas 17.2% 104

PAGE 105

Table 4-7. Prin Cluer 1 Interactios Gln O-Ag7, Leu, O-g7, Pro O-Tr8, Tyr OHArg7Gln O-Tr8, Not Themain ciHeo st Cluster 2 Cluster 3 n Pnt ti e Interactions Percentage 4,rH .2 h-7, 2% Tyr8, OH-Arg7, (SC) >65% 2ArHC 5 ro-, 1% Gln4, O-Arg7, H 58.32% 5,yH .9 eO7) 0% Pro5, O-Tyr8, H 47.83% 8,-, (SC) (SC) 34.24% Leu2, O-Arg7, HH (SC) 26.28% 4,yH 26.48% 7% Gln6, O-NH, 9 22.35% e: ntiineen. The adal Tyr-Arg interaction occurring in peptide II is shown in red. pal B intractins in peptide II (Clusters 1, 2, and 3) erceage Ineractons 707% Pe3, OArg H H (S) >0% P5, OTyr8H 352% Lu2, -Arg, (SC H 32.73% Tyr8, OH-Arg7, H Gln4, O-Tyr8, H HB iteracons NPFF are in grdition Percentag80.367.6 >515.7 105

PAGE 106

106 AA CC 7 Le u2 Arg BB Tyr8Pro5 e3 Arg7 Gln4Tyr8 Ph A rg7 Gln4 By observing the HB patterns for each cluster, we arrive at three important conclusions: (1) All principal HB interactionslositioeven and eight, as in the case of NPFF, (2) The hydroxyl group on the tyrosine side chain induces conformational changesthat may be important for loss in affinrge side c (3) The presence of the hydroxyl group on the tyrosine decreased the flarom ring at position eight. Figure 4-7. The pntfor pde II: representative structures for A) Clust Bteer 3 In NPFF, arginine side chain HB interactions primarily involve the epsilon hydrogen (H) with carboxyl anrbxyeu2 and Gln4 (Table 4-3). Amide hydrogens at the end of the side chain aesipating in HBs, leaving the C(NH2)2 group free to exert its et (Figure 4-8A): we speculate that the positive charge at the end of the arginine side chain is needed for accessing and binding to the receptor. When Phe8 is substituted by a tyrosine, the side chain of Arg7 is involved in a rearrangement of HB interactions: the H on invoityalelusyl ore l ve to texi cor 2 ges fr reshe abilnfiandn aeq iduity gur Ctomuen es ininof ati) Cs tly at pthe ons luston L partic ns shainaticepti. threeer 1,d caffec rev) Con

PAGE 107

107 BB TTyyrr88 AArrgg77 Leeuu22 L AA GGllnn44 AArrgg77 ++ HH Arg7 is occupied by the presence of the OH (aromatic ring), and the amide hydrogens at the endof the arginine side chain interact with the Leu2 oxygen (Figure 4-8B). otice that the presence of the hydroxyl induces a dual effect on the arginine side chain: it (slightl Figure 4-8. Arginine side chain interactions in A) NPFF and B) peptide II. The rearrangement of interactions may be responsible for the lowering in affinity. N y) perturbs the side chain conformationthe Leu2-Arg7 side chain interaction still occurs frequently and at the same time it prevents binding to the receptor by indirectly restricting thepositive charge of the chain. Solvent accessible surface area (SASA) calculations show that theArg7 side chain becomes more hydrophobic when Phe8 is replaced by Tyr. In Figure 4-9 wepresent the SASA histograms for the arginine in the full trajectories of peptides I (NPFF) and II. The equally populated histograms present similar distributions with an average SASA value of

PAGE 108

155 2 for peptide I and 144 2 for peptide II. The increase in hydrophobicity induced by the OH group, may affect the affinity for the receptor. PPoo ppuu ll tt II oonn aa SSAASSAA // 2 2 PPeeppttiiddee IIII NNPPFFFF 115544..99114433..55 Figure 4-9. SASA histograms for the arginine residue in NPFF (red) and peptide II (black). The small increase in hydrophobicity possibly contributes in the lower affinity for peptide II. The reduced flexibility of the aromatic ring may be an additional reason for low affinity. In NPFF, the phenylalanine ring does not participate in any HB interactions and its position can be approximated via the HB participation of the Phe8 amide hydrogen (primarily Gln4,O-Phe8,H, Table 4-2). As we have already seen, the PheTyr mutation resulted in a great loss of flexibility for the aromatic ring, mainly because of the Tyr8,OH-Arg7 interaction that stabilizes the ring conformation. Gouardres et al. have demonstrated the importance of the phenylalanine ring to binding affinity by performing a series of mutations; for example, removal of the benzyl ring by 108

PAGE 109

109 substituting the phenylalanine with an alanine, reduced the affinity dramatically. Similar effects were observed when the benzyl ring was extended fur PPhhee88 A A r r g g77 A A BB TT y y r r88 A A r r g g77 Leu2 Leu2 ther from the backbone (by adding one extra CH2 group to the side chain), or when the side chain was shortened by one CH2 group. It appears that the phenylalanine ring needs to have a specific positioning and length to access an approriate recWe hypothesize that the replacement of phenylalanine with a tyrosine alters the orientation of the ring by forcing it to acquire a position which is dictated by the hydrogen bond interaction with the side chain of the arginine (Figure 4-10). In that case, the tyrosie ring adopts a conformation that does not allow it to access the (hydrophobic) site of the recephydroxyl group changes the orientation of the ring and decreases its flexibility. Relative-SASA calculations on the complete NPFF trajectory and for the C-terminal residues 4-8 indicate that Phe8 (followed by glutamine at position six) is the mostly solvent-exposed residue in the peptide (Figure 4-11). Relative-SASA values were obtained after dividing p eptor site. n tor. Figure 4-10. The aromatic ring A) in NPFF and B) in peptide II. The introduction of the

PAGE 110

110 the SASA values of each amino acid in the peptide by the corresponding SASA values of the amino acid itself (x100). 2030405060elatve SASA 010 GlnPro5Gln6Arg7Phe8ResidueRi 4 Figure 4-11. Relative-SASA calculations for residues 4-8 in NPFF. The phenylalanine is mainly solvent exposed. In order to estimate the actual degree of solvent accessibility for the phenylalanine and to further examine the environment around the aromatic ring, SASA calculations for the phenylalanine (Peptide I) and for the tyrosine (peptide II) were compared. SASA results for the complete trajectories (Figure 4-12) and their corresponding clusters (Figure 4-13), are in average SASA values for the phenylalanine and the tyrosine. Similarly to clustering consideratio agreement for the two different residues. The full trajectories present practically identical ns, we also compared the SASA histograms for each individual cluster, to extract useful information from the enormous amount of data provided by the complete simulation trajectories. In Figure 4-13 only Clusters 1 and 3 for each peptide are presented. In both cases,

PAGE 111

Cluster 2 displays a distribution very similar to Cluster 3 and for the sake of simplicity it has been omitted. Note that the average SASA values are very similar between clusters. For exam) has an average SASA value of 1862, which is very close to the average value of 1872 for Cluster 3 in peptide II. The height of each distribution is different between clusters (even though they have similar widths), since the cluster populations are different. All distributions are relatively narrow, suggesting that the average SASA values represent the majority of the members of each cluster. The only exception is Cluster 1 in NPFF (Figure 4-13A, in black), where a slightly broader distribution is observed. For a very broad distribution, the calculation of an average value is not very useful because significant populations may be found towards the ends of the distribution. Nevertheless, NPFF Cluster 1 has a width comparable to Cluster 1 in peptide II, therefore, we assume that there is no significant population with extreme SASA values to affect considerably the binding affinity. The similarity of the SASA resutides suggests that significant solvent exposure of the aromatic ring is not alone sufficient for high affinity for the receptor; it seems that a free and flexible ring properl ple, NPFF Cluster 3 (Figure 4-13A lts for the two pep y oriented in a hydrophilic environment is needed to access a hydrophobic pocket of the receptor. 111

PAGE 112

PPoo pp uu ll aa tt ii oo n n SSAASSAA // 2 2 NNPPFFFF PPeeppttiiddee IIII 117711..66 117722..22 PP oo pp uu ll aa tt ii oo n n SSAASSAA // 2 2 CClluusstteerr 11 CClluusstteerr 33 115500..33 118866..11 115588..99 118877..44 CClluusstteerr 11 CClluusstteerr 33 AA BB Figure 4-12. SASA histograms for the phenylalanine residue in NPFF (black) and for the tyrosine residue in peptide II (red). Full trajectories for both peptides were analyzed. Figure 4-13. SASA histograms for the phenylalanine residue in Cluster 1 and Cluster 3 of A) NPFF and B) peptide II. 112

PAGE 113

113 o far, it is observed that factors such as the charge and conformation of the arginine side chain, as well as th affinity to the NPFF2 recepwever, the degree each factor is responsible for lowering the affinity is still unclear. Additional amo acid subsy help to elucidate the situation. 4.3.3 C-terminal Substitutions: The Role of Arginine e nexocused our interest in position seven and the importance of the arginine. Phe-Leu-Phe-Gln-Pro-Gln-Lys-Phe-NH2 (Peptide III) differs by NPFF in position seven, where a lysine has substituted the arginine residue. This single mutation reduces more than 700-fold the affinity of Peptide III for the NPFF2 receptor (Table 4-1). Hydrogen bonding and clustering analyses on the REMD trajectory revealed only one HB pattern (Table 4-8). Interestingly enough, the hydrogen bonding pattern resembles the case of NPFF, with backbone interactions among residues 4, 7 and 8 being interactions. The backbone dihedral shown in Figures 4-14 and 4-15, respectively. Interactions Percentage G4, O75.69% Gln4,O Phe8, H 48.27% P5, O,-Phe8, H 21.72% Leu2, O-Lys7, H (SC) >20% Gln4,OE1 (SC) Lys7, H (SC) >20% S e flexibility of the phenylalanine ring are determinant for high tor. Ho in titutions ma W t f the on e in the principal angles as a function of time and the corresponding Ramachandran plots for peptide III, are Table 4-8. Principal HB interactions in peptide III (full trajectory) ln Lys7, H ro

PAGE 114

114 Amino aF8,K7,P5,Q4,F8,F3,P5,20208080140140200200 cid poK7,Q4,-180Time (ns)Time (ns) L2,F3,Q6,L2,Q6,-180180180 sition Figure 4-14. Backbone dihedral angles for each residue in peptide III as a function of the simulation time. Leu2Phe3Gln4Pro5Gln6Lys7Phe8180 180 -180 -180 Figure 4-15. Ramachandran plots for each residue in peptide III.

PAGE 115

The necessary for good affinity phenylalanine ring is not engaged in any interactions and similar to NPFF, it is restricted in a region determined by the Gln4,O-Phe8,H and Pro5,O-Phe8,H interactions. SASA calculations show a change in solvent accessibility (1632 instead of 1722 in NPFF) that probably plays a minor role in the loss of affinity. Thus, the great lowering in affinity must be attributed on the importance of the side chain on residue 7. The orientation of the lysine side chain in peptide III is mainly dictated by the interactions between one of the hydrogens at the end of the chain and the carbonyl oxygen on Leu2, or the carboxylate Gln4 (Table 4-8). These interactions suggest that, compared to the Arg side chain in NPFF, the conformational changes of the lysine side chain are negligible (Table 4-2). This statement is further supported by the average SASA values for the lysine (1522 instead of 154 2 for the arginine in NPFF). As we observed in the previous section, there is an indication that the positive charge at the end of the arginine side chain is crucial for high affinity: in NPFF, main interactions between the arginine side chain and other residues involve mostly the H, leaving the positively charged end of the chain free to approach the receptor; moreover, even though the amide hydrogens at the end of the side chain participate in HBs, the peptides affinity to the receptor remains high. For peptide III, the lysine side chain is also positively charged and seems to acquire an appropriate orientation for binding, however, the affinity for the receptor reduces greatly. A reasonable explanation for this could be the charge delocalization on the arginine side chain: even if one of the hydrogens at the end of the Arg side chain participates in a HB, there is always a positive charge which enables binding to the receptor. On the other hand, the positive charge on the NH3 115

PAGE 116

Arginine Lysine group at the end of the lysine side chain is very localized and therefore, if that part participates in a HBFigurside chain may be needed for high binding affinity. delocalization allows the Arg side chain to interact with receptor site(s) for most of the time. glutamine is replaced by a glutamic acid (Glu6), followed by a profound loss in binding affinity it is excluded from being a binding site to the receptor (Figure 4-16). e 4-16. Amino acids arginine and lysine. Charge delocalization at the end of the arginine After the second mutation study, we conclude that an appropriate orientation of the phenylalanine ring may be necessary but not sufficient for good receptor affinity and it must be combined with the effect of the positively charged arginine side chain. Additionally, charge Finally, it is important to mention that the deactivation of the Arg side chain induced a greater effect to the binding affinity than the substitution of the phenylalanine with a tyrosine. 4.3.4 C-terminal Substitutions: The Role of Glutamine Glutamine at position six: we proceeded with a mutation at position six, where the 116

PAGE 117

117 y the carboxyl function of the glutamic acid, resulting in an interaction between the carboxylate of the glutamic acid (Glu6,OE2) and the arTable 4-9. Principal HB interactions in peptide IV (full trajectory) Interactions Percentage Glu6, OE2 (SC) Arg7, H 32.95% Pro5, O-Phe8, H 32.12% Gln4, O-Arg7, H 29.64% Phe3, O-Arg7, H 27.52% n interesting feature is that peptide IV presents a very broad distribution of HB shown in Appendix B, Table B-2). This suggests that the replacement of the glutamine by a glutamsupported by the clustering results: clustering the REMD trajectory of peptide IV results in a conformations very frequently (more frequently than NPFF does), namely it is more flexible. or ation; it rather suggests that specific, well dor they by a factor of ~900 (Table 4-1, Peptide IV: Phe-Leu-Phe-Gln-Pro-Glu-Arg-Phe-NH2 ). The amide effect induced by the glutamine side chain in NPFF is now replaced b ginine (Table 4-9). A interactions, with 26 different HBs having frequencies between 32.95% and 10% (data are ic acid rendered the peptide much more mobile than before, a feature that is further multitude of almost evenly populated clusters. This result implies that the peptide changes It can be concluded that the Glu-Arg HB, resulted in a destabilization of the peptide by adding flexibility to it. It is important to clarify that the speculation of increased flexibility f peptide IV does not imply that NPFF adopts a rigid conform efined HB patterns (present for a sufficient amount of time) in NPFF, now last less have been replaced by others which were not very dominant before. This destabilization induces

PAGE 118

changes in the orientation of crucial for affinity groups, such as the Phe8 ring and the Arg7 schain. The other important conditions (free Phe8 ring, free and positively charged Arg side cend) are now satisfied, yet the binding affinity drops greatly. Thus, we conclude that certain orientation for the ring and the arginine side chain is also req ide hain uired for high affinity. hen the side chain of glutamine is shortened by one CH2 group we observe a similar but 4-1, Peptide V: Phe-Leu-Phe-2) reduces 21-fold the affinity. The principal backbone interaction hain also increased the ulting in a reduced affinity (Tustering also produced ent of the amide end (arginine side chain) with a carboxylate group, increased the degree of flexibility more than the shorteTable 4-10. Principal HB interactions in peptide V (full trajectory) Interactions Percentage Gln4, O-Arg7, H 42.98% Asn6, OD1 (SC)-Phe8, H 19.16% Glutamine at position four: the role of residue four is next investigated, by substituting the glutamine with a glutamic acid residue (Table 4-1, Peptide VI: Phe-Leu-Phe-Glu-Pro-Gln-Arg-Phe-NH2). This resulted in a 3.5 times loss in affinity, a very minor effect compared to the corresponding substitution at position six (Table 4-1). As shown in Table 4-11, the principal HB W minor effect: an asparagine residue instead of the glutamine (Table Gln-Pro-Asn-Arg-Phe-NH between Gln4 and Arg7 remains dominant, however, the asparagine side c flexibility of the molecule, resable 4-10). Cl sever al sparse clusters. However, as the HB analysis indicates, the replacem ning of the chain did. Leu2, O-Arg7 (SC) >35% Gln4, O-Phe8, H 18.21% 118

PAGE 119

backbone interactions between residues 4, 7 and 8 coincide with the ones for NPFF. The carboxylate of the glutamic acid slightly increases the flexibility of the molecule by participating in HB interactions with the arginine side chain, however this does not affect significantly thebinding affinity. The clustering results for Peptide VI yielded three major clusters (A ppendix B, Table Peptide Gln4 NPFF (Glu/Gln)4, O-Arg7, H 59.46% 67.26% Glu4, OE (SC)-Arg7, HH (SC) ~40% 0 ontrary to position six, it can be seen that the amide function at position four did not duce substantial conformational changes and probably it is not very important for receptor ation of hydrogen bonds ith the Arg7 and Phe8 to provide appropriate orientations for the side chains to approach titutions As already mentioned, N-terminal modifications do not significantly alter the affinity. Shortening the sequence by deleting residues from the N-terminal part produced only slight decrease in binding affinity until the Gln4 is removed. Pro-Gln-Arg-Phe-NH2 (peptide VII) causes a 45-fold loss in binding affinity (Table 4-1). The HB interactions involved in that system are displayed in Table 4-12. The average HB representation of the molecule can be divided into s B-3, B-4 and B-5). Table 4-11. Principal HB interactions in peptide VI (full trajectory) Interactions Percentage (Glu/Gln)4, O-Phe8, H 52.20% 46.71% C in a ffinity. The importance of residue four is rather attributed to the form w specific receptor sites. 4.3.5 N-terminal Subs 119

PAGE 120

two different clusters almost evenly populated (Cluster 1: 51.1% and Cluster 2: 48.9%). Tclustering results are shown in Appendix B (Tables B-6 and B-7). Table 4-12. Principal HB interactions in peptide VII (full trajectory) Interactions he Percentage Gln2, O-Arg3, H (SC) 36.48% Gln2, OE1 (SC)-Arg3, H 23.21% .15% H (in he smaller size of peptide guarantees a great degree of mobility and an obvious HB rearrad ains g sites of the receptor. Thus, even though we have concluded that the l (Gln-Arg-Phe-NH2) is mainly responsible for high affinity, it is notide lower affinity for NPFF receptors. We observed that high binding affinity is primarily Gln2, OE1 (SC)-Arg3, H (SC) 19 Pro1, O-Phe4, H 17.74% Gln2, OE1 (SC)-Phe4 14.18% Pro1, O-Arg3, H 10.41% Note: the backbone interactions involving the C-terminal arginine and phenylalanine residues red) have been rearranged after the removal of the glutamine4. T ngement: the absence of Gln4 allows the phenylalanine and the arginine to interact mostly with the proline and with the carboxylate on Gln. Consequently, the orientations of the ring anthe Arg7 side chain are not determined by the backbone Gln4-Arg7 and Gln4-Phe8 interactions any more. This finding further indicates the importance of Gln4 to orientate the two side chtowards specific bindin four-residue end of the C-termina alone sufficient. 4.4 Conclusions The structural features that determine the high binding affinity of the opioid neuropeptNPFF towards specific GPCRs in rat were investigated. REMD methodologies were applied to NPFF and to a series of analogues (with modified C-terminal sequence) that present substantially 120

PAGE 121

determined by specific interactions among Gln4, Arg7 and Phe8. These interactions enable appropriate orientation of the side chains of Arg7 and Phe8 to acce ss and bind to the receptor. hen the Phe8 is substituted by a tyrosine, the hydroxyl on the side chain of Tyr8 alters the lvent. At the same time, the onformith the Arg7. Thus, educed greatly. We ic ring that does noty interactions must a receptor site. Additionally, an aubstitution at position arge on guarantee binding to the recepeeds to be combined with the positive charge effect at the end of the chain. Moreptor mation that enables the side chains of Arg7 and Phe8 to appropriately orientate for binding tomportance of position 4 in backbone stabil W orientation of the Arg8 side chain by rendering it less accessible to so aromatic ring acquires a cation which is dictated by the interaction w even though the ring remains mostly exposed to the solvent, its flexibility r concluded that a highly flexible aromat participate in an be properly oriented to access mino acid s 7 where a lycine has replaced the arginine, indicated the importance of the positive ch Arg7: a proper orientation of the Arg7 side chain alone does not tor; it rather n over, charge delocalization on the Arg7 side chain increases the likelihood of Arg7-receinteraction. Additional substitutions at position six, either by replacing the amide effect of Gln6 with the carboxyl function of Glu6, or by shortening the Gln6 side chain after introducing an asparagine, produced an increase in flexibility of the molecule that resulted in a profound lowering in affinity. The replacement of Gln4 by a Glu demonstrated the role of position 4. The necessary for high affinity backbone interactions between residue 4 and (Arg7/Phe8) are preserved, resulting in a minor decrease in binding affinity. The role of Gln4 is to interact with Arg7 and Phe8 to form a backbone confor the receptor. The i ization was further supported after the significant loss in affinity observed for the N-terminal truncated peptide VII. When the Arg4 is absent, the peptide adopts random configurations that may inhibit binding to the receptor. 121

PAGE 122

CHAPTER 5 AN OVERALL VIEW In order to understand the biological basis of consciousness and the mental processes by which our behavior is dictated, a unified approach of the study of behavior and neural science becomes essential. One step towards that direction would be the investigation of how neurons are organized into signaling pathways in the brain and how they communicate. Information from one neuron is conveyed to another neuron across a synapse. Chemical transmission occurs when a neurotransmitter molecule is released from one neural cell and next binds with receptor sites on another neuron. The animal nervous system employs many chemical neurotransmitters to relay messages from one neuron to the next. FLP neuropeptides are the largest and most diverse family of animal neuropeptides known. They are particularly interesting because they are present in almost every organism and are involved in a wide range of neural functions, including learning, feeding processes, pain modulation and memory. These neuropeptides usually have highly mobile structures, a feature that could be explored theoretically only, since experimental approaches provide average results over time and over many molecules. Replica-Exchange Molecular Dynamics (REMD) method is one of the most reliable computational techniques to study complex systems such as proteins and peptides. Our goal was to theoretically investigate the relationship between the structure of FLPs and their activity against specific receptors in the brain. By exploring the conformational diversity of several peptides, we assigned particular structure-activity relationships between FLP neuropeptides and their corresponding receptors that may be important in understanding a variety of biological processes. 122

PAGE 123

In particular, we performed REMD a series of peptides to account for the different potencies of DFDGAMPGVLRF-NH and EMPGVLRF-NH against the GPCR NPR-1, in the nematode C. elegans. DFDGAMPGVLRF-NHadopts mainly two conformations, forming either a C-terminal loop, or a combined bicyclic structure via N-terminal with C-terminal loop interactions. There was a strong indication that the bicyclic region of DFDGAMPGVLRF-NH is responsible for its lower potency on NPR-1. By performing hydrogen bonding, clustering and NMR calculations, we concluded that the DFDGAM region behaves as an independent unit, and the common C-terminal amino acid sequence, PGVLRF-NH2 is a unique conformation, present in all peptides investigated. Additionally, conformational features of the mammalian neuropeptide NPFF (Phe-Leu-Phe-Gln-Pro-Gln-Arg-Phe-NH) were associated with its high binding affinity for the NPFF2 receptor. Hydrogen bonding, clustering and solvent accessible surface area calculations were carried out for NPFF, as well as for a series of NPFF analogues. It was concluded that a precise length and positioning of a free C-terminal aromatic ring along with, the presence of a flexible and positively charged side chain at position 7 were needed for high binding to the receptor. Appropriate orientations for these side chains are obtained through backbone interactions among residues at positions 4, 7 and 8. The significance of these studies lies on the fact that our findings may provide important insight as to the structural diversity and conformational changes that dictate the biological activity of FLPs. simulations on 222 2 2 123

PAGE 124

APPENDIX A 445.8K, 492.2K, 543.5K and 600.0K. 460.1K, 501.2K, 545.9K, 594.7K, 647.8K, 705.6K, 768.6K and 837.3K. 428.8K, 468.9K, 512.7K, 560.6K, 612.9K, 670.2K, 732.8K and 801.3K. 457.4K, 497.7K, 541.5K, 589.2K, 641.1K, 697.5K, 758.9K and 825.7K. 528.8K, 592.2K, 663.3K, 743.0K and 832.1K. 645.9K, 734.0K and 834.0K. SUPPLEMENTAL INFORMATION ON PROJECT I The REMD temperatures used for each peptide are shown below: EMPGVLRF-NH2 (10 replicas): 246.1K, 271.7K, 300.0K, 331.2K, 365.7K, 403.8K, DFDGAMPGVLRF-NH2 (14 replicas): 275.4K, 300.0K, 326.8K, 356.0K, 387.8K, 422.4K, SGSGAMPGVLRF-NH2 (14 replicas): 250.9K, 274.4K, 300.0K, 328.0K, 358.7K, 392.2K, DFDGEMPGVLRF-NH2 (14 replicas): 275.7K, 300.0K, 326.4K, 355.1K, 386.4K, 420.4K, PGVLRF-NH2 (12 replicas): 239.1K, 267.8K, 300.0K, 336.0K, 376.3K, 421.5K, 472.1K, DFDGAM-NH2 (10 replicas): 264.0K, 300.0K, 340.9K, 387.4K, 440.2K, 500.2K, 568.4K, Table form of peptide III. Interactions Percentage Pro7, O-Leu10, H 76.2% Gly8,O Phe12, H 58.6% Asp3,O Met6, H 38.5% deprotonated form and the fully protonated form of the peptide. A-1. Principal HB interactions in D(H)FDGAMPGVLRF-NH2, the partially protonated Pro7, O,-Arg11, H 63.0% Phe2,O Ala5, H 52.0% Ala5,O Gly8, H 33.4% Note: The peptide adopts an average conformation, which is an intermediate between the 124

PAGE 125

APPENDIX B SUPPLEMENTAL INFORMATION ON PROJECT II 58.2K. Hybrid-REMD temperatures for NPFF (12 replicas): 282.8K, 300.0K, 318.2K, 337.5K, 358.0K, nterac The REMD temperatures used for each peptide are shown below: Peptide I (12 replicas): 273.4K, 300.0K, 329.1K, 361.1K, 396.2K, 434.7K, 476.9K, 523.3K, 574.1K, 629.9K, 691.1K and7 Peptide II (12 replicas): 273.5K, 300.0K, 329.0K, 360.9K, 395.9K, 434.2K, 476.2K, 522.3K, 572.9K, 628.4K, 689.2K and 756.0K. Peptides III, IV and VI (12 replicas): 273.3K, 300.0K, 329.3K, 361.6K, 396.9K, 435.8K, 478.4K, 525.2K, 576.6K, 633.0K, 694.9K and 762.9K. Peptide V (12 replicas): 273.2K, 300.0K, 329.4K, 361.8K, 397.3K, 436.3K, 479.1K, 526.2K, 577.8K, 634.5K, 696.8K and 765.2K. Peptide VII (8 replicas): 263.4K, 300.0K, 341.7K, 389.3K, 443.4K, 505.1K, 575.4K and 655.4K. 379.7K, 402.7K, 427.1K, 453.0K, 480.5K, 509.7K, and 540.6K. Table B-1. Principle HB itions in NPFF (Cluster 1) Interactions Percentage Gln4, O-Arg7, H 84.50% Gln4, O-Phe8, H 55.90% Gln4,OE1(SC)-Arg7, H (SC) 49.46% Leu2, O-Arg7, HH12 (SC) 26.68% Gln4,OE1(SC)-Arg7, HH11 (SC) 27.14% Note: Main interactions resemble the full trajectory. 125

PAGE 126

Table I G 32.12% G 27.52% P 23.56% G, OE1 (SC) Leu2, H 21.26% Glu6, OE1 (SC) Arg7, H (SC) 21.09% Phe1, O-Gln4, H 20.14% .47% Glu6, OE1 (SC) Phe3, H 17.02% Glu6, OE1 (SC) Arg7, HH (SC) 16.14% 15.03% Pro5, O-9HN2 13.56% 10.17% Leu2, O-Arg7, H (SC) 10.09% ote: A broad distribution is observed, with 26 different HB interactions having frequencies bove 10%. B-2. Principal HB interactions in peptide IV (full trajectory) nteractions Percentage lu6, OE2 (SC) Arg7, H 32.95% Pro5, O-Phe8, H ln4, O-Arg7, H 29.64% Phe3, O-Arg7, H he1, O-Phe3, H 25.40% Gln4, O-Phe8, H lu6, OE2 (SC) Arg7, H 22.12% Glu6 Gln4, OE1 (SC) Arg7, HH (SC) 18.78% Glu6, OE2 (SC) Arg7, HH (SC) 18 Glu6, OE1 (SC) Phe1, H1 Phe3, O-Arg7, HH (SC) 14.17% Gln4, OE1 (SC) Arg7, HH2 (SC) 13.59% Glu6, OE2 (SC) Arg7, HH2 (SC) 13.29% Phe3, O-Glu6, H 12.90% Leu2, O-Arg7, HH (SC) 12.04% Phe3, O-Gln4, H (SC) 11.97% Glu6, OE2 (SC) Phe1, H1 10.67% Gln4, OE1 (SC) Arg7, HH (SC) 10.38% Gln6, OE2 (SC) Phe3, H N a 126

PAGE 127

Table B-3. Principle HB interactions in peptide VI (Cluster 1) Pro5,O-Phe8, H 74.15% HH ctions in peptide VI (Cluster 2) Pe B interactions in peptide VI (Cluster 3) Pe Interactions Percentage Phe1,O-Phe3, H 46.48% Glu4,OE2 (SC)-Arg7, 42.00% Glu4,O-Arg7, H 28.77% Table B-4. Principle HB intera Interactions rcentage Glu4,O-Arg7, H 95.79% Glu4,O-Phe8, H 82.24% Glu4,OE (SC)-Arg7, H (SC) 55.00% Glu4,OE (SC)-Phe3, H 39.15% Table B-5. Principle H Interactions rcentage Phe3,O-Arg7, H 76.64% Phe3,O-Gln6, H 74.23% Glu4,O-Phe8, H 67.38% Glu4,OE(SC)-Arg7, HH11 (SC) >55% Glu4,O-Arg7, H 43.40% 127

PAGE 128

Table B-6. Principle HB interactions in peptide VII (Cluster 1) Pro1, O-Phe4, H 34.74% SC) interac Gln2,O-Arg3, HH (SC) >20% Phe4, O-Gln2, H (SC) 11.30% Interactions Percentage Gln2,OE1(SC)-Arg3, H (SC) 33.23% Gln2,O-Arg3, H ( 20.18% Pro1, O-Arg3, H 19.39% Gln2,OE1(SC)-Arg3, HH (SC) >20% G ln2,OE1(SC)-Arg3, H 14.05% Table B-7. Principle HBtions in peptide VI (Cluster 2) Interactions Percentage Gln2,O-Arg3, H (SC) 53.52% Gln2,OE1-Arg3, H 32.81% G ln2,OE1-Phe4, H 29.45% 128

PAGE 129

LIST OF REFERENCES fourth ed.; ., Bloom, F.E. and Roth, R.H. Tal Basis of ed.; Oxford University Press: New 5. (4) Edwards, R. H. Curr Opin Neurobiol 1992, 2, 586. (5) Unwin, N. Cell 1993, 72 Suppl, 31. E. Proc Natl Acad Sci U S A 985, 82, 7757. T.; Reale, V.; Chatwin, H.; Zachariah, C.; deBono, M.; Evans, P. D.; 5, 7586. Parasitol Today 996, 12, 343. (10) Greenberg, M. J.; Price, D. A. Prog Brain Res 1992, 92, 25. (11) Li, C.; Kim, K.; Nelson, L. S. Brain Res 1999, 848, 26. (12) Li, C.; Nelson, L. S.; Kim, K.; Nathoo, A.; Hart, A. C. Ann N Y Acad Sci 1999, 897, 239. (13) Maule, A. G.; Mousley, A.; Marks, N. J.; Day, T. A.; Thompson, D. P.; Geary, T. G.; Halton, D. W. Curr Top Med Chem 2002, 2, 733. (14) McVeigh, P.; Geary, T. G.; Marks, N. J.; Maule, A. G. Trends Parasitol 2006, 22, 385. (15) McVeigh, P.; Leech, S.; Mair, G. R.; Marks, N. J.; Geary, T. G.; Maule, A. G. Int J Parasitol 2005, 35, 1043. (16) Mercier, A. J.; Friedrich, R.; Boldt, M. Microsc Res Tech 2003, 60, 313. (17) Price, D. A.; Greenberg, M. J. Science 1977, 197, 670. (1) Kandel, E., Schwartz, J., and Jessell, T. Principles of Neural Science McGraw-Hill, 2000. (2) Cooper, J. Rhe Biochemic Neuropharmacology, 7th York, 1996. (3) Fisher, J. M.; Scheller, R. H. J Biol Chem 1988, 263, 1651 (6) Yang, H. Y., Fratta, W., Majane, E.A. and Costa 1 (7) Mazarguil, H.; Gouarderes, C.; Tafani, J. A. M.; Marcus, D.; Kotani, M.; Mollereau, C.; Roumy, M.; Zajac, J. M. Peptides 2001, 22, 1471. (8) Dossey, A Edison, A. S. Biochemistry 2006, 4 (9) Brownlee, D. J.; Fairweather, I.; Holden-Dye, L.; Walker, R. J. 1 129

PAGE 130

(18) Brownlee, D. J.; Fairi 1999, 22, 16. (19) Cowden, C.; Stretton, A. O. Peptides 1995, 16, 491. (21) Dockray, G. J. Exp Physiol 2004, 89, 229. lair, K. L.; Halton, D. .; Shaw, C.; Thompson, D. P. Invert Neurosci 1995, 1, 255. (23) Moffett, C. L.; Beckett, A. M.; Mousley, A.; Geary, T. G.; Marks, N. J.; Halton, W.;mp9. 1998, 281, 1686. 65, 304. m (28) Perry, S. J.; Huang, E. Y. K.; Cronk, D.; Bagust, J.; Sharma, R.; Walker, R. J.; ilsonBu(29) Perry, S. J.; Straub, V. A.; Schofield, M. G.; Burke, J. F.; Benjamin, P. R. J 1n, M.; Karlstedt, K.; ontinen, V. K.; Kalso, E.; States, B.; Panula, P.; Ziff, E. Mol Pharmacol 1999, 55, 804. ) Espinoza, E.; Carrigan, M.; Thomas, S. G.; Shaw, G.; Edison, A. S. Mol eurobiol 2000, 21, 35. C. J Neurosci 1999, 19, 6318. (34) Li, C. Parasitology 2005, 131 Suppl, S109. utz, K. K.; Dikler, S.; Messinger, L.; Li, L.; Stretton, A. O. J Comp eurol 2005, 488, 396. weather, I. Trends Neurosc (20)Davis, R. E.; Stretton, A. O. Peptides 2001, 22, 7. (22) Maule, A. G.; Geary, T. G.; Bowman, J. W.; Marks, N. J.; B W D Thoson, D. P.; Maule, A. G. Int J Parasitol 2003, 33, 85 (24) Nelson, L. S.; Rosoff, M. L.; Li, C. Science (25)Papaioannou, S.; Marsden, D.; Franks, C. J.; Walker, R. J.; Holden-Dye, L. J Neurobiol 2005, (26)Reinitz, C. A.; Herfel, H. G.; Messinger, L. A.; Stretton, A. O. Mol BiocheParasitol 2000, 111, 185. (27)Kivipelto, L.; Panula, P. Eur J Neurosci 1991, 3, 175. W, S.; rke, J. F. Febs Letters 1997, 409, 426. Neurosci 200, 21, 5559. (30) Vilim, F. S.; Aarnisalo, A. A.; Nieminen, M. L.; Lintune K (31 N (32) Edison, A. S.; Espinoza, E.; Zachariah, (33) Wilmot, C. M.; Thornton, J. M. J Mol Biol 1988, 203, 221. (35) Yew, J. Y.; K N 130

PAGE 131

(36) Zachariah, C.; Cameron, A.; Lindberg, I.; Kao, K. J.; Beinfeld, M C.; Edison, A. Biochemistry 2001, 40, 8790. (37) Rosoff, M. L.; Doble, K. E.; Price, D. A.; Li, C. Peptides 1993, 14, 331. (38) Husson, S. J.; Clynen, E.; Baggerman, G.; De Loof, A.; Schoofs, L. Biochem hem Biophys Res Commun 999, 254, 222. (41) Marks, N. J.; Shaw, C.; Halton, D. W.; Thompson, D. P.; Geary, T. G.; Li, C.; aule, B(42) Marks, N. J.; Shaw, C.; Maule, A. G.; Davis, J. P.; Halton, D. W.; Verhaert, P.; (43) Steiner, D. F. Curr Opin Chem Biol 1998, 2, 31. (44) Schinkmann, K.; Li, C. J Comp Neurol 1992, 316, 251. J eurochem 2004, 90, 129. C. J Comp Neurol 2004, 475, 540. (48) Schmitz, D.; Mellor, J.; Nicoll, R. A. Science 2001, 291, 1972. D. R.; Wu, Y. M.; Lin, S. F. Oncogene 2000, 19, 5548. aron, M. .; Taghert, P. H. J Biol Chem 2003, 278, 52172. (51) Kubiak, T. M.; Larsen, M. J.; Nulf, S. C.; Zantello, M. R.; Burton, K. J.; 03, 278, 33724. em 2003, 278, 42115. S Biophys Res Commun 2005, 335, 76. (39) Marks, N. J.; Maule, A. G.; Geary, T. G.; Thompson, D. P.; Li, C.; Halton, D. W.; Shaw, C. Biochem Biophys Res Commun 1998, 248, 422. (40) Marks, N. J.; Maule, A. G.; Li, C.; Nelson, L. S.; Thompson, D. P.; Alexander-Bowman, S.; Geary, T. G.; Halton, D. W.; Verhaert, P.; Shaw, C. Bioc 1 M A. G.iochem Biophys Res Commun 2001, 286, 1170. Geary, T. G.; Thompson, D. P. Biochem Biophys Res Commun 1995, 217, 845. (45) Han, M.; Park, D.; Vanderzalm, P. J.; Mains, R. E.; Eipper, B. A.; Taghert, P. H. N (46) Kim, K.; Li (47) Connolly, C. N.; Wafford, K. A. Biochem Soc Trans 2004, 32, 529. (49) Robinson, (50) Johnson, E. C.; Bohn, L. M.; Barak, L. S.; Birse, R. T.; Nassel, D. R.; C G Bowman, J. W.; Modric, T.; Lowery, D. E. J Biol Chem 20 (52) Kubiak, T. M.; Larsen, M. J.; Zantello, M. R.; Bowman, J. W.; Nulf, S. C.; Lowery, D. E. J Biol Ch 131

PAGE 132

(53) Mertens, I.; Meeusen, T.; Janssen, T.; Nachman, R.; Schoofs, L. Biochem BiophysRes Commun 2005, 330, 967. hman, R. ; De Loof, A.; Schoofs, L. FEBS Lett 2004, 573, 55. im, K.; Chatwin, H.; Li, C.; Evans, P.; de Bono, M. Nat eurosci 2003, 6, 1178. M. O. Mol Pharmacol 1996, 49, 7. oyer, D.; Clarke, D. E.; Fozard, J. R.; Hartig, P. R.; Martin, G. R.; Mylecharane, J.; Saxena, P. R.; Humphrey, P. P. Pharmacol Rev 1994, 46, 157. C. A.; Motoshima, H.; Fox, B. .; Le Trong, I.; Teller, D. C.; Okada, T.; Stenkamp, R. E.; Yamamoto, M.; Miyano, M. Science 000, 2(60) Rasmussen, S. G.; Choi, H. J.; Rosenbaum, D. M.; Kobilka, T. S.; Thian, F. S.; Fischetti, R. F.; Schertler, G. F.; eis, W. I.; Kobilka, B. K. Nature 2007, 450, 383. (61) Cherezov, V.; Rosenbaum, D. M.; Hanson, M. A.; Rasmussen, S. G.; Thian, F. S.; .; Rasmussen, S. G.; Thian, F. S.; obilka, T. S.; Choi, H. J.; Yao, X. J.; Weis, W. I.; Stevens, R. C.; Kobilka, B. K. Science 2007, 18, 12Nishiuchi, Y.; imura, T.; Ito, S. Nature 1998, 392, 286. (64) Corbett, A. D.; Henderson, G.; McKnight, A. T.; Paterson, S. J. Br J Pharmacol 006, 1p(65) Kenakin, T. Nat Rev Drug Discov 2002, 1, 103. (68) Wyman, J. J Mol Biol 1965, 11, 631. (54) Mertens, I.; Vandingenen, A.; Meeusen, T.; Janssen, T.; Luyten, W.; Nac J. (55) Rogers, C.; Reale, V.; K N (56) Nisoli, E.; Tonello, C.; Landi, M.; Carruba, (57) Chen, K.; Li, H. Z.; Ye, N.; Zhang, J.; Wang, J. J. Brain Res Bull 2005, 67, 310. (58) H E (59) Palczewski, K.; Kumasaka, T.; Hori, T.; Behnke, A 289, 739. Edwards, P. C.; Burghammer, M.; Ratnala, V. R.; Sanishvili, R.; W Kobilka, T. S.; Choi, H. J.; Kuhn, P.; Weis, W. I.; Kobilka, B. K.; Stevens, R. C. Science 2007, 318, 1258. (62) Rosenbaum, D. M.; Cherezov, V.; Hanson, M. A K 366. (63) Okuda-Ashitaka, E.; Minami, T.; Tachibana, S.; Yoshihara, Y.; K 247 Supl 1, S153. (66) Stephenson, R. P. Br J Pharmacol Chemother 1956, 11, 379. (67) Furchgott, R. F. Advances in Drug Research; Academic: London/New York1966; Vol. 3. 132

PAGE 133

(69) Weber, G. Biochemistry 1972, 11, 864. (70) Rodbell, M. Nature 1980, 284, 17. 56, 615. (73) Della Rocca, G. J.; Maudsley, S.; Daaka, Y.; Lefkowitz, R. J.; Luttrell, L. M. J (77) Amara, A.; Gall, S. L.; Schwartz, O.; Salamero, J.; Montes, M.; Loetscher, P.; Cihak, J.; Simmons, G.; Clapham, P. R.; ignoret, N.; Marsh, M.; Stangassinger, M.; Borlat, F.; Wells, T. N.; Schlondorff, D.; Proudfoot, E. Je Havlin, R. H.; Tycko, R. Proc Natl Acad Sci U S A 2005, 102, 3284. (81) Andrew, E. R.; Bradbury, A.; Eades, R. G. Nature 1959, 183, 1802. ers, M. F.; Adams, M. W.; Park, J. B.; Zhou, Z. H.; ax, A. J Biomol NMR 1992, 2, 527. hysical Chemistry 2004, 107, 213. 267, 249. (87) Lyubartsev, A. P.; Martsinovski, A. A.; Shevkunov, S. V.; Vorontsovvelyaminov, (71) Gilman, A. G. Annu Rev Biochem 1987, (72) Bourne, H. R. Curr Opin Cell Biol 1997, 9, 134. Biol Chem 1999, 274, 13978. (74) Gether, U.; Lin, S.; Kobilka, B. K. J Biol Chem 1995, 270, 28268. (75) Roettger, B. F.; Ghanekar, D.; Rao, R.; Toledo, C.; Yingling, J.; Pinon, D.; Miller, L. J. Mol Pharmacol 1997, 51, 357. (76) Cocchi, F.; DeVico, A. L.; Garzino-Demo, A.; Cara, A.; Gallo, R. C.; Lusso, P. Nat Med 1996, 2, 1244. Baggiolini, M.; Virelizier, J. L.; Arenzana-Seisdedos, F. J Exp Med 1997, 186, 139. (78) Mack, M.; Luckow, B.; Nelson, P. J.; S A Exp Md 1998, 187, 1215. (79) (80) Laws, D. D.; Bitter, H. M. L.; Jerschow, A. Angewandte Chemie-International Edition 2002, 41, 3096. (82) Blake, P. R.; Lee, B.; Summ B (83) Tai, K. Biop (84) Berg, B. A.; Neuhaus, T. Physics Letters B 1991, (85) Berg, B. A.; Neuhaus, T. Physical Review Letters 1992, 68, 9. (86) Mezei, M. Journal of Computational Physics 1987, 68, 237. P. N. Journal of Chemical Physics 1992, 96, 1776. 133

PAGE 134

(88) Marinari, E., and Parisi, G. Europhys. Le tt 1992, 19. hysical Review Letters 1995, 74, 2151. hysics 1999, 110, 1754. B-53. H.; Haasnoot, G. Acta Crystallographica Section B-Structural Science 2004, 60, 528. J. J. Journal of Chemical Physics 2000, 113, 1276. (96) Katzgraber, H. G. Computer Physics Communications 2002, 147, 439. Y.; Nishikawa, T.; Okamoto, Y. Chemical Physics Letters 001, 333, 199. (100) Shin, S.; Son, W. J.; Jang, S. Journal of Molecular Structure-Theochem 2004, 73, 10(101) Habeck, M.; Nilges, M.; Rieping, W. Physical Review Letters 2005, 94. (102) Merlitz, H.; Wenzel, W. Chemical Physics Letters 2002, 362, 271. onal Statistics & Data Analysis 2004, 46, 441. 281, 140. 141. G. Journal of tatistical Physics 1996, 82, 155. (107) Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, A. H.; Teller, E. he (89) Hesselbo, B.; Stinchcombe, R. B. P (90) Tsallis, C. Journal of Statistical Physics 1988, 52, 479. (91) Falcioni, M.; Deem, M. W. Journal of Chemical P (92) Zavalij, P. Y.; Yang, S. F.; Whittingham, M. S. Acta Crystallographica Section Structural Science 2003, 59, 7 (93) Dova, E.; Peschar, R.; Sakata, M.; Kato, K.; Stassen, A. F.; Schenk J. (94) Yan, Q. L.; de Pablo, (95) Bedrov, D.; Smith, G. D. Journal of Chemical Physics 2001, 115, 1121. (97) Katzgraber, H. G.; Young, A. P. Physical Review B 2002, 65. (98) Katzgraber, H. G.; Young, A. P. Physical Review B 2002, 65. (99) Ishikawa, Y.; Sugita 2 69. (103) Kozumi, H. Computati (104) Hansmann, U. H. E. Chemical Physics Letters 1997 (105)Sugita, Y.; Okamoto, Y. Chemical Physics Letters 1999, 314 (106)Tesi, M. C.; vanRensburg, E. J. J.; Orlandini, E.; Whittington, S S Journal of Cmical Physics 1953, 21, 1087. 134

PAGE 135

(108)Feig, M.; Karanicolas, J.; Brooks, C. L. Journal of Mo lecular Graphics & odelling 2004, 22, 377. (109) Karanicolas, J.; Brooks, C. L. Proceedings of the National Academy of Sciences (112) Gront, D.; Kolinski, A.; Skolnick, J. Journal of Chemical Physics 2000, 113, 065. (113) Rathore, N.; Chopra, M.; de Pablo, J. J. Journal of Chemical Physics 2005, 122. (114) Fukunishi, H.; Watanabe, O.; Takada, S. Journal of Chemical Physics 2002, 116, 058. Physics 1993, 98, 798. (117) Alper, H. E.; Bassolinoklimas, D.; Stouch, T. R. Journal of Chemical Physics 993, 9(118) Beglov, D.; Roux, B. Biopolymers 1995, 35, 171. 9) Beglov, D.; Roux, B. Journal of Chemical Physics 1994, 100, 9050. on and ioinformatics 2004, 55, 493. (121) Okur, A.; Wickstrom, L.; Layten, M. M.; Song, K.; Hornak, V.; Simmerling, C. g, W.; Yang, R.; ieplak, P.; Luo, R.; Lee, T.; Caldwell, J.; Wang, J.; Kollman, P. J Comput Chem 2003, 24, 999. 01. M of the United States of America 2003, 100, 3954. (110)Pitera, J. W.; Swope, W. Proc Natl Acad Sci U S A 2003, 100, 7587. (111)Irback, A.; Sandelin, E. Journal of Chemical Physics 1999, 110, 12256. 5 9 (115) Kofke, D. A. Journal of Chemical Physics 2002, 117, 6911. (116)Alper, H. E.; Bassolino, D.; Stouch, T. R. Journal of Chemical 9 19, 5547. (11 (120) Kentsis, A.; Mezei, M.; Gindin, T.; Osman, R. Proteins-Structure Functi B Abstracts of Papers of the American Chemical Society 2006, 231. (122) Duan, Y.; Wu, C.; Chowdhury, S.; Lee, M. C.; Xiong, G.; Zhan C 1 (123) Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; BerendsenH. J. J Comput Chem 2005, 26, 17 (124) Schuler, L. D.; Daura, X.; Van Gunsteren, W. F. Journal of Computational Chemistry 2001, 22, 1205. 135

PAGE 136

(125) Warshel, A.; Levitt, M. Journal of Molecular Biology 1976, 103, 227. (126) Jorgensen, W. L.; Maxwell, D. S.; TiradoRives, J. Journal of the American (127) Jorgensen, W. L.; Tiradorives, J. Journal of the American Chemical Society 1988, 10, 16an, S.; arplus, M. Journal of Computational Chemistry 1983, 4, 187. (129) Straatsma, T. P.; Mccammon, J. A. Journal of Chemical Physics 1991, 95, 1175. s 2003, 119, 5740. 002, 23, 548. 71. Methods for Analysis of Terahertz Spectral Signatures of Nucleic Acids; 5th EE Conference on Nanotechnology, 2005, Nagoya, Japan. (135) Yoda, T.; Sugita, Y.; Okamoto, Y. Chemical Physics Letters 2004, 386, 460. (136) Todorova, N.; Legge, F. S.; Treutlein, H.; Yarovsky, I. J Phys Chem B 2008, 112, (137) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. (138) Downs, G. M.; Barnard, J. M. Reviews in Computational Chemistry, Vol 18 2002, (139) Cornilescu, G.; Delaglio, F.; Bax, A. J Biomol NMR 1999, 13, 289. (141) Wishart, D. S.; Sykes, B. D.; Richards, F. M. J Mol Biol 1991, 222, 311. .; Watson, M. S.; Boyko, R. F.; Sykes, B. D. J Biomol NMR 1997, 0, 329. Chemical Society 1996, 118, 11225. 157. (128) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.; Swaminath K (130) Gu, W.; Rahi, S. J.; Helms, V. Journal of Physical Chemistry B 2004, 108, 5806. (131) Shirts, M. R.; Pitera, J. W.; Swope, W. C.; Pande, V. S. Journal of Chemical Physic (132) Villa, A.; Mark, A. E. Journal of Computational Chemistry 2 (133) Jayaram, B.; Sprous, D.; Beveridge, D. L. Journal of Physical Chemistry B 1998, 102, 95 (134) Li, X. B., A.; Gelmont, B.; Globus, T.; Woolard, D.; Bykhovskaia, M. Computational IE 11137. Proteins-Structure Function and Bioinformatics 2006, 65, 712. 1 (140) Iwadate, M.; Asakura, T.; Williamson, M. P. J Biomol NMR 1999, 13, 199. (142) Wishart, D. S 1 136

PAGE 137

(143) Ando, S.; Xue, X. H.; Tibbits, G. F.; Haunerland, N. H. Comp Biochem Physiol B iochem Mol Biol 1998, 119, 213. n, J. G.; Oldfield, E. Science 1993, 260, 1491. (150) Carlacci, L.; Edison, A. S. Proteins 2000, 40, 367. (151) Geary, T. G.; Kubiak, T. M. Trends Pharmacol Sci 2005, 26, 56. ) Rosoff, M. L.; Burglin, T. R.; Li, C. J Neurosci 1992, 12, 2356. III. 1, 717. 1, 98, 14000. (158) Sossin, W. S.; Fisher, J. M.; Scheller, R. H. Neuron 1989, 2, 1407. (159) Cottrell, G. A. J Exp Biol 1997, 200, 2377. 730. r, orld Scientific: Singapore, 1999; Vol. VI; pp 129. B (144) de Dios, A. C.; Pearso (145) Osapay, K.; Case, D. A. Journal of the American Chemical Society 1991, 113, 9436. (146) Sitkoff, D.; Case, D. A. Abstracts of Papers of the American Chemical Society 1997, 214, 234. (147) Lee, B.; Richards, F. M. Journal of Molecular Biology 1971, 55, 379. (148) Shrake, A.; Rupley, J. A. Journal of Molecular Biology 1973, 79, 351. (149) Jessell, T. M.; Kandel, E. R. Cell 1993, 72 Suppl, 1. (152 (153) Hyman, L. H. The invertebrates.; Mc Graw Hill: New York, 1951; Vol. (154) Johnigk, S. A.; Ehlers, R. U. Nematology 1999 (155) Watts, D. J.; Strogatz, S. H. Nature 1998, 393, 440. (156) Nathoo, A. N.; Moeller, R. A.; Westlund, B. A.; Hart, A. C. Proc Natl Acad Sci U S A 200 (157) Bargmann, C. I. Science 1998, 282, 2028. (160) Furukawa, Y.; Miyawaki, Y.; Abe, G. Pflugers Arch 2006, 451, 646 (161) Lingueglia, E.; Champigny, G.; Lazdunski, M.; Barbry, P. Nature 1995, 378 (162) de Bono, M.; Bargmann, C. I. Cell 1998, 94, 679. (163) Hansmann, U. H. E. a. O., Y. Annual Reviews of Computational Physics StauffeD., Ed.; W 137

PAGE 138

(164) D.A. Case, T. A. D., T.E. Cheatham, III, C.L. Simmerling, J. Wang, R.E. Duke, R. Luo, K.M. Merz, D.A. Pearlman M. Crowley, R.C. Walker, W. Zhang, B. Wang, S. Hayik, Roitberg, G. Seabra, K.F. Wong, F. Paesani, X. Wu, S. Brozell, V. Tsui, H. Gohlke, L. Yang, .S. Ross, nd P.A. Kollman AMBER 9; University of California, San Francisco, 2006. (165) G.D. Hawkins, C. J. C. a. D. G. T. J. Phys. Chem 1996, 100, 19824. (167) Case, V. T. a. D. A. Biopolymers (Nucl. Acid. Sci.) 2001, 56, 275. (168) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. Journal of Computational hysic W. F.; Dinola, A.; Haak, J. Journal of Chemical Physics 1984, 81, 3684. (170) Sindhikara, D.; Meng, Y. L.; Roitberg, A. E. Journal of Chemical Physics 2008, 28. ww.chpc.utah.edu/~cheatham/software.html. Last accessed Sept, 2008. (172) Simmerling, C., Elber, R., and Zhang, J. Moil-viewA program for visualization mputing stochastic aths.Kluwer, Netherlands, 1995. (173) Meier, B. H. a. E., W.L. J Am Chem Soc 1987, 109, 7937. t, D. S.; Sykes, B. D. Methods Enzymol 1994, 239, 363. 21, 321. Acids; John Wiley and Sons: New ork, 1986. (178) Schwarzinger, S.; Kroon, G. J.; Foss, T. R.; Chung, J.; Wright, P. E.; Dyson, H. J. P. E.; Dyson, H. J. J Biomol MR 2000, 18, 43. 338, 3. A C. Tan, J. Mongan, V. Hornak, G. Cui, P. Beroza, D.H. Mathews, C. Schafmeister, W a (166) G.D. Hawkins, C. J. C. a. D. G. T. Chemical Physics Letters 1995, 246, 122. Ps 1977, 23, 327. (169) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, R 1 (171) Cheatham, T. E., III. Software, Benchmarks, w of structure and dynamics of biomolecules and STOA program for co p (174) Wishar (175) Xu, X. P.; Case, D. A. J Biomol NMR 2001 (176) Bundi, A.; Wuthrich, K. FEBS Lett 1977, 77, 11. (177) Wuthrich, K. NMR of Proteins and Nucleic Y J Am Chem Soc 2001, 123, 2970. (179) Schwarzinger, S.; Kroon, G. J.; Foss, T. R.; Wright N (180) Wishart, D. S.; Case, D. A. Methods Enzymol 2001 138

PAGE 139

(181) Andersen, N. H.; Cao, B.; Chen, C. Biochem Biophys Res Commun 1992, 184, 1008. (182) Akil, H.; Watson, S. J.; Young, E.; Lewis, M. E.; Khachaturian, H.; Walker, J. MAnnu Rev Neurosci 1984, 7, 223. 99. G. 355. (188) Yaksh, T. L.; Noueihed, R. Annu Rev Pharmacol Toxicol 1985, 25, 433. (189) Yaksh, T. L.; Rudy, T. A. Science 1976, 192, 1357. la, P. Febs Letters 2006, 580, ova, E. V.; Quan, Y.; Vaysse, P. J.; Wetzel, J. M.; Branchek, A.; Gerald, C.; Borowsky, B. J Biol Chem 2000, 275, 39324. (192) Gouarderes, C.; Zajac, J. M. Neuroscience Research 2007, 58, 91. (193) Kotani, M.; Mollereau, C.; Detheux, M.; Le Poul, E.; Brezillon, S.; Vakili, J.; al of Pharmacology 2001, 33, 138. (194) Panula, P.; Aarnisalo, A. A.; Wasowicz, K. Progress in Neurobiology 1996, 49, (195) Roumy, M.; Zajac, J. M. European Journal of Pharmacology 1998, 345, 1. C.; Cheve, G.; McCurdy, C. R. Peptides 2006, 27, 990. (198) Gouarderes, C.; Tafani, J. A.; Zajac, J. M. Peptides 1998, 19, 727. (183) Koltzenburg, M.; McMahon, S. B. Trends Pharmacol Sci 1991, 12, 3 (184) Mansour, A.; Watson, S. J.; Akil, H. Trends Neurosci 1995, 18, 69. (185) Roberts, W. J. Pain 1986, 24, 297. (186) Talbot, J. D.; Marrett, S.; Evans, A. C.; Meyer, E.; Bushnell, M. C.; DuncanH. Science 1991, 251, 1 (187) Willis, W. D., Jr. Nature 1995, 373, 19. (190) Anko, M. L.; Ostergard, M.; Lintunen, M.; Panu6955. (191) Bonini, J. A.; Jones, K. A.; Adham, N.; Forray, C.; Artymyshyn, R.; Durkin, M. M.; Smith, K. E.; Tamm, J. A.; Boteju, L. W.; Lakhlani, P. P.; Raddatz, R.; Yao, W. J.; Ogozalek, K. L.; Boyle, N.; Kouran T Mazarguil, H.; Vassart, G.; Zajac, J. M.; Parmentier, M. British Journ 1 285. (196) Vyas, N.; Mollereau (197) Gelot, A.; Frances, B.; Gicquel, S.; Zajac, J. M. Eur J Pharmacol 1998, 358, 203. 139

PAGE 140

(199) Panula, P.; Kalso, E.; Nieminen, M.; Brandt, A.; Pertovaara, A. Brain Research1999, 848 A20. (201) Roussin, A.; Serre, F.; Gouarderes, C.; Mazarguil, H.; Roumy, M.; Mollereau, C.; a, N.; Matsumoto, Y.; Hosoya, M.; ujii, R.; Watanabe, T.; Kikuchi, K.; Terao, Y.; Yano, T.; Yamamoto, T.; Kawamata, Y.; 2, 703. Biophys Res Commun 1999, 256, 352. ord, J.; ouzens, M.; Sutherland, G. R.; Jenkins, N. A.; Herzog, H. Brain Res Mol Brain Res 2000, 77, 99. (207) Allen, Y. S.; Adrian, T. E.; Allen, J. M.; Tatemoto, K.; Crow, T. J.; Bloom, S. R.; (209) Spinazzi, R.; Andreis, P. G.; Rossi, G. P.; Nussdorfer, G. G. Pharmacol Rev 2006, 8, 46. (200) Payza, K.; Akar, C. A.; Yang, H. Y. J Pharmacol Exp Ther 1993, 267, 88. Zajac, J. M. Biochem Biophys Res Commun 2005, 336, 197. (202) Bonnard, E.; Mazarguil, H.; Zajac, J. M. Peptides 2002, 23, 1107. (203) Hinuma, S.; Shintani, Y.; Fukusumi, S.; Iijim F Habata, Y.; Asada, M.; Kitada, C.; Kurokawa, T.; Onda, H.; Nishimura, O.; Tanaka, M.; Ibata, Y.; Fujino, M. Nat Cell Biol 2000 (204) Cikos, S.; Gregor, P.; Koppel, J. Biochem (205) Parker, R. M.; Copeland, N. G.; Eyre, H. J.; Liu, M.; Gilbert, D. J.; Crawf C 1 (206) Colmers, W. F.; El Bahh, B. Epilepsy Curr 2003, 3, 53. Polak, J. M. Science 1983, 221, 877. (208) Morris, B. J. J Comp Neurol 1989, 290, 358. 5 (210) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. Journal of Chemical Physics 1983, 79, 926. 140

PAGE 141

BIOGRAPHICAL SKETCH Georgios Leonis was born in Athens, the capital of Greece. In 1996, he entered the niversity of Athens obtaining a bachelors degree in chemistry with a specialization in the graduate program at the .D. program, U physical/theoretical chemistry. In August 2002, he enrolled in University of Florida in Gainesville, and received his Ph.D. in 2008. During the Ph he was applying theoretical methodologies to investigate biological problems. 141