Molecular modelling of neurotrophin-receptor binding

FIELD OF THE INVENTION
The present invention relates to a computational method for identifying minimum-energy structures in molecular systems, and more particularly for determining the bioactive conformations of peptide domains. The invention relates to determination of the bioactive conformations of the amino and carboxyl termini of Nerve Growth Factor (NGF) and analogues thereof, and the 3-dimensional structures of their complexes with the 24-amino acid second leucine-rich motif (LRM) of TrkA which represents the binding site for NGF. The invention further relates to the theoretically identified geometries of the TrkB and the p75.sup.NTR receptor binding sites and to the mechanisms of neurotrophin binding.
BACKGROUND OF THE INVENTION
The neurotrophins are a family of structurally and functionally related proteins, including Nerve Growth Factor (NGF), Brain-Derived Neurotrophic Factor (BDNF), Neurotrophin-3 (NT-3), Neurotrophin-4/5 (NT-4/5) and Neurotrophin-6 (NT-6). These proteins promote the survival and differentiation of diverse neuronal populations in both the peripheral and central nervous systems (Hefti, 1986; Hefti and Weiner, 1986; Levi-Montalcini, 1987; Barde, 1989; Leibrock et al., 1989; Maisonpierre et al., 1990; Rosenthal et al., 1990; Hohn et al., 1990; Gotz et al., 1994; Maness et al., 1994) and are involved in the pathogenesis of diverse neurological disorders. Neurotrophins exert certain of their biological effects through specific interactions with a class of transmembrane receptor tyrosine kinases (TrkA, TrkB and TrkC) (Kaplan et al., 1991; Klein et al., 1991, 1992; Soppet et al., 1991; Squinto et al., 1991; Berkemeier et al., 1991; Escandon et al., 1993; Lamballe et al., 1991). Specificity of neurotrophin action results from their selective interactions with Trk. That is, TrkA only binds NGF (Kaplan et al., 1991; Klein et al., 1991); TrkB binds BDNF, NT-3 and NT-4/5 (Soppet et al., 1991; Squinto et al., 1991; Berkemeier et al., 1991; Escandon et al., 1993; Lamballe et al., 1991; Klein et al., 1992; Vale and Shooter, 1985; Barbacid, 1993); and TrkC exclusively binds NT-3 (Lamballe et al., 1991; Vale and Shooter, 1985). This is particularly evident when the Trk receptors are coexpressed with the common neurotrophin receptor p75.sup.NTR (For review see Meakin and Shooter, 1992; Barbacid, 1993; Chao, 1994; Bradshaw et al., 1994; Ibanez, 1995).
The common neurotrophin receptor p75.sup.NTR is a transmembrane glycoprotein structurally related to the tumor necrosis factor and CD-40 receptors (Meakin and Shooter, 1992; Ryden and Ibanez, 1996). As all neurotrophins bind to p75.sup.NTR with similar affinity (Rodriguez-Tebar et al., 1990; Hallbook et al., 1991; Rodriguez-Tebar et al., 1992; Ibanez, 1995), neurotrophin specificity is conventionally thought to be caused by the binding selectivity for Trk receptors which are differentially expressed in different neuronal populations (Ibanez, 1995). However, accumulated experimental data on neurotrophin activity reveal important functional aspects of p75.sup.NTR (Heldin et al., 1989; Jing et al., 1992; Herrmann, 1993; Barker and Shooter, 1994; Dobrowsky et al., 1994, Matsumoto et al., 1995; Marchetti et al., 1996; Washiyama et al., 1996). The common neurotrophin receptor enhances functions and increases binding specificity of Trk receptors (Barker and Shooter, 1994; Mahadeo et al., 1994; Chao and Hempstead, 1995; Ryden and Ibanez, 1996). In addition, p75.sup.NTR possesses unique, Trk-independent signalling properties which involve ceramide production through activation of the sphingomyelin cycle (Dobrowsky et al., 1994), apoptosis (cell death) (Van der Zee et al., 1996; Cassacia-Bonnefil et al., 1996; Frade et al., 1996), and activation of the transcription factor NFKB (Carter et al., 1996). Recently, p75.sup.NTR has been demonstrated to participate in human melanoma progression (Herrmann et al., 1993; Marchetti et al., 1996). Furthermore, NGF and NT-3 increase the production of heparin by 70W melanoma cells, which is associated with their metastatic potential (Marchetti et al., 1996). Although this effect has been shown to be mediated by the common neurotrophin receptor, neither BDNF nor NT-4/5 appeared to be active. Explicit modelling of the various neurotrophin receptors will be critical to the rational design of neurotrophin-based drugs.
NGF displays high and low affinity binding sites in sensory and sympathetic neurons and in pheochromocytoma PC12 cells (Sutter et al., 1979; Landreth and Shooter, 1980; Schechter and Bothwell, 1981). The coexpression of the common neurotrophin p75.sup.NTR receptor with TrkA is required to form the high affinity binding site (Hempstead et al., 1991; Barker and Shooter, 1994; Mahadeo et al., 1994; Chao and Hempstead, 1995). Several models of the TrkA-p75.sup.NTR interaction have been proposed to explain high affinity NGF binding (Bothwell, 1991; Chao, 1992b; Chao and Hempstead, 1995; Wolf et al., 1995; Ross et al., 1996; Ross et al., 1997). These models differ with respect to direct (conformational model) or indirect (ligand-presentation model) interaction of p75.sup.NTR with TrkA. Direct TrkA-p75.sup.NTR interaction is consistent with much of the existing experimental data.
NGF, a 118 amino acid protein, is an extremely important neurotrophin, being implicated in the pathogenesis of Alzheimers disease, epilepsy and pain (Ben and Represa, 1990; McKee et al., 1991; Leven and Mendel, 1993; Woolf and Doubell, 1994; Rashid et al., 1995; McMahon et al., 1995). The binding of NGF to its receptors is determined by distinct sequences within its primary amino acid structure. While several regions of NGF participate in the NGF/TrkA interaction, mutation studies suggest that relatively few key residues, namely those located in the NGF amino and carboxyl termini, are required for high affinity binding.
The hairpin loop at residues 29-35 is responsible for recognition by p75.sup.NTR (Ibanez et al., 1992; Radziejewski et al., 1992), while the amino and carboxyl termini are important binding determinants for recognition by the TrkA receptor (Shih et al., 1994; Moore and Shooter, 1975; Suter et al., 1992; Burton et al., 1992, 1995; Kahle et al., 1992; Luo and Neet, 1992; Drinkwater et al., 1993; Treanor et al., 1995; Taylor et al., 1991). Truncation of either the amino or carboxyl terminus of NGF produces less active NGF analogues; similarly most deletion or point mutations of the amino terminus also lead to NGF analogues with diminished activity (Shih et al., 1994; Burton et al., 1992, 1995; Kahle et al., 1992; Drinkwater et al., 1993; Treanor et al., 1995; Taylor et al., 1991). On the other hand, the NGF.DELTA.2-8 (NGF with residues 2-8 removed) and NGF.DELTA.3-9 deletion mutants are almost as active as wild type NGF (Drinkwater et al., 1993). These NGF structure-activity relationships in combination with the considerable species variability (mouse, human, guinea pig and snake) of the amino acid sequence of the NGF termini (McDonald et al., 1991) are of potential value in understanding the NGF/TrkA interaction.
NGF exerts its biological activity as a non-covalent dimer (Treanor et al., 1995; Burton et al., 1995; McDonald et al., 1991; Ibanez et al., 1993; Bothwell and Shooter, 1977). Two 118 residue NGF monomers are dimerized by hydrophobic and van der Waals interactions between their three anti-parallel pairs of .beta.-strands; consequently, the amino terminus of one NGF monomer and the carboxyl terminus of the other are spatially juxtaposed (McDonald et al., 1991). Furthermore, although a dimer has 2 pairs of termini, only one pair of termini is required for TrkA receptor recognition (Treanor et al., 1995; Burton et al., 1995). Accordingly, solving the conformation of the complex formed by the amino terminus of one monomer and the carboxyl terminus of the other in dimeric NGF is of fundamental relevance to understanding the interaction of NGF with TrkA.
The X-ray crystallographic 3-dimensional structure of a dimeric mouse NGF (mNGF) has been reported recently (McDonald et al., 1991). However, within this structure, the amino terminus (residues 1-11) and the carboxyl terminus (residues 112-118) remain unresolved for both pairs of termini. High flexibility of the NGF termini makes it difficult to experimentally determine their bioactive conformations, particularly since transition metal ions commonly used in X-ray crystallography (McDonald et al., 1991) have high affinity for His residues (Gregory et al., 1993) which are present in the NGF amino terminus (Bradshaw et al., 1994). Indeed, conformational alterations in the receptor binding domains of NGF caused by Zn.sup.2+ cations leading to its inactivation have been described recently (Ross et al., 1997). Since the amino and carboxyl termini are crucial for NGF bioactivity as mediated via TrkA and because of the significance of NGF in multiple neurologic disease processes, the determination of the biologically active conformation of these termini is an important and challenging problem for computational chemistry.
In contrast with NGF, little is known about the BDNF binding determinant for TrkB activation. Although there were several attempts to identify structural elements of BDNF that determine its receptor specificity (Ibanez et al.,1991, 1993,1995; Suter et al., 1992; Ilag et al., 1994; Kullander and Ebendal., 1994; Urfer et al., 1994; Lai et al., 1996), the results of mutational experiments are still controversial. Thus, the reported importance of residues 40-49 (variable loop region 11) for TrkB related activity of BDNF (Ibanez et al., 1991; Ilag et al., 1994) has not been confirmed by recent studies (Lai et al., 1996), in which it has also been shown that the key residues of the TrkB receptor binding domain of BDNF are situated between residues 80 and 109. The latter result is consistent with earlier observations that the BDNF termini are not involved in TrkB binding and that the key residues of the BDNF binding epitope are Arg.sup.81 and Gln.sup.84 (Ibanez et al., 1993). In addition, it has been found that receptor binding domains of BDNF and NT-3 are located in the same regions of the molecules, i.e. in their "waist" parts (Ibanez et al., 1993; Urfer et al., 1994). As far as the NT-3 receptor binding domain is concerned, detailed mutational studies carried out by Urfer et al. (1994) revealed that it comprises a surface that includes major parts of the central .beta.-strand stem, with the most important residues being Arg.sup.103, Thr.sup.22, Glu.sup.54, Arg.sup.56, Lys.sup.80 and Gln.sup.83. These residues correspond to Arg.sup.104, Thr.sup.21, Glu.sup.55, Lys.sup.57, Arg.sup.81 and Gln.sup.84 of BDNF, respectively (McDonald et al., 1991).
The location of the neurotrophin binding sites within the Trk receptors is the subject of debate (MacDonald and Meakin, 1996). Published data indicate the existence of two putative neurotrophin binding sites of Trk proteins: the "second immunoglobulin-like domain" (residues Trp.sup.299t -Asn.sup.365t, denoted IgC2) and the "second leucine-rich motif" (LRM-2A and LRM-2B, residues Thr.sup.97 -Leu.sup.120 of TrkA and TrkB, respectively) (Schneider and Schweiger, 1991; Kullander and Ebendal, 1994; Perez et al., 1995; Urfer et al., 1995; Windisch et al., 1995a, 1995b, 1995c; MacDonald and Meakin, 1996; Ryden and Ibanez, 1996).
p75.sup.NTR belongs to a family of cell surface proteins that share a common pattern of four repeated cysteine-rich domains (CRDs) in the extracellular portion (Yan and Chao, 1991; Baldwin and Shooter, 1994). X-ray crystallographic studies on the extracellular portion of the related protein p55.sup.TNFR revealed that CRDs fold independently of each other, and three conserved disulfide bonds maintain a specific geometry of each CRD which consists of two loops (Banner et al., 1993) (loops A and B, FIG. 1a). Very little is known about the p75.sup.NTR receptor functional epitope and, particularly, about residues directly participating in molecular recognition processes. All four CRD repeats of p75.sup.NTR are found to be required for binding, with the second CRD being most important (Baldwin and Shooter, 1995). Residue Ser.sup.50 of p75.sup.NTR seems to be essential for NGF binding (Baldwin and Shooter, 1995).
The results of site-directed mutagenesis suggest that the p75.sup.NTR receptor binding domains within neurotrophins are juxtaposed positively charged residues located in two adjacent hairpin loops which represent variable region I (residues 23-35) and V (residues 93-98) (Ibanez et al., 1992; Ryden et al., 1995; Ibanez, 1994; Ibanez, 1995; Ryden and Ibanez, 1996) (FIG. 1c). In NGF, residues Lys.sup.32, Lys.sup.34, and Lys.sup.95 have been found to be involved in p75.sup.NTR receptor binding, with Lys.sup.32 making the strongest contact, followed by Lys.sup.34 and Lys.sup.95. In addition, some role of Asp.sup.30, Glu.sup.35, Arg.sup.103, Arg.sup.100, Lys.sup.88, and Ile.sup.31 in p75.sup.NTR binding and biological activity has been demonstrated (Ibanez et a., 1992; Ibanez, 1994). Two residues of variable region I, namely Arg.sup.31 and His.sup.33, have been demonstrated to be essential for binding of NT-3 to p75.sup.NTR, whereas similarly located Arg.sup.34 and Arg.sup.36 mediate binding of NT-4/5 (Ryden et al., 1995). In contrast to NGF, a positively charged residue in variable region V is not critical for binding of NT-3 or NT-4/5 to p55.sup.NTR (Ryden et al., 1995). In BDNF, however, only residues of variable region V, Lys.sup.95, Lys.sup.96 and Arg.sup.97, bind to p75.sup.NTR and compensate for the lack of positively charged residues in variable loop region I (Ryden et al., 1995).
Therefore, it is an object of the invention to establish explicit atom-level models of the Trk and p75.sup.NTR recognition sites for neurotrophins.
It is a further object of the invention to provide atom-level models of the mechanism of interaction of neurotrophins and their receptors.
SUMMARY OF THE INVENTION
The inventors describe a method of characterizing the interaction between non-contiguous domains of a ligand and its receptor. The specific examples of the method described in detail below concern the interaction of nerve growth factor (NGF) with its TrkA receptor binding site, brain-derived neurotrophic factor (BDNF) with its TrkB receptor binding site, and the interaction of the neurotrophins (e.g., NGF, BDNF, NT-3, and NT-4/5) with the neurotrophin binding site of the common neurotrophin receptor p75.sup.NTR.
In one aspect of the invention there is provided a variable basis Monte-Carlo (VBMC) simulated annealing method for identifying an optimal structure in a molecular system. The method comprises:
(a) providing a Markov chain with an initial basis set of N configurational variables which define a structure for said molecular system;
(b) translating the N configurational variables simultaneously along a basis vector to produce a new structure for the molecular system, wherein magnitude of translation is chosen randomly within a preselected range;
(c) calculating potential energy of the new state of the molecular system;
(d) deciding whether to accept or reject the new structure using an effective temperature dependent transition function, and if the new structure is accepted, replacing the current structure with the accepted structure;
(e) repeating steps (b) to (d), each repetition having a new basis vector, for a preselected first number of repetitions at a preselected upper temperature; then
(f) decreasing temperature according to a preselected cooling schedule;
(g) repeating steps (b) to (d) and (f), each repetition having a new basis vector, for a preselected second number of repetitions; then
(h) storing current structure of the molecular system;
(i) repeating steps (g) and (h) until a preselected number of stored structures have been accumulated; then
(j) rotating the basis set of configurational variables according to an effective distribution of the accumulated structures so that basis vectors are directed along low-energy valleys of a potential energy hypersurface thereby accelerating conformational motions and structural transitions of the molecular system, and erasing accumulated structures from steps (h) and (i);
(k) repeating steps (i) and (j) until, after an effective third number of repetitions, a preselected lower temperature limit is reached, thereby producing an unrefined global minimum structure; and
(l) refining the global energy minimum structure so that at least one local minimum on the potential energy hypersurface is identified.
The present invention also provides a method of identifying structures of a peptide domain of a neurotrophin to bind with a binding site of a receptor. The method comprises providing a plurality of neurotrophin analogues wherein at least one of neurotrophin analogue binds the receptor, identifying an optimal or near-optimal structure for respective peptide domains of the neurotrophin analogues that bind the receptor by applying a preselected number of the variable basis Monte-Carlo (VBMC) annealing simulations described above to each of the respective peptide domains to identify minimum or near-minimum energy structures for the peptide domains, and comparing said minimum or near-minimum energy structures to identify common structural features shared by neurotrophin analogues exhibiting receptor mediated activity.
The step of providing neurotrophin analogues may include providing wild-type neurotrophin analogues or mutants thereof.
The present invention also provides a method of identifying structures of a domain of a ligand to bind with a binding site of a receptor, comprising providing a plurality of ligand analogues wherein at least one ligand analogue binds the receptor, identifying an optimal or near-optimal structure for respective domains of the ligand analogues that bind the receptor by applying a preselected number of the variable basis Monte-Carlo (VBMC) annealing simulations described above to each of the respective domains to identify minimum or near-minimum energy structures for the domains; and comparing said minimum or near-minimum energy structures to identify common structural features shared by ligands exhibiting receptor mediated activity.
As would be apparent to a person skilled in the art, methods of the present invention are not limited to neurotrophin-receptor interactions, but are applicable to the characterization of other ligand-receptor interactions. That is, an array of ligands that are either known to interact with a specific receptor or are candidate ligands that may potentially interact with this receptor may be computationally analyzed to identify optimal or near-optimal structures that provide the desired interaction. In some embodiments of the invention, the ligand may have a peptide domain, e.g., the ligand may be a peptide hormone, such as insulin, prolactin, or growth hormone. In some embodiments, the ligands or candidate ligands under study may be native ligands, mutants of native ligands, or isosteres of native ligands. They may be structurally different but have similar binding activity or functional effect(s). The optimal or near-optimal structure identified may act as an agonist or antagonist. "Isosteric" is an art-recognized term and, for purposes of this disclosure, a first moiety is "isosteric" to a second when they have similar molecular topography and functionality. This term should not, therefore, be restricted to mere similarity in chemical bond configuration. Two compounds may have different chemical bond configurations but nonetheless be isosteric due their 3-dimensional similarity of form and chemical nature, e.g. relevant centers such as heteroatoms are located in the same position, charges and/or dipoles are comparable, and the like.
The present invention provides a ligand for binding with TrkA, wherein TrkA comprises a leucine rich motif (LRM) comprising amino acid residues 93 to 117, the coordinates of the residues of said LRM given in Appendix 1, said LRM motif having five binding areas, area A comprising amino acid residue Phe.sup.105tA capable of hydrophobic bonding, area B comprising residues Phe.sup.111tA, Phe.sup.113tA, and Thr.sup.114tA capable of hydrophobic bonding, area C comprising residues Asp.sup.109tA and His.sup.112tA capable of ionic interaction, area D comprising residue Lys.sup.100tA capable of ionic interaction, and area E comprising residues Asn.sup.95tA to Ile.sup.98tA capable of multiple parallel N-strand type hydrogen bonding. The ligand includes at least three moieties each comprising effective atomic elements having spatial occupancy, relative atomic position, bond type and charge to define a three dimensional conformation of said ligand so that a first moiety can bind with a first binding area, a second moiety can bind with a second binding area, and a third moiety can bind with a third binding area.
Put another way, the ligand comprises at least three moieties effectively spaced relative to each other so that a first moiety can bind with a first binding area, a second moiety can bind with a second binding area, and a third moiety can bind with a third binding area. More specifically, the ligand comprises at least one moiety being capable of hydrophobic bonding and being present in an effective position in the ligand to hydrophobically bind to area A, at least one positively charged moiety being present in an effective position in said ligand to ionically bind to area C, and at least one negatively charged moiety being present in an effective position in said ligand to tonically bind to area D.
The present invention also provides a method of designing a ligand to bind with an LRM of TrkA. The method comprises providing a template of said LRM of TrkA, said LRM comprising amino acid residue sequence 93 to 117 inclusive, said amino acid residues having spatial Cartesian coordinates given in Appendix 1 wherein binding area A in the sequence comprises residue Phe.sup.105tA capable of hydrophobic bonding, binding area B comprises residues Phe.sup.111tA, Phe.sup.113tA, Thr.sup.114tA capable of hydrophobic bonding, binding area comprises residues Asp.sup.109tA and His.sup.121tA capable of ionic interaction, binding area D comprises Lys.sup.100tA capable of ionic interactions, and binding area E comprises residues Asn.sup.95tA to lle.sup.98tA capable of multiple parallel .beta.-strand type hydrogen bonding and computationally evolving a chemical ligand using any effective algorithm with preselected spatial constraints so that said evolved ligand comprises at least three effective moieties of suitable identity and spatially located relative to each other in the ligand so that a first of the moieties hydrophobically interacts with binding area A, a second of the moieties ionically interacts with binding area C, and a third of the moieties tonically interacts with binding area D of the LRM of TrkA. The algorithm may be a genetic algorithm and the LRM may be the second LRM of a plurality of LRMs.
The invention further provides a method of designing a ligand to bind with an LRM of TrkA, comprising providing a template of said LRM of TrkA, said LRM comprising amino acid residue sequence 93 to 117 inclusive, the amino acid residues having spatial Cartesian coordinates given in Appendix 1 wherein binding area A comprises Phe.sup.105tA capable of hydrophobic bonding, binding area B comprises Phe.sup.111tA, Phe.sup.113tA, Thr.sup.114tA capable of hydrophobic bonding, binding area C comprises Asp.sup.109tA and His.sup.112tA capable of ionic interaction, binding area D comprises Lys.sup.100tA capable of ionic interaction, and binding area E comprises Asn.sup.95tA to lle.sup.98tA capable of hydrogen bonding. The method comprises computationally evolving a ligand using an effective algorithm so that said evolved ligand comprises at least three effective moieties located relative to each other in the ligand so that a first moiety can bind with a first of the five binding areas, a second moiety can bind with a second of the five binding areas, and a third moiety can bind with a third of the binding areas.
In another aspect the invention provides a ligand for binding with TrkB, wherein TrkB comprises a leucine-rich motif (LRM) comprising amino acid residues 93 to 117, the coordinates of the residues of said LRM given in Appendix 2, said LRM having five binding areas, binding area F comprising amino acid residue Asp.sup.100tB capable of ionic interaction, binding area G comprising amino acid residue Lys.sup.109tB capable of ionic interaction, binding area H comprising amino acid residues Lys.sup.113tB capable of ionic interaction, binding area I comprising amino acid residue Arg.sup.94tB capable of ionic interaction, and binding area K comprising amino acid residue Lys.sup.104tB capable of hydrogen bonding or ionic interaction. The ligand comprises at least three moieties effectively spaced relative to each other so that a first moiety can bind with a first binding area, a second moiety can bind with a second binding area, and a third moiety can bind with a third binding area.
The present invention provides a method of designing a ligand to bind with TrkB, wherein TrkB comprises a leucine-rich motif (LRM) comprising amino acid residues 93 to 117, the coordinates of the residues of said LRM given in Appendix 2, said LRM having five binding areas, binding area F comprising amino acid residue Asp.sup.100tB capable of ionic interaction, binding area G comprising amino acid residue Lys.sup.109tB capable of ionic interaction, binding area H comprising amino acid residue Lys.sup.113tB capable of ionic interaction, binding area I comprising amino acid residue Arg.sup.94tB capable of ionic interaction, and binding area K comprising amino acid residue Lys.sup.104tB capable of hydrogen bonding or ionic interaction. The method comprises computationally evolving a ligand using an effective algorithm so that the evolved ligand comprises at least three effective moieties located relative to each other in the ligand so that a first moiety can bind with a first of said five binding areas, a second moiety can bind with a second of said five binding areas, and a third moiety can bind with a third of the binding areas. The algorithm may be a genetic algorithm.
The invention also provides a ligand for binding with the common neurotrophin receptor p75.sup.NTR, wherein p75.sup.NTR comprises a binding site including amino acid residues Cys.sup.39p to Cys.sup.94p inclusive, said residues of said binding site having spatial coordinates given in Appendix 3, said binding site having three binding areas including binding loop 2A comprising region Cys.sup.39p to Cys.sup.58p capable of attractive electrostatic interactions, binding loop 2B comprising region Cys.sup.58p to Cys.sup.79p capable of attractive electrostatic interactions, and binding loop 3A comprising region Cys.sup.79p to Cys.sup.94p capable of attractive electrostatic interactions. The ligand comprises at least two moieties effectively spaced relative to each other so that a first moiety can bind to a first loop and a second moiety can bind to a second loop.
The present invention also provides a method of designing a ligand for binding with the common neurotrophin receptor p75.sup.NTR, wherein p75.sup.NTR comprises a binding site including amino acid residues Cys.sup.39p to Cys.sup.94p inclusive. The residues of the binding site have spatial coordinates given in Appendix 3. The binding site has three binding areas including binding loop 2A comprising region Cys.sup.39p to Cys.sup.58p capable of attractive electrostatic interaction, binding loop 2B comprising region Cys.sup.58p to Cys.sup.79p capable of attractive electrostatic interaction, and binding loop 3A comprising region Cys.sup.79p to Cys.sup.94p capable of attractive electrostatic interaction. The method comprises computationally evolving a ligand using an effective algorithm so that the evolved ligand comprises at least two effective moieties located relative to each other in the ligand so that a first moiety can bind to a first of the three loops and a second moiety can bind to a second loop of the three loops.
The invention provides a method of identifying a ligand to bind with a leucine rich motif (LRM) of TrkA. The method comprises providing a three dimensional conformation for the LRM. The LRM comprises amino acid residue sequence 93 to 117 inclusive, the amino acid residues having spatial Cartesian coordinates given in Appendix 1. The LRM has five binding areas, area A comprising amino acid residue Phe.sup.105tA capable of hydrophobic interaction, area B comprising amino acid residues Phe.sup.111tA, Phe.sup.113tA, and Thr.sup.114tA capable of hydrophobic interaction, area C comprising amino acid residues Asp.sup.109tA and His.sup.112tA capable of ionic interaction, area D comprising amino acid residue Lys.sup.100tA capable of ionic interaction, and area E comprising Asn.sup.95tA to lle.sup.98tA capable of multiple parallel .beta.-strand type hydrogen bonding. The method includes providing a data base containing molecules coded for spatial occupancy, relative atomic position, bond type and/or charge; and screening the data base to select those molecules comprising moieties having a three dimensional conformation and effective charge that can bind with at least three of the five binding areas.
The invention provides a method of identifying a ligand to bind with a leucine rich motif (LRM) of TrkB. The method comprises providing a three dimensional conformation for the LRM. The LRM comprises amino acid residues 93 to 117, the coordinates of the residues of the LRM are given in Appendix 2, the LRM having five binding areas, binding area F comprising amino acid residue Asp.sup.100tB capable of ionic interaction, binding area G comprising amino acid residue Lys.sup.109tB capable of ionic interaction, binding area H comprising amino acid residue Lys.sup.113tB capable of ionic interaction, binding area I comprising amino acid residue Arg.sup.94tB capable of ionic interaction, and binding area K comprising amino acid residue Lys.sup.104tB capable of hydrogen bonding or ionic interaction. The method includes providing a data base containing molecules coded for spatial occupancy, relative atomic position, bond type and/or charge; and screening the data base to select those molecules comprising moieties having a three dimensional conformation and effective charge that can bind with at least three of the five binding areas.
The present invention also provides a method of identifying a ligand to bind with common neurotrophin receptor p75.sup.NTR. The method comprises providing a three dimensional conformation for the common neurotrophin receptor p75.sup.NTR, wherein p75.sup.NTR comprises a binding site including amino acid residues Cys.sup.39p to Cys.sup.94p inclusive. The residues of the binding site have spatial coordinates given in Appendix 3, the binding site having three binding areas including binding loop 2A comprising region Cys.sup.39p to Cys.sup.58p capable of attractive electrostatic interaction, binding loop 2B comprising region Cys.sup.58p to Cys.sup.79p capable of attractive electrostatic interaction, and binding loop 3A comprising region Cys.sup.79p to Cys.sup.94p capable of attractive electrostatic interaction. The method includes providing a data base containing molecules coded for spatial occupancy, relative atomic position, bond type and/or charge; and screening the data base to select those molecules comprising moieties that can bind with at least two of the three binding loops.
In one aspect of the invention, a ligand (e.g., small molecule) designed to interact with a receptor such as Trk or p75.sup.NTR may be bifunctional, such that a first portion of the ligand binds to one receptor monomer and a second portion of the ligand binds to a second receptor monomer. Thus, a receptor dimer is effectively cross-linked by the interaction of ligand with both receptor monomers.
According to the invention, detailed information is provided regarding those amino acid residues of neurotrophins and of neurotrophin receptors which are involved in neurotrophin-receptor binding. Such information includes the bioactive conformations of the respective peptide domains of the neurotrophins and receptors. This in turn permits the design of small molecules which interact with said amino acid residues or domains. Where such interaction block the residues or changes the conformation of the domain, the small molecule can inhibit neurotrophin-receptor binding, in certain cases abolishing it completely. As would be appreciated by a person skilled in the art, small molecule design according to this aspect of the invention is applicable to other ligand-receptor interactions in addition to neurotrophins, according to the methods disclosed herein.
As a person skilled in the art would appreciate, the invention not only provides molecules (ligands) that interact with receptors, but also molecules which, in their mimicry of a receptor's bioactive conformation, can bind to the receptor binding site of a ligand such as, for example, a neurotrophin. Such molecules, whether they interact with ligand or with receptor, can block native ligand-receptor interaction. In some embodiments of the invention, a small molecule which binds to a receptor may even trigger signal transduction similar to that of the native ligand, i.e., act as an agonist.

BRIEF DESCRIPTION OF THE DRAWINGS
The invention will be described, by example only, reference being had to the accompanying drawings, in which:
FIG. 1 shows the amino acid sequences of the conformationally variable termini of the dimeric NGF analogues under study. (First monomer, amino terminus: residues 1-15 of SEQ ID NO:1; residues 1-15 of SEQ ID NO:2; residues 1-8 of SEQ ID NO: 3; SEQ ID NO: 4; residues 9-15 of SEQ ID NO: 2; SEQ ID NO: 5. Second monomer, carboxyl terminus: SEQ ID NO: 6; SEQ ID NO: 7; SEQ ID NO: 8.) The residues are numbered from 1 to 118 for the first monomer, and from 1' to 118' for the second monomer. The fixed parts of the proteins simulated by the X-ray geometry of mouse NGF (McDonald et al., 1991) are underlined.
FIG. 2 shows the structure of the complex of the amino and carboxyl termini of (1-118)hNGF homodimer. The carboxyl terminus is shown in grey (residues 5-11 of SEQ ID NO: 6), the amino terminus in white, (residues 1-12 of SEQ ID NO: 1) and the interacting conservative region His.sup.75' -Asn.sup.77' of the "south" part of the NGF dimer, in black (residues 2-4 of SEQ ID NO: 10).
FIG. 3 shows the structure of the complex of the amino and carboxyl termini of (1-118)mNGF homodimer (grey: residues 5-11 of SEQ ID NO: 7, white: residues 1-12 of SEQ ID NO: 2, black: residues 2-4 of SEQ ID NO: 10).
FIG. 4 shows the structure of the complex of the amino and carboxyl termini of dimeric mNGF.DELTA.3-9 molecule grey: residues 5-11 of SEQ ID NO: 7, white: residues 1-5 of SEQ ID No: 3, black: residues 2-4 of SEQ ID NO: 10).
FIG. 5 shows the structure of the complex of the amino and carboxyl termini of dimeric hNGF-H4D molecule. (grey: residues 5-11 of SEQ ID NO: 6, white: residues 1-12 of SEQ ID NO: 4, black: residues 2-4 of SEQ ID NO: 10).
FIG. 6 shows the structure of the complex of the amino and carboxyl termini of dimeric mNGF.DELTA.1-8 molecule (grey: residues 5-11 of SEQ ID NO: 7, white: residues 9-12 of SEQ ID NO: 2, black: residues 2-4 of SEQ ID NO: 10).
FIG. 7 shows the structure of the complex of the amino and carboxyl termini of BDNF homodimer (grey: residues 5-11 of SEQ ID NO: 8, white: residues 1-10 of SEQ ID NO: 5, black: residues 2-4 of SEQ ID NO: 10).
FIG. 8 illustrates the TrkA leucine-rich motifs. Locations of parallel .beta.-strands are illustrated by the shaded areas designated by symbol .beta.. Fragments B1-B4 demonstrate NGF binding affinity, while N1-N4 do not (Windisch et al., 1995a, 1995b).
FIG. 9 shows sequences of the amino and carboxyl termini of fully active NGF analogues. Regions with rigid structures are underlined. (First monomer, amino terminus: SEQ ID NO: 1; SEQ ID NO: 2; SEQ ID NO: 3. Second monomer, carboxyl terminus: SEQ ID NO: 10; SEQ ID NO: 6; SEQ ID NO: 7; SEQ ID NO: 11; SEQ ID NO: 9).
FIG. 10 shows the predicted bioactive conformation of the NGF binding domain (a) and the structures of the NGF/LRM-2A complexes of hNGF (b), mNGF (c), mNGF.DELTA.3-9 (d), and hNGF:NT-4/5 (e). Only residues essential for discussion are designated. A, B, C, D and E represent specific NGF/LRM-2A binding areas. The LRM-2A is illustrated in grey, NGF in white.
FIG. 11 shows the TrkB receptor binding site of BDNF (a), predicted features of the BDNFILRM-2B complex (b), and geometry of the "improper" NGF/LRM-2B complex (c). Only residues essential for discussion are designated. F, G, H, I and K represent specific BDNF/LRM-2B binding areas. The LRM-2B is illustrated in grey, BDNF and NGF in white.
FIG. 12 is a detail of the LRM-2A (SEQ ID NO:12) conformation when bound to hNGF. The three groups of the .beta.-strand involved in H-bonding with the parallel N-strand motif of the NGF analogues are designated by NH or CO. Hydrophobic residues are illustrated as light, polar uncharged as dark, charged as black.
FIG. 13 presents principal residues of the NGF binding domain essential for TrkA receptor recognition. (Amino terminus: SEQ ID NO: 1; carboxyl terminus: SEQ ID NO: 10; SEQ ID NO: 6). Conserved residues of NGF molecules among all characterized species (McDonald et al., 1991) are designated by X, whereas those which are substituted in NGF molecules derived from different species by similar amino acids and NGF residues maintaining the bioactive conformation or interacting with specific binding areas A-E of the LRM-2A are indicated by X.
FIG. 14 presents principal residues of the LRM-2A (SEQ ID NO: 12). The residues which are specific for TrkA (Windisch et al., 1995a, 1995b) and critical for the present model are designated by X and X, respectively.
FIG. 15 is a schematic representation of the second leucine-rich motif of TrkA (residues 93tA to 116tA) (a: residues 1-24 of SEQ ID NO:12) and TrkB (residues 93tB to 116tB) (b: SEQ ID NO:12) as specific binding sites for NGF and BDNF, respectively. General consensus residues are shaded. Arrows designate distinct positions essential for molecular recognition processes.
FIG. 16 represents the X-ray crystallography-derived geometry of the .alpha.-carbon trace of the two loops (2A and 2B) of the second cysteine-rich domain (white bonds) and the first loop (3A) of the third cysteine-rich domain (grey bonds) of p55.sup.TNFR (Banner et al., 1993) (a), the preferred geometry of the .alpha.-carbon trace of the corresponding fragment of p75.sup.NTR (b), and the predicted bioactive conformation of mouse NGF (McDonald et al., 1991); (c). White and grey bonds in panel (c) illustrate different protomers. The p75.sup.NTR binding determinant of NGF (loops I and V) is denoted. Disulfide bonds are shown in black.
FIG. 17 shows the sequence of a putative binding site for neurotrophins within p75.sup.NTR aligned with the corresponding sequence of homologous p55.sup.TNFR (SEQ ID NO: 14; SEQ ID NO: 15; SEQ ID NO: 16; SEQ ID NO: 17). Conserved pattern of disulfide bonds is illustrated. Abbreviation (h) stands for human (Johnson et al., 1986), (r) for rat (Radeke et al., 1987), and (c) for chicken (Large et al., 1989) p75.sup.NTR.
FIG. 18 illustrates the most stable geometries of the complexes of the neurotrophins with the common neurotrophin receptor p75.sup.NTR : with NGF (a); with BDNF (b); with NT-3 (c); with NT-4/5 (d). The .alpha.-carbon traces of the neurotrophins are shown in white, while that of the p75.sup.NTR binding site is shown in grey. Disulfide bonds are illustrated in black. Only residues mentioned in the text are denoted.
FIG. 19 demonstrates geometric compatibility of the van der Waals surfaces of BDNF (light) and the p75.sup.NTR binding site (dark).
FIG. 20 illustrates conformational changes of the p75.sup.NTR binding site upon BDNF binding. Conformation of the p75.sup.NTR receptor binding site before binding is illustrated in white, after binding in grey. The two conformations have been superimposed using the equilibrium positions of the reference a-carbon atoms located in loops 2A and 3A. Straight lines denote linear approximations of the conformations of the a.-carbon traces of loop 2B in both structures.

DETAILED DESCRIPTION OF THE INVENTION
PART A
Theoretical Studies of the Bioactive Conformation of Nerve Growth Factor Using VBMC--A Novel Variable Basis Monte-Carlo Simulated Annealing Algorithm for Peptides
Background of the VBMC Algorithm
The structural optimization of the peptide domains under study required a rigorous search of conformational space. As with other theoretical peptide studies, such calculations are plagued by a multiple-minima problem (Gibson and Scheraga, 1988) which traditionally is addressed using either molecular dynamics, Monte Carlo approaches or other algorithms. The Monte Carlo method, although exceedingly powerful, is limited by the excessive CPU time requirements on existing computers. To remedy this problem, we have devised a novel Variable Basis Monte Carlo simulated annealing algorithm for peptides (VBMC).
The potential energy hypersurface of a peptide has an enormous number of local minima (Gibson and Scheraga, 1988; Piela et al., 1989, 1994), but usually only the global energy minimum structure is required. Despite efforts to elaborate an efficient method of global energy minimization (Gibson and Scheraga, 1988; Judson et al., 1994), the multiple-minima problem is far from being solved. Although simulated annealing (Kirkpatrick et al., 1983; Vanderbilt et al., 1984) is mathematically correct and guarantees convergence to the global minimum in principle, the suggested "cooling" schedule appears to be far too slow for practical purposes, and faster "cooling" has to be used (Gidas, 1985; Kushner, 1987). Within the simulated annealing approach, thermal molecular motion can be simulated either by Monte Carlo (Kirkpatrick et al., 1983; Nayeem et al., 1991; Wilson et al., 1988; Snow, 1992; Shamovsky et al., 1992) or molecular dynamics (Lelj et al., 1991; Villani and Tamburro, 1994) methods.
The major problem of the Monte Carlo technique is that motion of a molecular system along a Markov chain is very slow (Piela et al., 1994; Scheraga, 1989; Nayeem et al., 1991). On the other hand, the Monte Carlo technique has several considerable advantages over molecular dynamics. First, it does not require calculating forces. Second, potential energy functions do not have to be continuous. Third, much larger steps in conformational space can be utilized (Li and Scheraga, 1988; Kincaid and Scheraga, 1982). Fourth, the additional iterative self-consistent SHAKE routine (Ryckaert et al., 1977), which is used in molecular dynamics simulations to deal with holonomic constraints (Brooks III and Case, 1993), is not required in the Monte Carlo method. This has led to efforts to accelerate configurational motion within the Monte Carlo approach (Raj et al., 1994; Snow, 1992; Kincaid and Scheraga, 1982; Rao et al., 1979; Rossky et al., 1978). The main reason for inefficiency with the conventional Monte Carlo method (Metropolis et al., 1953) is a slow and cumbersome sequential change of configurational variables.
To overcome these problems, lessons can be learned from molecular dynamics. In molecular dynamics (Allen and Tildesley, 1987) configurational variables move simultaneously, and the soft modes of the system explicitly direct these motions. Soft modes are defined as low-energy valleys on the potential energy hypersurface. Although there have been attempts to direct Monte Carlo configurational motion along the soft modes (Raj et al., 1994; Rao et al., 1979; Rossky et al., 1978), these improvements require calculating forces and, consequently, result in losing the main advantages of the Monte Carlo approach. An algorithm which would retain the advantages of the Monte Carlo method and direct configurational changes along the soft modes is thus required. One possible solution has recently been suggested (Szentpaly et al., 1994).
The classical Monte Carlo-Metropolis method (Metropolis et al., 1953) consists of modelling a Markov chain containing subsequent conformations of the molecule under study: x.sub.0, x.sub.1, x.sub.2, . . . , x.sub.i, . . . , where x (x.sub.i ={x.sub.i.sup.1. . . x.sub.i.sup.m . . . x.sub.i.sup.N }) is the vector of independent structural parameters. Each vector x.sub.i+1, is derived from the preceding one, x.sub.i, by means of a random change of one randomly chosen variable:
x.sub.i+1.sup.m =x.sub.i.sup.m +.lambda..sub.m .multidot..xi.(1)
where .xi. is a random quantity from the interval (-1,1) and .lambda..sub.m is a maximal allowed increment of the given variable. N moves of the Markov chain (where N is the number of independent parameters) represent a Monte Carlo step. Whether the new state x.sub.i+1 is accepted or rejected is decided according to a transition probability function. The following functional form of transition probability is used in the present expression (2) (Gunningham and Meijer, 1976):
W(x.sub.i .fwdarw.x.sub.i+1)=1/[1+exp(.DELTA.E.sub.i /kT)] (2)
where .DELTA.E.sub.i is the difference between conformational energies of the states x.sub.i+1 and x.sub.i, k is the Boltzmann constant and T is absolute temperature.
Those skilled in the art will appreciate that other functional forms of transition functions may be used to screen state a new state x.sub.i+1. For example, another effective temperature dependent transition function for screening a new state x.sub.i+1 derived from a previous state x.sub.i is given by: ##EQU1## where .DELTA.E.sub.i is the difference between conformational energies of the states x.sub.i+1 and x.sub.i, k is the Boltzmann constant and T is absolute temperature.
The direct application of this conventional Monte Carlo technique to the conformational space study of protein structure is not efficient because the conformational variables of peptide chains appear to be mutually dependent, i.e., low-energy conformational motions involve several conformational variables. When applied to molecular systems such as protein secondary, tertiary or quaternary structure, equation (1), in which structural parameters are sequentially changed one after another, results in very small values of .xi. and, consequently, in a very slow conformational motion (Piela et al., 1994; Scheraga, 1989; Nayeem et al., 1991). In order to model a simultaneous change of the structural parameters, we have used Eq.(3) (instead of the conventional (1)),
x.sub.i+1 =x.sub.i +.lambda..sub.m .multidot..xi..multidot.g.sup.m (3)
where g.sup.m is a randomly or sequentially chosen basis vector from a basis G={g.sup.m }, .xi. is a random quantity from an interval (-1,1) and .lambda..sub.m is a maximal acceptable move in direction g.sup.m. In this study .lambda..sub.m was constant and equal to 10.degree..
A basis G is formed from linear combinations of independent conformational parameters x for their change to be realized along the soft modes of the molecule. These modes are revealed by dispersion analysis of the distribution of the sequential configurational points of the Markov chain. The soft modes of the molecular system are to be directed along the axes of the dispersion ellipsoid because they should correspond to the directions of maximal deviations made by the independent variables during a certain period of simulation. Initially, basis G coincides with a unit matrix, i.e., G=I, and Eq.(3) is equivalent to (1). The rotation of basis G is carried out after every 150 Monte Carlo steps based on the particular scattering of the previous 150 configurations x. Basis vectors g.sup.m are made equal to the eigenvectors of the covariant matrix D which has the following elements: ##EQU2## where M is the number of points in the distributions (M=150), and x.sub.n.sup.1 is the i-th structural parameter at the n-th Monte Carlo step. Vectors g.sup.m form the orthonormalized basis and are used in the Markov chain instead of independent conformational parameters. Periodic rotations of basis G in Monte Carlo simulations result in a simultaneous change of all the conformational variables along the valleys of a potential energy surface, which considerably accelerates conformational motions and transitions (Szentpaly et al., 1994).
The method disclosed herein is highly advantageous over known processes for several reasons. In the present process the preselected range from which the magnitude of translation is chosen is constant and does not depend on the basis set. Thus the present method is much more efficient than previous methods in part because the magnitude of translation along the basis vector is limited by a preselected range. The present method is also more efficient because rotation of the basis set of configurational variables is avoided at the upper temperatures and is implemented only at the lower temperatures.
Implementation of the VBMC Algorithm
VBMC was written in Fortran-77. The algorithm was then interfaced with the CHARMM force field (Brooks et al., 1983; The force field parameters are extracted from CHARMM release 23.1, Molecular Simulations, Inc., Waltham, Mass.). Torsional distortion, van der Waals and electrostatic interactions, and hydrogen bonding were thus explicitly considered. The united atom approach was used for all carbons with non-polar hydrogens. Point atomic charges of Weiner et al. (Weiner et al., 1984) were used. A dielectric constant equal to the interatomic separation in angstroms was utilized to implicitly simulate solvation effects (Weiner et al., 1984). All calculations were performed on an 8-node IBM Scalable POWERparallel 2 [SP2] high performance computer at Queen's University, Kingston, Canada. Each Monte Carlo simulated annealing computation required up to 120 hours of CPU time.
Appendix 4 is attached. Appendix 4 is a computer program written in Fortran-77 and is a non-limiting example of a program used to perform the VBMC simulated annealing method in accordance with the present invention.
Each run of the VBMC algorithm provides a unique final structure which would represent the global energy minimum structure, regardless of the initial geometry, if the "cooling" was performed infinitely slowly. However, since this is impossible, several Monte Carlo simulated annealing computations must be performed from different starting points x.sub.0, using different initial temperatures T.sub.0, "cooling" schedules and random sequences. An initial temperature of 1000.degree. K. is sufficient. The Markov chain is kept at a constant temperature T.sub.0 for 10.sup.4 Monte Carlo steps before the initiation of "cooling". The temperature T is then decreased at every Monte Carlo move (3) by a small increment such that after 2.multidot.10.sup.4 Monte Carlo steps it reaches 120.degree. K. producing an unrefined global energy minimum structure.
The global energy minimum structure is then refined to identify the nearest local minimum or stationary point on the potential energy hypersurface. Refinement may be accomplished using any one of several non-linear minimization routines to identify the local minimum on the potential energy hypersurface. For example, gradient descent minimization and conjugate gradient minimization may be used. Quasi-Newton minimization methods such as the Newton-Raphson method may also be used. Other non-linear methods include the Marquardt method, the Powell method and a variable metric method. Those skilled in the art will be aware of the various non-linear routines which may be used for refinement and those mentioned herein are more fully described in Nonlinear Programming: Theory, Algorithms, and Applications. 1983. New York: John Wiley & Sons.
Applying VBMC to NGF
The aim of this computational study was to ascertain the biologically active conformation of the TrkA receptor binding determinant of NGF. A comprehensive review of the literature identified an initial study set of 17 proteins which have been assessed for TrkA mediated activity using comparable experimental methods. From these 17 analogues, a subset of six (three active and three inactive) which reflects the range of structural diversity was selected for preliminary study. Accordingly, in the present invention the low-energy conformations of the amino and carboxyl termini of six NGF analogues were obtained by means of Monte Carlo simulated annealing calculations. Three of these analogues retain full TrkA mediated activity: human NGF (hNGF; Shih et al., 1994; Burton et al., 1992), mouse NGF (mNGF; Burton et al., 1992; Drinkwater et al., 1993), and mNGF.DELTA.3-9 (deletion 3-9 of mNGF; Drinkwater et a., 1993). The other three analogues are biologically inactive: hNGF-H4D (a point mutation of hNGF in which His [H] in position 4 is replaced by Asp [D]; Shih et al., 1994), mNGF&1-8 (Burton et al., 1992; Drinkwater et al., 1993; Taylor et al., 1991) and BDNF (Suter et al., 1992). The results and conclusions obtained from these six analogues were then validated by being applied to the remaining 11 analogues of the initial study set.
The amino acid sequences of the termini of the neurotrophin molecules under study are presented in FIG. 1. Residues are numbered from 1 to 118 for the first monomer, and from 1' to 118' for the second monomer. The mutual orientation of the termini of the different monomers is determined by the disulfide bonds Cys.sup.15 -Cys.sup.80 and Cys.sup.68' -Cys.sup.110' and hydrophobic interactions Val.sup.14 :Val.sup.14' and Val.sup.109 :Val.sup.109', which are absolutely conserved in all neurotrophins (McDonald et al., 1991). The conformational space of the peptide molecules consists of flexible torsional angles determining geometry of the amino terminus of one monomer and the carboxyl terminus of the other, while keeping the rest of the dimer fixed in the known geometry of mNGF (McDonald et al., 1991). Covalent bond lengths and bond angles of the termini are kept fixed. The geometry of each residue is determined only by flexible torsional angles of the peptide backbone (.phi. and .psi.) and the side chain (x.sub.1, x.sub.2, . . . ). All torsional angles of amino acid Pro except for .psi. are fixed. Torsional angles describing structural motifs with the essentially planar geometry (such as peptide bonds or conjugated regions in side chains) are kept fixed. Likewise, the configurations of the sp.sup.3 -carbons are fixed. With these constraints, the total conformational space of the flexible termini of hNGF, mNGF and BDNF consists of 74, 73 and 67 variables, respectively.
Results and Discussion
By diagonalizing the covariant matrix in the framework of the VBMC algorithm, rigid modes of a molecular system, which should slow down conformational changes, are supposed to be revealed simultaneously with soft modes of the system. To test the efficiency of the VBMC algorithm, its performance was compared with the conventional Monte-Carlo method at the equilibrium parts of the Markov chains generated at different constant temperatures. These comparisons showed that at high temperatures (700.degree.-1000.degree. K.) the basis variability results in decreasing acceptance rates. On the other hand, the temperature decrease reinforces the advantage of the VBMC approach, and at temperatures below 600.degree. acceptance rates of the VBMC Markov chains were significantly higher. Thus, at T=300.degree. K. the acceptance rate of the conventional Markov chain was found to be 0.22, which is very close to that recommended by Kincaid and Scheraga (1982) as optimal (0.20). The acceptance rate of the VBMC algorithm obtained at the same conditions was 0.26, i.e. 18% higher. These evaluations are in agreement with what one would expect. Indeed, at low temperatures only soft modes of molecules are involved in thermal motions, hence, they are especially populated in corresponding Monte-Carlo generated conformational ensembles and can be revealed from them. Higher acceptance rates of the low temperature VBMC Markov chains over conventional Markov chains mean that the average increase of the sizes of the Monte-Carlo moves along soft modes of the system are higher than their decrease along the rigid modes. Thus, the VBMC disclosed herein for evaluating peptide conformations not only changes directions of the Monte-Carlo moves, but also accelerates conformational changes by directing them along the soft -modes of the molecule. Consequently, the present VBMC algorithm is more efficient than conventional Monte-Carlo method only at moderate and low temperatures, when thermal motions tend to take place along the valleys of the potential energy surface. This particular temperature region causes serious problems in known simulated annealing algorithms, and the VBMC method disclosed herein provides a solution to these problems.
Seven independent VBMC simulations for each of the six molecular systems under study were performed. Simultaneously, local energy refinements were periodically carried out from a number of points generated during the "slow cooling" stage. All local minimizations resulted in higher energy stationary points on the potential energy surface than those produced by VBMC computations. This means that the VBMC technique as an approach to global minimization cannot be replaced by relatively inexpensive periodic energy refinements from the high-temperature conformational space; at high temperatures only high-energy conformers are sufficiently populated. Although VBMC runs did not converge to the same conformation, the final structures differed insignificantly; i.e., the peptide backbone conformation remained unchanged but there were limited differences in the side chain torsional angles. Multiple VBMC simulations identified rigid and flexible parts of the molecules. Minimum energy structures of all molecules are displayed in FIGS. 2-7 while Table 1 presents optimized conformational parameters of the amino and carboxyl termini of the considered NGF analogues.
Geometric Similarity of Bioactive NGF Conformers
Rigid regions of hNGF and mNGF which maintain the same conformation in all independent VBMC simulations are formed by residues 9-11 and 112'-118', creating a richly hydrogen bonded (H-bonded) 3-dimensional complex. Conformational variability within this rigid zone is primarily restricted to side chains and does not affect the geometry of the backbone. An interesting example of such side chain mobility occurs with Arg.sup.114', which is H-bonded to the carbonyl oxygen of the ninth amino acid of the amino terminus and which is therefore expected to have a fixed conformation. Nevertheless, Arg.sup.114' has two distinct conformers in which the H-bonding is realized by the hydrogen atoms attached to the different NH.sub.2 groups, with the energy difference being about 1 kcal/mol. In addition to local conformational variability, hNGF and mNGF have a region of "cooperative" flexibility which is made up of residues 1-8. Several distinct conformations exist within this region, with its conformational variability being caused by a number of possibilities for the H-bond network formed by residues 1-3.
Comparison of the most stable conformers of hNGF (FIG. 2) and mNGF (FIG. 3) reveals striking conformational similarity between their rigid regions in spite of several differences in the primary structure of the amino and carboxyl termini (FIG. 1). As shown in Table 1, both termini have the same conformation, except for the variable region consisting of residues 1-3. The most stable H-bond networks in this area differ considerably because of the mT3hS substitution.
The essential difference between the amino acid sequences of the termini of hNGF and mNGF is the mM9hR substitution, which does not change the geometry of the rigid region. Residues Met and Arg have different side chain types, i.e., the former is non-polar while the latter is polar and positively charged under physiological conditions. It is surprising that this substitution does not affect geometry since it introduces the positive charge of Arg.sup.9 which is located in an electrostatic field created by at least five neighbouring charged functional groups. Conformational stability of the rigid region of hNGF is maintained by the exact fit of the Arg.sup.9 side chain into a "hollow" in the rigid region, which is occupied by the Met.sup.9 side chain in mNGF. Furthermore, the Arg.sup.9 side chain, being in this "hollow", forms two additional interchain H-bonds with carbonyl oxygens of residues Val.sup.117' and Arg.sup.118'. This further stabilizes the rigid region of hNGF.
The physical reason for the geometric similarity of the rigid regions of hNGF and mNGF is that the structural features of these two complexes are determined by similar H-bonding and electrostatic interactions. The major stabilizing factor is the H-bonding and electrostatic interaction of charged side chains of residues Glu.sup.11 and Arg.sup.118'. In addition, these side chains are involved in other H-bonding: the carboxyl oxygens of Glu.sup.11 are bonded to the .delta.-NH group of His.sup.8, the hydroxyl group of Ser.sup.113' and the backbone NH group of Lys.sup.115' ; the side chain NH and NH.sub.2 groups of Arg.sup.118' are bonded to the backbone Phe.sup.7 and Lys.sup.115' carbonyl groups. There are some other common interchain H-bonds in hNGF and mNGF. The .epsilon.-amino group of the Lys.sup.115' side chain forms a H-bond with the .delta.-carbonyl oxygen of the Asn.sup.77' residue, situated at the "south" part of NGF structure (McDonald et al., 1991).
In addition to H-bonding, the particular geometry of the rigid region in both molecules is also maintained by electrostatic interactions of positively charged residues, specifically His.sup.4, His.sup.8, Arg.sup.114', Arg.sup.118' and His.sup.75'. An electrostatic repulsion between His.sup.4 and boundary residues His.sup.8 and Arg.sup.118' of the rigid region forces His.sup.4 to be located as far as possible from the rigid region, separating the 1-8 loop from the rest of the structure. This tendency is further reinforced in hNGF by the presence of a third positively charged boundary residue Arg.sup.9. The His.sup.75' amino acid located in the "south" part of NGF also participates in maintaining the structure of the rigid region. An electrostatic repulsion between His.sup.75' and Arg.sup.114' prevents H-bonding of the amino group of Arg.sup.114' with the Asn.sup.77' side chain and the Ser.sup.113' backbone: indeed, removing the .epsilon.-proton from His.sup.75' results in two H-bond contacts which change the global energy minimum structure of the complex.
The separation of flexible loop 1-8 from the rigid region is also determined by residue Pro.sup.5. This amino acid imposes restrictions on conformational motion of the loop, making it "bent" at this particular point, and thereby defining the location of the adjacent His.sup.4 residue at the most distant point from the rigid region. Thus, residues His.sup.4 and Pro.sup.5 are primarily responsible for the separation of the flexible loop from the rigid region.
Although mNGF.DELTA.3-9 does not have the flexible loop, the energy optimized geometric features of its termini are very similar to those of the hNGF and mNGF rigid regions (Table 1, FIGS. 2-4). There are only two structural elements which are specific for the rigid region of mNGF.DELTA.3-9. First, although Arg.sup.118' has a different conformation, the result is functionally the same, namely both NH.sub.2 groups of the Arg.sup.118' side chain are H-bonded to the Glu.sup.11 side chain. In addition, the .epsilon.-NH group of Arg.sup.118', which is H-bonded to the Phe.sup.7 carbonyl group of the flexible loop in both hNGF and mNGF, faces the carbonyl terminus in mNGF.DELTA.3-9. Second, the "hollow", which is occupied by the ninth amino acid side chain in hNGF and mNGF, is now filled with the protein backbone of the remainder of the loop, i.e., residues 1-2. Functionally, the structure of the rigid region of mNGF.DELTA.3-9 resembles that of hNGF because of the stabilizing interchain H-bond contacts of the moieties located in the "hollow".
Geometric Distinction of NGF Analogues Inactive of TrkA Binding
The energy minimum structures of the complexes formed by the amino and carboxyl termini of molecules, hNGF-H4D, mNGF.DELTA.1-8 and BDNF, which are inactive with respect to TrkA binding, do not possess the common structural features found in active molecules. Point mutation H4D in hNGF incorporates negative instead of positive charge into the 1-8 loop. As a result of this mutation, the loop is no longer separated from the rigid region, forming a united complex instead (FIG. 5). In this complex the negatively charged carboxyl group of Asp.sup.4 is surrounded by four positively charged NH.sub.2 groups of Arg.sup.9 and Arg.sup.114', and residue His.sup.8 is not H-bonded to the Glu.sup.11 carboxyl group. Thus, because of the H4D point mutation, the common structural features inherent in hNGF and mNGF are lost from hNGF-H4D.
Deletion 1-8 also changes dramatically the most stable structure of the rigid region of mNGF (FIG. 6). The newly created positively charged amino terminus is located between negatively charged carboxyl groups of Glu.sup.11 and Arg.sup.118', with charged side chains of Arg.sup.114', Lys.sup.118' and Arg.sup.118' being substantially shifted from their preferred positions in active molecules. Thus, the electrostatic field caused by the positive charge of the new amino terminus in mNGF.DELTA.1-8, located in the vicinity of the carboxyl group of Glu.sup.11, is responsible for the destruction of the geometry of the rigid region inherent in the active molecules.
The complex formed by the amino and carboxyl termini of BDNF is also significantly different from those formed by the molecules active for TrkA binding (FIG. 7, Table 1). There are several reasons for the considerable conformational changes in BDNF. First, negatively charged residue Asp.sup.5 eliminates the separation of the flexible loop from the rigid region, and interacts with the positively charged His.sup.3, Arg.sup.9 and amino terminus. Second, because of its long tail, Arg.sup.8 is unable to form the H-bond contact with the Glu.sup.11 side chain and forms bifurcated H-bonds with the carbonyl group of Gly.sup.10. Third, Arg.sup.116' is H-bonded to Glu.sup.66' located in the "south" part of the molecule, which imposes a specific restriction on conformational possibilities of the carboxyl terminus of BDNF.
Explanation for The Observed Structure-Activity Relationships
The principal result of the molecular simulations presented is that the NGF molecules with full TrkA mediated activity considered in this study have regions with similar 3-dimensional structure, whereas in all inactive analogues, the geometry of this region is altered, mainly by novel electrostatic interactions. It is therefore reasonable to attribute the TrkA mediated activity to a specific conformation of this particular region. The amino and carboxyl terminal regions therefore represent a "functional epitope" which contains relatively few contact residues mediating high affinity binding. The ability of a small number of residues to cause high affinity binding has previously been demonstrated in other systems and may represent a general property of protein-protein interfaces. Some experimental data consistent with this concept are presented below.
The VBMC computations show that the following amino acids are critical in determining the structure of the complex in wild type NGF dimers: His.sup.4, Pro.sup.5, His.sup.8, Arg.sup.9, Glu.sup.11, His.sup.75', Asn.sup.77', Lys.sup.115' and Arg.sup.118'. This is significant since residues His.sup.4, Pro.sup.5, Glu.sup.11, His.sup.75', Asn.sup.77' and Lys.sup.115' are conserved in NGF molecules across all species (McDonald et al., 1991). The overlap between these two sets reflects a tendency to maintain the particular geometry of the complex in natural NGF molecules. In addition, as demonstrated, the substitution of the important residue Arg.sup.9 by the essentially different Met.sup.9 does not result in geometric changes of the complex, which is consistent with equivalent bioactivity of hNGF and mNGF (Burton et al., 1992). Thus, because of multiple physical factors maintaining the geometry of the complex, some species variability observed in NGF termini (McDonald et al., 1991) is acceptable and does not cause geometric changes in the rigid region.
It is shown that each NGF terminus has several important residues, which determine the particular structure of the complex. This is consistent with high sensitivity of biological activity of NGF to the primary structure of both the amino and carboxyl termini (Klein et al., 1991; Shih et a., 1994; Burton et al., 1992, 1995; Kahle et al., 1992; Drinkwater et al., 1993; Treanor et a., 1995; Taylor et al., 1991). The lack of the TrkA mediated activity of the mNGF.DELTA.3-11, mNGF.DELTA.3-12, mNGF.DELTA.3-13, mNGF.DELTA.3-14, mNGF.DELTA.9-14 and mNGF.DELTA.112-118 deletion mutants (Drinkwater et al., 1993) is thus explained by the absence of the key residues Glu.sup.11 and Arg.sup.118'.
The separation of the flexible loop from the rest of the complex is crucial in permitting the rigid region to exist with definite geometric features. Residues His.sup.4 and Pro.sup.5 which cause this separation are absolutely conserved elements of the wild type NGF molecules derived from different species (McDonald et al., 1991), and their compulsory location in the middle of the flexible loop is well suited for facilitating the maximal separation. Point mutation P5A or H4D weakens or destroys the tendency of the loop to separate from the rigid region; accordingly, these mutations result in dramatic loss of TrkA related activity (Shih et al., 1994).
The flexible loop is not an essential structural element for TrkA activation since the mNGF.DELTA.3-9 deletion mutant, which does not contain the loop (FIG. 4), is as active as wild type mNGF (Drinkwater et al., 1993). However, complete elimination of all loop residues results in almost inactive mutants mNGF.DELTA.1-8 (Burton et al., 1992; Drinkwater et al., 1993; Taylor et al., 1991) and hNGF.DELTA.1-9 (Shih et al., 1994; Burton et al., 1992, 1995; Kahle et al., 1992; Luo and Neet, 1992). Our molecular simulation studies enable the explanation for this apparent contradiction. It is demonstrated that the backbone of residues Ser.sup.8 -Ser.sup.9 in the mNGF.DELTA.3-9 mutant is naturally incorporated into the structure of the rigid region because it matches the geometry of the "hollow" occupied by either Met.sup.9 or Arg.sup.9 side chain in wild type NGF dimers. On the contrary, total truncation of the flexible loop in the mNGF.DELTA.1-8 deletion mutant results in placing the positively charged amino end in the vicinity of the key H-bonded Glu.sup.11 and Arg.sup.118' charged side chains (FIG. 3), which abolishes the desired geometry of the rigid region (FIG. 6). The same applies to the hNGF.DELTA.1-9 deletion mutant which has neither the 1-8 loop nor the stabilizing Arg.sup.9 residue (FIG. 2). This being the case, biological activity of amino terminus deletions of NGF should be determined by the length of the remainder of the terminus rather than its particular sequence. Indeed, available experimental data confirm this point. Deletion mutant mNGF.DELTA.2-8 which has exactly the same number of residues in the amino terminus as the highly active mNGF.DELTA.3-9 mutant is also very active (Drinkwater et al., 1993). Some decrease in biological activity of mNGF.DELTA.2-8 is likely to be caused by unfavorable directing of the Met.sup.9 hydrophobic side chain away from the complex whereas in the mNGF.DELTA.3-9 deletion mutant this side chain is replaced by the hydrophilic hydroxyl group (FIG. 4). Deletion mutants mNGF.DELTA.3-8 (Drinkwater et al., 1993) and hNGF.DELTA.1-5 (Shih et al., 1994), which contain a greater number of amino terminus residues, are remarkably less active than wild type NGF because the remainder of the loop is too long to occupy the "hollow" but too short to separate from the rigid region. The present explanations for the observed structure-activity relationships enable the prediction that, for example, a S2G mutation in the mNGF.DELTA.3-9 deletion mutant would retain high biological activity.
Although the bioactive conformation of the NGF binding determinant has been disclosed above, the question of which NGF residues are involved in NGF/TrkA recognition was not yet addressed. One would reason, most probably, those residues which are primarily responsible for maintaining the particular structure of the NGF termini do not directly interact with the receptor. Thus, side chains of His.sup.4, Pro.sup.5, His.sup.8, Met.sup.9, Arg.sup.9, Glu.sup.11 and Arg.sup.118' are not expected to play a crucial role in molecular recognition. On the other hand, charged residues Arg.sup.114' and Lys.sup.115' which are conserved in NGF molecules across all species (McDonald et al., 1991) but not involved in principal intramolecular electrostatic interactions are candidates for the centres of intramolecular electrostatic recognition. Although both Arg.sup.114' and Lys.sup.115' participate in stabilizing intramolecular H-bonding within the bioactive conformations of the NGF termini, their electrostatic properties as charged residues are not utilized for this purpose.
The other structural element of the termini with potential for stereochemical fit to the TrkA receptor is a short exposed .beta.-sheet motif located between the .alpha.-carbons of residues 7 and 9 (FIGS. 2,3). On the one hand, H-bonding between .beta.-pleated sheets is rather common for protein-protein interactions (Kobe and Deisenhofer, 1993, 1994, 1995; Yoder et al., 1993; Baumann et al., 1993). On the other hand, a short exposed .beta.-sheet motif is generally inherent in the structure of a leucine-rich repeat (Kobe and Deisenhofer, 1993, 1994, 1995) which has been recently identified as the TrkA recognition site for NGF (Windisch et al., 1995a, 1995b). Other regions of NGF might also be involved in the NGF/TrkA recognition (Shih et al., 1994; Ibanez et al., 1993). This will be discussed in depth below.
Conclusions
The Variable Basis Monte Carlo simulated annealing technique disclosed herein enables analysis of the low-energy conformational space of peptide molecules. Although multiple VBMC calculations do not consistently converge to the global energy minimum structure, they do permit the biologically active conformation to be evaluated. Definite physical factors maintain and distinguish the bioactive conformation of a molecule, and these factors can be understood by analyzing the low-energy conformational space. The precisely defined global energy minimum structure is unnecessary, especially considering that any empirical force field functions are not ideal. In biologically relevant molecular modelling, exploring the bioactive conformation is more important than determining the global energy minimum. VBMC is a useful tool for facilitating this exploration.
The present VBMC simulations predict that the amino (1-11) and carboxyl (112'-118') termini of dimeric (1-118)NGF form a complex comprised of a rigid region and a conformationally variable loop. The rigid region is formed by residues 9-11 and 112'-118' while the variable loop is formed by residues 1-8. The major stabilizing factor of the rigid region is the ionic contact of the Glu.sup.11 and Arg.sup.118' side chains. The separation of the loop from the rigid region arises from the electrostatic repulsion between His.sup.4 and boundary residues of the rigid region His.sup.8 , Arg.sup.9 (in hNGF) and Arg.sup.118'. The geometry of the complex explains observed structure-activity relationships. Accordingly, it is the conformation of the rigid region which is recognized by the TrkA receptor, defining the biologically active conformation of NGF for TrkA activation.
The split of the complex of the NGF termini into two distinct moieties could have important biological implications. Although only the structure of the rigid region is recognized by the TrkA receptor, the flexible loop is able to control conformational stability of the biologically active conformation of the rigid region and, consequently, NGF-induced TrkA phosphorylation.
PART B
Theoretical Modelling of Trk-Receptor Recognition Sites for NGF and BDNF
Introduction
As mentioned above, the location of the neurotrophin binding sites within the Trk receptors is the subject of debate (MacDonald and Meakin, 1996). Based upon published data, there is reason to believe that Trk proteins have two putative neurotrophin binding sites: the "second immunoglobulin-like domain" (IgC2, residues Trp.sup.299 -Asn.sup.365) and the "second leucine-rich motif" (LRM-2A and LRM-2B, residues Thr.sup.97 -Leu.sup.120 of TrkA and TrkB, respectively) (Schneider and Schweiger, 1991; Kullander and Ebendal, 1994; Perez et al., 1995; Urfer et al., 1995; Windisch et al., 1995a, 1995b, 1995c; MacDonald and Meakin, 1996; Ryden and Ibanez, 1996). LRM containing proteins are a diverse group with different functions and cellular locations, and are known to mediate strong and selective protein-protein interactions (Kobe and Deisenhofer, 1993,1994,1995). The amino acid sequence of an LRM is characterized by conserved hydrophobic consensus residues (Schneider et al., 1988; Schneider and Schweiger, 1991; Kobe and Deisenhofer, 1993, 1994, 1995). Typically, LRMs exist as repetitive cassettes, with an individual LRM representing a right-handed .beta.-.alpha. hairpin unit (Kobe and Deisenhofer, 1993, 1994,1995). Recent crystallographic studies on porcine ribonuclease inhibitor indicate that the side chains of the consensus residues form the hydrophobic core of the LRM modules, thereby determining their geometric features (Kobe and Deisenhofer, 1993).
According to the invention, the inventors report explicit atom-level models of the receptor sites for NGF and BDNF based on the structures of the second LRMs of TrkA (LRM-2A) and of TrkB (LRM-2B), respectively. To deduce these models, we have investigated the interaction of the bioactive conformation of the NGF amino and carboxyl termini region with the LRM-2A and of the "waist" region of BDNF with the LRM-2B. A series of four active NGF analogues, namely human NGF (hNGF) (Burton et al., 1992; Shih et al., 1994), mouse NGF (mNGF) (Burton et al., 1992; Drinkwater et al., 1993), the 3-9 deletion of mNGF (mNGF.DELTA.3-9) (Drinkwater et al., 1993) and the hNGF:NT-4/5 heterodimer (Treanor et al., 1995), have been docked to the LRM-2A of rat TrkA. For all four NGF molecules studied, a stereochemical fit of the LRM-2A to the amino and carboxyl region was achieved. These results indicate that the bioactive conformation of the NGF termini is a functional epitope which mediates high affinity binding via an interaction with the LRM-2A. This approach has also been applied to BDNF/LRM-2B docking, and their stereochemical fit, based mainly on electrostatic complementarity, has been identified and is described below.
Methods
A sequence of three logical steps has been used to deduce the models of the interactions of NGF and BDNF with their respective Trk receptor binding sites: Step 1: The bioactive conformations of the amino and carboxyl termini of NGF analogues were determined. Step 2: The residues of the Trk proteins which form potential binding sites for NGF and BDNF were identified. Step 3: The interaction of the bioactive conformation of the NGF amino and carboxyl termini with the TrkA binding site and the BDNF .beta.-strand stem region with the TrkB binding site were modeled computationally. Finally, the predicted features of the NGF/TrkA and BDNF/TrkB interaction models were validated against available experimental data. Residues of the first monomer of a neurotrophin are designated 1 to 118, those of the second monomer are designated 1' to 118'. Residues of the TrkA LRM-2A and the TrkB LRM-2B binding sites are designated by abbreviations tA and tB, respectively. Residues of the BDNF monomer are designated 1 to 119.
Step 1: Revealing the Bioactive Conformative of the NGF Termini
The most energetically favored conformations of the amino and carboxyl termini of NGF analogues have been obtained by VBMC computations and are described above. Since the 3-dimensional structures of the termini of fully active NGF molecules were found to be different from those of TrkA-inactive analogues, the former conformations were predicted to be "bioactive" for TrkA-mediated activity of NGF. That is, those geometric features that are exclusively inherent to active NGF molecules are expected to be compatible with the structure of the TrkA binding site for NGF. Since the geometry of the central .beta.-strand stem of BDNF is rigid and has been established by X-ray crystallography (Radziejewski and Robinson, 1993), this logical step is bypassed for BDNF. Instead, it is assumed that the X-ray crystallography deduced conformation of the amino acid backbone of the TrkB binding determinant of BDNF represents its "bioactive" conformation.
Step 2: Selection of The LRM-2 Boundaries Within The TRK Receptors
Since there is controversy in the literature regarding the actual boundaries of LRMs (Schneider et al., 1988; Schneider and Schweiger, 1991; Kobe and Deisenhofer, 1993, 1994, 1995; Windisch et al., 1995a, 1995b), the following experimental background (FIG. 8) was used to define boundaries for this computational study. According to Schneider and colleagues (1988, 1991), the LRM domains of TrkA begin with Leu.sup.68tA, Leu.sup.93tA and Leu.sup.117tA, whereas those boundaries are slightly shifted in the crystallographic studies of Kobe and Deisenhofer (1993). Furthermore, the boundaries of the "minimal" LRM module of TrkA (designated B4 in FIG. 8), which demonstrates the same NGF binding affinity as wild type TrkA, are substantially shifted towards the carboxyl terminus (Windisch et al., 1995a, 1995b) with respect to those suggested by Schneider and colleagues (1988, 1991). The same boundary shift has been suggested for the LRM-2B module (Windisch et al., 1995a, 1995b, 1995c). However, a comparison of active (B1 and B4) and inactive (N4) LRM domains indicates that the presence of the "minimal" amino acid sequence (B4) does not ensure NGF binding. Understanding those factors which maintain the essential features of the LRM geometry in segments B1 and B4, and those factors which destroy them in N4, is of fundamental relevance to the correct assignment of the LRM boundaries.
.beta.-Strand motifs exist in proteins only if hydrogen bonded (H-bonded) to other .beta.-strands; otherwise their structure is unstable (Kobe and Deisenhofer, 1994). Tandem repeats of LRMs represent a prominent example of H-bonded parallel .beta.-strand motifs. The general location of the .beta.-strand motifs in the LRM, according to X-ray data, is illustrated in FIG. 8; however, in a single LRM their particular geometry is destroyed (Kobe and Deisenhofer, 1993, 1994, 1995). Consequently, the only possible way to exclude dramatic conformational changes in a protein domain upon disruption of the inter-strand H-bonding between .beta.-strand motifs is to avoid those motifs. Accordingly, both boundaries of segment B4 are located in the middle of the N-strand regions of the adjacent LRMs such that this segment appears to be free of any .beta.-strand. Nevertheless, the .beta.-strand motif is a characteristic structural unit of LRMs, and thus it is likely that this unit is involved in the NGF/TrkA recognition process. Additionally, a short exposed .beta.-strand motif has been detected in the bioactive conformation of the NGF termini, as described above. The incorporation of the .beta.-strand regions at both sides of the B4 segment (as in N4) results in cooperative H-bonding between these P-strands and a complete loss of NGF binding, which indicates that accessibility of the P-strand motif for intermolecular contacts is essential. On the other hand, the separation of the entire LRM segment from the adjacent LRMs within the B1 molecule as a result of conformational kinking is likely to be energetically more favorable, since the entire LRM module has the internal stabilizing factors. Moreover, kinking between adjacent LRMs does not disrupt the structure of their .beta.-strand motifs (since they remain H-bonded with their opposite sides), but makes them accessible for molecular recognition processes.
Accordingly, in the modelling of the present invention we used the conventional boundaries of the LRMs of TrkA and TrkB (Schneider et al., 1988; Schneider and Schweiger, 1991), which keep the .beta.-strand motif in the LRM, and we explicitly arranged H-bonding of this unit with the parallel .beta.-strand motif of the NGF amino terminus to preserve it from conformational reorganizations. Thus, the segments between residues Leu.sup.93tA and Leu.sup.117tA in TrkA, and Leu.sup.93tB and Leu.sup.117tB in TrkB, were respectively designated the LRM-2A and the LRM-2B.
Step 3: Computational Simulation of The Neurotrophin/Receptor Interactions
Docking interactions NGF/LRM-2A and BDNF/LRM-2B were simulated using a combined VBMC/CHARMM approach. As described above, the VBMC algorithm permits conformational space search; CHARMM (Brooks et al., 1983) is a force field which defines conformational energy. The same approach was also used to study improper docking interactions of the central .beta.-strand stem region of NGF with the LRM-2B, as a negative control.
The VBMC algorithm performs conformational and orientational space searches of a complicated molecular system with the aim of finding the structure which corresponds to the global minimum on the potential energy E.sub.tot surface.
E.sub.tot =E.sub.t +E.sub.w +E.sub.e +E.sub.hb +E.sub.qq +E.sub.hy +E.sub.c (4)
The following potential energy terms have been considered in the framework of the united atom CHARMM force field (Brooks et al., 1983; The force field parameters are extracted from CHARMM version 23.1, Molecular Simulations, Inc., Waltham, Mass.): torsional distortion E.sub.t, van der Waals interactions E.sub.w, electrostatic interactions E.sub.e, and hydrogen bonding E.sub.hb. Since the united atom approach is not appropriate to treat rigorously the quadrupole-quadrupole interactions of benzene rings (which are important for determining their mutual orientations in proteins (Hunter and Sanders, 1990; Hunter et al., 1991), single quadrupoles have been placed at the centers of the benzene rings of phenylalanine, tryptophan and tyrosine, and the energy of the quadrupole-quadrupole interactions, E.sub.qq, has been calculated as suggested by Schauer and Bernstein (1985). (The potential energy function of two interacting benzene molecules, comprising E.sub.qq and E.sub.w terms, has stationary points corresponding to both known electrostatically favorable "edge-to-face" and "offset stacked" orientations, with the dimerization energy of 1.94 kcal/mol agreeing well with ab initio quantum mechanical calculations (Hobza et al., 1994).) In addition, the conformationally dependent free energy of hydration, E.sub.hy, has been calculated according to the hydration shell model including both nonspecific and specific hydration terms (Hopfinger, 1973; Hodes et al., 1979). Point atomic charges of Weiner and colleagues (1984) have been used. A dielectric constant equal to the interatomic separation in angstroms has been used to implicitly simulate salvation effects (Weiner et al., 1984).
The E.sub.c term is used only for the NGF/LRM-2A complex and represents the distance constraints which restrain the H-bonding of residues Glu.sup.11 and Arg.sup.118' within the bioactive conformation of the NGF termini, as described above. In addition, two intermolecular H-bonds, arising from the alignment of complementary parallel .beta.-strand motifs of NGF and LRM-2A, are also constrained and contribute to E.sub.c. In the Monte Carlo conformational space search, E.sub.c is defined as follows:
E.sub.c =10.times..SIGMA.(d.sub.AB -1.8).sup.2, (5)
where d.sub.AB is the distance in angstroms between atoms A and B which are to be H-bonded. The E.sub.c term is neglected when performing final energy minimization following the VBMC simulated annealing. In addition, this term has not been utilized when studying the mNGF.DELTA.3-9/LRM-2A complex. Since, as described above, the mNGF.DELTA.3-9 deletion mutant does not have a .beta.-strand region, the geometry of its LRM-2A counterpart, namely Leu.sup.93tA -lle.sup.98tA, has been kept fixed in the crystallographic conformation (Kobe and Deisenhofer, 1993) during Monte Carlo simulations. The E.sub.c term, being "soft", does not preclude conformational and orientational changes to occur during Monte Carlo simulations. This term increases rigidity of the bioactive conformation of the ligand, which allows us to utilize the knowledge obtained at Step 1, instead of explicit considerations of inactive NGF analogues, in Step 3. In addition, this term keeps conformations of the .beta.-strands of both ligand and binding site from destruction, and thereby allows us to avoid an explicit consideration of the adjacent first LRM of TrkA. Furthermore, this term maintains the known orientation of the parallel .beta.-strand motifs in the NGF/LRM-2A complex.
The amino acid sequences of the binding determinants of the NGF molecules studied are presented in FIG. 9. The conformational space of the NGF analogues consists of torsional angles determining geometry of the amino terminus of one monomer (residues 1-11) and the carboxyl terminus of the other (residues 112'-118'), while keeping the rest of the dimer in the experimental geometry of mNGF (McDonald et al., 1991). The conformational spaces of the LRM domains of TrkA and TrkB consist of torsional angles representing fragments Arg.sup.94tA -Leu.sup.117tA and Arg.sup.94tB -Leu.sup.117tB, respectively. Covalent bond lengths and bond angles have been kept fixed. The geometry of each i-th residue is determined only by flexible torsional angles of the backbone (.phi..sub.i and .psi..sub.i) and the side chain (X.sub.i 1, X.sub.i 2, . . . ). All torsional angles for prolines, except .psi..sub.i have been fixed. Torsional angles describing structural motifs with planar geometries (such as peptide bonds or conjugated regions in side chains) were fixed. Likewise, the configurations of the sp.sup.3 -carbons were fixed. The conformational space of BDNF consists only of torsional angles determining side-chain geometries of those residues which are located in the vicinity of its central .beta.-strand stem and define a molecular surface (Table 2). In addition to the conformational variables of each complex, the six orientational variables (three Euler angles and three translations along the coordinate axes) were used to describe the ligand-receptor orientations.
In accordance generally with the simulated annealing approach (Kirkpatrick et al., 1983; Vanderbilt and Louie, 1984; Brunger, 1998; Nilges et al., 1988; Wilson et al, 1988; Wilson and Doniach, 1989), the molecular systems were kept at a constant temperature of 800.degree. K. for 2.times.10.sup.6 Monte Carlo moves before the initiation of "cooling". The temperature was then decreased at every Monte Carlo move by a small increment such that after 2.times.10.sup.6 Monte Carlo moves it reached 120.degree. K. At this time, a local energy refinement was performed to identify the nearest stationary point on the potential energy surface. According to the VBMC algorithm described above, the basis of independent structural parameters was recalculated after every 7.times.10.sup.4 Monte Carlo moves to direct them along the valleys of the potential energy surface, thereby accelerating conformational and orientational changes.
All calculations were performed on an Indigo-2 Silicon Graphics Workstation and an 8-node IBM Scalable POWER parallel 2 [SP2] high performance computer.
Results and Discussion
Conformational Flexibility within Receptor Environments
Five independent VBMC simulations from different initial configurations were performed for each of the six molecular complexes under study. The elimination of the distance constraints E.sub.c in the complexes with the LRM-2A did not result in any noticeable conformational changes within the complexes. Although independent VBMC simulations did not always converge to the same geometry for the same molecule, the final structures differed insignificantly; i.e., the conformations of the peptide backbones remained unchanged with the exception of residues 1-3 of the NGF amino terminus and the terminal residue Leu.sup.117tA of the LRM-2A. The differences between the VBMC generated structures were located in side-chain conformations of a limited number of residues, specifically Ser.sup.1, Val.sup.117' (Thr.sup.117' in mNGF), Thr.sup.97tA and Ser.sup.101tA. These residues are not involved in intermolecular or intramolecular interactions, and do not affect the overall structures of the complexes. No considerable differences were identified in the independent VBMC-generated conformations of the BDNF/LRM-2B and the NGF/LRM-2B complexes.
A grid search of the internal space of the complexes using the water molecule radius 1.4 .ANG. as a probe (Ponnuswamy and Manavalan, 1976) (as incorporated into the QUANTA program, Version 4.1, Molecular Simulations, Inc., Waltham, Mass.) showed that there is no internal cavity available for a solvent molecule in the hNGF/LRM-2A complex. This explains the rigidity of the complex and demonstrates the "perfect" stereochemical fit of the hNGF amino-carboxyl termini complex to the LRM-2A. Other complexes have some internal cavities, especially with the LRM-2B binding site.
Specific Binding Areas within Receptor Environments
The TrkA Receptor Binding Site
Comparison of the most stable structures of the different complexes with the LRM-2A binding site (FIG. 10) shows their striking conformational similarity. In all molecules, ligand-receptor binding is caused by similar hydrophobic, ionic, H-bonding and van der Waals interactions. There are five distinct areas (A-E) within the NGF/LRM-2A fit: areas A 30 and B have hydrophobic interactions, areas C and D have ionic H-bonding, and area E has cooperative H-bonding between parallel .beta.-strand motifs.
Area A represents the hydrophobic region at which residue Phe.sup.105tA interacts with Trp.sup.76', which in turn forms hydrophobic contacts with Phe.sup.12 and Leu.sup.112'. The aromatic rings of Phe.sup.12, Trp.sup.76' and Phe.sup.105tA are approximately perpendicular to each other, which is energetically favorable (Hunter and Sanders, 1990; Hunter et al., 1991; Hobza et al., 1994). This particular geometry would create a hole within the complex, if it was not filled by hydrophobic residue Leu.sup.112'.
In area B, Phe.sup.7 of NGF is placed within the hydrophobic core of the LRM-2A created by the consensus residues (Schneider et al., 1988; Schneider and Schweiger, 1991; Kobe and Deisenhofer, 1993, 1994, 1995). Similar to the favorable orientation of the aromatic residues in area A, the aromatic rings of Phe.sup.111tA and Phe.sup.113tA are perpendicular to that of Phe.sup.7. On the other hand, residues Phe.sup.111tA and Phe.sup.113tA are not perpendicular to each other; their orientation instead resembles the favorable "offset stacked" geometry described by Hunter and others (Hunter and Sanders, 1990; Hunter et al., 1991; Hobza et al., 1994). As in area A, a hole created by the three neighboring favorably oriented aromatic rings is filled by a fourth residue, Thr.sup.114tA. Interactions between aromatic rings in proteins is very common and represent an important conformational driving force (Warme and Morgan, 1978; Hunter et al., 1991). Consequently, the fact that all three phenylalanine residues of the LRM-2A find their counterparts within the NGF binding domain supports the calculated structure of the NGF/LRM-2A complex.
In ionic binding area C, the negatively charged residue Asp.sup.109tA exerts a central role. It is surrounded by the three positively charged amino acids His.sup.112tA, Arg.sup.114' and Lys.sup.115' in hNGF, mNGF and mNGF.DELTA.3-9, and by His.sup.112tA, Arg.sup.115' and Arg.sup.9 in hNGF:NT-4/5. Residue His.sup.112tA plays a secondary role: it creates the electrostatic dipole Asp.sup.109tA .fwdarw.His.sup.112tA in the LRM-2A, which imposes constraints on the spatial positions of the recognized positively charged residues of NGF, thereby reinforcing geometric selectivity in this area. Although the hNGF:NT-4/5 heterodimer has two substitutions in the C-area segment of the NT-4/5 carboxyl terminus, namely R114S and K115R, the geometry of the C area remains qualitatively the same, since residues Arg.sup.9 and Arg.sup.115' of hNGF:NT-4/5 functionally replace the absent Lys.sup.115' and Arg.sup.114' amino acids, respectively. Although Arg.sup.9 also exists in the hNGF homodimer, it is not involved in the ionic binding in this molecule, since it is separated from area C due to electrostatic repulsions. Similarly, the charged amino end of mNGF.DELTA.3-9, which functionally replaces Arg.sup.9, is also separated from area C. There is no place for the fourth positively charged residue in area C. Thus, in all molecules studied, the two positively charged amino acids of the ligand saturate the coordination possibilities of the receptor residue Asp.sup.109tA.
Area C is partially isolated from its surrounding environment by flanking hydrophobic residues Val.sup.106tA and Pro.sup.115tA. Similarly, the principal charged residues of NGF, namely His.sup.8, Glu.sup.11 and Arg.sup.118', are located between hydrophobic residues lle.sup.6 (Val.sup.6 in mNGF), Val.sup.111' (Leu.sup.113' in hNGF:NT-4/5) and Val.sup.99tA within the complexes. Thus, residues Val.sup.911tA, Val.sup.106tA and Pro.sup.115tA of the LRM-2A participate in "compartmentalization" of the principal ionic interactions of the complexes, thereby shielding them from the conformation destructive influence of salvation.
The organization of ionic binding area D is similar to that of area C. Positively charged residues Arg.sup.59 and Arg.sup.69 are located in the vicinity of negatively charged Asp.sup.16 within the NGF structure (McDonald et al., 1991; Bradshaw et al., 1994), such that the coordination of the latter is not full. The third positively charged residue, Lys.sup.100tA, comes from the LRM-2A to complete the Asp.sup.16 coordination. In addition to the ionic binding, residue Lys.sup.100tA forms an H-bond with the Ser.sup.13 side chain.
Area E consists of the H-bonded parallel .beta.-strand motifs of the LRM-2A and NGF (FIG. 10b, 10c and 10e), each of the .beta.-strands having three peptide bonds. The LRM-2A .beta.-strand motif is located between the .alpha.-carbon atoms of Asn.sup.95tA and Ile.sup.98tA, whereas the NGF .beta.-strand is located between the .alpha.-carbons of the sixth and ninth residues. The locations of the .beta.-strand regions in the uncomplexed LRM (Kobe and Deisenhofer, 1993) and NGF (see above) resemble those in the complexes considered. In the complexed LRM-2A, the .beta.-strand motif is shortened by one peptide bond at its amino end to accommodate the His.sup.4 side chain of NGF. On the other hand, the two peptide bond .beta.-strand of the uncomplexed NGF is lengthened by one peptide bond at its amino end in the complexes.
The TrkB Receptor Binding Site
FIG. 11a illustrates the central N-strand stem region of BDNF, and FIG. 11b presents the VBMC-generated structure of the BDNF/LRM-2B complex. In this complex, ligand-receptor binding is based mainly on intermolecular electrostatic interactions of either ionic or hydrogen bonding origin. There are five distinct binding areas (F-K) within the TrkB receptor environment: four salt bridges (areas F, G, H and 1) and one H-bonding area K. In contrast with the NGF/LRM-2A binding pattern, there is no participation of the LRM-2B .beta.-strand motif in cooperative H-bonding with the BDNF binding epitope.
In area F, the primary positively charged residue (Ibanez et al., 1993) of the BDNF binding domain, Arg.sup.81, is bound to Asp.sup.100tB, the only negatively charged residue of the LRM-2B. It should be noted that Arg.sup.81 is exclusively inherent in BDNF molecules while being substituted by electrostatically neutral Thr in NGF or by Lys in NT-3 or NT-4/5 molecules (Bradshaw et al., 1994), which indicates the importance of this area in neurotrophin recognition processes (see also Urfer et al., 1994). On the contrary, residue Gln.sup.84, identified by Ibanez and co-workers (1993) as important, is not directly involved in specific intermolecular interactions within the LRM-2B domain.
In binding area G, the negatively charged residue Glu.sup.55, which is absolutely conserved in all neurotrophins (Bradshaw et al., 1994), exerts a central role. It is surrounded by the three positively charged amino acids, Lys.sup.25, Lys.sup.57 and Lys.sup.109tB. Two of them, belonging to the neurotrophin, are also highly conserved (Bradshaw et al., 1994). The third one, Lys.sup.109tB, is specific for the TrkB receptor site (Windisch et al., 1995a, 1995b, 1995c). All BDNF residues of the G binding area, namely Lys.sup.25, Lys.sup.57 and Glu.sup.55, have been shown to be essential in NT-3 structure for TrkC receptor activation (Urfer et al., 1994), which is in agreement with observations that the TrkB and TrkC binding domains are located in the same region of these neurotrophins. This particular arrangement of the three positively charged residues around one with a negative charge seems to be very common in neurotrophin-receptor recognition processes, since the NGF/LRM-2A interaction also includes two binding areas of this type (C and D). Existence of the absolutely conserved negatively charged residue Asp.sup.106 in direct proximity to area G increases binding affinity of Lys.sup.109tB, thereby highlighting some importance of this area in the BDNF/TrkB recognition processes. Although detailed mutational experiments did not demonstrate importance of the equivalent residue Asp.sup.105 in NT-3 for TrkC related activities (Urfer et al., 1994), its direct involvement in the Zn.sup.2+ mediated regulation of neurotrophin bioactivities has been recently proven by multidisciplinary studies (Ross et al., 1997), and, therefore, its location within the binding determinants of BDNF and NT-3 is consistent with these observations.
In area H, conserved residue Asp.sup.24 of BDNF forms a salt bridge with specific TrkB residue Lys.sup.113tB. Although Asp.sup.24 has not been demonstrated to be involved in direct neurotrophin-receptor interactions, either for TrkB or TrkC related activities, it is highly conserved in neurotrophins (Bradshaw et al., 1994) and located in direct proximity to Thr.sup.22 in NT-3, which has been shown to be included in the TrkC receptor binding domain (Urfer et al., 1994). These observations are consistent with the present model of BDNF/TrkB binding, since substitution of the equivalent residue of BDNF, Thr.sup.21, by bulky Gln (Urfer et al., 1994) would sterically hinder proper location of the LRM-2B domain (FIG. 11b), whereas mutation D24A, although lacking one specific ligand-receptor interaction, would not significantly alter the BDNF/LRM-2B orientation resulting from multiple-point electrostatic complementarity. In addition, the absence of negatively charged residue Asp.sup.24 is shown to induce salt bridge interaction of LyS.sup.113tB with neighboring Glu.sup.18, with only minor conformational alterations of the LRM-2B.
In area I, Glu.sup.18, a specific residue of BDNF (Bradshaw et al., 1994), forms a salt bridge with the LRM-2B amino acid Arg.sup.94tB. Residue Glu.sup.18 is also involved in intramolecular electrostatic interactions with Lys.sup.57.
Area K is the only region of the distinct BDNF/LRM-2B interactions which is not of ionic nature. The positively charged residue Lys.sup.104tB forms multiple H-bond contacts with the BDNF loop I backbone carbonyl groups and the neutral Thr.sup.35 side chain. The energy of such H-bonding is much lower than that of salt bridge contact, and, therefore, a substitution of the uncharged amino acid Thr.sup.35 by a negatively charged residue would stabilize the particular structure of the BDNF/LRM-2B complex. In accordance with the present model, the experimental substitution of BDNF residues 34-41 by the corresponding segment of NGF, which includes the substitution of negatively charged residue Glu.sup.35 for Thr.sup.35, resulted in a significant increase in bioactivity of the resulting chimeric molecule relative to wild type BDNF in TrkB mediated survival (Lai et al., 1996).
In addition to these five distinct binding areas of the BDNF/LRM-2B complex, there is one more area which reinforce binding specificity. The side chain of Pro.sup.60, a specific residue of BDNF (Bradshaw et al., 1994), demonstrates a perfect geometric fit into the hollow in the "zigzag"-shaped .beta.-strand region of the LRM-2B (FIG. 11b). This geometric complementarity indicates that Pro.sup.60 "recognizes" the .beta.-strand motif of the receptor binding site. This interaction, obviously, does not significantly contribute to the total energy of the complex in the case of a perfect fit, but does represent a steric hindrance when this fit cannot be readily achieved. Accordingly, BDNF is very selective and does not bind non-cognate Trk receptors.
FIG. 11c illustrates the VBMC-generated geometry of the "improper" NGF/LRM-2B complex. In this complex, specific binding areas F and I are absent due to lack of properly charged residues at positions 20 and 81, which would correspond to BDNF residues Glu.sup.18 and Arg.sup.81, respectively. In addition, residue Pro.sup.60, determining the proper BDNF/LRM-2B orientation, is substituted in NGF molecules (Bradshaw et al., 1994). These substitutions cause dramatic conformational and orientational alterations in the complexed LRM-2B domain (compare FIG. 11b and 11c). Predicted Cartesian 3-dimensional coordinates of atoms of the TrkB LRM-2B binding site in the complex with BDNF are presented in Appendix 2.
Ligand-Receptor Interaction Energy
NGFITrkA Interactions
Table 3 presents the components of the intermolecular energies of the NGF/TrkA complexes. Stability of the NGF/LRM-2A complexes results primarily from attractive electrostatic and van der Waals interactions, with intermolecular H-bonding playing a secondary role. It should be noted, however, that in the framework of the molecular mechanical approach, the H-bonding energy is partially incorporated into the electrostatic interaction energy term, so the tabulated E.sub.hb values reflect only part of the H-bonding energy. Accordingly, intermolecular H-bonding is more important for the NGF/LRM-2A interaction than it appears in Table 3. On the other hand, the intermolecular component of the E.sub.qq term seems to be insignificant. The reason for the latter is that the aromatic moieties of Phe.sup.7, Trp.sup.76', Phe.sup.105tA, Phe.sup.111tA and Phe.sup.113tA are "preorganized" for intermolecular contacts with their counterparts in the complexes in electrostatically attractive geometry in which the quadrupole-quadrupole interaction energy term is not significant. In addition, hydrophobic ligand-receptor interaction energy is presumably extremely important for NGF/LRM-2A complex formation, but rigorous quantitative evaluation is currently not possible.
Since neither electrostatic interactions nor hydrogen bonding can significantly stabilize a ligand-receptor complex in a water environment, the stability of the complex generally results from van der Waals and hydrophobic interactions (Clackson and Wells, 1995). Considerable contribution of intermolecular van der Waals interaction energy E.sub.w (Table 3) arises from the geometric complementarity of the NGF binding domain and the TrkA recognition site, such that the high value of the E.sub.w term is due to a large number of intermolecular van der Waals contacts. This complementarity is a consequence of the right-handed twists of both the NGF amino-carboxyl termini complex and the LRM-2A, this geometric feature being characteristic of each of these domains in both uncomplexed (Kobe and Deisenhofer, 1993; Shamovsky et al., 1996) and complexed conformations. Thus, purely geometric ligand-receptor compatibility together with distinct hydrophobic contacts play a crucial role in NGF/TrkA recognition, while attractive electrostatic interactions and H-bonding further reinforce specificity of binding, which is in agreement with generalizations made by Clackson and Wells (1995).
BDNF/TrkB and NGF/TrkB Interactions
Table 4 presents the components of the potential energies of the complexes of BDNF and NGF with the LRM-2B in their minimum-energy conformations. Data presented in section A of Table 4 indicate that the difference in total energies (92.0 kcal/mol) is almost exclusively a consequence of more favorable electrostatic interactions in the BDNF/LRM-2B complex than in the NGFILRM-2B complex (negative control). Section B of Table 4 illustrates that the BDNF complex is also more favorable in terms of intermolecular electrostatic energy by 77.2 kcal/mol (85% of the difference in total electrostatic energies). On the contrary, intermolecular van der Waals interactions are more favorable in the "improper" NGF/LRM-2B complex. Furthermore, the value of intermolecular van der Waals energy is significantly lower for NGF/LRM-2B than for the NGF/LRM-2A complex, whereas electrostatic contribution is more than twice as high. Consequently, in contrast to the NGF/TrkA binding pattern, molecular recognition of BDNF by the TrkB receptor binding site is almost exclusively based on electrostatic complementarity, whereas the roles of van der Waals and hydrophobic interactions are insignificant. At the same time, such a perfect electrostatic fit could not be achieved without a purely geometric compatibility of the BDNF and the LRM-2B binding domains in size and shape (FIG. 11b).
Induced Conformational Changes
The NGF Amino and Carboxyl Termini
Conformations of the NGF amino and carboxyl termini remain almost the same after complexation with the LRM-2A (Table 5; compare FIG. 10a with 10b, 10c, 10d and 10e). A few local conformational changes occur in the Arg.sup.114' and the Lys.sup.115' side chains, and in the backbones of residues Ser.sup.2 and Phe.sup.7. As predicted above, conserved charged residues Arg.sup.114' and Lys.sup.115' do not participate in principal intermolecular interactions within the NGF termini but are expected to form ionic intermolecular contacts with the TrkA binding site. Indeed, these are the intermolecular ionic NGF/LRM-2A interactions in area C, which change considerably the side-chain conformations of Arg.sup.114' and Lys.sup.115' while keeping their backbone geometry intact. The LRM-2A-induced conformational changes in residue Phe.sup.7 are caused by the intermolecular H-bonding in area E while the geometry of Ser.sup.2 is slightly changed to accommodate the LRM-2A residue Thr.sup.114tA. The rest of the structure of the LRM-2A-complexed NGF termini retains the energy minimum conformation of the uncomplexed molecules because it is controlled by multiple intramolecular H-bonds. Consequently, the geometry of the NGF termini is "preorganized" for complexation with the LRM-2A, and only local conformational changes, primarily in residues Arg.sup.114' and Lys.sup.115', occur upon complexation. Geometric "preorganization" of the NGF termini as a ligand for binding to a leucine-rich motif is consistent with X-ray studies of Kobe and Deisenhofer (1995), who have demonstrated that ribonuclease A maintains a fixed geometry upon binding to the LRMs of ribonuclease inhibitor.
The LRM-2A Domain
FIG. 12 illustrates the isolated structure of the LRM-2A from the hNGF/LRM-2A complex. Table 6 presents the equilibrium torsional angles of the hNGF-bound LRM-2A. As seen, the side chains of the general consensus residues, Leu.sup.93tA, Leu.sup.96tA, lle.sup.98tA, Leu.sup.103tA and Phe.sup.111tA (Schneider et al., 1988; Schneider and Schweiger, 1991), form the hydrophobic core, similarly to their conformations in intact leucine-rich repeats (Kobe and Deisenhofer, 1993). Terminal consensus residue Leu.sup.117tA is solvent-exposed due to a boundary effect. Amino acids Ala.sup.107tA, Pro.sup.108tA, Ala.sup.110tA and Phe.sup.113tA are also incorporated into the LRM-2A hydrophobic core. On the other hand, residues Val.sup.99tA, Phe.sup.105tA, Val.sup.106tA and Pro.sup.115tA are not involved in hydrophobic interaction within the core, and apparently have specific roles in the recognition site of TrkA. The most hydrophilic residues of the LRM-2A, namely Arg.sup.94tA , Lys.sup.100tA, Arg.sup.104tA, Asp.sup.109tA, His.sup.112tA and Arg.sup.116tA are exposed from the hydrophobic core in the complexes. Amino acids Lys.sup.100tA, Asp.sup.109tA and His.sup.112tA participate in specific intermolecular interactions, while the other three are solvent-exposed. The reasonable locations of the hydrophobic and hydrophilic residues of the LRM-2A within the NGF/LRM-2A complexes with respect to water environment supports the calculated structures.
NGF/LRM-2A binding spans almost the whole LRM-2A domain, i.e., an interval between residues Leu.sup.93tA and Pro.sup.115tA comprising three distinct regions: .beta.-strand, .alpha.-helix, and loop connecting the carboxyl terminus of the .beta.-strand and the amino terminus of the .alpha.-helix. This is in agreement with recent crystallographic studies on the ribonuclease A--porcine ribonuclease inhibitor complex (Kobe and Deisenhofer, 1995), in which these particular regions of leucine-rich repeats of ribonuclease inhibitor are also involved in the molecular recognition process. While the .beta.-strand structure exists in the complexed LRM-2A, the .alpha.-helix motif is destroyed by intermolecular interactions with NGF. Accordingly, the LRM-2A .beta.-strand determines the proper orientation of the ligand with respect to the recognition site of TrkA, but does not undergo noticeable conformational changes. Consequently, the cooperative ligand-receptor H-bonding in area E cannot directly mediate those changes in receptor conformation which trigger receptor phosphorylation processes. Available experimental data confirm this point. First, the mNGF.DELTA.3-9 deletion mutant, which does not have the .beta.-strand motif and, consequently, cannot specifically bind area E, is as active as wild type mNGF (Drinkwater et al., 1993). Second, the 24-residue LRM fragment Thr.sup.97tA -Leu.sup.120tA lacking the .beta.-strand region (Kobe and Deisenhofer, 1993), demonstrates the same kinetics of binding to NGF as wild type TrkA (Windisch et al., 1995a, 1995b).
The LRM-2B Domain
Table 7 presents the equilibrium torsional angles of the LRM-2B in the complex with BDNF. Side chains of all general consensus residues of the LRM-2B, except for Phe.sup.111tB and terminal amino acid Leu.sup.117tB, form the hydrophobic core and thereby stabilize the characteristic hairpin shape of the LRM (FIG. 11b). In contrast with the LRM-2A, aromatic residues of the LRM-2B, namely Phe.sup.105tB, Tyr.sup.108tB and Phe.sup.111tB, do not seem to have any specific roles in BDNF/LRM-2B molecular recognition. They form a cluster in which their aromatic rings interact with each other and are about mutually perpendicular (FIG. 11b). On the contrary, all charged residues of the LRM-2B, namely Arg.sup.94tB, Asp.sup.100tB, Lys.sup.104tB, Lys.sup.109tB and Lys.sup.113tB, are involved in specific electrostatic interactions with BDNF.
Similar to the NGF/LRM-2A complex, almost the whole LRM-2B domain, from Leu.sup.93tB to Lys.sup.113tB, fits to the BDNF binding determinant (FIG. 11b). On the other hand, there are significant differences of the neurotrophin-induced conformational alterations in the two LRM domains. First, since the BDNF binding domain does not have a .beta.-strand motif which would fit to the .beta.-strand of the LRM-2B, the characteristic geometric features of the latter are lost. Second, region Tyr.sup.108tB to Asn.sup.116tB retains its .alpha.-helical structure in the BDNF-bound LRM-2B domain. The reason of this conformational stability is that the spacer between the two BDNF-bound residues, Lys.sup.1091tB and Lys.sup.113tB (3 residues), is consistent with the period of an .alpha.-helix (3.5 residues).
Principal Amino Acids Involved in Molecular Recognition
Principal NGF Residues
The amino acids of NGF which are primarily responsible for maintaining bioactive conformations (McDonald et al., 1991; Shamovsky et al., 1996) or which directly participate in NGF/TrkA recognition are specified in FIG. 13. Almost all residues of the amino and carboxyl termini of NGF as well as the His.sup.75' -Asn.sup.77' region are involved in specific intra- or intermolecular interactions. Only five amino acids of the 30-residue binding domain of NGF do not seem to be essential for either maintaining its bioactive conformation or binding to the recognition site of TrkA. These include Ser.sup.1, Ser.sup.2, Ser.sup.3, Ala.sup.116' and Val.sup.117'.
The observation that the important amino acids of the NGF binding domain are conserved across all species while the apparently unimportant five residues are variable (FIG. 13) (McDonald et al., 1991) supports the particular structural features of the NGF/LRM-2A complexes calculated herein. There are, however, exceptions which require explanation. In most cases, principal amino acids of NGF molecules derived from different species appear to be substituted by similar residues, e.g. 16V, F7L, L112I, S113T or R114K. In some special cases, the substituted amino acid is quite different but the particular substitution is still consistent with the present model, e.g. R9M, R9L or RI 18Q (McDonald et al., 1991; Shamovsky et al., 1996); in the case of the H8N substitution (McDonald et al., 1991), histidine and asparagine have the same topology with respect to H-bonding to the Glu.sup.11 carboxyl oxygen. The structure of snake NGF (sNGF) requires special consideration. Principal hydrophobic residues Phe.sup.7 and Phe.sup.12 are replaced by histidine residues in sNGF. Although these mutations could be accompanied by structural differences in snake TrkA, they are consistent with the present model. According to the survey of Warme and Morgan (1978), side-chain interactions of histidine with tryptophan or threonine are very common in natural proteins, which implies the existence of specific inter-residue affinities. This being the case, attractive interactions His.sup.7. . . Thr.sup.114tA and His.sup.12 . . . Trp.sup.76' in areas B and A, respectively, may replace corresponding hydrophobic interactions with Phe.sup.7 and Phe.sup.12.
Principal LRM-2A Residues
The principal residues of the LRM-2A are shown in FIG. 14. The general consensus residues (Schneider et al., 1988; Schneider and Schweiger, 1991) maintain the specific geometry of the LRM (Kobe and Deisenhofer, 1993, 1994, 1995). The amino acids which form the distinct binding areas are involved in specific intermolecular interactions with the NGF binding domain. Since NGF does not activate the TrkB receptor, those residues which are specific for the TrkA receptor have been postulated to be directly involved in NGF/LRM-2A recognition (Windisch et al., 1995a, 1995b). As shown, the LRM-2A segment has nine specific residues, seven of them directly participating in NGFILRM-2A recognition in the framework of the present model. Thus, Lys.sup.100tA binds Asp.sup.16 in area D, Pro.sup.108tA is completely buried inside the NGF/LRM-2A complex, residues ASp.sup.109tA, His.sup.112tA and Pro.sup.115tA form binding area C, and residues Phe.sup.131tA and Thr.sup.114tA are involved in forming area B. The other specific residues, Arg.sup.104tA and Arg.sup.116tA, do not seem to be required for specific NGF/LRM-2A interactions. It is seen, furthermore, that binding areas A and E are not specific for the recognition site of TrkA, indicating that binding specificity of the LRM-2A is determined by Lys.sup.100tA, Asp.sup.109tA and Phe.sup.113tA. The fact that these particular amino acids are replaced in the recognition site of TrkB by residues of a different chemical nature, namely Asp.sup.100tA, Lys.sup.109tA and Lys.sup.113', respectively (Windisch et al., 1995a, 1995b), making NGF/TrkB binding impossible, supports the herein predicted details of NGF/LRM-2A recognition. Principal residue Phe.sup.105tA, although common to both the TrkA and TrkB recognition sites (Windisch et al., 1995a, 1995b), is also important for the NGF/TrkA recognition process, since oxidation of its counterpart, Trp.sup.76', results in a complete loss of receptor binding (Merrell et al., 1975).
Principal BDNF Residues
Since BDNF/TrkB molecular recognition is predicted to be primarily based on electrostatic complementarity, principal BDNF and LRM-2B residues are identified by analysing their contribution to the intermolecular electrostatic energy. Table 8 presents a comparison of individual contributions of residues of BDNF, NGF and the LRM-2B to the intermolecular electrostatic energy for the BDNF/LRM-2B and the NGF/LRM-2B complexes, with cutoff being .+-.10 kcal/mol. BDNF/TrkB molecular recognition is achieved by those residues that make significantly more favourable contacts with their counterparts than do corresponding residues in the "improper" NGF/LRM-2B interaction. Accordingly, in section A of Table 8, residues of BDNF that make more favourable contacts with the LRM-2B than do those of NGF by at least 10 kcal/mol, are highlighted. They are: Glu.sup.18, Asp.sup.24, Glu.sup.55, Arg.sup.81 and Gln.sup.84. These residues make up the BDNF binding epitope for TrkB activation. Only two of them, Arg.sup.81 and Gln.sup.84, have been previously identified (Ibanez et al., 1993). The most unusual case is residue Gln.sup.84, which does not markedly participate in electrostatic interactions with the LRM-2B; however, its substitution by His.sup.84 results in considerable repulsive interactions with positively charged residues of the receptor site. Therefore, Gln.sup.84 cannot be referred to as a principal residue by itself in terms of the TrkB receptor binding, and its apparent importance (Ibanez et al., 1993) comes from unfavorable substitution by a positively charged His.sup.84. The only residue of NGF which binds to the TrkB receptor site significantly more strongly than does the corresponding residue of BDNF is Glu.sup.35, in agreement with observations (Lai et al., 1996).
Principal LRM-2B Residues
Section B of Table 8 identifies the LRM-2B residues which significantly contribute to specificity of binding. Amino acids Arg.sup.94tB, Asp.sup.100tB, Ser.sup.101tB, Lys.sup.109tB and Lys.sup.113tB make significantly more favorable contacts with their counterparts in BDNF than with those in NGF. Additionally, these residues are responsible for specific binding in areas F-K. The only residue of the LRM-2B that favours the NGF counterpart is Lys.sup.104tB, which is caused by the abovementioned effect of GIu.sup.35.
FIG. 15 illustrates schematically the geometry of the LRM-2A and the LRM-2B receptor sites. Those residues which are predicted to form specific bonds with their preferred ligand are shown by arrows. The similarity of the binding patterns of the two LRMs is striking, especially bearing in mind the differences pointed out in the nature of the intermolecular interactions, and that the binding patterns were identified in independent computations. Thus, binding specificity of TrkA and TrkB is determined by the residues located in distinct positions of their second LRMs, with their respective binding patterns being incompatible with each other.
It is shown that van der Waals and hydrophobic interactions play a more important role in TrkA than in TrkB recognition. As a part of this phenomenon, we have demonstrated that NGF/LRM-2A binding is characterized by greater geometric complementarity, such that the ligand-receptor complex does not contain marked internal cavities. A plausible explanation for these particular features of Trk receptor recognition is that the NGF binding determinant is flexible and, because of that, more is required to stabilize its bioactive conformation in solution. Accordingly, TrkA receptor binding involves physical factors that are more effective in an aqueous environment, namely van der Waals and hydrophobic interactions.
Explanation for Structure-Activity Relationships
TrkA mediated activity has been evaluated for 27 point and deletion mutants of the TrkA binding domain of NGF: hNGF (Burton et al., 1992; Shih et al., 1994), mNGF (Burton et al., 1992; Drinkwater et al., 1993), mNGF.DELTA.3-9 (Drinkwater et al., 1993), (1-120)hNGF (Schmelzer et al., 1992), (1-120)mNGF (Luo and Neet, 1992), (1-117)hNGF (Schmelzeret al., 1992) and hNGF:NT-4/5 (Treanor et al., 1995) retain full activity; mNGF.DELTA.2-8, mNGF.DELTA.61-66 (Drinkwater et al., 1993), hNGF-R9A, hNGF.DELTA.1-5 and hNGF-P5A (Shih et al., 1994) demonstrate intermediate potency; hNGF.DELTA.1-9 (Burton et al., 1992; Kahle et al., 1992; Luo and Neet, 1992; Drinkwater et al., 1993; Shih et al., 1994; Burton et al., 1995), mNGF.DELTA.1-8 (Taylor et al., 1991; Burton et al., 1992), mNGF.DELTA.1-9, mNGF.DELTA.9-13 (Woo et al., 1995), mNGF.DELTA.3-8, mNGF.DELTA.3-11, mNGF.DELTA.3-12, mNGF.DELTA.3-13, mNGF.DELTA.3-14, mNGF.DELTA.9-14, mNGF.DELTA.112-118 (Drinkwater et al., 1993), hNGF-H4A, hNGF-H4D (Shih et al., 1994), BDNF (Suter et al., 1992), and NT-4/5 (Treanor et al., 1995) are almost or completely inactive. The present model of NGF/LRM-2A recognition enables a comprehensive explanation for the structure-activity relationships.
As demonstrated, fully active hNGF, mNGF and hNGF:NT-4/5 fit to the five binding areas A-E of the recognition site of TrkA, while the mNGF.DELTA.3-9 deletion mutant fits only to A, C and D, without any specific binding taking place in areas B or E. Therefore, there are three common geometric features of the four fully active molecules which distinguish them from the less active analogues. First, conformations of the amino and carboxyl termini of these NGF analogues are "preorganized" for binding to the principal LRM-2A binding areas A, C and D. Second, there is no stereochemical incompatibility in the binding areas B and E. Third, There is no Hole Inside the NGF/LRM-2A Complexes
The structures of the other three fully active NGF analogues, (1-120)hNGF, (1-120)mNGF and (1-117)hNGF, are consistent with the above criteria. Indeed, since the NGF carboxyl end is located on the surface of the NGF/LRM-2A complexes (FIG. 10) and does not directly participate in principal intermolecular interactions, the lengthening of the NGF carboxyl terminus by two amino acids does not explicitly affect NGFILRM-2A compatibility. As far as the (1-1 17)hNGF molecule is concerned, terminal residue Arg.sup.118' of NGF has been shown to form inter-chain ionic H-bonds with Glu.sup.11 (see above) and, hence, seems to be important in maintaining the structure of the NGF bioactive conformation. On the other hand, there is a second factor stabilizing the bioactive conformation of the hNGF termini, namely inter-chain H-bonding of residue Arg.sup.9. Because of the latter binding, the above criteria are still fulfilled for (1-117)hNGF.
The deletion of principal residue Phe.sup.12, which participates in hydrophobic interactions in area A, results in a substantial loss of TrkA mediated activity of the mNGF.DELTA.3-12, mNGF.DELTA.3-13, mNGF.DELTA.3-14 and mNGF.DELTA.9-14 deletion mutants (Drinkwater et al., 1993). Likewise, the absence of principal residues Arg.sup.114' and Lys.sup.115', essential for the ionic binding in area C, results in dramatic loss of bioactivity of mNGF.DELTA.112-118 (Drinkwater et al., 1993). Similarly, geometric perturbations of the essential residues Arg.sup.59 and Arg.sup.69 in the mNGF.DELTA.61-66 deletion mutant result in incomplete NGF/LRM-2A compatibility in area D and, accordingly, a considerable loss of activity.
The specific location of the NGF ninth amino acid side chain in the "hollow" existing on the surface of the NGF rigid region has been predicted herein to be critical for high TrkA mediated activity. The present invention provides a detailed explanation for this prediction. That is, the "hollow" is buried inside the NGF/LRM-2A complex, and a moiety located in this "hollow" fills the cavity which otherwise would exist inside the complex and destroy NGF/LRM-2A compatibility. Since this moiety is located in the internal cavity of the complex (FIG. 10), severe limitations are imposed on its size. A short side chain does not fill this cavity; therefore, the hNGF-R9A mutant demonstrates noticeably diminished activity (Shih et al., 1994). The cavity preferably is filled either by a bulky amino acid side chain, such as arginine (e.g. hNGF), methionine (e.g. mNGF) or leucine (sNGF), or by the two residue peptide backbone (mNGF.DELTA.3-9 or mNGF.DELTA.2-8). A longer amino terminus tail (e.g. in mNGF.DELTA.3-8, mNGF.DELTA.9-13 or mNGF.DELTA.1-5) cannot be accommodated in the internal cavity of the NGF/LRM-2A complex and, hence, interferes with binding areas B and E.
Most deletion or point mutations of the NGF amino terminus destroy the exposed .beta.-strand motif Phe.sup.7 -Arg.sup.9 and change the proper orientation of the Phe.sup.7 side chain, thereby affecting NGF/LRM-2A compatibility in areas B and E. Thus, incorporation of a negatively charged residue into the relatively long NGF amino terminus, as in hNGF-H4D, BDNF, NT-3 and NT-4/5, or substitution H8R, which is characteristic for BDNF and NT-4/5, has each been demonstrated to cause dramatic conformational changes in the complex of the neurotrophin amino-carboxyl termini, making it incompatible with the LRM-2A. This stereochemical incompatibility of the geometry of the NGF amino terminus with binding areas B and E of the LRM-2A is the reason for the partial or complete loss of TrkA-related activity of mNGF.DELTA.3-8, mNGF.DELTA.l-5, hNGF-P5A, hNGF-H4A, hNGF-H4D, mNGF.DELTA.1-8, hNGF.DELTA.1-9, mNGF.DELTA.1-9, mNGF.DELTA.9-13, mNGF.DELTA.3-11, BDNF and NT-4/5.
As demonstrated, both NGF monomers are-directly involved in specific ligand-receptor interactions, i.e., the first monomer binds areas B, D and E, whereas the second one binds areas A and C. Consequently, the dimeric structure of NGF is necessary to retain high biological activity, which is also consistent with experimental data (Maness et al., 1994). On the contrary, only one BDNF protomer is essential for TrkB binding (FIGS. 11a, 11b).
In addition, recent reports of experiments have demonstrated the importance of residues Phe.sup.7 (Kullander and Ebendal, 1996), Leu.sup.112, Arg.sup.114 and Lys.sup.115 (Kruttgen et al., 1996) for TrkA mediated activity of NGF. This is consistent with their participation in the specific NGF/LRM-2A binding in the framework of the present model.
Conclusions
The complexes of the amino and carboxyl termini of the four fully active NGF analogues with the second leucine-rich motif of TrkA were obtained using Variable Basis Monte Carlo simulated annealing calculations. The same methodology was utilized to define the structure of the complex of the second leucine-rich motif of TrkB with the BDNF receptor binding determinant. The structural features of the resulting NGF/TrkA and BDNF/TrkB binding predictions are supported by the following: (i) geometric and electrostatic complementarity of the binding domains of NGF and BDNF to binding sites of TrkA and TrkB, respectively, without considerable conformational changes of the ligands; (ii) appropriate environments for hydrophobic and hydrophilic residues within the complexes; (iii) analysis of structural variability of both binding domains; and (iv) comprehensive explanation of currently published structure-activity relationships. Within these complexes, NGF/TrkA molecular recognition is based upon multiple-point intermolecular interactions of hydrophobic, van der Waals, ionic and hydrogen bonding type. Extensive intermolecular van der Waals contacts arise from complementarity of the NGF binding domain and the TrkA recognition site in size and shape. In addition, there are five specific binding areas within the receptor model: hydrophobic A and B, ionic C and D and cooperative hydrogen bonding E. The NGF/TrkA stereochemical compatibility in areas A, C and D is required for TrkA receptor activation, whereas areas B and E reinforce binding specificity, making the NGF molecule a very sophisticated key triggering the chain of biological events following receptor phosphorylation. The 3-dimensional structures of the binding domains of the fully active NGF analogues are "preorganized" for stereochemical fit to the second leucine-rich motif of TrkA. Five distinct binding areas F-K were also identified in the BDNF/TrkB complex, although, in contrast with NGF/TrkA binding, BDNF/TrkB recognition is based primarily on ionic and hydrogen bonding interactions. In spite of significant differences in the nature of ligand-receptor interactions involved in TrkA and TrkB recognition processes, binding patterns of their receptor binding sites for NGF and BDNF are almost identical. The predicted mechanism of high affinity NGF/TrkA and BDNF/TrkB binding is consistent with the conformational model of TrkA-p75.sup.NTR interactions (Chao and Hempstead, 1995).
PART C
Theoretical Modelling of the P75.sup.NTR Receptor Recognition Site for Neurotrophins
Introduction
As summarized above, the neurotrophins represent a family of structurally and functionally related proteins which play a crucial role in the development, survival and maintenance of the sympathetic and sensory neurons (Purves, 1988; Snider and Johnson, 1989; Barde, 1989; Thoenen, 1991; Korsching, 1993; Persson and Ibanez, 1993; Davies, 1994; Ibanez, 1995). The neurotrophin family consists of nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), neurotrophin-4/5 (NT-4/5), and more recently discovered neurotrophin-6 (NT-6) (Berkmeier et al., 1991; Barde, 1991; Barbacid, 1993; Bradshaw et al., 1994; Chao, 1994; Gotz et al., 1994). Each neurotrophin is capable of binding independently with the common neurotrophin receptor p75.sup.NTR, as opposed to specific Trk receptor interactions.
Methods
The geometric features of complexes of monomeric neurotrophins with the p75.sup.NTR receptor binding site have been obtained by multiple molecular dynamics runs of 200-300 ps at constant temperatures between 300 and 400.degree. K., followed by local energy refinements. The QUANTA V4.1 molecular modelling program (Molecular Simulations Inc., Waltham, Mass.) with the united atom CHARMM potential energy terms (Brooks et al., 1983) has been used. Residues of p75.sup.NTR are designated by abbreviation p.
According to the invention, we superimposed the binding domains of NGF, BDNF, NT-3 and NT-4/5 with a putative binding site of p75.sup.NTR and carried out multiple molecular dynamics computations of the ligand-receptor complexes. These calculations revealed the 3-dimensional structures of the binding domains of the neurotrophins and p75.sup.NTR, demonstrated their geometric and electrostatic complementarity, and identified principal residues involved in receptor recognition. All neurotrophins are believed to induce similar conformational changes in the p75.sup.NTR binding site, generally located in the orientation of the sequential loops within the cysteine-rich repeats.
Initial geometries of individual components of the complexes were based upon the X-ray crystallographic coordinates of neurotrophins NGF (McDonald et al., 1991), BDNF and NT-3 (Robinson et al., 1995) and the p55.sup.TNFR receptor extracellular domain (Banner et al., 1993) which is homologous to p75.sup.NTR, Geometry of the NT-4/5 monomer was constructed from the X-ray derived structure of NT-3 (Robinson et al., 1995). Since NGF and other neurotrophins fold in similar rigid conformations, their molecular motions were restricted by imposing the following atom constraints. First, peptide backbones of the neurotrophins were fixed at their crystallographic coordinates. Further, conformations of the amino acid side chains of the neurotrophins were also fixed except for those located within and in the vicinity of the variable loop regions I and V (the p75.sup.NTR binding determinant). Thus, amino acid side-chain motions were allowed for residues 23-35 and 93-98 (NGF numbering) as well as for eight additional neighboring residues, namely His.sup.84 (Gln in other neurotrophins), Phe.sup.86 (Tyr in other neurotrophins), Lys.sup.88 (Arg in other neurotrophins), Trp.sup.99, Arg.sup.100, Phe.sup.101 (Trp in NT-3 and NT-4/5), Arg.sup.103, and Asp.sup.105. The flexible amino and carboxyl termini of the neurotrophins were excluded to avoid artifactual spacial hindrances for the receptor binding.
FIG. 16a represents a fragment of the crystallographically resolved structure of the p55.sup.TNFR extracellular domain (Banner et al., 1993), corresponding to the second CRD (loops 2A and 2B) and the first loop of the third CRD (loop 3A). FIG. 17 illustrates the sequence of this fragment and its alignment with the three known p75.sup.NTR peptides (Johnson et al., 1986; Radeke et al., 1987; Large et al., 1989). FIG. 16b presents the minimum energy conformation of the corresponding fragment of human p75.sup.NTR. As is apparent, its size and shape are complementary to the exposed p75.sup.NTR binding determinant of NGF (FIG. 16c). Accordingly, this 58-amino acid fragment of p75.sup.NTR, including the second and the beginning of the third CRDs (FIG. 17), has been considered as a putative binding site for neurotrophins. The artefactual charges on the terminal amino and carboxyl groups of the peptide representing the p75.sup.NTR binding site were neglected.
Once the most stable equilibrium 3-dimensional structures of the complexes being studied were obtained, the residues contributing the most to the energy of intermolecular interactions were identified. Point atomic charges of Weiner et al. (1984) and a dielectric constant equal to the interatomic separation in angstroms (to implicitly simulate solvation effects) (Weiner et al., 1984) were utilized to calculate the electrostatic intermolecular terms. The van der Waals intermolecular terms were evaluated by the united atom CHARMM force field (Brooks et al., 1983). The calculations were performed on a 200 MHz Indigo-2 SiliconGraphics Workstation.
Results and Discussion
The most stable structures of the complexes of the neurotrophins with the putative p75.sup.NTR receptor binding site obtained by the molecular dynamics computations described herein are illustrated in FIG. 18. Table 9 presents net values of the electrostatic and van der Waals terms of the ligand-receptor interaction energy. Tables 9 and 10 list individual residues of the neurotrophins and the p75.sup.NTR receptor which significantly contribute to the intermolecular electrostatic interaction energy. The highlighted residues represent the functional epitopes of the ligands (Table 9) and the receptor (Table 10), whereas the other listed residues, although being located within the binding interfaces, apparently do not participate in major ligand-receptor electrostatic interactions in the particular complex. Predicted Cartesian 3-dimensional coordinates of atoms of the p75.sup.NTR binding sites in the complexes with NGF and BDNF are presented in Appendix 3.
Significant contributions of electrostatic interactions to the ligand-receptor interaction energy (Table 9) result from opposite net charges of the binding domains of neurotrophins (positively charged) and the common neurotrophin receptor (negatively charged), which effect is consistent with published predictions (Ryden et al., 1995). Data presented in Tables 9 and 10 demonstrate that, on the average, the interaction between the functional epitopes is responsible for 94% of intermolecular electrostatic energy, but only 43% of van der Waals interaction energy. Consequently, residues participating in strong and specific ligand-receptor interactions represent only a part of a bigger binding interface in which significant van der Waals contribution is accumulated from a large number of weak contacts, which is possible only as a consequence of geometric complementarity of the binding domains. FIG. 19 presents the structure of the BDNF/p75.sup.NTR complex as atomic van der Waals spheres and demonstrates the geometric fit of the binding domains. The van der Waals energy and, therefore, geometric complementarity of the binding interfaces in the NGF/p75.sup.NTR complex is significantly lower than in the other neurotrophin/p75.sup.NTR complexes (Table 9), which is consistent with its fastest rate of dissociation (Ibanez, 1995). However, the pure geometric factor cannot explain the slowest rate of dissociation observed for the BDNF/p75.sup.NTR complex (Ibanez, 1995). Thus, stability of the complex arises from a combination of geometric and electrostatic complementarities of the binding interfaces, consistent with the generalizations of Clackson and Wells (1995). A specific feature of the neurotrophin/p75.sup.NTR binding interfaces is that they do not include substantially hydrophobic residues, so they do not have a hydrophobic core, which has been found in many other protein-protein complexes (Clackson and Wells, 1995; Livnah et al., 1996). The predominant interactions within the binding domains of the complexes under study are of the salt-bridge type.
Specificity of binding of the neurotrophins to the p75.sup.NTR receptor is determined by charged residues within the binding domains (Ryden et al., 1995). As seen in FIG. 18, they create 3-dimensional networks in which most of them are in contact with more than one residue of the opposite charge. Therefore, any missing charge would destroy more than just one bond, but a significant portion of, if not the whole, network. That is, the electrostatic complementarity of the binding interfaces would be altered, which is consistent with the experimentally observed sensitivity of binding to mutations of charged residues (Ibanez et al., 1992; Ryden et al., 1995; Ryden and Ibanez, 1996). Both ligand and receptor provide the network with positive and negative charges (Table 9 and 10, FIG. 18), even though negative charges dominate in the receptor, and positive in the ligands. This particular spatial arrangement of the charges increases specificity of binding and prevents electrostatic repulsions within the same binding domains. For example, in all the complexes, positively charged residue Lys.sup.56p of the receptor forms salt-bridges with negatively charged residues Asp.sup.75p and Asp.sup.76p of the receptor, thereby occupying one site of their coordination, such that only the opposite sides of the latter residues remain available for intermolecular ionic interactions (FIG. 18). Negatively charged residue Asp.sup.93 of BDNF holds together positively charged residues Arg.sup.97 and Arg.sup.101 of BDNF, such that they are both able to interact with juxtaposed amino acids Asp.sup.75p and Asp.sup.76p of the receptor (FIG. 18b). Similarly, negatively charged residue Asp.sup.30 of the neurotrophins forms a salt-bridge interaction with Arg.sup.100, which directly interacts with p75.sup.NTR.
In spite of the differences in the binding interfaces of the respective complexes, binding patterns of the p75.sup.NTR receptor are similar. All three loops of the receptor binding site (2A, 2B and 3A) are involved in binding (FIG. 18). The principal residues of loop 2A are Asp.sup.47p and Lys.sup.56p ; those of loop 2B are Asp.sup.75p and Asp.sup.76p ; and those of loop 3A are Asp.sup.88p and Glu.sup.89p. As is seen in Table 10, loop 2B forms the strongest contacts. In addition to the principal binding pattern, the p75.sup.NTR binding domain comprises specific residues for ionic interactions with particular neurotrophins (Table 10, FIG. 18). Thus, Arg.sup.80p interacts only with Glu.sup.35 of NGF; residue Glu.sup.53p interacts only with Lys.sup.96 of BDNF, and Glu.sup.60p and Glu.sup.73p interact only with residues of NT-3 and NT-4/5.
Site-directed mutagenesis studies of the p75.sup.NTR binding site have suggested that mutation of any one of the residues Asp.sup.75p, Asp.sup.76p or Glu.sup.89p to Ala does not affect NGF binding, whereas mutation of Ser.sup.50p, which does not participate in major ionic interactions, results in the complete abolishment of binding (Baldwin and Shooter, 1995). The apparent discrepancy with the present theoretical studies can be explained, since the cited experimental observations are still controversial. Indeed, although both Asp.sup.75p and Asp.sup.76p are involved in ionic interactions in the loop 2B region with positively charged residues of neurotrophins, one of them might be enough to maintain binding to NGF. Likewise, the functional role of the missing residue Glu.sup.89p in loop 3A might be fulfilled by adjacent Asp.sup.88p. Residue Ser.sup.50p of p75.sup.NTR is predicted to be hydrogen bonded to Gln.sup.96 of NGF (FIG. 18a), and its mutation to Asn, Ala or Thr, which abolishes receptor binding (Baldwin and Shooter, 1995), may create steric hindrances in the NGF/p75.sup.NTR interface. The present molecular modelling studies suggest that Ser.sup.50p is not important for binding of any other neurotrophin (FIG. 18, b-d).
In contrast to the p75.sup.NTR receptor binding pattern, there are only two common residues of the binding interfaces of neurotrophins which directly interact with the receptor site, Arg.sup.100 and Arg.sup.103. Contribution of Arg.sup.103 is significant in all the complexes, while Arg.sup.100 seems to be less important for the NGF/p75.sup.NTR binding.
As is seen in Table 9, most important residues of the NGF binding domain are (in order of decreasing significance) Lys.sup.32, Lys.sup.34, Arg.sup.103, Lys.sup.95, and His.sup.84, all of them forming specific ionic contacts with negatively charged residues of p75.sup.NTR (FIG. 18a). Involvement of Lys.sup.32, Lys.sup.34, and Lys.sup.95 in p75.sup.NTR binding has been very well established experimentally, with the same order of significance being observed (Ibanez, 1994). The most important residues of NGF, Lys.sup.32 and Lys.sup.34, bind to the loop 2B residues of p75.sup.NTR (FIG. 18a). Here we demonstrate that residue Lys.sup.88 is not critical for p75.sup.NTR binding, whereas residues Arg.sup.103 and His.sup.84 make direct ionic contacts with the receptor binding site. Importance of the latter residue has not been previously reported in the context of p75.sup.NTR binding. The roles of Asp.sup.30, Glu.sup.35 and Arg.sup.100 seem to be secondary, since their contributions are relatively small. Hydrophobic residue lle.sup.31 is not directly involved in ligand-receptor interactions.
In BDNF, all three positively charged residues of variable region V, Lys.sup.95, Lys.sup.96 and Arg.sup.97 (Ryden et al., 1995), bind to negatively charged residues of the p75.sup.NTR binding site, namely Asp.sup.47p, Glu.sup.53p and Asp.sup.75p, respectively (FIG. 18b). In addition, the residues Arg.sup.88 and Arg.sup.101 play a more important role than in the NGF case (Table 9); they bind to the loop 2B residues Asp.sup.75p and Asp.sup.76p of p75.sup.NTR, thereby fulfilling the function of missing positively charged residues of loop 1. The increased importance of residue Arg.sup.88 in BDNF and other neurotrophins relative to that of Lys.sup.88 in NGF (Table 9) is permitted by a longer side chain of Arg, which results in a decrease of the separation between opposite charges (FIG. 18). Because of missing residue His.sup.84, only the positively charged residue Arg.sup.104 binds to the loop 3A residues Asp.sup.88p and Glu.sup.89p of p75.sup.NTR.
In NT-3, the loop 2B residues of the receptor Glu.sup.73p, Asp.sup.75p, and Asp.sup.76p bind to Arg.sup.31, His.sup.33, Arg.sup.100, and Arg.sup.87. Residue Asp.sup.47p of loop 2A of the receptor also binds to Arg.sup.100. Residue Lys.sup.95 of variable region V does not form any salt-bridge contact with the receptor, consistent with its reported insignificance for NT-3 binding (Ryden et al., 1995). Thus, the present results agree well with the established roles of Arg.sup.31 and His.sup.33 of NT-3 (Ryden et al., 1995; Ryden and Ibanez, 1996) and predict the increased importance of residue Arg.sup.100 in NT-3 for p75.sup.NTR binding (Table 9). The binding pattern of NT-4/5 is similar to that of NT-3 (FIG. 18, c,d). The increased electrostatic contribution of Arg.sup.36 with respect to corresponding NT-3 residue His.sup.33 (Table 9) is caused by the longer side chain of the former, which results in formation of a salt-bridge contact with residue Glu.sup.60p of the receptor (FIG. 18d).
Binding of a neurotrophin induces conformational changes in the p75.sup.NTR binding site. The conformational space of the three loops (2A, 2B and 3A) is limited by eight disulfide bonds; therefore, the most significant changes occur in the junction areas of those loops. FIG. 20 illustrates the superimposed equilibrium geometries of the free and the BDNF-complexed p75.sup.NTR binding site. The superimposition is performed by a least-square minimization technique using the a-carbon reference atoms of loops 2A and 3A. As is apparent, the orientation of loop 2B with respect to the two flanking loops is considerably changed upon binding. Because of striking similarities between the binding patterns of the p75.sup.NTR receptor with the different neurotrophins, a similar conformational change takes place in all four complexes under investigation. The driving force behind this particular conformational transition is ionic interactions of the principal residues of loop 2B, Asp.sup.75p and Asp.sup.76p, with positively charged residues of the neurotrophins, taking into account that spacial positions of the flanking loops 2A and 3A are fixed by ionic interactions of residues Asp.sup.47p, Lys.sup.56p, Asp.sup.88p, and Glu.sup.89p with corresponding residues of the neurotrophins (FIG. 18). Thus, structural variability between the neurotrophins does not preclude them from inducing similar conformational change in the common neurotrophin receptor.
The binding pattern of neurotrophins to the p75.sup.NTR receptor predicted by the present molecular modelling studies can be additionally verified by checking conservation of the identified principal residues. The sequences of the binding site of the p75.sup.NTR receptor from different species (FIG. 17) demonstrate that principal residues forming its common binding pattern, namely Asp.sup.47p, Lys.sup.56p, Asp.sup.75p, Asp.sup.76p, Asp.sup.88p, and Glu.sup.89p, are conserved in p75.sup.NTR. The known variability of regions I and V among neurotrophins (Ryden et al., 1995) implies that conservation of the principal residues of the p75.sup.NTR binding determinants may be checked only within the same neurotrophin derived from different species. Study of a recent sequence alignment of neurotrophins (Bradshaw et al., 1994) allowed us to conclude that the functional epitope of the neurotrophins (Table 9) is generally conserved within the same neurotrophin even though there are some exceptions, often represented by the substitution by a similar amino acid. Arg.sup.100 and Arg.sup.103 are the only two residues of the binding epitopes which are conserved throughout all neurotrophins (Bradshaw et al., 1994). Other residues within the functional epitopes maintain specificity of binding of different neurotrophins.
The present model also represents a key to understanding the conservation of residues which do not directly participate in receptor binding. For example, Gly.sup.33 is absolutely conserved in all neurotrophins, yet is not involved in any specific interactions with the receptor binding site. However, within the present model, its conservation can be understood, since it is located very close to loop 3A of the receptor and forms van der Waals contacts with it (FIG. 18), such that any other residue would impose steric hindrance which would alter or destroy the ligand-receptor geometric complementarity.
The principal NGF/TrkA binding areas necessary to trigger TrkA phosphorylation and subsequent biological events have been discovered by the inventors. Specifically, the inventors have discovered that essentially the full binding site of the second leucine rich motif (LRM) of TrkA, amino acid residues 93 to 117, forms the binding site for NGF, and comprises five distinct binding areas, A, B, C, D and E. Appendix 1 provides the Cartesian coordinates of the LRM-2A residues defining the areas A to E in PDB format. Binding areas A and B bind hydrophobically to ligands, areas C and D form ionic bonds with charged portions on ligands and area E forms hydrogen bonds with ligands.
The stereochemical fit to at least three of these binding sites, A, C and D by a ligand is necessary to activate TrkA receptor. Binding areas B and E reinforce stereoselectivity of the TrkA binding site by selecting an NGF analogue with a stereochemically compatible amino terminus.
Binding area A of the second LRM comprises hydrophobic interaction of residue Phe.sup.105tA. Binding area B comprises hydrophobic interaction of residues Phe.sup.111tA, Phe.sup.113tA and Thr.sup.114tA. Binding area C comprises attractive ionic interaction between a suitable portion on the ligand and Asp.sup.109tA. Binding area D comprises attractive ionic interaction between a suitable portion on the ligand and residue Lys.sup.100tA. Binding area E comprises multiple parallel .beta.-strand type hydrogen bonding with region Asn.sup.95tA -lle.sup.98tA.
Main structural features of the bioactive conformation of NGF include: 1) ionic hydrogen bonding between Glu.sup.11 in the amino terminus and Arg.sup.118' in the carboxyl terminus; 2) separation of the flexible loop 1-8 from the rigid region (9-11 and 112' to 118') because of electrostatic repulsions caused by His.sup.4 and conformational restrictions imposed by Pro.sup.5 ; 3) packing the side chain of Arg.sup.9 or Met.sup.9 in the "hollow" of the rigid region, this "hollow" being an indentation in the protein structure by residues Lys.sup.115' -Arg.sup.118' ; and 4) an embryonic .beta.-sheet structure present between the a-carbons of the 7th and 9th residues.
Within NGF, residues Trp.sup.76', Arg.sup.114', Lys.sup.115' and Asp.sup.16 comprise the pharmacophore which binds TrkA. Any synthetic molecule, whether peptide or non-peptide, which fits into binding areas A, C and D is contemplated to elicit a biological response (agonist) or block biological response (antagonist).
The present invention provides a ligand for binding with TrkA, wherein TrkA comprises a second leucine rich motif (LRM) region comprising amino acid residues 93 to 117, the coordinates of the residues of said LRM given in Appendix 1, said LRM motif having five binding areas, area A comprising hydrophobic interaction of amino acid residue Phe.sup.105tA, area B comprising hydrophobic interaction of amino acid residues Phe.sup.111tA, Phe.sup.113tA, and Thr.sup.114tA, area C comprising ionic interaction of amino acid residues Asp.sup.109tA and His.sup.112tA, area D comprising ionic interaction of amino acid residue Lys.sup.100tA, and area E comprising multiple parallel .beta.-strand type hydrogen bonding of region Asn.sup.95tA to lle.sup.98tA. The ligand comprises at least one functional group being capable of hydrophobic bonding and being present in an effective position in the ligand to hydrophobically bind to area A, at least one positively charged functional group being present in an effective position in the ligand to ionically bind to area C, and at least one negatively charged functional group being present in an effective position in the ligand to tonically bind to area D.
Secondary features of the ligand include at least one effective functional group present in a position in the ligand to hydrophobically bind to area B of TrkA, and a functional group present in an effective position in the ligand to hydrogen bind to area E of TrkA.
The invention further provides the ligand comprising amino acid residues, at least one of the residues being capable of hydrophobic bonding and being present in an effective position in the ligand to hydrophobically bind to Phe.sup.105tA in area A on the TrkA receptor, at least one of the residues being a positively charged residue present in an effective position in the ligand to tonically interact with Asp.sup.109tA and His.sup.112tA forming area C on the TrkA receptor, and one of the residues being a negatively charged residue present in an effective position in the ligand to tonically bind to Lys.sup.100tA in area D on the TrkA receptor.
The present invention also provides a method of designing a ligand to bind with an LRM binding site on TrkA. The method includes computationally evolving a chemical ligand using an effective genetic algorithm with preselected spatial constraints so that the evolved ligand comprises at least three effective functional groups of suitable identity and spatially located relative to each other in the ligand so that a first of the functional groups hydrophobically interacts with binding area A, a second of the functional groups ionically interacts with binding area C, and a third one of the functional groups ionically interacts with binding area D of the second LRM of TrkA. Those skilled in the art will be aware of the various techniques which may be used for de novo ligand design. For example, packages such as Quanta, Discover and Insight provide computational methodologies for generating ligands.
Some ligands according to the present invention may be peptides or peptidomimetics. Morgan and co-workers (Morgan et al., 1989) define peptide mimetics as "structures which serve as appropriate substitutes for peptides in interactions with receptors and enzymes". The mimetic must possess not only affinity but also efficacy and substrate function. For purposes of this disclosure, the terms "peptide mimetic" and "peptidomimetic" are used interchangeably according to the above excerpted definition. That is, a peptidomimetic exhibits function(s) of a particular peptide, without restriction of structure. Peptidomimetics of the invention, e.g., analogues of the neurotrophin and receptor binding sites described herein, may include amino acid residues or other chemical moieties which provide functional characteristics described herein. "Isosteric" is an art-recognized term and, for purposes of this disclosure, a first moiety is "isosteric" to a second when they have similar molecular topography and functionality. This term should not, therefore, be restricted to mere similarity in chemical bond configuration. Two compounds may have different chemical bond configurations but nonetheless be isosteric due their 3-dimensional similarity of form and chemical nature, e.g. relevant centers such as heteroatoms are located in the same position, charges and/or dipoles are comparable, and the like.
Known compounds may also be screened to determine if they satisfy the spatial and functional requirements for binding with the second LRM of TrkA. The molecules in the data bases are screened to determine if they contain effective moieties spaced relative to each other so that a first moiety hydrophobically interacts with binding area A, a second of moiety ionically interacts with binding area C, and a third moiety ionically interacts with binding area D of the second LRM of TrkA.
The contents of all references and appendices (see Appendices 1, 2, 3, and 4) cited throughout this application are hereby expressly incorporated by reference.
Equivalents
Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the invention described herein. Such equivalents are intended to be encompassed by the following claims.
TABLE 1__________________________________________________________________________Equilibrium variable torsional angles (in degrees) of the amino andcarboxyltermini of the most stable conformations of the NGF dimers andanalogues.*Active structures Inactive structuresNo. hNGF mNGF mNGFD3-9 hNGF-H4D mNGFDI-8 BDNF__________________________________________________________________________Amino terminus 11 .psi. E 82.9 E 85.9 E 89.2 E 80.9 E 63.7 E 68.2 .phi. 55.7 56.3 57.8 57.2 47.9 56.3 -69.8p.1 -72.0 -72.1 -75.0 -66.2 -76.9 X.sup.2 72.7 71.7 66.0 72.6 71.2 70.1 X.sup.3 -120.4 -111.7 -107.0 -104.8 -102.8 -97.7 10 .psi. G -89.1 G -95.1 G -88.0 G -129.5 G -81.6 G -134.1 .phi. -172.5 -178.6 -177.5 104.7 -80.8 155.0 9 .psi. -36.1 M -21.1 S 146.5 R -46.5 -66.6 R 71.4 .phi. -91.4 -92.8 -103.6 -80.9 157.5 -128.5 -77.2sup.1 -75.2 -66.7 -115.0 -142.4 -73.5 -173.9sup.2 -179.7 64.0 69.5 -69.7 -179.8 X.sup.3 -175.5 179.9 -- -141.9 178.3 -71.4 X.sup.4 134.3 -- - 98.1 -- -109.8 8 .psi. 118.6 H 110.4 S 106.1 H 151.6 105.2 .phi. -116.5 -110.4 52.8 -53.7 -- -100.1 -89.7sup.1 -90.9 -45.9 -62.0 -- 179.6 -73.8sup.2 -71.5 -172.3 -72.5 -- -105.5 X.sup.3 -- -- -- -- -- 162.1 X.sup.4 -- -- -- -- 133.7 7 .psi. F 135.7 F 129.4 -- 142.3 161.6 .phi. -89.0 -- -114.8 -- -166.6 -171.61 -163.9 -- -170.7 -- -- -96.3up.2 -91.5 -- -100.2 -- -- 6 .psi. 40.3 V 61.7 -- 40.0 114.4 .phi. 47.7 40.4 39.9 -- -- -159.7p.1 -177.3 -- -159.7 -- -- 87.1.sup.2 -- 88.6 -- -- 5 .psi. P 156.0 P 149.8 105.5 D- 82.8 .phi. -- -- -- -- 56.5 -- X.sup.1 -- -- -- -- -176.9 -- X.sup.2 -- -- -- -- -112.8 4 .psi. H 121.4 H 122.1 -- 117.9 -- S 136.2 .phi. -129.8 -- -102.8 -- -53.4 -162.5up.1 -163.3 -- -66.4 -- -172.9 -69.1up.2 -69.2 -- -97.5 -- -178.3 3 .psi. S 2.1 -1.9 -- -25.0 -- H 153.8 .phi. -91.9 -84.5 -- -68.2 -- 48.2 -171.8up.1 -174.9 -- 39.3 -- 82.1 -53.2sup.2 -51.2 -- -120.5 -- 69.8 2 .psi. S 138.5 S 139.4 -- 38.5 -- -- .phi. -128.9 -114.1 -- 50.8 -- -- .sup.1 173.3 169.7 -- -173.2 -- -- -141.9p.2 -139.8 -- -120.6 -- -- 1 .psi. S 133.0 S 127.9 -- 131.9 -- -- .phi. 49.4 48.6 -- -133.8 -- -- -43.2.sup.1 -44.1 -- -43.5 -- -- -171.5up.2 -172.4 -- --3.6 --Carboxyl terminus112' .phi. L -113.3 L -113.0 L -112.4 L -112.0 L -111.5 L -110.6 .psi. 98.7 98.8 99.2 97.7 89.9 100.1 -72.3sup.1 -71.6 -71.6 -71.7 -71.8 -72.1 163.7sup.2 163.0 162.6 162.2 161.9 161.1113' .phi. S -83.4 S -77.2 S -77.4 S -76.8 S -91.9 T -89.5 .psi. 146.0 136.1 143.5 141.7 149.4 134.0 177.0sup.1 179.4 176.3 175.6 169.6 57.4 65.2.sup.2 60.6 59.0 57.5 73.2 -179.7114' .phi. R -66.7 R -71.5 R -78.0 R -75.1 R -117.0 I -57.9 .psi. -47.0 -53.7 -59.3 -57.1 -31.9 -32.2 -99.1sup.1 -145.6 -93.4 177.5 -168.1 -162.1 -173.1p.2 -179.8 -171.5 -159.7 86.5 87.9 170.4 X.sup.3 -173.6 -172.9 58.1 -73.8 -- X.sup.4 140.3138.4 141.5 124.7 173.4 --115' .phi. K -163.2 K -162.1 K -167.1 K -163.0 K -133.4 K -167.7 .psi. 110.6 132.0 153.3 157.6 141.7 126.1 -139.8up.1 -143.3 -140.1 -137.8 -164.0 -142.2 78.7.sup.2 81.7 85.0 79.3 172.4 83.5 X.sup.3 -96.4 -97.9 -95.8 -97.4 133.9 -100.4 X.sup.4 129.5 120.9 131.0 133.6 -73.6 109.2 -39.1.sup.5 -35.6 -47.7 -44.1 9.6 -9.5116' .phi. A -60.5 A -69.6 A -58.2 A -63.9 A -51.6 R -61.4 .psi. -47.5 -40.7 -47.5 -42.4 -45.3 -39.3 -- X.sup.1 -- -- -- -- -80.9 -- X.sup.2 -- -- -- -- -173.4 -- X.sup.3 -- -- -- -- 174.3 -- X.sup.4 -- -- -- -- 129.6117' .phi. V -71.8 T -85.5 T -64.6 V -41.0 T -114.6 G -73.3 .psi. 112.0 116.4 108.2 130.2 168.4 77.3 -74.3sup.1 53.4 50.9 -66.2 57.2 -- -- X.sup.2 60.5 58.0 -- -170.5 --118' .phi. R -93.8 R -118.2 R -58.5 R -125.2 R --51.5 R -49.0 .psi. -57.9 -48.4 -37.6 141.3 -56.3 -40.3 -104.3up.1 -80.9 -109.5 -126.2 -153.8 -125.9 79.1.sup.2 109.1 73.3 106.7 88.0 98.0 X.sup.3 116.1 147.7 -76.3 -78.8 -74.9 -76.6 X.sup.4 116.1 147.0 -87.2 -98.5 -125.9 -111.5__________________________________________________________________________ *Torsional angles are designated according to the IUPACIUB rules (1970).
TABLE 2______________________________________Residues of BDNF and NGF defining the surfaces of their central.beta.-strand stems, the side-chain thermal motion of which is takeninto account when optimizing the structures of the complexes withthe LRM-2B.BDNF NGF______________________________________Ile.sup.16 Lys.sup.57 -- Lys.sup.57Glu.sup.18 Asn.sup.59 Arg.sup.598Val.sup.20 Met.sup.61 Ser.sup.610Thr.sup.21 Lys.sup.65 --l.sup.22Asp.sup.24 Gln.sup.79 Tyr.sup.794Lys.sup.25 Arg.sup.81 Thr.sup.815Lys.sup.26 Thr.sup.82 Thr.sup.826Thr.sup.27 Gln.sup.84 His.sup.847Val.sup.29 Tyr.sup.86 Phe.sup.869Thr.sup.35 Arg.sup.104 Arg.sup.103Glu.sup.55 Asp.sup.106 Asp.sup.105______________________________________
TABLE 3______________________________________Components of the molecular mechanical potential energy of theintermolecular NGF/LRM-2A interactions (in kcal/mol)Term hNGF mNGF mNGF.DELTA.3-9 hNGF:NT-4/5______________________________________E.sub.w -61.2 -57.9 -39.3 -65.1E.sub.e -96.2 -87.9 -99.0 -65.2E.sub.hb -7.4 -5.9 -3.1 -4.6E.sub.qq -0.3 -0.3 -0.1 -0.1Total -165.1 -152.0 -141.5 -135.1______________________________________
TABLE 4______________________________________Components of the molecular mechanical potential energy of theBDNF/LRM-2B and the NGF/LRM-2B complexes (in kcal/mol)BDNF/LRM-2B NGF/LRM-2B Difference*______________________________________A. Components of the potential energyE.sub.w 89.3 82.3 -7.0E.sub.el -183.8 -93.1 90.7E.sub.hb -31.0 9.5Et 114.9 -0.9E.sub.qq 0.4 0.0 -0.4Total -10.2 92.0B. Components of the intermolecular interactions energyE.sub.w -48.0 -64.9 -16.9E.sub.el -221.4 -144.2 77.2E.sub.hb -5.6 -4.3 1.3Total -275.0 -213.5 61.5______________________________________ *Difference of the energy terms within the minimumenergy structures of th NGF/LRM2B and the BDNF/LRM2B complexes. Underlined values demonstrate tha the difference in intermolecular electrostatic interactions dominates in the differences in potential energies of the complexes.
TABLE 5__________________________________________________________________________Equilibrium variable torsional angles (in degrees) of the amino andcarboxyltermini of the most stable conformations of the uncomplexed (UC) andcomplexed (C) NGF analogues.* hNGF mNGF mNGF.DELTA.3-9No. UC C UC C UC C__________________________________________________________________________Amino terminus 11 .psi. E 82.9 78.2 E 85.9 85.4 E 89.2 96.6 .phi. 60.3 55.7 56.3 56.3 57.8 60.3 .chi..sup.1 -69.8 -69.3 -72.0 -69.4 -72.1 -73.7 .chi..sup.2 70.2 72.7 71.7 71.3 66.0 67.0 .chi..sup.3 -117.6 -120.4 -111.7 -117.1 -107.0 -109.2 10 .psi. -89.1 -122.1 G -95.1 -103.2 G -88.0 -94.8 .phi. -172.5 -166.5 -178.6 174.1 -177.5 168.7 9 .psi. R -36.1 -25.7 M -21.1 -14.3 146.5 166.1 .phi. -84.2 -92.8 -91.4 -77.3 -103.6 -132.1 .chi..sup.1 -77.2 -92.9 -75.2 -84.9 -66.7 -179.1 .chi..sup.2 -174.1-173.9 -179.7 179.0 64.0 122.2 .chi..sup.3 158.1 -175.5 179.9 171.5 -- -- .chi..sup.4 136.4 134.3 -- -- -- 8 .psi. 118.6 98.3 110.4 96.1 106.1 156.1 .phi. -127.36.5 -110.4 -121.8 52.8 46.0 .chi..sup.1 -108.0 -89.7 -90.9 -104.1 -45.9 -35.5 .chi..sup.2 -87.9 -73.8 -71.5 -65.8 -172.3 -174.1 7 .psi. F 135.7 113.6 F 129.4 106.7 -- -- .phi. -122.4 -68.8 -89.0 -130.8 -- -- .chi..sup.1 -147.2 -171.6 -163.9 -143.9 -- -- .chi..sup.2 -85.8 -96.3 -91.5 -87.5 -- -- 6 .psi. 40.3 I 57.9 61.7 67.9 -- -- .phi. 38.5 40.4 47.7 48.3 -- -- .chi..sup.1 -159.7 -165.7 -177.3 -177.3 -- -- .chi..sup.2 88.3 87.1 -- -- -- 5 .psi. P 156.0 164.8 P 149.8 153.3 -- -- 4 .psi. H 121.4 125.4 H 122.1 122.9 -- -- .phi. -133.8 -129.2 -129.8 -125.9 -- -- .chi..sup.1 -163.2.5 -163.3 -161.8 -- -- .chi..sup.2 -71.9 -69.1 -69.2 -69.9 -- -- 3 .psi. 2.1 -4.9 -1.9 -5.1 -- -- .phi. -78.9 -91.9 -84.5 -82.3 -- .chi..sup.1 -172.3 -171.8 -174.9 -171.4 -- -- .chi..sup.2 -51.2 -53.2 -51.2 -52.3 -- -- 2 .psi. 138.5 135.7 S 139.4 139.4 -- -- .phi. -157.5 -128.9 -114.1 -160.3 -- -- .chi..sup.1 173.4.3 169.7 170.3 -- -- .chi..sup.2 -133.0 -141.9 -139.8 -134.5 -- -- 1 .psi. 133.0 129.5 S 127.9 128.6 -- -- .phi. 46.6 48.6 49.4 53.8 -- -- .chi..sup.1 -41.9.2 -44.1 -44.0 -- -- .chi..sup.2 -171.0 -171.5 -172.4 -170.9 -- --Carboxyl terminus112' .phi. L -113.3 -113.1 L -113.0 -114.1 L -112.4 -113.7 .psi. 95.5 98.8 98.7 99.3 99.2 103.9 .chi..sup.1 -72.1 -72.3 -71.6 -72.1 -71.6 -72.1 .chi..sup.2 163.2 163.7 163.0 164.1 162.6 163.8113' .phi. -83.4 -72.5 S -77.2 -70.5 -77.4 -63.0 .psi. 136.1 147.9 146.0 148.6 143.5 145.8 .chi..sup.1 179.1.0 179.4 -177.9 176.3 -177.0 .chi..sup.2 65.2 56.0 60.6 56.3 59.0 52.4114' .phi. -66.7 -101.6 R -71.5 -92.5 -78.0 -92.3 .psi. -41.9 -47.0 -53.7 -43.8 -59.3 -50.4 .chi..sup.1 -175.3 -145.6 -173.2 -93.4 -167.4 .chi..sup.2 -130.0 -173.1 -179.8 -114.8 -171.5 -101.4 .chi..sup.3 159.9 170.4 -173.6 151.4 -172.9 172.3 .chi..sup.4 -117.4 138.4 140.3 -139.0 141.5 -130.9115' .phi. K -163.2 -155.3 K -162.1 -152.2 K -167.1 -161.4 .psi. 110.6 111.2 132.0 111.0 153.3 132.6 .chi..sup.1 -139.8 -150.7 -143.3 -156.2 -140.1 -144.6 .chi..sup.2 -170.9 78.7 81.7 -170.0 85.0 -177.7 .chi..sup.3 177.2 -96.4 -97.9 -174.3 -95.8 -169.3 .chi..sup.4 -165.4 129.5 120.9 -173.7 131.0 175.4 .chi..sup.5 -63.2 -39.1 -35.6 -61.4 -47.7 -53.3116' .phi. A -60.5 -62.3 A -69.6 -61.7 A -58.2 -67.9 47.5 -43.9 -47.8 -47.5 -40.4117' .phi. V -71.8 -106.1 T -85.5 -106.0 T -64.6 -74.3 .psi. 107.0 112.0 116.4 98.5 108.2 107.7 .chi..sup.1 -75.0 -74.3 53.4 55.1 50.9 56.6 .chi..sup.2 -- -- 63.2.5 58.0 54.0118' .phi. R -93.8 -58.0 R -118.2 -58.4 R -58.5 -79.4 -57.9 -53.7 -48.4 -59.7 -37.6 -49.6 -104.3sup.1 -88.5 -80.9 -84.7 -109.5 -135.3 .chi..sup.2 79.1 74.4 109.1 79.7 73.3 108.0 .chi..sup.3 147.7 161.4 116.1 147.5 -76.3 -70.1 .chi..sup.4 119.7 147.0 116.1 118.7 -87.2 -135.3__________________________________________________________________________ *Torsional angles are designated according to the IUPACIUB rules (1970). Torsional angles undergoing considerable changes upon LRM2A binding are shown in bold.
TABLE 6______________________________________Optimized conformational variables (in degrees) ofhNGF-complexed LRM-2A.*No.Type .phi. .psi. .chi..sup.1 .chi..sup.2 .chi..sup.3 .chi..sup.4 .chi..sup.5______________________________________ 94tA Arg -161.6 -62.5 -169.2 -164.9 66.8 172.2 95tA Asn -114.1 132.0 33.0 -111.6 96tA Leu -131.9 114.2 -161.2 141.1 97tA Thr -145.0 125.0 -179.9 -176.6 98tA Ile -123.1 21.6 61.3 105.7 99tA Val -61.2 107.3 175.7100tA Lys 54.4 40.8 -75.6 161.7 -174.8 166.2 -41.3101tA Ser -97.4 -40.0 -176.4 167.7102tA Gly 144.4 46.5103tA Leu -135.6 155.5 -85.0 90.0104tA Arg -81.7 179.3 -33.6 165.0 -52.8 -144.7105tA Phe -68.6 -40.8 -159.4 93.3106tA Val -106.1 94.3 176.1107tA Ala -173.5 -54.8108tA Pro -26.2109tA Asp -85.5 64.7 -116.1 102.8110tA Ala -117.5 -57.2111tA Phe -79.4 135.3 38.4 89.8112tA His -132.1 78.5 -67.1 89.9113tA Phe -120.8 -29.0 54.6 95.4114tA Thr -59.9 -56.0 -23.4 -176.5115tA Pro -47.9116tA Arg -150.8 142.7 46.8 98.5 -100.4 155.7117tA Leu -76.3 -39.4 57.0 117.9______________________________________ *Torsional angles are designated according to the IUPACIUB rules (1970).
TABLE 7______________________________________Optimized conformational variables (in degrees) ofBDNF-complexed LRM-2B.*No.Type .phi. .psi. .chi..sup.1 .chi..sup.2 .chi..sup.3 .chi..sup.4 .chi..sup.5______________________________________ 94tB Arg -109.3 -54.2 -69.7 169.4 -171.6 119.4 95tB Asn -163.6 141.0 -129.4 -109.4 96tB Leu -81.9 86.5 -92.9 -58.2 97tB Thr -138.3 143.5 -177.9 169.6 98tB Ile -136.4 136.6 -164.6 86.9 99tB Val -107.6 151.8 -70.1100tB Asp -68.8 -72.8 23.6 95.2101tB Ser 36.4 36.0 -156.1 157.6102tB Gly 42.0 44.4103tB Leu -118.2 149.0 49.9 87.1104tB Lys -92.9 68.5 -147.2 173.5 91.3 -169.5 45.1105tB Phe 73.7 -78.5 -160.5 -111.0106tB Val 52.6 19.9 -174.2107tB Ala -113.9 26.4108tB Tyr -58.7 -20.3 -46.5 102.7 -74.2109tB Lys -71.4 -27.6 62.1 -113.2 176.6 176.6 -55.9110tB Ala -91.7 -32.6111tB Phe -63.5 -37.4 -168.3 83.6112tB Leu -60.3 -43.2 -169.7 91.0113tB Lys -61.5 -48.0 -78.9 152.7 -167.9 63.2 22.9114tB Asn -59.4 -36.6 -70.9 -107.4115tB Ser -68.6 -32.9 -60.0 68.6116tB Asn -80.7 -45.4 -83.4 -96.8117tB Leu -127.8 -58.8 -75.6 105.5______________________________________ *Torsional angles are designated according to the IUPACIUB rules (1970).
TABLE 8______________________________________Contributions of individual residues to intermolecular electrostaticinteraction energies of the complexes BDNF/LRM-2B andNGF/LRM-2B (in kcal/mol)*.BDNF/LRM-2B NGF/LRM-2B Difference.dagger-dbl.______________________________________A. Binding epitopes of neurotrophinsGlu.sup.18 -61.6 Val.sup.18 0.3 61.9Asp.sup.24 -65.1 Asp.sup.24 -44.9 20.2Lys.sup.25 25.3 Lys25 15.5 -9.8Gly.sup.33 -11.9 Gly.sup.33 -13.4 -1.5Thr.sup.35 -11.4 Glu.sup.35 -50.5 -39.1Glu.sup.55 -50.2 Glu.sup.55 -18.9 31.3Lys.sup.57 27.1 Lys.sup.57 22.5 -4.6Arg.sup.81 -32.2 Thr.sup.81 0.1 32.3Gln.sup.81 1.2 His.sup.84 11.5 10.3Arg.sup.104 -14.3 Arg.sup.103 -9.5 4.8Asp.sup.106 -39.7 Asp.sup.105 -46.4 -6.7Total -232.8 -133.7 99.1B. Binding epitopes of the LRM-2BArg.sup.94tB -40.9 4.2 45.1Asn.sup.95tB 0.2 -12.8 -13.0Asp.sup.100tB -42.1 -9.3 32.8Ser.sup.101tB -17.9 -0.1 17.8Gly.sup.102tB -1.3 -12.0 -10.7Lys.sup.104tB -18.5 -48.1 -29.6Lys.sup.109tB -46.4 -21.5 24.9Lys.sup.113tB -54.6 -37.2 17.4Total -221.5 -136.8 84.7______________________________________ *The .+-.10 kcal/mol cutoff of intermolecular electrostatic interaction energy has been used to identify binding epitopes within the complexes. .dagger-dbl.Difference of the intermolecular electrostatic interaction energies of the given residue within the minimumenergy structures of the NGF/LRM2B and the BDNF/LRM2B complexes. Residues within the binding epitopes of BDNF and the LRM2B which cause their binding specificity are bolded.
TABLE 9__________________________________________________________________________Contributions of individual residues of the neurotrophins to theintermolecularinteraction energy of their complexes with the common neurotrophinreceptor p75.sup.NTR(kcal/mol)*NGF BDNF NT-3 NT-4/5Res. EL VDW Res. EL VDW Res. * EL VDW Res. EL VDW__________________________________________________________________________D-30 17.9 -0.9 D-30 25.0 -1.8 D-29 24.6 -2.6 D-32 19.8 -1.8K-32 -94.9 -6.0 S-32 0.5 -4.5 R-31 -48.2 -8.0 R-34 -42.6 -9.3K-34 -88.8 -3.3 G-34 -1.8 -3.1 H-33 -35.0 -7.1 R-36 -57.2 -7.7E-35 -10.0 -2.2 T-35 1.7 -1.4 Q-34 0.5 -2.3 D-37 -0.8 -1.7H-84 -30.9 -1.0 Q-84 1.19 -2.4 Q-83 0.8 -0.7 Q-83 0.6 -1.0F-86 -0.2 -0.3 Y-86 -0.5 -0.5 Y-85 -13.4 0.3 Y-96 -9.2 -0.3K-88 -8.5 -0.1 R-88 -12.7 -0.2 R-87 -20.0 -0.3 R-98 -16.2 -0.2D-93 -9.0 -1.3 D-93 13.7 -2.2 E-92 24.1 -2.8 D-103 9.0 -2.1K-95 -63.7 -5.6 K-95 -82.9 -7.6 N-94 -0.1 -0.2 Q-105 -0.3 -0.2Q-96 -0.3 -3.7 K-96 -58.2 2.0 K-95 -49.6 -2.6 G-106 -0.6 -0.6A-97 0.2 -0.2 R-97 -79.3 -5.7 L-96 -0.5 -6.7 R-107 -28.0 -11.2R-100 -11.8 -0.7 R-101 -59.2 -1.1 R-100 -79.3 1.7 R-111 -57.8 -1.9R-103 -64.5 0.0 R-104 -64.2 -1.6 R-103 -51.5 0.6 R-114 -55.3 -2.9D-105 10.6 -0.2 D-106 8.3 -0.2 D-105 5.5 -0.1 D-116 4.8 -0.1Total -336.1 -15.6 -317.8 -18.2 -248.3 -15.6 -237.3 -35.0 55%00% 99% 39% 97% 32% 102% 61%Net -337.3 -32.3 -321.4 -46.7 -254.8 -48.9 -233.0 -57.5__________________________________________________________________________ *EL and VDW denote electrostatic and van der Waals terms of the ligandreceptor interaction energy, respectively. The .+-.10 kcal/mol cutoff of the intermolecular electrostatic interaction energy has been used to identify the functional epitopes (bolded). "Total" electrostatic and van der Waals contributions of the functional epitopes to the "net" values of the intermolecular interaction energy calculated over the whole binding interface are # presented. Residue numbering corresponds to the general alignment of neurotrophins (Bradshaw et al, 1994).
TABLE 10__________________________________________________________________________Contributions of individual residues of p75.sup.NTR to the intermolecularinteractionenergy of their complexes with neurotrophins (kcal/mol)*.NGF BDNF NT-3 NT-4/5Res. EL VDW Res. EL VDW ELes. VDW Res. EL VDW__________________________________________________________________________D-47 -46.0 0.2 D-47 -67.1 0.1 D-47 -50.5 -0.3 D-47 -26.7 -4.1E-53 -2.6 -0.2 E-53 -55.4 1.6 -6.63 -0.2 E-53 -3.6 -0.1P-54 -12.4 -2.8 P-54 -13.5 -2.5 P-54 -15.1 0.2 P-54 -5.3 -0.8C-55 -1.4 -0.9 C-55 -8.4 -1.3 C-55 -14.9 -1.0 C-55 -3.0 -1.2K-56 21.8 -4.5 K-56 13.2 -4.8 K-56 32.9 -5.0 K-56 14.8 -5.8E-60 -6.2 -0.3 E-60 -3.0 -0.1 E-60 -10.0 -0.2 E-60 -10.4 -0.4E-73 -4.2 0.0 E-73 -9.7 -0.2 E-73 -12.5 -0.3 E-73 -11.2 -0.3D-75 -56.5 0.6 D-75 -77.1 0.4 -65.5 -1.1 D-75 -59.4 -2.7D-76 -91.6 -2.0 D-76 -22.3 -8.4 D-76 -39.9 -3.7 D-76 -39.3 -5.1R-80 -18.5 -3.9 R-80 2.9 -2.1 R-80 5.3 -4.0 R-80 3.1 -4.3D-88 -49.1 0.4 D-88 -39.8 -0.4 D-88 -34.4 0.4 D-88 -33.1 -1.2E-89 -32.5 -1.6 E-89 -25.2 -6.2 E-89 -33.8 -3.8 E-89 -33.0 -5.9Total -285.4 -13.6 -287.2 -20.2 -243.7 -14.8 -198.3 -25.5 42%4% 89% 43% 96% 30% 85% 44%__________________________________________________________________________ *See footnote to Table 9.
REFERENCES
Allen, M. P., and D. J. Tildesley. 1987. Computer Simulation of Liquids. Claredon Press: Oxford.
Baldwin, A. N., and E. M. Shooter. 1994. J. Biol. Chem. 269:11456-11461.
Baldwin, A. N., and E. M. Shooter. 1995. J. Biol. Chem. 270:4594-4602.
Banner, D. W., A. D'Arcy, W. Janes, R. Gentz, H.-J. Schoenfeld, C. Broger, H. Loetscher, and W. Lesslauer. 1993. Cell 73:431-445.
Barbacid, M. 1993. Oncogene 8:2033-2042.
Barbacid, M. 1994. J. Neurobiol. 25:1386-1403.
Barde, Y.-A. 1989. Neuron 2:1525-1534.
Barker, P. A., and E. M. Shooter. 1994. Neuron 13:203-215.
Baumann, U.; Wu, S.; Flaherty, K. M.; McKay, D. B. 1993. EMBO J. 12: 3357-3364.
Ben Ari, Y.; Represa, A. 1990. TINS 13: 312-318.
Benedetti, M., A. Levi, and M. V. Chao. 1993. Proc. Natl. Acad. Sci. U.S.A. 90:7859-7863.
Berkemeier, L., J. Winslow, D. Kaplan, K. Nicolics, D. Goeddel, and A. Rosenthal. 1991. Neuron 7:857-866.
Bothwell, M. A. 1991. Cell 65:915-918.
Bothwell, M. A., Shooter, E. M. 1977. J. Biol. Chem. 23: 8532-8536.
Bradshaw, R. A., J. Murray-Rust, C. F. Ibanez, N. Q. McDonald, R. Lapatto, and T. L. Blundell. 1994. Protein Science 3:1901-1913.
Brooks, B. R., R. E. Bruccoleri, B. D. Olason, D. J. States, S. Swaminathan, and M. Karplus. 1983. J. Comput. Chem. 4:187-217.
Brooks III, C. L., and D. A. Case, 1993. Chem. Rev. 93: 2487-2502.
Brunger, A. T. 1988. J. Mol. Biol. 203:803-816.
Burton, L. E., C. H. Schmelzer, E. Szonyi, C. Yedinak, and A. Gorrell. 1992. J. Neurochem. 59:1937-1945.
Burton, G., C. Schmelzer, M. Sadic, F. Hefti, and J. Treanor. 1995. Soc. Neurosci. Abs. 21:1061.
Carter, B. D., C. Kaltschmidt, B. Kaltschmidt, N. Offenhauser, R. Bohm-Matthaei, P. A. Baeuerle, and Y.-A. Barde. 1996. Science 272:542-545.
Cassacia-Bonnefil, P., B. D. Carter, R. T. Dobrowsky, and M. V. Chao. 1996. Nature 383:716-719.
Chang, G., W. C. Guida, and W. C. Still. 1989. J. Am. Chem. Soc. 111:4379-4386.
Chao, M. V. 1992a. Cell 68:995-997.
Chao, M. V. 1992b. Neuron 9:583-593.
Chao, M. 1994. J. Neurobiol. 25:1373-1385.
Chao, M. V., and B. L. Hempstead. 1995. Trends Neurosci. 18:321-326.
Clackson, T., and J. A. Wells. 1995. Science 267:383-386.
Clary, D. O., and L. F. Reichardt. 1994. Proc. Natl. Acad. Sci. U.S.A. 91:11133-11137.
Davies, A. M. 1994. J. Neurobiol. 25:1334-1348.
Dobrowsky, R., M. H. Werner, A. M. Castellino, M. V. Chao, and Y. A. Hannun. 1994. Science 265:1596-1599.
Drinkwater, C. C., P. A. Barker, U. Suter, and E. M. Shooter. 1993. J. Biol. Chem. 268:23202-23207.
Escandon, E., L. E. Burton, E. Szonyi, and K. J. Nikolics. 1993. Neurosci. Res. 34:601-613.
Frade, J. M., A. Rodrigues-Tebar, Y.-A. Barde. 1996. Nature 383:166-168.
Gibson, K. D., and H. A Scheraga. 1988. In Structure and Expression: Vol. 1. From Proteins to Ribosomes; Sarma, M. H., Sarma, R. H., Eds.; Adenine Press: Guilderland, N.Y., pp. 67-94.
Gidas, B. 1985. J. Stat. Phys. 39: 73-131.
Glass, D.; Nye, S.; Hantzopoulos, P.; Macchi, M.; Squinto, S.; Goldfarb, M.; Yancopoulos, G. 1991. Cell 66: 405-413.
Gotz, R.; Koster, R.; Winkler, C.; Raulf, F.; Lottspeich, F.; Schard, M.; Thoenen, H. 1994. Nature 372: 266-269.
Gregory, D. S., A. C. R. Martin, J. C. Cheetham, and A. R. Rees. 1993. Protein Engineenng 6:29-35.
Gunningham, G. W.; Meijer, P. H. E. 1976. J. Comp. Phys. 20: 50-63.
Hallbook, F., C. F. Ibanez, and H. Persson. 1991. Neuron 6:845-858.
Hefti, F. H. 1986. J. Neurosci. 6:2155-2162.
Hefti, F. H., and Weiner, W. J. 1986. Annals of Neurology 20:275-281.
Heldin, C. H., A. Emlund, C. Rorsman, and L. Ronnstrand. 1989. J. Biol. Chem. 264:8905-8912.
Hempstead, B. L., D. Martin-Zanca, D. L. Kaplan, L. F. Parada, and M. V. Chao. 1991. Nature 350:678-683.
Herrmann, J. L., D. G. Menter, J. Hamada, D. Marchetti, M. Nakajima, and G. L. Nicolson. 1993. Mol. Biol. 4:1205-1216.
Hobza, P., H. L. Selzle, and E. W. Schlag. 1994. J. Am. Chem. Soc. 116:3500-3506.
Hodes, Z. I., G. Nemethy, and H. A. Scheraga. 1979. Biopolymers 18:1565-1610.
Hohn, A.; Leibrock, J.; Bailey, K.; Barde, Y.-A. 1990. Nature 344: 339-341.
Hopfinger, A. J. 1973. Conformational Properties of Macromolecules. Academic Press, New York-London. pp. 70-85.
Hunter, C. A., and J. K. M. Sanders. 1990. J. Am. Chem. Soc. 112:5525-5534.
Hunter, C. A., J. Singh, and J. M. Thornton. 1991. J. Mol. Biol. 218:837-846.
Ibanez, C. F., T. Ebendal, and H. Persson. 1991. EMBO J. 10:2105-2110.
Ibanez, C. F., T. Ebendal, G. Barbany, J. Murray-Rust, T. L. Blundell, and H. Persson. 1992. Cell 69:329-341.
Ibanez, C. F., L. Ilag, J. Murray-Rust, and H. Persson. 1993. EMBO J. 12:2281-2293.
Ibanez, C. F. 1994. J. Neurobiol. 25:1349-1361.
Ibanez, C. F. 1995. Trends Biotech. 13:217-227.
Ilag, L. L., P. Lonnerberg, H. Persson, and C. F. Ibanez. 1994. J. Biol. Chem. 269:19941-19946.
IUPAC-IUB Commission on Biochemical Nomenclature. 1970. Biochemistry 9:3471-3479.
Jing, S. Q., P. Tapley, and M. Barbacid. 1992. Neuron 9:1067-1079.
Johnson, D., A. Lanahan, C. R. Buck, A. Sehgal, C. Morgan, E. Mercer, M. Bothwell, and M. Chao. 1986. Cell 47:545-554.
Judson, R. S.; Jaeger, E. P.; Treasurywala, A. M. 1994. J. Mol. Struct. (Theochem) 308: 191-206.
Kahle, P., L. E. Burton, C. H. Schmeizer, and C. Hertel. 1992. J Biol. Chem. 267:22707-22710.
Kaplan, D. R.; Hempstead, B. L.; Martin-Zanca, D.; Chao, M. V.; Parada, L. F. 1991. Science 252: 554-558.
Kaplan, D. R., and R. M. Stephens. 1994. J. Neurobiol. 25:1404-1417.
Kincaid, R. H.; Scheraga, H. A. 1982. J. Comput. Chem. 3: 525-547.
Kirkpatrick, S., C. D. Gelatt, Jr., and M. P. Vecchi. 1983. Science 220:671-680.
Klein, R.; Jing, S.; Nanduri, V.; O'Rourke, E.; Barbacid, M. 1991. Cell 65: 189-197.
Klein, R.; Lamballe, F.; Bryant, S.; Barbacid, M. 1992. Neuron 8: 947-956.
Kobe, B., and J. Deisenhofer. 1993. Nature 366:751-756.
Kobe, B., and J. Deisenhofer. 1994. Trends Biochem. Sci. 19:415-421.
Kobe, B., and J. Deisenhofer. 1995. Nature 374:183-186.
Korsching, S. 1993. J. Neurosci. 13:2739-2748.
Kruttgen, A., J. V. Heymach, P. Kahle, and E. M. Shooter. 1996. Soc. Neurosci. Abs. 22:557.
Kullander, K., and T. Ebendal. 1994. J Neurosci. Res. 39:195-210.
Kullander, K., and T. Ebendal. 1996. Soc. Neurosci. Abs. 22:557.
Kushner, H. J. 1987. SIAM J. Appl. Math. 47: 169-185.
Lai, K.-O., D. J. Glass, D. Geis, G. D. Yancopoulos, and N. Y. Ip. 1996. J. Neurosci. Res. 46:618-629.
Lamballe, F.; Klein, R.; Barbacid, M. 1991. Cell 66: 967-970.
Landreth, G. E., and E. M. Shooter. 1980. Proc. Natl. Acad. Sci. U.S.A. 77: 4751-4755.
Large, T. H., G. Weskamp, J. C. Helder, M. J. Radeke, T. P. Misko, E. M. Shooter, and L. F. Reichardt. 1989. Neuron 2:1123-1134.
Leibrock, J.; Lottspeich, A. H.; Hohn, A.; Hofer, M.; Hengerer, B.; Masiakowski, P.; Thoenen, H.; Barde, Y.-A. 1989. Nature 341: 149-152.
Lelj, F.; Grimaldi, P.; Cristianziano, P. L. 1991. Biopolymers 31: 663-670.
Leven, G. R.; Mendel, L. M. 1993. TINS 16: 353-359.
Levi-Montalcini, R. 1987. EMBO J. 6:1145-1154.
Li, Z.; Scheraga, H. A. 1988. J. Mol Struct. (Theochem) 179: 333-352.
Livnah, O., E. A. Stura, D. L. Johnson, S. A. Middleton, L. S. Mulcahy, N. C. Wrighton, W. J. Dower, L. K. Jolliffe, and I. A. Wilson. 1996. Science 273:464-471.
Loeb, D. M.; Maragos, J.; Martin-Zanca, D.; Chao, M. V.; Parada, L. F.; Greene, L. A. 1991. Cell 66: 961-966.
Luo, Y., and K. E. Neet. 1992. J. Biol. Chem. 267:12275-12283.
MacDonald, J. I. S., and S. O. Meakin. 1996. Mol. Cell. Neurosci. 7:371-390.
Mahadeo, D., L. Kaplan, M. V. Chao, and B. L. Hempstead. 1994. J. Biol. Chem. 269:6884-6891.
Maisopierre, P. C.; Belluscio, L. S. S.; Squinto, S.; Ip, N.Y.; Furth, M. E.; Lindsay, R. M.; Yancopoulos, G. D. 1990. Science 247: 1446-1451.
Maness, L. M., A. J. Kastin, J. T. Weber, W. A. Banks, B. S. Beckman, and J. E. Zadina. 1994. Neurosci. Biobehav. Rev. 18:143-159.
Marchetti, D., McQuillan, D. J., W. C. Spohn, D. D. Carson, and G. L. Nicolson. 1996. Cancer Res. 56:2856-2863.
Matsumoto, K., R. K. Wada, J. M. Yamashiro, D. R. Kaplan, and C. J. Thiele. 1995. Cancer Res. 55:1798-1806.
McDonald, N. Q., R. Lapatto, J. Murray-Rust, J. Gunning, A. Wlodawer, and T. L. Blundell. 1991. Nature 354:411-414.
McKee, A. C.; Kosik, K. S.; Kowal, N. W. 1991. Ann. Neurol. 30:156.
McMahon, S. B.; Bennett, D. L. H.; Priestley, J. V.; Shelton, D. L. 1995. Nature Med. 1: 774-780.
Meakin, S. O., and E. M. Shooter. 1992. Trends Neurosci. 15:323-331.
Merrell, R., M. W. Pulliam, L. Randono, L. F. Boyd, R. A. Bradshaw, and L. Glaser. 1975. Proc. Natl. Acad. Sci. U.S.A. 72:4270-4274.
Metropolis, N.; Rosenbluth, A. W.; Rosenbluth, M. N.; Teller, A. H.; Teller, E. 1953. J. Chem. Phys. 21: 1087-1092.
Moore, J. B., and E. M. Shooter. 1975. Neurobiology 5:369-381.
Morgan et al. 1989. In Annual Reports in Medicinal Chemistry. Ed.: Vinick, F. J. Academic Press, San Diego, Calif., pp.243-252.
Nayeem, A.; Vila, J.; Scheraga, H. A. 1991. J. Comput. Chem. 12: 594-605.
Nilges, M., G. M. Clore, and A. M. Gronenborn. 1988. FEBS Lett. 229:317-324.
Paine, G. H.; Scheraga, H. A. 1985. Biopolymers 24:1391-1436.
Perez, P., P. M. Coll, B. L. Hempstead, D. Martin-Zanca, and M. V. Chao. 1995. Mol. Cell. Neurosci. 6:97-105.
Persson, H., and C. F. Ibanez. 1993. Curr. Opin. Neurol. Neurosurg. 6:11-18.
Piela, L.; Kostrowicki, J.; Scheraga, H. A. 1989. J. Phys. Chem. 93: 3339-3346.
Piela, L.; Olszewski, K. A.; Pillardy, J. 1994. J. Mol. Struct. (Theochem) 308: 229-239.
Ponnuswamy, P. K., and P. Manavalan. 1976. J. Theor. Biol. 60:481-486.
Purves, D. 1988. Body and Brain. A Trophic Theory of Neural Connectors. Cambridge, Mass.: Harvard University Press, pp. 1-1231.
Radeke, M. J., T. P. Misko, C. Hsu, L. A. Herzenberg, and E. M. Shooter. 1987. Nature 325:593-597.
Radziejewski, C., and R. C. Robinson. 1993. Biochemistry 32:13350-13356.
Radziejewski, C.; Robinson, R. C.; DiStefano, P. S.; Taylor, J. W. 1992. Biochemistry 31: 4431-4436.
Raj, N.; Morley, S. D.; Jackson, D. E. 1994. J. Mol. Struct. (Theochem) 308: 175-190.
Rao, M.; Pangali, C.; Berne, B. J. 1979. Mol. Phys. 37: 1773-1798.
Rashid, K.; van der Zee, C. E. E. M.; Ross, G. M.; Chapman, C. A.; Stanisz, J.; Riopelle, R. J.;
Racine, R. J.; Fahnestock, M. 1995. Proc. Natl. Acad. Sci. U.S.A. 92: 9495-9499.
Ripoll, D. R.; Scheraga, H. A. 1988. Biopolymers 27: 283-1303.
Robinson, R. C., C. Radziejewski, D. I. Stuart, and E. Y. Jones. 1995. Biochemistry 34:4139-4146.
Rodrigues-Tebar, A., G. Dechant, and Y.-A. Barde. 1990. Neuron 4:487-492.
Rodnigues-Tebar, A., G. Dechant, R. Gotz, and Y.-A. Barde. 1992. EMBO J. 11:917-922.
Rosenthal, A.; Goeddel, D. V.; Nguyen, T.; Lewis, M.; Shih, A.; Laramee, G. R.; Nikolics, K.; Winslow, J. W. 1990. Neuron 4: 767-773.
Ross, A. H., M.-C. Daou, C. A. McKinnon, P. J. Condon, M. B. Lachyankar, R. M. Stephens, D. R. Kaplan, and D. E. Wolf. 1996. J. Cell Biol. 132:945-953.
Ross, G. M., I. L. Shamovsky, G. Lawrance, M. Solc, S. M. Dostaler, S. L. Jimmo, D. F. Weaver, and R. J. Riopelle. 1997. Nature Med. 3:872-878.
Rossky, P. J.; Doll, J. D.; Friedman, H. L. 1978. J. Chem. Phys. 69: 4628-4633.
Rovelli, G.; Heller, R. A.; Canossa, M.; Shooter, E. M. 1993. Proc. Natl. Acad. Sci. USA 90:8717-8721.
Ryckaert, J.-P., Ciccotti, G.; Berendsen, H. J. C. 1977. J. Comput. Phys. 23:327-341.
Ryden, M., J. Murray-Rust, D. Glass, L. L. Ilag, M. Trupp, G. D. Yancopoulos, N. Q. McDonald, and C. F. Ibanez. 1995. EMBO J. 14:1979-1990.
Ryden, M., and C. F. Ibanez, C. F. 1996. J. Biol. Chem. 271:5623-5627.
Schauer, M., and E. R. Bernstein. 1985. J. Chem. Phys. 82:3722-3727.
Schechter, A. L., and M. A. Bothwell. 1981. Cell 24:867-874.
Scheraga, H. A. 1989. Prog. Clin. Biol. Res. 289: 3-18.
Scheraga, H. A. 1994. Pol. J. Chem. 68: 889-891.
Schmelzer, C. H., L. E. Burton, W.-P. Chan, E. Martin, C. Gorman, E. Canova-Davis, V. T. Ling, M. B. Sliwkowski, G. McCray, J. A. Briggs, T. H. Nguyen, and G. Polastri. 1992. J. Neurochem. 59:1675-1683.
Schneider, R., E. Schneider-Scherzer, M. Thurnher, B. Auer, and M. Schweiger. 1988. EMBO J. 7:4151-4156.
Schneider, R., and M. Schweiger. 1991. Oncogene 6:1807-1811.
Shamovsky, I. L.; Yarovskaya, I. Yu.; Khrapova, N. G.; Burlakova, E. B. 1992. J. Mol. Struct. (Theochem) 253: 149-159.
Shamovsky, I. L., G. M. Ross, R. J. Riopelle, and D. F. Weaver. 1996. J. Am. Chem. Soc. 118:9743-9749.
Shih, A., G. R. Laramee, C. H. Schmeizer, L. E. Burton, and J. W. Winslow. 1994. J. Biol. Chem. 269:27679-27686.
Simon, I.; Nemethy, G.; Scheraga, H. A. 1978. Macromolecules 11: 797-804.
Snider, W. D., and E. M. Johnson. 1989. Ann. Neurol. 26:489-506.
Snow, M. E. 1992. J. Comput. Chem. 13: 579-584.
Soppet, D.; Escandon, E.; Maragos, J.; Middlemas, D. S.; Reid, S. W.; Blair, J.; Burton, L. E.;
Stanton, B.; Kaplan, D. R.; Hunter, T.; Nikolics, K.; Parada, L. F. 1991. Cell 65: 895-903.
Squinto, S. P.; Stitt, T. N.; Aldrich, T. H.; Davis, S.; Bianco, S. M.; Radziejewski, C.; Glass, D. J.; Masiakowski, P.; Furth, M. E.; Valenzuela, D. M.; DiStefano, P. S.; Yancopoulos, G. D. 1991. Cell 65: 885-893.
Suter, U., C. Angst, C.-L. Tien, C. C. Drinkwater, R. M. Lindsay, and E. M. Shooter. 1992. J. Neurosci. 12:306-318.
Sutter, A., R. J. Riopelle, R. M. Harris-Warrick, and E. M. Shooter. 1979. J. Bio. Chem. 254:5972-5982.
Szentpsly, L. V.; Shamovsky, I. L.; Nefedova, V. V.; Zubkus, V. E. 1994. J. Mol. Struct. (Theochem) 308: 125-140.
Taylor, R. A. V., J. F. Kerrigan, F. M. Longo, M. deBoisblanc, and W. C. Mobley. 1991. Soc. Neurosci. Abs. 17: 712.
Thoenen, H. 1991. Trends Neurosci. 14: 165-170.
Treanor, J. J. S., C. Schmeizer, B. Knusel, J. W. Winslow, D. L. Shelton, F. Hefti, K. Nikolics, and L. E. Burton. 1995. J. Biol. Chem. 270: 23104-23110.
Ullrich, A., and J. Schlessinger. 1990. Cell 61: 203-212.
Urfer, R., P. Tsoulfas, D. Soppet, E. Escandon, L. F. Parada, and L. G. Presta. 1994. EMBO J. 13:5896-5909.
Urfer, R., P. Tsoulfas, L. O'Connell, D. L. Shelton, L. F. Parada, and L. G. Presta. 1995. EMBO J. 14:2795-2805.
Vale, R. D.; Shooter, E. M. 1985. Methods Enzymol. 109: 21-39.
Vanderbilt, D., and S. G. Louie. 1984. J. Comput. Phys. 56:259-271.
Van der Zee, C. E. E., G. M. Ross, R. J. Riopelle, and T. Hagg. 1996. Science 274: 1729-1732.
Villani, V.; Tamburro, A. M. 1994. J. Mol. Struct. (Theochem) 308: 141-157.
Warme, P. K., and R. S. Morgan. 1978. J. Mol. Biol. 118: 289-304.
Washiyama, K., Y. Muragaki, L. B. Rorke, V. M.-Y. Lee, S. C. Feinstein, M. J. Radeke, D. Blumberg, D. R. Kaplan, and J. Q. Trojanowski. 1996. Amer. J. Path. 148:929-940.
Weiner, S. J., P. A. Kollman, D. A. Case, U. C. Singh, C. Ghio, G. Alagona, S. Profeta Jr., and P. Weiner. 1984. J. Am. Chem. Soc. 106:765-784.
Weskamp, G.; Reichardt, L. F. 1991. Neuron 6: 649-663.
Wilson, C., and S. Doniach. 1989. Proteins 6:193-209.
Wilson, S. R., W. Cui, J. W. Moscowitz, and K. E. Schmidt. 1988. Tetrahedron Lett. 29:4373-4376.
Windisch, J. M., B. Auer, R. Marksteiner, M. E. Lang, and R. Schneider. 1995a. FEBS Lett. 374:125-129.
Windisch, J. M., R. Marksteiner, and R. Schneider. 1995b. J. Biol. Chem. 270: 28133-28138.
Windisch, J. M., R. Marksteiner, M. E. Lang, B. Auer, and R. Schneider. 1995c. Biochemistry 34:11256-11263.
Wolf, D. E., C. A. McKinnon, M.-C. Daou, R. M. Stephens, D. R. Kaplan, and A. H. Ross. 1995. J. Biol. Chem. 270:2133-2138.
Woo, S. B., D. E. Timm, and K. E. Neet. 1995. J Biol. Chem. 270:6278-6285.
Woo, S. B., and K. E. Neet. 1996. J Biol. Chem. 271:24433-24441.
Woolf, C. J.; Doubell, T. A. 1994. Current Opinions in Neurobiol. 4: 525-534.
Yan, H., and M. V. Chao. 1991. J. Biol. Chem. 266:12099-12104.
Yoder, M. D.; Keen, N. T.; Jurnak, F. 1993. Science 260: 1503-1507. ##SPC1##
__________________________________________________________________________# SEQUENCE LISTING- (1) GENERAL INFORMATION:- (iii) NUMBER OF SEQUENCES: 17- (2) INFORMATION FOR SEQ ID NO: 1:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 16 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..16 (D) OTHER INFORMATION:/not - #e= "N-terminal residues 1-16 of human NGF - #"#1: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Ser Ser Ser His Pro Ile Phe His Arg Gly Gl - #u Phe Ser Val Cys Asp# 15- (2) INFORMATION FOR SEQ ID NO: 2:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 16 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..16 (D) OTHER INFORMATION:/not - #e= "N-terminal residues 1-16 of mouse NGF - #"#2: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Ser Ser Thr His Pro Val Phe His Met Gly Gl - #u Phe Ser Val Cys Asp# 15- (2) INFORMATION FOR SEQ ID NO: 3:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 9 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..9 (D) OTHER INFORMATION:/not - #e= "N-terminus of deletion mutant de - #lta 3-9 of mouse NGF"#3: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Ser Ser Gly Glu Phe Ser Val Cys Asp1 5- (2) INFORMATION FOR SEQ ID NO: 4:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 15 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..15 (D) OTHER INFORMATION:/not - #e= "N-terminal residues 1-15 of mutant H4 - #D of human NGF"#4: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Ser Ser Ser Asp Pro Ile Phe His Arg Gly Gl - #u Phe Ser Val Cys# 15 10- (2) INFORMATION FOR SEQ ID NO: 5:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 13 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..13 (D) OTHER INFORMATION:/not - #e= "N-terminal residues 1-13 of BDNF"#5: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- His Ser Asp Pro Ala Arg Arg Gly Glu Leu Se - #r Val Cys# 10- (2) INFORMATION FOR SEQ ID NO: 6:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 11 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..11 (D) OTHER INFORMATION:/not - #e= "C-terminal residues 108-118#NGF" of human#6: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Val Cys Val Leu Ser Arg Lys Ala Val Ar - #g# 10- (2) INFORMATION FOR SEQ ID NO: 7:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 11 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..11 (D) OTHER INFORMATION:/not - #e= "C-terminal residues 108-118#NGF" of mouse#7: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Val Cys Val Leu Ser Arg Lys Ala Thr Ar - #g# 10- (2) INFORMATION FOR SEQ ID NO: 8:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 11 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..11 (D) OTHER INFORMATION:/not - #e= "C-terminal residues 108-118 of BDNF"#8: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Val Cys Thr Leu Thr Ile Lys Arg Gly Ar - #g# 10- (2) INFORMATION FOR SEQ ID NO: 9:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 11 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..11 (D) OTHER INFORMATION:/not - #e= "C-terminal residues 108-118 of NT-4/5 - #"#9: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Val Cys Thr Leu Leu Ser Arg Thr Gly Ar - #g# 10- (2) INFORMATION FOR SEQ ID NO: 10:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 4 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..4 (D) OTHER INFORMATION:/not - #e= "Residues 74-77 of human NGF, mous - #e NGF or delta 3-9 mutant of mouse NGF"#10: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Lys His Trp Asn- (2) INFORMATION FOR SEQ ID NO: 11:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 4 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..4 (D) OTHER INFORMATION:/not - #e= "Residues 74-77 of NT-4/5"#11: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Arg His Trp Val1- (2) INFORMATION FOR SEQ ID NO: 12:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 25 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..25 (D) OTHER INFORMATION:/not - #e= "Second leucine rich motif#93 to 117"LRM)of TrkA, residues#12: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Leu Arg Asn Leu Thr Ile Val Lys Ser Gly Le - #u Arg Phe Val Ala Pro# 150- Asp Ala Phe His Phe Thr Pro Arg Leu# 25- (2) INFORMATION FOR SEQ ID NO: 13:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 24 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..24 (D) OTHER INFORMATION:/not - #e= "Second Leucine rich motif(LRM)#residues 93 to 117"kB,#13: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Leu Arg Asn Leu Thr Ile Val Asp Ser Gly Le - #u Lys Phe Val Ala Tyr# 15- Lys Ala Glu Leu Lys Asn Ser Asn 20- (2) INFORMATION FOR SEQ ID NO: 14:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 58 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..58 (D) OTHER INFORMATION:/not - #e= "Identified neurotrophin binding s - #ite of human p75NTR, residues 39-96"- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..2 (D) OTHER INFORMATION:/not - #e= "cysteine 39 forms a#bond with cysteine 55"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:17..18 (D) OTHER INFORMATION:/not - #e= "cysteine 55 forms a#bond with cysteine 39"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:20..21 (D) OTHER INFORMATION:/not - #e= "cysteine 58 forms a#bond with cysteine 71"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:33..34 (D) OTHER INFORMATION:/not - #e= "cysteine 71 forms a#bond with cysteine 58"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:23..24 (D) OTHER INFORMATION:/not - #e= "cysteine 61 forms a#bond with cysteine 79"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:41..42 (D) OTHER INFORMATION:/not - #e= "cysteine 79 forms a#bond with cysteine 61"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:43..44 (D) OTHER INFORMATION:/not - #e= "cysteine 81 forms a#bond with cysteine 94"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:56..57 (D) OTHER INFORMATION:/not - #e= "cysteine 94 forms a#bond with cysteine 81"e- (xi) SEQUENCE DESCRIPTION: SEQ ID NO: - # 14:- Cys Leu Asp Ser Val Thr Phe Ser Asp Val Va - #l Ser Ala Thr Glu Pro# 15- Cys Lys Pro Cys Thr Glu Cys Val Gly Leu Gl - #n Ser Met Ser Ala Pro# 305- Cys Val Glu Ala Asp Asp Ala Val Cys Arg Cy - #s Ala Tyr Gly Tyr Tyr# 45- Gln Asp Glu Thr Thr Gly Arg Cys Glu Ala# 55- (2) INFORMATION FOR SEQ ID NO: 15:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 58 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..58 (D) OTHER INFORMATION:/not - #e= "Identified neurotrophin binding s - #ite of rat p75NTR, residues 39-96"- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..2 (D) OTHER INFORMATION:/not - #e= "cysteine 39 forms a#bond with cysteine 55"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:17..18 (D) OTHER INFORMATION:/not - #e= "cysteine 55 forms a#bond with cysteine 39"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:20..21 (D) OTHER INFORMATION:/not - #e= "cysteine 58 forms a#bond with cysteine 71"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:33..34 (D) OTHER INFORMATION:/not - #e= "cysteine 71 forms a#bond with cysteine 58"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:23..24 (D) OTHER INFORMATION:/not - #e= "cysteine 61 forms a#bond with cysteine 79"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:41..42 (D) OTHER INFORMATION:/not - #e= "cysteine 79 forms a#bond with cysteine 61"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:43..44 (D) OTHER INFORMATION:/not - #e= "cysteine 81 forms a#bond with cysteine 94"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:56..57 (D) OTHER INFORMATION:/not - #e= "cysteine 94 forms a#bond with cysteine 81"e#15: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Leu Asp Asn Val Thr Phe Ser Asp Val Va - #l Ser Ala Thr Glu Pro# 15- Cys Lys Pro Cys Thr Glu Cys Leu Gly Leu Gl - #n Ser Met Ser Ala Pro# 30- Cys Val Glu Ala Asp Asp Ala Val Cys Arg Cy - #s Ala Tyr Gly Tyr Tyr# 45- Gln Asp Glu Glu Thr Gly His Cys Glu Ala# 55- (2) INFORMATION FOR SEQ ID NO: 16:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 58 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..58 (D) OTHER INFORMATION:/not - #e= "Identified neurotrophin binding s - #ite of chicken p75NTR, residues 40-97."- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..2 (D) OTHER INFORMATION:/not - #e= "cysteine 40 forms a#bond with cysteine 56"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:17..18 (D) OTHER INFORMATION:/not - #e= "cysteine 56 forms a#bond with cysteine 40"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:20..21 (D) OTHER INFORMATION:/not - #e= "cysteine 59 forms a#bond with cysteine 72"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:33..34 (D) OTHER INFORMATION:/not - #e= "cysteine 72 forms a#bond with cysteine 59"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:23..24 (D) OTHER INFORMATION:/not - #e= "cysteine 62 forms a#bond with cysteine 80"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:41..42 (D) OTHER INFORMATION:/not - #e= "cysteine 80 forms a#bond with cysteine 62"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:43..44 (D) OTHER INFORMATION:/not - #e= "cysteine 82 forms a#bond with cysteine 95"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:56..57 (D) OTHER INFORMATION:/not - #e= "cysteine 95 forms a#bond with cysteine 82"e#16: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Leu Asp Ser Val Thr Tyr Ser Asp Thr Va - #l Ser Ala Thr Glu Pro# 15- Cys Lys Pro Cys Thr Gln Cys Val Gly Leu Hi - #s Ser Met Ser Ala Pro# 30- Cys Val Glu Ser Asp Asp Ala Val Cys Arg Cy - #s Ala Tyr Gly Tyr Phe# 45- Gln Asp Glu Leu Ser Gly Ser Cys Lys Glu# 55- (2) INFORMATION FOR SEQ ID NO: 17:- (i) SEQUENCE CHARACTERISTICS:#acids (A) LENGTH: 62 amino (B) TYPE: amino acid (C) STRANDEDNESS: single (D) TOPOLOGY: linear- (ii) MOLECULE TYPE: peptide- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..62 (D) OTHER INFORMATION:/not - #e= "Region of P55TNFR#to identified neurotrophin binding site of p75NTR, r - #esidues 41-102"- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:1..2 (D) OTHER INFORMATION:/not - #e= "cysteine 41 forms a#bond with cysteine 56"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:17..18 (D) OTHER INFORMATION:/not - #e= "cysteine 56 forms a#bond with cysteine 41"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:20..21 (D) OTHER INFORMATION:/not - #e= "cysteine 59 forms a#bond with cysteine 74"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:33..34 (D) OTHER INFORMATION:/not - #e= "cysteine 74 forms a#bond with cysteine 59"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:23..24 (D) OTHER INFORMATION:/not - #e= "cysteine 62 forms a#bond with cysteine 82"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:41..42 (D) OTHER INFORMATION:/not - #e= "cysteine 82 forms a#bond with cysteine 62"e- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:44..45 (D) OTHER INFORMATION:/not - #e= "cysteine 84 forms a disulfide bond with - # cysteine 100"- (ix) FEATURE: (A) NAME/KEY: Peptide (B) LOCATION:56..57 (D) OTHER INFORMATION:/not - #e= "cysteine 100 forms a#bond with cysteine 84"e#17: (xi) SEQUENCE DESCRIPTION: SEQ ID NO:- Cys Glu Ser Gly Ser Phe Thr Ala Ser Glu As - #n His Leu Arg His Cys# 15- Leu Ser Cys Ser Lys Cys Arg Lys Glu Met Gl - #y Gln Val Glu Ile Ser# 30- Ser Cys Thr Val Asp Arg Asp Thr Val Cys Gl - #y Cys Arg Lys Asn Gln# 45- Tyr Arg His Tyr Trp Ser Glu Asn Leu Phe Gl - #n Cys Phe Asn# 60__________________________________________________________________________

Molecular modelling of neurotrophin-receptor binding

Information

Patent Number

Date Filed

Date Issued

Inventors

Original Assignees

Examiners

Agents

CPC

US Classifications

Field of Search

US

International Classifications

Abstract

Description

Claims

Priority Claims (1)

US Referenced Citations (1)

Foreign Referenced Citations (1)

Non-Patent Literature Citations (39)