CA2190296A1 - Factors and methods for reducing biological activity of a multimeric protein and methods of screening therefor - Google Patents

Abstract

Factors and methods for disrupting or inhibiting the association of protomers of a multimeric protein are described. Such inhibition reduces the biological activity of the target protein, which is desirable in the case of certain physiological disorders.
Methods of screening candidate factors that mimic the interface between protomers of a multimeric protein, so as to identify a factor that disrupts or inhibits association of the protomers, are also described. In particular, peptides that inhibit the association of protomers of a growth factor having a cysteine knot, such as a neurotrophin, aredescribed.

Description

FACTORS AND METHODS FOR REDUCING BIOLOGICAL ACTIVITY OF A
MULTIMERIC PROTEIN AND METHODS OF SCREENING THEREFOR

This invention relates to the fieid of protein structure-function relationships.More specifically, the invention relates to perturbing a protein's native structure so as to reduce its biological activity. Still more specifically, the invention relates to methods of disrupting multimerization of a protein so as to reduce its biological activity. In 10 particular, the invention relates to methods of disrupting association of the protomers of a protein having a cysteine knot, such as a neurotrophin, so as to reduce its biological activity.

BACKGROUND OF THE INVENTION
Multimeric proteins are those composed of two or more polypeptide subunits, or protomers. A dimeric protein has two protomers, a tetramer has four protomers, and so on. In a particular multimeric protein, the component protomers associate in a specific way to produce the protein's native quaternary structure. For a dimer, such assembly or association is called dimerization. The protomers of a given multimeric protein may 20 be identical to ("homo-") or different from ("hetero-") each other. For example, the tetrameric protein hemoglobin has two identical a-subunits and two identical 3-subunits. Protomers may differ in function within the multimeric protein. For example, the heterodimeric protein cholera toxin has a first subunit that permits penetration of the plasma membrane by the toxin and a second, non-identical subunit that catalyzes 25 the covalent modification of a cytosolic G-protein.

It is well known in biochemistry that the activity of a multimeric protein oftendepends on association of its component subunits. For example, the individual protomers of a particular enzyme may display no catalytic activity in isolation, but only be able to function when in a multimeric state. Similarly, various signal transduction pathways require ligand-induced self-association of cell surface receptors such as EGF-R, PDGF-R and the neurotrophin receptors for signalling to occur. Various ligands involved in signal transduction are themselves known to exist in mu~imeric 5 form.

SUMMARY OF THE INVENTION
In the case of some physiological disorders, it may be desirable to disrupt the multimeric association of a particular protein so as to reduce or even eliminate its 10 biological activity and thereby produce a therapeutic effect. The present invention provides a method of reducing the biological activity of a multimeric protein having at least two protomers by perturbing the association of the protomers, i.e., disrupting multimer integrity. The method includes the step of preparing a factor that mimics a portion of an interface between the protomers. In this context, mimicking should be 15 understood in a functional sense, wherein the factor is characterized by being able to perturb association of the protomers. A subsequent step of the method is mixing the factor with the protein. The factor may be a peptide or a peptide derivative. The mixing step may include administering the factor to a human or an animal so that the factor interacts with the multimeric protein in situ, providing a therapeutic effect. The 20 multimeric protein may be an enzyme. It may be involved in signal transduction; for example, the multimer may be a ligand or a receptor in a signal transduction pathway.
In a preferred embodiment, the multimeric protein is a member of the cysteine knot family of growth factors, which family includes NGF, TGF~2 and PDGF. In a preferred embodiment, the multimeric protein is a neurotrophin.

The invention further provides a sensitive, rapid and convenient method of screening for and identifying factors that can disrupt the association of protomers of a multimeric protein. In a preferred embodiment, the multimeric protein is a member of the cysteine knot family of growth factors or a neurotrophin. The method includes the 2 1 902~6 steps of preparing a factor that mimics a portion of an interface between the protomers, mixing the factor with the protein, and separating different species of the protein, wherein the species are distinguished from each other by having different numbers of protomers. A factor that reduces the amount of a multimeric species of the protein in 5 favor of species having fewer protomers can be identified. Such a factor may be useful in reducing the biological activity of the multimeric protein, i.e., be an antagonist. The method may include, prior to the separation step, the step of subjecting the mixture to a cross-linking agent that covalently joins protomers within a multimeric species. The separation step may include any technique for separating different protein species that 10 would be known to a person skilled in the art. These techniques include gel filtration, HPLC, isoelectric focussing and gel electrophoresis. Electrophoresis may be non-denaturing, or in the case where a chemical cross-linking step has taken place, denaturing.

The invention further provides a factor that can disrupt the association of protomers of a multimeric protein, as described above.

The invention additionally provides use of a factor according to the invention to treat a human being or an animal. A kit for a screening system according to the 20 invention could be assembled. The screening system of the invention could also be automated.

Factors and methods according to the invention can thus provide a solution to problems presented in certain physiological conditions, whère the reduction or 25 inhibition of the activity of a multimeric protein would be advantageous and desired.

DETAILED DESCRIPTION OF THE FIGURES
For a better understanding of the present invention and to show more clearly how it may be carried into effect, reference is made by way of example to the ,~ .

accompanying figures, which illustrate aspects and features according to preferred embodiments of the present invention. The legends below clarify aspects, including materials and methods, of working examples of the practice of the invention.

Figure 1 Ball and stick diagram of a single NGF protomer viewed from a face-on orientation (A) and a 90~ rotation (B), where the right edge of the protomer in B
represents the dimer interface5. Amino acid residues 13-111 are illustrated as hollow spheres (location of a-carbon) with residues 58~8 and 108-110 highlighted as larger filled balls. Stick diagram (C) of residues 58-68 and 108-110 in isolation from the crystal structure, showing the location of non-hydrogen atoms and the four sulfur atoms (large spheres) participating in two of three disulfide bonds in NGF (formed from Cys58-Cys108 and Cys68-Cys110). Peptide R11 was designed from the residues 15 indicated in C and comprises a linear sequence NH2-Cys-Gly-Ser-Glu-Val-Pro-Asn-Ser-Ala-Arg-Cys-Cys-Val-Cys-OH (R11 (linear) or R11 (I)) that is constrained by a single disulfide bond in R11 (monocyclic) (R11 (m)) or two disulfide bonds (R11), shown in (D).
The structures of BR11 and SR11 used in some assay systems are also indicated. All linear peptides and peptide intermediates in this study were synthesized with an ABI
20 model 420 automated peptide synthesizer using standard 9-fluorenylmethoxycarbonyl chemistry and solid state peptide synthesis methods'8. Cysteine residues were introduced as either trityl- or acetamidomethyl-protected species. (The latter is indicated in the figure by "ACM".) The first intrachain disulfide bond formed between deprotected Cys residues was accomplished by dissolving the peptide in 0.1 M
25 ammonium bicarbonate at a concentration of 0.1 mg/ml and stirring the solution while exposed to air. At various times, the solution was sampled and the reaction products analyzed by HPLC separation'9. The HPLC separation was performed on a C,8 reverse phase column (5 ,u particle, 300 A pore; Vydac) using a 1 %/min linear gradient of 0.1 %
trifluoroacetic acid in H2O to 0.1 %trifluoroacetic acid in acetonitrile. The oxidized 2 1 ~0296 product containing the disulfide (which eluted at a lower retention time than did the reduced starting material) was purified by HPLC. The second disulfide bond was formed between the two acetamidomethyl-protected Cys residues by slowly introducing 0.1 mmol of the peptide R1 1(monocyclic) dissolved in 1.5 ml of methanol int~ 2.5 ml of 1.0 M 12 in methanol. The product of the 12-oxidized acetamidomethyl-protected peptide also eluted earlier in the HPLC gradient and was purified using the method described above. The structure of all peptides was confirmed by amino acid analysis and mass spectroscopy.

Figure 2 The ability of R11 to inhibit the activity of NGF in vifro using a biological assay system. R11 inhibited neurite extension with an apparent IC50 of 10 ,uM, displaying a shallow inhibition curve over a wide concentration range. Neither of the less constrained intermediates R11(1) or R11 (m) were effective in blocking NGF-dependent growth when tested in concentrations up to 250 I~M. Dissociated cells enriched for sensory neurons were prepared from embryonic day 8 (ED8) chick dorsal root ganglion (DRG) as described'2. Neurons were seeded into wells of Terasaki plates treated sequentially with poly-D-lysine and laminin at a density of 800-1000 cells/well in "synthetic" Dulbecco's modified Eagle medium20 containing 1% fetal calf serum and NGF at 10 pM. The cells were incubated with the additives indicated at 37 ~C in a 5%
C02 atmosphere. At 18-20 hours, the cells were fixed in 4% formaldehyde in phosphate-buffered saline and scored for neurite growth. The cells on the entire lower horizontal surface of the well were counted using an inverted microscope fitted with phase contrast optics. A neurite was scored if its caliber from origin to terminal was constant and its length was equal to or greater than 1.5 cell body diameters. Neurite growth was corrected for background (no NGF) growth in the presence and absence of peptide. Survival of neurons at 24 hours was not influenced in the presence of 500 ,uM
R11.

21 ~02~6 Figure 3 The ability of R11 (linear), R11 (monocyclic) and R11 to influence binding of NGF
to trkA and p75 was determined by examining the effects of these peptides on chemical cross-linking of '251-NGF. The trkA and p75 receptors were cross-linked and identified 5 by immunoprecipitation (A). At peptide concentrations of 250,uM, R11 was most effective at preventing the covalent incorporation of '251-NGF into either receptor, followed by R11 (monocyclic) and R11 (linear) (B). The same concentration of R11 was also effective in preventing NGF-dependent phosphorylation of trkA (C). Receptorcross-linking studies were performed on PC12 cells2' which were maintained in RPMI
1640 supplemented with 10% fetal calf serum. Cells were harvested by incubation and trituration in Ca2' and Mg2+ free Gey's balanced salt solution, washed and suspended in 10 mM HEPES buffer (pH 7.35) containing 125 mM NaCI, 4.8 mM KCI, 1.3 mM CaCI2, 1.2 mM MgSO4,1.2 mM KH2PO4, 1 g/l sucrose and 1 g/l bovine serum albumin (HEPES
Krebs Ringer buffer, HKR buffer) at 106 cells/ml. Subsequent procedures were carried 15 out at 4 ~C unless otherwise noted. For receptor cross-linking, '251-NGF was prepared by the method of Sutter et al.'2 from NGF obtained from Cedarlane. The radioiodinated NGF obtained had a specific activity of 60-100 cpm/pg, was stored at 4 ~C and used within one week of preparation. PC12 cells in HKR buffer were incubated with 0.1 nM
'251-NGF and the indicated peptides (one ml total volume) for 2 hr at 4 ~C. Receptor 20 proteins were cross-linked with either 0.2 mM bis(sulfosuccinimidyl)suberate (BS3) or a combination of 2.5 mM 1-ethyl-3(3-diethylaminopropyl)carbodiimide (EDC) and 1.0 mM
N-hydroxysulfosuccinimide (S-NHS) as indicated. All cross-linking agents (from Pierce) were dissolved in water and added in 20 ,ul aliquots and the reaction allowed toproceed for 30 min at 25 ~C. Upon completion of the cross-linking reaction, cells were 25 washed three times in HKR buffer and the pellets dissolved in sodium dodecyl sulfate (SDS) reducing sample buffer for electrophoresis. For positive receptor identification, samples of 107 cells in 1 ml HKR buffer were cross-linked under the conditions described above, washed in HKR buffer and solubilized in non-denaturing Iysis buffer (1% NP40, 10% glycerol, 1 mM phenylmethylsulfonylfluoride, 10,ug/ml leupeptin and 2 1 902~6 0.5 mM o-vanadate in Tris-buffered saline)'5. The solutions were clarified by centrifugation and immunoprecipitated with either mAb 192 (monoclonal antibody for p75; Cedarlane) or rabbit anti-trkA, which had been raised against a COOH-terminal peptide of trkA. The '251-NGF cross-linked receptor-antibody complexes were isolated using 50 ,ul of a 50% solution of rabbit anti-mouse agarose (Sigma; for mAb 192) or protein A-Sepharose (Pharmacia; for anti-trk) equilibrated in Iysis buffer.
Immunoprecipitates were washed three times in Iysis buffer and the pellets denatured in SDS reducing sample buffer. Radiolabelled receptor preparations were electrophoresed using a discontinuous gradient gel where the separating gel gradient varied from 4.0% acrylamide/18% urea to 10% acrylamide/50% urea. Gels were subsequently fixed, dried and exposed to X-Omat XAR film (Kodak) for autoradiography. The ability of R11 to influence trkA phosphorylation was determined by incubating 107 PC12 cells (prepared as described above) in one ml HKR buffer with 50 ng NGF and the indicated concentrations of peptides for 15 min at 37 ~C. The cells were isolated by centrifugation, solubilized in Iysis buffer and the trkA receptor isolated as described above. The trkA immunoprecipitate was dissociated in SDS reducing sample buffer and electrophoresed using a 6% polyacrylamide gel. The gel was then transferred to PVDF membrane (BioRad), probed with horseradish peroxidase-conjugated anti-phosphotyrosine (RC20; Transduction) and developed using ECL
(Amersham). All procedures were carried out according to the manufacturers' instructions.

Figure 4 R11 inhibited the covalent cross-linking of NGF protomers in a cell-free system.Radiolabelled '251-NGF (0.5 nM) was incubated in the presence (R11 +) or absence(R11-) of 100 ,uM R11 in HKR buffer at 0 ~C for 2 hr in the absence of cells. Cross-linking was performed as described in the legend to Figure 3 using BS3 (BS3+).
Proteins were electrophoresed using SDS-polyacrylamide gel electrophoresis (12%
linear gel) and autoradiographed as described. R11 inhibited the cross-linking of NGF

protomers as evident by the loss of NGF dimer (at 26 kDa). In the presence of R11, NGF protomer cross-linking to BSA (66 kDa) increased dramatically.

Figure 5 The calculated most stable conformation of molecule Vl (CAMAMCCVC).

Figure 6 The calculated most stable conformation of molecule XV (CVCCAMMAC).

Figure 7 A likely binding interaction of molecule V (CAAAAACCVC) with mouse NGF
monomer.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS
Practice of the invention will be described in detail below with reference to a particular example, namely disruption of the dimerization of a neurotophin, nerve growth factor (NGF). However, a person skilled in the art would be able to employ the principles of the invention in regard to other growth factors having a cysteine knot, to other neurotrophins and to multimeric proteins of various sizes, subunit compositions and functions, some examples of which were given above. Although the invention is described in detail below with reference to certain preferred embodiments thereof, these embodiments are offered to illustrate but not to limit the invention. It is possible to make other embodiments that employ the principles of the invention and fall within its spirit and scope.

A family of structurally and functionally related neurotrophic factors exist which are collectively known as neurotrophins. The family of neurotrophins include the nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT-3), neurotrophin4 (NT4) and neurotrophin-5 (NT-5).

The neurotrophins exhibit similar structural conformations, including three surface ~-hairpin loops, a ~-strand, an internal reverse turn region, and NH2 and COOH
termini. With respect to sequence similarities, the neurotrophins share approximately 50% amino acid identity. The neurotrophins are also functionally similar in that they 5 each exhibit low affinity binding to a receptor known as the "p75 nerve growth factor receptor" or p75NGFR. Each neurotrophin also exhibits binding to a receptor of the tyrosine kinase (trk) family which is of higher affinity than the binding to the p75 receptor.

Neurotrophin-mediated biological activities include, for example, neurotrophin binding to the p75NGFR receptor, neurotrophin binding to one of the trk receptors, neuron survival, neuron differentiation including neuron process formation and neurite outgrowth, and biochemical changes such as enzyme induction.

Nerve growth factor (NGF) acts as a survival and differentiation factor for a variety of target neurons in both peripheral and central nervous systems. It is known to promote growth (neuritic sprouting) in target neurons. NGF is also known to bind to the cell surface receptor proteins p75NGFR and trkA. Binding of NGF to trkA induces trkA
autophosphorylation. For neurological disorders where excessive neuritic sprouting is implicated in pathogenesis (Alzheimer's disease', epilepsy2, pathological pain syndromes34), antagonists to neurotrophins such as NGF may have therapeutic utility.

Mature NGF is composed of two identical subunits of length 118 residues.
Variable domains within the NGF family of neurotrophins include the NGF NH2- andCOOH-terminal residues (1-9 and 111-118, respectively) and certain internal residues that are revealed by the crystal structure of NGF5 to form four loop structures. These loops are three 13 sheet structures, residues 25-35 (Loop 1, L1), 40-50 (Loop 2, L2) and 90-100 (Loop 4, L4); and a twisted loop (Loop 3, L3) residues 62-68.

2 1 902~6 It has been suggested previously that the variable regions mediate the biological effects of the neurotrophins, via specific trk family receptors and the common neurotrophin receptor p75. This suggestion led to the employment of site-directed mutagenesis and recombinant chimeric protein techniques to demonstrate that specific 5 residues within L2, L4 and the NH2 and COOH termini are required for for trk activation~8 and that domains of L1 and L4 are involved in p75 binding9. Monoclonal antibodies against antigenic determinants encompassing the NH2 and COOH termini and the L3 region of NGF implicate these domains in trkA receptor signalling'~.

Residues 58, 67, 68, 108, 109 and 110 are included in those residues that are conserved among the neurotrophins. The specific requirements of these residues with respect to NGF binding to receptors have not been examined using recombinant protein techniques, as they are required for protein structural integrity. The participation of NGF residues 60-67 in mediating interaction of NGF with either p75 or trkA has been excluded by deletion mutagenesis studies8.

Two reports regarding the crystal structure of NGF5 " implicate specific residues within the conserved regions of the molecule that participate in interprotomericinteractions responsible for dimer integrity of this neurotrophin at physiological 2 0 concentrations. The present inventors have noted that these include certain residues within the sequences from residue Cys58 to residue Cys68 and from residue Cys108 to residue Cys110; this information was exploited as described below. Intramolecular interactions within the NGF monomer have been demonstrated for two conserved amino acids (67 and 109) and three variable residues (59, 61 and 64). Residues Val109 and Cys110 of NGF have been demonstrated to have interprotomeric interactions at the dimer interface5 ".

Using the atomic coordinates of the 2.3 ~ crystal structure of NGF5, the inventors have designed and synthesized a conformationally constrained peptide, designated R11, that incorporates amino acid residues from two domains that appear to be local with respect to each other in native NGF. Linear peptide R11 is a 14-mer having the following primary sequence: NH2-Cys-Gly-Ser-Glu-Val-Pro-Asn-Ser-Ala-Arg-Cys-Cys-Val-Cys-OH. The first eleven residues are identical to residues 58 to 68 of ~NGF, but in the reverse order. (That is, residue 68 of NGF is Cys, residue 67 Gly, residue 66 Ser, and so on.) The COOH-terminal three residues of R11 correspond to residues Cys1 08-Val1 09-Cys110 of NGF.

In native NGF, residues 58-68 and 108-110 include residues that are a portion of the dimer interface. The conformation of amino acid residues 58-68 (L3) of native NGF was recognized by the inventors as a target peptide that could be synthesized in a configuration mimicking the conformation of this domain in NGF. In the NGF monomer, Cys58 and Cys68 take part in a cysteine knot motif by forming covalent (disulfide) bonds with Cys108 and Cys110, respectively. This occurs both within the NGF
monomer, as shown in Figures 1A and 1 B, and as an isolated structure, as shown in Figure 1 C. As shown in Figure 1 C, the a-amine of Cys58 and the a-amine of Cys108 are in fairly close proximity to each other (4.6 A).

As illustrated in Figure 1 D, peptide R11 was cyclized by disulfide bond formation between the side chain sulfhydryls of its NH2-terminal and COOH-terminal Cys residues, and between the sulfhydryls of the two adjacent internal Cys residues at positions eleven and twelve. These two disulfide bonds mimic the two disulfide bonds of the NGF cysteine knot described above. In addition, R11 is constrained by theadditional peptide bond between residues eleven and twelve, which has no counterpart in NGF. R11 is the subject of U.S. Patent Application Serial No. 08/241,462, filed May 11, 1994, and International Patent Application No. PCT/CA95/00603, filed October 25, 1 995.

21 ~02q6 A molecular dynamics study of R11 indicates that the bridge formed by COOH-terminal residues Cys-Val-Cys would yield a conformation in which the interatomic distances between the R11 NH2- and COOH-terminal Cys residues are very similar to the distances observed between residues Cys68 and Cys110 of the NGF crystal 5 structure. The study also indicates that the first eleven residues of R11 would have considerable mobility in solution, consistent with the predicted mobility of residues Cys58-Cys68 in NGF5 ". The four Cys residues of R11 that participate in the two disulfide linkages are much more constrained, however, and molecular dynamics suggests that they would likely exist in a conformation with limited mobility similar to 10 that of the cysteine knot motif in NGF.

We have investigated the ability of two analogues of R11 (SR11 and BR11:
Figure 1 D) with differing orientation of the two peptide domains in the NGF dimer disruption assay. The analogue SR11, where the Cys58 to Cys68 domain is inverted15 with respect to the same domain of R11, was equally effective in reducing crosslinking of NGF protomers as R11. The analogue BR11, where the Cys-Val-Cys residues are located at the NH2 end of the peptide, was not effective in influencing the crosslinking of NGF protomers and did not inhibit NGF-mediated neurite growth at concentrations of 500 ~M. Taken together, this data would suggest specific requirements of the 20 orientation of the Cys residues in R11 with respect to the interaction with NGF, while the orientation of the residues within the Cys58 to Cys68 domain is less critical.

Peptide R11 effectively inhibited the neurite growth of embryonic day 8 (ED8) chick dorsal root ganglion (DRG) neurons in vitro12 with an IC50 of 10 ,uM, as shown in 25 Figure 2. The dose response profile for the peptide displayed a shallow inhibition curve over a wide concentration range. Such a profile is not typical of a competitive inhibitor. That is, a different dose response profile would be expected if R11 were simply competing with NGF for binding to a receptor. Neither of the less constrained synthetic intermediates of R11, i.e., R11 (linear) or R11 (monocyclic), was as effective in blocking NGF-dependent neurite growth at concentrations up to 250 ,uM. This indicates that the R11 peptide with its two disulfide bridges has differences in its conformation that lead to differences in its function. R11 was also shown to inhibit both seizure and mossy fibre sprouting in an animal model of epilepsy whereby repeated 5 subconvulsive electrical stimulation of the forebrain leads to a progressive and permanent amplification of seizure activity (kindling).22 Indeed we have demonstrated that R11 is an effective antagonist of BDNF and NT-3 in vitro.22 The ability of the peptides under study to interfere with binding of NGF to NGF
10 receptors was evaluated by analysis of inhibition of the chemical cross-linking of '251-NGF to the common neurotrophin receptor p75'3 and the specific NGF receptor trkA'4.
Immunoprecipitation and polyacrylamide gel electrophoresis of cross-linked receptors allowed identification of both receptors, as shown in Figure 3A. At a concentration of 250 IJM, R11 effectively prevented the covalent attachment of '251-NGF to each receptor 15 (Figure 3B); this concentration is similar to that required to block virtually all NGF-mediated neurite growth. At concentrations of 250 ,uM, neither R11(1) nor R11 (m) were able to inhibit greater than 50% of the observed '251-labelling.

Rapid NGF-dependent phosphorylation of the trkA receptor'5 is mediated by 20 receptor homodimerization-induced autophosphorylation'6. Phosphorylation of the trkA
receptor could be detected in NGF-treated PC12 cells within 15 minutes (Figure 3C) in the absence of R11, but was blocked in the presence of 250 ,uM R11. A possible explanation is that ligand-induced receptor clustering did not occur. In view of the '251-labelling results discussed above, a reduction in or absence of receptor clustering 25 appears to result directly from a reduction in NGF (ligand) binding to its receptor.

Previously, L3 of NGF (residues 62-68) and the NGF NH2- and COOH-terminal domains were implicated in trk~ binding and signalling8'0. Whereas R11 could in principle compete directly with L3 for binding to trkA, thereby producing the results 2 1 qo2s6 shown in Figure 3B, in contrast L3 and the NH2- and COOH-terminal domains of NGFhave not been implicated in p75 binding'7, and such a mechanism cannot thereforeexplain the inhibition by R11 of the interaction of NGF with p75, also shown in Figure 3B. It should be noted that the same results for p75 were achieved with tw~ different 5 cross-linking reagents, bis(sulfosuccinimidyl)suberate (BS3) and a combination of 1-ethyl-3(3-diethylaminopropyl)carbodiimide (EDC) and N-hydroxysulfosuccinimide (S-NHS).

The ability of the R11 peptide to influence protomer interactions of NGF was 10 examined, and results are shown in Figure 4. In the absence of R11, NGF incubated with the protein-protein cross-linking reagent BS3 and subsequently electrophoresed on a polyacrylamide gel under denaturing conditions exhibited a novel band corresponding to two NGF monomers. That is, under denaturing conditions, NGF
normally dissociates and electrophoreses as a monomer (Figure 4, lane 1). In contrast, 15 native NGF dimers that had been covalently cross-linked by BS3 were unable todissociate and were resolved from free monomer as a more slowly migrating band on the gel (Figure 4, lane 2). Such cross-linking of NGF protomers was reduced almost to the point of elimination when NGF was incubated in the presence of 100 ,uM R11 (a concentration that blocked 70% of NGF-mediated neurite growth) and then subjected to 20 BS3, as shown in Figure 4, lane 3. This result can be explained by a mechanism in which R11 perturbs or disrupts the association of NGF protomers to form the native, functional NGF multimer. This mechanism also explains the biochemical and biological effects of R11 discussed above.

The above-described reactions were performed in buffer containing 1 mg/ml bovine serum albumin (BSA), which should sponge up nonspecific (and presumably low affinity) protein-protein interactions. It can be noted that lane 3 of Figure 4 shows yet another, still more slowly migrating '251-labelled band, whose size is compatible with the sum of the molecular weights of an NGF protomer and BSA. This supports the proposed mechanism wherein R11 disrupts NGF dimers, thereby resulting in failure of BS3 to chemically cross-link NGF protomers, and excludes a nonspecific effect of R11 on the efficiency of the BS3 cross-linking reaction. The observations are consistent with an R1 1-induced conformational alteration in NGF protomers that would be 5 inappropriate for receptor recognition (Figure 3) but appropriate for chemically-mediated cross-linking to other proteins (e.g., BSA, Figure 4, lane 3). In the absence of R11, NGF protomer conformation would favor interprotomer cross-linking (Figure 4, lane 2).

The experimentally observed structure-activity relationships of R11 analogues suggested to the inventors that: i) the Cys-Val-Cys domain may be responsible for biological activity of R11; and ii) the COOH-terminal attachment of the Cys-Val-Cys domain may cause it to adopt bioactive conformation. Accordingly, the inventors analysed low-energy conformations of R11 analogues with both NH2- and COOH-15 terminal attachment of the Cys-Val-Cys domain, with the aim to determine the geometric features inherent in active molecules.

The low-energy conformations of the following 18 R11 analogues were obtained by Variable Basis Monte Carlo simulated annealing computations combined with a ring 20 closure algorithm.

I-IX: C(A)nCCV~ n = 1-9 X-XVIII: CVCC(A)nC n = 1-9 .

The potential energy function of the molecules included terms for angle bending,torsional distortion, hydrogen bonding, van der Waals interactions and electrostatic interactions as defined in the united atom CHARMM force field.

The main result of the molecular modelling is that, beginning with n=5 in the molecules having the Cys-Val-Cys domain at the COOH terminus (i.e., in molecules V-IX), the Cys-Val-Cys domain adopts the same conformation, namely a ~-strand conformation. None of molecules X-XVIII with the NH2-terminal attachment of the Cys-Val-Cys domain can adopt this particular conformation. The reasons behind the geometric features demonstrated by the molecules in this molecular simulation are apparent in Figures 5 and 6. In the molecules having the Cys-Val-Cys domain at the COOH terminus, two disulfide bonds result in formation of a cavity inside the molecule, immediately adjacent to the Cys-Val-Cys domain, and the terminal charged amino group occupies this cavity if the molecule is sufficiently flexible (i.e., if n is sufficiently large). This particular location of the amino group is stabilized by its hydrogen bonding with 4 backbone carbonyl groups which turn inside the cavity. This orientation of the carbonyl groups surrounding the cavity results in both NH and CO groups of the Val residue being exposed in the same direction, which makes the conformation of the Cys-Val-Cys domain consistent with that of a ,B-strand (Figure 5). On the other hand, the NH2-terminal attachment of the Cys-Val-Cys domain results in a different H-bond pattern producing a different conformation of this domain (Figure 6).

Figure 7 represents a likely explanation for how the geometric features of molecules V-IX could account for biological activity. According to this model, the ~B-strand motif of such an R11 analogue forms a hydrogen bond to a parallel ,B-strand of NGF at Val111. In addition to two H-bonds, there are two hydrophobic interactions of the Val residue of the R11 analogue with Val111 and Val109 of NGF, which stabilizes the R1 1-NGF complex. This complex alters the dimer interaction of NGF monomers,which decreases biological activity of NGF.

To summarize the conclusions resulting from practice of the invention with respect to NGF, perturbation of NGF dimer structure by R11 is sufficient explanation for lack of chemical cross-linking of '251-NGF to either p75 or trkA receptors, inhibition of NGF-induced stimulation of neurite growth and absence of NGF-mediated trkA
5 phosphorylation. The inventors believe that the Cys-Val-Cys sequence resembling NGF residues 108-110 at the dimer play a key role in such perturbation.

The identification of a peptide antagonist of NGF that perturbs the dimeric structure of this neurotrophin has important implications for therapeutic strategies 10 involving antagonists of other members of the dimeric neurotrophin family, as well as other multimeric proteins where biological activity depends on association of protomers.
The crystal structure atomic coordinates of the protomers in the native protein can be used as reference in the design of an antagonist species that interferes with protomer association. Rather than the antagonist competing directly with the multimeric protein 15 for binding to a substrate, a receptor or the like, the antagonist would perturb protomer association and multimer assembly, presumably in a competitive fashion. Since a functional multimer would be unable to assemble, its biological activity would be reduced or even eliminated. Such approaches can be used to develop antagonists of other members of the cysteine knot family of growth factors notwithstanding the fact 20 that, unlike NGF, both TGF~2 and PDGF contain interprotomeric disulfide bonds23.

REFERENCES

1. McKee, A.C., Kosik, K.S. and Kowall, N.W. Ann. Neurol. 30, 156 (1991).
2. Ben Ari, Y. and Represa, A. TINS 13, 312-318 (1990).

3. Woolf, C.J. and Doubell, T.A. Current Opinion in Neurobiol. 4, 525-534 (1994).

4. Leven, G.R. and Mendel, L.M. TINS 16, 353-359 (1993).

5. McDonald, N.Q, Lapatto, R., Murray-Rust, J., Gunning, J., Wlodawer, A. and Blundell, T.L. Nature 354, 411-414 (1991).

6. Kahle, P., Burton, L.E., Schmelzer, C.H. and Hertel, C. J. Biol. Chem. 267, 22707-22710 (1992).

7. Kullander, K. and Ebendal, T. J. Neurosci. Res. 39195-210 (1994).

8. Drinkwater, C.C., Barker, P.A., Suter, U. and Shooter, E.M. J. Biol. Chem. 268, 23202-23207 (1993).

9. Ibànez, C.F., llag, L.L., Murray-Rust, J. and Persson, H. EMBO J. 12, 2281- 2293 ( 1993).

10. Nanduri, J., Vroegop, S.M., Buxser, S.E. and Neet, K.E. J. Neurosci. Res. 37, 433-444 (1994) 11. Holland, D.R., Cousene, L.S., Meng, W. and Matthews, B.W. J. Mol. Biol. 239, 385-400 (1994).

12. Sutter, A., Riopelle, R.J., Harris-Warrick, R.M. and Shooter, E.M. J. Biol. Chem.
254, 5972-5982 (1979).

13. Chao, M. J. Neurobiol. 25, 1373-1385 (1994).

14. Barbacid, M. J. Neurobiol. 25, 1386-1403 (1994).

15. Kaplan, D.R., Martin-Zanca, D. and Parada, L.F. Nature 350, 158-160 (1991).16. Jing, S., Tapley, P. and Barbacid, M. Neuron 9, 1067-1079 (1992).
17. Ibànez, C.F. J. Neurobiol. 25, 1349-1361 (1994).
18. Atherton, E. and Sheppard, R.C. Solid Phase Peptide Synthesis IRL Press, Ox~ord (1989).

~ 21 ~0296 -19. Lee, K.K., Black, J.A. and Hodges, R.S. High Pefformance Liquid Chromatograp~y of Peptides and Proteins CRC Press 389-398 (1991).
20. Bottenstein, J.E. and Sato, G.H. Proc. Nafl. Acad. Sci. USA 76, 514-517 (1979).
21. Greene, L.A. and Tischler, A.S. Proc. Natl. Acad. Sci. USA 73, 2424 2432 (1976).
22. Rashid, K., Van der Zee, C.E.E.M., Ross, G.M., Chapman, C.A., Stanisz, J., Riopelle, R.J., Racine, R.J. and Fahnestock, M. Proc. Natl. Acad. Sci. USA 92, 9495-9499 (1995) 23. McDonald, N.Q. and Hendrickson, W.A. Cell 73, 421-424 (1993).

Claims (12)

1. A method of reducing the biological activity of a multimeric protein having at least two protomers, comprising the steps of:
preparing a factor that mimics a portion of an interface between the protomers, wherein the factor perturbs association of the protomers, and mixing the factor with the protein.

2. A method of screening for a factor that disrupts the association of protomers of a multimeric protein, comprising the steps of:
preparing a candidate factor that mimics a portion of an interface between the protomers, mixing the candidate factor with the protein, separating different species of the protein, wherein the species are distinguished from each other by having different numbers of protomers, and identifying a factor that reduces the amount of a multimeric species of the protein in favor of a species having fewer protomers.

3. The method of claim 1, wherein the multimeric protein comprises a cysteineknot.

4. The method of claim 1, wherein the multimeric protein is a neurotrophin.

5. The method of claim 4, wherein the neurotrophin is nerve growth factor (NGF).

6. The method of claim 3 or 4, wherein the factor comprises amino acid residues resembling at least a portion of a cysteine knot.

7. The method of claim 6, wherein the factor comprises a Cys-Val-Cys sequence resembling NGF residues 108-110.

8. The method of claim 2, wherein the multimeric protein comprises a cysteine knot.

9. The method of claim 2, wherein the multimeric protein is a neurotrophin.

10. The method of claim 9, wherein the neurotrophin is nerve growth factor (NGF).

11. The method of claim 8 or 9, wherein the factor comprises amino acid residuesresembling at least a portion of a cysteine knot.

12. The method of claim 9, wherein the factor comprises a Cys-Val-Cys sequence resembling NGF residues 108-110.