J Biol Chem, Vol. 275, Issue 14, 9946-9956, April 7, 2000

Genuine Monovalent Ligands of TrkA Nerve Growth Factor Receptors Reveal a Novel Pharmacological Mechanism of Action^*

Sergei Maliartchouk^§, Thomas Debeir^¶, Natalia Beglova, A. Claudio Cuello, Kalle Gehring^**, and H. Uri Saragovi^§§

From the Departments of  Pharmacology and Therapeutics,  Biochemistry, and  Oncology and the Cancer Center, McGill University, Montréal, Québec H3G 1Y6, Canada

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Developing small molecule agonistic ligands for tyrosine kinase receptors has been difficult, and it is generally thought that such ligands require bivalency. Moreover, multisubunit receptors are difficult to target, because each subunit contributes to ligand affinity, and each subunit may have distinct and sometimes opposing functions. Here, the nerve growth factor receptor subunits p75 and the tyrosine kinase TrkA were studied using artificial ligands that bind specifically to their extracellular domain. Bivalent TrkA ligands afford robust signals. However, genuine monomeric and monovalent TrkA ligands afford partial agonism, activate the tyrosine kinase activity, cause receptor internalization, and induce survival and differentiation in cell lines and primary neurons. Monomeric and monovalent TrkA ligands can synergize with ligands that bind the p75 subunit. However, the p75 ligands used in this study must be bivalent, and monovalent p75 ligands have no effect. These findings will be useful in designing and developing screens of small molecules selective for tyrosine kinase receptors and indicate that strategies for designing agonists of multisubunit receptors require consideration of the role of each subunit. Last, the strategy of using anti-receptor mAbs and small molecule hormone mimics as receptor ligands could be applied to the study of many other heteromeric cell surface receptors.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Nerve growth factor (NGF)¹ is a dimeric hormone composed of two identical protomers. NGF binds to either or both of two receptors termed TrkA and p75. Cells expressing TrkA bind NGF with intermediate affinity (K_d ~10¹⁰ M) (1-3), and cells expressing p75 bind NGF with lower affinity (K_d ~10⁹ M) (4). Co-expression of TrkA and p75 creates high affinity NGF binding sites (K_d ~10¹² M) (3), indicative of physical and functional interactions (5-8).

Agonists that activate TrkA afford protection from apoptotic cell death and neuronal differentiation and axonal growth (9). The p75 receptor mediates apoptosis in some neuronal and nonneuronal cells (reviewed in Refs. 10 and 11)), but it is unclear whether p75-mediated death is constitutive, induced by agonistic p75 ligands, or can be antagonized by other ligands. Culture studies where Trk-specific ligands were mixed with p75-specific ligands have shown synergy and reciprocal regulation of function (6-8, 12).

TrkA is a tyrosine kinase receptor that transduces NGF signals. The dimeric NGF protein induces TrkA dimerization leading to activation of the kinase (13), as expected by analogy with other receptor tyrosine kinases (14). However, dimeric ligands do not always lead to receptor activation (15-17). Hence, the possibility that monomeric ligands could induce conformational changes leading to receptor dimerization or activation remains an attractive hypothesis (18). Since no biological studies have been done with defined and genuine monomeric ligands of TrkA or any other tyrosine kinase receptor, this is one aim of the present study.

Functional synergy between bivalent Trk ligands and bivalent p75 ligands, leading to enhanced Trk activation and cell survival, have been reported (6, 8). However, no functional studies of synergy have been done with defined and genuine monovalent ligands of p75 and TrkA. This is another aim of the present study.

To answer both aims, we used defined monovalent and monomeric ligands that bind to the extracellular domain of TrkA and p75 receptors. Specifically, we asked (i) whether monovalent and monomeric ligands of TrkA can act as partial agonists, (ii) whether monomeric ligands of TrkA can synergize with ligands of p75, and (iii) what the valency requirement is for p75 ligands to synergize with TrkA ligands. Three sets of ligands that bind the extracellular domain of NGF receptors were available. Each ligand was used in its bivalent or monovalent state, alone or in combinations, to probe receptor function in biological and biochemical assays.

Anti-TrkA mAb 5C3 is an agonist ligand of TrkA (19). Anti-p75 mAb MC192 is a p75 ligand that can synergize with TrkA ligands (6). Peptides termed C(92-96) and C(92-97) bind TrkA in vitro and target TrkA-expressing cells in vivo (20-22). When TrkA is engaged by these peptide analogs, binding of the natural ligand NGF is antagonized (21), but a possible intrinsic activity of the peptide analogs upon binding TrkA had not been studied.

Biophysical characterization of C(92-96), described herein, defines it as a genuine monomeric and monovalent TrkA ligand. We report that genuine monovalent TrkA ligands are partial agonists and induce TrkA activation and internalization and cell survival and differentiation. Expectedly, bivalent TrkA ligands afford more robust signals. These data challenge the exclusive notion of ligand bivalency postulated for activation of tyrosine kinase receptors. For p75, only bivalent ligands afford signals that synergize with TrkA-mediated signals. This suggests that TrkA and p75 differ in their requisites for ligand activation. Last, oligomerizing ligands afford the same signals as homodimerizing ligands.

The insight that monovalent small molecule ligands can be partial agonists will be useful for screening and designing pharmacological agents, and the approach described can be adapted to the study of other receptors.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Cell Lines

Rat PC12 cells express low levels of rat TrkA and 40,000-100,000 p75 receptors/cell (TrkA⁺ p75⁺⁺⁺). B104 rat neuroblastomas express ~50,000 p75 receptors/cell but do not express Trks (TrkA p75⁺⁺⁺). The 4-3.6 cells are B104 cells stably transfected with human trkA cDNA and express equal surface levels of p75 and TrkA (TrkA⁺⁺⁺ p75⁺⁺⁺) (23). The 6-24 cells are PC12 cells stably transfected with human trkA cDNA and overexpressing TrkA (TrkA⁺⁺⁺ p75⁺⁺⁺). Cell surface expression of each of the NGF receptors was routinely controlled in all cells by quantitative FACScan assays (Becton Dickinson) (data not shown). These cells do not express detectable mRNA for neurotrophins (data not shown; Ref. 23) and undergo apoptosis when neurotrophins or serum are withdrawn.

Dissociated Neuronal Dorsal Root Ganglia Cultures

Fetal rat dorsal root ganglia (DRG) primary cultures were established essentially as described (24) from Harlan Sprague-Dawley day 17 rat embryos. All ganglia were dissected and dissociated first enzymatically with trypsin and then mechanically. Dissociated cells were cultured (100,000 cells/well) in 96-well plates precoated with collagen and grown for a total of 8 days in Neuro Basal Medium containing N2 supplement (Life Technologies, Inc.), antibiotics, and L-glutamine. These DRG cultures are ~85% TrkA-expressing and are heavily dependent on TrkA signals for survival (25, 26).

Antibody and Fragment Preparation

The activities of anti-human TrkA IgG mAb 5C3 and anti-rat p75 IgG mAb MC192 have been described (6, 19). mAbs 5C3 and MC192 do not cross-block each other's binding. Purified IgGs were digested with papain (Life Technologies, Inc.) to yield monovalent fragments (Fabs). For further purification, first papain was inactivated; second, the Fc fragments were removed in protein G-Sepharose columns (HiTrap; Amersham Pharmacia Biotech); and third, the Fabs containing  light chains were purified to >98% purity in KappaLock-Sepharose columns (Upstate Biotechnology, Lake Placid, NY) and by preparative FPLC with sizing columns (Amersham Pharmacia Biotech). No IgG was detected in Fab preparations. FPLC spectrometry and size exclusion analysis under native conditions did not reveal the presence of aggregates, even at 40 µM Fab concentrations (bioassays use nanomolar concentrations). The conditions used would have detected <0.2% of Fab aggregates. Binding competition assays between 5C3 Fabs and the intact antibody indicated that the affinity of Fabs (K_d 10 nM) is within 5-fold of the intact IgG (K_d 2 nM). The affinities of the MC192 Fabs were not measured directly. However, FACScan assays demonstrated that MC192 Fabs and MC192 IgG (Fig. 1, A and B) and 5C3 Fabs and 5C3 IgG (Fig. 1, C and D) bound their cellular targets in a specific and saturable fashion indistinguishable from each other (Fig. 1).

View larger version (20K): [in this window] [in a new window]   Fig. 1.   Binding profile of mAb versus Fab fragments. 4-3.6 cells expressing equal levels of TrkA and p75 were analyzed by FACScan binding as described under "Materials and Methods." Binding of intact IgG was revealed with fluorescein isothiocyanate-coupled goat anti-mouse IgG and binding of Fabs with fluorescein isothiocyanate-coupled goat anti-mouse Fab. Controls excluded the specific primary. Saturating concentrations are achieved at 50 nM Fabs and at 25 nM IgG. These concentrations have equal number of receptor-binding units because IgGs are bivalent. Decreasing the saturating dose 5-fold results in similar immunostaining patterns for IgG and Fabs (compare A versus B and C versus D), suggesting similar binding properties. Note that FACScan immunostaining conditions (105 cells stained/test, binding primary for 30 min at 4 °C) are not identical to the conditions used for survival assays; hence, saturating concentrations for the latter cannot be extrapolated.

Cyclic NGF Mimics

The NGF mimic C(92-96) is an N-acetylated (N-Ac) cyclic peptide with primary sequence N-Ac-YCTDEKQCY. The NGF mimic C(92-97) is N-Ac-YCTDEKQACY. The C(92-96) and C(92-97) peptides are cyclized by intrachain disulfide bonds (indicated by the underlines) (21). These peptides are structural mimics of the C-D -turn of NGF (27). Linear peptides with the same sequences do not bind TrkA and were prepared as controls by substituting Cys with Met (primary sequence YMTDEKQMY). The linear peptides do not cyclize, and NMR spectroscopy indicated a lack of conformation (data not shown). The C(92-97)_dimer is a tethered covalently linked dimer of C(92-97). High pressure liquid chromatography and mass spectroscopy analysis confirmed the expected retention time and mass for a dimer.

Peptide Synthesis and Characterization

N-Ac peptides were synthesized by Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. Purification, quality control, and characterization of the peptides were done as described (21) and by NMR diffusion studies (this report). More convincingly, the full NMR spectra of N-Ac-C(92-96) were analyzed (27). Assignment of all resonances and distances and resolution of the structure showed the peptide to be monomeric. Therefore, it is extremely unlikely that N-Ac-C(92-96) is a noncovalent dimer. With respect to a possible covalent dimer, mass spectroscopy (API III MS System, Sciex, Thornhill, Ontario) by electrospray ionization quadrupole (data points every 0.1 Da) verified the chemical composition and monomeric state of N-Ac-C(92-96) with 1192.3 ± 0.3 atomic mass units measured, which is the theoretical mass (1192.2) for an oxidized monomer. No trace of a covalent peptide dimer was detected even after prolonged signal averaging, using conditions that would have detected 1% of dimer. Therefore, it is extremely unlikely that N-Ac-C(92-96) is a covalent dimer.

NMR Spectroscopy

NMR samples contained 5 mM N-Ac-C(92-96) in distilled water at pH 5.7 containing 10% (v/v) of D₂O for the deuterium lock. When D₂O was used as a solvent, the peptide was twice lyophilized and redissolved in D₂O. Spectra were acquired at 500-MHz proton frequency on a three-channel Bruker DRX500 spectrometer equipped with pulsed field gradients. Standard experimental protocols were used for the acquisition of NMR spectra and spectral assignments. Isotropic self-diffusion measurements used NMR pulse field gradients at different peptide concentrations (28, 29). Seventeen one-dimensional assays were done at each concentration with gradient strengths from 0.67 to 63.65 G/cm, gradient duration of 3.5 ms, and a diffusion time of 150 ms. Peptide signal decay was measured at nine different frequencies. Data were fit to the equation I = I°exp((G)²(  /3)), where I is the experimentally measured signal intensity attenuated by diffusion,  is the ¹H gyromagnetic ratio,  is gradient duration, G is the gradient strength,  is time between gradient pulses, and  is diffusion coefficient. Results were averaged.

Ligand Concentrations and Valency

Antibodies are defined as "artificial receptor ligands" because they are specific, they bind with high affinity, with saturable and reversible kinetics, and they are bioactive. Responses to a full dose range (from picomolar to high micromolar) were studied previously for some ligands, and responses for the same dose range for all ligands were studied herein (data not shown). Usually, only optimal concentrations of NGF mimics, mAbs, or Fabs that afford trophic signals are shown for clarity. The NGF mimic C(92-96) is monomeric and monovalent. It is water-soluble and does not aggregate even at 18 mM (this paper). The C(92-97)_dimer is a covalent dimer and bivalent analog of the C(92-97) NGF mimic. Intact IgGs are dimeric and bivalent, and Fabs are monomeric and probably monovalent. Where indicated, Fabs were cross-linked with goat anti-mouse Fab antibody (-Fab; Sigma) at a 2:1 ratio of Fab to -Fab. This cross-linking ratio affords optimal dimerization (one -Fab can bind two Fabs). Higher cross-linking of Fabs using Fab/-Fab ratios of 1:1 or 1:5 (each Fab bound by many -Fabs), leading to ligand oligomerization, achieved results comparable with dimerizing ratios of 2:1 Fab/-Fab (data not shown).

Protection from Apoptotic Death

Primary DRG Cultures-- After a total of 8 days of culture with NGF (Prince Labs, Toronto, Canada) or the indicated test or control ligands, cell survival were studied using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay and by microscopic observation.

Cell Lines-- 5,000 cells/well in protein-free medium (PFHM-II; Life Technologies, Inc.) containing 0.2% bovine serum albumin (crystalline fraction V; Sigma) were seeded in 96-well plates (Falcon, Mississauga, Canada). The cultures were untreated or were treated with the indicated test or control ligands. Cell viability was quantitated using the MTT assay after 56-72 h of culture, and apoptotic death was confirmed by analysis of DNA fragmentation patterns. Percentage of protection was standardized from MTT OD readings relative to optimal NGF (1 nM) = 100%. The OD values of untreated cells were subtracted and were <15% for cell lines and <30% for primary cultures. The higher survival of untreated primary cultures is probably due to endogenous production of limiting amounts of growth factors.

Tyrosine Phosphorylation Assays

TrkA tyrosine phosphorylation was assayed after a 15-min treatment of intact cells with the indicated agent(s) and revealed by Western blotting of whole cell lysates as described (6). Anti-phosphotyrosine (-Tyr(P)) mAb 4G10 (Upstate Biotechnology, Inc., Lake Placid, NY) or antiserum -Tyr(P)490 against the Tyr(P)⁴⁹⁰ of TrkA (within the Shc docking site) (30) was used as a primary antibody. Bands in x-ray films were quantified by densitometry, and intensities were standardized relative to 1 nM NGF. Densitometry of four to five independent gels was analyzed statistically by paired Student's t tests with Bonferroni corrections.

TrkA Internalization Measurements

Live 4-3.6 cells were treated as indicated for 45 min at 4 °C in the presence or absence of 0.25% sodium azide. Cells were maintained at 4 °C or shifted to 37 °C for another 20 min to allow ligand-induced receptor internalization (9). Then, cells were washed and immunostained with mAb 5C3 at 4 °C (phosphate-buffered saline, 0.5% bovine serum albumin, 0.1% sodium azide), for analysis of surface TrkA expression by FACScan immunofluorescence as described (9). In each assay, 5,000 cells were acquired, and the mean channel fluorescence of bell-shaped histograms was analyzed (LYSIS II, Becton Dickinson, CA). Percentage inhibition of mAb 5C3 binding was calculated as a change in mean channel fluorescence with respect to control untreated cells. Rapid loss of surface TrkA is interpreted as receptor internalization, which is delayed or inhibited by low temperatures or sodium azide (9).

Chemical Cross-linking

Live 4-3.6 single cell suspensions were bound by the indicated ligand(s) for 45 min at 4 °C. Cells were then washed in phosphate-buffered saline, cross-linked with 1 mM disuccinimidyl suberate (Pierce) for 15 min at 15 °C as described (31). Unreacted disuccinimidyl suberate was quenched with 5 mM ammonium acetate, and whole cells were lysed directly in SDS sample buffer. Equal amounts of protein for each sample were resolved in a 5-10% SDS-polyacrylamide gel electrophoresis gradient, transferred to nitrocellulose, and Western blotted with anti-Trk polyclonal antibody 203 (a gift of Dr. David Kaplan, McGill University) that recognizes the intracellular domain of Trk. This antibody was selected because of high specificity toward Trk in Western blots and because its epitopes are not affected by disuccinimidyl suberate cross-linking.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Cell Survival Induced by Monovalent and Bivalent TrkA Ligands-- Previously, we showed that anti-human TrkA mAb 5C3 significantly protected cells from apoptosis when cultured in serum-free medium (SFM), but anti-p75 mAb MC192 did not promote cell survival. Combinations of anti-TrkA mAb 5C3 and anti-p75 mAb MC192 synergized to protect cells optimally to levels comparable with 1 nM NGF, as did combinations of mAb MC192 and 10 pM NGF (6). Therefore, we tested 4-3.6 cells (human TrkA⁺⁺⁺ rat p75⁺⁺⁺) (Fig. 2A) or PC12 cels (rat TrkA⁺ rat p75⁺⁺⁺) (Fig. 2B) in the same paradigm but using putative monovalent ligands.

View larger version (29K): [in this window] [in a new window]   Fig. 2.   Trophic protection by monovalent and bivalent TrkA or p75 ligands. Cells were cultured in SFM supplemented with the indicated ligands for ~68 h. Cross-linking of Fabs was achieved with a 2-fold molar excess of -Fab antibody. Cell survival was measured in MTT assays. Protection from apoptotic death was calculated relative to that of optimal NGF (1 nM, 100% protection). Results shown are the average ± S.E., n = 4, and representative from at least three experiments. A, 4-3.6 cells. *, significant protection compared with control mouse IgG. **, significantly higher than 0.5 nM mAb 5C3. B, PC12 cells. *, significantly higher than 10 pM NGF. **, not significantly different from 10 pM NGF. p < 0.01.

Significant protection was afforded by 5C3 Fabs, in a dose-dependent manner. 5C3 Fabs at 1-10 nM afford protection comparable with 10 pM NGF. More robust protection was afforded by 1-10 nM 5C3 Fab·-Fab complexes. Negative control -Fab or mouse IgG did not afford survival. Positive control bivalent mAb 5C3 at the optimal concentration of 0.5 nM protected ~50% of the cells. Interestingly, 10 nM 5C3 Fab·-Fab complexes afforded significantly higher protection than 0.5 nM 5C3 IgG, possibly because 5C3 Fab·-Fab complexes are more flexible than IgG or oligomerize TrkA more efficiently.

Monovalent p75 Ligands Do Not Potentiate NGF Signals-- 1 nM NGF protects PC12 cells (expressing rat TrkA⁺ rat p75⁺⁺⁺) from apoptosis induced by culture in SFM. Low concentrations of NGF (10 pM) as a high affinity ligand afforded ~30% survival. Monovalent MC192 Fabs failed to synergize with 10 pM NGF, and protection was not significantly different than 10 pM NGF alone. Synergy did occur with MC192 Fab·-Fab complexes plus 10 pM NGF. The effect was dependent on the concentration of MC192 Fab·-Fab complexes (data not shown) and was optimal at 1 nM cross-linked MC192 Fab.

As positive control, bivalent MC192 synergized with 10 pM NGF increasing protection from ~30 to ~90%. Synergy was dependent on the concentration of mAb and was optimal at 0.5 nM MC192 (data not shown). Controls -Fab, mouse IgG, bivalent MC192, monovalent MC192 Fabs, and MC192 Fab·-Fab complexes alone did not protect PC12 cells substantially from apoptosis. In all permutations of these experiments, apoptotic cell death was confirmed by analysis of DNA fragmentation patterns (data not shown).

Synergy of Bivalent and Monovalent Ligands of NGF Receptors-- To analyze the valency requirement of each NGF receptor, combinations of bivalent and monovalent antibody-based ligands were tested on 4-3.6 cells for synergy in protection of apoptotic cell death (Fig. 3). In these assays, it was encouraging to observe that comparable biological responses by different ligands (e.g. 1 nM cross-linked 5C3 Fabs afford the same protection as 0.5 nM 5C3 bivalent IgG) also result in equivalent receptor occupancy (e.g. 1 nM Fabs bind the same number of receptors as 0.5 nM IgG). Positive controls of bivalent 5C3 combined with bivalent MC192 were synergistic and afforded 100% protection. Negative controls -Fab alone and mouse IgG alone did not afford cell survival in SFM (data not shown).

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Synergistic trophic protection by p75 and TrkA Ligands. Experiments using 4-3.6 cells were as described in Fig. 2. Cross-linking of Fabs was achieved with a 2-fold excess of -Fab antibody. Only optimal doses for each of the TrkA or p75 ligands (0.5 nM IgG and 1 nM Fabs) are shown. Protection from apoptotic death was calculated relative to that of optimal NGF (1 nM, 100% protection). Results shown are average ± S.E., n = 4, and are representative from at least three independent experiments. *, not significantly higher than control mouse IgG. **, significantly higher than control mouse IgG; ***, not significantly higher than 5C3 Fabs alone. ****, not significantly higher than mAb 5C3 alone. *****, significantly lower than all dimeric combinations of MC192 and 5C3. p < 0.01.

A combination of monovalent 5C3 Fab with either monovalent MC192 Fab or with bivalent MC192 did not result in synergy; the ~25% protection was not significantly different from that seen with monovalent 5C3 Fab alone. However, 5C3 Fab·-Fab complexes synergized with bivalent MC192. A combination of monovalent MC192 Fab with bivalent 5C3 did not result in synergy; the ~50% protection was not significantly different than that afforded by bivalent 5C3 alone. In contrast, MC192 Fabs·-Fab complexes synergized with bivalent 5C3 and afforded ~110% protection. One bias in this assay is that -Fab cross-linking does not occur exclusively at MC192 Fabs but also occurs upon bivalent 5C3, resulting in some multivalent 5C3 oligomers. However, -Fab cross-linking of bivalent 5C3 IgG does not enhance its activity (data not shown), therefore the biological effect of cross-linking occurs at the MC192 Fabs.

Last, -Fab cross-linking of 1 nM MC192 Fab plus 5C3 Fab afforded ~65% protection. This activity can be ascribed to four theoretical ligand mixtures: 5C3 dimers (25%), MC192 dimers (25%), and 5C3/MC192 heterodimers (50%). If the 5C3/MC192 heterodimers are indeed formed, they seem to be inactive, because we observed that reducing the concentration of bivalent mAbs 5C3 and MC192 to 0.125 nM (the concentrations in the theoretical mixtures above) results in synergy and ~65% protection (data not shown).

While the date is suggestive that heterodimers are inactive, this is an unclear issue, because we have no evidence that the heterodimers indeed form. The putative heterodimeric ligands cannot be isolated and analyzed because they dissociate and reassociate during purification; nor can they be stabilized, because binding activity is lost upon chemical cross-linking. Additionally, it is noteworthy that more extensive cross-linking with higher ratios of -Fab did not increase protection, although higher oligomerization of ligands is expected (data not shown).

For clarity, only nearly optimal concentrations of ligands are presented, but responses from micromolar to picomolar concentrations were studied (see "Materials and Methods"). Thus, optimal protection is afforded by combinations that result in homodimerizing ligands, and no increased protection is seen with oligomerizing ligands. In all permutations of these experiments, apoptotic cell death was confirmed by analysis of DNA fragmentation patterns (data not shown). Moreover, the ligands mediate trophic effects in a TrkA-dependent manner, because no concentration or combination of NGF and antibody could induce significant protection of B104 cells (TrkA, p75⁺⁺⁺) (data not shown).

Ligand Valency and TrkA Tyrosine Phosphorylation-- TrkA tyrosine phosphorylation (TrkA-Tyr(P)) was studied as a biochemical correlate of cell survival in SFM (Fig. 4). Analysis was done by Western blotting with mAb 4G10 against phosphotyrosine (total Tyr(P)), or with antibodies against phosphorylated tyrosine 490 of TrkA (Tyr(P)⁴⁹⁰), which is the Shc binding site of TrkA. A representative Western blot of total Tyr(P) is shown in Fig. 4A. Statistical analysis of densitometry for several blots of total TrkA-Tyr(P) (Fig. 4B, upper panel), and for several blots of TrkA-Tyr(P)⁴⁹⁰ (Fig. 4B, lower panel) were used to quantify the TrkA-Tyr(P) data. Western blotting with anti-TrkA antibodies, done in parallel, demonstrated that all lanes contained the same amount of receptor (data not shown).

View larger version (24K): [in this window] [in a new window]   Fig. 4.   Bivalent TrkA ligands induce optimal TrkA tyrosine phosphorylation. Equal amounts of protein from whole cell lysates were resolved by SDS-polyacrylamide gel electrophoresis and analyzed by Western blotting with anti-Tyr(P) mAb 4G10 (total Tyr(P)) or with an antibody recognizing Tyr(P)490 within the Shc binding site of TrkA (TrkA Tyr(P)490) (not shown). Blot shown is representative of at least three independent total Tyr(P) experiments. A, 4-3.6 cells were untreated (lane 1) or treated with indicated ligands for 15 min at 37 °C. In lanes 3, 5, and 7, monovalent Fabs were cross-linked with -Fab antibodies as indicated (+). Homodimeric binding of p75 and TrkA (lane 11) with intact IgGs enhances TrkA Tyr(P) over each IgG alone (lanes 9 and 10). Ligand concentrations are as in Table I. *, significant difference from untreated samples (paired t tests, n = 6, p < 0.05). B, densitometric scanning quantification of TrkA total Tyr(P) (upper panel) and Tyr(P)490 (lower panel) intensities relative to optimal NGF treatment (average ± S.E., n = 6). Filled bars indicate anti-Fab cross-linking. *, significant difference from untreated samples (paired t tests, n = 6, p < 0.05).

Monovalent 5C3 Fabs induced small but significant increases in TrkA-Tyr(P) (Fig. 4A, lane 2) and TrkA-Tyr(P)⁴⁹⁰ compared with untreated control (Fig. 4A, lane 1) or MC192 Fabs (Fig. 4A, lane 4). Much higher signals were induced by 5C3 Fab·-Fab complexes (Fig. 4A, lane 3). Quantification showed that ~80% of total TrkA-Tyr(P) and ~55% of TrkA-Tyr(P)⁴⁹⁰ were induced compared with optimal NGF-induced signals (Fig. 4B). In contrast, no significant TrkA-Tyr(P) or TrkA-Tyr(P)⁴⁹⁰ was induced by p75 ligands bivalent MC192, MC192 Fab·-Fab complexes, or monovalent MC192 Fabs (Fig. 4A, lanes 10, 5, and 4). For quantitative statistical analysis of these data, see Fig. 4B. All of these findings are consistent with the survival data.

There were no significant differences between treatments with 5C3 Fabs plus MC192 Fabs (Fig. 4A, lane 6) versus 5C3 Fabs alone (Fig. 4A, lane 2), indicating a lack of synergy. Cross-linking of 5C3 Fabs plus MC192 Fabs with -Fab afforded an increase in total Tyr(P) (Fig. 4A, lane 7).

Approximately 85% of total Tyr(P) and ~65% of Tyr(P)⁴⁹⁰ of TrkA were induced compared with optimal NGF-induced levels (Fig. 4B). However, the increases in TrkA Tyr(P) and Tyr(P)⁴⁹⁰ induced by 5C3 Fab·MC192 Fab·-Fab complexes were not statistically different from increases induced by 5C3 Fab·-Fab complexes (Fig. 4B). All of these findings are consistent with the survival data.

Thus, 5C3 Fabs are partial agonist monovalent ligands of TrkA that induce receptor activation and lead to trophic cell survival. Although the evidence that mAb 5C3 Fabs are indeed monomeric is compelling (see "Materials and Methods"), it is possible that these large Fab molecules of ~50 kDa could aggregate. Hence, other ligands were tested.

Characterization of Small Molecule Monomeric TrkA Ligands-- C(92-96) is a small molecule (~1 kDa), cyclic and conformationally constrained peptide analog of the C-D -turn region of a single NGF protomer. Therefore, the C(92-96) mimic of NGF was studied as a candidate genuine monovalent and monomeric TrkA ligand.

To address the valency of C(92-96), we determined the solution structure of the pharmacophore to better than 0.5 Å root mean square deviation. Nuclear Overhauser effect and total correlation spectroscopy spectra were consistent with a monomeric, nonaggregated state and a five-residue pharmacophore within a -turn (27). Five chemical moieties are too few to bind two receptors simultaneously as a bivalent agent, hence this ligand is monovalent.

The following criteria indicate that C(92-96) is monomeric. First, mass spectroscopy of C(92-96) demonstrated that there were no covalent dimers or oligomers (see "Materials and Methods"). Second, the aggregation state of the peptide at millimolar concentrations in solution was resolved by high resolution proton NMR spectroscopy (Fig. 5A). Third, natural ¹³C abundance NMR relaxation parameters were measured for the -carbon atom, heteronuclear NOEs, and the molecular correlation time of C(92-96) was assessed. The overall correlation time detected of 1.76 ns at 5 °C is expected for a monomer. Fourth, the translational self-diffusion constant in solution unequivocally identified C(92-96) as monomeric.

View larger version (16K): [in this window] [in a new window]   Fig. 5.   NMR spectroscopy of N-Ac(92-96) peptide mimetic. A, one-dimensional spectrum of the monomeric Nac-C(92-96) peptide in H2O solution, pH 5.6, 278 K. B, the absence of concentration-dependent aggregation as measured by the translational diffusion rate. Measurements were made by the pulsed field gradient technique of Stejskal and Tanners (reviewed in Ref. 29) at three different peptide concentrations. The dashed line indicates the expected diffusion rate for a dimeric peptide of the same shape as N-Ac-C(92-96). The plot shows the attenuation of the signal intensity (vertical axis) as a function of molecular size and shape (monomer versus dimer) and as a function of gradient duration (), strength (G), diffusion delay (), and gyromagnetic ratio (). The diffusion constant for the monomer is 1.02 ± 0.06 × 106 (cm2 s1), and for the dimer it is predicted to be 0.81 × 106 (cm2 s1).

Pulse field gradient NMR measurements of the self-diffusion coefficient () were determined at various peptide concentrations of 2, 6, or 18 mM; T = 278 K. Values of  = 1.01 ± 0.07 (10⁶ cm²/s);  = 1.00 ± 0.06 (10⁶ cm²/s); and  = 1.04 ± 0.05 (10⁶ cm²/s) were measured for 18, 6, and 2 mM samples, respectively (Fig. 5B). These  values are essentially the same, indicating an identical state for the peptide. Thus, the samples remain monomeric, and peptide aggregates are undetectable in solution even at concentrations as high as 18 mM. We estimate that the self-association constant for any putative aggregate cannot be larger than 10 M¹. Thus, a 10 µM solution of C(92-96) (used hereafter) could not contain >1 nM self-aggregated dimers, if any aggregate at all.

Small Molecule Monomeric and Monovalent Agonists of TrkA-- Four questions were addressed. First, the genuine monovalent and monomeric TrkA ligand C(92-96) was tested for trophic support of cells in SFM. Second, a covalent dimeric analog of C(92-96) termed C(92-97)_dimer was also evaluated to directly compare the potency and efficacy of monomeric versus dimeric small molecule TrkA ligands. Third, to study whether surface density of TrkA receptors influences trophic signals, these agents were assayed in parallel on cell lines that differ only in TrkA density (PC12 versus 6-24, and B104 versus 4-3.6 cells). Fourth, to study whether the ligands activate receptors in normal neurons, primary cultures of dissociated dorsal root ganglia from day 17 rat embryos were tested. These cells express TrkA and p75 receptors, and their survival and differentiation are dependent on TrkA activation. Growth and survival were studied first in MTT assays (Table I). Differentiation was studied morphometrically (Fig. 6).

View larger version (97K): [in this window] [in a new window]   Fig. 6.   Monovalent TrkA ligands induce the differentiation of embryonic DRG cultures. Primary neuronal cultures from embryonic day 17 rat DRGs were treated with the indicated ligands for 8 days, and cell differentiation was studied morphometrically. Magnification was × 60. Data are representative of three independent experiments.

The trophic response was dependent on NGF dose and was optimal to 1 nM NGF in all cell types (Table I, row 1). Better survival was seen at 10 pM NGF for 6-24 and 4-3.6 cells (Table I, row 3), suggesting that high TrkA expression affords better efficacy when ligand concentrations are limiting. B104 cells did not respond to any dose of NGF (data not shown). Negligible survival to 10 pM NGF in DRG cultures is due to the fact that DRG cultures are heterogeneous and secrete growth factors, masking the effect of low concentrations of exogenous NGF.

The C(92-96) NGF mimic did not afford significant survival of PC12, 6-24, or 4-3.6 cells compared with control linear peptides, but it did afford significant survival of DRG cultures (Table I, row 4 versus row 6). This effect was dose-dependent, and C(92-96) at 1 µM afforded ~10% growth of DRG (data not shown). In contrast, the C(92-97)_dimer peptide afforded good trophic support for 6-24 and 4-3.6 cells and very low but statistically significant support for PC12 cells (Table I, row 5). The 6-2.4 and 4-3.6 cells express comparable numbers of TrkA receptors, suggesting a TrkA density-dependent response.

MAb MC192 alone afforded very low or insignificant trophic support of cell lines (Table I, row 7); but as a bivalent p75 ligand, it synergizes with TrkA ligands (e.g. see Fig. 2). High DRG survival in response to mAb MC192 alone (Table I, rows 7 and 10) is explained by the mAb potentiating endogenously produced growth factors. Furthermore, bivalent MC192 potentiated the activity of C(92-96) (Table I, row 8). In cell lines the combination is synergistic, while in DRG cultures the combination is additive due to high protection afforded by each ligand alone.

As a control, bivalent MC192 did not synergize with linear peptides (Table I, row 10). Further controls using B104 cells (TrkA, p75⁺⁺⁺ parental to 4-3.6) demonstrated no protection by the peptide NGF mimics, alone or in combination with mAb MC192 (data not shown), suggesting that the activity requires TrkA expression.

Monovalent TrkA Ligands Induce the Differentiation of Embryonic DRG Cultures-- The differentiation of dissociated primary DRG cultures was studied (Fig. 6). Untreated DRG cultures had sparse, bipolar, and poorly differentiated neurons (Fig. 6A). At 20 pM NGF the increase in the number and the length of neurites and branches was very low (Fig. 6B); at 1 nM NGF the increase was optimal (Fig. 6C). Treatment with control linear peptide did not induce differentiation (not shown). Treatment with 0.5 nM MC192 alone (Fig. 6D) or with 10 µM C(92-96) alone (Fig. 6E) induced substantial differentiation. However, treatment with a combination of 10 µM C(92-96) plus 0.5 nM MC192 (Fig. 6F) induced higher differentiation, comparable with that induced by 1 nM NGF. These differentiation data are consistent with synergy in survival seen for cell lines and primary cultures (see Table I).

Monovalent TrkA Ligands Induce TrkA Tyrosine Phosphorylation in Synergy with Bivalent p75 Ligands-- To further assess whether the signals induced by small cyclic peptides are mediated by TrkA, tyrosine phosphorylation of the receptor was studied in 4-3.6 cells (Fig. 7). Representative anti-Tyr(P) Western blots are shown in Fig. 7A. A summary of densitometric analysis from several blots is given in Fig. 7B.

View larger version (26K): [in this window] [in a new window]   Fig. 7.   Monovalent ligands induce TrkA tyrosine phosphorylation. Treatment and analysis of 4-3.6 cells was as in Fig. 4; ligand concentrations are as per Table I. A, cells were untreated (lane 1) or treated with 1 nM NGF (lane 2), 10 pM NGF (lane 3), 10 µM C(92-96) (lane 4), 10 µM C(92-97)dimer (lane 5), 10 µM control linear peptide (lane 6), 1 nM MC192 mAb (lane 7), 10 µM C(92-96) plus 1 nM MC192 (lane 8), 10 µM C(92-97)dimer plus 1 nM MC192 (lane 9), or 10 µM control linear peptide plus 1 nM MC192 (lane 10). B, densitometric scanning quantification of band intensities relative to optimal NGF treatment (average ± S.E., n = 6). *, significant difference from untreated samples (paired t tests, n = 6, p < 0.05).

C(92-96) alone did not induce an increase in TrkA-Tyr(P) compared with untreated cells or cells treated with control linear peptide or bivalent MC192 (Fig. 7A, lane 4 versus lanes 1, 6, and 7). Significant TrkA-Tyr(P) was induced by treatment with C(92-97)_dimer (Fig. 7A, lane 5), representing ~30% of that induced by 1 nM NGF (Fig. 7A, lane 2). Combinations of mAb MC192 and C(92-96) peptide (Fig. 7A, lane 8) or mAb MC192 and C(92-97)_dimer peptide (Fig. 7A, lane 9) afforded notable increases in TrkA-Tyr(P), comparable with those induced by 10 pM NGF (Fig. 7A, lane 3). These results are consistent with the survival data. In contrast, treatment with a combination of bivalent MC192 and linear peptide controls (Fig. 7A, lane 10) did not result in significant increases in TrkA-Tyr(P). For statistics of densitometric analysis, see Fig. 7B.

Small Molecule Monovalent Ligands Induce TrkA Receptor Homodimerization-- TrkA tyrosine phosphorylation leading to trophic and differentiative signals require TrkA homodimerization. To study whether monovalent C(92-96) peptide induces TrkA homodimerization, chemical cross-linking studies of the receptor were done in 4-3.6 cells (Fig. 8). Cells were treated as indicated, followed by chemical cross-linking, and then were detergent-solubilized and analyzed by Western blotting with anti-Trk polyclonal antibody 203. 

View larger version (70K): [in this window] [in a new window]   Fig. 8.   Monovalent ligands induce or stabilize putative TrkA homodimers. 4-3.6 cells were untreated (lane 1) or treated with 2 nM NGF (lane 2), 1 nM MC192 (lane 3), 10 µM C(92-96) (lane 4), or 10 µM C(92-96) plus 1 nM MC192 (lane 5). Cells were then chemically cross-linked with disuccinimidyl suberate, lysed, resolved by SDS-polyacrylamide gel electrophoresis, and analyzed by Western blotting with anti-TrkA polyclonal antiserum.

A doublet consistent with previously reported TrkA monomers of p110 and p140 was seen in all samples (Fig. 8, arrows). Samples from NGF-treated cells and in C(92-96) plus MC192-treated cells had a band of ~280 kDa, consistent with the molecular mass of TrkA-TrkA homodimers (Fig. 8, lanes 2 and 5). This band was also detected, albeit weakly, in samples from cells treated with C(92-96) alone (Fig. 8, lane 4). A second band of ~220 kDa (that may be consistent with cross-linked p140-p75 heterodimers or p110 homodimers) was seen in samples from NGF-treated cells and more weakly in C(92-96) plus MC192-treated cells (Fig. 8, lanes 2 and 5). The 280-kDa and 220-kDa bands were not seen in untreated cross-linked cells (Fig. 8, lane 1), in cells treated with MC192 alone (Fig. 8, lane 3), or in linear peptide control with or without MC192 treatment (data not shown). Similar data were obtained whether whole cell lysates were analyzed or cell lysates were immunoprecipitated with anti-TrkA antibodies prior to Western blotting (data not shown).

Given that the efficiency of chemical cross-linking is <5% of the TrkA expressed on the cell surface, we have not studied the individual components of the 280- and 220-kDa bands other than the fact that they contain TrkA. However, it is unlikely that these bands comprise NGF, because they are detected in the C(92-96) plus MC192-treated cells.

Small Molecule Monovalent Ligands Induce TrkA Receptor Internalization-- Next, we assessed whether the monomeric C(92-96) peptide binds to a receptor domain that overlaps with mAb 5C3. This study was done by attempting to block mAb 5C3 binding with C(92-96). Moreover, since agonistic ligands are expected to cause receptor internalization, it was of interest to study whether monomeric ligands such as C(92-96) can induce TrkA internalization.

We studied ligand-dependent receptor internalization as a decrease of surface receptor density, which can be inhibited by low temperature or by poisons such as sodium azide. 4-3.6 cells were treated with NGF, C(92-96) peptide, or control peptides in the presence or absence of sodium azide at different temperatures. Surface TrkA receptors were quantitated by FACScan analysis with mAb 5C3 before and after 20 min of internalization (Table II). This time was selected because the t for ¹²⁵I[NGF] internalization is ~10 min (9).

View this table: [in this window] [in a new window]   Table II NGF peptide mimics induce TrkA internalization 4-3.6 cells were treated with the indicated ligands or controls at 4 °C in the presence (+) or absence () of sodium azide. Cells were then maintained at 4 °C or shifted to 37 °C to allow internalization. Analysis was done by FACScan, immunostaining surface TrkA with mAb 5C3. Percentage loss of binding was calculated as a change in mean channel fluorescence with respect to untreated cells. Results are average ± S.E., n = 3, 5000 cells were acquired for each assay.

C(92-96) did not block surface mAb 5C3 binding sites at 4 °C (Table II, row 2). Treatment with C(92-96) at 37 °C caused a ~23% loss of surface mAb 5C3 binding sites (Table II, row 2). This effect was sensitive to sodium azide (Table II, row 3). A 23% loss of surface TrkA represents ~11,000 receptors that presumably internalized out of ~50,000 expressed at the surface. Similar results were obtained with C(92-97) (data not shown). Negative control linear peptides did not affect the number of mAb 5C3 binding sites (Table II, rows 4 and 5).

In positive control studies, treatment with NGF at 4 °C blocked ~21% of the surface mAb 5C3 binding sites (Table II, row 1; also published in Ref. 19), suggesting that NGF partially blocks mAb 5C3. Treatment with NGF at 37 °C increased loss of surface mAb 5C3 binding sites from ~21 to ~46% (Table II, row 1), probably because of TrkA internalization.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

We demonstrate that artificial ligands selective for subunits of receptor complexes can be used to study receptor structure-activity relationships in systems where each subunit has distinct or unclear functions. For NGF receptors, the main novel findings of this study are as follows: (i) genuine monomeric and monovalent ligands of TrkA can be partial agonists, suggesting that bivalent ligands are not the sole mechanism for dimerizing and/or activating tyrosine kinase receptors; (ii) the monovalent p75 ligands used in this study do not enhance TrkA-mediated signals, suggesting that p75 ligands may require bivalency; and (iii) bivalent ligands that induce TrkA-TrkA and p75-p75 homodimers afford optimal signals. Putative ligand-induced TrkA-p75 heterodimers do not seem to afford signals, and receptor oligomerization does not result in enhanced signals.

While it is well established that physiological ligands activate signal transduction by inducing receptor dimerization, allosteric models of receptor activation have also been proposed (15, 32-36). There are two major obstacles to studying allosteric models experimentally. First, there is a paucity of monovalent ligands that activate receptor tyrosine kinases (18). Second, the role of each subunit must be considered in the analysis of heteromeric receptors, and agents that bind to and affect the activity of each subunit must be available. We postulate that the strategy of using growth factor-derived and antibody-based artificial ligands can be easily adapted to study other multisubunit receptors, or for receptors where a subunit has unclear function or no defined ligand. Also, strategies that develop monovalent small molecule agonists that bind to the extracellular domain of receptors will be useful for the discovery of pharmacological agents.

Valency, Avidity, and Aggregation State of Agonistic Ligands-- Monovalent ligands 5C3 Fabs and C(92-96) are partial agonists. For T cell receptor complexes and G-coupled receptors, it has been shown that sometimes ligands with high affinity can overcome low avidity and lead to activation or to conformational changes (37-39). This does not seem to be the case for the ligands 5C3 Fabs or C(92-96) because their affinity ranges from nanomolar to micromolar. Hence, N-Ac-C(92-96) is a ligand of relative low affinity and avidity that can induce or stabilize putative TrkA homodimers. Since receptor dimerization alone does not necessarily cause activation (15, 17), the simplest interpretation is that the TrkA ligands induce allosteric or conformational changes, as shown for other receptors (16, 18, 35). However, this report would be a case where monomeric and monovalent ligands induce allosteric or conformational changes.

Arguably, 5C3 Fabs could aggregate in solution as do other large peptides (40), but this is unlikely to occur at nanomolar Fab concentrations in 0.2% bovine serum albumin and did not occur at micromolar Fab concentrations in related studies (41). Hence, we conclude that Fabs are monomeric. It is also unlikely that 5C3 Fabs are bivalent, and sequence analysis of the variable complementarity-determining regions of mAb 5C3 excluded this possibility.²

More conclusively, the genuine monomeric small peptide C(92-96) affords trophic signals, although the dimeric C(92-97)_dimer is more efficient. Monomeric C(92-96) is monovalent, because it has a 5-residue pharmacophore (27) and could not interact with two receptors simultaneously. The intriguing possibility that C(92-96) could dimerize after docking is unlikely (see below), but it remains unexplored and would require structural analysis of receptor-ligand complexes.

Ligand Density at the Cell Surface-- The optimal functional concentration of the peptide N-Ac-C(92-96) and of mAb 5C3 and 5C3 Fabs approximates their K_d (10 µM, 2 nM, and 5 nM, respectively). In contrast, the optimal functional concentration of 1 nM NGF is 2-3 orders of magnitude above its K_d (~10¹¹ to 10¹² M). It is unlikely that these differences reflect requirements for receptor occupancy. It is more likely that NGF, the mAbs, or the peptides have different half-lives in solution at 37 °C or that they are bound by matrix proteoglycans or carrier proteins.

In some cases, the local concentration of a ligand at the cell surface can be higher than in solution, making the ligands more likely to dimerize or to aggregate. While the mobility of a ligand is reduced to two dimensions when bound to a receptor, ligand mobility within the plane of the membrane is still exclusively dependent on the receptor. Therefore, monovalent ligands could only dimerize subsequent to inducing their receptors to dimerize. In addition, we estimate that a cell expressing 50,000 receptors out of which 5% are bound would achieve a local ligand concentration of ~10 mM. This concentration was tested by NMR for monovalent C(92-96) with no evidence of aggregation. Last, we demonstrated that the monovalent ligands induce rapid receptor internalization, which will effectively reduce possible high local concentrations of ligand at the cell surface. It is noteworthy that self-aggregation of C(92-96) beyond the detection of NMR is unlikely to account for the effects, because C(92-97)_dimer at high nM concentrations did not afford signals.

Mechanism of Action of TrkA and p75 Ligands-- How can conformational changes induced by monovalent ligands lead to receptor dimerization? While the function of the monovalent ligands is defined by comparison with the natural ligand NGF, their mechanism of action may differ. We hypothesize three possible mechanisms: (i) conformational changes that favor direct dimerization; (ii) a reduction of the rate at which preformed receptor dimers disengage; and (iii) increased receptor mobility with a consequent increase in spontaneously dimerized receptors.

The most attractive explanation is that the monovalent ligands could be inverse antagonists (32, 33). Inverse antagonists can stabilize receptor conformation(s) that are favorable to subsequent TrkA-TrkA interactions. Indeed, one criterion for defining inverse antagonists is that their potency increases with increased receptor density on the cell surface, and this was observed for the activity of C(92-96) and C(92-97)_dimer in PC12 cells that have low TrkA numbers versus 6-24 and 4-3.6 cells that have high TrkA numbers. Furthermore, an inverse antagonist would antagonize the natural agonist NGF, and NGF blocking properties have been shown previously for C(92-96) (20, 21). Last, as expected, the C(92-97)_dimer affords higher signals, which would be predicted for a bivalent ligand that induced receptor dimerization directly.

With respect to p75 receptors, bivalent ligands potentiate the effects of all bivalent TrkA ligands (NGF, mAb 5C3, 5C3 Fab·-Fab complexes, and C(92-97)_dimer). However, bivalent p75 ligands did not potentiate all monovalent TrkA ligands. mAb MC192 potentiated the activity of C(92-96), but it did not potentiate the activity of 5C3 Fabs. These data suggest that these monovalent TrkA ligands probably have different mechanisms of activation, and this possibility is supported by the fact that C(92-96) and 5C3 Fabs bind to nonoverlapping sites.

We speculate that a small molecule like C(92-96) docks onto a small pocket of TrkA and may therefore be sensitive to a documented p75-mediated regulation (6, 7) or internalization (9, 42, 43) of TrkA. Consequently, C(92-96) may be sensitive to ligands binding p75 concomitantly, whereas larger molecules like 5C3 Fabs possess more extended TrkA binding surfaces and do not exhibit this sensitivity. Hence, two classes of partial agonism (or inverse antagonism) by monomeric TrkA ligands may have been uncovered in this study.

Agonistic Ligand Binding Sites-- MAb 5C3 and the NGF analog C(92-96) do not block each other's binding; hence, they bind to nonoverlapping sites of TrkA. Ligands docking onto restricted receptor pockets or "hot spots" are presumed to be more efficient at mediating (ant)agonistic function (34, 44). Reportedly, there are at least two activating TrkA "hot spots" (45-48) encompassing the IgC-2-like domain and the leucine-rich motif.

mAb 5C3 binds an epitope within the IgC-2-like domain of TrkA, and the epitope is stabilized by disulfide bonds (19). TrkA and other tyrosine kinase receptors have a dimer interface stabilized by disulfide bonds (49, 50). The agonistic "hot spot" of mAb 5C3 and 5C3 derivatives may be at the dimer interface of this receptor. However, 5C3 and C(92-96) do not block each other; hence, the data suggest that C(92-96) binds elsewhere. NGF may utilize both sites to fully activate the receptor.

Conclusions-- The screening of functional small molecule ligands that bind multisubunit receptors may require testing ligand combinations that target all subunits. It would be of interest to test other NGF receptor ligands or activators in this paradigm of synergy, including peptidic small molecule p75 ligands (21) or similar peptides reproduced by others (51, 52), organic p75 ligands, gangliosides, and alkaloid derivatives that activate TrkA (53-55).

With respect to NGF receptors, our results support the hypothesis that functional receptors consist of TrkA homodimers and p75 homodimers. Our results also demonstrate that genuine monovalent and monomeric ligands of TrkA tyrosine kinase receptors can be functionally agonistic. Recently, two small molecule ligands of other receptors were shown to be agonistic. In one, a mimic of granulocyte colony-stimulating factor activated the granulocyte colony-stimulating factor receptor (56), but no studies of the aggregation state of the ligand were performed. In another, a small molecule activated the insulin receptor tyrosine kinase (57). However, this insulinomimetic ligand is a symmetrical lipophylic agent, in principle capable of dimerizing the receptor as shown for similar ring structures (16). Hence, our study is the first formal proof, to our knowledge, of genuine monovalent ligands of the extracellular domain of a tyrosine kinase acting as partial agonists by inducing or stabilizing receptor homodimers.

Neurotrophins and their receptors play a role in neurodegenerative diseases, pain, and neoplasias (44). Internalizing TrkA ligands could be exploited to deliver radioligands, toxins, oligonucleotides, or membrane-impermeable molecules selectively to receptor-expressing cells. This study has implications for the design and screening of small molecules with pharmacological, diagnostic, or targeting activity for neurotrophin receptors.

	FOOTNOTES

^* This work was supported by grants from the Medical Research Council of Canada (MRCC) (to H. U. S.) and the National Cancer Institute of Canada (to K. G.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

^§ Recipient of a GlaxoWellcome studentship.

^¶ Recipient of a Lavoisier fellowship.

^** Recipient of a fellowship from the Fonds de Recherche en Santé de Québec.

^§§ Scholar of the MRCC and the Pharmaceutical Manufacturer's Association of Canada. To whom correspondence should be addressed: McGill University, Pharmacology and Therapeutics, 3655 Drummond St. #1320, Montréal, Québec H3G 1Y6, Canada. Tel.: 514-398-3628; Fax: 514-398-6690; E-mail: Uri@pharma.mcgill.ca.

² H. U. Saragovi, unpublished data.

	ABBREVIATIONS

The abbreviations used are: NGF, nerve growth factor; mAb, monoclonal antibody; DRG, dorsal root ganglia; FPLC, fast protein liquid chromatography; N-Ac, N-acetyl; SFM, serum-free medium; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide.

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES