Studies showing that neurotrophin binding to p75NTR can promote cell survival in the absence of Trk (tropomyosin-related kinase) receptors, together with recent structural data indicating that NGF may bind to p75NTR in a monovalent manner, raise the possibility that small molecule p75NTR ligands that positively regulate survival might be found. A pharmacophore designed to capture selected structural and physical chemical features of a neurotrophin domain known to interact with p75NTR was applied to in silico screening of small molecule libraries. Small, nonpeptide, monomeric compounds were identified that interact with p75NTR. In cells showing trophic responses to neurotrophins, the compounds promoted survival signaling through p75NTR-dependent mechanisms. In cells susceptible to proneurotrophin-induced death, compounds did not induce apoptosis but inhibited proneurotrophin-mediated death. These studies identify a unique range of p75NTR behaviors that can result from isolated receptor liganding and establish several novel therapeutic leads.
- small molecules
- nerve growth factor
- brain-derived neurotrophic factor
- neurotrophin receptor
- hippocampal neurons
Neurotrophins [NGF, BDNF, neurotrophin-3 (NT-3), and NT-4/5] are secreted proteins that have potent effects on neuronal survival, differentiation, and synaptic function (Chao, 2003; Huang and Reichardt, 2003; Teng and Hempstead, 2004; Lu et al., 2005). Mature neurotrophins bind preferentially to tropomyosin-related kinase (Trk) and p75NTR, whereas proneurotrophins, which contain an N-terminal domain proteolytically removed in “mature” forms, interact with p75NTR and through their N-terminal domains, with the sorting receptor sortilin (Fahnestock et al., 2001; Harrington et al., 2004; Nykjaer et al., 2004). p75NTR interacts with Trks and modulates Trk signaling but is also independently coupled to various prosurvival and proapoptotic signaling systems (Salehi et al., 2000; Roux et al., 2001; Mamidipudi et al., 2002). Depending on the operative ligands, coexpression of Trk or other receptors, and expression of downstream signaling elements, p75NTR promotes cell survival or death and modulates neurite outgrowth (DeFreitas et al., 2001; Chao, 2003). proNGF induces death of superior cervical ganglion neurons and oligodendrocytes through p75NTR, and its concomitant binding to p75NTR and sortilin has been shown to activate death pathways (Lee et al., 2001; Beattie et al., 2002; Nykjaer et al., 2004). In other systems, p75NTR signaling through phosphatidylinositol 3-kinase AKT (PI3K/AKT) or interleukin-1 receptor-associated kinase/nuclear factor κB (IRAK/NFκB) promotes survival (Carter et al., 1996; Roux et al., 2001; Mamidipudi et al., 2002; Reddypalli et al., 2005).
It has been suggested that the unliganded monomeric form of p75NTR is proapoptotic and that homodimerization induced by neurotrophin binding eliminates that effect (J. J. Wang et al., 2000). This is consistent with studies showing no effects on survival of monomeric p75NTR ligands, including monovalent Fabs (Maliartchouk et al., 2000b) and monomeric cyclic peptides (Longo et al., 1997), whereas related bivalent forms in each study promoted cell survival. However, it is unlikely that these monomeric ligands engage the receptor in the same way as the natural ligands. Moreover, although active NGF is a homodimer containing two potential p75NTR binding sites, recent structural evidence suggests it engages only one p75NTR molecule, disallowing the binding of another (He and Garcia, 2004). These findings raise the possibility that small molecules mimicking a single p75NTR-interacting domain of NGF might be found that are capable of competing with NGF or proNGF binding and/or modifying p75NTR signaling and regulating cell survival.
Materials and Methods
Computational modeling, pharmacophore generation, and virtual and functional screening.
Computational studies were performed using the Accelrys (San Diego, CA) Catalyst and InsightII systems. Three-dimensional (3D) multiconformer virtual libraries were generated using catDB and catShape (Accelrys) from molecule files obtained from numerous sources including: The Chapman and Hall/CRC Combined Chemical Dictionary (Boca Raton, FL), the National Cancer Institute library of anticancer compounds, and libraries from Maybridge (Cornwall, UK), Asinex (Moscow, Russia), Interbioscreen (Moscow, Russia), Chemstar (Minneapolis, MN), Comgenex (Budapest, Hungary), Timtec (Newark, DE), and Sigma (rare chemicals dictionary; St. Louis, MO). Molecular dynamics simulations were performed using the Discover module within InsightII, using the CFF91 force field and explicit water solvent. Pharmacophore placement and library searching were performed with Catalyst, as illustrated in Figure 1 and further described in Results.
Antibodies and proteins.
Polyclonal rabbit anti-NGF antibody (Ab) was obtained from Chemicon (Temecula, CA). Monoclonal anti-phospho-AKTS473, polyclonal anti-AKT, polyclonal anti-phospho-NFκB-p65S536, and site-specific polyclonal anti-TrkY490 (recognizes TrkBY515) were purchased from Cell Signaling Technology (Beverly, MA). Monoclonal anti-NFκB-p65 was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-actin was purchased from Sigma. Polyclonal anti-TrkB antibodies were purchased from Upstate (Charlottesville, VA). p75NTR polyclonal rabbit antibodies 9651 (Huber and Chao, 1995) and 9650 raised against the neurotrophin-binding region (residues 43–161; cysteine repeat regions II, III, and IV) of the extracellular domain of recombinant p75NTR were generously provided by Dr. Moses Chao (Skirball Institute, New York University, New York, NY). Mouse 2.5 S NGF was purchased from Invitrogen (Carlsbad, CA). Recombinant human BDNF was from Sigma. Polyclonal anti-IRAK was obtained from Santa Cruz Biotechnology, and polyclonal anti-p75NTR antibody was from Promega (Madison, WI). p75NTR/Fc and TrkA/Fc chimeras were obtained from R & D Systems (Minneapolis, MN). Furin-resistant proNGF was prepared as described previously (Beattie et al., 2002).
Hippocampal neurons were prepared from embryonic day 16 (E16) to E17 mouse embryos as described previously (Yang et al., 2003). Cell survival was examined by microscopic analysis of morphology and 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) metabolism, or terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) and 4′,6′-diamidino-2-phenylindole (DAPI) staining. For morphology/MTT studies, low-density cultures were initiated in poly-l-lysine-coated A/2 plates by adding 25 μl of cell suspension (2000 neurons per well; 12,500 cells/cm2), 25 μl of DMEM containing 10% FBS, and different concentrations of recombinant BDNF, NGF, or NGF loop 1 mimic (LM11A) compounds to each well. After 48 h in culture, cell survival was assessed as described previously (Longo et al., 1997) using a combination of standard morphological criteria along with visual determination of whether a given cell converted MTT to its blue formazan product. The number of surviving neurons was determined by counting the total number of cells in each well that were both morphologically intact and filled with blue product. For each neurotrophin or compound concentration, duplicate wells were counted and the resulting values averaged. Activity of each compound was confirmed by blinded counts. Counts were normalized to survival achieved with 0.95 nm BDNF or to baseline survival. Fitting of dose–response curves was performed with SigmaPlot (Systat Software, Richmond, CA).
For TUNEL/DAPI analysis, E16–E17 mouse hippocampal neurons were plated on poly-l-lysine-coated cover glasses in 24-well plates (Fisher Scientific, Pittsburgh, PA) by adding 0.25 ml of cell suspension (20,000 cells/cm2), 0.25 ml of DMEM containing 10% FBS, recombinant BDNF, or NGF, and different concentrations of LM11A compounds to each well. The final concentrations of BDNF and NGF were 1.9 nm. After 48 h in culture, cell death was assessed using the fluorescein-12-dUTP, DeadEnd Fluorometric TUNEL system (Promega) and Vectashield plus DAPI (Vector Laboratories Burlingame, CA). Cells were visualized with a fluorescence microscope (DM IRE2; Leica, Nussloch, Germany) using 520 nm (TUNEL) and 460 nm (DAPI) filters. For each condition, images of 24 preselected fields (totaling 300–500 cells) were captured in a nonbiased manner from each of three separate assays. The fraction of dead cells was determined by dividing the number of nuclei exhibiting TUNEL staining by the number of DAPI-positive nuclei. Counting was performed in a blinded manner.
For studies using p75NTR+/+ and p75NTR−/− neurons, mice carrying a mutation in exon 3 of the p75NTR gene (encoding the extracellular domain) (Lee et al., 1992) were bred onto a B6 congenic background (>10 B6 backcrosses).
For signaling pathway inhibitor studies, 2-(4-morpholinyl)-8-phenyl-4H-1-benzopyran-4-one (LY294002), 2-(2-diamino-3-methoxyphenyl-4H-1-benzopyran-4-one (PD98059) (Calbiochem, La Jolla, CA), and SN50 (Alexis, Lausen, Switzerland) were added to cultures at final concentrations of 25 μm, 50 μm, and 2.5 μg/ml respectively, concomitantly with BDNF, NGF, or LM11A compounds. For antibody inhibition studies, p75NTR antisera and control nonimmune serum were used at a final dilution of 1:100 in the presence of BDNF, NGF, or LM11A compounds. For all studies applying signaling inhibitors, p75NTR antibodies, or p75NTR−/− neurons, survival was assessed at 48 h.
Protein extraction and Western blot analysis.
For assays of Trk levels and Trk, AKT, and NFκB activation, hippocampal neurons derived from E16–E17 mice were cultured in poly-l-lysine-coated six-well plates (1.5 × 106 cells per well; Corning, Corning, NY) in DMEM containing 10% FBS, followed by incubation in serum-free DMEM for 2 h before addition of neurotrophins or compounds. At the indicated time points, neurons were harvested in lysis buffer consisting of: 20 mm Tris, pH 8.0, 137 mm NaCl, 1% Igepal CA-630, 10% glycerol, 1 mm PMSF, 10 μg/ml aprotinin, 1 μg/ml leupeptin, and 500 μm orthovanadate (Zhou et al., 1995). Lysates were centrifuged, the supernatant collected, and protein concentrations determined using the BCA protein assay reagent (Pierce, Rockford, IL). Western blots were performed as described previously (Yang et al., 2003). Western blot signals were detected using the ECL chemiluminescence system (Amersham Biosciences, Arlington Heights, IL).
NGF and proNGF displacement from p75NTR or TrkA.
NGF ELISA was performed as described previously (Longo et al., 1999). Briefly, 96-well plates were incubated with 0.1 pmol (at 1 nm) of p75NTR/Fc or TrkA/Fc recombinant protein (R & D Systems) overnight at 4°C followed by incubation with blocking buffer for 1 h at room temperature. proNGF at 1.8 nm or different concentrations of NGF and LM11A compounds were diluted in sample buffer, added to wells, and incubated for 6 h with shaking at room temperature. Plates were then washed five times with Tris-buffered saline (TBS) containing 0.05% Tween 20 and incubated with anti-NGF rabbit polyclonal antibody overnight at 4°C. After five washes with TBS, wells were incubated for 2.5 h at room temperature with anti-rabbit IgG horseradish peroxidase (HRP) conjugate and then washed five times again. 3,3′,5,5′-tetramethyl-benzidine substrate was added and optical density measured at 450 nm. Data showing compound-mediated displacement of NGF from p75NTR were fit using the Nonlinear Regression module of SigmaPlot to the following modification of the Gaddum-Schild equation [adapted from Motulsky and Christopoulos (2003)]: where top is the maximal signal, bottom is the basal signal, [C] is compound concentration, S is the Schild coefficient, Hill slope is the Hill coefficient, [NGF] is NGF concentration, EC50 is the concentration of NGF resulting in 50% maximal (top − bottom) binding, and A2 is the concentration of compound resulting in a doubling of the EC50 from the curve obtained in the absence of the compound. In these studies, the calculated Hill slope ranged from 1.0 to 1.6 and was generally not significantly different from 1.
p75NTR antibody competition.
NIH3T3 fibroblasts expressing either null vector or p75NTR (Huang et al., 1999) were obtained from Dr. William Mobley (Stanford University, Stanford, CA). Cells were grown in monolayers, harvested in PBS with 2 mm EDTA, pelleted, and resuspended in ice-cold DMEM HEPES with 1 mg/ml BSA. Six to nine × 106 cells from one confluent six-well plate were used for each experimental point. For binding analysis, p75NTR antibody (1:100) was allowed to bind in the presence or absence of 100 nm LM11A compounds for 90 min at 4°C with gentle rotation, followed by four washes in PBS. The final cell pellet was resuspended in lysis buffer. Western blots were performed as described above. To detect the presence of p75NTR antibody, blots were probed with horseradish peroxidase-linked goat anti-rabbit IgG (Amersham Biosciences). Signals were detected by the ECL chemiluminescence system. To control for variation in protein loading, the blot was stripped and reprobed with β-actin monoclonal antibody (Sigma).
Pheochromocytoma 12 (PC12) cells (generously provided by Dr. W. Mobley) (Beattie et al., 1996) were grown on plates coated with poly-l-lysine (Sigma) in DMEM containing 5% fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. NGF (1.9 nm) or LM11A compounds (5 nm) were added to the same medium for 1 min before harvesting. Coimmunoprecipitations were conducted as described previously (Wooten et al., 2001; Mamidipudi et al., 2002), with 400 μg of cell extract incubated overnight at 4°C with 2 μg of polyclonal anti-p75NTR antibody followed by 2 h incubation with 30 μl of protein-A beads (Pierce). After incubation, the beads were washed five times with protein lysis buffer as described above. Protein was eluted by boiling for 10 min in 30 μl of 2× NuPAGE LDS sample buffer. Aliquots were submitted to NuPAGE 4–12% Bis-Tris Gels (Invitrogen) and transferred to polyvinylidene difluoride membranes. Western blots were incubated with antibody against IRAK (1:1000) or p75NTR (1:2000), then with HRP-linked secondary antibodies. Signal was detected using the ECL chemiluminescence system.
Oligodendrocyte culture, proneurotrophin treatment, and cell death assay.
Cortical oligodendrocytes from rat pups were prepared as described previously (Yoon et al., 1998; Harrington et al., 2002). Cells were treated with recombinant, cleavage-resistant proNGF at 0.05 nm. Controls were treated with equivalent volumes of proNGF purification buffer containing 350 mm imidazole. Twenty-four hours after treatment, cells were fixed and processed for MBP and TUNEL staining as described previously (Beattie et al., 2002). Two hundred to 250 cells were counted per well, for a minimum of 600 cells per experimental condition.
Tandem in silico and in vitro screening yields chemically distinct neurotrophic compounds with high efficiency
Initial modeling of active dimeric cyclic peptides mimicking NGF β-hairpin loop 1, a domain that interacts with p75NTR (Longo et al., 2005), suggested that energetic and structural constraints would disallow simultaneous β-hairpin folding of both peptide subunits, and this implied that the peptides might act in a monomeric manner (S. M. Massa, unpublished data). Based on early virtual screening of 3D conformer libraries, it also seemed unlikely we would find many functionally dimeric molecules that were nonpeptides and <500 Da molecular weight. We therefore focused on finding compounds emulating a single loop 1 structure. Compounds were selected for screening in cell survival assays by virtual screening of a compound library with a pharmacophoric loop 1 model, as outlined in Figure 1. An average of 35 conformers of each of >800,000 compounds were screened, yielding ∼800 compounds with conformers that fit with a calculated internal energy of <10 kcal/mol. This number was further reduced to ∼60 compounds by visual inspection on the bases of likely steric compatibility with a hypothetical shallow receptor binding pocket and maximal flexibility of the functional groups. Thirty-five compounds were obtained from commercial sources, of which 23 were soluble in water. In preliminary studies using a chicken DRG neuronal survival assay, 4 of 23 compounds tested showed significant neurotrophic activity (Massa et al., 2002, 2003).
Neurotrophic effects of NGF and LM11A compounds are similar in hippocampal neuron cultures
To extend the evaluation of the compounds from chick DRGs to mammalian CNS neurons, further biologic analysis of the compounds was performed using embryonic mouse hippocampal neurons. Under conditions of low density and lack of glial support, the survival of these neurons is dependent, in part, on the addition of neurotrophins (Lindholm et al., 1996). Screening of the 23 previously tested compounds in these cultures demonstrated activity of three of the compounds (LM11A-7, -24, and -31) previously identified as having activity in DRG cultures and identified a fourth active compound, LM11A-28 (structures in Fig. 2).
To begin to understand the mechanisms of action of these compounds and test the conjecture that they work via the targeted receptor p75NTR we examined the dose–response relationships of the survival-promoting activities of LM11A compounds compared with NGF and BDNF using embryonic hippocampal neurons in culture conditions in which NGF promotes neuronal survival. Similar cultures have been reported to express little or no TrkA (Ip et al., 1993; Marsh et al., 1993; Brann et al., 1999), but consideration of differences in species, timing, media, and the presence or absence of a glial feeder layer prompted examination of Trk expression in the cultures used in the present studies (Fig. 3A). Western analysis demonstrated abundant TrkB and no detectable TrkA expression. Although the expression of TrkA at trace levels cannot be entirely ruled out, these findings suggest that in these cultures, neurotrophic activity is mediated by BDNF principally through TrkB and p75NTR, whereas NGF functions primarily through p75NTR.
Addition of LM11A-7, -24, -28, and -31 increased the number of GAP-43-positive neurite-bearing cells, consistent with increased neuronal survival (Fig. 3B, photomicrographs). Dose–response profiles of the active compounds (Fig. 3C) demonstrated EC50 values in the range of 100–300 pm and intrinsic activities 80–100% of the NGF response. Maximal NGF and compound responses were less than that of BDNF. Although it seems likely this lower level of efficacy was caused by the simultaneous activation of TrkB and p75NTR within a cell by BDNF that does not occur with the other ligands, it remains possible that it is attributable to differential effects on neuronal subpopulations expressing different complements of receptors. Independently synthesized preparations of LM11A-24 and LM11A-31 gave similar results (data not shown). LM11A-36, a compound structurally similar to LM11A-24 (Fig. 2), showed little or no neurotrophic activity. Using TUNEL/DAPI staining, a similar profile of effects on preventing cell death were found for BDNF, NGF, and LM11A-24, -31, and -36 (Fig. 4A,B), thus confirming the findings of the morphological/MTT assays. Of note, in both assays, at compound concentrations >5 nm, survival response was reduced (Figs. 3C, 4B). This reduction could have resulted either from mechanisms unrelated to p75NTR or from modulation of receptor multimer formation and survival signaling at higher receptor occupancies, even by a compound without neurotrophic activity (e.g., LM11A-36). Neurotrophic responses similar to that of NGF in this hippocampal neuron system are consistent with activation of survival signaling through the NGF binding region of p75NTR.
Of the four active compounds initially identified, two (LM11A-24, a derivative of caffeine, and LM11A-31, an amino acid derivative), which were predicted to have “drug-like” character by Lipinski criteria (Lipinski, 2000) and blood–brain barrier calculations (Clark, 1999; Fu et al., 2001), were chosen for more detailed mechanistic study. We further prioritized LM11A-31, because preliminary studies (data not shown) indicated that it exhibits significant oral uptake and blood–brain barrier penetration. The relatively inactive compound LM11A-36 was chosen as a negative control because of its structural similarity to LM11A-24 (Fig. 2). The hypothesis that the LM11A compounds work through p75NTR was examined by investigating the influence of the compounds on (1) the binding of ligands or antibodies to p75NTR, (2) neuronal survival with blockade or in the absence of p75NTR, (3) maximal cell survival mediated by NGF, and (4) the activation of signaling systems associated with p75NTR-mediated survival. To further explore possible functional differences between the compounds and neurotrophins, we examined the effects of compounds in oligodendrocyte cultures, preparations in which neurotrophins cause, rather than prevent, cell death.
LM11A compounds interact with p75NTR but not TrkA
To assess the interactions of the LM11A compounds with neurotrophin receptors, we examined the effects of increasing concentrations of compounds on NGF binding to the recombinant chimeric proteins p75NTR-Fc and TrkA-Fc. These chimeric proteins contain the extracellular regions of the p75NTR and TrkA receptors and thus provide a useful tool for determining whether compounds affect neurotrophin binding. LM11A-31 (Fig. 5A) and LM11A-24 (Fig. 5B), but not LM11A-36 (Fig. 5C), shifted the NGF/p75NTR-Fc binding curve significantly to the right. The inhibition of NGF binding caused by each active compound was reversed with increasing NGF concentration, consistent with a mechanism that is, at least in part, competitive in nature. When these data were fit to the Gaddum–Schild equation that describes ligand binding in the presence of an inhibitor (Motulsky and Christopoulos, 2003; Neubig et al., 2003) (see Materials and Methods), the resulting Schild coefficients were significantly <1.0 for both active compounds (LM11A-31, 0.58 ± 0.04; LM11A-24, 0.26 ± 0.01), suggesting a more complex model (e.g., multiple sites within a ligand interacting with a receptor) (Lutz and Kenakin, 1999; Neubig et al., 2003). Gaddum–Schild analysis also yields a measure of potency that is independent of the mechanism of inhibition known as A2 (i.e., the concentration of compound that shifts the EC50 twofold to the right). The derivation of the Kd (that is, the equating of A2 with Kd) for an inhibitor in Gaddum–Schild analysis requires a competitive interaction between the antagonist and the ligand, as indicated by a Schild coefficient of 1.0. When this is not the case, A2 nonetheless remains a useful quantitative measure for comparing compounds within a given system and provides a basis for comparison with other parameters such as biologic potency. For example, LM11A-24 (A2, 31.6 + 1.3 nm) and LM11A-31 (A2, 1192 + 1.2 nm) have similar biologic potencies (Fig. 3) despite the ∼35-fold greater potency of LM11A-24 in the NGF displacement assay; this suggests that properties in addition to their ability to disrupt NGF binding contribute to the biological activity of each ligand. It could also suggest different mechanisms of action between the compounds.
Additional support for the interaction of LM11A-31 with p75NTR was obtained through the analysis of LM11A compound effects on Ab 9651 binding to p75NTR-expressing fibroblasts. Ab 9651 was raised against the extracellular cysteine-rich repeat regions II, III, and IV, implicated in neurotrophin binding (Huber and Chao, 1995). Ab 9651 binding to NIH3T3 cells was dependent on the expression of p75NTR (Fig. 5D). In p75NTR-expressing cells, coincubation of a high concentration of LM11A-31 (100 nm) with antibody, in the absence of NGF, at 4°C (which blocks receptor uptake) (Schlessinger et al., 1978), caused a significant decrease in antibody binding, whereas LM11A-36 (the analog inactive in binding or functional assays) had no effect (Fig. 5D). The specificity of LM11A compounds for p75NTR was supported by additional receptor binding studies showing that the active compounds had no effect on NGF binding to TrkA (Fig. 5E,F). Together, these results suggest that the compounds interact with the extracellular part of p75NTR in a way that can interfere with the function of the ligand-binding domain.
LM11A compound activity is p75NTR dependent
We then determined whether compound survival-promoting effects were dependent on the presence of p75NTR. Ab 9651, previously shown to block neurotrophic activity of NGF and NGF loop 1 peptide mimetics in mouse dorsal root ganglion neurons (Longo et al., 1997), partially blocked the neurotrophic activity of BDNF and NGF (Fig. 6A) and completely blocked the neurotrophic activity of LM11A-24 and LM11A-31, whereas nonimmune antibody had no effect. Another independently derived rabbit polyclonal anti-p75NTR antibody (Ab 9650) gave virtually identical results (data not shown), further corroborating the specificity of p75NTR blockade. Neither of the antibodies produced changes in baseline survival in these cultures. In addition, we examined the response of p75NTR −/− cells to neurotrophins and the LM11A compounds (Fig. 6B). Baseline survival under these culture conditions was similar in p75NTR+/+ and p75NTR−/− cells. p75NTR−/− cultures demonstrated decreased responsiveness to BDNF and the LM11A compounds in a pattern similar to that found in the p75NTR antibody studies, but differed in that the response to NGF was completely lost. These results are consistent with BDNF acting through both TrkB and p75NTR, whereas the LM11A compounds act primarily through p75NTR. The ability of p75NTR antibody to completely disrupt activity of LM11A compounds while only partially disrupting that of NGF may be a result of differences between NGF and compounds in ligand-receptor affinity or steric and/or allosteric mechanisms. Finally, treatment with 5 nm LM11A-24 or -31 along with 1.9 nm NGF, concentrations of each which induce a maximal response, produced no additive effect on survival (data not shown), further supporting the hypothesis that LM11A compounds act through p75NTR.
If the LM11A compounds work through interactions with p75NTR, we expect that they should affect its ability to signal. One of the most proximal NGF-induced effects on p75NTR associated with cell survival is the rapid recruitment to the receptor of IRAK (Mamidipudi et al., 2002, 2004). In PC12 cell cultures, addition of NGF has been shown to induce, within 1 min, the recruitment of IRAK, with subsequent downstream activation of NFκB (Mamidipudi et al., 2002). We found that LM11A-24 and -31, but not LM11A-36, induced recruitment of IRAK to p75NTR in PC12 cells (Fig. 7A).
Despite the observation that LM11A-24 and LM11A-31 did not affect NGF-TrkA/Fc binding, it remained possible that LM11A compounds might activate TrkB, the principal Trk expressed by these hippocampal neurons, or the nominally expressed TrkA, as their primary mechanism for promoting survival. Also, ligand binding to p75NTR might influence Trk activation. With these considerations, it was of interest to determine whether LM11A compounds would promote Trk activation. LM11A-24 and LM11A-31 were assessed for their ability to activate Trk autophosphorylation, as indicated by TrkAY490 (or the analogous TrkBY515) phosphorylation, a well established marker of Trk activation. In hippocampal cultures, BDNF exposure resulted in robust Trk activation (Fig. 7B), whereas no activation could be detected with NGF or LM11A compounds. The lack of signal with NGF confirms that these cultures produce little or no TrkA and supports the idea that the trophic effects of NGF are mediated principally by p75NTR. In 3T3–TrkA cells, NGF exposure produced the expected TrkA autophosphorylation response, whereas the LM11A compounds again showed no activity (Fig. 7C). These results suggest that activation of Trk receptors does not play a primary role in promotion of cell survival by LM11A compounds. In addition, they argue against a mechanism in which the compounds produce their neurotrophic effects by inducing secretion of endogenous neurotrophins. In particular, if compounds induced BDNF secretion, we would expect to detect activation of TrkB in compound-treated hippocampal cultures, which was not observed (Fig. 7B,C). The possibility that the compounds selectively induce secretion of other factors remains although seems unlikely.
Signaling intermediates activated by LM11A compounds
Prosurvival signaling pathways with well established associations with p75NTR actions include activation of NFκB (Carter et al., 1996; Foehr et al., 2000; Gentry et al., 2000; Mamidipudi et al., 2002; Reddypalli et al., 2005) and PI3K/AKT (Roux et al., 2001; Lachyankar et al., 2003; Reddypalli et al., 2005). These signaling intermediates have been shown to be capable of being independently regulated by Trk and p75NTR through pathways with varying degrees of overlap and cross-talk and with different kinetics. NFκB activation was implicated in the survival-promoting activities of the LM11A compounds by the finding that application of an inhibitor of NFκB translocation, SN50 (Lin et al., 1995), at a relatively low concentration targeted to minimize non-NFκB interactions (Kolenko et al., 1999) significantly reduced cell survival promoted by both LM11A compounds and the neurotrophins, with no effect on baseline survival (Fig. 8A). Trichodion, a compound that inhibits NFκB activation through the stabilization of IκB (Erkel, 2000), gave nearly identical results (data not shown). The PI3 kinase inhibitor LY294002, which inhibits AKT activation, markedly decreased survival in all cases, including baseline survival under conditions of no treatment or exposure to LM11A-36 (Fig. 8A). In addition, the ERK (extracellular signal-regulated kinase) inhibitor PD98059 significantly decreased BDNF-stimulated survival and had a small but significant effect on NGF activity but produced no significant decrement in survival promoted by either LM11A-24 or LM11A-31 (Fig. 8A). Although these inhibitors are not completely specific, these studies suggested that ERKs were likely not significantly involved in LM11A compound activity, although the possibility of their action through the PI3K/AKT and NFκB pathways warranted further characterization.
Consistent with the recruitment of IRAK to p75NTR induced by the compounds (Fig. 7A) and the effect of SN50, treatment of hippocampal neurons with 20 nm LM11A-24 or LM11A-31 (a concentration in the plateau range for acute signaling activation; see below) led to an ∼1.5-fold increase in NFκB-p65 phosphorylation (Fig. 8B), indicative of activation of the NFκB pathway (Sakurai et al., 1999), similar in extent and time course to that induced by both neurotrophin proteins. LM11A-36 at the same concentration did not induce NFκB-p65 phosphorylation. Examination of compound dose-activation response (Fig. 8C) demonstrated that LM11A-31 induced activation of NFκB over a concentration range of 0.5–3 nm, which is similar to that required for promotion of survival and concordant with a role for this signaling mechanism in mediating compound-induced survival. In studies of AKT activation, LM11A-24 and -31 at 20 nm showed similar degrees of activation to that found with NGF at 30 min, whereas BDNF induced a substantially greater response. In addition, the onset of activation stimulated by LM11A-24 and -31 was slower than that of NGF (Fig. 8D). Similar to the observations of NFκB, the half-maximal response of AKT activation to LM11A-31 was at ∼1 nm (Fig. 8E). Finally, we found that NGF and LM11A compound failed to activate AKT in cultures of p75NTR−/− neurons (Fig. 8F). Together with evidence for the p75NTR dependence of compound-induced survival (Fig. 6A,B), these findings suggest that LM11A compounds induce survival of hippocampal neurons in culture, at least in part, through interactions with p75NTR that produce activation of survival-promoting signaling pathways involving AKT and NFκB. The possibility that LM11A compounds promote survival by blocking death-promoting activity of endogenous proneurotrophins that might be present in hippocampal cultures is made less likely by the finding that baseline survival of hippocampal neurons was not affected by the absence of p75NTR (Fig. 6B).
LM11A-31 and LM11A-24 do not promote death of mature oligodendrocytes but inhibit proNGF-mediated death
Although NGF and the LM11A compounds promoted survival in the hippocampal cultures used in these studies, liganding of p75NTR by mature NGF or by proNGF has been associated with cell death rather than promotion of survival in certain cell types (Casaccia-Bonnefil et al., 1996; Lee et al., 2001). To determine whether LM11A compounds would promote survival or cause death in systems in which neurotrophins promote cell death, we examined the survival of mature oligodendrocytes treated with LM11A compounds and proNGF. Mature oligodendrocytes express p75NTR but not TrkA and undergo apoptotic death on treatment with NGF or proNGF (Casaccia-Bonnefil et al., 1996; Yoon et al., 1998; Beattie et al., 2002). Unlike NGF or proNGF, LM11A-24, -31, and -36 alone did not promote cell death (Fig. 9A). In addition, proNGF-induced cell death was significantly inhibited by LM11A-24 and -31 over a concentration range of 1–10 nm, but not by LM11A-36, which appeared to decrease survival somewhat at 10 nm (Fig. 9A).
To determine whether LM11A compounds block proNGF binding to p75NTR, proNGF binding to p75NTR was assessed over a concentration range of 1500–10,000 nm. LM11A-24 and -31 inhibited proNGF binding equally, up to a ∼30% decrement at the highest concentration (Fig. 9B).
Using knowledge of NGF loop 1 peptide activities (Longo et al., 1990, 1997), along with available sequence, mutation analysis, and x-ray crystallographic structure, we developed a pharmacophoric virtual screening method that efficiently identified [four active of 23 tested compounds (17%)] several chemically diverse small molecules with potent neurotrophic activity that interact with p75NTR. Although attempts to identify nonpeptide small molecule inhibitors of protein–protein interactions have recently been fruitful (Berg, 2003), there are relatively few reports of small molecules, interacting with receptors whose principal ligands are proteins, that achieve one or more ligand-like effects. Activators of receptors for thrombopoietin (c-Mpl) (Kimura et al., 1998), erythropoietin (Goldberg et al., 2002), insulin (Zhang et al., 1999), and the granulocyte colony stimulating factor (Tian et al., 1998) were identified by empiric, high throughput screening. In addition, Maliartchouk et al. (2000a) and Zaccaro et al. (2005) have developed low-molecular-weight TrkA agonists through progressive modification of small cyclic peptides emulating a β-turn. Virtual (also called in silico) screening of chemical libraries based on the structure of protein–protein interaction sites, ligand–receptor interaction sites, or a series of known small ligands is a potentially powerful approach for the identification of novel lead compounds (for review, see Kurogi and Guner, 2001; Mason et al., 2001; Guner, 2002; Singh et al., 2002; Guner et al., 2004). To date, in the area of modulation of protein–protein interactions, virtual screening has yielded only inhibitors, including compounds inhibiting the interactions of Bak with Bcl-2 (J. L. Wang et al., 2000), Bak with Bcl-x (Enyedy et al., 2001), major histocompatibility complex class II proteins with CD4 (Li et al., 1997), and very late antigen-4 with vascular cell adhesion molecule-1 (Singh et al., 2002). As in the present work, each of these screening studies was predicated on identification of an oligopeptide with the desired function. The results reported here differ from all previous studies in that we have identified compounds with agonist-like activity. Although binding anywhere within a pocket may lead to inhibition, it seems likely the structural requirements for emulating a functional protein ligand–receptor interaction will generally be more stringent.
Several lines of evidence support the hypothesis that LM11A compounds work principally through direct interactions with p75NTR: (1) active compounds fail to elicit prosurvival activity and AKT signaling in hippocampal neurons from p75NTR−/− mice; (2) active compounds appear to interact with p75NTR, as demonstrated by their ability to displace NGF from p75NTR-Fc but not from TrkA-Fc; and (3) Ab 9651, an antibody targeted to the neurotrophin binding domain of p75NTR, inhibits compound prosurvival effects and is displaced from cells by active LM11A compounds. Additional support for this contention is provided by the findings that active compounds induce rapid recruitment of IRAK to p75NTR, that they do not activate Trks, and that at maximum doses, NGF and active compounds have no additive effects. Thus, p75NTR appears to be necessary for LM11A compound activity; however, it remains possible that the compounds also work in some measure through low-level or alternative modes of Trk activation, or through other mechanisms that interact with p75NTR. Finally, it was of interest to note that although they are ostensibly related only by the property of containing a pharmacophore designed to mimic the NGF loop 1 p75NTR binding domain, the otherwise chemically diverse LM11A compounds have similar neurotrophic activities and potencies. This observation is also consistent with the possibility that these compounds act through p75NTR.
The apparent monomeric nature of these compounds is consistent with our preliminary computational studies suggesting that peptide dimerization might promote the proper folding of only one of the peptide subunits. Furthermore, these findings suggest that cyclic monomeric peptides (Longo et al., 1997) may be inactive as a result of an inability to maintain a proper binding conformation. The subsequent finding that NGF may be disallowed from simultaneous interaction with two p75NTR molecules (He and Garcia, 2004) is consistent with this possibility as well.
The hypothesis that compound activity occurs through binding to p75NTR raises the question of whether that mechanism is consistent with the finding that a measure of the ligand-displacement potency, A2, of LM11A-31 is nearly four orders of magnitude greater than the EC50 for its biologic effect (compare Figs. 3C, 5A). We suggest that this finding is not unexpected, because large differences between biologic potency of small molecules and apparent binding affinity, as estimated by ligand displacement, are common (Lutz and Kenakin, 1999). Potential causes of these differences include: differences between receptor states in binding versus functional assays; postreceptor signal amplification, such that maximal biologic effects are seen at low receptor occupancies (Lutz and Kenakin, 1999); and partial displacement of a macromolecular ligand by a smaller antagonist. In Gaddum–Schild analysis, a Schild coefficient different from 1.0 suggests a noncompetitive interaction between the ligand and the antagonist; this may occur if the mechanism of antagonism is through allosteric effects, or if there are multiple ligand binding sites available on which the antagonist has differential effects. The latter situation can occur if there are multiple distinct receptors for the ligand, or, as we propose for the present case, the ligand binds at several sites on the same receptor. The current data are consistent with a model in which the compounds are relatively ineffective in displacing NGF from p75NTR because they interfere at only one of multiple sites of interaction between the neurotrophin and receptor. This mechanism is also consistent with recently reported crystallographic structure (He and Garcia, 2004) of a complex between NGF and p75NTR, suggesting that there are at least two interacting patches, with one centered on NGF loop 1. A purely allosteric model seems less likely (although it is not ruled out), given the virtually complete reversal of the compound binding at high levels of NGF. The incomplete displacement of NGF by the compounds is similar to that observed in the analysis of small chemokine receptor antagonist binding (Sabroe et al., 2000).
The high concentrations required to inhibit proNGF binding compared with those blocking proNGF-induced death of oligodendrocytes suggest the possibilities that: (1) in the cell-based assay, with native receptor conformation in the presence of coreceptors (e.g., sortilin), the proNGF-p75NTR interaction may be more susceptible than in the in vitro p75NTR-Fc assay to disruption by the compounds; (2) at low concentrations, the compounds qualitatively alter proNGF binding to decrease the induction of cell death but do not decrease the total amount of proNGF binding; or (3) that the compounds induce preferential activation of prosurvival signaling by p75NTR without affecting proNGF binding. Preferential survival pathway activation by LM11A compounds could result from differences in the ways the compounds modulate receptor structure, as well as lack of binding to the sortilin coreceptor. Indeed, previous studies suggest that engagement of both sortilin and p75NTR by proNGF promotes efficient ligand binding, receptor complex activation, and apoptotic actions (Nykjaer et al., 2005). Future analyses of signaling in these cells in response to compounds in the presence or absence of proNGF may help to further elucidate mechanisms by which compounds block proNGF-induced oligodendrocyte death.
The potential for differences between activities generated by a native ligand and small molecule mimetics of a specific domain of that ligand has important therapeutic implications. By targeting only one member of a group of receptors that interact with a given ligand, we may expect to generate compounds that activate a pattern of receptor-mediated effects that may or may not occur naturally. In the present case, in targeting p75NTR, we seem not to activate Trk, but likely activate pathways involving NFκB and PI3K. Such differences in signaling patterns may eventually prove clinically useful. For example, the LM11A compounds may promote survival under conditions in which neurotrophins promote death (as in oligodendrocyte death) and may be less likely to induce excessive sympathetic fiber sprouting and upregulation of pain transmission occurring, likely via Trk signaling, with neurotrophin treatments (Bergmann et al., 1998; Walsh et al., 1999).
Future systematic studies of these compounds with respect to the structural details of the binding interface, kinetics of activation, and pathways involved should enhance our general understanding of neurotrophin action and promote further development of specific neurotrophin system modulators for therapeutic use. Recent evidence suggests that p75NTR, alone or interacting with other proteins such as proNGF, might mediate cell death in various neuropathological conditions including Alzheimer’s disease, spinal cord trauma, axotomy, and retinal dystrophy (Nykjaer et al., 2005). This raises the possibility that LM11A compounds will constitute important leads for therapeutic development. In current studies in our laboratories, the structural diversity of active LM11A compounds has provided a basis for synthesizing novel compounds designed to incorporate optimal medicinal chemistry properties.
This work was supported by the Institute for the Study on Aging (F.M.L.), the Alzheimer’s Association (F.M.L.), the French Foundation (F.M.L.), National Institutes of Health Grants NS30687 (B.L.H.) and NS39472 (S.O.Y.), and the Veteran’s Administration (S.M.M.). We thank Dr. Saunak Sen (Research Enhancement Award Program Data Analytic Laboratory, San Francisco Veterans Affairs Medical Center and Department of Epidemiology and Biostatistics, University of California, San Francisco) for advice on statistical analysis and Dr. Richard Mailman (Departments of Psychiatry, Pharmacology, and Neurology, University of North Carolina) for critical review of this manuscript.
*S.M.M. and Y.X. contributed equally to this work.
- ↵Correspondence should be addressed to either of the following: Dr. Frank M. Longo, Department of Neurology and Neurological Sciences A343, Stanford University, 300 Pasteur Drive, Stanford, CA 94305, Email: or Dr. Stephen M. Massa, Department of Neurology (127), San Francisco Veterans Affairs Medical Center, 4150 Clement Street, San Francisco, CA 94121,