The Paradoxical Signals of Two TrkC Receptor Isoforms Supports a Rationale for Novel Therapeutic Strategies in ALS

Full length TrkC (TrkC-FL) is a receptor tyrosine kinase whose mRNA can be spliced to a truncated TrkC.T1 isoform lacking the kinase domain. Neurotrophin-3 (NT-3) activates TrkC-FL to maintain motor neuron health and function and TrkC.T1 to produce neurotoxic TNF-α; hence resulting in opposing pathways. In mouse and human ALS spinal cord, the reduction of miR-128 that destabilizes TrkC.T1 mRNA results in up-regulated TrkC.T1 and TNF-α in astrocytes. We exploited conformational differences to develop an agonistic mAb 2B7 that selectively activates TrkC-FL, to circumvent TrkC.T1 activation. In mouse ALS, 2B7 activates spinal cord TrkC-FL signals, improves spinal cord motor neuron phenotype and function, and significantly prolongs life-span. Our results elucidate biological paradoxes of receptor isoforms and their role in disease progression, validate the concept of selectively targeting conformational epitopes in naturally occurring isoforms, and may guide the development of pro-neuroprotective (TrkC-FL) and anti-neurotoxic (TrkC.T1) therapeutic strategies.

Motor neurons degenerate in Amyotrophic lateral sclerosis (ALS), spinal muscular atrophy (SMA), and spinal cord injury (SCI) [4], and neurotrophic strategies that delay or prevent motor neuron death and rescue motor neuron function may be beneficial. Experimental ALS therapy required very high neurotrophin protein concentrations by intrathecal injections or expression by viral vectors [5] or a combination of at least two growth factors [6][7][8].
While the rationale of using BDNF and NT-3 as drugs for ALS seems strong, they have failed clinically [3]. There are several reasons postulated for the failure. One problem is that neurotrophins have a short half-life in vivo and may not reach the target tissue. A second problem relates to the vast expanse of tissue that needs to be reached, from peripheral nerve terminals to spinal cord, because TrkB and TrkC targets are distributed throughout the motor neuron. Activation of receptors at the neuromuscular junction in the periphery versus receptors at the neuronal cell body in the spinal cord result in transduction of different signals (neuritogenic versus neurotrophic pathways respectively) [9]. A neuroprotective therapy would require activation of both receptors pools.
A third problem is that NT-3 not only binds to pro-survival receptor full length TrkC (TrkC-FL), but also binds to receptors that have pro-inflammatory or neurodegenerative functions: p75 NTR and truncated isoforms of TrkC (TrkC.T1). The unintended p75 NTR target is expressed in most neurons, glia, and many other cell types. The p75 NTR functions to execute the axonal pruning and cell death required during embryonic development [10], and it is upregulated in ALS [11]. When neurotrophins are used as drugs it is essential to circumvent activation of p75 NTR , and this has been achieved in vivo using neurotrophin mutants or p75-blocking strategies [12][13][14]. The unintended TrkC.T1 truncated isoform target is an mRNA splice variant that lacks the kinase intracellular domain, but retains the ectodomain and transmembrane primary sequence identical to TrkC-FL [12]. While TrkC-FL has trophic activity, multiple lines of evidence (biochemical, cell biological, genetic, and pharmacological) indicate that TrkC.T1 activates Rac1 and can be deleterious to neurons [15,16]. However, discriminating between TrkC-FL and TrkC.T1 activity is difficult because NT-3 binds to both isoforms with equal affinities [17].
To circumvent the problems associated with the use of NT-3 as a therapeutic (binding p75 NTR and TrkC.T1, short half-life, not reaching all receptor pools), we used antibody-based ligands [3,[17][18][19] to develop a selective agonist that can discriminate between TrkC-FL and TrkC.T1 isoforms even though these receptors have identical ectodomain primary sequences. Selectivity was achieved by targeting disulfide-stabilized ectodomain secondary structures in TrkC-FL, structures which are absent in TrkC.T1 due to the influence of an intracellular neoepitope of TrkC.T1.
We also demonstrate that TrkC.T1 is up-regulated in mouse and human ALS, due to decreased miR-128, a miR that destabilizes TrkC.T1 mRNA. TrkC.T1 protein and mRNA are present in activated spinal cord astrocytes, and we show ex vivo that in an NT-3-dependent manner TrkC.T1 mediates up-regulation of TNF-α. Therefore, using the TrkC-FL selective agonist, and validate ligand-dependent activation of this target for therapy in a mouse model ALS.
Together, these data elucidate biological paradoxes of growth factor receptors and their isoforms in disease progression, validate the concept of selective targeting of conformational epitopes present in naturally occurring receptor isoforms of identical primary sequence, and validate selective activation of TrkC-FL as a therapeutic target in model of ALS. HEK293-TrkC-FL or HEK293-TrkC.T1 cells were studied by FACScan. HEK293 wild type cells are negative control. MAb 2B7 (red histograms) binds to cell surface TrkC-FL protein, at mean channel fluorescence~300. Several isotype-matched controls are shown for background, all at mean channel fluorescence~10. No significant binding to the cell surface was detected using mAb 2B7 on NIH-TrkC.T1 cells, mean channel fluorescence~15. (B) Non-reducing western blots of HEK293-TrkC-FL or HEK293-TrkC.T1 cells. MAb 2B7 only recognizes lysates form TrkC-FL cells. A control antibody 750 (against an intracellular neo-epitope that appears due to mRNA splicing) only recognizes TrkC. T1 and demonstrates that the cells express TrkC.T1 protein. (C) Fluorescent microscopy of 293-TrkC-FL or 293-TrkC.T1 cells immunostained with primary mAb 2B7, or antibody 750, or with non-binding mouse IgG control (red). DAPI counterstain for nuclei. Pictures were taken using a Leica confocal microscope with 63x magnification. MAb 2B7 only recognizes full length TrkC, whereas antibody 750 only recognizes TrkC.T1. (D) Lysates were prepared from HEK293-TrkC-FL or HEK293-TrkC.T1 cells and split as equal samples. For surface was measured on live HEK293 cells stably transfected to express equal numbers of either TrkC-FL or TrkC.T1 (cDNAs were expressed so there is no possible mRNA truncation). 2B7 binding to TrkC-FL and not to TrkC.T1 was evaluated after transfection of human and rat cDNAs, with similar results.
These FACScan studies are quantitative, as they use a homogeneous population expressing a single receptor isoform, they yield single bell-shaped histograms that can detect~1,000 receptors/cell, with data from 5,000 to 10,000 cells acquired per experiment (n>10 independent experiments), and in each independent experiment the TrkC-FL signal shifted >25-fold above background and the TrkC.T1 signal shifted <2-fold above background).
FACScan studies using rat rMC-1 cells endogenously expressing TrkC.T1 mRNA and protein showed undetectable binding by mAb 2B7, and previous studies demonstrated 2B7 binding to TrkC-FL endogenously expressed in rodent primary neuronal cultures [3]. The 2B7 epitope is at the juxtamembrane in the ectodomain, and both TrkC-FL and TrkC.T1 have the primary sequence of the epitope (Fig 1B). Previously, we showed that in western blots mAb 2B7 bound to TrkC-FL under non-reducing conditions but not under fully reducing conditions [21,22]. Here, we show in western blot studies using non-reduced samples that mAb 2B7 binds to 293-TrkC-FL whole cell lysates but it does not bind to 293-TrkC.T1 whole cell lysates. In contrast, a control 750 antibody against the TrkC.T1 intracellular neo-epitope binds to 293-TrkC.T1 whole cell lysates but not 293-TrkC-FL whole cell lysates (Fig 1B). In immunofluorescence studies mAb 2B7 binds to 293-TrkC-FL and antibody 750 binds to 293-TrkC.T1 (Fig 1C) without significant cross-reactivity. These studies were replicated in cells expressing TrkC-FL or TrkC.T1 after transfection with human, mouse or rat cDNAs, with identical results (data not shown).
Together, these data indicate that mAb 2B7 binds specifically to the ectodomain of cell surface TrkC-FL from mouse, rat, and human species, but it does not bind to any TrkC.T1 species whether it is expressed after transfection or expressed endogenously.

Receptor conformation determines ligand selectivity
MAb 2B7 binding is sensitive to full reduction of TrkC-FL (75 mM DTT, 5 min 100°C), suggesting that a disulfide bond-stabilized secondary structure is required for proper display of the epitope [23]. Both TrkC-FL and TrkC.T1 proteins have disulfide bonds. TrkC-FL migrates in SDS-PAGE as M r~1 5 kDa smaller than when fully reduced (M r~1 20 kDa and~135 kDa bands respectively), and TrkC.T1 migrates in SDS-PAGE as M r~2 5 kDa smaller than when fully reduced (M r~1 00 kDa and~125 kDa bands respectively), as detected using antibody C44H5 that binds to all TrkC isoforms (Fig 1D). TrkC.T1 migrates faster due to the truncation, and both TrkC-FL and TrkC.T1 migrate more slowly in their reduced state due to the "relaxed" state of the proteins. The minor bands detected by C44H5 likely correspond to differentially glycosylated receptors [23].
Both TrkC-FL and TrkC.T1 were subjected to mild to intermediate reducing conditions (0, 0.1, 1.0 and 10 mM DTT ± 1 min 90°C). These conditions preferentially reduce disulfide bonds exposed on the protein surface. MAb 2B7 detects non-reduced TrkC-FL as a single band of M r 120 kDa (Fig 1E lane 1), and mild reduction generates a "relaxed" protein M r~1 25 kDa, and intermediate reduction yields a M r~1 35 kDa band (Fig 1E, lanes 2-5). These mild/intermediate reducing conditions do not affect the mAb 2B7 epitope in TrkC-FL, and therefore could be used to study "gain of epitope" in TrkC.T1. MAb 2B7 does not detect non-reduced TrkC.T1 (Fig 1E lane 6). Mild and intermediate reduction affords progressively improved detection of TrkC.T1 by mAb 2B7 (Fig 1E, lanes 7-10).
Together, the data show that 2B7 selectively binds to TrkC-FL, at an epitope sensitive to disulfide bond arrangements which differ between TrkC-FL and TrkC.T1, and which can be resolved by mild reduction of exposed Cys-Cys bonds at the protein surface.
A differential folding observed for TrkC-FL and TrkC.T1 (as defined by mAb 2B7 binding) suggest that either the loss of the intracellular TrkC-FL kinase domain or the gain of the 84 amino acid neo-epitope in the TrkC.T1 cytoplasmic tail affects how each protein is processed. These possibilities were evaluated (Fig 2) by expression of a TrkC.T1 cDNA construct where the 84 amino acid neo-epitope was deleted (TrkC.T1 Δinsert ). Expression of TrkC.T1 Δinsert resulted in proteins recognized by mAb 2B7 at the cell surface in FACScan assays (Fig 2A) and in non-reducing western blots (Fig 2B).
Together, these data demonstrate that a naturally occurring truncated isoform TrkC.T1 containing an intracellular neo-epitope has a new and different folding/conformation than TrkC-FL, not recognized by mAb 2B7. Deletion of the intracellular neo-epitope generates a structure that is recognized by mAb 2B7, like the TrkC-FL folding/conformation.
The three species we studied (human, rat and mouse) preserve the mRNA splicing that generates TrkC.T1, the neo-epitope of the intracellular domain of TrkC.T1, and the specific conformation of the ectodomain of each isoform (as defined by mAb 2B7). These data suggest an important function for TrkC.T1. The TrkC.T1 isoform transduces signals [23] linked to neurodegeneration [15], but its mechanism of action remained unclear. Given that NT-3 binds equally to TrkC-FL and to TrkC.T1, while mAb 2B7 is a selective ligand of TrkC-FL, we carried out functional studies to test the ligand-dependent mechanisms of action.
TrkC.T1 induces TNF-α mRNA and protein in a ligand-dependent manner A rat glial cell line (rMC-1) endogenously expressing TrkC.T1 was studied after treatment with TrkC agonists. NT-3 significantly increased the secreted form of TNF-α (Fig 3A, ELISA of conditioned media, p<0.004 versus untreated cells), and increased the cell-bound form of TNF-α (Fig 3B, cell immunofluorescence). NT-3 up-regulated TNF-α to levels similar to positive control LPS (a ligand of Toll-like receptors), and to levels higher than positive proNGF, a control p75 NTR agonist known to increase TNF-α [23]. Several doses of mAb 2B7 did not increase TNF-α in conditioned media (Fig 3A) or in the cellular biomass (Fig 3B). The failure of mAb 2B7 to induce TNF-α via TrkC.T1 is consistent with mAb 2B7 not binding to TrkC.T1. Note that a full range of concentrations were tested, but to simplify the graphics only the optimal or biologically active concentrations of ligands (e.g. 100 nM mAb 2B7 verified to act as agonist of TrkC-FL) are shown.
Together, the data indicate that TrkC.T1 regulates production of TNF-α in an agonistdependent manner, and this could be a confounding factor when using NT-3 for ALS therapy. Therefore we studied the expression and the role of each TrkC isoform in ALS.
TrkC.T1 up-regulation in spinal cord astrocytes in a mouse ALS model and human sporadic ALS is due to miR128 downregulation In the spinal cord of symptomatic G93A ALS mice TrkC.T1 mRNA expression is significantly elevated, compared to wild type mice (p< 0.04). In contrast, expression of full-length TrkC (TrkC-FL) mRNA does not differ between ALS mice and wild type mice (Fig 4A). Standardization versus β-tubulin (a neuronal marker, to account for neuronal loss), or versus GAPDH (a ubiquitous marker) yields comparable results.
TrkC-FL and TrkC.T1 have non-overlapping 3'UTRs, and they are differentially regulated at the post-transcriptional level. The spliced TrkC.T1 mRNA is destabilized by micro-RNA miR128 [24]. In healthy wild type spinal cord miR128 is expressed. Pre-symptomatic ALS spinal cords (~100 days of age) have lower miR128, and symptomatic ALS spinal cords (~140 days of age) have even lower miR128 levels (p<0.003). In contrast, miR-151-3p that regulates full length TrkC-FL mRNA remains unchanged in wild type versus ALS mice ( Fig  4B) and this miRNA serves as internal control. A miR128-promoted degradation would explain why healthy spinal cords have low or undetectable TrkC.T1 mRNA and protein, whereas reduced miR128 levels would lead to increased TrkC.T1 mRNA in the mouse model of ALS.
Similar results were obtained in spinal cords obtained post-mortem from human sporadic ALS (SALS, n = 4) versus non-ALS control (n = 3). In SALS there is a significant 2.84 ± 0.35 fold-increase in TrkC.T1 mRNA and a small non-significant elevation of TrkC-FL mRNA (standardized versus β-tubulin) (Fig 4C). Elevated TrkC.T1 mRNA in human SALS is associated with significantly reduced levels of miR128 (a known disruptor of TrkC.T1 mRNA) versus non-ALS control (Fig 4D) normalized to the reference RNU6 small nuclear RNA. In human SALS there was also a reduction in miR-151-3p (a known disruptor of TrkC-FL) versus non-ALS control and this resulted in a small and non-significant elevation of TrkC-FL mRNA.
The cells that express TrkC.T1 protein and mRNA in spinal cord were identified. Immunohistochemical studies using a specific TrkC.T1 antibody 750, showed expression in GFAP + cells (astrocytes) of symptomatic ALS mice (Fig 4E), but not in Iba-1 + cells (microglia) or NeuN + cells (neurons) (data not shown). In control wild type healthy spinal cords TrkC.T1 protein is virtually undetectable (Fig 4E), consistent with the very low number of GFAP + activated astrocytes. These data were confirmed by in situ mRNA hybridization combined with immunohistochemical studies in spinal cord of symptomatic ALS mice (Fig 4F). The GFAP + phenotype and the morphology of TrkC.T1 mRNA-expressing cells indicate that they are activated astrocytes.
Together, the data demonstrate that elevated TrkC.T1 mRNA and reduced miR128 are detected in the mutant SOD1 mouse model, and also in humans with ALS unrelated to SOD1 mutations (which represent the majority of clinical cases). In ALS, up-regulation of TrkC.T1 mRNA and protein occurs in astrocytes rather than neurons. The neurotoxic nature of elevated TNF-α in ALS has been documented, and our data provide a novel degenerative mechanism where NT-3 acting through TrkC.T1 can drive production of TNF-α and cause neurotoxicity. Previous work using neuronal cell lines and primary neuronal cultures showed that mAb 2B7 and 2B7 Fabs activate endogenously expressed TrkC-FL [17]. Here we show a detailed study of the kinetics of activation by these agonists. Biological (cell survival assays, Fig 5A) and biochemical (signal transduction assays, Fig 5B) show comparable activation by control NT-3, mAb 2B7, and 2B7 Fabs, which is sustained for up to 4 hr. In Fig 5B some of the exposures are not linear, this was intentional to show the longer time kinetics of some ligands. Quantifications were done from properly exposed gels and show that mAb 2B7 or 2B7 Fabs sustain the p-AKT signal significantly longer (p< 0.0002) than NT-3, with the signal persisting >24 hours (quantification in Fig 5C, 5D and 5E). In cellular controls, the mAb 2B7 and its Fabs do not activate signals in cells transfected to express a receptor family member TrkA or TrkB (data not shown).
Sustained activity is a known characteristic of the Trk receptor family and it is required for neuronal survival [25,26].

2B7 and 2B7 Fabs are stable in vivo
The poor specificity of NT-3 (binding to multiple receptors TrkC-FL, p75 NTR , TrkC.T1), poor pharmacokinetics (circulation half-life under 2 minutes), and the inability to reach the different pools of receptors make NT-3 a poor therapeutic agent for ALS. On the other hand, mAb 2B7 has high TrkC-FL selectivity, does not bind to TrkC.T1 or p75 NTR , and it does not induce increased TNF-α. Moreover, mAbs are known to be relatively stable in circulation.
For a pharmacokinetic study, 2B7 or 2B7 Fabs were administered intraperitoneally, and serum was prepared from blood samples collected over time. The presence of 2B7 or 2B7 Fabs in serum was quantified by ELISA against the original peptide epitope, as described [27,28]. Extrapolation from binding curves indicate that the half-lives of circulating and bioavailable 2B7 Fab and 2B7 mAb that are able to bind to the target were respectively~36 hours and~80 hours.

2B7 is present in spinal cord neurons and activates pro-survival signals in spinal cord in vivo
We studied whether 2B7 administered at the periphery can reach the motor neuron cell body in the spinal cord, a key feature for neurotrophic activity [29]. Labeled 2B7 was administered intraperitoneally to wild type mice, and after 48 hours spinal cords were studied for the presence of labeled 2B7. The 2B7 signal was found in the cell bodies of putative ventral motor neurons of lumbar spinal cord (Fig 6A), and there was no label in any other cells within the spinal cord.
Then, we asked whether activation of TrkC-FL signals is detected in spinal cord after 2B7 treatment. Wild type mice received single intraperitoneal injection of 2B7 Fab or control (~30 μg/mouse), and~18 hours later the mice were saline-perfused, spinal cord tissues were collected, and activation of trophic signals were quantified by western blots (n = 3 mice per group). From densitometric quantification, the data shows that 2B7 treatment activated a sig-nificant~35% increase in p-AKT and a significant~50% increase in p-PLCγ compared to control (n = 3) or SALS (n = 4). Data are expressed as the mean + SEM. (F) From the same human samples the levels of miR-128 (regulator of TrkC.T1) and miR-151-3p (regulator of TrkC-FL) were measured by SYBR Green qRT-PCR (Quanta) and results were normalized to the reference RNU6 (small nuclear RNA). Data are expressed as the mean + SEM. (A-F) t-test was applied for statistical analysis. * p<0.05, ** p< 0.01 versus control. doi:10.1371/journal.pone.0162307.g004 Selective TrkC-FL Agonist as Therapy for Motor Neuron Dystrophy control treatment (Fig 6B). Note that these studies were done in freely moving wild type healthy mice, which have normal motor neuron activity. Therefore the significant increases reported are above normal signals.
To complement this study in wild type mice, a study evaluated activation of signals in the spinal cord of ALS mice at post-natal day 102 (symptomatic, n = 5 mice per group). Data shows activated p-Akt and p-MAPK in 2B7-treated mice, when compared to control IgGtreated mice (Fig 6C) at~18 hours post-injection. From these 2B7-treated or control-treated ALS spinal cords, cryosections were prepared for immunohistochemical studies with anti-p-Akt and anti-p-MAPK. The representative fluorescent microscopy data shows that 2B7 activates (i) p-Akt in neuronal cell bodies in the gray matter (co-localized with the NeuN marker) and in neuronal fibers of the white matter; and (ii) p-MAPK almost exclusively in neuronal cell bodies, but not in neuronal fibers of the white matter. In contrast, control-treated ALS mice studied in parallel had lower or absent neuronal Akt and MAPK activation (Fig 6D).
Taken together, these data demonstrate that 2B7 treatment activates trophic signals significantly above normal homeostatic levels in healthy wild type mice and in spinal cord neurons of symptomatic ALS mice. This suggests that therapy may be initiated after the onset of symptoms. Moreover, 2B7 treatment benefits from the key features of a relatively simple systemic administration.

Criteria for therapeutic dose
A dose and frequency for a therapeutic study in ALS mice was estimated from the pharmacokinetic study. A relatively low dose and frequency of drug were chosen from the pharmacokinetic data. The rationale was based on the following reasons. First, the schedule has to be acceptable for pharmacological treatment of chronic diseases such as ALS. Second, biochemical studies in cultured cells confirmed that TrkC activation by the agonists was long-lived. Third, biochemical studies in vivo showed that within 18 hours of injection there was activated TrkC in the spinal cord, and imaging studies showed that at 48 hours after injection the reagents were still present in spinal cord neuronal cell bodies. Fourth, a receptor agonist should not require saturation of all the receptor sites to activate signals, and a low level of agonist in the circulation could activate TrkC without causing receptor desensitization or down-regulation (which would be counterproductive). Fifth, pharmacokinetic experiments showed reasonable stability in circulation after intraperitoneal injection.
2B7 Fabs extend life span, improve motor neuron performance, body weight, and motor neuron phenotype in a mouse model of ALS.
A 0.5 mg/kg dose of 2B7 Fab or control non-binding IgG at a frequency of three times a week was injected intraperitoneally in transgenic SOD1 G93A mutant mice. For the therapeutic study, we elected to use 2B7 Fabs (lacking the Fc domain) to eliminate the potential for immune activation that could lead to misinterpretation. For example, we previously reported [15] that mAb 2B7 IgG could cause cell death in vivo but without detectable binding to TrkC.T1. In retrospect neuronal death was likely caused by Fc receptor-binding, activation of the complement cascade, and/or recruitment of macrophages.
In a realistic therapeutic paradigm treatment was initiated at postnatal day~100 (after disease onset, when hind leg reflexes were compromised). Treatment was continued until postnatal day~168. At postnatal day 160 the control group had 50% surviving mice whereas the 2B7 Fab group had 80% surviving mice. At postnatal day 168 (when treatment was discontinued) the control group had 35% surviving mice whereas the 2B7 Fab group had 60% surviving mice. At postnatal day 175 the control group had 0% surviving mice whereas the 2B7 Fab group had 60% surviving mice (Fig 7A). We note that the surviving mice in the 2B7 Fab group died precipitously between days 177 and 180 (9 days after discontinuation of treatment), and this is coherent with drug elimination pharmacokinetics.
In addition to significantly increased survival (Fig 7A), the 2B7 Fab-treated group exhibited significant (p < 0.01) improvement to motor performance in the rotarod test (Fig 7B), and maintained body weight (Fig 7C). In addition, their spinal cord neurons had enhanced ChAT and VAChT phenotype compared to control-treated ALS mice (Fig 7D).
In sum, 2B7 Fab therapy was effective in vivo in a mouse model of ALS, even when administered after disease onset, demonstrating that selective TrkC-FL agonists can be significantly and meaningfully effective in vivo, improving the key endpoints of life-span, motor function, weight, and motor neuron phenotype. To our knowledge, no reported therapeutic treatment has achieved such efficacy in these endpoints in this animal model.

Discussion
This work addresses the concept that functional ligands can selectively target the ectodomain of receptors that have identical primary sequences and which are not discriminated by natural ligands. Extension of this work to other receptors known to generate intracellular isoforms would generalize the concept and help to validate important therapeutic targets.

Structural differences in TrkC-FL and TrkC.T1 ectodomain
To our knowledge, this is the first example of a mAb differentiating between naturally occurring RTK ectodomains with identical primary structure. Moreover, the antibody acts as a A single intraperitoneal injection of 2B7 was done in wild type C57BL/6J mice (n = 3 per group), spinal cords were collected after 18 hrs, and detergent lysates of soft tissues were prepared and studied by Western blots. Quantification of p-Akt or p-PLCγ was done by densitometric analysis versus actin to standardize loading. The data shows activated p-Akt, and p-PLCγ in 2B7-treated mice, when compared to control IgG-treated mice. t-test * p<0.05 versus control. (C) The quantitative data in (B) was validated using symptomatic ALS mice (post-natal day 102). ALS mice received a single intraperitoneal injection of 2B7 or non-binding control IgG (n = 5 per group). After~18 hrs spinal cords were collected for biochemical and immunohistochemical studies. Detergent lysates of soft tissues were studied by Western blots using with anti-p-Akt, anti-p-MAPK, or anti-actin to standardize loading. The data suggests activated p-Akt, and p-MAPK in 2B7-treated ALS mice, when compared to control IgGtreated ALS mice. t-test * p<0.05 and ** p<0.01 versus control. (D) Immunohistochemical studies of ALS spinal cords. Cryosections of 2B7-treated or control-treated ALS mice (as in panel C) were immunostained with anti-p-Akt and anti-p-MAPK antibodies. Fluorescent microscopy data shows that 2B7-treatment induced Akt activation in neuronal cell bodies in the gray matter (co-localized with the NeuN marker) and in neuronal fibers of the white matter; and MAPK activation almost exclusively in neuronal cell bodies, but not in neuronal fibers of the white matter. Akt and MAPK activation in neurons are not seen in control IgG-treated ALS samples. functional agonist of TrkC-FL, while the natural ligand NT-3 does not discriminate between isoforms. Our strategy is therefore distinct from other mAb-based ligands that target unique structures generated by activating mutations in a receptor ectodomain [23].
The 2B7 epitope may be defined as a "receptor hot spot", and it was surprising that it would be expressed as a "cryptic epitope" in the TrkC.T1 isoform. The functional relevance of conformation and "cryptic epitopes" has been documented for Trk-receptors. For example, it is possible to activate Trk with monvalent/monomeric ligands through putative conformational mechanisms [26,30], rather than by directly inducing or stabilizing Trk-Trk dimers [23]. "Agonistic hot spots" in Trk receptors can be unmasked by homotypic co-expression of p75 NTR The spinal cords of wild type control, and 2B7-treated or Ig-treated ALS mice were dissected at day 160 (n = 2 per group). Cryosections were studied using fluorescent microscopy after immunostaining with anti-ChAT antibodies (a marker of motor neuron cell body) or anti-VAchT antibodies (a marker of motor neuron cell body and fibers). The ChAT and VAchT phenotype appears to be improved in 2B7-treated G93A ALS mice.
doi:10.1371/journal.pone.0162307.g007 [31], and these hot spots can be regulated functionally by antibody-based ligands [32]. Overall, these previous results support the notion of conformational regulation, through mechanisms that can be ligand-dependent.
The present work also demonstrates a third mechanism: that the gain of a neoepitope in the intracellular domain may be instructive of extracellular domain protein folding. It will be important to identify how the 2B7-recognized structure (TrkC-FL; TrkC.T1 Δinsert ) and the 2B7-unrecognized structure (TrkC.T1) are formed, because it may be relevant to many receptors that have splicing isoforms. Current work addresses this issue, under the hypothesis of differential folding and S-S formation during Golgi-ER transport.
We show that ligand-dependent signals differ between TrkC-FL and TrkC.T1, and it is also possible that functional differences may be independent of a conventional ligand. Synaptogenesis during development is regulated by physical interactions of TrkC.T1 with PTPsigma and TrkC-FL appears to be less effective at this role [33]. Potentially the structural differences at the ectodomains that we report here may contribute to specific physical interactions with PTPsigma and we are currently exploring this hypothesis.
A strategy that simultaneously affords intracellular truncation together with conformational changes to the ectodomain may be an efficient use of a single gene to generate proteins with different signal transduction properties (e.g. lacking a tyrosine kinase domain) and which can also differentially interact with non-conventional "ligands" or co-receptors through homotypic interactions (e.g. p75 NTR on the same cell membrane) or heterotypic interactions (e.g. PTPsigma on other cells), or to act as a conventional "dominant negative" inhibitor.

Receptor isoforms and disease
We used TrkC-FL agonist mAb 2B7 to validate selective agonist-dependent activation of this target as a potential therapeutic strategy for ALS. Moreover, we showed the mechanism of action by which TrkC.T1 plays a neurotoxic role in ALS, and suggest that TrkC.T1 antagonists or silencing may be potential therapeutics. Our results raise caution about the therapeutic use of growth factors that can activate receptor isoforms, and specifically show the significant barriers to the use of neurotrophins as drugs.

p75 NTR receptors as therapeutic barriers
We conclude that avoiding engagement of p75 NTR and TrkC.T1 seems essential in any therapeutic strategy. Neurotrophins bind to p75 NTR , which has multiple actions that include prodeath and pro-inflammatory signals [13]. p75 NTR has widespread expression in motor neurons [34], sensory and sympathetic neurons [35][36][37], and non-neuronal subtypes such as Schwann cells [38,39]. The p75 NTR is re-expressed during/following injury to recapitulate its developmental role in axonal pruning and cell death [32], and the cleaved extracellular domain of p75 NTR is a biomarker in ALS [40,41].
In sum, p75 NTR expression in cells not meant to be targeted and up-regulation in disease would compromise the use of wild type neurotrophins. Indeed, p75 NTR antagonists have modest efficacy at delaying symptoms in ALS models, and engineered neurotrophins that do not bind to p75 NTR are neuroprotective whereas wild type neurotrophins are not [13,14].

TrkC.T1 receptors as therapeutic barriers
Biochemical and cell biological data indicate that TrkC.T1 signals differ from TrkC-FL [42] and can activate RAC1 in a ligand-dependent manner [43]. Genetic and pharmacological data indicate that truncated TrkC.T1 is deleterious to neurons but the mechanism of action was unclear [15]. Here we extend those studies to show that TrkC.T1 causes a paracrine neurotoxic damage by directly up-regulating TNF-α, most likely through RAC1 activation, which is reported to up-regulate TNF-α.
Given the known role of TNF-α in human ALS [10], our data showing the role of TrkC.T1 at increasing neurotoxic TNF-α, and the correlation of TrkC.T1 up-regulation associated with down-regulation of miR-128 and disease progression in human sporadic ALS and in an ALS animal model, we suggest that TrkC.T1 may be a potential target for ALS therapy. This could be achieved by using TrkC.T1 antagonists, Rac1 inhibitors, by silencing TrkC.T1 expression using our pLKO.1 TrkC-T1 vectors, or by re-expressing miR-128 to promote degradation of TrkC.T1 mRNA. Current work is addressing these possibilities.

Poor pharmacokinetics and delivery as therapeutic barriers
Neurotrophins have very short half-lives in circulation [11]. The poor pharmacokinetics and poor specificity of NT-3, combined with the inability of NT-3 to reach the vast expanse of target receptors (from neuromuscular junctions to motor neuron cell bodies in spinal cord behind the blood brain barrier) makes them unsuitable as systemic drugs, and poor therapeutic agents for ALS. Experimental treatments using NT-3 protein or viral vectors expressing NT-3 (as well as therapies using many other growth factors) have generally failed for long-term therapy of ALS and failed primary endpoints in human clinical trials [44,45].
Although NT-3 delivered to spinal cord could spare neurons in spinal cord, it did not improve function or survival [14]. Likewise, adenoviruses expressing high levels of NT-3 injected in muscle protected the innervation of that specific muscle but did not protect centrally [6]. Another growth factor, IGF-1, also failed in ALS when injected subcutaneously [17] or when expressed in muscle [17], but had some efficacy after intraspinal delivery using viral vectors [46].
These results are consistent with the fact that signals at the neuromuscular junction differ from signals at the cell body [47] and emphasize the importance of biodistribution and pharmacokinetics to regulate the "location and tempo" of receptor activation required for neuronal survival [48]. Direct and repeated injection of therapeutics in each muscle or in spinal cord is problematic because of the large expanse that would need to be injected. A suitable therapy would have to reach all target tissues at anatomically distant places, from a single delivery route. MAbs as therapeutics are spared most or all of these problems.

TrkC-FL as a therapeutic target
MAbs are known to be relatively stable in circulation and mAb 2B7 and Fabs have a relatively long circulating half-life, are bioavailable and can be detected in motor neurons of the PNS and the CNS after a single systemic delivery. Additionally, 2B7 and Fabs are selective agonists for the target TrkC-FL. Hence, we used a challenging therapeutic paradigm with treatment initiated after first signs of motor decline were detected. This is a realistic model because patients would present with existing symptoms.
The therapeutic doses and frequency of 2B7 Fabs utilized are acceptable for treatment of chronic diseases such as ALS. We rationalized a dose and frequency suitable to achieve and maintain low-levels of circulating agonist, which would be sufficient to maintain a constant pro-survival signal without causing desensitization.
While antibodies have been shown to be high-quality therapeutics for cancer and immune disorders, the use of an agonist of an RTK also requires caution because of the potential for oncogenicity. Indeed, TrkC-FL has roles in tumor development, including breast cancer and adenoid cystic carcinoma, which could be a safety concern. On the other hand, targeting TrkC-FL therapeutically could potentially extend to neurodegenerative pathologies other than ALS where TrkC-FL and NT-3 are relevant, such as proprioceptive neurons, the repair of inner ear neurons damaged by noise, amongst others. Hence, it will be crucial to carry out detailed pharmacokinetic and pharmacodynamic studies to optimize intervention for chronic neurodegeneration.
The results presented here may guide the development of pro-neuroprotective (TrkC-FL) and anti-neurotoxic (TrkC.T1) therapeutic strategies in ALS and related diseases.

Antibodies and Proteins
The mAb 2B7 is an agonistic anti-TrkC IgG-C2 (D5) domain antibody. Production and purification of mAB 2B7 and its Fabs were performed as previously reported [7,50]. The epitope is conserved in mouse, rat and human TrkC and the mAb is known to bind to these species. mAb 2B7 does not bind to p75 NTR , TrkA, TrkB, and many other cell surface receptors tested. MAb 2B7 binds to cells transfected to express TrkC, but not to untransfected parental cells [6].
The mouse polyclonal 750 antibody was developed in our laboratory against a peptide comprising a neo-epitope which is present only in TrkC.T1 due to mRNA splicing (residues 595-611 sequence GIYVEDVNVYFSKGRHG). The anti-GFAP is a rabbit polyclonal from Millipore. NGF, proNGF and NT-3 were purchased (Alomone Labs and ProSpec-Tany). The rabbit anti-VAChT antibody was a gift from Dr. Claudio Cuello (McGill University) and the rabbit anti-ChAT antibody was purchased (ThermoFisher).
Expression of TNF-α, and TrkC isoforms ALS mice. For mRNA expression, total RNA, including short RNAs, were extracted from whole spinal cords (WT mice or ALS pre-symptomatic or symptomatic mice) using the miR-Neasy1 Mini Kit (QIAGEN). One μg of extracted RNA was reverse transcribed with the SuperScript III First-Strand Synthesis System for RT-PCR (Invitrogen). mRNA levels of TrkC isoforms and TNF-α were analyzed by real-time quantitative PCR using SYBR Green I (Quanta) in an 7500 FAST real-time PCR thermocycler (Applied Biosystems) and thermal cycling was performed as follows: 94°C for 30 sec, followed by 40 cycles at 94°C for 5 sec, 60°C for 30 sec. Specific primers were designed for the two TrkC isoforms and TNF-α. GAPDH or β-tubulin III were used as housekeeping genes ( Table 1). Data are expressed as the mean relative quantification (RQ) ± SEM (n = 8 individual mice per group). WT samples were used as a calibrator (RQ = 1).
HEK293 cells. For quantification of TrkC protein, detergent lysates of 293-TrkC or 293-TrkC.T1 or TrkC.T1 Δinsert cells were analyzed by Western blotting with MAb 2B7, or antibody 750 specific for TrkC.T1 or Ab C44H5 for all isoforms. For analyses of cell surface expression, 293-TrkC or 293-TrkC.T1 cells were fixed and immunostained with primary mAb 2B7, or antibody 750, or with non-binding mouse IgG control. DAPI was used to stain nuclei. Immunofluorescent pictures were taken using a Leica confocal microscope with 63x magnification.
RMC-1 cells. Non infected rMC-1 or cells infected with lentiviral pLKO.1 scrambled or pLKO.1 TrkC.T1 were treated with 2B7 100 nM or NT-3 10 nM or LPS 1 μg/ml for 6 h. Quantification of TNF-α mRNA was performed by real-time quantitative PCR as described previously and primers used for rat TNF-α and rat GAPDH are reported ( Table 1). Quantification of TNF-α in supernatant (conditioned media) was done using commercial rat TNF-α Standard ELISA (PeproTech). Data are expressed as the mean + SEM relative to the untreated (3 independent experiments, each in triplicate). Cell-associated levels of TNF-α were studied in the same cultures of rMC-1 cells treated as above, by immunofluorescence. Cells were fixed and immunostained with primary rabbit anti-mouse TNF-α (Millipore) followed by incubation with secondary anti-rabbit Alexa 488 (Invitrogen). Immunofluorescent pictures were taken using a Leica confocal microscope with 63x magnification to reveal cell-associated TNF-α protein (surface, intracellular, and vesicular).

MicroRNA expression
MicroRNA cDNA synthesis was performed with the qScript microRNA cDNA synthesis kit (Quanta) using 1 μg of total spinal cord RNA. Then, a real time SYBR Green qRT-pCR amplification of miRNAs were performed using 200 nM of each Perfecta microRNA assay primers for miR-128-1 (destabilizes TrkC.T1), miR-151-3p (destabilizes TrkC-FL) [51], or RNU6 (as reference) and universal primers (Quanta). qRT-PCR was performed using an 7500 FAST real-time PCR thermocycler (Applied Biosystems) and thermal cycling was performed as follows: preincubation/activation; 95°C for 2 min, followed by 40 cycles at 94°C for 5s, 60°C for 30s. The WT mRNA sample was used as a calibrator (RQ = 1). Data are expressed as the mean + SEM from 6 individual mouse spinal cords of ALS or WT.

Human spinal cord samples
Samples were collected post-mortem under a protocol from the Institutional Review Ethic Board (Brain Research Centre, UBC), with written informed consent from the donor or the next of kin. The Institutional Review Board specifically approved this study. Samples from four individual spinal cords of non-SOD1-related sporadic ALS (SALS) and three individual non-ALS controls were used. Information on human donors is listed in Table 2. Total RNAs were purified form the cervical region, and studies were done by RT-PCR as above. Data are expressed as the mean + SEM with control mRNA samples as a calibrator (RQ = 1).

Fluorescent Activated Cell Scan (FACS) analysis
FACs analyses were performed as described [52]. Briefly, cells were resuspended in 0.1 mL of binding buffer were incubated with mAb 2B7 or control mIgG for 20 min at 4°C, washed in binding buffer to remove excess primary antibody, and immunostained with FITC-mIgG secondary antibody for 20 min at 4°C. Cells were acquired and analyzed on a FACScan-BD Sciences using the Cell Quest program. As negative controls, no primary (background fluorescence) or irrelevant mouse IgG (Sigma) were used followed by secondary antibody. Specificity was gauged using various cells expressing different receptors.

Cell Metabolism/Survival Assays
The growth/survival profile of the cells were quantified in 96-well plates using the tetrazolium salt reagent 4-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; Sigma) 48 hr after plating; by reading the optical density (OD), as previously described [46]. Cells cultured in serum-free-medium (SFM) die by apoptosis, but they can be rescued if they express TrkC-FL and are supplemented with NT-3. Cells received vehicle, or were supplemented with 2B7 or 2B7 fabs (tests) or mIgG (negative control) or NT-3 (positive control) at the indicated concentrations. Cellular controls used cells expressing a related receptor TrkA, and its ligand NGF was the positive control. TrkA-expressing cells do not respond to 2B7 (data not shown). All assays were in quadruplicate and were repeated n>3 independent times. MTT data are standardized to optimal dose of neurotrophin = 100% survival, and serum-free medium (SFM) = 0% survival, using the formula ((ODtest − ODSFM) x 100 / (ODoptimal NTF − ODSFM)).

Activation of TrkC-FL in spinal cord in vivo
A single intraperitoneal injection of 0.5 mg/kg 2B7 (~10 μg) was done in wild type C57-BL6 mice (n = 3 per group) or ALS mice (n = 5 per group), and tissue was collected~18 hours postinjection of 2B7 or control IgG. Detergent lysates were analyzed by Western blotting with Anti-phospho-MAPK, anti-phospho-Akt, or anti-p-PLCγ (all from Cell Signaling). After stripping, membranes were re-probed with anti-actin (Sigma) to standardize loading. Quantification was done by densitometric analysis.

Pharmacokinetics and detection of 2B7 in the spinal cord
The pharmacokinetics of 2B7 or 2B7 Fabs were studied after intraperitoneal administration in a 6-week old C57BL/6J (B6). The circulating levels of 2B7 or 2B7 Fabs able to bind TrkC were quantified by ELISA based on specific binding to the known peptide epitope. Blood was collected at different times after a single IV injection (100 μg), plasma was prepared in Heparincoated tubes, and assayed in ELISA. Pre-bleed was used as background control.
For studies of 2B7 reaching motor neuron cell bodies in spinal cord, labeled mAb (2B7-ATTO) was injected intraperitoneally into a 6-week old C57BL/6J (B6) mice in two doses of 75 μg, 12 hr apart. After 48 hr, mice were perfused, and sections of their spinal cord were examined by confocal microscopy.
2B7 Fabs therapeutic efficacy in SOD1 mice B6.Cg-Tg(SOD1 Ã G93A)1Gur/J (JAX Labs, stock # 004435) mice over expressing the human SOD1 gene with the G93A mutation, crossed into a CBL/black 6 background, an average 50% survival at 157.1 ± 9.3 days [1]. The agonist was injected intraperitoneally three times a week at 0.5 mg/kg 2B7 Fab (~10 μg/mouse/injection), in transgenic SOD1G93A mutant mice (n = 12 mice per group, 6 males and 6 females). Treatment was initiated after disease onset was measured (e.g. when hind leg reflexes were compromised, at postnatal day~100). Treatment was continued until postnatal day 168. These experiments and the quantification of experimental endpoint were done double-blinded.

Experimental endpoints in ALS
Hindlimb leg reflex (HLR) is measured as a reduction in hindlimb extension when animals are lifted by the tail, and is an early deficit observed in mutant SOD1 Tg mice. Animals are lifted by the base of the tail to score hindlimb extension and postural reflexes [9]. Experimental endpoints of body weight, performance in the rotarod (measurements weekly starting at day~100, when hind-leg reflexes were compromised), and overall survival (extension of life-span) were done as described [53] for mutant SOD1 Tg mice ± mAb 2B7 ± control treatment. The health of the animals was monitored daily, and there were no unexpected events.

Study Approval
All animal protocols and endpoints in these studies were specifically approved by the University of British Columbia and McGill University IACUC. Humane endpoints were used during the animal survival study as described [23], and mice were euthanized when they met the specified criteria of high loss of body weight, immobility, or Hind-leg reflex level 4 as defined in the IACUC approved protocols. Euthanasia was done by CO 2 and cervical dislocation.