Single Cell Analysis of SMN Reveals Its Broader Role in Neuromuscular Disease

Abstract

The mechanism underlying selective motor neuron (MN) death remains an essential question in the MN disease field. The MN disease Spinal Muscular Atrophy (SMA) is attributable to reduced levels of the ubiquitous protein SMN. Here we report that SMN levels are widely variable in MNs within a single genetic background and that this heterogeneity is not only seen in SMA MNs, but also in MNs derived from controls and Amyotrophic Lateral Sclerosis (ALS) patients. Furthermore, cells with low SMN are more susceptible to cell death. These findings raise the important clinical implication that some SMN-elevating therapeutics might be effective in MN diseases besides SMA. Supporting this, we found that increasing SMN across all MN populations using a Nedd8-activating enzyme inhibitor promotes survival in both SMA and ALS-derived MNs. Altogether, our work demonstrates that examination of human neurons at the single cell level can reveal alternative strategies to explore for treatment of degenerative diseases.

Graphical abstract

Introduction

Motor neuron (MN) disorders are a clinically heterogeneous group of neurological diseases characterized by progressive loss of MNs resulting in muscle atrophy. Whereas ALS is a late-onset, rapidly progressing neurodegenerative disease, in which about 90% of the cases are sporadic and only 10% are inherited, SMA is a genetic, early-onset, degenerative disorder caused by low levels of Survival of Motor Neuron (SMN) protein. In the last few years, stem cell technologies have created the opportunity to generate large numbers of human MNs and other types of neurons from induced pluripotent stem cells (iPSCs). This has stimulated great deal of interest in the “disease in a dish” idea, i.e., applying this type of system to understanding more about disease mechanisms and identifying better treatments. However, few papers have used this method to go beyond recapitulating disease to providing more information about poorly understood aspects of specific degenerative processes. In this paper, we take advantage of our ability to study the behavior of individual MNs of different genetic backgrounds to achieve an insight into common mechanisms that regulate the death of diseased MNs.

SMA is caused by mutation or deletion of the Survival of Motor Neuron1 (SMN1) gene (Lefebvre et al., 1995, Lefebvre et al., 1997). Most humans harbor one or more copies of SMN2, a paralog to SMN1. SMN2 differs from SMN1 at a single nucleotide, which leads to an altered splicing pattern (Lorson et al., 1999, Monani et al., 1999) and the production of an unstable protein that lacks exon 7 (SMNΔ7) (Cho and Dreyfuss, 2010, Le et al., 2005). Importantly, each copy of SMN2 does produce a small amount of functional full-length SMN protein. The severity of SMA, inversely correlates with the number of copies of SMN2 that the patients retain (Harada et al., 2002). ALS patients do not carry SMN1 gene mutations, and the vast majority of them have approximately the same number of SMN2 copies (1-2) as the general population (Blauw et al., 2012). Many questions remain unanswered about the role that SMN plays in controlling MN survival, but we are beginning to acquire more information about why low levels of SMN lead to MN death, at least in SMA. By analyzing purified populations of SMA patient MNs, our lab discovered that there are molecular differences between MNs and other types of neurons, in particular their preferential activation of an ER stress response, that help explain why MNs die when compared to these other neurons (Ng et al., 2015).

A poorly understood aspect of MN diseases is why MNs carrying the same mutations and apparently exposed to the same stressors respond differently. This is particularly true for SMA, characterized by an acute phase and a chronic phase (Swoboda et al., 2005). The acute phase of the disease is accompanied by a wave of MN dysfunction and death, and the more chronic stage can be accounted for by the prolonged survival of a more resistant subset of MNs. To address the issue of selective cell death further, we performed the current study, where we analyze large numbers of individual MNs prepared from mice and human patients. We show that there is wide diversity of SMN protein levels per cell, with low SMN expressors and high expressors coexisting in the same culture. Even severe SMA patient MN cultures have a population of cells with levels of SMN similar to those found in cultures from unaffected patients. Those cells survive relatively normally although the majority of MNs have low SMN and die quickly. In addition, we have found that heterogeneity of SMN levels is not restricted to SMA cells; rather, a similar diversity of SMN levels exists in MNs derived from control patients and from ALS patients. Thus, we hypothesized, and work in this paper confirms, that selective death is determined, at least in part, by variation in SMN levels per cell. Specifically, individual MNs with low levels of SMN have a higher probability of dying than do the ones with higher levels, regardless of the origin of the cells. The implications of this observation are significant: we suggest that SMN has a broader role in promoting MN survival than currently appreciated and propose that types of drugs that are capable of elevating SMN in MNs of both SMA and ALS origin, i.e., those that can elevate SMN levels in MNs without SMN1 mutations, will enhance their survival.

Results

MNs are diverse in terms of their SMN levels

To probe the mechanistic underpinning for selective survival of some MNs in MN diseases, we focused first on SMA. It is commonly accepted that the factor that best determines SMA severity is SMN protein levels. We have shown that MNs produced from type I SMA patients die faster than those from types II and III (Ng et al., 2015). This prompted us to ask if SMN levels themselves might be heterogeneous in populations of MNs all made from the same starting population of embryonic stem cells (ESCs) or iPSCs. Most prior studies of SMA had measured SMN only at the whole tissue or whole cell population level, but we hypothesized that there might be differences in an individual MN's SMN expression that might affect its chances of survival, with cells with low SMN levels being more prone to death.

We first studied MNs produced from ESCs obtained from Wild-type (Wt) and SMA model mice (A2 (Smn-/-;SMN2+/+)) (Monani et al., 2000). Levels of SMN in MN cultures from these mice are reduced by 90% compared to control (Figure 1A), with only a small percentage of MNs surviving the first 24 hours after differentiation and plating (Figure 1B). The MNs in both mouse strains stably express GFP under the Hb9 MN-specific promoter. As described in the methods section, we used an imaging technique to quantify SMN levels in a defined portion of the MN cell body that produces reliable and consistent results, as we have previously shown (Makhortova et al., 2011, Ng et al., 2015). We observed that not all the MNs in the culture expressed the same amount of SMN protein (Figure 1C). We determined the distribution of single cell SMN protein levels by examining hundreds to thousands of cells per well. This distribution is represented in histograms, where the intensity of SMN immunostaining quantified in arbitrary units is binned on the X axis, and the number of MNs falling into each of those bins is shown on the Y axis. We observed that in Wt MN cultures, cells displayed more than a 3-fold variation in the level of SMN protein, with low, medium, and high SMN expressors in the same cultures (Figure 1D). Importantly, we found that mouse SMA MNs also presented a wide SMN distribution profile, with the histogram being shifted overall towards lower levels, but including MNs with normal levels of SMN.

Figure 1. Motor neurons in vitro and in vivo show remarkable heterogeneity in SMN levels.

(A) Lysates of Wt and Smn-/-;SMN2+/+;Hb9-GFP (SMA A2) MN cultures at dissociation were immunoblotted with SMN and Actin antibodies. SMN levels are extremely reduced in Smn-/-;SMN2+/+;Hb9∷GFP cells. (B) Quantification of MN number 24 hours after dissociation in Wt and Smn-/-;SMN2+/+;Hb9-GFP (A2) cultures reveals that SMN deficient MNs have reduced survivability in this early period (**p < 0.01, Student's T-test, n=11). (C) Mouse Hb9-GFP (Wt and A2) MNs fixed and immunostained with the SMN antibody (red). DNA dye bisbenzimide (Hoechst 33258) was used to stain the nucleus (blue) and GFP marks MNs (green). Dotted lines outline MNs. SMN expression is indicated with closed triangles marking high, open triangles marking medium, and the v-shape marking low levels in individual cells. Scale bar = 50μm. (D) Histogram analysis showing the percentage of MNs falling into each SMN intensity bin. (E) Representative western blot of lysates from MNs cultures differentiated from iPSCs derived from a control, type III, II and I SMA patients showing SMN levels. The quantification of 4 biological repeats from independent sets of MN differentiations is shown in (F) (One-way ANOVA test followed by Dunnet's analysis; **p<0.01, ***p<0.001, n=4). (G) Histogram analysis from MNs derived from iPSC from SMA patients with different disease severities. (H) Histogram analysis from MNs derived from ALS iPSC patient lines with different mutations. (I) Representative immunostaining on cervical cryosections from a P10 Wt untreated mouse showing MNs labeled with Hb9-GFP and SMN in red in the ventral horns of the spinal cords (nuclei stained with DAPI, blue). Scale bar = 20 μm. (J) Histogram analysis from 3 P10 Wt and SMNΔ7 untreated mice showing the MNs populations based on their SMN levels.

In order to confirm these results and to exclude the possibility that MN diversity only occurred in mouse MN cultures, we performed a similar study in human MNs derived from a mild (type III) SMA cell line (I-39C), a severe (type II) line (I-51C), and a very severe (type I) line (I-38G) (Table S1). First, we confirmed by western blot that, as expected, diseased cultures expressed markedly reduced SMN protein levels compared to healthy control ones, type I MN cultures having the lowest levels (Figure 1E-F). Histograms of SMN levels in MNs from all of these SMA patients, as well as from an unaffected control, show similar results to those we observed in mouse MNs, with even type I MNs being surprisingly diverse in their SMN protein levels and containing some individual MNs harboring similar SMN levels to those from unaffected patients (Figure 1G). Having found that this SMN heterogeneity is not restricted to SMA MNs but also applies to Wt cultures, we sought to discover if MN SMN heterogeneity also occurred in ALS iPSC-derived MNs. We obtained iPSC lines from patients with well-described types of ALS-causing mutations (Table S2) (Dimos et al., 2008), derived MNs from each of these iPSC lines, and measured SMN levels. The single-cell histogram analysis revealed that ALS MNs also display a marked variability in their SMN levels, with up to a 4-fold difference between low and high expressors (Figure 1H). In order to confirm that this observation is indeed real and not an artifact of cell culture, we also studied mouse spinal cords in vivo and observed a similar MN diversity in the amount of SMN expressed per cell (Figure 1I-J). These data indicate that, contrary to what would be expected, individual MNs derived from the same subject, normal or diseased, or present in the same MN pool within a given spinal cord segment, express very different SMN protein levels.

MNs expressing low SMN protein levels are more prone to die

These results along with our earlier study showing SMN's role in regulating stress-induced apoptosis in MNs (Ng et al., 2015), suggested that high SMN protein levels might be associated with lower rates of MN death in any condition in which MNs are stressed. To further correlate changes in SMN distribution with MN survival, we utilized a short hairpin RNA (shRNA) targeting SMN previously used in our laboratory to reduce SMN levels in Wt MNs by about 75% (Makhortova et al., 2011). We then monitored survival over a four-day period. We found that 45% of SMN knockdown MNs died while only 15% of the MNs expressing a non-silencing (NS) control shRNA did (Figure 2A-B). At the conclusion of the observation period, we performed an analysis of SMN levels in individual surviving MNs. Among these surviving cells, we found that the histogram profile for those MNs was left-shifted but not to values lower than in the starting population (Figure 2C), suggesting that the MNs with lower levels of SMN had died and therefore that there might be a threshold of SMN expression required for MN survival. As a second approach, we determined how the levels of SMN protein change over time in culture, again correlating levels with survival. MNs of both types die with increased time in culture (with a much more pronounced rate in the SMA MNs), but the average levels of SMN actually increased among the surviving cells (Figure 2D-E and S1A-B). This observation is again consistent with the idea that higher levels of SMN are correlated with improved survival.

Figure 2. MNs expressing low SMN protein levels are more prone to die.

(A) Mouse MNs expressing mCherry under the Hb9 promoter (Hb9∷mCherry) infected with pGIPZ lentivirus expressing GFP and SMN shRNA or NS shRNA control were imaged each day as indicated. Representative images are shown. Scale bar = 100 μm. (B) Quantification of the percentage MN survival reveals MN death induced by SMN knockdown (***p < 0.001, Two-way ANOVA with Dunnett's multiple comparison test, n = 5). (C) Histogram analysis reveals that SMN levels across the MN population are quite variable. SMN knockdown reduces SMN levels and cells with sub-threshold SMN levels die. (D) Quantification of the number or Hb9-GFP mouse Wt ESC-derived MNs after dissociation showing their survival profile. The MN cultures were live-imaged taking advantage of the Hb9-GFP endogenous labeling at the indicated time points (Oneway ANOVA test followed by Dunnet's analysis; *p<0.05, **p<0.01, ***p<0.001, n=3). (E) Histogram analysis from Wt mouse ESC-derived MNs show the percentage of MNs falling into each SMN intensity bin after the indicated number of days in culture. (F) Human SMA type II MNs derived from iPSC were treated for three days with the indicated compounds. Quantification of the SMN protein levels in the Islet1⁺ MNs related to the DMSO-treated cells is shown. The total number of cells per well after the treatment with the indicated compounds is quantified and shown in (G), and the number of Islet1⁺ MNs related to DMSO-treated wells is shown in (H) (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p <0.01 between each treatment compared to the DMSO. Two-way ANOVA with Dunnett’s multiple comparison test, n = 3). Based on the DMSO-treated cells, arbitrary SMN thresholds to bin the Islet1⁺ MNs into three categories were established: high, the 10% higher SMN expressors; low, the 40% expressing the lowest SMN amount; and medium, the remaining 50%. Quantification of the percentage of Islet1⁺ MNs derived from type II SMA iPSC harboring high, medium and low SMN levels after the three-day treatment is shown in (I) (*p < 0.05 between the “High SMN” population found for each treatment compared to the DMSO. Two-way ANOVA with Dunnett's multiple comparison test, n = 3). (J) Quantification of the number of Islet1⁺ MNs derived from type II SMA iPSC harboring high, medium and low SMN levels after the treatment (***p < 0.001 between the “High SMN” population found for each treatment compared to the DMSO; ###p < 0.001 between the “Medium SMN” population found for each treatment compared to the DMSO; §§ p < 0.01 and §§§ p < 0.001 between the “Low SMN” population found for each treatment compared to the DMSO. Two-way ANOVA with Dunnett's multiple comparison test, n = 3). (K) Representative images of the type II MN cultures treated with the indicated compounds. Scale bar = 50 μm.

MNs with high SMN have a survival advantage when exposed to toxic compounds

Over the years, we have found that there are many types of compounds that have the unexpected ability to increase average SMN levels across a population of MNs while, at the same time, reducing the number of MNs. We hypothesized that these effects could occur because of selective compound toxicity –i.e., being more toxic to MNs with low, than to those with high, SMN. Among this category of compounds are trichostatin A (TSA), the HDAC inhibitors M344 and parabinostat, the Na+/K+ ATPase pump inhibitor ouabain and the proteasome inhibitor MG132 (Figure 2F-K). As a control, we used C3, an SMN2 splicing modulator previously shown to be effective at elevating SMN in vitro and in vivo (Naryshkin et al., 2014). We applied these compounds to human SMA type II MN cultures for three days, fixed them and measured SMN by single cell imaging. We confirmed that a number of these compounds increased the average SMN in treated wells while decreasing MN numbers. This was not seen with C3, which increases average SMN, but does not do so at the expense of MN survival (Figure 2F-K). We established arbitrary SMN thresholds to bin the MNs into three categories: high (the 10% highest SMN expressors), low (the 40% expressing the lowest SMN amount) and medium (the remaining 50%). While the percentage of low and even medium SMN expressors dramatically decreased after treatment with toxic compounds, the percentage of high SMN expressors markedly increased (Figure 2I-J). Thus, when challenged with compounds that are cytotoxic, low SMN MNs are more prone to die, whereas cells with higher SMN levels have a survival advantage, explaining how average SMN levels can increase while cell number decreases.

SMN protein has a relatively long half-life in MNs

These results suggest a causative relationship between supra-threshold SMN levels and MN survival. This relationship depends, to some degree, on SMN protein being relatively constant over the period of the experiments. To investigate this possibility, we performed an additional set of experiments. First, we measured SMN protein half-life by treating the cultures with cycloheximide (CHX) to block protein synthesis, (Lorson and Androphy, 2000, Vitte et al., 2007, Locatelli et al., 2015). We found that in HEK293T cells SMN half-life is about 19 hours (Figure S1C-D) and in human MN cultures is at least 48 hours (Figure S1E-F). Thus, SMN is a rather stable protein in MNs.

Next, we hypothesized that if we treated MNs with a cytotoxic compound that would “select” for MNs with high SMN (as a result of selective death of MNs with low levels), those would be more resistant to a second round of a cytotoxic compound exposure. To test this, we treated MN cultures in two different ways: 1) either we exposed them for 24 hours to cytotoxic concentrations of TSA, removed the medium, let neurons recover in control medium for another 24 hours, then retreated the cultures with the same concentration of TSA for another 24 hours, or 2) exposed the MN cultures to the same concentration of TSA during just the last 24 hours of the culture. As expected, very few of the previously treated MNs died (“TSA 1+2” in Figure S1G-J), compared to those only exposed to TSA for the second round of treatment (“TSA 1”). A single-cell analysis of the SMN levels per MN showed that the first 24-hour treatment shifted the histograms towards the right, resulting in a larger percentage of MNs with higher SMN (Figure S1K-L). A similar shift was also observed when the cells were subjected to just the second cycle of treatment (Figure S1K-L). This study was done over a short period of time compared to the turnover of SMN that we previously determined, and combined with our other studies, is again consistent with the idea that higher levels of SMN confer resistance to environments that are stressful for MNs.

SMN overexpression promotes MN survival, but does not affect the survival of iPSC-derived cortical neurons

To show directly that the correlation between higher SMN protein levels and increased MN capability to cope with stress is functionally significant, we treated Wt mouse MNs with a lentivirus encoding doxycycline (DOX)-inducible SMN or, as a control, RFP and initiated cell death by withdrawing trophic factors (TFs) (Yang et al., 2013). This paradigm allowed us to explore the relationship between SMN and survival in an experimentally rigorous setting. Indeed, compared to expression of the RFP control, overexpression of SMN in these Wt MNs led to a dose-dependent increase in MN number (Figure 3A-D) associated with populations of MNs being increasingly shifted into higher SMN-expression categories (Figure 3B). These results demonstrate that the survival of MNs even in the absence of SMN1 mutations, is improved by augmenting SMN expression.

Figure 3. SMN overexpression promotes MN survival, but does not affect the survival of iPSC-derived cortical neurons.

(A) Representative images of mouse Wt MNs (green) fixed and immunostained with a SMN antibody (red). Cells were infected with a doxycycline(DOX)-inducible lentivirus carrying SMN or RFP as control, then treated with increasing concentrations of DOX or vehicle (water) (no DOX and 30 ng/mL DOX -treated cells are shown) and subjected to trophic factor withdrawal for the last seven days of the culture. Scale bar = 50μm. (B) Histogram analysis showing the percentage of MNs falling into each SMN intensity category after SMN overexpression induced by different DOX concentrations. (C) Quantification of the surviving MNs after SMN overexpression relative to the number of MNs remaining after RFP overexpression (**p < 0.01, One-way ANOVA followed by Dunnett's multiple comparison test, n = 4). (D) Quantification of SMN protein levels after SMN overexpression relative to RFP overexpression (*p < 0.05, ANOVA followed by Dunnett's multiple comparison test, n = 4). (E) Human MNs differentiated from iPSCs derived from ALS patients with the indicated mutations were infected with a DOX-inducible lentivirus carrying SMN or RFP at day 1 after plating, 0.5μg/ml DOX added one day after the infection and kept for 5 days, when the cultures where fixed and stained for Islet1. Scale bar = 50µM. (F-H) In each case the number of Islet1⁺ MNs remaining at the end of the culture was related to each RFP control infected, with no DOX cells (“RFP – dox”) (*p < 0.05, **p < 0.01, ***p<0.001, One-way ANOVA followed by Dunnett's multiple comparison test, n = 3). Healthy control (1016A and BJ) and types III, II and I SMA iPSCs were differentiated into cortical neurons by forcing the expression of neurogenin-2 (Zhang et al., 2013). Subsequently, cells were transduced or SMN-overexpressing DOX-inducible lentivirus or RFP. Six days later trophic factors (TF) were removed for nine days followed by fixation and staining of the cultures with anti-Brn2 and SMN antibodies. (I-M): The quantification of the number of Brn2⁺ cortical neurons remaining after TF withdrawal (-TF) for each over-expressing virus (RFP or SMN) was related to the “plus TF” (+TF) condition and shown for each line: (I) 1016A, (J) BJ, (K) I-39C, (L) I-51N, and (M) I-39G. (No statistical significance found by One-way ANOVA, n = 3).

These experiments along with the ones shown in Figure 2 led us to postulate that higher SMN levels could be also protective in MNs subjected to other types of endogenous stress. It is well known that cellular stress is a hallmark of MN degeneration in ALS. Several studies have reported connections between ALS-inducing mutations and SMN function (Sun et al., 2015, Groen et al., 2013, Kariya et al., 2012, Yamazaki et al., 2012; however, there is controversy about whether increasing SMN levels would have any therapeutic effect in ALS patients. To determine if SMN might play a role in the survival of ALS MNs, we again performed SMN overexpression experiments. We first characterized the MN death rate in these cultures and identified the TDP43-47A (G298S) and SOD1 lines as having the highest basal level of MN death (data not shown). We infected these MNs, together with an unaffected control line, with either DOX-inducible RFP control or SMN expressing lentiviruses and treated the cells with DMSO or 0.5μg/ml DOX for 5 days. We observed that the number of surviving MNs significantly increased in both lines, as well as in the control line, when SMN was overexpressed (Figure 3E-H). Thus, increasing SMN protein prevents MN death in cultures subjected to the cellular stress associated with ALS, again in the absence of any SMN1 gene mutations.

In spite of SMA being considered a disease primarily affecting spinal MNs, recent studies have suggested that other neuronal types, such a pyramidal neurons of the motor cortex (d'Errico et al., 2013) and hippocampal neurons (Wishart et al., 2010) might be affected. Therefore, we wondered if increasing SMN might also exert a protective effect on cortical neurons derived from control and SMA iPSCs under conditions in which the cells are induced to die rapidly by withdrawing trophic factors. In order to address this question, we employed a recently described protocol to differentiate iPSCs into cortical neurons by overexpressing Ngn2 (Zhang et al., 2013). As before, we increased SMN levels in these cells using lentiviral vectors and then subjected them to TF withdrawal. We observed that cortical neurons in each line died when TF were removed, but the amount of death did not track with SMA severity (as it did in the case of MNs), except that the type I line did have more death (Figure 3I-M). We next determined the percentage of Brn2⁺ neurons that died due to TF withdrawal. We did not find a cytoprotective effect of the increased SMN on the survival of cortical neurons in any lines studied (Figure 3I-M), despite achieving similar SMN overexpression levels to the ones achieved in the MN cultures and despite the fact that these cortical neurons also present a wide heterogeneity in their SMN expression profile (Figure S2). These data indicate that the protective effect of increasing SMN may not apply to all types of neurons under all conditions in which cells are stressed.

Therapeutic implications associated with SMN having a broader role in regulating MN survival

Having shown that increasing SMN protein levels may be a broadly useful strategy for promoting MN survival independently of SMN1 mutations, we aimed to identify small molecules that could exert this effect and could therefore be used to treat other MN disorders, particularly ALS. Our laboratory previously demonstrated that SMN can be targeted for degradation following its phosphorylation by GSK-3β, and that certain inhibitors of that kinase are able to support the survival of SMN-deficient mouse MNs (Makhortova et al., 2011). Cullin ring ligases (CRLs) regulate the ubiquitination and degradation of the preponderance of proteins destabilized by GSK-3β phosphorylation (Makhortova et al., 2011). Thus, we hypothesized that MLN4924, an inhibitor of Nedd8-activating enzyme that acts on the Cullin proteins and is currently in the clinic for hematological and other malignancies, may stabilize SMN and increase its intracellular levels. To test this, we first performed biochemical analyses to validate whether SMN is degraded by a CRL dependent mechanism. We expressed equal amounts of hemagglutinin (HA)-tagged forms of the full-length SMN (FL) protein or a truncated SMN protein carrying a deletion of exon 7 (Δ7) in HEK293T cells and found that treatment with MLN4924 significantly increased the levels of both the FL and Δ7 SMN proteins (Figure 4A-B) by blocking its ubiquitination (Figure 4C). We then found that exposing mouse Wt ESC-derived MNs to MLN4924 led to the accumulation of SMN protein (Figure S3A-B) without affecting SMN RNA levels (Figure S3C). We confirmed these results in dose-response experiments on the severely affected SMA A2 mouse MNs and human SMA type I MNs, where we observed that MLN4924 was surprisingly efficient at increasing the average SMN levels (Figure 4D, F). Single-cell histogram results showed further that MLN4924 treatment shifted the population of MNs, both mouse and human, into the higher SMN categories (Figure 4E, G; Figure S3D-E). Similar results were obtained using type II and III human MNs (Figure S3F-G). Importantly, the increase in SMN produced by MLN4924 restored the levels of SMN close to those of control MNs, at least in the type II line (Figure S3H). We also tested the effects on SMN levels of MLN4924, which acts post-translationally, together with C3, which acts on SMN2 RNA splicing. As expected, the combination of MLN4924 and C3 led to an increase in SMN levels that was substantially greater than the effect of either single compound (Figure S3I-J; results confirmed by immunofluorescence, data not shown). These data confirm that the two compounds have different mechanisms of action and may form the basis of a combination therapy that could be employed in the future.

Figure 4. Therapeutic implications associated with SMN having a broader role in regulating MN survival.

(A) Lysates of 293T cells transfected with expression plasmids encoding HA-tagged SMN-FL or -Δ7 and treated with 0.3μM MLN4924 or DMSO for 3 days in 0.5% serum containing media were immunoblotted with the HA and tubulin (Tub) antibodies. Light and dark exposures are shown for HA. (B) Quantification of fold intensity change of HA-SMN-FL and HA-SMN-Δ7 protein levels normalized to tubulin protein levels analyzed as in (A). Numbers reflect the fold change effect of MLN4924 compared to the DMSO control for each SMN form (*p < 0.05, **p < 0.01, Student's T-test, n=4). (C) Lysates of 293T cells transfected with the expression plasmid including HA-SMN-FL and treated with 0.3μM MLN4924 or DMSO for 3 days were incubated with 2% SDS to disrupt non-covalent interactions, diluted to 0.2% SDS and immunoprecipitated with the HA antibodies. 10% of the total lysates and immunoprecipitates were immunoblotted with the HA, Tub, and ubiquitin (ub) antibodies, respectively. MLN4924 prevents ubiquitination of SMN protein. (D) Quantification of average SMN levels reveals that MLN4924 increases SMN protein in SMA A2 (Smn-/-;SMN2+/+;Hb9-GFP) MNs (**p < 0.01, One-way ANOVA followed by Dunnett's multiple comparison test, n = 11). (E) Histogram analysis showing the distribution of Hb9-GFP MNs according to their SMN protein levels after MLN4924 treated. (F) Quantification of average SMN protein levels after MLN4924 treatment in MNs derived from SMA type I iPSC (**p < 0.01, One-way ANOVA followed by Dunnett's multiple comparison test, n = 8). (G) Histogram analysis showing the distribution of type I Islet1⁺ MNs according to their SMN protein levels after MLN4924 treatment. (H) Mouse SMA A2 MNs (green) treated with increasing doses of MLN4924 or DMSO for four days were subjected to immunocytochemistry with the SMN antibody (red) and Hoechst (blue). Selected images are shown for 0.3μM MLN4924 and DMSO. Dotted lines represent tracing of the MNs. Scale bar = 50μm. (I) Quantification of the percentage of surviving MNs treated as in (G) and expressed as a comparison of MNs surviving after 4 days relative to 1 day reveals that MLN4924 treatment significantly promotes survival of SMA A2 MNs (*p < 0.05, **p < 0.01, One-way ANOVA followed by Dunnett's multiple comparison test, n = 11). (J) Representative images of Islet1⁺ MNs differentiated from iPSCs derived from SMA type I patients and treated with 1.25μM DMSO, MLN4924. Scale bar = 50μm. (K) Quantification of Islet1⁺ SMA type I MNs treated with DMSO or increasing concentrations of MLN4924 for 3 days, fixed, and subjected to immunocytochemistry (*p<0.05, **p < 0.01 Oneway ANOVA followed by Dunnett's multiple comparison test, n=4). (L) Quantification of the increase of SMN levels in SMNΔ7 mouse MNs infected with lentivirus expressing the dominant- negative (DN) form of the indicated Cullin for 8 days as compared to the empty vector infected cells (*p < 0.05, One-way ANOVA followed by Dunnett's multiple comparison test, n=6). (M) Hb9∷GFP SMNΔ7 mouse MNs fixed and immunostained 8 days after being infected with empty vector or Cul5DN lentivirus (SMN in red, scale bar = 50 μm). (N) Quantification of the number of SMNΔ7 MNs 8 days after infection with a lentivirus carrying the Cul5DN compared to empty vector treated ones (**p < 0.01, two-tailed T-Test, n = 6).

This prompted us to test the effects of MLN4924 on MN survival. We first used differentiated SMA A2 mouse MNs and, by performing live time-lapse imaging of Hb9-GFP MNs during a 3-day period of time, we observed that the treatment prevented about 33% of the death in the severe SMA mouse model, something unexpected given that these SMA MNs die very quickly (Figure 4H-I). We performed similar experiments in SMA human MNs and, found that MLN4924 very efficiently rescued, in a dose-dependent manner, MN death in all three SMA lines, even the most severe type I (Figure 4J-K and S3K-N). Because the amount of death is higher in the type I lines, the amount of increase in MN survival achieved by MLN4924, calculated in this fashion, was also higher in type I cells. These results demonstrate that MLN4924 efficiently prevents MN death even in patient cells that carry a low number of SMN2 copies.

Inhibition of Cullin5 ubiquitin ligase activity by MLN4924 prevents SMN degradation in vitro and in vivo

We investigated MLN4924's mechanism of action and its effects in vivo. First, we sought to identify the specific Cullin E3 ubiquitin ligase whose inhibition mediates the effects of MLN4924 in MNs. In this case, we derived MNs from ES cells obtained from the SMNΔ7 mouse, where SMN protein is reduced by 70% compared to control and the basal MN death occurs over a more prolonged period of time compared to A2 MNs, which makes these neurons more compatible with viral infection (Figure S4A-B). We infected these MNs with lentiviral vectors encoding the dominant-negative (DN) form of the Cullins whose neddylation and subsequent activation is prevented through treatment with MLN4924 (Cullin1-5) (Soucy et al., 2009, Emanuele et al., 2011) and observed that SMNΔ7 cultures treated with the DN form of Cullin5 (Cul5DN), but not with DN forms of the other cullins, showed significantly increased SMN levels (Figure 4L-M) and MN survival (Figure 4N) compared to the vector control. Overexpressing Cul5DN in Wt MNs produced similar results (Figure S4C-E). Further evidence supporting the idea that the effects of MLN4924 on MN survival reflect its ability to increase SMN were obtained from experiments with MLN4924 treatment and Cul5DN overexpression were combined. These produced overlapping SMN histograms with no additive effect on either SMN levels or MN survival (data not shown). These data indicate that at least part of the mechanism by which MLN4924 promotes SMN stability leading to MN survival is through inhibition of the Cullin5 ubiquitin ligase.

We next investigated the ability of MLN4924 to increase SMN levels and ameliorate some of the hallmarks of the disease in vivo. Since this compound is not blood brain barrier penetrant, we first delivered it directly to the central nervous system (CNS) by injecting Wt (FVB) pups at postnatal days 0 and 3 (P0 and P3) (Porensky et al., 2012, Hua et al., 2011), and confirmed that MLN4924 treatment could increase SMN compared to the vehicle-treated mice (Figure S4F). We then treated SMNΔ7 mice adding a third and last intra-CNS injection at P6. Encouragingly, weight, one of the most measured parameters to determine the general health of SMNΔ7 mice, was 25% higher in mice treated with MLN4924 compared to the vehicle-treated ones (Figure S4G-H) and their lifespan showed a statistically significant increase of 3 days (Figure S4I).

MLN4924 increases SMN levels and promotes robust survival in ALS patient iPSC-derived MNs whereas an SMN2 splicing modulator does not

Our results suggest that compounds like MLN4924, working across all MN populations and even in the absence of SMN1 mutations, might be successful in treating ALS as well. We decided to carry out a side-by-side comparison between MLN4924 and the splicing modulator C3 given its proven efficacy at increasing SMN protein levels in SMA cells (Naryshkin et al., 2014). Since cells derived from ALS patients have normal levels of SMN protein expressed from the SMN1 gene and a normal distribution of SMN2 gene copies (1-2 copies, Table S2) (Blauw et al., 2012), we did not expect SMN2 splicing modulators to have a substantial effect on SMN protein levels in ALS MNs. On the other hand, MLN4924 should still be effective in those cells based on its mechanism of action. To confirm this, we derived MNs from each of the mutant ALS iPSC lines described previously (Figure 1H, 3E-H), treated the cultures with MLN4924 or C3 for 3 days and quantified SMN levels in MNs. We observed that MLN4924 very effectively increased SMN levels in the four lines in a dose-dependent manner, whereas C3 only produced a small increase in the severe SOD1 mutant line and showed no effect in the other lines (Figure 5A-F). When analyzing MN survival, we observed that MLN4924 robustly rescued MNs in the lines where extensive MN death occurs, TDP43-47A and the SOD1 mutants, whereas C3 showed at best a small effect in the severe SOD1 line that did not reach statistical significance (Figure 5G-H).

Figure 5. MLN4924 increases SMN levels and promotes robust survival in ALS patient iPSC-derived MNs whereas an SMN2 splicing modulator does not.

(A-D)Human MNs differentiated from iPSCs derived from ALS patients with the indicated mutations were treated with DMSO, MLN4924 or C3 at indicated concentrations for 3 days, fixed, and subjected to immunocytochemistry with SMN and Islet1 antibodies. SMN protein levels measured in Islet1⁺ MNs are shown as a fold change related to the DMSO-treated MNs. Concentration-dependent data were fitted to a non-linear regression equation using Prism (*p < 0.05, **p < 0.01, ***p < 0.001, Two-way ANOVA followed by Dunnett's multiple comparison test, n = 4). (E-F) Representative images from DMSO, 1.25μM MLN4924 or 1.25 μM C3 treated ALS MNs (SMN in red, Islet1 in green and Hoechst in blue. Scale bar = 50 μm). (G-H) Quantification of the number of Islet1⁺ MNs derived from TDP43-47A and SOD1 ALS iPSC patient lines after MLN4924 or C3 treatment. Concentration-dependent data were fitted to a non-linear regression equation using Prism (*p < 0.05, **p < 0.01, ***p < 0.001, Two-way ANOVA followed by Dunnett's multiple comparison test, n = 4).

Discussion

SMA and ALS are MN diseases with very different causes that share some molecular mechanisms like disrupted RNA processing and a depleted number of SMN-containing nuclear gems (Cauchi, 2014, Sun et al., 2015), but the relationship between the two disorders has not been clarified. In the case of SMA, it is universally accepted that the disease is caused by low levels of SMN protein, and this has led to a significant effort in academia and industry to discover SMN-elevating therapeutics. On the other hand, how increasing SMN levels impacts the survival of MNs affected by ALS-causing mutations is a question that has been addressed by other groups before but without resolution (Turner et al., 2009, Wang et al., 2014, Blauw et al., 2012, Turner et al., 2014).

SMA patients suffer from an acute phase marked by an intense MN loss that is followed by a more chronic stage where the remaining MNs stay alive even in patients with severe forms of the disease. How some MNs manage to survive even in severe SMA is completely unclear. One possibility is that death is random, while another is that it reflects additional molecular heterogeneity across MN populations that specifies which particular cells are more, or less, likely to die. To address this question, we combined the use of mouse and human pluripotent stem cells with single cell imaging. We found that a critical molecular difference among MNs from the same subject is the single cell levels of SMN, which, in fact, vary widely. This is highly significant because our work now demonstrates that, in a given population of MNs, the lowest SMN expressors are the most likely to die when exposed to the basal stress of the culture or to various chemical stressors. Conversely, even in cultures made from the most severe types of SMA patient cells, we found that there were individual cells with relatively normal levels of SMN, and these had improved survival. This leads us to postulate that it is variation in SMN expression itself that explains the existence of the acute and chronic SMA phases. That is, the lowest expressors die in the acute phase leaving behind MNs with levels of SMN above the threshold needed to support longer term survival. The implication of this finding is that SMA actually ceases to be a MN disease, at least exclusively, once the most vulnerable MNs have died.

Our results suggest that SMN is an important and general regulator of MN survival independent of the existence of SMN1 mutations. In order to validate this hypothesis, we performed SMN overexpression experiments and other studies and confirmed that increased SMN protein levels indeed protects against the basal MN death occurring in cell culture, stress-induced MN death (triggered by trophic factor withdrawal or addition of cytotoxic compounds), and also against the MN death induced by the stress known to be associated with ALS mutations. However, this cytoprotective effect of SMN does not apply to all types of neurons under all conditions since increasing SMN by lentiviral overexpression in human control and SMA iPSC-derived cortical neurons did not improve their survival, despite the fact that these neurons display similar profiles of SMN heterogeneity than the MN cultures and that similar levels of overexpression were achieved

This study provides a justification for using compounds that increase SMN protein levels to promote MN survival in diseases, beyond SMA, in which MNs die. As SMN seems to be involved in the cellular response to oxidative or ER stress (Strasswimmer et al., 1999, Ng et al., 2015), low SMN expressors may be especially prone to death in the context of ALS, where disruption of stress response pathways plays a major role in disease pathogenesis (Cozzolino and Carri, 2012, Bernard-Marissal et al., 2012, Bogdanov et al., 2000). In this regard, it is important to note that most of the therapeutic approaches that have shown promising results in in vitro and in vivo SMA models either up-regulate SMN2 gene expression or modify its splicing. However, SMN2 copy number is normally not high in ALS patients, and, therefore, these compounds are unlikely to be broadly effective in settings in which SMN1 mutations are not involved.

To pursue this idea, we followed up on our previous work in which we found that GSK3β is an important regulator of SMN stability as phosphorylation of SMN by this kinase targets it for degradation (Makhortova et al., 2011). Because nearly all substrates of GSK3β that are subsequently degraded are targeted for ubiquitination by CRL family members, we hypothesized that SMN might also fall in this category. As MLN4924 is a potent and specific regulator of CRL E3 ubiquitin ligases, we assessed its effect on SMN protein. Interestingly, SMN was identified as a putative CRL substrate in a global survey of proteins whose ubiquitination was inhibited by MLN4924 (Emanuele et al., 2011), and we indeed found that MLN4924 stabilizes SMN by blocking its ubiquitination and degradation. In addition, we discovered that in particular Cullin5 is partially responsible for the ubiquitination and proteasomal targeting of SMN.

MLN4924 increased SMN levels and MN survival in every mouse or human SMA line tested and it did so at a magnitude similar to that achieved by the potent SMN2 splicing modulator C3, as shown by us and our collaborators (Naryshkin et al., 2014). MLN4924 also ameliorated the disease phenotype of a severe SMA mouse model in vivo and, used in combination with C3 produced additive effects on SMN protein levels. However, in a side-by- side study comparing the efficacy of MLN4924 and C3 in promoting MN survival, MLN4924 is effective regardless of patient genotype whereas C3 only acts on SMA MNs.

Taken together, our data support three significant conclusions: 1) MNs display a wide heterogeneity of SMN expression even within a single genetic background and in absence of SMN1 mutations, with low SMN expressors being more likely to die under all conditions we studied, suggesting that SMN exerts a broader control on MN survival than had been recognized; 2) there are additional druggable targets for SMA that provide alternatives to the splicing modulators – small molecules and anti-sense oligonucleotides – that are currently in the clinic; 3) there is now a clear therapeutic link between SMA and ALS in that certain drugs that can increase SMN levels might be therapeutic for both. From a clinical perspective, having a single drug able to treat both a classical, late-onset, heterogeneous disease, such as ALS, along with a childhood, early onset, genetic disease, such as SMA, may offer significant advantages during the long and difficult drug approval process.

Finally and fundamentally, our work demonstrates that unique insights into disease can be achieved using an iPSC-based approach. This method uniquely provides investigators both with large numbers of differentiated cells, as well as the ability to study their properties at the single cell level. We applied this method to SMA, an early-onset, and to familial ALS, a late onset disease, but anticipate similar progress made with MN disorders associated with multiple initiating factors.

Experimental Procedures

Mouse treatments

All animal studies were approved by Harvard University Institutional Animal Care and Use Committee and performed in accordance with institutional and federal guidelines. Intra-ventricular injections of MLN4924 or vehicle were performed using an ultrasound back-scatter microscope and injection guidance system (Vevo 770, VisualSonics) as described before (Ozdinler and Macklis, 2006). A second injection within the spinal cord parenchyma at the C2 level was also performed. MLN4924 was prepared as a 10 mM stock solution in DMSO and diluted fresh in phosphate buffered saline (PBS) to obtain a final concentration in the cerebrospinal fluid of approximately 50 μM. Mice were randomly assigned to the vehicle or MLN4924 groups and treated at postnatal days 0, 3, and 6. They were weighed every day from postnatal day 2 and sacrificed 8 hours after the last injection to perform SMN ELISA analysis of spinal cord tissue. At least 5 mice per group were studied.

Spinal MN histology

The spinal cord was dissected from Wt and SMNΔ7 mice euthanized at postnatal day 10 and fixed in 4% paraformaldehyde in PBS overnight (o/n). The spinal cord was rinsed in PBS, transferred to 15% sucrose o/n, followed by 30% sucrose o/n, then embedded in OCT for cryosectioning. For MN measurements and SMN quantification, sixteen sections per sample were imaged blindly at 20× magnification using a Nikon Eclipse Ti microscope, keeping the settings constant for each of the channels. Subsequent image quantification was performed automatically using the Columbus Image Data Storage and Analysis System (PerkinElmer). To determine MN numbers, area and SMN levels, an average number of 700 MNs per experimental group were analyzed.

Biochemical Assays

Human Embryonic Kidney (HEK) 293T cells were transfected with Lipofectamine 2000 (Thermo Fisher Scientific) with the HA-tagged forms of the FL SMN or Δ7 protein. For ubiquitination assays, 1,10-phenanthroline (VWR) was included in the lysis buffer to inhibit deubiquitinating enzymes. Lysates were pre-cleared with protein A/G Sepharose beads prior to immunoprecipitation with HA antibodies. For protein turnover assays, human MNs were incubated with the indicated concentrations of cycloheximide (CHX, Sigma Aldrich) for 6 to up to 48 hours. Immunoblot analysis was performed as previously described (Makhortova et al., 2011). Densitometric analysis was performed on scanned autoradiographs using the Quantity One software (Bio-RAD).

Mouse MN Differentiation and Survival Assays

Mouse Wt hb9∷GFP, SMN-/-;SMN2+/+;hb9∷GFP, SMN-/-;SMN2+/+;SMNΔ7;hb9-GFP and hb9∷mCherry ES cells were maintained and differentiated as described previously (Yang et al., 2013). For MN survival assays, MLN4924 was added after plating and GFP+ MN numbers were quantified at day 4 and normalized relative to the number at day 1. To assess the effect of SMN knockdown on MN survival, hb9∷mCherry MNs were infected with lentiviral constructs expressing GFP and non-silencing shRNA or SMN shRNA at day 1-3, and compound was added after 2 days. For the cullin dominant negative (CulDN) experiments, SMN-/-;SMN2+/+;SMNΔ7;hb9∷GFP MNs were similarly infected with the lentiviral constructs at day 1 after plating, and SMN levels and MN survival determined 9 days after infection. When indicated, MLN4924 was added to the medium 5 days after plating, 4 days after infection with the CulDN lentivirus, when the expression of the CulDN is not yet detected.

SMN overexpression assay

Mouse Wt MNs were infected with the RFP or SMN dox-inducible constructs encoded in lentivirus at day 1 after plating and treated with multiple concentrations of doxycycline (Clontech) 1 day later. Infected neurons were selected with puromycin (1 μg/ml, Thermo Fisher Scientific) five days after the infection, and, two days later, trophic factors (BDNF, GDNF and CNTF (R&D Systems)) were withdrawn from the medium in order to induce cellular stress and cell death. Seven days later the cells were fixed and stained, and SMN levels and number of MNs quantified. For the human iPSC-derived ALS MN experiments, cells were infected 1 day after plating in presence of 8 μg/ml polybrene, and a day later the media was changed and 0.5 μg/ml doxycycline added when applicable. Cells were fixed at day 7, 5 days after doxycycline treatment.

Human MN Differentiation and Survival Assays

The information regarding SMA and ALS iPSC lines is summarized in Table S1 and S2, respectively, and detailed in Supplemental Experimental Procedures. Human iPSC were derived, maintained, and differentiated into MNs following previously described protocols (Yang et al., 2013, Rigamonti et al., 2016) and specific modifications are included in Supplemental Experimental Procedures. For astrocyte co-cultures, mouse astrocytes were prepared from P0-P2 CD1 mouse cortices as previously described (Schildge et al., 2013). Four days prior to dissociation of 30-day differentiated human MNs, astrocytes were seeded at 10,000/well in 384-well plates. Papain-dissociated single cells were plated at 10,000/well in medium containing Neurobasal with 2% B27 and 1% N2 (Thermo Fisher Scientific) supplemented with 20 ng/ml of BDNF, GDNF, and CNTF and Ara-C. Survival of human MNs was determined by counting Islet1⁺ cells.

Human Cortical Neuron Differentiation and Survival Assays

A recently reported method (Zhang et al., 2013) was followed to produce human cortical neurons from human iPSCs by forcing expression of neurogenin-2. RFP control or SMN dox-inducible lentiviruses were added 1 day after infection with Ngn2 and rtTA lentiviruses and puromycin after 1 additional day (0.5 μg/ml). CD-1 mouse (Charles River) cortical astrocytes were added to the culture 2 days later at density of 4,000 cells per well in Neurobasal based medium containing: B27 Supplement, Glutamax, Penicillin-Streptomycin, human BDNF (10 ng/ml), human NT-3 (10 ng/ml), and doxycycline (0.2 μg/ml). No puromycin was used after this point. Two days later, 2 μM Ara-C was added to stop cell proliferation. Trophic factor withdrawal was performed by removing B27 Supplement, human BDNF, and human NT-3 after complete medium change and from this point onwards half of the medium was changed every 3 days for an extra 9 days, when cells were fixed and stained for Brn2 and SMN. Survival of human cortical neurons was determined by counting Brn2⁺ cells.

Immunocytochemistry and Image Analysis

Cells were fixed, permeabilized and immunostained as previously described (Makhortova et al., 2011). Images of cells were captured using an automated Operetta or Opera confocal microscope at 20× magnification and subsequent image quantification was performed using the Columbus Analysis System. Specific details on how MNs were identified and immunofluorescence signal quantified can be found in Supplemental Experimental Procedures.

Enzyme Linked Immunosorbent Assay (ELISA)

In vivo SMN levels were determined by ELISA (Enzo Life Sciences). FVB Wt mice treated with DMSO or MLN4929 were euthanized 8 hours after the last injection, and the spinal cords were dissected. The cervical sections containing C1-C8 were isolated and individually homogenized, and measurements were made using 10 μg of protein (n=7). The assay was performed following the manufacturer's protocol.

Statistical Analysis

Statistical significance was determined by two-tailed Student's T test for groups of two and by analysis of variance (ANOVA) for groups of 3 or more. Data are presented as mean + SEM. A confidence interval of 95% was used for all comparisons. All experiments represent the result of at least 3 biological replicates. The statistical significance for the Kaplan-Meier analysis was determined by Mantel-Cox and Wilcoxon tests.