Astrocytes from Familial and Sporadic ALS Patients are Toxic to Motor Neurons

Amyotrophic lateral sclerosis (ALS), commonly referred to as Lou Gehrig’s disease, is a fatal neurodegenerative disease characterized by loss of motor neurons (MNs) in the motor cortex, brain stem, and spinal cord, resulting in muscle paralysis and ultimately death due to respiratory failure. Approximately 90% of all cases are classified as sporadic ALS (sALS), defined as having no family history of the disease⁶. The remaining cases are inherited in a dominant fashion termed familial (fALS) of which 20% are associated with mutations within the superoxide dismutase 1 (SOD1) gene⁶. Numerous hypotheses have been proposed to account for MN loss, however the precise molecular mechanisms leading to ALS remain unknown⁷. Studies performed in fALS mouse models have implicated non-neuronal cells such as microglia and astrocytes in the progression phase of fALS^{2, 8–11}. In particular, in vitro co-culture systems have shown that MNs perish in the presence of astrocytes harboring SOD1 mutations^{1, 3–5}. However, all of these in vitro and in vivo studies have been conducted using models that highly overexpress mutant SOD1, which may not fully mimic the actual disease. Despite significant knowledge gained from these studies, no efficacious drugs or therapies have been realized in human clinical trials to date. Indeed, models of ALS have come under scrutiny, given that drugs and therapies are tested in models of fALS, only accounting for a small fraction of ALS patients. Thus, efforts have been focused on developing in vitro cell based models for sALS, representing the majority of the patient population. Currently, there is a major impetus to develop in vitro cell-based models for ALS, utilizing induced pluripotent stem cell (iPSC) technology^{12, 13}, although no reports have been published to date on sALS models. The generation of sALS disease models will be important to decipher whether astrocytes influence disease progression in sALS. There is precedent that glia are involved in the pathogenesis of sALS, since fALS and sALS patients show strikingly similar spinal cord pathologies with pronounced glial activation and upregulation of pro-inflammatory mediators and cytokines¹⁴. Indeed, in our analysis of ALS patient spinal cord samples, we observed increased glial cell numbers, which is indicative of astrocyte activation in the presence of atrophied MNs in both fALS and sALS patients (Supplementary Fig. 1,2). What remains to be determined is whether sALS astrocytes play an active role in the death of MNs, and whether this non-cell autonomous effect can be recapitulated in vitro. Here, we show that sALS and fALS astrocytes, derived from post-mortem spinal cord neural progenitor cells (NPCs), share a common non-cell autonomous toxicity, selectively killing MNs in a co-culture model system. Upon co-culture with MNs, we observed that inflammatory gene expression is upregulated within fALS and sALS astrocytes. In addition, we demonstrate that suppression of wild-type SOD1 in sALS astrocytes is neuroprotective in our model system, further implicating a role of wild-type SOD1 in sALS.

In order to test whether sALS astrocytes are similarly neurotoxic as fALS astrocytes, we derived astrocytes from adult NPCs isolated from post-mortem lumbar spinal cord tissue from fALS and sALS patients. We hypothesized that isolation of NPCs from these patients would provide a renewable source of multipotent cells that could be differentiated into astrocytes that were then used for co-culture studies with MNs, as previously described¹⁵. Sample collection was coordinated by The National Disease Research Interchange (NDRI) and all tissues were received within a time window of 48 to 72 hours post-mortem for immediate processing in our laboratory. Samples were non-identifiable with the exception of critical information regarding clinical diagnosis, gender, age at death, and length of time from diagnosis to death (Supplementary Table 1). To purify NPCs from these spinal cords, we utilized an enzymatic digestion protocol to dissociate the cells, followed by percoll gradient NPC enrichment, as previously described (Supplementary Fig. 3a)^{16, 17}. Within 3–4 weeks, NPC cultures were established, and marker analysis showed robust levels of the NPC marker; NESTIN (Supplementary Fig. 3b,c). Furthermore, the NPCs proved to be tri-potent, capable of differentiating into neurons, oligodendrocytes, and astrocytes (Supplementary Fig. 3d–f). NPC cultures were successfully established from one fALS patient harboring a SOD1 A4V mutation (Supplementary Fig. 4), 7 sALS patients, and one non-ALS control. Further DNA sequencing verified that our cohort of sALS patients did not harbor mutations in common ALS gene loci (Supplementary Table 2). To increase the rigor of our analyses, we utilized two additional non-ALS controls: SCP-27; a fetal human NPC line, and 1800; a fetal human primary astrocyte cell line (Supplementary Table 1).

After initial expansion of all NPC lines, cells were differentiated into astrocytes by medium supplementation with 10% fetal bovine serum. Since NPCs can give rise to multiple cell types in these differentiation conditions, we performed an extensive marker analysis of our established astrocyte cell lines to evaluate the purity of our cultures. Indeed, we found high levels of expression of well-known astrocyte markers including VIMENTIN, GFAP, and the astrocyte precursor marker CD44 (Fig. 1a and Supplementary Table 3), without significant contamination of other cell types. We found no detectable oligodendroglial progenitors or microglia in our cultures as assessed by immunohistochemistry for NG2 and CD11B, respectively (Fig. 1a and Supplementary Table 3). The absence of microglia in the cultures was also confirmed by quantitative RT-PCR for CD11B and IBA1 (Fig. 1b and Supplementary Fig. 5). Additionally, we evaluated the maturity of astrocytes by measuring expression of a panel of genes known to be enriched in astrocytes in vivo¹⁸. NPC-derived astrocytes shared a similar gene profile to primary astrocytes isolated directly from the human spinal cord (Fig. 1c). All of these results suggest that highly enriched astrocyte lines can be derived from NPCs isolated from postmortem spinal cord tissues, which represent a renewable source of human fALS and sALS astrocytes for subsequent analysis.

We next explored whether NPC-derived astrocytes from fALS and sALS patients were toxic to MNs by co-culturing mouse embryonic stem cell derived MNs with the differentiated astrocytes from each patient. We utilized an embryonic stem cell line containing an Hb9-eGFP reporter to allow for easy visualization of these MNs in co-culture with astrocytes¹⁹. For reproducibility of experiments, we enriched for Hb9-eGFP+ MNs by fluorescence activated cell sorting (Supplementary Fig. 6). Twenty-four hours following plating of the MNs onto a confluent layer of astrocytes, we found no significant difference in the number of MNs present between ALS and control astrocyte cultures (Fig. 2a), nor any significant difference in MN soma diameter or length of neuritic extensions up to 48 hours following co-culture initiation (Fig. 2b,c). However, after 96 hours in culture, MNs that were cultured on top of fALS astrocytes began to degenerate, evidenced by atrophy of the MN soma and shortened neurites (Fig. 2b, c). Indeed, quantification of MNs at 120 hours following co-culture showed that the number of MNs present in fALS astrocyte co-cultures was statistically reduced by 50% compared to non-ALS astrocyte control co-cultures (Fig. 2a, d). Similar results were observed when either SOD1 A4V or SOD1 G93A transgenes were overexpressed in non-ALS control astrocyte lines (Supplementary Fig. 7), supporting previous co-culture studies showing toxicity of astrocytes overexpressing mutant SOD1 ^{1, 3–5, 15, 20}.

Next, we shifted our focus to examine whether sALS astrocytes exhibit a similar non-cell autonomous toxicity towards MNs. Similar to our previous experiments utilizing fALS astrocytes, we noticed no significant changes in MNs co-cultured with sALS astrocytes for up to 48 hours. However, by 96 hours of co-culture, MNs co-cultured with all lines of sALS astrocytes began to degenerate (Fig. 2b). Quantification of MNs remaining after 120 hours of co-culture with sALS astrocytes showed a reduction of 45–70% compared to non ALS-controls (Fig. 2a, d). The MN damage elicited by the sALS patient-derived astrocytes was indistinguishable from fALS astrocytes, suggesting a shared mechanism for MN death in both types of co-cultures. To verify that the observed MN toxicity was caused by astrocyte-specific factors, we co-cultured MNs with ALS patient-derived fibroblasts and observed no significant decline in MN survival compared to MNs co-cultured with control fibroblasts (Supplementary Fig. 8a). To determine whether the fALS or sALS astrocytes affect survival of other neuronal cell types in vitro, we co-cultured GABAergic (gamma aminobutyric acid positive) neurons with ALS patient-derived astrocytes. Neither the fALS nor sALS astrocytes triggered death of GABAergic neurons, suggesting the ALS astrocyte-derived toxicity is specific toward MNs, thereby adding confidence to our assay (Supplementary Fig. 9a). Collectively, these results demonstrate that sALS astrocytes are also toxic towards MNs, and support the idea of a shared disease mechanism between fALS and sALS.

Previous work has shown that astrocytes isolated from the fALS SOD1 mouse model can cause MN death in vitro via secreted factors ^{1, 5}. To determine whether our patient-derived fALS and sALS astrocytes also release neurotoxic factors, we treated MNs with astrocyte-conditioned media and measured MN survival. MNs cultured in conditioned media from sALS or fALS astrocytes died ~50% faster than MNs treated with conditioned media from non-ALS control astrocytes (Fig. 2e, f). Conditioned media collected from ALS patient-derived fibroblasts did not cause MN death, demonstrating the specificity of the astrocyte-derived toxicity (Supplementary Fig. 8b). Likewise, GABAergic neurons were unaffected by treatment with conditioned media prepared from fALS or sALS astrocytes indicating that MNs are selectively sensitive to factors present in the ALS astrocyte-conditioned media (Supplementary Fig. 9b). These results suggest that fALS and sALS astrocytes either secrete factors which are toxic to MNs, or alternatively do not provide proper MN supportive factors, resulting in cell death.

In order to investigate potential causes of the toxicity originating from fALS and sALS astrocytes, we examined genes involved in inflammatory signaling, since previous reports have documented heightened inflammatory gene activation in fALS astrocytes^{1, 3}. We utilized quantitative Reverse Transcriptase polymerase chain reaction (qRT-PCR) array technologies to assay the expression of a functionally diverse set of 84 genes that includes both cytokine and chemokine ligands and receptors, and other intermediary signaling proteins all involved in inflammatory gene regulation. Gene expression analysis across several fALS and sALS astrocyte lines revealed that 35–60% of total inflammatory genes assayed were upregulated compared to non-ALS control astrocytes (Fig. 3 and Supplementary Table 4). We observed a degree of heterogeneity between our ALS astrocytes as expected with a varied disease etiology. However, clustering analysis revealed a set of 22 upregulated genes encompassing chemokines, proinflammatory cytokines, as well as components of the complement cascade, many of which have been previously implicated in ALS²¹ (Fig. 3b). We next performed a network-based pathway analysis using Ingenuity Pathway Analysis software, which identified 4 networks ranked by a statistical scoring method (Supplementary Table 5). The NF-κB signaling complex was identified as a major interactor within the highest ranked network (Score=33) based upon the high number of connections and its centralized location within the interactome (Supplementary Fig. 10). Additionally, IFN-α was also identified within this gene network interactome. Within the second ranked network (Score=12), pathways involving kinase signaling such as MAPK, JNK, and AKT, which establish a high number of interactions with the inflammatory genes in this cluster were identified (Supplementary Fig. 11). Many of these signaling networks have been previously implicated in fALS, and we now show evidence for their potential involvement in mediating inflammatory responses in sALS astrocytes^{22, 23}.

To increase the confidence of our assays, we determined whether suppression of mutant SOD1 in the fALS astrocytes would rescue MN death using a lentivirus encoding a shRNA ²⁴ to specifically knockdown SOD1 levels. Suppression of mutant SOD1 expression in our fALS astrocytes resulted in a full rescue of astrocyte-mediated MN toxicity as compared to the non-ALS control line (Fig. 4a, b). As expected, the missense shRNA control did not alter the fALS astrocyte-mediated toxicity towards MNs (Fig. 4b). The indistinguishable pathological and clinical features of fALS and sALS have long insinuated a shared disease mechanism between these forms of ALS and recently SOD1 has been proposed as a common link. Several studies have suggested that wild-type SOD1 can adopt conformational changes in sALS patients that are similar to the known pathological forms of SOD1 mutated in fALS ^25–27. These findings led us to test whether shRNA-based suppression of wild-type SOD1 in the sALS astrocytes would confer neuroprotection to MNs. In order to test this hypothesis, we performed experiments similar to our SOD1 knockdown in fALS astrocytes. All MNs co-cultured with sALS astrocytes expressing the missense shRNA control showed significant cell death, consistent with previous experiments. Remarkably, when we suppressed SOD1 in our sALS astrocytes, we observed statistically significant neuroprotective effects in 4 of the 6 sALS astrocyte lines (Fig. 4b). In the remaining 2 sALS astrocytes lines, we observed a trend of neuroprotection, although these results were not significant. This lower level of neuroprotection may reflect the observation that SOD1 was not efficiently suppressed in these lines, as quantified by ELISA analysis (Fig 4c), even though the level of lentiviral transduction efficiency was equivalent among all the astrocyte lines. Collectively, these results suggest that suppression of SOD1 in sALS astrocytes is beneficial and can confer significant MN protection from ALS astrocyte-derived toxicity, when greater than 50% of wild-type SOD1 knockdown is achieved. These results support previous findings ^25–27 that wild-type SOD1 should be considered a possible therapeutic target for sALS.

A major challenge in the field of ALS research has been to evaluate the similarities and differences between fALS and sALS. Here, we tested whether astrocytes derived from the more common form of the disease, sALS, showed a similar glial-mediated toxicity towards MNs as demonstrated in various fALS models. Our results indicate a shared mechanism leading to MN death between fALS and sALS through astrocyte-mediated toxicity. In addition, we provide evidence that wild-type SOD1 is involved in this toxicity, suggesting that therapies targeting SOD1 in astrocytes may be beneficial not only to fALS patients with SOD1 mutations, but also to sALS patients as well. Recently, novel vectors have been shown to efficiently target astrocytes of adult brain and spinal cord²⁸ and constitute an attractive therapeutic option. Alternatively, astrocyte replacement has also been proposed as a potential therapy for slowing disease progression in ALS, and stem cell transplants are currently in human clinical trials²⁹.

Although glia represents a potential therapeutic target in ALS, an ultimate goal in the ALS research field is to understand why MNs perish in the disease. Of note, our studies presented here investigated only ALS astrocyte-derived toxicity on non-ALS derived MNs. The derivation of patient-specific MNs will be important to further evaluate pathological abnormalities in these cell types. To aid in this endeavor, we recently demonstrated that human MNs can be efficiently and rapidly generated from pluripotent stem cells³⁰. This novel strategy should be advantageous in generating MNs from patient-specific iPS cells. In sum, this work introduces the first model system to investigate molecular disease mechanisms and evaluate potential therapies for sALS, and highlights the great potential to use patient-specific samples for studying complex diseases.