https://orcid.org/0000-0001-9676-8548
Abstract
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease characterized by the loss of upper and lower motor neurons (MNs). The loss of MNs in ALS leads to muscle weakness and wasting, respiratory failure, and death often within two years of diagnosis. Glial cells in ALS show aberrant expression of pro-inflammatory and neurotoxic proteins associated with activation and have been proposed as ideal therapeutic targets. In this study, we examined astrocyte-targeted treatments to reduce glial activation and neuron pathology using cells differentiated from ALS patient-derived iPSC carrying SOD1 and C9ORF72 mutations. Specifically, we tested the ability of increasing interleukin 10 (IL-10) and reducing C-C motif chemokine ligand 2 (CCL2/MCP-1) signaling targeted to astrocytes to reduce activation phenotypes in both astrocytes and microglia. Overall, we found IL10/CCL2NAb treated astrocytes to support anti-inflammatory phenotypes and reduce neurotoxicity, through different mechanisms in SOD1 and C9ORF72 cultures. We also found altered responses of microglia and motor neurons to astrocytic influences when cells were cultured together rather than in isolation. Together these data support IL-10 and CCL2 as non-mutation-specific therapeutic targets for ALS and highlight the role of glial-mediated pathology in this disease.
Introduction
Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disease. The primary phenotype of ALS is progressive loss of both upper and lower motor neurons resulting in muscle weakness, paralysis, and death within 2–5 years of diagnosis [1]. Though the vast majority of ALS cases are sporadic in nature with no genetic cause, there are inherited cases of familial ALS. Mutations in over 40 causative genes have been identified in familial ALS, the most common of which are mutations in superoxide dismutase 1 (SOD1) and chromosome 9 open reading frame 72 (C9ORF72) [2]. The mechanisms of how these mutations contribute to disease progression and lead to motor neuron degeneration are still unclear. Previous studies have identified the ability of both astrocytes and microglia to contribute to motor neuron toxicity in ALS [3, 4], and glial activation can be localized to degenerating motor neurons (MNs) and associated with disease progression in ALS patients [5]. ALS patient-derived induced pluripotent stem cells (iPSCs) differentiated into astrocytes, microglia, and MNs have allowed for characterization of mutation-specific ALS phenotypes. Clear differences have been identified in the transcriptional profiles of iPSC-differentiated MNs carrying SOD1 and C9ORF72 mutations [6], setting a precedence for mutation-specific mechanisms of pathology in ALS.
These mutation-specific phenotypes may clarify contrasting theories regarding whether glial pathology is initiated by astrocytes or microglia in ALS. In support of astrocyte-driven pathology, MNs exposed to C9ORF72 astrocytes have decreased firing [7], increased oxidative stress [4], and increased cell death [4, 8]. There were early reports of SOD1 astrocyte contribution to MN loss [3] and improvements to lifespan when mutant SOD1 protein was knocked down in astrocytes [9]; however, these data were from in vivo studies where microglia are also present. Interestingly, SOD1 microglial activation was found to occur before disease onset in SOD1 mice [10], and targeting these pro-inflammatory microglia [10, 11] or their production of astrocyte-activating factors [12] has shown to extend survival. A recent transcriptomic study comparing across human and mouse models of ALS attributed production of reactive oxygen species and overall senescent phenotypes to C9ORF72 astrocytes, whereas SOD1 astrocytes were found to upregulate pathways involved in reactivity to stimuli [13]. These data support differing astrocyte-driven dysfunction in C9ORF72 models while microglia appear to initiate aberrant activation in SOD1 models.
Increased inflammation is shared characteristic across ALS mouse models and patient samples [14], and C-C motif chemokine ligand 2 (CCL2/MCP-1) has been specifically identified in patient CSF [15–18], and its expression localized to astrocytes in patient spinal cord [19]. Importantly, this CCL2 upregulation is found in both familial and sporadic ALS patients; therefore, it may be an ideal candidate for therapeutics. We have previously found microglia secreting high levels of the anti-inflammatory cytokine interleukin 10 (IL-10) capable of reducing astrocyte-induced ALS phenotypes in a sporadic ALS patient iPSC model [20]. Anti-inflammatory microglia expressing high levels of IL-10 were also found to delay disease onset and extend survival in an SOD1 mouse model [21]. In this study, we differentiated astrocytes, microglia, and motor neurons from ALS patient-derived iPSCs carrying SOD1 and C9ORF72 mutations in order to compare the effects of astrocyte-targeted anti-inflammatory treatments. Specifically, we hypothesized that increasing IL-10 and reducing CCL2 signaling in astrocytes would reduce glial-driven inflammation and motor neuron dysfunction in both SOD1 and C9ORF72 cultures. IL-10 treated astrocytes have shown to reduce microglial activation better than treating microglia directly [22], so this treatment paradigm was designed to address both microglial and astrocyte dysfunction. We confirmed the overexpression of CCL2 by C9ORF72 astrocytes and found that SOD1 astrocytes express CCL2 after exposure to SOD1 microglia. IL10/CCL2NAb treatment induced less reactive phenotypes for both C9ORF72 and SOD1 conditions, although specific signaling pathways altered by the treatment and the downstream impact on microglia and MNs diverged between SOD1 and C9ORF72 backgrounds. These data reveal the importance of paracrine signaling between glial cells and neurons in ALS and support the potential for astrocyte-targeted anti-inflammatory treatments to reduce ALS pathology.
Results
Spinal cord patterned astrocytes were differentiated using established protocols [20, 23, 24] from ALS patient-derived induced pluripotent stem cells (iPSCs) diagnosed with familial ALS due to a mutation of superoxide dismutase 1 (SOD1, A4V) or C9ORF72 hexanucleotide repeat expansion. Mutated SOD1 and C9ORF72 astrocytes appeared morphologically similar (Fig. 1A) and expressed consistent transcript levels for astrocyte markers including glial fibrillary acidic protein (GFAP) and aldehyde dehydrogenase 1 family member L1 (ALDH1L1, Fig. 1B). There were no differences in transcript expression for the pro-inflammatory mediator NFκB. As NFκB is known to drive expression of pro-inflammatory signaling molecules, we assessed interleukin 1 beta (IL-1β) and interleukin 6 (IL-6) as well as complement components C3 and C1q and found no difference in their expression levels (Fig. 1C–E, 2-way ANOVAs, ns). Transcripts for glial-derived and brain-derived neurotrophic factors (GDNF, BDNF) were also similarly expressed in baseline cultures of SOD1 and C9ORF72 astrocytes (Fig. 1F, 2-way ANOVA, ns).
iPSC-differentiated astrocytes from ALS patients with SOD1 and C9ORF72 mutations have similar inflammatory phenotypes. (A) ALS patient-derived iPSCs carrying either SOD1 (left) or C9ORF72 (right) mutations differentiated into astrocytes have similar morphology (representative image, 20× objective) and expression of astrocyte marker transcripts GFAP and ALDH1L1 (B, 2-way ANOVA, ns). (C) Transcripts for NFκB and downstream pro-inflammatory ligands IL-1B and IL-6 (D) are readily detected in both SOD1 and C9ORF72 astrocytes with no differences in expression (2-way ANOVAs, ns). (E) Transcripts for complement components C1q and C3 as well as neurotrophic factors (F, GDNF, BNDF) have no differences in expression between SOD1 and C9ORF72 astrocytes (2-way ANOVA, ns). (G) Grouped ALS astrocyte conditioned media (ACM) was found to contain 20-fold higher expression of pro-inflammatory ligand CCL2 compared to healthy control (HC) ACM. (H) Though SOD1 ACM had similar expression of CCL2 to HCs at baseline, increased secretion of CCL2 was induced to the level of C9ORF72 ACM when SOD1 astrocytes were exposed to media conditioned by SOD1 iPSC-derived microglia (1-way ANOVA, **P < 0.005). Alt text: Panel A shows a microscopic image of astrocytes from both of the patient lines used. Panels B–F indicate that there are no differences in the baseline characteristics between the two patient lines. Panels G and H show that the cytokine CCL2 is highly upregulated in patient astrocytes compared to control astrocytes.
To explore the secretome profiles of ALS astrocytes, we examined astrocyte conditioned media (ACM) via multiplex human cytokine array. To first assess the top dysregulated cytokines based on ALS disease status, values from baseline (untreated) SOD1 and C9ORF72 ACM were pooled with ACM from a sporadic ALS line and compared to ACM generated from healthy control (HC) astrocytes. We identified C-C motif chemokine 2 (CCL2, also known as MCP-1) as the top dysregulated cytokine with approximately 20-fold increased expression in ALS ACM compared to HC ACM (Fig. 1G). Analysis of ALS samples by specific mutation revealed that while baseline C9ORF72 astrocytes had significantly higher expression of CCL2, baseline SOD1 astrocytes had similar levels of CCL2 expression to HC astrocytes (Fig. 1H, 1-way ANOVA, HC vs SOD1 P = 0.8004, HC vs C9ORF72 **P = 0.0021). However, SOD1 astrocytes could be induced to CCL2 expression levels observed in C9ORF72 astrocytes when SOD1 astrocytes were exposed to media conditioned by SOD1 microglia (Fig. 1H, 1-way ANOVA, SOD1C9ORF72HC vs SOD1 MCM treated **P = 0.0020). Importantly, these data are consistent with previous reports of high CCL2 expression in ALS patients [15–18].
Because of the proposed role of microglia in ALS pathology [10], we next investigated the influence of ACM on activation phenotypes and phagocytic ability of SOD1 and C9ORF72 iPSC-differentiated microglia (Fig. 2A). Microglia were exposed to 100 ng/ml lipopolysaccharide (LPS), HC ACM, mutation-specific ALS ACM (SOD1 ACM or C9ORF72 ACM), or no treatment (UTX) for 24 h and assessed for changes in soma size as an indication of priming [20, 25–27]. SOD1 microglial soma size increased in response to pro-inflammatory LPS and decreased in response to HC ACM showing a possible benefit of healthy astrocyte-secreted factors on SOD1 microglial priming (Fig. 2B, 1-way ANOVA, UTX vs LPS ****P < 0.0001, UTX vs HC ACM ****P < 0.0001). SOD1 ACM did not reduce microglial soma size (HC ACM vs SOD1 ACM, **P = 0.0011). C9ORF72 microglia were also reactive to pro-inflammatory LPS (Fig. 2C, 1-way ANOVA, UTX vs LPS *P = 0.0213), but they showed a decreased soma size in response to both HC and C9ORF72 ACM (UTX vs HC **P = 0.0023, UTX vs C9ORF72 ACM **P = 0.0019). To examine the effects of ACM on microglial phagocytosis, live imaging of SOD1 and C9ORF72 microglia was performed to quantify phagocytosis of pH-indicator beads (pHrodo) for 24 h after treatment (Fig. 2D). In SOD1 microglia, LPS induced a trend for higher phagocytosis (Fig. 2E, 1-way ANOVA, UTX vs LPS P = 0.0815), which was significantly higher than both HC and SOD1 ACM treatment conditions (LPS vs HC ACM**P = 0.0050, LPS vs SOD1 ACM *P = 0.0288). However, neither HC nor SOD1 ACM treatments significantly altered phagocytosis compared to untreated SOD1 microglia (UTX vs HC P = 0.2563, UTX vs SOD1 P = 0.6418). Despite effects of LPS and ACM on C9ORF72 microglial soma size, LPS, HC ACM, and C9ORF72 ACM treatments did not have any effects on C9ORF72 microglial phagocytosis (Fig. 2F, 1-way ANOVA, ns).
SOD1 and C9ORF72 microglial activation and motor neuron function are not improved by exposure to ACM. SOD1 and C9ORF72 iPSC-differentiated microglia (A, representative image, 40× objective) have altered responses to ACM. (B) SOD1 microglia increase soma size when treated with pro-inflammatory lipopolysaccharide (LPS) and decrease soma size when exposed to HC ACM whereas SOD1 ACM does not induce this effect (1-way ANOVA, **P < 0.005, ****P < 0.0001). (C) C9ORF72 microglia increase soma size in response to LPS and decrease soma size in response to both HC and C9ORF72 ACM (1-way ANOVA, *P < 0.05, **P < 0.005). (D) Phagocytosis of pH-indicator beads (pHrodo, 40× objective) is significantly decreased in SOD1 microglia by both HC (**P < 0.005) and SOD1 ACM (*P < 0.05) compared to LPS-treated but not to untreated SOD1 microglia (E, 1-way ANOVA, ns). (F) C9ORF72 microglial phagocytosis was not influenced by exposure to LPS, HC ACM, or C9ORF72 ACM (1-way ANOVA, ns). (G) Neither SOD1 or C9ORF72 iPSC-derived motor neurons (representative images, 20× objective) increased calcium flux in response to depolarization with KCl after exposure to HC or ALS ACM (H, I, 1-way ANOVAs, ns). Alt text: Panel A shows a microscopic image of patient microglia, and panels B and C show that patient astrocyte medium impacts microglial size for the SOD1 sample. Panel D shows a microscopic image of the fluorescent beads internalized by patient microglia, and panels E and F quantify how microglia engulfment in impacted by SOD1 astrocyte medium. Panel G shows a microscopic image of patient motor neurons, and panels H and I show that motor neuron calcium function is unaffected by patient astrocyte medium.
SOD1 and C9ORF72 iPSCs were next differentiated in motor neurons (MNs) to determine the influence of astrocyte secreted factors on MN function (Fig. 2G). Several groups have shown iPSC-derived SOD1 and C9ORF72 MNs have increased cell death [28–30] and abnormal function [28, 31, 32] compared to control iPSC-derived MNs. To assess the direct effects of astrocyte-secreted factors on these phenotypes, MNs were treated with HC ACM, ALS ACM (SOD1 or C9ORF72 ACM), or no treatment (UTX). After 48 h of treatment, SOD1 and C9ORF72 MNs were assessed for calcium flux in response to depolarization with KCl. In these MN monocultures, neither HC nor disease ACM was able to significantly alter the calcium response of SOD1 or C9ORF72 MNs compared to untreated cultures (Fig. 2H and I, 1-way ANOVAs, ns). There was a trend for increased calcium flux in SOD1 MNs with HC ACM compared to UTX (P = 0.1524), but this was not found with SOD1 ACM treatment.
Based on the recent finding that microglial activation is better reduced by anti-inflammatory astrocytes than direct microglial treatments [22], we proposed that astrocyte-directed treatments would appropriately reduce both astrocyte and microglial-driven pathology in ALS. To test this hypothesis, we treated SOD1 and C9ORF72 astrocytes with 430 pg/ml of recombinant human interleukin 10 (IL-10). IL-10 is shown to increase anti-inflammatory phenotypes through inhibition of NFκB and by NFκB-independent mechanisms [33–35] as well as provide direct trophic support to neurons and reduce microglial pro-inflammatory cytokine production [36, 37]. To measure normalization of microglial phenotypes, we combined IL-10 treatment with 2 ng/ml of CCL2 neutralizing antibodies (NAb) (IL10/CCL2NAb, Fig. 3A). A 48-h wash condition was also included to control for IL-10 or CCL2 NAb remaining in ACM and to assess non-continuous treatment on astrocyte phenotypes. We evaluated transcript levels of multiple factors (Supplemental Fig. S1) following treatment and found that BDNF transcripts were increased in C9ORF72 IL10/CCL2NAb-treated astrocytes (Fig. 3B, 2-way ANOVA, UTX vs IL10/CCL2NAb *P = 0.0285) and a trend toward increase was also seen in treated SOD1 astrocytes (UTX vs IL10/CCL2NAb P = 0.1036). IL10/CCL2NAb C9ORF72 astrocytes also had decreased C1q transcripts that remained decreased 48-h after treatment (Fig. 3C, 2-way ANOVA, UTX vs IL10/CCL2NAb *P = 0.0208, UTX vs 48wash *P = 0.0420), whereas SOD1 astrocytes had no changes to C1q transcript production (ns). A decrease (SOD1, *P = 0.0495) or trend for decrease (C9ORF72, P = 0.0602) in IL-1β transcripts was seen in all treated astrocytes 48-h after IL10/CCL2NAb treatment (Fig. 3D, 2-way ANOVA). Interestingly, IL10/CCL2NAb treated C9ORF72 astrocytes increased transcripts for IL-6, which could be either neuroprotective or pro-inflammatory in ALS [38], though this effect was not found in the 48wash condition (Fig. 3E, 2-way ANOVA, UTX vs IL10/CCL2NAb **P = 0.0015, UTX vs 48wash P = 0.2698) nor in treated SOD1 astrocytes (ns).
Increase of IL-10 and neutralization of CCL2 differently but beneficially alter SOD1 and C9ORF72 astrocyte phenotypes. (A) Schematic of ALS glial-mediated activation and motor neuron damage (left) and hypothesized effects of astrocyte-targeted treatments with 430 pg/ml recombinant human IL-10 and 2 ng/ml of neutralizing CCL2 antibodies (right, IL10/CCL2NAb). (B) C9ORF72 astrocytes increased expression of BDNF transcripts with IL10/CCL2NAb and decreased transcripts for C1q (C) both during treatment and 48 h after treatment was removed (48wash), while SOD1 astrocytes had no changes in expression of these targets (2-way ANOVAs, *P < 0.05, ns). (D) SOD1 astrocytes did show decreased transcript expression for IL-1β 48 h after IL10/CCL2NAb (2-way ANOVA, *P < 0.05) E. C9ORF72 astrocytes also had a surprising increase in IL-6 expression during IL10/CCL2NAb, though this effect was ameliorated in 48wash condition (2-way ANOVA, **P < 0.005, *P < 0.05). (F) Western blot analyses identify increased anti-inflammatory and decreased pro-inflammatory protein expression after IL10/CCL2NAb in SOD1 and C9ORF72 astrocytes. IL10/CCL2NAb treatment increased protein expression of IL-10 in SOD1 astrocytes and TGFb in C9ORF72 astrocytes (G, 2-way ANOVA, *P < 0.05, **P < 0.005). Both SOD1 and C9ORF72 astrocytes had decreased protein expression of pro-inflammatory mediator NFκB 48 h after IL10/CCL2NAb treatment, though no changes to phosphorylated NFκB (phNFκB, 2-way ANOVA, **P < 0.005, *P < 0.05, ns). Alt text: Panel A is a cartoon overview of the experimental hypothesis, and the remaining panels show how transcripts and protein are altered in patient astrocytes following treatment with IL-10 over-expression and CCL2 neutralization.
We next evaluated intracellular changes in anti- and pro-inflammatory protein expression via Western blot on collected cell pellets of IL10/CCL2NAb treated astrocytes (Fig. 3F). Production of IL-10 by SOD1 astrocytes was increased with IL10/CCL2NAb treatment and had a trend for sustained increased in 48wash condition (Fig. 3G, 2-way ANOVA, UTX vs IL10/CCL2NAb *P = 0.0207, UTX vs 48wash P = 0.0624). C9ORF72 astrocytes did not significantly increase IL-10 protein production, but treatment did increase anti-inflammatory TGF-β protein (Fig. 3G, 2-way ANOVA, UTX vs IL10/CCL2NAb **P = 0.0050). Both SOD1 and C9ORF72 astrocytes had decreased total NFκB protein 48-h after treatment (Fig. 3G, 2-way ANOVA, SOD1 UTX vs 48wash **P = 0.0096, C9ORF72 UTX vs 48wash *P = 0.0274) and no changes to phosphorylated NFκB (ns). Although signaling pathways activated by IL10/CCL2NAb treatment may be mutation-specific, patterns of decreased pro-inflammatory and increased anti-inflammatory signaling are apparent in both SOD1 and C9ORF72 IL10/CCL2NAb-treated astrocytes.
To assess the efficacy of astrocyte-targeted IL10/CCL2NAb treatment on microglial properties, we treated SOD1 and C9ORF72 microglia with ACM from IL10/CCL2NAb and 48wash astrocytes for 24 h (Fig. 4A). SOD1 microglia responded to SOD1 IL10/CCL2NAb ACM and 48wash ACM with a decrease in soma size compared to SOD1 ACM (Fig. 4B, 1-way ANOVA, p****P < 0.0001), whereas the already small C9ORF72 microglia soma size was not negatively impacted by IL10/CCL2NAb treatment (Fig. 4C, 1-way ANOVA, ns). SOD1 microglia had a decreased trend for phagocytosis of pH-indicator beads with SOD1 48wash ACM (Fig. 4D and E, 1-way ANOVA, UTX vs IL10/CCL2NAb P = 0.2219, UTX vs 48wash P = 0.0589). Although not quite significant, the separation of C9ORF72 microglial phagocytosis after exposure to IL10/CCL2NAb and 48wash is notable compared to the LPS and HC ACM treated groups in 2F (Fig. 4E, 1-way ANOVA, UTX vs IL10/CCL2NAb ACM P = 0.1458, IL10/CCL2NAb ACM vs 48wash P = 0.0553).
IL10/CCL2NAb-treated astrocytes may influence microglial activation but not motor neuron function in mono-cultures. (A) Representative image of SOD1 and C9ORF72 microglia exposed to untreated or IL10/CCL2NAb ACM (20x objective). (B) SOD1 microglia treated with SOD1 IL10/CCL2NAb ACM do show an improvement in priming as measured by soma size (1-way ANOVA, ****P < 0.0001) while C9ORF72 microglia remain in unprimed status regardless of ACM treatment (C, 1-way ANOVA, ns). (D) Both IL10/CCL2NAb and 48wash ACM from SOD1 astrocytes show trends for decreased phagocytosis of pH-indicator beads by SOD1 microglia compared to untreated (1-way ANOVA, P > 0.05), while only IL10/CCL2NAb ACM has a trend for decreased phagocytosis in C9ORF72 microglia (E, 1-way ANOVA, P > 0.05). Direct treatment of microglia with IL-10 and CCL2 neutralizing antibodies did have a similar effect on SOD1 microglia as IL10/CCL2NAb ACM, while it did not change phagocytic ability of C9ORF72 microglia (1-way ANOVAs, SOD1: IL10/CCL2NAb vs direct treatment P = 0.7820, C9ORF72: UTX vs direct treatment P = 0.9439). Neither SOD1 (F) nor C9ORF72 (G) motor neurons had increased calcium flux after exposure to IL10/CCL2NAb ACM, 48wash ACM, or direct IL10CCL2 treatments compared to untreated ACM (1-way ANOVAs, ns). SOD1 ACM and C9ORF72 ACM values in B, C, F, and G repeated from Fig. 2 for ease of comparison. Alt text: Panel A shows microscopic images of fluorescent bead internalization in patient microglia in the presence and absence of IL-10 expression and CCL2 neutralization treatment, and panels B-E quantify the impact of astrocyte targeted treatment on microglia size and function. Panels F and G show that motor neuron calcium function is not altered in response to IL-10 and CCL2 astrocyte modulation.
To compare astrocyte-targeted vs microglial-targeted treatments, we also applied 430 pg/ml IL-10 with 2 ng/ml CCL2 NAbs directly onto microglia with ALS ACM treatments. SOD1 microglia with direct IL10/CCL2NAb treatments had similar rates of fluorescent bead phagocytosis to SOD1 I IL10/CCL2NAb ACM, but only half the effect of SOD1 48wash ACM (Fig. 4D mean differences: SOD1 ACM vs IL10/CCL2NAb ACM = 145 beads, SOD1 ACM vs direct IL10/CCL2NAb = 92.44 beads, SOD1 ACM vs 48wash = 210.8 beads). Direct IL10/CCL2NAb treatments on C9ORF72 microglia were about one third as effective as astrocyte-targeted treatments (Fig. 4E mean differences: C9ORF72 ACM vs IL10/CCL2NAb ACM = 235.8 beads, C9ORF72 ACM vs direct IL10/CL2NAb = 88.64 beads). Application of IL10/CCL2NAb ACM, 48wash ACM, or direct IL10/CCL2NAb treatments of SOD1 or C9ORF72 MN monocultures did not increase calcium flux compared to untreated ACM (Fig. 4F and G, 1-way ANOVAs, ns). Together, these data suggest that astrocyte-targeted IL10/CCL2NAb treatment can reduce activation of SOD1 and C9ORF72 microglia and do so more effectively than microglia-directed treatments. However, improvements to ALS MN function may require influences greater than short term exposure to astrocyte secreted factors.
To further address the impacts of glial modulation on MNs, we examined the effects of astrocyte-targeted treatments in co-cultures of MNs and microglia to assess the importance of dynamic cell interactions on therapeutic effects. To allow for live cell imaging, we established stable fluorescent lines of SOD1 and C9ORF72 iPSCs using lentiviruses driving expression of either GFP or RFP under the elongation factor 1 (EF-1) promoter. These were then separately differentiated into GFP+ or RFP+ microglia and MNs before microglia were introduced to MN cultures (Fig. 5A). A trend for increased calcium flux was found in SOD1 MN + SOD1 microglia co-cultures when treated with SOD1 IL10/CCL2NAb ACM compared to untreated or SOD1 ACM treated cultures (Fig. 5B, 1-way ANOVA, UTX vs IL10/CCL2NAb P = 0.1006, SOD1 vs IL10/CCL2NAb P = 0.1362). In C9ORF72 co-cultures, C9ORF72 ACM significantly reduced calcium flux, an effect that was ameliorated by C9ORF72 IL10/CCL2NAb ACM (Fig. 5B, 1-way ANOVA, UTX vs C9ORF72 ACM *P = 0.0224, UTX vs IL10/CCL2NAb ACM ns). This is likely due to astrocyte-driven microglial activation, as these effects were not seen in earlier MN monocultures treated with C9ORF72 ACM.
IL10/CCL2NAb treated astrocytes reduce phagocytosis and apoptosis of SOD1 and C9ORF72 motor neurons in co-cultures with microglia. (A) GFP- and RFP-labeled MNs and microglia allow for live-imaging analyses of co-cultures (20× objective). (B) Exposure of SOD1 co-cultures to SOD1 ACM does not alter calcium flux, but a trend for increased flux can be found after IL10/CCL2NAb ACM exposure compared to untreated and SOD1 ACM co-cultures (UTX vs IL10/CCL2NAb P = 0.1006, SOD1 ACM vs IL10/CCL2NAb P = 0.1362). C9ORF72 co-cultures significantly decrease flux after C9ORF72 ACM treatment, with a trend for amelioration with IL10/CCL2NAb ACM treatment (UTX vs IL10/CCL2NAb P = 0.2145, C9ORF72 ACM vs IL10/CCL2NAb P = 0.2566). (C) Phagocytosis of co-cultures quantified using RFP+/GFP+ colocalization in microglia (yellow box—GFP/RFP double positive, black box—not double positive) after exposure to ACM via live imaging paradigm. (D) SOD1 ACM increases phagocytosis in SOD1 co-cultures which is ameliorated by IL10/CCL2NAb treatment and 48-h afterwards (1-way ANOVA, ****P < 0.0001). C9ORF72 ACM does not increase already high levels of phagocytosis observed in untreated C9ORF72 co-cultures, but this rate is significantly decreased by both IL10/CCL2NAb and 48wash ACM (1-way ANOVA, ***P < 0.0005, ****P < 0.0001). (E) TUNEL assay to assess MN apoptosis in SOD1 and C9ORF72 MN + microglia co-cultures (63× objective). (F) IL10/CCL2NAb ACM decreases apoptosis compared to ALS ACM in both SOD1 and C9ORF72 co-cultures (1-way ANOVAs, *P < 0.05, ***P < 0.0001). Alt text: Panel A shows a microscopic image of motor neuron-microglia co-cultures with panel B showing that motor neuron responses to astrocyte medium with and without IL-10 and CCL2 modulation are altered in C9ORF72 conditions. Panel C also shows microscopic images of these co-cultures and indicates fluorescent labels engulfed by microglia with panel D quantifying the amount of engulfment. Panel E shows microscopic images, and panels F and G quantify the amount of cell death occurring in motor neuron-microglia co-cultures in the presence and absence of astrocyte medium with and without IL-10 and CCL2 modulation.
To test this hypothesis, we quantified phagocytosis in these co-cultures as microglia became double-positive for RFP and GFP upon engulfment of fluorescent MN elements (Fig. 5C). 48 h after co-cultures were treated with ACM, GFP+/RFP+ microglia were significantly increased in SOD1 co-cultures with SOD1 ACM and significantly reduced in both SOD1 IL10/CCL2NAb and 48wash ACM (Fig. 5D, 1-way ANOVA, UTX vs SOD1 ****P < 0.0001, SOD1 vs IL10/CCL2NAb ****P < 0.0001, SOD1 vs 48wash ****P < 0.0001). C9ORF72 co-cultures had a high rate of phagocytosis that was not increased by C9ORF72 ACM, but both C9ORF72 IL10/CCL2NAb and 48wash ACM reduced these values (Fig. 5D, 1-way ANOVA, UTX vs C9ORF72 ACM P = 0.4659, C9ORF72 ACM vs IL10/CCL2NAb ****P < 0.0001, C9ORF72 ACM vs 48wash ***P = 0.0006). Further, we found that addition of microglia without astrocyte influence did not significantly alter calcium flux in either SOD1 or C9ORF72 MNs compared to MN monocultures (Supplemental Fig. S2, t-tests, ns).
Finally, we examined the effects of IL10/CCL2NAb astrocyte-targeted treatments on MN cell death in co-cultures using a TUNEL stain (Fig. 5E, Supplemental Fig. S3). We identified high baselines of MN cell death in untreated co-cultures with microglia in both SOD1 and C9ORF72 cultures (Fig. 5F). Exposure of SOD1 co-cultures to SOD1 ACM increased TUNEL+ MNs compared to SOD1 IL10/CCL2NAb ACM (1-way ANOVA, UTX vs SOD1 ACM P = 0.0732, SOD1 ACM vs IL10/CCL2NAb *P = 0.0147). C9ORF72 co-cultures had reduced MN death when treated with C9ORF72 IL10/CCL2NAb ACM compared to UTX, though this effect was not sustained in 48wash ACM condition (1-way ANOVA, UTX vs IL10/CCL2NAb **P = 0.0054, IL10/CCL2NAb vs 48wash ***P = 0.0007). Together, these data highlight convergent and divergent effects of glial-neuron interactions in both SOD1 and C9ORF72 iPSC cultures.
Discussion
Although only 10% of ALS patients have familial cases that can be linked to mutations in causative genes, these have historically been the focus of research [39, 40]. This is primarily due to the logistics of creating animal models for sporadic ALS without a specific genetic mutation to target. However, growing evidence supports the idea that the specific mechanisms of motor neuron loss may not be shared across ALS inherited mutations and sporadic cases [6, 41]. iPSC models of disease have allowed for more exploratory and comparative studies to begin addressing this. For example, SOD1 and C9ORF72 mutation-carrying iPSC-differentiated motor neurons have been shown to have unique transcriptional profiles, with defects in RNA processing and transport unique to C9ORF72 MNs and SOD1 MNs downregulating TGFβ, SMAD, and MAPK signaling [6]. Nevertheless, there are commonalities across ALS backgrounds. In addition to loss of upper and lower motor neurons in ALS, one shared hallmark of disease across familial and sporadic cases is glial-mediated inflammation [3–5]. Though it may be either microglia or astrocytes that initiate this activation, it has been proposed that converting these glial cells towards a more neuroprotective phenotype can prevent disease onset and symptom progression [42–46]. Based on this concept, we hypothesized that astrocyte-targeted therapeutics would prevent both astrocyte-driven neurotoxicity and promote anti-inflammatory microglial phenotypes.
In this study, we examined IL-10 as an anti-inflammatory therapeutic targeted to astrocytes carrying SOD1 and C9ORF72 mutations. We combined IL-10 expression with a neutralizing antibody treatment against CCL2—a cytokine capable of inducing microglial activation [47] that we and others have found highly upregulated by ALS astrocytes [15–19] (Fig. 1). Previous studies have tried directly targeting activated microglia with IL-10 treatments and found it ineffective without other anti-inflammatory treatments [48] or astrocyte-secreted factors [22], so we focused on astrocyte-directed anti-inflammation. Overall, we saw reduced pro-inflammatory factors, increased anti-inflammatory signaling, and increased neurotrophic support from IL10/CCL2NAb treated astrocytes (Figs 2 and 3), though the specific pathways activated were different between SOD1 and C9ORF72 cultures. At baseline, C9ORF72 astrocytes appeared to reside in a more pro-inflammatory state than SOD1 astrocytes, whereas SOD1 microglia were more reactive to astrocytic influences than C9ORF72 microglia (Fig. 2). Though microglia do express C9ORF72, microglia with mutations in C9ORF72 have recently been reported to be very similar to healthy controls unless exposed to extrinsic factors [49], thereby supporting an astrocyte-driven activation in C9ORF72-associated ALS. Though the exact function of the C9ORF72 is unclear, it is thought to interact with Rab proteins to regulate endosomal trafficking, autophagy, and lysosomal biogenesis [50]. Mutations in C9ORF72 in astrocytes are associated with abnormal RNA metabolism, increased NFκB, and altered glutamate regulation all of which can contribute directly to increased neurotoxicity [7, 51].
Fitting with in vivo datasets of upregulated CCL2 by SOD1 astrocytes [19], we found that SOD1 astrocytes could be induced to a pro-inflammatory status after exposure to SOD1 microglial conditioned media (Fig. 1H). A recent study found SOD1 astrocytes more reactive to pro-inflammatory stimuli than other ALS astrocytes [52], which supports the idea of microglia-driven activation in SOD1-associated ALS. This microglial activation could result from both gain and loss of function mechanisms due to SOD1 protein, as SOD1 is an antioxidant enzyme responsible for regulation of reactive oxygen species (ROS) [53]. Microglia monitor ROS [54] and increase Toll-like receptor (TLR)-mediated inflammatory responses in response to insufficient degradation of ROS [55, 56]. This mechanism, coupled with the gain of protein aggregates that microglia are tasked with clearing [57], may explain why SOD1 mutations are associated with activated microglia over activated astrocytes.
After IL10/CCL2NAb treatment (Fig. 3), SOD1 astrocytes decreased NFκB and increased production of IL-10—an ideal combination to reduce microglial-mediated inflammation. In support of this, application of IL10/CCL2NAb treated SOD1 ACM onto SOD1 microglia did reduce activation phenotypes (Fig. 4). Treated C9ORF72 astrocytes increased TGFβ production, which early in ALS disease course is found to increase neuroprotection and decrease excitotoxicity [58]. Combined with the increased BDNF and reduced tagging of synapses for microglial engulfment through decreased C1q, IL10/CCL2NAb induced C9ORF72 phenotype appears calibrated to specifically target astrocyte-driven neurotoxicity. Expectedly, treated C9ORF72 ACM again had little impact on the homeostatic status of C9ORF72 microglia (Fig. 4). Further, applying IL10/CCL2NAb directly onto microglia was not as able to recapitulate the effects of IL10/CCL2NAb ACM, validating the astrocyte-targeted paradigm and confirming that effects on microglia were not due to compounds remaining in treated ACM. We also confirmed the importance of dynamic signaling between cells on activation and function in disease contexts through the use of co-cultures (Fig. 5). Although we found no effect of ACM on MN calcium flux in either SOD1 or C9ORF72 monocultures, co-cultures of motor neurons and microglia had altered calcium flux compared to MNs cultured alone.
A change in MN calcium flux can be partially attributed to microglial activation and phagocytosis, particularly in untreated SOD1 co-cultures where MN response to depolarizing stimuli is reduced compared to SOD1 MNs alone (Fig. 5B vs 2H). However, in C9ORF72 co-cultures which have a high rate of phagocytosis with or without ACM treatment (Fig. 5D), calcium flux is significantly decreased in co-cultures treated with C9ORF72 ACM compared to untreated co-cultures (Fig. 5B). This indicates the detrimental effects of C9ORF72 astrocytes towards MNs are likely not prevented by C9ORF72 microglia, and that C9ORF72 microglia are likely more activated in the presence of MNs than in monocultures. We hypothesize that this increased microglial response to astrocyte secretions is influenced by diseased C9ORF72 MNs providing activating stimuli to C9ORF72 microglia in our co-culture paradigm. In support of this, we found that both untreated and ACM-treated C9ORF72 co-cultures had high levels of neuron death that could be reduced by astrocyte-targeted IL10/CCL2NAb treatment. The need for an activating stimulus to enable C9ORF72 microglial response to astrocytic influences may also help explain the discordance between soma size and phagocytosis phenotypes for the C9ORF72 microglia, as the phagocytosis assay provides the pHrodo Zymosan bioparticles as a pro-inflammatory activating stimulus after priming with LPS or ACM, whereas the soma size analyses are conducted after exposure to LPS or ACM only. In SOD1 co-cultures, a strong trend for increased neuron death was associated with increased phagocytosis but not with decreased neuron calcium flux. This finding could indicate an increase in excitability of surviving MN populations [59, 60] as a compensatory mechanism for those lost due to microglial engulfment or cytotoxicity.
These co-culture experiments highlight the importance of dynamic communication between glial cells and neurons in the context of ALS. iPSC-differentiated monocultures remain a valuable tool to identify mechanisms of intrinsic dysfunction in diseased cells. However, therapeutics designed to correct dysfunction should be examined in both mono- and co-cultures to assess how paracrine interactions with other cell types may influence phenotypes and pathology. For example, the high level of TUNEL+ C9ORF72 MNs when these cells are co-cultured with C9ORF72 microglia is unchanged by C9ORF72 ACM, which is unexpected based on previous reports of C9ORF72 astrocyte-induced neurotoxicity [7]. This may be due to a maximal effect of neurotoxicity induced by distress signals secreted directly from the C9ORF72 MNs such as nitric oxide [61] or complement factors [62, 63] that activate the C9ORF72 microglia. This MN-induced microglial neurotoxicity could be reduced by exposure of these co-cultures to IL10/CCL2NAb treated astrocytes, perhaps due to increased neurotrophic support by the astrocytes rather than a direct astrocyte-microglial mechanism. Overall, this in vitro study provides insight into the mechanisms of glial-driven pathology in SOD1 and C9ORF72 associated models of ALS. Importantly, astrocyte-targeted IL-10 and reduction of CCL2 was found beneficial in both models, indicating a potential for a broad therapeutic effect. In considering translation to in vivo applications, the in vitro 48wash data support a sustained treatment paradigm, such as through gene therapy-based approaches, as removal of treatment allowed the return of some neuroinflammatory and neurotoxic phenotypes. Together, these data support IL-10 and CCL2 as therapeutic targets for ALS patients across mutations.
Materials and methods
Cell culture
Two healthy control (21.8, 4.2) and three ALS patient (ALS 71 [SOD1 A4V mutation], CS29i [C9ORF72 HRE mutation], AB34.12 [sporadic ALS]) iPSC lines were utilized in these experiments [20, 64–66]. The healthy control (HC) lines were generated from individuals with no known neurodegenerative disorders. All pluripotent stem cells were maintained on Matrigel (Corning) in Essential 8 (Gibco) and passaged every 4–6 days. iPSCs and differentiated cells were confirmed mycoplasma negative. SOD1 and C9ORF72 lines were made into GFP- and RFP-expressing stable lines by infecting cells with LentiBrite GFP Control Lentiviral Biosensor (Millipore, #17-10 387, titer 7.34 × 108 IFU/ml) or LentiBrite RFP Control Lentiviral Biosensor (Millipore, #17-10 409, titer 4.59 × 108 IFU/ml) at MOI of 20 for 24 h. Virus was removed and cells were allowed to expand for 1 week. GFP+ and RFP+ cells were isolated using WOLF Cell Sorter (Nanocellect) and expanded to create purified stable lines.
Astrocyte, microglia, and motor neuron differentiations
Spinal cord patterned astrocytes were generated from iPSCs as previously described [20, 23, 24]. Differentiation reagents were purchased from ThermoFisher unless otherwise noted. Briefly, iPSCs were differentiated into neural progenitor cells (NPCs) using dual SMAD inhibition (SB 431542 and LDN 1931899) and patterned towards ventral-caudal spinal cord using retinoic acid (RA) and hedgehog smoothened agonist (SAG). NPCs were passaged every 6 days and on passage 3 were differentiated into astrocytes using ScienCell Astrocyte Medium (ScienCell Research Laboratories, Carlsbad, CA, USA) containing 1% astrocyte growth supplement, 1% penicillin-streptomycin, and 2% B27. After passage 4, astrocytes were used for ACM generation and collection.
Microglia were differentiated using the commercially available differentiation kit (StemCell Technologies #05310, #100-0019, #100-0020, Vancouver, BC, Canada) based on a previously published protocol [20, 23, 67]. As previously described [20, 23], iPSCs were differentiated into hematopoietic progenitor cells (HPCs) using the STEMdiff Hematopoietic Kit (StemCell Technologies, Vancouver, BC, Canada). Floating HPCs were collected and plated at 50 000 cells/ml in STEMdiff microglia differentiation media (StemCell Technologies, Vancouver, BC, Canada) for 24 days, followed by rapid maturation in STEMdiff microglia maturation media (StemCell Technologies, Vancouver, BC, Canada) for a minimum of 4 days.
Spinal motor neurons were differentiated based on the Maury et al. (2015) protocol [68]. Briefly, embryoid bodies were generated from iPSCs and patterned in the presence of Chir- 99 021 with dual SMAD inhibition (SB 431542 and LDN 1931899) followed by treatment with retinoic acid (RA), smoothened agonist (SAG), and DAPT. Spinal motor neuron progenitor cells were then dissociated and plated on Matrigel-coated glass coverslips or 96-well plates for terminal differentiation and maturation in growth factor supplemented medium for 21–42 days in vitro.
Quantitative real-time polymerase chain reaction (qRT-PCR)
RNA was isolated from cell pellets using the RNeasy Mini Kit (Qiagen) following manufacturer’s instructions, quantified using a Nanodrop Spectrophotometer, treated with RQ1 Rnase-free Dnase (Promega), and converted to cDNA using the Promega Reverse Transcription system (Promega). SYBR green RT-qPCR was performed in triplicate using cDNA and run on the Bio-Rad CFX384 real time thermocycler. Primer sequences shown in Table S1. Cq values for each target were normalized to GAPDH and calculated using the 1/dCt method. A minimum of three differentiations for each line were collected and run in technical triplicates. Individual data points represent the average of technical triplicates for each experiment.
Multiplex human cytokine array on conditioned media
Eve Technologies (Calgary, AB, Canada) performed the 48 multiplex cytokine array assay from duplicate differentiations using conditioned medium samples generated from iPSC-derived astrocytes. Data were analyzed for fold change difference of each cytokine expression level in ALS ACM compared to HC ACM.
Microglia soma size and pHrodo phagocytosis assays
As previously described [20, 69], microglia were treated with microglia maturation media (UTX), 1:2 ACM, or 100 ng/ml lipopolysaccharide (LPS, Sigma Aldrich, L2018) and placed in Incucyte (Sartorius) to allow for live cell imaging with 20× objective during 24-h treatment. Incucyte software was used to calculate average soma area for all cells in one well (5000 microglia per well of 24-well plate) at 24-h timepoint for soma size analyses. Points in soma size graphs represent the average of 3 technical well replicates and error bars indicate SEM. After 24-h ACM treatment and soma measurement, 1 ug/ml pHrodo Red Zymosan Bioparticles (ThermoFisher, #P35364) were added to microglia cultures. Plates were returned to Incucyte and imaged for 24 h using brightfield and red fluorescent channels at 20×. Images were analyzed using Incucyte software for total number of RFP+ microglia in each well and averaged across three technical replicates. Data points on graph represent mean and standard error of the mean for experimental replicates.
Fluo-4NW calcium flux assay
Calcium flux was measured in MNs seeded in a 96 well plate using the Fluo-4 NW Calcium Assay Kit (ThermoFisher, #F36206) per the manufacturer’s instructions. Growth medium was removed, and 100 μl of dye loading solution was added to each well and incubated for 30 min at 37C followed by 30 min at room temperature. Just before measuring fluorescence, 25 μl of the 50 mM KCl agonist or PBS (control) were spiked into each well. Fluorescence (excitation 494 nm, emission 516 nm) was immediately measured using a GloMax microplate reader. As previously described [69], fluorescence (% above PBS baseline) was calculated by subtracting the fluorescence of PBS-stimulated wells from each test well then dividing by PBS-stimulated well value and multiplying by 100. Individual data points in relative fluorescence graphs represent the average of three technical replicates.
In vitro astrocyte treatments
Treatments of SOD1 and C9ORF72 astrocytes were performed with 430 pg/ml recombinant IL-10 (PeproTech, #200-10) and 2 ng/ml CCL2 neutralizing antibodies (Bio-Techne, #AF-279-NA) in supplemented Astrocyte media for 48 h before collection of treatment media and cell pellets. For 48 wash conditions, astrocytes were rinsed with PBS and fed with supplemented Astrocyte media. Media were collected from 48wash astrocytes after 48 h. Treatment values were set based on doubling the amount of IL-10 secreted by microglia previously found capable of reducing astrocyte-driven pathology [20] and doubling the amount of CCL2 secreted by ALS astrocytes as determined by multiplex cytokine array (Fig. 1H).
Western blot
Cell pellets were lysed by sonication with Triton X-100 and protein concentration was determined using a BCA assay (ThermoFisher). Equal amounts of protein were loaded onto 10% or 12% pre-cast Tris-HCl Mini-PROTEAN gels (Bio-Rad, Hercules, CA, USA) and proteins separated by electrophoresis, then transferred to PVDF membranes (Bio-Rad). Membranes were blocked for 1 h in Odyssey TBS blocking buffer (LI-COR), followed by overnight primary antibody incubation and 30 min secondary antibody incubation. Quantification was performed with FIJI (ImageJ, National Institutes of Health, Bethesda, MD, USA) and normalized to REVERT total protein stain (LI-COR). Primary antibodies used were IL-10 (abcam, ab133575), mouse anti-TGFβ (ThermoFisher, #MAB-16949), mouse anti-NFκB p65 (Cell Signaling, #6956) and rabbit anti-phNFκB (Cell Signaling, #3033) all at 1:1000 dilutions. Secondary antibodies used were anti-rabbit IRDye 800CW (LI-COR, 1:5000 dilution, Lincoln, NE, USA) and anti-mouse IRDye 680RD (LI-COR, 1:5000 dilution, Lincoln, NE, USA).
Microglia and MN co-cultures
After a minimum of 4 days in STEMdiff maturation media, SOD1 and C9ORF72 microglia were added at a ratio of 1:4 to SOD1 or C9ORF72 MNs (between 21–36 days of MN maturation) with either 25% ALS UTX, IL10/CCL2NAb, or 48wash ACM from appropriate astrocyte line, direct IL10/CCL2NAb treatment with 25% ALS UTX ACM, or 25% fresh Astrocyte media (UTX). Treated co-cultures were placed in Incucyte for 48 h and imaged every 2 h in brightfield, RFP, and GFP channels at 20×. After 48 h, co-cultures were fixed for TUNEL assay or analyzed for calcium flux using Fluo4-NW assay. Due to the semi-adherent nature of microglia in these co-cultures, many microglia are lost during the staining process. As such, these images were not used to quantify microglia or phagocytosis. Phagocytosis was calculated using Incucyte software to identify the number of GFP+/RFP+ double positive microglia at the 48-h timepoint. Data points on co-culture phagocytosis graph represent the average of three technical replicates for double positive GFP+/RFP+ microglia normalized to total number of cells in the field of view.
TUNEL assay
Plated cells were fixed in 4% PFA for 20 min at room temperature, rinsed with PBS, and then stained using a Click-iT TUNEL Alexa Fluor 647 Assay Kit (ThermoFisher, #C10247) following manufacturer’s instructions. As previously described [69], cells were permeabilized as described for ICC, washed, and incubated with DNA labeling solution for 1 h at 37C. Cells were washed and optional DAPI nuclear counterstain was then applied for 30 min at room temperature. Coverslips were imaged with standard fluorescent microscopy. Three images were acquired from randomly selected fields for each coverslip. Images were analyzed for total fluorescence in either channel using FIJI (ImageJ) software. Relative expression for each condition (total TUNEL stain divided by total 7AAD stain) was analyzed to account for variable number of cells in each ROI as previously described [23]. Representative images were acquired on a Zeiss confocal microscope using 63× oil objective and are displayed as a maximum intensity projection of z-stack image series.
Statistical analyses
Experimental conditions within each experiment were performed in technical triplicates for a minimum of three independent experiments unless otherwise noted. Data were analyzed using GraphPad Prism software and the appropriate statistical tests including the Student’s t test, 1-way ANOVA, and 2-way ANOVA followed by Tukey’s post hoc analysis of significance. Changes were considered statistically significant when P < 0.05.
Acknowledgements
We thank Emily Welby for technical assistance. C9ORF72 iPSCs were obtained from the Cedar Sinai Stem Cell Core, and the fibroblasts used for the original generation of the HC iPSCs and the SOD1 iPSCs were obtained from Coriell Cell Repository. Figures 3A and graphical abstract created using BioRender.
Author contributions
R.L.A. performed and analyzed experiments. All authors designed experiments and interpreted data. A.D.E. supervised the study and provided funding. R.L.A. wrote the manuscript and created figures. All authors edited and approved the manuscript.
Conflict of interest statement: The authors declare no competing interests.
Funding
This project was supported by the Medical College of Wisconsin Center for Immunology (R.L.A.), and the Neuroscience Research Center (A.D.E.). We thank the Pick family and the family of Mary Alice Schleicher for their support of ALS research.
Data availability
For original data, please contact [email protected].
