Antisense Dipeptide Repeat Proteins Drive Widescale Purine Metabolism Aberration in C9orf72 Amyotrophic Lateral Sclerosis via ADA

Abstract

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterised by the death of motor neurons leading to paralysis and death, generally 3–5 years post-symptom onset. The most frequent genetic cause of ALS is a hexanucleotide repeat expansion (HRE) in the chromosome 9 open reading frame 72 (C9orf72) gene, that has three major hypothesised pathological mechanisms including the production of dipeptide repeat proteins (DPRs). Our laboratory has previously identified purine metabolism dysfunction in induced neural progenitor cell-derived astrocytes (iAstrocytes) from C9orf72 ALS (C9-ALS) cases (C9-iAstrocytes), driven by loss of the enzyme adenosine deaminase (ADA). Here, we have demonstrated that loss of ADA along with changes to ecto-5′-nucleotidase and hypoxanthine-guanine phosphoribosyl transferase led to disruption in purine metabolite levels including purine dNTP output. These changes were recapitulated in patient CSF, whilst loss of ADA was recapitulated in patient white matter. Immunofluorescence also demonstrated purinosome formation dysfunction in C9-iAstrocytes. These changes are likely driven by DPRs as ADA loss was recapitulated in in vitro and in vivo DPR models. Finally, ADA levels could be recovered by reducing DPR levels either by inhibiting serine/arginine-rich splicing factor 1 or overexpressing RuvB-like 2. Our data demonstrate that DPR production negatively affects purine function in C9-ALS suggesting a potentially pivotal role for purine metabolism dysfunction in C9-ALS pathology.

1. Introduction

Amyotrophic lateral sclerosis (ALS) is an age-onset neurodegenerative disorder driven by the death of motor neurons in the brain stem and spinal cord, leading to paralysis and subsequently death, generally from respiratory failure 3–5 years post-symptom onset [1]. Global prevalence of ALS continues to rise, and it therefore presents a growing burden on healthcare services worldwide [2,3,4]. The main genetic cause of ALS is a hexanucleotide repeat expansion (HRE) in the chromosome 9, open reading frame 72 (C9orf72) gene [5,6] (C9-HRE), which accounts for 7–8% of all ALS cases in Western populations [7,8]. The C9-HRE leads to a more aggressive disease progression, with disease duration shorter than other forms of ALS, increased prevalence of bulbar and cognitive deficit, and an earlier age-of-onset [9,10,11,12,13,14]. There are three major hypothesised mechanisms arising from the C9-HRE that contribute to ALS pathology. Firstly, haploinsufficiency caused by the C9-HRE leads to reduced protein expression and therefore a loss-of-function in the C9orf72 protein [5,15,16,17]. C9orf72 is a key regulator of autophagy, forming a complex with the SMCR8-C9orf72 complex subunit and tryptophan-aspartic acid repeat-containing protein that coordinates ras-related protein Rab-1A-dependent trafficking of Unc-51-like autophagy activating kinase 1 to the phagophore, initiating autophagy [18,19,20,21]. C9orf72 has also been linked to mitochondrial function as it translocates to the inner mitochondrial membrane which prevents degradation of translocase of inner mitochondrial membrane domain-containing protein 1, a process required for complex I assembly [22]. Transcription of the C9-HRE produces RNA foci and subsequent translation of repeat-containing transcripts produces dipeptide repeat proteins (DPRs) [5,23,24,25,26]. RNA foci pathologically have been shown to accumulate in the nucleus of cells and are thought to bind and sequester RNA binding proteins [27,28,29,30]. However, accumulation has been shown to be a relatively low-frequency event, there is conflicting data on correlations between RNA foci accumulation and age-of-onset, and there is no correlation between RNA foci production and cognitive decline [31,32,33,34]. RNA foci accumulation has even been suggested as a protective mechanism, an attempt by the cell to prevent translation of the C9-HRE into DPRs [35]. DPRs are produced when transcripts containing the C9-HRE are exported from the nucleus through the action of the RNA-binding protein serine/arginine-rich splicing factor 1 (SRSF1), which facilitates interaction between C9orf72 transcripts and nuclear export factor 1 [27,36]. Once exported to the cytoplasm, C9-HRE containing transcripts can be translated through repeat associated non-AUG translation, allowing for the translation of both sense (GGGGCC) and antisense (CCGGGG) strands and leading to five individual DPRs. Glycine-alanine (poly-GA) and glycine-arginine (poly-GR) are translated from the sense strand, proline-alanine (poly-PA) and proline-arginine (poly-PR) are translated from the antisense strand, and glycine-proline (poly-GP) can be translated from both [23,24,25,26]. Poly-GP and poly-PA are uncharged, form flexible coil structures, and do not form aggregates [37,38]. This means they have fewer protein interactions, rendering them relatively non-toxic compared to other DPRs [39,40], though poly-GP may become toxic under proteasome inhibition [41]. Poly-GA, though uncharged, forms amyloid-like aggregates that impair proteostasis, dendritic branching, and induce ER stress [42,43,44]. Mouse models expressing poly-GA display motor and cognitive impairment [45], but its toxicity remains lower than that of arginine-rich DPRs. Poly-GR and poly-PR, unlike other DPRs are highly charged and polar, with flexible coil structures [37] making them highly toxic due to their strong electrostatic interactions with low-complexity domains in RNA-binding proteins [38,46,47,48,49,50] that allows them to disrupt a wide range of cellular processes including nucleolar function, nucleocytoplasmic transport, stress granule formation, and mitochondrial activity [38,46,47,49,50,51,52,53,54,55,56,57]. Neuronal RNA foci and DPR inclusions have been observed in the brains and spinal cords of those with C9orf72 ALS (C9-ALS) [23,24,25,26] suggesting they are a key trigger of motor neuron death in ALS pathology. As ALS is a non-cell autonomous disease, extensive research has also focused on the effect of the C9-HRE on various cell types, including astrocytes. Astrocytes derived from C9orf72 patients have been shown to induce toxicity, neurite network maintenance dysregulation and electrophysiological dysfunction in control motor neurons via direct and indirect interaction [58,59,60,61,62]. This may in part be driven by DPR production as SRSF1 inhibition in astrocytes was able to ameliorate toxicity to motor neurons in co-cultures [36,63], or alternatively, extracellular vesicle dysfunction, as restoring production of miR-494-3p also ameliorated toxicity [61]. The presence of RNA foci has been demonstrated in astrocytes derived post-mortem from C9-ALS patients [33,64,65], but the presence of DPR inclusions clinically in astrocytes has not previously been observed. In vitro, however, poly-GP has been detected in iPSC-derived astrocytes [62], whilst DPRs expressed in neurons were shown to be taken up by astrocytes through both direct and indirect cell-to-cell transmission [66] and synthetic poly-GA and poly-PA were shown to be taken up by induced neural progenitor cell (iNPC)-derived astrocytes (iAstrocytes) and transmitted to neurons in co-culture [67]. Poly-GR and poly-PR peptides were also taken up by healthy human astrocytes and induced mis-splicing of the excitatory amino acid transporter 2 (EAAT2) glutamate transporter [53], whilst human primary astrocytes expressing poly-GR exhibit significantly impaired glutamate uptake [17]. This suggests DPRs are the probable trigger for astrocytic EAAT2 dysfunction and glutamate excitotoxicity that has been observed in C9-ALS patients [6,68]. Further, both depletion and inhibition of SRSF1 in C9orf72 iAstrocytes (C9-iAstrocytes) (preventing SRSF1-mediated nuclear export of C9-HRE transcripts to block DPR production) increased the survival of healthy motor neurons in co-cultures [36,63]. Therefore, gain-of-function mechanisms arising from the C9-HRE, particularly DPRs, likely play a role in the myriad aberrations that have been observed in astrocyte homeostasis in ALS, including in purine metabolism. We have previously shown that C9-iAstrocytes were less able to metabolise the purine nucleoside adenosine, which was ostensibly caused by downregulation in the enzyme adenosine deaminase (ADA) [69]. This is a key pathway in the context of ALS as it has roles in the regulation of energy metabolism, antioxidant production, DNA methylation and the DNA damage response [70,71,72,73,74], all pathways previously implicated in the pathology of ALS [75]. In this follow up study, we have expanded our understanding of how the C9-HRE affects purine metabolism, de novo purine biosynthesis (DNPB; for pathway maps see Figure 1) and the mechanistic triggers involved. Here, we demonstrate that key enzymes involved in the purine metabolism pathway are dysregulated at both the protein and RNA level, which along with loss of ADA, induce significant aberration in metabolite output, which is recapitulated in patient CSF and post-mortem motor cortex from C9-ALS cases. Moreover, using multiple gain-of-function models, including a novel Neuro-2a (N2a) gain-of-function model, we have found that DPR production leads to loss of ADA at the RNA level, which is also observed in vivo using 1000-repeat DPR fly models of C9-ALS. In addition, we have for the first time, assessed purinosome formation in C9-ALS and found that purinosome formation is compromised in C9-iAstrocytes, resulting in disrupted growth and metabolic function, potentially due to direct DPR binding to key DNPB enzymes. Finally, we have shown that removal of DPRs via SRSF1 knockdown or upregulation of RuvB-like 2 (RuvBL2) restores ADA, providing novel positive consequences for DPR amelioration in C9-ALS.

2. Results

2.1. ADA Loss Leads to Purine Metabolism Alterations in C9orf72 iAstrocytes

To follow up on our previous work highlighting loss of ADA at the RNA and protein level in C9-iAstrocytes, we performed in depth characterisation of ADA function and purine metabolism in the same model. To assess the wider impact of the C9–HRE and/or reduced ADA expression, we initially examined the expression of several other enzymes involved in purine metabolism (Figure 2). Western blot characterisation initially allowed us to reproduce our original findings, demonstrating that ADA was significantly downregulated in C9-iAstrocytes compared to controls (Figure 2B). Our analysis of the enzymes up and downstream of ADA subsequently showed that two key purine enzymes were also aberrantly expressed (Figure 2D): ecto-5′nucleotidase (CD73), which catalyses the dephosphorylation of AMP to adenosine, is significantly reduced, and hypoxanthine-guanine phosphoribosyl transferase (HGPRT), which catalyses purine salvage by converting hypoxanthine to inosine monophosphate (IMP), was significantly increased (Figure 2G). This was confirmed at the RNA level for CD73 (Figure 2H). This suggests that purine metabolism from adenosine/deoxyadenosine to urate via inosine/deoxyinosine and hypoxanthine is altered in C9-iAstrocytes. We then combined several techniques to examine the effect on intermediate output at different stages in the pathway. Initially, we generated a metabolite profile using untargeted LC/MS on control and C9-iAstrocytes (Figure 3A–G and Supplementary Figure S1A–N). This indicated alterations in levels of deoxyinosine and the ratio between deoxyadenosine and deoxyinosine (Figure 3B,C). A trend for reduced hypoxanthine and xanthosine was also observed (Figure 3E,F; adenosine, inosine and urate were not robustly detected in this untargeted analysis), the latter feeds carbon into methylxanthine and ultimately caffeine, which was significantly reduced in C9-iAstrocytes (Supplementary Figure S1I). Moreover, the analysis indicated ADP levels were increased in C9-iAstrocytes suggesting further bioenergetic imbalances (Figure 3G). Lower ADA would result in a specific imbalance in the purine dNTP pool (dATP, dGTP). dATP levels were significantly increased in C9-iAstrocytes (Figure 3D), whilst dGTP levels were unchanged (Supplementary Figure S1A). As we did not detect adenosine, inosine and urate with an untargeted approach, we utilised targeted biochemical assays to measure ADA activity, inosine output and urate output in iAstrocytes. As expected, loss of ADA at the RNA and protein level led to a significant reduction in ADA activity and concomitantly, inosine output in C9-iAstrocytes (Figure 3H,I). This ultimately led to reduced urate levels in two out of the three C9orf72 lines measured (Figure 3J). We also validated aberrant dNTP pool balance using a coupled click chemistry and DNA polymerase-based assay that uses biotin-labelled dNTP specific primers to measure individual dNTPs from iAstrocytes. Here, we found an increase in the total purine dNTP pool, driven by dATP, leading to a purine imbalance (Figure 3K–M). These data suggested that the C9-HRE results in specific purine metabolism dysfunction, which is driven by ADA, leading to increased purine salvage via HGPRT, reduced carbon flow to urate and an altered purine dNTP balance. We then utilised lentivirus expressing ADA (LV-ADA) or an empty vector (LV-EV) in iAstrocytes, confirming upregulation via Western blot (Supplementary Figure S2E–H). We demonstrate that LV-ADA treatment functionally increased ADA activity, also significantly increasing inosine output (Figure 3N–Q), demonstrating that loss of ADA likely triggers the reduced inosine output we observed. Endogenous inosine levels are usually cited in the micromolar range in the CSF [76] though levels in excess of 1 mM in the CSF in hypoxic conditions have been reported [77]. Our previous work demonstrated that low millimolar levels of inosine (1–13 mM) were able to improve bioenergetic output in iAstrocytes [69]. In line with this previous work, we therefore supplemented iAstrocytes with 7.5 mM inosine to model restoration of downstream metabolite output and again performed untargeted LC/MS. Inosine supplementation increased hypoxanthine levels (Figure 3R), suggesting inosine output was able to stimulate downstream metabolite production in the pathway. Interestingly, we also observed a reduction in ADP levels in C9-iAstrocytes after inosine supplementation (Figure 3T), suggesting that inosine output may also at the very least correlate with upstream metabolite output.

2.2. ADA Mediated Purine Metabolism Alterations Are Evident in CSF and Post-Mortem Tissue from C9orf72 Cases

To confirm whether these findings were cell model-dependent, we performed UPLC-MS/MS in the CSF and plasma of healthy control and age/sex matched C9orf72 cases. Analysis in the CSF of C9orf72 cases matched the expected impact of reduced ADA expression, namely reduced inosine, hypoxanthine and a trend for a loss of xanthine and urate (Figure 4A–E). Interestingly, no alterations were found in the plasma of C9orf72 cases, suggesting a CNS-specific effect (Figure 4F–J). To confirm these CNS findings, we measured ADA levels in the post-mortem grey and white matter motor cortex of control and C9orf72 cases. Here, we were able to recapitulate our findings in vitro, demonstrating a significant reduction in ADA in the white matter of C9orf72 cases, which was not evident in grey matter (Figure 4K,L). These data demonstrated that the ADA-mediated purine dysfunction observed in C9-iAstrocytes is also observed in both CNS patient tissue and biofluid, suggesting that the C9-HRE dysregulates purine metabolism in the CNS.

2.3. The C9-HRE Reduces Purinosome Formation in iAstrocytes

Another key aspect of purine metabolism is DNPB which is particularly important for IMP production in times of high purine demand [78,79,80]. This crucial pathway had never previously been investigated in the context of C9-ALS. As we have demonstrated significant aberration in other areas of purine metabolism, we sought to uncover whether this aberration extended to DNPB regulation. We initially found that expression of the six DNPB enzymes (adenylosuccinate lyase, ADSL; 5-Aminoimidazole-4-carboxamide ribonucleotide formyltransferase, ATIC; formylglycinamide ribonucleotide amidotransferase, FGAMS; phosphoribosylglycinamide formyltransferase/phosphoribosylglycinamide synthetase/phosphoribosylaminoimidazole synthetase, GART; phosphoribosylaminoimidazole carboxylase/phosphoribosylaminoimidazole succinocarboxamide synthetase, PAICS; and amidophosphoribosyltransferase, PPAT) was unaltered in C9-iAstrocytes (Supplementary Figure S3A–L), suggesting carbon flow through DNPB was unaffected. This was supported by our metabolite profiling analysis which indicated no changes in the DNPB intermediates 5-aminoimidazole-4-carboxylate ribonucleotide (CAIR), succinylaminoimidazole carboxamide ribonucleotide (SAICAR) and the end-product of DNPB IMP (Supplementary Figure S1D,K–L). However, as HGPRT levels were higher in C9-iAstrocytes, we could not rule out that levels of IMP were being salvaged to overcome DNPB dysfunction in C9-iAstrocytes. DNPB enzymes undergo liquid–liquid phase separation (LLPS) to form purinosomes, enhancing DNPB under times of purine metabolism dysfunction and metabolic stress [81,82]. We therefore measured purinosome formation in iAstrocytes under physiological and purine depletion conditions by measuring puncta formation of PAICS by immunofluorescence. This analysis demonstrated that C9-iAstrocytes had significantly reduced purinosome levels (Figure 5A–C). Purine depletion in culture mimics high purine demand, inhibiting cells from relying on purine metabolism/salvage and instead being driven towards DNPB. When iAstrocytes were depleted of purines during differentiation and forced to rely on DNPB, a reduction in biomass was observed suggesting reduced cell proliferation (Figure 5D). This was particularly evident in C9-iAstrocytes, suggesting decreased DNPB efficiency. Purinosome formation has been previously linked with mitochondrial function and is mTOR dependent [83]. We therefore tested whether loss of purinosome formation affects metabolic function in C9-iAstrocytes by blocking mTOR activity with rapamycin under purine rich and purine depleted conditions and measuring mitochondrial and glycolytic flux (Figure 5E–K). Aside from a trend for increased glycolysis, (Figure 5I), purine depletion alone had little metabolic effect on iAstrocytes. However, mTOR inhibition under purine depletion conditions reduced mitochondrial coupling and increased mitochondrial uncoupling in iAstrocytes, with C9orf72 lines displaying increased susceptibility to uncoupling compared to controls, leading to a significant reduction in mitochondrial respiration (Figure 5E–G). Moreover, mTOR inhibition under purine depletion conditions decreased glycolytic reserve in C9-iAstrocytes, suggesting a metabolic reliance on DNPB under times of stress (Figure 5K). These data suggest that the potential ramifications of purine metabolism dysfunction in C9-iAstrocytes may be exacerbated by an inability to overcome high purine demand through purinosome formation, which may lead to reduced cell proliferation that could also be aggravated by mitochondrial and glycolytic deficiencies.

2.4. Gain-of-Function C9orf72 Mechanisms Are Responsible for Loss of ADA In Vitro and In Vivo

We next aimed to understand how the C9-HRE induced the purine metabolism dysfunction we observed. To do this, we modelled loss-of-function of C9orf72 and the expression of DPRs. Initially, we utilised HeLa cells with CRISPR/Cas9-mediated C9orf72 knockout (as confirmed by Western blot; Supplementary Figure S4A) to demonstrate that ADA was unaffected by loss of C9orf72 function (Figure 6A,B), with CD73 and HGPRT also unaffected (Supplementary Figure S4B,C). We next generated a gain-of-function model by building on our previous work in N2a cells [36] generating stable 38-repeat sense or 39-repeat antisense HRE expressing N2a lines that exhibit DPR pathology (Supplementary Figure S4F–H). Furthermore, these lines recapitulated the metabolic defects we observed in C9-iAstrocytes in our previous work [69] including alterations in mitochondrial coupling, decreased mitochondrial spare respiratory capacity and decreased glycolytic capacity (Supplementary Figure S4M–S). We were also able to repeat the reduced ADA expression we observed in C9-iAstrocytes in this model, which reached significance in antisense expressing N2a cells (Figure 6C,D). As expected, loss of ADA at the protein level led to a concomitant reduction in both ADA activity and inosine output to similar levels for both sense and antisense expressing lines (Figure 6I,J). We next explored individual DPR contribution by expressing 36-repeat constructs of poly-GA, poly-GR and poly-PR in HeLa lines, here demonstrating that poly-PR alone was sufficient to significantly reduce ADA expression, with a similar reduction observed in poly-GA and poly-GR expressing lines that did not reach significance (Figure 6E,F). The same was observed for CD73, however no observable effect was seen with HGPRT (Supplementary Figure S4T–W), suggesting that the salvage effect may be iAstrocyte or CNS specific. These data suggested that DPR production was the C9-HRE mechanism driving purine metabolism dysfunction. Our previous work showed that in C9-iAstrocytes, ADA was reduced at the RNA level. To assess whether this was the case in our mechanistic models, we measured mRNA levels in HEK293T after transfection of 36-repeat poly-PR. Poly-PR transfection was sufficient to significantly downregulate ADA mRNA expression in HEK293T, reproducing our observations in C9-iAstrocytes (Figure 6G). To confirm this in vivo, we utilised the only in vivo model expressing DPRs at a length similar to the longest repeats reported in patients, a Drosophila > 1000-repeat model [84]. 1000-repeat poly-PR and poly-GR were expressed pan-neuronally in flies, using nSyb-Gal4, and whole brain ADA and HGPRT mRNA levels were assessed. Driving poly-GR and poly-PR production in vivo resulted in a loss of ADA in the brain (Figure 6H), whilst poly-PR led to an increase in HGPRT (Supplementary Figure S4X). In a further attempt to align this with our iAstrocyte data, we drove poly-PR and poly-GR DPR expression in glia only, using repo-Gal4, but did not see significant changes in ADA or HGPRT (Supplementary Figure S4Y,Z). This was likely due to glia only representing 10% of the total population of the brain, and as we were looking at total brain levels, any changes were masked. As our data pointed towards loss of ADA at the transcriptional level, we also sought to rule out post-translational mechanism contribution to ADA reduction as well. To do this, we utilised cycloheximide (CHX) assays, which demonstrated that the stability of ADA was unaffected by sense or antisense expression in N2a cells (Supplementary Figure S5A–D), suggesting that increased turnover was not a factor in ADA loss. Subsequently, we attempted to assess direct binding between ADA and DPRs at the protein level using immunofluorescence, but found no conclusive evidence of co-localisation between ADA and any DPRs (Figure 7A), which was confirmed by the fact that ADA was not present at a higher level than controls in the insoluble fraction of sense or antisense N2a lysates (Supplementary Figure S5E,F) demonstrating that ADA is likely not drawn into insoluble DPR inclusions. These data suggest that loss of ADA was not being driven at the protein level either via direct or indirect mechanisms and confirm, alongside our iAstrocyte and Drosophila data, that loss of ADA was a transcriptional event.

2.5. Poly-PR Interacts with DNPB Enzymes

As DNPB enzyme expression was unaffected across the board in C9-iAstrocytes, we next wanted to uncover how the C9-HRE was able to induce a reduction in purinosome formation. Direct DPR interaction impairing LLPS is well reported [38], we therefore investigated whether direct interaction between DPRs and DNPB enzymes was a possible factor in the reduction in purinosome formation we had observed in C9-iAstrocytes. To do this, we utilised Split-TurboID [85]. This process involved subcloning 36-repeat poly-PR and the six DNPB enzymes, into the C-terminal fragment (TbC) and N-terminal fragment (TbN) of Split-TurboID respectively. These constructs were co-expressed in HEK293T cells and biotin added to cultures. If the two proteins interact, then TbC and TbN fragments of TurboID would reconstitute and regain biotin ligase function. Therefore, by staining for biotin after Western blot, we could demonstrate that TurboID was reconstituted within the cell and that poly-PR co-localises with any of the DNPB enzymes also present in respective samples. Using this technique, we were able to demonstrate that poly-PR at the very least proximally locates to four of six DNPB enzymes, implying direct interaction (Figure 7B). This aligns with findings by Lee et al. [38] and suggests that the mechanism by which purinosome formation is reduced in C9-iAstrocytes is through direct interaction.

2.6. Inhibiting DPR Production Rescues ADA Levels

In further confirmation of DPR involvement in purine metabolism dysfunction, we assessed how blocking DPR production in C9-HRE expressing lines would affect ADA expression. To achieve this, we inhibited nuclear export of repeat-containing transcripts by knocking down SRSF1. Here, we co-transfected V5-tagged 45-repeat sense or 43-repeat antisense HRE constructs into HEK293T with a control shRNA or SRSF1 shRNA. As expected, we observed a significant reduction in ADA expression in antisense HRE expressing lines (Figure 8A,B), confirming our findings in C9-iAstrocytes, N2a, HeLa and Drosophila. By then quantifying V5 expression in these lines, we observed that knockdown of SRSF1 was sufficient to completely eradicate DPR production (Figure 8C,D), which has been established previously [36]. Reduction in antisense HRE DPRs led to the restoration of ADA expression to a level comparable to controls (Figure 8E,F). Interestingly, when measuring ADA at the mRNA level, SRSF1 knockdown alone was sufficient to significantly increase ADA levels in HEK293T, suggesting SRSF1 may regulate ADA levels directly (Supplementary Figure S6A). To assess whether SRSF1 was a binding partner of ADA, we performed RNA immunoprecipitation with ADA and FLAG-tagged SRSF1 in wild-type N2a. Immunoprecipitation of SRSF1 led to the known SRSF1 binder survival motor neuron protein (SMN) but not the intron-less control transcript jun proto-oncogene, AP-1 transcription factor subunit (JUN) being immunoprecipitated (Supplementary Figure S6B). Using primers designed to recognise ADA, we found significant immunoprecipitation of ADA mRNA by SRSF1 in N2as indicating that SRSF1 binds ADA transcripts (Supplementary Figure S6B). To the best of our knowledge this is the first time ADA and SRSF1 interaction has been implicated and warrants further investigation in terms of ADA transcriptional regulation in the CNS. Due to this unexpected result, we could not completely rule out a direct SRSF1 effect as the cause for ADA restoration. Therefore, to confirm the DPR specific effects, we used an alternative approach to remove DPRs by overexpressing the AAA+ family member RuvBL2, recently shown to reduce sense DPR levels and related phenotypes in in vitro and in vivo models expressing 45-repeat sense constructs [86]. RuvBL2 overexpression alone had no effect on ADA levels (Supplementary Figure S6C,D). However, RuvBL2 expression in the context of DPR production restored ADA levels in transfected HeLa (Figure 8G,H). This data confirmed that inhibition of DPRs restores ADA levels, bolstering our hypothesis that DPR production is the main driver of ADA mediated purine metabolism dysfunction in C9-ALS. To further understand the effect of DPR expression and subsequent inhibition on purine metabolite output, we expressed the 45-repeat sense construct in HEK293T and again co-transfected with either control or SRSF1 shRNA before performing untargeted LC/MS. Here, we observed that sense DPR expression significantly increased adenosine and deoxyadenosine levels (Figure 8I,J), as well as inducing a trend for an increase in dATP recapitulating our observations in iAstrocytes (Figure 8M). SRSF1 inhibition was able to ameliorate the significant increase in adenosine and deoxyadenosine levels (Figure 8I,J).

3. Discussion

In this study, we have continued our previous work which identified loss of ADA in C9-iAstrocytes and have shown that loss of ADA drives widescale purine metabolism dysfunction which is not only observed in vitro, but in vivo, ex vivo and in C9orf72 patient-derived biofluids. We initially utilised Western blots to demonstrate that C9-iAstrocytes also exhibit a downregulation in the enzyme CD73 and upregulation in HGPRT (Figure 2D,G), two key enzymes required for purine metabolism function. CD73 ordinarily dephosphorylates adenosine nucleotides to produce adenosine, though its loss in the context of purine metabolite output is unclear as we were unable to clearly detect adenosine via metabolite profiling in iAstrocytes. It could be hypothesised that reduced CD73 is a protective mechanism that prevents toxic adenosine accumulation in light of ADA downregulation. If this was the case, carbon would be shuttled via AMP to ADP, which was elevated in our C9-iAstrocytes (Figure 3G). HGPRT converts hypoxanthine to IMP or guanosine monophosphate and is the key enzyme in purine salvage and thus the principal source of purine nucleotides. Upregulation of HGPRT may therefore be a response to reduced upstream metabolite output and/or a drive for higher purine output to account for enhanced DNA repair in ALS iAstrocytes. We further demonstrated that these changes in enzyme expression have tangible functional implications that manifest in decreased ADA activity and reduced capacity to produce both inosine and urate (Figure 3H–J). These findings have important implications in ALS due to the roles that inosine and urate play in energy generation and the antioxidant response respectively. The breakdown of inosine by purine nucleoside phosphorylase (PNP) produces ribose-1-phosphate, which can contribute to glycolysis via the pentose phosphate pathway, importantly bypassing the ATP-consuming hexokinase enzyme and making the production of inosine key in times of bioenergetic stress, as observed in C9-ALS. Urate contributes to the response to oxidative stress both directly as a potent free radical scavenger [70] and indirectly as an activator of the oxidative stress response coordinator Nrf2 [87], whilst urate levels in patients have previously been shown to correlate with disease progression and outcomes in ALS [88,89,90,91,92,93]. We have also observed a significant increase in dATP, ADP and total dNTP levels, a significant decrease in deoxyinosine levels and a significant shift in deoxyadenosine:deoxyinosine and dATP:dGTP ratios (Figure 3), all of which would be expected to be concomitant with reduced ADA activity and may have pathological consequences for ALS. We have previously hypothesised that reduced ADA could lead to elevated dATP [75]. This is one of the major mechanisms leading to ADA-deficient severe combined immunodeficiency (SCID), in which elevated dATP inhibits ribonucleotide reductase in immune cells leading to impaired DNA synthesis and repair [73]. Interestingly, ADA-deficient SCID (uniquely amongst all other forms of SCID) leads to severe neurological impairment, including in motor function [94,95,96,97,98,99,100], highlighting the importance of ADA for the maintenance of neuronal homeostasis. Here, we demonstrate that, similar to ADA-deficient SCID, reduced ADA expression and activity lead to a significant elevation of dATP (Figure 3D) and a significant shift towards dATP in dGTP:dATP ratio (Figure 3M) indicating that iAstrocyte dNTP balance is impaired which could have detrimental effects on their ability to respond to DNA damage. We subsequently demonstrated that elevating ADA expression significantly increases inosine output, and supplementing cells with inosine significantly increased hypoxanthine and reduced ADP output showing that restoring ADA could be beneficial for restoring metabolite output in the pathway (Figure 3P–R,T). Finally, we confirmed that the purine metabolism changes that we observe in vitro are exhibited clinically are not limited to cell models, demonstrating significant downregulation in inosine and hypoxanthine in C9orf72 patient CSF (Figure 4B,C) and loss of ADA in white matter motor cortex (Figure 4K,L).

We also then sought to identify the underlying cause of purine metabolism dysfunction in the context of the C9-HRE, specifically whether the dysfunction we had observed could be attributed to gain-of-function pathological mechanisms. Here, we were able to recapitulate loss of ADA protein expression in three independent gain-of-function models, along with loss of CD73 in DPR models (Figure 6C–F and Figure 8A,B; Supplementary Figure S4T–U), and aberrant purine metabolite output in the presence of DPR expression (Figure 8I–M). We also demonstrated reduced RNA expression in a Drosophila DPR model (Figure 6H). Interestingly, HGPRT upregulation was only able to be replicated at the RNA level in Drosophila (Supplementary Figure S4X), suggesting that HGPRT expression may be either repeat-length dependent (as Drosophila and iAstrocyte models carry much longer repeat lengths than 36-repeat DPR expressing in vitro lines), or could be a response to accumulated stress over time (as DPR expression in vitro would represent acute expression as opposed to the chronic expression of DPRs in Drosophila and iAstrocyte lines). We then ruled out post-translational mechanisms as drivers of reduced ADA expression, as protein stability of ADA remained intact and direct interaction between ADA and DPRs was not apparent (Figure 7 and Supplementary Figure S5). This allowed us to pinpoint downregulation at the transcriptional level as the likely causative factor in loss of ADA (Figure 6G,H). Inhibiting DPR production either via SRSF1 knockdown or RuvBL2 overexpression both restored ADA expression (Figure 8A–H), and SRSF1 inhibition was able to restore normal metabolite output (Figure 8I,J), further reinforcing DPRs as the likely drivers of a reduction in ADA. Interestingly, our findings suggest that SRSF1 itself directly regulates ADA expression (Supplementary Figure S6A,B). This interaction is likely physiological as the models used were not disease specific. Furthermore, DPR production alone does not alter SRSF1 levels [36]; therefore, the DPR-mediated effects on ADA transcription that we observed are likely independent of ADA regulation by SRSF1. Consistent with this, RuvBL2 overexpression to inhibit DPRs and restore ADA expression further confirms that DPRs suppress ADA via a distinct pathway. However, the restoration of metabolite output that we observed upon SRSF1 inhibition may have resulted from either the direct upregulation of ADA, inhibition of DPR translation or a combination of those two mechanisms. Mechanistically, DPRs have been shown to interfere with several stages of transcription. Directly binding and disrupting nucleolar proteins involved in mRNA processing and transcriptional regulation [53], impairing nuclear import of transcription factors by interacting with importins [101], and perturbing heterochromatin structure thus inducing widespread transcriptional dysregulation [102]. It is also noteworthy that the only consistent finding across all models was ADA downregulation, suggesting that this effect is a general consequence of DPR expression and is not cell line or disease context (i.e., not repeat length) dependent, which is further confirmed by our previous finding that neurons derived from the same C9orf72 iNPC lines also exhibit significant ADA downregulation [69]. Nonetheless, astrocytes are the main source of cerebral purines [103] which may render astrocytes particularly sensitive to the DPR-induced purine metabolism dysfunction we observe here and have wider pathogenic implications in the CNS. Together, these findings support DPR-induced transcriptional dysregulation as the central driver of purine metabolism dysfunction and highlight loss of ADA in particular as a possibly key pathogenic mechanism in ALS, especially in light of the finding that DPRs are the major driver of ALS pathogenesis over other C9orf72 disease mechanisms [104].

We also demonstrated that purinosome formation was significantly reduced in C9-iAstrocytes (Figure 5A–C), which may lead to reduced cell proliferation under purine depletion (Figure 5D) and impact on metabolic function under purine depletion and mTOR inhibition (Figure 5E–K). Mechanistically, this is likely driven again by DPRs but not transcriptionally. Here, we highlighted a possible dual effect of DPRs on purine metabolism with poly-PR co-localising with four of the six DNPB enzymes, likely disrupting purinosome formation. This recapitulates data presented by Lee et al. [38] who similarly found that 20-repeat poly-PR and poly-GR peptides had direct interactions with four of six DNPB enzymes. Lee et al. also demonstrated that poly-GR and poly-PR peptides directly interact with the nucleolar protein, nucleophosmin 1. At high concentrations of peptide, this interaction prevented its LLPS and LLPS with binding partner surfeit locus protein 6, demonstrating a precedent for direct DPR interactions inhibiting LLPS and suggesting that the direct DPR interaction we observed here may be the trigger for the reduction in purinosome formation in C9-iAstrocytes. Interestingly, however, purinosome regulation is also linked to mTOR activity, which itself is regulated by C9orf72, thus C9orf72 haploinsufficiency induced by the C9-HRE may also contribute to the reduced purinosome formation we observe. To assess the functional implication of reduced purinosome formation, we subjected iAstrocytes to purine depleted conditions, forcing them to utilise DNPB over purine salvage. iAstrocytes cultured in these conditions exhibited significantly lower proliferation than iAstrocytes cultured in standard conditions, which was exacerbated in C9-iAstrocytes (Figure 5D). As purinosomes enhance DNPB efficiency, reduced formation would likely result in reduced metabolic flexibility, leaving iAstrocytes more susceptible to purine deficit, which is reflected in our findings here and our previous observation in C9-iAstrocytes [105].

In this article, we have built upon our previous work on purine metabolism in C9-iAstrocytes and have uncovered wide-ranging purine metabolism dysfunction that is repeated in several in vivo and in vitro models and patients. We have also demonstrated that the driver of these dysfunctions is through DPR expression both via direct and indirect interactions. Our data therefore demonstrates a robust and translatable purine metabolism dysfunction that could play a fundamental role in ALS pathology. Work is underway in our lab to further elucidate the ramifications of these dysfunctions upon disease, and potential therapeutics targeting this pathway.

4. Materials and Methods

4.1. Human Biosamples

Skin biopsies used for iNPC reprogramming were taken from three controls with an average age at time of skin biopsy of 58 (±15.7), and three C9-ALS cases with an average age at biopsy of 60.7 (±9.2) and average disease duration of 26 (±6.2) months (Supplementary Table S1). CSF was taken from seven controls with an average age of 63.6 (±8.3) and seven C9orf72 patients with average age at onset of 62.8 (±7.5) and average disease duration of 34.4 (±9.9) months (Supplementary Table S2). Plasma was taken from ten controls with an average age of 63.6 (±8.3), and ten C9orf72 patients with average age at onset of 62.8 (±7.7) and average disease duration of 35.8 (±14.3) months (Supplementary Table S2). Post-mortem white and grey matter was taken from six controls with an average age at autopsy of 68.2 (±5.9). C9orf72 white matter was taken from five patients with an average age at autopsy of 63.2 (±8.5) and average disease duration of 34.6 (±14.4) months. C9orf72 grey matter was taken from seven patients with an average age at autopsy of 64.6 (±7.3) and average disease duration of 34.6 (±12.7) months (Supplementary Table S3).

4.2. Ethical Approval

For patient-derived cell cultures, all ethical approvals are in place, subject informed consent was obtained (study numbers STH16573, STH16350, Research Ethics Committee (REC) reference: 12/YH/0330). All ethics were in place for the use of the patient plasma/CSF data under the study title: A Multicentre Biomarker Study in Neurodegeneration (REC reference: 16/LO/2136, IRAS project ID: 204405). All samples were collected at the Sheffield Teaching Hospital. All investigators were blinded to personal donor information outside clinical disease data. Post-mortem brain tissue was released under Sheffield Brain Tissue Bank (REC reference: 08/MRE00/103+5, IRAS project ID: 140226). Drosophila melanogaster are not legislated and are outside the NC3Rs restriction; however, all applicable international, national, and/or institutional guidelines for the care and use of animals were followed. Further information can be found in the Institutional Review Board Section.

4.3. Tissue Culture

All iNPCs were reprogrammed from fibroblasts and have been extensively characterised and published by the PI team et al. in Sheffield [58,69,105]. iNPCs were maintained at 37 °C, 5% CO₂ and 95% humidity in DMEM containing 1% N2 supplement (Life Technologies, Carlsbad, CA, USA), 1% B27 and 20 ng/mL fibroblast growth factor-2 (Preprotech, Cranbury, NJ, USA). iAstrocytes were differentiated through the addition of DMEM containing 25 mM glucose and supplemented with 10% heat-inactivated FBS (Sigma-Aldrich, St. Louis, MO, USA) and 0.2% N2 supplement. Purine depletion was achieved by dialysing FBS at 4 °C in 0.9% NaCl using 25 kDa MWCO Pur-A-Lyser dialysis tubes (Merck, Darmstadt, Germany) (30 min/0.1 mL). iAstrocytes were cultured in purine depleted FBS from differentiation day 2 to 7 prior to assaying. Cultures were supplemented with 7.5 mM inosine 24 h prior to harvest. HEK293T and HeLa cells were cultured in DMEM, supplemented with 10% FBS and 5 mM Pen-Strep (Lonza, Basel, Switzerland). N2a cells were cultured as above with the addition of 5 mM sodium pyruvate (Sigma-Aldrich).

4.4. Plasmids

C9orf72 synthetic 45-repeat sense and 43 repeat antisense HREs in a pcDNA3.1 plasmid were generated as described previously [36]. For lentiviral transduction of N2a cells, repeat sequences were subcloned into a SIN-PGK-cPPT-GDNF-WHV lentiviral backbone as described previously [63], which caused a contraction of the sense and antisense repeat sequences from 45-repeat and 43-repeat to 38-repeat and 39-repeat respectively. Additionally, 36-repeat poly-GA, poly-GR and poly-PR in pcDNA3.1, originally described in Mizielenska et al. [33], were a gift from Adrian Isaacs and were subsequently subcloned into a pCI-Neo-V5-N as described in Hautbergue et al. [36] and Bauer et al. [106]. SRSF1 shRNA was generated as described previously [36]. FLAG-tagged SRSF1 plasmids were generated as described previously [107]. HA-tagged RuvBL2 plasmid was generated as described previously [86]. Transfections were performed using either jetPRIME transfection kit (Polyplus, Strasbourg, France) per the manufacturer’s instructions, Xfect transfection reagent (Takara Bio, Kusatsu, Japan) per the manufacturer’s instructions, polyethyleneimine (PEI) (2.5 µL/µg plasmid) with Opti-MEM (50 µL/0.7 µg plasmid) or 0.5 M calcium chloride with HBS buffer (280 mM NaCl, 100mM HEPES, 1.5mM Na₂HPO₄, pH 7.1).

4.4.1. TbN-DNPB Enzyme and TbC-PR36 Plasmids

The TbN and TbC were amplified from V5-TurboID-NES_pCDNA3 (Addgene # 107169) [108]. PPAT-TbN, GART-TbN, TbN-FGAMS, TbN-PAICS, TbN-ADSL, TbN-ATIC, PPAT-P2A-TbN, GART-P2A-TbN, TbN-P2A-FGAMS, TbN-P2A-PAICS, TbN-P2A-ADSL, and TbN-P2A-ATIC were generated from PPAT-SYFP21-215, GART-SYFP2 1-215, SYFP21-215-FGAMS, SYFP21-215-PAICS, SYFP21-215-ADSL, and SYFP21-215-ATIC, PPAT-P2A-SYFP21-215, GART-P2A-SYFP21-215, SYFP21-215-P2A-FGAMS, SYFP21-215-P2A-PAICS, SYFP21-215-P2A-ADSL, and SYFP21-215-P2A-ATIC [109], respectively, by replacing SYFP21-215 with TbN. TbC-PR36 and TbC-P2A-PR36 were derived from SYFP21-215-PAICS and SYFP21-215-P2A-PAICS, respectively, by substitution of SYFP21-215 with TbC and PAICS with 36-repeat poly-PR.

4.4.2. Cloning of ADA into a Lentiviral Vector

LV-ADA was generated by amplifying ADA coding sequence (pOTB7/ADA inAGE 3629376) by PCR using BclI and XhoI restriction sites. PCR products were resolved by agarose gel electrophoresis, purified and digested overnight with BclI and XhoI. SIN-PGK-cPPT-GDNF-WHV vector was digested with BamHI and XhoI treated with calf intestinal phosphatase to prevent self-ligation. Insert and vector DNA were purified by phenol-chloroform extraction and ethanol precipitation. Ligation reactions were performed using T4 DNA ligase and incubated overnight at 16 °C and subsequently transformed into DH5α cells that were then plated on ampicillin-containing LB agar. Colonies were screened by restriction digest, and a positive clone was expanded using QIAGEN Plasmid Plus Midi Kit (QIAGEN, Düsseldorf, Germany). Correct insertion and orientation were confirmed by restriction digest and Sanger sequencing (Source BioScience, Nottingham, UK) using LV reverse and pGK forward primers. LV-EV consisted of SIN-PGK-cPPT-GDNF-WHV vector without ADA coding sequence insertion.

4.4.3. CRISPR/Cas9 C9orf72 Knockout Cell Lines

C9orf72-targeted pSpCas9n D10A nickase plasmids were generated according to the Zhang lab protocol [110]. Briefly, C9orf72 targeted DNA oligo pairs were annealed and cloned into pSpCas9n(BB)-2A-Puro via BbsI sites to produce pSpCas9n(BB)-2A-Puro nickase plasmids targeting the C9orf72 sense strand, and C9orf72 antisense strand. C9orf72 targeting oligos were as follows: Sense: 5′-taacacatataatccggaa-3′; Antisense: 5′-acacactctatgaagtggg-3′. Each oligo was immediately followed by an NGG PAM site on the target sequence, allowing for the production of single strand nicks on opposing DNA strands. All plasmids were confirmed by sequencing.

To generate HeLa C9orf72 knockout clones, cells were co-transfected with the sense and antisense C9orf72-targeting pSpCas9n(BB)-2A-Puro nickase plasmids in a 1:1 ratio; 24 h post-transfection cells were trypsinised and re-plated at 50% confluency. Upon re-plating, cells were cultured in media containing 3 µg/mL puromycin to select transfected cells; 72 h post selection cells were re-plated in serial and limiting dilutions across 96-well plates to select for single clones.

4.5. Lentivirus

Here, 13 µg SIN-PGK-cPPT-GDNF-WHV, 13 µg pCMVDR8.92, 3 µg pRSV-Rev and 3.75 µg M2G plasmids per 10 cm dish, were transfected into HEK293T/17 cells and incubated for three days. Media was then filtered and centrifuged at 19,000× g for 1 h 30 m. Media was removed and pelleted virus incubated in 1% BSA on ice for 1 h prior to resuspension. Concentration was calculated by initially transducing HeLa cells with 1 × 10⁻², 10⁻³ or 10⁻⁴ dilutions of virus alongside a known virus. RT-qPCR with WPRE primers was used to then calculate viral titre based on the 2^−ΔΔCT method [111]. Transductions were performed using the following formula: Vol. of virus (µL) = (multiplicity of infection × cell number)/viral titre.

4.6. Sense and Antisense N2a Line Generation

N2a cells were transduced with either 38-repeat sense or 39-repeat antisense C9-HRE expressing lentivirus. Following transduction, cells were serially diluted across 96-well tissue culture plates to obtain single-cell-derived colonies. Individual colonies were expanded to generate clonal N2a lines stably expressing both HRE lentiviral constructs.

4.7. Drosophila Husbandry

Drosophila were raised on cornmeal–yeast medium (8% w/v cornmeal, 1.8% w/v yeast, 1% w/v soya flour, 8% w/v malt extract, 4% w/v molasses 0.8% w/v agar, 0.25% w/v Methyl paraben, 0.4% v/v propionic acid) at 25 °C on a 12 h light:dark cycle. Neuronal Synaptobrevin (nSyb)-Gal4 (RRID: BDSC_51635), Upstream activator sequence (UAS)-mCD8-GFP (RRID: BDSC_32184) and Repo-Gal4 (RRID: BDSC_7415) stocks were obtained from the Bloomington Drosophila Stock Center (BDSC). All 1000-repeat DPR stocks were described by us previously [84,86,112,113,114]. All experiments were performed using flies from at least 3 independent crosses.

4.8. Western Blotting

Pelleted cells were lysed in an appropriate volume of lysis buffer (Radio-Immunoprecipitation Assay buffer with 10% protease inhibitor cocktail) and incubated on ice for 30 min prior to centrifugation at 13,000 rpm for 30 min at 4 °C to remove insoluble material. Concentration was calculated by diluting samples 1/500 in Bradford reagent, measuring absorbance on a spectrophotometer. For Western blots on insoluble material, the insoluble pellet was collected and resuspended in lysis buffer prior to 5–10 s sonication and subsequently following remaining steps. Samples were then diluted in Laemmli buffer prior to being denatured by heating at 95 °C for 5 min. Then, 10–30 µg of protein was loaded into 5% stacking and 8–15% resolving SDS polyacrylamide gel and electrophoresis performed using Mini-PROTEAN Tetra Handcast power packs (Bio-Rad, Hercules, CA, USA). Protein was then transferred to nitrocellulose or polyvinylidene membranes in transfer buffer at 250 mA for 1 h. Membranes were incubated in blocking solution (5% BSA in TBST) for 1 h at RT prior to incubation with primary antibody overnight at 4 °C, washed in TBST then incubated for 1 h in secondary antibody and imaged with EZ-ECL HRP chemiluminescence kit on Odyssey Fc imaging system. Protein signal was quantified with Image Studio Lite 5.2 (both LI-COR). Details of antibodies used can be found in Supplementary Materials and methods (Supplementary Tables S4 and S5).

4.9. ADA Activity/Inosine Output Assay

ADA activity was assessed with an ADA activity assay kit (Abcam, Cambridge, UK) as per the manufacturer’s instructions. Briefly, iAstrocyte and N2a cell pellets were lysed in assay buffer and flash frozen prior to storage at −80 °C. Protein concentrations were determined using the Bradford assay as described previously. A standard curve was generated using an inosine standard (1 in 2 serial dilutions 0–320 nmol/well, prepared in assay buffer) was loaded into a 96-well white-walled plate. Then, 1–4 µg protein was loaded into sample wells along with a reaction mix containing 40 µL assay buffer, 2 µL ADA converter, 2 µL developer, 1 µL ADA probe and 5 µL ADA substrate. Background mix that was comprised of the reaction mix with ADA substrate replaced with assay buffer was added to remaining sample wells and standard wells. A PHERAstar (BMG Labtech, Ortenberg, Germany) plate reader was used to measure fluorescence (Ex/Em = 535 nm/587 nm) at 2-min intervals for 90 min. Inosine output was calculated by normalising RFU at an individual time point to protein content in each well. ADA activity was calculated by taking the RFU from two timepoints over the course of the kinetic read and entering those values into the following formulae:

ΔT = T₂ − T₁

ΔRFU = (RFU₂ − RFU_2BG) − (RFU₁ − RFU_1BG)

ADA activity = (ΔRFU/ΔT × µg protein) × Dilution factor = pmol/min/µg

where:

• T₁ = Timepoint 1
• T₂ = Timepoint 2
• RFU₁ = RFU at timepoint 1
• RFU₂ = RFU at timepoint 2
• RFU_1BG = Background RFU at timepoint 1
• RFU_2BG = Background RFU at timepoint 2

4.10. Urate Assay

The urate assay was carried out as described previously [69]. Briefly, six days post-differentiation media was replaced with DMEM containing 5 mM glucose, 10% FBS, 0.2% N-2 supplement and 0.3 mM glutamine. Cells were incubated for 24 h then harvested and stored at −80 °C. On day seven of assay, samples were lifted and lysed in 125 µL cold uric acid assay buffer for 20 min on ice and centrifuged at 13,000× g for 10 min at 4 °C. Protein concentration was quantified using Bradford assay as described above. Urate standards (0–40 nmol/well) in 50 µL assay buffer were added to the assay plate in duplicate along with 50 µL samples. A master mix containing 46 µL assay buffer, 2 µL urate probe and 2 µL enzyme mix were added to each well. Plates were incubated at 37 °C for 30 min prior to absorbance measurement at 570 nm on PHERAstar plate reader (BMG LABTECH, Ortenberg, Germany).

4.11. RT-qPCR

4.11.1. iAstrocytes

RNA was extracted by adding an appropriate volume of TRIzol either directly to cells in plate or to pelleted cells. Cells were subsequently centrifuged at 4 °C and 12,000 rpm for 5 min. Then, 1:5 volumes chloroform was added to a tube and tubes were shaken for 20 s. Tubes were then incubated at RT for 10 min before being centrifuged for 10 min at 13,300 rpm and 4 °C. Upper phase was then collected and RNA precipitated by adding equal volumes isopropanol and 1 µL glycogen and incubated on ice for 10 min. After this, RNA was pelleted by centrifuging for a further 20 min at 13,300 rpm and 4 °C and supernatant discarded. Pellet was washed with 200 µL 70% ethanol and centrifuged again for 10 min at 13,300 rpm and 4 °C before discarding ethanol and airdrying for 15 min. Pellet was DNase treated by resuspending in a master mix containing 22 µL ultra-pure H₂O, 2.5 µL DNase buffer and 0.5 µL DNase enzyme (Roche, Basel, Switzerland), incubated for 30 min at 37 °C, followed by 10 min at 75 °C. RNA concentration was quantified using Nanodrop 1000 spectrophotometer (Thermo Fisher, Waltham, MA, USA). For cDNA synthesis, a priming mix containing 2 µg RNA with 1 µL 40 µM random hexamers, 1 µL 10 mM dNTPs and made up to a final volume of 13.5 µL with ultra-pure H₂O was made. This mixture was heated to 50 °C for five minutes before being cooled in ice water. This solution was then added to a reaction mix comprising 4 µL 5× first strand buffer, 2 µL 0.1 M DTT and either 1 µL M-MLV reverse transcriptase (Thermo Fisher, Waltham, MA, USA) or 1 µL ultra-pure H₂O (as control). This mix was then incubated at 25 °C for 10 min, 42 °C for 60 min and 85 °C for 5 min on a PCR cycler (G-storm, St. Neots, UK). RT-qPCR was then performed by adding 1 µL cDNA to 96-well white-walled conical bottom PCR plates (Bio-Rad) followed by a mix of 1 µL 5 µM mix of relevant forward and reverse primers (for details see Supplementary Table S6), 3 µL ultra-pure H₂O and 5 µL Brilliant III Ultra-Fast SYBR Green (Agilent, Santa Clara, CA, USA) before cycling on CFX96™ or CFX384™ RealTime System C1000 Touch™ Thermal Cycler (Bio-Rad) and analysis using CFX Maestro v2.2 (Bio-Rad).

4.11.2. Drosophila

For Drosophila, RT-qPCR RNA was isolated from 40–50 fly heads per sample using 200 µL of Trizol (Invitrogen, Waltham, MA, USA), followed by a 5-min incubation and addition of 1:5 volumes of chloroform. Tubes were incubated for a further 5 min and then centrifuged at 14,000 rpm for 15 min at 4 °C. The upper phase was collected, and 1:2 volumes of Trizol and 1:5 volumes of chloroform were added, followed by a 5-min incubation. The tubes were then centrifuged at 14,000 rpm for 5 min at 4 °C. The upper phase was transferred, and 3 volumes of 100% ethanol were added for RNA precipitation. After the RNA was pelleted, the supernatant was discarded, and the pellet was washed with 70% ethanol. DNase treatment was carried out as described above. cDNA was synthesised using Random Hexamers (Thermo Fisher, N8080127), dNTPs (APEXxBIO, Hsinchu, Taiwan), 5× M MLV reaction buffer, and M MLV reverse transcriptase (Promega, Madison, WI, USA) as described above. Gene expression quantification by RT-qPCR and analysis were performed as described above.

4.12. Metabolic Flux Assay

On day 5 post-differentiation, iAstrocytes were plated at 15,000 cells/well in a XF cell culture plate in 5 mM glucose, 1 mM glutamine, 0.2% N2, 10% FBS ± purines and incubated for 48 h. On day 7, iAstrocyte media was changed to XF Medium (Agilent) supplemented with 1 mM L-glutamine ± rapamycin at 1 µM and incubated at 37 °C, 0% CO₂ for 60 min. Metabolic flux was measured in an Agilent XF96e bioanalyser under basal conditions and in the presence of glucose (5 mM) and mitochondrial inhibitors oligomycin, FCCP and rotenone/antimycin A 1.3 µM, 1.75 µM, 1.3 µM respectively. Following the metabolic flux assay, all media was removed from the cells, which were then frozen at −80 °C, prior to Cyquant analysis (Invitrogen) as per the manufacturer’s instructions. Fluorescence measurements were taken 485/538 nm Ex/Em on a Fluorostar plate reader (BMG Labtech). N2as were plated at 30,000 cells/well in a XF cell culture plate in 25 mM glucose, 4 mM glutamine, 1 mM pyruvate and 10% FBS for 24 h. On day 7, iAstrocyte media was changed to XF Medium (Agilent) supplemented with 4 mM L-glutamine, 1 mM pyruvate (or 0.4 glutamine, 0.1 mM pyruvate for the starvation assay) and incubated at 37 °C, 0% CO₂ for 60 min. Metabolic flux was measured in an XF96e bioanalyser (Agilent) under basal conditions, 25 mM glucose, 4 mM glutamine, 1 mM pyruvate, or starvation conditions 2.5 mM glucose, 0.4 mM glutamine, 0.1 mM pyruvate and the mitochondrial inhibitors oligomycin, FCCP and rotenone/antimycin A at 1.0 µM, 0.5 µM, 0.75 µM respectively. Following the metabolic flux assay, all media was removed from the cells, which were then frozen at −80 °C, prior to Cyquant analysis as described above.

4.13. Immunofluorescence

iAstrocytes were initially plated in 96-well black-walled plates (Greiner, Kremsmünster, Austria) before fixing in 3.7% paraformaldehyde at RT for 20 min. Cells were then permeabilised with 0.2% Triton for 3 min and subsequently blocked with 1% BSA or 4% goat serum for 30 min. Cells were then incubated in primary antibody for either 1 h at RT or overnight at 4 °C before being incubated in secondary antibody for 1 h at RT. Cells were then treated with Hoechst for 10 min. Imaging was performed with Opera Phenix High-Content Screening System (PerkinElmer, Waltham, MA, USA) and images were analysed and processed using Harmony High-Content Imaging and Analysis software (PerkinElmer, v5.2). HeLa cells were plated in 24-well plates on cover slips which were fixed, permeabilised, blocked and blotted as above. Cover slips were then mounted on slides with Immunomount before imaging on BX53 fluorescent microscope (Olympus, Tokyo, Japan) and images processed in Fiji (ImageJ, v1.52g). Details of antibodies used can be found in Supplementary Materials and methods (Supplementary Tables S4 and S5).

4.14. iAstrocyte Untargeted Metabolite Profiling

Day 7 iAstrocytes were harvested in ice-cold methanol before being dehydrated under vacuum in Genevac Vacuum Concentrator (Thermo Fisher). Samples were stored at −80 °C until analysis. Prior to the analysis, the samples were re-suspended in 50 µL of 50:50 methanol:water (LC/MS grade), vortexed for 1 min, placed in an ultrasonic bath for 10 min and then left on ice for 30 min. Samples were centrifuged for 5 min at 1200 rpm, and 80 µL of the resulting solution was loaded into a HPLC vial and run automatically on a Synapt G2Si high resolution TOF-MS coupled to a Acquity UPLC (Waters, Wilmslow, UK). The column was a Waters BEH C18 column (1.7 µm particle size, 2.1 × 50 mm dimensions). The mobile phases consisted of 0.1% formic acid and acetonitrile + 0.1% formic acid. The flow rate was 0.3 mL/min, and the injection volume was 2 µL. The gradient programme was as follows: 99–65% A (0–3 min), 65–1% A (3–6 min), 99% A (7 min) with a total run time of 7 min per sample. Samples were run using electrospray ionisation in both positive and negative mode with a capillary voltage of 4 V, sampling cone voltage of 25 V, desolvation gas 730 L/h, and a source temperature of 100 °C in positive mode and with a capillary voltage of 4.5 V, sampling cone voltage of 33 V, desolvation gas 610 L/h, and a source temperature of 100 °C in negative mode. Data was processed for further analysis following the untargeted metabolomics procedure outlined by Parker et al. [115] using XC-MS Online [116] to provide a table of m/z, retention time and intensity. All data were normalised to the weight of the initial cell pellet. Metabolomics data analysis was performed using the web-based tool Metaboanalyst 6.0 [117]. Metaboanalyst provides a streamlined workflow for analysing untargeted metabolomics data, supporting both statistical and functional assessments as well as pathway analysis. The platform provides a comprehensive overview of metabolites that are up- or down-regulated within a system. For pathway analysis, the platform uses the mummichog algorithm developed [118] which enables direct pathway prediction from untargeted metabolomics peak list datasets. The algorithm prioritises metabolite prediction in untargeted datasets, thus identifying significant pathways in one step without the prior annotation of the metabolites. The prediction can potentially match to several metabolite candidates, and so adduct ions are limited in the software to the most common forms. This allows for the overall prediction of the functional purpose of the different systems. All data is freely available on the University of Sheffield data repository.

4.15. dNTP Measurement by Coupled Click Chemistry and DNA Polymerase-Based Assay

Day 7 iAstrocytes were harvested in ice-cold methanol before being dehydrated under vacuum in Genevac Vacuum Concentrator. Samples were stored at −80 °C until analysis. Purine dNTPs (dATP, and dGTP) were quantified using the fluorescence-based method described in [119]. First, a general biotinylated oligonucleotide ([Btn]CCGCCTCCACCGCC or Biotin Primer 1 and [Btn]CCGCCTCCACCGCCG or Biotin Primer was annealed with a specific oligonucleotide AAAAAGAAAAAGAAAAAGAAAAAGCGGCGGTGGAGGCGG or Poly(dG-dA5) and GGGAGGGAGGGAGGGAGGCGGTGGAGGCGG or Poly(dA-dG3)) at 10 μM in annealing buffer (300 mM NaCl, 50 mM Tris-HCl, 10 mM MgCl₂, 100 μg/mL BSA, pH 7.9). After heating at 95 °C for 5 min, the oligonucleotide mix was cooled to 25 °C at a slow ramp rate (1 °C/min). Then, cell extracts were rehydrated in 40 μL nuclease-free water and incubated for 40 min at 60 °C in a final 100 μL volume of DNA Polymerase Reaction Buffer Mixture 1 (0.04 U/μL Taq DNA Polymerase, Taq DNA Polymerase Buffer 1×, 0.6 μM annealed-oligonucleotides, 40 μM EdUTP) for dATP, or dGTP determination. To pull-down biotin-labelled DNA, samples of 1 × 10⁶ cells (corresponding to 50 μL) were transferred to a black microplate with clear bottom containing 50 μL of Streptavidin beads diluted 1:10 in PBS. After incubation for 30 min in agitation, DNA was denatured with NaOH 0.1 M for 5 min and beads were washed twice with 200 μL buffer (10 mM Tris-HCl pH 7.5, 1 mM EDTA and 0.1% SDS) and four times with 200 μL ddH₂O. Conjugation of the TAMRA fluorophore to the beads was performed through copper(I)-catalysed alkyne-azide cycloaddition (CuAAC) or click reaction. Thus, 100 μL of the click Buffer Mix (10 μM TAMRA-Azide, 0.5 mM CuSO₄, 2.5 mM THPTA, 5 mM sodium dNTP determination ascorbate in TBS) was added and incubated for 1 h in agitation. After the washing steps already described, samples were diluted in 50 μL TBS 1× and fluorescence was measured with 529/575 nm Ex/Em wavelengths. Each sample was assayed with technical duplicates and standard curves were also included at each experiment in duplicate. Fluorescence read was subjected to 200 flashes and bottom reading using Tecan Spark 10M (Tecan Group Ltd., Männedorf, Switzerland).

4.16. Plasma and CSF Metabolomics

CSF and plasma samples were collected as described previously [120] as part of the AMBROSIA (A Multicentre Biomarker Resource Strategy In ALS) cohort. Metabolomics using ultra-high performance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) was performed by Metabolon, Inc. Samples and derived aliquots were stored at −80 °C until analysis. Sample preparation was conducted using the automated MicroLab STAR^® system (Hamilton Company, Reno, NV, USA). Recovery standards were added prior to extraction for quality control. Proteins were precipitated using methanol under vigorous agitation (Glen Mills GenoGrinder 2000, Clifton, NJ, USA), followed by centrifugation. Extracts were divided into aliquots for four UPLC-MS/MS analyses: two reverse-phase methods using positive electrospray ionisation, one reverse-phase method with negative electrospray ionisation, and one hydrophilic interaction chromatography method with negative electrospray ionisation. Organic solvents were evaporated (TurboVap^®, Zymark, Hopkinton, MA, USA), and extracts were stored under nitrogen overnight prior to analysis. Control samples analysed alongside experimental samples included pooled matrix samples (technical replicates), extracted water blanks (process controls), and internal QC standards (recovery and internal standards). QC metrics included median relative standard deviation calculations for both instrument (spiked standards) and overall process variability (endogenous metabolites in pooled samples). Sample injection was randomised, with QC samples interspersed at regular intervals.

Metabolomics was performed using a Waters ACQUITY UPLC system coupled to a Thermo Scientific Q-Exactive Orbitrap mass spectrometer with a heated electrospray ionisation source, operated at 35,000 resolution. Reconstituted extracts were analysed by four complementary methods [121]. Metabolon, Inc’s informatics pipeline comprised laboratory information management system, proprietary peak-identification software, and visualisation tools. Raw spectral data were processed and deconvoluted using Metabolon, Inc-developed applications. Compounds were identified by matching retention indices, accurate mass (±10 ppm), and MS/MS fragmentation spectra against a proprietary library. Compounds were quantified by calculating the area under the curve. Run-day normalisation (“block correction”) adjusted compound medians to 1.0 to correct for instrument variability.

4.17. Post-Mortem Tissue Collection and Processing

Post-mortem white and grey matter was donated to the Sheffield Brain Tissue Bank after consent from the donor or next of kin. Brain and spinal cord samples were snap-frozen at autopsy using liquid nitrogen. Samples were processed as described previously [122] prior to Western blot.

4.18. Split-TurboID

HEK293T cells were cultured in purine-depleted media overnight. One well of cells was co-transfected with 2 µg of TbC-PR36 and 2 µg of PPAT-TbN, GART-TbN, TbN-FGAMS, TbN-PAICS, TbN-ADSL, or TbN-ATIC. In controls, one well of cells was co-transfected with 2 µg of TbC-P2A-PR36 and 2 µg of PPAT-P2A-TbN, GART-P2A-TbN, TbN-P2A-FGAMS, TbN-P2A-PAICS, TbN-P2A-ADSL, or TbN-P2A-ATIC. Biotin was added 19 h post-transfection at a concentration of 50 µM to initiate in vivo labelling reaction. After 4 h, incubation cells are washed 3 times with PBS and pelletised. The cell pellet is lysed in lysis buffer (50 mM Tris, pH 7.8, 135 mM KCl, 15 mM NaCl, 5 mM MgCl₂, 5% glycerol, 1% Triton X-100, supplemented with 1 mM DTT, 2 mM ATP, 1 mM L-glutamine, protease inhibitor cocktail without EDTA, and phosphatase inhibitor) and cell debris was pelleted by centrifugation at 10,000× g for 10 min at 4 °C. Protein concentration was determined by Bradford assay. Then, 30 µg of each cell lysate are subjected to Western blot analysis.

4.19. Statistical Methodology

Statistical analysis was carried out in Graphpad Prism Software (v10.0.3). All data underwent normality distribution analysis by D’Agostino−Pearson omnibus normality test, Anderson−Darling test, Shapiro−Wilk normality test and Kolmogorov−Smirnov normality test with Dallal−Wilkinson−Lilliefor p value and an F test to assess variance (where applicable) prior to parametric or non-parametric analysis. The statistical test used is outlined in figure legends. All error bars indicate standard deviation. In this study, our iAstrocyte lines were derived from three controls and three C9orf72 patients. The authors acknowledge that often the convention is to take an average of assays from each line and treat these as individual biological repeats for the purposes of statistical analyses. For our work, we have treated each assay as a biological repeat. This is to transparently capture the variance between replicates, even from individual lines, as has been the convention for ours and others’ previous work.