Ubiquitous expression of MATR3 in Drosophila leads to neuromuscular junction defects followed by motor dysfunction
To investigate the role of MATR3 in an in vivo model system, we generated transgenic Drosophila lines with site-specific integration of either wild type (WT) human MATR3 or MATR3 with ALS-associated mutations Ser85Cys (S85C) and Phe115Cys (F115C) with an N-terminal FLAG tag. We confirmed similar transgene expression levels between groups before further analysis (Fig. 1b, c). We then utilized the well-established UAS/GAL4 ectopic expression system to systematically assess the effect of MATR3 expression in target tissues. Constitutive ubiquitous expression of WT or mutant MATR3 in Drosophila caused developmental lethality, leading to complete loss of egg-to-adult viability (Fig. 1d). In contrast, neither the non-transgenic control (standard w1118 strain crossed with Tub-Gal4 driver) nor ubiquitous expression of an irrelevant transgene, EGFP, caused any lethality, confirming that the phenotype was exclusively due to MATR3 expression (Fig. 1d).
To assess our model for specific motor neuron defects, we utilized the GeneSwitch (GS) expression system induced by a mild dose of RU486 for controlled conditional expression of MATR3 through development to larval stages (Fig. 1e). We labeled the third-instar larval neuromuscular junction (NMJ), a well-established model for investigating neurodegeneration, for the presynaptic marker HRP to assess synaptic bouton morphology. Expression of MATR3 WT and F115C, but not S85C, reduced the number of synaptic boutons (normalized to muscle area) in the NMJ compared to controls (Fig. 1f, g). To determine if ubiquitous expression of MATR3 results in motor dysfunction among adult animals, we used the inducible expression system to bypass expression in developmental stages and instead conditionally express MATR3 in adults, induced by RU486 (Fig. 1e). Consistent with neuromuscular junction defects in larvae, conditional expression of MATR3 in adults resulted in statistically significant reduction in their locomotion ability compared to controls (Fig. 1i). Additionally, these flies had a significantly shortened lifespan compared to controls (Fig. 1h). We observed a strong nuclear WT and mutant MATR3 signal with a granular expression pattern (Additional file 7: Fig S1) in the Drosophila ventral nerve chord (VNC), suggesting that pathogenic mutations do not perturb the subcellular distribution of MATR3 protein in our model.
We next performed a pulse-chase experiment to analyze MATR3 degradation kinetics in vivo. We briefly induced MATR3 expression using RU486, then chased protein turnover using immunoblotting. Since MATR3 associates with both detergent (NP40)-soluble and insoluble fractions, we analyzed protein stability in both fractions. NP40-soluble wild type and F115C MATR3 exhibited similar degradation kinetics (Fig. 1j), with the F115C mutant protein exhibiting a slightly longer half-life (t1/2 = 44 h) compared to WT (t1/2 = 42 h). The NP40-soluble S85C mutant, on the other hand, accumulated for the first 12 h then gradually degraded, resulting in an overall increase in half-life of the protein (t1/2 = 51 h) (Fig. 1j). However, NP40-insoluble MATR3 exhibited different kinetics: insoluble F115C and S85C mutants accumulated in the first 12 h and thus took longer to degrade compared to the WT protein, resulting in overall increase in half-life of insoluble species of F115C (t1/2 = 56 h) and S85C (t1/2 = 50 h) compared to that of WT (t1/2 = 36 h) (Fig. 1k). This suggests that pathogenic mutations in MATR3 result in decreased protein turn-over, leading to more stable insoluble forms of the protein that might disrupt its functions.
Muscles exhibit higher vulnerability to the ALS/myopathy-linked MATR3 S85C mutation
To dissect the tissue-specific roles of MATR3, we expressed MATR3 in disease-relevant tissues, including motor neurons and muscles. While not as potent as in the ubiquitous expression paradigm, expression of MATR3 in either motor neurons (Fig. 2a, Additional file 8: Fig S2A) or muscles (Fig. 2b, Additional file 8: Fig S2B) significantly shortened the lifespan compared to controls, indicating that tissue-specific MATR3 expression results in neuronal and muscular degeneration, respectively. Furthermore, expression of MATR3 in Drosophila eyes did not cause external or internal eye degeneration, even when aged for 30 days (Additional file 9: Fig S3). These finding suggest that MATR3-mediated toxicity is not systemic, and that motor neurons and muscles might be specifically susceptible to MATR3 expression.
To examine whether reduced longevity is supported by concurrent motor dysfunction, we tested the locomotion abilities of flies expressing MATR3 in muscles. At day 2, MATR3-expressing flies do not exhibit any obvious motor deficit (Fig. 2c, d). However, at day 15 of the adult lifespan, while MATR3 flies retained their climbing ability, the speed at which they climbed was significantly impaired compared to controls (Fig. 2d). Interestingly, the motor dysfunction in S85C mutant flies was exacerbated in an age-dependent manner; by day 30, approximately 25% of animals entirely lost their climbing ability (Fig. 2c). Among the flies that did climb, their climbing velocity remained significantly lower compared to that of controls (Fig. 2d). This suggests that while reduced lifespan induced by tissue-specific MATR3 expression occurs at a later age, climbing defects start to manifest as early as day-15, suggesting a slow and progressive deterioration.
To investigate if these locomotion defects are caused by morphological defects in the muscles, we performed hematoxylin and eosin (H&E) staining on cross-sections of the dorsal longitudinal muscles (DLMs) in the thorax. DLMs in MATR3-expressing flies appeared smaller compared to those of controls, most likely due to muscle atrophy in these animals (Fig. 2e, f). Complementary to motor dysfunction, the atrophy phenotype was particularly severe in S85C mutant flies, which exhibited enlarged gaps between muscle fibers, indicating muscle degeneration (Fig. 2e). F115C mutant flies showed similar degeneration, but to a lesser degree than the S85C mutant flies (Fig. 2e). This finding demonstrates that muscle-targeted expression of MATR3 leads to adult-onset degeneration that is progressive in the pathogenic S85C mutant.
We next utilized this model to identify biochemical alterations caused by MATR3 mutations that lead to acquisition of toxic properties, particularly solubility of MATR3 in NP40-containing buffers. In Drosophila muscles, MATR3 WT and F115C mutant proteins were distributed almost evenly between the soluble and insoluble fractions (Fig. 2g, h). However, the S85C mutation drastically reduced the solubility of the protein, with most of the protein accumulating in the insoluble fraction (Fig. 2g, h). Next, we took advantage of the age-dependent onset of toxicity in the muscle-expression paradigm to assess for any potential correlation between toxicity and solubility. Soluble-insoluble fractionation of MATR3 in Drosophila muscles in day 1-old and day 20-old flies showed significantly higher enrichment of MATR3 protein in the insoluble fraction at day 20 compared to day 1 in both WT and mutant groups (Fig. 2i). Interestingly, at day 20, relative insolubility of S85C mutant remained higher to that of WT and F115C, suggesting that S85C mutant is possibly more prone to become insoluble (Fig. 2i).
RNA-binding domains of MATR3 drive toxicity in vivo
We sought to determine which functional domain(s) of MATR3 is responsible for mediating toxicity. We generated transgenic MATR3 lines with deletion mutations in each of the four functional domains: ΔRRM1, ΔRRM2, ΔZNF1 and ΔZNF2 (Fig. 3a). When ubiquitously expressed, deletion of the RRM2 domain completely rescued developmental lethality, yielding adults expressing MATR3 (Fig. 3b). RRM1 deletion also partially rescued developmental toxicity (Fig. 3b). However, deletion of either of the zinc-finger domains (ZNF1 or ZNF2) did not suppress toxicity (Fig. 3b). To determine if RRM1/2 deletion similarly modulates the NMJ defect, we induced conditional MATR3 expression with RU486 and labeled the third instar larval NMJs for HRP. Importantly, RRM2 deletion was sufficient to strongly rescue the NMJ defects caused by MATR3 (Fig. 3c), restoring the number of synaptic boutons to near-control levels (Fig. 3d). We then moved to an adult expression paradigm to evaluate if RRM2 deletion retains its rescue ability. In adults, deletion of RRM1 and RRM2, independently, significantly extended the lifespans of flies expressing MATR3 (Fig. 3e). However, unlike in developmental toxicity, ΔRRM2 adults retain some toxicity, with a shorter lifespan compared to controls (Fig. 3e). Taken together, adult longevity analyses suggest that MATR3 toxicity may be mediated by both RRM1 and RRM2 RNA-binding domains. ZNF1/2 deletion, on the other hand, exacerbated toxicity in the adults by significantly shortening their lifespan (Fig. 3f).
These observations prompted us to examine the role of RRM1 and RRM2 domains in mediating toxicities associated with disease-causing mutations in MATR3. To test this, we generated mutant transgenic lines including F115C-ΔRRM1, F115C-ΔRRM2, S85C-ΔRRM1 and S85C-ΔRRM2. RRM2 deletion was equally successful in rescuing developmental toxicity in flies expressing pathogenic mutant MATR3 (Fig. 4a). Interestingly, ΔRRM1, while mildly protective on its own, did not have any rescue effect on mutant MATR3 developmental toxicity (Fig. 4b). A possible explanation for this could be that F115C and S85C mutant MATR3 exert higher developmental toxicity compared to WT MATR3 in vivo, and thus, any mild protective effect of RRM1 deletion in developmental stages is negated, or perhaps irreversible. However, in adults, both RRM1 and RRM2 deletion significantly extended the lifespan of MATR3 mutant flies (Fig. 4c, d).
We then assessed the effect of RRM1 and RRM2 deletion on total levels of MATR3 WT and mutants. We observed that, especially in the WT background, the total levels of ΔRRM1 and ΔRRM2 were significantly higher compared to full-length MATR3 (Additional file 10: Fig S4), indicating that mitigated toxicities in ΔRRM1 and ΔRRM2-expressing flies is not a consequence of reduced protein expression. Next, we asked if removal of either RRM1 or RRM2 domains has any effect on modulating solubility of the MATR3 protein. Analysis of insoluble and soluble fractions showed that RRM2 deletion significantly increased MATR3 solubility (Fig. 4e–h). Deletion of the RRM1 domain, however, had a milder impact on increasing MATR3 solubility. Interestingly, while RRM2 deletion decreased the insolubility of both WT and mutant MATR3, the relative decrease in insolubility was higher in the WT background compared to the mutant background (Fig. 4f–h). This correlated with our longevity analyses, where RRM2 deletion in the WT background extended the lifespan of MATR3 flies more than in the mutant background (Figs. 3e, 4c, d). Overall, our results highlight an imperative role of the MATR3 RNA-binding domains, particularly the RRM2 domain, in mediating MATR3 WT and mutant toxicity.
Rump, a homolog of hnRNPM, is a strong modifier of MATR3 toxicity in vivo
To identify modifiers of MATR3 toxicity, we focused on proteins that have been identified as high-confidence interactors of MATR3 in two or more high-throughput studies [26, 36, 37]. Most protein–protein interactors were RNA-binding proteins involved in multiple aspects of RNA metabolism, primarily splicing. We screened the interactome for proteins previously implicated in neurodegenerative diseases, including hnRNP family proteins. We obtained publicly available RNAi lines for the selected candidate genes and screened for those that do not cause any intrinsic toxicity with ubiquitous expression (Additional file 11: Table S1). To perform the screen, we combined MATR3 WT and mutants with each candidate RNAi line, crossed them with the ubiquitous driver and looked for viable adults in the progeny (Fig. 5a).
We identified Rumpelstiltskin (rump), the Drosophila homolog of human HNRNPM, as a strong modifier of MATR3 toxicity (Fig. 5a, b) (Additional file 11: Table S1). hnRNPM is an RNA-binding protein that binds to pre-mRNA and splicing regulator complexes to regulate alternative splicing [38, 39]. Furthermore, hnRNPM is one of the strongest interactors of MATR3, identified through multiple high-throughput protein–protein interaction studies [26, 36, 37, 40]. We observed that knocking down rump (Additional file 12: Fig S5A, B) significantly rescued S85C toxicity, as evident from rescue of egg-to-adult viability in S85C flies also expressing rump RNAi (Fig. 5b). Interestingly, among all hnRNPs tested, only rump suppressed the MATR3 toxicity, suggesting a potentially prominent role for hnRNPM in mediating MATR3 toxicity in vivo (Additional file 11: Table S1). To validate this interaction, we knocked down rump in adults conditionally expressing MATR3 and assessed longevity. rump knockdown (KD) significantly increased the lifespans of flies expressing F115C- or S85C-mutant MATR3 (Fig. 5c, d, f). rump KD on its own did not exert intrinsic toxicity during development (Fig. 5b), however it did promote decreased longevity in adults compared to controls (Additional file 12: Fig S5C). Interestingly, rump KD did not have any obvious effect on WT lifespan (Fig. 5e, f), indicating a role for hnRNPM in mediating toxicity caused by pathogenic mutations in MATR3. This evidence further highlights the disease relevance of a MATR3 and hnRNPM genetic interaction in vivo. Furthermore, rump KD decreased the insolubility of MATR3 S85C in muscles (Fig. 5g, h), suggesting a potential mechanism for alleviating MATR3 toxicity in vivo.
hnRNPM genetically and physically interacts with MATR3 via its RRM2 domain in mammalian cells
To identify functional interactions between MATR3 and hnRNPM, we turned to mammalian cell systems. Mouse myoblast C2C12 cells are particularly susceptible to MATR3 overexpression, as ectopic expression of MATR3 results in cytoplasmic mislocalization of MATR3 and accumulation into cytoplasmic granules in a subset of cells (Fig. 6a, b). We also observed that overexpression of the S85C mutation led to significantly increased accumulation of MATR3 into cytoplasmic granules compared to WT overexpression (Fig. 6b). Moreover, we observed clear colocalization between MATR3 and endogenous hnRNPM in the cytoplasmic granules, suggesting that MATR3 overexpression concurrently leads to mislocalization of hnRNPM and sequestration into cytoplasmic granules (Fig. 6a). Interestingly, overexpression of F115C- and S85C-mutant MATR3 caused a higher degree of sequestration of hnRNPM into the cytoplasmic granules compared to WT (Fig. 6c). This correlates with selective suppression of MATR3-mutant toxicity by hnRNPM KD in our Drosophila models, supporting the hypothesis that hnRNPM is important for suppressing MATR3 toxicity in a mutation-dependent manner.
While hnRNPM was shown to interact with wildtype MATR3 through mass spectrometric studies [26, 36, 37], we sought to assess the physical interaction between hnRNPM and mutant MATR3. Co-immunoprecipitation revealed a physical interaction between hnRNPM and WT MATR3 as well as F115C and S85C mutants (Fig. 6d). Interestingly, deletion of the RRM2 domain interrupted this interaction, suggesting that the RRM2 domain is required for mediating interaction between MATR3 and hnRNPM (Fig. 6d). To investigate if this interaction is RNA-dependent, we treated the immunoprecipitated lysate with RNaseA to degrade the RNA. Treatment with RNaseA decreased the interaction between MATR3 and hnRNPM in mammalian cells (Fig. 6e), suggesting that these proteins interact, at least partially, through binding to shared RNA targets.
MATR3 and hnRNPM share common transcriptomic targets
We hypothesized that MATR3 and hnRNPM functionally interact to regulate metabolism of shared RNA targets and that dysregulation of these transcripts possibly leads to disease pathogenesis. We employed an in silico approach to further elucidate this functional interaction. We compared published eCLIP datasets [34] for MATR3 and hnRNPM from two different cell types, K562 (lymphocytes) and HepG2 (hepatocytes), and mined transcripts bound by both proteins (Additional file 4: Tables S2, Additional file 5: Tables S3). Motif discovery across significantly enriched eCLIPs peaks for MATR3 and hnRNPM revealed unique but distinct motifs for MATR3 and hnRNPM. MATR3 binding sites were enriched in AGAAG and UCUUC motifs (Fig. 7a), while hnRNPM binding sites were enriched in UGUUG and ACAAC motifs (Fig. 7b), indicating that these proteins bind to unique motifs within their respective transcript targets. Importantly, we found that MATR3 and hnRNPM share appreciable overlap in the transcripts that they bind (Fig. 7c, d) (Additional file 9: Tables S3). In K562 cells ~ 46% of MATR3-bound transcripts are also bound by hnRNPM (Fig. 7c); in HepG2 cells, ~ 26% of MATR3-bound transcripts are also bound by hnRNPM (Fig. 7d). We then compared the binding patterns of MATR3 and hnRNPM using their read density across the commonly shared transcripts. Considering a window of 300 nucleotides around significantly enriched (p < 0.05 and fold-change ≥ 4) MATR3 peaks from K562 and HepG2 cells, we observed that hnRNPM read density is highly enriched around MATR3 peak centers in both cell types (Fig. 7e, d). The vice versa was also observed, in that, MATR3 read densities were highly enriched around hnRNPM peak centers (Additional file 13: Fig S6A, B), suggesting that MATR3 and hnRNPM bind in close proximity to each other on the shared transcript.
To further investigate the nature of MATR3 and hnRNPM interaction with shared transcripts, specifically in the context of ALS/myopathy-linked mutations in MATR3, we performed RNA-immunoprecipitation of MATR3 and hnRNPM in HEK293T cells expressing either MATR3-WT or disease-linked mutations MATR3-F115C and MATR3-S85C (Fig. 8a). RT-qPCR for rationally-selected targets from the shared transcriptome, particularly those that show high binding affinity to both MATR3 and hnRNPM, and additionally, have been previously associated to neurodegeneration in ALS including DYRK1A, SMYD3 and ZNF644, revealed that F115C and S85C mutants exhibit significantly higher binding to these transcripts compared to WT (Fig. 8b–d). Importantly, we found that hnRNPM also exhibited significantly higher binding to the same transcripts in cells expressing F115C and S85C compared to cells expressing MATR3-WT (Fig. 8b–d). This suggests that disease-causing mutations in MATR3 lead to aberrant RNA-binding to MATR3 and, concurrently, to its interacting-partner, hnRNPM. Gene ontology analysis of all common targets (Additional file 6: Tables S4) revealed that the top 20 most enriched unique biological processes shared between the two cell types included significant processes such as neurogenesis, proteasomal protein ubiquitination, histone modification & chromosome organization (Fig. 8h, i). Furthermore, gene ontology analysis on the basis of disease associations indicated that the shared targets are enriched in genes associated with neurodevelopmental disorders (Additional file 6: Tables S4). These results indicate that MATR3 and hnRNPM both bind to, and thus may co-regulate, transcripts that have important functions in nervous system development and maintaining cellular/neuronal health. Thus, it is likely that aberrant regulation of these transcripts, and consequently these processes, caused by mutations in MATR3 result in disease pathogenesis.
We sought to further validate this hypothesis in our in vivo model. We rationally-selected candidate targets from the MATR3-hnRNPM shared transcriptome, focusing on candidates that have previously been shown by other studies to be regulated by MATR3 in mammalian cells [20, 27, 34], and additionally exhibit disease-relevance in ALS and other neurodegenerative disorders. Assessment of levels of candidate transcripts, including Dystrophin (Dys), Ataxin-1 (Atx-1), Adenylate kinase 3 (Adk3) and Sialyltransferase (SiaT), revealed significantly increased mRNA levels in mutant MATR3-expressing flies compared to control (Additional file 14: Fig S7), while it remained unchanged in the WT compared to control. Furthermore, RNA-immunoprecipitation of MATR3 and hnRNPM in human cells expressing either MATR3-WT or F115C and S85C mutations, showed significantly higher binding of UTRN mRNA (Dys homolog) to both F115C and S85C mutants compared to WT (Additional file 15: Fig S8). Interestingly, UTRN mRNA also showed significantly higher binding to hnRNPM in cells expressing either F115C or S85C mutations compared to cells expressing MATR3-WT (Additional file 15: Fig S8). These results further support our idea that hnRNPM modulates mutant MATR3 toxicity, and both proteins genetically and physically interact to regulate common targets.