Nucleotide repeat expansions can elicit neurodegeneration as RNA by sequestering specific RNA-binding proteins, preventing them from performing their normal functions. Conversely, mutations in RNA-binding proteins can trigger neurodegeneration at least partly by altering RNA metabolism. In Fragile X-associated tremor/ataxia syndrome (FXTAS), a CGG repeat expansion in the 5′UTR of the fragile X gene (FMR1) leads to progressive neurodegeneration in patients and CGG repeats in isolation elicit toxicity in Drosophila and other animal models. Here, we identify the amyotrophic lateral sclerosis (ALS)-associated RNA-binding protein TAR DNA-binding protein (TDP-43) as a suppressor of CGG repeat-induced toxicity in a Drosophila model of FXTAS. The rescue appears specific to TDP-43, as co-expression of another ALS-associated RNA-binding protein, FUS, exacerbates the toxic effects of CGG repeats. Suppression of CGG RNA toxicity was abrogated by disease-associated mutations in TDP-43. TDP-43 does not co-localize with CGG RNA foci and its ability to bind RNA is not required for rescue. TDP-43-dependent rescue does, however, require fly hnRNP A2/B1 homologues Hrb87F and Hrb98DE. Deletions in the C-terminal domain of TDP-43 that preclude interactions with hnRNP A2/B1 abolish TDP-43-dependent rescue of CGG repeat toxicity. In contrast, suppression of CGG repeat toxicity by hnRNP A2/B1 is not affected by RNAi-mediated knockdown of the fly TDP-43 orthologue, TBPH. Lastly, TDP-43 suppresses CGG repeat-triggered mis-splicing of an hnRNP A2/B1-targeted transcript. These data support a model in which TDP-43 suppresses CGG-mediated toxicity through interactions with hnRNP A2/B1 and suggest a convergence of pathogenic cascades between repeat expansion disorders and RNA-binding proteins implicated in neurodegenerative disease.
Fragile X-associated tremor/ataxia syndrome (FXTAS) is an inherited neurodegenerative disorder characterized by late-onset, progressive gait difficulties, dementia and tremors which affects 1/3000 men over age 50 (1). FXTAS results from a modest expansion of a trinucleotide (CGG) repeat in the 5′ UTR of the fragile X mental retardation gene (FMR1) (2). Expansions to >200 CGG repeats silence FMR1 transcription and cause fragile X syndrome, a frequent cause of autism and intellectual disability (3,4). In contrast, ‘pre-mutation’ range expansions (55–200 CGG repeats) cause FXTAS, characterized by enhanced FMR1 transcription and the accumulation of CGG repeat-containing FMR1 mRNA in nuclear foci (5–7).
The mechanism by which CGG repeat expansion elicits neurodegeneration is unclear (8). Based on similarities to the molecular basis of myotonic dystrophy, it is suggested that expanded CGG RNA repeats bind to and sequester-specific RNA-binding proteins, leading to a perturbation of their normal functions (2,9,10). In myotonic dystrophy, expanded CUG repeats in the 3′UTR of DMPK or expanded CCUG repeats in an intron of ZNF9 are each capable of binding to Muscleblind-like (MBNL) and CELF1 splicing factors, leading to altered RNA splicing and a convergence of clinical phenotypes despite disparate genomic locations and gene functions (11). In mouse models of DM1, certain aspects of the pleiotropic phenotypes are suppressed by co-expression of MBNL or recapitulated by knocking out MBNL proteins (12–14).
In FXTAS, evidence for a primary RNA toxic gain-of-function mechanism is less complete. A number of different proteins have been identified that interact directly or indirectly with CGG repeats including Purα, CELF1, hnRNP A2/B1, Sam 68, hnRNP-G and the Drosha/DGCR8 miRNA biogenesis complex (15–18). One of these, hnRNPA2/B1, was independently identified as a component of pathological inclusions isolated in FXTAS patient brain tissue by unbiased mass spectroscopy (19). Overexpression of hnRNP A2/B1 suppresses degeneration in CGG repeat toxicity in Drosophila; moreover, the expression of CGG repeats in mammalian neurons and in Drosophila disturbs hnRNP A2/B1 functions, including dendritic mRNA transport and suppression of retrotransposon activity (20,21).
In parallel, a growing number of neurodegenerative disorders have been linked to mutations and/or aggregation of RNA-binding proteins (9,22). Specifically, ubiquitinated, cytoplasmic inclusions of TAR DNA-binding protein (TDP-43) are characteristic of most sporadic and familial forms of amyotrophic lateral sclerosis (ALS) and some forms of frontotemporal dementia (FTD) (23). Additionally, mutations in TDP-43, TAF-15, and FUS (fused in sarcoma/translocated in liposarcoma) are known to cause ALS and/or FTD (24–28). More recently, mutations in hnRNP A2/B1 and hnRNPA1 were found to cause multisystem proteinopathy, a disorder that includes muscle, brain and peripheral nervous system involvement (29). Thus, the mis-localization and dysfunction of RNA-binding proteins is capable of eliciting neurodegeneration in multiple contexts.
Recently, an intronic GGGGCC repeat expansion in C9ORF72 was identified as the most common known inherited cause of ALS and FTD (30,31). C9FTD/ALS patients exhibit both TDP-43-positive cytoplasmic aggregates and GGGGCC repeat nuclear RNA foci in both autopsy tissues and in iPS-derived patient neurons (30,32). These findings suggest a potential convergence of RNA-mediated and RNA-binding protein-mediated neurodegenerative pathogenic cascades. To explore the potential interaction between RNA repeat toxicity and RNA-binding proteins implicated in neurodegenerative disorders, we utilized a Drosophila model of CGG repeat toxicity (a model of FXTAS) to examine for genetic interactions with TDP-43 and FUS, two RNA-binding proteins implicated in neurodegenerative disease. We demonstrate that co-expression of the ALS-FTD-associated protein TDP-43 suppresses CGG repeat expansion-associated toxicity. TDP-43-mediated suppression of CGG repeat toxicity is post-transcriptional, occurs independent of TDP-43 RNA-binding capacity, and requires interaction with hnRNP A2/B1 homologues. These data suggest roles for TDP-43 and hnRNP A2/B1 in CGG RNA repeat toxicity and support a disease model in which complex interactions between RNA-binding proteins and toxic RNA modulate RNA homoeostasis and contribute to neurodegenerative disease.
Overexpression of TDP-43 suppresses CGG repeat toxicity
We utilized a previously described Drosophila model of CGG repeat toxicity where the FMR1 5′UTR containing a 90 CGG repeat expansion was placed upstream of eGFP under the control of a UAS promoter (CGG90-GFP) (10). When transgene expression is targeted to the Drosophila eye using a GMR-Gal4 driver, these flies demonstrate substantial eye degeneration compared with flies expressing eGFP alone [(10), Fig. 1A]. Several RNA-binding proteins are known to associate specifically with CGG repeats in vitro and in cell culture-based systems (15–18) and a subset of these are capable of modulating CGG repeat toxicity when their expression is altered in Drosophila (15,16,18).
We evaluated the impact of co-expressing human myc-TDP-43 or FUS in CGG90-GFP flies. As reported (33,34), expression in the eye of wild-type TDP-43 alone elicited a very modest rough eye phenotype, and expression of wild-type FUS alone causes mild eye depigmentation with retention of normal eye structure (Fig. 1D and F). Unexpectedly, overexpression of TDP-43 in CGG90-GFP expressing flies ameliorated the rough eye phenotype (Fig. 1C). The suppression of expanded CGG repeat toxicity by TDP-43 was consistent and statistically significant over multiple crosses (Fig. 1G). In contrast, overexpression of WT FUS exacerbated CGG90-GFP-associated rough eye phenotypes and elicited significant depigmentation (Fig. 1E and G).
CGG repeat RNA levels and RAN translation are not altered by TDP-43
TDP-43 is a multifunctional protein, first identified as a DNA-binding protein capable of suppressing HIV-1 transcription (35). TDP-43 binds to both RNA and DNA to affect transcription, RNA stability and alternative splicing (36–38). In the cytoplasm, it contributes to stress granule dynamics (39,40) and RNA transport (41). We, therefore, serially evaluated components of TDP-43 function to determine how its overexpression triggers CGG repeat toxicity suppression.
Co-expression of HDAC6 reduces CGG90-GFP mRNA expression (42) and loss of TDP-43 leads to down-regulation of dHDAC6 mRNA (36,43). To determine whether TDP-43 suppression of CGG toxicity is a result of reduced transcription of CGG repeats either directly as a repressor or indirectly by stabilizing HDAC6 mRNA, we assessed CGG90-GFP mRNA expression in the presence or absence of TDP-43. Overexpression of myc-TDP-43 in CGG90-GFP flies, however, had no impact on HDAC6 mRNA by qRT–PCR (Fig. 2A). Expression of HDAC3 and HDAC11 mRNA was similarly unaltered by overexpression of TDP-43 (data not shown). Consistent with this, CGG90-GFP mRNA levels were unaffected by co-expression of TDP-43 (Fig. 2A). Thus, the effects of TDP-43 on CGG repeat toxicity appear to be post-transcriptional.
CGG repeat-containing RNAs form foci in FXTAS patient tissues and in cellular and animal models of disease (7,17,44). One model of pathogenesis in FXTAS posits that the CGG repeat RNA binds to and sequesters specific RNA-binding proteins into proteinaceous inclusions. Consistent with this model, CGG repeat-binding proteins Purα, hnRNP A2/B1, Sam68 and DROSHA/DGCR8 co-localize with CGG RNA in FXTAS patients (15–18). In contrast, TDP-43 does not to interact with CGG RNA in vitro (17) and CGG90-GFP expression in isolation did not alter expression of the TDP-43 fly orthologue TBPH (Supplementary Material, Fig. S1). To examine whether expressed TDP-43 might indirectly associate with CGG RNA foci in vivo or alter RNA foci numbers, we combined in situ hybridization to CGG repeats with immunofluorescence detection of TDP-43 (Fig. 2B). Myc-TDP-43 expressed alone was present in both the nucleus and cytoplasm (33). CGG90-GFP expressing flies, but not flies expressing TDP-43 alone, exhibit nuclear and cytoplasmic CGG repeat RNA foci (Fig. 2B). Co-expression of TDP-43 protein with CGG90-GFP did not alter the distribution of TDP-43 or the distribution or number of CGG RNA foci (Fig. 2B and C). Moreover, there is little co-localization (<30% total co-staining) of TDP-43 to CGG RNA foci and no evidence for sequestration based on re-distribution of TDP-43 (Fig. 2B).
Recently, we demonstrated that CGG repeat expansions in the 5′ UTR of FMR1 are capable of eliciting Repeat Associated Non-AUG (RAN) translation to produce a polyglycine-containing protein (FMRpolyG) (44). This unconventional form of translation is toxic in cell culture systems and contributes to neurodegeneration in Drosophila models of FXTAS (44). In CGG90-GFP expressing flies, RAN translation triggers formation of bright FMRpolyG-GFP insoluble aggregates in expressing tissues (44). To test whether overexpression of TDP-43 affects RAN translation, we examined the formation of FMRpolyG-GFP aggregates in flies expressing only CGG90-GFP alone or together with myc-TDP-43. Co-expression of TDP-43 did not alter the level of RAN translation product FMRpolyG-GFP as detected by western blot (Fig. 2D and E) or change the number or distribution of FMRpolyG-GFP aggregates (Fig. 2F and G). Moreover, we observed no significant co-localization of TDP-43 with GFP aggregates in CGG90-GFP expressing flies (data not shown). Together, these results suggest that TDP-43 modulates CGG repeat-associated RNA-mediated toxicity, rather than by altering CGG repeat transcription or through enhanced RAN translation of FMRpolyG. However, our current data cannot rule out a contribution to the observed phenotypes from CGG RAN translation in other reading frames.
TDP-43 expression in FXTAS patient-derived samples
We next examined whether the expression or distribution of TDP-43 was altered in FXTAS patient samples. We first performed immunohistochemical analysis of TDP-43 distribution in human autopsy samples. In both control and FXTAS brain autopsy samples, TDP-43 was diffusely localized within the nucleus without any cytoplasmic aggregation (as is typically seen in FTD) or accumulation into prominent nuclear inclusions (Fig. 3A). As we had limited access to autopsy tissue, we were not able to determine whether there were any significant differences in TDP-43 mRNA or protein expression in human patient brain samples. However, we did evaluate the expression of TDP-43 within fibroblast cell lines from control and premutation carriers (42,45). These studies demonstrated no significant differences in TDP-43 RNA or protein expression between premutation carriers, FXTAS patients and controls. Moreover, we found no correlation between CGG repeat size and either TDP-43 mRNA or protein levels (Fig. 3B and C).
Functional requirements for TDP-43 to suppress CGG repeat toxicity
TDP-43 protein possesses an N-terminal domain containing a nuclear localization signal (NLS), a central region containing two RNA recognition motifs (RRMs), and a C-terminus containing a prion-like, glycine-rich domain that is important for protein–protein interactions including functional oligomerization during RNA granule assembly (schematically represented in Fig. 4A) (46,47). The RRM domains allow RNA recognition and binding, while the prion-like domain is the site of most ALS and FTD disease-causing mutations (46,48). TDP-43 predominantly resides in the nucleus (49), but it actively shuttles between the nucleus and the cytoplasm where it associates with and regulates stress granule dynamics (39,40). To test whether the cellular localization of TDP-43 affects its ability to suppress CGG repeat toxicity, we utilized UAS-TDP-43 ΔNLS (K82A/R83A/K84A) and UAS-TDP-43 ΔNES (L248A/I249A/I250A) mutant transgenic Drosophila lines (33). As previously reported (33), GMR-Gal4-driven expression of TDP-43-ΔNES alone elicits no rough eye phenotype, while expression of TDP-43-ΔNLS alone elicits a severe rough eye phenotype (Fig. 4G and 4I). When these two mutant lines were crossed to CGG repeat expressing flies, the TDP-43-ΔNES did not modulate CGG repeat toxicity (Fig. 4B and F), whereas co-expression of TDP-43-ΔNLS with CGG90-eGFP proved to be lethal even when transgene expression was largely limited to the eye (Fig. 4H).
To test whether RNA-binding properties or protein–protein interaction, domains are required for TDP-43 to suppress CGG repeat-associated toxicity, we generated new fly lines carrying p-element insertions under a UAS promoter of HA-tagged wild-type TDP-43, HA-TDP-43 containing mutations that prevent RNA binding [FFLL, see (50)], or HA-TDP-43 containing a deletion of a critical region (Δ321–366) for interactions with other RNA-binding proteins—most notable of which is hnRNP A2/B1 (46). These lines were first crossed to flies expressing GMR-Gal4 alone. Expression of either WT HA-TDP-43 or TDP-43-FFLL by themselves elicited very little toxicity (Fig. 4E and K), consistent with published reports (33), whereas the expression of TDP-43-Δ321–366 exhibited a mild rough eye phenotype (Fig. 4M), despite similar transcript levels by qRT–PCR for all three TDP-43 variants (Fig. 4O).
To assess the impact of these mutations on TDP-43's ability to suppress CGG repeat toxicity, we evaluated the impact of these same lines crossed to flies expressing both GMR-Gal4 and UAS-CGG90-GFP. Expression of WT HA-TDP-43, similar to the myc-TDP-43 line used in previous experiments, showed partial suppression of CGG repeat toxicity (Fig. 4D). Expression of TDP-43-FFLL-HA, which lacks RNA-binding capacity, also suppressed CGG repeat toxicity, though to a lesser extent then the WT protein (Fig. 4J). In contrast, C-terminal mutant TDP-43 (TDP-43-Δ321–366-HA) did not suppress the CGG repeat toxicity and instead slightly enhanced the rough eye phenotype (Fig. 4L and O).
We also assessed the ability of three ALS disease-causing mutations in TDP-43 to suppress CGG repeat toxicity. Two of these mutations, TDP-43 M337V and TDP-43 Q331K, reside within the 321–366 protein–protein interaction domain (24). The third mutation,G294A, lies outside of this region and is both less aggregate prone and less toxic in yeast than the other two mutations (24,51). Like TDP-43-Δ321–366, TDP-43 M337V expression in isolation elicits a mild rough eye phenotype (Fig. 5F and H, 33). However, when either of two TDP-43 M337V lines was crossed to CGG90-GFP expressing flies, there was a decrease in viable progeny from flies carrying both CGG and TDP-43 mutant transgenes. In viable progeny, rough eye phenotypes were significantly enhanced (Fig. 5E, G and Q). Results for TDP-43 G294A were very similar to M337, with only very mild phenotypes in isolation but a decrease in viable progeny and enhanced rough eye phenotypes in flies expressing both CGG repeats and the mutant TDP-43 transgene (Fig. 5M–Q). Multiple fly lines expressing TDP-43 Q331K mutations exhibited minimal toxicity on their own but were unable to suppress CGG repeat-associated toxicity (Fig. 5I–L and Q). These differential effects were not easily explained by differences in TDP-43 transgene expression [Supplementary Material, Fig. S2, (52)]. Taken together, these findings implicate the C-terminal domain of TDP-43 as critical for suppression of CGG repeat-associated phenotypes and suggest that mutations in TDP-43 preclude suppression of CGG RNA toxicity.
TDP-43 suppression of CGG repeat toxicity is dependent on hnRNP A2/B1 homologues
TDP-43 predominantly interacts with two types of proteins: nuclear splicing factors and cytoplasmic RNA-binding proteins involved in protein translation (53). One particularly well-studied interactor is the RNA-binding protein hnRNP A2/B1 (54). HnRNP A2/B1 interacts directly with TDP-43 through the C-terminal domain and this interaction is critical to many aspects of TDP-43 function, including its roles in splicing (46). Recently, mutations in hnRNPA2/B1 and the closely related paralogue hnRNPA1 were found to cause multisystem proteinopathy and ALS (29). Interestingly, hnRNP A2/B1 is a component of ubiquitin-positive inclusion in FXTAS patients (19), and it directly binds to CGG repeat RNA in vitro (15,16). Both human hnRNP A2/B1 and the fly homologues Hrb87F and Hrb98DE are able to suppress CGG repeat toxicity in Drosophila (16). To test whether Drosophila hnRNP A2/B1 homologues were important for TDP-43-mediated suppression of CGG repeat toxicity, we utilized RNAi fly lines targeting Hrb87F (Hrp36) and Hrb98DE (Hrp38) (Fig. 6A and B). Of note, CGG90-GFP flies exhibit normal expression of both hnRNP A2/B1 homologues (Supplementary Material, Fig. S3). Expression of these RNAi lines in isolation or in the presence of myc-TDP-43 had no apparent effect on eye morphology (Fig. 6D–H). Consistent with their known roles in CGG repeat toxicity (16), knockdown of Hrb87F or Hrb98DE slightly exacerbated CGG90-GFP rough eye phenotypes (Fig. 6K, M and O). When either Hrb87F or Hb98DE expression was reduced, the ability of TDP-43 to suppress CGG repeat toxicity was completely blocked, suggesting that hnRNP A2/B1 homologues are critical for TDP-43 to suppress CGG repeat toxicity (Fig. 6L, N and O).
Normal TBPH expression is not required for hnRNP A2/B1 suppression of CGG toxicity
To test whether hnRNP A2/B1 and TDP-43 act interdependently to suppress CGG repeat toxicity, we examined whether the normal expression of the TDP-43 orthologue TBPH is required for hnRNP A2/B1-mediated suppression of CGG toxicity. Neither RNAi directed against TBPH (Fig. 7A) or a heterozygous loss of function TPBH alleles had obvious adverse effects on eye morphology, either in isolation or in the context of CGG90-GFP co-expression [Fig. 7D and E; Supplementary Material, Fig. S4, (16)]. Consistent with published reports (16), overexpression of hnRNP A2 suppressed CGG repeat toxicity (Fig. 7F and J). However, unlike the corollary experiment where hnRNP A2/B1 is required for TDP-43's effects on CGG repeats, RNAi knockdown of TBPH did not prevent hnRNP A2/B1-mediated suppression of CGG RNA repeat toxicity (Fig. 7H and J). These results suggest that normal amounts of TBPH are not required for hnRNP A2/B1-mediated suppression of CGG repeat-associated toxicity. However, because of lethality associated with loss of TBPH, the effects of the complete absence of TBPH are less clear.
TDP-43 restores alternative splicing of EPH, which is perturbed by CGG repeat expression
HnRNP A2/B1 binds to its target RNAs and modulates alternative RNA splicing of these targets (55). In Drosophila, the Hrb87F and Hrb98DE splicing targets have been systematically investigated (56). One target shared by both Hrb87F and Hrb98DE is EPH, which encodes a protein tyrosine kinase receptor important for nervous system development and function (57,58). Exon 8b of the EPH gene is either included or excluded to produce EPH isoforms with or without this exon (Fig. 8A). Both isoforms were present in fly eyes as shown by qualitative RT–PCR using primers common to both isoforms (Fig. 8B), or specific primers for individual isoforms (Fig. 8C). Previously, it was reported that RNAi knockdown of Hrb87F and Hrb98DE in S2 cells can modulate the ratios of such isoforms (56). Using a quantitative RT–PCR-based approach, we determined the relative ratios of this particular splicing event involving inclusion or exclusion of exon 8b. Hrb87F and Hrb98DE act to favour the inclusion of exon 8b from the transcript (Fig. 8A). Consistent with this, RNAi-based knockdown of either Hrb87F or Hrb98DE leads to decreased inclusion of exon 8b compared with control lines (Fig. 8D), similar to published studies in Drosophila S2 cells (56). In contrast, ectopic expression of TDP-43 in isolation had no effect on EPH splicing (Fig. 8D). To test whether CGG repeat expression alone might sequester Hrb87F and Hrb98DE to a significant enough degree to impact their normal function, we assessed EPH splicing in CGG90-GFP expressing flies. CGG90-GFP expression resulted in even greater suppression of Exon 8b inclusion than seen with knockdown of either Hrb87F or Hrb98DE alone, consistent with a significant functional impairment of both proteins (Fig. 8D). If TDP-43 imparts its phenotypic rescue on CGG repeat toxicity through its interactions with hnRNP A2/B1 homologues, then we reasoned its overexpression might correct these splicing phenotypes. Consistent with this reasoning, TDP-43 co-expression leads to a restoration of EPH exon 8b splicing in CGG90-eGFP flies (Fig. 8B and E).
Our data define a novel role for TDP-43 in CGG repeat toxicity in Drosophila. The ability of TDP-43 to suppress CGG RNA-associated phenotypes is dependent on its ability to interact with two Drosophila homologues of hnRNP A2/B1, a protein that itself can directly bind CGG repeat RNA and modulate CGG repeat-associated phenotypes. We further demonstrate that an hnRNP A2/B1 alternative splicing target is dysregulated in CGG expressing flies and that co-expression of TDP-43 corrects this mis-splicing event. Taken together, these findings are consistent with a model, where TDP-43 acts to prevent functional sequestration of hnRNP A2/B1 by CGG repeats and thus alleviate toxicity. Moreover, the two RNA-binding proteins studied here, TDP-43 and hnRNP A2/B1, are both implicated in a wide range of neurodegenerative disorders, including ALS, FTD, and multisystem proteinopathy (22,29,47). This work implicates the normal functions and interactions of these two proteins in RNA repeat disorders, suggesting a convergence of pathogenic cascades in these two classes of neurodegenerative disease.
TDP-43 is essential in development in mammals, and homozygous knockouts are lethal in both mice and Drosophila (59,60). RNAi knockdown of TDP-43 in cultured tumour cells and in differentiated neurons causes cell death (61,62), while global knockdown of TBPH is lethal in adult flies, suggesting critical roles for the protein after development (63,64). Indeed, postnatal loss of neuronal TDP-43 function causes neurodegeneration in mice (65,66). In contrast, overexpression of human TDP-43 can itself elicit neurodegeneration, with greater toxicity observed with disease causing mutations in mice (67,68), rats (69,70) and in flies (33,63). In addition, cytoplasmic inclusions of TDP-43 are widely observed in various neurodegenerative disorders including many subtypes of ALS, FTD and other dementias (as reviewed in 22,47). Thus, the finding that TDP-43 overexpression suppresses CGG repeat toxicity is surprising, and suggests that some native functions of TDP-43 in RNA homoeostasis may impinge on critical pathways in FXTAS pathogenesis.
Our data do not support a model where CGG repeats directly sequester TDP-43. TDP-43 is not an abundant component of ubiquitin-positive inclusions in FXTAS patient tissue (19), and TDP-43 does not bind to CGG repeat RNA in silico, in vitro or in cell culture (17). Furthermore, when TDP-43 and CGG repeat RNAs are co-expressed in flies, they do not co-localize (Fig. 2). Lastly, neither RNAi-based knockdown of TBPH nor TPBH heterozygous null mutations exacerbate CGG repeat-associated toxicity [Fig. 6 and (16)]. Thus, the suppressive effects of TDP-43 on CGG repeat toxicity appear to reflect a protective gain-of-function rather than a loss of TDP-43 function as is proposed for some other RNA-binding proteins such as hnRNP A2/B1 and Drosha/DGCR8 (16,18).
Similarly, the suppressive effects of TDP-43 do not appear to be mediated transcriptionally. Knockdown of TBPH or TDP-43 leads to decreased HDAC6 mRNA levels in human cells and Drosophila (43). We thus initially hypothesized that the effects of TDP-43 on CGG repeat toxicity might be mediated by an enhancement of HDAC6 expression, as HDAC6 suppresses transcription of CGG repeat-containing mRNAs (42). However, in our experimental system, overexpression of TDP-43 did not impact either CGG90-EGFP mRNA levels or HDAC6 mRNA levels in flies expressing CGG repeats. The lack of effect on HDAC6 mRNA levels with TDP-43 overexpression may reflect differences in an experimental design with previous studies, which utilized TBPH knockdown or heterozygous null fly lines. Alternatively, the presence of the CGG repeat mRNA in these flies might interfere with this process.
HnRNP A2/B1 is a member of a larger group of heterogeneous nuclear RNA-binding proteins that modulate RNA transcription, splicing and transport (71). It interacts with both normal length (30 repeat) and pathogenic length (90 repeat) CGG repeat RNA (16). Given that knocking down expression of either hnRNP A2/B1 homologue prevents TDP-43-mediated suppression of CGG repeat toxicity, it appears likely that hnRNP A2/B1 acts as a mediator for TDP-43's protective roles. This is reminiscent of findings related to another RNA-binding protein, CUGBP1/CELF1. CELF1 does not bind CGG repeat RNA directly, yet it suppresses CGG repeat-mediated toxicity in Drosophila. (16). As with TDP-43, CELF1 suppression of CGG repeat toxicity requires adequate expression of Drosophila hnRNP A2/B1 homologues (16).
Taken together, we propose that co-expression of RNA-binding proteins such as CELF1 and TDP-43 with CGG repeats alters the composition, behaviour and distribution of multimeric RNA-binding protein complexes within cells. These alterations somehow prevent sequestration of hnRNP A2/B1 by CGG repeats, allowing it to perform its normal functions. Consistent with this model, TDP-43 expression corrects a CGG repeat-triggered hnRNP A2-mediated mis-splicing event. In this context, the finding that co-expression of ALS mutant forms of TDP-43 either exacerbate or fail to rescue CGG repeat toxicity (Fig. 5) is particularly interesting. These TDP-43 mutants can still bind to hnRNP A2 in vitro (46), but their dynamics in RNA protein complexes may be altered based on studies of stress granule dynamics in transfected cells (40).
TDP-43 and hnRNP A2/B1’s roles in CGG toxicity raise interesting questions related to the relevance of these findings to other repeat disorders, most notably the intronic GGGGCC repeat implicated in FTD and ALS. The primary sequences of CGG and CCGGGG repeats are highly similar, with shared predicted secondary structures and a mutual tendency to form G-quadruplexes (72,73). HnRNP A2/B1 and its paralogue hnRNP A1 are capable of interacting with GGGGCC repeats (74,75). Given the TDP-43-positive cytoplasmic inclusions observed in C9 FTD/ALS, it will be important to evaluate whether similar genetic interactions in these proteins might alter GGGGCC repeat toxicity in model systems.
MATERIALS AND METHODS
Drosophila stocks and genetics
All flies were maintained with standard food and culture conditions at 25°C unless otherwise stated. Fly lines from the Bloomington Stock Center are: GMR-Gal4 (#8605), TBPHKG08578 insertion mutation line (#14737), RNAi lines against TBPH (#29517), against Hrb87F (#31244) and Hrb98DE (#31303), UAS-GFP (multiple lines). GMR-Gal4: UAS-(CGG)90-eGFP (CGG90-GFP) recombinant line was a kind gift from Peng Jin (10). UAS-myc-TDP-43-WT, UAS-TDP-43-ΔNES, UAS-TDP-ΔNLS, UAS-Myc-M337V were described (33), UAS-myc-hnRNP A2 was made as in Kim et al. (29). UAS-FUS fly line was a kind gift from Udai Pandey (34). N-terminally myc-tagged TDP-43 G294A was cloned into the pUAST vector (the starting methionine of TDP-43 was removed to avoid dual start site activity) and injected commercially (Bestgene, Chino Hills, CA, USA).
The UAS-TDP-43-WT-HA, UAS-TDP-43-FFLL-HA, UAS-TDP-43-Δ321–366-HA, UAS-TDP-43-Q331K-HA transgenic flies were made as follows: the initial plasmids pFLAG-TDP-43-WT, pFLAG-TDP-43-Δ321–366 and pFLAG-TDP-43-Q331K were from Buratti and Baralle (46), and TDP-43-FFLL from Voigt (50). TDP-43 cDNA fragments from these vectors were PCR amplified out (TDP-43 forward 5′-CACCATGTCTGAATATATTCGGGTAACCGAAG-3′, and TDP-43 reverse 5′-CATTCCCCAGCCAGAAGACTTAG-3′) and then were cloned to pEntry-TOPO vector (Life Technologies). Individual pEntr-TDP-43 vectors were recombined with Gateway Plasmid pTWH (Drosophila Genomic Resource Center). The resulting plasmids were used to generate transgenic flies by standard P-element insertion with the exception of TDP-43 Q331K, which was recombined to plasmid pUASg-HA-AttB (76) and injected into AttP line 9723 (Bestgene, Chino Hills, CA, USA). Transgenic lines were crossed to GMR-Gal4 to test TDP-43 mRNA expression level. Those lines with TDP-43 mRNA levels that were closest to UAS-myc-TDP-43 were selected for further crosses.
Human fibroblast cell lines
Nine fibroblast cell lines were obtained from Dr Paul Hagerman and colleagues (University of California at Davis). The clinical state of these patients has been described previously (45). All cell lines were from male subjects. Nomenclature used follows that used in the publication where these lines were first published (45). Briefly, there are three control lines: C1 (CGG31), C4 (CGG22) and C5 (CGG30), Six cell lines from CGG expanded lines :P3 (CGG81), P4 (CGG70), P5 (CGG67), F1 (CGG122), F2 (CGG105) and F3 (CGG97). Fibroblasts were cultured and maintained as previously described (45) and repeat size was confirmed by PCR for all lines using C and F primers (42).
Eye phenotype examination and scoring
Eye morphology of 1–3 day post-eclosion flies was quantitatively scored as previously described (77). Briefly, over 100 eyes from each genotype were scored based on the following criteria: supernumerary inter-ommatidial bristles, abnormal bristle orientation, ommatidium fusion, ommatidium pitting, disorganization of ommatidial array, retinal collapse. The presence of each feature is given 1 point, and additional 2 points were given if >5% of the eyes were affected, or 4 points if >50% of eyes were affected. Higher score means the eyes are more degenerated. The scores were calculated and presented as means ± SEM and non-parametric analysis (Mann–Whitney U-test) was applied to determine the statistical differences between groups. The representative fly eye images were taken by Leica M125 stereomicroscope and photographed with a Leica DFC425 digital camera as previously described (44).
RNA isolation and qRT–PCR
Total RNA was extracted from 12–15 fly heads of each genotype using Trizol (Life Technologies) and quantitated by nano-drop. Total RNA from fibroblast cells was extracted by lysis in Trizol buffer following manufacturer's manual; 1 μg of total RNAs was then reverse-transcribed to cDNA by iScript cDNA synthesis kit (Bio-Rad). The primer sets used in the quantitative PCR step for the individual genes were as follows:
GFP FORWARD: 5′-TCTTCTTCAAGGACGACGGCAACTAC-3′
GFP REVERSE: 5′-GTACTCCAGCTTGTGCCCCAGGATGT-3′
HDAC6 FORWARD: 5′-CTGACGGAGCGCTGCCTG-3′
HDAC6 REVERSE: 5′-GCAGGGAGAGCTCGAATGTGG-3′
HRB87F FORWARD: 5′-GGCCACCACCTCCGGTGTTG-3′
HRB87F REVERSE: 5′-CCAGCAGAGCTACGGCGGC-3′
HRB98DE FORWARD: 5′-CGGCAACCAGAATGGTGGCGG-3′
HRB98DE REVERSE: 5′-CCACCACCGAAGCTGTTGTTGCC-3′
TBPH FORWARD: 5′-GCAGCGCAGACGGGTCTCG-3′
TBPH REVERSE: 5′-GTCCGGCTGGGGCGAGTAC-3′
EPH SET 1 FORWARD: 5′- GCCATTACCACTTGACTACGCCAGC-3′
EPH SET 1 REVERSE: 5′- CTCGAGCGAATTCCCTTATAGCTTGATTAGG-3′
EPH SET 2 FORWARD: 5′-AGCAACGAAGTTAATTCTATGGATACTACTCC-3′
EPH SET 3 FORWARD: 5′-AGCAACGAAGTGACGACGCC-3′
EPH SET 2 and 3 REVERSE: 5′-CACATCAATGTCCTGTACAAAGTTTGGGGG-3′
RPL32 FORWARD: 5′-GTTGTGCACCAGGAACTTCTTGAATCCG-3′
RPL32 REVERSE: 5′-CTTCCAGCTTCAAGATGACCATCCGC-3′
TDP-43 FORWARD: 5′-TGTCTTCATCCCCAAGCCATTCAGG-3′
TDP-43 REVERSE: 5′-CCAAATCCACCCTGATTCCCAAAGC-3′
β-ACTIN FORWARD: 5′-GGCATCCTCACCCTGAAGTA-3′
β-ACTIN REVERSE: 5′-AGAGGCGTACAGGGATAGCA-3′
PCR analysis was performed using the iQ SYBR Green Supermix in a myIQ Single Color RTPCR system (Bio-Rad). All runs included a standard dilution curve representing 10× to 0.01× of the RNA concentration utilized for all primer sets to insure linearity. Equivalent efficiency of individual primer sets was confirmed prior to data analysis. The levels of individual mRNAs were normalized to those of RPL32 mRNA for each sample run and expressed as a ratio of levels of GMR-Gal4 driver lines (fold control expression) unless otherwise stated. For human fibroblast lines, TDP-43 level was normalized to β-actin mRNA. All samples were run in triplicate and all data represent at least three independent experiments.
Western blotting for GFP and tubulin were performed as previously described (44). Twenty-five fly heads with each genotype from 0 to 3 day post-eclosion flies were cut, isolated and then homogenized in RIPA buffer with protease inhibitors (Roche), and passed three times through a 27-gauge needle to shear DNA. Homogenates were then centrifuged at 13 000g for 15 min. Supernatant was collected as protein lysate and quantitated with the BCA assay (Bio-Rad); 25 μg of protein lysate for each genotype was resolved with 12% SDS–PAGE, transferred to PVDF membrane, and incubated with the following antibodies: monoclonal mouse-anti-tubulin (DHSB E7, 1:5000), monoclonal mouse-anti-GFP (Roche 7.1 and 13.1, 1 : 1000). Blots then were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (Jackson ImmunoResearch) and then detected by the standard film. At least three independent experiments were performed. FMRpolyG-GFP and GFP proteins were quantitated by the ImageJ software and were normalized to tubulin level in the same sample.
Western blotting of TDP-43 and β-actin were performed as follows: ∼1 × 107 fibroblast cells at confluency were directly lysed in Laemmli buffer containing 2% 2-mercaptoethanol, and equal amount of cell lysate from each cell line were boiled for 5 min at 95°C and resolved with 12% SDS–PAGE, transferred to PVDF membrane and incubated with the following antibodies: rabbit polyclonal anti-TDP-43 (Proteintech Group, Chicago, IL, USA; 1 : 1500), mouse monoclonal anti-β-actin (Sigma, 1 : 10 000). Blots then were incubated with HRP-conjugated secondary antibodies (Jackson ImmunoResearch) and then detected by the standard film. At least three independent experiments were done and quantitation of TDP-43 was performed by the ImageJ software and normalized to level of β-actin from the same sample.
In situ hybridization, immunofluorescence and confocal microscopy
About 30 1–3 day post-eclosion fly heads from each genotype were isolated, immediately frozen in OCT media, and then cryosectioned to 10 μm. Transverse sections were then fixed with 4% paraformaldehyde in 1× PBS for 15 min, washed 3× in PBS, and permeabilized with 0.1% triton × 100 in PBS for 5 min. In situ hybridization was performed as in Mankodi et al. (78) with minor modifications. Sections were pre-hybridized with 30% formamide in 2× SSC diluted with PBS for 30 min at room temperature and then hybridized with 500 μl probe solution in a light proof box for 2 h in 37°C. The probe solution is as follows: 1 ng/ml Cy5 labelled 2′-O-Me-(CCG)6 probe, 0.02% BSA, 66 mg/ml yeast RNA (Sigma), 2 μl RNAse inhibitor (Sigma), 30% formamide in PBS-diluted 2× SSC. After hybridization, slides was washed with 1× SSC for 30 min and incubated with mouse-anti-myc primary antibody 9E10 (Sigma) diluted in 5% normal goat serum (Vector labs) in 1× PBS at 4°C overnight. Slides were then washed with PBS and incubated with Alexa 568 goat anti-mouse secondary antibody (in 5% normal goat serum in PBS) for 30 min at room temperature. After further washes, samples were dried for 5 min, incubated with 100 μl prolong gold with DAPI for 1 h, and examined on an Olympus FV1000 confocal microscope with identical laser settings for each slide. Images were overlaid in the ImageJ software.
RNA foci were quantified as follows: a 50 × 50 μm regions of interest (ROI) within fly retina were randomly selected from slides, >20 ROI were counted for each genotype, and total numbers of RNA foci were counted and quantitated as foci number per mm2. RNA foci overlapping with 30% of a DAPI-stained nucleus were considered as nuclear-associated RNA foci, and the percentile of nuclear-associated foci over total foci were calculated from the same ROI as for the total RNA foci counting. For flies expressing both CGG90-GFP and TDP-43, RNA foci overlapped with TDP-43 staining were also quantitated as to the percentile of total RNA foci.
GFP aggregates from the sample prepared above were examined directly using an Olympus FV1000 confocal microscope. No GFP antibody was utilized. 50 × 50 μm ROI within fly retina were selected on each slide, and the number of GFP aggregates was counted. A total of 20 ROI for each genotype were counted and quantitated as GFP aggregates per mm2.
Human sample immunohistochemistry
Clinical autopsy tissue from FXTAS and control patients was from previously described cases (44,79). Paraffin embedded sections through regions with known significant ubiquitin inclusion burden and matching regions in control brains were processed in parallel without antigen retrieval. Rabbit polyclonal TDP-43 (ProteinTech Group, 1 : 1000) was used as the primary antibody.
Eye phenotypes were scored on an integer (i.e. non-continuous) scale. Therefore, non-parametric statistical tests were used for all comparisons of eye phenotype. When multiple comparisons were performed, a Kruskal–Wallis one-way ANOVA was first used, followed by the Mann–Whitney U-test comparisons with a Bonferroni correction. For other statistics, standard parametric measures were used, with Student's t-test for single comparisons or a one-way ANOVA with post hoc Student's t-test for multiple comparisons. All error bars are standard error of the mean.
This work was funded by the Veterans Administration (VAMC/BLRD I21BX001841), the Nathan Shock Center (P30-AG13283) and the NIH (R01NS086810 and K08NS069809) to P.K.T. F.H. was supported by a NAF postdoctoral fellowship.
The authors thank Peng Jin for providing (CGG)90-GFP fly lines, Udai Pandey for providing FUS fly lines. TDP-43 wt and TDP-43-Δ321–366 TDP-43 constructs were a kind gift from Francisco Baralle and Emanuele Buratti. TDP-43-FFLL constructs were a kind gift from Aaron Voigt and Jorg B. Schulz. UASg-HA-AttB plasmid was a gift from Johannes Bischof and Konrad Basler. Human FXTAS tissue was a kind gift from Elan Louis and control tissue was obtained from the University of Michigan Brain Bank. Fibroblast cell lines were obtained as a kind gift from Dr Paul Hagerman and colleagues (University of California at Davis). We thank Michelle Frazer for technical assistance and Henry Paulson for feedback on the manuscript.
Conflict of Interest statement. None declared.