Learn more: PMC Disclaimer | PMC Copyright Notice
TDP-43 promotes microRNA biogenesis as a component of the Drosha and Dicer complexes
Associated Data
Abstract
Although aberrant microRNA (miRNA) expression is linked to human diseases including cancer, the mechanisms that regulate the expression of each individual miRNA remain largely unknown. TAR DNA-binding protein-43 (TDP-43) is homologous to the heterogeneous nuclear ribonucleoproteins (hnRNPs), which are involved in RNA processing, and its abnormal cellular distribution is a key feature of amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD), two neurodegenerative diseases. Here, we show that TDP-43 facilitates the production of a subset of precursor miRNAs (pre-miRNAs) by both interacting with the nuclear Drosha complex and binding directly to the relevant primary miRNAs (pri-miRNAs). Furthermore, cytoplasmic TDP-43, which interacts with the Dicer complex, promotes the processing of some of these pre-miRNAs via binding to their terminal loops. Finally, we show that involvement of TDP-43 in miRNA biogenesis is indispensable for neuronal outgrowth. These results support a previously uncharacterized role for TDP-43 in posttranscriptional regulation of miRNA expression in both the nucleus and the cytoplasm.
MicroRNAs (miRNAs), small noncoding RNAs of ∼20–22 nt, have emerged as novel regulatory factors of gene expression (1). The expression of each individual miRNA is tightly regulated in a development- and cell-specific manner through transcriptional or posttranscriptional control. As a result, miRNAs can act as regulatory switches for development, organogenesis, and cellular differentiation and control distinct functions that are required for the maintenance of different cell subtypes. Moreover, the altered expression of certain miRNAs is involved in the pathogenesis of developmental abnormalities and human diseases such as cancer and Parkinson's disease (2, 3). As a consequence, understanding the mechanisms that regulate the expression of each individual miRNA is essential to elucidate the molecular pathogenesis of human diseases.
MiRNAs are generated from long primary transcripts, termed pri-miRNAs, which consist of a short dsRNA region and a loop. The nuclear Drosha complex cleaves pri-miRNAs to release intermediate precursors that are termed pre-miRNAs. Pre-miRNAs are then transported by Exportin-5 into the cytoplasm where they are cleaved further by the Dicer complex to generate mature miRNAs. Finally, mature miRNAs are incorporated into the RNA-induced silencing complex (RISC), after which they can hybridize to the 3′-untranslated region (UTR) of their target mRNAs to repress translation or degrade these mRNAs (4). DGCR8 is a partner protein that is indispensable for the processing of pri-miRNAs by Drosha (5). In addition, recent work has identified multiple proteins that modulate the processing of specific miRNAs by interacting with the Drosha complex or by binding directly to pri-miRNAs or both (6, 7). Compared with the number of regulatory proteins that are involved in Drosha cleavage, only a few proteins have been identified that regulate the processing of pre-miRNAs by Dicer (6, 7). Among these proteins, transactivation-responsive RNA-binding protein (TRBP) is an integral component of the Dicer complex and is required not only for the cleavage of pre-miRNAs by Dicer but also for the recruitment of Argonaute 2 (Ago2), the catalytic engine of RISC, to miRNA bound by Dicer (8, 9).
Recently, mutations have been identified in the genes that encode the RNA-binding proteins TAR DNA-binding protein-43 (TDP-43) and fused in sarcoma (FUS) [also known as translocated in liposarcoma (TLS)] in patients with amyotrophic lateral sclerosis (ALS) or frontotemporal lobar degeneration (FTLD) (10). Normally, both TDP-43 and FUS are localized predominantly in the nucleus and they are homologous to heterogeneous nuclear ribonucleoproteins (hnRNPs), which regulate various aspects of RNA processing and metabolism (10). It has also been demonstrated that TDP-43 is accumulated in cytoplasmic aggregates in the affected regions of the spinal cord and brain in patients with ALS and FTLD, respectively, whether or not they carry mutations in TDP-43 (11). Similarly, FUS is a major component of cytoplasmic inclusions in patients with FUS mutations (12). However, the inclusions in these patients are not immunoreactive for TDP-43 (12), which raises the issue of whether the disease processes that are driven by FUS mutations are independent of TDP-43. The accumulation of TDP-43 in cytoplasmic aggregates results in the abnormal intracellular localization of TDP-43, for example, the clearance of TDP-43 from the nucleus (11). Thus, it is likely that the loss of normal TDP-43 function, which leads to defects or alterations in RNA processing and metabolism, plays, at least in part, a causative role in the pathogenesis of ALS and FTLD. However, the physiological functions of TDP-43, especially related to the disease pathogenesis, have not been characterized in detail (10). It has been shown previously that both TDP-43 and FUS are involved in the Drosha complex (5, 13), but the role of these proteins in miRNA processing has not been elucidated so far.
In this study, we identified TDP-43, but not FUS, as a component of nuclear Drosha complexes that contain DGCR8, which is indispensable for pri-miRNA processing. TDP-43 facilitated the binding of the Drosha complex to a subset of pri-miRNAs, which results in their efficient cleavage into pre-miRNAs. More surprisingly, cytoplasmic TDP-43 also associated with the Dicer complex that contained TRBP. This interaction facilitated the processing by Dicer of the specific pre-miRNAs, namely a subset of the pre-miRNAs whose production in the nucleus is regulated by TDP-43, via direct binding to their terminal loops. Finally, we showed that TDP-43 promotes neuronal outgrowth by facilitating miRNA production. Taken together, these findings suggest that the maturation of a subset of miRNAs is modulated at multiple steps by TDP-43, which reveals a unique function of TDP-43 not only in the nucleus but also in the cytoplasm.
Results
TDP-43 Is Involved in the Nuclear Drosha Complex Containing DGCR8.
To confirm the association between TDP-43 and the nuclear miRNA processing machinery, first we used immunoprecipitation with an anti-FLAG antibody to isolate nuclear complexes that contained FLAG-Drosha from a cell lysate of HEK293T-derived cell lines that stably expressed FLAG-Drosha (5). Western blot analysis demonstrated that FLAG-Drosha interacted with TDP-43, but not with LSD1 and p84, in addition to DGCR8, p68, and FUS as previously described (Fig. 1A) (5, 13). The involvement of TDP-43 in Drosha complexes was also confirmed by immunoprecipitation of endogenous Drosha complexes (SI Appendix, Fig. S1). To analyze this association reciprocally, we established HEK293T-derived cell lines that stably expressed FLAG-TDP-43 at a similar level to that of endogenous TDP-43 (SI Appendix, Fig. S2). After isolation of nuclear complexes that contained FLAG-TDP-43, we analyzed them by Western blot analysis. The results demonstrated that FLAG-TDP-43 interacted with Drosha, DGCR8, and Exportin-5, in addition to other previously reported proteins such as endogenous TDP-43, p68, and FUS (Fig. 1B) (14–16). Therefore, we confirmed the association between TDP-43 and the Drosha complex in a bidirectional manner. To investigate the involvement of TDP-43 in the Drosha complex that contains DGCR8, which is required for pri-miRNA processing, we analyzed alternate fractions from gel-filtration chromatography of the complexes that contained FLAG-Drosha. The results reproduced those reported previously (5) and demonstrated that there were two major complexes, a large complex that did not contain DGCR8, which peaked in fraction 17, and a small complex, which peaked in fractions 27–35 (Fig. 1C). In the small complex, we found that whereas DGCR8 was present in fractions 25–33, Drosha and TDP-43 were highly concentrated in fractions 31–33. In contrast, FUS was identified predominantly in the large complex and was detected faintly, which did not contain DGCR8 (Fig. 1C), suggesting that FUS is not involved in miRNA processing. Considering that TDP-43 associated with DGCR8 (Fig. 1B), we concluded that TDP-43 was involved in the nuclear Drosha complex that contained DGCR8, which is indispensable for pri-miRNA processing.
Association between TDP-43 and the Drosha complex. (A and B) Nuclear extracts (NE) from HEK293T-derived stable cell lines that expressed FLAG-Drosha (A) or FLAG-TDP-43 (B) and from nontransfected HEK293T cells were subjected to immunoprecipitation (IP) with anti-FLAG antibody followed by immunoblot (IB) analysis. Samples that corresponded to 5% of the NE used in the assay were loaded as Input. (A) FLAG-Drosha, whose expression was approximately fivefold higher than that of endogenous Drosha, was represented by two forms (band a and band b) as previously described (5). (B) Band a, full-length FLAG-TDP-43; band b, endogenous TDP-43; band c, cleaved fragment of FLAG-TDP-43. (C) Complexes that contained FLAG-Drosha were subjected to Superose 6 size-exclusion chromatography and analyzed by immunoblotting (IB). The fractions from the column are indicated at the bottom. (D) Immunoprecipitates of nuclear complexes that contained FLAG-Drosha were digested with the indicated amount of either RNase V1 or RNase A (0.1–10 units), separated into supernatant (Sup.) and bead (Beads) fractions, and subjected to IB analysis. (E) Schematic diagram of wild-type TDP-43 tagged with Myc at the N terminus and the regions deleted in each construct. The regions that were indispensable for the association with the Drosha complex (amino acids 316–402) and the Dicer complex (amino acids 261–402) are indicated in red. Gly-rich, glycine-rich domain; A2B, hnRNP A2-binding domain; M9-like, M9-like domain; NES, nuclear export signal; NLS, nuclear localization signal; RRM, RNA recognition motif (32). (F) A wide range of deletion mutants of Myc-TDP-43, which are indicated at the top, were transfected into HEK293T-derived stable cell lines that expressed FLAG-Drosha. Immunoprecipitates of nuclear complexes that contained FLAG-Drosha were subjected to IB analysis.
TDP-43 Associates with the Drosha Complex in an RNA-Dependent and -Independent Manner.
To elucidate the molecular interaction between TDP-43 and the Drosha complex, the immunoprecipitates of the nuclear complexes that contained FLAG-Drosha were treated with various amounts of RNase V1 or RNase A or both. The RNase treatment decreased the association between TDP-43 and the Drosha complex in a dose-dependent manner, but most of the TDP-43 remained bound even in the presence of a 10-fold excess of RNase (Fig. 1D and SI Appendix, Fig. S3), which indicated that TDP-43 interacted with the Drosha complex only partially through certain RNA species and mainly through a protein–protein interaction.
To determine the regions in TDP-43 that are responsible for the association with the Drosha complex, we created a wide range of deletion mutants of TDP-43 tagged with Myc (Fig. 1E and SI Appendix, Fig. S4). The expression constructs for the TDP-43 mutants were then transfected into HEK293T-derived stable cell lines that expressed FLAG-Drosha and nuclear complexes that contained FLAG-Drosha were isolated. Western blot analysis demonstrated that the association of TDP-43 with the Drosha complex was dramatically reduced when the region in the C terminus of TDP-43 that corresponded to amino acids 316–401 was deleted (Fig. 1 E and F). The RNA-binding domain (amino acids 105–174) and glycine-rich domain (amino acids 274–314) were dispensable for the association (Fig. 1F). An in vitro pull-down assay demonstrated that the region corresponding to amino acids 316–401 was sufficient for the association of TDP-43 with the Drosha complex (SI Appendix, Fig. S5). Although the Drosha complex contains multiple proteins that might mediate the association of TDP-43 with the complex, DGCR8, at least, was dispensable for this association (SI Appendix, Fig. S6). Taken together, the results demonstrate that TDP-43 associates with the Drosha complex through certain RNA species and through a region of the C-terminal tail.
Identification of miRNAs Whose Expression Is Altered by TDP-43 Knockdown.
Given that TDP-43 was not associated with all of the Drosha complexes that contained DGCR8, next we investigated how many miRNAs are regulated by TDP-43. We chose three different cell lines, namely SH-SY5Y, HeLa, and Neuro2a cells, for TDP-43 knockdown experiments to identify TDP-43–regulated miRNAs that are conserved among mammals and to ensure that we did not overlook any cell-specific miRNAs. We also used two different siRNAs against TDP-43 (siRNA1 for SH-SY5Y and HeLa cells and siRNA2 for Neuro2a cells) to exclude the possibility of an off-target effect. Seventy-two hours after the siRNAs were transfected, TDP-43 had almost entirely disappeared (SI Appendix, Fig. S7). Therefore, we performed miRNA microarray analysis using the RNA extracted at this time. Among the 847 human miRNAs analyzed, TDP-43 depletion significantly reduced (>1.5-fold) the expression of 11 and 12 miRNAs in SH-SY5Y and HeLa cells, respectively (SI Appendix, Fig. S8 A and B), whereas 12 miRNAs among the 609 mouse miRNAs analyzed were significantly down-regulated in Neuro2a cells (SI Appendix, Fig. S8C). Among these candidate miRNAs, ∼10 miRNAs, which included miR-132-5p and miR-335-5p, were down-regulated regardless of the cell line and regardless of species, although some of the miRNAs, such as miR-143-3p and miR-145-5p, were expressed in a cell-specific manner (SI Appendix, Fig. S8D). To validate the results obtained by the global survey of microarray data, we then performed quantitative RT-PCR (qRT-PCR). For most of the candidate miRNAs that were identified on the basis of decreased expression, the partner miRNA strand (for example, the partner strand of miR-143-3p is miR-143-5p) could not be quantified using the microarray most likely due to low expression. Therefore, we also included the partner miRNA strands in the validation by qRT-PCR. We verified that TDP-43 knockdown caused a significant reduction in the expression level of miR-558-3p and miR-574-3p and both strands of miR-132 and miR-143 in SH-SY5Y and HeLa cells (Fig. 2A and SI Appendix, Fig. S9A). The same set of miRNAs, except for human-specific miR-558-3p, was also down-regulated in Neuro2a cells (SI Appendix, Fig. S9B). Northern blot analysis demonstrated that the production of pre-miR-132 and pre-miR-143 as well as corresponding mature miRNAs was significantly decreased (SI Appendix, Fig. S10 A–G). In contrast, a reduction in the expression levels of pri-miRNAs was not observed after the depletion of TDP-43 in SH-SY5Y and HeLa cells (SI Appendix, Fig. S11 A and B), which suggested that the reduced levels of mature miRNA were not caused by a reduction in transcription. Taken together, the results demonstrate that knockdown of TDP-43 caused a reduction in the expression level of at least six miRNAs, i.e., miR-558-3p, miR-574-3p, and both strands of miR-132 and miR-143, and this reduction was most likely due to the impairment of posttranscriptional regulation.
Identification of TDP-43–regulated miRNAs. (A) The expression levels of the mature forms of the indicated miRNAs in SH-SY5Y cells transfected with either TDP-43 siRNA or control siRNA are shown (n = 5). (B) An RIP assay was performed in SH-SY5Y cells whereby RNA fragments were immunoprecipitated with anti–TDP-43 antibody or control IgG and subsequently subjected to quantitative RT-PCR analysis for the indicated pri-miRNAs (n = 3). (C) Nuclear extracts (NE) and cytoplasmic extracts (CE) from HEK293T-derived stable cell lines that expressed FLAG-TDP-43 (T) and from nontransfected HEK293T cells (C) were subjected to immunoprecipitation (IP) with anti-FLAG antibody, after which the RNA fragments were purified. RT-PCR amplification was performed using RT primers specific for the pri-miRNAs (lanes 2–7) or that recognized both pri-miRNAs and pre-miRNAs (lanes 8–13). EMSA was performed with the indicated amounts of recTDP-43 and radiolabeled wild-type (WT) pre-miR-574 (D and E) or mutant pre-miR-574 RNAs (MUT1 and MUT2) (E). The percentage amounts of RNA that bound to the indicated amounts of recTDP-43 protein were then quantified (n = 3). (D) (Upper) The sequence of human pre-miR-574 is shown. The regions known to be processed into mature miR-574-5p and miR-574-3p are shown in green. (E) (Upper) The expected hairpin structures of MUT1 and MUT2 are shown, and the mutated sites are highlighted in red.
TDP-43 Binds to Selected Pri-miRNAs in the Nucleus.
To investigate whether the biogenesis of the subset of miRNAs that was down-regulated after TDP-43 knockdown is directly controlled by TDP-43, an RNA-immunoprecipitation (RIP) assay with anti–TDP-43 antibody was performed using lysate derived from SH-SY5Y cells. This analysis showed that TDP-43 bound directly to pri-miR-132, pri-miR-143, pri-miR-558, and pri-miR-574, but not to pri-miR-17, pri-miR-423, and pri-miR-766 (Fig. 2B). Next, we purified RNAs that were bound to the nuclear FLAG-TDP-43 and subjected them to RT-PCR, which confirmed that TDP-43 bound directly to the same set of pri-miRNAs (Fig. 2C, lanes 3 and 9). To confirm the direct binding of TDP-43 to these pri-miRNAs further, we performed an electrophoretic mobility shift assay (EMSA) with various amounts of recombinant TDP-43 (recTDP-43) in vitro. GST-recTDP-43 was expressed in Escherichia coli and purified, after which the GST tag was cleaved off to yield recTDP-43 (SI Appendix, Fig. S12A). Consistent with a previous report (17), recTDP-43 bound to an RNA that consisted of 12 UG repeats (12-UG–repeat RNA) with high affinity (SI Appendix, Fig. S12B). Moreover, recTDP-43 bound efficiently to a portion of the terminal exon of the SC35 gene, termed SC35 IIB, which does not contain canonical UG repeats but was recently shown to bind to TDP-43 (SI Appendix, Fig. S12C) (18). These findings confirmed that purified recTDP-43 retained the ability to bind RNA. Therefore, we performed EMSA with 32P-labeled pri-miRNAs and various amounts of recTDP-43. Consistent with the in vivo TDP-43–binding results (Fig. 2 B and C), pri-miR-558, pri-miR-132, pri-miR-143, and pri-miR-574, but not pri-miR-423 (SI Appendix, Fig. S8D), bound efficiently to recTDP-43. Taken together, the results show that TDP-43 binds selectively and directly to at least four pri-miRNAs in the nucleus.
Cytoplasmic TDP-43 Binds to the Terminal Loops of the Pre-miRNAs.
Given that TDP-43 shuttles continuously between the nucleus and the cytoplasm (19), we examined the possible involvement of TDP-43 in miRNA processing in the cytoplasm. First, we extracted the RNAs that were bound to the cytoplasmic complexes that contained FLAG-TDP-43. As expected, when RT-PCR was performed with RT primers specific to the pri-miRNAs, no PCR products were amplified (Fig. 2C, lane 6). In contrast, RT primers designed to hybridize to both the primary and precursor forms of each miRNA could amplify sequences that corresponded to pre-miR-143 and pre-miR-574, but not to pre-miR-132, pre-miR-558, and pre-miR-423 (Fig. 2C, lane 12), which suggested that pre-miR-143 and pre-miR-574 bound selectively to TDP-43 in the cytoplasm. Consistent with the in vivo results, EMSA with 32P-labeled pre-miRNAs demonstrated that recTDP-43 did not bind to pre-miR-132, pre-miR-558, or pre-miR-423 (SI Appendix, Fig. S13 A–C). These findings suggested that TDP-43 bound to a pyrimidine-rich sequence and a UG-rich sequence, both of which have been reported recently to be preferential binding targets of TDP-43 (20–22), in the flanking region of pri-miR-132 and pri-miR-558, respectively (SI Appendix, Fig. S12D). In contrast, EMSA demonstrated that recTDP-43 bound to pre-miR-574 (Fig. 2D), which is in contrast to a previous report, which demonstrated that pre-miR-574 did not bind to GST-TDP-43 in vitro (23). To investigate the site within pre-miR-574 that is recognized by TDP-43, first we mutated two nucleotides within the terminal loop to extend the dsRNA region (Fig. 2E, MUT1). Although recTDP-43 bound more efficiently to miR-574-5p RNA that contained the UG-rich sequence than to the 12-UG–repeat RNA (SI Appendix, Fig. S12B), significant binding of TDP-43 to MUT1 RNA was not observed (Fig. 2E). Therefore, next we mutated a UGUGUC sequence in the terminal loop to UGCGCC, which did not affect the hairpin structure (Fig. 2E, MUT2). Once again, no significant binding of TDP-43 to MUT2 RNA was observed (Fig. 2E). These results suggested that the sequence within the terminal loop, but not the dsRNA region, is critical for the recognition of pre-miR-574 by TDP-43. We confirmed that TDP-43 also bound to the terminal loop, but not the sequences in the dsRNA region of pre-miR-143 (SI Appendix, Fig. S14). Taken together, the results indicate that cytoplasmic TDP-43 binds to pre-miR-143 and pre-miR-574 through their terminal loops.
TDP-43 Is Involved in the Cytoplasmic Dicer Complex Containing TRBP.
The selective binding of cytoplasmic TDP-43 to a subset of pre-miRNAs allowed us to investigate the possible interaction between TDP-43 and the Dicer complex that contains TRBP, which is required for miRNA processing in the cytoplasm. First, we isolated cytoplasmic complexes that contained FLAG-Dicer from lysates from HEK293T-derived stable cell lines expressing FLAG-Dicer (8, 9). Western blot analysis demonstrated that FLAG-Dicer interacted with TDP-43 in addition to TRBP and Ago2, whereas no association was observed between FUS and Dicer (Fig. 3A). The involvement of TDP-43 in Dicer complexes was also confirmed by immunoprecipitation of endogenous Dicer complexes (SI Appendix, Fig. S15). To confirm the association between TDP-43 and the Dicer complex reciprocally, we isolated cytoplasmic complexes that contained FLAG-TDP-43 from lysates from HEK293T-derived stable cell lines expressing FLAG-TDP-43. Western blot analysis demonstrated that Dicer, Ago2, and FUS were cytoplasmic TDP-43–interacting proteins, whereas no association was observed between TDP-43 and TRBP, Exportin-5, or Optineurin, which was recently reported to be accumulated in TDP-43–positive intracytoplasmic inclusions in patients with sporadic ALS (Fig. 3B) (24). Taken together, we confirmed the association between TDP-43 and the Dicer complex in a bidirectional manner. To investigate the involvement of TDP-43 in the cytoplasmic miRNA processing machinery, we used Western blotting to analyze serial fractions from gel-filtration chromatography of the complexes that contained FLAG-Dicer. Consistent with previous reports (8, 9), this analysis resulted in the coelution of a fraction of FLAG-Dicer with TRBP and Ago2 as components of a complex of ∼500 kDa (Fig. 3C). Whereas Dicer was highly enriched in fractions 30–32, TRBP and TDP-43 were concentrated in fractions 29–31 and 31–33, respectively (Fig. 3C), which suggested that Dicer complexes containing TRBP, which is indispensable for pre-miRNA processing, could be further divided into at least two populations, i.e., complexes that lacked or contained TDP-43.
Association between TDP-43 and the Dicer complex. (A and B) Cytoplasmic extracts (CE) from HEK293T-derived stable cell lines that expressed FLAG-Dicer (A) or FLAG-TDP-43 (B) and from nontransfected HEK293T cells were subjected to immunoprecipitation (IP) with anti-FLAG antibody followed by immunoblot (IB) analysis. Samples that corresponded to 5% of the CE used in the assay were loaded as Input. (A) The expression of FLAG-Dicer was approximately fourfold higher than that of endogenous Dicer. (B) Band a, full-length FLAG-TDP-43; band b, endogenous TDP-43; band c, cleaved fragment of FLAG-TDP-43. (C) Complexes that contained FLAG-Dicer were subjected to Superose 6 size-exclusion chromatography and analyzed by immunoblotting (IB). The fractions from the column are indicated at the bottom. (D) Immunoprecipitates of cytoplasmic complexes that contained FLAG-Dicer were digested with the indicated amounts of either RNase V1 or RNase A (0.1–10 units), separated into supernatant (Sup.) and bead (Beads) fractions, and subjected to IB analysis. (E) A wide range of deletion mutants of TDP-43 were transfected into HEK293T-derived stable cell lines that expressed FLAG-Dicer. Immunoprecipitates of cytoplasmic complexes that contained FLAG-Dicer were subjected to IB analysis. The deleted regions in each Myc-TDP-43 mutant are indicated at the top.
TDP-43 Associates with the Dicer Complex in an RNA-Dependent and -Independent Manner.
To elucidate the molecular interaction between TDP-43 and the Dicer complex, the immunoprecipitates of the cytoplasmic complexes that contained FLAG-Dicer were treated with various amounts of RNase V1 or RNase A or both. The RNase treatment decreased the association in a dose-dependent manner, but approximately half of the TDP-43 remained bound to the complex even in the presence of a 10-fold excess of RNase (Fig. 3D and SI Appendix, Fig. S16), which indicated that TDP-43 associated with the Dicer complex not only through certain RNA species but also through a protein–protein interaction. To determine the regions in TDP-43 that were responsible for the association with the Dicer complex, a wide range of deletion mutants of TDP-43 (Fig. 1E) were transfected into HEK293T-derived stable cell lines that expressed FLAG-Dicer, and then the cytoplasmic complexes that contained FLAG-Dicer were isolated. Western blot analysis demonstrated that the association was dramatically reduced when the region in the C terminus of TDP-43 that corresponded to amino acids 261–401 was deleted (Fig. 3E). This region is longer than that required for the association with the Drosha complex (Fig. 1E). Deletion of the RRM1 domain (amino acids 105–174) slightly increased the amount of association, which suggested that the RNA-binding domains, at least, were dispensable for the association (Fig. 3E). An in vitro pull-down assay demonstrated that the region corresponding to amino acids 261–401 was sufficient for the association with the Dicer complex (SI Appendix, Fig. S17). Taken together, the results demonstrate that TDP-43 interacts with the Dicer complex through certain RNA species and through a region of the C-terminal tail.
Nuclear TDP-43 Promotes the Cleavage of Selected Pri-miRNAs by Drosha.
To examine the effect of the binding of TDP-43 to pri-miRNAs on the production of pre-miRNAs, we prepared Drosha complexes that lacked or contained TDP-43 (Fig. 4A). Selective depletion of TDP-43 did not affect the amount of DGCR8 in the complex (Fig. 4A). Then, we performed EMSA to compare the binding affinity of these two complexes for pri-miRNAs. The Drosha complex that lacked TDP-43 bound to all of the pri-miRNAs examined, but the binding affinity of the complex for pri-miR-132, pri-miR-143, pri-miR-558, and pri-miR-574 was significantly reduced in the absence of TDP-43 (SI Appendix, Fig. S18 A–F). In contrast, the affinity of the complex for pri-miR-423 was not altered by TDP-43 depletion (SI Appendix, Fig. S18 E and F), which suggested that TDP-43 promoted efficient binding of the Drosha complex to selected pri-miRNAs. Next, we performed an in vitro miRNA processing assay, which showed that the reduction in binding affinity caused by the depletion of TDP-43 proportionally reduced the production of the relevant pre-miRNAs (Fig. 4B and SI Appendix, Fig. S19 A–E). In addition, a pri-miRNA processing assay using the gel-filtration fractions revealed that the TDP-43–enriched fraction (fraction 33) had higher processing activity for pri-miR-558, but not for pri-miR-423, than fraction 27, which contained less TDP-43 (SI Appendix, Fig. S20 A–C). Overall, our data clearly demonstrated that involvement of TDP-43 in the Drosha complex facilitates the binding of the complex to at least four pri-miRNAs, which results in their efficient cleavage into pre-miRNAs.
Analysis of the miRNA processing activity with or without TDP-43. HEK293T-derived stable cell lines that expressed either FLAG-Drosha or FLAG-Dicer were transfected with TDP-43 siRNA1 or control siRNA. (A and C) Subsequently, nuclear extracts (NE) (A) or cytoplasmic extracts (CE) (C) were subjected to immunoprecipitation (IP) with anti-FLAG antibody followed by immunoblot (IB) analysis. (B) The relative cleavage efficiency for Drosha complex that lacked TDP-43 against Drosha complex that contained TDP-43 is plotted for each pri-miRNA (n = 3). (D) The relative cleavage efficiency for the Dicer complex that lacked TDP-43 against the Dicer complex that contained TDP-43 is plotted for each pre-miRNA (n = 3). (E) Each siRNA-resistant Myc-tagged TDP-43 construct shown at the top was cotransfected with TDP-43 siRNA1 into SH-SY5Y cells. The expression of the Myc-tagged TDP-43 proteins and the knockdown of endogenous TDP-43 were confirmed by IB analysis. (F) Total RNA was extracted from SH-SY5Y cells transfected with TDP-43 siRNA1 and each siRNA-resistant Myc-tagged TDP-43 construct and was subjected to quantitative RT-PCR analysis for the indicated miRNAs (n = 5).
Cytoplasmic TDP-43 Promotes the Cleavage of the Specific Pre-miRNAs by Dicer.
We examined whether the involvement of TDP-43 in the Dicer complex also affects miRNA maturation by using Dicer complexes that lacked or contained TDP-43 (Fig. 4C). Selective depletion of TDP-43 did not affect the amount of TRBP in the complex (Fig. 4C). EMSA showed that the Dicer complex bound to all of the pre-miRNAs examined regardless of the presence of TDP-43, but the affinity of the complex for pre-miR-143 and pre-miR-574 was significantly reduced in the absence of TDP-43 (SI Appendix, Fig. S21 A–E). In contrast, the affinity of the complex for pre-miR-132 and pre-miR-423 was not altered by TDP-43 depletion (SI Appendix, Fig. S21 A–E), which suggested that the direct binding of TDP-43 through the terminal loop promoted the efficient binding of the Dicer complex to selected pre-miRNAs. In accordance with this, the conversion of pre-miR-143 and pre-miR-574 to the respective mature miRNAs was significantly reduced when TDP-43 was depleted from the Dicer complex (Fig. 4D and SI Appendix, Fig. S22 A–D). Taken together, these results suggested that cytoplasmic TDP-43 facilitates the binding of the Dicer complex to a subset of pre-miRNAs and thereby promotes the production of mature miRNAs. Finally, we transfected siRNA-resistant TDP-43 constructs together with TDP-43 siRNA1 into SH-SY5Y cells (Fig. 4E). The expression of wild-type TDP-43, but not the Δ357–414 deletion mutant, could fully restore the miRNA processing activities (Fig. 4F), which indicated that the association of TDP-43 with the Drosha and Dicer complexes through its C terminus is indispensable for the efficient processing of the subset of miRNAs in vivo.
TDP-43 Promotes Neuronal Outgrowth Through the Regulation of miRNA Processing.
To address the biological importance of TDP-43 involvement in processing of the subset of miRNAs, we knocked down TDP-43 in Neuro2a cells and induced differentiation. Consistent with a previous report (25), TDP-43 depletion significantly attenuated neurite outgrowth (Fig. 5 A and B). Therefore, next we performed qRT-PCR, which demonstrated that TDP-43–regulated miRNA expression, i.e., miR-132-3p, miR-132-5p, and miR-143-3p, was up-regulated in addition to non-TDP-43–regulated miRNA expression, i.e., miR-423-3p during differentiation (SI Appendix, Fig. S23 A–D). This up-regulation of miR-423-3p expression was not affected by TDP-43 depletion, whereas that of the TDP-43–regulated miRNAs was significantly inhibited (SI Appendix, Fig. S23 A–D). Considering that miR-132-3p is known to promote neuronal outgrowth in vitro and in vivo (26, 27), we examined whether cotransfection of miRNAs could rescue the inhibited neurite outgrowth induced by TDP-43 depletion (Fig. 5 C–E). The overexpression of pri-miR-132 per se had no significant ability to promote neurite outgrowth further in cells transfected with control siRNA (Fig. 5E). In contrast, the overexpression of pri-miR-132, but not pri-miR-143 or pri-miR-423, partially but significantly restored neurite outgrowth in cells transfected with the TDP-43 siRNA (Fig. 5 C–E). Taken together, these results suggested that TDP-43 regulates neuronal outgrowth by facilitating miRNA production at least in part.
Requirement of TDP-43–regulated miRNAs for neuronal differentiation. (A) Images of Neuro2a cells 48 h after differentiation that was induced 24 h after transfection with either TDP-43 siRNA or control siRNA. (B) The average length of the longest neurites 48 h after differentiation is plotted (n = 101). (C and D) Differential interference-contrast (DIC) images of Neuro2a cells (Left) 48 h after differentiation that was induced 24 h after cotransfection of TDP-43 siRNA with either pmR-ZsGreen1-pri-miR-132 (C) or pmR-ZsGreen1-pri-miR-143 (D). Plasmid transfection was ensured by ZsGreen1 expression (Right). (E) The average length of the longest neurites 48 h after differentiation is plotted (n = 101). (Scale bars, 20 μm.)
Discussion
Although a number of roles have been identified for TDP-43 in RNA metabolism and processing in the nucleus (20–22, 28, 29), the distinct biological function of TDP-43 in the cytoplasm has not been identified so far. In this study, we demonstrated that TDP-43 facilitates the posttranscriptional processing of a subset of miRNAs not only in the nucleus but also in the cytoplasm. This sequential facilitation was achieved both by the direct binding of TDP-43 to the primary and precursor forms of the miRNAs and by a protein–protein interaction between TDP-43 and the nuclear Drosha and cytoplasmic Dicer complexes that might be mediated by other associated proteins.
A recent study using deep sequencing identified RNA targets of TDP-43 and revealed two preferential binding sites for TDP-43, namely UG- and pyrimidine-rich sequences (20–22, 28). In contrast, another study identified TDP-43 mRNA per se as a target of TDP-43 (29). Although the potential binding site in the 3′-UTR of TDP-43 mRNA contains no canonical UG repeats, TDP-43 recognizes this site with relatively low affinity in vitro (29). Furthermore, it was reported that UG repeats are neither necessary nor sufficient to specify a TDP-43 binding site (20). In this study, we demonstrated that TDP-43 binds in vitro to a portion of the terminal exon of the SC35 gene and the terminal loop of pre-miR-143, neither of which contains UG repeats. More importantly, we showed that TDP-43 could not bind to UG repeats in the dsRNA region of pre-miR-574, which indicated that the secondary structure also plays a critical role in the determination of the targets of TDP-43.
In this study, we demonstrated that processing of certain miRNAs is under the posttranscriptional control of TDP-43. Unfortunately, we could not detect the reduced expression of let-7b, which has recently been reported as the only microRNA that was down-regulated after TDP-43 knockdown in human Hep-3B cells (23). Although this can be attributed to the difference in the expression levels of each miRNA among the cell lines examined, it also remains to be elucidated how TDP-43 affects the biogenesis of let-7b. In contrast, recent deep sequencing analysis of RNAs bound to TDP-43 mapped certain sequencing outputs to genomic regions where a number of known miRNAs are annotated, including a UG-rich sequence in the flanking region of pri-miR-558, as shown here (21, 28). Although the biological consequences of the binding of TDP-43 to most of these pri-miRNAs remain to be determined, further studies should reveal more miRNAs whose expression is regulated by TDP-43.
Certain physiological functions of TDP-43 are lost in regions of the spinal cord and brain that are affected in ALS and FTLD, respectively, but it remains unclear which functions of TDP-43 are associated with the pathogenesis of ALS and FTLD. However, mislocalization of TDP-43 due to accumulation in cytoplasmic aggregates probably reduces the processing of TDP-43–regulated miRNAs by Drosha and Dicer. The knockout of Dicer specifically in postmitotic postnatal motor neurons in mice induces locomotor dysfuction due to the presence of a reduced number of motor neurons (30), indicating that a subset of miRNAs is essential for long-term survival of spinal motor neurons. Among the TDP-43–regulated miRNAs identified in this study, miR-132-3p is highly enriched in neurons and promotes neuronal outgrowth in vitro and in vivo by reducing the levels of the GTPase-activating protein, p250GAP (26, 27). In this study, we demonstrated that the attenuation of neuronal outgrowth induced by TDP-43 depletion in differentiated Neruo2a cells can be attributed, at least in part, to the reduced expression of miR-132-3p and miR-132-5p induced by the loss of nuclear TDP-43 function. The identification of TDP-43–regulated miRNAs that are highly expressed specifically in motor neurons and the further elucidation of the significance of each individual miRNA, including miR-132, in motor neuron survival will contribute to our understanding of the pathogenesis of ALS and FTLD and the development of novel therapeutic strategies for the treatment of both diseases.
Methods
The immunoprecipitation assay, RNA-binding assay, and in vitro miRNA processing assay were carried out as previously described (5, 31). See SI Appendix, SI Methods for details. All of the statistical analyses were carried out using the Mann–Whitney U test. All values are displayed as the mean ± SEM. Statistical significance is displayed as *P < 0.05, **P < 0.01, or ***P < 0.001.
Acknowledgments
We thank Professors K. Nishikura and R. Shiekhatter for providing many cell lines. We also thank all members of the Y. Yoneda laboratory for sharing reagents. This work was supported by the Osaka University Life Science Young Independent Researcher Support Program through the Special Coordination Program to Disseminate Tenure Tracking System from Ministry of Education, Culture, Sports, Science and Technology-Japan (MEXT); by Grants-in-Aid for Scientific Research (B) and for Scientific Research on Innovative Areas “RNA regulation” (20112006) from MEXT; and by grants from The Sumitomo Foundation, The Tokyo Biochemical Research Foundation, Nagao Memorial Fund, and Takeda Science Foundation (to Y.K.).
Footnotes
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1112427109/-/DCSupplemental.