circGlis3 is elevated in the islets of obese mouse models
To identify circRNAs potentially contributing to the development of obesity-associated disturbances in β-cell dysfunction and apoptosis, we performed global circRNAs expression profiling in pancreatic islets obtained from islets of two mouse models of obesity: HFD-fed mice compared to normal chow diet (NCD) fed mice and mice homozygous for the obesity ob mutation of the leptin (Lepob/ob) compared to wild type littermates. Theses obese mouse models are capable of long-term compensatory insulin hypersecretion and β-cell mass 13. The body weight, blood glucose and insulin levels of these mice were listed in Figure S1A-F. Out of 3646 circRNA-specific probe sets, 2738 and 2534 circRNAs were detected in islets of NCD/HFD and control/Lepob/ob, respectively (Figure S1G and H). In the islets of HFD-fed mice, expression of 302 circRNAs was significantly altered compared to circRNAs in NCD controls, of which expressions of 130 circRNAs increased and those of 172 circRNAs decreased (selection criteria: Log2 fold change >2 or <-2, p < 0.05; Figure S1G, Table S1). In Lepob/ob islets, expressions of 767 circRNAs were significantly changed, of which expressions of 413 circRNAs increased and expressions of 354 decreased (selection criteria: Log2 fold change >2 or <-2, p < 0.05; Figure S1H, Table S2). Amongst the differentially expressed circRNAs, 45.2% overlapped with recently published transcripts by RNA-seq (Table S3) 14.
The changes observed in the circRNA levels were confirmed by qRT-PCR analysis of the 20 most dysregulated circRNAs, including 10 up-regulated and 10 down-regulated ones. Nearly 90% of the results of qRT-PCR were consistent with those of RNA-seq, indicating that the results of the RNA-seq were credible (Figure S1I and J). We chose the circRNA, mmu_circRNA006170 (circBase ID: mmu_circ_0000943, termed circGlis3 in the subsequent study because its host gene is Glis3) for further study. The reasons were as follows: (1) circGlis3 was one of the most abundant circRNAs of the differentially expressed ones according to its Reads Mean in our RNA-seq (Table S4). (2) circGlis3 was enriched in pancreas, and was also detectable in spleen, lung and kidney where the expression was more than 10 times lower compared to that in the pancreas (Figure 1A). (3) circGlis3 was significantly upregulated in pancreas of HFD-fed mice, although the mGlis3 (mRNA of Glis3) and protein levels of GLIS3 were downregulated (Figure 1B and C) as previously reported 15. This finding indicated that the higher expression of circGlis3 in obesity was not simply a by-product of splicing and was suggestive of functionality.
Then, we measured the expression of circGlis3 in the islets of Lepob/ob mice aged from 4 to 12 weeks, and observed the increase of circGlis3 expression at the beginning of 6-week of age with the onset of insulin resistance (Figure 1D). We also observed a similar increase of expression of circGlis3 in the islets of mice treated with 15-week HFD and young Leprdb/db mice, while the circGlis3 expression was significantly decreased accompanied by glucose raising (Figure 1E and F), all of which showed that this observation is not limited to one mouse model of obesity and insulin resistance. Expressive abundance of circGlis3 was also increased in other tissues of obese mice compared to normal mice, such as in the kidney, liver, lung, and brain (Figure S1K and L), although to a lesser extent than that observed in the islets. We also compared circGlis3 expression in primary islet versus exocrine glands isolated from normal mice, revealing that circGlis3 expression was 4.5-fold higher in islets versus exocrine glands, indicating that islets represent the main source of circGlis3 expression in pancreas (Figure 1G). Moreover, we quantified circGlis3 expression in the sera of a cohort of human individuals, the results shown that human individuals with impaired glucose tolerance performed higher circGlis3 levels generally (Figure 1H). Interestingly, circGlis3 levels were significantly increased in human serum from obese and moderately diabetic individuals with compensated β-cell function, but the circGlis3 expression decreased in decompensated stage (Figure 1I, the concrete data were listed in Table S6). Taken together, the expression of circGlis3 in the islets was increased in dietary and genetic mouse models of obesity, as well as in the sera of overweight with insulin resistance human subjects.
Splicing factor QKI regulates formation of circGlis3
Mouse circGlis3 derived from Exon3 (Chr19:28530873-28531986) of the Glis3 gene (Gene ID: 226075; ENSMUSG00000052942) with a length of 1114 nt on chromosome 19 (Figure 2A). The sequence is consistent with circBase database annotation (http://www.circbase.org/). Sequencing analysis confirmed its back-spiced junctions (Figure 2B). There are more than 80% homologous between human circGlis3 and mouse circGlis3 (Fig S2A). Divergent primers and convergent primers for circGlis3 and linear transcript were designed. The cDNA and genomic DNA (derived from MIN6 cells) was amplified and analyzed using agarose gel electrophoresis (Figure 2C). We then investigated the stability and localization of circGlis3 in MIN6 cells. Resistance to digestion with RNase R exonuclease confirmed that this RNA specie is circular inform (Figure 2D). qRT–PCR analysis of nuclear and cytoplasmic circGlis3 and fluorescence in situ hybridization (FISH) against circGlis3 demonstrated that the circular form of Glis3 Exon3 preferentially localized in the cytoplasm (Figure 2E and F).
We next investigated the mechanism by which circGlis3 is formed. Previous studies have shown that splicing factors may participate in regulating circRNA biogenesis 16. We postulated that if a protein factor contributes to circGlis3 biogenesis, it would also be regulated by obesity. Based on this hypothesis, we investigated the expression level of a candidate panel of splicing factors in the islets of HFD-fed mice. We found that the QKI increased in obese mice at mRNA and protein levels (Figure 2G and H), and its expression level was positively correlated with circGlis3 (Figure 2I, the concrete data were listed in Table S6). We also got the same pattern in glucose and palmitate stimulated MIN6 cells (Figure S2B and C). The effect of ectopic expression of QKI on circGlis3 formation was confirmed by qRT–PCR analysis (Figure 2J), the transfection efficiency as shown in Figure 2SD. QKI usually works as a dimer, capable of binding two well separated regions of a single RNA molecule and promotes the circle-forming exons into close proximity and circRNA biogenesis 17. To estimate whether QKI binding sites in the Introns flanking the circRNA-forming Exons of Glis3, we referenced the experimental scheme from Conn SJ 18, and searched for sequences that match potential QKI response elements (QREs) in the vicinity of the QKI RIP-enriched regions. We obtained four instances of a bipartite motif that contains the sequence UAAY in conjunction with a relaxed version of the canonical QKI hexamer 17 as previously reported 19. There are two putative elements that located on upstream and two are located on downstream of the circRNA-forming splice sites, as shown in Figure 2K. To assess whether QKI binds the Glis3 pre-mRNA, we performed RNA-immunoprecipitation (RIP) assays, using qRT-PCR to quantify QKI occupancy within the Introns adjacent to the circGlis3 forming Exon3 (Figure 2L, Figure S2E). Pull-down assay also confirmed the binding interaction between the sites (which in the Introns flanking of Glis3 Exon3) and QKI protein (Figure 2M). Ectopic expression of QKI in MIN6 cells, and more RIP assays were used to evaluate that QKI expression levels changed the binding interaction between motif sites (QRE1&2, QRE3, QRE4) (Figure 2N-P, Figure S2F-H). Subsequently, mutation of the single putative binding sites individually had little effect on circRNA formation, but mutation of both members of the upstream pair and the downstream pair substantially reduced circRNA formation, mutation of all four sites was more effective (Figure 3Q). All these data indicate that QKI binds upstream and downstream of the circRNA-forming Exon3 in Glis3 to promote circGlis3 formation.
Upregulation of circGlis3 promotes insulin transcription and alleviates β-cell apoptosis in vitro.
To explore the potential role of circGlis3 in regulating β-cell function, we used short hairpin RNA targeting knockdown of circGlis3 (sh-circGlis3), which is a group of three shRNAs to knock down circGlis3 in MIN6 cell and primary islets respectively, and a plasmid with scramble sequence (sh-NC) as control. In addition, an overexpression plasmid (oe-circGlis3) was used to upregulate circGlis3. MIN6 cells and primary islets were transfected with sh-circGlis3, sh-NC, oe-circGlis3, or empty vector (pEx3ciR) and then were collected at 48 h. The efficiency of knockdown and overexpression were about 60% and 40-fold, respectively (Figure S3A and B). oe-circGlis3 significantly increased the mRNA for the insulin genes (Ins1 and Ins2) (Figure 3A and B) and the insulin content (Figure 3C and D), while knockdown of circGlis3 induced the opposite result. Glucagon content was not affected by the change in circGlis3 levels (Figure S3C). Next, we performed glucose challenge experiments using primary islets and MIN6 cells with knockdown or overexpression of circGlis3. Suppression of circGlis3 expression decreased insulin secretion after exposure to high glucose (16.7 mM glucose), and insulin secretion was markedly increased after circGlis3 overexpression (Figure 3E and F). To explain how upregulation of circGlis3 promotes insulin production, we detected both mRNA and protein expression of selected transcriptional factors with well-established roles in insulin transcription including pancreatic and duodenal homeobox factor 1 (Pdx1), neurogenic differentiation factor 1 (NeuroD1), Nk6 homeobox 1 (Nkx6.1), cAMP responsive element binding protein 1 (CREB1), and v-maf musculoaponeurotic fibrosarcoma oncogene homologue A (MafA). circGlis3 overexpression selectively increased CREB1 and NeuroD1 both in mRNA and protein levels (Figure 3G and H). However, neither the mRNA nor the protein levels of Pdx1, Nkx6.1 and MafA were affected by circGlis3 ectopic expression (Figure S3D and E). These results suggest that events responsible for the increase in insulin content may take place at the transcriptional level.
To identify the effect of circGlis3 on maintaining β-cell mass, we modified circGlis3 expression in MIN6 cells. CCK-8 assay showed that cell proliferation was slightly affected by regulating circGlis3 (Figure S3F), and Ki67 immunofluorescent staining also echoed this result (Figure S3G). circGlis3 overexpression in MIN6 cells resulted in a striking reduction in palmitate-induced apoptosis, as assessed by annexin V staining and by counting the cells displaying nuclei, further suggesting that circGlis3 upregulation acts as a resister of the obesity-induced apoptosis of β-cell (Figure 3I). Then we performed TUNEL assay in MIN6 cells and primary mice islets, even in primary human islets, the results shown that the downregulation of circGlis3 resulted in a rise in the number of apoptotic cells, but overexpression of circGlis3 shown the opposite results (Figure 3J-L). Furthermore, western blot indicated that apoptosis proteins were altered in abnormal circGlis3 treated cells, consistent with the flow cytometric and TUNEL analysis. Suppression of circGlis3 expression induced Cleaved-CASPASE 3 and BAX expression while decreased BCL-2 expression, overexpression induced the opposite result (Figure 3M). These findings indicate that the upregulation of circGlis3 can alleviate β-cell apoptosis during obesity.
Overexpression circGlis3 protects against β-cell dysfunction and apoptosis in vivo.
To test whether overexpression of circGlis3 can alleviate obesity-induced β-cell dysfunction in vivo, 1 × 1012 adeno-associated virus particles expressing circGlis3 (oe-circGlis3) was injected into male C57BL/6 mice aged at 8 weeks via pancreatic ductal infusion, and then fed the mice with HFD over 16 weeks. The time pattern of oe-circGlis3 injection and HFD feeding is shown in Figure 4A. We observed about 9-fold upregulation of islets circGlis3 expression in mice that had received the oe-circGlis3 compared to those receiving adeno-associated virus particles with empty vector (oe-vector) even 20 weeks after injection (Figure S4A). The expression level of circGlis3 was not significantly changed in other organs (Figure S4B). oe-circGlis3 treatment had no effect on cumulative energy intake (Figure S4C) and body weight (Figure S4D). Overexpression of circGlis3 as well as had no effect on fasting glycemia, although it was slightly lower than that in the control mice (Figure S4E). However, the fasting blood glucose was lower in circGlis3 overexpression mice after 3 months of HFD feeding (Figure S4F). Homeostatic model assessment of insulin resistance (HOMA-IR) indices of mice overexpressing circGlis3 were significantly decreased (Figure 4B). In accordance with this, glucose tolerance tests revealed an improvement of glucose tolerance upon circGlis3 overexpression (Figure 4C), insulin sensitivity was also improved upon upregulation of circGlis3 (Figure 4D). Moreover, we isolated islets of oe-circGlis3 treated and control mice after 8- and 16-week HFD feeding and performed GSIS test. The results revealed that insulin release was both markedly improved in oe-circGlis3 treated mice when islets exposed to 16.7 mM glucose (Figure 4E and F). CREB1 and NeuroD1 protein levels were also increased in oe-circGlis3 treated mice (Figure 4G). Morphometric analysis of the pancreas section revealed that β-cell mass was 1.7-fold higher in oe-circGlis3 treated mice than in control mice (Figure 4H). TUNEL assays were also performed after 8- and 20-week HFD feeding. Counts of TUNEL-positive β-cell were lower by 67% and 56% in circGlis3 overexpression mice after 8- and 20-week HFD feeding respectively, as compared to control animals (Figure 4I).
To confirm the effect of circGlis3 on β-cell protection, the same virus particles were injected into Leprdb/db mice aged at 4 weeks. We found that overexpression of circGlis3 prevented the increase of blood glucose at 10-week of age (Figure 4J). Further investigations demonstrated that the activation of circGlis3 led to decrease in the number of TUNEL-positive β-cell (Figure 4K) and maintain the β-cell mass (Figure 4L). Corresponding to the above results, the upregulation of circGlis3 resulted in significantly decreased the Cleaved-CASPASE 3 and BAX expression while increased BCL-2 expression (Figure 4M). Collectively, these results indicate that overexpression of circGlis3 in dietary and genetic mouse models of obesity results in improvement of β-cell function and inhibited to β-cell apoptosis.
circGlis3 regulates insulin transcription by sponging miR-124-3p
circRNAs may function as competing endogenous RNAs (ceRNAs) to sponge miRNAs, thereby modulating the derepression of miRNA targets and imposing an additional level of post-transcriptional regulation 14. To identify the potential miRNA targets of circGlis3, in silico analysis was performed by using Starbase, miRand, RNAhybrid and Targetscan databases, and jointly predicted that 4 miRNAs may act as biological targets of circGlis3 (Figure 5A). According to the assumption of ceRNA, the expression of circRNA and miRNA should show a negative correlation 20. We then tested the expression patterns of the above miRNA in the islets of Lepob/ob mice and HFD-fed mice. There were three miRNAs (miR-124-3p, miR-298-5p and miR-3104-3p) decreased in the islets of obese mice as opposed to the expression of circGlis3, while another one (miR-3113-5p) increased (Figure 5B and C). Although the expression pattern of miR-3104-3p is opposite to circGlis3, it showed very low abundance and hardly detected by qRT-PCR. The direct binding between the high abundance miRNAs (miR-124-3p and miR-298-5p) and circGlis3 was validated by affinity pull-down of endogenous miRNAs associated with circGlis3 using in vitro transcribed biotin-labeled circGlis3 sense or anti-sense and demonstrated via qRT-PCR analysis. First, the efficiency of the biotin-labeled circGlis3 binding on beads was detected (Figure S5A). The biotin-labeled circGlis3 sense in MIN6 cells was significantly captured endogenous miR-124-3p compared to blank (Beads) and circGlis3 anti-sense (Figure 5D). However, it slightly enriched for miR-298-5p (Figure S5B). FISH analysis was also detected in MIN6 cells, we found that miR-124-3p was co-localized with circGlis3 in the cytoplasm (Figure 5E). Based on the above results we selected miR-124-3p with the highest enrichment into further studies.
The predicted binding sites of miR-124-3p to circGlis3 were illustrated in Figure S5C. For further confirmation, we constructed a dual-luciferase reporter by inserting the wild type (WT) or mutant (MT) linear sequence of circGlis3 into pMIR-REPORT luciferase vector. We found that overexpression of miR-124-3p reduced the luciferase activities of the WT reporter vector but not mutant reporter vector (Figure 5F). Moreover, RNA pull-down assay using biotin-labeled miR-124-3p was also performed. The result revealed that endogenous circGlis3 was pulled down by miR-124-3p, whereas the negative control with a disrupting putative binding sequence failed to co-precipitate out circGlis3 (Figure 5G). As expected, upregulation of circGlis3 reduced the level of miR-124-3p, and knockdown of circGlis3 increased the level of miR-124-3p (Figure. 5H). However, we found no significant difference in circGlis3 levels after overexpression or suppression of miR-124-3p in MIN6 cells (Figure S5D). This data demonstrated that miR-124-3p bound to circGlis3 but did not induce the degradation of circGlis3. All these data demonstrated that circGlis3 physically associated with miR-124-3p and may function as a ceRNA.
To investigate whether circGlis3 regulates β-cell function through sponging miR-124-3p, the rescue experiment was performed. Overexpression of circGlis3 in MIN6 cell and primary islets resulted in upregulated Ins1 and Ins2 transcript, while overexpression of miR-124-3p abrogated this activation (Figure 5I and J, Figure S5E). The promotion effects of insulin secretion were also abolished by overexpression of miR-124-3p (Figure 5K and L). To validate whether circGlis3 may function as ceRNAs to modulate the derepression of miR-124-3p targets and imposed an additional level of post-transcriptional regulation, prediction of target genes of miR-124-3p was performed by TargetScan, Starbase, miRDB and miRWalk. Creb1 and NeuroD1 were in the intersection of predictions, which functions as major islet specific transcription factors (Figure S5F). The predicted binding sites of miR-124-3p to Creb1 and NeuroD1 were illustrated in Figure S5G and H. Luciferase reporter assay was applied through pMIR-REPORT vector holding wild-type or mutant 3’-UTR of Creb1 and NeuroD1, respectively. We found that overexpression of miR-124-3p reduced the luciferase activities of the WT reporter vector but not mutant reporter vector (Figure 5M and N), and the protein level also changed accordingly (Figure 5O). The restraint effect of miR-124-3p on Creb1 and NeuroD1 were reversed with the addition of circGlis3 in vitro (Figure. 5P and Q) and in vivo (Figure. 5R and S). However, palmitate-induced cell apoptosis induced via downregulation of circGlis3 could not be reversed by inhibition of miR-124-3p (Figure S5I). Altogether, these results suggest that circGlis3 regulates β-cell function by competitively binding miR-124-3p, while this interaction did not affect β-cell apoptosis.
circGlis3 Interacts with SCOTIN to prevent β-cell apoptosis
We next investigated the mechanisms by which circGlis3 prevented β-cell apoptosis during obesity. In addition to sponging for miRNAs, circRNAs could also regulate the functionality of RNA-binding proteins through direct binding 21. To investigate this possibility for circGlis3, we performed experiments in which we biotinylated the circGlis3 through assays of in vitro transcription and cyclization reaction (Figure S6A and B), pulled it down using anti-biotin beads, and subjected the pull-downs to mass spectrometry. Several proteins were identified from this analysis as the potent circGlis3 interacting protein (Table S5). Among the top candidates identified, SCOTIN and FUS most sparked our interest (Figure 6A, Figure S6C and D). SCOTIN is a pro-apoptotic factor, which is induced upon DNA damage or cellular stress in a p53-dependent manner 22,23. We hypothesized that circGlis3 could arrest β-cell apoptosis through the modulation of the SCOTIN.
We further confirmed the interaction between circGlis3 and SCOTIN by running additional independent experiments. First, we performed RIP assays, where we confirmed that endogenous circGlis3 interacts with SCOTIN in MIN6 cells (Figure 6B). Second, we pulled down the endogenous SCOTIN by biotinylated probes specific for circGlis3, and subjected the pull-down product to western blot analysis to detect the SCOTIN protein (Figure 6C). Finally, we coupled the circFISH protocol for circGlis3 visualization with an immunofluorescence staining for SCOTIN in MIN6 cells and islets, and results demonstrated that circGlis3 co-localized with the SCOTIN in the cytoplasm (Figure 6D). Taken together, these complementary set of experiments consistently showed that endogenous circGlis3 can interact with SCOTIN. These interactions take place mainly in the cytoplasm, where SCOTIN mostly reside.
Based on these findings, we then aimed to investigate the possibility that circGlis3 could function through the regulation of SCOTIN. We found that the SCOTIN was increased in the islets of 20-week HFD and Leprdb/db mice (Figure 6E and F), as well as in MIN6 cell which incubated with palmitate and glucose (Figure S6E and F), indicating a increase expression of SCOTIN in obese conditions. However, ectopic expression of circGlis3 has no effect on the transcription and translation of SCOTIN (Figure S6G). Overexpression of Scotin can induce β-cell apoptosis in MIN6 cells and primary islets, while upregulation of circGlis3 abrogated this effect (Figure 6G and H). The change trend of apoptosis detected by flow cytometry is consistent with the results of TUNEL experiment (Figure 6I). Further investigations demonstrated that overexpression of Scotin in HFD treated mice led to largely increase in the number of TUNEL-positive β-cells compared with control, while upregulation of circGlis3 abrogated this effect (Figure 6J). Of note, overexpression of Scotin induces apoptosis of β-cell accompanied by sequestration of circGlis3 in the cytoplasm (Figure 6K), and increased the Cleaved-CASPASE 3 and BAX expression while decreased BCL-2 expression (Figure 7L). It suggests that circGlis3 prevents β-cell apoptosis in a Caspase-dependent manner during obesity by interacting with SCOTIN and decreasing SCOTIN activity.
FUS sequestrates circGlis3 to reduce its abundance in diabetes
Finally, we attempt to explain why the expression level of circGlis3 decreased when diabetes occurred and lost its protective effect on β-cells. As mentioned earlier, the production of circGlis3 is regulated by QKI. However, the expression level of QKI did not decrease prominently with the occurrence of diabetes (Figure 7A and B, the concrete data were listed in Table S6). It suggests that there may be additional mechanisms to regulate the level of circGlis3 in diabetes. Using RNA pull-down assay and mass spectrometry analysis as we previously mentioned, we identified FUS that were pulled down by biotinylated probes specific for circGlis3, but not by probes for anti-sense (Figure S7A). FUS is a 70kDa RNA-binding protein with multiple functions, which can shuttle between the nucleus and cytoplasm in responses to stress 24,25. Interestingly, we found that the mRNA and protein levels of FUS were slightly up-regulated in HFD treated mice (Figure 7C) and MIN6 cells stimulated by glucose and palmitate (Figure S7B and C). Consistent with the results of previous studies 26, FUS mainly appears in the cytoplasm under stress (Figure 7D). Western blot analysis with anti-FUS antibody indicated the existence of FUS within the circGlis3 sense RNA probe pull-down samples in MIN6 cells (Figure 7E). Meanwhile, a RIP assay with FUS antibody showed that endogenous FUS directly bound to circGlis3 in MIN6 cells (Figure 7F). Ectopic expression of circGlis3 have no direct effect on transcription and translation of FUS (Figure S7D and E). However, we observed that overexpression of Fus reduced the free copies of endogenous circGlis3 in MIN6 cells and primary islets (Figure 7G), and the efficiency of overexpression as shown in Figure S7F.
In order to verify the regulatory effect of FUS on the downstream of circGlis3, we designed a series of competitive inhibition experiments. We found that overexpression of FUS accompanied by palmitate stimulation could reverse the inhibitory effect of circGlis3 on miR-124-3p, which was specifically reflected in the increase of miR-124-3p and the downregulation of its target genes (Figure 7H). Similarly, the combination of FUS and SCOTIN with circGlis3 is also competitive. The results of SCOTIN-RIP showed that FUS overexpression could competitively inhibit the enrichment of circGlis3 by SCOTIN in the presence of palmitate (Figure 7I), and accompanied with the increased of TUNEL-positive β-cells in HFD mice islets (Figure S7G). Conversely, after overexpression of Scotin, FUS combined with circGlis3 also decreased (Figure 7J).
It has been identified that the nuclear protein FUS quickly assembles cytoplasmic Stress Granules (SGs) during cellular stress 24. Zhang’s results supported that palmitate may disrupt nucleocytoplasmic transport by inducing SGs formation 27. We then immunostained normal MIN6 cells and NCD-fed mouse pancreatic tissue for circGlis3 and FUS and found that normal condition caused dispersed distribution of FUS in cytoplasm and nucleus with a few SGs assembly (Figure S7H). Given that SGs assembly may restricted diffusion of mRNAs and ncRNAs 28, we checked whether decrease of circGlis3 is recruited into SGs assembly of FUS in palmitate-stimulated MIN6 cells and 20-week HFD-fed mice pancreas. As shown in Figure 7K and L, upregulation of FUS accelerated the SGs formation (white arrows), and abolished the circGlis3 expression in cytoplasm when circGlis3 was over-expressed. The above results indicate that when diabetes occurs, FUS competes with miR-124-3p and SCOTIN to bind circGlis3, resulting in the decrease of circGlis3 by restricting diffusion in cytoplasm via recruitment function of FUS-formed SGs.