FUS binding to RNA prevents R-loops

The protein FUS (FUSed in sarcoma) is a metazoan RNA-binding protein that influences RNA production by all three nuclear polymerases. FUS also binds nascent transcripts, RNA processing factors, RNA polymerases, and transcription machinery. We explored the role of FUS binding interactions for activity during transcription. In vitro run-off transcription assays revealed FUS enhanced RNA produced by a non-eukaryote polymerase. Activity also reduced the formation of R-loops between RNA products and their DNA template. Analysis by domain mutation and deletion indicated RNA-binding was required for activity. We interpret that FUS binds and sequesters nascent transcripts to prevent R-loops forming with nearby DNA. DRIP-seq analysis showed that a knockdown of FUS increased R-loop enrichment near expressed genes. Prevention of R-loops by FUS binding to nascent transcripts has potential to affect transcription by any RNA polymerase, highlighting the broad impact FUS can have on RNA metabolism in cells and disease.


INTRODUCTION
A prominent member of the heterogeneous nuclear ribonucleoprotein (hnRNP) family is FUS (FUsed in Sarcoma). FUS is best known for gene mutations leading to amyotrophic lateral sclerosis or pediatric sarcomas (1)(2)(3). FUS is conserved throughout metazoan species and among the highest expressed proteins in human tissues (3). FUS has a structure that is largely intrinsically disordered (4). The disordered domains contained in FUS are its low complexity (LC) domain and three arginine and glycine-rich (RGG/RG) domains. Evidence of FUS activity on transcription is reported for all three RNA polymerases in the nucleus (5)(6)(7)(8)(9). FUS binds directly to nascent RNA transcripts, RNA Pol II, RNA Pol III, and other transcription and RNA-processing factors (4,(10)(11)(12)(13). Outside of these, FUS also interacts with mRNA, DNA repair machinery, DNA loops formed during recombination, and snoRNAs (12,(14)(15)(16).
FUS affinity for RNA involves an RNA-recognition motif (RRM), Zinc finger domain (ZnF), and the three disordered RGG/RG domains (17)(18)(19). FUS binds a large number and variety of messenger and noncoding RNAs with a degenerate specificity but prefers G-rich sequences, RNA stem-loops, and more complex RNA structures (18)(19)(20). RNA binding drives FUS to oligomerize into large assemblies, which enhance binding to RNA Pol II and other protein partners (21,22). In cells, the assembly behavior of FUS is found in protein particles that contain FUS and RNA Pol II and form in a transcriptiondependent manner (5,23). FUS is also enriched near transcription start sites (TSS) of most expressed genes encoding mRNA and prevents phosphorylation of Ser-2 (Ser2P) in the heptad repeat of the Cterminal domain (CTD) in RNA Pol II (8,24).
We set out to better define the role that binding the RNA polymerase has in FUS activity on transcription. We anticipated direct binding to the polymerase and RNA processing factors would be required for FUS activity (7,8,17). However, FUS showed activity on the transcription of a noneukaryotic RNA polymerase intended to be a negative control. We decided to refocus our study on this unexpected activity observed in the absence of previously known protein:protein interactions.

Recombinant protein expression and purification
FUS constructs used in this study are available upon request. All recombinant proteins were expressed and purified from E. coli BL21(DE3) cells. All FUS constructs were N-terminally fused to 6x His and maltose-binding protein (MBP) to improve purification and solubility. Protein expression was induced after cell growth reached an OD600 of 0.8 and then allowed to continue overnight at 17 ºC with shaking (200 rpm). Purification was made using 1 to 2 g of frozen pellets of induced E. coli, lysates incubated with Ni-NTA Sepharose beads (Cytiva Lifesciences, 17531802), and eluted in FUS-SEC buffer (1 M urea, 0.25 M KCl, 50 mM Tris-HCl pH 8.0) with 250 mM imidazole and either 1.5 mM -mercaptoethanol or 1 mM DTT added (5,18,20).
Plasmids for protein expression of T7(P266L) polymerase fused with a 6xHis tag were provided by A. Berman (University of Pittsburgh) (25). Expression in E. coli BL21(DE3) cells and purification were done essentially as previously published (26). Protein expression was induced by 0.5 mM Isopropyl -D-1-thiogalactopyranoside (IPTG) when transformed cells grew to an OD600 of 0.6 then continued for 3 hours at 37 ºC. Induced cells stored at -80 ºC were thawed, lysed in T7 Buffer (50 mM Tris-HCl pH 8.0, 100 mM NaCl, 5% v/v glycerol, and 5 mM -mercaptoethanol added just before use) with 1 mM imidazole then incubated with Ni-NTA beads. T7 Buffer with increasing imidazole was used to wash beads 4 times (1 mM imidazole), 2 times (10 mM imidazole), and to elute protein (100 mM imidazole).
Size-exclusion chromatography (SEC) was performed with a 10/300 Superdex® 200 column (Cytiva Lifesciences, 17517501) for FUS using FUS-SEC buffer and for T7 Pol with T7-SEC buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 0.1 mM EDTA, and 1 mM DTT). FUS protein was stored at room temperature for up to 4 weeks or until activity appeared lost. T7 Pol activity was assessed by titrating protein into transcription assays as described below. The absence of nuclease activity was confirmed by comparing plasmid DNA and transcribed RNA incubated overnight with or without protein and electrophoresis using a Novex™ TBE-urea, 6% polyacrylamide gel (Invitrogen, EC68655BOX) and stained with SYBR-Gold (Invitrogen, S11494).

DRIP-seq assay and data analysis
DRIP-seq assays were performed essentially as a previously published protocol (29 Sample libraries were prepared and sequenced by Novogene Corporation Inc. using a NovoSeq6000 with 150 base paired-end reads (8 G raw data per sample). Reads were trimmed using TrimGalore (version 0.6.6) and aligned to GRCh38 using STAR (version_2.7.8a), essentially as previously described. Bam files were processed using the deepTools2.0 suite (version 3.5.1) to normalize to 1x depth (reads per genome coverage) with bamCoverage, merge replicates by bigwigCompare, and tags binned by computeMatrix, which was then used to compute scaling factors and generate heat maps and profiles. Merged bigwig files were converted to bedgraph and then peaks by Macs2 (version 2.1.1.20160309). HOMER (version v4.11.1) was used to calculate GC-content and annotate peaks. Data was visualized, and figures were generated using Integrated Genomics Viewer (version 2.5.0). DRIP-seq data is available (GSE206740) from the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/).

FUS increases RNA yields from T7 Pol transcription
In our previous work, we observed FUS at concentrations found in the cell, >2 M, greatly increased RNA Pol II transcription from a naked DNA template (5). We asked whether the specific protein-protein interactions FUS makes with RNA Pol II were required for activity. We chose the T7 phage RNA polymerase (T7 Pol) as a negative control that lacked binding sites present in mammalian RNA polymerases (28). Recombinantly expressed FUS with an N-terminal maltose binding protein (MBP) tag and T7 Pol were purified to homogeneity for in vitro run-off transcription of a linear DNA template (Supplemental Figure 1A).
Accumulated products of run-off transcription were observed by polyacrylamide gel electrophoresis (PAGE) and SYBR staining ( Figure 1A). Two stained bands could be seen: a larger band that was confirmed by DNase I digestion to be linear DNA template (Supplemental Figure 1B-C), and a smaller band confirmed by RNaseA digestion to be the RNA transcript (Supplemental Figure 1D). The DNA template detected offered a normalization control for quantitative analysis of PAGE images.
Contrary to our expectation, T7 Pol produced more RNA transcript in the presence of FUS (2 µM) ( Figure 1B, Supplemental Figure 1D-E). Titrating buffer constituents present in the assay eliminated these as potential sources of activity (Supplemental Figure 1F). Increasing concentrations of FUS during T7 Pol transcription increased RNA products observed by PAGE ( Figure 1C, Supplemental Figure 1G). Sufficient concentration of FUS produced a shift of RNA to the well, which was reversed by proteinase K treatment to eliminate FUS aggregates formed (Supplemental Figure 1H). FUS tendency to aggregate has also been noted by previous studies (21,30).
We investigated whether a direct interaction of FUS and T7 Pol could be observed through a coimmunoprecipitation (co-IP). The protocol used for co-IP was successful in detecting FUS binding to protein interactors, including RNA Pol II and III, in previous studies (5,9,11,30,31). T7 Pol incubated with immobilized FUS remained in the supernatant fraction and undetectable in the elution for negative controls and all replicates but one (N = 4, Figure 1D, Supplemental Figure 1I). Robust binding of T7 Pol could not be provoked by the linear DNA template, or an RNA known to cause FUS to bind RNA Pol II ( Figure 1D, Supplemental Figure 1I) (8,21). We concluded that a protein:protein interaction, like that seen for FUS and RNA Pol II, was unlikely to offer a compelling mechanism for FUS to influence transcription by T7 Pol.

Requirements for FUS activity on T7 Pol transcription
We next considered whether high protein concentrations affected transcription by non-specific molecular crowding. Increases to T7 Pol activity by molecular crowding with glycerol or PEG has been observed by previous studies (32)(33)(34). Activities were compared FUS or a well-folded soluble protein, bovine serum albumin (BSA), present at equal concentration during run-off transcription. BSA had no effect on RNA products observed (Figure 2A). Fluorescence-based quantification with an RNA-specific dye also detected no activity from BSA ( Figure 2B). We also noted FUS activity was diminished after storage at -80 ºC, but stored at room temperature, FUS activity remained stable up to 4 weeks, then no activity was seen at 12 weeks (Supplemental Figure 2A). The lack of effect on transcription by inactive FUS added support to the conclusion that the activity observed was specific. We considered if RNA in solution might alter FUS activity. Such an effect by RNA or DNA in solution could make the activity of FUS observed during run-off transcription difficult or impossible to observe in a cell, where concentrations of nucleic acids are high. We investigated two RNA molecules, a nonspecific competitor RNA, yeast tRNA, and an RNA, TET456, previously used to induce FUS binding to RNA Pol II (21). FUS activity was unaffected by tRNA at the same concentration of transcript the assay produced ( Figure 2C). Titrating tRNA or TET456 RNA concentrations had no effect on T7 Pol or FUS activity (Supplemental Figure 2B-D). A repeat of these tests with single-stranded DNA also found the activity of both proteins unaffected (Supplemental Figure 2E).

FUS prevents RNA:DNA hybrids formed during transcription
Because FUS is an RNA-binding protein, we considered a mechanism for nascent RNA to affect transcription by binding its complementary template DNA to form an RNA:DNA hybrid. Together with the displaced non-template DNA strand, the structure is termed an R-loop (35,36). R-loops are found in cells at low but consistent levels, even though they can pose a threat to DNA stability, replication, and transcription (9,35,37,38). For reasons like this, cells possess multiple mechanisms to prevent or remove them (35,39).
By dot blot assay using the S9.6 antibody, RNA:DNA hybrids were found to arise early during runoff transcription and their level increased over time ( Figure 3A) (29,40). A nuclease that cleaves RNA:DNA hybrids, RNaseH, could eliminate signals detected, confirming specificity for the S9.6 antibody in this assay. Additional controls showed no signal was produced by RNA alone or when NTP or T7 Pol was omitted from the run-off transcription assay (Supplemental Figures 3A).   We focused on the role of RNA-binding for activity by substituting serine for arginine residues in one (RS1, RS2, RS3) or all (RS4) of the RGG/RG domains in FUS ( Figure 4A). Compared to FUS affinity for RNA, characterization of these substitutions by previous studies found RS1 and RS2 maintained similar affinity, RS3 affinity was reduced by 7-fold, and RS4 affinity was reduced by >30fold (18). In the run-off transcription assays, RS1 or RS2 increased RNA yields up to 2-fold but did not significantly reduce RNA:DNA hybrids (Figure 4C, Supplemental Figure 4D). RS3 increased T7 Pol transcripts by >2-fold, but less than that for FUS. RNA:DNA hybrids were reduced by RS3 a small but significant amount (p < 0.05, student t-test assuming equal variances, Figure 4C, Supplemental Figure 4D). RS4 activity on transcript production matched that of RS1 and RS2, and a small reduction of hybrids could be detected (Figure 4C, Supplemental Figure 4D).
Finally, inspection by microscopy was performed to assess any self-assembled particles made by FUS or RS proteins. A necessary change from conditions in run-off transcription assays was for proteins to be allowed to incubate at room temperature for 24 hours. FUS, RS1, RS2, and RS3 proteins (2 µM) all produced visible particles up to 0.5 µm in diameter, but RS4 yielded no evidence of assemblies (Supplemental Figure 4E). RS4 particles were not produced by incubation of RS3 and RS4 at 10 µM (Supplemental Figure 4F). Lastly, FUS and RS4 were tested with protein concentrations up to 100 µM, incubation at room temperature or 4 ºC, and addition of RNA, but RS4 did not yield unambiguous particles, while FUS achieved particles ≥1 µm in diameter (Supplemental Figure 4G-H).

FUS expression lowers R-loop abundance in human cells
We hypothesized the activity observed for T7 Pol transcription would also be found in human cells. To test this in human HEK293T/17 cells, we used an siRNA approach to knockdown FUS expression and measured R-loop enrichment genome-wide by DRIP next-generation sequencing approach, DRIP-seq, that immunoprecipitates R-loops with the S9.6 antibody (29,41,42). Knockdown of FUS protein by siRNA (siFUS) was found to be >90% relative to control siRNA (SCR) (Supplemental Figure 5A) (8).
Antibody detection of RNA:DNA hybrids and dsDNA in cell lysates was assessed by dot blot assay (Supplemental Figure 5B). Real-time PCR analysis confirmed that positive control genomic sites, RPL13A and TFPT, known to contain R-loops was enriched by S9.6 pulldown, relative to a negative control site, EGR1. No enrichment was observed for lysates incubated with RNaseH (Supplemental Figure 5C) (29). Analysis also suggested R-loop enrichment for RPL13A and TFPT increased in cells with FUS knocked down relative to control (Supplemental Figure 5C).

DRIP-seq analysis detected enrichment of R-loops at transcription start sites (TSS) of expressed
genes, which was higher in siFUS-treated cells relative to SCR controls ( Figure 5A). Increased enrichment of >2-fold was observed for 29% (N=5260) of protein coding genes (N=18188) (Figure 5B).
An example typical of RNA Pol II genes, GAPDH, R-loop enrichment near its promoter was substantially greater in siFUS samples relative to input, SCR control, and RNaseH treated samples ( Figure 5C) (8).
The RNA Pol I genes encoding ribosomal RNA (rRNA) contained the largest increase to R-loop enrichment found in siFUS samples ( Figure 5D). On average, most effects on R-loop enrichment due to FUS knockdown were in GC-rich DNA sequences (Supplemental Figure 5D). Also, peaks for R-loop enrichment were primarily in intergenic and satellite DNA in SCR-treated cells (N=5045 peaks), and mostly in protein-coding genes and promoters for siFUS-treated cells (N=7247 peaks) (Supplemental Figure 5E). Peaks produced by siFUS treatment were noted at repetitive DNA features (Supplemental Figure 5F). Finally, inspection of RNA Pol III transcribed genes found increased R-loop enrichment at 5S rRNA genes, but not RMRP, RN7SK, or tRNA gene clusters on chromosomes 1 and 6 (Supplemental Figure 5G).
In conclusion, the analysis of DRIP-seq indicated widespread FUS activity preventing R-loops in human HEK293T/17 cells. FUS activity on R-loops was not limited to transcription by RNA Pol I, II, or III. The changes to R-loop enrichment that could be measured by DRIP-seq revealed a close similarity of FUS activity in human cells and that observed during in vitro run-off transcription.

DISCUSSION
Here we explore FUS activity on transcription that involves the potential for RNA transcripts to bind their DNA template, forming an R-loop. Our study followed the unexpected result that FUS increased the RNA yield from T7 Pol transcription. Subsequent analysis determined FUS reduced RNA:DNA hybrids that formed during transcription. Our model is that FUS increases RNA yield by binding nascent transcripts to prevent or delay R-loop formation, which would slow or block the RNA polymerase. In human cells, a reduction of FUS protein increased R-loop abundance. The R-loop enrichment was most prominent at expressed genes, notably those transcribed by RNA Pol I and II. These results indicate FUS bound to nascent RNA transcripts can influence transcription by preventing R-loops.
Previous studies indicate FUS influences transcription by RNA Pol I, II, and III (3,7,9,43). We have added to this list T7 Pol from bacteriophage, which shares no unambiguous sequence similarity to eukaryote RNA polymerases (44,45). Our results suggest that by preventing R-loops, FUS lowers a barrier that accumulates over the course of transcription. It is important to note that R-loops have many effects on transcription and chromatin stability besides the simple stalling of the polymerase that we have studied here (35,36,(46)(47)(48). This activity for FUS also differs from those previously studied in that direct binding to the polymerase would not be a requirement for the mechanism we have proposed (21,22). The likelihood is low that a binding site is conserved between human and phage RNA polymerases, and our co-IP assay did not provide evidence of FUS binding to T7 Pol (Figure 1D).
Binding to RNA rather than the polymerase also provides the simplest explanation for FUS to affect formation of R-loops at sites transcribed by RNA Pol I, II, and III ( Figure 5).
Ordinarily, an increase in transcription would favor RNA:DNA hybrid formation, but the prevention of R-loops by an RNA-binding protein is not unprecedented (49). Failure to recruit snRNPs, SR proteins, and hnRNPs can allow R-loops to form (41,(50)(51)(52)(53)(54). Protection against R-loops by SRSF1 is one example that has been reproduced in run-off transcription by T7 Pol (53). The effect seen in DRIP-seq from a FUS knockdown indicates that FUS is not easily replaced by another RNA-binding protein in the cell. This may be due in part to the contribution of the LC domain ( Figure 4B). By comparing activity of FUS and other RNA-binding proteins with LC domains, a future study may reveal what features in a LC domain can contribute most to boost activity that prevents R-loops.
In conclusion, the ability of FUS to influence transcription can be extended to include preventing Rloops by binding nascent RNA transcripts. FUS and R-loops have complex relationships with transcription and DNA stability. The mechanism we describe would not be expected to conflict with FUS activities that involve binding the polymerase, transcription factors, or RNA-processing machinery, but this will remain to be seen by future studies. It is noteworthy that FUS and R-loops are both known to contribute to the same or related neurodegenerative diseases (9,(55)(56)(57)(58). Since FUS is highly expressed protein in human cells, its activity likely has significant implications to R-loop function and the contributions these enigmatic nucleic acid structures make to biology and disease.

SUPPLEMENTARY DATA
Supplementary Data are available at NAR online.

CONFLICT OF INTEREST
There are no conflicts of interest to report relating to this work.       C In S1 S2 E M In S1 S2 E M In S1 S2 E + DNA + RNA In S1 S2 E M In S1 S2 E M In S1 S2 E C + DNA

Anti-FUS IP
In S1 S2 E IgG IP In S1 S2 E F

MBP-FUS T7 Pol
IgG IP In S1 S2 E M MBP-FUS T7 Pol + RNA C In S1 S2 E M In S1 S2 E M In S1 S2 E + DNA + RNA In S1 S2 E M In S1 S2 E M In S1 S2 E C

B.
A.

C.
A.