SRSF3-TRIM28-MDC1 prevents DNA damage caused by R-loops in fatty liver disease in mice

Abstract

Serine-rich splicing factor 3 (SRSF3) is crucial for the metabolic functions of the liver. The genetic deletion of SRSF3 in mouse hepatocytes impairs hepatic lipid and glucose metabolism and leads to fibrosis and formation of hepatocellular adenoma that progresses to hepatocellular carcinoma. SRSF3 protein is proteosomally degraded in metabolic-dysfunction associated fatty liver disease (MAFLD) and metabolic-dysfunction-associated steatohepatitis (MASH). We show here that depleting SRSF3 protein in hepatocytes promoted R-loop accumulation and increased DNA damage in the liver. Prevention of SRSF3 degradation in vivo protected hepatocytes from DNA double-strand breaks in mice with MASH. This protection extended to other DNA-damaging agents such as camptothecin, palmitic acid, or hydrogen peroxide when tested on HepG2 cells in vitro. SRSF3 interacted with TRIM28 and MDC1, which are components of the ATM DNA-damage repair complex, and knockdown of any of these 3 proteins reduced the expression of the other 2 proteins, suggesting they form a functional complex. Lastly, by preventing degradation of SRSF3, we were able to reduce tumors in a diethyl-nitrosamine-induced (DEN-induced) model of cirrhotic HCC. These findings suggest that maintenance of SRSF3 protein stability is crucial for preventing DNA damage and protecting liver from early metabolic liver disease and progression to HCC.

Keywords: DNA repair; Endocrinology; Hepatology; Liver cancer; RNA processing.

Figure 1. Preventing SRSF3 degradation reduces hepatocyte DNA damage in vivo.

(A) Immunoblotting of γH2ax, 53BP1, BRCA1, and SRSF3 in livers from mice on high-fat diet (MAFLD) or Western diet (MASH), or lean mice on normal chow (control). Graph shows quantification of γH2ax, 53BP1, BRCA1, and SRSF3 protein levels normalized to β-actin (n = 3–4/group). Lean mice are shown in white, MAFLD mice in yellow, and MASH mice in red. (B) Immunoblotting of γH2ax and SRSF3 from normal or MAFLD human livers. Graph shows quantification of γH2ax and SRSF3 protein levels normalized to β-actin (n = 6–7/group). Normal is shown in white and MAFLD in red. (C) Immunohistochemical staining for γH2ax on FFPE sections from livers of lean, MAFLD, and MASH mice. Arrows indicate representative positive nuclei. Graph shows quantification of γH2ax⁺ nuclei per field (n = 3/group). Scale bars: 100 μm. (D) Immunohistochemical staining for γH2ax on FFPE liver sections lean mice on normal chow (control) or mice on Western diet infected with AAV8 expressing GFP (MASH-GFP), WT SRSF3 (MASH-WT), or the degradation-resistant K11R-mutant SRSF3 (MASH-K11R). Arrows indicate representative positive nuclei. Graph shows quantification of γH2ax⁺ nuclei/field (n = 3/group). Scale bars: 100 μm. All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001 by 1-way ANOVA.

Figure 2. Preventing SRSF3 degradation reduces HepG2 cell DNA damage in vitro.

(A) Immunofluorescence staining for γH2ax in HepG2 cells (green) treated with control or SRSF3 siRNA (20 nM, 48 hours). DAPI was strained to visualize the nuclei (blue). Panels show γH2AX fluorescence alone or merged with DAPI. Scale bar: 100 μm. (B) Immunoblotting of SRSF3, γH2ax, 53BP1, and BRCA1 from human hepatocytes with or without SRSF3 knockdown. Gels were run in parallel, and individual actin control blots are shown. (C) Graphs showing percentage of γH2ax⁺ nuclei by immunofluorescence (n = 3/group) or SRSF3, γH2ax, 53BP1, and BRCA1 protein levels normalized to β-actin by Western blot (n = 3/group). (D–F) HepG2 cells were infected with AAV8 expressing GFP, SRSF3-WT, SRSF3-K11R directly at MOI 500,000 for 48 hours. γH2ax was detected by immunofluorescence (red) following induction of DNA damage with 0.1 μM CPT for 1 hour (D), 500 μM PA for 12 hours (E), or 200 μM H₂O₂ for 1 hour (F). Nuclei were counterstained with DAPI (blue). Scale bars: 100 μm. Graphs show quantification of γH2ax⁺ nuclei/field by immunofluorescence (n = 3/group). Control is shown in white, GFP in green, WT in yellow, and K11R in red. All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001 by 1-way ANOVA.

Figure 3. Knockdown of SRSF3 causes R-loop accumulation.

(A) Dot blot of R-loops using antibody S9.6 in HEK293 cells treated with control siRNA (si-Control) or SRSF3 siRNA (si-SRSF3). Bots were stripped and reblotted for double-stranded DNA (dsDNA). Graph shows quantification of R-loop levels normalized to dsDNA (n = 2/group). (B) Dot blot of R-loops in genomic DNA from Flox and SRSF3-KO hepatocytes (SRSF3-HKO). Graph shows quantification of R-loop levels normalized to dsDNA (n = 3/group). (C) HEK293 cells were transfected by GFP, Flagged-SRSF3-WT, or Flagged-SRSF3-K11R plasmids for 48 hours, before being treated with 0.1 μM CPT or DMSO (vehicle control) for 1 hour and R-loops detected by dot blot. Graph shows quantification of R-loop levels normalized to dsDNA (n = 3/group). (D) HepG2 cells were infected by AAV8 expressing GFP or Flagged-SRSF3-K11R, before being treated with 0.1 μM CPT or DMSO for 1 hour, and R-loops were detected by dot blot. Graph shows quantification of R-loop levels normalized to dsDNA (n = 3/group). (E) Dot blot of R-loops in genomic DNA from livers of lean mice on normal chow (control) or mice on a Western (MASH) diet 7 weeks after infection with AAV8 expressing GFP (MASH-GFP), WT SRSF3 (MASH-WT), or the degradation-resistant K11R-mutant SRSF3 (MASH-K11R). Graph shows quantification of R-loop levels normalized to dsDNA (n = 3/group). (F) Immunoblotting of SRSF3, γH2ax from HEK293 cells with or without SRSF3 knockdown. Five units of RNase H protein were transfected in selected wells for 4 hours to digest R-loops. Graph shows quantification of SRSF3 and γH2ax protein levels normalized to β-actin (n = 3/group). (G) Immunofluorescent staining for γH2ax in HEK293 cells treated as above. DAPI was stained to visualize the nuclei. ce staining for TRIM28 and SRSF3. In all cases DAPI was used to visualize the nuclei. Scale bar: 10 μm. Original magnification, ×630. Graph shows quantification of γH2ax⁺ nuclei/field (n = 3/group). All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001 by 1-way ANOVA.

Figure 4. Spatial and genomic effects of SRSF3 on R-loops.

(A) Immunofluorescence staining for R-loops with antibody S9.6 in HEK293 cells with siSrsf3 knockdown. In some samples, cells were treated with RNASEH1 before S9.6 staining to digest RNA-DNA hybrids. Nuclei were counter stained with DAPI and S9.6 staining was quantified specifically in nuclei. Scale bars: 10 μm. (B) Genomic R-loops were profiled in primary hepatocytes with acute deletion of SRSF3 using Cut&TAG and short-read sequencing. Reads were aligned to the mm10 genome and peaks assessed using the HOMER suite. Graph shows the total number of peaks in the control and KO hepatocytes, compared by 2-tailed t test. (C) Adjacent peaks were merged into larger R-loop regions. Graph shows total number of R-lop regions in control and KO hepatocytes. (D) The size of the R-loop regions was quantified. The KO hepatocytes showed an increase in R-loom size by nonparametric Kolmogorov-Smirnov test. (E) R-loops were annotated to known genomic features. The left pie chart shows the proportion of total R-loops assigned to different genomic features. The right pie chart shows the distribution of R-loops that were different between control and KO hepatocytes. (F) Three representative genes (Rcc1, Ppp1r21, and Tcf12) that showed differential R-loops in the KO hepatocytes. Gene structure is shown at top with KO and control R-loops and SRSF3 CLIP profile below. (G) The scatter plot shows the correlation of SRSF3 CLIP tags at sites that have differential R-loop peaks. Dotted line indicates separation of peaks into those with background levels of CLIP tags versus peaks with associated SRSF3 binding. (H) Motif analysis on the differential R-loops. Chart shows enriched de novo motifs with enrichment P value, percent of targets and background containing the motif, and best match to known factors. *P < 0.05, **P < 0.01, ****P < 0.0001. Enrichment probabilities calculated using cumulative hypergeometric distribution.

Figure 5. SRSF3 interacts with TRIM28 and MDC1.

(A) Overlap of SRSF3-interacting proteins and R-loop–interacting proteins from MS data. (B) STRING protein-protein interaction network of proteins involved in DNA repair (green) and R-loop resolution (yellow) with SRSF3. (C–E) Immunoblots of SRSF3, TRIM28, and MDC1 proteins in HepG2 cells transfected with siRNA to SRSF3, TRIM28, or MDC1 (40 nM, 48 hours). Gels were run in parallel, and individual actin control blots are shown. Graph shows quantification of protein levels normalized to β-actin (n = 3/group). (F–H) Immunoblots of SRSF3, TRIM28, and MDC1 proteins in primary human hepatocytes transfected with siRNA to SRSF3, TRIM28, or MDC1 (20 nM, 48 hours). SRSF3 knockdown efficiency for F is shown in Figure 2B. Graph shows quantification of protein levels normalized to β-actin (n = 3/group).

Figure 6. TRIM28 and MDC1 proteins are reduced by lipid overload.

(A) Immunoblots for TRIM28, TRIM28(pSer473), and MDC1 in hepatocytes from mice on high-fat diet (MAFLD), Western diet (MASH), or normal chow (control). Actin loading control for TRIM28 is same as in Figure 1A for SRSF3. Graph shows quantification of protein levels normalized to β-actin (n = 3–4/group). Lean mice are shown in white, MAFLD mice in yellow, and MASH mice in red. (B) Immunoblots of TRIM28, TRIM28(pSer473), MDC1, and ATM from HepG2 cells treated with methyl-β-cyclodextrin (MBCD; 1 mM) as control or palmitic acid (500 μM) complexed to MBCD (1 mM) for 12 hours. Graph shows quantification of protein levels normalized to β-actin (n = 3/group). (C) Immunoblots of TRIM28, TRIM28(pSer473), MDC1, and SRSF3 in HepG2 cells infected with AAV8 expressing GFP, SRSF3-WT, SRSF3-K11R (MOI 500,000 for 48 hours) followed by 500 μM PA treatment for 12 hours. Graph shows quantification of protein levels normalized to β-actin (n = 3/group). Control group is shown in white, GFP in green, WT in yellow, and K11R in red. All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by 1-way ANOVA.

Figure 7. Loss of TRIM28 causes DNA damage.

(A) Immunoblots of γH2ax in primary human hepatocytes treated with control and TRIM28 siRNA. Graph shows quantification of γH2ax protein levels normalized to β-actin (shown for MDC1 in Figure 5G) (n = 3/group). (B) Immunoblots of γH2ax and TRIM28 in hepatocytes from flox mice and TRIM28-HKO mice on high-fat diet (MAFLD) for 16 weeks. Graph shows quantification of protein levels normalized to β-actin (n = 4–5/group). Flox mice are shown in white and TRIM28-HKO mice in red. (C) Immunoblots of γH2ax and TRIM28 in hepatocytes from Flox mice and TRIM28-HKO mice on Western diet (MASH) for 12 weeks. Graph shows quantification of protein levels normalized to β-actin (n = 6-7/group). Flox mice are shown in white and TRIM28-HKO mice in red. 33. (D) Immunoblots of γH2ax, TRIM28, and SRSF3 in HEK cells transfected with SRSF3 siRNA or/and HA-TRIM28 plasmid. Graph shows quantification of protein levels normalized to β-actin (n=3/group). (E) Whole liver images and H&E-stained sections from 15- to 18-month Flox mice and TRIM28-HKO mice. Black line indicates tumor-normal border. Scale bar for liver pictures: 1 cm. Scale bar for sections: 250 μm. Graphs show total tumor number and liver to body weight ratio (n = 7-8/group). Flox mice are shown in white and TRIM28-HKO mice in red. All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by 2-tailed t test or 1-way ANOVA.

Figure 8. SRSF3 and TRIM28(pSer824) colocalize to sites of DNA damage.

HepG2 cells were treated with DMSO or CPT (5 μM) for 1 hour. (A) Immunofluorescence staining for γH2ax and SRSF3. (B) Immunofluorescence staining for TRIM28(pSer824) and γH2ax. (C) Immunofluorescence staining for TRIM28(pSer824) and SRSF3. (D) Immunofluorescence staining for TRIM28 and SRSF3. In all cases DAPI was used to visualize the nuclei. Scale bars: 10 μm. Graph shows Pearson’s correlation coefficient (n = 3/group) derived from overlap image. DMSO (control) group is shown in white and CPT group shown in red. All quantified results are presented as mean ± SD; ***P < 0.001 by 2-tailed t test.

Figure 9. Preventing SRSF3 degradation reduces HCC.

(A) Schematic of the chemically induced mouse HCC model. DEN (25 mpk, i.p.) was administered to mice at 2 weeks of age, mice were placed on Western diet starting at 4 weeks of age and injected with TAA (300 mpk, i.p.) twice a week for a further 20 weeks. AAV8 infection was performed by tail vein injection on week 6. (B) Whole liver images and H&E and Sirius red–stained sections from lean mice (control), HCC mice (DEN+TAA+NASH), HCC mice infected with AAV8-GFP (AAV8-GFP), and HCC mice infected with AAV8-SRSF3-K11R (AAV8-K11R). Scale bar for liver pictures: 1 cm. Scale bar for sections: 250 μm. Graphs show liver to body weight ratio (n = 4–7/group) and quantification of fibrosis by Sirius red staining (n = 4/group). Control mice are shown in white, DEN+TAA+MASH-treated (DEN) in yellow, GFP-infected in green, and K11R-infected in red. (C) Sirius red–stained whole liver sections from mice infected with AAV8-GFP or AAV8-K11R (n = 4/group) showing pale steatotic tumors (arrows). Graph shows quantification of tumor number. (D) Representative reticulin staining of liver sections from AAV8-GFP– or AAV8-K11R–infected mice. T, tumor; NL, adjacent nontumor liver. Scale bar for reticulin-stained sections: 250 μm. Scale bar for magnified sections: 50 μm. (E) Immunohistochemical staining for γH2ax (brown) and quantification of positive nuclei/field (n = 3/group). Black arrows indicate representative positive cells. Scale bar: 250 μm. All quantified results are presented as mean ± SD; *P < 0.05, **P < 0.01, ****P < 0.0001 by 1-way ANOVA or 2-tailed t test.