Non-transcriptional activity of IRF3 promotes liver fibrosis and stress granule assembly via a dsRNA-TLR3-dependent pathway in hepatic stellate cells

Abstract

Background & Aims:

Interferon regulator factor 3 (IRF3) executes multiple transcriptional and non-transcriptional functions which regulate development of metabolic liver diseases. However, the contribution of the transcriptional and non-transcriptional activities of IRF3 to hepatic fibrosis are unknown.

Approach & Results:

Chronic carbon tetrachloride (CCl₄) and choline-deficient L-amino-defined-diets both induced fibrosis in C57BL/6J (wild-type) and mice expressing only non-transcriptional IRF3 activity (Irf3^S1/S1) but not Irf3-deficient (Irf3^−/−) mice. HSC-specific deletion of Irf3 (Lrat-Irf3) also attenuated CCl₄-induced fibrosis. Molecular pathways associated with regulation of extracellular matrix were up-regulated during differentiation in primary HSCs from wild-type and Irf3^S1/S1, but not Irf3^−/−, mice. Fibrogenic gene expression was also reduced in primary HSCs from Irf3^−/−, but not Irf3^S1/S1, mice upon TGFβ stimulation. IRF3 gene silencing in LX2 cells suppressed TGFβ-induced fibrogenic gene expression and collagen production. Proteomic analysis identified stress granule (SG) associated proteins within the IRF3 interactome, and IRF3 localized to SGs in LX2 cells upon stimulation with poly(I:C) or TGFβ. Mechanistically, differentiation of primary HSCs or stimulation of LX2 cells with TGFβ induced accumulation of endogenous double-stranded RNA (dsRNA), which triggered assembly of IRF3-containing SGs in a TLR3-dependent mechanism. TGFβ stimulated SG assembly in primary HSCs from wild-type and Irf3^S1/S1, but not Irf3^−/−, mice, and SG were detected in fibrotic human livers.

Conclusion:

The non-transcriptional activity of IRF3 is a key driver of hepatic fibrogenesis. Non-transcriptional activity of IRF3 is activated via accumulation of endogenous dsRNA and TLR3-dependent assembly of IRF3-containing SGs and acts as a central regulator of HSC activation and hepatic fibrogenesis.

INTRODUCTION

Fibrosis is a hallmark of advanced liver disease and is driven by dysregulated wound healing responses to chronic liver injury that result in imbalanced deposition and resorption of extracellular matrix (ECM)¹. A complex interplay of stimuli, including the multifunctional cytokine transforming growth factor-β (TGFβ) and activation of innate immune signaling pathways triggered by pathogen- or damage-associated molecular patterns (PAMPs/DAMPs)², induces excessive transdifferentiation of quiescent hepatic stellate cells (HSCs) into proliferative, myofibroblast-like cells that are essential for hepatic fibrogenesis. The stage of fibrosis is the most important prognostic factor in steatotic liver diseases, including alcohol-associated liver disease (ALD) and metabolic dysfunction-associated steatotic liver disease (MASLD)^3,4. Current treatment strategies are limited; therefore, it is important to elucidate the mechanisms involved in hepatic fibrosis to develop effective therapeutics.

Interferon regulatory factor 3 (IRF3) is ubiquitously expressed and critically regulates host defense against viral infection^5,6. Upon activation, IRF3 translocates to the nucleus and induces transcription of type I interferons and other anti-viral genes^5,7. IRF3 also executes non-transcriptional functions⁸, restraining nuclear factor κ-B (NFκB) activity by binding to the p65 subunit in the cytoplasm^9,10, and also translocating to the mitochondrial outer membrane where it promotes apoptosis through direct interaction with Bax^11,12. There is emerging evidence from murine models implicating both the transcriptional and non-transcriptional functions of IRF3 in the pathogenesis of metabolic liver diseases, including ALD^13–15 and MASLD^6,10,16–19. However, the contributions of IRF3 to the development of liver fibrosis are not clear, as global Irf3^−/− mice are reported to have either enhanced²⁰ or diminished²¹ carbon tetrachloride (CCl₄)-induced liver fibrosis. Furthermore, while the transcriptional functions of IRF3 have been the primary focus of investigation in models of fibrosis, its non-transcriptional activity has not been explored^10,13,16,18.

Cytoplasmic stress granules (SG) are large, organized, membrane-less conglomerates of untranslated mRNAs, stalled ribosomes, and RNA-binding proteins that form in response to various conditions of cellular stress²², including RNA virus infection²³. SGs serve as a form of liquid-liquid phase separation to limit translation until the stressful event has passed, at which point they disassemble. SG assembly requires the actions of core nucleating proteins such as G3BP SG assembly factor 1 (G3BP1), its functional homolog G3BP2, and cell cycle associated protein 1 (CAPRIN1)²³. Cytoplasmic SGs are most well-studied in the context of anti-viral defense and are implicated in the pathogenesis of multiple cancers and neurodegenerative diseases²⁴ but their potential involvement in fibrosis is unknown.

Herein, we interrogated the distinct contributions for the transcriptional and non-transcriptional activities of IRF3 and HSC-intrinsic IRF3 to HSC activation and CCl₄-induced hepatic fibrosis using in vivo and ex vivo approaches in Irf3-deficient (Irf3^−/−) mice, mice expressing only non-transcriptional IRF3 activity (Irf3^S1/S1), and mice with HSC-specific deletion of Irf3 (Lrat-Irf3), paralleled by complementary mechanistic investigations with immortalized human HSCs. We identified that the non-transcriptional activity of IRF3 enhances assembly of pro-fibrotic cytoplasmic SGs to drive HSC activation and ECM synthesis. We further provide evidence implicating upstream recognition of accumulated endogenous double-stranded RNA (dsRNA) by Toll-like receptor 3 (TLR3) in HSC activation and SG assembly. Finally, we found enhanced SG assembly within multiple hepatic cell types, including HSCs, in fibrotic human livers. Our findings demonstrate a critical role for the non-transcriptional activity of IRF3 in mediating profibrotic responses and provide a novel connection between intracellular stress responses, innate immune signaling pathways and fibrotic responses in HSCs.

METHODS

Murine models

All procedures using animals were approved by the Cleveland Clinic Institutional Animal Care and Use Committee. Male and female C57BL/6 (used as wild-type controls for all experiments), Tlr3^−/− and Irf3^fl/fl mice were from Jackson Laboratory (Bar Harbor, ME). Irf3^−/− and Irf3^S1/S1 mice, on a mixed C57BL/6J/N background, were obtained from Ganes Sen (Cleveland Clinic), Irf3^S1/S1 mice only express an Irf3 gene mutated at amino acids SS388/390AA, the essential phosphorylation sites required for IRF3 transcriptional activity^11,25. Lrat-CRE mice were provided by Robert Schwabe (Columbia University)²⁶. Mice were used in either the CCl₄-induced or the choline-deficient L-amino acid-defined (CDAA) model of liver fibrosis or to isolate primary hepatic stellate cells.

For the CCl₄-induced model of fibrosis, 10–12 week-old male and female C57BL/6, Irf3^−/− and Irf3^S1/S1 mice received intraperitoneal injections with CCl₄ or olive oil as the vehicle control twice-weekly for five weeks. CCl₄ was diluted 1:4 in olive oil. The initial dose was 0.25 μL/g body weight, the second was 0.5 μL/g body weight, followed by 6–8 additional doses at 1 μL/g body weight. Mice were euthanized for tissue collection 24, 48, and 72 hours after the final CCl₄ exposure.

For the choline-deficient L-amino acid-defined (CDAA) sufficient model of liver fibrosis, 5–6-week-old male C57BL/6, Irf3^−/− and Irf3^S1/S1 mice were housed 5 mice/cage in standard micro-isolator cages and maintained on a 14h:10h light–dark cycle. Mice and food intake per cage was measured weekly. Mice were weight matched and randomly distributed to Choline Sufficient Iron Supplemented L-amino acid defined (CSAA – 518754, Dyets Inc, Bethlehem, PA), or Choline Deficient Iron Supplemented L-amino acid defined (CDAA – 518753, Dyets Inc, Bethlehem, PA) diet groups and fed for 22 weeks. At the end of the 22-week feeding protocol, mice were fasted for 6 hours and then euthanized.

Primary murine hepatic stellate cell isolation and culture

Primary murine hepatic stellate cells (HSC) were isolated from adult chow-fed C57BL/6, Irf3^−/−, Irf3^S1/S1 and Tlr3^−/− mice (at least 24 weeks old), as previously described²⁷.

Human liver samples

De-identified paraffin-embedded liver tissue sections were provided by the NIAAA supported Clinical Resource for Alcohol Hepatitis Investigations at Johns Hopkins University (R24AA025017). Sections were obtained from explanted livers of patients with severe alcohol-associated hepatitis during liver transplantation, or wedge biopsies from the donor livers (normal control). Tissue collection from explanted livers or biopsies from donor livers was approved by Institutional Review Boards at Johns Hopkins Medical Institutions (IRB00107893 and IRB00021325).

Statistical analysis

Values in all figures represent mean ± standard error of the mean (SEM) unless otherwise indicated. Analysis of variance was performed using the general linear models procedure (SAS, Cary, NC). Data were log-transformed as necessary to obtain a normal distribution. Follow-up comparisons were made by least square means testing with p-values < 0.05 considered to be significant.

RESULTS

Contribution of the non-transcriptional activity of IRF3 in fibrotic liver injury in mice

While IRF3 is highly expressed in immune cells, it is also expressed by multiple cell types, including hepatocytes, endothelial cells, and HSCs (Human Protein Atlas proteinatlas.org). Data in the Single-cell Atlas of the Human Liver²⁸ showed increased IRF3 expression within hepatocytes, scar-associated macrophages, and myofibroblast subsets in cirrhotic livers compared to healthy controls, suggesting that IRF3 may play a multifaceted role in response to liver injury in different cell populations.

Phospho-IRF3 was increased in livers of C57BL/6 mice exposed to chronic CCl₄, detected in both hepatocytes (blue arrows) and non-parenchymal cells (red arrows) in livers (Figure 1A). In contrast, no signal was detected in either Irf3-deficient mice or Irf3^S1/S1 mice, which express an Irf3 gene mutated at amino acids (SS388/390AA), the essential phosphorylation sites required for IRF3 transcriptional activity (Figure 1A). Alpha-smooth muscle actin (α-SMA), a marker of HSC activation (Figure 1B), and Picrosirius red, which stains collagen fibrils (Figure 1C), accumulated in both C57BL/6 and Irf3^S1/S1 mice livers harvested 72 hours after the last chronic CCl₄ injection. In contrast, both α-SMA and Picrosirius red staining were lower in Irf3^−/− mice compared to the other genotypes. Genetic deficiency of Irf3 also protected against choline-deficient L-amino-defined (CDAA) diet-induced liver fibrosis, as indicated by reduced Picrosirius red, α-SMA, and type I collagen (COL1A1) staining as compared to both C57BL/6 and Irf3^S1/S1 mice (Supplemental Figure 1).

Fig. 1. The non-transcriptional activity of IRF3 mediates hepatic fibrogenesis in mice.

C57BL/6, Irf3^−/−, and Irf3^S1/S1 mice were exposed to CCl₄ or olive oil for five weeks with liver harvest 48 or 72 hours after the last dose. (A-C) Immunohistochemistry for (A) phospho-IRF3 in hepatocytes (blue arrows) and non-parenchymal cells (red arrows), (B) α-SMA or (C) Picrosirius Red staining was performed on paraffin-embedded liver sections (n = 3–4 for Olive Oil, n= 5–20 for CCl₄). (D) Hepatic expression of indicated mRNAs, with normalization to 18s rRNA, was assessed by qRT-PCR. Values represent means ± SEM of fold change relative to olive oil controls within each genotype (n = 9–22). Values with different alphabetical subscripts are significantly different from each other (p < 0.05), assessed by ANOVA.

To further investigate the contributions of the transcriptional and non-transcriptional activity of IRF3 to hepatic fibrosis, expression of fibrotic and inflammatory genes was measured. 48h after the last CCl₄ injection, expression of Acta2 (encodes α-SMA) and Col1a1 mRNA was higher in livers of C57BL/6 and Irf3^S1/S1 mice compared to Irf3^−/− mice. Similarly, expression of inflammatory markers associated with liver injury, including Tnfa, Il6, and Cxcl5, was higher in livers of C57BL/6 and Irf3^S1/S1 mice compared to Irf3^−/− mice. In contrast, expression of the ECM-degrading enzyme Mmp8 was elevated in Irf3^−/− mice compared to C57/BL6 or Irf3^S1/S1 mice (Figure 1D). In summary, these findings suggest that the non-transcriptional activity of IRF3 contributes to hepatic fibrogenesis in two murine models, chronic CCl₄- and CDAA-induced, potentially through enhanced HSC activation.

HSC-intrinsic IRF3 function in CCl₄-induced liver fibrosis

Activation of HSC is a critical initiating step in the development of fibrosis. Since the Irf3-deficient mice failed to increase expression of α-SMA in response to chronic CCl4 or CDAA diet-induced fibrosis, we hypothesized the HSC-intrinsic activity of IRF3 contributed to HSC activation. To test this hypothesis, HSC-specific Irf3-deficient mouse (Lrat-Irf3) were generated and exposed to chronic CCl₄. Compared with Irf3^fl/fl controls, Lrat-Irf3 exhibited attenuated fibrosis, with decreased staining of both α-SMA (Figure 2A) and Picrosirius red (Figure 2B). These results support an important role for IRF3 expression in HSCs during the progression of fibrotic injury.

Fig. 2. HSC-intrinsic IRF3 promotes hepatic fibrogenesis in mice.

Irf3^fl/fl and Lrat-Irf3 mice were exposed to CCl₄ or olive oil for five weeks with liver harvest 48 or 72 hours after the last dose. (A-B) Immunohistochemistry for (A) α-SMA and (B) Picrosirius Red staining was performed on paraffin-embedded liver sections. All values represent means ± SEM (n=3–10 for Olive Oil, n=3–5 for CCl₄). Values with different alphabetical subscripts are significantly different from each other (p<0.05), assessed by ANOVA.

Transcriptional and non-transcriptional functions of IRF3 in primary murine HSC

Our discovery of the importance of HSC-intrinsic activity of IRF3 to fibrogenesis prompted us to examine the transcriptional and non-transcriptional functions of IRF3 specifically in HSCs. Primary HSCs were isolated from C57BL/6, Irf3^−/−, Irf3^S1/S1 mice and differentiated on plastic over 10 days in culture (Supplemental Figure 2). On D5 and D10 of differentiation, bulk RNA-seq was conducted; Reactome based pathway enrichment analysis revealed multiple ECM-related pathways (highlighted in red boxes) were upregulated in C57BL/6 HSC on D10 compared to D5, consistent with differentiation in culture (Figure 3A). Notably, several pathways related to ECM organization and collagen chain trimerization, biosynthesis, and formation were downregulated in HSCs from Irf3^−/− mice as compared to C57BL/6 mice (Figure 3B). In contrast, these pathways were upregulated in HSC from Irf3^S1/S1 mice compared to both Irf3^−/− and C57BL/6 mice (Figure 3C and D). In particular, the ECM organization pathway, which includes genes involved in synthesis, crosslinking and modulation of ECM components, was consistently ranked at the top among all significantly regulated pathways. Interestingly, several ECM degradation and remodeling pathways (highlighted in blue boxes) were also upregulated in Irf3^S1/S1 HSCs compared to both Irf3^−/− and C57BL/6 HSCs (Figure 3C and D), suggesting a coordinated activation of both ECM production and turnover processes, as well as dynamic remodeling in Irf3^S1/S1 cells.

Fig. 3. The non-transcriptional activity of IRF3 is pro-fibrotic in primary HSCs.

(A-D) RNA was isolated from D5 or D10 cultured genotype-specific primary HSCs for bulk RNA-seq analysis (n = 3). Reactome pathway enrichment analysis of the top 20 significantly affected pathways (p < 0.05) are illustrated for (A) pathways increased in D10 C57BL/6 compared to D5 C57BL/6, (B) pathways decreased in D10 Irf3^−/− compared to D10 C57BL/6, (C) pathways increased in D10 Irf3^S1/S1 compared to D10 Irf3^−/− and (D) pathways increased in D10 Irf3^S1/S1 compared to D10 C57BL/6. Pathways highlighted in red are associated with fibrogenesis; pathways in blue with resolution of fibrosis. (E-H) Expression of indicated mRNAs in primary HSCs was assessed by qRT-PCR and normalized to 18s rRNA. Values represent means ± SEM of fold change relative to D2 cells within each genotype (n = 4–25). Values with different alphabetical subscripts are significantly different from each other (p < 0.05), assessed by ANOVA.

The results from bulk RNA-seq provide an unbiased perspective of the contributions of the Irf3 gene to HSC differentiation. We further examined the changes in expression of ECM production-related mRNAs during differentiation of primary HSCs isolated from C57BL/6, Irf3^−/− and Irf3^S1/S1 mice. Expression of Acta2 (Figure 3E), and Col3a1 (Figure 3G), a major structural component of ECM, mRNA was increased during differentiation in all three genotypes. However, by D10, HSC from Irf3^−/− mice expressed less Acta2 (Figure 3F) and Col3a1 (Figure 3H) mRNA compared to HSCs from C57BL/6 or Irf3^S1/S1 mice. Taken together, these results indicate that the non-transcriptional activity of IRF3 contributes to increased HSC activation and ECM production during differentiation.

Identification of a pro-fibrotic dsRNA-TLR3-IRF3 axis in primary HSCs and LX2 cells

We next assessed for genotype-specific differences in responses of primary HSCs to TGFβ, a prototypical pro-fibrotic cytokine. TGFβ increased expression of Acta2 and Col1a1 mRNA in primary HSCs from C57/BL6 and Irf3^S1/S1mice, but not Irf3^−/− mice (Figure 4A). Because IRF3 is activated via multiple upstream nucleic acid sensors, including TLR3, primary HSCs were challenged with the double-stranded RNA (dsRNA) analogue polyinosinic-polycytidylic acid [poly(I:C)], a prototypical activator of TLR3. Poly(I:C) increased expression of Acta2 mRNA in HSCs from C57BL/6 and Irf3^S1/S1, but not Irf3^−/−, mice (Figure 4B). As expected, poly(I:C) increased expression of Ifnb1 (encodes interferon-β) mRNA in HSCs from C57BL/6 mice but not from Irf3^−/− or Irf3^S1/S1 mice (Figure 4B). Further, poly(I:C)-stimulated expression of Acta2 and Ifnb1 mRNA was diminished in HSCs isolated from Tlr3^−/− mice compared to HSCs from C57BL/6 mice (Figure 4C). Interestingly, expression of Acta2 and Col1a1 mRNA was also decreased in HSCs from Tlr3^−/− mice even without stimulation with exogenous ligand (Figure 4D). This suggested that endogenous dsRNA was accumulating during HSC differentiation. Indeed, dsRNA increased in HSC from C57BL/6 during differentiation, detected using a sequence agnostic anti-dsRNA monoclonal antibody (J2) (Figure 4E). Taken together, these results indicate sensing of accumulated endogenous dsRNA by TLR3 induces activity of IRF3 and enhances HSC activation.

Fig. 4. The dsRNA-TLR3-IRF3 axis promotes fibrotic responses in primary HSCs and LX2 cells.

(A-D) Genotype-specific expression of indicated mRNAs was assessed by qRT-PCR in primary HSCs that were either stimulated for 24 hours with 5 ng/mL TGFβ on D10 of culture (A) or for 5 hours with poly(I:C) on D5 of culture (B,C) or unstimulated (D). Fold change is relative to unstimulated D10 C57BL/6 cells (A; n=5–8) or D2 cells within each genotype (B-D, n=5–25). (E) Immunofluorescence with anti-dsRNA antibody (J2) and DAPI nuclear stain was performed in unstimulated D2 and D10 primary HSCs from C57BL/6 mice. Top-down views of 3D reconstructions of Z-stacks and fold change of dsRNA fluorescence intensity normalized to nuclear volume relative to unstimulated D2 cells are shown (n=3). (F-I) LX2 cells transfected with scrambled control siRNA or indicated siRNA were stimulated with 5 ng/mL TGFβ for 24h, then expression of indicated mRNAs was assessed by qRT-PCR (F,H,I; n=3–6) and Type I collagen in conditioned media by Western blot (G; n=3). (J) Immunofluorescence with anti-dsRNA antibody (J2) and DAPI nuclear stain was performed in LX2 cells stimulated with 5 ng/mL TGFβ for 16h. Top-down views of Z-stacks analyzed as in E are shown (n=11). Fold change is relative to unstimulated (E,F,H-J) or TGFβ-stimulated (G) control cells. All qRT-PCR data were normalized to 18S/18s rRNA. All values depict means ± SEM, and values with different alphabetical subscripts are significantly different from each other (p<0.05), assessed by ANOVA.

In order to dissect the molecular mechanisms for the profibrogenic role of IRF3 in HSCs, we interrogated the role of IRF3 in the fibrogenic activity of LX2 cells. Gene silencing of IRF3 with short interfering RNA (siRNA) in LX2 cells decreased TGFβ-induced expression of fibrogenic genes, including COL1A1 and ACTA2 (Figure 4F), and accumulation of type I collagen in conditioned media (Figure 4G), as compared to control cells transfected with scrambled siRNA. In contrast, siRNA knock down of the extracellular receptor for type I interferons (IFNAR) did not impact TGFβ-induced expression of COL1A1 mRNA (Figure 4H). Effective IFNAR gene-silencing was achieved as expression of CCL2, a known IFN receptor-dependent gene, was reduced (Figure 4H). These data are consistent with an important contribution of the non-transcriptional activity of IRF3 in the fibrogenic response to TGFβ.

One previous report implicated cyclic GMP-AMP synthase(cGAS)-stimulator of interferon genes (STING) DNA-sensing pathway in IRF3-mediated fibrotic responses to TGFβ in LX2 HSCs²⁹, yet our data in primary murine HSCs indicated that TLR3 was also a key nucleic acid sensor in HSCs. Therefore, we next assessed the potential for engagement of multiple nucleic acid sensors upstream of IRF3 in HSCs. LX2 cells were transfected with siRNA directed against 1) TLR3, 2) mitochondrial antiviral signaling protein (MAVS), the convergent downstream sensor for dsRNA sensing by retinoic acid-inducible gene I (RIG-I) and melanoma differentiation-associated protein 5 (MDA5), 3) myeloid differentiation factor 88 (MYD88), the downstream adaptor for single-stranded RNA sensing by TLR7/8, and 4) stimulator of interferon genes (STING), the downstream adaptor for double-stranded DNA sensing by cGAS. Gene silencing of STING, TLR3, and MAVS, but not MYD88, reduced TGFβ-induced COL1A1 mRNA expression (Figure 4I). However, gene silencing of TLR3, but not MAVS or STING, inhibited TGFβ-induced accumulation of type I collagen in conditioned media (Figure 4G). Notably, TGFβ stimulation increased accumulation of dsRNA in LX2 cells (Figure 4J), similar to the accumulation of dsRNA during differentiation of primary HSCs (Figure 4E) and consistent with a role for dsRNA sensors mediating TGFβ-induced fibrogenic responses.

Localization of IRF3 to dsRNA-induced cytoplasmic SGs in LX2 cells

Untargeted proteomics of immunoprecipitated IRF3 from unstimulated LX2 cells revealed over 500 proteins (Supplemental Table 1) in the immunoprecipitate, including many DNA and RNA-binding proteins. The most abundant proteins included histones, polyadenylate-binding proteins, and the DNA polymerase delta catalytic subunit. Notably, none of the known IRF3 binding partners, such as Bax¹² or p65⁹, involved in IRF3’s non-transcriptional functions were detected. Gene ontology enrichment analysis identified pathways involved in metabolism of RNA and cellular responses to stress (Figure 5A). Unexpectedly, several cytoplasmic SG component proteins were detected, including CAPRIN1, G3BP1, and G3BP2 (Figure 5A), suggesting that IRF3 is present within cytoplasmic SGs in HSCs.

Fig. 5. TGFβ induces assembly of IRF3-containing SGs in LX2 cells.

(A) Gene Ontology enrichment analysis of IRF3 binding proteins identified by mass spectrometry in lysates of unstimulated LX2 cells. Dark orange bars indicate P<10⁻²⁰, and light orange bars indicate p < 10⁻¹⁰. Example SG component proteins are highlighted. (B-J) Immunofluorescence was performed in LX2 cells stimulated with 5 ng/mL of TGFβ for 16h. 3D reconstructions of Z-stacks were created with IMARIS, and the Surfaces module was used to depict nuclei (stained with DAPI) and G3BP1⁺ or CAPRIN1⁺ SGs. Subpanels B1a/2a, E1a/2a, and H1a/2a are representative top-down views showing DAPI (blue), G3BP1 or CAPRIN1 (red), and total dsRNA or IRF3 (green), and subpanels B1b/2b, E1b/2b, and H1b/2b show nuclei (blue), G3BP1⁺ or CAPRIN1⁺ SGs (yellow), and dsRNA or IRF3 inside (magenta) and outside (gray) of G3BP1⁺ or CAPRIN1⁺ SGs. Panels C, F, and I are higher-magnification top-down and oblique views of insets from subpanels B2b, E2b, and H2b. For orientation purposes, certain nuclei are labeled with asterisks. (D,G,J) Quantification of dsRNA (D), G3BP1 (G), CAPRIN1 (J), and IRF3 (G,J) fluorescence inside G3BP1⁺ (D,G) or CAPPRIN1⁺ (J) SGs normalized to nuclear volume. All values depict means ± SEM of fold change relative to unstimulated (D) or TGFβ-stimulated (G,J) cells (n=4–8). Values with different alphabetical superscripts are significantly different from each other (p<0.05), assessed by ANOVA.

Focal cytoplasmic G3BP1 aggregates, indicative of SG assembly²², markedly increased with transfection with poly(I:C), a prototypical inducer of SG assembly³⁰, but not in mock-transfected LX2 cells (Supplemental Figure 3A subpanels 1a/2a,C). Poly(I:C) transfection also induced punctate cytoplasmic aggregation of IRF3 (Supplemental Figure 3A subpanels 1a/2a), which was confirmed to be inside G3BP1⁺ SGs (Supplemental Figure 3A subpanels 1b/ 2b,C) using 3D reconstructions and masking of the IRF3 fluorescence channel with IMARIS software. Furthermore, IRF3 had a punctate distribution within G3BP1⁺ SGs (Supplemental Figure 3B subpanels 2c/2d). Notably, only occasional cytoplasmic G3BP1⁺ or CAPRIN1⁺ SGs were identified at baseline (Figure 5B/E/H subpanels 1a/1b), whereas challenge with TGFβ strongly induced cytoplasmic aggregation of both G3BP1 and CAPRIN1 (Figure 5B/E/H subpanels 2a/2b,G,J). G3BP1⁺ SGs contained dsRNA (Figure 5B–D). Furthermore, IRF3 was detectable throughout G3BP1⁺ and CAPRIN1⁺ SGs (Figure 5E/H subpanels 1b/2b,G,J). Punctate areas of higher IRF3 aggregation were visible both inside and immediately surrounding SGs in high-magnification views (Figure 5F,I). TGFβ-induced G3BP1+ SGs also expressed high amounts of the canonical SG marker Fragile X messenger ribonucleoprotein 1 (FMR1) (Supplemental Figure 3D), and cytoplasmic IRF3⁺ aggregates contained T-cell intracellular antigen-1 (TIA1) (Supplemental Figure 3E). In total, these findings identify TGFβ as a novel inducer of the assembly of IRF3-containing SGs.

Role of IRF3 and TLR3 in TGFβ-induced SG assembly in primary HSCs and LX2 cells

Because IRF3 was detected within TGFβ-induced SGs, we next determined the requirement for IRF3 in their assembly. Gene silencing of IRF3 with siRNA decreased TGFβ-induced assembly of G3BP1⁺ SGs (Figure 6A,B). Notably, TLR3 gene silencing in LX2 cells phenocopied IRF3 gene silencing (Figure 6A,B), consistent with TLR3 being the predominant IRF3 activator under these conditions.

Fig. 6. The TLR3-IRF3 pathway mediates assembly of pro-fibrotic SGs in LX2 cells.

(A,B) Immunofluorescence was performed in LX2 cells transfected with scrambled control siRNA or indicated siRNA and stimulated with 5 ng/mL TGFβ for 16h. 3D reconstructions of Z-stacks were created with IMARIS and analyzed using the Surfaces module as in Fig. 4. The first and third columns are representative top-down views showing DAPI (blue) and G3BP1 (red). The second and fourth columns are representative top-down views showing nuclei (blue), and G3BP1⁺ SGs (yellow). (B) Quantification of G3BP1 fluorescence inside G3BP1⁺ SGs normalized to nuclear volume (n=3–4). (C-F) LX2 cells were pre-treated for one hour with 25 μM EGCG or vehicle (C-D) or transfected with scrambled control siRNA or siRNAs against G3BP1 and G3BP2 (E,F) and then stimulated with 5 ng/mL TGFβ for 24h. Expression of indicated mRNAs was assessed by qRT-PCR with normalization to 18S rRNA (C,E; n=3) and Type I collagen and fibronectin in conditioned media by Western blot (D,F; n=3). All values represent means ± SEM of fold change relative to unstimulated (C,E) or TGFβ-stimulated (B,D,F) control cells, and values with different alphabetical subscripts are significantly different from each other (p<0.05), assessed by ANOVA.

We next determined whether SG formation was required to mediate the fibrotic responses to TGFβ using both pharmacologic and genetic approaches. Pharmacologic inhibition of G3BP1 activity with (−)-epigallocatechin gallate (EGCG)^31–33 decreased mRNA expression of the fibrogenic genes COL1A1 and FN1 in LX2 cells (Figure 6C). Dual gene silencing of G3BP1 and its functional homolog G3BP2 also suppressed TGFβ-induced ACTA2, COL1A1 and FN1 mRNA expression in LX2 cells (Figure 6E). Moreover, both EGCG pre-treatment and G3BP1/2 gene silencing decreased TGFβ-induced accumulation of type I collagen and fibronectin in conditioned media in LX2 cells (Figure 6D,F).

After demonstrating that IRF3 mediate SG assembly and that SGs promoted TGFβ-induced fibrotic responses in LX2 cells, we assessed whether these findings generalized to primary HSCs and tested the requirement for the non-transcriptional activity of IRF3 in SG assembly. Pre-treatment with EGCG decreased TGFβ-induced Acta2, Col1a1, and Fn1 in primary murine HSCs culture from C57BL/6 mice (Figure 7A). Furthermore, HSCs cultured from Irf3^−/− mice exhibited decreased SG assembly and accumulation of dsRNA inside of SGs (Figure 7B,C), similar to the impact of IRF3 gene silencing in LX2 cells (Figure 6). In contrast, TGFβ induced robust SG assembly and accumulation of dsRNA inside the SGs in primary HSCs cultured from Irf3^S1/S1 mice (Figure 7B,C), equivalent to the response of HSCs from C57BL/6 mice. Taken together, these results identify the non-transcriptional activity of IRF3 as an important mediator of TGFβ-induced SG assembly.

Fig. 7. The non-transcriptional activity of IRF3 mediates SG assembly.

(A) D10 primary HSCs cultured from C57BL/6 mice were pre-treated for one hour with 25 μM EGCG or vehicle and then stimulated with 5 ng/mL TGFβ for 24h. Expression of indicated mRNAs was assessed by qRT-PCR with normalization to 18s rRNA (n=3). (B,C) Immunofluorescence was performed in D10 primary HSCs cultured from C57BL/6, Irf3^−/−, and Irf3^S1/S1 mice stimulated with 5 ng/mL of TGFβ for 16h. 3D reconstructions of Z-stacks were created with IMARIS and analyzed using the Surfaces module as in Fig. 4. The first and third columns are representative top-down views showing DAPI (blue), CAPRIN1 (red), and total dsRNA (green). The second and fourth columns are representative top-down views showing nuclei (blue), CAPRIN1⁺ SGs (yellow), and dsRNA inside (magenta) and outside (gray) of CAPRIN1⁺ SGs. (C) Quantifications of CAPRIN1 (top) and dsRNA fluorescence inside CAPRIN1⁺ SGs (bottom) normalized to nuclear volume (n=3). All values depict means ± SEM of the fold change relative to unstimulated (A) or TGFβ-stimulated (B) C57BL/6 cells. Values with different alphabetical superscripts are significantly different from each other (p<0.05), assessed by ANOVA.

TGFβ-induced SG assembly in human induced pluripotent stem cells-derived HSCs

Differentiation of human induced pluripotent stem cells (iPSCs) into HSCs (hiHSCs) is a cutting-edge experimental modality to faithfully model human HSC response in vitro. Therefore, we assessed whether TGFβ triggers the assembly of IRF3-containing SGs in hiHSCs. TGFβ-induced robust G3BP1⁺ SG assembly and IRF3 aggregation inside the G3BP1⁺ SGs (Figure 8A–C) in hiHSCs. This data is consistent with our results obtained in LX2 cells (Figure 5).

Fig. 8. SGs are induced in human iPSC-derived HSCs and livers under fibrotic conditions.

(A-C) Immunofluorescence was performed in iPSC-derived HSCs stimulated with 5 ng/mL of TGFβ for 16h. 3D reconstructions of Z-stacks were created with IMARIS and analyzed using the Surfaces module as in Fig. 4. Subpanels A1a/2a are representative top-down views showing DAPI (blue), G3BP1 (red), and IRF3 (green), and subpanels A1b/2b show nuclei (blue), G3BP1⁺ SGs (yellow), and IRF3 inside (magenta) and outside (gray) of G3BP1⁺ SGs. Panel B shows higher-magnification top-down and oblique views of inset from subpanel A2b. For orientation purposes, certain nuclei are labeled with asterisks. (C) Quantification of G3BP1 and IRF3 fluorescence inside G3BP1⁺ SGs normalized to nuclear volume. All values depict means ± SEM of fold change relative to TGFβ-stimulated cells (n=3–4). (D) Representative images of immunofluorescence performed with anti-α-SMA (green) and anti-G3BP1 (red) antibodies in fixed liver tissue obtained from patients with severe alcohol-associated hepatitis (AH; n=4) or healthy donors (HC; n=4). White arrowheads indicate co-localization of α-SMA and G3BP1. Values with different alphabetical superscripts are significantly different from each other (p<0.05), assessed by ANOVA.

Enhanced SG assembly in fibrotic human livers

Given the increased formation of SGs in isolated HSCs, we next tested the hypothesis that enhanced SG assembly would be present in fibrotic human livers. Immunostaining for G3BP1 and α-SMA was performed in liver samples from patients with severe alcohol-associated hepatitis (sAH) undergoing liver transplant and healthy donors. G3BP1 aggregation, indicative of SG assembly, was increased across the parenchyma in livers from patients with sAH (Figure 8D). In contrast, minimal G3BP1 aggregation was observed in the livers from healthy donors. Consistent with our cell based data, G3BP1 immunostaining co-localized with α-SMA (Figure 8D). These findings demonstrate that SG assembly occurs in multiple hepatic cell types, including HSCs, in fibrotic liver disease.

DISCUSSION

Hepatic stellate cells are the primary driver of hepatic fibrogenesis in response to chronic liver injury. The pathways that regulate HSC activation and fibrogenesis are not completely understood. In particular, there is a growing appreciation that TLR-mediated innate immune signaling drives HSC activation in response to injury². For example, TLR4 agonists enhance HSC activation³⁴ and contribute to progression of fibrosis in patients with HCV³⁵. Our data build upon prior findings of a critical role for TLR3 in CCl₄-induced liver fibrosis by identifying TLR3 as the predominant sensor of accumulated dsRNA in HSCs, leading to IRF3 activation. Mechanistically, the non-transcriptional activity of IRF3 stimulated SG assembly and fibrogenic gene expression.

Although the role of IRF3 is well described in viral hepatitis, limited studies have examined the role of IRF3 in non-viral mediated liver fibrosis, and the specific contributions of the transcriptional and non-transcriptional activities of IRF3 to fibrotic injury have not been delineated. Irf3^S1/S1 mice, which lack the essential phosphorylation sites required for the transcriptional function of IRF3, provide a unique opportunity to interrogate the role of the non-transcriptional activity of IRF3 to hepatic fibrosis. Unlike Irf3^−/− mice, Irf3^S1/S1 mice developed similar levels of hepatic inflammation and fibrosis as C57BL/6 mice in two murine models of liver fibrosis. Moreover, our molecular dissection of the fibrotic responses of both primary HSCs and LX2 cells confirmed the critical role for the non-transcriptional activity of IRF3 in fibrogenic responses of HSC. Results from both the primary HSCs and LX2 cells are consistent with a non-transcriptional mechanism of action for IRF3 in fibrotic responses in HSCs. These results, together with our in vivo studies using Lrat-Irf3 mice, define an important role for HSC-intrinsic IRF3 in hepatic fibrogenesis.

Our investigation also identified an important dsRNA-TLR3-IRF3-SG pro-fibrogenic axis in HSCs. SG are increasingly appreciated as important contributors to diverse physiological and pathological processes²³. However, to the best of our knowledge, this is the first report of a pro-fibrotic role for SGs and of TGFβ as an inducer of their assembly. Both pharmacologic and genetic blockade strategies were employed in our study to demonstrate that stress granules promote in vitro fibrotic responses to TGFβ. The in vitro effects of EGCG were phenocopied by dual gene silencing of both G3BP1 and its functional homolog G3BP2. While we show that HSC-intrinsic IRF3 is important for in vivo hepatic fibrogenesis, we have not specifically assessed the contribution of SG in this process because global SG-deficiency is embryonically lethal. Generation of conditional G3BP1-deficient mice will be necessary to delineate the specific contributions of SGs to hepatic fibrogenesis. Further investigation is also necessary to clarify the specific mechanisms by which SGs promote the fibrotic response to TGFβ. One mechanism could be the stabilization of fibrogenic mRNAs as SGs can modulate mRNA stability²³. However, when TGFβ-stimulated LX2 cells were exposed to a pharmacologic inhibitor of SGs that rapidly dissolves preformed SGs³⁶ (G3Ia) in combination with Actinomycin D to inhibit transcription, the half-life of fibrogenic mRNAs was not changed (Supplemental Figure 6). Possible alternative mechanisms for the contribution of SGs to increased fibrogenic responses include serving as a platform for enhanced activation of nucleic acid sensors and sequestering intracellular signaling molecules.

TGFβ-induced SG assembly required IRF3 expression in both primary HSCs and LX2 cells. Importantly, SG assembly in primary HSCs from Irf3^S1/S1 mice was equivalent to HSC from wild-type mice, demonstrating that the non-transcriptional activity of IRF3 was required for SG assembly. Further study will be required to determine whether these findings represent a novel application of the known non-transcriptional functions of IRF3 related to interaction with NFκB and/or Bax or due to a different molecular mechanism. Of note, phospho-IRF3 can directly interact with the TGFβ signaling molecule mothers against decapentaplegic homolog 3 (SMAD3)³⁷ in other cell types to suppress its activity, but the strong fibrogenic responses seen in primary HSCs cultured from Irf3^S1/S1 mice to TGFβ make this potential interaction unlikely to be relevant.

Our mass spectrometry-based proteomic analysis of IRF3-interacting proteins adds SG component proteins to the list of known extranuclear IRF3 binding partners like Bax^11,12, the p65 subunit of NFκB⁹, and SMAD3³⁷. The presence of IRF3 within SGs, rather than focal non-SG aggregates like G3BP1-cGAS complexes^32,33, is supported by the findings that (1) IRF3 was localized to focal aggregates of multiple bonafide SG proteins and (2) TGFβ-induced G3BP1 aggregates contained dsRNA.

Previous work established the importance of the cGAS-STING pathway in hepatic fibrosis and provides a framework for the efficient pharmacological targeting of IRF3^14,18,20,29. Our findings broaden this insight to include contributions of the TLR3-IRF3 axis in HSCs to fibrogenesis. While previous work demonstrated that TLR3-mediated activation of IRF3 in HSC in response to poly(I:C) regulates viral replication in hepatocytes³⁸, our data identifies an important role for TLR3-IRF3 signaling in HSCs in response to non-viral fibrotic signals. Our results are also consistent with prior studies demonstrating that Tlr3^−/− mice exhibit decreased CCl₄-induced liver fibrosis³⁹ and that exposure of cultured HSCs to hepatocyte-derived exosomes³⁹ or neutrophil extracellular traps⁴⁰ activates TLR3 signaling and fibrogenic gene expression.

Importantly, we demonstrate that dsRNA accumulated both during differentiation of primary HSCs in culture or in response to TGFβ in LX2 cells. Accumulated dsRNA was the trigger for TGFβ-induced SG assembly, and TLR3 was the critical RNA sensor. Further investigation is necessary to clarify whether this dsRNA originates from the nucleus as a results of transcription of repetitive elements or activation of endogenous retroviruses⁴¹, and/or from mitochondria⁴². Notably, dsRNA derived from both sources can activate TLR3^43,44. TGFβ also induces accumulation of mitochondrial DNA to activate the cGAS-STING-IRF3 axis in HSCs²⁹, and we find that siRNA knock-down of STING decreased fibrogenic gene expression in LX2 cells. Taken together, these data suggest that multiple nucleic acid sensing pathways are activated in parallel in LX2 cells in response to TGFβ.

Taken together, our investigation led to the discovery that the non-transcriptional activity of IRF3 contributes to fibrotic injury by modulating HSC activities through enhanced SG assembly in response to accumulated endogenous dsRNA. Given the ubiquitous nature of IRF3 and SGs, it will be interesting in future studies to determine the contributions of this novel dsRNA-TLR3-IRF3-SG axis to other fibrotic and non-fibrotic disease states.

Impact and implications.

In this study, we defined that IRF3 promotes sterile hepatic fibrogenesis independently of canonical type I interferon production and uncovered a novel profibrogenic dsRNA-TLR3-IRF3-SG axis. Our findings clarify the relative contributions of transcriptional and non-transcriptional IRF3 activities to hepatic fibrogenesis and identify SG assembly as a potential therapeutic target for prevention of IRF3-mediated liver fibrosis.

Highlights.

• The non-transcriptional activity of the anti-viral protein IRF3 is a key driver of hepatic fibrosis in murine models

• Accumulation of endogenous dsRNA in HSCs signals via TLR3 to activate IRF3 and stimulate the formation of cytoplasmic stress granules

• Formation of stress granules is required for TGFβ-stimulated pro-fibrotic responses in HSCs

• Targeting the non-transcriptional activity of IRF3 and SG assembly are potential therapeutic strategies for fibrotic liver diseases.