IRF3 activates RB to authorize cGAS-STING–induced senescence and mitigate liver fibrosis

Abstract

Cytosolic double-stranded DNA surveillance by cyclic GMP-AMP synthase (cGAS)–Stimulator of Interferon Genes (STING) signaling triggers cellular senescence, autophagy, biased mRNA translation, and interferon-mediated immune responses. However, detailed mechanisms and physiological relevance of STING-induced senescence are not fully understood. Here, we unexpectedly found that interferon regulatory factor 3 (IRF3), activated during innate DNA sensing, forms substantial endogenous complexes in the nucleus with retinoblastoma (RB), a key cell cycle regulator. The IRF3-RB interaction attenuates cyclin-dependent kinase 4/6 (CDK4/6)–mediated RB hyperphosphorylation that mobilizes RB to deactivate E2 family (E2F) transcription factors, thereby driving cells into senescence. STING-IRF3-RB signaling plays a notable role in hepatic stellate cells (HSCs) within various murine models, pushing activated HSCs toward senescence. Accordingly, IRF3 global knockout or conditional deletion in HSCs aggravated liver fibrosis, a process mitigated by the CDK4/6 inhibitor. These findings underscore a straightforward yet vital mechanism of cGAS-STING signaling in inducing cellular senescence and unveil its unexpected biology in limiting liver fibrosis.

Directly activating RB, IRF3 mediates cGAS-STING signaling–induced cellular senescence and regulates liver fibrosis.

INTRODUCTION

Senescence is a cellular state that irreversibly and stably arrests the cell cycle in response to various intrinsic or extrinsic stresses, such as telomere shortening, chromosomal instability, DNA damage, oxidative stress, and oncogenic and mitogenic stimulations (1, 2). Senescent cells characterize morphological and biochemical phenotypes such as dilated shapes, high lysosomal enzyme β-galactosidase, and elevated expression of cell cycle inhibitors p16^INK4a and p21^Cip1/Waf1 (3, 4). Another notable change associated with cellular senescence is senescence-associated secretory phenotype (SASP), the secretory program ranging from proinflammatory cytokines, chemokines, and growth factors to metalloproteinases (5). These nonproliferative cells can modify cellular microenvironments via SASPs to cause chronic and low-grade inflammation. Senescent cells are usually briefly present, cleared by immune cells attracted by SASPs (1, 6). However, their accumulation in tissues and organs leads to the deterioration of organ function and probably contributes to the progression of age-associated diseases, including osteoarthritis, atherosclerosis, pulmonary fibrosis, sarcopenia, glaucoma cataracts, and type 2 diabetes mellitus (7–9). Pathologies associated with cellular senescence are also involved in neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, and amyotrophic lateral sclerosis (10–12), probably cause anxiety, and impair neurogenesis (13). Therefore, senolytics (senescence-destroying) candidates, which trigger the death of senescent cells, have shown phenotypes of improving age-related disease and increasing longevity in mice (14, 15). However, cellular senescence could function as a beneficial factor in a variety of human diseases. For example, the attenuation of senescence for activated hepatic stellate cells (HSCs) exacerbates liver fibrosis (16–18). Therefore, understanding their molecular bases and beneficially manipulating these processes can be a promising avenue in treating some severe human diseases.

DNA damage response, p53-p21^Cip1/Waf1, and p16^INK4a–retinoblastoma (RB) serve as key regulatory pathways underlying cellular senescence (19). Briefly, DNA damage response activates p38^MAPK and ataxia telangiectasia–mutated that stabilizes the tumor suppressor p53, leading to an up-regulation cyclin-dependent kinase inhibitor CDKN1A/p21^Cip1/Waf1 and CDK2 inhibition. Most senescent cells also activate the CDKN2A/p16^INK4a locus, which up-regulates the expression of adenosine 5′ddiphosphate ribosylation factor and p16^INK4a and thus inhibits CDK4/6 (19, 20). By contrast, suppression of CDK2/4/6 results in a hypophosphorylation and active state of RB, which associates with and inactivates the E2 family (E2F) transcription factors, arresting cell cycle at the G₁ (pre–DNA synthesis) to S (DNA synthesis) checkpoint (21–23). The regulation of SASPs might be further elusive. Diverse SASP programs are detected in senescent cells with distinct origins, primarily driven by stress-induced p38^MAPK and nuclear factor κB (NF-κB) pathways, and probably modulated at mRNA translation level by mammalian target of rapamycin (24, 25).

Recently, an innate immune pathway that senses mislocalized double-stranded DNA, the cyclic GMP-AMP synthase (cGAS)–Stimulator of Interferon Genes (STING) (cGAS-STING) cascade, was reported to drive cellular senescence robustly (26–29). Innate nucleic acid sensor cGAS monitors invading microbes and tissue injury by detecting cytosolic double-stranded DNA (30), regardless of whether it originates from pathogens, chromosomes, or mitochondria (31). Via the second messenger 2′3’-cyclic guanosine monophosphate-adenosine monophosphate synthesized by cGAS (32), endoplasmic reticulum–associated protein STING [also known as Mediator of IRF3 Activation (MITA) or Endoplasmic Reticulum IFN Stimulator (ERIS)] (33) is activated and translocated to the Golgi apparatus, where it assembles classical STING signalosomes to phosphorylate and mobilize interferon regulatory factor 3 (IRF3) (34, 35), a central signaling mediator of the cGAS-STING cascade. IRF3 then translocates as a homodimer into the nucleus, where it functions as a transcription factor to transcribe type I interferons (IFN-Is) and numerous IFN-stimulated genes (ISGs) (36–38), directly or indirectly and coordinating with simultaneously activated NF-κB (39). Innate DNA sensing by cGAS-STING-IRF3 signaling establishes an immune state in immune and nonimmune cells to restrict microbial infection, modulate adaptive immunity, and trigger tissue repair and regeneration (40, 41). Either compromisation or hyperactivation of cGAS-STING signaling frequently causes infectious diseases, autoimmune and autoinflammatory diseases, and cancers. Dysregulation of cellular senescence might partially contribute to STING-triggered pathologies (26). Besides, transcription-independent roles of IRF3 have been well recognized now, such as intervening in transforming growth factor–β (42) and NF-κB signaling (43), triggering protein condensation (44), and engaging in apoptotic cascade (45). The discovery of cGAS-STING–driven cellular senescence represents one of the substantial advances in senescence research. However, the underlying mechanism for STING-induced senescence, albeit firmly, is less known, except for our recent observation of a noncanonical cGAS-STING–PKR-like endoplasmic reticulum kinase (PERK) pathway that controls cellular senescence at the mRNA translation level (46). Deciphering the molecular network of senescence resulting from cGAS-STING signaling might provide an intriguing probe to understand this fundamental cellular process.

We found that IRF3, a classic signaling mediator of cGAS-STING signaling, is critical in developing cellular senescence in a variety of cells. The C-terminal phosphorylated IRF3 formed endogenous complexes with tumor suppressor RB in the nucleus, which disrupted CDK4/6-cyclin-RB complexes and thus maintained RB at a hypophosphorylation and active state. Analyses of fibrotic models from global or conditional IRF3 knockout (KO) mice revealed a profound pathological role of this cGAS-STING-IRF3-RB axis in the senescence of HSCs and the development of liver fibrosis. These findings collectively indicate the presence of an endogenous IRF3-RB complex that dominates RB’s function in cellular senescence and its pathological implication in fibrotic liver disease.

RESULTS

IRF3 is critical in initiating cellular senescence

The cGAS-STING pathway is recognized as a crucial trigger for initiating cellular senescence (27–29), yet its underlying mechanism remains unresolved. To explore this, we first investigated the role of key cGAS-STING signaling components, cGAS, STING, TANK-binding kinase 1 (TBK1), and IRF3, in the senescence process. Treatment of gut epithelial DLD1 cells with hydroxyurea (HU), a well-defined inducer of cGAS-STING signaling and cellular senescence (26, 46), activated STING signaling and led to phosphorylation of TBK1 and IRF3 (Fig. 1A). Notably, IRF3 remained persistently active during extended observation (Fig. 1A). Cellular senescence phenotypes, characterized by dephosphorylation of RB, expression of p16^INK4a and p21^Cip1/Waf1, positive senescence-associated β-galactosidase (SA-β-Gal) staining, and SASP factors including interleukin-6 (IL-6) and IL-1α, were evident on the 6th day following treatment with HU or doxorubicin (DOX) (Fig. 1, A to C). Similar to previous reports, we noticed an accumulation of cGAS in cytoplasmic DNA foci of senescent cells (fig. S1A) and attenuated senescence in cGAS KO cells in response to DNA damage (fig. S1, B and C), suggesting an essential role of cGAS in sensing injury and driving senescence. Similarly, knocking out of the adapter protein STING or the kinases TBK1/inhibitor of NF-κB kinase ε (IKKε) also reduced DNA damage–induced senescence (fig. S1, D and E). IRF3 genetic ablation markedly inhibited DNA damage–induced senescence in DLD1 cells, an effect reversible by reintroducing IRF3 (Fig. 1, B and C). Diminished senescence following HU treatment was also observed in mouse embryonic fibroblasts (MEFs) (fig. S1, F and G). These observations suggest a notable role of cGAS-STING-IRF3 signaling in the induction of cellular senescence following DNA damage.

Fig. 1. IRF3 deletion attenuates cellular senescence phenotypes.

(A) Treatment of DNA damage inducer HU for 24 hours induced the activation of DNA sensing in DLD1 gut epithelial cells at day 1, revealed by the activating phosphorylation of TBK1 and IRF3. Phospho-IRF3, p16^INK4a, and p21^Cip1/Waf1 were accumulated over time (days 4 to 7). IB, immunoblotting. (B and C) DNA damage, triggered by treatment of HU or DOX, induced robust senescence phenotypes in DLD1 cells at day 6, revealed by SA-β-Gal staining (B), senescence marker p16^INK4a, and SASPs (C). Senescence phenotypes were markedly suppressed in DLD1 cells with KO of IRF3, which was largely restored by reintroduction of IRF3. (D) mRNA-seq assay was performed in HU-induced senescent WT or IRF3 KO DLD1 cells, revealing by the heatmap depicting relative mRNA levels. (E and F) The gene set enrichment analysis plot (E) and heatmap (F) showed that genes associated with senescence were enriched in HU-treated WT but not in IRF3 KO DLD1 cells. NES, normalized enrichment score. (G) Human embryonic fibroblasts MRC-5 at passages 28 and 35 were stained for SA-β-Gal and quantified to assess the replicative senescence in WT and IRF3-deficient human fibroblasts. (H) Primary MEFs at indicated days from Irf3^+/+ or Irf3^−/− mice were analyzed by SA-β-Gal staining to evaluate IRF3’s role in spontaneous immortalization. Scale bars, 100 μm. n = 3 independent biological repeats unless specified. *P < 0.05; **P < 0.01; ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction. The statistics source data are provided in table S1, and the scanned films of each immunoblotting are provided in fig. S7.

To further characterize the effect of IRF3 in HU-induced senescence, we conducted an mRNA sequencing (mRNA-seq) transcriptomic analysis. Substantial gene expression changes were observed following HU treatment, particularly when comparing wild-type (WT) and IRF3 KO DLD1 cells (Fig. 1D). Gene set enrichment analysis revealed a notable decrease in genes related to established cellular senescence markers in IRF3 KO cells (Fig. 1E). Specifically, the expression of mRNA encoding for SASP factors such as CCL2, CXCL1, and IL-8 was markedly increased in WT but not IRF3 KO DLD1 cells after HU treatment (Fig. 1F).

Given the diverse triggers of cellular senescence, we also examined the effects of IRF3 deletion in different cellular contexts, including MRC-5 human embryonic fibroblasts for assessing replicative senescence (Fig. 1G and fig. S1H) (47) and spontaneously immortalized primary MEFs from WT and Irf3^−/− mice for evaluating senescence induced by oxygen stress (Fig. 1H and fig. S1I) (27, 48). IRF3 deletion in MRC-5 or MEFs resulted in substantially reduced senescence, as evidenced by SA-β-Gal staining and decreased SASPs. These findings suggest that IRF3 is a critical regulator in senescence processes induced by various upstream stimuli.

IRF3 regulates cellular senescence via the p16^INK4a-RB pathway

Supplementation into DLD1 cells of IFN-α, a primary product of canonical cGAS-STING-IRF3 signaling known to activate downstream Janus kinase–signal transducers and activators of transcription signaling (fig. S2, A and B), failed to reverse the senescence impairment in IRF3 KO cells (Fig. 2, A and B). Considering that IRF3 KO also reduces SASP factors like IL-6 and IL-1α, which are known to promote senescence (49), we treated IRF3 KO cells with medium from senescent cells (MS) or recombinant IL-6. While this treatment partially mimicked the senescence induced by IRF3, it failed to effectively restore the senescence phenotype in IRF3 KO cells (Fig. 2, C and D, and fig. S2, C to E). These observations suggest an essential regulatory function of IRF3 in the senescence induction process beyond its role in cytokine production.

Fig. 2. IRF3 controls senescence through the p16^INK4a-RB pathway.

(A and B) Treatment of human IFN-α1b (1000 U/ml, every 24 hours) failed to restore senescence in HU-treated DLD1 cells with IRF3 deficiency, as evaluated by SA-β-Gal staining (A) and SASPs (B). n.s., not significant. (C and D) Four-day treatment of MS failed to restore senescence in HU-treated IRF3 KO DLD1 cells, as indicated by SA-β-Gal staining (C) and quantitative reverse transcription polymerase chain reaction (qRT-PCR) assays for senescence marker p21^Cip1/Waf1 and SASP factor IL-1α (D). (E and F) WT and IRF3 KO DLD1 cells were treated with 1 μM palbociclib, a Food and Drug Administration–approved specific inhibitor of CDK4/6, for 3 days and examined for SA-β-Gal staining (E) and qRT-PCR assays of SASPs (F). (G and H) HU-induced senescence of WT and RB KO DLD1 cells was examined by SA-β-Gal staining (G) and SASPs (H); RB KO DLD1 cells were generated by a CRISPR-mediated approach and verified by sequencing and immunoblotting. Scale bars, 100 μm. n = 3 independent biological repeats unless specified. *P < 0.05; ***P < 0.001, by analyses of variance (ANOVA) with Bonferroni correction.

CDK4/6 kinases are pivotal for phosphorylating RB in the p16^INK4a-RB pathway, and their inhibition can lead to RB hypophosphorylation and trigger senescence (28). Markedly, the introduction of palbociclib, a Food and Drug Administration–approved CDK4/6 inhibitor, notably reversed the dampened senescence phenotypes in IRF3 KO DLD1 cells (Fig. 2, E and F). Furthermore, the genetic ablation of RB nearly completely inhibited senescence induced by HU (Fig. 2, G and H), underscoring the crucial role of the p16^INK4a-RB pathway in senescence triggered by DNA damage. Therefore, these findings establish a connection between cGAS-STING-IRF3 and p16^INK4a-RB signaling pathways in regulating cellular senescence.

IRF3 associates with RB, the key regulator of cellular senescence

To elucidate the molecular mechanisms behind IRF3-driven cellular senescence, we generated DLD1 stably expressing IRF3 and performed mass spectrometry to identify potential IRF3-interacting proteins. Notably, RB, a central regulator of senescence, emerged as a prominent IRF3 associate (Fig. 3A and raw data available in a public database). We validated the interaction between endogenous IRF3 and RB in DNA damage–induced senescent cells through coimmunoprecipitation (Fig. 3B). Further, immunofluorescence and nucleocytoplasmic fraction assay revealed that, under DNA damage–induced senescence conditions, IRF3 was accumulated in the nucleus overlapping with RB cellular distribution (fig. S3, A and B). The proximity ligation assay (PLA) was then used, confirming the association of endogenous IRF3 and RB proteins in senescent cells (Fig. 3C). Expectedly, the activated form of IRF3 (IRF3 5D) exhibited a strong interaction with RB compared to its WT (Fig. 3D). RB also interacted with IRF7 (Fig. 3D), an IFN-I–induced homolog of IRF3 that amplifies interferon signaling (50). Our analyses demonstrated a specific interaction between RB and IRF3, distinct from other cGAS-STING pathway components (fig. S3C) and other senescence regulators (fig. S3D). This association between a central mediator of cGAS-STING signaling and a critical regulator of cellular senescence suggests an intriguing link that may illuminate biological processes.

Fig. 3. IRF3 interacts directly with RB, a key regulator of cellular senescence.

(A) Mass spectrometry analysis of the IRF3 immunoprecipitations (IP) from DLD1 cells indicated an elevated interaction between IRF3 and endogenous RB at day 6 upon HU-induced cellular senescence. Flag-tagged IRF3 was stably reintroduced into IRF3 KO DLD1 cells in this setting. (B) The endogenous interaction of IRF3 and RB was evaluated by coimmunoprecipitation assay in HU/DOX-induced senescent DLD1 cells. (C) PLAs in DLD1 cells demonstrated in situ signals for the IRF3-RB complex during HU/DOX-induced senescence. Scale bars, 10 μm. (D) Coimmunoprecipitation assays were performed in transfected human embryonic kidney (HEK) 293 cells to evaluate the interactions between RB and WT/activated IRF3 or IRF7. IRF3 5D (S396D/S398D/S402D/S405D/S427D) and IRF7 7D (S475D/S476D/S477D/S479D/S483D/S484D/S487D) are phosphomimetics for activated IRF3 and IRF7, respectively. (E to G) Domain mapping assays by coimmunoprecipitation assays indicated an interaction between the interferon-activating domain (IAD) of IRF3 5D (E) and the RB-B (F), as depicted by a schematic of their interacting domains (G). (H and I) The coimmunoprecipitation assays in HEK293 cells revealed two IRF3 mutants defective to RB interaction (S221A and R255A/R262A/H263A) (H). The transcriptional potentials of IRF3 phosphomimetics and mutants (S221A and R78A/R86A, R255A/R262A/H263A) were individually analyzed by IRF3-responsive reporters of IFNβ and 5×interferon-stimulated response element (ISRE) (I). n = 3 independent biological repeats unless specified. ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction.

To determine the specific domains responsible for the interaction between RB and IRF3, we used a series of RB and IRF3 truncation mutants for domain mapping assays (fig. S3E). Our studies identified the interferon-activating domain (IAD) of IRF3 and the pocket B domain of RB (RB-B) as essential for their interaction (Fig. 3, E to G). Investigating further the IRF3 mutants that we generated previously aiming for probing IRF3 structural features and modification residues (51), we found that the IRF3 S221A mutation, which failed to bind RB (Fig. 3H), retains the canonical transcriptional ability to induce IFN genes (Fig. 3I). Conversely, the IRF3 R78A/R86A mutation, which affects the DNA binding domain and impairs IFN signal activation (52), still retained the ability to bind RB, while IRF3 R255A/R262A/H263A mutation served as a control, losing IFN activation and RB-binding capabilities (Fig. 3, H and I). The IRF3 C222A/E224A mutation analysis also indicated that IRF3 did not bind RB via the I(L)xCxE motif (fig. S3F). These findings suggest that the interaction with RB is independent of the classical transcriptional function of IRF3.

IRF3 attenuates RB hyperphosphorylation and initiates cellular senescence

RB is pivotal in cell cycle regulation, particularly its interactions with the E2F family and other transcription factors. The phosphorylation of RB at multiple sites, such as T826 by CDK4 and T821 by CDK6, is key to its inhibition and interaction with E2Fs (53). The RB mutant lacking these inhibitory phosphorylations (T821A/T826A, referred to as “activated RB”), which has a high affinity for E2F proteins, showed a strong association with phosphomimetic IRF3 (Fig. 4A). Furthermore, a substantial reduction in RB phosphorylation levels were seen in senescent cells (fig. S4A). Conversely, IRF3 ablation prevented the stress-induced down-regulation of phospho-RB at T826 in human embryonic fibroblasts MRC-5 (Fig. 4B) and immortalized MEFs (Fig. 4C). Reintroducing IRF3 into IRF3 KO cells restored the diminished phospho-RB levels observed during DNA damage–induced senescence (Fig. 4D). Moreover, DLD1 IRF3 KO cells expressing the IRF3 S221A mutant, which is incapable of forming the IRF3-RB complex, showed diminished ability to suppress RB hyperphosphorylation (Fig. 4E). As expected, knocking out of upstream activators of IRF3 such as cGAS or TBK1/IKKε resulted in elevated levels of phospho-RB (fig. S4, B and C). During extended cellular senescence, the number of senescent cells in IRF3 deletion cells did not notably increase, and a marked decrease in RB phosphorylation was not observed (fig. S4, D and E), indicating that IRF3 loss diminishes but does not delay cellular senescence. These data indicate a robust regulation of RB hypophosphorylation by IRF3 and the importance of IRF3 in the entry of senescence.

Fig. 4. IRF3 attenuates RB hyperphosphorylation.

(A) Coimmunoprecipitation assays showed an enhanced interaction between IRF3 5D and RB mutant T821/826A, the mutation RB at a constant state of hypophosphorylation. (B) Compared with WT cells, MRC-5 cells with IRF3 deficiency displayed a higher phospho-RB (T826) level, as measured in passage 34. (C) Immunoblotting showed an increase in phosphorylated RB in MEFs isolated from Irf3^−/− mice. (D) IRF3 deficiency in DLD1 cells prevented the DNA damage–induced senescence markers, including decreased RB phosphorylation at T826 and up-regulation of p16^INK4a and p21^Cip1/Waf1, which were restored by reintroducing IRF3 via a lentiviral delivery. (E) Flag-tagged IRF3 WT and mutants were stably reconstituted into IRF3 KO DLD1 cells via a lentiviral delivery. Differences in RB phosphorylation, p16^INK4a, and p21^Cip1/Waf1 protein levels upon HU-induced cellular senescence were measured by immunoblotting in cells harboring distinct IRF3 mutants. (F) Doxycycline-induced expression of STING R281Q, a constitutively active form of STING, reduced RB phosphorylation levels at T826 but increased p16^INK4a and p21^Cip1/Waf1 expression, as measured by immunoblotting. (G) Constitutively activation of STING promoted senescence entry, as revealed by SA-β-Gal staining and quantifications. Scale bars, 50 μm. (H) Induced expression of STING in A549 cells by a Tet-On system triggered IRF3 activation and attenuated phospho-RB at T826, a process blocked by CRISPR-mediated IRF3 KO. (I) IRF3 deficiency blocked STING-induced senescence in A549 cells, as assessed by SA-β-Gal staining. Scale bars, 100 μm. (J) A clone formation assay was performed in WT, STING Tet-On, and IRF3 KO A549 cells and observed on day 10. (K) IRF3 deficiency diminished STING-induced cell growth arrest, as assessed by Cell Counting Kit-8 assay from day 0 to day 4. OD, optical density. n = 3 independent biological repeats unless specified. ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction.

By contrast, the inducible expression of STING R281Q, a constitutively active form of STING (54), using a Tet-On system, led to a gradual reduction of phospho-RB levels (Fig. 4F) and an increase in cellular senescence, as evidenced by SA-β-Gal staining (Fig. 4G) and elevated levels of SASPs (fig. S4F). To explore this regulation further, we used an STING-sensitive system in A549 cells, a lung epithelial carcinoma cell line lacking STING expression (55). Inducible STING expression in A549 cells activated TBK1-IRF3 signaling, resulting in C-terminal phosphorylation of IRF3 and notably decreased RB T826 phosphorylation (fig. S4G). Conversely, IRF3 KO in A549 cells inhibited the reduction of phospho-RB (Fig. 4H) and prevented the induction of senescence, as indicated by SA-β-Gal staining (Fig. 4I).

RB phosphorylation typically facilitates cell cycle progression, while its dephosphorylation leads to cell proliferation arrest by associating and inhibiting E2F transcription factors (1, 2). A colony formation assay demonstrated substantial growth arrest in A549 cells with induced STING expression (Fig. 4J), further confirmed by cell counting assay and 5-ethynyl-2-deoxyuridine (EdU) staining (Fig. 4K and fig. S4H). Notably, the KO of IRF3 in STING-Flag Tet-On A549 cells prevented this proliferation inhibition, as shown in comparison with compared with cells of IRF3 negative KO (Fig. 4, J and K, and fig. S4H). In contrast, p53 appeared not to play a role in this process (fig. S4, I and J). These data suggest that the STING-TBK1-IRF3 axis controls the entry into cellular senescence, leading to proliferation arrest.

IRF3-RB interaction prevents CDK-mediated RB phosphorylation, thus activating RB

To fully understand how IRF3 activates RB, we generated stable DLD1 cell lines expressing various forms of IRF3, including WT, S221A, and R78A/R86A. IRF3 S221A expectedly did not interact with RB (Fig. 5, A and B), which correlated with a reduced entry into senescence, evidenced by SA-β-Gal staining (Fig. 5C) and SASP levels (fig. S5A). However, this mutant maintained its ability to activate downstream IFN signaling (fig. S5B). These observations from point mutagenesis emphasize the importance of the IRF3-RB axis in senescence regulation in response to damage, independent of the transcriptional activity of IRF3.

Fig. 5. IRF3 activates RB by attenuating CDK-induced RB phosphorylation.

(A) Flag-tagged IRF3 WT and mutants were stably reconstituted into IRF3 KO DLD1 cells. Coimmunoprecipitation assays evaluated the associations of endogenous RB and IRF3 mutants in HU/DOX-induced senescent DLD1 cells. (B) PLA assays were performed in IRF3 KO DLD1 cells stably expressed IRF3 or mutants, which revealed in situ cellular signals for the IRF3-RB complex with or without HU treatment. 4′,6-diamidino-2-phenylindole (DAPI) staining was used to visualize the cell nuclei. Scale bars, 10 μm. (C) Roles of IRF3 mutants on HU-induced cellular senescence were assessed by SA-β-Gal staining. Scale bars, 100 μm. (D) An in vitro kinase assay of CDK4/6 and RB was performed. RB proteins were expressed in HEK293 cells and pulled down using Myc-tag antibody and pretreated with λ-protein phosphatase (λPPase) to remove its phosphorylation. CDK4, CDK6, and cyclin D1 were expressed in HEK293 cells and pulled down using Flag-tag antibody, incubated with separately purified RB and IRF3 5D for kinase assay in the absence or presence of palbociclib, with adenosine 5′-triphosphate (ATP). (E) An in vitro kinase assay used RB and Smad3 as phosphorylation substrates, which were expressed in HEK293 cells and purified by Myc-tag antibody with λPPase for pretreatment. (F) The RB–cyclin D1 complex was detected by coimmunoprecipitation, by which IRF3 5D dissociated but not the IRF3 mutant defective for RB interaction. (G) The association of RB and E2F1 was enhanced in the presence of IRF3 5D, as revealed by coimmunoprecipitation assays. (H) Regular or senescent WT and IRF3 KO DLD1 cells were subjected to coimmunoprecipitation and immunoblotting, which revealed their diverse interactions of RB and E2F1. Besides, an endogenous complex of IRF3, E2F1, and RB was detected by coimmunoprecipitation in senescent DLD1 cells. n = 3 independent biological repeats unless specified. ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction.

In addition, we established an inducible system for expressing constitutively activated IRF3 (IRF3 5D) in DLD1 cells, observing a dose-dependent suppression of RB phosphorylation (fig. S5C). Next, we used an in vitro kinase assay system, using separately expressed and purified CDK4, CDK6, cyclin D1, RB, and IRF3 5D proteins, which further elucidated the effect of IRF3. In this system, CDK4/6 directly phosphorylated RB at T826 and T821 in the presence of cyclin D1, a process inhibited by the CDK4/6 inhibitor palbociclib (Fig. 5D). Notably, the addition of IRF3 5D blocked CDK4-mediated RB phosphorylation in a dose-dependent manner (fig. S5D), suggesting a crucial IRF3 in preventing CDK4/6-mediated RB inactivation and a key mechanism for cGAS-STING-IRF3 signaling–induced cellular senescence. To distinguish whether IRF3 5D affects CDK4/6 kinase activity or interferes with CDK4/6-RB phosphorylation through the RB-IRF3 interaction, we analyzed CDK4-mediated phosphorylation of Smad3, another CDK4 substrate (56). The selective inhibition of CDK4-mediated RB (but not Smad3) phosphorylation (Fig. 5E) underscores the importance of the IRF3-RB interaction. The IRF3 5D compromised the RB–CDK4/6–cyclin D1 interaction, unlike the mutant lacking RB interaction capability (Fig. 5F). These insights suggest a direct mechanism by which IRF3, through its interaction with RB, prevents CDK-mediated RB phosphorylation.

We next assessed the role of IRF3 in forming the RB-E2F complex, a critical inhibitory complex that regulates the transcription function of E2F proteins (57). Notably, the activated form of IRF3 substantially facilitated the association between RB and E2F1 (Fig. 5G), forming a tripartite complex consisting of IRF3, RB, and E2F1 (fig. S5E). Moreover, we observed the interaction of endogenous E2F1 and IRF3 with RB in HU-induced senescent cells (Fig. 5H), confirming the presence of an endogenous RB-IRF3-E2F1 complex in response to DNA damage. Furthermore, we evaluated the expression of IFN genes such as Interferon-Induced Protein with Tetratricopeptide Repeats 1 (IFIT1) and ISG15 in DLD1 senescent cells, showing a heightened activation of IFN signaling (fig. S5, F to H). These observations indicate that cGAS-STING–induced interferon responses and senescence are interlinked and mutually reinforcing, potentially because of complex interactions within cellular microenvironments. Our findings delineate a direct molecular mechanism through which STING-induced cellular senescence is initiated, involving the formation of an IRF3-RB-E2F complex in the nucleus that inhibits CDK-mediated RB phosphorylation and interferes with E2F-mediated transcription.

IRF3 facilitates the senescence of HSCs during liver fibrosis

Previous research has highlighted the crucial role of cellular senescence in mitigating hepatic fibrosis (16–18). To investigate the physiological impact of the STING-IRF3-RB axis in this context, we examined mouse liver fibrosis induced by carbon tetrachloride (CCl₄) or bile duct ligation (BDL). After CCl₄ or BDL treatment, senescent cells and collagen fibers were notably present in mouse livers (Fig. 6, A and B), aligning with recent findings (58). However, cellular senescence was noticeably reduced in Irf3^−/− livers, a phenotype that was reversed with the CDK4/6 inhibitor palbociclib, as shown by SA-β-Gal staining and SASP analysis (Fig. 6, A and B, and fig. S6A), suggesting a critical role of IRF3 in modulating senescence in liver fibrosis.

Fig. 6. The IRF3-RB axis regulates senescence during liver fibrosis.

(A) The carbon tetrachloride (CCl₄; 0.5 mg/kg) was injected twice a week for 4 weeks to induce liver fibrosis in Irf3^+/+ and Irf3^−/− 8-week-old C57BL/6 mice. Mouse livers after CCl₄ treatment exhibited an accumulation of SA-β-Gal–positive cells, which were less in Irf3^−/− mice. Scale bars, 100 μm. (B) The bile duct of Irf3^+/+ and Irf3^−/− 8-week-old C57BL/6 mice was ligated to induce liver fibrosis. Palbociclib (100 mg/kg) was used on days 7, 8, and 9 after the surgery. Representative images and quantification of SA-β-Gal activity in the liver of Irf3^+/+ and Irf3^−/− mice showed the percentage of SA-β-Gal–positive cells. Senescence is raised after treatment with palbociclib in Irf3^−/− mice after BDL. Scale bars, 100 μm. (C) Representative Sirius Red–stained liver sections in sham/CCl₄ treatment revealing collagen deposition and bridging fibrosis. The degree of collagen deposition was increased in Irf3^−/− mice. Scale bars, 100 μm. (D) qRT-PCR assays of collagen mRNA level of livers were shown. (E) Representative Sirius Red–stained liver sections in BDL surgery and palbociclib treatment. Scale bars, 100 μm. (F) Serum ALT and AST levels markered the degree of liver injury. Palbociclib administration markedly reduced serum ALT and AST in Irf3^−/− mice. (G) Hepatic parenchymal cells (HCs) and primary hepatic NPCs were isolated from CCl₄-treated Irf3^+/+ or Irf3^−/− mouse livers. qRT-PCR assays of mRNA level of senescence makers p16^INK4a, p21^Cip1/Waf1, and SASPs such as CXCL1 and IL-6 were shown. (H) Primary HCs and NPCs were isolated from CCl₄-treated Irf3^+/+ or Irf3^−/− mouse livers. Cellular senescence and the activation of cGAS-STING-IRF3 signaling were analyzed by immunoblotting of phospho-RB (T826), p16^INK4a, p21^Cip1/Waf1, and phospho-IRF3 (S396) antibodies. (I) qRT-PCR analyzed cGAS, STING, and IRF3 expression in HCs and NPCs after CCl₄-induced liver fibrosis. (J) Immunofluorescence assay to costain α-SMA and p21^Cip1/Waf1 in WT and Irf3^−/− mouse livers during liver fibrosis. α-SMA–positive cells in Irf3^−/− mouse livers showed fewer p21^Cip1/Waf1 signals, which was reversed by palbociclib treatment. Scale bars, 20 μm. n = 5 independent biological repeats unless specified. *P < 0.05; **P < 0.01; ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction.

Blocking IFN-I signaling with IFNAR1 neutralizing antibodies did not substantially decrease senescence during fibrosis progression (fig. S6B). Fibrosis progression, visually confirmed by Sirius Red staining, revealed larger collagen deposition areas and higher collagen mRNA levels in Irf3^−/− mice in the CCl₄-induced hepatic fibrosis model (Fig. 6, C and D). A similar pattern was also observed in the BDL-induced model (Fig. 6E and fig. S6, C and D). Disease parameters, including serum alanine aminotransferase (ALT) and aspartate transaminase (AST) levels (59), were elevated in BDL-induced liver fibrosis and further increased in Irf3^−/− mice (Fig. 6F). BDL surgery also caused liver enlargement, as indicated by a notable rise in the hepatosomatic ratio, which was exacerbated by IRF3 deficiency (fig. S6E). Hematoxylin and eosin (H&E) staining showed an accumulation of more nonparenchymal cells (NPCs) (58) with IRF3 ablation (fig. S6, F and G). Notably, these changes were mitigated by the palbociclib administration. These findings suggest that the STING-IRF3-RB axis suppresses liver fibrosis by promoting cellular senescence.

To pinpoint the primary cell types in liver fibrosis exhibiting SA-β-Gal positivity, we isolated primary hepatocytes (HCs) and NPCs from WT and Irf3^−/− mouse livers after CCl₄ treatment. Analysis of senescence markers p16^INK4a, p21^Cip1/Waf1, and SASPs indicated that IRF3 deletion predominantly affected senescence in NPCs, not HCs (Fig. 6G). IRF3 phosphorylation was noted in primary NPCs from WT mice following CCl₄ induction, and elevated phospho-RB levels were observed in IRF3 KO NPCs (Fig. 6H). NPCs showed higher mRNA and protein levels of cGAS and STING than HCs (Fig. 6, H and I), elucidating the enhanced IRF3 activation and more pronounced senescence in NPCs during liver fibrosis.

Given that HSC activation and Kupffer cell proliferation are key events in hepatic fibrosis (16, 60), we performed immunofluorescence analyses to identify further the cells involved in IRF3-triggered senescence. Notably, over half of the HSCs expressed the senescence marker p21^Cip1/Waf1 (Fig. 6J), and most cells with elevated p21^Cip1/Waf1 expression coexpressed the HSC marker α–smooth muscle actin (α-SMA) (fig. S6H), in line with recent findings (61). In contrast, only about 10% of p21⁺ cells were Kupffer cells (fig. S6H), and there was no notable difference in the proportion of F4/80⁺ cells between WT and Irf3^−/− mice (fig. S6I). Costaining analyses revealed a correlation between p21⁺ and p53⁺ cells but not with the proliferation marker Ki67 (fig. S6J). Moreover, cellular senescence in HSCs was reduced in Irf3^−/− mice, and palbociclib treatment increased p21⁺ HSC levels in IRF3 KO mice (Fig. 6J), paralleling with SA-β-Gal staining results (Fig. 6B). These observations support a key role for the IRF3-RB complex in suppressing HSC-driven liver fibrosis and imply that CDK4/6 inhibition could be a promising therapeutic approach for fibrotic liver diseases.

The STING-IRF3-RB axis reduces liver fibrosis through HSC senescence

To explore the in vivo role of the STING-IRF3-RB axis in regulating cellular senescence of HSCs during liver fibrosis, we used an HSC-specific promoter-driven adeno-associated virus serotype-8 (AAV8)–glial fibrillary acidic protein (GFAP) (promoter)–enhanced green fluorescent protein (EGFP)–Cre (62) to selectively delete IRF3 in HSCs of Irf3^flox/flox mice (Fig. 7A). The targeted deletion was verified by increasing Cre mRNA and EGFP expression in HSCs (Fig. 7B), along with a reduction in IRF3 protein levels in isolated HSCs (Fig. 7C). Upon CCl₄ treatment, IRF3 activation and RB hypophosphorylation were observed in HSCs (Fig. 7C). Similar to the global IRF3 KO, targeted deleting IRF3 in HSCs resulted in substantially diminished senescence phenotypes, evident from the sustained RB hyperphosphorylation (Fig. 7C), reduced senescence markers such as IL-6 and p21^Cip1/Waf1 (Fig. 7D), and a decrease in senescent cells (Fig. 7E). Moreover, the selective deletion of IRF3 in HSCs led to increased collagen fiber aggregation (Fig. 7F) and elevated expression of collagen-related genes (Fig. 7G) in livers during CCl₄-induced fibrosis. In addition, we detected a noticeable increase in NPCs near bile ducts and blood vessels (Fig. 7H) but not inflammatory factors in the livers between WT and conditional IRF3 KO mice (Fig. 7I). These findings align with results from the BDL model in IRF3 KO mice, highlighting cGAS-STING-IRF3-RB signaling as a critical suppressive role in HSC-mediated liver fibrosis.

Fig. 7. Selective deletion of IRF3 in HSCs attenuates HSC senescence and promotes liver fibrosis.

(A) A flow chart of the CCl₄-induced hepatic fibrosis model in IRF3 conditional KO mice was shown. The mice with IRF3 conditional KO in HSC were established by tail vein injection of AAV8–GFAP (promoter)–EGFP–Cre twice at the indicated time to Irf3^flox/flox mice. (B) qRT-PCR assays of Cre mRNA and immunofluorescence of α-SMA and GFP indicated that upon AAV8-GFAP-EGFP-Cre injection in mouse livers, epidermal growth factor receptor–Cre was expressed and localized in HSCs. Scale bars, 20 μm. (C and D) RB phosphorylation and IRF3 activation were evaluated by immunoblotting in primary HSCs isolated from livers of CCl₄-treated WT or IRF3 conditional KO (Cre) mice. (C). SASPs and senescence marker p21^Cip1/Waf1 were evaluated by qRT-PCR assays (D). (E) Representative SA-β-Gal staining showed fewer SA-β-Gal positive signals from IRF3 conditional KO mice in liver sections. Scale bars, 100 μm. (F to I) Liver sections from CCl₄-treated control or IRF3 conditional KO (Cre) mice were evaluated by Sirius Red staining (F), qRT-PCR of collagens, and α-SMA (G), hematoxylin and eosin (H&E) staining (H), and qRT-PCR of tumor necrosis factor–α (TNF-α) and IL-1β, markers of inflammation (I). Scale bars, 100 μm. (J) This model illustrates the regulatory role of the STING-IRF3-RB axis in HSCs during liver fibrosis. The process initiates with the activation of cGAS-STING signaling, leading to the phosphorylation and nuclear translocation of IRF3. Inside the nucleus, IRF3 forms complexes with RB. This interaction inhibits the CDK4/6-cyclin-RB complex, resulting in the activation of RB. The activated RB then binds to and inhibits E2F transcription factors, effectively triggering cellular senescence in activated HSCs. Consequently, this pathway mitigates liver fibrosis by promoting senescence in HSCs. n = 5 independent biological repeats unless specified. **P < 0.01; ***P < 0.001, by analysis of variance (ANOVA) with Bonferroni correction.

DISCUSSION

RB is central to cell cycle progression by controlling the activity of the E2F family and other transcription factors, conferring tumor suppressor and senescent roles in cancers, aging, and embryogenesis (22, 57). Endogenous complexes in the nucleus of various senescent cells, comprising C-terminal phosphorylated IRF3 and hypophosphorylated RB and formed by their interactions between RB-B and IAD domains, were revealed in this report. Markedly, binding of activated IRF3 prevents hyperphosphorylation of RB caused by CDK4/6 kinases, probably due to the competitive association of the pocket B domain between IRF3 and CDK4/6 kinases. This straightforward mechanism of shifting RB complex activates RB to sequester E2F transcriptional factors, thus arresting the cell cycle (Fig. 7J).

Biological significance of the IRF3-RB interaction

Cell fate determination is a primary function of cGAS-STING signaling, resulting from its master controls over cytokine production (27, 29), autophagy (63), mRNA translation (46), and other cellular processes such as condensation (44). IRF3 receives information from pattern-recognition receptors, including Toll-like receptors, retinoic acid–inducible gene I–like receptors, and DNA sensor cGAS (64). Viral proteins such as human papillomavirus E7, SV40 T antigen (Tag), and adenovirus E1A also interact with RB through its pocket domain, which prevents RB from associating with physiological partners (65) or induces RB degradation (66), leading to RB dysfunction and an active transcription state of cells. Notably, activation of IRF3 by severe acute respiratory syndrome coronavirus 2 infection can accelerate the senescence of lung tissue, revealed by multidimensional omics analyses of patient lung tissues (67). The observations may be associated with the various sequelae contributing to heightened mortality and affect prognosis in older adults following severe acute respiratory syndrome coronavirus 2 infection (68, 69). Consequently, strategies targeting senescence could offer potential preventative measures to mitigate tissue damage resulting from viral infections, particularly given the frequent activation of IRF3 in numerous patients with COVID-19 (70). As IRF3 plays a pivotal role in antiviral responses, the inherent interplay between IRF3 and RB might also act as a noncanonical and competitive mechanism to regulate the cell cycle and combat viral infections.

Cellular senescence in diverse cell types exhibits multifaceted and probably controversial effects on liver fibrosis (16, 71). Our findings reveal that, during liver fibrosis, NPCs exhibit substantially higher mRNA and protein levels of cGAS and STING than HCs, in line with recent observations for compromised cGAS-STING signaling in HCs (72, 73). Similarly, protein levels of senescent regulators such as RB, p16, and p21 are more prominent in NPCs, emphasizing the major senescent events occurring in NPCs but not HCs during liver fibrosis (16). The role of cellular senescence in HSCs, crucial in liver damage response and extracellular matrix production in fibrotic scars, is more apparent. Senescence in HSCs attenuates their proliferation, diminishes extracellular matrix secretion, and stimulates SASPs to attract natural killer cells for cleansing, thereby preventing excessive proliferation and transdifferentiation of HSCs into fibroblasts (74, 75). This senescence process in HSCs prevents liver fibrosis (16–18).

In our study, using the CCl₄ and BDL murine models of liver fibrosis, we observed that IRF3 deficiency in HSCs exacerbates liver fibrosis symptoms, implicating the selective role of STING-IRF3 signaling in avoiding HSC-induced liver fibrotic diseases. Conversely, recent findings indicate a profibrotic role of cGAS-STING signaling in macrophages in nonalcoholic fatty liver disease, likely due to increased liver inflammation (71, 76). Therefore, contrasting liver fibrosis phenotypes from HSC-specific IRF3 KO and macrophage-specific STING KO suggests a context-dependent, disease-type, and stage-specific modulation of liver fibrosis by cGAS-STING signaling. Furthermore, cGAS-STING signaling has been linked to promoting precancerous cell senescence, introducing additional dimensions to its antitumor capabilities (77). Many tumor cells exhibit down-regulated or absent cGAS-STING signaling (26, 78–80). CDK4/6 inhibitor administration attenuated the IRF3’s impact on liver fibrosis, confirming the in vivo relevance of RB-IRF3 regulation. Considering RB’s interaction with chromatin-remodeling enzymes and its role in gene expression modulation (81), it is tempting to speculate on the potential transcriptional functions of the RB-IRF3 complex, particularly to genes regulated by cGAS-STING signaling: This aspect and other biological events following the formation of the IRF3-RB complex warrant further investigation.

Molecular mechanisms underlying STING-induced cellular senescence

Our research has uncovered an unconventional function of IRF3 beyond its known role as a transcription factor. We found that IRF3 functions as a critical regulator of cell cycle progression and cellular senescence by interacting with and activating RB in the nucleus. This previously unidentified STING-IRF3-RB pathway effectively regulates STING-induced cellular senescence, augmenting the recently found STING-PERK signaling (46). Our findings suggest that the STING-PERK–eukaryotic initiation factor 2 alpha (eIF2α) pathway may trigger the onset of cellular senescence, while the STING-TBK1-IRF3-RB pathway seems to reinforce senescence symptoms. Both pathways are vital for robust senescence induction via cGAS-STING signaling. This research enhances our understanding of the p53-p21^Cip1/Waf1 and p16^INK4a-RB mechanisms in controlling senescence and broadens the scope of cGAS-STING signaling in the context of senescence-related diseases.

A recent study has highlighted that a Smad-RB complex influences the p21-activated secretory phenotype (82). Notably, the IRF3 IAD and the Mad Homology 2 (MH2) domain of R-Smads show notable structural and electrostatic similarities (42, 83), indicating a potential conserved action pattern among these signaling mediators. Given that IRF3’s IAD interacts with transcriptional comodulators, it is plausible that the RB-IRF3 interaction might similarly drive an IRF3-related secretory phenotype. The interaction between C-terminal phosphorylated IRF3 and the RB-B is notably robust, likely because of the structural attributes from multiple residue phosphorylations in the IAD by TBK1 and other kinases. In vitro studies show that activated IRF3 effectively disrupts the RB-CDK complex, substantially inhibiting RB hyperphosphorylation. This phenomenon aligns with observed decreases in phospho-RB levels in MEFs with STING activation (84).

Senescent cells display morphological and functional changes over time (1, 85). Proteomic analyses have revealed a dynamic SASP pattern, with various signaling components peaking at different senescence stages (86). We observed that the RB-IRF3 complex formation spans a more extended period in various primary and cultured cells and organ tissues, likely triggered by ongoing cGAS-STING signaling in response to cellular and tissue damage. Therefore, exploring further how the IRF3-RB complex in the nucleus is balanced and regulated is intriguing.

In summary, our research uncovers the widespread presence of a previously unidentified IRF3-RB complex in various senescent cells, highlighting its marked influence on RB activity. This discovery is pivotal in understanding the regulation of damage-induced cellular senescence and fibrotic liver disease and elegantly complements our recent findings on STING-PERK-eIF2α signaling, collectively presenting a dual-faceted molecular basis for cGAS-STING–induced cellular senescence.

MATERIALS AND METHODS

Expression plasmids, reagents, and antibodies

Expression plasmids encoding Flag-, Myc-, or hemagglutinin (HA)–tagged WT, or mutations, or truncations of human IRF3, TBK1, IKKε, STING, Mitochondrial Antiviral Signaling Protein (MAVS), Smad3, and the reporters of IFNβ_Luc and 5×ISRE_Luc have been described previously (26, 87). HA-E2F1 was from Addgene (24225). Open reading frames (ORFs) of human IRF7, p53, CDK4, CDK6, cyclin D1, and cyclin E1 were obtained from the Invitrogen ORF Lite Clone Collection cDNA library by polymerase chain reaction (PCR). HA-Flag-tagged p53, CDK4, CDK6, cyclin D1, and cyclin E1 were constructed on the mammalian expression vector pDEST. Myc-or Flag-tagged WT, mutations, and truncations of human RB were constructed on the mammalian expression vector pRK5. Site-directed mutagenesis to generate expression plasmids encoding RB K530A, N757A, T821A/T826A, and IRF3 S221A, R78A/R86A, and R255A/R262A/H263A were performed using a kit from Stratagene. IRF3 5D, STING, and STING R281Q were constructed on lentiviral vector pCMV, using them to generate the Tet-On system in various cells. Flag-IRF3, Flag-IRF3 S221A, and Flag-IRF3 R78A/R86A were constructed on lentiviral pBobi, using them to generate stable cell lines of DLD1 cells. All coding sequences were verified by DNA sequencing, and detailed information about plasmids is listed in table S4 and upon request.

The pharmacological reagents HU (Sigma-Aldrich), DOX (Sigma-Aldrich), palbociclib (Selleck), doxycycline (Sangon Biotech), G418 (Yeasen), etoposide (Sigma-Aldrich), λ-protein phosphatase (λPPase) (New Englad Biolabs), and IFNα-1b (Sangon Biotech) were purchased and used according to the manufacturer’s instructions.

Detailed information about antibodies applied in immunoblotting, immunoprecipitation, and immunofluorescence is provided in table S2. Monoclonal anti-TBK1 (1:5000 dilution), anti-IKKε (1:3000 dilution), anti-pTBK1 (S172) (1:3000 dilution), anti-IRF3 (1:2000 dilution), anti-pIRF3 (S396) (1:1000 dilution), anti-STING (1:3000 dilution), anti-p53 (1:3000 dilution), anti-E2F1 (1:1000 dilution), anti-Myc Tag, and anti-HA Tag antibodies were purchased from Cell Signaling Technology. Anti-pIRF3 (S386) (1:2000 dilution), anti-IRF3 (1:2000 dilution), anti-pRB (T821) (1:1000 dilution), anti-pRB (T826) (1:3000 dilution), anti–collagen I (1:3000 dilution), and anti–α-SMA (1:2000 dilution) antibodies were purchased from Abcam. Anti-p16^INK4a (1:2000 dilution) and anti-p21^Cip1/Waf1 (1:2000 dilution) were purchased from Santa Cruz Biotechnology. Anti-pSmad3 (T179) (1:2000 dilution) was purchased from Abclonal. Anti-RB (1:2000 dilution) was purchased from BD Biosciences.

Mice

Irf3^−/− and Irf3^flox/flox mice with C57BL/6 background were a gift from X. Wang. Both male and female littermates were used for all the experiments. HSC-specific deletion of IRF3 was obtained by injection of AAV8 bearing GFAP (promoter)–EGFP-Cre into Irf3^flox/flox mice through tail vein injection at a dose of 1 × 10¹¹ virus genome (PackGene Biotech), twice in 2 weeks. AAV8–GFAP (promoter)–EGFP was used as a control. All the mice were bred and maintained in a pathogen-free animal facility at the laboratory animal center of Zhejiang University. The care of experimental animals was approved by the Committee of Zhejiang University and followed Zhejiang University guidelines.

Cell lines

DLD1, HeLa, human embryonic kidney (HEK) 293, A549, and MRC-5 cell lines were obtained from American Type Culture Collection. No cell lines used in this study were found in the database of commonly misidentified cell lines that International Cell Line Authentication Committee and National Center for Biotechnology Information BioSample maintain. STING KO, TBK1/IKKε dKO, IRF3 KO, p53 KO, and RB KO cells were generated with the CRISPR technology. The IRF3, IRF3-5D, STING, and STING-R281Q inducible expressing cells were generated by a lentiviral vector containing the inducible Tet-On system, followed by ORF of the target genes and selected by G418 antibiotic at a concentration of 1500 μg/ml for 5 to 7 days. DLD1 cells bearing IRF3 stable expression were generated by the lentiviral vector pBobi and selected by puromycin at 1 mg/ml for 3 days. All guide RNAs used are listed in table S3.

Cell culture, transfections, and infections

HEK293, DLD1, and MEFs were cultured in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum (FBS) at 37°C in 5% CO₂ (v/v). A549 cells were cultured in RPMI 1640 medium, and MRC-5 cells were cultured in minimum essential medium with 10% FBS at 37°C in 5% CO₂ (v/v). Lipofectamine 3000 (Invitrogen) or polyethyleneimine (Polysciences) transfection reagents were used for plasmid transfection. Lentiviral infection was performed by directly applying a virus-containing medium into target cells, which was later replaced by the fresh medium containing 10% FBS.

CCl₄-induced liver fibrosis model

As previously reported, the modified CCl₄-induced mouse fibrosis model was performed (58). The mice were randomly divided into sham and CCl₄ groups. Mice from the sham group were treated with olive oil, and mice from the CCl₄ group were treated with CCl₄ (0.5 ml/kg of body weight; diluted 1:9 with olive oil; Sigma-Aldrich) intraperitoneal injection twice a week for 4 weeks to induce liver fibrosis. Mice were euthanized, and the livers were harvested 48 hours after the final CCl₄ injection.

BDL-induced liver fibrosis model

BDL was performed as previously described with the modifications (88). The mice were randomly divided into sham and BDL groups. Mice were anesthetized with 2% pentobarbital sodium, and their left and right hepatic ducts and hepatic portal site of the common bile duct were isolated and ligated. Control mice (sham-operated) underwent laparotomy but had no ligation. Both sham and BDL groups were treated with saline or palbociclib (100 mg/kg, every day since day 7) through intragastric administration or with anti-IFNAR1 (10 mg/kg, every other day since day 4) through intraperitoneal injection. Serum ALT and AST levels were measured to evaluate hepatocellular damage.

MEFs for spontaneous immortalization assay

MEFs were generated from embryonic day 12.5 (E12.5) to E13.5 embryos of WT and IRF3 KO mice at 8 to 10 weeks of age and under conventional culture conditions, including 20% oxygen and 5% CO₂. We followed a 3T3 protocol (27) for spontaneous cellular senescence by seeding 1 × 10⁶ cells in a 10-cm dish every 3 days. Aliquots of the cells at indicated passages were expanded for continued passages, frozen in liquid nitrogen, and analyzed by the SA-β-Gal staining, Western blot, or quantitative reverse transcription PCR (qRT-PCR) assays.

CRISPR-Cas9–mediated generation of KO cells

CRISPR-Cas9 genomic editing for gene deletion was performed as described (89). Guide RNA sequences targeting STING, TBK1, IKKε, IRF3, p53, and RB genomic sequences were cloned into the pX330 plasmids. These constructs, together with the GFP vector, were transfected into cells. Thirty-six to 48 hours after transfection, cells were selected by flow cytometry, and single clones were obtained by serial dilution and validated by immunoblotting with indicated antibodies. All primers used in the CRISPR-Cas9 methods are also listed in table S3.

Isolation of liver NPCs and HSCs

As described previously, NPCs and HSCs were isolated from murine livers using a two-stage collagenase perfusion approach (58, 90). Livers of euthanized mice were perfused in situ through a balanced salt solution, followed by tissue lysis solution containing collagenase and deoxyribonuclease. Then, livers were excised, mashed, further digested, and filtered. Filtered cells were centrifuged twice at 50g for 2 min to remove HCs; the remaining NPC fraction was collected.

HSC enrichment was conducted by an optimized isolation method (91), and the remaining NPC fraction was resuspended in 11.5% OptiPrep (Sigma-Aldrich) and put between a bottom cushion of 15% OptiPrep and a top layer of phosphate-buffered saline (PBS). After centrifugation at 1400g for 15 min at 4°C, the HSC fraction was at the interface between the top and intermediate layer and collected. Cell viability was examined by Trypan blue exclusion.

Luciferase reporter assay

HEK293 cells were transfected with indicated reporters (50 ng) bearing an ORF coding Firefly luciferase, along with the pRL-Luc with Renilla luciferase coding as the internal control for transfection, as well as other expression vectors specified in the results section. Twenty-four hours after transfection, cells were lysed by passive lysis buffer (Promega), and luciferase assays were performed using a dual luciferase assay kit (Promega), quantified with POLARstar Omega (BMG Labtech) and normalized to the internal Renilla luciferase control.

Senescence assays

Cellular senescence was induced by DNA damage, replication stress, spontaneous immortalization, activation of cGAS-STING signaling, or liver fibrosis. Cell senescence β-galactosidase staining kit (Yeasen, 40754ES60) was used to identify senescent cells or frozen sections of mouse livers. Images were acquired using an inverted fluorescence microscope (Nikon). For DNA damage–induced senescence, DLD1, HeLa, or MEFs were treated with 10 mM HU, 1 μM DOX, or 1 μM etoposide at ~60 to 70% confluence for 24 hours and changed to standard medium for another 5 days. MEFs at days 3, 9, 18, and 27, and MRC-5 at passages 28 and 35 were harvested and fixed for SA-β-Gal staining for replication stress or spontaneous immortalization–induced senescence. In the experiments to detect the SASPs, cells were lysed after 6 days of stimulations and subjected to RNA extraction and qRT-PCR assays as described in the previous section to measure the expression of cytokines, including IL-6, IL-1α, p16^INK4a, and p21^Cip1/Waf1.

qRT-PCR assay

Mouse livers, DLD1, MEFs, and A549 cells stimulated with HU, DOX, or doxycycline were lysed, and total RNA was extracted using the RNeasy extraction kit (Axygen). cDNA was generated by All-in-One cDNA Synthesis SuperMix (Bimake), and qRT-PCR was performed using the EvaGreen qPCR MasterMix (Abcam) and CFX384 Real-Time PCR System (Bio-Rad). Relative quantification was expressed as 2 − ΔC_t, where C_t is the difference between the main C_t value of the triplicate of the sample and that of an endogenous L19 or GAPDH (glyceraldehyde-3-phosphate dehydrogenase) mRNA control. The human and mouse primer sequences used are listed in table S3.

Colony formation assay

A549 cells were planted in 6-cm dish with a concentration of 5 × 10³ cells per well and cultured in a 5% CO₂-humidified incubator at 37°C for 10 days. The colonies were stained with crystal violet for observation.

Library preparation, sequencing, and data analysis for mRNA-seq

Collected RNA was used to construct RNA-seq libraries using the Smart-seq2 method (92). Briefly, a priming buffer containing deoxynucleotide triphosphates and oligo(dT) primers was added to RNA and denatured at 72°C for 3 min, followed by cDNA synthesis and preamplification. According to the manufacturer’s instructions, sequencing libraries were constructed from 1 ng of preamplified cDNA using a TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme, TD503). mRNA-seq was performed with biological replicates for all samples. Barcoded libraries were pooled and sequenced on the Illumina HiSeq X Ten platform in paired-end mode. Raw sequencing reads were trimmed to 50 base pairs and mapped to the human genome (hg19) using TopHat v2.1.1 with default parameters. Only uniquely mapped reads were used for subsequent analysis. The RNA abundance of each gene was quantified as fragments per kilobase of exon per million mapped fragments (FPKM) using Cufflinks v2.2.1. Genes with an FPKM of <1 in all samples were excluded, and for the remaining genes, FPKM values smaller than 1 were set to 1 in subsequent analyses. A summary of the mRNA-seq data generated in this study is provided in table S6.

Coimmunoprecipitations and immunoblottings

DLD1 or A549 cells with indicated treatments, or HEK293 cells transfected with plasmids, were lysed using Myc Lysis Buffer (MLB) (1) [20 mM tris-Cl, 200 mM NaCl, 10 mM NaF, 1 mM NaV₂O₄, 1% NP-40, 20 mM β-glycerophosphate, and protease inhibitor (pH 7.5)]. Cell lysates were then subjected to immunoprecipitation using anti-RB (Santa Cruz Biotechnology, sc-102; 1:200 dilution) antibodies for endogenous proteins or tags of FLAG (Sigma-Aldrich, F3165; 1:200 dilution), Myc (GNI, 4110-MC; 1:300 dilution), or HA (Cell Signaling Technology, 3724S; 1:200 dilution) for transfected proteins. After three washes with the cold MLB, adsorbed proteins were analyzed by SDS–polyacrylamide gel electrophoresis (PAGE) (Bio-Rad) and immunoblotting with the indicated antibodies. Cell lysates were also analyzed by SDS-PAGE and immunoblotting to control protein abundance.

λPPase assay

HEK293 cells were transfected with His-Myc–tagged RB and Myc-tagged Smad3. Cells were lysed in the modified Myc-lysis buffer [25 mM tris-HCl, 300 mM NaCl, and 1% Triton X-100 (pH 7.5)] after 24 hours of transfection. Samples with New England Biolabs buffer for protein metallophosphatases, MnCl₂, and λPPase were incubated at 30°C for 60 min at Thermo-Shaker and then washed three times with New England Biolabs buffer for subsequent in vitro kinase assay.

In vitro kinase assay

HEK293 cells were transfected with the plasmids of indicated proteins, and immunoprecipitations with antibodies were performed after 24 hours of transfection. His-Myc–tagged RB and Myc-tagged Smad3 were pulled down with Myc-tagged antibody and pretreated using λPPase assay. With three washes, immunoprecipitated proteins were incubated in kinase assay buffer [20 μM adenosine 5′-triphosphate (ATP), 20 mM tris-HCl, 1 mM EGTA, 5 mM MgCl₂, 0.02% 2-mercaptoethanol, 0.03% Brij-35, and bovine serum albumin (BSA; 0.2 mg/ml) (pH7.4)] at 30°C for 60 min on a Thermo-Shaker. Adding an SDS loading buffer stopped the reaction, and the samples were subjected to SDS-PAGE and specified immunoblotting.

Immunofluorescence, PLAs, and microscopy

In the experiments to visualize subcellular localization indicated proteins, DLD1 or HeLa cells were fixed in 4% paraformaldehyde, treated with 0.5% Triton X-100 in PBS, and blocked in 2% BSA in PBS for 1 hour. Samples were then incubated sequentially with primary antibodies, including anti-Flag (M2) (Sigma-Aldrich, F3165; 1:300 dilution), anti-RB (BD Biosciences, 554145; 1:100 dilution), anti-λH2AX (Abcam, ab2893; 1:200 dilution) and Alexa-labeled secondary antibodies (the Jackson laboratories, 111-095-003, 115-095-003, 111-025-003, and 115-025-003; 1:500 dilution) with extensive washes. Frozen liver sections were fixed in 4% paraformaldehyde, treated with 0.5% Triton X-100 in PBS, and blocked in 2% BSA in PBS for 1 hour. Samples were then incubated sequentially with primary antibodies including anti-p21^Cip1/Waf1 (Santa Cruz Biotechnology, sc-6246; 1:100 dilution), anti–α-SMA (Abcam, ab7817; 1:500 dilution), anti-F4/80 (Bio-Rad, MCA497RT; 1:500 dilution), anti-Ki67 (Abcam, ab15580; 1:100 dilution), anti-p53 (Cell Signaling Technology, 4947; 1:300 dilution), and Alexa-labeled secondary antibodies (the Jackson laboratories, 111-095-003, 115-095-003, 111-025-003, and 115-025-003; 1:500 dilution) with extensive washes. The PLAs were performed using the Duolink Detection Kit with PLA PLUS and MINUS probes for rabbit and mouse antibodies (Duo92002 and Duo92004) according to the manufacturer’s protocol. Samples incubated with primary antibodies including anti-IRF3 (Abcam, ab68481; 1:50 dilution), anti-RB (BD Biosciences, 554145; 1:50 dilution), and anti-Flag (M2) (Sigma-Aldrich, F3165; 1:100 dilution). The PLA signals were recognized as red fluorescent dots. Slides were mounted with VECTORSHIELD and stained with 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories). Immunofluorescence images were obtained and analyzed using a Nikon Eclipse Ti inverted microscope, a Zeiss LSM710 confocal microscope, or a Zeiss LSM880 confocal microscope. The images were processed using ImageJ software.

EdU labeling of proliferating cells

A549 cells were labeled with EdU (10 μM) using the EdU Cell Proliferation Kit (Sangon Biotech, E607204). After 2 hours of incubation with EdU, cells were fixed for 10 min in 4% paraformaldehyde in PBS and permeabilized with 0.5% Triton X-100 in PBS for 15 min at room temperature. Blocking by 2% BSA in PBS, the cells were incubated for 45 min at room temperature with EdU reaction mixtures. After washing with PBS, nuclei were labeled by staining with DAPI, and images were captured using a Zeiss LMS710 confocal microscope.

Histopathological analysis

For histologic examination, mouse liver samples were dissected, fixed in 4% paraformaldehyde for 12 hours at 4°C, and divided into two parts for paraffin and freezing embedding. For paraffin embedding, the samples were dehydrated in a graded series of ethanol, embedded in paraffin, and sliced into 6-μm sections before staining by H&E or Sirius Red. For freezing embedding, the samples were dehydrated with 30% sucrose overnight at 4°C, embedded in optimal cutting temperature compound, and immediately frozen at −80°C. Sectioned samples of 10 μm in thickness were washed twice with PBS for SA-β-Gal staining and immunofluorescence.

Nano–liquid chromatography/tandem mass spectrometry analysis

Phoenix National Proteomics Core services performed nano–liquid chromatography/tandem mass spectrometry analysis for protein identification and characterization and label-free quantification. Tryptic peptides were separated on a C18 column and analyzed by LTQ-Orbitrap Velos (Thermo Fisher Scientific). Proteins were identified using the National Center for Biotechnology Information search engine against the human RefSeq protein databases. The mass spectrometry results for immunoprecipitations from normal and HU-induced senescent DLD1 Flag-IRF3 stable cells are also listed in table S5. Because of its similar molecular weight to immunoglobulin G, IRF3 was in both sections, and the Peptide-Spectrum Match (#PSM) of IRF3 only served as qualitative rather than quantitative. The mass spectrometry proteomics data of IRF3 interacting proteins during senescence have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the iProX partner repository with the dataset identifier IPX0007514000.

Statistics and reproducibility

Quantitative data are presented as the means ± SEM from at least three independent experiments. When appropriate, statistically significant differences between multiple comparisons were analyzed using the one-way analysis of variance (ANOVA) test with Bonferroni correction. Differences were considered significant at P < 0.05. If preserved and properly processed, then all samples were included in the analyses, and no samples or animals with conventional injection damage were excluded. No statistical method was used to predetermine the sample size. All experiments except those involving animals were not randomized. Immunoblottings, reporter assays, and qRT-PCR experiments have been independently repeated a minimum of three times to ensure reproducibility. The investigators were not blinded to allocation during experiments and outcome assessment.

Study approval

The study is compliant with all relevant ethical regulations regarding animal work. All animal experiments followed the guidelines and were approved by the Zhejiang University Laboratory Animal Committee. All cell lines are available at American Type Culture Collection.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Legends for tables S1, S5 and S6

Tables S2 to S4

Other Supplementary Material for this manuscript includes the following:

Tables S1, S5 and S6