Multiple Roles of IL6 in Hepatic Injury, Steatosis, and Senescence Aggregate to Suppress Tumorigenesis

Chronic injury and inflammation are recognized as key pathogenic factors in the development of hepatocellular carcinoma (HCC), which is one of the most common forms of liver cancer and currently the third leading cause of cancer mortality worldwide (1). In its most typical etiologic setting, HCC presents in cirrhotic patients following decades of active hepatitis arising from common risk factors, such as viral infections, alcohol-induced liver disease, and nonalcoholic steatohepatitis (NASH; ref. 2).

Diverse factors integral to the hepatic microenvironment in chronic hepatitis have been hypothesized to promote the initiation and development of liver cancer (3). In addition to chronic inflammation (4–6), factors such as fibrosis and stellate cell activation (7, 8), steatosis (9), and also hormonal signaling (10) have also been hypothesized to contribute either directly and indirectly to HCC development. Thus, it is not surprising that the mechanisms underlying HCC development appear to be diverse, and not fully understood (11).

The role of the pleiotropic, inflammatory cytokine, IL6 in hepatocarcinogenesis is a case in point. IL6 expression has been reported to be increased in patients with chronic hepatitis, for instance following HBV and HCV infections, and NASH (12–14), and strongly correlates with enhanced risk of HCC in patients with human chronic hepatitis (15). Moreover, observations in experimental animal models lend support to the notion of a direct linkage between IL6 together with signal transducer and activator of transcription 3 (STAT3), a key mediator of IL6 signaling, and the development of inflammation-associated cancers, including HCC (5). We too have reported that IL6 directly contributes to genomic instability in hepatocytes while enhancing accelerated hepatocarcinogenesis in the context of liver regeneration (16).

On the other hand, numerous studies in animal models involving chronic hepatitis demonstrate that IL6 signaling is crucial for ameliorating liver injury and fibrosis (17, 18). Moreover, recent evidence from both human patients and animal models suggests a clear linkage between reduced IL6 signaling and obesity (19–21), now considered a major independent risk factor in HCC development. Thus, an unresolved paradox arises in which the benefit of IL6 in preventing liver injury, fibrosis, and obesity appears to contradict its apparent linkage to the development of HCC. In addition, while the role of IL6 in hepatocarcinogenesis on a background of acute liver injury is well established (10), its role in hepatocarcinogenesis within the context of chronic liver injury remains unclear.

To explore the role of IL6 within the context of chronic liver injury we have utilized the multidrug-resistant gene 2 knockout (Mdr2^−/−) mouse model, which is a prototype of inflammation-associated cancer (4, 22). Targeted disruption of the Mdr2 (Abcb4) gene leads to regurgitation of toxic bile from leaky ducts into the portal tracts, resulting in periductal inflammation, sclerosing cholestatic hepatitis, and fibrosis from an early age (22). Importantly, the Mdr2 model uniquely reflects the patterns of chronic injury and hepatic senescence typically associated with chronic hepatitis in human patients (23, 24). Ultimately all Mdr2^−/− mice spontaneously develop HCC while progressing through distinct phases of inflammation, dysplasia, development of dysplastic nodules, and carcinoma (4, 22), similar to the development of HCC in humans (25). Moreover, adult human patients carrying mutations in the MDR3/ABCB4 gene, the human ortholog of the murine Mdr2 gene, also show an increased risk for dysplasia with subsequent development of HCC and cholangiocarcinoma (26).

Our findings demonstrate that IL6 signaling in Mdr2^−/− mice, mediated in part via its soluble receptor (sIL-6R) and hepatocyte-specific STAT3, strongly suppresses liver injury, fibrosis, inflammation, hepatic steatosis, and hepatocarcinogenesis in a mechanism associated with reinforcement of senescence and the senescence-associated secretory phenotype (SASP) in the liver. These observations point to the existence of diverse tumor-suppressive roles of IL6 and STAT3 signaling in the development of chronic hepatitis–associated liver cancer.

Mice were maintained under specific pathogen–free (SPF) conditions in an animal facility with a temperature of approximately 23°C in a 12-hour light–dark cycle, and received sterile commercial rodent chow and water ad libitum. Maintenance of mice and all experimental procedures were performed in accordance with the Institutional Animal Care and Use Committee approved animal treatment protocols (license number OPRR-A01–5011).

Wild-type (WT) C57BL/6 mice were purchased from Harlan Laboratories. Mdr2^−/− (C57BL/6) mice were derived from Mdr2^−/− (FVB/NJ) as described previously (6). IL6 knockout (IL6^−/−; C57BL/6) mice were purchased from The Jackson Laboratory. Sgp130Fc transgenic mice (C57BL/6) were as described previously (27). Homozygosity of the sgp130Fc^+/+ and sgp130Fc^−/− (WT) alleles was determined by ELISA for soluble sg130Fc protein levels in the serum using antihuman sgp130 ELISA (R&D Systems, catalog no. DY228). Floxed-STAT3 (Stat3^floxP; C57BL/6) mice (kindly provided by E. Razin, Hebrew University, Jerusalem, Israel; ref. 28) were crossed with Alb-Cre^+/+ (C57BL/6) mice (kindly provided by D. Wallach, Weizmann Institute of Science, Rehovot, Israel; ref. 29) to form Stat3^floxP/+Alb-Cre^+/− mice, the offspring of which were then crossed to generate Stat3^floxPAlb-Cre (Stat3^ΔHep) and Stat3^floxP strains. Double-mutant Mdr2^−/−IL6^−/−, Mdr2^−/−sgp130Fc, and Mdr2^−/−Stat3^floxPAlb-Cre (Mdr2^−/−Stat3^Δhep) mice were generated by crossing Mdr2^−/− with either IL6^−/− or sgp130Fc, or Stat3^floxP Alb-Cre mice and their progeny were identified by PCR analysis and ELISA (sgp130Fc). For all strains, male and female mice were born with normal Mendelian frequency and were indistinguishable from parental mice with respect to skin discoloration, fertility, and general behavior. Blood samples were collected at the time of sacrifice for analysis of alanine aminotransferase (ALT) determined using the Reflotron system (Roche). Visual inspections and photography for liver tumors were performed upon sacrifice and tumors with diameter greater than 2 mm were counted and measured. Liver specimens were either fixed in 4% buffered formalin or snap-frozen in liquid nitrogen for further analysis. For in vivo IL6 neutralization experiments, rat anti-mouse IL6 mAb (MP5-20F3; BioXCell; 100 μg, i.p.) or control rat anti-horseradish peroxidase IgG (HRPN; BioXCell) was administrated to mice.

Livers samples were fixed in 4% buffered formaldehyde for 24 hours followed by 80% ethanol and embedded in paraffin blocks. Sections (5 μm) were stained by hematoxylin and eosin (H&E) or Masson trichrome for analysis of liver fibrosis by standard procedures. Morphologic analysis and dysplastic nodules were performed by an accredited pathologist in a blinded fashion and liver fibrosis scored according to the following criteria: grade II, septal formation and beginning of bridging fibrosis; grade III to IV, incomplete to complete bridging fibrosis. Fibrosis quantification was performed on twenty randomly photomicrographs of selected high power fields (×40) per liver and the fibrotic area (blue staining) was quantified using ImageJ software. Macrophages were stained using rat anti-mouse F4/80 antigen (Serotec), followed by anti-Rat HRP (Histofine) and developed with a DAB kit (Zymed). Phosphorylated STAT3 was stained using anti-p-STAT3 using monoclonal rabbit anti-mouse p-STAT3 (Tyr 705; Cell Signaling Technology) as described previously (30). Oil red O (ORO; Sigma) staining was performed on liver frozen sections (10 μm) fixed in 0.5% glutaraldehyde (Sigma) and counterstained with hematoxylin (Emmonya Biotech). Images of stained sections were quantified as percentage area stained positively per high power field was quantified using ImageJ software (ImageJ, RRID: SCR_003070) in 8 to 20 random fields per sample. See Supplementary Material for details on the mouse genotyping, MRI, RNA, antibodies and ELISA, SA-β-gal staining, hepatocyte isolation, and Western blot analysis.

Transcriptional profiling data have been deposited in the Gene Expression Omnibus database at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE37468.

Data were evaluated for significance by two-tailed Student t test or Mann–Whitney test, as indicated and unless otherwise noted. P ≤ 0.05 was considered significant for all analyses. Calculations were performed using GraphPad Prism 6.02 software (GraphPad Software, Inc).

To examine the association of IL6 expression in chronic hepatitis we confirmed and extended previous observations according to which chronic liver injury and inflammation in Mdr2^−/− mice is associated from an early age with elevated hepatic IL6 expression (Supplementary Fig. S1A; refs. 31, 32). Analysis of IL6 mRNA levels in isolated hepatocytes indicates that elevated IL6 expression in the Mdr2^−/− mice originates in part from hepatocytes. Western blot analysis confirmed that chronic liver injury and IL6 expression in livers from the Mdr2^−/− mice were also associated with strong upregulation of phosphorylated Stat3 (Supplementary Fig. S1B; ref. 32), which by immunostaining was found to localize both to infiltrating inflammatory cells and to hepatocytes in the proximity of the inflammatory foci concentrated within the periportal areas (Supplementary Fig. S1C).

To determine the importance of IL6 expression in chronic hepatitis in Mdr2^−/− mice we crossed Mdr2^−/− (C57BL/6) mice with IL-6^−/− (C57BL/6) mice to generate double-mutant Mdr2^−/−IL6^−/− mice. Importantly, in agreement with the previous observations of Mair and colleagues (32), IL6 deficiency exacerbated chronic liver injury, indicated by elevated aminotransferase levels (Fig. 1A), and was accompanied by a substantial reduction in phosphorylated Stat3 levels (Fig. 1B). Mdr2^−/−IL6^−/− mice also displayed elevated hepatic fibrosis with an overall degenerating histopathologic profile, which deteriorated from grade II in Mdr2^−/− mice to predominantly grade III to IV in Mdr2^−/−IL6^−/− mice (Fig. 1C and D). Similar effects of IL6 deficiency on liver injury and fibrosis were also observed in male mice (Supplementary Fig. S2A–S2C).

To determine the effect of IL6 deficiency on liver tumorigenesis in Mdr2^−/− mice we allowed single- and double-mutant mice to age and followed tumorigenesis by periodic assessment of liver cancer by MRI analysis at the ages of 7, 10, 12, and 14 months, and by visual scoring for liver tumors upon sacrifice at the age of 14 months (Fig. 1E and F). This analysis revealed significantly accelerated tumor appearance in the female Mdr2^−/−IL6^−/− mice, leading to a 2-fold increase in tumor incidence (P < 0.01, log rank) by 14 months of age compared with Mdr2^−/− controls (Fig. 1F). Visual scoring confirmed the MRI-based assessment of tumor incidence and further showed that IL6 deficiency increased the average tumor load by nearly 4-fold (P = 0.002), with a trend of increased tumor size (Fig. 1G and H). Importantly, an inspection of H&E-stained sections from livers of the 14-month-old mice revealed that preneoplastic, dysplastic nodules characteristic of Mdr2^−/− mice (4) appeared to be larger and more abundant in Mdr2^−/−IL6^−/− mice than in single-mutant controls (Fig. 1I). Treatment of Mdr2^−/− mice with anti-IL6 mAb at 3 months of age also increased tumor load compared to control IgG (Fig. 1J), recapitulating the effect of Il6 gene ablation. Male Mdr2^−/−IL6^−/− mice aged 14 months presented a similar trend of increased tumorigenesis but without reaching statistical significance by this age (Supplementary Fig. S2D), perhaps due to the slower development of liver cancer typical in male Mdr2^−/− strains (33). However, hepatic IL6 mRNA levels in older Mdr2^−/− mice did not vary substantially between female and male Mdr2^−/− mice (Supplementary Fig. S2E), suggesting that other gender-related factors may underlie the more aggressive tumorigenesis observed in the female mice.

In its classical form, IL6 signals on-target cells via a receptor complex consisting of the membrane-bound IL6 receptor (IL6R) and its co-receptor (gp130). However, the IL6R also exits in a soluble form (sIL6R) that when complexed with IL6 can target cells expressing gp130 alone in a mechanism called trans-signaling (34). This form of IL6 signaling is specifically blocked by the soluble form of gp130 (sgp130; ref. 34). To determine whether IL6 trans-signaling contributes to inflammation-induced hepatocarcinogenesis, we crossed Mdr2^−/− (C57BL/6) mice with transgenic sgp130Fc (C57BL/6) mice (27, 35) to generate Mdr2^−/−sgp130Fc mice. Assessment of liver injury (ALT) and Stat3 phosphorylation by Western blot analysis in young (3 months old) female mice demonstrated that blockade of trans-signaling in Mdr2^−/− mice increased chronic liver injury by about 2-fold (P < 0.001) while simultaneously reducing Stat3 phosphorylation levels by roughly 4-fold (P < 0.05; Fig. 2A and B). Long-term follow-up for tumor development by visual scoring upon the sacrifice of the mice at 14 months of age revealed significantly increased tumor incidence and tumor load in Mdr2^−/−sgp130Fc mice compared with Mdr2^−/− controls, together with a trend of increased tumor size (Fig. 2C and D). Evaluation of H&E-stained sections showed that dysplastic nodules in the livers of Mdr2^−/−sgp130Fc mice also appeared to be more abundant and larger than in Mdr2^−/− controls (Fig. 2E). This exacerbated phenotype of increased tumor incidence and tumor load was also observed in male Mdr2^−/−sgp130Fc mice, which were analyzed at both 14 and 16 months of age (Supplementary Fig. S2F). Importantly, loss of IL6 signaling in Mdr2^−/− mice did not affect mRNA levels of leukemia inhibitory factor (LIF), oncostatin M (OSM), or ciliary neurotrophic factor (CNTF; Supplementary Fig. S2G); thus, ruling out compensatory increases in these IL6 family member cytokines as likely contributing factors to the aggravated phenotype in these mice. These observations indicate that IL6 signaling, supported by IL6 trans-signaling, functions to suppress chronic injury and tumorigenesis in Mdr2^−/− mice.

In contrast to our findings in Mdr2^−/− mice, Naugler and colleagues (10) and Bergmann and colleagues (36) have shown that IL6 signaling and trans-signaling are crucial for HCC development in mice treated with diethylnitrosamine (DEN), a carcinogen-induced liver cancer model that provokes acute hepatic injury together with a strong compensatory regenerative response, but without chronic liver injury. Considering this different experimental outcome and the close linkage of IL6 expression with Stat3 phosphorylation observed also within hepatocytes in Mdr2^−/− mice, we postulated that in chronic liver injury Stat3 may act via its hepatoprotective function (32), rather than through its mitogenic function (37) to mediate the tumor-suppressive function of IL6. As such, we reasoned that Stat3 ablation in hepatocytes in Mdr2^−/− mice would recapitulate the effect of IL6 deficiency.

Mair and colleagues (32) have previously shown that conditional ablation of Stat3 in both hepatocytes and cholangiocytes of Mdr2^−/− mice leads to severe morbidity and premature mortality after weaning (32), thus precluding an assessment of its long-term effect on hepatocarcinogenesis. To avoid this outcome, we chose a strategy of hepatocyte-targeted Stat3 ablation, with the expectation that exclusion of cholangiocytes from the target population would diminish the severe consequences of its ablation on cholangitis and enable long-term survival. To this end, we crossed Mdr2^−/− mice with mice carrying the Stat3^floxP and hepatocyte-specific Alb-Cre recombinase (Cre) alleles (28, 29) to produce Mdr2^−/−Stat3^floxPAlb-Cre (Mdr2^−/−Stat3^Δhep) mice. Assessment of liver injury at an early age (3 months) in these mice confirmed the previous observations (32) to which conditional Stat3 ablation aggravated bile acid–induced liver injury (Fig. 3A). Female Mdr2^−/−Stat3^Δhep mice also subsequently displayed markedly increased tumorigenesis with increased tumor loads of about 2-fold (P = 0.039 and P = 0.033 at 14 and 16 months, respectively) in comparison with Mdr2^−/−Stat3^floxP control mice (Fig. 3B and C). Immunostaining confirmed the presence of the deleted Stat3 allele and the loss of IL6-induced Stat3 phosphorylation in the Mdr2^−/−Stat3^Δhep mice (Supplementary Fig. S3A and S3B), thus excluding the possibility that survival of the Mdr2^−/−Stat3^Δhep mice and tumor development was due to a failure to eliminate the floxed Stat3 allele. Intriguingly, the tumors in the Mdr2^−/−Stat3^Δhep mice appeared slightly smaller (P < 0.05) than in control mice (Fig. 3C), suggesting that while Stat3 signaling protects hepatocytes from tumor initiation, it nevertheless contributes toward tumor growth. Consistent with their increased tumor load, dysplastic nodules in the livers of Mdr2^−/−Stat3^Δhep mice appeared to be more abundant and larger (Fig. 3D), suggesting that the dysplastic nodules, which are considered to be preneoplastic lesions, are more closely associated with tumor initiation than tumor growth. Male Mdr2^−/−Stat3^Δhep mice displayed similar increases in both ALT levels and tumorigenesis, the latter being delayed in comparison with the female mice, and consistent with the delayed tumorigenic phenotype in male Mdr2^−/− mice (Supplementary Fig. S4A and S4B). Tumorigenesis in Mdr2^−/−Stat3^floxP control mice was similar to that of the Mdr2^−/−Alb-Cre control mice (Supplementary Fig. S4C), indicating that expression of the Cre transgene per se did not contribute to the tumorigenic phenotype. qPCR analysis showed that IL6 mRNA levels in Mdr2^−/−Stat3^Δhep livers were not diminished by hepatocyte-targeted Stat3 ablation, indicating that the effects of Stat3 signaling deficiencies were due to the reduction in hepatocyte-targeted Stat3 signaling and not to a reduction in IL6 levels (Supplementary Fig. S4D). IL6 and Stat3 are involved in a wide range of pivotal cellular processes that can affect tumorigenesis, including apoptosis. Apoptosis was marginally increased in livers of IL6-deficient Mdr2^−/− strains as indicated by levels of cleaved caspase-3 by Western blot analysis (Supplementary Fig. S5). This finding is in agreement with the role of IL6 in preventing cell death, but inconsistent with a hypothesis of apoptosis-mediated suppression of tumorigenesis in Mdr2^−/− mice (36). Moreover, because the effects Stat3 ablation in Mdr2^−/−Stat3^Δhep mice are hepatocyte targeted, as opposed to those in IL6^−/− and sgp130Fc mice, which affect total IL6 signaling, these findings indicate that Stat3 signaling within hepatocytes is a crucial mediator of the IL6 hepatoprotective function.

Current clinical and experimental evidence points to hepatic lipid accumulation and inflammation, as in NAFLD and NASH, as aggravating factors in the development of liver cancer (9). Bioinformatics analysis and QIAGEN Ingenuity Pathway Analysis (IPA) of the differential gene expression patterns in nontumorous liver samples from female Mdr2^−/− and control WT mice at 14 months of age revealed a close association between tumorigenesis in Mdr2^−/− mice with similar processes linked to hepatocarcinogenesis in humans and mice including fatty acids metabolism and lipid accumulation, inflammation, and hormone metabolism (Supplementary Table S1). Because previous studies have shown that tumorigenesis in Mdr2^−/− mice is also inherently associated with hepatic steatosis and inflammation (4), we assessed the effect of IL6/Stat3 signaling deficiencies on hepatic lipid accumulation and inflammation in Mdr2^−/− mice. Quantification of lipid content by ORO staining revealed strikingly increased hepatic steatosis in Mdr2^−/− mice possessing either IL6 or Stat3 signaling deficiencies, particularly in Mdr2^−/−sgp130Fc mice (Fig. 4). Increased lipid accumulation appears to be due in large part to the disruption in IL6 and Stat3 signaling, because hepatic steatosis is also significantly elevated in single-mutant IL6^−/− and Stat3^Δhep mice after the age of about 6 months (19, 38). Thus, aggravated hepatic steatosis in the double-mutant strains is likely an additive outcome of independent attributes linked independently to both Mdr2 and IL6/Stat3 signaling deficiencies and appears closely linked to later events associated with tumor development in the mature mice.

Because we and others have demonstrated that active inflammation involving monocytes/macrophages and T cells contributes to hepatocarcinogenesis in Mdr2^−/− mice (4, 6, 39), we also assessed the effect of IL6 deficiency on hepatic inflammation levels. Inspection of H&E-stained liver sections suggested the appearance of increased inflammation within the periportal spaces of Mdr2^−/−IL6^−/− mice at both 3 and 14 months of age (Supplementary Fig. S6A). F4/80 immunostaining showed that macrophages levels were significantly increased in the livers of Mdr2^−/−IL6^−/− and Mdr2^−/−sgp130Fc mice and marginally increased also in Mdr2^−/−Stat3^Δhep mice relative to single-mutant Mdr2^−/− controls (P = 0.01, P < 0.001, and P = 0.03, respectively; Fig. 5A). In this respect, we noted that IL6/Stat3 signaling–deficient Mdr2^−/− mice generally displayed elevated levels of mRNA encoding the chemokine, Ccl2, in line with its role in attracting macrophage infiltration (Fig. 5B). Macrophage infiltration was not increased in the noninflamed livers of control IL6^−/− and Stat3^Δhep mice (Supplementary Fig. S6B), suggesting that the increased macrophage infiltration in the double-mutant Mdr2^−/− mice was cholangitis linked, and related to their increased hepatic injury. While T cells including CD4⁺ Th cells and FoxP3⁺ Treg cells have also previously been proposed to alternatively suppress or promote hepatocarcinogenesis in humans and mice (39–41), we found increased T-cell populations in Mdr2^−/− mice compared with WT controls, but no differences between Mdr2^−/− and Mdr2^−/−IL6^−/− mice (Fig. 5C). Also, CD4⁺ helper T-cell populations, although significantly elevated in all Mdr2^−/− strains relative to WT controls, showed no significant differences between Mdr2^−/− mice and double-mutant strains (Fig. 5D), as were FoxP3⁺ Treg levels, with the exception of Mdr2^−/−Stat3^Δhep mice, in which FoxP3⁺ Treg cells appeared to be substantially elevated (Supplementary Fig. S6C and S6D). Thus, the absence of consistent meaningful shifts in T-cell populations was inconsistent with a hypothesis of their involvement in the IL6 deficiency–enhanced tumorigenesis. Together, these observations show that IL6/Stat3 signaling deficiency–enhanced tumorigenesis is associated with an underlying phenotype of aggravated NASH, consisting of increased hepatic steatosis and macrophage infiltration.

Hormone signaling by estrogens has previously been identified as an important protective factor against liver injury and hepatocarcinogenesis (10). Because our bioinformatics analysis suggested that primary cholangitis in female Mdr2^−/− mice is associated with aberrant hormone metabolism and signaling (Supplementary Table S1), we postulated that reduced hormone-mediated protection may also contribute to hepatocarcinogenesis in Mdr2^−/− mice. Quantitative analysis of mRNAs encoding genes related to estrogen signaling and metabolism in livers of female Mdr2^−/− mice by qPCR, including mRNAs for estrogen receptor (Esr1), Cyp17a1 and Sult1e1, confirmed the bioinformatics assessment and suggested that IL6 deficiency may aggravate aberrant hormonal metabolism in the Mdr2^−/− mice (Supplementary Fig. S7A). Thus, expression of both Esr1 and Cyp17a1, a key enzyme in the steroidogenic pathway that produces estrogens, appears strongly downregulated in the female Mdr2^−/− mice; while Sult1e1, a key enzyme in estrogen homeostasis and metabolism, was significantly upregulated. This picture of reduced hormonal metabolism in female Mdr2^−/− mice appeared to be disrupted further in mice with IL6 deficiency (Supplementary Fig. S7A).

To test whether hormonal signaling by estrogens protects against hepatocarcinogenesis in Mdr2^−/− mice, we subjected mice aged 2 months to either ovariectomy or sham operations and assessed the mice for liver injury and tumor development at the age of 14 months. Surprisingly, these analyses showed that ovariectomy of the Mdr2^−/− mice significantly reduced chronic liver injury (ALT), hepatomegaly, and hepatic IL6 mRNA expression compared with sham-treated controls (Supplementary Fig. S7B–S7D). Moreover, ovariectomized mice ultimately displayed marginally reduced tumor loads, as well as fewer and smaller dysplastic nodules (Supplementary Fig. S7E–S7G). These findings are inconsistent with a hypothesis of estrogen-mediated suppression of liver injury and tumorigenesis in female Mdr2^−/− mice, thus ruling out the loss of hormonal signaling as a driver of enhanced tumorigenesis in the IL6-deficient mice.

Previous findings by our lab and confirmed by others demonstrate that cellular senescence is an inherent feature of the Mdr2^−/− hepatic microenvironment manifest by upregulation of senescence markers including SA-β-gal, and p16^INK4a, associated with stellate cell activation and fibrosis (24, 42). Consistent with these observations, bioinformatics analysis revealed significant upregulation in Mdr2^−/− livers of genes associated with “negative regulation of cell proliferation” (Supplementary Table S1), including numerous factors constituting the SASP (43), such as Tgfb2, Timp1, Timp2, Ccl2, and Cxcl14 (Supplementary Table S2). Because senescence can act to impede the development of liver cancer and IL6 is a crucial element of the SASP (43–45), we next assessed the effect loss of IL6 signaling on senescence and SASP markers in livers of Mdr2^−/− mice.

Real-time qPCR analysis confirmed that expression of mRNAs encoding the senescence marker, p16^INK4a and SASP markers Pai-1, Ccl2, Timp1, and Timp2 were significantly elevated in livers of young (3 months) Mdr2^−/− mice relative to WT controls (Fig. 6A). Levels of mRNAs encoding cGAS and Sting were also significantly elevated in livers of Mdr2^−/− mice (Fig. 6A), consistent with the increased presence of senescence and micronuclei in Mdr2^−/− hepatocytes (16) and the role of cGAS and Sting in detecting senescence-related cytoplasmic DNA and also micronuclei (46). In contrast, IL6-deficient strains, Mdr2^−/−IL6^−/− and Mdr2^−/−sgp130Fc, showed striking reductions in the expression of cGAS and Sting and SASP mRNAs, most of which remained at or near WT levels. These reductions were accompanied by dramatic reductions in senescence as indicated by levels of p16^INK4a mRNA and SA-βgal positive–stained cells in double-mutant mice (Fig. 6A and B). Moreover, consistent with the role of the SASP in reinforcing senescence-imposed cell-cycle arrest (44), levels of compensatory hepatocyte proliferation (Ki-67) increased by about 50% in young IL6 signaling–deficient strains relative to Mdr2^−/− controls (P < 0.05; Fig. 6C). Thus, aggravated hepatocarcinogenesis upon IL6 deficiency associates with a profound disruption of the SASP and senescence in Mdr2^−/− mice.

Collectively, these findings point to broad functional roles of IL6 and Stat3 signaling in both the suppression of protumorigenic phenotypes in chronic hepatitis, including hepatic steatosis and inflammation, and in the reinforcement of senescence-mediated suppression of compensatory hepatocytes proliferation, which together may function to suppress tumorigenesis in Mdr2^−/− mice.

IL6 supported injury, inflammation, and regeneration have been described as microenvironmental drivers of liver cancer development, particularly involving acute liver injury (5, 10). However, our findings demonstrate that within the context of chronic liver injury IL6 and Stat3 signaling play crucial roles in suppressing hepatocarcinogenesis. This conclusion is supported by observations derived from three independent molecular genetic approaches in which IL6 and Stat3 signaling were eliminated or substantially reduced in the inflammation-induced HCC model, the Mdr2^−/− mouse. Importantly, all perturbations of IL6 and Stat3 signaling in the mice resulted in uniformly increased chronic liver injury, a key coindication commonly associated with the development of liver cancer in humans, thus confirming the protective role of IL6 in liver injury and fibrosis (17, 18, 32). In this setting of chronic hepatitis, perturbations in IL6/STAT3 signaling also increased the appearance of dysplastic nodules, which are bona fide preneoplastic lesions (4), and ultimately both increased and accelerated the appearance of HCC, thus supporting the notion of a causal link between chronic injury, wound healing, and hepatocarcinogenesis.

How does IL6/STAT3 signaling suppress hepatocarcinogenesis in the context of chronic liver injury? Our findings indicate that total IL6 and hepatocyte-targeted Stat3 deficiencies in Mdr2^−/− mice strongly aggravate chronic inflammation and hepatic steatosis, which are prominent protumorigenic conditions in humans (9). Mechanistically, this observation is substantiated by independent observations demonstrating a crucial role of IL6 and Stat3, in preventing adiposity and hepatic steatosis (19, 38).

However, our findings also reveal a close relationship between low-level IL6 expression associated with chronic liver injury and senescence-mediated suppression of hepatocarcinogenesis. Through induction of durable cell-cycle arrest and immune-mediated elimination of cells with genomic instability and/or oncogene activation, senescence is recognized as one of the most important cell-intrinsic suppressive mechanisms in the prevention of neoplastic transformation (47). Indeed, Kang and colleagues (45) have shown that senescence surveillance of premalignant hepatocytes acts to impede the development of liver cancer. Our current findings demonstrate that IL6 deficiency in these mice leads to a frank collapse in the fabric of the SASP, and is associated with increased hepatocyte proliferation. This observation is also consistent with studies showing that IL6 and SASP factors play crucial roles in the reinforcement of cellular senescence (44). In some respects, this senescence-mediated role of IL6 signaling resembles that demonstrated by Pencik and colleagues (48) in hampering prostate cancer development. Together, this suggests that loss of IL6 in Mdr2^−/− mice may release hepatocytes from paracrine senescence-imposed cell-cycle arrest.

These conclusions contrast sharply with the commonly accepted notion of the protumorigenic roles of IL6 and STAT3 signaling in models of chemical-induced liver cancer, which are associated with acute liver injury (10, 36). How, then, does IL6/Stat3 signaling promote hepatocarcinogenesis in chemical-induced liver cancer, but does the opposite in the context of chronic injury–associated liver cancer?

It is now appreciated that molecular and cellular mechanisms underlying acute liver injury–associated hepatocarcinogenesis can differ greatly from those underlying liver cancer arising on a background of chronic injury (11). Unlike models in which neonatal-administered hepatocarcinogens, such as DEN, induce acute liver injury and regeneration followed by rapid HCC development, in the Mdr2^−/− model liver cancer develops slowly on a background of chronic, low-level liver injury together with notable inflammation and fibrosis. Importantly, the Mdr2 model is unique in that it recapitulates the presence of hepatic senescence typically associated with chronic hepatitis present in humans during hepatocarcinogenesis, thus closely mimicking the events leading to HCC development in human patients (23, 24). However, at the same time, we have shown that in Mdr2^−/− mice IL6 is also crucial for partial hepatectomy-induced accelerated hepatocarcinogenesis (16). Thus, it is evident that IL6 can also play a decisive role in promoting liver cancer when appearing on the background of chronic liver injury.

The key to understanding this apparent contradiction may lie in the differential roles that IL6 and Stat3 play in liver regeneration versus senescence. In autocrine and paracrine senescence inherent to chronic cholangitis, SASP factors including TGFβ and IL6 act to reinforce senescence and suppress cell division (43, 44, 49). On the other hand, upon acute loss of liver parenchyma IL6 expressed at high levels crucially facilitates the hepatic regenerative response mediated by growth factors, including hepatocyte growth factor (HGF; ref. 50). During liver regeneration hepatocytes become refractory to TGFβ signaling (51) perhaps supported by high IL6 expression, which, as we have shown previously, strongly suppresses hepatic SMAD2 phosphorylation and TGFβ signaling, thus contributing to the HGF-induced DNA synthetic response in hepatocytes (52).

DEN is a strong mutagen that induces hepatocyte injury followed by a short wave of regeneration associated with hepatocyte proliferation. Similarly, bile acids also induce hepatocyte injury and DNA damage together with genomic instability that increases substantially when placed in the context of resection-induced liver regeneration (16, 53). Under these circumstances, IL6 promotes hepatocyte proliferation and thus is associated with “propagating” DEN-induced point mutations and with generation of chromosomal aberrations, some of which are carcinogenic. As such, when the number of tumors are assessed months after DEN administration in WT mice or after liver resection in Mdr2^−/− mice (16), the absence of IL6 is associated with a decrease in tumors as a result of fewer replicating hepatocytes many months prior, and thus fewer initiated hepatocytes and the carcinomas that would develop from them. In contrast, in the presence of chronic liver injury, senescence-associated low-level IL6 expression suppresses hepatocyte proliferation (49). Because IL6 at low levels appears to support senescence in chronic cholangitis, the absence of IL6 would be associated with an increase in tumors resulting from replicating hepatocytes containing chromosomal aberrations.

Thus, it is conceivable that in hepatocarcinogenesis IL6 may display opposing, context-dependent functions, that is, initiating and promoter functions when associated with acute tissue loss and regeneration, and suppressor functions when expressed in the context of chronic injury and senescence. However, evidence shows that circumstances exist under which senescence can also promote tumorigenesis (54). Ironically then, such a protumorigenic view of senescence-associated IL6 expression may also be relevant to liver cancer, in particular, for cells such as cholangiocytes, for which IL6 acts as a growth factor per se (55).

Finally, a growing number of clinical studies have found the presence of elevated serum sgp130 levels associated with the development of obesity, diabetes, and insulin resistance in some adult patients (21). An extrapolation of these clinical findings together with the findings of this study predicts that elevated sgp130 serum levels and related genetic polymorphisms may also be associated with and serve as diagnostic markers for the severity of NAFLD and NASH, and perhaps for increased odds of developing liver cancer in some patients.

Collectively, our findings demonstrate that in the context of chronic liver injury IL6 likely plays an inherent role in impeding hepatocarcinogenesis by suppressing the protumorigenic phenotypes such as NAFLD and NASH, and by reinforcing the antitumorigenic effects of the SASP. This conclusion may also have implications for therapeutic strategies based on inhibition of IL6 or STAT3 signaling in the treatment of certain forms of chronic inflammation, in particular for the over 1 million patients treated in the past decade with anti-IL6R antibodies.

No disclosures were reported.

A. Shriki: Investigation, writing–original draft. T. Lanton: Investigation, writing–original draft, writing–review and editing. A. Sonnenblick: Funding acquisition, investigation, writing–review and editing. O. Levkovitch-Siany: Validation, investigation. D. Eidelshtein: Investigation. R. Abramovitch: Formal analysis, investigation. N. Rosenberg: Investigation. O. Pappo: Formal analysis. S. Elgavish: Formal analysis. Y. Nevo: Data curation, formal analysis. R. Safadi: Resources. A. Peled: Conceptualization, resources, writing–review and editing, data review. S. Rose-John: Conceptualization, resources, funding acquisition, writing–review and editing. E. Galun: Conceptualization, resources, supervision, funding acquisition, writing–review and editing. J.H. Axelrod: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, writing–review and editing, project administration.

The authors are grateful to Deborah Olam for excellent technical assistance, to Neta Barashi for helpful discussions and advice, to Tali Bdolah-Abram for expert advice on statistical evaluations, and to the Core Research Facility (CRF) and staff members at The Faculty of Medicine, the Hebrew University of Jerusalem. J.H. Axelrod was supported by grants to from the Israel Science Foundation (ISF 923/14) and the Israel Cancer Research Fund (8004906). The work of E. Galun was supported by grants from the ISF collaboration with Canada (2473/2017), the personal ISF (486/2017) and the ISF and ICORE (ISF 41/2011), the Deutsche Forschungsgemeinschaft, Bonn, Germany (SFB841), and the ERC advance - GA number 786575 – RxmiRcanceR (to E. Galun) and the Israel Ministry of Science and Technology (MOST). The work of E. Galun was also supported by the Kron, Raskin, and Robert Benson foundations. The work of S. Rose-John was supported by the Deutsche Forschungsgemeinschaft, Bonn, Germany (SFB877, Project A1; SFB841, Project C1) and by the Cluster of Excellence “Inflammation at Interfaces.” A. Sonnenblick was supported by grants from the Kass Research Award-American Physicians Fellowship for Medicine in Israel, the Israel Cancer Association, Israel Cancer Research Foundation Fellowship, Hadassah Medical Center Internal Grant, and the Morasha Program of the Israel Science Foundation grant (ISF1728/11).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.