To test whether the increased NASH observed in Asc- and caspase-1-deficient mice was mediated by IL-1β or IL-18, we performed similar experiments using mice deficient in either the IL-1 receptor (Il1r^−/−), or IL-18 (Il18^−/−). Il1r^−/− mice did not show any changes in the severity of NASH when compared to wt mice when fed MCDD (Supplementary fig. 1a–b). In contrast, but similar to Asc^−/− and caspase-1^−/− mice, MCDD-fed Il18^−/− animals featured a significant exacerbation of NASH severity (Fig. 1g–h; Supplementary fig. 1c).

To assess the role of the NLRP3 inflammasome in NASH progression, we fed singly-housed Nlrp3^−/− and wt animals MCDD for 24 days and evaluated disease progression. Nlrp3^−/− mice developed exacerbated NASH compared to wt mice as judged by increased levels of serum ALT and AST, plus NAFLD activity inflammation scores (Fig. 1e–f; Supplementary fig. 1c). Remarkably, bone marrow (BM) chimeric mice in which NLRP3 and ASC deficiency was limited to the hematopoietic compartment did not show any increase in the severity of NASH when compared to wt mice reconstituted with wt BM (Supplementary fig. 3a–f). Likewise, knock-in mice that specifically express a constitutively active NLRP3 inflammasome in CD11c+ myeloid cells (Nlrp3KI;CD11c⁺-Cre) or hepatocytes (Nlrp3KI;Albumin-Cre)¹⁶ did not feature any significant differences in MCDD-induced NASH severity as compared to wt mice (Supplementary figure 3g–l). These results suggest that aberrations in inflammasome function in cells other than hepatocytes or myeloid cells are key determinants of the enhanced disease progression in inflammasome-deficient mice.

We recently discovered that inflammasomes act as steady-state sensors and regulators of the colonic microbiota, and that a deficiency in components of two inflammasomes, NLRP6¹⁷ and NLRP3 (unpublished), both of which include ASC, caspase-1, and involve IL-18 but not IL-1R, results in the development of an altered transmissible, colitogenic intestinal microbial community¹⁷. This microbiota is associated with increased representation of members of Bacteroidetes (Prevotellaceae) and the bacterial phylum TM7, and reductions in representation of members of the genus Lactobacillus in the Firmicutes phylum¹⁷. Moreover, electron microscopy (EM) studies disclosed aberrant colonization of crypts of Lieberkuhn with bacteria with morphologic features of Prevotellaceae¹⁷. Therefore, we sought to investigate whether enhanced NASH severity in inflammasome-deficient mice is driven by their altered microbiota. Strikingly, co-housing of Asc^−/− and Il18^−/− mice with wt animals for 4 weeks (beginning at 4–6 weeks of age), prior to induction of NASH with MCDD resulted in significant exacerbation of NASH in the wt cagemates [which we will refer to as wt(Asc^−/−) and wt(Il18^−/−), respectively, in the following text], as compared to singly-housed, age- and gender-matched wt controls (n=5–7 mice/genotype/housing condition). In co-housed wt mice, disease severity reached comparable levels to that of co-housed Asc^−/− and Il18^−/− mice (Fig. 2a–h). Moreover, significantly increased numbers of multiple inflammatory cell types were present in the liver of wt(Asc^−/−) compared to wt mice (Supplementary fig. 2a). Similar findings were observed in wt mice co-housed with caspase-1^−/−, Nlrp3^−/−, and Nlrp6^−/− mice (Supplementary fig. 4a–f). To exclude the possibility that aberrant microbiota represented in all mice maintained in our vivarium, we co-housed wt mice with other strains of NLR-deficient mice that were either obtained from the same source as Asc^−/− and Nlrp3^−/− mice (Nlrc4^−/−, Nlrp12^−/−), or generated in our laboratory (Nlrp4c^−/−). None of these strains featured a similar phenotype (Supplementary fig. 4g–l). These results indicate that the transmissible colitogenic microbiota present in inflammasome-deficient mice is a major contributor to their enhanced NASH. In agreement with this, combined antibiotic treatment with Ciprofloxacin and Metronidazole, previously shown to abrogate the colitogenic activity of the microbiota associated with inflammasome-deficient mice associated microbiota¹⁷, significantly reduced the severity of NASH in Asc^−/− mice, and abolished transmission of the phenotype to wt(Asc^−/−) animals (Supplementary fig. 5).

To ascertain the effects of MCDD on the gut microbiota, we performed a culture-independent analysis of amplicons generated by primers directed against variable region 2 of bacterial 16S rRNA genes of fecal samples collected from wt mice co-housed with Asc−/− animals [wt(Asc−/−)], their Asc−/− cagemates [Asc−/−(wt)] as well as singly-housed wt controls 1d and 12d prior to, and 7, 14 and 19 days after initiation of this diet (n=20 animals; 8 singly-housed wt, 6 co-housed wt and 6 Asc−/− mice). The structures of bacterial communities were compared based on their phylogenetic content using unweighted UniFrac. The results are illustrated in Figure 3. Table 1 provides a list of all phylotypes that, based on criteria outlined in Methods, discriminate co-housed wt(Asc−/−) from their singly-housed wt counterparts. Prior to MCDD, and consistent with our previous findings¹⁷, the fecal microbiota of wt(Asc−/−) mice adopted a configuration similar to Asc−/− cagemates, including the appearance of Prevotellaceae (Table 1 and Fig. 3a–c). There was also a significant increase in proportional representation of members of the family Porphyromonadaceae (primarily in the genus Parabacteroides) in wt(Asc−/−) mice compared to their singly-housed wt counterparts (Fig. 3d,e). The representation of Porphyromonadaceae was greatly increased in both the co-housed wt and Asc−/− mice (but not in singly-housed wt) when they were switched to a MCDD diet (p<0.01; t-test; Fig. 3d). A dramatic increase in the family Erysipelotrichaceae (phylum Firmicutes) also occurred with MCDD in co-housed animals, to a level that was >10% of the community (Fig. 3f). Although the Prevotellacaeae decreased when co-housed wt(Asc−/−) mice were placed on MCDD, their relative abundance remained significantly higher than in singly-housed wt animals (Fig. 3c).

Together, these results pointed to the possibility that members of the altered intestinal microbiota in inflammasome-deficient MCDD-treated mice may promote a signaling cascade in the liver upon translocation, resulting in progression to NASH in susceptible animals. Toll Like Receptors (TLR) play a major role in NAFLD pathophysiology due to the liver’s exposure to relatively large amounts of PAMPs derived from the intestine and delivered via the portal circulation^18–20. Therefore, we hypothesized that TLR signaling mediates the increased susceptibility to progression to NASH in mice exposed to the gut microbiota of Asc^−/− animals. Myd88^−/−;Trif^−/−(Asc^−/−) mice are devoid of all TLR signaling pathways. When co-housed with Asc^−/− mice between 5 and 9 weeks of age, they exhibited decreased severity of NASH after exposure to MCDD for 24 days, compared to wt(Asc^−/−) mice (Supplementary fig. 6a–b). To define which specific TLRs was responsible for the inflammatory response, we co-housed Tlr4-, Tlr9-, or Tlr5-deficient mice with Asc^−/− animals and induced NASH with MCDD as previously described. Similar to wt mice, Tlr5^−/− mice co-housed with Asc^−/− mice ((Tlr5^−/−(Asc^−/−)) featured a statistically significant exacerbation of hepatic injury, steatosis, and inflammation, when compared to singly-housed Tlr5^−/− controls (Fig. 3c; supplementary fig. 6g–h), indicating that TLR5 does not mediate the microbiota-mediated exacerbation in disease severity. In contrast, Tlr4^−/−(Asc^−/−) and Tlr9^−/−(Asc^−/−) mice did not show the customary increase in disease severity when compared to their singly-housed Tlr4^−/− and Tlr9^−/− counterparts (Fig. 3a–b; supplementary fig. 6c–f).

These observations suggest that intact bacteria or bacterial products derived from the intestine trigger TLR4 and TLR9 activation, which results in an increased rate of disease progression in mice that house a colitogenic gut microbiota associated with inflammasome deficiency (i.e. Asc^−/−, wt(Asc^−/−) mice). Efforts to sequence 16S rRNA genes that might be present in total liver DNA, microbial qPCR assays of portal vein blood DNA, histologic analysis of intact liver, and aerobic and anaerobic cultures of liver homogenates did not reveal any evidence of intact bacteria in wt or Asc^−/− mice fed MCDD (data not shown). Notably, transmission electron microscopy studies of colon harvested from wt and Asc^−/− mice revealed an abundance of electron dense material, suggestive of some black-pigmented bacterial species, in colonic epithelial cells and macrophages located in the lamina propria of Asc^−/− mice but not in wt animals (Fig. 3e; supplementary fig. 7c). In agreement with previous results, we did not detect any translocation of intact bacteria (Fig. 3e; supplementary fig. 7c).

These observations provide evidence for the uptake of bacterial products from locally invasive gut microbes in Asc^−/− mice (Fig. 3e; supplementary fig. 7c). If microbial components, rather than whole organisms were transmitted to the liver then they should be detectable in the portal circulation. Indeed, levels of TLR4 and TLR9 agonists but not TLR2 agonists (assayed by their ability to activate TLR reporter cell lines), were markedly increased in the portal circulation of MCDD-fed wt(Asc^−/−), and Asc^−/− mice as compared to wt controls (n= 13–28 mice/group; fig. 3d; supplementary figure 7a–b). Altogether, these results indicate a mechanism whereby TLR4 and TLR9 agonist efflux from the intestines of inflammasome-deficient mice or their co-housed partners, through the portal circulation, to the liver where they trigger TLR4 and TLR9 activation that in turn results in enhanced progression of NASH.

We next explored the downstream mechanism whereby microbiota-induced TLR signaling enhances NASH progression. Pro-inflammatory cytokines, and in particular tumor necrosis factor (TNF)-α, a downstream cytokine of TLR signaling, are known to contribute to progression of hepatic steatosis to steatohepatitis and eventually hepatic fibrosis in a number of animal models and in human patients^21,22. Following induction of NASH by the MCDD, hepatic Tnfα mRNA expression was significantly upregulated in Asc^−/− and Il18^−/− mice, which exhibit exacerbated disease, but not in Il1r^−/− mice, which do not (Supplementary fig. 8a–c). Moreover, Tnfα mRNA levels were significantly increased in wt mice that had been previously co-housed with Asc^−/− or Il18^−/− mice and then fed MCDD (Supplementary fig. 8d–e), suggesting that its enhanced expression was mediated by elements of the microbiota responsible for NASH exacerbation. In contrast, we did not observe any changes in Il6 or Il1β mRNA levels in the livers of Asc^−/−, Il18^−/−, or Il1r^−/− mice compared to wt controls (Supplementary fig. 8a–c). Furthermore, while MCDD-administered singly-housed Tnfα^−/− mice had comparable NASH severity to singly-housed wt animals (Fig. 3f–h; supplementary fig. 8f) co-housing with Asc-deficient mice prior to MCDD induction of NASH resulted in increased liver injury, hepatic steatosis, and inflammation in wt mice but not in Tnfα^−/− mice (Fig. 3f–h; supplementary fig. 8f). These results indicate that TNF-α mediates the hepatotoxic effects downstream of the transmissible gut microbiota present in Asc^−/− mice.

The aberrant gut microbiota in NLRP3 and NLRP6 inflammasome-deficient mice induces colonic inflammation through epithelial induction of CCL5 secretion¹⁷. To test whether this colon inflammation influences TLR agonist influx into the portal circulation and NASH progression, we induced NASH in wt and Ccl5^−/− mice that had been either singly-housed or co-housed. MCDD-fed, singly-housed wt and Ccl5^−/− mice exhibited equivalent levels of NASH severity (Supplementary fig. 9a–c), suggesting that CCL5 does not play a role in the early stages of NAFLD/NASH in the absence of the inflammasome-associated colitogenic microbiota. However, we documented significantly increased levels of liver injury, inflammation, and steatosis in wt(Asc^−/−) but not Ccl5^−/−(Asc^−/−) mice (Fig. 5a–c), which led us to conclude that CCL5 is required for the exacerbation of disease through cohousing with inflammasome-deficient mice. Moreover, Ccl5^−/−(Asc^−/−) animals exhibited significantly reduced levels of TLR4 and TLR9 agonists in their portal vein blood than wt(Asc^−/−) mice (Supplementary fig. 9d–f). Together, these results suggest that microbiota-induced subclinical colon inflammation is a determining factor in the rate of TLR agonist influx from the gut, and in NAFLD/NASH progression.

The MCDD system is a common model for studying inflammatory processes associated with progression from NAFLD to NASH, yet it lacks many of the associated metabolic phenotypes of NAFLD, such as obesity and insulin resistance²³. As such, our results in this model might conceivably be limited to the way dysbiosis can influence NASH progression in patients with enhanced intestinal permeability, such as those with IBD²⁴, but not for the majority of patients who suffer from NASH in the context of metabolic syndrome. To test whether alterations in the gut microbiota of inflammasome-deficient mice may affect the rate of progression of NAFLD and other features associated with metabolic syndrome, we extended our studies to genetically obese mice and mice fed with high fat diet (HFD).

Leptin-receptor deficient (db/db) animals develop multiple metabolic abnormalities, including NAFLD and impaired intestinal barrier function²⁵, that closely resemble the human disease²⁶. However, significant hepatocyte injury, inflammation, and fibrosis are not observed in the absence of a “second hit”²⁷. Upon cohousing of db/db mice with Asc^−/− (db/db(Asc^−/−)) or WT mice (db/db(wt)) for a period of twelve weeks, and as previously shown for Asc^−/− mice¹⁷, the colon and ileum of all db/db(Asc^−/−) mice exhibited mild to moderate mucosal and crypt hyperplasia (Fig. 5d–f), that was not seen in db/db(wt) mice.

Strikingly, co-housed db/db(Asc^−/−) mice also showed increased levels of hepatocyte injury as evidenced by higher levels of ALT and AST in their sera, and significantly exacerbated steatosis and hepatic inflammation scores when compared with db/db(wt) mice (Fig. 5g–i). In addition to a parenchymal inflammatory exudate, patchy areas of markedly degenerated hepatocytes and hepatocytes undergoing necrosis were observed, but only in db/db(Asc^−/−) animals (Fig. 5f). Furthermore, some areas of congestion were seen in the centro-lobular zone as well as in the hepatic parenchyma – features that resemble Peliosis Hepatis, a condition observed in a variety of pathological settings including infection (data not shown). In accord with our MCDD results, hepatic Tnfα mRNA levels were significantly higher in co-housed db/db(Asc^−/−) mice than in db/db(wt) animals (Fig. 5j). Again, no significant differences were observed in hepatic Il6 or Il1β mRNA levels (Fig. 5j).

Interestingly, db/db(Asc^−/−) mice developed significantly more weight gain as compared to db/db(wt) after 12 weeks of co-housing (Figure 6a), suggesting that the inflammasome-associated gut microbiota could exacerbate additional processes associated with the metabolic syndrome, such as obesity. To address this possibility, we monitored multiple metabolic parameters in wt, wt(Asc^−/−), and Asc^−/− mice fed a high fat diet (HFD) for 12 weeks. Strikingly, Asc^−/− mice gained body mass more rapidly and featured enhanced hepatic steatosis (Fig. 6b–c; supplementary fig. 11f). Asc^−/− mice also showed elevated fasting plasma glucose and insulin levels, and decreased glucose tolerance as compared to singly-housed weight matched wt mice (Fig. 6d–f). Interestingly, wt(Asc^−/−) mice recapitulated the same increased rate of body mass gain and steatosis when compared to singly-housed wt controls, although they did not show significant alterations in glucose homeostasis (Fig. 6d–f). Nevertheless, antibiotic treatment (Ciprofloxacin and Metronidazole) abrogated all these effects, including altered rate of gain in body mass, glucose intolerance, and fasting plasma insulin levels in Asc^−/− mice as compared to wt mice (Fig. 6g–j). Alterations of these metabolic parameters were not caused by changes in feeding behavior between the antibiotic-treated and untreated groups (data not shown). These results suggest different levels of microbiota-mediated regulation of the various manifestations of the metabolic syndrome: i.e., some features (obesity, steatosis) are pronounced and transmissible by co-housing, while others (glycemic control) are affected by alterations in the microbiota but not readily transferable by co-housing. Additionally, we performed a 16S rRNA-based analysis of the fecal microbiota of Asc−/− and wt animals that were treated with or without Ciprofloxacin and Metronidazole (4 weeks) prior to switching to HFD for 4 additional weeks. We also observed an overrepresentation of Porphyromonadaceae in Asc^−/− as compared to wt mice (p=0.052) at week eight of HFD in the absence of antibiotics. Importantly, the analysis demonstrated that Prevotellaceae and Porphyromonadaceae, two family level taxa, were undetectable 8 weeks after antibiotic treatment (Supplementary fig. 12a–c; Table 2).

This ‘gut-liver axis’, driven by alterations in gut microbial ecology, may offer an explanation for a number of long-standing albeit poorly understood clinical associations. One example is the occurrence of primary sclerosing cholangitis (PSC) in patients with inflammatory bowel disease, particularly those with inflammation along the length of the colon. Celiac disease, another inflammatory disorder with increased intestinal permeability, is associated with a variety of liver disorders, ranging from asymptomatic transaminasemia, NAFLD, to primary biliary cirrhosis (PBC). In fully developed cirrhosis, complications associated with high mortality such as portal hypertension, variceal bleeding, spontaneous bacterial peritonitis, and encephalopathy are triggered by translocation of bacteria or bacterial components, providing another important example of the importance of the interplay between the microbiome, the immune response and liver pathology²⁸.

Recent reports suggest a complex role of inflammasome function in multiple manifestations of the metabolic syndrome. Activation of IL-1β, mainly through cleavage by the NLRP3 inflammasome, promotes insulin resistance ^29,30, atherosclesotic plaque formation ³¹, and β cell death ^32,33. Moreover, caspase-1 activation seems to direct adipocytes toward a more insulin-resistant phenotype³⁴. Conversely, Il18-deficient mice are prone to develop obesity, hyperphagia and insulin resistance³⁵. These discrepancies most likely reflect a hierarchical contribution of multiple inflammasome components in different metabolic processes, tissues, and mouse models. In agreement with previous studies, we found increased obesity and insulin resistance in Il18-deficient mice fed with a HFD (data not shown). However, and in contrast to two previous reports^30,34, we showed that Asc^−/− mice are prone to obesity induction, hepatosteatosis, as well as impaired glucose homeostasis when fed a HFD. We propose that alterations in intestinal microbiota communities associated with multiple inflammasome deficiencies could account for these discrepancies and it should be added to the list of major environmental/host factors affecting manifestations and progression of metabolic syndrome in susceptible populations.

1