Hepatocyte-specific IL11 cis-signaling drives lipotoxicity and underlies the transition from NAFLD to NASH

IL11 is important for fibrosis in non-alcoholic steatohepatitis (NASH) but its role beyond the stroma in liver disease is unclear. Here, we investigate the role of IL11 in hepatocyte lipotoxicity. Hepatocytes highly express IL11RA and secrete IL11 in response to lipid loading. Autocrine IL11 activity causes hepatocyte death through NOX4-derived ROS, activation of ERK, JNK and caspase-3, impaired mitochondrial function and reduced fatty acid oxidation. Paracrine IL11 activity stimulates hepatic stellate cells and causes fibrosis. In mouse models of NASH, hepatocyte-specific deletion of Il11ra1 protects against liver steatosis, fibrosis and inflammation while reducing serum glucose, cholesterol and triglyceride levels and limiting obesity. In mice deleted for Il11ra1, restoration of IL11 cis-signaling in hepatocytes reconstitutes steatosis and inflammation but not fibrosis. We found no evidence for the existence of IL6 or IL11 trans-signaling in hepatocytes or NASH. These data show that IL11 modulates hepatocyte metabolism and suggests a mechanism for NAFLD to NASH transition. IL11 contributes to the development of non-alcoholic steatohepatitis (NASH) through incompletely understood mechanisms. Here, the authors report that lipotoxicity-driven autocrine IL11 activity underlies hepatocyte metabolic dysfunction and death via a NOX4/ERK-mediated mechanism while paracrine IL11 activity stimulates hepatic stellate cells contributing to fibrosis and inflammation in the context of NASH.

I nterleukin 11 (IL11) is a fibrogenic factor 1-4 that is elevated in fibrotic precision-cut liver slices across species 5 . IL11 has recently been shown to have negative effects on hepatocyte function after toxic liver insult 6 and, directly or indirectly, contributes to nonalcoholic steatohepatitis (NASH) pathologies 7 . At the other end of the spectrum, a number of earlier publications suggest that IL11 is protective in mouse models of ischemic-, infective-, or toxin-induced liver damage [8][9][10][11][12][13] . However, it is now apparent that the recombinant human IL11 (rhIL11) reagent used in these earlier studies has unexpected effects in the mouse 6 and the question as to the true biological effect of IL11 in the liver, specifically in hepatocytes, remains open.
IL11 is a member of the interleukin 6 (IL6) cytokine family and binds to its cognate alpha receptor (IL11RA) and then to glycoprotein 130 (gp130) to signal in cis. IL6 itself has been linked to liver function and publications suggest an overall beneficial effect [14][15][16][17][18][19] . Aside from cis-signaling, IL6 can also bind to soluble IL6 receptor (sIL6R) and signal in trans. IL6 trans-signaling is considered maladaptive in the context of metabolic and autoimmune disease but, somewhat paradoxically, beneficial for liver regeneration 16 . It is possible that IL11, like IL6, also signals in trans but experiments to date have found no evidence for this in tumors or reproductive tissues 20,21 .
The factors underlying the transition from nonalcoholic fatty liver disease (NAFLD) to NASH are multifactorial but lipid loading of hepatocytes is of central importance 22 . Certain lipid species are toxic for hepatocytes and lipotoxicity leads to cytokine release causing hepatocyte death along with activation of hepatic stellate cells (HSCs) and immune cells 22,23 . Lipotoxicity, such as that due to palmitate 24 , is an early event in NASH and represents a linkage between diet, NAFLD, and NASH. While genetic or pharmacological inhibition of IL6 cis-signaling worsens steatosis phenotypes 17,18,25 , a role for IL11 in hepatic lipotoxicity has not been described.
In the current study, we used a range of in vitro and in vivo approaches to address key questions regarding IL11 in hepatocyte biology, NAFLD, and NASH: (1) Defining the role of IL11 signaling in human hepatocytes, (2) examining whether lipotoxicity is related to IL11 activity in hepatocytes, (3) establishing whether IL11 (or IL6) trans-signaling contributes to NASH, (4) dissecting the interrelationship between IL11 cis-signaling in hepatocytes and the development of steatosis, hepatocyte death, inflammation, and fibrosis. These studies demonstrate a detrimental effect of lipotoxicity-associated IL11 signaling in hepatocytes that appears to be apical pathology in the aetiology of NASH.

Results
High levels of IL11RA expression in primary human hepatocytes. We first assessed the expression levels of IL6R and IL11RA in healthy human or mouse liver by immunohistochemistry. In both human and mouse liver sections there was limited staining of IL6R but robust expression of IL11RA, which appeared mostly localized to hepatocytes ( Fig. 1a; Supplementary Fig. 2a). This is consistent with staining data from the human protein atlas using two additional antibodies (CAB032830 and HPA036652; https:// www.proteinatlas.org). Interestingly, as compared to control livers, IL11RA expression was increased in liver biopsies from patients with NASH and also in livers from mice with NASH on a Western Diet supplemented with fructose ( Supplementary Fig. 2b and c).
Flow cytometry studies confirmed that both IL11RA and gp130 are highly expressed in primary human hepatocytes whereas IL6R was expressed at much lower levels ( Fig. 1b; Supplementary  Fig. 3a). In contrast, IL6R was highly expressed on immune cells where IL11RA expression was low that is in keeping with previous data showing reciprocal expression of these two receptors on different cells (Fig. 1b) 1,19 . RNA-seq and Ribo-seq studies confirmed IL11RA and gp130 transcripts to be highly expressed and actively translated in hepatocytes, whereas IL6R transcripts were low and there was little eveidence of IL6R translation (Fig. 1c-e; Supplementary Fig. 3b and c). Immunofluorescence staining of primary human and mouse hepatocytes and some of the most commonly used hepatocyte-like cell lines (HepG2 and AML12) revealed that all of these cells consistently had high IL11RA but low IL6R expression ( Supplementary Fig. 3d). Overall, these data show strong co-expression of both IL11RA and gp130 in hepatocytes across species.
We next probed for the existence of IL11 trans-signaling in a physiological context, in the absence of preformed, synthetic/alien protein complexes. We stimulated cells with IL11 in the presence of either soluble gp130 (sgp130, to inhibit putative transsignaling) or soluble IL11RA (sIL11RA, to potentiate putative trans-signaling). IL11-induced caspase-3 activation, NOX4 upregulation, ERK and JNK signaling, and hepatocyte cell death were unaffected by sgp130 or sIL11RA (1 µg/ml) ( Fig. 1k and l; Supplementary Fig. 4b). Furthermore, IL11 dose-dependently caused hepatocyte cell death, which was unaffected by the addition of sgp130 or sIL11RA ( Supplementary Fig. 4c). Reciprocally, increasing doses of sgp130 or sIL11RA had no effect on ALT release from IL11-stimulated hepatocytes (Supplementary Fig. 4d). These data argue against the existence of IL11 trans-signaling in hepatocytes.
Reactive oxygen species and caspases are implicated, together or alone, in lipotoxic cell death 26 . To probe the mechanisms underlying IL11-regulated hepatocyte death we inhibited NOX4 with GKT-13781 or DPI and caspases with Z-VAD-FMK. NOX4 inhibitors reduced IL11-induced ERK and JNK activation and robustly protected hepatocytes from IL11-induced cell death (Supplementary Fig. 4e and f). Pan-caspase inhibition, while protective, was not as effective as NOX4 inhibition in preventing cell death and did not reduce either NOX4 induction or ERK activation (Supplementary Fig. 4g and h). This places NOX4 activity upstream of late-stage (24 h) ERK and caspase-3 activation in IL11-stimulated hepatocytes and suggests that apoptotic cell death is only one mode of cell death in this context.    Palmitate-loaded hepatocytes secreted large amounts of IL11 (28fold higher than control, P < 0.0001) (Fig. 2a), produced more IL6, CCL2, and CCL5 ( Fig. 2b-d), and exhibited cell death and ALT release ( Fig. 2e-g).
To test if IL11 secretion from lipid-laden hepatocytes was mechanistically related to lipotoxicity we incubated cells with neutralizing anti-IL11RA antibody (X209) 7 or sgp130. X209 reduced the secretion of all cytokines, including IL11 itself, whereas sgp130 had no effect ( Fig. 2a-g). This suggests the importance of autocrine loop of IL11 cis-signaling for hepatocyte lipotoxicity. Using hyperIL11 stimulation, which is not detected by IL11 enzyme-linked immunosorbent assay 1 , we then established the existence of feed-forward autocrine IL11 signaling in hepatocytes ( Supplementary Fig. 5a). The production of reactive oxygen species (ROS) from damaged mitochondria is important for lipotoxicity 23 and ROS from NOX4 are also pertinent 28 . Consistent with IL11 cis-driven effects on ROS in palmitateloaded hepatocytes, we found that X209, but not sgp130, partially restored total and reduced glutathione (GSH) levels, and this was accompanied by reduction in ROS ( Lipotoxicity is strongly associated with activation of JNK, which contributes to caspase-3 activation, lipoapoptosis, and dysfunctional mitochondria function. Accordingly, palmitateloaded hepatocytes exhibited an increase in JNK activation along with ERK phosphorylation, caspase-3 cleavage, and NOX4 upregulation (Fig. 2j). Treatment of lipotoxic cells with Z-VAD-FMK prevented caspase-3 cleavage, as expected, and reduced hepatocyte death (Supplementary Fig. 5c and d). Notably, inhibition of IL11 signaling with X209 reduced palmitateinduced NOX4 upregulation along with JNK, ERK, and caspase-3 activation (Fig. 2j). While STAT3 was activated by palmitate loading, this effect appears unrelated to IL11-driven lipoapoptosis as it was unaffected by X209 (Fig. 2j).
Lipid loading of hepatocytes was confirmed by Oil Red O staining ( Supplementary Fig. 5e) and quantitative analysis revealed a reduction in hepatocyte triglyceride (TG) levels with X209 treatment (Supplementary Fig. 5f). Inhibition of IL11 signaling also resulted in increased mitochondrial O 2 consumption rates, and maximal and spare respiratory capacity levels (Supplementary Fig. 5g and h). This effect was mediated by IL11 cis-signaling as sgp130 had no effect (Supplementary Fig. 5g and h). The fact that TGs were reduced suggested increased beta-oxidation may play a role 29 . We tested for this and found that inhibition of IL11 cis-signaling improved fatty acid oxidation in lipid-laden hepatocytes (Fig. 2k).
In the context of fatty liver disease, hepatocytes release factors to activate HSCs and activate/recruit immune cells, which are key events in the progression from compensated steatosis to NASH. As IL11 is secreted by lipotoxic hepatocytes (Fig. 2a), it could in theory act in paracrine on HSCs to drive HSC-to-myofibroblast transformation 7 . We cultured HSCs with conditioned media from either control or palmitate-treated hepatocytes and found that media from lipotoxic hepatocytes strongly induced ACTA2 and Collagen expression in HSCs ( Fig. 2l; Supplementary Fig. 5i and j). Addition of X209 to the conditioned media blocked ACTA2 and Collagen ( Fig. 2l; Supplementary Fig. 5i and j). These data demonstrate that lipotoxic hepatocytes release IL11 that acts in a paracrine fashion to activate IL11 signaling in HSCs.
No evidence for IL11 or IL6 trans-signaling in two NASH models. We then tested whether trans-signaling underlies NASH in vivo using two preclinical mouse NASH models: The Western Diet supplemented with fructose (WDF) model and the methionine-and choline-deficient high fat diet (HFMCD) model. The WDF model is associated with obesity, hyperlipidemia, high glucose levels, and insulin resistance and is seen as translatable to common forms of human NASH, as in diabetic patients. The HFMCD model stimulates rapid onset NASH, specifically driven by hepatocyte lipotoxicity, that is associated with weight loss in the absence of insulin resistance. Lipotoxicity is common to both models whereas obesity and insulin resistance are not.
Three weeks prior to starting either the WDF or HFMCD diet, mice were injected with an AAV8 virus encoding either albumin promoter-driven sgp130 (AAV8-Alb-sgp130), which contains the whole extracellular domain of mouse gp130 protein (amino acid 1-617), or albumin promoter alone (AAV8-Alb-Null) ( Fig. 3a; Supplementary Fig. 6a, 7a). AAV8-Alb-sgp130 administration induced high levels of sgp130 in the liver, which was also detectable in the peripheral circulation, suitable for both local and systemic inhibition of putative IL6 or IL11 trans-signaling ( Fig. 3b; Supplementary Figs. 6b, 7b and c).
After 16 weeks of WDF, IL11 levels were strongly upregulated in the liver and the periphery but IL6 expression was unaffected ( Fig. 3b-d). Mice on WDF became obese ( Supplementary Fig. 6c), an approximate twofold increase in liver mass and developed severe steatosis and fibrosis by gross morphology, histology, and quantitative analysis of liver triglycerides ( Fig. 3e-g). These phenotypes were unaffected by high levels of sgp130 expression ( Fig. 3b-g). Similarly, mice on WDF had elevated levels of ALT, AST, collagen, and peripheral cardiovascular risk factors (fasting Fig. 1 IL11RA is highly expressed in hepatocytes and IL11 cis-signaling is hepatotoxic. a Immunohistochemistry staining of IL11RA and IL6R in healthy human liver sections (scale bars, 20 µm, n = 1 independent experiment, due to limited amount of human liver section). b Flow cytometry forward scatter (FSC) plots of IL11RA, IL6R, and gp130 staining and fluorescence intensity plots of IL11RA and IL6R staining on hepatocytes and THP-1. c Abundance of IL11RA1 and IL6R reads in hepatocytes at baseline based on RNA-seq (left) and Ribo-seq (right) (transcripts per million, TPM) (n = 3). d, e Read coverage of d IL11RA1 and e IL6R transcripts based on RNA-seq (gray) and Ribo-seq (red) of primary human hepatocytes (n = 3). f Western blots showing ERK, JNK, and STAT3 activation status and g ALT secretion (n = 4) by hepatocytes following a dose range stimulation of either hyperIL11 or hyperIL6. h ALT levels in the supernatants of hepatocytes stimulated with hyperIL11 alone or in the presence of increasing amounts of soluble gp130 (sgp130) (n = 4). i, j Western blots of hepatocyte lysates showing i phosphorylated ERK and JNK and their respective total expression in response to hyperIL11 stimulation alone or with sgp130 and j phospho-STAT3 and total STAT3 in response to hyperIL6 stimulation with and without sgp130. k Representative FSC plots of propidium Iodide (PI) staining of IL11-stimulated hepatocytes in the presence of sgp130 or soluble IL11RA (sIL11RA). l Western blots showing phospho-ERK, phospho-JNK, cleaved caspase-3, and their respective total expression, NOX4, and GAPDH in hepatocytes in response to IL11 stimulation alone or in the presence of sgp130 or sIL11RA. i, j, l Representative data of n = 2 independent experiments. b-l Primary human hepatocytes; f-l 24 h stimulation; hyperIL11, hyperIL6, IL11 (20 ng/ml), sgp130, sIL11RA (1 µg/ml). c, g, h Data are shown as box-and-whisker with median (middle line), 25th-75th percentiles (box), and min-max values (whiskers). g, h One-way ANOVA with Dunnett's correction. Source data are provided as a Source data file.    blood glucose, serum triglycerides, and serum cholesterol), along with depleted levels of GSH but none of these parameters were affected by sgp130 (Fig. 3h-n). Livers from mice on WDF showed increased expression of pro-inflammatory and fibrosis genes and this signature was unaffected by sgp130-mediated inhibition of putative trans-signaling ( Fig. 3o; Supplementary Fig. 6d and e).
In a second set of experiments we induced NASH using the HFMCD diet ( Supplementary Fig. 7a). HFMCD diet increased IL11 levels in liver and serum, whereas IL6 levels were slightly lower in the liver albeit mildly increased in the periphery ( Supplementary Fig. 7b, d and e). Mice on HFMCD diet developed rapid and profound steatosis by gross morphology, histology, and molecular assays, which was unaltered by sgp130 expression (Supplementary Fig. 7f and g). Hepatocyte damage markers (ALT and AST) and collagen expression were elevated and GSH levels were depleted by HFMCD diet, irrespective of sgp130 expression ( Supplementary Fig. 7f, h-k). Similarly, the HFMCD diet was associated with dysregulated expression of inflammation and fibrosis genes and these molecular phenotypes were unaffected by sgp130 expression (Supplementary Fig. 7l and m). At the signaling level, both WDF and HFMCD diets stimulated ERK and JNK activation, consistent with elevated IL11 cissignaling ( Fig. 3p; Supplementary Fig. 7n). In contrast, phospho-STAT3 levels in the liver were not elevated by WDF (Fig. 3p) and were mildly increased in mice on the HFMCD diet (Supplementary Fig. 7n). In all instances, there was no effect of sgp130 on diet-induced signaling events. Overall, these data suggest that neither IL6 nor IL11 trans-signaling plays a role in NASH, which is consistent with other studies where IL6 family trans-signaling has not been detected 20,21,30,31 .
Hepatocyte-specific IL11 cis-signaling is required to initiate NASH. While we found no evidence to support IL11 trans-signaling in NASH models, our in vitro data showed evidence of pathological IL11 cis-signaling in lipotoxic hepatocytes. To test the effects of IL11 cis-signaling in heptocytes in vivo, we administered AAV8-Alb-Cre to Il11ra1 loxP/loxP mice to delete Il11ra1 specifically in hepatocytes (CKO mice). CKO mice were then fed either normal chow (NC), HFMCD diet or WDF (Figs. 4a, 5a). Liver IL11RA protein was greatly diminished in the CKOs following AAV8-Alb-Cre injection, showing the model to be effective and suggesting that hepatocytes are the largest hepatic reservoir of IL11RA (Figs. 4b, 5b). Both WT and CKO mice had similar levels of serum IL11 after 4 weeks of HFMCD and 16 weeks of WDF ( Supplementary Figs. 8a, 9a).
In addition to rapidly stimulating lipotoxicity-driven NASH, the HFMCD diet causes weight loss 32 . Surprisingly, weight loss in mice on the HFMCD diet was initially limited and later reversed in CKO mice ( Fig. 4c; Supplementary Fig. 8b). Mice on WDF gained weight and fat mass throughout the experimental period, as expected. However, and equally surprising, these obesity phenotypes were mitigated in CKO mice ( Fig. 5c and d;  Supplementary Fig. 9b). These data suggest that inhibition of IL11 signaling is permissive for weight homeostasis, with contextspecific anti-cachectic or anti-obesity effects, which requires further study.
By gross morphology, histology and quantitative triglyceride analysis, the CKO mice on either HFMCD or WDF diet were protected from steatosis (Figs. 4d and e, 5e, and f) and those on WDF had less liver mass (Fig. 5g). Liver damage markers were markedly reduced in CKO mice fed with either HFMCD diet (reduction: ALT, 99%; AST, 97%; P < 0.0001 for both) or WDF (reduction: ALT, 98%; AST, 98%; P < 0.0001 for both) and found to be comparable to NC control levels ( Figs. 4f and g, 5h and i). In both models, GSH levels were diminished in WT mice on both NASH diets but normalized in CKOs (Figs. 4h, 5j).
Liver fibrosis was greatly reduced in CKO mice on either NASH diet as compared to WT (reduction: HFMCD, 87%; WDF, 64%) (Figs. 4d, 4i, 5e, 5k). Upregulation of pro-inflammatory and fibrosis genes in mice on either the HFMCD or WDF diets was also diminished in the CKOs (Figs. 4j, 5l; Supplementary Figs. 8c and d, 9c and d). This suggests that transformation of HSCs to myofibroblasts and activation of immune cells are, in part, secondary to upstream, IL11-driven events in hepatocytes that are consistent with the paracrine effects we detected in vitro ( Fig. 2l; Supplementary Fig. 5i and j). At the signaling level, both HFMCD diet and WDF resulted in elevated ERK and JNK phosphorylation. This was prevented in CKO mice, consistent with inhibition of IL11 signaling in hepatocytes (Figs. 4k, 5m).
Mice on WDF are known to develop hyperglycemia, hypertriglyceridemia, and hypercholesterolemia, all of which were improved in the CKOs (Supplementary Fig. 9e-g). Furthermore, as compared to controls, CKOs on WDF had elevated serum levels of β-hydroxybutyrate, a peripheral marker of liver fatty acid oxidation and ketone production ( Supplementary  Fig. 9h). Overall these data suggest improvement of liver metabolism in CKOs, in keeping with our finding that inhibition of IL11 signaling promotes fatty acid oxidation in lipotoxic hepatocytes (Fig. 2k).
Reconstitution of IL11 cis-signaling in hepatocytes in IL11ra1 null mice restores steatohepatitis but not liver fibrosis. To complement our loss-of-function experiments using the CKO mice we employed in vivo gain-of-function experiments. To do so, we assessed whether restoring IL11 cis-or trans-signaling specifically in hepatocytes in mice with global Il11ra1 deletion (Il11ra1 −/− knockouts (KOs)) resulted in disease. KO mice were injected with AAV8 encoding either the full-length, membranebound Il11ra1 (mbIl11ra1; to reconstitute cis-signaling) or a secreted/soluble form of Il11ra1 (sIl11ra1, which constitutes the extracellular portion of Il11ra1; to enable trans-signaling) or a control construct, and the animals were then fed with NC, HFMCD diet, or WDF ( Fig. 6a; Supplementary Fig. 10a, 11a).
KO mice injected with AAV8-Alb-mbIl11ra1 re-expressed IL11RA1 on hepatocytes and KO mice injected with AAV8-Alb-sIl11ra1 had increased expression of sIL11RA1 in both the liver and the periphery (Fig. 6b; Supplementary Fig. 10b, 11b and c). As expected, WT mice receiving control AAV8 constructs (AAV8-Alb-Null) on NC had normal livers and developed steatosis, inflammation, liver damage, and liver fibrosis when on either HFMCD diet or WDF (Fig. 6c-j; Supplementary Fig. 10c  and d, 11d-k). KO mice injected with control virus and fed either HFMCD or WDF diets were protected from NASH phenotypes, although protection from NASH with germline global deletion of Fig. 3 Inhibition of IL6 family cytokine trans-signaling has no effect on NASH or metabolic phenotypes in mice on Western Diet supplemented with fructose. a Schematic of WDF feeding in mice with hepatocyte-specific expression of sgp130 for data shown in (b-p). Three weeks following AAV8-Alb-Null or AAV8-Alb-sgp130 virus injection, mice were fed WDF for 16 weeks. b Western blots showing hepatic levels of sgp130, IL11, IL6, and GAPDH as internal control (n = 4 mice/group). c Serum IL11 levels. d Serum IL6 levels. e Representative gross anatomy, H&E-stained (scale bars, 50 µm), and Masson's Trichrome (scale bars, 100 µm) images of livers. Representative dataset from n = 8 mice/group is shown for gross anatomy; representative dataset from n = 4 mice/group is shown for H&E-stained and Masson's Trichrome images. f Liver weight. g Hepatic triglycerides content. h Serum ALT levels. i Serum AST levels. j Hepatic collagen levels. k Fasting blood glucose levels. l Serum triglycerides levels. m Serum cholesterol levels. n Hepatic GSH content. o Hepatic pro-inflammatory and fibrotic genes expression heatmap (values are shown in Supplementary Fig. 6d and e). p Western blots of hepatic phospho-ERK, ERK, phospho-JNK, JNK, phospho-STAT3, and STAT3 (n = 4 mice/group). c, d, f-o n = 8 mice/group. c, d, f-n Data are shown as box-andwhisker with median (middle line), 25th-75th percentiles (box), and min-max values (whiskers); one-way ANOVA with Tukey's correction. Source data are provided as a Source data file.
Restoration of IL11 cis-signaling in KO mice using mbIl11ra1 recapitulated hepatic steatosis and inflammation that was evident from gross morphology to molecular patterns of gene expression and signaling (Fig. 6c-h, j and k; Supplementary Fig. 10c, 11d-h,  11j, 11l). However, hepatic collagen content and fibrotic gene expression were not restored (Fig. 6c, i and j; Supplementary   Fig. 10d, 11d, 11i, 11k) presumably because IL11 cis-signaling in HSCs, important for HSC-to-myofibroblast transformation 7 , was unaffected by the albumin-driven Il11ra1 expression (i.e., HSCs remain deleted for Il11ra1 in this model).
In contrast, expression of the sIL11RA in hepatocytes of KOs, which would theoretically activate trans-signaling as IL11 levels are already elevated, had no effect despite very high IL11RA levels ( Fig. 6b; Supplementary Fig. 10b, 11b and c) and mice remained  Fig. 4 Hepatocyte-specific inhibition of IL11 cis-signaling protects mice against HFMCD diet-induced NASH. a Schematic of HFMCD feeding regimen for AAV8-Alb-Cre injected Il11ra1 loxP/loxP (conditional knockout; CKO) mice for experiments shown in (b-k). Il11ra1 loxP/loxP mice were intravenously injected with either AAV8-Alb-Null or AAV8-Alb-Cre to delete Il11ra1 specifically in hepatocytes 3 weeks prior to the start of HFMCD diet. b Western blots of hepatic IL11RA and GAPDH (n = 3 mice/group). c Body weight (shown as a percentage (%) of initial body weight). d Representative gross anatomy, H&Estained (scale bars, 50 µm), and Masson's Trichrome (scale bars, 100 µm) images of livers. Representative dataset from n = 5 mice/group is shown for gross anatomy; representative dataset from n = 4 mice/group is shown for H&E-stained and Masson's Trichrome images. e Hepatic triglycerides content. f Serum ALT levels. g Serum AST levels. h Hepatic GSH content. i Hepatic collagen levels. j Heatmap showing hepatic mRNA expression of proinflammatory markers (Tnfα, Ccl2, Ccl5) and fibrotic markers (Col1a1, Col1a2, Col3a1, Acta2). Values are shown in Supplementary Fig. 8c and d. k Western blots showing hepatic ERK and JNK activation status (n = 3 mice/group). c, e-j NCD (n = 5 mice/group), HFMCD (n = 6 mice/group). c Data are shown as mean ± SD, two-way ANOVA with Tukey's correction, statistical significance (P values) are shown for comparison between WT HFMCD and CKO HFMCD; e-i data are shown as box-and-whisker with median (middle line), 25th-75th percentiles (box), and min-max values (whiskers); two-way ANOVA with Tukey's correction. Source data are provided as a Source data file. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20303-z protected from NASH (Fig. 6c-j; Supplementary Fig. 10c, 10d, 11d-k). Signaling changes were consistent in that mIL11RA expression restored pathological ERK and JNK activation in KOs on either diet, whereas sIL11RA1 did not ( Fig. 6k; Supplementary  Fig. 11l). Similarly, in the WDF model, restoration of hepatocytespecific IL11 cis-signaling (mIL11RA) in KO mice caused hyperglycemia, hypertriglyceridemia, and hypercholesterolemia but expression of sIl11ra1 did not (Fig. 6l-n).

Discussion
Metabolic liver disease commonly occurs in the context of obesity and type 2 diabetes and manifests initially as NAFLD that can progress to NASH 22,33 . A key underlying pathology in the progression to NASH is "substrate overload", whereby a nutritional abundance overruns the hepatocyte's ability to process fat, causing lipotoxicity. Cytokines are key NASH factors secreted from lipotoxic hepatocytes 22 and here we establish IL11 as an important component of the lipotoxic milieu and a driver of the NAFLD-to-NASH transition. A large body of evidence supports the idea that IL6 cis-and trans-signaling in the liver is beneficial 16,17,19,25 . However, at the same time, a pathogenic role for IL6 trans-signaling in hepatic steatosis has also been proposed 30,34 . Using man-made, artificial protein constructs we found that IL11 trans-signaling is toxic in hepatocytes, whereas hyperIL6 appears protective. However, we found no evidence to support a meaningful role for IL6 or IL11 trans-signaling in a biologically relevant context either in vitro or in vivo, using both gain-and loss-of-function approaches. Notably, sgp130, a therapeutic agent that inhibits IL6 trans-signaling, had no effect on lipotoxicity, NAFLD or NASH. Thus, we suggest that IL6 family member trans-signaling has no role in hepatocytes or NASH, which is in agreement with studies outside the liver 20,21 .
Our data show a central importance of IL11 cis-signaling in hepatocytes for multiple NASH pathologies. This effect was established using both hepatocyte-specific loss-of-function on a wild-type genetic background and also hepatocyte-specific gainof-function on an Il11ra1 null background. This overturns the suggestion in the literature that IL11 is protective for hepatocytes based on the use of recombinant human IL11 in murine models of liver disease 8,10-12 . While restoration of IL11 cis-signaling in hepatocytes causes steatohepatitis in mice globally deleted for Il11ra1, lipotoxicity-associated fibrosis was still prevented. In contrast, hepatocyte-specific Il11ra1 deletion protected mice from both steatohepatitis and fibrosis. This places hepatocyte dysfunction upstream of HSC activation, which is consistent with the observed paracrine effects of IL11 from lipotoxic hepatocytes.
We repeatedly documented metabolically advantageous effects associated with the inhibition of IL11 signaling. In vitro, inhibition of IL11 improved mitochondrial function and increased beta-oxidation along with a reduction of intracellular triglycerides and ROS production. The effects on ROS are likely complex as IL11 induces NOX4 but also impacts mitochondria function and perhaps also the endoplasmic reticulum, directly or indirectly. Interestingly, inhibition of NOX4 was more effective than caspase inhibition in reducing IL11-stimulated cell death. Thus IL11induced NOX4/ROS lies upstream of late-stage ERK and caspase activation in hepatocytes and apoptosis is not the only form of cell death in this context. In vivo, deletion of Il11ra1 in hepatocytes limited WDF-induced fat accumulation and body weight gain while reducing serum glucose, triglyceride, and cholesterol levels. This was associated with lower liver fat, lesser hepatic oxidative stress, and increased serum levels of beta-hydroxybutyrate, thought metabolically beneficial in itself 35 .
Our studies have limitations and pose questions. The published literature suggests IL6R is highly expressed in hepatocytes 16 and it was surprising that primary human hepatocytes express very little IL6R. The differential activation of ERK/JNK by IL11 as compared to STAT3 by IL6 at 24 h in hepatocytes was also notable but the underlying mechanisms are unknown. While we show consistent effects of IL11 inhibition on pro-inflammatory factors we did not specifically address effects on immune cells themselves. Metabolic effects appear closely related to IL11-mediated NOX4-derived ROS but ROS from other sources also likely contributes. Temporal relationships between ERK and NOX4 activation require further evaluation. We found a beneficial effect of hepatocyte-specific IL11 inhibition on reducing fat deposition and body weight gain in mice on an obesogenic NASH diet, which while notable was not studied in depth. We surmise that IL11 hepatocyte biology is a nascent field and that these various matters require further study.
In conclusion, we propose a model for lipotoxicity-driven NAFLD-to-NASH transitions whereby lipid-laden hepatocytes secrete IL11 leading to autocrine hepatocyte metabolic dysfunction and cell death along with paracrine activation of neighbouring HSCs and other cells (Fig. 7). We rule out IL6 or IL11 trans-signaling as relevant for hepatocyte biology or liver pathology in lipotoxicity. We suggest that inhibiting IL11 signaling in hepatocytes targets an initiating nexus for dietinduced steatohepatitis that impacts subsequent liver fibrosis and inflammation. Hence, therapeutic inhibition of IL11-induced lipotoxicity may be beneficial in metabolic liver diseases, such as NASH.

Methods
Ethics statements. All experimental protocols involving human subjects (commercial primary human cell lines and human liver sections) have been performed in accordance with the ICH Guidelines for Good Clinical Practice. As written in their respective datasheets, ethical approvals have been obtained by the relevant parties and all participants gave written informed consent: commercial human liver sections (by Abcam); liver sections from healthy control and NASH patients (by Fibrofind); primary human hepatocytes and hepatic stellate cells (by ScienCell); HepG2 and THP-1 (by ATCC).
Animal studies were carried out in compliance with the recommendations in the Guidelines on the Care and Use of Animals for Scientific Purposes of the National Advisory Committee for Laboratory Animal Research (NACLAR). All experimental procedures were approved (SHS/2014/0925 and SHS/2019/1482) and conducted in accordance with the SingHealth Institutional Animal Care and Use Committee.
AAV8 vectors. All Adeno-associated virus serotype 8 (AAV8) vectors used in this study were synthesized by Vector Biolabs. AAV8 vector carrying a mouse membrane-bound Il11ra1 cDNA (NCBI accession number: BC069984), a mouse soluble Il11ra1 cDNA, and a mouse soluble gp130 cDNA driven by Albumin (Alb) promoter is referred to as AAV8-Alb-mbIl11ra1, AAV8-Alb-sIl11ra1, and AAV8-Alb-sgp130, respectively. AAV8-Alb-sgp130 and AAV8-Alb-sIl11ra1 were constructed by removing the transmembrane and cytoplasmic regions of mouse gp130 sequence (NCBI accession number: BC058679) and mouse Il11ra1 sequence, respectively. AAV8-Null vector was used as vector control. To specifically delete Fig. 5 Hepatocyte-specific inhibition of IL11 cis-signaling protects mice against WDF-induced obesity and NASH. a Schematic of WDF-fed control and CKO mice for data shown in (b-m). Three weeks following AAV8-Alb-Null or AAV8-Alb-Cre virus injection, CKO mice were fed WDF for 16 weeks. b Western blots showing hepatic levels of IL11RA and GAPDH (n = 3 mice/group). c Body weight (shown as a percentage (%) of initial body weight). d Fat mass. e Representative gross anatomy, H&E-stained (scale bars, 50 µm), and Masson's Trichrome (scale bars, 100 µm) images of livers. Representative dataset from n = 5/group is shown for gross anatomy; representative dataset from n = 4 mice/group is shown for H&E-stained and Masson's Trichrome images. f Hepatic triglycerides content. g Liver weight. h Serum ALT levels. i Serum AST levels. j Hepatic GSH content. k Hepatic collagen levels. l Hepatic pro-inflammatory and fibrotic genes expression on heatmap (values are shown in Supplementary Fig. 9c and d). m Western blots showing activation status of hepatic ERK and JNK (n = 3 mice/group). c, d, f-l n = 5 mice/group. c, d Data are shown as mean ± SD, two-way ANOVA with Tukey's correction, statistical significance (P values) are shown for comparison between WT WDF and CKO WDF; f-k data are shown as box-and-whisker with median (middle line), 25th-75th percentiles (box), and min-max values (whiskers); two-way ANOVA with Tukey's correction. Source data are provided as a Source data file.
Antibodies. ACTA2 (ab7817, Abcam, 1:1000 for western blot and 1:500 for operetta assay), Albumin (ab207327, Abcam, 1:100 for IF and flow cytometry),   6 Hepatocyte-specific IL11 cis-signaling but not IL11 trans-signaling drives steatohepatitis in mice on WDF. a Schematic showing WDF feeding regimen of Il11ra1 +/+ (WT) and Il11ra1 −/− (KO) mice for experiments shown in (b-n). AAV8-Alb-Null, AAV8-Alb-mbIl11ra1 (full-length membrane-bound Il11ra1), and AAV8-Alb-sIl11ra1 (soluble form of Il11ra1)-injected KO mice were given 16 weeks of WDF feeding, three weeks following virus administration. b Western blots showing hepatic levels of IL11RA and GAPDH (n = 2 mice/group). c Representative gross anatomy, H&E-stained (scale bars, 50 µm) and Masson's Trichrome (scale bars, 100 µm) images of livers. Representative dataset from n = 6 mice/group is shown for gross anatomy; representative dataset from n = 4 mice/group is shown for H&E-stained and Masson's Trichrome images. d Liver weight. e Hepatic triglycerides content. f Serum ALT levels. g Serum AST levels. h Hepatic GSH content. i Hepatic collagen content. j Hepatic pro-inflammatory and fibrotic genes expression heatmap (values are shown in Supplementary Fig. 10c and d). k Western blots showing activation status of hepatic ERK and JNK (n = 2 mice/group). l Fasting blood glucose levels. m Serum triglycerides levels. n Serum cholesterol levels. d-j, l-n n = 6 mice/group. d-i, l-n Data are shown as box-and-whisker with median (middle line), 25th-75th percentiles (box), and min-max values (whiskers); one-way ANOVA with Tukey's correction. Source data are provided as a Source data file.  Fig. 7 Proposed mechanism for IL11 in lipotoxicity-driven NASH transition. Excessive lipid accumulation in hepatocytes stimulates IL11 protein secretion and autocrine IL11 activity, which upregulates NOX4 and increases reactive oxygen species production. Subsequently, hepatocyte mitochondrial oxidative capacity and fatty acid metabolism impaired and steatosis established. ERK, JNK, and caspase-3 become activated and this leads to lipoapoptosis, along with other forms of cell death. IL11 also acts in paracrine to drive hepatic stellate cell-to-myofibroblast transformation and fibrosis. Cytokines and chemokines released from lipotoxic hepatocytes and HSCs activate and recruit immune cells causing inflammation.
Cell culture. All the experiments performed with primary human hepatocytes, primary adult mouse hepatocytes, and primary adult human hepatic stellate cells were carried out at low cell passage (≤P3). Palmitate (saturated fatty acid) treatment in vitro. Palmitate (0.5 mM) conjugated in fatty acids free BSA in the ratio of 6:1 was used to treat cells as described in figure legends; 0.5% BSA solution was used as control.
Flow cytometry. For surface IL11RA, IL6R, and gp130 analysis, primary human hepatocytes and THP-1 cells were stained with IL11RA, IL6R, or gp130 antibody, and the corresponding Alexa Fluor 488 secondary antibody. Omission of primary antibody staining was used as negative control. Cell death analysis was performed by staining primary human hepatocytes with Dead Cell Apoptosis Kit with Annexin V FITC and PI (V13242, Thermo Fisher). PI +ve cells were then quantified with the flow cytometer (Fortessa, BD Biosciences) and analyzed with FlowJo version X software (TreeStar): the preliminary FSC/SSC gates of the starting cell population was 10,000 events. Debris (SSC-A vs FSC-A) and doublets (FSC-H vs FSC-A) were excluded. Boundaries between "positive" and "negative" staining were set at 10 3 for PI staining. A figure exemplifying the gating strategy is provided in the Supplementary Fig. 1. Reactive oxygen species (ROS) detection. Primary human hepatocytes were seeded on 8-well chamber slides (1.5 × 10 4 cells/well). For this experiment, cells were not serum-starved prior to palmitate treatment. Twenty-four hours following palmitate stimulation, cells were washed, incubated with 25 µM of DCFDA solution (ab113851, Abcam) for 45 min at 37°C in the dark, and rinsed with the dilution buffer according to the manufacturer's protocol. Live cells with positive DCF staining were imaged with a filter set appropriate for fluorescein (FITC) using a fluorescence microscope (Leica).
Seahorse assay. Primary human hepatocytes were seeded into the Seahorse XF Cell Culture Microplate (1 × 10 4 cells/well) and serum-starved overnight prior to stimulations. Seahorse measurements were performed on Seahorse XFe96 Extracellular Flux analyzer (Agilent). XF Cell Mito Stress Test kit (103015-100, Agilent) was used to measure the mitochondrial oxygen consumption rate as per the manufacturer's protocol. Briefly, stimulation media were removed and replaced with 180 μl of Mitostress assay medium at 37°C. Oligomycin (1 μM; ATP synthase inhibitor) was injected following basal OCR measurements followed by injection of FCCP (1 μM; an uncoupling agent that collapses the proton gradient and disrupts the mitochondrial membrane potential), and finally by injection of a mixture of rotenone (1 μM; a complex I inhibitor) and antimycin A (1 μM; a complex III inhibitor). The percentage of fatty acid oxidation analysis was performed by using Seahorse XF Mito Fuel Flex Test kit (103260-100). Acute injections of CPT1 alpha inhibitor Etomoxir (4 µM) was used to inhibit mitochondrial FAO, whereas BPTES (3 µM) and UK5099 (2 µM) were used to inhibit mitochondrial glutamine and glucose oxidation, respectively, to inhibit 100% mitochondrial fuel oxidation. Seahorse Wave Desktop software was used for report generation and data analysis.
RNA-sequencing (RNA-seq) and ribosome profiling (Ribo-seq) Generation of RNA-seq libraries. Total RNA was extracted from human hepatocytes using RNeasy columns (Qiagen). RNA was quantified using a Qubit RNA High-Sensitivity Assay kit (Life Technologies) and its quality was assessed on the basis of their RNA integrity number using the LabChip GX RNA Assay Reagent Kit (PerkinElmer). TruSeq Stranded mRNA Library Preparation kit (Illumina) was used to measure transcript abundance following standard instructions from the manufacturer.
Generation of Ribo-seq libraries. Hepatocytes were grown to 90% confluence in a 10 cm culture dish and lysed in 1 ml cold lysis buffer (formulation as in TruSeq® Ribo Profile Mammalian Kit, RPHMR12126, Illumina) supplemented with 0.1 mg/ ml cycloheximide. Homogenized and cleared lysates were then footprinted with Truseq Nuclease (Illumina) according to the manufacturer's instructions. Ribosomes were purified using Illustra Sephacryl S400 columns (GE Healthcare), and the protected RNA fragments were extracted with a standard phenol:chloroform: isoamylalcohol technique. Following ribosomal RNA removal (Mammalian Ribo-Zero Magnetic Gold, Illumina), sequencing libraries were then prepared out of the footprinted RNA by using TruSeq® Ribo Profile Mammalian Kit according to the manufacturer's protocol.
The final RNA-seq and ribosome profiling libraries were quantified using KAPA library quantification kits (KAPA Biosystems) on a StepOnePlus Real-Time PCR system (Applied Biosystems) according to the manufacturer's protocol. The quality and average fragment size of the final libraries were determined using a LabChip GX DNA High Sensitivity Reagent Kit (PerkinElmer). Libraries with unique indexes were pooled and sequenced on a NextSeq 500 benchtop sequencer (Illumina) using NextSeq 500 High Output v2 kit and single-end 75-bp sequencing chemistry.
Data processing and analyses for RNA-sequencing and ribosome profiling. Raw sequencing data were demultiplexed with bcl2fastq V2.19.0.316 and the adaptors were trimmed using Trimmomatic 36 V0.36, retaining reads longer than 20 nt post-clipping. Ribo-seq reads were aligned using bowtie 37 to known mtRNA, rRNA, and tRNA sequences (RNACentral 38 , release 5.0) and only unaligned reads were retained as Ribosome protected fragments (RPFs). Alignment to the human genome (hg38) was carried out using STAR 39 . Gene expression was quantified on the CDS (coding sequence) regions for Ribo-seq and exonic regions for RNA-seq using uniquely mapped reads (Ensembl database release GRCh38 v86) with feature counts 40 . TPM was calculated and visualized using boxplot to compare baseline expression of IL11RA (ENSG00000137070), IL6R (ENSG00000160712), and gp130 (ENSG00000134352). Read coverage using Ribo-seq and RNA-seq reads for IL11RA, IL6R, and gp130 was visualized using Gviz R package 41 with strandspecific alignment files.