Amphiregulin from regulatory T cells promotes liver fibrosis and insulin resistance in non-alcoholic steatohepatitis

SUMMARY

Production of amphiregulin (Areg) by regulatory T (Treg) cells promotes repair after acute tissue injury. Here, we examined the function of Treg cells in non-alcoholic steatohepatitis (NASH), a setting of chronic liver injury. Areg producing Treg cells were enriched in the livers of mice and humans with NASH. Deletion of Areg in Treg cells, but not in myeloid cells, reduced NASH-induced liver fibrosis. Chronic liver damage induced transcriptional changes associated with Treg cell activation. Mechanistically, Treg cell–derived Areg activated pro-fibrotic transcriptional programs in hepatic stellate cells via epidermal growth factor receptor (EGFR) signaling. Deletion of Areg in Treg cells protected mice from NASH-dependent glucose intolerance, which also was dependent on EGFR signaling on hepatic stellate cells. Areg from Treg cells promoted hepatocyte gluconeogenesis through hepatocyte detection of hepatic stellate cell–derived interleukin 6. Our findings reveal a maladaptive role for Treg cell–mediated tissue repair functions in chronic liver disease and link liver damage to NASH-dependent glucose intolerance.

Graphical Abstract

eTOC

Non-alcoholic fatty liver disease and its progressive form, non-alcoholic steatohepatitis (NASH), are a prevalent cause of chronic liver disease. Savage et al. demonstrate that regulatory T (Treg) cells are enriched in mouse and human NASH and find that production of the EGFR ligand amphiregulin by Treg cells promotes NASH–induced liver fibrosis and glucose intolerance through direct signaling to hepatic stellate cells.

INTRODUCTION

Fibrosis is an end stage disease and a leading cause of morbidity and mortality worldwide, with a paucity of effective therapeutic options. Identifying principles underlying the development of tissue fibrosis may therefore offer new therapeutic targets. Non-alcoholic fatty liver disease (NAFLD) and its progressive form, non-alcoholic steatohepatitis (NASH), currently affect up to 30% of the world population and will soon become the most prevalent cause of chronic liver disease^1,2. The progression to NASH involves hepatocyte damage and liver inflammation, with the development of liver fibrosis being the key determinant of patient outcomes^3–9. As yet, however, there are no effective pharmacological therapies for NASH or NASH-mediated fibrosis¹⁰, stemming at least in part from the limited understanding of cellular and molecular mechanisms driving disease progression.

NAFLD and NASH are closely associated with insulin resistance. Over half of patients with type 2 diabetes have fatty liver disease, and more than 40% of patients with NASH have type 2 diabetes, increasing proportionally with greater severity of NASH and fibrosis^2,11–13. Insulin resistance contributes to the development of NAFLD and its progression to NASH, but the role for NASH in promoting insulin resistance, specifically the molecular and cellular interactions underlying NASH-induced metabolic decline, is poorly understood and thus therapeutically unexplored¹⁴. Hepatic stellate cells – liver-resident fibroblasts – are central to liver fibrosis¹⁵ and interplay with hepatocytes via cell–cell contact and soluble mediators that support hepatocyte development and homeostasis^16,17. However, the interplay between hepatic stellate cells and hepatocytes in settings of metabolic disease is relatively unexplored. In metabolic disease, pro-inflammatory cytokines such as interleukin (IL)-6 from adipose tissue macrophages circulate systemically and promote hepatocyte insulin resistance¹⁸. However, whether cells in closer proximity (e.g. hepatic stellate cells) produce mediators that promote hepatocyte insulin resistance is unclear.

Regulatory T (Treg) support tissue homeostasis by limiting inflammation and promoting repair¹⁹. Treg cell activity is considered a key component of the adaptive response to tissue damage, and necessary to support effective tissue repair and restoration¹⁹. Beneficial Treg cell functions are apparent in the uncontrolled inflammation and fatal tissue damage observed in Foxp3^−/− mice^20,21, a phenotype that can be rescued by transfer of T cells expressing Foxp3^20,21. Furthermore, Treg cell depletion results in increased insulin resistance in the setting of obesity²² and worsened liver injury and fibrosis in chronic toxin injury and following bile duct ligation^23,24. Treg cell production of amphiregulin (Areg), a low-affinity ligand for the Epidermal Growth Factor Receptor (EGFR)-family^25–28, is critical to repair following acute muscle, brain, and lung injury^25–28. In these settings, Treg cell–derived Areg is notably independent of Treg cell–mediated immunosuppression, and acts on tissue-resident mesenchymal cells and stem cells – not immune cells^25,26,28. Together, these studies suggest that Treg cells protect from epithelial cell injury by both interacting with immune cells to suppress inflammation and by producing epithelial growth factors that signal to non-immune cells. However, the impact of Treg cells in settings of chronic tissue damage may be distinct.

Here, we examined the effects of Treg cell–derived Areg in chronic liver disease. We found that Areg from Treg cells promotes liver fibrosis and NASH-dependent glucose intolerance. These findings reveal potential therapeutic avenues for treating fibrotic disease by targeting interactions between Treg cells and tissue-resident fibroblasts.

RESULTS

Areg production by Treg cells promotes liver fibrosis

To study the tissue reparative role of Treg cells in the liver, we first analyzed liver T cell subsets for expression of Areg. In healthy wildtype mice, liver Treg cells produced substantial amounts of Areg protein relative to spleen Treg cells and to other liver T cell subpopulations (Fig. 1A), suggesting that Treg cell–derived Areg may contribute to liver homeostasis. Consistent with previous work²³, following chronic carbon tetrachloride (CCl₄)–induced injury – a toxin-based model that leads to robust liver fibrosis and in which germline deletion of Areg is protective²⁹ – increased frequencies of liver Treg cells were detected (Fig. S1A). In addition, the overall Treg cell number and the number of Areg-producing liver Treg cells were increased (Fig. S1B–S1C). To investigate the role of Areg-producing Treg cells in a more clinically-relevant setting, we utilized two diet-induced mouse models of NASH. Mice fed the fructose, palmitate and cholesterol-rich (FPC) NASH diet for 16 weeks, which induces moderate liver fibrosis as well as insulin resistance³⁰, had increased frequencies and overall number of liver Treg cells as compared to animals fed normal chow (Fig. 1B–1C). The number of Areg-producing Treg cells was also increased in NASH livers (Fig. 1D), suggesting that Treg cell–derived Areg may play a role in protection from and/or the development of NASH. Similar results were seen in the choline-deficient, L-amino acid-defined high fat diet (CDAA-HFD) model of NASH, which induces substantial fibrosis (Fig. S1D–S1G). However, liver Treg cells were not enriched in the high fat diet (HFD) model that induces robust steatosis but fails to progress to NASH (Fig. S1D–S1G). Additionally, liver Treg cells were not enriched at an earlier timepoint (8 weeks) on the FPC diet when liver steatosis is evident but NASH has yet to develop³⁰ (Fig. S1D–S1G). Together, these findings are consistent with hepatocyte injury contributing to an enrichment of Treg cells within inflamed and damaged, but not solely steatotic, livers. To assess the clinical relevance of our findings, we examined Treg cells in human liver disease, first utilizing samples from healthy and NASH human liver tissue (Table S1). Treg cells in human NASH liver samples were increased compared to normal liver (Fig. 1E, S1H). Further, Treg cells were enriched relative to total CD3⁺ T cells in human NASH (Fig. 1F, S1I), suggesting a disproportionate increase in Treg cells in human NASH livers. We expanded the analyses to include liver specimens from different etiologies of human liver fibrosis. Treg cells were enriched in human alcoholic steatohepatitis (ASH), autoimmune hepatitis, and drug/toxin-induced liver fibrosis (Fig. S1J–S1L), consistent with our murine findings in different damage models (Fig. S1A–S1C).

Fig. 1. Treg cell–derived amphiregulin promotes liver fibrosis.

(A) Flow cytometric analysis of Areg protein expression by the indicated mouse T cell subsets. Tconv.: TCRβ⁺CD4⁺Foxp3⁻. Data are representative of 2 independent experiments, each with n = 4 mice.

(B–D) Liver Treg cell frequency and number, and number of liver Areg-producing Treg cells, as assessed by flow cytometric analysis from mice fed normal chow or FPC diet. n ≥ 3 mice per group. Data are representative of 2 independent experiments.

(E–F) Quantification of Treg cell staining in human healthy or NASH liver specimens. n = 9–10 per group.

(G–J) Sirius Red staining of liver sections from mice lacking Treg cell–derived Areg (Areg^Fl/FlFoxp3^YFP-cre) or littermate controls (Areg^Fl/Fl) fed (G–H) FPC diet for 16 weeks or (I–J) the CDAA-HFD. n ≥ 6 mice per group. Data are representative of 2–4 independent experiments.

(K) Expression of indicated genes in liver tissue as assessed by RT-qPCR of Areg^Fl/FlFoxp3^YFP-cre or littermate Areg^Fl/Fl mice fed the FPC diet for 16 weeks. Data representative of 4 independent experiments.

(B–F, H, J–K) two-sided Student’s t-test: *p < 0.05, **p < 0.01, ***p < 0.001. Data are expressed as mean ± s.e.m.

See also Figure S1–S2.

We next sought to functionally assess the role of Treg cell–derived Areg in liver disease and fibrosis. We aimed to clarify whether Areg from Treg cells offers a protective function – as has been suggested by prior studies employing total depletion of Treg cells in chronic CCl₄ and bile duct ligation murine models^23,24. Following chronic CCl₄ injection, Areg^+/− and Areg^−/− mice had similar levels of liver fibrosis, as measured by Sirius Red staining, whereas Areg^+/+ animals were protected from liver fibrosis compared to Areg^−/− mice (Fig. S2A and data not shown). Further, Areg^+/− mice had less liver Areg protein than Areg^+/+ mice, as low as in Areg^−/− mice (Fig. S2B–S2C), suggesting haploinsufficiency in Areg^+/− mice. Thus, we used lineage-specific Cre drivers to ablate Areg production from all hematopoietic cells (Vav1-Cre), myeloid cells (Lyz2^wt/cre), all T cells (CD4-Cre), or Treg cells (Foxp3^YFP-cre)²⁵. In the chronic CCl₄ model, we observed reduced liver fibrosis, as measured by Sirius Red staining, in the absence of Areg derived from all hematopoietic cells (Areg^Fl/FlVav1-Cre), but not from myeloid cells (Areg^Fl/FlLyz2^wt/cre), relative to littermate controls (Areg^Fl/Fl) (Fig. S2D–S2E). Deletion of Areg in all CD4⁺ and CD8⁺ T cells (Areg^Fl/FlCD4-Cre), and specifically in Treg cells (Areg^Fl/FlFoxp3^YFP-cre), offered protection from CCl₄-induced fibrosis (Fig. S2F–S2G). Deletion of Areg from all hematopoietic cells and all CD4⁺ and CD8⁺ T cells, but not myeloid cells, inhibited the development of NASH-mediated fibrosis (Fig. S2H–S2J). Furthermore, less Sirius Red staining was observed in liver sections from FPC diet–fed Areg^Fl/FlFoxp3^YFP-cre animals compared to littermate controls (Fig. 1G–1H). Mice lacking T cell– or Treg cell–derived Areg also had less Sirius Red staining in liver sections than littermate controls in the CDAA-HFD model of NASH (Fig. 1I–1J, S2K). The protection from fibrosis observed in the absence of Treg cell–derived Areg was not attributable to alterations in body weight, hepatocyte cell damage or death, liver fat accumulation, or liver inflammation (Fig. S2L–S2R). Loss of Treg cell–derived Areg did not affect overall Areg protein levels in the livers of mice fed the FPC diet (Fig. S2S–S2T), suggesting a localized effect. Finally, the expression of genes reflecting hepatic stellate cell activation was reduced in the livers of FPC-fed Areg^Fl/FlFoxp3^YFP-cre mice compared to littermate controls (Fig. 1K). Taken together, these data demonstrate a specific expansion of liver Areg-producing Treg cells in the setting of chronic liver damage and identify a maladaptive role for Treg cell–derived Areg – promoting liver fibrosis following toxin-induced chronic liver injury and diet-induced NASH.

Treg cells in liver fibrosis exhibit a transcriptionally distinct activation profile

Treg cells in tissues have distinct transcriptional signatures³¹. To define transcriptional features of Treg cells in the context of liver fibrosis, we examined Treg cells from CCl₄-treated mice by RNA-Seq (Fig. S2G). Liver Treg cells clustered distinctly from spleen Treg cells and from liver and spleen conventional CD4⁺Foxp3⁻ T cells (Tconv), with Treg cells from fibrotic livers clustering distinctly from those at steady state (Fig. S3A–S3B). Differential gene expression analysis revealed transcriptional alterations occurring specifically within liver Treg cells in the context of chronic CCl₄ injury, as compared to liver Tconv cells or splenic counterparts (Fig. S3C). Further, fibrotic liver Treg cells upregulated pro-inflammatory cytokine receptors and tissue-homing chemokine receptors (Fig. S3D). Treg cells isolated from fibrotic versus normal livers displayed increased metabolic pathway activation, cell cycle progression, and increased cytokine receptor signaling (Fig. S3E). To study human Treg cells in liver fibrosis, we examined previously published single-cell RNA sequencing data from healthy and cirrhotic human livers³², analyzing a cluster of Treg cells identified based on the expression of FOXP3 (Fig. S3F). Notably, human Treg cells isolated from cirrhotic livers upregulated genes including TIGIT, CXCR4, CXCR3, IL10, CTLA4 and CCR5, and were enriched for transcriptional pathways reflecting increased activation and proliferation (Fig. S3G–S3H), consistent with observations for murine hepatic Treg cells in the context of chronic CCl₄-induced injury (Fig. S3E). Together, these data suggest that Treg cells sense alterations in the hepatic environment to mediate their localization and activation in the setting of fibrosis.

To provide a more detailed view of Treg cells at steady state and in NASH, we performed single cell RNA-Seq of Treg cells isolated from the liver, spleen and visceral adipose tissue (VAT) of mice fed normal chow or the NASH-inducing CDAA-HFD. VAT Treg cells were analyzed due to their documented unique transcriptome and T cell receptor (TCR) repertoire, as well as their role in protecting from metabolic disease^22,33–36. Four clusters of Treg cells were observed: (1) a naïve and/or circulating population characterized by expression of Ccr7 and Sell (which encodes CD62L; hereafter referred to as “naïve”), (2) an activated tissue-resident cluster expressing Ccr8, (3) a memory population expressing Ccr2, Ccr4 and Rorc (hereafter referred to as the “Ccr2” cluster), and (4) a cluster expressing Nkg7 (Fig. 2A–2B). Areg was predominantly expressed by the Ccr2 and Ccr8 Treg cell clusters (Fig. 2B). At steady state, liver Treg cells were enriched for the Ccr2 cluster, while the Nkg7 population was greatest in the VAT, and the spleen was enriched for naïve Treg cells (Fig. S4A). Liver Treg cells expressed distinct chemokine and cytokine receptors compared those isolated from spleen and VAT (Fig. 2C–2D). Liver Treg cells expressed higher levels of Itgae (encoding CD103) compared to VAT Treg cells, while Cd69 was more highly expressed by VAT compared to liver Treg cells (Fig. 2D), suggesting differential tissue retention molecules between the tissue Treg cell populations. Furthermore, gene ontology analysis revealed heightened responsiveness to microenvironmental cues for Treg cells isolated from liver, as compared to those from spleen or VAT (Fig. S4B–S4C). To identify the transcription factors responsible for the observed liver-specific transcriptional profile, we utilized the ARACNe-AP algorithm^37,38 to reconstruct Treg cell transcriptional networks utilizing mutual information estimators from bulk RNA-Seq of Treg cells, which was then input into msVIPER³⁹ to score transcription factor activity in each sample (Fig. 2E). These analyses revealed that transcription factors such as T-bet, RORα, STAT5, PPARγ, AHR, GATA3 and Maf had increased activity in liver Treg cells compared to spleen Treg cells at steady state, while transcription factors such as Smad2, TCF1 and LEF1 had reduced activity in liver Treg cells compared to spleen Treg cells (Fig. 2F). Transcription factors whose activity were increased in both liver and VAT Treg cells compared to spleen Treg cells included PPARγ (known to regulate VAT Treg cell development⁴⁰), AHR, AR, HIF1α, and RORγt, while transcription factors with increased activity in liver Treg cells but not VAT Treg cells compared to spleen Treg cells included HOXB4, IRF4, IRF8, and STAT5 (Fig. 2G). Taken together, these analyses suggest that liver Treg cells share critical transcription factors with VAT Treg cells but that a distinct set of transcription factors were also responsible for the unique liver Treg cell transcriptional program.

Fig. 2. Liver Treg cells have a unique transcriptional profile in fibrosis.

(A) Uniform Manifold Approximation and Projection (UMAP) plot of single cell RNA-Seq analysis of Treg cells from liver, spleen and VAT of mice fed normal chow or CDAA-HFD.

(B) Expression of indicated gene as displayed by a feature plot among all Treg cells.

(C–D) Differentially expressed genes between Treg cells at steady state in (C) the liver and spleen and (D) liver and VAT.

(E) Overview of ARACNe-AP and msVIPER analyses to score transcription factor activity in Treg cells of each sample.

(F) Differential transcription activity in normal chow Treg cells from liver compared to spleen Treg cells, as analyzed by msVIPER.

(G) Transcription factors with increased activity in normal chow Treg cells from liver versus spleen that also had increased activity in VAT Treg cells compared to spleen (top) or had increased activity in liver versus spleen Treg cells but not in VAT versus spleen Treg cells (bottom).

(H–I) Differential gene expression and representative gene ontology (GO) terms between liver Treg cells from NASH compared to normal chow.

(J) Differential transcription activity in liver Treg cells from normal chow compared to NASH, as analyzed by msVIPER.

(K) Frequency of liver Treg cells in each cluster per mouse from normal chow or NASH.

(L) Proportion of Treg cell TCR clonotypes with each level of clonal expansion among liver Treg cells from Ccr8 cluster.

(M) Normalized entropy score of Treg cell TCR clonotypes from liver Ccr8 cluster.

See also Figure S3–S4.

We then considered the effect of NASH on Treg cells. NASH induced substantial transcriptional changes in liver Treg cells, including increased expression of genes such as Il2ra, Klrg1, Areg, chemokine and cytokine receptors including Il1rl1 (encoding ST2) and Il18r1, and reduced expression of genes such as the T cell stemness marker, Tcf7 (Fig. 2H). Gene ontology analyses revealed an increased activation signature for Treg cells in NASH livers (Fig. 2I). We next utilized the ARACNe-AP/VIPER approach (Fig. 2E) to analyze transcription factor activity in liver Treg cells. Transcription factors such as PPARγ, GATA3, Maf, NFAT, and RORγt – each with known functions in Treg cells – had increased activity in liver Treg cells from NASH compared to normal chow, while SMAD family members and LEF1 had reduced activity in liver Treg cells in NASH (Fig. 2F). Taken together, these data suggest that a mixture of transcription factors drive the liver Treg cell transcriptional profile in NASH. We next considered whether specific liver Treg cell subsets were uniquely altered. Only the Ccr8-expressing cluster of liver Treg cells was expanded in diet-induced NASH mice (Fig. 2K); this cluster displayed gene expression changes reflective of increased activation (Fig. S4D–S4E). Taken together, these data revealed that an activated subset of hepatic Treg cells isolated from fibrotic liver environments expressed unique transcriptional signatures reflective of their increased activation in the setting of chronic tissue damage and inflammation.

As TCR signaling is thought to contribute to Treg cell accumulation in the brain and VAT^27,35, we sought to examine the potential role of antigen-driven expansion of Treg cells in the setting of NASH. A greater proportion of Treg clonotypes were expanded in NASH livers than in normal chow, a pattern specific to the Ccr8-expressing cluster (Fig. 2L, S4F). Further, there was a trend towards a reduction in the normalized Shannon entropy in this Ccr8-expressing cluster in NASH samples (Fig. 2M), suggesting reduced clonal diversity within this cluster in NASH. However, the sum frequency of top clones was unchanged between normal chow and NASH liver samples (Fig. S4G). Moreover, we utilized the GLIPH algorithm that is designed to identify motifs predicted to bind the same MHC/peptide antigen^41,42. Across the normal chow and NASH liver samples, the GLIPH algorithm identified 279 motifs, the vast majority of which were found in both chow and NASH TCR repertoires (Fig. S4H), suggesting that Treg cell enrichment in NASH livers is not due to a select set of antigens and consistent with prior studies suggesting Areg production by Treg cell is independent of TCR signaling²⁵.

Treg cell–derived Areg directly activates hepatic stellate cells to promote liver fibrosis

To examine the tissue-specific cell types responding to Treg cell–derived Areg to promote liver fibrosis, we investigated whether Treg cell–derived Areg directly or indirectly activates hepatic stellate cells, the predominant (>95%) source of collagen-producing cells and myofibroblasts in liver fibrosis^15,43. We hypothesized that Treg cells may localize near the specific cell type that senses Treg cell–derived Areg, reflecting a direct signaling interaction between these cells. Using a tracing approach to identify hepatic stellate cells (Rosa26R^tdTomatoLrat-Cre)¹⁵, Treg cells, but not CD4⁺Foxp3⁻ Tconv cells, colocalized with tdTomato⁺ hepatic stellate cells during chronic CCl₄-induced injury (Fig. S5A–S5B) and in the FPC and CDAA-HFD models of NASH (Fig. 3A–3D). We next sought to validate these imaging findings with functional manipulations of hepatic stellate cells. In response to recombinant murine Areg stimulation in vitro, quiescent hepatic stellate cells increased expression of Lif, an EGFR signaling target gene⁴⁴, and genes indicative of hepatic stellate cell activation (Acta2, Timp1) (Fig. 3E, S5C). In addition, an increase in the expression of genes reported to be involved in hepatic insulin resistance, including Il6^45–48, Spp1^49,50 and Igfbp1⁵¹ (Fig. 3E, S5C), were also observed following Areg treatment. Areg induced expression of transcripts corresponding to EGFR signaling, cell proliferation and differentiation, and tissue repair and collagen biosynthesis gene expression signatures (Fig. 3F, S5C), consistent with Treg cell–derived Areg acting directly on hepatic stellate cells to promote their activation and liver fibrosis development.

Fig. 3. Treg cell–derived Areg directly activates hepatic stellate cells to promote liver fibrosis in NASH.

(A–D) Hepatic stellate cell lineage trace mice (Rosa26R^tdTomatoLrat-Cre) were fed the (A–B) FPC diet for 16 weeks or (C–D) CDAA-HFD for 8 weeks. (A, C) Treg cell (CD4⁺Foxp3⁺) and CD4⁺Foxp3⁻ T cell localization relative to hepatic stellate cells (tdTomato⁺) was assessed by confocal imaging analysis, with (B, D) quantification of the distance of Treg cells (Foxp3⁺ cells) and CD4⁺Foxp3⁻ Tconv (CD4⁺ Tconv) to the nearest hepatic stellate cell across multiple sections stained from n = 5 mice. Representative images (A, C), yellow arrowheads indicate Foxp3⁺ cells.

(E–F) RNA-Seq analysis of purified quiescent hepatic stellate cells left untreated, or treated with recombinant murine AREG (rmAREG) in vitro. n = 5 samples per group. (E) Selected genes significantly upregulated following Areg treatment and (F) representative gene ontology terms significantly enriched by Areg treatment.

(G–J) Sirius Red staining of livers from Egfr^Fl/FlLrat-Cre and littermate control Egfr^Fl/Fl mice fed (G–H) FPC diet for 16 weeks (n ≥ 7 mice per group) or (I–J) CDAA-HFD (n ≥ 4 mice per group). Data representative of 2–3 independent experiments.

(K) Expression of indicated genes in liver tissue as assessed by RT-qPCR of Egfr^Fl/FlLrat-Cre or littermate Egfr^Fl/Fl mice fed the FPC diet for 16 weeks. Data representative of 3 independent experiments.

(B, D, F, H, K): two-sided Student’s t-test. **p < 0.01, ***p < 0.001. Data are expressed as mean ± s.e.m.

See also Figure S5.

To functionally study the interplay of Treg cell–derived Areg with hepatic stellate cells in vivo, we generated mice with hepatic stellate cell–specific Egfr deletion⁵². Animals lacking EGFR on hepatic stellate cells (Egfr^Fl/FlLrat-Cre) exhibited no quantifiable steady state abnormalities when compared to Egfr^Fl/Fl littermate controls (Fig. S5D–S5H); however, in the setting of chronic CCl₄ injury (Fig. S5I–S5J), hepatic stellate cell–specific Egfr deletion resulted in reduced liver fibrosis. Moreover, in both the FPC diet and CDAA-HFD models of NASH, mice lacking EGFR on hepatic stellate cells developed less fibrosis (Fig. 3G–3J). Finally, loss of EGFR on hepatic stellate cells resulted in reduced expression of genes indicative of hepatic stellate cell activation in the livers of mice fed the FPC diet (Fig. 3K). The protection observed in these models was similar to that seen following loss of Treg cell–derived Areg (Fig. 1G–1K, S2G), offering in vivo functional evidence complementing our imaging studies (Fig. 3A–3D, S5A–S5B). Together, these findings supported the hypothesis that Treg cell–derived Areg activates EGFR signaling on hepatic stellate cells to promote liver fibrosis.

Treg cell–derived Areg promotes the activation of quiescent hepatic stellate cells in NASH

We next sought to examine the effect of Treg cell–derived Areg on hepatic stellate cell gene expression in NASH in vivo. Accordingly, we performed single cell RNA-Seq analysis of hepatic stellate cells enriched from livers of Areg^Fl/FlFoxp3^YFP-cre or littermate control Areg^Fl/Fl mice fed the CDAA-HFD. Four clusters of hepatic stellate cells were observed, characterized by expression of: (1) Fcna (a marker of quiescent hepatic stellate cells⁵³), (2) Mmp2 and (3) Cxcl5 (markers of intermediately activated hepatic stellate cells), and (4) Acta2 (representing highly activated myofibroblast-like hepatic stellate cells) (Fig. 4A–4B). The Acta2 subset of hepatic stellate cells expressed Timp1 and was the only observed cell population to express the cytokine Il6 (Fig. 4B). Treg cell–derived Areg induced gene expression changes predominantly in the Fcna- and Mmp2-expressing clusters (Fig. 4C). In the Fcna- and Mmp2-expressing clusters, Treg cell–derived Areg induced expression of genes encoding collagens and extracellular matrix–modifying enzymes and transcriptional signatures reflecting cellular proliferation (Fig. 4C–4D). Together, these data confirm the role for Areg from Treg cells in activating and transforming more quiescent hepatic stellate cell populations into pro-fibrotic Acta2-expressing cells in the setting of NASH.

Fig. 4. Treg cell–derived Areg promotes the activation of quiescent hepatic stellate cells in NASH.

(A) UMAP plot displaying clusters of cells in single cell RNA-Seq of liver cells enriched for hepatic stellate cells from Areg^Fl/Fl or Areg^Fl/FlFoxp3^YFP-cre mice (n = 2 mice per genotype) fed the CDAA-HFD.

(B) Feature plots displaying expression of indicated genes across cell clusters.

(C) For each indicated hepatic stellate cell cluster, differential gene expression analysis between cells from Areg^Fl/Fl compared to Areg^Fl/FlFoxp3^YFP-cre mice. Number of significantly up- and down-regulated genes (average Log₂ fold change > 0 or < 0, respectively; P_adj < 0.05) indicated.

(D) Representative gene ontology terms with differential enrichment between Areg^Fl/Fl compared to Areg^Fl/FlFoxp3^YFP-cre Fcna-expressing hepatic stellate cells (top) and Mmp2-hepatic stellate cells (bottom). Red: enriched in Areg^Fl/Fl cells; blue: enriched in Areg^Fl/FlFoxp3^YFP-cre cells.

(E) Overview of ARACNe-AP and VIPER analyses to determine relative transcription factor activity in hepatic stellate cells.

(F) Distribution of transcription factor activity in hepatic stellate cells from Areg^Fl/Fl compared to Areg^Fl/FlFoxp3^YFP-cre mice, with fibrosis activity score indicated by color. Fibrosis activity score based on output from ARACNe-AP.

(G) Volcano plot of transcription factor activity of hepatic stellate cells from Areg^Fl/Fl compared to Areg^Fl/FlFoxp3^YFP-cre mice. Fibrosis activity score for each transcription factor indicated by color of label.

To identify transcription factors responsible for the transcriptional changes observed in hepatic stellate cells in response to Treg cell–derived Areg, we utilized the ARACNe-AP algorithm^37,38 to reconstruct hepatic stellate cell transcriptional networks then incorporated these networks into msVIPER³⁹ to score transcription factor activity within our hepatic stellate cell single cell RNA-Seq dataset (Fig. 4E). Further, we utilized the ARACNe-AP output to generate a fibrosis activity score for each transcription factor, which quantifies their ability to regulate the expression of pro-fibrotic genes in hepatic stellate cells. Transcription factors with a high fibrosis activity score had increased activity in hepatic stellate cells from Areg^Fl/Fl mice compared to Areg^Fl/FlFoxp3^YFP-cre mice (Fig. 4F). Several transcription factors functionally shown to promote fibrotic gene expression in hepatic stellate cells had significantly greater activity in hepatic stellate cells from Areg^Fl/Fl mice compared to Areg^Fl/FlFoxp3^YFP-cre mice (Fig. 4G), such as GLI2⁵⁴, RUNX1 (that upregulates Timp1⁵⁵ and is thought to promote liver fibrosis⁵⁶), and CREB3L1, which was previously identified as a potential master regulator of the fibrotic response in hepatic stellate cells⁵⁷. STAT6, which downregulates Il6 expression⁵⁸, had enhanced activity in hepatic stellate cells from Areg^Fl/FlFoxp3^YFP-cre compared to Areg^Fl/Fl mice (Fig. 4G). Together, these data suggest that Treg cell–derived Areg is sensed by quiescent hepatic stellate cells in NASH and promotes activity of pro-fibrotic transcription factors that induce hepatic stellate cell activation and liver fibrosis.

Treg cell–derived Areg promotes glucose intolerance in a NASH-dependent manner through EGFR signaling on hepatic stellate cells

NASH and insulin resistance are closely related manifestations of metabolic disease. Treg cell immunosuppressive function plays a critical role in protecting HFD-fed mice (which do not progress to NASH) from developing insulin resistance^22,36; therefore, we sought to explore the potential contribution of Treg cell–derived Areg in the nexus of NASH and glucose intolerance using the FPC diet (which induces both NASH and insulin resistance³⁰). On the FPC diet, mice deficient for Areg production by all T cells (Areg^Fl/FlCD4-Cre), or specifically by Treg cells (Areg^Fl/FlFoxp3^YFP-cre), were significantly more glucose tolerant than corresponding littermate control (Areg^Fl/Fl) animals (Fig. 5A, S6A). This protection appeared to be liver-specific, as no differences in systemic glucose levels following injection of insulin were observed in the setting of FPC diet–induced NASH (Fig. S6B–S6C), consistent with insulin primarily inducing glucose uptake in the fat and muscle, but not liver⁵⁹. Further, no observable changes in fasting serum insulin levels were detected in the absence of either T cell– or Treg cell–derived Areg (Fig. S6D–S6E). In the setting of FPC diet–induced NASH, loss of EGFR on hepatic stellate cells in Egfr^Fl/FlLrat-Cre mice resulted in significantly greater glucose tolerance than in Egfr^Fl/Fl littermate controls (Fig. 5B), with no differences in systemic glucose levels observed following insulin injection (Fig. S6F) – consistent with a role for Treg cell–derived Areg, acting via EGFR signaling on hepatic stellate cells, in promoting hepatic glucose intolerance.

Fig. 5. Treg cell–derived Areg promotes glucose intolerance in a NASH-dependent manner through EGFR-signaling on hepatic stellate cells.

(A–B) Glucose tolerance tests (GTT) in (A) littermate Areg^Fl/Fl (black) and Areg^Fl/FlFoxp3^YFP-cre (blue) and (B) littermate Egfr^Fl/Fl (black) and Egfr^Fl/FlLrat-Cre (light blue) mice fed NASH-inducing FPC diet for 16 weeks. Data are representative of 3–4 independent experiments, each with n ≥ 4 mice per group.

(C–D) GTT of (C) littermate Areg^Fl/Fl (black) and Areg^Fl/FlFoxp3^YFP-cre (blue) and (D) littermate Egfr^Fl/Fl (black) and Egfr^Fl/FlLrat-Cre (light blue) mice fed a HFD that induces insulin resistance without NASH. Data are representative of 2–3 independent experiments, each with n ≥ 4 mice per group.

(E) GTT of mice with hepatocyte–specific Egfr deletion (Egfr^Fl/Fl - AAV8-TBG-Cre) or littermate controls (Egfr^Fl/Fl - AAV8-TBG-NULL) fed the FPC diet for 16 weeks. Adult Egfr^Fl/Fl mice were administered AAV8-TBG-Cre or AAV8-TBG-null viral particles 1 week prior to beginning diet experiment. n = 8 mice per group. Data are representative of 2 independent experiments.

(F) GTT of mice with hepatocyte–specific Egfr deletion or littermate controls (as in (E)) fed a HFD that does not induce NASH for 8 weeks. n ≥ 5 mice per group. Data are representative of 2 independent experiments.

(A–F): 2-way ANOVA with Holm Sidak post-hoc test. (A–F): two-sided Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns: not significant. Data are expressed as mean ± s.e.m.

See also Figure S6.

We next explored whether increased glucose tolerance in the absence of Treg cell–derived Areg also occurred in the absence of NASH, utilizing the standard HFD that induces insulin resistance systemically and in the liver⁶⁰ but does not induce NASH⁶¹. Loss of Areg from T cells, including specifically Treg cells, or EGFR on hepatic stellate cells, did not affect glucose tolerance in HFD-fed mice (Fig. 5C–5D, S6G). Furthermore, loss of Treg cell–derived Areg did not affect glucose sensitivity at an earlier timepoint (8 weeks) on the FPC diet (Fig. S6H), when liver steatosis is evident but prior to the development of NASH³⁰. To examine whether Treg cell–derived Areg promotes glucose intolerance by directly signaling to hepatocytes, we specifically deleted Egfr in hepatocytes via tail vein injection of AAV8-TBG-Cre viral particles in adult Egfr^Fl/Fl mice, an approach that leads to >99% Cre-mediated recombination specifically in hepatocytes (Fig. S6I)^62,63. Deletion of Egfr in hepatocytes did not alter glucose tolerance with or without NASH (Fig. 5E–5F) and did not affect NASH-induced liver fibrosis (Fig. S6J–S6K). Although EGFR signaling on hepatocytes is essential to liver regeneration⁶⁴, these data suggested EGFR signaling on hepatic stellate cells, not hepatocytes, promotes liver fibrosis and glucose intolerance in NASH. Taken together, these data demonstrated that Treg cell–derived Areg, via EGFR-mediated activation of hepatic stellate cells, promotes glucose intolerance in a NASH-dependent manner, in contrast to previous studies which utilized Treg cell–transfer and –depletion models to demonstrate a protective role for Treg cells in preventing insulin resistance in the setting of HFD feeding^22,36.

Treg cell–derived Areg promotes hepatocyte gluconeogenesis in NASH through hepatic stellate cell–derived IL-6

To study how Treg cell–derived Areg contributes to glucose intolerance, we considered the role of the liver in glucose homeostasis; specifically, that the liver serves as a key site of gluconeogenesis, a major therapeutic target in insulin resistance. In mice fed the FPC diet, but not the HFD, we observed reduced fasting blood glucose levels in the absence of Treg cell–derived Areg (Fig. 6A). For functional assessment of hepatic gluconeogenesis in NASH, a pyruvate tolerance test (PTT) was performed in FPC diet–fed mice. Loss of Treg cell–derived Areg led to reduced systemic blood glucose levels following pyruvate injection (Fig. 6B). Moreover, deletion of Egfr in hepatic stellate cells led to reduced gluconeogenesis in NASH, as assessed by fasting blood glucose levels (Fig. 6C) and the PTT (Fig. 6D), which, in sum, was consistent with Treg cell–derived Areg promoting hepatic gluconeogenesis in a NASH-dependent manner via EGFR signaling–mediated activation of hepatic stellate cells.

Fig. 6. Treg cell–derived Areg promotes hepatocyte gluconeogenesis in NASH through hepatic stellate cell–derived IL-6.

(A) Blood glucose levels following 16-hour fasting in littermate Areg^Fl/Fl (black) and Areg^Fl/FlFoxp3^YFP-cre (blue) mice fed (left) FPC diet for 16 weeks or (right) HFD for 8 weeks. Data are representative of 3–4 independent experiments, each with n ≥ 4 mice per group.

(B) Pyruvate tolerance test of littermate Areg^Fl/Fl (black) and Areg^Fl/FlFoxp3^YFP-cre (blue) mice fed the FPC diet for 16 weeks. Data are representative of 3 independent experiments, each with n ≥ 4 mice per group.

(C) Blood glucose levels following a 16-hour fast in littermate Egfr^Fl/Fl and Egfr^Fl/FlLrat-Cre mice fed (left) FPC diet for 16 weeks or (right) HFD for 8 weeks. Data are representative of 2–3 independent experiments, each with n ≥ 4 mice per group.

(D) Pyruvate tolerance test in mice with hepatic stellate cell–specific Egfr deletion or littermate controls on the FPC diet for 16 weeks. n ≥ 7 mice per group. Data are representative of 3 independent experiments.

(E–G) Basal (black) and cAMP/dexamethasone–induced (cAMP + dex; red) glucose production by in vitro–cultured primary mouse hepatocytes. (E) Hepatocytes plated in wells coated with indicated collagen densities. Data are representative of 5 independent experiments. (F) Hepatocytes treated for 18 hours with Medium 199 alone (with or without the addition of rmAREG; “cell-free media”), or treated with conditioned Medium 199 harvested following 23-hour culture of hepatic stellate cells with or without rmAREG. Data are representative of 3 independent experiments. (G) Hepatocytes treated for 18 hours with indicated recombinant protein. Data are representative of 3–5 independent experiments.

(H–I) Expression of Il6 in bulk liver tissue of mice on the FPC diet for 16 weeks. (H) Comparison of littermate Areg^Fl/Fl and Areg^Fl/FlFoxp3^YFP-cre mice. (I) Comparison of littermate Egfr^Fl/Fl and Egfr^Fl/FlLrat-Cre mice. Data are representative of 2–4 independent experiments, each with n ≥ 6 mice per group.

(J–K) GTT of littermate Areg^Fl/Fl (black) and Areg^Fl/FlFoxp3^YFP-cre (blue) mice fed FPC diet for 16 weeks and treated with (J) isotype or (K) IL-6R blocking antibody 48 hours prior to fasting for GTT (see methods). Data are representative of 3 independent experiments, each with n ≥ 4 mice per group.

(L–M) GTT of littermate Egfr^Fl/Fl and Egfr^Fl/FlLrat-Cre mice fed FPC diet for 16 weeks and treated with (L) isotype or (M) IL-6R blocking antibody 48 hours prior to fasting for GTT (see methods). Data are representative of 2 independent experiments, each with n ≥ 5 mice per group.

(A–D, H–M): two-sided Student’s t-test. (B, D–G, J–M): 2-way ANOVA with Holm Sidak post-hoc test. *p < 0.05, **p < 0.01, ***p < 0.001. ns: not significant. Data are expressed as mean ± s.e.m.

We next investigated how Treg cell–derived Areg promotes hepatic gluconeogenesis in NASH. We first examined whether tissue fibrosis influences hepatocyte glucose production, hypothesizing that Treg cell–derived Areg could contribute to hepatocyte gluconeogenesis indirectly via promoting NASH-mediated fibrosis by activation of hepatic stellate cells (Fig. 1, 3). Using an in vitro primary mouse hepatocyte glucose production assay^65,66, hepatocytes cultured on higher collagen density had significantly greater glucose production (Fig. 6E), suggesting Treg cell–derived Areg may promote hepatic gluconeogenesis by increasing the extent of liver fibrosis in the setting of NASH (Fig. 1G–1K). Alternatively, Treg cell–derived Areg could promote the production of a soluble mediator by hepatic stellate cells that directly influences hepatocyte glucose production. Thus, we adapted the hepatocyte glucose production assay to treat hepatocytes with media conditioned by Areg-stimulated hepatic stellate cells prior to assessing glucose production. We observed increased glucose production with media conditioned by Areg-treated hepatic stellate cells relative to cell-free media controls (with or without the addition of Areg), and media conditioned by hepatic stellate cells without Areg stimulation (Fig. 6F), suggesting Areg promotes the production of soluble factor(s) by hepatic stellate cells that stimulate hepatocyte glucose production. We then focused on assessing potential hepatic stellate cell–derived mediators that were upregulated in response to Areg stimulation (Fig. 3E, S5C), screening hepatic stellate cell–derived molecules by treating hepatocytes with corresponding recombinant versions of each candidate. Primary murine hepatocytes treated with recombinant IL-6 produced significantly greater glucose (Fig. 6G), consistent with previous studies that indicate IL-6 promotes hepatic insulin resistance through IL-6/STAT3-induced inhibition of insulin receptor signaling^18,45,46, and work demonstrating that IL-6 receptor signaling on hepatocytes promotes hepatic gluconeogenesis in mouse models of acute stress⁶⁷. Several other candidate mediators (Fig. 3E, S5C), including connective tissue growth factor (CTGF/CCN2), stanniocalcin-1 (STC-1), osteopontin, and insulin like growth factor binding protein 1 (IGFBP-1) did not affect glucose production when added directly to hepatocyte cultures (Fig. 6G).

We next interrogated whether Treg cell–derived Areg promotes glucose intolerance through IL-6 in vivo. Il6 expression was reduced in the livers of mice lacking Treg cell–derived Areg compared to littermate Areg^Fl/Fl controls when fed the FPC diet (Fig. 6H). Similarly, Il6 expression in mice with hepatic stellate cell–specific Egfr deletion was reduced compared to Egfr^Fl/Fl littermate controls in liver tissue from mice fed the FPC diet (Fig. 6I). As above, Il6 was expressed solely by activated hepatic stellate cells in NASH (Fig. 4B), and enhanced activity of STAT6 – known to downregulate Il6 expression⁵⁸ – was seen in hepatic stellate cells from NASH mice lacking Treg cell–derived Areg compared to controls (Fig. 4G). These findings were consistent with Treg cell–derived Areg, and EGFR-signaling on hepatic stellate cells, promoting Il6 expression by hepatic stellate cells in NASH. To functionally assess whether Treg cell–derived Areg promotes glucose intolerance via IL-6 in vivo, we blocked IL-6/IL-6R signaling prior to the glucose tolerance testing in mice on the FPC diet. Injection of an antibody that blocks IL-6/IL-6R signaling, but not an isotype control antibody, restored glucose tolerance in Areg^Fl/Fl mice to levels similar as those lacking Treg cell–derived Areg (Areg^Fl/FlFoxp3^YFP-cre) (Fig. 6J–6K), consistent with IL-6R signaling being responsible for the observed glucose intolerance in the setting of FPC diet–induced NASH. Similarly, blockade of IL-6/IL-6R signaling restored glucose tolerance in Egfr^Fl/Fl mice to levels similar as those lacking EGFR on hepatic stellate cells (Egfr^Fl/FlLrat-Cre) (Fig. 6L–6M). Taken together, these data supported a mechanism whereby Treg cell–derived Areg drives hepatic glucose intolerance through EGFR-mediated activation of hepatic stellate cells – exacerbating liver fibrosis and increasing production of hepatic stellate cell–derived IL-6 – which in turn stimulate hepatocyte gluconeogenesis and raise serum glucose levels.

DISCUSSION

In settings of infection, inflammation, and tissue damage, Treg cell activity, recruitment, and mobilization is part of an adaptive response to challenge that restores and/or preserves tissue function. In cancer, Treg cells are implicated in promoting tumor growth⁶⁸, which can be viewed as a misreading of cues that would normally support tissue homeostasis. The production of Areg from Treg cells is critical to protection from acute lung, brain and muscle injuries^25–28. Here, we find that Treg cell–derived Areg promotes disease progression and severity, highlighting a maladaptive role for Treg cell–derived Areg in driving liver fibrosis and metabolic dysfunction.

Our study is consistent with prior work of Treg cell–derived Areg, in which early damage-associated inflammatory mediators released upon tissue injury such as IL-18 and IL-33, activate Areg production in Treg cells, whereas TCR signaling is dispensable for Areg production²⁵. Further, our data suggest that hepatocyte injury leads to the expansion of liver Treg cells, which subsequently interact with quiescent hepatic stellate cells to promote their activation. Across hepatocyte injury etiologies, liver Treg cells are expanded, but in NAFLD without NASH, Treg cell frequency and number were unchanged. The activation of quiescent hepatic stellate cells by Treg cell–derived Areg leads to the development of liver fibrosis and the production of IL-6 that promotes glucose intolerance in hepatocytes.

Treg cell depletion in models of liver fibrosis demonstrate a protective role for Treg cells^23,24. The tissue repair function of Treg cells is distinct from immunosuppressive effects²⁵, and our findings here are consistent with that framework. Treg cell depletion fails to distinguish Treg cell immunomodulatory activity from tissue repair properties and thus is unable to interrogate the role of each of these distinct functions. Although Treg cell immunosuppressive function may prevent aberrant inflammation and thus dampen fibrosis – in contrast to the effect of Areg – studies to examine such activity would have to include Treg cell–specific deletion of numerous immunosuppressive genes (e.g., Il2ra, Ctla4, Il10, among others).

Additionally, while earlier theories have posited that insulin resistance contributes to the progression of liver disease, increasing evidence points to a bidirectional relationship; however, thus far few growth factors and their sources have been identified in the interaction between NASH and insulin resistance. Previous studies have identified that depletion or dysfunction of visceral adipose tissue (VAT) Treg cells contributes to insulin resistance and sensitivity in the setting of HFD-induced obesity – a model devoid of NASH development – by increasing local inflammatory parameters^22,36. Our study is consistent with Treg cells in metabolic organs modulating glucose homeostasis and finds that Treg cells in the injured liver are activated and expand. However, rather than their reported immunosuppressive role being important to suppress insulin resistance in the setting of HFD-induced obesity, we demonstrate maladaptive sensing of factors associated with their tissue-protective function – specifically, Areg derived from liver Treg cells – as an upstream driver of glucose intolerance in NASH. We find that Treg cell–derived Areg promotes IL-6 expression in activated hepatic stellate cells. Although cytokines such as IL-6 have previously been reported to promote insulin resistance, studies suggest that IL-6 from adipose tissue macrophages is released systemically to signal to hepatocytes¹⁸; given their proximity and cell–cell contact interactions with hepatocytes, in the setting of NASH, activated hepatic stellate cells represent more likely candidates for inducing high local concentrations of IL-6 to signal to hepatocytes.

Although Cre may have toxicity to certain proliferating cells^69,70, in this report and prior studies^25,28, Treg cell frequencies and numbers are the same between Areg^Fl/Fl and Areg^Fl/FlFoxp3^YFP-cre mice, suggesting that Cre is not toxic to Treg cells at steady state or in injured tissue when they may be more proliferative. Although a prior report suggests that the Foxp3^YFP-cre allele is a hypomorph⁷¹, prior work^25,28 and data in this study find that Areg^Fl/Fl and Areg^Fl/FlFoxp3^YFP-cre mice do not differ in terms of immune infiltration in the setting of tissue damage and Treg cells from these mice exhibit similar functionalities in vitro. Furthermore, in this study, we utilized additional Cre drivers independent of Foxp3^YFP-cre, specifically the CD4-Cre and Vav1-Cre systems, in which we observed the same conclusions as with the Foxp3^YFP-cre allele, suggesting that the Foxp3^YFP-cre allele does not affect immune responses or overall Treg cell function in these mice and that this allele is not a hypomorph in this context.

In sum, we find that liver injury drives the expansion of Treg cells, and that the production of Areg by Treg cells drives a key set of cellular interactions, first by promoting hepatic stellate cell activation to contribute to liver fibrosis – the key prognostic factor for patients with liver disease – and the production of soluble mediators, including IL-6, by hepatic stellate cells to promote hepatocyte-driven glucose intolerance. These findings highlight potential therapeutic targets in chronic liver disease – broadening our understanding of how NASH contributes to insulin resistance – and identify a maladaptive role for tissue-protective Treg cells.

Limitations of the Study

Our study has several limitations. While our data suggest that Treg cells have a close interaction with hepatic stellate cells in the setting of hepatocyte injury, hepatic stellate cell–derived factors that mediate this response, such as chemokines, remain to be identified. Further, the effect of Treg cells in different stages of human liver fibrosis is unclear and will require analysis of many additional samples. Although we offer in vitro and in vivo data suggesting that IL-6 from hepatic stellate cells promotes hepatocyte gluconeogenesis, we do not provide direct functional evidence supporting this claim. As elevated circulating IL-6 is a proposed biomarker of HCC⁷² and IL-6 has been implicated in hepatocellular carcinoma development^73–75, whether the circuit identified in this study contributes to the development of HCC warrants further study.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Nicholas Arpaia (na2697@cumc.columbia.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

This paper does not report original code. All raw sequencing data have been deposited in the Gene Expression Omnibus (GEO) under accession GSE208706. Any additional information required to reanalyze the data reported in this paper are available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

Foxp3^GFP, Areg^Fl/Fl, Egfr^Fl/Fl, Lyz2^wt/cre (JAX stock #004781), Areg^Fl/FlFoxp3^YFP-cre, Areg^Fl/FlCD4-Cre, Rosa26R^tdTomato (JAX stock #007914), Vav1-Cre (JAX stock #008610) and Lrat-Cre transgenic mice have been previously described^{15,25,52,76–81}. Areg^Fl/FlLyz2^wt/cre, Areg^Fl/FlVav1-Cre, Egfr^Fl/FlLrat-Cre and Rosa26R^tdTomatoLrat-Cre were generated by breeding Areg^Fl/Fl, Egfr^Fl/Fl or Rosa26R^tdTomato mice as applicable with indicated Cre recombinase-expressing mice. For some experiments, C57BL/6 male mice – either 10 weeks old or retired breeders – were purchased from Charles River. Mice were screened for maintenance of the C57BL/6N Nnt allele. Areg^Fl/FlFoxp3^YFP-cre and Areg^Fl/Fl mice were screened for germline deletion, and mice with germline deletion were excluded from experimental cohorts. To delete Egfr on hepatocytes, 1 × 10¹¹ viral particles of AAV.TBG.PI.Cre.rBG (AAV8-TBG-Cre) or control AAV.TBG.PI.Null.bGH (AAV8-TBG-NULL) (Addgene, 107787-AAV8, 105536-AAV8, respectively) were diluted in sterile saline and administered via tail vein injection to 8–10 week old male Egfr^Fl/Fl mice. Experiments were begun 1 week following viral particle injection, and successful deletion was confirmed at the termination of the experiment. All experiments involving mice were performed under protocols AC-AAAQ8474 and AC-AABD8554 approved by the Columbia University Irving Medical Center (CUIMC) Institutional Animal Care and Use Committee. All mouse strains were maintained in the CUIMC animal facility in accordance with institutional guidelines.

METHOD DETAILS

Mouse liver injury models

For induction of CCl₄-mediated fibrosis, 8–10-week-old male mice were treated twice weekly with 0.5 μL/g CCl₄ (Sigma, Cat. # 319961) for 9 weeks. CCl₄ was diluted 4-fold in corn oil (Sigma, Cat. # C8267) prior to intraperitoneal injection. For acute CCl₄ injury, tissues were processed 48 hours after an acute injection of 0.5 μL/g CCl₄. For NASH-mediated liver injury, 10–12-week-old male mice were fed the FPC diet (Envigo, TD.190142) for 16 weeks, and the drinking water was supplemented with 23.1 g/L fructose (Sigma, Cat. # F2543) and 18.9 g/L glucose (Sigma, Cat. # 49159), or 10-12-week-old male mice were fed the choline-deficient, amino acid supplemented, high fat diet (CDAA-HFD) for 8 weeks (Research Diets, Cat. # A06071302). For insulin resistance without NASH, male mice were fed the high fat diet (HFD, 60% kcal from fat, TestDiet, Cat. # 58Y1) for 8 weeks. At the end of each experiment, mice were euthanized, and the livers were removed for further processing. In diet experiments, mice were fasted for 6 hours prior to euthanasia. In some experiments, blood was collected via cardiac puncture for serum analysis. For fibrosis quantification, matched liver lobes were fixed in 10% formalin, paraffin-embedded, and stained with Sirius Red. For Sirius Red staining quantification, 6–10 random images (10X magnification) were taken of each liver section and quantification of Sirius Red positive area was performed using ImageJ. Serum ALT levels were measured on a Heska Element HT5 as per the manufacturer’s protocol. Liver triglycerides were extracted using the Folch Extraction method. Briefly, liver tissue was homogenized in PBS. A chloroform methanol mixture (12 mL) was added to the homogenate (3 mL) to create an 8:4:3 ratio, the mixture was centrifuged, and the organic layer was isolated. The remaining mixture was again extracted, with addition of 10 mL solution of chloroform, methanol and water (86:14:1). Following centrifugation, the organic layer was again isolated. The combined organic layer extract was then dried via nitrogen gas two times, the second following addition of 15% Triton X-100 in chloroform. The lipids were reconstituted in water and quantified using the Infinity triglyceride reagent (Thermo Scientific), as per the manufacturer’s protocol. Serum insulin levels were measured via ELISA (Mercodia) as per the manufacturer’s protocol.

Metabolic testing was performed after 16 weeks on the FPC diet or 8 weeks on the HFD. For glucose and pyruvate tolerance tests, mice were fasted for 16 hours, then administered 1.5–2 g/kg glucose or 2 g/kg pyruvate via intraperitoneal injection. Blood glucose was measured prior to injection, and at 15-, 30-, 60-, 90-, and 120 minutes post-injection. For IL-6/IL-6R signaling blockade prior to the glucose tolerance testing of mice fed FPC diet, 48 hours prior to fasting, an isotope control (Bio X Cell, Clone LTF-2, 10 mg/kg) or anti-mouse IL-6R antibody (Bio X Cell, Clone 15A7, 10 mg/kg) was injected intraperitoneally. For insulin tolerance tests, mice were fasted 6 hours, then injected with 0.75 U/kg insulin (Sigma, Cat. # 11376497001) via intraperitoneal injection. Blood glucose was measured prior to insulin injection, and 20-, 40-, 60-, 90-, and 120 minutes post-insulin injection.

Murine AREG ELISA

Liver tissue was weighed and then snap frozen. Tissue was then homogenized in tissue lysis buffer (100 mM Tris pH 7.4, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, 0.5% sodium deoxycholate in water) with protease inhibitor and EDTA (Thermo Scientific). Total protein was quantified by the Bradford assay (Thermo Scientific). Areg protein concentration of tissue homogenate was determined using an Areg ELISA kit (R&D Systems) as per the manufacturer’s protocol.

Immune cell isolations

Splenic cell suspensions were prepared and red blood cells were lysed in ammonium chloride (ACK) buffer. For some experiments, CD4⁺ T cells were then enriched using the Dynabeads FlowComp mouse CD4 kit (Life Technologies) prior to staining. For isolation of liver leukocytes, the portal vein was cut and mice were perfused with ice cold PBS. The liver was then minced and digested in RPMI1640 supplemented with 5% fetal bovine serum (FBS, Corning), HEPES, GlutaMAX and Pen/Strep, with collagenase A (1 mg/mL, Sigma) and DNAse I (0.5 μg/mL, Sigma) at 37°C for 45–60 minutes with constant shaking. Following a wash in RPMI1640, lymphocytes were enriched via Percoll (GE Healthcare) density centrifugation (44% layered over 67%). After centrifugation, the lymphocyte layer was collected and washed, and following lysis of red blood cells with ACK buffer, cells were stained (see below). For flow cytometric analysis of myeloid cells, cells were stained directly following digestion without Percoll enrichment. For analysis of VAT lymphocytes, VAT was minced and digested in DMEM supplemented with 2% FBS and 1.5 mg/mL Collagenase type II (Sigma) for 25 minutes. Red blood cells were then lysed with ACK buffer, and cells were then stained (see below).

Flow cytometry analysis

For flow cytometry analysis, cells were labeled with Ghost Dye cell viability reagent (Tonbo Biosciences). The following antibodies were used for cell surface staining: anti-CD45, -Ly-6C (BioLegend), -CD19, -CD8a, -F4/80, -Ly-6G, -CD11b, -CD11c (Tonbo Biosciences), -CD3e, -TCRß, -CD4, -NK1.1, -MHCII (BD Biosciences). The following antibodies were used for intracellular staining in conjunction with the Foxp3/transcription factor staining buffer set (Tonbo Biosciences) as per the manufacturer’s protocol: anti-Foxp3 (eBioscience), -AREG (R&D). Secondary staining was performed following primary intracellular staining of Areg using strepdavidin-BV421 (BioLegend). Samples were acquired on an LSR Fortessa Flow Cytometer (BD Biosciences) and data were analyzed using FlowJo software (BD Biosciences).

Immunofluorescence microscopy studies

To induce liver damage, hepatic stellate cell lineage trace mice (Rosa26R^tdTomatoLrat-Cre) were either fed the FPC diet (as above) for 16 weeks or the CDAA-HFD for 8 weeks, or injected with CCl₄ twice weekly for 3–4 weeks (as above). Liver lobes were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) in PBS for 1 hour at room temperature, followed by dehydration in 30% sucrose in PBS overnight at 4°C. Liver samples were snap frozen in optimized cutting temperature compound (Fisher Healthcare), and 10 μm sections were cut and fixed for 10 minutes with ice cold acetone. For cleaved caspase-3 staining, liver sections were stained with anti-Cleaved Caspase-3 (Asp175) monoclonal antibody (Cell Signaling) as per the manufacturer’s protocol. For Treg localization studies, after antigen retrieval (Antigen Unmasking Solution, Tris-Based, Vector Laboratories), slides were blocked with horse serum (Gibco) in PBS + 0.3% Triton X-100. Primary staining with rabbit anti-CD4 (Abcam, ab183685) and rat anti-Foxp3 (Invitrogen, 14577382) was performed overnight in PBS + 0.3% Triton-X100 with 1% horse serum at 4°C. Slides were washed in PBS + 0.3% Triton-X100, followed by secondary staining with donkey anti-rat AF488 and anti-rabbit Cy5 (Jackson Immunoresearch) in PBS + 0.3% TritonX100 with 1% horse serum. Tile scan images were acquired on a Nikon Ti Eclipse confocal microscope using 20X magnification. Quantification of the distance between Foxp3⁺ and CD4⁺ cells to tdTomato⁺ cells was performed using a nearest neighbor search algorithm in R.

Hepatic stellate cell isolation and culture in vitro

Quiescent hepatic stellate cells were isolated as described previously⁸², with 200,000 cells plated per well in a 24-well plate in Medium 199 (Gibco; Cat. # 11150059) supplemented with PenStrep and 10% FBS. For in vitro conditioned media hepatocyte glucose production assays (see below), culture media was also placed in cell-free wells. After 5 hours, the cell culture media was changed to 0.1% FBS in Medium 199 with PenStrep and recombinant murine Areg (R&D Systems; 500 ng/mL) was added. For hepatocyte glucose production assays, conditioned media was harvested 23 hours after Areg-treatment began (see below for subsequent steps). For analysis of gene expression changes, after 4 hours, RNA was extracted using Trizol Reagent (Invitrogen). For gene expression analysis by RT-qPCR, cDNA was synthesized using qScript cDNA SuperMix (QuantaBio) and qPCR was performed using Maxima SYBR Green qPCR Master Mix (Thermo Scientific). Ct values were normalized to 18s levels. A list of qPCR primers is included in Table S2. RNA-Seq of hepatic stellate cells was performed by the Columbia Genome Center as follows: a poly-A pull-down was performed to enrich mRNAs from total RNA samples, followed by library construction using Illumina TruSeq chemistry. Libraries were then sequenced using an Illumina NovaSeq 6000 at the Columbia Genome Center. RTA (Illumina) was used for base calling and bcl2fastq2 (version 2.19) was used for converting BCL to fastq format, coupled with adaptor trimming. Pseudoalignment was performed to a kallisto index created from transcriptomes (Ensembl v96; Mouse:GRCm38.p6) using kallisto (0.44.0).

Hepatic stellate cell single cell RNA-Seq in NASH

For hepatic stellate cells single cell RNA-sequencing, male Areg^Fl/Fl or Areg^Fl/FlFoxp3^YFP-cre mice were fed the CDAA-HFD as above for 8 weeks. Hepatic stellate cells were enriched as described previously⁸², then viable nucleated cells were sorted using a FACS Aria cell sorter (BD Biosciences). At the Columbia Genome Center, libraries were prepared using the 10x Genomics Single Cell Gene Expression 3’ workflow and sequencing was performed on an Illumina NovaSeq 6000. CellRanger 6.1.2 was used to process the sequencing data. Analysis was subsequently performed using Seurat 4.2.0, using standard commands. Network analysis approaches were also used to analyze transcription factor activity in the hepatic stellate cell clusters of the single cell RNA-Seq. ARACNe-AP was used to identify the transcription factor regulon utilizing bulk RNA-Seq of hepatic stellate cells^37,38, as described previously (100 bootstraps; p value threshold of 10⁻⁸)⁸³. msVIPER was then used to infer the relative transcription factor activity in each sample³⁹, as described previously⁸³. The fibrosis activity score for each transcription factor was generated by calculating the sum of the product of the TF mode (direction of the interaction; −1 to 1) and likelihood (strength of the interaction; 0 to 1) outputs of the ARACNe-AP analysis for a set of pro-fibrotic genes (Col1a1, Col1a2, Col3a1, Col5a1, Acta2, Timp1, Lox).

In vitro metabolic assays

The primary hepatocyte glucose production assay was adapted from previous studies^65,66. To isolate primary hepatocytes, the liver was digested in situ via retrograde perfusion of 20 mL of EGTA-based solution⁸² followed by 50 mL enzyme buffer solution⁸² containing collagenase IV (Worthington, LS004188; 0.4 mg/mL). The liver was removed and dissociated to a single cell suspension in HBSS. The single cell suspension was washed twice with HBSS, and dead cells were removed via 35% Percoll density gradient centrifugation. Hepatocytes were resuspended in Medium 199 with Pen/Strep and 10% FBS and plated. To generate collagen-coated plates, collagen (Corning, 354236) was diluted in water for the indicated densities and plated for 1 hour at 37°C. Wells were washed with PBS prior to plating hepatocytes. For collagen density and recombinant protein assays, 100,000 cells were plated per well in a collagen-coated (for recombinant protein assays, 12.6 μg/cm²) 24 well plate. For hepatic stellate cell conditioned media experiments, 50,000 hepatocytes were plated per well in a collagen-coated (12.6 μg/cm²) 48 well plate. Two hours after plating hepatocytes, the seeding medium was replaced. Four hours later, the hepatocytes were placed in starving medium overnight: Medium 199 with Pen/Strep for collagen density experiments, Medium 199 with PenStrep supplemented with recombinant murine IL-6 (R&D systems; 406-ML; 10 ng/mL), CTGF (R&D systems; 9190-CC; 50 ng/mL), IGFBP-1 (R&D systems; 1588-B1–025; 100 ng/mL), STC-1 (R&D systems; 9400-SO; 50 ng/mL), osteopontin (R&D systems; 441-OP; 10 ng/mL), or hepatic stellate cell–conditioned media as described above for conditioned media experiments. To neutralize the effect of the supplemented recombinant Areg on hepatocytes, Areg neutralizing antibody (R&D Systems; AF989; 1 μg/mL) was added to conditioned media prior to hepatocyte culture. 24 hours after initial plating, hepatocytes were washed twice with PBS and glucose production medium was added to each well. Glucose production medium consisted of phenol red-free, glucose-free DMEM (ThermoFisher, A1443001) supplemented with 10 mM HEPES, 2 mM GlutaMAX, 2 mM sodium pyruvate (Sigma, P2256) and 20 mM sodium lactate (Sigma, L7022). To assess glucose production, pCPT-cAMP (Sigma, C3912; 100 μM) and dexamethasone (ThermoFisher, A13449; 1 μM) were added to glucose production media. Following a 6-hour incubation, glucose content in the supernatant was determined by a glucose-oxidase/peroxidase reaction (Sigma, GAGO20).

Treg cell RNA-sequencing and analysis

Foxp3^GFP male mice were intraperitoneally injected twice weekly for 4 weeks with CCl₄ or fed the CDAA-HFD for 8 weeks, then liver, spleen and/or VAT cell suspensions were prepared and stained as described above. For bulk RNA-Seq analysis, Treg (CD4⁺GFP⁺) and Tconv (CD4⁺GFP⁻) cells were sorted from liver and spleen samples on a FACS Aria cell sorter (BD Biosciences) into Trizol Reagent (Invitrogen). RNA-sequencing was performed at the Columbia Genome Center. Clontech Ultra Low v4 kit was used for cDNA synthesis followed by Nextera XT library preparation, followed by sequencing using an Illumina NovaSeq 6000. RTA (Illumina) was used for base calling and bcl2fastq2 (version 2.19) for converting BCL to fastq format, coupled with adaptor trimming. Pseudoalignment was performed to a kallisto index created from transcriptomes (Ensembl v96, Mouse:GRCm38.p6) using kallisto (0.44.0). Differential gene expression analyses were performed using the DESeq2 package⁸⁴. Human single cell RNA-sequencing data were previously generated and were accessed from Gene Expression Omnibus (GEO) accession GSE136103.

For single cell RNA-Seq, liver, VAT and spleen cells from Foxp3^GFP mice fed normal chow or the CDAA-HFD for 8 weeks were prepared as above. Prior to fluorescent antibody staining, cells from each mouse were stained with a distinct TotalSeq-C hashtag antibody (BioLegend), as per the manufacturer’s protocol. Treg cells were then sorted as above (CD4⁺GFP⁺). At the Columbia Genome Center, libraries were prepared using the 10x Genomics Single Cell 5’ workflow for gene expression, cell hashing, and full-length V(D)J T cell receptor sequencing. Sequencing was performed on an Illumina NovaSeq 6000 and CellRanger 6.1.2 was used to process the sequencing data. Analysis was subsequently performed using Seurat 4.2.0, using standard commands. Network analysis approaches were also used to analyze transcription factor activity in the Treg cell clusters of the single cell RNA-Seq. ARACNe-AP was used to identify the transcription factor regulon utilizing bulk RNA-Seq of Treg cells^37,38, as described previously (100 bootstraps; p value threshold of 10⁻⁸)⁸³. msVIPER was then used to infer the relative transcription factor activity in each population of interest³⁹, as described previously⁸³.

Human liver staining

Formalin-fixed, paraffin embedded (FFPE) liver samples were obtained from core needle biopsies (NASH patients) or non-cancerous tissue from surgical resections (normal liver). Consecutive 3 μm thickness sections from FFPE tumor tissues were cut and transferred onto positively charged glass slides (Superfrost Plus). Slides were dried overnight and stored at 4°C until use. The Ventana Discovery Ultra staining platform was used. FOXP3 (clone SP97, Abcam, 1:50 dilution) and aSMA (clone 1A4, Dako, 1:200 dilution) were double stained using the DISCOVERY ChromoMAP DAB kit (Roche) and DISCOVERY Yellow kit (Roche), respectively, as per the manufacturer’s protocol. CD3 (polyclonal, Dako, 1:100 dilution) and CD8 (clone SP16, Abcam, 1:100 dilution) were double stained using the DISCOVERY ChromoMAP DAB kit and DISCOVERY Purple kit (Roche), respectively, as per the manufacturer’s protocol. Cell counting of FOXP3⁺ and CD3⁺ cells was performed manually by a trained pathologist blinded from clinical data and results were expressed as a ratio of number of cells per 10 mm² of liver tissue. The whole tissue area was considered for biopsies and a representative area was selected for the surgical specimens.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analyses were performed in GraphPad Prism 9. Two-sided unpaired t-tests or 2-way ANOVA with post-hoc Holm-Sidak tests were used, as applicable. p < 0.05 was considered statistically significant. Relevant procedures are indicated in figure legends throughout the manuscript, with the number of replicates indicated in the corresponding figure legend for each experiment. Blinding and randomization of subjects was used throughout this study as appropriate, and sample size estimation was based upon previous studies. All data, samples, and subjects were included in the study.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rat Anti-Mouse CD45	BD	564225
Anti-Mouse CD19	Tonbo	65-0193-U100
anti-mouse CD3e	Tonbo	65-0031-U100
anti-mouse TCRb	BD	563135
Rat Anti-Mouse CD4	BD	612843
anti-mouse CD8a	Tonbo	50-0081-U100
anti-mouse/rat Foxp3	Thermo Scientific	11-5773-82
anti-mouse NK-1.1	BioLegend	108714
anti-Mouse Amphiregulin, Biotin, Polyclonal	R&D Systems	BAF989
anti-mouse Ly-6C	BioLegend	128033
anti-mouse F4/80	Tonbo	20-4801-U025
Anti-Mouse Ly-6G	Tonbo	60-1276-U100
anti-mouse CD11b	Tonbo	20-0112-U025
anti-mouse CD11c	Tonbo	35-0114-U100
anti-mouse MHC Class II (I-A/I-E)	Tonbo	75-5321-U100
InVivoMAb anti-mouse IL-6R	Bio X Cell	15A7
InVivoMAb rat IgG2b isotype control, anti-keyhole limpet hemocyanin	Bio X Cell	LTF-2
Rabbit Recombinant Monoclonal CD4 antibody	Abcam	ab183685
FOXP3 Monoclonal Antibody	Thermo Fisher	14-5773-82
Cleaved Caspase 3 (Asp175) Rabbit monoclonal antibody	Cell Signaling Technology	9579S
Mouse Amphiregulin Antibody	R&D Systems	AF989
TotalSeq-C0301 anti-mouse Hashtag 1 Antibody	BioLegend	155861
TotalSeq-C0302 anti-mouse Hashtag 2 Antibody	BioLegend	155863
TotalSeq-C0303 anti-mouse Hashtag 3 Antibody	BioLegend	155865
TotalSeq-C0304 anti-mouse Hashtag 4 Antibody	BioLegend	155867
TotalSeq-C0305 anti-mouse Hashtag 5 Antibody	BioLegend	155869
TotalSeq-C0306 anti-mouse Hashtag 6 Antibody	BioLegend	155871
TotalSeq-C0307 anti-mouse Hashtag 7 Antibody	BioLegend	155873
Recombinant Anti-FOXP3 antibody	Abcam	SP97
Mouse monoclonal Anti-aSMA	Dako	1A4
Rabbit polyclonal Anti-CD3E (human)	Dako	A0452
Rabbit recombinant monoclonal CD8 alpha	Abcam	SP16
Bacterial and virus strains
AAV.TBG.PI.Cre.rBG	Addgene	107787-AAV8
AAV.TBG.PI.Null.bGH	Addgene	105536-AAV8
Biological samples
Chemicals, peptides, and recombinant proteins
Carbon tetrachloride	Sigma	319961
Corn oil	Sigma	C8267
D-Fructose	Sigma	F2543
Glucose monohydrate	Sigma	49159
Recombinant human insulin	Sigma	11376497001
Percoll	Cytiva	17089101
Collagenase A	Sigma	11088793001
Collagenase type II	Sigma	C6885
Collagenase IV	Worthington	LS004188
Trizol Reagent	Invitrogen	15-596-018
Recombinant mouse amphiregulin	R&D Systems	989AR100CF
Recombinant murine IL-6	R&D systems	406-ML
Recombinant murine CTGF	R&D systems	9190-CC
Recombinant murine IGFBP-1	R&D systems	1588-B1-025
Recombinant murine STC-1	R&D systems	9400-SO
Recombinant murine osteopontin	R&D systems	441-OP
pCPT-cAMP	Sigma	C3912
Dexamethasone	ThermoFisher	A13449
Detergent Compatible Bradford Assay	Thermo Scientific	23246
Halt Protease and Phosphatase Inhibitor Cocktails	Thermo Scientific	78442
Critical commercial assays
Infinity triglycerides Reagent	Thermo Scientific	TR22421
Mouse Insulin ELISA kit	Mercodia	NC9440604
Dynabeads FlowComp Mouse CD4 kit	Life Technologies	11461D
Ghost Dye Red 780	Tonbo Biosciences	13-0865-T100
Foxp3/transcription factor staining buffer kit	Tonbo Biosciences	TNB-0607-KIT
Antigen Unmasking Solution, Tris-Based	Vector Biosciences	H-3301
qScript cDNA SuperMix	QuantaBio	95048
Maxima SYBR Green qPCR Master Mix	Thermo Scientific	K0253
glucose-oxidase/peroxidase reaction	Sigma	GAGO20
DISCOVERY ChromoMAP DAB kit	Roche	05266645001
DISCOVERY Yellow kit	Roche	07698445001
DISCOVERY Purple kit	Roche	07053983001
Mouse Amphiregulin DuoSet ELISA	R&D systems	DY989
Deposited data
Raw and analyzed RNA-Seq data	This paper	GEO: GSE208706
Experimental models: Cell lines
Experimental models: Organisms/strains
Mouse: C57BL/6N	The Jackson Laboratory	JAX: 005304
Mouse: Foxp3GFP: B6.129-Foxp3tm2Ayr/J	Fontenot et al.76	MGI:3574964
Mouse: AregFl/Fl: Aregtm2c(EUCOMM)Hmgu	Arpaia et al.25	MGI:5806473
Mouse: Foxp3YFP-cre: B6.129(Cg)-Foxp3tm4(YFP/icre)Ayr/J	Rubtsov et al.81	MGI:3790499
Mouse: CD4-Cre: Tg(Cd4-cre)1Cwi/BfluJ	Sawada et al.80	JAX: 017336
Mouse: Lyz2wt/cre: B6.129P2-Lyz2tm1(cre)Ifo/J	The Jackson Laboratory	JAX: 004781
Mouse: Vav1-Cre: B6.Cg-Commd10Tg(Vav1-icre)A2Kio/J	The Jackson Laboratory	JAX: 008610
Mouse: EgfrFl/Fl: Egfrtm1Dwt	Lee et al.52	MGI:3513096
Mouse: Rosa26RtdTomato: B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J	The Jackson Laboratory	JAX: 007914
Mouse: Lrat-Cre: Tg(Lrat-cre)1Rshw	Mederacke et al.15	MGI:5545650
Oligonucleotides
Primers for qPCR, see Table S2	This paper	N/A
Recombinant DNA
Software and algorithms
ARACNe-AP	Lachmann et al.37	https://github.com/califano-lab/ARACNe-AP
msVIPER	Alvarez et al.39	https://www.bioconductor.org/packages/release/bioc/html/viper.html
DESeq2	Love et al.84	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Flowjo	BD	v10
Prism	GraphPad	Prism 9.0
Other
FPC diet	Envigo	TD.190142
CDAA-HFD	Research Diets	A06071302
High fat diet	TestDiet	58Y1

Highlights.

• Treg cells are enriched and activated in human and mouse chronic liver disease

• Treg cell–derived Areg promotes liver fibrosis and glucose intolerance

• Areg from Treg cells activates hepatic stellate cells via EGFR signaling

• Activated hepatic stellate cells promote hepatocyte gluconeogenesis via IL-6