Defective autophagy in CD4 T cells drives liver fibrosis via type 3 inflammation

Abstract

Conventional CD4 T cells represent a major source of inflammatory mediators that drive progression of chronic liver disease to fibrosis and to end-stage cirrhosis. Identification of T cell pathways that limits the inflammatory response could thus have therapeutic relevance. Here we show, using both human samples and mouse models, that autophagy is deficient in CD4 T cells from patients with advanced fibrosis, and that loss of autophagy following genomic deletion of ATG5 in T cells is associated with the emergence of pathogenic IL-17A + IFN-γ + Th17 T cells that drive liver fibrosis in mice. Mechanistically, liver CD4 T cells lacking autophagy display a Th17 glycolytic phenotype associated with enhanced type 3 cytokine (i.e., IL-17A and GM-CSF) release, shifting hepatic myofibroblasts, hepatocytes and macrophages toward a proinflammatory phenotype. We also show that autophagy can be rescued in CD4 T cells from patients with extensive liver fibrosis, leading to decreased frequency of pathogenic Th17 cells and reduced GM-CSF levels; in addition, limited fibrosis is observed in mice in which Rubicon, a negative regulator of autophagy, is deleted specifically in their T cells. Our findings thus implicate autophagy in CD4 T cells as a key therapeutic target to control inflammation-driven fibrosis during chronic liver injury.

Liver fibrosis is a consequence of the sustained inflammatory processes underpinning chronic liver disease. Here authors show that autophagy in CD4 T cells is an important process in preventing the emergence of pathogenic Th17 cells, which are causal to the progression of liver fibrosis.

Introduction

Chronic liver diseases, including metabolic, alcoholic, viral, or autoimmune diseases, are characterized by persistent inflammation that constitutes a key pathogenic link between parenchymal cell death and progression of fibrosis to end-stage cirrhosis¹. In response to repetitive liver injury, cross-talk between distinct populations of recruited or resident myeloid and lymphoid cells promotes the activation of the fibro-inflammatory properties of hepatic stellate cells (HSC), the key fibrogenic cells of the liver^1,2. Consequently, HSC acquire a myofibroblastic phenotype (hepatic myofibroblasts, MF), accumulate at the site of injury, produce extracellular matrix of altered composition and release cytokines and chemokines that perpetuate the inflammatory process^2,3. Within the last few years, advances in single-cell multi-omics have revealed the plasticity of immune cells and their heterogeneity in phenotypes and functions within subsets that can vary during fibrosis progression^1,4. However, as of today, antifibrogenic therapies are still lacking, emphasizing the need to better understand the mechanisms that govern the dialog of immune with fibrogenic cells and hepatocytes. Although recent studies have extensively focused on the innate immune response and the role of macrophage subsets in the fibrogenic process, studies in experimental models and human samples have also demonstrated that selective T cell subsets, in particular Th17 are a major source of inflammatory and fibrogenic mediators^1,5. Interestingly, Th17 are an heterogeneous and plastic population, with the capacity to express and produce several cytokines, including interferon (IFN)-γ and granulocyte-macrophage colony stimulating factor (GM-CSF) in addition to interleukin (IL)-17A, that further define a subset with a higher pathogenic potential^6,7. Although pathogenic Th17 have been well characterized in the context of inflammatory and autoimmune diseases, their involvement in the pathogenesis of liver fibrosis has not been described.

Macroautophagy (referred as to autophagy) is an essential cellular recycling pathway that engulfs cellular contents, including organelles and macromolecules, in a double-membrane autophagosome for lysosomal degradation⁸. Autophagy has emerged as an immunometabolic pathway that regulates the inflammatory response originating from innate, innate-like and adaptive immune cells^9,10, through its capacity to clear intracellular pathogens and damaged organelles, to enhance antigen presentation to MHC class II or thymic selection. In the liver, autophagy is a central regulator of chronic liver injury^11–13 and these data have recently been reinforced by the identification of loss of function mutations in the autophagic gene ATG7, that predispose to cirrhosis and hepatocellular carcinoma¹⁴. In T cells, autophagy constitutes a major regulatory mechanism that controls T cell function and fate, by maintaining T cell homeostasis and survival, as well as regulating their activation, differentiation, and metabolic activity^15–19. Consequently, dysregulation of autophagy in T cells has been linked to the development of inflammatory diseases, such as inflammatory bowel diseases or metabolic disorders, with different beneficial or deleterious outcomes^9,20,21.

In the present study, we identify an underappreciated role of T cell autophagy in the control of liver fibrosis. We describe that patients with extended fibrosis are characterized by defective autophagy in hepatic and circulating CD4 T cells. Modeling T cell autophagy defects observed in human samples, we show in mice models that autophagy deficiency in CD4 T cells is a major component of the inflammatory/fibrogenic response. Mechanistically, autophagy deletion shifts CD4 T cells toward a pathogenic Th17 glycolytic phenotype that drives a type 3 cytokine dialog with fibrogenic and other microenvironmental cells, including hepatocytes and macrophages. Finally, we demonstrate that autophagy can be restored in CD4 T cells from patients with extended fibrosis upon exposure to an autophagy activator, resulting in a reduction in the frequency of pathogenic Th17 and decrease of type 3 cytokines, and that selective T cell autophagy activation limits inflammation-driven liver fibrosis in mice. Our data identify autophagy in CD4 T cells as a target for antifibrotic therapy.

Results

Patients with extended fibrosis/cirrhosis and fibrotic mice show defective autophagy in CD4 T cells

We first evaluated whether autophagy gene expression is modulated in intrahepatic T cells from patients with cirrhosis. To that aim, we used publicly available single-cell RNA sequencing (scRNA-seq) datasets from livers of uninjured individuals and patients with cirrhosis of different etiologies (alcoholic liver disease n = 2, Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD, n = 2), and primary biliary cholangitis n = 1)⁴. We performed Gene Set Enrichment Analysis (GSEA)²² to evaluate the differential enrichment of an autophagy-related gene set defined by Kyoto Encyclopedia of Genes and Genomes (KEGG), on the pseudo-bulk differential expression of the pooled CD4 or CD8 T cells. GSEA of CD4 T cells revealed a significantly lower expression of the autophagic pathway in cirrhotic compared to uninjured livers (Fig. 1a). Detailed analysis of the genes in the leading edge of the GSEA showed downregulation of core genes involved in the autophagy pathway (Fig. 1b). In contrast, no significant downregulation of the autophagic pathway was observed in CD8 T cells from cirrhotic vs uninjured livers (Fig. 1a).

Fig. 1. Patients with extended fibrosis/cirrhosis show defective autophagy in CD4 T cells.

a GSEA from publicly available single-cell data⁴ of autophagy pathway genes in CD4 and CD8 T cells subsets from healthy and cirrhotic human livers (n = 5 individuals/group) (b) Heatmap of pseudo-bulk autophagic gene expression from the leading edge of the GSEA graph of the CD4 T subset. c Representative histograms and Geometric Mean Fluorescence Intensity (geoMFI) of LC3II + CD4 T cells in human intrahepatic leukocytes from patients with extended fibrosis/cirrhosis (F3-F4, METAVIR score) (n = 5) and controls (n = 8), incubated with 30 μM chloroquine (CQ) or vehicle (#p = 0.0156 by two-tailed Wilcoxon matched-pairs signed rank test and *p = 0.0186 by two-tailed Mann–Whitney test). d Representative confocal immunofluorescent images and percentage of LC3 (green) colocalized with LAMP1 (red) over total LC3 in CD4 T cells from liver (n = 9/group) and blood isolated from patients with extended fibrosis/cirrhosis (n = 13) compared to controls (n = 12) *p = 0.037. Scale bar is 1.99 μm. e Representative western blot analysis and quantification normalized to β-actin of autophagy-related proteins in CD4 T cells from blood of healthy and patients with mild or extended fibrosis/cirrhosis. Results are expressed as a fold over healthy except for pULK1 Ser⁷⁵⁷ (n = 12 healthy, n = 8 F1-F2, n = 16 F3-F4 and *p = 0.049 for Rubicon; n = 11 healthy, n = 8 F1-F2, n = 14 F3-F4 and **p = 0.005 for P62; n = 15 healthy, n = 8 F1-F2, n = 10 F3-F4 for ATG5-ATG12; n = 9 healthy, n = 10 F3-F4 and *p = 0.026 for pULK Ser⁷⁵⁷). (d-e) Data are mean ± S.E.M. Statistical analysis was performed by (a) GSEA (d) two-tailed Mann–Whitney for liver CD4 samples and (d–e) two-tailed univariate analysis for blood samples in which age of controls and patients differed statistically. Each variable achieving a p-value < 0.05 (i.e., only Rubicon) was then introduced into a two-tailed bivariate model (Table S2). Source data are provided as a Source Data file.

Further studies were performed to compare the variations in the components of the autophagic pathway in CD4 T cells from patients with extended fibrosis/cirrhosis (F3/F4 METAVIR score) and controls, both in liver and blood. No difference in age between patients and controls for liver samples was observed, whereas age differed significantly in blood samples (Table S1, S2). Therefore, and because autophagy has been reported to decline with age, in particular in CD4 T cells²³, bivariate analysis of the data adjusted on age was performed when appropriate. FACS analysis of the autophagy marker LC3-II showed a decrease in the average fluorescence intensity of LC3-II in intrahepatic CD4 T cells from patients with extended fibrosis/cirrhosis compared to control livers in the presence of chloroquine (CQ), demonstrating a reduction of the autophagic flux (Fig. 1c). Immunofluorescence analysis of autolysosome formation showed a lower number of LC3+ puncta colocalizing with the lysosomal marker LAMP1 in intrahepatic and circulating CD4 T cells from patients with extended fibrosis/cirrhosis (F3-F4, METAVIR score), compared to controls (Fig. 1d). In addition, blood CD4 T cells from patients with extended fibrosis/cirrhosis showed an increase in the expression of the inhibitory autophagy protein Rubicon^24,25, as well as of P62, the level of which is inversely correlated with autophagy levels¹⁰, whereas there was no significant difference in the expression of the ATG5-ATG12 complex (Fig. 1e). These alterations were also associated with an increase in the phosphorylation of the autophagy initiation complex protein ULK1 on Ser⁷⁵⁷, a phosphorylation site that prevents ULK1 activation and its interaction with AMPK²⁶ (Fig. 1e). Alterations in the autophagy proteins were observed at earlier (F1-F2, METAVIR score) stages of fibrosis but did not reach significance (Fig. 1e). We also analyzed autophagy flux in liver CD4 T cells from CCl₄-exposed mice, compared to that of spleen. There was no increase in LC3-II levels in liver CD4 T cells from fibrotic mice in the presence of CQ, indicating blockade of LC3-II degradation, whereas an increase was observed in the spleen, reflecting functional autophagy (Figure S1a). Taken together, these data demonstrate that autophagy is impaired in CD4 T, both in patients with extended fibrosis/cirrhosis and in experimental models.

Autophagy-deficient CD4 T cells show a pathogenic Th17 phenotype in fibrotic mice

We next investigated the effects of autophagy deficiency on T cell functions and the consequences on liver fibrosis. To that aim, we developed mice lacking ATG5 in T cells (ATG5^{TLymph−/−}) by crossing ATG5^fl/fl,27 with transgenic mice expressing the recombinase Cre under the Lck promoter. The efficiency of the deletion was confirmed by the decreased ATG5 protein expression and LC3-I accumulation in spleen T cells from ATG5^{TLymph−/−} mice (Figure S1b).

We evaluated the consequences of autophagy deletion on T cell activation in mice exposed to chronic toxic injury, as induced by repeated administration of carbon tetrachloride (CCl₄). FACS analysis of the intrahepatic T cell profile from ATG5^{TLymph−/−} vs ATG5^fl/fl mice showed enhanced T cell activation selectively in autophagy-deficient CD4, as reflected by increased frequency of CD25 + , CD69+ among CD4+ but not among CD8+ or CD4-CD8- T cells (Fig. 2a, b, and S1c). To better characterize the molecular changes in CD4 T cells resulting from autophagy deficiency, we performed RNA sequencing of sorted hepatic CD4 T cells from ATG5^{TLymph−/−} and ATG5^fl/fl CCl₄-exposed mice. A total of 2591 genes were differentially regulated between the 2 groups (Figs. 2c, and S1d), with 1508 genes up-regulated and 1083 down-regulated (adj p ≤ 0.05) (Fig. 2c). Among the up-regulated KEGG pathways with normalized enrichment score (NES) ≥ 1.8, most were related to “inflammation” “apoptosis/cell cycle” and “metabolism” (Fig. 2d, e), whereas most of the down-regulated genes were related to ribosomal pathways (Supplementary data 1).

Fig. 2. Hepatic T cell immune profiling reveals a shift toward a pathogenic Th17 phenotype in ATG5^TLymph-/- fibrotic mice.

Mice were (a–g) injected CCl₄ for 5 weeks or (h) subjected to BDL. a, b Flow cytometry analysis of activation markers CD25 (n = 7 ATG5^fl/fl and n = 6 ATG5^Tlymph-/- mice) and CD69 (n = 8 ATG5^fl/fl and n = 4 ATG5^Tlymph-/- mice) among (a) CD4+ **p = 0.0012 for CD25 and **p = 0.0081 for CD69 and (b) CD8 + T cells. c–e RNA-seq analysis of purified CD4 + T cells isolated from CCl₄-injected mice (n = 5 ATG5^fl/fl and n = 6 ATG5^Tlymph-/- mice). c Volcano plot indicating number of significantly up- and down-regulated genes (adj p ≤ 0.05). d Top KEGG pathways significantly enriched in ATG5^TLymph-/- compared to ATG5^fl/fl CD4 T cells. KEGG pathways are ordered by NES. e Selected genes significantly upregulated in ATG5^TLymph-/- compared to ATG5^fl/fl CD4 T cells. f Absolute numbers of non-pathogenic (IL-17A + IFN-γ- and IL−17A + IL-22 + , *p = 0.0478), pathogenic Th17 (IL-17A + IFN-γ+, *p = 0.0344), and GM-CSF+ among CD4 + T cells (n = 7 mice/group) in CCl₄ injected mice. g–j Frequencies and dot plots of non-pathogenic and pathogenic Th17 among CD4 + T cells in (g–h) CCl₄ model (n = 23 ATG5^fl/fl and n = 22 ATG5^Tlymph-/- mice, pooled data from 4 experiments, ****p < 0.0001 for IL-17A + IFN-γ- and IL-17A + IFN-γ+; n = 13 ATG5^fl/fl and n = 14 ATG5^Tlymph-/- mice, pooled data from 2 experiments, **p = 0.0045 for IL-17A + IL-22+ and n = 7 mice/group, ***p = 0.0006 for GM-CSF + ) and (i-j) BDL model (n = 10 ATG5^fl/fl and n = 9 ATG5^Tlymph-/- mice for IL-17A + IFN-γ-, IL-17A + IFN-γ + *p = 0.0289 and GM-CSF+ *p = 0.0295, pooled data from 3 experiments and n = 6 mice/group for IL-17A + IL-22 + , pooled data from 2 experiments). a, b, f, g, i Data are mean ± S.E.M. Statistical analysis was performed by (a, f, g, i) two-tailed Mann–Whitney, (c, e) DESeq2 and p-values were adjusted using the Benjamini & Hochberg method (d) GSEA and p-values were adjusted using Benjamini-Hochberg method. Source data are provided as a Source Data file.

Increased expression of inflammatory pathways in ATG5-deficient CD4 T cells included genes related to cytokine-cytokine receptor interactions and chemokine signaling (Il21, Ccr5, Cx3cr1) and Th17 differentiation pathways (Cxcr3, Rora, AhR, Irf4, Nfatc1, Gzmb) (Fig. 2d, e). FACS analysis further confirmed that hepatic ATG5^{TLymph−/−} CD4 T cells from CCl₄-exposed mice display a Th17 phenotype (Fig. 2f–h). The absolute number of non-pathogenic (IL-17A + IFN-γ-, IL-17A + IL-22 + ) and pathogenic Th17 (IL-17A + IFN-γ+) was increased, while there was no change in the number of GM-CSF + CD4 + T cells (Fig. 2f). In terms of frequency, similar increases in pathogenic Th17 were obtained both in the CCl₄ and the bile duct ligation (BDL) models of fibrosis, in addition to an increase in GM-CSF + CD4 T cell frequency (Fig. 2g–j). In contrast, the absolute number of IFN-γ + (Th1) and IL-4+ (Th2) CD4 T cells was not modified (Figure S1e). These data were confirmed by secretome analysis of activated spleen CD4 T cell from ATG5^{TLymph−/−} and wild type (WT) that showed enhanced levels of Th17-related cytokines, including IL-17A, IL-22 and GM-CSF (Figure S1f). Although, IL-17A may originate from innate-like T cells^1,28, the increase in the frequency of IL-17A + T cells was restricted to CD4 + T cells, with no modification among CD4-CD8- T cells between CCl₄-exposed ATG5^{TLymph−/−} and ATG5^fl/fl groups (Figure S1g); in addition, there were no change in the frequencies of IFN-γ + and TNF-α + among CD8 + T cells (Figure S1h).

ATG5 deficiency in CD4 T cells also resulted in an increase in genes related to cell death, survival and exhaustion. These data were supported by FACS analysis showing a decrease in the number and frequency of the intrahepatic CD4 + T cells among CD45+ immune cells observed in ATG5^{TLymph−/−} mice compared to WT counterparts (Figure S2a, S3). As previously reported in other organs^15–19, a decrease in the frequency of CD8 + T cells among CD45+ cells was also observed (Figure S2b). However, we found no difference in the frequency of CD4-CD8- T, Ly6G + , CD11c+ and an increase in TCRγδ + T cells in ATG5^{TLymph−/−} mice (Figure S2b). Enhanced apoptosis and exhaustion in autophagy-deficient CD4 T cells was further demonstrated by an increased frequency of cleaved caspase-3+ and PD-1+ among CD4 + T cells in ATG5^{TLymph−/−} vs WT mice exposed to CCl₄ (Figures S2a, S3), that was not observed in the other T cell subsets (Figures S2c, S3).

Finally, among metabolic pathways deregulated between CCl₄-exposed ATG5^{TLymph−/−} vs ATG5^fl/fl CD4 T cells, up-regulation of some genes related to oxidative phosphorylation (Ugqcr11, Ndufa13) and of a large number of glycolytic genes (Hk1, Hk2, Pkm GPi1, Aldoa, Pfkp, Pgk1, Gapdh, Eno1 and Hif1a) was observed (Fig. 2d, e). These data are in line with the reported role of glycolysis and oxidative phosphorylation in Th17 functions²⁹.

Altogether, these data demonstrate that in response to chronic liver injury, autophagy-deficient mice show reduced CD4 + T cell number and a shift toward an activated pathogenic Th17 phenotype, associated with a glycolytic/oxidative phosphorylation profile and enhanced exhaustion and apoptosis.

Genetic disruption of autophagy in T cells promotes liver fibrosis in mice

We next investigated the consequences of autophagy deletion in T cells on liver fibrosis in two classical models, i.e chronic toxic injury, as induced by repeated administration of CCl₄, and biliary fibrosis following bile duct ligation (BDL). ATG5^{TLymph−/−} exposed to chronic administration of CCl₄ for either 2 or 5 weeks showed enhanced liver fibrosis compared to their WT counterparts (ATG5^fl/fl), reflected by increased collagen accumulation in Sirius Red stained liver tissue sections (Fig. 3a) and enhanced density of α-SMA + fibrogenic cells (Fig. 3b), with no modification of liver injury (Figs. S4a, b). Enhanced liver fibrosis was associated with increased hepatic expression of inflammatory genes in ATG5^{TLymph−/−} mice compared to their WT littermates, including Tgfb1, Il1a, Il1b, Tnfa, Ccl2 and Ccr5 (Fig. 3c, Table S3). Administration of neutralizing IL-17A antibody to ATG5^{Tlymph−/−} mice reduced fibrosis triggered by CCl₄ administration, to the level observed in ATG5^fl/fl counterparts (Fig. 3d). When combining anti-IL-17A and anti-GMCSF, there was a tendency to further decrease fibrosis compared to IL-17A antibody alone (p = 0.06, Fig. 3d). Worsening of liver fibrosis following autophagy deletion in T cells was also observed following BDL (Fig. 3e, f) and was associated with enhanced hepatic expression of inflammatory genes (Fig. 3g) with no impact on liver injury (Fig. S4c). Taken together, these data demonstrate that autophagy deficiency in T cells exacerbates liver inflammation and fibrosis through IL-17A and GM-CSF-mediated pathway.

Fig. 3. Autophagy deletion in T cells exacerbates liver fibrosis in mice.

a, b Representative images and quantification of (a) Sirius red (SR) (b) α-SMA-positive areas in liver tissue sections from ATG5^fl/fl and ATG5^TLymph-/- mice injected CCl₄ for 2 (n = 5 ATG5^fl/fl and n = 6 ATG5^Tlymph-/- mice, **p = 0.0087 for SR and *p = 0.0455 for α-SMA) or 5 weeks (n = 21 ATG5^fl/fl and n = 22 ATG5^Tlymph-/- mice for SR, pooled data from 3 experiments (***p = 0.0006 ATG5^fl/fl 2 weeks vs ATG5^fl/fl 5 weeks and ***p = 0.0002 ATG5^fl/fl vs ATG5^TLymph-/-) and n = 13 ATG5^fl/fl and n = 9 ATG5^Tlymph-/- mice for α-SMA, pooled data from 2 experiments, **p = 0.0035). c Hepatic gene expression of inflammatory cytokines, chemokines and chemokine receptors from CCl₄-injected mice for 5 weeks (n = 13 mice/group, pooled data from 2 experiments, ***p = 0.0006 Tgfb1, *p = 0.0399 Il1a, *p = 0.0158 Il1b, *p = 0.0140 Tnfa, *p = 0.0440 Ccr5). d Representative images and quantification of SR of liver sections from ATG5^TLymph-/- mice injected for 2 weeks with CCl₄+ isotypes, CCl₄+ anti-IL-17A alone or CCl₄+ anti-IL-17A and anti-GM-CSF (n = 11 for ATG5^fl/fl+isotype, n = 12 for ATG5^Tlymph-/-+isotype and n = 7 for ATG5^Tlymph-/-+αIL-17A and ATG5^Tlymph-/-+αIL-17A + αGM-CSF, *p = 0.0255, **p = 0.0016, ****p < 0.0001). e, f Representative images and quantification of (e) SR and (f) α-SMA-positive areas in liver tissue sections from mice subjected to BDL (*p = 0.0435, **p = 0.0057). g Hepatic gene expression of inflammatory cytokines, chemokines and chemokine receptors from BDL mice (*p = 0.0207 Il1a, **p = 0.0094 Il1b, *p = 0.0182 Tnfa, *p = 0.0329 Ccl3, *p = 0.0295 Ccr2). e–g n = 10 ATG5^fl/fl and n = 9 ATG5^Tlymph-/- mice. Scale bar is 100 μm. Data are mean ± S.E.M. Statistical analysis was performed by two-tailed Mann-Whitney. Source data are provided as a Source Data file.

Autophagy-deficient CD4 T cells enhance the inflammatory functions of hepatic myofibroblasts and hepatocytes via an IL-17A-dependent pathway

In order to investigate the underlying fibrogenic mechanisms, functional co-culture experiments were performed. We first focused on the interplay of autophagy-deficient CD4 T cells with fibrogenic cells (Fig. 4a). In their myofibroblastic activated phenotype (MF), hepatic stellate cells are characterized by a high proliferative capacity, and contribute to the hepatic inflammatory response by producing pro-inflammatory mediators following activation by cytokines such as IL-17^1,28. Co-culture experiments of mouse hepatic myofibroblasts with CD4 T cells isolated from ATG5^{TLymph−/−} or WT mice did not affect their proliferative capacity, as assessed by BrdU incorporation (Fig. 4b). In contrast, IL-6, KC/IL-8, CCL2 and CXCL10 secretion was enhanced in hepatic myofibroblast/ATG5^{TLymph−/−} -CD4 T cell coculture; this increase was already observed upon co-culture with non-activated ATG5-deficient CD4 (Fig. 4c), but strongly potentiated by activated ATG5^{TLymph−/−} CD4 T cells (Fig. 4c, e). Similar results were obtained in conditioned media experiments (Fig. 4d), suggesting a contact-independent pathway. Production of CCL2 and KC/IL-8 was exclusively restricted to hepatic myofibroblasts, since no secretion was detected in activated CD4 T cells, whether from WT or ATG5-deficient mice, and CXCL10 and IL-6 were produced by CD4 T cells but at very low levels (Fig. S5). Neutralizing IL-17A in the conditioned media efficiently abrogated CCL2, CXCL10, IL-6 and KC/IL-8 production by hepatic myofibroblasts (Fig. 4e), in line with the reported presence of IL-17 receptors in these cells^30–32.

Fig. 4. Autophagy-deficient CD4 T cells enhance the inflammatory properties of hepatic myofibroblasts (MF) and hepatocytes.

a Experimental co-culture protocol of MF with either non-activated or activated spleen ATG5^TLymph-/- or WT CD4 T cells (b) BrdU incorporation in hepatic MF following direct co-cultures with n = 4 non-activated or activated CD4 T cell preparations. c–e ELIZA quantification of cytokine/chemokine levels in supernatant from MF (c) cocultured with non-activated or activated CD4 T cells (n = 8 CD4 + T cell isolations; n = 3 experiments for MF alone, *p = 0.0134, **p = 0.0047 for CCL2, **p = 0.0065, ***p = 0.0002 for IL-6 by two-tailed Mann–Whitney and ##p = 0.0078 by two-tailed Wilcoxon matched-pairs signed rank test for CCL2), (d) following direct contact or exposure to the conditioned medium (CM) of activated CD4 T cells (n = 6 different CD4 T cell isolations, *p = 0.0411, **p = 0.0022 by two-tailed Mann–Whitney) and (e) exposed to CM of activated ATG5^Tlymph-/- or WT CD4 T cells, following neutralization with 10 μg/ml IL-17A antibody or isotype (n = 12 different CD4 T cell isolations for isotypes group and n = 10 for ATG5^TLymph-/- CD4+anti-IL-17A, **p = 0.0029 for CCL2, *p = 0.0261, ***p = 0.0004 for CXCL10, ***p = 0.0004 for IL-6, ****p = < 0.0001for KC/IL-8 by two-tailed Mann–Whitney and ##p = 0.0059 for CCL2, ##p = 0.0098 for IL-6, ##p = 0.0020 for KC/IL-8 by two-tailed Wilcoxon matched-pairs signed rank test). Results are expressed as fold over MF +activated ATG5^fl/fl CD4 T cells. f Experimental co-culture protocol of primary hepatocytes with CM from activated CD4 T cells from either ATG5^TLymph-/- or WT mice. When indicated 10 μg/ml anti-IL-17A neutralizing antibody or control isotype was added. g mRNA gene expression of chemokines and cytokines in hepatocytes (n = 9 different CD4 T cell isolations, **p = 0.0019 for Cxcl1, ***p = 0.0005 (ATG5^fl/fl CD4+isotype vs ATG5^TLymph-/-CD4+isotype), ***p = 0.0003 (ATG5^fl/fl CD4+isotype vs ATG5^TLymph-/-CD4+anti-IL-17A) for Cxcl10, ***p = 0.0004 for Ccl2, **p = 0.0084 for Il6 by two-tailed Mann–Whitney and ##p = 0.0078 for Cxcl1, #p = 0.0391 for Cxcl2, #p = 0.0156 by two-tailed Wilcoxon matched-pairs signed rank test). Results are expressed as fold over hepatocytes +activated ATG5^fl/fl CD4 T cells. Data are mean ± S.E.M. ns: not significant. Source data are provided as a Source Data file. a, f Created in BioRender. Gilgenkrantz, H. (2025) https://BioRender.com/l88g711.

We next focused on the interactions between CD4 T cells and hepatocytes (Fig. 4f), since in response to chronic liver injury, damaged hepatocytes produce a large variety of inflammatory mediators that contribute to fibrosis progression through cross-talk with hepatic myofibroblasts^1,33. Exposure of hepatocytes to the conditioned media of ATG5^{TLymph−/−} CD4 T cells led to an increase in Cxcl1, Cxcl2, Ccl2 and Il6 mRNA expression, that was no longer observed in the presence of a neutralizing IL-17A antibody (Fig. 4g). These data are in line with the reported role of IL-17A on chemokine production by hepatocytes, including CXCL1, CXCL2 and CCL2, all of which playing a major role in HSC activation, neutrophil accumulation, myeloid cell recruitment and liver fibrosis^1,34,35.

These data demonstrate that autophagy deletion in CD4 T cells shifts hepatic myofibroblasts and hepatocytes, at least in part, toward a pro-inflammatory phenotype via an IL-17A-mediated pathway.

Autophagy-deficient CD4 T cells alter macrophage phenotype via a GM-CSF-dependent mechanism

Recruitment of monocytes that differentiate into pro-inflammatory/profibrogenic (scar-associated) macrophages is also a characteristic feature of chronic liver injury, and their role in the pathogenesis of liver fibrosis, has been extensively described^1,4. FACS analysis of the frequency of CD11b + F4/80+ monocyte-derived macrophages (MoMac) among CD45+ cells showed no difference in CCl₄-exposed ATG5^{TLymph−/−} vs WT mice. However, the frequencies of IL-1α + and IL-1β + MoMac were significantly increased in ATG5^{TLymph−/−} mice compared to ATG5^fl/fl, whereas that of TNF-α + MoMac was not affected (Fig. 5a, and S6). These data were further confirmed in co-culture experiments between LPS-stimulated naïve bone marrow-derived macrophages (BMDMs) and activated CD4 T cells isolated from ATG5^fl/fl or ATG5^{TLymph−/−} mice (Fig. 5b). As observed in vivo, autophagy-deficient CD4 T cells caused a significant increase in the frequency of IL-1α and IL-1β-producing macrophages, as well as a marginal but significant increase in CCL2 concentration, without modifications in TNF-α and IL-6 levels (Fig. 5c, and S7). These effects were preserved when transwell inserts prevented contact between autophagy-deficient CD4 T cells and BMDMs (Fig. 5d), suggesting a contact-independent interaction. Interestingly, whereas addition of an IL-17A neutralizing antibody did not modify the pro-inflammatory BMDMs phenotype induced by autophagy-deficient CD4 T cells (Fig. 5d), neutralizing GM-CSF in the co-culture media efficiently abrogated IL-1α and β production by macrophages (Fig. 5d).

Fig. 5. Autophagy-deficient CD4 T cells shift macrophages toward a pro-inflammatory phenotype.

a Flow cytometry analysis of the frequency and cytokine production of intrahepatic CD11b + F4/80+ MoMac from CCl₄-exposed ATG5^TLymph-/- and ATG5^fl/fl mice (n = 13 ATG5^fl/fl and n = 12 ATG5^TLymph-/- mice, pooled data from 2 experiments, *p = 0.0257 for IL-1α, *p = 0.0457 for pro-IL-1β). b Experimental protocol of bone marrow-derived macrophages (BMDMs)/CD4 T cell co-culture created in BioRender. Gilgenkrantz, H. (2025) https://BioRender.com/l88g711. c (left) Intracellular staining of 1 ng/ml LPS-stimulated BMDMs for IL-1α and pro-IL-1β. (right) ELIZA on the co-culture supernatant (n = 8 different isolations, *p = 0.0145, **p = 0.0045, ***p = 0.002). d Representative dot plots and mean quantification of IL-1α and pro-IL-1β in BMDMs co-cultures (n = 4 mice/group), transwell (n = 4 mice/group, *p = 0.0286), or in the presence of 10 μg/ml control isotype, IL-17A (n = 4 mice/group, *p = 0.0286)or GM-CSF neutralizing antibody for 24 hrs (n = 8 mice/group, **p = 0.0028, *p = 0.0390 and ##p = 0.0078 by two-tailed Wilcoxon matched-pairs signed rank test). Data are mean ± S.E.M. Statistical analysis was performed by two-tailed Mann-Whitney, unless otherwise indicated. Source data are provided as a Source Data file.

These data demonstrate that autophagy-deficient CD4 T cells shift macrophages toward an IL-1-producing phenotype through a GM-CSF-mediated pathway.

Activation of autophagy in T cells inhibits inflammation and reduces fibrosis

We next evaluated whether autophagy can be restored in CD4 T cells from patients with fibrosis. Activating autophagy in CD4 T cells from patients by AZD8055, a pharmacological ATP-competitive mTOR inhibitor³⁶, strongly decreased the phosphorylation of ULK at the inhibitory Ser⁷⁵⁷ site and resulted in an increase in the autophagic flux in the presence of chloroquine (CQ) (Fig. 6a). Restoring autophagy in CD4 T cells from patients with fibrosis led to a decrease in the frequency of both total Th17 and the pathogenic IL-17A + IFN-γ+ Τh17 cells, and to a reduction of GM-CSF production with no modifications in that IFN-γ+ cells (Fig. 6b). The consequences of autophagy activation in T cells on liver fibrosis were also investigated in mice with specific T cell deletion of the autophagy inhibitory protein, Rubicon (RUBCN^{TLymph−/−}). The efficiency of the deletion was confirmed by the decrease in Rubicon expression in CD3 T cells from RUBCN^{TLymph−/−} mice at the protein levels (Fig. S8a), and the resulting increase in autophagic flux in T cells from RUBCN^{TLymph−/−} mice compared to RUBCN^fl/fl (Fig. S8b). The extent of liver necrosis or serum transaminase levels did not differ between RUBCN^fl/fl and RUBCN^{TLymph−/−} following chronic exposure to CCl₄ (Figs. S8c, d). However, compared to RUBCN^fl/fl, RUBCN^{TLymph−/−} showed reduced Sirius red and α-SMA+ staining areas (Fig. 6c). In addition, hepatic expression of inflammatory genes including Tgfb1, Il1b, Tnfa, Ccl2, Cxcl10 and Ccr1 was down-regulated in RUBCN^{TLymph−/−} mice (Fig. 6d). FACS analysis showed that there was no modification in the frequency of pathogenic Th17 (IL-17A + IFN-γ+), IL-4+ or IFN-γ+ in CD4 T cells from CCl₄-injected RUBCN^{TLymph−/−}mice or of GM-CSF levels in isolated activated CD4 T cells from RUBCN^{TLymph−/−} (Fig. 6e and S8e, f). However, the frequency of non-pathogenic IL-17A + IL-22+ subset was increased in RUBCN^{TLymph−/−} mice compared to WT counterparts (Fig. 6e).

Fig. 6. Activation of autophagy in T cells inhibits inflammation and reduces fibrosis.

a Western blot analysis of CD4 T cell lysates from patients with extended fibrosis/cirrhosis exposed to 1 µM AZD8055 or vehicle for 3 hrs. LC3 expression was measured in the presence of 30 µM CQ (n = 7 patients, #p = 0.0156). Lanes have been rearranged; non-adjacent lanes are separated by a vertical black line. b Flow cytometry analysis of the frequency of IL-17A + IFN-γ + and IFN-γ + cells among CD4 + T cells (n = 8 patients, ##p = 0.0078, #p = 0.0312) and ELIZA quantification of GM-CSF secretion in PBMCs from patients with extended fibrosis/cirrhosis exposed to 10 μM AZD8055 (n = 7 patients). c, d RUBCN^Tlymph-/- and their WT counterparts mice were injected with CCl₄ twice a week for 5 weeks. c Representative images and quantification of SR (n = 18 RUBCN^fl/fl and n = 14 RUBCN^TLymph-/-mice, pooled data from 4 experiments, **p = 0.0065) and α-SMA + areas in liver tissue sections (n = 10 mice/group, pooled data from 2 experiments, ***p = 0.0006). d Hepatic expression of inflammatory cytokines, chemokines, and chemokine receptors genes (n = 18 RUBCN^fl/fl and n = 14 RUBCN^TLymph-/-mice, pooled data from 4 experiments, *p = 0.0282 for Il1b, *p = 0.0205 for Tnfa, *p = 0.0139 for Ccl2 and *p = 0.040 for Cxcl10). e Flow cytometry analysis of IL-17A + IFN-γ-, IL-17A + IFN-γ + (n = 17 mice/group, pooled data from 3 experiments), and IL-17A + IL-22+ (n = 12 RUBCN^fl/fl and n = 14 RUBCN^TLymph-/-mice, pooled data from 2 experiments, *p = 0.0322) CD4 T cell frequencies in the liver of RUBCN^TLymph-/- and WT littermates exposed to CCl₄. Scale bar is 100 μm. b–e Data are mean ± S.E.M. Statistical analysis was performed by (a, b) two-tailed Wilcoxon matched-pairs signed rank and (c–e) by two-tailed Mann-Whitney test. Source data are provided as a Source Data file.

These data show that defects in autophagy can be rescued in CD4 T cells from patients with extended fibrosis/cirrhosis, resulting in a decrease in type 3 cytokine release, and demonstrate that mice with selective autophagy activation in T cells show limited fibrosis development.

Discussion

Major advances have been made in the understanding of mechanisms underlying liver fibrosis progression and immune cells have been identified as key actors of the inflammatory-driven fibrogenic response. Conventional T cell subsets, including Th17, actively contribute to the fibrotic process by triggering the release of pro-inflammatory and fibrogenic mediators by hepatic stellate cells and epithelial cells, as well as through interactions with other immune cells such as macrophages and neutrophils, leading to amplification and perpetuation of the inflammatory response¹. Therefore, identification of pathways modulating T cell phenotype may serve as the basis for antifibrotic strategies. Our data identify autophagy as a central regulator of CD4 T cell phenotype during chronic liver injury, and demonstrate that autophagy defect in CD4 T cells is a major component of the inflammatory/fibrogenic response.

Autophagy is a key regulator of T cell activation, proliferation, survival and differentiation, along with controlling T cell metabolic capacity^37,38. A major finding of our study is that the autophagy machinery is selectively disrupted in intrahepatic CD4 T cells in experimental models and in patients with extended fibrosis, as illustrated by reduced autophagic flux, low autophagosome formation, decreased expression of core autophagic genes and enhanced expression of inhibitory autophagic proteins such as Rubicon. Although the initiating signal leading to T cell reduction of autophagy remains to be identified, these data are consistent with a more global impairment of autophagy in patients and mouse models of chronic liver injury, reported in liver sinusoidal endothelial cells and hepatocytes^11,13,39,40. The consequences of autophagy deficiency in CD4 T cells, explored by RNA sequencing profiling ATG5-deficient CD4 T cells from fibrotic mice, identifies deregulated pathways related to inflammation, apoptosis and metabolism. Genes related to Th17 differentiation and/or functions are strongly up-regulated, in particular those characteristic of the pathogenic Th17 subset, including Cxcr3, Rora, Irf4, Il12r1 and Gzmb^6,7,41. These data are corroborated in functional studies showing that in fibrotic mice, ATG5-deficient CD4 T cells shift toward a Th17 phenotype and produce higher levels of IL-17A and GM-CSF, both described as profibrogenic^30–32,42 and required for conferring pathogenicity to Th17^6,43. In line, functional studies show that ATG5-deficient mice exhibit an increase in the frequency of pathogenic Th17 in two models of fibrosis. ATG5 deficiency is also associated with the emergence of a unique CD4 transcriptional glycolytic signature (Hk1, Hk2, Pkm, Gpi1, Aldoa, Pkfp, Pgk1, Gapdh, Eno1 and Hif1a) and to up-regulation of genes from oxidative phosphorylation, described to support metabolic reprogramming through glycolysis and to direct lineage commitment of pathogenic Th17 cells⁴⁴. These data corroborate with the increased glycolytic capacity of pathogenic Th17 cells described in various inflammatory diseases^6,45. Moreover, they are in line with the reported role of a pathogenic Th17 subset that drives chronic metabolic liver injury, and is controlled by the glycolytic enzyme PKM2⁷. The link between defective autophagy and the emergence of pathogenic hepatic Th17 subset in patients with fibrosis remains to be further evaluated. However, our findings demonstrate that autophagy activation decreases the frequency of pathogenic Th17 in patient with extended fibrosis. They are further reinforced by the reduction in fibrosis when administrating neutralizing IL-17A antibodies to CCl₄-injected ATG5^{TLymph−/−} mice, an effect potentiated when combining with anti-GM-CSF.

The liver fibrogenic process is orchestrated by a coordinated dialog between fibrogenic and microenvironmental cells, including parenchymal and immune cells. Importantly, besides professional immune cells, hepatocytes and hepatic stellate cells have also emerged as key players in the immunoregulation of liver fibrosis, since they perpetuate the inflammatory process through their immunoregulatory properties, including their capacity to secrete cytokines and chemokines that sustains immune cell recruitment¹. A driving force of liver fibrosis is the development of type 3 inflammation, characterized by production of IL-17A and GM-CSF, that targets both hepatic myofibroblasts and hepatocytes and shifts macrophages toward a profibrogenic phenotype^1,46. In vitro studies demonstrate that increased release of distinct type 3 cytokines by ATG5-deficient CD4 T cells directs their interactions with hepatic myofibroblasts, hepatocytes and macrophages, leading to enhanced cytokine/chemokine secretion. Indeed, CD4 T cells interactions with hepatic myofibroblasts or hepatocytes are driven by IL-17, in line with the expression of IL-17R by both cell types and the reported increase in chemokine and cytokine release by both cell types upon exposure to IL-17A^{30–32,35,46}. Accordingly, autophagy-deficient CD4 T cells license hepatic myofibroblasts to secrete CCL2, CXCL10, KC/IL-8 and IL-6, and hepatocytes to express Cxcl1, Cxcl2, Ccl2 and Il6, all of which were previously reported to drive HSC activation, inflammatory cell recruitment and activation. In addition, autophagy-deficient T cells promote the emergence of macrophages with an IL-1α/IL-β pro-inflammatory/profibrogenic signature, in keeping with type 3 cytokines, in particular GM-CSF, as drivers of scar-associated macrophages differentiation during liver fibrosis⁴², whereas the contribution of IL-17A is more disputed³⁵ and rather predominantly targets non-hematopoietic cells^35,47. Indeed, our data demonstrate that interactions between autophagy deficient CD4 T cells and macrophages are promoted by GM-CSF and do not involve IL-17A. Finally, an additional amplification of the inflammatory reaction may also originate from the release of IL-1 and IL-6 by macrophages, hepatocytes and hepatic myofibroblasts via a positive feedback loop, in light of their role in Th17 activation⁶.

Impaired autophagy has been implicated in the development of various chronic inflammatory diseases, including liver diseases^{11,13,39,40,48–51}, and preclinical studies have shown that systemic administration of autophagy inducers, such as anti-DBI or Rapamycin, holds therapeutic potential to treat liver diseases due to their demonstrated anti-inflammatory effects^52,53. Our data in human samples show that autophagy defect can be rescued in CD4 T cells from patients with extended fibrosis, following inhibition of the autophagy negative regulator mTOR with AZD8055, and results in a decrease in the frequency of pathogenic IL-17A + IFN-γ + CD4 T cells and a reduction of GM-CSF production. These findings are corroborated by data in experimental models in which autophagy in T cells is activated following specific T cell invalidation of Rubicon in mice, leading to protection against liver fibrosis. Interestingly, beneficial effects on fibrosis in these mice are associated with an increase in non-pathogenic IL-17A + IL-22 + CD4 T cells, and are in line with the antifibrogenic properties of T cell-derived IL-22^32,54. Although Rubicon is recognized as a negative regulator of canonical autophagy^24,25, it also acts as a positive regulator of non-canonical autophagic pathways such as LC3-associated endocytosis and phagocytosis^51,55. However, modulation of the ULK1 initiation complex, specific for canonical autophagy, together with the mirror pro- and antifibrogenic effects upon ATG5 and Rubicon T cell deletion, respectively, strongly argue for a role of T cell canonical autophagy in the fibrogenic process.

In conclusion, our study provides new insights into the potential therapeutic benefits of enhancing autophagy in CD4 T cells and contributes to the growing evidence supporting the positive impact of autophagy induction as a therapeutic strategy for liver fibrosis.

Methods

Gene set enrichment and heatmap analysis

scRNA-seq data were from publicly available data, obtained from 5 patients with cirrhosis (alcohol-related liver disease (n = 2), Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) (n = 2), and primary biliary cholangitis (n = 1)), and 5 uninjured individuals⁴. Pseudo-bulk RNA counts were merged on a per patient basis and by merging all CD4 clusters (identified in http://www.livercellatlas.mvm.ed.ac.uk, as CD4 + T cell 1, 2 and 3) and all CD8 clusters (CD8 + T cell 1 and 2). Differential expression on pseudo-bulk was performed using limma trend⁵⁶. Upper-quartile normalization of the pseudo-bulk was used for heatmap representation.

Human liver samples

Liver samples (n = 11) were obtained from patients with extended fibrosis (F3-F4 Metavir score) undergoing resection or transplantation of the liver at the digestive surgical department of Beaujon Hospital (Clichy, France). Control liver samples were taken from the non-tumoral part of liver resection from patients with no cirrhosis (n = 14) (Table S1). Liver samples were obtained with informed consent from patients. The study was conducted in accordance with the Helsinki Declaration of 1975 and was approved by the Comité d’Evaluation de l’Ethique des projets de Recherche Biomédicale (CERB) Paris Nord (IRB 00006477 n°2017-013). Fresh liver specimens were examined by a pathologist and samples were collected at a distance from the tumor (when present) and surgical margins.

Human blood samples

Blood samples were collected following written informed consent from 42 patients hospitalized at Hôpital Beaujon (Clichy, France) with biopsy-proven fibrosis or cirrhosis, due to alcohol or MASLD (Table S2). The study complies with all relevant ethical regulations and was approved by local Ethics committee (Comité de protection des personnes Ile de France III N°3193 and Comité de protection des personnes Ile de France VI N°5018). Blood from 18 healthy donors was obtained through a formalized agreement with French Blood Agency (Etablissement Français du Sang, agreement n° 2015012778).

Human CD4 T cell isolation

CD4 T cells were isolated either from human peripheral blood mononuclear cells (PBMCs) or human intrahepatic leukocytes (IHLs). PBMCs were prepared from human blood samples using density gradient centrifugation as previously described⁵¹ on Ficoll-Paque Plus (Cytiva, cat#17144002), and were frozen in fetal bovine serum (FBS) with 10% DMSO and stored at -150 °C. Human IHLs were isolated as previously described⁵¹. Briefly, human liver pieces were mechanically dissociated, and the homogenate subjected to 33% Percoll (Cytiva, cat#17089101) density-gradient centrifugation at 690 g for 20 mins. CD4 T cells were isolated from frozen PBMCs or IHLs using CD4 microbeads (Miltenyi Biotec, cat#130-045-101), according to manufacturer’s instructions.

Mouse intrahepatic leukocytes (IHLs) and CD3 T cell isolation

Mouse IHLs were isolated as described in ref. ⁵⁷. Briefly, livers were digested using the Liver Dissociation Kit, mouse (Miltenyi Biotec, cat#130-105-807). After filtration through a 100 µm cell strainer and centrifugation at 300 g, the resulting cells were subjected to a 33% Percoll density gradient in RPMI 1640 complete medium (ThermoFisher) containing 10% FBS and centrifuged at 690 g for 20 mins. Red blood cells in the pellet were lysed in RBC lysis buffer (BioLegend, cat#420301). After washing, the isolated IHL were used for flow cytometry.

Mouse CD3 T cells were isolated using the Pan T Cell Isolation Kit, mouse (Miltenyi Biotec, cat# 130-095-130).

Immunofluorescent staining

Liver or blood human CD4 T cells were seeded at 0.1 × 10⁶ onto 0.1% Poly-L-lysine (Sigma)-coated glass coverslips in 24 well-plate, fixed with 4% PFA and incubated with rabbit anti-LC3 (Novus Biologicals, cat#NB100-2220, 1:200) and rat anti-LAMP1 (Abcam, cat#ab25245, 1:200), followed by goat anti-rat Alexa555 (ThermoFisher Scientific, cat#A21434, 1:500), or goat anti-rabbit Alexa488 (ThermoFisher Scientific, cat#A11034, 1:1000). Representative images were acquired using confocal microscope (Confocal Zeiss LSM 780), equipped with a 63X oil immersion objective. Ten fields were randomly selected, and the percentage of LC3 colocalized with LAMP1 over total LC3 *100 analyzed by Imaris x64 (Oxford Instruments, software 7.4.2) was averaged from around 50 cells per individual. Results were expressed as mean percentage of colocalization per individual. No staining signal was observed when omitting the primary antibody.

Immunoblotting

Human CD4 and mouse CD3 T cells were solubilized in RIPA lysis buffer (Thermofisher Scientific, cat#89901). Immunoblotting was performed with either rabbit anti-Rubicon (Cell signaling, cat#8465S, 1:1000), guinea pig anti-P62 (Progen, cat#GP62-C, 1:2000), rabbit anti-phosopho-ULK1 (ser⁷⁵⁷) (Cell signaling, cat#14202S, 1:1000), rabbit anti-ATG12 (Novus Biologicals, cat#NBP2-15501, 1:2000), rabbit anti-ATG5 ((Novus Biologicals, cat#NB110-53818, 1:2000), rabbit anti-LC3 (Novus Biologicals, cat#NB100-2220, 1:1000) and mouse anti-β-actin (Sigma-Aldrich, cat#A5441, 1:10000) followed by incubation with peroxidase-conjugated donkey anti-rabbit IgG (cat#711-035-152, 1:2500), donkey anti-guinea pig (cat#706-035-148, 1:2500) or rabbit anti-mouse IgG (cat#315-035-003, 1:10000) secondary antibodies (Jackson ImmnoResearch). Band intensity was quantified using the ImageJ software and normalized to the loading control for quantification. Uncropped and unprocessed scans are provided in the Supplementary information file.

Animals

ATG5-loxP/loxP²⁷ mice were obtained from Dr. Noboru Mizushima (University of Tokyo, Japan) and RUBCN-loxP/loxP⁵⁸ from Dr. T Yoshimori (Osaka University, Japan). Lymphocyte cell-specific RUBCN and ATG5-deficient mice were generated by crossing RUBCN-loxP/loxP or ATG5-loxP/loxP to Lck-Cre mice (Jackson Laboratory, Charles River France, L’Arbresle, France). The resulting double heterozygotes (Lck-Cre + /-, RUBCN + /loxP or ATG5 + /loxP) were backcrossed with RUBCN-loxP/loxP or ATG5-loxP/loxP mice, respectively, to produce lymphoid-specific RUBCN (Lck-Cre + /-, RUBCN-loxP/loxP, RUBCN^{TLymph−/−}) or ATG5 (Lck-Cre + /-, ATG5-loxP/loxP, ATG5^{TLymph−/−}) knockout mice. The wild-type littermates (Lck-Cre−/−, RUBCN-loxP/loxP, RUBCN^fl/fl and Lck-Cre−/−, ATG5-loxP/loxP, ATG5^fl/fl) were used as controls. C57BL/6 J mice were purchased from Janvier Labs (C57BL/6JRj, stock# RRID:MGI:2670020). All mice were bred separately and housed in different cages in specific pathogen-free (SPF) animal facilities and had ad libitum access to food and water. Dark/light cycles were 12 h/12 h, with temperature of 21 °C and humidity of 50%. Cervical dislocation was used for mice euthanasia. Males were used for animal studies since it has been extensively reported that they are more susceptible to developing liver fibrosis than females (10.1152/ajpendo.00427.2019).

Experimental models of liver fibrosis

Experiments were performed in accordance with all relevant ethical regulations and protocols were approved by the Ministère de l’Enseignement Supérieur, de la Recherche et de l’Innovation (APAFIS #30284-2021030912208867).

CCl₄-induced fibrosis

Liver fibrosis was induced in male mice aged 10-12 weeks, by intraperitoneal injection of 0.6 ml/kg body weight of carbon tetrachloride (CCl₄, Sigma-Aldrich cat#289116) diluted 1:10 in mineral oil (MO, Sigma-Aldrich, cat#M-5310) twice a week for 2 or 5 weeks, as indicated. MO was administered to the control group. Mice were sacrificed 24 hrs after the last injection of CCl₄.

Biliary fibrosis

was induced by ligating the common bile duct after midline laparotomy, in male mice aged 12–14 weeks. Mice were sacrificed 12 days after the surgery.

IL-17A & GM-CSF neutralization

in vivo ATG5^fl/fl and ATG5^{TLymph−/−} mice were injected with CCl₄ for 2 weeks as described above. ATG5^{TLymph−/−} mice were injected intraperitoneally every other day with either 200 μg/mouse of anti-mouse/rat IL-17A (InVivoMab, cat.# BE0173) alone, or combined with anti-mouse GM-CSF (InVivoMab, cat#BE0259), or the corresponding isotype controls (InVivoMab, mouse IgG1 cat#BE0083 and rat IgG2a cat#BE0089). Injections of the antibodies started one day before the first CCl₄ injection.

Histological analyzes

were performed on formalin-fixed paraffin-embedded tissue sections. Quantification from ten fields for each mouse (×10 magnification) was performed using ImageJ software in a blinded manner.

Immunohistochemical detection of α-SMA was performed on mouse liver tissue sections as previously described⁵¹ using the MOM immunodetection kit, peroxidase (Vector Laboratories, cat#BMK-2202) and a mouse monoclonal anti-α-SMA antibody (1:5000, Sigma-Aldrich, cat#A2547).

Flow cytometry

Mice intrahepatic T cells were analyzed by flow cytometry using combinations of the following fluorochrome-conjugated anti-mouse antibodies: CD3 (17A2, 1:50), CD4 (RM4-5, 1:200), CD8 (53-6.7, 1:200), TCRγδ (GL3, 1:500), CD25 (PC61, 1:50), CD69 (H1.2F3, 1:100) from BioLegend, CD45 (30-F11, 1:200) from BD Biosciences, and PD-1 (J43, 1:160) from ThermoFisher Scientific (Table S4).

Intrahepatic macrophages were identified using the following anti-mouse fluorochrome-conjugated antibodies: CD3 (17A2, 1:50), CD11c (N418, 1:100), Ly6G (IA8, 1:100), F4/80 (BM8, 1:100), Ly6C (HK1.4, 1:100) from BioLegend, CD45 (30-F11, 1:200) and CD11b (M1/70, 1:200) from BD Biosciences (Table S4).

For all experiments, dead cells were eliminated from the analysis using the eBioscience^TM fixable viability dye eFluor^TM 506 (ThermoFisher Scientific cat#65-0866-14).

Intracellular staining of mouse IHLs or human PBMCs

For detection of cytokine production in mouse T cells, IHLs were isolated and cells were stimulated for 4 hrs at 37 °C with 25 ng/ml PMA (Sigma-Aldrich, cat# P8139) and 1 μg/ml ionomycin (Sigma-Aldrich, cat# I0634), in the presence of 10 μg/ml Brefeldin A (BioLegend, cat#420601). For detection of cytokines produced by macrophages, IHL were stimulated with 10 ng/ml lipopolysaccharide (LPS) (Sigma-Aldrich, cat#L3024) for 3 hrs in the presence of Brefeldin A.

Cytokine production by human PBMCs was performed in cells exposed to anti-CD3/ anti-CD28 and 10 μM of AZD8055 or vehicle overnight and stimulated with PMA/ionomycin for further 4 hrs. Cells were stained with surface monoclonal antibodies for 30 mins at 4 °C, fixed, permeabilized using Cytofix/Cytoperm buffer (BD Biosciences, cat#554714), and labeled with antibodies against intracellular cytokines for 45 mins at 4 °C (Table S4). All data were acquired using the BD Biosciences Fortessa X20 flow cytometry and analyzed using FlowJo software (Tree star, 10.7.1 version).

Detection of LC3-II in human and mice samples was performed using the Guava Autophagy LC3 antibody-based detection kit (Luminex, cat#FCCH100171). Cells were incubated with 30 μM of either CQ or vehicle for 3 or 6 hours.

Liver function

Alanine aminotransferase (ALT), aspartate aminotransferase (AST) and direct bilirubin serum levels were measured at the Plateforme de Biochimie, INSERM U 1149, Paris, France.

RNA preparation and Real-time PCR

Pieces of liver tissue from two different lobes were collected from each mouse and were snap-frozen in liquid nitrogen. Total RNA was extracted using Qiazol lysis reagent (Qiagen, cat#79306) and RNeasy mini columns (Qiagen, cat#74104), as previously described⁴⁹ and cDNA prepared using High-Capacity cDNA Reverse Transcriptase Kit (ThermoFisher Scientific, cat#4368813). Real-time quantitative polymerase chain reaction (RT-qPCR) was performed with Absolute Blue QPCR SYBR Green mix (ThermoFisher Scientific, cat#AB-4166B) using Light cycler 96 Instrument (Roche). Primers for the different mouse genes are listed (Table S3). Gene expression was normalized to 18S, and relative expression was calculated using the (2^-ΔΔCt) method.

RNA-sequencing and analysis

CD4 + T cells were sorted from livers of ATG5^fl/fl and ATG5^{TLymph−/−} CCl₄-injected mice using BD FACSMelody^TM Cell Sorter (BD Biosciences). RNA was extracted using RNeasy Micro kit (Qiagen, cat#74004), and qualified with AGILENT Tapestation 2200 on High Sensitivity RNA chip. RNA library preparation was performed using Smarter stranded V3; cDNA and final libraries were controlled with AGILENT Tapestation 2200 on High Sensitivity DNA chip. Final equimolar pooled library preparations were sequenced using NovaSeq 6000 with SP Reagent Kit (800 Millions of 134Gbases reads). Reads were mapped to Ensembl human GRCh38 genome with v107 transcriptome annotation. Gene-level counts were obtained using featureCount. Differential gene expression analysis was performed on raw counts using DESeq2, and results were considered statistically significant for adjusted p values ≤ 0.05. Gene set enrichment analysis (GSEA) was performed using the fast GSEA implementation (10.18129/B9.bioc.fgsea; Bioconductor, open source software), pre-ranked by statistical value from the studied comparison by DESeq2, on the Mus musculus published pathways from the Kyoto Encyclopedia of Genes and Genomes (KEGG). Volcano plots were built using the -log10 of adjusted p value in function of the log2 fold changes. The RNA sequencing datasets generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE292735. [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE292735].

Co-culture experiments

Spleen CD4 T cells were isolated using CD4 (L3T4) MicroBeads, mouse (Miltenyi Biotec, cat#130-117-043) from either ATG5^fl/fl or ATG5^{TLymph−/−} mice were activated overnight with 2.5 μg/ml plate-bound anti-CD3 (BioLegend, clone 145-2C11, cat#100340) and 1 μg/ml suspension anti-CD28 (BioLegend, clone 37.51, cat#102116) antibodies in RPMI complete medium.

Mouse hepatic myofibroblasts (MF) were isolated as previously described³⁰. Cells were cultured to 80% confluency in DMEM supplemented with 10 % FBS, 2% penicillin-streptomycin, 5 ng/ml EGF and used between the fourth and ninth passage. Myofibroblasts were co-cultured with non-activated or pre-activated CD4 T cells for 24 hrs at a ratio of 1:10. For neutralizing antibody experiments, myofibroblasts were exposed for 24 hrs to the conditioned media from pre-activated CD4 T cells in the presence of 10 μg/ml anti-IL-17A (BioLegend, clone TC11-18H10.01, cat#506945) or isotype (BioLegend, cat#400432).

Bone marrow-derived macrophages (BMDMs) were isolated from the femur and tibia bones of ATG5^fl/fl mice. After flushing the bones with RPMI complete medium, bone marrow cells were cultured in RPMI complete medium supplemented with 10% L929 cell-conditioned medium containing M-CSF for 7 days. Adherent macrophages were co-cultured with pre-activated CD4 T cells at a 1:1 ratio in the presence of 1 ng/ml LPS. For neutralizing antibody experiments, BMDMs were co-cultured for 24 hrs with pre-activated CD4 T cells in the presence of either 10 μg/ml anti-IL-17A, 10 μg/ml anti-GM-CSF (BioLegend, clone MP1-22E9, cat#505402) or control isotype (BioLegend, cat#400432 and cat#400543).

Hepatocytes were isolated as previously described⁵⁹. Hepatocytes were seeded into 12-well plates with 50,000 cells/well in Williams medium with 10% FBS for 3 hrs, and serum-deprived for 24 hrs. After washing, cells were incubated for 48 hrs with conditioned media from activated CD4 T cells in the presence of 10 μg/ml anti-IL-17A or control isotype.

Transwell experiments

CD4 T cells (upper compartment) and macrophages (lower) were cultured in transwell plates in RPMI complete medium containing 1 ng/ml LPS at 37 °C for 24 hrs (Corning, cat# 3413, 0.4 µm pore membrane).

DNA synthesis

was assessed in MF following co-culture with CD4 T cells for 24 hrs. BrdU was added to the co-culture medium for the last 18 hrs. Following removal of CD4 T cells DNA synthesis was evaluated using the Cell Proliferation Eliza BrdU kit (Roche, cat#11296736001).

Cytokine and chemokine assay

The levels of cytokines and chemokines were quantified using either a customized kit from Meso Scale Discovery (MSD) or standard ELIZA assays (mouse CCL2 (cat#88-7391-88), TNF-α (cat#88-7324-88), IL-6 (cat#88-7064-88) and human GM-CSF (cat#88-8337-88)).

Statistical analysis

All data presented in graphs are expressed as mean ± standard error of the mean (S.E.M) or median (interquartile range), as indicated. Statistical analyzes were performed using two-tailed Mann-Whitney test to calculate significant levels between two groups. For comparison of means from multiple groups against one control group, Kruskal-Wallis with Dunn’s multiple comparison post-test analysis was employed. Two-tailed Wilcoxon matched-pairs signed rank test was used to assay the statistical significance when indicated. The potential relationship between age and autophagy markers was analyzed by linear regression univariate analysis. Each variable achieving a p-value < 0.05 (i.e., only Rubicon) was then introduced into a bivariate model. Analyzes were performed using GraphPad Prism 9.5.1 and SPSS 28.0.1.1 software. All p-values are two-sided and a p-value < 0.05 was considered statistically significant. Sample sizes were adequate to detect effects between groups, as determined by the reproducibility and variability of each experiment and limited by the availability of animal samples.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.A.S. designed, conducted, analyzed data from human, mice and cell experiments, compiled the figures and wrote the manuscript. J.W. and M.G. conducted and analyzed data from human CD4 experiments. C.C., M.A., S.B., M.A.C., J.K., M.C., M.M. and C.P. contributed to the acquisition of data for in vivo and in vitro experiments. T.T.S. and A.H. provided human liver and blood samples. R.N. supervised and analyzed scRNA-seq results. V.P. and E.W. helped with clinical insights and discussion. H.G. and E.W. contributed to the design and analysis of the data. L.S., H.G. and E.W. provided feedback. S.L. conceived, designed, provided financial support, wrote the manuscript, and supervised the project.

Peer review

Peer review information

Nature Communications thanks Samuel Huber, Frank Tacke and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA sequencing datasets generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE292735. [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE292735]. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-59218-y.