Wildtype, Il15^–/– and Il15ra^–/– mice in the C57Bl/6 background generated in our colony have previously been described (27, 34). Il15^–/– and Il15ra^–/– mice were back-crossed to C57BL/6J mice (Charles River, Canada) every four generations, and the three genotypes used in this study were established from these crosses. Analyses of gut microbiota of the three genotypes showed minimal differences (35). Mice were bred and housed in ventilated cages in the same housing unit throughout the experiment. The experimental protocols were approved by Animals Ethics committee of the Faculty of Medicine and Health Sciences, Université de Sherbrooke (2020-2732).

Liver fibrosis was induced in 8-10 week-old mice by intraperitoneal (i.p.) injection of CCl₄ as previously described (36). Male mice were used for fibrosis induction as female sex hormones reduced production of inflammatory cytokine in the liver (37). Briefly, CCl₄ (Sigma-Aldrich, Oakville, ON) diluted in corn oil (1:3) was injected via i.p. route (0.5 mL CCl₄ per g of body weight) twice a week for five weeks. Three to four days after the last treatment, mice were sacrificed, blood collected by cardiac puncture and liver and spleen tissues resected. Serum was separated and kept frozen at -80°C. Liver pieces were snap frozen and stored at -80°C for gene expression studies. For histopathology analyzes, tissue sections from the different liver lobes were fixed for 12-16 hours in buffered 4% paraformaldehyde solution and embedded in paraffin.

Quantitative determination of serum Alanine Aminotransferase (ALT) was performed using a kinetic assay according to the manufacturer’s instruction (Pointe Scientific, MI, USA).

Formalin-fixed paraffin-embedded tissue sections of 4 μm thickness were deparaffinized, rehydrated, and stained with hematoxylin and eosin (H&E), Sirius red/Fast green in Picric Acid (Fast green FCF, EMD Millipore Corporation, MA, USA) or Masson’s trichrome stain following standard procedures (36). Digital images of the stained sections were acquired using a Nanozoomer Slide Scanner (Hamamatsu Photonics, Japan) and analyzed by the Nanozoomer Digital Pathology software NDPview 2.7.52 (Hamamatsu Photonics). Sirius red staining positive areas were quantified using the NIH ImageJ software (version 1.53n). For immunohistochemistry, rehydrated liver tissue sections were immersed in Tris-EDTA Buffer (10mM Tris Base, 1mM EDTA Solution, 0.05% Tween 20, pH 9.0) and maintained at 90°C, 20 min in a steamer for antigen retrieval. Sections were then blocked with 5% BSA in Tris-buffered saline (TBS) containing 20% Tween-20 (TBS-T) followed by overnight incubation with primary antibody at 4°C. After washing, tissue sections were incubated with appropriate secondary antibody for 2h, washed and nuclei were stained using Hoechst 33342 DNA staining dye (Thermo Fisher Scientific, Canada). After washing, tissue sections were mounted with a coverslip. The list of antibodies used are listed in Supplementary Table 1 . Images were acquired in Nanozoomer and were analyzed using NDP-View2 and ImageJ software (Hamamatsu Photonics, Japan).

Quantitative RT-PCR was carried out as described before (36). Briefly, Total RNA from frozen tissues was extracted using QIAzol Lysis Reagent (Qiagen, Toronto, Ontario, Canada). cDNA was synthetized from 1μg of purified RNA using QuantiTect^® reverse transcription kit (Qiagen, Toronto, Ontario, Canada). Quantitative RT-PCR amplification reactions was carried out in QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, Canada) using SYBR Green Supermix (Bio-Rad, Mississauga, Ontario, Canada). The expression of indicated genes was measured using primers listed in Supplementary Table S2 . Gene expression levels between samples were normalized based on the Cycle threshold (Ct) values compared to housekeeping gene 36B4 (Rplp0). Statistical analyzes were performed using GraphPad Prism 9 software (San Diego, CA). The values are presented as mean ± standard error of the mean (SEM). The statistical significance (p value) was calculated as indicated in figure legends.

Intrahepatic lymphocytes were isolated as described (38). At sacrifice, liver tissues were collected, rinsed with Krebs–Ringer–Buffer (KRB; 154 mM NaCl, 5.6 mM KCl, 5.5 mM Glucose, 20.1 mM HEPES, 25 mM NaHCO3, pH 7.4) and digested in pre-warmed (37°C) KRB supplemented with 2 mM CaCl₂, 2 mM MgCl2, 300 CDU (casein digestion units)/mL Collagenase IV (Worthington) and 150 U/mL DNase I (Sigma) using the GentleMACS™ Dissociator (Miltenyi Biotec, Cambridge, MA, US) following the manufacturer’s instructions. The homogenized liver samples were incubated for 30 min at 37°C under gentle agitation, followed by a second round of digestion using the GentleMACS™ Dissociator. The digested liver tissues were passed through a 70 μm mesh size cell strainer and rinsed with cold PEB buffer (0.5% bovine serum albumin, 2 mM EDTA in Phosphate buffered saline). Samples were centrifuged at 50g for 5 min, at 4°C to eliminate contaminating hepatocytes. The supernatant was centrifuged at 300g for 10 min at 4°C to collect the leukocytes. Cells were further purified using a Percoll gradient as described (38), rinsed with PEB buffer and used for flow cytometry. Data were acquired on Cytoflex using CytExpert software (BeckMan Coulter, Indianapolis, IN) and analyzed using the FlowJo software (TreeStar Inc, Ashland, OR). Antibody panels used for identifying T cells and myeloid subsets are listed in Supplementary Table 3 .

To understand the role of IL-15 signaling in hepatic fibrogenic response, male wildtype (WT), Il15^–/– and Il15ra^–/– mice were treated with CCl₄ for 5 weeks to induce of liver fibrosis. We first examined the expression of the Il15 gene in CCl₄-treated livers. Wildtype mice showed a marked increase Il15 gene expression following CCl₄ treatment, whereas such an upregulation was not evident in the livers of Il15ra^–/– mice ( Figure 1A ). Serum alanine transferase (ALT) levels were higher in CCl₄-treated wildtype mice when compared to vehicle-treated controls ( Figure 1B ). In contrast, no difference in ALT levels were observed between control and CCl₄-treated Il15^–/– and Il15ra^–/– mice, suggesting that CCl₄-induced liver damage is amplified by IL-15 signaling via IL-15Rα. Sirius red staining of liver sections for collagen fibers revealed extensive liver fibrosis in CCl₄-treated wildtype mice when compared to the vehicle-treated controls with marked portal to portal bridging fibrosis ( Figure 1C ). Liver sections of CCl₄-treated Il15^–/– and Il15ra^–/– mice showed significantly reduced collagen deposition compared to CCl₄-treated wildtype mice, with collagen deposition in portal areas and occasional portal to portal bridging ( Figure 1C ). Quantification of the Sirius red-staining area showed significantly less fibrosis in CCl₄-treated Il15ra^–/– mice than in Il15^–/– mice ( Figure 1D ). Masson’s trichrome staining corroborated with the Sirius red staining pattern and intensity ( Figure 1E ). These results indicate that IL-15 plays a profibrogenic role in liver fibrosis pathogenesis, and this is mediated by IL-15Rα-dependent IL-15 signaling.

Next, we examined the expression of genes associated with hepatic fibrogenic response in the livers of CCl₄-treated and wildtype, Il15^–/– and Il15ra^–/– mice. CCl₄-treated wildtype mice livers showed increased expression of Col1a1 and Col1a3 genes coding for collagens ( Figure 2 ), corroborating with increased collagen deposition observed by Sirius Red and Masson’s trichrome staining ( Figures 1C–E ). Livers of CCl₄-treated Il15^–/– and Il15ra^–/– mice also showed upregulation of Col1a1 and Col1a3 genes that was comparable to wildtype mice ( Figure 2 ), although collagen deposition in IL-15 or IL-15Rα deficient mice was not increased to the extent observed in wildtype mice ( Figures 1C–E ). Notably, the induction of the pro-fibrogenic cytokine TGFβ (Tgfb) was increased in CCl₄-treated wildtype livers compared to the corresponding control livers, but not in the livers of Il15^–/– and Il15ra^–/– mice ( Figure 2 ). Induction of the antifibrotic matrix metalloproteinase 2 (Mmp2) (39), was increased in wildtype mice but not in Il15^–/– mice. Even though Mmp2 was induced in Il15ra^–/– mice, this increase was not statistically significant. Mmp14 was also induced only in wildtype mice, whereas Mmp9 was upregulated in Il15^–/– mice though not in Il15ra^–/– mice ( Figure 2 ). Wildtype mice also showed upregulation of tissue inhibitor of MMP gene Timp1, whereas Il15^–/– and Il15ra^–/– mice did not show significant change in Timp1 or Timp2 expression. Overall, livers of Il15^–/– and Il15ra^–/– mice exposed to CCl₄ upregulate collagen genes similarly to wildtype mice but show blunted induction of Tgfb and variable induction of matrix modifying enzyme genes and that the latter two may contribute to reduced liver fibrosis in IL-15 or IL-15Rα deficient mice compared to wildtype controls.

Examination of hematoxylin/eosin-stained liver sections revealed notable infiltration of mononuclear cells in CCl₄-treated livers of wildtype mice in periportal areas that was markedly reduced in Il15^–/– mice and further reduced in Il15ra^–/– mice ( Figure 3A ). When we assessed the distribution CD45⁺ cells that labels all the leukocytes, along with collagen I (COL1A1) expression in the liver sections, the livers of oil-treated control mice showed COL1A1 localized to the portal area with only a few CD45⁺ cells. The livers of CCl₄-treated wildtype mice showed increased COL1A1 deposition and marked infiltration of CD45⁺ cells that accumulated along the collagen fibers ( Figures 3B–D ). CCl₄-treated Il15^–/– mice showed significantly reduced COL1A1 deposition and CD45⁺ cell infiltration, confirming the profibrogenic role of IL-15, and suggesting the involvement of hepatic leukocytes in the pathogenic process. Notably, CCl₄-treated Il15ra^–/– mice showed even less staining of COL1A1 and CD45⁺ cell infiltration compared to Il15^–/– mice, clearly indicating that IL-15 signaling via IL-15Rα-dependent trans-presentation promotes liver fibrosis via promoting inflammatory cell recruitment. Moreover, the differences observed between IL-15-deficient and IL-15Rα-deficient mice suggest a possible antifibrogenic role for IL-15Rα-independent IL-15 signaling.

As myeloid cells play a key role in liver fibrosis (11) and IL-15Rα-independent IL-15 signaling impacts macrophage functions (34), we examined macrophage infiltration into the livers of CCl₄-treated Il15^–/– , Il15ra^–/– and wildtype mice. CCl₄-treated wildtype mice showed a significant increase in the transcript levels of Cd68, a marker for macrophages, that coincided with the upregulation of Ccl2 chemokine gene that codes for macrophage chemotactic protein-1 (MCP-1) ( Figure 4A ). CCl₄-treated Il15^–/– mice also showed an increase in Cd68 and Ccl2 expression, whereas these genes were not upregulated in CCl₄-treated Il15ra^–/– mice ( Figure 4A ). CCl₄-treated wildtype mice livers also upregulated Ccl5 and Cx3Cl1 genes, which encode the chemokines CCL5/RANTES and CX3CL1/Fractalkine, respectively, that are also chemotactic for monocytes and lymphocytes ( Figure 4A ). Neither Il15^–/– nor Il15ra^–/– mice upregulated Ccl5 and Cx3Cl1 genes following CCl₄ treatment. We next examined the staining of CD68⁺ macrophages in the livers of CCl₄-treated mice. Consistent with Cd68 gene expression, CCl₄-treated wildtype and Il15^–/– mice showed increased numbers of CD68⁺ cell infiltration into the liver that localized to the areas of α smooth muscle cell actin (αSMA)-staining myofibroblasts ( Figure 4B ). In contrast, Il15ra^–/– mice displayed markedly less αSMA staining and very few CD68⁺ cells in the liver. In the normal liver, CD68 is expressed in only a half of liver resident KCs, which are damaged during liver injury and are replenished by monocyte-derived macrophages that express CD68 (40–44). Hence, the increase in the number of CD68⁺ cells in the livers of CCl₄ treated mice injury results mainly from infiltrating macrophages. Quantitation of the αSMA and CD68 staining areas showed significantly less staining in Il15^–/– mice compared to wildtype mice that were further reduced in Il15ra^–/– mice ( Figures 4C, D ). These data further confirm the requirement of IL-15 signaling via IL-15Rα-dependent trans-presentation in promoting the accumulation of macrophages during hepatic fibrogenesis.

To characterize the immune cells that infiltrated the liver during CCl₄-induced liver fibrosis, we isolated intrahepatic leukocytes (IHL) and analyzed them by flow cytometry ( Figure 5A ). The proportion and total numbers of CD45⁺ IHLs was significantly higher in the CCl₄-treated wildtype mice when compared to oil-treated controls ( Figures 5A, B ). Even though the proportion of CD45⁺ IHLs increased in CCl₄-treated Il15^–/– mice, their absolute numbers did not increase ( Figures 5A, B ). CCl₄-treated Il15ra^–/– mice did not show an increase in proportion CD45⁺ IHLs but their total numbers were significantly higher than in the corresponding oil-treated controls. The absolute number of CD45⁺ IHLs recovered in CCl₄-treated Il15^–/– and Il15ra^–/– mice were comparable between them but were significantly lower than that of wildtype mice ( Figure 5B ), indicating that IL-15 signaling via IL-15Rα promotes leukocyte infiltration during hepatic fibrogenic response. As IL-15 signaling is required for the homeostasis of NK and CD8⁺ T cells (14, 15), we examined their numbers in Il15^–/– and Il15ra^–/– mice after fibrosis induction. The numbers of NK cells, CD3⁺ T cells and CD8⁺ T cells were not upregulated by CCl₄ treatment, but they were significantly reduced in the livers of both vehicle-treated and CCl₄-treated Il15^–/– and Il15ra^–/– mice when compared to the corresponding wildtype mice ( Figures 5A, B ). CCl₄ treatment increased B cell numbers in the liver, but this increase was significantly reduced in Il15ra^–/– mice ( Figure 5B ).

Phenotypic segregation of CD45⁺ IHL-myeloid cell subsets into neutrophils (CD11b⁺Ly6G⁺), eosinophils (CD11b⁺CD11c^–Ly6G^–Ly6C^Lo) and monocytes (CD11b⁺CD11c^–Ly6G^–) showed comparable frequencies of these myeloid cell subsets in the three genotypes within the non-treated and CCl₄-treated groups ( Figures 6A, B ). However, CCl₄-treated wildtype mice showed a significant increase in the absolute numbers of neutrophils, eosinophils and CD11b⁺ monocytes ( Figures 6C–E ). The increase in these cell numbers was minimal in Il15^–/– and Il15ra^–/– mice, and the number of monocytes in these mice were significantly lower than in wildtype mice after CCl₄ treatment. Even though the hepatic CD11b⁺CD11c^–Ly6G^– population includes both resident KCs and monocyte-derived macrophages, the increase in their numbers during CCl₄-induced liver fibrosis results mainly from infiltrating monocyte derived macrophages (11, 41). Next, we segregated CD11b⁺CD11c^–Ly6G^– macrophages into inflammatory (Ly6C^HiCCR2⁺) and anti-inflammatory (Ly6C^LoCCR2^-CX3CR1⁺) subsets ( Figure 7A ). In wildtype mice, the proportion of the inflammatory subset appear decreased following CCl₄ treatment, and the anti-inflammatory subset increased, although total numbers for both subsets showed a net increase ( Figures 7A–C ). Absolute numbers of pro-inflammatory monocytes did not increase in either Il15^–/– or Il15ra^–/– mice following CCl₄ treatment compared to vehicle-treated controls, whereas anti-inflammatory monocytes were increased in CCl₄-treated Il15^–/– mice but not Il15ra^–/– mice ( Figures 7B, C ). Most of the CCR2⁺ monocytes were also positive for CX3CR1 and expressed high levels of Ly6C ( Figure 7A , last row and Figure 7D ; Supplementary Figure S1 ). These Ly6C^hiCCR2⁺CX3CR1⁺ cells may represent pro-inflammatory cells transitioning to anti-inflammatory cells (11, 13). The number of these transitional cells increased in CCl₄-treated wildtype mice, but not in Il15^–/– or Il15ra^–/– mice ( Figure 7D ). Ly6C^Hi monocytes expressed invariably both CCR2 and CX3CR1, whereas Ly6C^Lo monocytes expressed variable levels of CCR2 and CX3CR1 ( Supplementary Figure S1 ). Collectively, these data indicate that IL-15 signaling via IL-15Rα promotes monocyte recruitment and their differentiation towards proinflammatory phenotype during hepatic fibrogenesis.