Identification of a Kupffer cell subset capable of reverting the T cell dysfunction induced by hepatocellular priming

1The liver is peculiarly biased toward inducing immune tolerance, as exemplified by the acceptance of liver allografts across complete major histocompatibility complex (MHC) mismatch barriers or the propensity of hepatitis B virus (HBV) and other hepatotropic viruses such hepatitis C virus (HCV) to establish life-long persistent infections. Liver tolerance involves a complex array of coordinated events that ultimately hinder the effector functions of intrahepatic lymphocytes. For example, the unique anatomy and hemodynamics of the fenestrated and basement membrane-less liver capillaries (i.e., sinusoids), through which about one-third of all blood cells transit slowly every minute, allow circulating, intravascular T cells to sense MHC-antigen (Ag) complexes displayed by the non-professional Ag-presenting hepatocytes. Using mouse models of HBV infection, it has been recently shown that hepatocellular priming of virus-specific naive CD8+ T cells induces local activation and initial vigorous proliferation but eventually leads to the development of dysfunctional cells devoid of cytotoxic and antiviral activity. The transcriptional signature of these cells does not obviously overlap with that of other known dysfunctional CD8+ T cell states such as exhaustion, and accordingly, CD8+ T cells primed by hepatocytes are not readily responsive to in vivo anti-PD-L1 treatment. The notion that in vivo interleukin-2 (IL-2) administration overcomes this dysfunction not only illustrates that efficient hepatocellular priming can occur under specific conditions but also provides the opportunity to identify which cellular and molecular determinants drive immunogenic responses within the tolerogenic liver microenvironment.

2Several formulations of IL-2 have been variably used in the past >25 years as a therapy to augment T cell responses against viral or tumor Ags, and at the moment of writing, more than 40 clinical trials are evaluating the immune stimulatory potential of this cytokine in different oncological indications (https://www.clinicaltrials.gov). As the IL-2 functional pleiotropy has often driven undesired toxicity in the clinical setting, deconvoluting the biology responsible for its efficacy may help improve the therapeutic potential of IL-2-based strategies.

3The IL-2 receptor consists of a heterocomplex of up to three subunits: α (CD25), β (CD122), and the common γ chain (CD132). Although each receptor subunit can independently bind IL-2 with low affinity (Kd ~ 10−8 to 10−7 M), only the intermediate-affinity βγ dimeric (Kd ~ 10−9 M) and the high-affinity αβγ trimeric (Kd ~ 10−11 M) receptors mediate intracellular signal transduction. In addition to T cells and natural killer (NK) cells, myeloid cells have been reported to express the intermediate-affinity βγ receptor, with some dendritic cell (DC) subtypes displaying the three subunits of the IL-2 receptor. However, the significance of IL-2 receptor expression by myeloid cells in vivo is controversial. For instance, DCs may supply the α chain in trans, thus supporting high-affinity binding of IL-2 to naive T cells undergoing priming. Other studies have suggested that DC production of CD25 quenches IL-2 in the outer T cell area of lymph nodes, thus guiding T cell differentiation. Whether the IL-2-mediated reversal of the T cell dysfunction induced by hepatocellular priming is due to an exclusive effect of this cytokine on T cells or whether myeloid cells are involved is currently unknown.

4To shed light on the immune mechanisms underpinning the IL-2-mediated reinvigoration of intrahepatically primed T cells, we initially took advantage of transgenic mice that express a non-secretable version of the particulate HBV core protein under the transcriptional control of the hepatocyte-specific mouse major urinary protein (MUP) promoter (hereafter MUP-core mice). These animals, like the HBV replication-competent transgenic mice described below, never develop spontaneous liver pathology, as the hepatocellular expression of the viral gene products occurs non-cytopathically, and endogenous T cells specific for these products are profoundly tolerant. As controls for proper CD8+ T cell differentiation into effector cells, we used wild-type (WT) mice transduced with recombinant, replication-defective lymphocytic choriomeningitis (LCMV)-based vectors targeting the HBV core and envelope proteins (rLCMV-core/env) to intrahepatic professional Ag-presenting cells (APCs) (i.e., Kupffer cells [KCs] and hepatic DCs) that are not natural targets of HBV. Both groups of mice were injected with naive CD8+ TCR transgenic T cells (TN) specific for epitopes contained within the core and envelope proteins of HBV (Cor93 and Env28 TN, respectively) (Figure 1A). One day after TN injection, selected MUP-core mice received IL-2 immune complexes (IL-2c) consisting of IL-2 coupled with non-neutralizing IL-2-specific monoclonal antibodies (S4B6) that enhance the half-life of IL-2 in vivo (Figure 1A). To test whether IL-2c treatment had exclusively a direct effect on TN or whether it required the presence of additional cells, we performed depletion experiments. We initially focused on KCs, as these cells are capable of inducing full effector differentiation of CD8+ T cells upon in vivo rLCMV transduction. KCs were depleted through clodronate liposome (CLL) injection 2 days prior to T cell injection (Figure 1A). This treatment effectively depletes KCs while sparing hepatic DCs (Figures 1B–1F). Consistent with previously published results, Cor93 and Env28 TN transferred to WT mice injected with rLCMV-core/env differentiated into bona fide effector cells that formed tight clusters scattered throughout the liver lobules; in contrast, Cor93 T cells transferred to MUP-core mice generated dysfunctional cells devoid of IFN-γ-producing ability that coalesced around portal tracts (Figures 1G–1I). IL-2c administration improved the capacity of Ag-specific Cor93 T cells to expand, differentiate into IFN-γ-producing cells and accumulate in clusters scattered throughout the liver lobules, but it had no effect on irrelevant Env28 TN (Figures 1G–1I). Optimal in vivo reinvigoration of intrahepatically primed Cor93 T cells required the presence of KCs, as IL-2c treatment failed to improve T cell expansion, effector differentiation, and intraparenchymal cluster formation in CLL-treated mice (Figures 1G–1I). Similar results were obtained when recombinant IL-2 was used in place of IL-2c and when HBV replication-competent transgenic mice, which express all viral proteins in hepatocytes and secrete enveloped virions containing the HBV particulate core protein into the bloodstream, were used in place of MUP-core recipients (data not shown).

5To confirm that hepatic DCs are not necessary for the optimal in vivo response to IL-2, we depleted this cell population by diphtheria toxin (DT) injection in MUP-core mice reconstituted with CD11cDTR bone marrow (Figure 1J). This treatment significantly decreased the number of hepatic DCs while sparing KCs (Figures 1K–1O). DC depletion did not affect the capacity of IL-2 to promote expansion, effector differentiation, and intraparenchymal cluster accumulation of intrahepatically primed Cor93 T cells (Figures 1P–1R). Similarly, other phagocytic cells such as neutrophils and monocytes were found not to be involved in the response to IL-2, as neutrophil depletion (via anti-Ly6G Abs) or combined neutrophil and monocyte depletion (via anti-Gr1 Abs) did not affect the in vivo reinvigoration of intrahepatically primed T cells by IL-2 (Figure S1). Taken together, these results indicate that KCs are required for optimal in vivo reinvigoration of intrahepatically primed T cells by IL-2.

6Flow cytometric analyses revealed that a fraction of KCs expresses all three subunits of the IL-2 receptor (CD25, CD122, and CD132) (Figures 2A and 2B). We therefore investigated the effect of IL-2 treatment on KCs. To this end, we isolated liver non-parenchymal cells (LNPCs), including KCs, from C57BL/6 mice and stimulated them ex vivo with recombinant IL-2 (Figure 2C). We observed a dose-dependent increase in STAT5 phosphorylation in KCs but not in liver sinusoidal endothelial cells (LSECs) (Figure 2D). Similar results were obtained when IL-2c was used in place of IL-2, and STAT5 phosphorylation in KCs was confirmed by immunoblot analysis (Figure 2E). Of note, the IL-2-dependent fold change in STAT5 phosphorylation observed in KCs was ~10% than that observed in CD4+FoxP3+ splenic T regulatory cells (data not shown). Nevertheless, these data indicate that KCs express a functional IL-2 receptor capable of responding to IL-2 in vitro. To assess the consequences of IL-2 treatment on KCs in vivo, we treated C57BL/6 mice with IL-2c and then performed RNA sequencing (RNA-seq) analysis on flow cytometry-sorted KCs 48 h later (Figures 2F and 2G). A total of 4,073 differentially expressed genes (DEGs), 1,515 up- and 2,558 downregulated, were identified as significantly regulated by IL-2c (Table S1). Functional enrichment analysis of upregulated genes showed an increased transcription of genes involved mainly in Ag presentation and proteasomal processing, ribosomal RNA processing and splicing, DNA replication and cell cycle, and mitochondrial oxidative metabolism (Figure 2H; Figure S2; Table S1). Among the upregulated gene clusters, we focused on the Ag presentation pathway, which includes several macromolecular complexes composed of ubiquitins, chaperones, MHC-I, and proteasome subunits (Figures 2I–2K; Figures S3A–S3E). Genes encoding for these protein families—specifically MHC-I-related proteins, immunoproteasome subunits, the transcription regulator of MHC-I genes Nlrc5 and the transporter associated with Ag processing 1 (Tap1)—were induced in KCs upon IL-2c treatment (Figures 2I–2K; Figures S3B–S3F). The upregulation of MHC-I and co-stimulatory molecules in KCs isolated from mice treated with IL-2c was confirmed at the protein level (Figure 2L). On the basis of these results, we reasoned that in vivo treatment with IL-2c might increase the cross-presentation ability of KCs. To test this possibility, we measured the capacity of in vitro differentiated Cor93-specific effector CD8+ T cells (Cor93 TEFF) to produce IFN-γ (as an indirect measure of Ag recognition) upon incubation with KCs isolated from control and IL-2c-treated HBV replication-competent transgenic mice (Figure 2M). Consistently with previously published data, baseline KC cross-presentation of the core protein in this experimental system at steady state was negligible (Figures 2N and 2O), despite KCs being constantly exposed to abundant HBV virions in the circulation. Cor93 TN remained dysfunctional even when isolated from the liver of HBV replication-competent transgenic mice previously transferred with highly pathogenic Env28-specific effector CD8+ T cells (data not shown). This indicates that KC cross-presentation remains insignificant during acute liver inflammation, even though the inflammatory conditions potentially favor not only the uptake of HBV virions but also the phagocytosis of damaged hepatocytes containing the particulate HBV core protein. In spite of this, treating HBV replication-competent transgenic mice with IL-2c slightly but significantly increased the cross-presentation capacity of KCs incubated in vitro with Cor93 TEFF cells (Figures 2N and 2O). We also assessed the ability of KCs isolated from IL-2-treated C57BL/6 mice to cross-prime HBV-specific naive CD8+ T cells exposed to the serum of HBV replication-competent transgenic mice in vitro (Figure 2P). Compared with KCs isolated from PBS-treated mice, KCs exposed to IL-2 in vivo induced a higher proliferation of Cor93 TN in in vitro culture (Figures 2Q and 2R). Finally, to evaluate the in vivo relevance of our findings, we took advantage of MUP-core mice, which express only a non-secretable, particulate form of the HBV core protein and in which KC cross-presentation should depend on the uptake of the few hepatocytes that are known to be injured by Cor93 TN transfer. We generated MUP-core mice whose hematopoietic cells (including KCs) lack Tap1 and therefore cannot express MHC-I and present Ags to CD8+ T cells. This was achieved by injection of either WT or Tap1−/− bone marrow into irradiated MUP-core mice, followed by CLL treatment to deplete the residual radio-resistant KCs and allow the complete reconstitution of the entire KC compartment with bone marrow-derived cells (Figure 2S). Cor93 TN injected into MUP-core mice whose hematopoietic cells (including KCs) lacked MHC-I had a much lower response to IL-2c than did Cor93 TN injected into mice carrying Ag presentation-competent KCs (Figures 2T and 2U), suggesting that Cor93 T cells interacted with IL-2-stimulated KCs that cross-presented core protein-derived epitopes after the uptake of damaged hepatocytes. Taken together, these results indicate that optimal reinvigoration of intrahepatically primed CD8+ T cells by IL-2 requires the capacity of KCs to cross-present HBV Ags, possibly derived from circulating virions and/or damaged hepatocytes.

7Next, we asked whether the IL-2-responsive KCs represent a distinct subpopulation. To this end, we used high-dimensional single-cell RNA-seq (scRNA-seq) to characterize KC heterogeneity at steady state. We flow cytometry-sorted live CD45+ Lineage− CD64+ F4/80+ liver macrophages from C57BL/6 mice (Figure 3A), isolated RNA, and generated transcriptional profiles for each cell (n = 169) using the Smart-seq2 pipeline. This dataset was analyzed using Seurat, and four main cell clusters were identified and visualized using uniform manifold approximation and projection (UMAP) (Figure 3B). Cluster 0 (n = 68) and cluster 1 (n = 59) cells showed higher expression of classical KC-associated gene markers, such as Clec4f, Lyz2, and Csf1r (Figures 3C and 3D; Table S2). Pathway analysis of their respective gene markers yielded immunological pathways and processes typical of macrophages and professional APCs and were thus considered bona fide KCs (Figures 3E and 3F; Table S3). Cells in cluster 2 (n = 30) expressed genes such as Cd34, Cd209c, and Fgd4 but low amounts of macrophage genes (Figures 3C and 3D), while among their specific markers we found a large number of ribosomal and non-coding genes. They also showed a smaller number of transcripts detected per cell and a higher percentage of mitochondrial genes, indicating a high fraction of apoptotic cells in this population, and hence were excluded from subsequent analyses (Figure 3E and data not shown). Cells in cluster 3 (n = 12) expressed genes associated with endothelial cells, including Pecam1 (CD31), Clec4g, Lyve1 (Figures 3C and 3D), and Kdr (VGFR2) (Table S2); in addition, their specific markers were enriched in endothelial cell processes (Figure 3F; Table S3), arguing for contamination of sorted cells with LSECs (Figure 3E). Although both cluster 0 and cluster 1 showed expression of KC markers, they were clearly distinguished by the expression of many genes (Figure 3C). Of note, compared with cells in cluster 0, we found that cells in cluster 1 were enriched in genes associated with Ag processing, cross-presentation, and IL-2 signaling pathway (Figure 3F; Tables S2 and S3). Among the DEGs, we initially used Mrc1 (CD206) and Lamp2 (CD107b) (Figures S4A and S4B; Table S2) as a first approach to identify and flow cytometry-sort the two KC populations. An ad interim bulk RNA-seq analysis of the two populations revealed Esam (ESAM) as highly differentially expressed (Figure S4C).

8On the basis of these data, we designed a panel of markers for use in conventional flow cytometry to identify these KC subpopulations and validate the aforementioned high-throughput approach. The CD45+ F4/80+ CD11bint TIM-4+ KC population split into CD206−ESAM− (hereafter KC1; ~70%–85% of total KCs) and CD206+ESAM+ (hereafter KC2; ~15%–30% of total KCs) cells (Figures 4A and 4B). Imaging analyses confirmed the presence of two distinct KC subpopulations (Figure 4C; Video S1). Importantly, RNA-seq analyses on bulk KC1 and KC2 sorted from C57BL/6 mice confirmed that KC2 are enriched in IL-2 signaling components (IL-2 receptor subunits and molecules implicated in intracellular signal transduction) (Figures 4D and 4E; Table S4). Higher expression of the IL-2 receptor subunits, MHC-I, and co-stimulatory molecules in KC2 was confirmed at the protein level using flow cytometry analysis (Figures 4F–4J). Together, the data suggest that KC2 are better equipped than KC1 to respond to IL-2 and increase their capacity to cross-present hepatocellular Ags. Thus, one might predict that IL-2 treatment might render KC2 more sensitive than KC1 to CD8+ T cell-mediated killing. To test this hypothesis, we treated HBV replication-competent transgenic mice with IL-2c 24 h after Cor93 TN cell injection and checked the KC1/KC2 ratio 4 days later (Figure 4K). Consistent with the hypothesis that IL-2 preferentially increased the capacity of KC2 to cross-present hepatocellular Ags and thus rendered them more sensitive to CD8+ T cell-mediated killing, we found that KC2 almost completely disappeared in Cor93 T cell-injected HBV transgenic mice treated with IL-2c (Figures 4L–4N). Notably, neither IL-2c treatment alone (in the absence of Cor93 TN cell transfer) nor severe liver inflammation (induced by Cor93 TEFF) altered the KC1/KC2 ratio (Figure S5).

9We next sought to generate a model in which KC2 could be selectively depleted to assess their role in the cross-presentation of hepatocellular Ags upon in vivo IL-2 treatment. We took advantage of the observation that KC2 (but not KC1) express the endothelial cell marker VE-cadherin (encoded by Cdh5) (Figures 5A–5D) to establish a system allowing inducible depletion of KC2 but not endothelial cells. This was achieved by (1) injecting Cdh5creERT2; Rosa26iDTR bone marrow into irradiated MUP-core mice, (2) depleting the residual radio-resistant KCs by CLL to allow the complete reconstitution of the entire KC compartment with bone marrow-derived cells, (3) inducing DTR expression in KC2 by tamoxifen administration, and finally (4) depleting KC2 by DT injection prior to Cor93 TN transfer followed by IL-2c treatment (Figure 5E). DT treatment caused a ~75% decrease in KC2 (Figures 5G and 5H) and resulted in a lower ability of Cor93 T cells to proliferate and differentiate into cytotoxic effector cells clustered throughout the liver lobule in response to IL-2c (Figures 5I–5L). These data indicate that KC2 are required for the optimal reinvigoration of intrahepatically primed T cells by IL-2.

10Here, we have delineated the mechanisms by which hepatocellularly primed HBV-specific CD8+ T cells acquire antiviral and pathogenic effector functions following the exogenous administration of IL-2. These mechanisms rely on KCs and, in particular, on a hitherto unidentified subset of KCs, referred to as KC2, that is poised to respond to IL-2 and cross-present viral Ags contained within circulating virions or within hepatocytes.

11The observation that DCs, generally regarded as the main cross-presenting APCs in vivo, are dispensable for the optimal reinvigoration of intrahepatically primed T cells by IL-2 remains to be explored but is consistent with the lack of a functional IL-2 receptor expression by DCs.

12The results reported here are noteworthy considering that steady-state KC cross-presentation of HBV Ags is an inefficient process that cannot be increased by liver inflammation, hepatocellular death, or the administration of therapeutic monoclonal antibodies directed against HBsAg leading to the generation of circulating immune complexes. Our data do not rule out a direct effect of IL-2 on T cells; however, they indicate that optimal in vivo reinvigoration of intrahepatically primed T cells by IL-2 depends on the presence of KC2. The extent to which the reported increase in cross-presentation of hepatocellular Ags by KCs in response to in vivo IL-2 administration is the result of direct IL-2 sensing by KCs or an indirect effect remains to be determined.

13The location of CD8+ T cell priming during natural HBV infection is unclear. Immunological dogma holds that naive CD8+ T cells initially encounter cognate Ag in secondary lymphoid organs where cross-priming by professional APCs of circulating virions and subviral particles promotes the differentiation into effector cells endowed with liver homing potential. Yet there is no direct experimental evidence indicating that secondary lymphoid organs are the priming site for CD8+ T cells in HBV-infected humans or chimpanzees, and the idea behind their involvement has been supported by the assumption that the intrahepatic priming of naive CD8+ T cells (especially in the presence of high Ag load) should promote functional impairment of these cells. On the basis of these considerations, it is generally believed that priming of HBV-specific naive CD8+ T cells in the liver cannot promote viral clearance but, rather, contribute to the establishment of a persistent infection. However, on the basis of the results presented here, it is tempting to speculate that during acute HBV infection, the liver is fully competent to sustain the priming of HBV-specific effector CD8+ T cells endowed with antiviral activity, provided it occurs in the presence of high local concentrations of IL-2 that increase KC2 cross-presentation. A putative source of IL-2 in this scenario might be intrahepatic Ag-specific effector CD4+ T cells. In support of this hypothesis, which cannot be tested in humans because of the limitations of collecting liver biopsies in acutely infected patients, are the observations that (1) CD4+ T cell depletion in chimpanzees prior to infection precludes effective T cell priming and causes persistent infection with minimal immunopathology, and (2) detection of CD8+ T cells in the liver of HBV-infected chimpanzees coincides with the hepatic detection of CD4+ T cells.

14In parallel to the present study, we have used high-dimensional single-cell sequencing, mass cytometry, and flow cytometry, coupled with in vivo fate-mapping models to perform an in-depth characterization of KC2 at steady state (, this issue of Immunity). These analyses have revealed a specific metabolic role for KC2 in regulating glucose homeostasis and oxidative stress. Future studies should be directed at identifying the signals required for KC2 development and maintenance and explore the potential role of these cells in other diseases affecting the liver.

15We envision that strategies aimed at targeting IL-2 to KC2 should be considered for the treatment of chronic HBV infection and for other conditions similarly requiring to overcome the tolerogenic potential of the hepatic microenvironment. Such strategies might include liposome or nanoparticle-based formulations targeting KC2-expressed surface Ags as well as integrase-defective third-generation lentiviral vectors exploiting combinations of transcriptional and post-transcriptional microRNA-mediated control.

16The therapeutic implications of our study rely on the existence of an IL-2-responsive, cross-presenting KC population in humans. Of note, recent publications suggest the existence of a human KC2-like subset that expresses at least some of the human orthologs of the KC2-specific genes identified here. Future studies are certainly warranted to dissect human KC heterogeneity and function in the healthy as well as in the diseased liver.

17Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Matteo Iannacone (iannacone.matteo@hsr.it).

18This study did not generate new unique reagents.

19C57BL/6, CD45.1 (inbred C57BL/6), BALB/c, Thy1.1 (CBy.PL(B6)-Thya/ScrJ), β-actin-GFP [C57BL/6-Tg(CAG-EGFP)1Osb/J], Ai14(RCL-tdT)-D [B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J ], β-actin-DsRed [B6.Cg-Tg(CAG-DsRed∗MST)1Nagy/J], Tap1−/− (B6.129S2-Tap1tm1Arp/J), CD11cDTR [B6.FVB-1700016L2RikTg(Itgax-DTR/EGFP)57Lan/J], ROSA26iDTR [C57BL/6-Gt(ROSA)26Sortm1(HBEGF)Awai/J], Cdh5CreERT2 [Tg(Cdh5-cre/ERT2)1Rha] mice were purchased from Charles River or The Jackson Laboratory. MUP-core transgenic mice (lineage MUP-core 50 [MC50], inbred C57BL/6, H-2b), that express the HBV core protein in 100% of the hepatocytes under the transcriptional control of the mouse major urinary protein (MUP) promoter, have been previously described. HBV replication-competent transgenic mice (lineage 1.3.32, inbred C57BL/6, H-2b), that express all of the HBV Ags and replicate HBV in the liver at high viral copies without any evidence of cytopathology, have been previously described. In indicated experiments, MUP-core and HBV replication-competent transgenic mice were used as C57BL/6 x BALB/c H-2bxd F1 hybrids. Cor93 TCR transgenic mice (lineage BC10.3, inbred CD45.1), in which > 98% of the splenic CD8+ T cells recognize a Kb-restricted epitope located between residues 93-100 in the HBV core protein (MGLKFRQL), have been previously described. Env28 TCR transgenic mice (lineage 6C2.36, inbred Thy1.1 BALB/c), in which ~83% of the splenic CD8+ T cells recognize a Ld-restricted epitope located between residues 28–39 of HBsAg (IPQSLDSWWTSL), have been previously described. For imaging experiments Cor93 transgenic mice were bred against β-actin-GFP, while Env28 transgenic mice were bred against β-actin-DsRed mice (inbred BALB/c). Bone marrow (BM) chimeras were generated by irradiation of MUP-core or C57BL/6 mice with one dose of 900 rad and reconstitution with the indicated BM; mice were allowed to reconstitute for at least 8 weeks before experimental manipulations. Mice were housed under specific pathogen-free conditions and entered experiments at 8-10 weeks of age. In all experiments, mice were matched for age, sex and (for the 1.3.32 animals) serum HBeAg concentration before experimental manipulations. All experimental animal procedures were approved by the Institutional Animal Committee of the San Raffaele Scientific Institute and are compliant with all relevant ethical regulations.

20Replication-incompetent LCMV-based vectors encoding HBV core and envelope proteins (rLCMV-core/env) were generated, grown and titrated as previously described. Mice were injected intravenously (i.v.) with 2.5 × 105 infectious units of rLCMV vector 4h before CD8+ T cell injection. All infectious work was performed in designated BSL-2 or BSL-3 workspaces, in accordance with institutional guidelines.

21Mice were adoptively transferred with 5 × 106 or 1 × 106 HBV-specific naive CD8+ TCR transgenic T cells isolated from the spleens of Cor93 and/or Env28 TCR transgenic mice, as described. IL-2/anti-IL-2 complexes (IL-2c) were prepared by incubating 1.5 μg of rIL-2 (R&D Systems) with 50 μg anti-IL-2 mAb (clone S4B6-1, BioXcell) per mouse, as previously described. Mice were injected with IL-2c intraperitoneally (i.p.) one day after T cell transfer, unless otherwise indicated. In indicated experiments, naive CD8+ T cells from the spleens of Cor93 TCR transgenic mice were differentiated in vitro for 7-9 days into effector cells prior to adoptive transfer (1 × 107 cells), or in vitro co-culture, as described. In indicated experiments, Kupffer cells (KCs) were depleted by intravenous injection of 200 μL of clodronate-containing liposomes (Liposoma) 2 days prior to T cell injection, as described, unless otherwise indicated. In indicated experiments, mice were injected i.p. with 200 μg of anti-Ly6G depleting antibody (clone 1A8, BioXcell) one day before and one day after T cell transfer. In indicated experiments, mice were injected intravenously (i.v.) with 200 μg of anti-Gr1 depleting antibody (clone RB6-8C5, BioXcell) every 48h starting from 3 days before T cell transfer. In indicated experiments, C57BL/6 or MUP-core mice were lethally irradiated and reconstituted for at least 8 weeks with BM from CD11c-DTR mice; dendritic cells were subsequently depleted by injecting i.p. 20 ng per gram of mouse of diphtheria toxin (Millipore) every 48h starting from 3 days before T cell transfer. In indicated experiments, MUP-core mice were lethally irradiated and reconstituted for at least 8 weeks with BM from C57BL/6 or Tap1−/− mice. To achieve full reconstitution of Kupffer cells from donor-derived BM, mice were injected with 200 μL of clodronate-containing liposomes 28 and 31 days after BM injection. In indicated experiments, MUP-core mice were lethally irradiated and reconstituted for at least 8 weeks with BM from Cdh5CreERT2; Rosa26iDTR; Rosa26tdTomato; CX3CR1GFP mice. To achieve full reconstitution of Kupffer cells from donor-derived BM, mice were injected with 200 μL of clodronate-containing liposomes 28 and 31 days after BM injection. To induce the expression of the Cre recombinase, mice were treated with 5 mg of Tamoxifen (Sigma) by oral gavage in 200 μL of corn oil one week before further manipulations. KC2 were depleted subsequently by injecting i.p. 20 ng per gram of mouse of diphtheria toxin (Millipore) 3 days and 1 day prior to T cell transfer.

22Single-cell suspensions of liver, spleen and blood were generated as described. Kupffer cell isolation was performed as described. All flow cytometry stainings of surface-expressed and intracellular molecules were performed as described. Cell viability was assessed by staining with Viobility 405/520 fixable dye (Miltenyi, #130-109-816), LIVE/DEAD Fixable Far-Red dye (Invitrogen, # L34973) or DAPI (Invitrogen, # D1306). Abs used included: anti-CD3 (clone: 145-2C11, Cat#562286, BD Biosciences), anti-CD4 (clone: RM4-5, Cat #48-0042-82, eBioscience), anti-CD8a (clone: 53-6.7, Cat# 558106, BD Biosciences), anti-CD11b (clone: M1/70, Cat#101239), anti-CD19 (clone: 1D3, Cat#562291 BD Biosciences), anti-CD25 (clone: PC61, Cat#102015), anti-CD31 (clone: 390, Cat#102427), anti-CD45 (clone: 30-F11, Cat#564279 BD Biosciences), anti-CD64 (clone: X54-5/7.1, Cat#139311), anti-F4/80 (clone: BM8, Cat#123117), anti-I-A/I-E (clone: M5/114.15.2, Cat#107622), anti-TIM4 (clone: RTM4-54 Cat#130010), anti-TIM4 (polyclonal, Cat#orb103599 Biorbyt), anti-CD69 (clone: H1.2F3, Cat# 104517), anti-CD45.1 (clone: A20, Cat#110716), anti-IFN-γ (clone: XMG1.2, Cat# 557735 BD Biosciences), anti-CD11c (clone: N418, Cat# 117308), anti-I-Ab (clone: AF6-120.1, Cat# 116420), anti-Stat5 pY694 (clone: 47, Cat# 612599 BD Biosciences), anti-Foxp3 (clone FJK-16 s, Cat#25-5773-82 Thermofisher), anti-CD122 (clone TM-B1 Cat#123210), anti-CD132 (clone TUgm2 Cat#132306), anti-CD40 (clone 3/23 Cat#558695 BD Biosciences), anti-CD80 (clone 1610A1 Cat#553769 BD Biosciences), anti-H2-Kb (clone AF6-88.5 Cat#742861 BD Biosciences), anti-ESAM (clone 1G8/ESAM, Cat#136203), anti-CD206 (clone C068C2, Cat#141712), anti-Ly6G (clone 1A8, Cat #562700 BD Biosciences), anti-Ly6C (clone HK1.4, Cat# 128008), anti-CD49b (clone DX5, Cat#562453 BD Biosciences), anti CD107b (clone M3/84, Cat #12-5989-82 eBioscience). All Abs were purchased from BioLegend, unless otherwise indicated. Recombinant dimeric H-2Ld:Ig and H-2Kb:Ig fusion proteins (BD Biosciences) complexed with peptides derived from HBsAg (Env28-39) or from HBcAg (Cor93-100), respectively, were prepared according to the manufacturer’s instructions. Dimer staining was performed as described. Flow cytometry staining for phosphorylated STAT5 was performed using Phosflow Perm Buffer III (BD Bioscience), following the manufacturer’s instructions. Flow cytometry staining for Foxp3 was performed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience), following the manufacturer’s instructions. In indicated experiments, cells were stained with CellTrace™ Violet cell proliferation kit (CTV, Invitrogen), following manufacturer’s instructions. All flow cytometry analyses were performed in FACS buffer containing PBS with 2 mM EDTA and 2% FBS on a FACS CANTO II (BD Bioscience) or CytoFLEX LX (Beckman Coulter) and analyzed with FlowJo software (Treestar).

23For the experiment described in Figure 2, KCs were sorted from liver non-parenchymal cells as live, lineage negative (CD3-, CD19-, Ly6G-, CD49b-), CD45+, CD11bint, F4/80+, CD64+, MHCII+, TIM4+ cells. For the experiment described in Figure 3, single cells were sorted from liver non-parenchymal cells as live, CD45+, lineage negative (CD3-, CD19-, Ly6G-, CD49b-), F4/80+, CD64+ cells. For the experiment described in Figure 4, KCs were sorted from liver non-parenchymal cells as live, CD45+, CD11bint, F4/80+, MHCII+, TIM4+ cells. Among total KCs, KC1 were sorted as CD206- ESAM- cells and KC2 as CD206+, ESAM+ cells. Total KCs, KC1 and KC2 were flow cytometry-sorted with a 100 μm nozzle at 4°C on a FACSAria Fusion (BD) cell sorter in a buffer containing PBS with 2% FBS. Cells were always at least 98% pure (data not shown). In indicated experiments, F4/80+ cells were purified from liver non-parenchymal cells by positive immunomagnetic separation (Miltenyi Biotec, #130-110-443), according to the manufacturer’s instructions. In indicated experiments, CD8+ T cells were purified from splenocytes using EasySep™ kit (StemCell # 19858), according to the manufacturer’s instructions.

24Single cells were sorted on a 96-well plate and cDNA libraries were generated using the Smart-seq v2 protocol with the following modifications: i) 1mg/ml BSA Lysis buffer (Ambion® Thermo Fisher Scientific, Waltham, MA, USA); ii) use of 200 pg cDNA with 1/5 reaction of Illumina Nextera XT kit (Illumina, San Diego, CA, USA). The length distribution of the cDNA libraries was monitored using a DNA High Sensitivity Reagent Kit on the Perkin Elmer Labchip (Perkin Elmer, Waltham, MA, USA). All samples were subjected to an indexed paired-end sequencing run of 2x151 cycles on an Illumina HiSeq 4000 system (Illumina, San Diego, CA, USA) (298 samples/lane). The RSEM tool was used to perform Transcript Per Million (TPM) normalization starting from FASTQ files.

25Single cell data analysis was performed using Seurat (v3.2.2). 169 cells were obtained after applying a filter to the TMP matrix of at least 200 genes expressed per cell and only genes expressed in at least 3 cells were retained. TPM expression was further normalized and scaled using the SCTransform function, and Umap reduction was then applied on first 12 Principal Components after running PCA. Unbiased clustering was made using the FindClusters function in Seurat with default parameters and a resolution value of 1. Specific markers for the different unbiased clusters were found using the function FindAllmarkers or FindMarkers in Seurat with default parameters and were then used for functional enrichment analysis with the online tool Enrichr.

26The plots showing normalized expression values with a color scale on top of Umap plots (on Figure S4) and the Violin plots of specific genes were produced with FeaturePlot and VlnPlot Seurat functions, respectively.

27Bulk RNA-seq on CD206-CD107b- and CD206+CD107b+ cells (shown in Figure S4): between 20,000 and 50,000 cells were flow cytometry-sorted using CD206 (Mrc1) and CD107b (Lamp2) to identify CD206-, CD107b- and CD206+, CD107b+ cells. Total RNA was extracted using Arcturus PicoPure. RNA Isolation kit (Arcturus. Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer’s protocol. All Mouse RNAs were analyzed on Agilent Bioanalyser (Agilent, Santa Clara, CA, USA) for quality assessment with RNA Integrity Number (RIN) range from 5.8 to 6.7 and median of RIN 6.4. cDNA libraries were prepared using 2 ng of total RNA and 1ul of a 1:50,000 dilution of ERCC RNA Spike in Controls (Ambion. Thermo Fisher Scientific, Waltham, MA, USA) using the Smart-Seq v2 protocol with the following modifications: i) addition of 20 μM TSO; ii) use of 200 pg cDNA with 1/5 reaction of Illumina Nextera XT kit (Illumina, San Diego, CA, USA). The length distribution of the cDNA libraries was monitored using a DNA High Sensitivity Reagent Kit on the Perkin Elmer Labchip (Perkin Elmer, Waltham, MA, USA). All samples were subjected to an indexed paired-end sequencing run of 2x151 cycles on an Illumina HiSeq 4000 system (Illumina) (25 samples/lane)

28Bulk RNA-seq experiment on total KCs (shown in Figure 2) and bulk RNA-seq experiment on sorted KC1 and KC2 (shown in Figure 4): flow cytometry-sorted KCs, KC1 and KC2 were lysed in ReliaPrep RNA Cell Miniprep System (Promega #Z6011) and total RNA was isolated following manual extraction. DNA digestion was performed with TURBO DNA-free Kit (Invitrogen #AM1907). RNA was quantified with Qubit RNA HS Assay Kit (Invitrogen # Q32852) and analysis of its integrity was assessed with Agilent RNA 6000 Pico Kit (Agilent #5067-1513) on a Bioanalyser instrument. 6 RNA samples of sorted KC1 and KC2, were processed with the “SMART-seq Ultra Low Input 48” library protocol in order to obtain 30.0M clusters of fragments of 1x100nt of length through NovaSeq 6000 SP Reagent Kit (100 cycles).

29Bulk RNA-seq experiment on CD206-, CD107b- and CD206+, CD107b+ cells (shown in Figure S4): raw reads were obtained and mapped to the mouse genome build GRCm38. Gene counts were generated using featureCounts (part of the R subread package) with GENCODE version M9 annotations. Differential Expression Analysis genes (DEGs) and MA plots were performed using the R package edgeR.

30Bulk RNA-seq experiment on total KCs (shown in Figure 2) and on sorted KC1 and KC2 (shown in Figure 4): raw reads were aligned to mouse genome build GRCm38 using STAR aligner. Read counts per gene were then calculated using featureCounts based on GENCODE gene annotation version M16. Read counts were subject to log2 TPM (transcript per million) normalization to account for transcript length and library size.

31Only genes with a TPM value higher than 1 in at least 4 (for the total KC experiment in Figure 2) or 3 (for the KC2 versus KC1) samples were considered for following analysis. Differentially Expressed Genes (DEGs) between groups treated with IL-2c and PBS were identified by generating a linear model using LIMMA R package. Only DEGs with an adjusted P value < 0.05 (using Benjamini Hochberg correction method) were selected for further analysis. For the final KC2 versus KC1 comparison an additional |logFC| > 1 filter was applied.

32Bulk RNA-seq analysis of the experiment described in Figure 2: of the 4073 significant (FDR < 0.05) identified DEGs between control (PBS) and treated (IL-2c) samples, 1515 were upregulated and 2558 were downregulated. Those were subject to a functional enrichment analysis using the EnrichR R package. Both the up- and the downregulated DEGs were checked for any biological signature enrichment in both the Gene Ontology Biological Process Database and the Kyoto Encyclopedia of Genes and Genomes for Mouse. After merging the results for the two databases, 858 significant (FDR < 0.05) Terms were identified, of which 428 were derived from the upregulated DEGs and 430 from the downregulated ones. In order to select the top enriched terms, only those with a high Combined Score (-log(p value) ∗ Odds Ratio) were considered. Based on the distribution of the Combined Score in the upregulated terms and in the downregulated ones, a threshold of 100 was chosen for the former, while a threshold of 30 for the latter.

33For visualization and analysis, both upregulated and downregulated terms were subject to a clustering algorithm, in order to identify the most prominent biological signatures. Briefly, a Jaccard Index Similarity score was calculated for each pair set of terms, based on the DEGs annotated for each term, using an in-house developed script. Next, terms were clustered using a hierarchical clustering method, using as distance measure the Pearson correlation between the calculated Jaccard Index Similarity scores. An arbitrary number of clusters was selected and manually annotated based on the terms present. To visualize the result, the pheatmap R package was used.

34Radar plots were generated using the fmsb R package. Different sets of genes were selected based on literature analysis, defining different biological processes. For each category, the mean TPM expression for each gene within samples (separately for control and treated samples) was calculated. Next, the mean between all the genes belonging to a category was calculated and used as the value to represent the dimension in the radar plot.

35Network plot (Figure S3F) was built using Cytoscape software (V 3.8.0 for MacOS). Briefly, starting from EnrichR tables (Table S1), a matrix defining every pair of term-gene was generated, and used as a node list input for Cytoscape.

36Gene Set Enrichment Analysis (GSEA) from bulk RNA-seq of KC1 and KC2 (Figure 4) was performed using the GseaPreranked Java tool using pre-ranked Log2 fold changes between KC2 and KC1 populations in expressed genes. HALLMARK_IL2_STAT5_SIGNALING Gene Set contained in MsigDB (Broad Institute), Version 6. Since the gene set is based on human genes, mouse orthologs in humans were identified using the homologene R package (https://cran.r-project.org/web/packages/homologene/index.html).

37Immunoblot on plated KCs was performed as described. Primary Abs include anti-STAT5 and anti-pSTAT5 (Tyr694) (rabbit; Cell Signaling Technology #8215) and β-actin (polyclonal; Abcam ab228001). As secondary Ab horseradish peroxidase-conjugated goat anti-rabbit IgG (Jackson ImmunoResearch, Cat# 111-035-003) was used. Reactive proteins were visualized using a Clarity Western ECL substrate kit (Bio-Rad), and exposure was performed using UVItec (Cambridge MINI HD, Eppendorf). Images were acquired by NineAlliance software.

38Confocal microscopy analysis of livers was performed as described. For confocal images of KC1 and KC2, C57BL/6 mice were injected i.v. with 2 μg of anti-F4/80 Alexa Fluor 488 (BioLegend, #123120) and 2 μg of anti-CD206 APC (BioLegend, #141708) 10 minutes before harvesting the liver. The liver was fixed overnight in PBS with 4% paraformaldehyde and subsequently incubated for 24h in PBS with 30% sucrose. Next, liver lobes were embedded in O.C.T (Killik Bio-Optica) and cut at −14°C into 60 μm thick sections with a cryostat. Sections were blocked for 15 min with blocking buffer (PBS, 0.5% BSA, 0.3 % Triton) and stained for 1h at room temperature (RT) with anti-CD38 Alexa Fluor 594 (BioLegend, #102725) in wash/stain buffer (PBS, 0.2% BSA, 0.1% triton). Sections were then washed twice for 5 min, stained with DAPI (Sigma) for 5 min, washed again and mounted for imaging with FluorSave Reagent (Millipore). For additional confocal imaging, the following primary Abs were used for staining: anti-CD45.1 AF647 (110720, BioLegend), anti-F4/80 (BM8, Invitrogen), anti-Lyve-1 (NB600-1008, Novus Biological), anti-CD38 (102702, BioLegend). The following secondary Abs were used for staining: Alexa Fluor 488-, Alexa Fluor 514-, Alexa Fluor 568-conjugated anti-rabbit or anti-rat IgG (Life Technologies). Image acquisition was performed with a 63x oil-immersion or 20x objective on an SP5 or SP8 confocal microscope (Leica Microsystem). To minimize fluorophore spectral spillover, the Leica sequential laser excitation and detection modality was used. Where necessary to compensate for uneven slide illumination, fluorescent intensity of layers was normalized using Imaris normalize Layers tool. Where necessary, autofluorescence was filtered from the image by channel subtraction of a deep red autofluorescent channel from APC signal with the Imaris Channel Arithmetics tool.

39The extent of hepatocellular injury was monitored by measuring serum alanine aminotransferase (sALT) activity at multiple time points after treatment, as previously described. Serum HBeAg was measured by enzyme-linked immunosorbent assays (ELISA), as previously described. Blood cell counts were measured by Vet abc (scil).

40Results are expressed as mean ± s.e.m. All statistical analyses were performed in Prism (GraphPad Software), and details are provided in the figure legends. Normality of data distribution was tested in all graphs with a Shapiro-Wilk or D’Agostino & Pearson normality test and parametric tests were chosen only when normality could be confirmed for each dataset. One-tailed test were chosen over two-tailed test when basic biology dictates that the change between the control and treatment group can only occur into one direction (e.g., in cell depletion experiments, where the number of cells will be decreased in the treatment versus the control group). Comparisons are not statistically significant unless indicated.