IFN-γ facilitates liver fibrogenesis by CD161⁺CD4⁺ T cells through a regenerative IL-23/IL-17 axis in chronic hepatitis B virus infection

Accumulation of circulating CD161⁺CD4⁺ T cells in HBV cohorts

The association between CD161⁺CD4⁺ T cells and HBV-related disease parameters has not been investigated previously. Here, using the gating strategy shown in Figure 1a, the frequencies of circulating CD161⁺CD4⁺ T cells were determined in different HBV cohorts. First, a significant accumulation of CD161⁺CD4⁺ T cells was documented in the circulation of CHB and LC compared with the HC group (Figure 1b and c). Second, higher frequencies of CD161⁺CD4⁺ T cells were detected in HBeAg^- than in HBeAg⁺ patients with CHB (Figure 1d). Third, the frequencies increased along with more advanced scores of intrahepatic inflammation and fibrosis in the biopsied HBV cohort (Figure 1e). Intriguingly, there was a closer correlation of the frequencies with HBV-related disease parameters in HBeAg^- than in HBeAg⁺ patients with CHB (Figure 1f), being positive for ALT or liver stiffness, negative for HBV DNA and not significant for AST. Unexpectedly, the statistical analysis of the LC cohort based on disease parameters including Child Pugh score, HBeAg status and HBV DNA yielded no significant findings (Supplementary figure 1a–c).

Higher frequencies of circulating CD161⁺CD4⁺ T cells are associated with increased HBeAg seroconversion in HBeAg⁺ patients undergoing antiviral treatment

Treatment-induced HBeAg seroconversion represents a partial immune control of chronic HBV infection and is associated with a reduced risk of disease progression. Here, the role of CD161⁺CD4⁺ T cells in HBeAg seroconversion was determined using a longitudinal cohort of HBeAg⁺ patients (containing 50 CHB and 34 LC) who had achieved HBeAg seroconversion (responders) or not (non-responders) after a period of 52-week antiviral treatment with entecavir. Notably, responders had significantly higher baseline frequencies in both the CHB (Figure 2a) and LC (Supplementary figure 1d) groups compared to non-responders. However, the serial data (at treatment weeks 0 and 52) of responders only documented a significant increase in circulating CD161⁺CD4⁺ T frequencies in CHB (Figure 2b) rather than in the LC (Supplementary figure 1e) group. In addition, based on the mean value of baseline frequencies, we divided CHB and LC (Supplementary figure 1f, h) groups into ‘low’ and ‘high’ subgroups, respectively, and found that the responder rate was much higher in the ‘high’ subgroup (‘high’ vs. ‘low’, CHB: 42.11% vs. 9.68%, LC: 27.78% vs. 6.25%). Meanwhile, ROC curves were generated using CHB and LC (Supplementary figure 1g, i) data, demonstrating the statistical value of the baseline frequency to predict HBeAg seroconversion at week 52 (AUC = 0.784, P = 0.004). Moreover, the expression of markers of exhaustion (CD39, PD-1, CTLA-4 and Tim-3) and proliferation (Ki-67) was determined in circulating CD161⁺CD4⁺ T cells. Serial data documented a decreased exhaustion with an enhanced proliferation of cells from responders; however, there were no significant findings in non-responders (Figure 2c  and  d).

Phenotypic analysis of circulating CD161⁺CD4⁺ T and homologous CD161^-CD4⁺ T cells from patients with CHB

Using peripheral blood mononuclear cells (PBMCs) from patients with CHB (n = 18), the phenotypic analysis of CD161⁺CD4⁺ T cells was first carried out by the flow cytometric co-staining of CD4/CD161 with the following indicators: intracellular cytokines (pro-inflammatory IFN-γ, IL-17A, TNF-α, IL-22 and anti-inflammatory IL-4 and IL-10) and membrane proteins (Foxp3, the lineage-specific transcription factor of regulatory T cells; CD25, a T-cell activation marker; CD45RO, an immune memory marker). Statistical data indicated that CD161⁺CD4⁺ T vs. CD161^-CD4⁺ T cells expressed much higher levels of IFN-γ, IL-17A, TNF-α and CD45RO, along with similar levels of IL-22, IL-10, IL-4, Foxp3 and CD25, displaying a pro-inflammatory and memory-like phenotype (Figure 3a  and  b). In addition, data documented a significant liver-homing ability of CD161⁺CD4⁺ T vs. CD161^-CD4⁺ T cells, as reflected by the much higher expression levels of chemokine receptors, including CCR6, CXCR6 and CX3CR1 (Figure 3c  and  d).

To explore the role of CD161 signalling in CD4⁺ T cells, sorted CD161⁺CD4⁺ T cells from the circulation of patients with CHB (n = 10) were treated for 72 h with or without imipramine (the inhibitor of CD161 signalling via suppression of acid sphingomyelinase (ASM)),¹⁷ anti-CD161 Ab (the activator of CD161 signalling) or recombinant IL-23 (rIL-23, the upstream signal of Th17 differentiation).^{18, 19} Cytometric beads array (CBA) analysis demonstrated that imipramine significantly decreased the in vitro production of IFN-γ, IL-17A, TNF-α and IL-22 by CD161⁺CD4⁺ T cells rather than CD161^-CD4⁺ T cells (Supplementary figure 2b). Furthermore, imipramine downregulated and anti-CD161 Ab or rIL-23 upregulated the activation/phosphorylation level of STAT3 (a transcription factor associated with IL-17 production) in CD161⁺CD4⁺ T cells and the supernatant concentration of IL-17 (Supplementary figure 2c–e).

Circulating CD161⁺CD4⁺ T cells in HBV cohorts express a pattern of cytokines and liver-homing chemokine receptors

The expression of CD161 in T cells is closely linked with the production of IFN-γ and IL-17, both of which play a role in liver inflammation and fibrosis. Thus, we next focused on analysing IFN-γ/IL-17 or liver-homing receptors expressing CD161⁺CD4⁺ T cell subsets in the circulation. First, we found that IL-17⁺, IFN-γ⁺ and IL-17⁺IFN-γ⁺ subsets were significantly increased in the CHB and LC groups, with the most dramatic changes detected in IL-17⁺IFN-γ⁺ subsets (Figure 4a  and  b), which also significantly increased with more advanced histological scores (Figure 4c). Second, we observed that CCR6, CXCR6 or CX3CR1 expressing subsets accumulated not only in the CHB and LC groups (Figure 4d  and  e), but also in the biopsied HBV cohort with elevated histological scores (Figure 4f). Third, we roughly quantified the intrahepatic CD161⁺ cells (containing T, NK and MAIT) in liver biopsy samples by immunohistochemical staining and documented a significant increase in the cell number along with more advanced histological scores (Figure 4g  and  h). Moreover, we established positive correlations between the intrahepatic CD161⁺ cell number and frequencies of peripheral CCR6, CXCR6 or CX3CR1 expressing CD161⁺CD4⁺ T cell subsets, using paired data analysis (Figure 4i), suggesting liver recruitment as a possible way to expand intrahepatic CD161⁺ cells in disease conditions.

Associations between circulating and intrahepatic CD161⁺CD4⁺ T cells in HBV-mediated cirrhosis

To dissect the potential associations between circulating and intrahepatic CD161⁺CD4⁺ T cells in chronic HBV setting, paired (blood and liver) specimens were collected from HHA (normal) and HCC (cirrhotic) cohorts for subsequent flow cytometric analysis. Initially, an enrichment of total CD161⁺CD4⁺ T cells was observed in cirrhotic livers, despite poor correlations between the paired cell frequencies (Figure 5a). Notably, CD161⁺CD4⁺ T cell subsets co-expressing IL-17 and IFN-γ were enriched in cirrhotic livers, with a positive correlation between the paired subset frequencies (Figure 5b). Moreover, although a similar enrichment of CD161⁺CD4⁺ T cell subsets expressing CCR6, CXCR6 or CX3CR1 was detected in cirrhotic livers, the positive correlations between the paired subset frequencies were only established in CCR6⁺ and CXCR6⁺ instead of the CX3CR1⁺ subset (Figure 5c  and  d). These findings collectively indicate that circulating CD161⁺CD4⁺ T cell subsets that co-express IL-17/IFN-γ, CCR6 or CXCR6 may greatly contribute to the increased number of their intrahepatic counterparts through liver infiltration.

Intrahepatic chemokine expression for circulating CD161⁺CD4⁺ T cells in HBV-mediated cirrhosis

The tissue-homing process is triggered by the interaction between chemokines and chemokine receptors. To obtain more direct evidence of liver infiltration of circulating CD161⁺CD4⁺ T cells, immunohistochemistry was performed to determine intrahepatic chemokine expression and observe spatial contacts with CD161⁺ cells. First, statistical data documented the increased chemokine expression of CCL20, CXCL16, and CX3CL1, the respective ligands for chemokine receptors, including CCR6, CXCR6, and CX3CR1, in cirrhotic livers (Figure 6a  and  b). Next, the increased number of CD161⁺CD4⁺ T cells was confirmed in cirrhotic livers by double staining of CD161 and CD4 (Figure 6c). In addition, positive correlations were established especially in cirrhotic conditions between the intrahepatic CD161⁺CD4⁺ T cell number and chemokine expression (including CCL20, r = 0.59; CXCL16, r = 0.62; CX3CL1, r = 0.53) (Figure 6d). Moreover, the intrahepatic expression of CD161 not only increased with enhanced chemokine expression in cirrhotic livers, but also exhibited a close proximity with each chemokine in spatial position, as reflected by double staining of CD161 and CCL20, CD161 and CXCL16, or CD161 and CX3CL1 (Figure 6e).

The pro-fibrogenic features of circulating CD161⁺CD4⁺ T cells in chronic HBV setting

In this part of the study, the effects of circulating CD161⁺CD4⁺ T cells in liver fibrosis were determined after in vitro co-cultures with pHSCs, which were then collected for analysis of the well-established genes in liver fibrogenesis, including α-SMA, collagen type I, TGF-β1 and tissue inhibitor of metalloproteinase 1 (TIMP1). To begin with, in contact co-cultures, LC rather than CHB-sourced CD161⁺CD4⁺ T cells, at a cell ratio of 1:1 with pHSCs, were found to have pro-fibrogenic effects (Supplementary figure 4c, d). Subsequent co-cultures were conducted using CD161⁺CD4⁺ T cells from the LC cohort at a cell ratio of 1:1. After co-culturing pHSCs with homologous CD161⁺CD4⁺ T or CD161^-CD4⁺ T cells in the contact or transwell (non-contact) manner, CD161⁺CD4⁺ T cells rather than CD161^-CD4⁺ T cells exhibited pro-fibrogenic effects in both manners, which were much more remarkable in the contact manner, suggesting a role for membrane proteins in direct cell contacts (Figure 7a). CD161⁺CD4⁺ T cells can produce both anti-fibrotic IFN-γ and pro-fibrotic IL-17, and IL-23 is an important trigger of IL-17 production by IL-23R⁺CD161⁺CD4⁺ T subsets.^{18, 19} Thus, IL-17, IFN-γ and IL-23 were suspected to formulate a functional axis to regulate liver fibrogenesis, which was then explored with the assistance of neutralising anti-IL-17, anti-IL-23 and anti-IFN-γ Abs. In the transwell assay, only the neutralisation of IL-17 significantly inhibited liver fibrogenesis, demonstrating the direct pro-fibrotic effects of IL-17 (Figure 7b). In contrast, in the contact manner, the neutralisation of each cytokine (IL-17, IL-23 or IFN-γ) more or less inhibited liver fibrogenesis, with the neutralising anti-IL-17 Ab eliciting the strongest effect, collectively suggesting the pro-fibrotic effects of the axis, if present in the co-culture system, and IL-17 as the most downstream effector (Figure 7c). Meanwhile, in the single culture system, pHSCs were also treated with neutralising anti-IL-17, anti-IL-23 or anti-IFN-γ Ab, and statistical data demonstrated that only the neutralisation of IFN-γ showed pro-fibrogenic effects, confirming the anti-fibrogenic role of IFN-γ in an autocrine manner (Supplementary figure 4e). Taken together, the ‘anti-fibrotic’ IFN-γ may exert major effects as an upstream regulator of the ‘pro-fibrotic’ axis.

Because IFN-γ can facilitate HSCs to acquire the phenotype of APCs, as demonstrated by the expression of molecules involved in antigen presentation (MHC-II, CD1d and CD1b) and T-cell activation (CD80, CD86),^{9, 10} while IL-23 is mainly produced by APCs,^{18, 19} IFN-γ was suspected to exist as an upstream regulator of IL-23 production by pHSCs. As expected, contact co-culture with CD161⁺CD4⁺ T cells, comparable to 500 U mL⁻¹ IFN-γ, significantly enhanced the APC phenotype of pHSCs, as reflected by increased mRNA levels of CD80, CD86, MHC-II and CD1d, which was diminished by the neutralisation of IFN-γ (Figure 7d). Flow cytometry was also performed to confirm the phenotypic characteristics of APCs in IFN-γ-treated pHSCs (Supplementary figure 4g). Notably, contact co-cultures with CD161⁺CD4⁺ T cells as compared to IFN-γ had a much greater ability to induce IL-23 expression by pHSCs, which could be partly downregulated by the neutralising anti-IFN-γ Ab, suggesting a mechanism besides IFN-γ for IL-23 production (Figure 7d). Subsequently, the IL-23/IL-17 axis was confirmed by treating homologous CD161⁺CD4⁺ T cell and CD161^-CD4⁺ T cells from the LC cohort (n = 6) in the presence or absence of recombinant IL-23 for further analysis of IL-17 production in the supernatant. As shown in Figure 7e, rIL-23 showed dose-dependent effects in triggering IL-17 production by CD161⁺CD4⁺ T cells rather than CD161^-CD4⁺ T cells, demonstrating the requirement of CD161 signalling in the IL-23/IL-17 axis. In addition, IL-17 also exhibited dose-dependent effects in promoting liver fibrogenesis by pHSCs (Figure 7f). To date, the regenerative and pro-fibrotic IFN-γ/IL-23/IL-17 axis has been established in the contact culture of CD161⁺CD4⁺ T cell subsets with pHSCs.

Evidence of direct contact between CD161⁺ and α-SMA⁺ cells in HBV-mediated cirrhotic livers

To obtain further evidence of direct contact between intrahepatic CD161⁺CD4⁺ T cells and HSCs, the expression of lectin-like transcript 1 (LLT1, the ligand for CD161) was first determined in pHSCs and cirrhotic livers. Significantly elevated expression levels of LLT1 were observed not only in culture-activated vs. quiescent pHSCs (Supplementary figure 4b and c), but also in cirrhotic vs. normal livers (Figure 8a). In addition, there was a positive correlation between intrahepatic LLT1 and α-SMA expression in cirrhotic livers (Figure 8b). Further, by immunohistochemical double staining of α-SMA/CD161 and CD161/LLT1, both the spatial contacts and the remarkably increased area of positive co-staining were documented in cirrhotic livers, suggesting that CD161⁺ cells may directly interact with HSCs to be involved in liver fibrogenesis (Figure 8c  and  d). Moreover, immunofluorescent triple staining of CD161/IL-23R/α-SMA revealed a significant increase in α-SMA⁺ in direct contact with CD161⁺IL-23R⁺ (being responsive to IL-23 for IL-17 production) cells in cirrhotic livers compared to normal livers (Figure 8e).

Subjects

A total of 94 patients with CHB, 73 with HBV-mediated LC and 28 healthy controls (HC) were enrolled in this study. The phenotypic analysis of circulating CD161⁺CD4⁺ T cells was carried out before antiviral treatment (clinical parameters of subjects are summarised in Supplementary table 1). As disease controls, chronic hepatitis and cirrhotic patients with non-HBV aetiologies including autoimmune, non-alcoholic steatohepatitis (NASH) and alcohol were also analysed. Of the HBV cohorts, HBeAg-positive patients (CHB, n = 50; LC, n = 34) were followed up for a period of 52-week entecavir monotherapy. At 3, 6, 9 and 12 months after treatment, the serological responses of HBeAg were evaluated. Here, responders were defined as those who achieved HBeAg seroconversion, referring to HBeAg loss with or without the appearance of anti-HBe antibody (Ab). The baseline characteristics of the responders and non-responders are summarised in Supplementary table 2. In addition, liver biopsy specimens from 62 patients with chronic HBV infection were also examined (Supplementary table 3). The degree of inflammation (‘A’ for activity) and fibrotic stage (“F’ for fibrosis) were graded according to the Metavir scoring system by two senior pathologists.²⁵ The patients were diagnosed and classified as previously described.²⁶ The exclusion criteria for HBV cohorts included drug abuse, other viral infections, coexisting serious illness and pregnancy.

To explore discrepancies between circulating and intrahepatic CD161⁺CD4⁺ T cells, 15 patients with HCC secondary to HBV-associated cirrhosis and 15 with surgery-requiring hepatic haemangioma (HHA) (clinical parameters summarised in Supplementary table 4) were enrolled to analyse the paired blood and liver samples. Here, cirrhotic and normal liver specimens were collected as tumor-free para-tumorous tissues (at least 1 cm away from the tumor) from HCC and HHA cohorts.

Ethics statement

All individuals enrolled in this study provided written informed consent. The study protocol was approved by the Research Ethics Committee of Zhongshan Hospital (B2013-068, B2016-059) and Tongji Hospital (2015-KYSB-11).

Reagents

For flow cytometry, anti-human antibodies (Abs) for CD3-percp-cy5.5 (SK7), CD161-Percp-cy5.5 (HP-3G10), CD45RO-PE (UCHL1), and IFN-γ-PE (4S.B3), IL-17-APC (64DEC17), TNF-α-PE (MAb11), IL-22-eFlour450 (22URTI) and IL-10-PE (JES3-9D7) were purchased from eBioscience (Thermo Fisher Scientific, Waltham, USA). Antibodies against PD-1-BV510(EH12.2H7), TIM-3-PE(F38-2E2), CD39-Percp-cy5.5 (A1), CTLA-4-PE (L3D10) and Ki-67-BV421 (Ki-67) were purchased from Biolegend (San Diego, CA, USA). Antibodies against CD3-BV421 (SK7), CD4-BB515 (RPA-T4), CD161-APC (DX12), CD25-PE (M-A251), CCR6-PE-Cy7 (11A9), CXCR6-BV421 (13 B 1E5), CX3CR1-PE (2A9-1), Foxp3-PE (236A/E7) and IL-4-Percp-Cy5.5 (8D4-8) were purchased from BD Biosciences. Stimulatory anti-CD161 antibody was purchased from BD Bioscience (Becton). Imipramine was purchased from Sigma-Aldrich (St. Louis, MO, USA), and recombinant human cytokine IL-23 (rIL-23) was purchased from R&D Systems (Bio-Techne, Minnesota).

Preparation of circulating and intrahepatic cell samples

Peripheral blood mononuclear cells were isolated from ethylenediaminetetraacetic acid (EDTA)-treated blood samples (10 ml) using the Ficoll density gradient centrifugation method (GE Healthcare, USA). Circulating CD161⁺CD4⁺ and CD161^-CD4⁺ T cells were sorted from PBMCs by FACS Aria II (Becton Dickinson) after sterile surface staining for CD3, CD4 and CD161, and the cell purity was confirmed to be > 95% (Supplementary figure 2a).

A single-cell suspension of mouse liver tissue was prepared using the gentle MACS Dissociator (Miltenyi Biotec) with subsequent digestion by collagenase type I (Sigma-Aldrich) and DNase (Gibco). Liver-infiltrating lymphocytes (LILs) were recovered by centrifugation over 40% and 70% Percoll gradient (GE Healthcare). Using peritumoral liver specimens from the HHA cohort (n = 5), primary HSCs (pHSCs) were isolated through 8% Nycodenz (Axis-Shield) gradient centrifugation and then cultured in high-glucose DMEM supplemented with 10% FBS. The cell purity of HSCs was demonstrated by positive staining for α-smooth muscle actin (α-SMA) (Supplementary figure 4a).

Flow cytometry

Peripheral blood mononuclear cells and LILs were prepared for phenotypic analysis of circulating and intrahepatic CD161⁺CD4⁺ T cells. Human Fc Block™ (BD Bioscience, USA) was used to block non-specific binding. Dead cells were excluded using a fixable viability dye (FVD)-780 (eBioscience). Negative controls were set appropriately. FMO (fluorescence minus one) staining was performed to set a proper gate for indicators without an obvious bimodal distribution of expression. For staining of intracellular cytokines, cells were pre-treated with a cell stimulation cocktail containing phorbol-12-myristate-13-acetate (PMA) and ionomycin (2 μL mL^-1) (eBioscience) for 5 h. A minimum of 1 × 10⁵ cells were collected by LSRFortessa^TM (BD Bioscience) for further analysis using FlowJo V10 software (Tree Star).

Immunohistochemistry and immunofluorescence

For immunohistochemistry, liver tissue specimens (4 μm, paraffin-embedded, formalin-fixed) were dewaxed, hydrated and subjected to heat-induced antigen retrieval. Sections were blocked with 5% BSA (Sigma-Aldrich). A standard two-step protocol (Novolink Polymer Detection System) was applied with the following primary antibodies: rabbit anti-α-SMA (1:1000) (ab124964), mouse anti-CD161 (1:200) (556079; BD Biosciences), rabbit anti-IL-23R (1:100) (ab175072), rabbit anti-LLT1 (1:100) (ab197341), rabbit anti-CCL20 (1:100, ab136904), rabbit anti-CXCL16 (1:100, #500-P200, Peprotech) and rabbit anti-CX3CL1 (1:400) (ab89229). Immunohistochemical double staining was performed using a DouMaxVision TM kit (MXB). Immunofluorescence triple staining was performed by Servicebio Technology Co., Ltd (Wuhan, China). The positively stained area was quantified by ImageJ (NIH), while the number of positively stained cells was calculated using Image-Pro Plus version 6.0 (Media Cybernetics, Rockville, MD, USA). Six high-power fields (hpf, 400× magnification) of each specimen were selected for the statistical analysis.

Cell cultures and co-cultures

T-cell activity was maintained in vitro using anti-CD3/CD28 beads at a bead-to-cell ratio of 1:1. At different cell ratios, pHSCs (5 × 10⁴ per well in 24-well plates) were co-cultured with sorted circulating CD4⁺ T subsets in the contact or non-contact manner for 72 h, with or without addition of the neutralising anti-IL-17 (10 μg mL^-1) (AF-317-NA; R&D Systems), anti-IL-23p19 (100 μg mL^-1) (AF1716; R&D Systems) or anti-IFN-γ (50 μg mL^-1) (MAB285; R&D Systems) Ab. The non-contact co-culture was carried out using a transwell system (Corning) with a polycarbonate membrane (pore size: 0.4 µm, not allowing cells to pass through) between the upper and lower compartments.

Quantitative reverse transcription PCR (qRT-PCR)

Approximately 1 × 10⁵ cells were collected for qRT-PCR analysis. Total RNA was extracted using Trizol (Invitrogen), followed by reverse transcription and quantitative real-time PCR using commercial kits (Takara Bio). The list of primers used are shown in Supplementary table 5. The mRNA levels were calculated using the ▵▵Ct method after normalisation with GAPDH and quantified as fold change.

Detection of cytokines in supernatants

To determine in vitro cytokine production, supernatant samples were collected after treating cells (a total of 1 × 10⁶) with anti-CD3/CD28 beads at a bead-to-cell ratio of 1:1 for 72 h. CBA analysis was then applied to detect cytokine concentrations using the commercial Human Th Cytokine Panel kit (IFN-γ/TNF-α/IL-17/IL-21/IL-22, LEGENDplex^TM; Biolegend) and LSRFortessa^TM. The concentration of IL-17 was also analysed using a human enzyme-linked immunosorbent assay (ELISA) kit (eBioscience).

Western blot analysis

Approximately 1 × 10⁵ cells were collected for cell lysis, protein extraction and western blot analysis. PhosSTOP (Roche) was added before analysing the phosphorylated forms of the proteins. Protein levels were assessed using primary rabbit anti-human Abs: STAT3 (88 kDa) (10253-2-AP; Proteintech), phospho-STAT3 at Tyry⁷⁰⁵ (79, 86k Da) (#9145; CST) and LLT1 (34 kDa) (ab151738; Abcam). Relative protein expression was normalised with an anti-β-actin antibody (42 kDa) (Proteintech) and quantified using ImageJ (NIH).

Statistical analysis

Results are presented as the mean ± SD or mean (range). Statistical analyses were performed using GraphPad Prism version 7.0 (San Diego, CA, USA). The Student’s t-test, analysis of variance (ANOVA) with the Bonferroni test and the Pearson correlation analysis were performed as appropriate. Statistical significance was set at P ≤ 0.05.