All individuals were recruited according to the protocol that was approved by the clinical Ethics Committees of Shanghai Changhai Hospital (CHEC2017-118) and registered on Clinicaltrials.gov (NCT03734783). The study was conducted between July 2017 and November 2019. Written informed consent was obtained from all participants.

Patients with CHB were all sero-positive for hepatitis B surface antigen (HBsAg) for at least 6 months and negative for other hepatotropic viruses, such as hepatitis D virus (HDV), hepatitis C virus (HCV), hepatitis E virus, hepatitis A virus, and human immunodeficiency virus 1/2. Healthy controls (HCs) did not have any previous history and current evidence of liver disease. Serum alanine aminotransferase (ALT) values were normal, and HBsAg and other hepatitis virus markers were negative. Furthermore, all participants were sero-negative for markers such as ceruloplasmin, anti-nuclear antibodies and anti-mitochondrial antibodies for co-existent autoimmune and metabolic liver diseases.

Peripheral blood mononuclear cells (PBMCs) were isolated using lymphocyte separation medium (Ficoll-Hypaque density gradient, Axis-shield, Germany). The antibodies and fluorochromes used in this work are shown in Table S1 .

For Breg subset and B10 or IL-35+B cell detection, PBMCs (2 × 10⁶) were stimulated with CD40L (5 µg/mL, Purified Anti-Human CD154 (CD40L), Tonbo Biosciences, USA) plus CpG-ODN (1.5 µM, TLR9 Agonist-Stimulatory Class B tlrl-2006, InvivoGen) and lipopolysaccharide (LPS, 1 µg/mL, eBioscience) for 48 h. Furthermore, cells were activated for another 4 h at 37°C in 5% CO₂ with 50 ng/mL phorbol myristate acetate, 1 mmol/L ionomycin (both from Sigma, St. Louis, MO, USA), and 10 mg/mL brefeldin A (Tocris Cookson, Bristol, UK) in complete RPMI-1640 (Invitrogen, Carlsbad, CA, USA) supplemented with 10% heat-inactivated fetal bovine serum (Gibco, Grand Island, NY, USA). The cells were then collected and washed with PBS once and stained for surface markers. Then, they were permeabilized with Perm/Fix solution (Cytofix/Cytoperm™, BD biosciences) and stained intracellularly with specific fluorescent conjugated anti-human IL-10, eBi3/IL-12/IL-35p35 (IL-35). For B cell subsets, we analyzed two classical Breg subsets: CD24^hiCD38^hi, CD24^hiCD27+, the total IL-10 secreting B cells and IL-35 secreting B cells, which were all gated on CD19+ cells ( Figure S1B ). We also determined the frequency of IL-35 producing Breg: CD24^hiCD38^hiIL-35+, CD24^hiCD27+IL-35+, and IL-10 secreting Breg: CD24^hiCD38^hiIL-10+, CD24^hiCD27+IL-10+, frequency was determined into CD24^hiCD38^hi and CD24^hiCD27+ gates, respectively ( Figure S1 ).

For T cell subsets detection, freshly isolated PBMCs (2 × 10⁶) were stimulated for 5 h with 50 ng/mL phorbol myristate acetate,1 mmol/L ionomycin (both from Sigma, St. Louis, MO, USA), and 10 mg/mL brefeldin A (Tocris Cookson, Bristol, UK) in complete RPMI-1640 (Invitrogen, Carlsbad, CA, USA) supplemented with 10% heat-inactivated fetal bovine serum (Gibco, Grand Island, NY, USA). Upon harvest, cells were first stained with surface markers (CD3-APC-Cy7-A, CD4-FITC-A, and CD8-PerCP-Cy5-5-A) and then intracellularly stained with IL-17-APC-A, IFN-γPE-Cy7-A, and IL-4-PE-A. All these T cell subsets were gated on PBMC cells ( Figure S2 ).

Flow cytometry was performed using a FACSCalibur (Becton Dickinson, San Jose, CA). FACS data were analyzed using CellQuest software (Becton Dickinson Rutherford, NJ).

Serum concentrations of inflammatory cytokines, including IL-4, IL-17A, IL-21, and interferon-γ (IFN-γ), were measured using commercially available Luminex MAP kits (ProcartaPlex 5 plex, 1 plate 5-plex, eBioscience) in accordance with the manufacturer’s instructions. Samples were two times diluted with 1×universal assay buffer and tested in triplicate.

Serum and PBMC culture supernatant concentrations of IL-35 and IL-10 were measured by using commercially available ELISA kits (XpressBio, USA; Proteintech, USA) in accordance with the manufacturers’ instructions. All samples were not diluted and assessed in triplicate.

Purified B cells isolated from PBMC using anti-human CD19 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) were treated with CpG/CD40L/LPS with or without HBVcore (1-183 a.a) (ProSpec, USA) for 48 h. After stimulation, B cells were washed with PBS twice and then co-cultured with autologous CD19-depleted PBMCs (50,000 cells) at the ratio of 2:1 in anti-CD3/CD28 (working concentration of anti-CD3 and anti-CD28 was 10μg/ml and 2μg/ml respectively, Tonbo Biosciences, San Diego, CA)-coated 96-well plate in complete RPMI-1640 (Invitrogen, Carlsbad, CA, United States) supplemented with 10% heat-inactivated fetal bovine serum (Gibco,Grand Island, NY, United States) for another 48 h. After 48 h of co-culture, the frequency of IL-4-, IFN-γ-, and IL-17A-secreting T cells was evaluated by flow cytometry.

Transwells of 0.4 µm pore size were used (Millicell, Merck Millipore, Billerica, MA, USA). CD19-depleted PBMCs (50,000 cells) were added to the lower chambers of 24-well plates and subsequently stimulated with anti-CD3/anti-CD28 as described above, whereas non-treated or stimulated B cells (100,000) were added to the upper chambers in the transwells. As controls for the transwell assay, CD19-depleted PBMCs (50,000 cells) and stimulated B cells (100,000 cells) were also co-cultured in the same plate in the lower chambers with anti-CD3/anti-CD28. After 48 h of co-culture, CD19-depleted PBMCs were gathered, and the frequency of IFN-γ-secreting T cells was tested by flow cytometry. For cytokine blocking study, anti-IL-12/IL-35 p35 (1 µg/mL, R&D Systems) and anti-IL-10 (5 µg/mL, Invitrogen) were added to the co-culture system to investigate the potential mechanism of Breg on T cells.

Serum HBV DNA levels were quantified using fluorescent quantitative PCR with commercially available kits (Sansure Biotech, China). The detection range was 1 × 10² to 5 × 10⁸ IU/mL.

The levels of HBsAg, HBeAg, anti-HBs, anti-HBc, anti-HBe, anti-HCV, anti-HDV, anti-HGV, anti-HIV-1, and anti-HIV-2 were measured using commercially available kits (Abbott Laboratories, North Chicago, IL) in our clinical laboratory. The dynamic range of serum HBsAg was 0.05–250 IU/mL. The samples were diluted to 1:500 or 1:1000 using the ARCHITECT HBsAg Manual Diluent (Abbott Diagnostics) if >250 IU/mL.

Normally distributed continuous quantitative data were expressed as the mean ± SEM, and non-normally distributed continuous quantitative data were expressed as median (interquartile range,IQR). The non-parametric Mann–Whitney U test was used to evaluate the differences between two independent samples, and the non-parametric Kruskal–Wallis ANOVA test was used to evaluate the differences in more than two groups. The non-parametric Friedman ANOVA test was used to evaluate effector T cell proportions in co-culture experiments. The statistical correlation between variables was calculated by Spearman rank correlation analysis. P-value < 0.05 was considered statistically significant. All analyses were performed using SPSS software (version 21.0.0; Chicago, IL). The graph was made using GraphPad Prism 5 (San Diego, CA, USA).

The clinical and biochemical characteristics of the studied patients are listed in Table 1 . All 68 CHB patients were treatment-naïve at enrollment. Approximately 57.35% (39/68) of patients had ALT levels greater than the upper limit of normal (ULN) at enrollment, and higher proportions of patients in the abnormal ALT group were HBeAg positive compared with those in the normal ALT group. Similarly, HBeAg-positive patients at enrollment had higher median ALT levels than HBeAg-negative patients. Twenty-six age- and gender-matched HCs were enrolled.

To determine the role of IL-35 in persistent HBV infection, we first determined the serum IL-35 concentration in 68 patients with persistent HBV infection. Compared with HCs, CHB patients had significantly increased serum IL-35 levels. Stratified analysis found that the serum IL-35 levels were significantly increased in 39 patients with liver flare ( Figure 1A ); patients with active viral replication, including 39 HBeAg+ CHB patients; and 49 patients with viral load more than 2E4 IU/mL ( Figures 1A, B ).

Breg response was defined by functional cytokines (23), and many researches have demonstrated that IL-35+B cells play a suppressive regulatory role in autoimmune and infectious diseases (22–24). However, the role of IL-35+B cells in chronic HBV infection is still unclear. To evaluate the role of circulating mononuclear cells in IL-35 secretion in persistent HBV infection, we quantified the frequency of IL-35+B cells in total B cells in 35 patients with chronic HBV infection. Compared with that in HCs, the frequency of both IL-35+B cells and IL-10-secreting B (B10) cells in total B cells in patients with chronic HBV infection was not increased significantly. ( Figure 2A ).

To analyze the relationship between IL-35+B cells and liver inflammation degree and viral replication level, we divided the treatment-naïve patients into two groups according to the ALT levels: patients with high ALT levels (ALT ≥ 200 U/L, n=25) and patients with low ALT levels (ALT < 200 U/L, n=10). Then, we also divided the patients into two groups according to HBeAg status and HBV DNA levels: patients with high virus replication (HBeAg positive, n=19; or HBV DNA ≥ 2E4 IU/mL, n=17) and patients with low virus replication (HBeAg negative, n=16 or HBV DNA < 2E4 IU/mL, n=18). Compared with HCs and patients with low ALT levels, patients with high ALT levels had the highest frequency of IL-35+B and B10 cells ( Figure 2B ). Among patients with different HBeAg status or viral loads and HCs, no significant difference was found in the frequency of IL-35+B cells ( Figures 2C, D ) in the peripheral blood total B cells. However, in patients with low viral replication, the frequency of B10 cells was higher than that of IL-35+B cells ( Figures 2C, D ). According to Spearman correlation analysis, a moderate correlation was observed between the frequency of IL-35+B cells and ALT levels ( Figure 2E ), but only mild correlation was found between the frequency of IL-35+B cells and HBV DNA levels ( Figure 2F ). The frequency of B10 cells was only moderately correlated with serum ALT levels in these patients ( Figure S3 ).

To quantify the frequency of various Breg subsets, the PBMCs of 33 CHB patients (17 HBeAg-positive patients and 16 patients in the abnormal ALT group) and 11 HCs were isolated and stimulated with CpG plus CD40L and LPS in vitro for 48 h, which was proved to potently induce Breg differentiation (22, 25). Then, the frequency of classical human Breg subsets CD19+CD24^hiCD38^hi and CD19+CD24^hiCD27+ was quantified using flow cytometry. Compared with HCs, the frequency of CD19+CD24^hiCD38^hi (19.76% ± 3.64% vs. 8.43% ± 3.46%) and CD19+CD24^hiCD27+ (11.42% ± 2.37% vs. 6.28% ± 2.18%) increased in patients with CHB ( Figure 3A ). Further stratified analysis of the frequency of Breg subsets in patients according to HBV replication and liver inflammation degree showed that patients with active liver inflammation (ALT ≥ 2*ULN) and virus replication (HBeAg positive) had higher frequency of Breg subsets, and in these patients, the Breg subset was dominated by CD19+CD24^hiCD38^hi ( Figures 3B, C ).

As IL-35 is one of the markers of phenotypic changes in B cells, we analyzed the relationship of the frequency of IL-35+B cells with human classical Breg subsets during persistent HBV infection. As shown in Figures 3D, E , the frequency of IL-35+B cells was strongly correlated with that of two classical human Breg subsets during persistent HBV infection (Spearman r = 0.749 and 0.782, P < 0.001). Furthermore, we found that patients with persistent HBV infection had higher frequency of IL-35-positive Breg subsets compared with HCs, but no difference was observed between the two classical Bregs ( Figure 3F ).

The main function of Bregs is to regulate the differentiation of Th cells and exert suppressive function on T cell proliferation through their effector factors, such as IL-10, IL-35, and TGF-β1. To evaluate the function of IL-35+B cells during persistent HBV infection, the frequency of IL-35+B cells and homologous effector T cell subsets was measured, and the correlation of IL-35+B cells with various T cell subtypes in CHB patients was analyzed using Spearman’s rank correlation analysis. We found that the frequency of IL-35+B cells was negatively correlated with that of IFN-γ-producing CD4+ and CD8+ cells, but positively correlated with that of IL-4-producing T cells ( Table 2 ). However, a weaker correlation was found between serum IL-35 levels and different T cell effector cytokines ( Table S2 ).

To determine the suppressive effect of Bregs on T cells, B cells isolated from PBMCs using anti-CD19 microbeads were stimulated with CpG/CD40L/LPS with or without HBVcore peptide for two days. Then, autologous CD19-depleted PBMCs were co-cultured with HBV-exposed B cells or stimulated B cells and subsequently activated with anti-CD3/anti-CD28 for another 48 h. Because of the importance of Th1/Th2 imbalance in CHB disease progression, we aimed to determine the regulatory effect of Bregs on IL-4- and IFN-γ-producing T cells. As shown in Figures 4A–E , ECs were stimulated significantly in vitro by anti-CD3/CD28, and Bregs significantly decreased the frequency of IFN-γ-producing CD4+ and CD8+ T cells ( Figures 4A, D ). However, the frequency of IL-4-producing T cells and IL-17-producing T cells were not significantly affected by Bregs ( Figures 4B, C, E ).

Das et al. (17) had shown that Bregs exert their immunoregulatory role by secreting IL-10 during persistent HBV infection. We aimed to determine whether the IL-35 pathway is also involved in Bregs’ suppressive function during persistent HBV infection. As shown in Figure 5A , IL-35 neutralizing antibody could significantly reverse the suppressive effect of Bregs on EC cytokine production, but no synergistic function of IL-35 and IL-10 neutralizing antibody was found. Furthermore, we found that the suppressive function of Bregs on ECs was significantly decreased in the trans-well co-culture system ( Figure 5B ).