Research ResourcesHEPATITIS

Multifactorial heterogeneity of virus-specific T cells and association with the progression of human chronic hepatitis B infection

See allHide authors and affiliations

Science Immunology  08 Feb 2019:
Vol. 4, Issue 32, eaau6905
DOI: 10.1126/sciimmunol.aau6905

Mapping the T cell response to hepatitis B

Unlike evaluating antibody responses to infections, studying T cell responses has always been more complicated. A significant hurdle here continues to be the development of reagents to identify and characterize pathogen-specific T cells. Here, Cheng et al. have generated a comprehensive panel of peptide class I major histocompatibility complex tetramers to study CD8+ T cells that recognize hepatitis B virus (HBV). Using HBV-specific tetramers in conjunction with mass cytometry, they have documented functional states of HBV-specific CD8+ T cells in individuals with chronic HBV infection. The tetramer panels generated here to study HBV-specific T cells should be of broad utility to the HBV research community.


Associations between chronic antigen stimulation, T cell dysfunction, and the expression of various inhibitory receptors are well characterized in several mouse and human systems. During chronic hepatitis B virus (HBV) infection (CHB), T cell responses are blunted with low frequencies of virus-specific T cells observed, making these parameters difficult to study. Here, using mass cytometry and a highly multiplexed combinatorial peptide–major histocompatibility complex (pMHC) tetramer strategy that allows for the detection of rare antigen-specific T cells, we simultaneously probed 484 unique HLA-A*1101–restricted epitopes spanning the entire HBV genome on T cells from patients at various stages of CHB. Numerous HBV-specific T cell populations were detected, validated, and profiled. T cells specific for two epitopes (HBVpol387 and HBVcore169) displayed differing and complex heterogeneities that were associated with the disease progression, and the expression of inhibitory receptors on these cells was not linearly related with their extent of T cell dysfunction. For HBVcore169-specific CD8+ T cells, we found cellular markers associated with long-term memory, polyfunctionality, and the presence of several previously unidentified public TCR clones that correlated with viral control. Using high-dimensional trajectory analysis of these cellular phenotypes, a pseudo-time metric was constructed that fit with the status of viral infection in corresponding patients. This was validated in a longitudinal cohort of patients undergoing antiviral therapy. Our study uncovers complex relationships of inhibitory receptors between the profiles of antigen-specific T cells and the status of CHB with implications for new strategies of therapeutic intervention.


Chronic hepatitis B virus (HBV) infection (CHB) remains a major health issue and is the leading causative agent of hepatocellular carcinoma worldwide. Despite an effective vaccine, CHB has no cure and many patients are only diagnosed during the later stages of the disease where treatment efficacy is limited. The estimated mortality due to viral hepatitis has escalated by more than 50% in the past decade (1). CHB develops into a series of stages (2) that are defined by a few clinical parameters with limited associated immunological evidence. How adaptive immunity changes as young patients progress from the immune tolerant stage (IT; HBeAg+ chronic infection, high viremia but limited liver inflammation) (2, 3) to progressive immune active stage (IA; HBeAg+ chronic hepatitis, high viremia and high liver inflammation) and, for some, spontaneously become HBeAg inactive carriers (InA; HBeAg chronic infection, low to undetectable viral load and limited liver inflammation) is not thoroughly understood. Some argue that these definitions could affect early treatment opportunity and should be revisited (4, 5).

Despite their low frequency in most CHB patients (6, 7), the virus-specific T cell response has been of much interest for HBV immunologists. Seminal experiments on CHB patients and animal models such as HBV-challenged chimpanzee have shown the indispensable function of virus-specific CD8+ T cells in viral control (6, 810). Historically, mapping for potential virus-specific CD8+ T cells against HBV has mainly focused on HLA-A*02:01–restricted epitopes. For instance, numerous studies on a single epitope targeted by A*02:01-restricted HBVcore18–27-specific CD8+ T cells have provided many implications for immunotherapy (11, 12). However, in Asia, which has high prevalence of HBV infection, the predominant allele in common East Asian ethnicities is A*11:01, whose immunogenicity against chronic HBV is less-well defined. Hence, there is an unmet need to investigate A*11:01-restricted HBV-specific CD8+ T cells in CHB. Furthermore, regardless of major histocompatibility complex (MHC) restriction, because of the very low frequencies of HBV-specific T cells, information about their unmanipulated phenotypes are lacking.

In murine chronic lymphocytic choriomeningitis virus (LCMV) infection, prolonged and elevated viral antigenic exposure coinciding with the up-regulation of multiple inhibitory receptors on virus-specific T cells has led to the definition of the state of T cell exhaustion (13, 14). The resemblance of exhausted T cells (TEX) observed in human chronic viral infection not only provides an explanation for the functional failure of immune response but also pinpoints a valuable target to boost host immunity. Evidence suggest that such TEX arise from an altered path of memory T cell development (15, 16). Besides several well-described defects (13, 14, 1619), a hallmark of T cell exhaustion is a progressive loss of functional capacity (20), which correlates with the cumulative expression of inhibitory receptors over the course of persistent antigen stimulation (13, 21). This functionally impaired T cell subset has been described in HIV and hepatitis C virus (HCV) infection; however, results from other studies do not necessarily fit this model (2226). Therefore, by simultaneously measuring a wide range of inhibitory receptors and memory-associated markers, we aimed to evaluate the extent to which the profiles of HBV-specific T cells fit with such a model of “hierarchical T cell exhaustion” in human CHB (13, 27, 28).

Here, to overcome challenges associated with identifying and deep-profiling unmanipulated HBV-specific T cells, we used mass cytometry together with a highly multiplexed combinatorial peptide-MHC (pMHC) tetramer strategy to simultaneously screen and interrogate 562 A*11:01-restricted T cell candidate epitopes. Using a self-validated automatic tetramer deconvolution and unsupervised high-dimensional analyses, we found that virus-specific CD8+ T cells targeting HBVpol387 and HBVcore169 displayed complex phenotypic profiles and T cell receptor (TCR) sequence usages that covaried with the HBV infection status. On the basis of a high-dimensional trajectory analysis, we also found that the profiles of HBV-specific T cells from blood are indicative of the degree of viral control in patients from two separately analyzed cohorts.


Comprehensive HBV epitope mapping

To generate a comprehensive HBV targeting pMHC library, viral DNA was isolated and deep-sequenced from serum samples of 15 longitudinal CHB patients to determine viral consensus sequences and common variants (Fig. 1A and fig. S1A; see Materials and Methods). The sequences were loaded onto the NetMHC platform to predict possible A*11:01-restricted binders. Four hundred eighty-four unique putative HBV epitopes above the predictive “weak binding” threshold were combined with 78 known epitopes derived from other common antigens to arrive at a total of 562 peptides, listed in table S1. Sequence homology was analyzed to group relatively similar peptides into the same cluster by a pairwise matching algorithm (fig. S1B and table S1). The resulting 284 peptide clusters were randomly assigned unique combinations of quadruple streptavidin-metal (SAv-metal) coding (Fig. 1A and fig. S1C). This approach avoided false interpretation of the combinatorial pMHC tetramer strategy (29) that would result from T cells expected to cross-react with multiple minor variants of the same peptide. The coded 562-plex pMHC tetramer library was pooled and simultaneously probed on each patient’s peripheral lymphocytes (fig. S1C) (30). To increase confidence of detection, we evenly divided and independently interrogated patient’s cells by the same 562-plex pMHC tetramer library of two entirely different SAv-metal coding configurations (figs. S1, C and D, and S2A). Together with >26 cellular markers (table S2), the signals of 562-plex pMHC tetramers were determined by mass cytometry. A self-validated automatic combinatorial tetramer deconvolution algorithm was used to identify tetramer-positive events in an unbiased way (Materials and Methods; fig. S1D). The correspondences between matching tetramers from two coding configurations were calculated using a bootstrapping statistical analysis (see Materials and Methods, Fig. 1A, and fig. S1D). Last, we enumerated the validated antigen-specific CD8+ T cells for those passing all the deconvolution criteria between two configurations, and such approach can be verified by the correlation of control epitopes [e.g., cytomegalovirus (CMV) and Epstein-Barr virus (EBV)] detected between the configurations (figs. S1 and S2A). Using this objective strategy, we were able to detect many unidentified candidate epitopes and their variants in CHB patients, as well as the previously identified epitopes (table S1).

Fig. 1 Comprehensive epitope mapping against HBV using highly multiplexed combinatorial pMHC tetramer strategy.

(A) Experimental workflows. The 562-plex pMHC tetramer library was generated from the deep sequencing of virus and epitope prediction. The library included 484 putative A*11:01-restricted HBV peptides and 78 known control peptides derived from other common virus or self-antigens. One thousand one unique combinations of quadruple SAv-metal codings were used to code the entire library. A self-validated tetramer deconvolution algorithm automatically identified the signals on patient’s T cells with statistical measurements. Validated antigen-specific CD8+ T cells targeting four viral epitopes were shown. (B) Mean frequency of HBV-specific CD8+ T cells from all patients tested across four different viral proteins. Plot only shows the detectable epitopes. Numbers at the bottom indicate the numbers of epitopes detected/screened for each viral protein. (C) Epitope nomenclature and annotation used in this report are shown. “*” indicates a peptide cluster that contained more than one peptide (table S1). Peptide sequences in boldface are previously unpublished sequences based on Immune Epitope Database. (D) Frequencies of four antigen-specific CD8+ T cells across various patient groups (color-coded). (E) Expression of cellular markers on four HBV-specific CD8+ T cells in heatmaps. Boxes highlight the discriminative markers for each patient group.

We hypothesized that the antigen-specific T cell responses could vary across different clinical stages and reflect CHB disease progression. Therefore, we applied this strategy to map potential T cell epitopes across three CHB patient groups (IT, IA, and InA) and one group of acutely resolved patients (R) (table S3). There was no difference in the overall magnitudes of detected antigen specificities between patient groups (fig. S3). However, across all tested patients, we detected T cells specific for more epitopes derived from polymerase (P) and core (C) compared with envelope (S) and x (X) proteins, including four epitopes with the highest frequencies observed (Fig. 1B). These were HBV-P-282 (cluster 090, four peptide variants), HBV-P-387 (cluster 106, one unique peptide), HBV-C-169 (cluster 178, seven peptide variants), and HBV-C-195v2 (cluster 283, one unique peptide) (Fig. 1, B and C; fig. S2; and table S1).

Several experiments were performed to validate and assess the HBV relevance of these four epitopes. For three of four of these epitopes, antigen-specific T cells were detected in some healthy donor (HD) or cord blood (CB) samples (Fig. 1D) but not in human leukocyte antigen (HLA)–mismatched patients (fig. S4). The detection of these cells was further reproduced and confirmed by fluorescence flow cytometry pMHC tetramer staining, which showed consistent results (fig. S5, A and B). These T cells could also proliferate upon the stimulation by corresponding viral peptides (fig. S6), except the seemingly unresponsiveness of cells from IT and HD. Although our results were clear in confirming the specificity of T cells stained with HBV-P-282 and HBV-C-195v2 peptide-loaded pMHC tetramers, these cells displayed a mostly naïve-like phenotype (Fig. 1E). It is likely that T cells specific for these epitopes could be naïve T cells with relatively high precursor frequencies due to the nonrandom nature of TCR recombination (31) and were similar to HCV-specific CD8+ naïve precursors that were previously described (32, 33).

A previously identified epitope, HBV-P-387 (34, 35), was observed with a relatively high frequency in CHB and also in half of the HDs tested, but to a lesser extent in CB samples. As described further in subsequent sections, the heterogeneous phenotypes of these cells were highly variable between patients (Fig. 1E), suggesting their relevance to the HBV-specific immune response. Given the high prevalence of HBV in Singapore, where the HDs were collected, we postulate that such unexpected detection of HBV-specific T cells could be due to the high coverage of vaccination or subclinical infection without the development of anti-HBc antibody (HBcAb), as reported in HBV-exposed health workers (Fig. 1D and fig. S5C) (36, 37). This remains to be determined. Last, among the validated epitopes, T cells specific for HBV-C-169 were only detected in HBV-infected individuals and elevated in patients with viral control (InA and R) (Fig. 1D) compared with patients with high viral load (IT), where these cells were undetectable. Such HBV-specific T cells displayed phenotypic profiles that differed according to the status of HBV infection (Fig. 1E). These results prompted us to further evaluate the profiles of HBVpol387- and HBVcore169-specific CD8+ T cells.

High-dimensional phenotypic profiling of HBV-specific CD8+ T cells

We directed further analyses at HBVpol387- and HBVcore169-specific CD8+ T cells because of their higher degrees of phenotypic heterogeneity observed across patients at various stages of CHB. Although both HBV-P-387 (LVVDFSQFSR) (34, 35) and one of the peptides within the HBV-C-169 (STLPETTVVRR) have been previously reported (34, 35, 3840), the phenotypes of these reactive T cells have not been investigated. Unique to HBVpol387- and HBVcore169-specific CD8+ T cells is the higher expression of TIGIT compared with other HBV-specific CD8+ T cells (fig. S7), as well as elevated programmed cell death protein 1 (PD-1) expression on HBVcore169-specific CD8+ T cells (fig. S7).

Unsupervised high-dimensional t-distributed stochastic neighbor embedding (t-SNE) visualization (41) and Phenograph cellular clustering (42) were applied to describe the phenotypes of virus-specific CD8+ T cells across individuals from one large batch of samples run in parallel (Fig. 2). On the basis of the expression levels of markers indicative of T cell activation, differentiation, trafficking, and inhibitory receptors typically associated with T cell exhaustion, 19 cellular clusters were objectively identified and annotated (Fig. 2A).

Fig. 2 Multifactorial memory atlas of HBVpol387- and HBVcore169-specific CD8+ T cells linked to HBV clinical stages.

(A) Unsupervised Phenograph clustering of cellular subsets on all detected antigen-specific CD8+ T cells across patient groups. n = 20, 4 patients per group. Nineteen cellular clusters objectively identified by Phenograph were color-coded as indicated, and the expression levels of probed cellular proteins are shown. (B) Visualization of the Phenograph clustering of nine major cellular clusters of HBVpol387-specific CD8+ T cells. The proportion of cellular clusters within HBVpol387-specific CD8+ T cells in individuals across various patient groups is shown. (C) Same analytical strategy for HBVcore169-specific CD8+ T cells. (D) Bar graph indicates the discrepancy of T cell memory–associated markers (CD27, CD28, CD45RO, CD127, and CXCR3), inhibitory receptors (PD-1 and TIGIT), and CD57 expressed on HBVcore169-specific CD8+ T cells. Error bars are median and range, and values from individuals were imposed. (E) Representative contour plots show the expression level of markers on HBVcore169-specific CD8+ T cells between patient groups. Patients were color-coded as indicated. (F) Top: Logistic regression of eight phenotypic markers showing that significant difference between patient groups was stacked against pseudotime imputed using Scorpius. Bottom: Logistic regression (black solid line) was used to visualize the trend of these eight cellular markers. Dots are individuals color-coded by clinical stages and showed the expression levels of these markers on HBVcore169-specific CD8+ T cells.

From this analysis, remarkable heterogeneity of HBVpol387-specific CD8+ T cells was observed. In many instances, several distinct populations specific for this one epitope could be seen even within individual patients. Despite such diverse phenotypes, this epitope sequence was highly conserved across all patients and time points subjected to HBV viral sequencing over a decade (fig. S8 and table S4). Quantification of cellular clusters within each T cell antigen specificity was performed across all batches of experiments using cluster-specific gating strategies (fig. S9) to test for compositional differences associated with the status of HBV infection. Besides cluster 8 (C8), all of the cellular clusters occupied by HBVpol387-specific CD8+ T cells expressed 2B4 with heterogeneous phenotype indicative of different T cell memory status, whereas only C9 showed elevated expression of PD-1 (Fig. 2). In terms of relationships with the status of infection, we observed a significant enrichment of cells with C13-like phenotypes (~80%) in IT patients, expressing CXCR3+CD27hiCD127hi (Fig. 2, A and B, and fig. S10, A to C). Significant differences were also seen for C8 in IA patients, which was highlighted by the coexpression of CD45ROhiCCR4+HVEM+ with a central memory T cell (TCM) phenotype (43). The phenotypically similar region C6 + C17 was preferentially occupied by InA, which were notably absent from IT patients (Fig. 2B and fig. S10, A to C). Such subset was similar to terminal effector memory RA (TEMRA) (43) but expressed CD127Int, suggesting an ongoing T cell response for InA but not IT patients. Perhaps related to their variable expression of coinhbitory receptors, expansion of in vitro stimulated HBVpol387-specific CD8+ T cells was moderate and only observed for a minority of patients tested (fig. S6).

HBVcore169-specific CD8+ T cells, composed of seven peptide sequence variants (Fig. 1C and table S1), had significantly greater frequencies (Fig. 1D) and pMHC tetramer staining intensity in patients with viral control (InA and R) (fig. S10D). No T cells specific for this epitope were detected in IT patients or HDs (Fig. 1D), except that one IT patient that had detectable cells (0.00193%) just below the imposed cutoff frequency (0.002%) (fig. S10E). t-SNE and Phenograph analysis showed that C8 and C11 were significantly enriched in InA and R, whereas C1 and C4 tended to be more prevalent in IA and InA patients. This is in line with hierarchical clustering (Fig. 1E) that segregated individuals into different clinical stages based on the phenotypes of HBVcore169-specific CD8+ T cells (Fig. 1E), which was not observed for EBV-specific CD8+ T cells analyzed in parallel (fig. S10F). In contrast to the elevated expression of CD57, PD-1, and TIGIT seen for these cells in IA patients, HBVcore169-specific CD8+ T cells from InA and R patients significantly expressed CD27, CD28, CD45RO, CD127, and CXCR3 (Fig. 2, D and E), suggesting that they were long-lived memory T cells that were associated with a high degree of viral control. However, these cells from R patients expressed relatively high levels of PD-1 and TIGIT despite their presumed clearance of virus and differed in the expression of other memory-associated markers and CD57 compared with IA patients. The PD-1–expressing HBVcore169-specific CD8+ T cells from R patients exhibited high level of interleukin-7 receptor (IL-7R) (CD127), indicating that they were not TEX (13, 14, 16) but likely were long-lived memory cells. It is plausible that such PD-1 expression denoted a sign of activation (44, 45) or adaptation (28), rather than exhaustion, and somehow related to their relatively high affinity for peptide-MHC (fig. S10D) (46, 47). It is also possible that the R patients were not completely cleared of the virus, and these PD-1–expressing HBVcore169-specific CD8+ T cells still actively inhibit the virus (48). In contrast, HBVcore169-specific CD8+ T cells from InA patients displayed intermediate levels of CD127 and CXCR3 and similar levels of CD27, CD28, and CD45RO compared with R patients. These cells in InA also had diminished expression of PD-1 compared with other groups (Fig. 2, D and E). This can be explained by the low to undetectable viral load during the stage of InA, leading to lesser TCR–viral antigen engagement compared with IA patients who had high viral load (46, 47).

Last, we applied Scorpius (49), a trajectory inference method, to compute the trajectory of HBVcore169-specific CD8+ T cells across three clinical stages using the patient-wise expression of eight phenotypic markers that showed statistical significances (Fig. 2F). Our analysis indicated that decreased expression of PD-1, TIGIT, and CD57, together with increased expression of CD27, CD28, CXCR3, CD45RO, and CD127, was associated with the inferred status of infection. Furthermore, patients associated with viral control (InA and R) were separated toward the end of the trajectory, whereas IA patients were on the opposite side. Thus, we demonstrated the highly heterogeneous HBV-specific CD8+ T cells during the progression of CHB using several high-dimensional analytical approaches, and such multifactorial cellular responses targeting HBVpol387 and HBVcore169 were able to delineate patients into their clinical stages.

Multifactorial interrelations of inhibitory receptors on virus-specific T cells

We next aimed to evaluate the relationships between various categories of cellular markers expressed by each of the antigen-specific T cells analyzed, with a special focus on nine different inhibitory receptors (16). To directly assess these relationships, one-dimensional soli-expression by nonlinear stochastic embedding (One-SENSE) (50) was used. One-SENSE works by reducing dimensionality of each category of markers into a one-dimensional (1D) t-SNE map that can be plotted in conjunction with alternative categories of markers mapped into additional 1D t-SNE maps. In this way, cells are separately arranged on the basis of their categorical expression, and then relationships between the categories can be intuitively visualized and described.

In this case, the categories we assigned (table S2) were “Differentiation + TNFR” (markers of differentiation and tumor necrosis factor receptor superfamily), “Inhibitory” (inhibitory receptors), and “Trafficking” (chemokine receptors). The three derived axes called out the cellular subsets objectively with all possible protein coexpression (Fig. 3A). We plotted and compared the profiles of T cells specific for four different epitopes, including HBVpol282, HBVpol387, and HBVcore169, and one that was derived from EBV (EBVEBNA3B). In general, of the observed combinations of inhibitory receptors expressed by the antigen-specific T cells across patients, one subset displayed HVEMint2B4+TIGIT+CD160+PD-1lo and was mainly contributed by subpopulations of HBVpol387- and EBVEBNA3B-specific CD8+ T cells. One-SENSE analysis also showed that 2B4 and HVEMlo were expressed by most of the antigen-specific T cells but limited expression of LAG-3, TIM-3, and CTLA-4 (Fig. 3A and fig. S7). In addition, we found that most of the PD-1+ cells expressed 2B4 and TIGIT but not CD160.

Fig. 3 Unsupervised analyses uncovered the complex model of inhibitory receptors (exhaustion markers) in CHB.

(A) One-SENSE objectively related three different T cell categories (Differentiation + TNFR, Inhibitory, and Trafficking) in 2D plots with visualization of the expression levels of cellular proteins. Dots are selected virus-specific CD8+ T cells (color-coded). Boxes annotated the epitopes that were enriched in the given regions. (B) Average fractions of the numbers of coexpressed inhibitory receptors on four HBV-specific CD8+ T cells across patient groups. Plots were from a representative experiment with all nine inhibitory receptors. (C) Average coexpressed inhibitory receptors on four HBV-specific CD8+ T cells across patient groups. Plots were composed of three experiments with simultaneous measurements of eight inhibitory receptors (without TIGIT). Each dot represents an individual.

By plotting the cells based on the patient groups, we noted that the phenotypes of HBVpol387- and HBVcore169-specific CD8+ T cells (blue and red) were most heavily influenced by status of infection, and the distinguishing features of these highly diverse cells are labeled. This heterogeneity can be best presented by HBVpol387- and HBVcore169-specific CD8+ T cells. HBVpol387 from IT had a relatively homogeneous phenotype in terms of memory-associated and trafficking receptors but varied in four distinct coexpressions of inhibitory receptors (Fig. 3A). In other clinical stages, greater diversity was observed for HBVpol387-specific CD8+ T cells in terms of memory- versus effector-associated markers, and these also had complex relationships with the patterns of inhibitory receptors coexpressed (Fig. 3A). Similar examination of HBVcore169-specific CD8+ T cells was limited because of the low cell numbers but also showed a greater degree of heterogeneity than might be expected. Nonetheless, this representation was consistent with that described above (Fig. 2).

We further quantified the numbers of coexpressed inhibitory receptors on these cells using a Boolean gating strategy (Fig. 3, B and C). In line with One-SENSE analysis, we failed to detect any HBV-specific cellular subsets that accumulated all inhibitory receptors (Fig. 3, A to C, and fig. S11). Although the different coexpression of inhibitory receptors on HBVpol387-specific CD8+ T cells can be shown by One-SENSE, there were no differences in terms of the numbers accumulated on the cells across patient groups. Conversely, IA patients had significantly higher numbers of inhibitory receptors on HBVcore169-specific CD8+ T cells compared with patients with viral control (Fig. 3C). Together with the visualization of One-SENSE, our data suggested a highly heterogeneous antigen-specific T cell phenotypes rather than a simple accumulation of so-called exhaustion markers during CHB, and the diverse coexpression of inhibitory receptors had different relationships with cellular differentiation and trafficking profiles that may associate with disease stages.

Functional capacity of HBV-specific CD8+ T cells

To address the relationships of functional capacity and inhibitory receptors, we pulsed and expanded patient’s cells using short-term in vitro peptide stimulation and then assessed their functional responses using intracellular cytokine staining (fig. S6). One-SENSE analysis was used to delineate virus-specific CD8+ T cells into five major heterogeneous functional subsets on the basis of the relationships between categories: “Functions”, “Inhibitory”, and “Differentiation + TNFR” (Fig. 4A, movie S1, and table S2). For each category (axes of One-SENSE plots), all possible cellular subsets were described by heatplots and descriptive labels. Biaxial plots of the most relevant markers for these five functionally distinct subsets were also represented (Fig. 4B), and the relative composition of these subsets was quantified (Fig. 4C). Overall, this analysis highlights nonlinear relationships between inhibitory receptors and functional capacity on virus-specific CD8+ T cells upon antigen recall in CHB.

Fig. 4 Nonlinear correlations of multifunctionalities and inhibitory receptors by One-SENSE.

(A) Patient’s PBMCs were stimulated with corresponding viral peptides for 10 days of in vitro culture to measure the functional capacity. Categorical (Function, Inhibitory, and Differentiation + TNFR) analysis of One-SENSE revealed the diverse multifunctional virus-specific CD8+ T cells subsets and their corresponding coexpression of inhibitory receptors. Dots are different virus-specific CD8+ T cells as annotated. Five different major functional subsets were labeled on the basis of the aligned heatplots and color-coded as indicated. (B) The expression levels of T cell functions, inhibitory receptors, and TNFR costimulatory receptors were compared between these five functional subsets. GM-CSF, granulocyte-macrophage colony-stimulating factor. (C) Bar graphs showed the proportion of each functional subset in HBVpol387- and HBVcore169-specific CD8+ T cells across patient groups. n = 5 per group, except for IT = 4.

Regardless of patient groups, a multifunctional subset was present (Fig. 4A, green box) and mainly contributed by HBVpol387- and HBVenv304-specific CD8+ T cells expressing MIP-1β+GrzA+GrzKloPerforin+ but CD107alo (Fig. 4, A and B, and fig. S12A). Another population of these cells (blue box) with otherwise similar phenotypic profiles was distinctly less able to produce cytokines but was GrzA+GrzK+PerforinInt without the degranulation marker CD107a, suggesting an alternative form of dysfunctional T cells that was associated with the expression of 2B4 and TIGIT but not PD-1 (Fig. 4, A and B, and fig. S12B). These cells were mostly composed of HBVenv304-specific CD8+ T cells from CHB patients (Fig. 4A and fig. S12A), which had expanded greatly in response to in vitro peptide stimulation (fig. S6). Unlike the other inhibitory receptors, CD160 and HVEM were less expressed by cells producing effector functions upon TCR stimulation. The sustained HVEM was mostly expressed by the nonfunctional (black box) subset of HBVpol282-specific cells (Fig. 4, A and B, and fig. S12, A and B). Such naïve-like and unresponsive T cells present a different type of dysfunctional T cells, perhaps associated with their expression of BTLA and CD160 before antigen recall (fig. S7) (51, 52). We also found that a fraction of HBVenv304-specific CD8+ T cells from R patients had similar functional profiles as EBV and influenza A virus (IAV)–specific CD8+ T cells that were PD-1LAG-3TIM-3lo, expressing IFN-γ+TNF-αhiMIP-1βhiGM-CSFhi (yellow box) (Fig. 4A and fig. S12, A and C). Moreover, HBVcore169-specific CD8+ T cells were in the unique plurifunctional (red box) subset coproducing various noncytolytic (8, 9) and cell-recruiting factors (GrzAGrzKIFN-γ+TNF-αloMIP-1β+GM-CSFintCD107a+) despite reciprocally coexpressing five inhibitory receptors including PD-1 (Fig. 4, A and B). This subset exhibited high levels of TNFR costimulatory receptors (OX40, GITR, 4-1BB, and CD27), suggesting greater activation and memory status (53). In this analysis, the major differences observed between patient groups were in the profiles of HBVcore169-specific CD8+ T cells. Patients with better viral control (R > InA > IA) displayed significantly higher frequencies of plurifunctional subset of HBVcore169-specific CD8+ T cells. In contrast, the opposite trend was observed for the frequencies of HBVcore169-specific CD8+ T cells within the MIP-1β+ multifunctional subset (Fig. 4C).

Together with the abovementioned data, we concluded that the immune responses of HBVcore169-specific CD8+ T cells were linked to viral control. Our analysis also displays complex nonlinear relationship between inhibitory receptors and functional capacity on virus-specific CD8+ T cells during CHB.

Clinical stage–dependent landscapes of virus-specific TCR

How TCRs are selected over the course of CHB is largely unknown. Hence, we sorted the pMHC tetramer–stained cells (fig. S5B) and sequenced the β chain of epitope-specific TCRs (HBVpol282, HBVpol387, HBVcore169, and HBVcore195) across various clinical stages. To map the TCR landscapes of these epitopes, we applied TCRdist, an algorithm that generates a distance matrix to quantify and obtain the relative motif similarity of TCR based on their sequences of amino acid (54). We used this TCRdist distance matrix to cluster similar TCRs using the unsupervised Phenograph clustering algorithm (42) and then visualized by the t-SNE dimensionality reduction algorithm (Fig. 5A). Thereafter, the sequence motifs that formed the basis of each TCR sequence cluster can be presented (Fig. 5B and fig. S13A), and the composition of these clusters in terms of the antigen specificity of each sequence derived from patients can be quantified (Fig. 5C). We found that TCR C15 and C27 were significantly increased in HBVpol282-specific TCRs compared with other epitopes (Fig. 5, C and D), suggesting that these motifs were important determinants for recognition of this HBV epitope and may associate with naïve-like phenotypes. In addition, the relative usage of each of these TCR sequence clusters differed significantly between patients grouped by status of HBV infection. Dominated by TRBV5-6, TCR C15 could be observed in all patients except for IA patients, which were instead enriched for the TRBV3-2+ TCR C12. TCR C27 was solely joined by TRBJ2-6 with highly conserved CDR3 (fig. S13A). Clusters of TCR sequence usages were more diverse for T cells specific for HBVcore169 and HBVpol387. Nonetheless, HBVcore169-specific TCR sequences also differed between patient groups. The TRBV3-2+ TCR C12 (also enriched in HBVpol282-specific TCRs) was similarly enriched in IA patients within HBVcore169-specific TCRs. In addition, HBVcore169-specific TCRs of R and InA had significant higher usages of TRBV6-6+ C5 and TRBV28+ C9, respectively (Fig. 5C).

Fig. 5 Unsupervised quantifications of HBV-specific TCR were associated with disease stages in an epitope-dependent manner.

(A) TCRdist measurements of epitope-specific TCRs were clustered by unsupervised Phenograph analysis and then projected by t-SNE. Each dot represents one TCR clone. Twenty-eight TCR sequence clusters on t-SNE map were labeled. (B) Sequence motifs (dashed boxes labeled with size) of representative TCR sequence clusters. Average linkage dendrogram for each TCR in the given cluster was presented and color-coded by generation probability. TCR logos display the frequency of V and J segments with CDR3β sequence in the middle. Bottom bars are source regions as indicated. Light gray is V region. Red is N insertion. Black is D for diversity. Dark gray is J region. (C) Percentages of the receptor in total epitope-specific TCRs for 28 TCR sequence clusters. (D) Proportion of TCR C27 and C15 in four different epitope-specific TCRs. (E) TCRdiv diversity measures for each epitope-specific TCR across patient groups. (F) Stacked bar charts show the top 11 TCR clones in individual patient. The frequencies and sequences of public TCR clones were presented. (G) 3D PCA projection delineated patient’s clinical stage using the epitope-specific TCR repertoires, tetramer response, and cellular profiles from the same individuals. (H) Correlation between the frequency and TCRdiv diversity measures of HBVcore169-specific CD8+ T cells.

We calculated repertoire diversity and density for each TCR sequence from various cell populations using a TCR diversity metric (TCRdiv) (54). Various patterns of this measure were observed among the epitope and patient groups (Fig. 5E). Of note, the TCRdiv scores for HBVpol387-specific TCRs have striking similar pattern with the proportion of cellular C13 in HBVpol387-specific CD8+ T cells across patient groups (Fig. 2B), which may be attributed to their relative enrichment in IT and HD groups. Both epitope-specific and bulk TCRs from HDs had overall greater diversity (fig. S13B), which separated them from CHB patients at the molecular level. In contrast to the uniformity of total CD8+ T cells (fig. S13C), CDR3 lengths were skewed between epitopes and patient groups. Overall, these findings fit with the previously described phenotypic differences observed in different patient groups. Thus, our analysis indicated that the biased TCR repertoire usages during CHB were epitope and clinical stage dependent.

Shared HBVcore169-specific TCR clones in patients with viral control

Focusing on the HBVcore169-specific response and inquiring a recently curated TCR database (55), we discovered several previously unidentified public HBVcore169-specific TCRβ clones that were shared between individuals in a clinical stage–dependent manner (Fig. 5F). In particular, a general public clone, CASGDSNSPLHF, was within the top three TCR clones in all three InA patients tested. Two other special public clones, CASSGGQIVYEQYF and CSARGGRGGDYTF, were each identified in two InA patients. An additional special public clone, CASSQDWTEAFF, was found at low frequencies in two acute resolved patients. That these public TCR clones were not shared across patient groups further highlights differences in the qualities of T cell responses that occur in acute versus chronic viral infection. Failure to detect public clones in IA patients suggested that the presence of public TCRs were essential for HBV viral control. By using principal component analysis (PCA) to combine the characteristics of HBVpol387- and HBVcore169-specific TCR repertoire and cellular response in the same donors, patient’s clinical status can be delineated (Fig. 5G). Last, we found that the observed frequency of HBVcore169-specific CD8+ T cells inversely correlated with their TCR repertoire diversity (Fig. 5H), suggesting the selective expansion of T cell clonotypes after viral clearance.

Phenotypic dynamics of HBV core169-specific CD8+ T cells

To assess the characteristic changes of antigen-specific T cells over the course of infection, we examined selected HBV epitopes in a longitudinally studied patient cohort who received entecavir (ETV) over the course of several years (n = 14, HLA-A*1101+ patients). ETV is a nucleotide analog that inhibits viral replication and leads to improved viral control in most patients. Although it does not inhibit HBeAg production by infected hepatocytes, it can also lead to HBeAg seroconversion and established anti-HBeAg antibody (HBeAb) in some patients, and this is a serological marker of further improved viral suppression. For 10 of 14 patients who had detectable HBVcore169-specific CD8+ T cells (Fig. 6A), we compared the profiles of these cells in patients who later lost HBeAg and produced HBeAb (HBeAg, n = 6) over the course of the study versus those who did not (HBeAg+, n = 4). Consistent with a previous report (56), decreased frequencies of HBVcore169-specific CD8+ T cells were observed in some patients (Fig. 6B and fig. S14A). Further experiments were performed to dissect the longitudinal viral mutation and tetramer response on this given epitope (fig. S14B), showing that these HBV-specific CD8+ T cells can recognize different variants beyond the database consensus peptide sequence. On the basis of the analysis of these cells at multiple time points, the detailed features could be tracked over time and major changes in the phenotypes of these cells were often observed (Fig. 6C). For instance, at early time points, HBVcore169-specific CD8+ T cells from patient HBeAg+03 (a patient who did not lose HBeAg during the study time frame) displayed a terminally differentiated effector phenotype (CD57+CD45RA+CCR7), whereas, at later time points, more than half of this phenotype was shifted. In contrast, HBVcore169-specific CD8+ T cells from a HBeAg patient (HBeAg01, a patient who lost HBeAg and established HBeAb) displayed a memory T cell phenotype (CD27+CD127+CD45RO+) at all time points tested that was maintained for more than 6 years after treatment. Similar trends were observed for the other patients studied, and detectable HBVcore169-specific CD8+ T cells were described quantitatively (Fig. 6D).

Fig. 6 The phenotypic dynamics and the machine learning–aided modeling of HBVcore169-specific CD8+ T cells.

(A) A total of 14 patients (n = 8 for HBeAg and n = 6 for HBeAg+) were included in the longitudinal cohort. The average frequency of all detectable HBVcore169-specific CD8+ T cells across different time points was shown. Each dot represents one patient who had detectable HBVcore169-specific CD8+ T cells. (B) Dynamics of HBVcore169-specific CD8+ T cells in two representative patients. (C) Phenotypic dynamics of HBVcore169-specific CD8+ T cells using One-SENSE. Numbers are frequencies, and boxes are annotated as indicated. (D) Fractions of memory (blue boxes) and terminal effector (red boxes) cells in individual patients across longitudinal time points. (E) Plot showed the changes of selective cellular marker expression on HBVcore169-specific CD8+ T cells across patient’s longitudinal time points [early and late, thick stacked bars in (D)]. Two time points (early and late) were picked to roughly match the time points between patients based on the drug intervention. “Early” are pretreatment time points besides one patient (HBeAg+04, whose earliest time point was 3 months after treatment), and “Late” are roughly 30 months after treatment. Plots showed the patients who had detectable HBVcore169-specific CD8+ T cells in both early and late time points. Statistical analysis was used to compare the cellular marker expression between two time points (early and late, solid lines) or patient groups (HBeAg+ and HBeAg, dashed lines). (F) Logistic model (dashed gray lines) of cellular marker expression (dependent variable) against SVM-predicted pseudotime (independent variable). Dots represent the expression levels of cellular markers on HBVcore169-specific CD8+ T cells across different patient’s longitudinal time points. (G) Statistical analysis of SVM-predicted pseudotime during the progression of patient’s longitudinal time points. A nonparametric paired t test was used.

In addition to summarizing the composition of these cells in terms of their memory versus effector phenotypes over time (Fig. 6D), we focused on the same markers we had found that vary the most across patients in our cross-sectional cohort (Figs. 1E and 2, D and E). In general, the fraction of memory cell subset in HBVcore169-specific CD8+ T cells was associated with lower HBeAg level over time (Fig. 6D). This was supported by the increased expression of T cell memory–associated markers (CD27, CD127, CD45RO, and CXCR3) (Fig. 6E) and the decreased expression of CD57 on HBVcore169-specific CD8+ T cells, in line with similar observation in InA and R patients from the cross-sectional cohort (Figs. 1E and 2D). In patients who sustained high HBeAg level, HBVcore169-specific CD8+ T cells displayed significantly lower levels of CD57 and higher levels of CD27 and CD127 after the virus was suppressed, whereas the cells derived from HBeAg patients already had such a cellular profile (CD57loCD27hiCD127hiCD45RO+) before the treatment, and this was maintained for several years. To further quantify these changes, we applied trajectory analysis using Scorpius to these samples. To examine the reproducibility of the trajectory detection in this longitudinal cohort as it compared with the cross-sectional cohort (Fig. 2F), we used support vector machines (SVMs) to map data from the longitudinal samples onto the pseudotime metric developed using Scorpius from the cross-sectional cohort (Fig. 2F; see Materials and Methods). That is, using the data from the cross-sectional cohort as a training set, we computed the pseudotime across patient’s time points in the testing set (longitudinal cohort; Fig. 6F). Of these seven cellular markers, the trajectories of HBVcore169-specific CD8+ T cells between the two independent patient cohorts were consistent (even when Scorpius was run independently without the use of SVM). In addition to validating the trajectory model on this independent cohort, this allowed us to test the hypothesis that the HBV-specific T cell phenotypic evolution would track with the extent of viral control. The phenotypes of HBVcore169-specific CD8+ T cells did show the expected progression along this pseudotime metric over the course of treatment for all patients (Fig. 6G). This implies that the virus-specific T cell response associated with viral control improved for each patient over the course of antiviral therapy, as would be expected. In addition, patients who lost HBeAg and established HBeAb had a more progressed (higher expression of T cell memory markers and lower PD-1 and TIGIT expression) phenotype at earlier time points than non-HBeAg seroconverting patients, suggesting that these patients had better virus-specific T cell response at the start of treatment. Thus, our data showed that HBVcore169-specific CD8+ T cells expressing increased cellular markers associated with long-term T cell memory development but decreased CD57 and two inhibitory receptors, PD-1 and TIGIT, were linked to viral control, and such machine learning–aided model could have predictive value for prognosis in CHB.


By fully leveraging a highly multiplexed combinatorial pMHC tetramer staining strategy, mass cytometry, and unsupervised high-dimensional analyses, we investigated 562 unique A*11:01-restricted candidate epitopes during the progression of HBV. Analysis of HBV-specific T cell responses is difficult because of the very low frequencies of these cells. In this regard, we show the importance of investigating both the specificity and phenotypic profiles of the antigen-specific T cells to verify their involvement in the HBV-specific immune response. Beyond this, our data highlight the heterogeneity of the virus-specific T cell response that was associated with disease stages and provide quantifiable analyses of HBV-specific TCRβ repertoires that corresponded to cellular phenotypes during chronic viral infection.

Host defense against HBV weighs on immune response driven largely by virus-specific T cells (6, 8). The number of A*11:01-restricted epitopes detected by this comprehensive approach was relatively limited compared with the reported epitopes in the context of A*02:01. It is possible that some epitope-specific T cells were only detectable in the liver but not in the periphery. Future investigation on HBV-specific intrahepatic lymphocytes is needed. The presence and frequencies of well-described A*02:01-restricted HBVcore18–27-specific CD8+ T cells (11, 12, 35, 57) have been shown to associate with viral control. Many therapeutics have been therefore developed on the basis of this T cells, including the blockade of overexpressed PD-1 to reinstate T cell function (5860), adoptive transfer of engineering virus-specific T cells (61), and TCR-like antibody (62) to deliver interferon-α (IFN-α) directly onto infected hepatocytes. Here, we presented evidence that the specific responses and characteristics of A*11:01-restricted HBVcore169-specific CD8+ T cells were linked to viral control, with public TCR clones used by these T cells. Comparative analysis showed that these cells had differing profiles across clinical stages. Furthermore, high-dimensional trajectory analysis allowed us to use the profiles of HBVcore169-specific CD8+ T cells to assign each patient a value along an objective pseudotime metric. The underlying features of HBVcore169-specific T cells along this trajectory were consistent across two independent patient cohorts, both showing correlations with viral control. On the basis of this metric, it is also conceivable that these antiviral-treated patients and having T cells with the most progressed features of viral control could be eligible for safe discontinuation of antiviral drug once they mounted HBVcore169-specific memory T cell response, in line with our recent report showing the predictive utility of HBV-specific T cells (63). This is important because even the most advanced serological measures are unable to accurately predict such outcomes. Collectively, our findings should affect HBV immunotherapy design and could be useful to predict patient’s clinical outcome based on the phenotypic response of HBVcore169-specific CD8+ T cells. We also anticipate that the utility of this approach could be extended to other epitopes associated with viral control derived from HBV core or other proteins that are restricted to other HLA alleles.

By comprehensively probing HBV epitopes on numerous HBV-infected patients, we failed to identify TEX expressing all inhibitory receptors or overt evidence for hierarchical T cell exhaustion. Instead, unsupervised visualization using One-SENSE showed complex nonlinear relationships between the expression of inhibitory receptors. Moreover, the dysfunctionality of HBV-specific T cells did not correlate to the linear accumulation of inhibitory receptors, indicating that these cells were not completely functionally inert (16). One interpretation is that these HBV-specific T cells do not fit the definition of TEX as reported in LCMV-specific T cells, but instead a different type of subset that was mostly absent, with the remaining dysfunctional T cells expressing various combinations of inhibitory receptors. On the basis of our data, we proposed that these T cell profiles, at least in the peripheral blood, could fit better with the description of functional adaption (28) in CHB. Nonetheless, our functional assessment relied on in vitro peptide stimulation because of the rare detection of HBV-specific T cells, and this might limit its relevance to in vivo response. Unlike during chronic LCMV infection, where maintenance of the TEX phenotype requires persistent and high antigen level (13, 20, 64), HBVcore169-specific CD8+ T cells expressing high level of CD27 and IL-7Rα (CD127) found in InA patients were not TEX and are likely to be maintained for decades with a limited amount of viral antigen (65, 66). HBV-specific T cells from these patients were PD-1intTIGITint with elevated expression of memory-associated markers and functional capacities that seem to be associated with control of the virus. On the other hand, for IA patients who have high and fluctuating viremia, HBV-specific T cells were detected that better matched the expected features of TEX. These included a strong coexpression of PD-1 and TIGIT and limited expression of memory-associated markers such as CD127. In addition, we also found substantial expression of CD57 on HBV-specific T cells from IA patients, a cellular marker that is indicative of more differentiated effector cells with low proliferative capacity, suggesting that these cells were not long-lived memory cells. Overall, cellular profiles of HBV-specific T cells in IA stage are consistent with persistent high antigen exposure. At later stages (InA) in patients with better viral control, HBV-specific T cells had a memory phenotype expressing lower CD57 but elevated CXCR3, CD45RO, CD27, and CD127. The lack of the detectable HBVcore169-specific CD8+ T cells in IT patients who have high viremia suggests that they might be largely deleted, and the absence of such particular T cells could contribute to the minimum liver inflammation in this stage. Future studies involving larger volume of blood samples from IT patients or using more sensitive approaches may help to address this issue. Further investigation of the expression level of EOMES and T-bet (16, 18) or the epigenetic modification (19) that better defined the bona fide TEX is important to address this aspect in CHB. It is also important to note that this report is limited to the analysis of circulating HBV-specific CD8+ T cells, and further examination of the exhaustion profile in intrahepatic lymphocytes is needed to address this question.

Despite the TCR sequence diversity of T cells specific for the HBVcore169 epitope, several public clones at relatively high abundance were detected in multiple patients. In line with the CMV-specific TCR repertoire (67), we noted that these previously unidentified public TCRs were different when derived from patients showing viral control (R and InA) versus viremia (IA), suggesting the functional importance of these T cells. The public virus-specific TCR clones may be selected over the course of viral clearance (i.e., from IA into InA) and contraction of the effector response (27). As previously reported for CMV and EBV infection (68, 69), virus-specific CD8+ T cells do not express IL-7Rα (CD127) until T cell memory had been established, and such selection is thought to be driven by high-affinity TCR–viral epitope binding (68). This is consistent with the characteristics of HBVcore169-specific CD8+ T cells in InA and R patients who carried public TCRs and elevated expression of T cell memory–associated markers including CD127, indicating their long-lived and self-renewing ability to maintain memory T cells pool after the reduction of viral antigen (65, 70). Analogously, intrahepatic and peripheral public TCR clones have been linked to viral clearance in HCV-infected chimpanzees (71).

Despite the challenges we have highlighted associated with detecting HBV-specific T cells because of their low prevalence, our report explores the previously unappreciated complexity of virus-specific T cells in lifelong human HBV viral infection. The cellular responses of HBVcore169-specific CD8+ T cells and TCR sequences used were associated with the status of HBV infection and could be used as an indicator of the relative extent of viral control. Thus, the results provided here could have important implications for the development of new biomarkers, treatment strategies, and immunotherapy aiming at HBV cure.


Patient samples and PBMC isolation

Patients (table S3) with HBV infection were recruited with fully informed written consent from the Division of Gastroenterology and Hepatology at National University Health System, Singapore. The respective local ethical institutional review boards approved the study, and the recruitment and sampling of suitable patients were completed at hospital. Up to 60 ml of blood was taken, and peripheral blood mononuclear cells (PBMCs) were further isolated by using Ficoll separation (Ficoll-Paque PLUS, GE Healthcare). All patients had clinical, serological, and virological evidence of chronic hepatitis B infection with detection of HBsAg and HBV DNA, and no positive result for the presence of HIV-1, HIV-2, and HCV. Three CHB patient groups—IT [HBV DNA > 2000 IU/ml, alanine aminotransferase (ALT) < 40 IU/ml, HBeAg+), IA (HBV DNA > 2000 IU/ml, ALT > 40 IU/ml, HBeAg+], and InA (HBV DNA < 200 IU/ml or undetectable, ALT < 40 IU/ml, HBeAg)—and one group of acute resolved patients (R, undetectable HBV DNA, HBsAg, and anti-HBcAb+) were enrolled in this study. Each CHB patient had at least three adjacent time points indicating consistent virological and serological evidence for the referring clinical stages. All patients were treatment free from any antiviral drug or clinical intervention at the time of blood draw. Patients who received ETV were followed up longitudinally and enrolled. Serological and virological scores (serum HBV DNA, HBeAg, and HBsAg) and liver function test were determined by clinical laboratory assay or enzyme-linked immunosorbent assay. Blood samples from anonymous HDs were recruited under Singapore Immunology Network (SIgN) institutional review board. Healthy CB samples were purchased from Singapore Cord Blood Bank under the institutional review board, without detection of HIV-1, HIV-2, human T-lymphotropic virus (HTLV)–I, HTLV-II, HCV, CMV, HBsAg, and anti-HBcAb. HLA-A*11:01 was confirmed by typing service from BGI Genomics.

562-Plex combinatorial (quadruple/triple SAv-metal–coded) pMHC tetramers

Fourteen different SAv-metals were made by labeling streptavidin with 14 different metal isotopes. Similar to a previous report (30), each SAv-metal was diluted into 20 μg/ml in EDTA-free W-buffer on the same day of tetramerization of pMHC. Two different configurations of quadruple SAv-metal coding for 562-plex pMHC tetramers were generated using an R-based script for a 14-choose-4 scheme (for 1001 combinations). The script was then loaded onto a TECAN Freedom EVO200 automatic liquid distribution robot to prepare the designed combinations of quadruple SAv-metal mixtures (each mixture contains four different SAv-metals) in 2-ml 96-well deep-well plates. To form pMHC tetramers, each peptide-loaded HLA-A*1101 monomer (562 different pMHC monomers) was randomly assigned for four different SAv-metals. To reach a 1:4 ratio of streptavidin to pMHC, quadruple SAv-metal mixtures were added to the corresponding pMHC monomer in a stepwise manner of four additions; each has 10 min of incubation at room temperature. d-biotin (10 μM) was added into the reaction at the end for another 10 min at room temperature to saturate unbound streptavidins. The 562-plex pMHC tetramers were combined and concentrated down to 5 μg/ml per pMHC tetramer in 10% fetal bovine serum (FBS) CyFACS buffer using an Amicon 50-kDa cutoff concentrator (Millipore). The total amount of protein in each concentrator was limited to 300 μg. To reach the desired concentration and volume, multiple spins were performed at 700g for 5 min. The 562-plex tetramers were then filtered by a 0.1-μm filter tube (Millipore) at 2000g for 25 min. The cocktail of tetramers was kept on ice and then spun at 14,000g for 1 min in a 1.5-ml Eppendorf tube to remove the remaining aggregates before the staining. The combinatorial streptavidin codings were rescrambled for every independent 562-plex combinatorial pMHC tetramer staining experiment.

For selected experiments, a 9-choose-3 (84 combinations) or 8-choose-3 (56 combinations) scheme was used to cover 120-plex (40 peptide clusters) or 50-plex (17 peptide clusters) combinatorial triple-coded pMHC tetramers staining preferentially selected (table S1) for more phenotypic analysis or in vitro peptide stimulation (table S2).

Highly multiplexed pMHC tetramer, antibody staining, and CD8 T cell enrichment

Cryopreserved PBMCs were thawed and washed with complete RPMI (10% FBS + 1% penicillin/streptomycin/l-glutamine +1% 1 M Hepes) (Gibco, Invitrogen) and rested for 3 hours at 37°C. After the recovery, cells were harvested and seeded on a nontreated 96-well plate, and about 10 million cells per patient were used and split evenly in two separated wells for two configurations of 562-plex combinatorial pMHC tetramer staining. Dasatinib (50 μM) was incubated with cells for 30 min at 37°C, 5% CO2, to prevent the down-regulation of TCR (30). Cells were washed with CyFACS buffer [2 mM EDTA + 0.05% sodium azide + 4% FBS in phosphate-buffered saline (PBS)] and incubated with 200 mM cisplatin (Pt-195) for 5 min on ice or with rhodium (Rh-103) for 20 min at room temperature (table S2) for viability measurement. After washing once with CyFACS buffer, cells from the same donor in separated wells were stained with 50 μl of cocktail containing the same 562-plex pMHC tetramers but completely different SAv-metal coding configurations for 1 hour at room temperature in the presence of 1:100 Fc block (BioLegend). Cells were washed twice with CyFACS buffer after incubation and resuspended in 50 μl of T Cell or CD8 T Cell Enrichment Kit (STEMCELL) antibody cocktail in 1:10 in CyFACS buffer for 30 min on ice. Cells were then washed and stained with 50 μl of primary antibody cocktail (table S2 and fig. S15) for 30 min on ice. Excessive antibodies were removed by washing the cells twice with CyFACS buffer, and cells were resuspended with 4 μl of enrichment beads (STEMCELL) and 46 μl of CyFACS buffer for 15 min on ice. After the staining, cells were washed with PBS and fixed with 200 μl of 2% paraformaldehyde (PFA; Electron Microscopy Sciences) overnight at 4°C. On the next day, PFA was removed and cells were incubated with permeabilization buffer (BioLegend) at room temperature for 10 min and then resuspended with 50 μl of intracellular antibody cocktail for 30 min at room temperature. For subsequent dual mass-tag cellular barcoding, 2 mM bromoacetamidobenzyl-EDTA (BABE; Dojindo) with 0.5 mM PbCl2 was dissolved in Hepes buffer, and each sample was given a unique combination of metal barcode (BABE–Pd-102, BABE–Pd-104, BABE–Pd-106, BABE–Pd-108, and BABE–Pd-110) on ice for 30 min. After 5-min incubation with CyFACS buffer on ice, cells were labeled with an iridium DNA intercalator (Ir-191/193, Fluidigm DVS) in 2% PFA at room temperature for 20 min. Cells were then washed with CyFACS buffer, and CD8 T cells were negatively selected using EasySep Magnet (STEMCELL) according to the manufacturer’s instruction. Enriched cells were washed twice by Milli-Q water and were ready for mass cytometry acquisition.

Statistical analysis

Nonparametric analysis of variance (ANOVA) was used for group comparison unless indicated elsewhere. P < 0.05 by nonparametric ANOVA allowed the subsequent multiple comparison test. P values were calculated using Prism software (GraphPad). All error bars are median and SEM.


Materials and Methods

Fig. S1. Comprehensive epitope mapping strategy and experimental workflow.

Fig. S2. Quality and detection of antigen-specific CD8+ T cells using highly multiplexed combinatorial pMHC tetramer staining and mass cytometry.

Fig. S3. Overall magnitudes of antigen-specific CD8+ T cell response in various clinical stages during HBV infection.

Fig. S4. Validation of highly multiplexed combinatorial pMHC tetramer strategy in HLA-A*11:01 and non–HLA-A*11:01 donors.

Fig. S5. Validation and reproducibility of antigen-specific CD8+ T cells using flow cytometry and the serological measurement of HDs.

Fig. S6. In vitro expansion of antigen-specific CD8+ T cells upon peptide stimulation.

Fig. S7. Expression levels of nine different inhibitory receptors on antigen-specific CD8+ T cells.

Fig. S8. Epitope frequencies in longitudinal patient cohorts of HBeAg seroconverters.

Fig. S9. Cellular profiles of subset clusters of HBV-specific CD8+ T cells identified by Phenograph and the enrichment strategy.

Fig. S10. Unsupervised Phenograph clustering analysis identified multifactorial T cell heterogeneity of HBV-specific CD8+ T cells.

Fig. S11. Coexpression of inhibitory receptors on virus-specific CD8+ T cells.

Fig. S12. Heterogeneous multifunctional subsets of virus-specific CD8+ T cells.

Fig. S13. Diverse characteristics of epitope-specific TCRβ repertoire using TCRdist.

Fig. S14. Dynamics of cellular response and viral mutation of HBVcore169-specific CD8+ T cells in a longitudinal patient cohort.

Fig. S15. Staining quality of cellular markers including nine inhibitory receptors using mass cytometry.

Table S1. List of screened HLA-A*11:01–restricted epitopes and the detected frequency.

Table S2. List of the antibody staining panels used for mass cytometry and high-dimensional cytometric data analysis.

Table S3. List of the patient samples and the clinical and serological information.

Table S4. Frequency of viral mutation on selective epitopes in longitudinal patient cohort across HBeAg seroconversion.

Movie S1. 3D image of categorical analysis of HBV-specific CD8+ T cells using One-SENSE.

References (7279)


Acknowledgments: We are deeply indebted to J. Crawford from St. Jude Children’s Research Hospital for the help in installing TCRdist. We are grateful to W. W. Phyo, A. T. Y. Ling, and J. Wang from the National University Health System, Singapore, for their support and coordination of patient samples. We also thank all the members of Newell laboratory and the Flow Cytometry core facility in Singapore Immunology Network. Funding: This study was fund by the Translational & Clinical Research Flagship program (NMRC/TCR/014-NUHS/2015) from the National Medical Research Council of Singapore (M.L.H., A.B., S.G.L., and E.W.N.); Singapore Immunology Network core funding (E.W.N.) of Agency for Science, Technology and Research (A*STAR), Singapore; and the A*STAR Singapore International Graduate Award (SINGA) (Y.C.). The flow cytometry and CyTOF platforms are part of the SIgN Immunomonitoring platform, supported by BMRC (IAF 311006 grant and transition fund H16/99/b0/011). Author contributions: Y.C. and E.W.N. conceived the study, design the experiments, and wrote the manuscript. Y.C. made the protein, completed the experiments, and analyzed the data. Y.O.Z., P.A., and P.F.d.S. performed the viral sequencing. Y.O.Z. analyzed the viral sequencing data and wrote the manuscript. E.B. wrote the scripts for the t-SNE visualization of TCRdist-analyzed TCR sequencing data, the Scorpius trajectory, and SVM modeling and wrote the manuscript. J.C. and E.W.N. did the epitope prediction. A.B. and S.G.L. recruited and provided the patient samples. Y.C., M.P., M.L.H., A.B., S.G.L., and E.W.N. reviewed and edited the manuscript. E.W.N. supervised the study. Competing interests: Y.C. and E.W.N. are inventors on U.S. patent application (pending) held by A*STAR, which covers the uses of epitope sequences (HBVcore169) and epitope-reactive TCRs for the treatment and diagnosis against HBV infection. A.B. receives research resources for the collaboration with Gilead Sciences. A.B. is a consultant and member of the advisory board of Gilead Sciences, MedImmune, Janssen-Cilag, IONIS, Abivax, and HUMABS BioMed. A.B. is a cofounder of LION TCR Pte. Ltd., which develops engineered T cells for the treatment of virus-related liver cancers. S.G.L. received research grants from Gilead Sciences, Merck, and Abbott Diagnostics. S.G.L. served as the advisory board member of Gilead Sciences, AbbVie, Abbott Diagnostics, Merck, Springbank, and Roche. E.W.N. is a cofounder, shareholder, advisor, and on the board of directors of ImmunoSCAPE Pte. Ltd. Data and materials availability: The viral sequencing data have been deposited into NCBI BioProject with the submission number PRJNA479. The HBV-specific TCR sequencing data can be found on Requests for reagents should be addressed to enewell{at} and cheng_yang{at}

Stay Connected to Science Immunology

Navigate This Article