Heterogeneity of exhaustion profiles indicates bias towards severe exhaustion of HCV-specific CD8⁺ T cells

We set out to understand how metabolic profiles of HBV-specific and HCV-specific CD8⁺ T cell responses in chronic infection are connected to their exhaustion profiles. Virus-specific CD8⁺ T cells were identified by ex vivo tetramer staining which revealed a higher frequency of HBV core_18-27-specific compared with HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells (figure 1A). HCV-infected patients had more variable liver inflammation with more severe hepatitis identified by transaminase elevation compared with our cohort of HBeAg-negative cHBV patients, as expected (online supplemental tables 1 and 2).6 9 We characterised the expression of exhaustion markers that inform about early differentiated ‘progenitor’ exhausted T cells with homeostatic properties and more severely exhausted T cell populations.7 18 PD-1 expression was shared by HBV-specific and HCV-specific CD8⁺ T cells but expression of the interleukin 7 receptor α-chain (CD127) was significantly reduced in HCV-specific CD8⁺ T cells, resulting in an enrichment of the PD-1⁺CD127⁺ subset in HBV core_18-27-specific CD8⁺ T cells (figure 1B–D). Of note, HCV-specific CD8⁺ T cells also expressed higher levels of CD38, CD39 and CD57, indicating stronger activation and a more severely exhausted phenotype, while no major differences in TIGIT, CD27, CD28 and CD122 expression were observed (figure 1E–G). Together, these data indicate significant heterogeneity of the exhaustion profiles of HBV core_18-27-specific and HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells in our cohort.

Hepatitis virus-specific CD8⁺ T cells exhibit different metabolic programmes

Our exhaustion analysis indicated that the comparison of HBV core_18-27-specific CD8⁺ T cells and HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells also reflected a comparison between mild and severe exhaustion programmes. The comparison of HBV-specific and HCV-specific CD8⁺ T cell responses thus may not only reflect differences in viral aetiology but also serves as a model to study differences in T cell exhaustion programmes. To understand the metabolic profiles of virus-specific CD8⁺ T cells in chronic HBV and HCV infection, we first performed transcriptome analysis of sorted virus-specific CD8⁺ T cells in untreated chronic infection. Interestingly, gene set enrichment analysis (GSEA) of KEGG metabolic pathways indicated significant differences in the metabolic programmes of mildly exhausted HBV-specific compared with severely exhausted HCV-specific CD8⁺ T cells (figure 2A). In particular, the gene sets for OXPHOS and citric acid cycle (TCA cycle) were significantly enriched in HBV-specific CD8⁺ T cells, while expression of genes involved in glycolysis had a non-significant trend towards higher expression in HCV-specific CD8⁺ T cells (figure 2B). These transcriptional analyses suggested differences in energy metabolism of HBV-specific and HCV-specific CD8⁺ T cells. We thus explored the functional consequences of these altered transcriptomes and functionally interrogated T cell metabolism using metabolism-directed flow cytometry. Glucose uptake was measured using 2-NBDG assay. HBV-specific and HCV-specific CD8⁺ T cells incorporated more 2-NBDG than naïve T cells, in line with higher bioenergetic requirements of activated antigen-specific CD8⁺ T cells (figure 2C,D).

We next assessed the mitochondrial properties of virus-specific CD8⁺ T cells using several mitochondrial dyes that differ in their ability to stain mitochondria depending on their membrane potential. TMRE staining as a direct correlate of ΔΨm electronegativity was significantly higher in HBV-specific compared with HCV-specific CD8⁺ T cells, indicating elevated mitochondrial membrane potential (figure 2E). In contrast, elevated mitochondrial mass, as indicated by MitoTracker Green (MTG) staining was observed in HCV-specific CD8⁺ T cells despite similar total cellular uptake of the MitoTracker Deep Red (MTDR) that correlates with mass and potential (figure 2F,G). The MTDR/MTG ratio (informing about relative mitochondrial polarisation as another estimate of mitochondrial membrane potential) was reduced in HCV-specific CD8⁺ T cells, fitting to the TMRE results (figure 2H). We also observed higher mitochondrial ROS in HBV-specific and HCV-specific CD8⁺ T cells compared with naïve T cells, however, there was a clear enrichment of ROS in HCV-specific compared with HBV-specific CD8⁺ T cells (figure 2I). These data inform about a pronounced alteration of mitochondrial function in HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells and, together, indicated significant heterogeneity in the metabolic programming of hepatitis virus-specific CD8⁺ T cells during chronic infection.

To understand if these surrogates of mitochondrial function translate into a differential ability to respire, we performed Seahorse analysis of (n=16) in vitro expanded HBV-specific and HCV-specific CD8⁺ T cell lines and control cell lines. We observed significant metabolic heterogeneity across these cell lines (online supplemental figure 1A,B). Interestingly, HBV-specific CD8⁺ T cells had a higher spare respiratory capacity compared with HCV-specific CD8⁺ T cell lines (online supplemental figure 1C). In glucose-starved flux assays, there was a trend towards higher glycolytic capacity in HBV-specific CD8⁺ T cell lines (online supplemental figure 1D). The difference in the glycolytic capacity was significant in the flux analysis of non-starved cells (online supplemental figure 1E). These data suggest that a higher relative metabolic fitness of HBV-specific CD8⁺ T cells can be observed in vitro, correlating to the ex vivo profiling. The ability of HBV-specific CD8⁺ T cells to respond to oligomycin-mediated inhibition of ATP synthase with increased ECAR indicates higher metabolic flexibility to alternatively use glycolysis for ATP production (online supplemental figure 1F). This is in contrast to HCV-specific CD8⁺ T cells, which in some cell lines could not increase their ECAR at all, despite effects on OCR (online supplemental figure 1G,H). In summary, these experiments illustrate significant metabolic heterogeneity in HBV-specific and HCV-specific CD8⁺ T cells with a bias to more significant mitochondrial impairment in severely exhausted HCV-specific CD8⁺ T cells.

The mitochondrial profile of HBV-specific and HCV-specific CD8⁺ T cells is associated with their differential PD-1/CD127 expression

We next examined the extent to which the mitochondrial polarisation depends on the exhausted subpopulations in cHBV and cHCV infection. Analysis of mitochondrial mass, potential and polarisation between PD-1⁺CD127⁺ and PD-1⁺CD127^- virus-specific CD8⁺ T cells in cHBV and cHCV infection revealed significantly reduced mitochondrial polarisation in the more exhausted PD-1⁺CD127^- subset, independent of aetiology of viral infection (figure 3A–C). In line with this observation, linear regression analysis also revealed a direct correlation of mitochondrial polarisation and the PD1/CD127 phenotype. Mitochondrial polarisation (MTDR/MTG) inversely correlated with the PD-1⁺CD127^- subset frequency, in contrast to a positive correlation with the mildly exhausted subset (PD-1⁺CD127⁺) in HBV-specific and HCV-specific CD8⁺ T cells (figure 3D). These analyses demonstrate a connection of the mitochondrial polarisation with the distribution of different exhaustion subsets of HBV core_18-27-specific and HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells and suggest that the metabolic differences are connected to the differentiation of exhausted subsets rather than differences inherent to the different hepatitis viruses.

Antigen clearance improves mitochondrial fitness of HCV-specific CD8⁺ T cells linked to accumulation of PD-1⁺CD127⁺ populations

Since we had identified severely impaired mitochondrial metabolism in exhausted HCV-specific CD8⁺ T cells associated with a terminally exhausted phenotype (PD-1⁺CD127⁻) during chronic infection that is associated with high levels of antigen stimulation and liver inflammation, we sought to understand if direct-acting antiviral (DAA) therapy would reprogramme HCV-specific CD8⁺ T cell metabolism. Previous studies had shown controversial results on a bulk virus-specific analytic level after overnight culture.11 22 Therefore, we performed a longitudinal metabolic analysis of HCV-specific CD8⁺ T cells ex vivo in cHCV patients treated with DAA therapy. Interestingly, ex vivo analysis identified a significant increase of virus-specific cells at week 2 after therapy initiation (figure 4A), suggesting an augmented virus-specific response, that occurred when transaminase levels significantly declined (figure 4B).

Metabolic analysis revealed an increase in the polarisation of mitochondria after 2 weeks of therapy (figure 4C). Interestingly, we observed a negative association between mitochondrial fitness and ALT levels (figure 4D). Thus, our ex vivo analyses indicated a metabolic response to DAA therapy. We next wondered if the improved mitochondrial fitness was associated with a change in the distribution of exhaustion subsets. Indeed, our monitoring showed an enrichment of PD-1⁺CD127⁺ cells during DAA therapy (figure 4E). Clearly, the increased frequency of the PD-1⁺CD127⁺ subset correlated with higher mitochondrial polarisation (figure 4F). We, thus, wondered if DAA therapy augmented the metabolism in severely exhausted cells or if the increase in mitochondrial metabolism is due to an accumulation of the metabolically fit subset. We did not observe significant changes in mitochondrial metabolism in the PD-1⁺CD127⁺ cells during therapy (figure 4G–I). In contrast, there was an increase in polarisation in the remaining severely exhausted PD-1⁺CD127⁻ cells during therapy (figure 4G–I). The augmentation of mitochondrial metabolism in the severely exhausted cells compared with the PD-1⁺CD127⁺ subset resulted in a similar mitochondrial polarisation of the remaining PD-1⁺CD127⁻ cells at week 2 (figure 4J,K). In sum, our data link the enhanced metabolic fitness of HCV-specific CD8⁺ T cells after therapy to an expansion of the PD-1⁺CD127⁺ population and improvement of the metabolic fitness of PD-1⁺CD127⁻ cells.

The metabolic profile of HBV core_18-27-specific and polymerase_455-463 -specific CD8⁺ T cell responses significantly differs and correlates with their exhaustion phenotype

In chronic HBV infection, ex vivo HBV-specific CD8⁺ tetramer responses are preferentially identified in HBeAg-negative infection.6 Comparison of immunodominant epitope responses recently revealed phenotypic and functional differences between HBV core_18-27-specific and polymerase_455-463-specific CD8⁺ T cells in patients despite similar low viral loads and inflammation.9 We, thus, wondered if these responses differed in their metabolic profile. Interestingly, HBV polymerase_455-463-specific CD8⁺ T cells had significantly reduced mitochondrial polarisation and increased mitochondrial ROS production compared with HBV core_18-27-specific CD8⁺ T cells (figure 5A), resembling the metabolic programme enriched in severely exhausted T cells and typical for HCV infection. Consistent with our previous analysis, HBV polymerase_455-463-specific CD8⁺ T cells had reduced frequencies of the PD1⁺CD127⁺ subset compared with HBV core_18-27-specific CD8⁺ T cells (figure 5B) and higher expression of severe exhaustion-associated markers CD38, CD39, CD57 and reduced expression of CD28 (figure 5C,D). These data indicate that HBV polymerase_455-463-specific CD8⁺ T cells have a more severe mitochondrial dysfunction that is connected to their exhaustion programme. Overall these data are in line with the connection of metabolism to exhaustion state also observed in chronic HCV infection.

Activation and metabolism of HBV polymerase_455-463-specific CD8⁺ T cell responses correlate with qHBsAg levels

Since the differences in metabolism and exhaustion programme of HBV-specific responses could not be explained by the degree of liver inflammation, we wondered if they were connected to antigen levels. We thus determined qHBsAg levels and performed correlation analyses. Linear regression analysis revealed a correlation between CD39 expression, mitochondrial polarisation and quantitative HBsAg (qHBsAg) levels for HBV polymerase_455-463-specific responses, while no such correlation was found for HBV core_18-27-specific CD8⁺ T cell responses (figure 5E,F). This suggests that antigen recognition drives the metabolic and exhaustion programme in polymerase_455-463-specific CD8⁺ T cell responses but that this differs in HBV core_18-27-specific CD8⁺ T cell responses.

Increased transaminase levels correlate with mitochondrial depolarisation in PD-1⁺CD127^- virus-specific CD8⁺ T cells in chronic HCV infection

The differentiation of exhausted T cells is strongly influenced by viral antigen but is also subject to inflammatory cues. In DAA therapy during HCV infection, we had observed a negative correlation of mitochondrial polarisation with ALT (figure 4D), however, the control of viral replication and resolution of liver inflammation are intertwined. Therefore, we next investigated the connection of the metabolic T cell phenotype with the clinical activity in chronic HBV and HCV infection. Patients with cHCV infection displayed a diverse range of liver inflammation as indicated by aspartate aminotransferase (AST) and alanine aminotransferase (ALT) enzyme activity and viral loads, while in our cohort, HBV patients had milder hepatic inflammation (online supplemental figure 2A,B, online supplemental tables 1 and 2). We observed an inverse relationship of liver transaminases and mitochondrial polarisation and activation/exhaustion state of HCV-specific CD8⁺ T cells but not in HBV infection (online supplemental figure 2C–L). There was no clear correlation with viral load as measured by PCR, however, it is unclear if that readout is a good measure of T cell antigen recognition (online supplemental figure 3A–D). There was no direct correlation of metabolic T cell features and transaminase levels in HBeAg-negative chronic infection (ENCI) and HBeAg-negative chronic hepatitis B (ENCHB) patients (online supplemental figure 3E–H). These data show a link between the mitochondrial impairment of more severely exhausted virus-specific CD8⁺ T cells and the degree of hepatic inflammation in cHCV infection.

Enolase expression is reduced in severely exhausted CD8⁺ T cells and correlates with mitochondrial membrane potential

Our data demonstrate a prominent role for mitochondrial depolarisation in severely exhausted CD8⁺ T cells. However, this depolarisation occurred despite high glucose uptake by HCV-specific CD8⁺ T cells (figure 2D) and despite reduced utilisation of glucose for glycolysis in the metabolic flux analysis (online supplemental figure 1). We, therefore, sought to understand whether glycolytic flux was altered due to dysregulation of enzymes required for glycolysis (figure 6A). MRNA analysis identified variable expression of glycolytic genes between HBV-specific, HCV-specific and CMV-specific CD8⁺ T cells (online supplemental figure 4A). We focused on the analysis of Eno1 (encoding enolase 1) (online supplemental figure 4B). Intracellular staining for ENO1 revealed highest expression in naïve CD8⁺ T cells, and reduced expression in HCV NS3_1073-1081-specific and NS3_1406-1415-specific CD8⁺ T cells compared with HBV core_18-27-specific and polymerase_455-463-specific CD8⁺ T cells (figure 6B).

We next investigated the relationship between exhausted T cell subsets and ENO1 expression. ENO1 was significantly more highly expressed in the metabolically fit PD-1⁺CD127⁺ subset compared to PD-1⁺CD127^- virus-specific CD8⁺ T cells in chronic HBV and HCV infection (figure 6C). ENO1 expression was inversely correlated with the expression of activation/exhaustion marker CD38, but not with CD39 (figure 6D, online supplemental figure 4C). However, ENO1 expression positively correlated with mitochondrial polarisation (figure 6E), and there was a trend towards decreased mitochondrial ROS with increasing ENO1 expression (online supplemental figure 4D) in line with a model in which enolase levels more directly impact mitochondrial polarisation and low levels of enolase downstream predispose the cell to mitochondrial ROS production. Interestingly, elevated ALT levels also inversely correlated with ENO1, whereas AST levels, viral loads and qHBsAg did not (figure 6F, online supplemental figure 4E-G). These data indicated that ENO1 was expressed by T cells with less severe exhaustion and better mitochondrial fitness, and provoked the hypothesis that ENO1 could support the mitochondrial function of virus-specific CD8⁺ T cells, counteracting exhaustion.

Enolase enzymatic activity controls virus-specific CD8⁺ T cell metabolism by regulating glycolytic flux

To investigate the functional role of ENO1, we performed extracellular flux analysis of expanded HBV-specific and HCV-specific CD8⁺ T cells and analysed their glycolytic capacity in the presence of enolase inhibitor23 (online supplemental figure 5A–G) and/or after addition of the downstream metabolite pyruvate. Metabolic flux analysis showed that enolase inhibition reduced glycolysis in HBV-specific and HCV-specific CD8⁺ T cells (online supplemental figure 5B,E). However, the drop in glycolysis could be overcome by simultaneous injection of the downstream metabolite pyruvate (online supplemental figure 5B,E). Interestingly, enolase inhibition did not cause changes in oxygen consumption, but pyruvate addition caused an immediate small drop in OCR, suggesting that the effects on glycolysis were dominant (online supplemental figure 5C,F). Comparison of the effect of enolase inhibition on glycolysis showed that HBV-specific CD8⁺ T cells were enriched in strong responders to enolase inhibition compared with HCV-specific CD8⁺ T cells, suggesting a stronger role of enolase for the glycolytic flux in these cells. This notion is in line with the higher expression levels of enolase observed in HBV-specific CD8⁺ T cells (figure 6B). The dominant effect of enolase inhibition on glycolysis was also observed when cells were exposed to inhibitor pretreatment before the extracellular flux analysis (online supplemental figure 5H–I). Thus, enolase can act as a glycolytic checkpoint in HBV-specific and HCV-specific CD8⁺ T cells and pyruvate can bypass enolase inhibition for enhanced glycolysis.

Enolase 1 inhibition can drive mitochondrial depolarisation in primary human CD8⁺ T cells

Since short-term effects of enolase inhibition on respiration in in vitro metabolic flux assays were limited, but the correlations of enolase and mitochondrial respiration were strong in ex vivo analysis, we next asked if prolonged regulation of glycolysis via ENO1 would also impact on mitochondrial polarisation of freshly isolated primary human CD8⁺ T cells and treated CD8⁺ T cells from healthy individuals overnight with an enolase-specific inhibitor. We observed that ENO1 inhibition resulted in significantly reduced mitochondrial polarisation, more depolarised mitochondria and trends towards reduced cytokine production (online supplemental figure 4H,I). Thus, ENO1 inhibition resulted in a mitochondrial phenotype typical for severely exhausted virus-specific CD8⁺ T cells.

Enolase enzymatic activity controls virus-specific CD8⁺ T cell effector function

Due to the prominent role of glycolysis for T cell effector function, we next wondered if enolase would represent a metabolic regulator of exhausted T cell function in virus-specific CD8⁺ T cells. We, therefore, examined cytokine production of HBV core_18-27-specific and HCV NS3_1073-1081-specific, NS3_1406-1415-specific and NS5B_2594-2602-specific CD8⁺ T cells after overnight treatment with AP-III-a4 hydrochloride or in combination with pyruvate. Treatment did not affect viability (online supplemental figure 4J). After treatment with the ENO1 inhibitor, HBV-specific and HCV-specific CD8⁺ T cells showed a significant drop in IFN-γ production (figure 6G). In contrast, the addition of pyruvate enhanced cytokine production (figure 6H–L, online supplemental figure 4K-M). The effects of enolase inhibition on IFN-γ production were stronger than on TNF-α production, however, pyruvate supplementation also enhanced the production of this cytokine, in particular in HCV-specific CD8⁺ T cells (figure 6J).

Enolase 1 expression is reduced in exhausted versus memory-like CD8⁺ T cells in HBV infection in mice

To understand the role of differential ENO1 expression in hepatitis virus-specific CD8⁺ T cell exhaustion in the same virus model and intrahepatic environment, and to understand differences between acute and chronic infection, we used a recently developed mouse model of acute or chronic HBV infection.24 In this model, C57Bl/6 mice were infected with different doses of an adenoviral vector encoding for a 1.3-overlength HBV genome (Ad-HBV-Luc) after transfer of naïve HBV Cor₉₃-specific CD8⁺ T cells. Depending on the dose of Ad-HBV-Luc administered, mice develop either acute self-limiting (10⁷ pfu) or chronic HBV infection (10⁸ pfu) (figure 7A). At 15 dpi and 30 dpi, respectively, liver-associated lymphocytes were isolated and Cor₉₃-specific CD8⁺ T cells were analysed by single-cell RNA-seq. UMAP analysis identified different antigen-specific T cell transcriptomes in acute and chronic infection (figure 7B). Clustering analysis revealed four separate clusters with differential expression of exhaustion markers (figure 7C). HBV-specific CD8⁺ T cells from mice with acute HBV infection were predominantly enriched in cluster 1 (C1). Chronic HBV-specific CD8⁺ T cells were enriched in C4 (figure 7D). Markers associated with severe exhaustion, such as Pdcd1, Cd38, Cd39, Tox and Fli1 showed increased expression in C4 (figure 7E). In contrast, C1 highly expressed Il7r, Tcf7, Slamf6 and Cxcr5 that indicate homeostatic and memory-like characteristics. Interestingly, Eno1 expression was significantly higher in acute compared with chronic HBV-specific CD8⁺ T cells, indicating a loss of enolase expression during chronic infection (figure 7F). GSEA of exhaustion signatures revealed a significant enrichment of exhaustion genes in HBV-specific CD8⁺ T cells isolated from chronic condition, as expected. Moreover, direct comparison of C1 versus C4 showed a significant enrichment of exhaustion-associated gene expression in C4 and an enrichment of memory-associated features in C1 (figure 7G).25 26 Moreover, C1 was significantly enriched for Eno1 expression compared with other clusters (figure 7H). These results are in line with our previous findings on ENO1 expression in less severely exhausted virus-specific CD8⁺ T cells in cHBV and cHCV patients (figure 6C). Moreover, we observed significantly elevated glycolytic gene expression in acute versus chronic cells, with C1 exhibiting highest expression of glycolytic genes among all clusters, as well as an enrichment of genes responsible for OXPHOS in acute versus chronic HBV infection. In sum, these data support a role for enolase in metabolically regulating differential exhaustion profiles of intrahepatic virus-specific CD8⁺ T cells (figure 7I).