Early Transcriptional Divergence Marks Virus-Specific Primary Human CD8+ T Cells in Chronic versus Acute Infection

Abstract

Distinct molecular pathways govern the differentiation of CD8⁺ effector T cells into memory or exhausted T cells during acute and chronic viral infection, but these are not well studied in humans. Here, we employed an integrative systems immunology approach to identify transcriptional commonalities and differences between virus-specific CD8⁺ T cells from patients with persistent and spontaneously resolving hepatitis C virus (HCV) infection during the acute phase. We observed dysregulation of metabolic processes during early persistent infection that was linked to changes in expression of genes related to nucleosomal regulation of transcription, T cell differentiation, and the inflammatory response and correlated with subject age, sex, and the presence of HCV-specific CD4⁺ T cell populations. These early changes in HCV-specific CD8⁺ T cell transcription preceded the overt establishment of T cell exhaustion, making this signature a prime target in the search for the regulatory origins of T cell dysfunction in chronic viral infection.

Keywords: CD4 T cell help; CD8 T cells; T cell dysfunction; adaptive immunity; hepatitis C virus; metabolism; network analysis; nucleosome; transcriptional regulation; viral escape.

Figure 1. Multivariate analysis of adaptive immunity during early HCV infection

(A) Gating strategy for sorting of HCV-specific CD8⁺ T cells, defined as CD8 and pMHC multimer positive lymphocytes negative for CD4, CD14, CD19, CD16, CD56, Annexin V, and LiveDead viability stain. (B) Sampling and assay overview. HCV-specific CD8⁺ T cells from 43 subjects with different outcomes and varying levels of antigen recognition (Chronic (C), Escape (E), and Resolver (R)) were sampled at multiple time points (dots) within the first 36 weeks of infection (n_total = 78, STAR Methods). CD4⁺ T cell status (magnitude of HCV-specific CD4⁺ T cell response, dot color) and neutralizing antibodies (titers from HCV pseudoparticle neutralization assay, white bars) were assessed for a subset of these samples (n_help = 37 and n_{neutralization}=32). See also Table S1.

Figure 2. Virus-specific CD8⁺ T cells display distinct transcriptional profiles of resolution, inflammation and persistent antigen exposure

(A) Heatmap of 258 differentially expressed genes across C, E, and R groups during early and late acute phase of HCV infection (<18 and >18 weeks post infection). Transcriptional profiles of HCV-specific CD8⁺ T cells were analyzed for differential gene expression between C and R in early or late acute infection or between early and late infection within C or R. Shown are means of scaled expression levels (expression scores) by group. Expression levels for E samples are shown but not used for detection of differential expression. (B) Venn diagrams of distinct transcriptional signatures defined by patterns of shared differential expression between C, E, and R groups and in early and late acute phase of infection. See also Tables S1, S2, and S3.

Figure 3. Modules of tightly co-regulated genes give rise to distinct patterns of transcriptional regulation in different infection outcomes

(A) Detection of gene co-expression modules in C, E, and R groups. Weighted gene co-expression networks were constructed using WGCNA and modules of highly correlated genes were detected by hierarchical clustering. Colored bars indicate detected modules and their relative size, similar colors across groups do not imply overlap. (B) Heatmap of scaled topological overlap matrices for isolated network modules from C, E, and R groups. Lighter colors indicate stronger positive correlation between genes within and across modules. (C and D) Chord diagrams of gene module correspondence between C, E, and R groups. Diagrams show significance of gene overlap between two modules (core score, blue connection) or a module and unconnected genes (isolation score, red connections) for individual group pairs of (C) C/R or (D) C/E and E/R. Shown are scores that passed a significance threshold of FDR <0.01. (E) Correlation, overlap and community networks for modules detected in C, E, and R groups. Correlation and overlap networks were constructed using within-group correlation significance (FDR <0.01) and across-group overlap significance (FDR <0.01) and merged for detection of module communities (indicated by color) (STAR Methods). Significance of correlation and overlap were used for network layout and shown as width of the edges connecting modules within networks. See also Table S4.

Figure 4. Differential expression of network modules is preserved across cohorts of acute HCV infection

(A and B) Barcode plots of network module signature enrichments for HCV-specific CD8⁺ T cells from (A) C and R patients during the early and late phase of acute infection and (B) a separate validation cohort of chronic and resolving patients sampled during the early acute phase of infection (≤ 21 weeks post infection). Ranked lists of log2 fold changes from analysis of differential expression between C and R groups were tested for enrichment of network modules (STAR Methods). Shown are representative examples of modules that were significantly enriched in early samples of the original data set (FDR < 0.001) (r2, r6 and r5) as well as a universally non-enriched module (c5). (C) Overview of enrichment results for all three sample subsets and all modules. Dot size indicates FDR corrected p-values of enrichment. Statistically significant enrichments (FDR < 0.001) are highlighted in color according to directionality of enrichment in respect to up-regulation (red) or down-regulation (blue) in the R group. See also Tables S4, S5, and S6.

Figure 5. Correlation and enrichment maps link gene co-expression modules from different immunological states to clinical traits, immune signatures and biological function

(A) Trait correlation map for C, E and R groups in acute HCV infection. Eigengenes (first principal components) of network modules were correlated with clinical traits (colors indicate r value and sign of correlations (red = positive, blue = negative). Shown are correlations with FDR < 0.1. (B and C) Enrichment maps for C, E, and R groups, showing enrichments scores for all significant enrichments (FDR < 0.05). (B) Modules were tested for enrichment of CD8⁺ T cell differential expression signatures from HCV (Figure 2), HIV (Gaiha et al., 2014), and mouse models of T cell memory and exhaustion (Doering et al., 2012; Singer et al., 2016; Subramanian et al., 2005) (C) Modules were tested for enrichment of CD8⁺ T cell WGCNA module signatures from acute (AM) and chronic (CM) LCMV infection (Doering et al., 2012). (D) Functional enrichment maps for C, E, and R groups. Detected modules were tested for enrichment of immune pathways (Langfelder and Horvath, 2008), GO terms, and KEGG pathways. Shown are enrichment scores for top-scoring immune pathways (top 5 per group per module by enrichment score and number of enrichment-driving genes) with FDR <0.05 and KEGG pathway/GO term enrichments below a significance threshold of FDR <1e-10. Enriched processes were grouped into clusters of functionally related and co-regulated terms related to metabolism, immune, and nucleosome clusters as indicated by colored dots in grey panels to the right of enrichment maps. See also Tables S3, S7 and S8.

Figure 6. A distinct temporal pattern of metabolic regulation marks chronic HCV infection

(A) Metabolism Cluster Network for C, E, and R, groups. Co-expression networks for each group were subset to genes annotated to the metabolism cluster and visualized as networks with components of genes grouped by exclusive or shared regulation between groups (colored bubbles). Nodes are sized by intramodular connectivity and colored by group expression score during the early phase of infection. (B) KEGG pathway enrichment in metabolism cluster network. Individual components from (A) were tested for enrichment of KEGG pathways. All pathways that were significantly enriched (FDR <0.05) in at least one component are shown. (C) Chronic-exclusively regulated genes annotated for OXPHOS. Genes (shown as nodes), colored by their module membership in the C group. Genes that overlap with a leading-edge signature of metabolic processes associated with T cell exhaustion (Bengsch et al., 2016) are encircled by a dashed line (12/17 genes). (D) Expression kinetics of hub genes for modules c3, c5, c6. Log2-transformed gene expression levels for the most highly connected gene of each module are plotted over time for each group (dots = individual samples, color = module membership in C). Longitudinal trends in expression are visualized using a local regression based (LOESS) smoother (grey shading = 95% confidence interval). See also Figures S1 and S2, and Tables S4, S8, and S9.

Figure 7. Distinct regulatory interactions link dysregulation of metabolic, nucleosomal, and immune processes in CD8⁺ T cells from early chronic HCV infection

(A) TR-target coverage for CD8⁺ T cells in chronic HCV infection. A large network of regulatory interactions was assembled from the literature (STAR Methods) and subset to TR-target interactions with matching module assignments. TRs are plotted by number (left) or percentage (right) of within-module regulatory interactions and intramodular connectivity. Transcription factors with intramodular connectivity > 0.5 and/or target coverage count > 20 or percentage >5% are labeled with their respective gene symbols. TR module membership is indicated by color. (B) Regulatory networks for select TRs from modules c3, c5, c6, and r2. Weighted gene correlation networks were restricted to strong interactions (weight > 0.025) between selected TRs and known targets (k_Within >0.25). For transcription co-factors and chromatin remodeling genes, only interactions targeting transcription factors are shown. TRs and targets are labeled with gene symbols and colored by regulatory function or annotation in metabolism, nucleosome, or immune cluster networks. Interaction types are indicated by color of edges connecting TRs and targets. (C) Representative expression kinetics of Chronic regulatory network genes. Shown are log2-transformed expression levels (dots represent individual samples) of genes annotated for regulation of transcription (Histones, RNA & DNA polymerases), metabolism (ATP Synthase, NADH dehydrogenase, Cytochrome C oxidase, and Oxidoreductase activity), and T cell function and homing (T cell maintenance & activation, Cell adhesion & migration). Color and number indicate C group module membership. Longitudinal trends in expression are visualized using a local regression based (LOESS) smoother (grey shading = 95% confidence interval). See also Figures S3 and S4, and Tables S8 and S10.