Research ArticleT CELLS

TOX is expressed by exhausted and polyfunctional human effector memory CD8+ T cells

See allHide authors and affiliations

Science Immunology  03 Jul 2020:
Vol. 5, Issue 49, eaba7918
DOI: 10.1126/sciimmunol.aba7918

The TOX profiles of T cells

Transcription factors TOX and TCF-1 have emerged as key drivers of exhaustion and stemness programs in CD8+ T cells. Using bulk and single-cell transcriptome analyses and flow cytometric analyses, Sekine et al. have generated a detailed map of TOX and TCF-1 expression in human CD8+ T cells. TOX is generally expressed by effector memory CD8+ T cells and is not restricted to exhausted T cells, whereas TCF-1 demarcates early-differentiated, memory CD8+ T cells. Using tetramers to examine the specificity of antigen-specific CD8+ T cells, they found that cytotoxic memory CD8+ T cells targeting both pathogenic and well-controlled chronic infections are more likely to express TOX. Their results propose that TOX-dependent transcriptional wiring is not restricted to exhausted CD8+ T cells.

Abstract

CD8+ T cell exhaustion is a hallmark of many cancers and chronic infections. In mice, T cell factor 1 (TCF-1) maintains exhausted CD8+ T cell responses, whereas thymocyte selection-associated HMG box (TOX) is required for the epigenetic remodeling and survival of exhausted CD8+ T cells. However, it has remained unclear to what extent these transcription factors play analogous roles in humans. In this study, we mapped the expression of TOX and TCF-1 as a function of differentiation and specificity in the human CD8+ T cell landscape. Here, we demonstrate that circulating TOX+ CD8+ T cells exist in most humans, but that TOX is not exclusively associated with exhaustion. Effector memory CD8+ T cells generally expressed TOX, whereas naive and early-differentiated memory CD8+ T cells generally expressed TCF-1. Cytolytic gene and protein expression signatures were also defined by the expression of TOX. In the context of a relentless immune challenge, exhausted HIV-specific CD8+ T cells commonly expressed TOX, often in clusters with various activation markers and inhibitory receptors, and expressed less TCF-1. However, polyfunctional memory CD8+ T cells specific for cytomegalovirus (CMV) or Epstein-Barr virus (EBV) also expressed TOX, either with or without TCF-1. A similar phenotype was observed among HIV-specific CD8+ T cells from individuals who maintained exceptional immune control of viral replication. Collectively, these data demonstrate that TOX is expressed by most circulating effector memory CD8+ T cell subsets and not exclusively linked to exhaustion.

INTRODUCTION

In the setting of many cancers and chronic viral infections, ongoing antigen exposure can lead to CD8+ T cell exhaustion, a phenomenon characterized by profound epigenetic and transcriptional changes associated with a progressive loss of effector functions and the up-regulation of various inhibitory receptors (IRs) (13). This process likely evolved to prevent excessive immune activation, but it remains a major stumbling block in the quest to develop more effective immunotherapies and vaccines. The functional consequences of exhaustion can be overcome to some extent via the administration of checkpoint inhibitors, which block signals transduced by IRs, such as CTLA-4 and PD-1 (46). Although therapeutic interventions based on this concept have revolutionized the treatment of various cancers in recent years, clinical responses are unpredictable and vary considerably among individuals with the same diagnosis. A more detailed understanding of CD8+ T cell exhaustion is therefore required to inform translational efforts in the fields of immunoprophylaxis and immunotherapy (6, 7).

It is widely held that the process of exhaustion drives CD8+ T cells into a transcriptionally distinct lineage (7) that encompasses “precursor exhausted” and “terminally exhausted” subsets within a unique differentiation spectrum (8). Precursor exhausted CD8+ T cells are thought to self-renew and differentiate into terminally exhausted CD8+ T cells, which display reduced effector functionality. Direct comparison of these subsets has further shown that precursor exhausted CD8+ T cells less commonly express IRs and more commonly express certain transcription factors, including T cell factor 1 (TCF-1) (8). TCF-1 is a high-mobility group (HMG) box transcription factor that plays a critical role in the differentiation and survival of memory CD8+ T cells under physiological conditions (9, 10). Moreover, TCF-1 maintains stemness and sustains CD8+ T cell responses against chronic viral infections (1113).

There is no definitive marker that identifies exhausted CD8+ T cells. However, recent work has shown that the transcription factor thymocyte selection-associated HMG box (TOX) distinguishes exhausted from memory CD8+ T cells in various mouse models (1418). These studies have demonstrated that TOX is absolutely required for the epigenetic remodeling and survival of exhausted CD8+ T cells (1418). In the classic dichotomy, infection with lymphocytic choriomeningitis virus (LCMV) clone 13 leads to chronic antigen stimulation and generates exhausted TOX+ virus-specific CD8+ T cell populations, which are clearly distinct from the memory CD8+ T cell populations that emerge in response to infection with the nonpersistent strain LCMV Armstrong (14, 15). Tumor-specific CD8+ T cells similarly express TOX and exhibit functional impairment associated with the up-regulation of IRs (15, 16). It has nonetheless remained unclear how these data apply to acute and chronic viral infections in humans. This is a highly pertinent issue in the context of global health, because immune control of viral replication has been associated with functional attributes of CD8+ T cells (1923).

HIV infection is characterized by sustained viral replication and immune activation, leading to a progressive loss of CD4+ T cells and the eventual development of AIDS in the absence of antiretroviral therapy (ART). CD8+ T cells are essential for immune control of HIV (24). In a majority of cases, however, immune control is partial at best, and HIV-specific CD8+ T cells display the hallmarks of exhaustion, including dysfunctionality, impaired proliferation, and the expression of various IRs, such as PD-1, TIGIT, 2B4, CD39, and Tim-3 (19, 21, 2528). In contrast, rare individuals maintain highly functional (polyfunctional) HIV-specific CD8+ T cell populations, which associate with effective viral suppression (19, 29, 30). Specific transcription factors, such as Eomes and BATF, have previously been linked with CD8+ T cell exhaustion in the context of HIV infection (21, 31). It has nonetheless remained unclear to date how this process relates to the expression of TOX and TCF-1.

We here investigated the expression of TOX and TCF-1 as a function of differentiation and specificity in the human CD8+ T cell landscape. TOX was expressed primarily in effector memory CD8+ T cells, whereas TCF-1 was expressed primarily in naive and early-differentiated memory CD8+ T cells. Progressive HIV disease was associated with increased expression of TOX, together with various activation markers and IRs, and decreased expression of TCF-1. However, polyfunctional CD8+ T cells specific for cytomegalovirus (CMV) or Epstein-Barr virus (EBV), which are controlled effectively by the immune system, also expressed to a high degree TOX. A similar pattern was observed for HIV-specific CD8+ T cells in cases where viral replication was controlled in the absence of ART. Collectively, these findings suggest a nuanced model in humans, wherein highly functional effector memory subsets recognizing persistence viruses can also express TOX.

RESULTS

TOX and TCF-1 are differentially expressed among circulating CD8+ T cells

Recent studies have examined the roles of TOX and TCF-1 in mouse models of chronic antigen-driven CD8+ T cell exhaustion (1418). However, the expression patterns of these transcription factors, especially in combination, are less well defined in humans. To address this issue, we compared the transcriptomes of naive T (TN), central memory T (TCM), effector memory T (TEM), and effector memory RA T (TEMRA) cells flow-sorted directly ex vivo as phenotypically distinct CD8+ subsets from peripheral blood samples donated by HIV individuals (fig. S1A). Analysis of the RNA-sequencing (RNA-seq) data revealed a core signature of genes that were differentially expressed among these CD8+ T cell subsets (table S1). In particular, Tox and Tcf7, which encodes TCF-1, were among the top identified genes that distinguished naive and non-naive (memory) CD8+ T cells (Fig. 1A). Tox transcript levels were significantly elevated in TCM, TEM, and TEMRA cells relative to TN cells, whereas Tcf7 transcripts were significantly elevated in TN cells relative to TEM and TEMRA cells (Fig. 1B). We then used the Assay of Transposase-Accessible Chromatin using sequencing (ATAC-seq) to characterize the epigenetic landscape of resting TN, TEM, and TEMRA cells. The Tox and Tcf7 loci both contained open chromatin clusters that distinguished TN cells from TEM and TEMRA cells (Fig. 1C). TCF-1 binding motifs were more readily accessible in the open chromatin regions (OCRs) of TN cells relative to the OCRs of TEM and TEMRA cells (Fig. 1D). Equivalent sequence-based analyses were not possible for TOX, which is thought to bind DNA in a structure-dependent manner (32).

Fig. 1 Expression of Tox and Tcf7 in CD8+ T cell subsets.

(A) RNA-seq heatmap showing differentially expressed genes (fold change > 2; P < 0.05) among TN, TCM, TEM, and TEMRA cells (n = 3 healthy donors). (B) Volcano plots comparing gene expression between TN cells and each subset of memory CD8+ T cells (n = 1 healthy donor). (C) ATAC-seq tracks showing enrichment of OCRs adjacent to Tox (top) and Tcf7 (bottom) for TN, TEM, and TEMRA cells (n = 2 healthy donors). (D) Enrichment of the Tcf7 motif in OCRs comparing TN versus TEM cells and TN versus TEMRA cells (n = 1 healthy donors). (E) Western blot analysis of TOX expression in TN versus memory (non-naive) CD8+ T cells. Actin was included as a loading control. (F) Representative histograms and donor-matched graphs showing percent expression ex vivo of TOX and TCF-1 in TN, TCM, TEM, and TEMRA cells. (G) Expression of TOX and TCF-1 in TN, TCM, TEM, and TEMRA cells after stimulation with ImmunoCult Human CD3/CD28 T Cell Activator for 5 days (aCD3/CD28). NC, negative control. (H) Details as in (G). Proliferating cells were identified using CellTrace Violet (CTV). *P < 0.05, **P < 0.01, and ***P < 0.001.

Differential expression of TOX and TCF-1 between naive and memory CD8+ T cells was validated at the protein level using Western blots (Fig. 1E and fig. S1, B and C). Flow cytometric analyses further confirmed that TOX was predominantly expressed in TEM and TEMRA cells, whereas TCF-1 was predominantly expressed in TN and TCM cells (Fig. 1F). A previous study found that tonic antigen-driven stimulation triggered the expression of TOX (15). We corroborated this result and traced the expression patterns of TOX and TCF-1 in flow-sorted TN, TCM, TEM, and TEMRA cells labeled with a stable fluorescent dye. Most of the proliferating cells in each subset expressed TOX, but not TCF-1, in response to stimulation via CD3 and CD28 (Fig. 1, G and H). Collectively, these results show that TOX and TCF-1 are differentially expressed and regulated among naive and memory subsets of human CD8+ T cells under physiological conditions.

Expression of Tox and Tcf7 separates CD8+ T cells into distinct trajectories

To extend these findings, we analyzed a publicly available single-cell RNA-seq (scRNA-seq) dataset acquired from CD8+ T cells (10× Genomics Repository). Nonlinear relationships among individual CD8+ T cells were assessed using Uniform Manifold Approximation and Projection (UMAP). Tox and Tcf7 transcripts were readily detectable and distinctly expressed at the single-cell level (Fig. 2A and fig. S2A). We then binned CD8+ T cells into Tox+ and Tcf7+ events (Fig. 2B). Tox+ cells expressed certain IR genes, including Pdcd1, Tigit, Cd244, and Lag3, whereas Tcf7+ cells expressed genes associated with naive and early-differentiated memory T cells, including Ccr7, Il7r, Sell, Nell2, Lef1, Bach2, Myc, and Id3 (Fig. 2B and fig. S2B). Multiple effector memory genes were also differentially expressed between Tox+ and Tcf7+ cells, including Gzmb, Gzma, Prf1, Cx3cr1, Tbx21, Eomes, Zeb2, and Prdm1 (Fig. 2B and table S2).

Fig. 2 Expression of Tox and Tcf7 in the CD8+ T cell landscape.

(A) scRNA-seq analysis of circulating CD8+ T cells (n = 2 healthy donors). The UMAP plots illustrate the distribution of Tox+ (more than one read) and Tcf7+ cells (more than one read). (B) Violin plots showing immune-related genes that were differentially expressed between Tox+ and Tcf7+ cells (n = 2 healthy donors). Differentially expressed genes were identified using a threshold of P < 0.05. (C) Clustering analysis of the same dataset using Seurat. (D) Heatmap showing genes that were differentially expressed among the clusters identified in (C). Highlighted genes encode proteins with known differential expression between naive and memory T cell subsets.

Further analysis of the same dataset using Seurat identified eight different clusters on the basis of differential gene expression (Fig. 2C). Two naive-like clusters (0 and 1) with the highest expression levels of Tcf7 also expressed high levels of Ccr7, Lef1, and Sell (table S3). In contrast, three effector memory–like clusters (5, 6, and 7) with the highest expression levels of Tox were enriched for effector-related genes, including Ccl5, Gzma, Nkg7, Gzmh, Klrd1, Prf1, Gnly, and Fgfbp2 (Fig. 2D and table S3). Although not a part of the core signature, certain IR genes, including Pdcd1, Tigit, Lag3, and Lilrb1, were also preferentially expressed among these Tox+ clusters (fig. S3). Collectively, these findings highlight an association between Tox and the expression of genes that encode various immune cell effector proteins and IRs.

TOX and TCF-1 protein expression in CD8+ T cell clusters

To validate the scRNA-seq data acquired from CD8+ T cells, we designed a 28-color flow cytometry panel to assess the expression of differentially regulated transcripts at the protein level in relation to TOX and TCF-1. Data were again concatenated and analyzed using UMAP (fig. S4, A and B). Manual gating identified TN, TCM, TEM, and TEMRA clusters in the UMAP space (Fig. 3A). Distinct topographical regions were then delineated by overlaying the expression of TOX and TCF-1 (Fig. 3B). Phenograph analysis revealed 27 unique cell clusters in the UMAP space (Fig. 3C and fig. S4C). TCF-1 was coexpressed with CCR7, CD45RA, and CD28 (Fig. 3D). In contrast, TOX was expressed in clusters with increased expression of cytolytic proteins, namely, granzyme B (GzmB) and perforin, and certain IRs, including PD-1, TIGIT, 2B4, and CD39 (Fig. 3E). Moreover, specific TOX clusters exhibited high expression levels of cytolytic proteins (cluster 5), IRs (cluster 27), or both cytolytic proteins and IRs (cluster 26; Fig. 3E). Manual gating analysis further confirmed that almost every single TOX+ cell expressed GzmB, perforin, PD-1, TIGIT, or 2B4 (fig. S4D).

Fig. 3 Expression of TOX and TCF-1 in CD8+ T cell clusters.

(A) Left: Flow cytometric gating strategy for the identification of TN, TCM, TEM, and TEMRA cells. Right: Subset distribution on the UMAP plot generated from bulk CD8+ T cells after data concatenation (n = 4 healthy donors). (B) Left: Expression of TOX and TCF-1 in bulk CD8+ T cells. Right: Expression intensities projected on the corresponding UMAP plots. (C) Same UMAP display with subpopulations colored using Phenograph (n = 27 clusters). (D) UMAP plots showing expression patterns for the indicated markers. (E) Hierarchical clustering of expression intensity (z-score) for each of the indicated markers in each cluster derived using Phenograph. (F) Top left: Flow cytometric gating strategy (CD95 was shown to distinguish non-naive cells) for the separation of TOX+ cells into TCF-1+ and TCF-1 populations. Top right: Histograms showing expression of the indicated markers among TOX+TCF-1+ (red) and TOX+TCF-1 cells (blue). Bottom: Scatter plots showing expression frequencies for the indicated markers among TOX+TCF-1+ and TOX+TCF-1 cells from all healthy donors. *P < 0.05, **P < 0.01, and ***P < 0.001.

In further analysis, we identified a progenitor-like TOX+TCF-1+ population and a highly differentiated TOX+TCF-1 population (Fig. 3F). TOX+TCF-1+ cells expressed higher levels of the early differentiation markers CCR7 and CD28, as well as some IRs (PD-1 and TIGIT), whereas TOX+TCF-1 cells expressed higher levels of perforin, GzmB, CD38, 2B4, and Tim-3 relative to TOX+TCF-1+ cells (Fig. 3F). GzmB+perforin+ cells were also more prevalent than either GzmB+perforin or GzmBperforin cells in the TOX+TCF-1 compartment (fig. S4D). TOXTCF-1+ cells were largely composed of a mixture of cells with a TN and TCM phenotype (fig. S4E). Collectively, these observations demonstrate the existence of distinct TOX+ subpopulations in the CD8+ T cell landscape, some of which lack expression of IRs.

HIV infection is associated with increased expression of TOX

To understand how an ongoing systemic infection affects the balance between TOX and TCF-1 expression in human memory CD8+ T cells, we recruited HIV+ viremic individuals and HIV+ aviremic individuals on ART (table S4). In line with previous studies (21, 25, 28, 31), we found that some IRs, including PD-1, TIGIT, and 2B4, were expressed at higher frequencies on memory CD8+ T cells from HIV+ viremic donors relative to HIV donors (Fig. 4A). Circulating memory CD8+ T cells also more commonly expressed TOX and less commonly expressed TCF-1 in the context of HIV infection (Fig. 4B). Increased frequencies of TOX+TCF-1 and TOXTCF-1 cells were detected in HIV+ viremic donors relative to HIV and HIV+ aviremic donors (Fig. 4B). Moreover, TOX+TCF-1+ cells persisted at elevated frequencies in HIV+ aviremic donors relative to HIV donors, whereas HIV donors harbored increased frequencies of TOXTCF-1+ cells relative to HIV+ viremic and HIV+ aviremic donors (Fig. 4B). Multiple studies have demonstrated that HIV infection is associated with premature aging of the immune system, where specifically CD8+ T cells are skewed toward a terminally differentiated phenotype (33, 34). In line with these observations, we found a close association between the frequencies of TCM and TOXTCF-1+ cells, as well as TEMRA and TOX+TCF-1 cells (fig. S5, A and B), suggesting that alterations in the TOX/TCF-1 profile are interlinked to a skewed differentiation state in HIV infection.

Fig. 4 Expression of TOX and TCF-1 in donors infected with HIV.

(A) Expression of the indicated IRs on non-naive CD8+ T cells from a representative HIV donor (blue boxes) and a representative HIV+ viremic donor (red boxes). (B) Left: Expression of TOX and TCF-1 in non-naive CD8+ T cells. Right: Scatter plots showing TOX+/−TCF-1+/− population frequencies among non-naive CD8+ T cells from HIV (blue), HIV+ viremic (red), and HIV+ aviremic donors (yellow). (C) Left: Coexpression of TOX versus PD-1, TIGIT, and 2B4 among non-naive CD8+ T cells from a representative HIV donor (top) and a representative HIV+ viremic donor (bottom). Stack bars showing the frequency of TOX and TOX+ cells among IR+ non-naive CD8+ T cells (n = 30 HIV donors in blue and n = 17 HIV+ viremic donors in red). Right: Coexpression of PD-1 versus GzmB among CD8+ T cells pregated on TOX. Stack bars showing the frequency of IR and IR+ cells among TOX+ non-naive CD8+ T cells (n = 30 HIV donors in blue and n = 17 HIV+ viremic donors in red). Right: Bar graphs showing the distribution of TOX+ cells among PD-1+/−GzmB+/− populations of CD8+ T cells (n = 30 HIV donors and n = 17 HIV+ viremic donors). (D) Left: Flow cytometric gating strategy for the identification of non-naive CD8+ T cells, designed to include stem cell–like clusters. Right: UMAP plots generated from non-naive CD8+ T cells after data concatenation. The colored plots show the distribution of non-naive CD8+ T cells in the UMAP space for HIV (n = 2; blue) and HIV+ viremic donors (n = 2; red). (E) Left: UMAP plots showing expression intensities for TOX and TCF-1. Right: Same UMAP display with subpopulations colored using Phenograph (n = 24 clusters). (F) UMAP plots showing expression patterns for the indicated markers. (G) Hierarchical clustering of expression intensity (z-score) for each of the indicated markers in each cluster derived using Phenograph. (H) Distribution of clusters between HIV (n = 2; blue) and HIV+ viremic donors (n = 2; red). *P < 0.05, **P < 0.01, and ***P < 0.001.

Memory CD8+ T cells expressing PD-1, TIGIT, or 2B4 mostly coexpressed TOX, in the absence or presence of actively replicating HIV (Fig. 4C). However, PD-1+TIGIT+2B4+ cells were more heavily skewed toward a TOX+TCF-1 phenotype in HIV+ viremic donors relative to HIV donors (fig. S6A). We also noted that many TOX+ cells expressed cytolytic molecules and that not all TOX+ cells expressed IRs (Fig. 4C).

In further analyses, we concatenated the multiparametric data and compared memory CD8+ T cells from HIV donors with memory CD8+ T cells from HIV+ viremic donors via UMAP (Fig. 4D). An equal TOX versus TCF-1 expression pattern was observed in HIV donors, whereas a skewed TOX+ phenotype was found in HIV+ viremic donors (Fig. 4, D and E). Specific markers that distinguished TOX+ and TCF-1+ clusters (Fig. 3) were then analyzed using Phenograph (Fig. 4E and fig. S6B). Again, TCF-1 was coexpressed with CCR7 and CD28, whereas TOX was expressed in clusters with increased expression of cytolytic proteins, terminal differentiation markers, and/or IRs (Fig. 4F). Memory CD8+ T cells from HIV donors were mostly found not only in clusters with high expression levels of TCF-1 (clusters 17 and 9) but also in several clusters with detectable expression of TOX (clusters 18 and 10) (Fig. 4, G and H, and fig. S6C). TOX+ clusters in HIV donors nonetheless tended to express lower levels of PD-1, TIGIT, and CD38 relative to TOX+ clusters in HIV+ viremic donors (clusters 20, 2, 4, 6, 13, 3, 11, 23, 1, and 21), whereas the dominant TOX+ clusters in HIV+ viremic donors expressed higher levels of cytolytic proteins and IRs relative to the dominant TOX+ clusters in HIV donors (Fig. 4, G and H). Together, these data show that memory CD8+ T cells are skewed toward a continuum of TOX+ effector memory phenotypes in the setting of chronic HIV infection.

Memory CD8+ T cells specific for chronic viral antigens express TOX

Studies in mice have investigated the expression of TOX and TCF-1 in LCMV infection models (14, 15, 18). However, it has remained unclear whether viral infections induce the expression of TOX in human CD8+ T cells, which are similarly prone to antigen-driven exhaustion via a process that is not necessarily associated with the expression of IRs (23, 35). We therefore investigated the expression of TOX and TCF-1 in human CD8+ T cell populations specific for influenza virus (Flu), which establishes acute infections, CMV or EBV, which establishes persistent infections controlled by the immune system, or HIV, which establishes persistent infections and continues to replicate vigorously in the absence of ART (Fig. 5A and fig. S7A). In line with studies of mice infected with LCMV (14, 15, 18), we found that Flu-specific CD8+ T cells rarely expressed TOX and commonly expressed TCF-1, whereas CD8+ T cells specific for CMV, EBV, or HIV commonly expressed TOX with or without TCF-1 (Fig. 5B and fig. S7B). The expression intensity [median fluorescence intensity (MFI)] of TOX was elevated in CMV- and HIV-specific CD8+ T cells (fig. S7C) and associated with increased GzmB and perforin levels (fig. S7D). Likewise, TOX expression was higher on effector memory subsets for Flu-, EBV-, CMV-, and HIV-specific CD8+ T cells (fig. S7E), again suggesting that TOX expression is interlinked to a more mature differentiation state.

Fig. 5 Expression of TOX and TCF-1 in virus-specific CD8+ T cells.

(A) Left: Flow cytometric identification of non-naive CD8+ T cells specific for Flu (blue), EBV (orange), CMV (green), or HIV (red) using MHC class I tetramers. Right: Overlays showing the distribution of each virus-specific population in the TOX+/−TCF-1+/− landscape. (B) Percent expression of TOX (left) and TCF-1 (right) in virus-specific non-naive CD8+ T cells. (C) Distribution of virus-specific non-naive CD8+ T cells on the UMAP plot derived in Fig. 4 (total n = 4 healthy donors). (D) Top: Distribution of virus-specific non-naive CD8+ T cells among clusters derived using Phenograph. Bottom: Summed distribution of virus-specific non-naive CD8+ T cells in each cluster derived using Phenograph. Total n = 4 healthy donors. (E) Percent expression of the indicated markers among virus-specific non-naive CD8+ T cells (total n = 45 healthy donors). (F) Left: Coexpression of TOX versus TNF, IL-2, and perforin among CMV-specific CD8+ T cells pregated on IFN-γ. Right: Stack bars showing the frequencies of TOX and TOX+ cells among IFN-γ+ virus-specific CD8+ T cells (Flu, n = 25 healthy donors; EBV, n = 27 healthy donors; CMV, n = 30 healthy donors; HIV, n = 25 healthy donors). *P < 0.05, **P < 0.01, and ***P < 0.001.

Data concatenation and UMAP visualization revealed specificity-based clustering of CD8+ T cells (Fig. 5C). Flu-specific CD8+ T cells were found mainly in cluster 9, which displayed high expression levels of TCF-1 and CCR7 and low to intermediate expression levels of TOX, cytolytic proteins, and IRs (Figs. 4H and 5, D and E). Many EBV-specific CD8+ T cells also localized to cluster 9, but additional clusters were identified with detectable expression of TOX and variable expression levels of cytolytic proteins and IRs (clusters 12, 10, 2, 4, 6, 7, and 8; Figs. 4H and 5, D and E). In contrast, CMV-specific CD8+ T cells predominated in clusters with high expression levels of TOX and cytolytic proteins and intermediate to high expression levels of IRs (clusters 11, 18, 21, and 23), whereas HIV-specific CD8+ T cells were most prevalent in clusters 1, 2, 3, 4, 5, and 6, which displayed high expression levels of TOX, low to intermediate expression levels of TCF-1, and variable expression levels of cytolytic proteins and IRs (Fig. 4H and 5, D and E).

To determine the functional relevance of TOX, we stimulated peripheral blood mononuclear cells (PBMCs) with pools of peptides corresponding to known optimal epitopes derived from Flu, CMV, EBV, and HIV and measured GzmB and perforin alongside surface mobilization of CD107a and the intracellular production of interferon-γ (IFN-γ), tumor necrosis factor (TNF), and interleukin-2 (IL-2). Virus-specific CD8+ T cells that produced IL-2, an early-differentiated function, rarely expressed TOX, whereas virus-specific CD8+ T cells that were loaded with cytolytic proteins and produced the effector cytokines IFN-γ and TNF commonly expressed TOX (Fig. 5F). More detailed analysis confirmed that IL-2 expression was more commonly produced within the TOXTCF-1+ compartment, while GzmB and perforin were primarily expressed by TOX+TCF-1 virus-specific memory CD8+ T cells (fig. S7F).

To understand the TOX and TCF-1 expression dynamics in vivo, we longitudinally tracked reactivation of CMV-specific CD8+ T cells in individuals with successful bone marrow transplantation. CMV-specific CD8+ T cells play a critical role in virus control during CMV reactivation, as evidenced in clinical trials using infusion of CMV-specific T cells to prevent CMV-related organ dysfunction (36) and in immune-deficient patients that experience uncontrolled viral replication and end-organ disease as a consequence of impaired T cell responses (37, 38). CMV-specific CD8+ T cells were readily detected (fig. S8A) and displayed a stable TOX+TCF-1/TOX+TCF-1+ profile before, recently, and long-term after bone marrow transplantation (fig. S8B), implying a stable TOX imprint on these cells also following high CMV antigen burden. We also recruited patients during the symptomatic acute phase of Flu infection and observed an expanded CD38hiKi-67+ population indicative of an early expanded antigen-specific population (3941). CD38hiKi-67+ demonstrated a TOX+TCF-1 and TOX+TCF-1+ profile (fig. S8C), suggesting that Flu-specific CD8+ T cells experience a more dynamic range of TOX expression similar to what has been observed in acute LCMV infection (15). Collectively, these observations demonstrate the dynamic nature of TOX and TCF-1 expression in virus-specific CD8+ T cells that is directly associated with different functional profiles.

HIV-specific memory CD8+ T cells from elite controllers exhibit increased expression of TCF-1

To correlate these findings with immune efficacy, we recruited HIV+ elite controllers (ECs), defined as individuals who maintain undetectable levels of plasma viral RNA in the absence of ART (table S4). As expected, we found that PD-1, CD38, and CD39 were expressed at higher frequencies among HIV-specific CD8+ T cells from HIV+ viremic donors relative to HIV-specific CD8+ T cells from ECs (Fig. 6A). Irrespective of disease status, a vast majority of HIV-specific CD8+ T cells expressed TOX (Fig. 6B). However, TCF-1+ cells were present at higher frequencies among HIV-specific CD8+ T cells from ECs relative to HIV-specific CD8+ T cells from HIV+ viremic donors (Fig. 6B). Further analysis revealed not only an increased fraction of TOXTCF-1+ but also lower TOXTCF-1 frequencies of HIV-specific CD8+ T cells from ECs relative to HIV-specific CD8+ T cells from HIV+ viremic donors (fig. S9A).

Fig. 6 Expression of TOX and TCF-1 in functional HIV-specific CD8+ T cells.

(A) Percent expression of the indicated markers among tetramer-defined HIV-specific CD8+ T cells from ECs (n = 11; light red), HIV+ aviremic donors on ART (n = 15; red), and HIV+ viremic donors (n = 16; dark red). (B) Percent expression of TOX and TCF-1 in tetramer-defined HIV-specific CD8+ T cells from the donor groups in (A). (C) Colored graphs: Percent frequencies of IL-2+ (left) and CD107a+ cells (right) among all responsive HIV-specific CD8+ T cells from the donor groups in (A) after stimulation of PBMCs with pools of peptides corresponding to known optimal epitopes derived from HIV. Donor-matched graphs: Expression intensities of TOX and TCF-1 in IL-2 versus IL-2+ (left) and CD107a versus CD107a+ HIV-specific CD8+ T cells (right). (D) Expression of TOX and TCF-1 in GzmB+perforin+IFN-γ+ (red) and IL-2+ HIV-specific CD8+ T cells (blue). *P < 0.05, **P < 0.01, and ***P < 0.001.

In response to cognate peptide stimulation, higher frequencies of HIV-specific CD8+ T cells from ECs produced TNF and IL-2 relative to HIV-specific CD8+ T cells from either HIV+ viremic donors or HIV+ aviremic donors on ART (Fig. 6C and fig. S9B). Moreover, TNF+ HIV-specific CD8+ T cells expressed lower levels of TOX on a per-cell basis relative to TNF HIV-specific CD8+ T cells (fig. S9B), IL-2+ HIV-specific CD8+ T cells expressed lower levels of TOX and higher levels of TCF-1 on a per-cell basis relative to IL-2 HIV-specific CD8+ T cells, and CD107a+ HIV-specific CD8+ T cells expressed lower levels of TCF-1 on a per-cell basis relative to CD107a HIV-specific CD8+ T cells (Fig. 6C). In addition, most GzmB+perforin+IFN-γ+ HIV-specific CD8+ T cells expressed TOX without TCF-1, whereas most IL-2+ HIV-specific CD8+ T cells expressed TCF-1 (Fig. 6D). Collectively, these results identify a functional dichotomy associated with the differential expression of TOX and TCF-1 that correlates with immune control of HIV.

DISCUSSION

In this study, we investigated the extent to which two HMG box transcription factors, TOX and TCF-1, shape the processes of exhaustion and memory differentiation among subpopulations of human CD8+ T cells. Under homeostatic conditions, effector memory CD8+ T cells primarily expressed TOX, whereas naive and early-differentiated memory CD8+ T cells primarily expressed TCF-1. Cytolytic gene and protein expression signatures among human CD8+ T cells were also defined by the expression of TOX. In the context of ongoing viral replication, dysfunctional HIV-specific CD8+ T cells commonly expressed TOX, which clustered with various activation markers and IRs, and rarely expressed TCF-1. However, functionally competent memory CD8+ T cells specific for CMV or EBV also expressed TOX. A similar phenotype was observed among HIV-specific CD8+ T cells from ECs.

TOX and TCF-1 were originally described as transcriptional regulators of lymphocyte development and maturation (4245). More recent studies have shown that TOX is expressed in most exhausted CD8+ T cells specific for various cancers and chronic viruses (1418), whereas TCF-1 is up-regulated in precursor exhausted CD8+ T cells (1113). TOX and TCF-1 are both required for the survival of murine CD8+ T cells that recognize chronic viral antigens (14, 18). In humans, TCF-1 has been associated with early-differentiated CD8+ T cells (46, 47), whereas TOX has been associated with exhausted CD8+ T cells (14, 15, 48, 49). However, these correlative studies did not assess both transcription factors simultaneously in relation to different aspects of CD8+ T cell immunobiology. We identified three discrete memory CD8+ T cell populations in humans based on the expression of TOX and TCF-1. Early-differentiated TCM and TEM cells displayed a TOXTCF-1+ phenotype with negligible expression of IRs. A vast majority of Flu-specific CD8+ T cells were encapsulated within this population, consistent with the acute memory phenotype described in mice infected with LCMV (15). Most healthy donors also harbored a progenitor-like TOX+TCF-1+ subset, which commonly expressed CD28, and an effector-like TOX+TCF-1 subset, characterized by the expression of cytolytic proteins. Moreover, IRs were expressed primarily in conjunction with TOX, but not all TOX+ cells expressed IRs. We found that TOX was frequently expressed in CMV-specific CD8+ T cells, which are highly cytolytic and polyfunctional (21, 50, 51). These effector properties have also been shown to confer protection against simian immunodeficiency virus in nonhuman primate models (52, 53). Our study is lacking tissue data, and given that previous studies have demonstrated increased expression levels of TOX in PD-1hiCD8+ T cells from tumors (15, 16), it might still be possible that forced or increased expression intensity of TOX might mark terminally exhausted CD8+ T cells. Our observations nevertheless suggest that exhaustion per se is not necessarily defined solely by TOX expression. Accordingly, balanced TOX expression appears to define biologically protective human effector memory CD8+ T cells, driven to a functional optimum by, for instance, CMV.

In contrast to laboratory mice, humans are commonly infected with many different chronic viruses, including herpesviruses, such as CMV and EBV, polyomaviruses, and endogenous retroviruses (54). Ongoing antigen exposure is therefore likely universal across various specificities in healthy donors, potentially explaining the widespread occurrence of TOX+ memory CD8+ T cells. These less virulent chronic viruses have evolved in synergistic equilibrium with the host such that the human immune system, as it exists today, represents the unfinished product of countless adaptations over millions of years (55). Accordingly, balanced TOX expression may serve as a positive regulator of survival that maintains highly functional and protective CD8+ T cell responses directed against ineradicable pathogens (23, 56, 57).

Previous studies have shown that exhaustion is stably imprinted at the epigenetic level (23, 57). Our data support the notion that a similar phenomenon applies in the case of TOX. For example, we found that HIV-specific CD8+ T cells expressed TOX more commonly than Flu-specific CD8+ T cells, even in donors with low plasma burdens of viral RNA. In addition, TOX was expressed less frequently in CD8+ T cells from HIV donors relative to CD8+ T cells from HIV+ aviremic donors on ART. Of particular note, we also found that TCF-1 was expressed at relatively high frequencies alongside TOX in HIV-specific CD8+ T cells from ECs. These cells typically produce multiple cytokines and proliferate vigorously in response to antigen encounter (19, 22). Moreover, we noted that IL-2 production was associated with increased expression levels of TCF-1 and decreased expression levels of TOX, whereas most cytolytic HIV-specific CD8+ T cells displayed a TOX+TCF-1 phenotype. As such, a functional spectrum of HIV-specific CD8+ T cells can therefore be defined on the basis of the transcription factors TOX and TCF-1.

Our collective dataset provides an expression atlas for TOX and TCF-1 in the human CD8+ T cell landscape under physiological conditions and under challenge from a relentless pathogen. We found that TOX demarcates cytolytic effector CD8+ T cells, which reside primarily in the TEM and TEMRA compartments, as well as exhausted CD8+ T cells. On the basis of these findings, we propose that TOX is a universal regulator of human memory CD8+ T cells specific for chronic viruses, such as CMV, EBV, and HIV.

MATERIALS AND METHODS

Samples

Peripheral blood was collected from individuals classified as HIV (n = 30), HIV+ viremic [viral load (VL) > 1000 RNA copies/ml; n = 23], HIV+ aviremic on ART (VL < 50 RNA copies/ml for >1 year; n = 17), and HIV+ ECs (VL < 50 RNA copies/ml for at least three consecutive visits in the absence of ART; n = 10). Blood was also collected from four adult patients with confirmed influenza A virus infection presenting with symptoms for 2 to 4 days and at 6 weeks later when patients had cleared their infection [as confirmed by lack of symptoms and negative polymerase chain reaction (PCR)]. In addition, blood was also collected from six individuals undergoing allogeneic hematopoietic stem cell transplantation. Blood was collected from the recipients at three time intervals: before day 0, days 14 to 60, and days 90 to 360 after transplantation. All grafts from the donors were CMV+, and all recipients experienced CMV reactivation that was eventually controlled. Recruitment occurred at four sites under protocols approved by the relevant Institutional Review Boards: Karolinska University Hospital (Stockholm, Sweden), Asociación Civil Impacta Salud y Educación (Lima, Peru), University of Alabama at Birmingham (Birmingham, AL, USA), and Massachusetts General Hospital (Boston, MA, USA). All samples were collected in accordance with the principles of the Declaration of Helsinki. Donor groups and clinical parameters are summarized in table S4.

Flow cytometry

PBMCs were isolated from whole blood via standard density gradient centrifugation and cryopreserved in liquid nitrogen. Cryopreserved PBMC samples were thawed, resuspended at 2 × 106 cells/ml, and rested overnight at 37°C in complete medium (RPMI 1640 medium supplemented with 10% fetal bovine serum, 1% l-glutamine, and 1% penicillin/streptomycin) in the presence of deoxyribonuclease (DNase) I (10 U/ml) (Roche) as described previously (21, 58, 59). Cells were then washed in fluorescence-activated cell sorting (FACS) buffer [phosphate-buffered saline (PBS) supplemented with 2% fetal bovine serum and 2 μM EDTA] and stained with major histocompatibility complex (MHC) class I tetramers and/or a directly conjugated antibody specific for CCR7 (clone G043H7, BioLegend) for 10 min at 37°C. Other surface markers were detected via the subsequent addition of an optimized panel of directly conjugated antibodies for 20 min at room temperature, and viable cells were identified by exclusion using the LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher Scientific). Cells were then washed again in FACS buffer and fixed/permeabilized using a FoxP3/Transcription Factor Staining Buffer Set (eBioscience). Intracellular markers were detected via the subsequent addition of an optimized panel of directly conjugated antibodies for 1 hour at room temperature. Stained cells were fixed in PBS containing 1% paraformaldehyde (Biotium) and stored at 4°C. All samples were acquired within 24 hours using a FACSymphony A5 (BD Biosciences). Data were analyzed with FlowJo software version 10.6.1 (FlowJo LLC). Gating strategies are similar to previous publications (58, 59), and all gates are based on fluorescent minus one (FMO) stainings or negative controls (specific populations).

Antibodies

Directly conjugated antibodies with the following specificities were used in flow cytometry experiments: CD3-BV605, CD3-BV650, or CD3-BUV805 (clone UCHT1), CD8-BUV395 (clone RPA-T8), CD14–allophycocyanin (APC)–Cy7 (clone MϕP9), CD15-BUV395 (clone W6D3), CD16-A700 (clone 3G8), CD19-BUV737 (clone SJ25C1), CD28-BUV563 (clone CD28.2), CD38-BUV496 (clone HIT2), CD49a-BUV615 (clone SR84), CD69-BV750 (clone FN50), CD95-BB630 (clone DX2), CD107a–phycoerythrin (PE)–CF594 (clone H4A3), CXCR5–APC-R700 (clone RF8B2), GzmB-A700 or GzmB-BB790 (clone GB11), IFN-γ–fluorescein isothiocyanate (FITC) or IFN-γ–PE (clone B27), IL-2–APC-R700 (clone MQ1-17H12), Ki-67–BB660 (clone B56), Lag-3–BUV661 (clone T47–530), perforin-BB700 (clone dG9), TCRγδ–PE-CF594 (clone B1), TIGIT-BUV737 (clone 741182), and TNF-BV650 (clone Mab11) from BD Biosciences; CCR7–APC-Cy7 (clone G043H7), CD8-BV570 or CD8-BV605 (clone RPA-T8), CD14-BV510 (clone M5E2), CD19-BV510 (clone HIB19), CD45RA-BV570 or CD45RA-BV650 (clone HI100), CD39-BV711 (clone A1), CD56-BV711 (clone HCD56), CD103-BV605 (clone Ber-ACT8), CD127-BV421 or CD127-BV785 (clone A019D5), CD161-BV605 (clone HP-3G10), PD-1–PE-Cy7 (clone EH12.2H7), perforin-BV421 (clone B-D48), Tim-3–BV650 (clone F38-2E2), TCR Vα7.2–PE (clone 3C10), and 2B4–PE/Dazzle 594 (clone C1.7) from BioLegend; TCF-1–AF488 or TCF-1–PE (clone C63D9) from Cell Signaling Technology; TOX-AF647 (clone REA473) from Miltenyi Biotec; and CD4–PE-Cy5.5 (clone S3.5) and CD25–PE-Cy5 (clone CD25-3G10) from Thermo Fisher Scientific.

Peptides

Pools of peptides corresponding to known optimal epitopes derived from CMV (n = 14), EBV (n = 26), Flu (n = 17), and HIV (n = 22) were purchased from Peptides & Elephants GmbH. All peptides were synthesized at a purity of >95%. Lyophilized peptides were reconstituted at 100 mg/ml in dimethyl sulfoxide (DMSO) and further diluted to 100 μg/ml in PBS.

Tetramers

MHC class I tetramers conjugated to BV421 or PE were used to detect CD8+ T cells with the following specificities: CMV NLVPMVATV (NV9/HLA-A*0201), EBV GLCTLVAML (GL9/HLA-A*0201), Flu GILGFVFTL (GL9/HLA-A*0201), HIV FLGKIWPSHK (FK10/HLA-A*0201), HIV ILKEPVHGV (IV9/HLA-A*0201), HIV SLYNTVATL (SL9/HLA-A*0201), EBV RVRAYTYSK (RK9/HLA-A*0301), Flu RVLSFIKGTK (RK10/HLA-A*0301), CMV QYDPVAALFL (QL10/HLA-A*2402), EBV TYGPVFMCL (TL9/HLA-A*2402), HIV KYKLKHIVW (KW9/HLA-A*2402), HIV RYPLTFGW (RW8/HLA-A*2402), CMV TPRVTGGGAM (TM10/HLA-B*0702), HIV GPGHKARVL (GL9/HLA-B*0702), CMV QIKVRVDMV (QV9/HLA-B*0801), EBV RAKFKQLL (RL8/HLA-B*0801), HIV EIYKRWII (EI8/HLA-B*0801), HIV KRWIILGLNK (KK10/HLA-B*2705), HIV ISPRTLNAW (IW9/HLA-B*5701), HIV KAFSPEVIPMF (KF11/HLA-B*5701), and HIV QASQEVKNW (QW9/HLA-B*5701). All tetramers were generated in house as described previously (60).

Functional assays

PBMCs were seeded in complete medium at 2 × 106 cells/ml in 96-well V-bottom plates (Corning) with unconjugated anti-CD28 (clone L293) and anti-CD49d (clone L25; each at 3 μl/ml; BD Biosciences), brefeldin A (1 μl/ml; Sigma-Aldrich), monensin (0.7 μl/ml; BD Biosciences), anti-CD107a–PE-CF594 (clone H4A3; BD Biosciences), and the relevant viral peptides (each at a final concentration of 0.5 μg/ml). Unstimulated negative controls were included in each assay. Cells were analyzed by flow cytometry after incubation for 6 hours at 37°C.

Proliferation assay

CD8+ T cell subsets were flow-sorted using the MA900 Multi-Application Cell Sorter (Sony Biotechnology), labeled with CellTrace Violet (0.5 μM; Thermo Fisher Scientific), and resuspended in complete medium in 96-well U-bottom plates (Corning) with IL-2 (100 IU/ml; PeproTech) and ImmunoCult Human CD3/CD28 T Cell Activator (5 μl/ml; STEMCELL Technologies). IL-2 was replenished on day 3. Cells were analyzed by flow cytometry after incubation for 5 days at 37°C.

Activation of the calcineurin pathway

Naive CD8+ T cells were cultured in complete medium at various densities with ImmunoCult Human CD3/CD28 T Cell Activator (5 μl/ml; STEMCELL Technologies) or ionomycin (100 ng/ml Sigma-Aldrich) and/or phorbol 12-myristate 13-acetate (10 ng/ml; Sigma-Aldrich). Unstimulated negative controls were included in each assay. Cells were analyzed by flow cytometry after incubation for 24 hours at 37°C.

UMAP and Phenograph

FCS3.0 data files were imported into FlowJo software version 10.6.0 (FlowJo LLC). All samples were compensated electronically. Dimensionality reduction was performed using the FlowJo plugin UMAP version 2.2 (FlowJo LLC). The downsample version 3.0.0 plugin and concatenation tool was used to visualize multiparametric data from comparable numbers of total CD8+ T cells per healthy donor (n = 4). Comparable numbers of memory CD8+ T cells from HIV (n = 2) and HIV+ individuals (n = 2) were similarly applied to UMAP. The following parameters were used in these analyses: metric = euclidean, nearest neighbors = 15, and minimum distance = 0.5. Clusters of phenotypically related cells were then detected using Phenograph version 0.2.1. The following markers were considered in each case: CCR7, CD25, CD28, CD38, CD39, CD45RA, CD49a, CD95, CD103, CXCR5, GzmB, Ki-67, Lag-3, perforin, PD-1, TCF-1, TIGIT, Tim-3, TOX, and 2B4. Phenograph clustering was performed using the coordinates from UMAP. Plots were generated using Prism version 8.2.0 (GraphPad Software Inc.).

RNA sequencing

Naive and memory CD8+ T cells (250 cells per subset) were flow-sorted directly into lysis buffer using FACSAria II (BD Biosciences). RNA-seq libraries were prepared from snap-frozen lysates using the SMART-Seq v4 Ultra Low Input RNA Kit (Takara). Briefly, 3′-oligo(dT) primers were hybridized to the poly(A) tails of mRNA molecules, and complementary DNA (cDNA) was generated using SMARTScribe Reverse Transcriptase (Takara) and preamplified using SeqAmp DNA Polymerase (Takara). Cleanup was performed using Agencourt AMPure XP Beads (Beckman Coulter). cDNA was quantified using Qubit 3.0 (Thermo Fisher Scientific), and fragment sizes were evaluated using 2100 BioAnalyzer (Agilent). PCR products were then indexed using a Nextera XT DNA Library Prep Kit (Illumina). Briefly, transcripts were tagmented using Amplicon Tagment Mix (Illumina) and indexed using a Nextera Index Kit (Illumina). Cleanup was performed again using Agencourt AMPure XP Beads (Beckman Coulter). Libraries were pooled, quantified, and sequenced across 75 base pairs (bp) using a paired-end approach with a 150-cycle high-output flow cell on NextSeq 550 (Illumina). Three biological replicates were sequenced per experiment. Fastq files from replicate sequencing runs were concatenated and aligned using STAR software version 2.5.2a and hg38. Depth ranged from 8 million to 13.6 million reads per sample. Raw counts were normalized using DESeq2 (Bioconductor). Normalized data were then transformed to Z scores and analyzed for differential expression using the limma package in R.

Assay of Transposase-Accessible Chromatin using sequencing

Naive and memory CD8+ T cells (30,000 to 50,000 cells per subset) were flow-sorted directly into complete medium using FACSAria II (BD Biosciences). ATAC-seq was carried out as described previously (59). Briefly, cells were pelleted, washed in PBS, and treated with lysis buffer [10 mM tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl2, 0.1% IGEPAL CA-630]. Nuclear pellets were resuspended in transposition buffer containing Tn5 transposase (Illumina) for 30 min at 37°C. Tagmented DNA was purified using the MinElute Reaction Cleanup Kit (Qiagen). Amplified libraries were purified using the QIAQuick PCR Purification Kit (Qiagen) and subjected to paired-end sequencing (38 bp + 37 bp) on NextSeq 550 (Illumina). Two biological replicates were sequenced per experiment. ATAC-seq peaks were called out in each replicate using the (-p 1e-7 --nolambda --nomodel) function in MACS2. A read count table was generated and used to identify differentially enriched regions among subsets using DESeq2 with the parameters fold change > 2 and P < 0.05. De novo Tcf7 and Tox motif analyses were carried out using HOMER. Peaks were visualized using Integrative Genomics Viewer software version 2.3.93.

Single-cell RNA-seq

scRNA-seq data were obtained from the public 10× genomics repository (www.10xgenomics.com/resources/application-notes/a-new-way-of-exploring-immunity-linking-highly-multiplexed-antigen-recognition-to-immune-repertoire-and-phenotype/). CD8+ T cell gene expression data from two healthy donors were normalized with SCTransform and integrated with Seurat v3. Tox+ (1909) and Tcf+ (9430) cells were identified, and differentially expressed genes (DEGs) were determined using DESeq2. The full script used for processing the data is made publicly available on Github (https://github.com/Jb-Gorin/Sekine-et-al.-2020).

Western blot

Naive and memory CD8+ T cells were sorted either directly or after stimulation into complete medium using the MA900 Multi-Application Cell Sorter (Sony Biotechnology). Cells were then pelleted, washed twice in ice-cold PBS, and lysed in RIPA Lysis and Extraction Buffer (Thermo Fisher Scientific). Protein content was quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Normalized amounts of protein in PBS were mixed with 4× loading buffer [62.5 mM tris-HCl (pH 6.8), 2% (w/v) SDS, 10% glycerol, 0.04 M dithiothreitol (DTT), and 0.01% (w/v) bromophenol blue] and heated for 10 min at 90°C. Equal volumes were loaded onto 10% Precast Polyacrylamide Gels (Bio-Rad). Gel electrophoresis was performed for 60 min at 130 V. Proteins were transferred to a polyvinylidene fluoride membrane over 1 hour at 100 V using the Mini-PROTEAN Tetra Cell and Mini Trans-Blot Module (Bio-Rad). Membranes were blocked with 5% skimmed milk in TBS-T [150 mM NaCl, 10 mM tris-HCl (pH 7.5), 0.1% Tween 20] and incubated with the relevant antibodies overnight at 4°C. The following antibodies were used in these experiments: anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (clone 6C5) diluted at 1:2000 (Abcam), anti–TCF-1 (clone 655202) diluted at 1:1000 (BioLegend), and anti-TOX (rabbit polyclonal ab155768) diluted at 1:700 (Abcam). Membranes were then washed in TBS-T, incubated with secondary antibodies conjugated to horseradish peroxidase (diluted at 1:4000) for 1 hour at room temperature, and developed using West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific). Antibody dilutions were prepared using 2% milk in TBS-T. Blots were imaged using the G:BOX Imaging System (Syngene).

Statistics

Differences between unmatched groups were compared using an unpaired t test or the Mann-Whitney U test, and differences between matched groups were compared using a paired t test or the Wilcoxon signed-rank test. Correlations were assessed using the Pearson correlation or the Spearman rank correlation. Nonparametric tests were used if the data were not distributed normally according to the Shapiro-Wilk normality test. All analyses were performed using R studio or Prism 7.0 (GraphPad). Phenotypic relationships within multivariate datasets were visualized using FlowJo software version 10.6.1 (FlowJo LLC).

Correction (19 April 2021): Tables S1 to S4 have been added to the Supplementary Materials PDF.

SUPPLEMENTARY MATERIALS

immunology.sciencemag.org/cgi/content/full/5/49/eaba7918/DC1

Fig. S1. Gating strategy and expression of TOX and TCF-1 in resting and activated CD8+ T cell populations.

Fig. S2. Tox and Tcf7 expression in CD8+ T cells.

Fig. S3. IR expression and distribution among different Seurat clusters.

Fig. S4. UMAP analysis and Phenograph clustering of CD8+ T cell populations in relation to TOX and TCF-1.

Fig. S5. UMAP analysis and Phenograph clustering of memory CD8+ T cells from HIV and HIV+ viremic donors and expression of TOX and TCF-1 in virus-specific memory CD8+ T cells.

Fig. S6. UMAP analysis and Phenograph clustering of memory CD8+ T cells from HIV and HIV+ viremic donors and expression of TOX and TCF-1 in virus-specific memory CD8+ T cells.

Fig. S7. Expression of TOX and TCF-1, functions, and memory markers in virus-specific memory CD8+ T cells.

Fig. S8. Longitudinal assessment of TOX and TCF-1 in virus-specific memory CD8+ T cells.

Fig. S9. Analysis of HIV-specific CD8+ T cell functionality in relation to TOX and TCF-1.

Table S1. List of RNA-seq core signature between TN, TCM, TEM and TEMRA.

Table S2. List of all significant scRNA-seq genes between TCF7+ and TOX+ cells.

Table S3. List of enriched genes in Seurat clusters.

Table S4. Cohort characteristics.

REFERENCES AND NOTES

Acknowledgments: We express our gratitude to all donors, health care personnel, study coordinators, administrators, and laboratory managers involved in this work. Funding: M.B. was supported by the Swedish Research Council (VR), the Karolinska Institutet, the Swedish Society for Medical Research (SSMF), the Jeansson Stiftelser, the Åke Wibergs Stiftelse, the Swedish Society of Medicine, the Cancerfonden, the Barncancerfonden, the Magnus Bergvalls Stiftelse, the Hedlunds Stiftelse, the Lars Hiertas Stiftelse, the Swedish Physician against AIDS Foundation, the Jonas Söderquist Stiftelse, and the Clas Groschinskys Minnesfond. Additional support was provided by R01 and R56 grants from the NIH (AI076066, AI118694, and AI106481 to M.R.B.) and the Penn Center for AIDS Research (AI045008). D.A.P. is a Welcome Trust Senior Investigator (100326/Z/12/Z). C.B. is funded by the European Union’s Horizon 2020 research and innovation program under grant agreement 681137-EAVI2020 and by NIH grant P01-AI131568. Author contributions: M.B. conceived the project. T.S., A.P.-P., S.N., J.-B.G., V.H.W., S.F.-J., S.V., M.Y., A.S.-S., A.G., D.A.P., M.R.B., and M.B. designed and performed experiments. T.S., A.P.-P., S.N., J.-B.G., V.H.W., Q.H., M.U., and M.B. analyzed data. J.K.S., C.B., P.N., E.G., S.L.-L., P.A.G., D.A.P., M.R.B., and M.B. provided critical resources. D.A.P., M.R.B., and M.B. supervised experiments. T.S., A.P.-P., S.N., M.R.B., and M.B. drafted the manuscript. C.B., P.A.G., D.A.P., M.R.B., and M.B. edited the manuscript. Competing interests: The authors declare that they have no competing financial interests, patents, patent applications, or material transfer agreements associated with this study. Data and materials availability: The sequence datasets reported in this paper have been deposited in the Gene Expression Omnibus under accession no. GSE148234.

Stay Connected to Science Immunology

Navigate This Article