CD4⁺ T cell help is required for the formation of a cytolytic CD8⁺ T cell subset that protects against chronic infection and cancer

Summary

Although CD4⁺ T cell “help” is crucial to sustain antiviral immunity, the mechanisms by which CD4⁺ T cells regulate CD8⁺ T cell differentiation during chronic infection remain elusive. Here, using single-cell RNA-sequencing we show that CD8⁺ T cells responding to chronic infection were more heterogeneous than previously appreciated. Importantly, our findings uncovered the formation of a CX₃CR1-expressing CD8⁺ T cell subset that exhibited potent cytolytic function and was required for viral control. Notably, our data further demonstrate that formation of this cytotoxic subset was critically dependent on CD4⁺ T cell help via IL-21, and that exploitation of this developmental-pathway could be used therapeutically to enhance the killer function of CD8⁺ T cells infiltrated into the tumor. These findings uncover additional molecular mechanisms of how “CD4⁺ T cell help” regulates CD8⁺ T cell differentiation during persistent infection, and have implications towards optimizing the generation of protective CD8⁺ T cells in immunotherapy.

Graphical Abstract

Introduction

During chronic viral infection and cancer, CD8⁺ T cells undergo a differentiation process commonly referred to as T cell exhaustion (Wherry and Kurachi 2015). This process is traditionally defined by a stepwise loss of effector functions, eventually leading to cell death (Wherry et al. 2003; Kahan, Wherry, and Zajac 2015). Despite their inability to completely clear the infection, exhausted T cells remain indispensable for viral control (Schmitz et al. 1999; Jin et al. 1999), indicating that some CD8⁺ T cells retain their cytotoxic potential. CD4⁺ T cell help has long been known to be essential for sustaining CD8⁺ T cell function during chronic viral infection (Battegay et al. 1994; Matloubian, Concepcion, and Ahmed 1994; Zajac et al. 1998). More recently, several studies have identified a critical role for CD4⁺ T cell-produced IL-21 in mediating this protective response (Elsaesser, Sauer, and Brooks 2009b; Froöhlich et al. 2009; Yi, Du, and Zajac 2009b; Xin et al. 2015). However, the underlying mechanism(s) by which CD4⁺ T cell–derived IL-21 coordinates this antiviral CD8⁺ T cell program remains incompletely understood.

Previous research indicates that CD8⁺ T cells responding to persistent infection can exist in two distinct transcriptional states, with differential expression of the transcription factors T-bet and Eomes coordinating this cell-fate decision process (Paley et al. 2012). In addition, several recent studies identify that virus-specific CD8⁺ T cells are non-homogenous and can be compartmentalized into at least two major subsets, with a CXCR5^hiTCF-1^hi subset serving as a self-renewing progenitor population that can give rise to a more terminally exhausted CXCR5^loTCF-1^lo subset (Leong et al. 2016; He et al. 2016; Im et al. 2016; Utzschneider et al. 2016). Moreover, CXCR5^hiTCF-1^hi CD8⁺ T cells are implicated as being the major responders to PD-1 checkpoint blockade (He et al. 2016; Im et al. 2016; Siddiqui et al. 2019; Kurtulus et al. 2019), a breakthrough immunotherapy demonstrated to enhance CD8⁺ T cell responses during chronic infections and cancer (Barber et al. 2006; Wherry and Kurachi 2015). These observations prompted us to ask whether additional heterogeneity exists amongst CD8⁺ T cells responding to persistent viral infection and whether CD4⁺ T cells could regulate this multifaceted process of CD8⁺ T cell differentiation to better meet the needs of a chronic infection.

Single cell RNA-seq (scRNA-seq) has emerged as a powerful technique to explore cellular diversity and differentiation states (Papalexi and Satija 2017; Villani et al. 2017). Thus, we sought to apply scRNA-seq to this biological process in order to assess the transcriptional profiles and developmental pathways of CD8⁺ T cells during persistent viral infection, and to determine how CD4⁺ T cell help regulates this intricate process of differentiation.

Results

scRNA-seq reveals broad transcriptional heterogeneity amongst LCMV-specific CD8⁺ T cells

To fully characterize the heterogeneity of CD8⁺ T cells responding to chronic viral infection, we performed scRNA-seq on CD8⁺ T cells specific for the GP_33-41 peptide of LCMV at days 8 and 30 post-infection (Figure 1A, S1A-B). Notably, virus-specific T cells grouped distinctly into nine clusters when visualized by t-distributed stochastic neighbor-embedding analysis (t-SNE) (Figure 1A) (Satija et al. 2015; Waltman and van Eck 2013), which was performed using the top nine principal components (Figure S1C-D). Each cluster appeared to be predominantly specific for either day 8 or day 30, with the exception of Naïve (cluster 1) and Slamf6⁺ (cluster 2) cells, which were present at both time points in either relatively low or high frequencies respectively (Figure 1A-C). In addition, our analyses identified the presence of three other transcriptionally distinct clusters on day 8 p.i. (clusters 7-9), two of which could be distinguished by their active cell cycle phase, with clusters 8 and 9 expressing S phase and G2/M phase markers, respectively (Figure 1A-C, S2A-B). Moreover, our analyses also identified the existence of three major subsets of CD8⁺ T cells on day 30 p.i.: Slamf6⁺ cells (cluster 2), Pdcd1⁺ cells (cluster 4), and Cx3cr1⁺ cells (cluster 6) (Figure 1A-C). Of note, the Slamf6⁺ subset, which was present on both day 8 and day 30, shared similarities with the previously described self-renewing progenitor population of CD8⁺ T cells, particularly in its expression of Tcf7, Icos, Il7r (encoding IL-7Ra), and Id3 (Im et al. 2016; Leong et al. 2016) (Figure 1D, S2C-D). Additionally, expression of Batf, a transcription factor we previously showed to be a critical transducer of IL-21 signaling in CD8⁺ T cells, was increased in this group (Figure S2C-D), (Xin et al. 2015). Expression of Cxcr5 was also enriched within this cluster, although Tcf7 expression seemed to correlate more strongly with Slamf6 expression (Figure S2C-D), possibly indicating that Slamf6 might mark a larger progenitor population that contains the previously identified CXCR5⁺TCF-1⁺ subset (Utzschneider et al. 2016). Together, the Pdcd1⁺ and Cx3cr1⁺ clusters comprise what was previously thought to be a homogeneous group of terminally exhausted cells. Both of these subsets express high amounts of Gzma and Gzmb, supporting their status as functional cytotoxic cells (Figure S2G). They differred, however, in their expression of numerous inhibitory receptors, with the Pdcd1⁺ group expressing increased amounts of Cd244 (encoding 2B4), Cd160, Lag3, Havcr2 (encoding Tim3), and of course, Pdcd1 (encoding PD-1) (Figure 1D, S2E). The Pdcd1⁺ group could also be distinguished by their expression of Cxcr6 and Cd7, as well as the transcription factors Nr4a2, Maf, and Eomes (Figure 1D, S2E). Conversely, Cx3cr1⁺ cells displayed high expression of many killer cell lectin-like receptors (Klre1, Klra9, (Figure 1D, S2F); Klrd1, Klrg1, data not shown). The Cx3cr1⁺ subset was also marked by high expression of the transcription factors Tbx21, Zeb2, Klf2 (Figure 1D, S2F), Id2, and Runx1 (data not shown). Overall, the transcriptional profile of Cx3cr1⁺ cells shares many similarities with short-lived effector cells found during acute infections (Dominguez et al. 2015; Kaech and Cui 2012; Omilusik et al. 2015).

Figure 1. scRNA-seq reveals transcriptional heterogeneity amongst CD8 T cells responding to persistent viral infection.

(A) t-SNE plot displaying clusters identified in a mixture of GP₃₃⁺ CD8 T cells from day 8 and day 30 p.i. with LCMV Cl13. (B) t-SNE plots showing CD8 clusters from day 8 or day 30 time points. (C) Summary dot plot depicting relative frequency of day 8 and day 30 virus-specific CD8 T cell populations. (D). Heatmap showing Z scores for the average expression of given genes within a particular cluster. See also Figures S1 and S2.

We next determined how the largest subset of effector CD8⁺ T cells on day 8 p.i. (cluster 7, D8 effector,) compared to the two major subsets of day 30 effector cells (Pdcd1⁺ and Cx3cr1⁺ clusters). To do this, we utilized two distinct computational approaches. First, we visualized the top 30 markers of each of the three clusters relative to the other two (Figure S3B). Second, a score was given to each cell based on its overall expression of the top 100 markers from each group (Figure S3C). Notably, both sets of analyses demonstrated a high degree of transcriptional divergence between the three-effector clusters. Our analyses did however identify that day 8 effector cells did resemble each of the two day 30 subsets in their expression of a few key genes (Figure S3D-E). Expression of the chemokine receptor Cxcr6, transcription factor Nfatc1, and inhibitory receptors Pdcd1, Cd160 and Lag3 were all shared between day 8 effector cells and the day 30 Pdcd1⁺ cluster (Figure S3D-E). Day 8 effector cells also expressed several markers of day 30 Cx3cr1⁺ effector cells, such as Klrc1, Itgb1, and Cx3cr1 (Figure S3D-E; and not depicted). Therefore, though transcriptionally distinct from day 30 Pdcd1⁺ and Cx3cr1⁺ effector cells, day 8 effector cells have begun to acquire features of both of these subsets.

Differential expression of CX₃CR1 and Ly108 can be used to distinguish 3 major subsets of CD8⁺ T cells during chronic viral infection

We next tested whether we could use the surface markers identified from our scRNA-seq analyses to separate the three predominant populations of virus-specific CD8⁺ T cells that form during chronic viral infection (Figure 2A). To do this, we infected mice with LCMV Cl13 and compared expression of CX₃CR1 (a cluster 6 marker) and Ly108 (encoded by Slamf6, a cluster 2 marker) on virus-specific CD8⁺ T cells. Using these markers, three clearly distinct subsets of CD8⁺ T cells emerged: Ly108⁺, CX₃CR1⁻Ly108⁻, and CX₃CR1⁺, representing clusters 2, 4, and 6 from our scRNA-seq analysis, respectively (Figure 2A-B, S4A). In the spleen, approximately half of GP_33-41-specific CD8⁺ T cells belonged to the CX₃CR1⁺ subset, and this frequency remained relatively stable from day 8 to day 45 p.i. (Figure 2B-C,S4A). On the other hand, Ly108⁺ and CX₃CR1⁻Ly108⁻ subsets comprised between 10-40% of the virus-specific CD8⁺ T cell compartment in the spleen depending on the time of analysis (Figure 2B-C,S4A). Given the differences in expression of chemokine receptors amongst these three subsets, we hypothesized that the frequency of each subset might vary by anatomical location. Notably, CX₃CR1⁺, Ly108⁺ and CX₃CR1⁻Ly108⁻ subsets were found to preferentially reside in distinct anatomical sites, with Ly108⁺ cells being widely restricted to secondary lymphoid tissues (Figure 2D-E). By contrast, the CX₃CR1⁺ subset was the dominant population in the circulation, lungs, and inguinal lymph nodes, whereas the CX₃CR1⁻Ly108⁻ subset was enriched in the bone marrow and the liver (Figure 2D-E).

Figure 2. Expression of Ly108 and CX₃CR1 define three subsets of CD8⁺ T cells during chronic viral infection.

(A) tSNE plot highlighting the three main subsets of CD8⁺ T cells present after establishment of chronic viral infection. (B) Representative flow plots depicting Ly108⁺, CX₃CR1⁻Ly108⁻ and CX₃CR1⁺ GP₃₃⁺ splenic CD8 subsets on day 14 p.i. (C) Kinetics (mean +/− SEM) of subsets in the spleen. Viremia (mean +/− SEM) kinetics overlayed in shaded yellow. (D) Representative flow plots depicting GP₃₃⁺ CD8⁺ T cell subsets in indicated tissues on day 14 p.i. (E) Frequency of CD8⁺T cell subsets in different anatomical locations. (F-G) Representative flow plots (F) and summary data (G) showing relative expression of surface molecules in GP₃₃⁺ subsets on days 21-30 p.i. (H-I) Representative flow plots (H) and summary data (I) showing relative expression of TFs in GP₃₃⁺ subsets (J) Representative flow plots and summary data showing proportion of Ly108⁺, CX₃CR1⁻Ly108⁻ and CX₃CR1⁺ subsets degranulating (CD107a⁺) and producing IFN-γ or co-producing TNF-α and IFN-γ⁺ upon ex vivo GP_33-41 stimulation. (K) Flow plot and summary data showing granzyme B expression in GP₃₃⁺ subsets. (L) Summary data depicting relative cytotoxicity of Ly108⁺, CX₃CR1⁻ Ly108⁻ and CX₃CR1⁺ subsets against peptide-pulsed target EL4 cells. GP₃₃-specific subsets were sort-purified on day 21 p.i. and cultured with target cells at a 5:1 ratio. Data (Mean+/− S.D. in E,G, I-K or Mean+/− S.E.M. in (L)) are from 3-8 mice and are representative of at least 3 independent experiments.*p<0.05 **p<0.01, ***p<0.0001. See also Figure S4.

To determine whether gene expression of surface molecules and transcription factors that we observed in our scRNA-seq analyses correlated with protein expression, we co-stained LCMV-specific CD8⁺ T cells for CX₃CR1, Ly108, and a representative set of markers. Consistent with our scRNA-seq data, we identified that the CX₃CR1⁺ subset displayed increased expression of KLRG1, KLRa9, and T-bet, whereas Ly108⁺ cells exhibited high expression of CXCR5, CD127, and TCF-1 (Figure 2F-I). Importantly, the CX₃CR1⁻Ly108⁻ subset shared a similar expression profile to that of cluster 4 (Pdcd1⁺) cells and displayed elevated expression of CXCR6 and Eomes, and had increased expression of multiple co-inhibitory receptors, including PD-1, Tim3, and 2B4 (Figure 2F-I). Ex vivo functional analyses of these 3 major subsets indicated that Ly108⁺ CD8⁺ T cells exhibitted an enhanced capacity to co-produce IFN-γ and TNF-α upon GP_33-41 peptide stimulation, whereas the CX₃CR1⁻Ly108⁻ subset displayed the lowest potential to produce pro-inflammatory cytokines, consistent with their exhausted phenotype (Figure 2J,S4C). Notably, we further identified that CX₃CR1⁺ CD8⁺ T cells displayed increased expression of granzyme B (Figure 2K), which correlated with their augmented cytotoxicity against peptide-pulsed target cells (Figure 2L). These data suggest that the CX₃CR1⁺ subset has enhanced cytolytic function and may be a potent killer of virus-infected cells. Collectively, and in line with our scRNA-seq analyses, these findings demonstrate that differential expression of CX₃CR1 and Ly108 can be used to distinguish 3 major subsets of CD8⁺ T cells during chronic viral infection, each of which displays distinct properties in their phenotype, transcriptional profile, anatomical distribution, and effector function.

Cx3cr1⁺ CD8⁺ T cells are required to control chronic viral infection

As the CX₃CR1⁺ CD8⁺ T cell subset displayed enhanced cytotoxicity ex vivo (Figure 2L), we next sought to determine whether CX₃CR1⁺ CD8⁺ T cells play a critical role in controlling viral replication during persistent infection. To do this, we generated Cx3cr1^DTR dLck-cre mice, which allows for the selective depletion of CX₃CR1⁺ T cells in vivo upon administration of diphtheria toxin (DT) (Figure 3A) (Zhang et al. 2005). As expected, upon sustained treatment with DT, we observed 55-70% decreases in the proportion of circulating and splenic CX₃CR1⁺GP33⁺ CD8⁺ T cells in LCMV Cl13-infected Cx3cr1^DTR dLck-cre⁺ mice as compared to their Cx3cr1^DTR dLck-cre⁻ counterparts (Figure 3B-C) Notably, abrogation of the CX₃CR1⁺ CD8⁺ T cell subset was further accompanied by approximately 4-5-fold increases in viral copy numbers detected in the blood and spleen respectively (Figure 3D). Moreover, we identified that viral load significantly and inversely correlated with the magnitude of virus-specific CX₃CR1⁺ CD8⁺ T cells (Figure 3E). Of note, we observed minimal expression of CX₃CR1 on CD4⁺ T cells residing in either the blood or spleen, and treatment with DT did not result in any significant changes in the proportion of CX₃CR1⁺ CD4⁺ T cells (Figure S5B). Thus, these data indicate that the impaired viral control observed in Cx3cr1^DTR dLck-cre⁺ mice upon DT treatment was likely due to a loss in protective CX₃CR1⁺ CD8⁺ T cells. Collectively, these data demonstrate a requirement for the CX₃CR1⁺ CD8⁺ T cell subset in limiting viral replication during persistent viral infection.

Figure 3. CX₃CR1⁺ CD8⁺ T cells are critical for control over chronic viral infection.

(A). Experimental design. (B) Representative flow plots and (C) summary data showing proportion of circulating or splenic CX₃CR1⁺, Ly108⁺ and CX₃CR1⁻Ly108⁻ CD8⁺T cell subsets in Cx3cr1^DTR dLck-cre⁻ or Cx3cr1^DTR dLck-cre⁺ mice on day 23 p.i. (D) Summary data showing GP copy number in the sera or spleen of experimental mice on d23 p.i. (E) Summary data showing inverse correlations between the proportion of CX₃CR1⁺CD8⁺ T cells and viral burden in the spleen or blood. Data (Mean+/− S.E.M.) in (C-D) are pooled and are from 3-5 mice/group/experiment and are representative of at least 2 independent experiments.*p<0.05 **p<0.01 ***p<0.0001. See also Figure S5.

Slamf6⁺ CD8⁺ T cells give rise to both Pdcd1⁺ exhausted and Cx3cr1⁺ effector CD8⁺ T cell subsets

Recently, it has been proposed that CD8⁺ T cells responding to chronic viral infection exist as two major subsets: a CXCR5^hiTCF-1^hi progenitor population, and a more terminally exhausted CXCR5^loTCF-1^lo population (He et al. 2016; Leong et al. 2016; Utzschneider et al. 2016; Im et al. 2016). However, after uncovering additional heterogeneity using scRNA-seq, we sought to determine the lineage relationship between the three major CD8⁺ T cell subsets that formed during persistent infection. To do this, we first used Monocle 2, which employs a machine-learning algorithm known as reverse graph embedding (RGE), to construct cell differentiation trajectories from our scRNA-seq data (Trapnell et al. 2014; Qiu et al. 2017). For this analysis, cells from the three major subsets present on day 30 p.i. (Slamf6⁺, Pdcd1⁺, and Cx3cr1⁺ cells), as well as clusters 3 (Slamf6⁺Pdcd1⁺) and 5 (Pdcd1⁺Cx3cr1⁺), which may represent intermediate states, were included (Figure 4A), and the top 100 marker genes for the Slamf6⁺, Pdcd1⁺, and Cx3cr1⁺ clusters were used to construct the trajectory. We found that the algorithm predicted a differentiation trajectory with one major branch point, in which Slamf6⁺ cells could form both Pdcd1⁺ and Cx3cr1⁺ cells as terminal endpoints (Figure 4A). In this trajectory, cluster 3 (Slamf6⁺Pdcd1⁺) cells were found mostly between Slamf6⁺ cells and the branch point, and also between the branch point and Pdcd1⁺ cells. Conversely, cluster 5 (Pdcd1⁺Cx3cr1⁺) cells were found mostly between the branch point and the Pdcd1⁺ and Cx3cr1⁺ subsets (Figure 4A). As expected, when visualizing expression of marker genes for the three major subsets over the course of pseudotime, a measure of progress through the differentiation trajectory, we found that genes expressed by both Pdcd1⁺ and Cx3cr1⁺ CD8⁺ T cells appeared later than those expressed by Slamf6⁺ cells, suggesting that Pdcd1⁺ and Cx3cr1⁺ effector subsets represent more terminally differentiated populations (Figure 4B). Next, we scored each cell based on the overall expression of the top 100 marker genes for the Slamf6⁺, Pdcd1⁺ and Cx3cr1⁺ CD8⁺ T cells to determine if clusters 3 and 5 may, in fact, be intermediate states in the differentiation process. We found that Slamf6⁺ markers were also highly expressed by most cluster 3 (Slamf6⁺Pdcd1⁺) cells, while a small subset of cluster 3 cells had already turned down expression of these genes (Figure S6A). Additionally, expression of genes enriched in Pdcd1⁺ cells was turned on in a small subset of cluster 3 cells, and cluster 5 (Pdcd1⁺Cx3cr1⁺) cells expressed an intermediate level of these genes (Figure S6A, middle). Markers of Cx3cr1⁺ cells were also intermediately expressed in cluster 5 cells (Figure S6A, bottom). Collectively, our scRNA-seq trajectory analyses support a model in which Slamf6⁺ CD8⁺ T cells can differentiate into either Pdcd1⁺ or Cx3cr1⁺ cells, with clusters 3 (Slamf6⁺Pdcd1⁺) and 5 (Pdcd1⁺Cx3cr1⁺) cells potentially existing as transition states between these three subsets.

Figure 4. Lineage relationship between CD8⁺ T cell subsets.

(A) Single cell trajectory of CD8⁺ T cell subsets present on day 30 p.i. (clusters 2, 3, 4, 5, and 6). (B) Heatmap showing expression of clusters 2, 4, and 6 markers used for trajectory analysis over pseudotime. (C-H) Experimental design (C) 60,000 CD45.2⁺CD44^hi CD8 T cells from each subset (Ly108⁺, CX₃CR1⁻ Ly108⁻ and CX₃CR1⁺) were transferred to CD45.1 recipient mice. These mice were then infected with LCMV Cl13 one day later. (D-E) Representative flow plots showing expansion (left) and phenotype (right) of cells from the blood on day 9 p.i. (D) or spleen and liver (E) on day 21 p.i. (F-H) Summary of transferred cell frequency (F) and phenotype (G-H) in either the blood (G) or spleen and liver (H). Data (Mean+/− S.D. in (F-H)) are from 3-4 mice/group and are representative of 3 independent experiments.*p<0.05 **p<0.01 ***p<0.0001. See also Figure S6.

To determine the in vivo differentiation trajectory, proliferative potential, and phenotypic stability of the CX₃CR1⁺, Ly108⁺, and CX₃CR1⁻Ly108⁻ (PD-1^hi) subsets, we performed adoptive transfer experiments using congenically marked CD8⁺ T cells. To do this, we sort-purified effector CD8⁺ T cell subsets from infected CD45.2⁺ donor mice and transferred equal numbers of cells to naive CD45.1⁺ mice, which were then subsequently challenged with LCMV Cl13 one day later (Figure 4C, S6B). Notably, compared to CX₃CR1⁻Ly108⁻ cells or CX₃CR1⁺ cells, we found that Ly108⁺ cells displayed vigorous expansion in the blood, spleen and liver (Figure 4D-F), which was associated with their increased expression of Ki67 and incorporation of BrdU relative to the CX₃CR1⁺ subset (Fig S6C). Moreover, and consistent with our Monocle 2 predictions, Ly108⁺ CD8 T cells were found to give rise to both CX₃CR1⁺ and CX₃CR1⁻Ly108⁻ subsets (Figure 4D-E, G-H), further supporting the identity of this subset as the recently described progenitor population with stem cell-like characteristics (He et al. 2016; Im et al. 2016; Leong et al. 2016; Utzschneider et al. 2016). We also identified that CX₃CR1⁺ cells accumulated to a greater extent than CX₃CR1⁻Ly108⁻ donor cells in all tissues assessed (Figure 4D-F), although this did not reach the level of statistical significance. Importantly, the majority of donor-derived cells from the CX₃CR1⁺ subset retained high CX₃CR1 and T-bet expression, and did not differentiate into Ly108⁺ or CX₃CR1⁻Ly108⁻ CD8⁺ T cells (Figure 4D-E, G-H, S6D), further indicating their terminal differentiation status. By contrast, although CX₃CR1⁻Ly108⁻ (PD-1^hi) donor cells accumulated to the lowest extent, approximately 40-50% of their recovered progeny acquired high CX₃CR1 and T-bet expression (Figure 4D-E, G-H, S6D). This observation could potentially be explained by additional heterogeneity residing among the CX₃CR1⁻Ly108⁻ population, such as their encompassing either Slamf6⁺Pdcd1⁺ or Pdcd1⁺Cx3cr1⁺ intermediate populations (clusters 3 and 5 (shown in Figure 1), that are in transitional differentiation states, and which may have distinct developmental potential from the PD-1^hi terminally exhausted subset. Overall, our data are likely suggestive of a bifurcation model wherein Ly108⁺ cells can directly form either CX₃CR1⁻Ly108⁻ or CX₃CR1⁺ cells, with a small fraction of cells within the CX₃CR1⁻Ly108⁻ subset maintaining their potential to develop into CX₃CR1⁺ cells.

CD4⁺ T cell help via IL-21 is critical for the generation of the CX₃CR1⁺ CD8⁺ T cell subset

Given this unexpected heterogeneity in the antiviral CD8⁺ T cell pool, coupled to the critical role for CD4⁺ T cell help in sustaining CD8⁺ T cell responses during persistent viral infection, we next sought to determine whether CD4⁺ T cell help regulates this multifaceted process of CD8⁺ T cell differentiation. Notably, CD8⁺ T cell intrinsic IL-21R signaling has previously been demonstrated to be essential for sustaining CD8⁺ T cell responses during chronic viral infection, and CD4⁺ T cells are implicated as being the major source of IL-21 (Elsaesser, Sauer, and Brooks 2009b; Fröhlich et al. 2009; Yi, Du, and Zajac 2009b; Xin et al. 2015). In line with this, our data identified that CD4⁺ T cells comprised over 90% of IL-21-producing lymphocytes in the spleen during LCMV Cl13 infection (Figure S7A). Nevertheless, how CD4⁺ T cell-derived IL-21 and CD8⁺ T cell-intrinsic IL-21R signaling regulate CD8⁺ T cell differentiation during chronic viral infection remains unclear. To gain insight into how the IL-21R signaling pathway may impact CD8⁺ T cell differentiation during chronic viral infection, we first compared the relative expression of Il21r on the three main day 30-LCMV-specific effector subsets. At least moderate amounts of Il21r expression could be detected in each CD8⁺ T cell subset (Figure 5A), although our scRNA-seq analyses identified that CX₃CR1⁻Ly108⁻ (PD-1^hi) cells displayed the highest expression of Il21r, possibly suggesting a role for CD4⁺ T cell help in regulating the differentiation of this phenotypically and functionally exhausted subset. To test this hypothesis, we employed three different experimental approaches. First, we used a transient CD4⁺ T cell-depletion model that is known to result in sustained viremia and severe CD8⁺ T cell dysfunction during chronic LCMV infection (Battegay et al. 1994; Matloubian, Concepcion, and Ahmed 1994; Zajac et al. 1998). As expected, mice depleted of CD4⁺ T cells one day before LCMV Cl13 infection exhibited significantly impaired viral control (Figure S7B). Notably, depletion of CD4⁺ T cells resulted in ~50-75% decreases in the proportion of CX₃CR1⁺ LCMV-specific CD8⁺ T cells circulating in the blood and inhabiting the spleen (Figure 5B-C). While this decrease in the CX₃CR1⁺ subset was accompanied predominantly by an increase in the CX₃CR1⁻Ly108⁻ subset in the blood, the frequency of both Ly108 and CX₃CR1⁻Ly108⁻ subsets were significantly increased in the spleen in the absence of CD4⁺ T cell help (Figure 5B-C). These differences in the subset distribution of LCMV-specific CD8⁺ T cells were further associated with markedly reduced expression amounts of T-bet to Eomes (Figure S7E) indicating that changes in the overall transcriptional program may be responsible for these alterations in subset skewing. Secondly, we used a co-adoptive transfer model wherein WT and Il21r^−/− P14 cells, which express a transgenic TCR specific for GP_33-41 (Pircher et al. 1990), were transferred into congenically marked recipient mice at a 1:1 ratio before initiating LCMV Cl13 infection the next day (Figure 5D,S7F-I). Notably, use of this experimental design helps control for WT and Il21r^−/− P14 cells to experience a similar degree of exposure to both viral load and inflammation in the same host. Importantly, deletion of IL-21R signaling in P14 transgenic CD8⁺ T cells strongly abrogated the development of the CX₃CR1⁺ CD8⁺ T cell subset, as witnessed by ~75% decreases in the frequency of splenic CX₃CR1⁺ P14 cells (Figure 5D). In accord with our CD4⁺ T cell depletion experiments, we observed substantial reductions in the relative expression of T-bet to Eomes in Il21r^−/− P14 cells (Figure S7H). Of note however, despite this skewing in CD8⁺ T cell subset distribution, we observed a similar capacity between WT and IL-21R-deficient P14 cells to degranulate and produce IFN-γ upon ex vivo peptide stimulation (Figure S7I). This may possibly be due to IL-21R-deficient P14 cells retaining a high proportion of Ly108⁺ progenitor CD8⁺ T cells, which we demonstrate exhibit a high capacity to degranulate and produce effector cytokines upon ex vivo peptide stimulation (Figure 2). Lastly, to assess whether CD8⁺T cell-intrinsic IL-21 signaling was important for the formation of virus-specific CX₃CR1⁺ effector cells originating from a polyclonal repertoire, we generated WT and Il21r^−/− CD8⁺ T cell mixed bone marrow chimera mice (Figure S7J) and challenged them with LCMV Cl13. Notably, and consistent with our P14 adoptive transfer experiments, abrogation of IL-21 signaling in the CD8⁺ T cell compartment resulted in a severe impairment in the development of GP_33-41⁺ CX₃CR1⁺ CD8⁺ T cells during chronic LCMV infection (Figure S7K), which was further associated with decreased control over viral replication (Figure S7L). Taken together, these data indicate a critical role for CD4⁺T cell-derived IL-21 in promoting the formation of CX₃CR1⁺ cytotoxic CD8⁺ T cells, likely by facilitating the differentiation of Ly108⁺ progenitor cells along a CX₃CR1⁺ developmental pathway, while simultaneously limiting CD8⁺ T cell differentiation towards the PD-1^hi exhausted subset.

Figure 5. CD4⁺ T cell help via IL-21 is critical for the generation of CX₃CR1⁺ CD8⁺ T cells.

(A) tSNE plot or violin plots highlighting the three major subsets of CD8 T cells during chronic LCMV infection, and their relative expression of Il21r. (B-C). Representative flow plots (B) and summary data (C) showing the CX₃CR1⁺, Ly108⁺ or CX₃CR1⁻Ly108⁻ subset distribution of GP₃₃⁺CD8 T cells in the blood or spleen of control or CD4-depleted mice on day 21 p.i. (D) Experimental design (top); representative flow plots (bottom) and summary data (right) depicting the subset distribution of WT and Il21r^−/− P14 cells on day 21 p.i. Data (Mean+/− S.E.M.) in (C-D) are pooled and are from 3-5 mice/group/experiment and are representative of at least 3 independent experiments.*p<0.05 **p<0.01 ***p<0.0001. See also Figure S7.

PD-L1 Blockade is insufficient to rescue CX₃CR1⁺ differentiation among un-helped CD8⁺ T cells

Given the extreme defect in the formation of the CX₃CR1⁺ subset in the absence of CD4⁺ T cell help, and the ensuing increase in the proportion of CX₃CR1⁻Ly108⁻ exhausted cells, we next questioned whether PD-L1 blockade, an immunotherapy known to partially reverse T cell exhaustion during persistent infection and cancer (Barber et al. 2006; Wherry and Kurachi 2015), could rescue the development of the CX₃CR1⁺ subset in LCMV Cl13-infected mice. Notably, although PD-L1 blockade in CD4⁺ T cell-depleted mice augmented both the magnitude of the GP_33-41⁺ CD8⁺ T cell response and the capacity of LCMV-specific cells to degranulate and secrete IFN-γ, it was not sufficient to rescue the formation of the CX₃CR1⁺ subset (Figure 6A-C,F). Similarly, α-PD-L1 treatment in CD4⁺ T cell-intact mice also had no appreciable effect on the subset distribution of GP_33-41⁺ CD8⁺ T cells (Figure 6A,C), despite resulting in an 80-90% reduction in viremia compared to control mice (Figure 6G). Additionally, despite displaying enhanced accumulation and augmented cytokine production, the majority of LCMV-specific CD8⁺ T cells from combinatorial α-CD4 and α-PD-L1-treated mice were actually found to reside within the CX₃CR1⁻Ly108⁻ subset and coordinately expressed increased amounts of PD-1 and Tim3, and diminished T-bet to Eomes levels as compared to all other experimental groups (Figure 6A,D-E). Moreover, α-PD-L1 treatment (beginning week 3 of LCMV Cl13 infection) in CD4⁺T cell-depleted mice had only a minor (and non-statistically significant) impact on viral control (Figure 6G), bringing to question the efficacy and sustainability of PD-L1 checkpoint blockade in the absence of CD4⁺ T cell help (and consequently the CX₃CR1⁺ subset). Taken together, these data indicate that the PD-1 inhibitory pathway is unlikely to play a major role in regulating the differentiation of CX₃CR1⁺ CD8⁺ T cells during chronic viral infection. Moreover, these data further demonstrate that blockade of the PD-1-PD-L1 axis is insufficient to rescue the CX₃CR1-differentiation defect of “un-helped” CD8⁺ T cells.

Figure 6. Blockade of PD-1-PD-L1 axis is insufficient to rescue the CX₃CR1-differentiation defect of un-helped CD8⁺ T cells.

(A-F) Separate groups of control or CD4-depleted mice were infected with LCMV Cl13 and treated with isotype or anti-PD-L1 antibodies on days 16, 19, and 22 p.i. GP₃₃⁺CD8 T cell responses were assessed on day 23 p.i. (A-B) Representative flow plots (A) and summary data (B-C) showing the proportion of splenic CD8 T cells (B) and their subset distribution (C). (D) Relative expression (gMFI) of T-bet to Eomes in GP₃₃⁺CD8⁺ T cells from indicated experimental groups. (E) Summary graph showing the relative expression of inhibitory molecules on GP₃₃⁺ cells from experimental mice. (F) Representative flow plots and summary data depicting the proportion of CD107a⁺IFN-γ⁺CD8 T cells upon ex vivo GP_33-41 stimulation. (G) Summary data showing viremia in experimental mice on day 23 p.i. Data (Mean+/− S.E.M.) in (B-G) are pooled and are from 3 mice/group/experiment and are representative of at least 2 independent experiments.*p<0.05 **p<0.01 ***p<0.0001

Provision of CD4⁺ T cell help in the form of IL-21 enhances CX₃CR1⁺ CD8⁺ T cell formation and tumor control

Lastly, given the striking resemblance between CD8⁺ T cells responding to chronic viral infection and tumor-infiltrating T cells (TILs) (Ahmadzadeh et al. 2009; Barber et al. 2006; Blackburn et al. 2009; Sakuishi et al. 2010), we next sought to investigate whether CD8⁺ T cells invading the tumor microenvironment also displayed a similar pattern of differentiation as virus-specific cells during LCMV Cl13 infection. To do this, we employed the well-characterized B16-F10 melanoma model (Overwijk and Restifo 2001), which is known to result in an extreme form of T cell dysfunction and for which current immunotherapy strategies are rather ineffective (Overwijk et al. 2003; Wang et al. 1998; Ahmadzadeh et al. 2009; Moroz et al. 2004). Importantly, even after preconditioning with sublethal irradiation (Klebanoff et al. 2005; Gattinoni et al. 2005; Hanada et al. 2019), the majority of CD8⁺ tumor infiltrating lymphocytes ( TILs) lacked expression of both CX₃CR1 and Ly108 (Figure 7A), consistent with a high level of T cell exhaustion occurring in the tumor microenvironment. Moreover, and similar to that observed in our persistent infection model, CX₃CR1⁻Ly108⁻ cells displayed uniformly high expression of PD-1 (Figure 7A). Of note, we detected minimal Ly108⁺ cells in the tumor, possibly due to their preferential residence in secondary lymphoid tissues (Figure 2). Previous reports have identified that adoptive cell therapy (ACT) with CD4⁺ T cells, can improve CD8⁺ T cell responses and enhance control over tumor progression (Lu et al. 2014; Muranski et al. 2008; Purwar et al. 2012; Vegran et al. 2014; Sondergaard et al. 2010). Many of these studies have further identified that the therapeutic benefit of this CD4⁺ T cell-mediated therapy is associated with CD4⁺ T cell secretion of IL-21 (Moroz et al. 2004; Vegran et al. 2014; Sondergaard et al. 2010), although it remains unclear how CD4⁺ T cell help regulates the differentiation trajectory of CD8⁺ T cells. Thus, we next asked whether provision of IL-21 via CD4⁺ T cell help could enhance the development of the CX₃CR1⁺ subset and its infiltration into the tumor, and whether this could confer protection against tumor growth. To generate tumor-specific IL-21⁺ CD4⁺ T cells, we used an in vitro activation approach wherein CD4⁺ T cells from IL-21 turbo red fluorescent protein (IL-21-tRFP) reporter mice (Shulman et al. 2014) were primed with tumor-pulsed dendritic cells (DCs) under T helper-17 (Th17) cell conditions. Next, IL-21-tRFP⁻ and IL-21-tRFP⁺ Th17 cells were sort-purified and ACT was administered to mice that were previously injected with B16-F10 melanoma cells 7-10 days prior (Figure 7B). We then proceeded to measure tumor burden in these experimental mice over time and after 8 days post-ACT, we assessed polyclonal T cell responses in the tumor. Remarkably, recipients of IL-21-tRFP⁺ CD4⁺ T cells displayed over a 2-fold increase in the proportion of tumor infiltrating CX₃CR1⁺ CD8⁺ T cells, which was highly associated with their enhanced control over tumor growth (Figure 7C-D). Moreover, the proportion of CX₃CR1⁺ CD8⁺ T cells strongly correlated with the magnitude of the IL-21⁺ CD4⁺ T cell helper response, which in turn was also found to inversely correlate with tumor burden (Figure 7E-F). Thus, our data indicate that provision of IL-21⁺ CD4⁺ T cell help in the form of ACT is sufficient to augment the expansion of CX₃CR1⁺ CD8⁺ TILs, which is mechanistically linked to improved tumor control. To determine whether the observed enhancement in CX₃CR1⁺ CD8⁺ TIL formation and tumor control afforded by IL-21⁺ CD4⁺ T cell ACT is dependent on direct IL-21 signaling in CD8⁺ T cells, we generated WT and Il21r^−/− CD8⁺ T cell mixed bone marrow chimera mice, and repeated our B16-F10 melanoma experiments (Figure 7G). Importantly, the expansion of CX₃CR1⁺ CD8⁺ TILs upon treatment with IL-21⁺ CD4 ACT was largely abolished when IL-21 signaling in CD8⁺ T cells was abrogated (Figure 7H). Additionally, the magnitude of CX₃CR1⁺ CD8⁺ TILs once again inversely correlated with tumor burden (Figure 7H), indicating that CD8⁺ T cell intrinsic IL-21 signaling is required for the therapeutic efficacy of IL-21⁺ CD4⁺ T cell ACT. Taken together, our findings demonstrate that CD4⁺ T cell help plays a critical role in the generation of protective cytotoxic CX₃CR1⁺ CD8⁺ T cells, and that this differentiation checkpoint may potentially be exploited for the purpose of ACT.

Figure 7. Provision of IL-21⁺CD4 help enhances CX₃CR1⁺ CD8⁺ T cell development and control over tumorgenesis.

(A) Representative flow plots showing the %CX₃CR1+, Ly108+ or CX₃CR1⁻Ly108⁻ CD44^hiCD8⁺ TILs responding to B16-F10 tumor and their relative expression of PD-1. (B) Experimental design. (C) Representative flow plots and summary data displaying the proportion and total number of CX₃CR1⁺ TILs on day 8 post ACT. Summary data showing negative correlation (far right) between %CX₃CR1⁺ TILs and tumor burden. (D) Tumor growth curves; of note summary data is from (n=6) IL-21⁻ Th17 cell recipients and (n=8) IL-21⁺ Th17 cell recipients. IL-21⁻ Th17 cell recipient mice that developed >30% of an IL-21⁺tRFP⁺ response (n=3) were excluded from the analysis (E) Representative flow plots and summary data displaying the %IL-21-tRFP⁺ CD4 T cells in the tumor at day 8 post-ACT. Summary data (right) showing inverse correlation between IL-21-tRFP⁺ CD4 T cells and tumor size. (F) Summary data showing positive correlation between IL-21⁺CD4⁺ T cell response and %CX₃CR1⁺CD8⁺ TILs. (G) Experimental design. (H) Representative flow plots and summary data displaying the proportion and total number of CX₃CR1⁺CD8⁺ TILs on day 8 post ACT. Summary data (bottom) showing inverse correlation between %CX₃CR1⁺ TILs and tumor burden. Data (Mean+/− S.E.M.) in (C-E,H) are pooled and are from 3-5 mice/group/experiment and are representative of at least 2 independent experiments.*p<0.05 **p<0.01 ***p<0.0001

Discussion

Collectively, our work has revealed the development of three major CD8⁺ T cell subsets during persistent viral infection, with each subset differing in their transcriptional program, phenotype, functional features, and tissue distribution. Thus, our findings add an additional layer of complexity to the previously proposed model of CD8⁺ T cell differentiation during chronic infection. Additionally, our study highlights a critical role for CD4⁺ T cell help via IL-21 in the generation of a transcriptionally and functionally distinct CX₃CR1⁺ cytotoxic CD8⁺ T cell subset that is essential for viral control. Thus, our study addresses a long-standing question within the field, and identifies a previously unrecognized mechanism by which CD4⁺ T cell help sustains protective antiviral CD8⁺ T cell responses during chronic viral infection. Overall, our data indicate that the self-renewing progenitor population developed early and remained in a transcriptionally similar state during the later phase of chronic viral infection. As the infection progressed, these Slamf6⁺ progenitor cells could potentially give rise to two distinct terminally differentiated effector CD8⁺ T cell subsets: a Pdcd1⁺ population and a CX₃CR1⁺ subset. These two late-phase effector subsets displayed unique properties, with Pdcd1⁺ cells appearing to be highly dysfunctional, as exemplified by their poor recall potential, and diminished cytokine production and cytolytic activity. Conversely, Cx3cr1⁺ cells exhibitted markedly superior cytolytic function, and a heightened capacity to degranulate and secrete effector cytokines. This transcriptional switch, from the bifurcation of Slamf6⁺ progenitor cells into either the Pdcd1⁺ or Cx3cr1⁺ cell subsets, likely marks the start of the true chronic infection differentiation program. Moreover, we speculate that this division of labor is likely essential for maintaining viral control, while simultaneously limiting excessive immunopathology.

T cell exhaustion is classically defined by sustained expression of co-inhibitory receptors, poor effector function, and a transcriptional state that is distinct from functional effector or memory T cells (Wherry and Kurachi 2015; Wherry et al. 2003; Kahan, Wherry, and Zajac 2015). Additionally, exhausted T cells are implicated as being terminally differentiated (Wherry and Kurachi 2015). In line with these studies, our scRNA-seq analyses and functional characterization of LCMV-specific CD8⁺ T cells identified a major population of CX₃CR1⁻Ly108⁻ (PD-1^hi) cells that appear highly dysfunctional and exhibit an exhausted phenotype. Moreover, our adoptive transfer experiments demonstrate that this CX₃CR1⁻Ly108⁻ CD8⁺ T cell subset displays a reduced capacity to expand upon secondary recall. However, our recall experiments also demonstrated that some CX₃CR1⁻1y108⁻ cells do retain their capacity to differentiate into the CX₃CR1⁺ subset. This observation could potentially be explained by the possibility that CX₃CR1⁻Ly108⁻ cells may also contain Slamf6⁺Pdcd1⁺ or Pdcd1⁺Cx3cr1⁺ intermediate populations (clusters 3 and 5 (Figure 1) respectively), that could have distinct recall properties. Alternatively, although unlikely, it may also be possible that some exhausted CD8⁺ T cells may retain the potential to redirect their differentiation towards the CX₃CR1⁺ cytolytic subset. By contrast, the majority of CX₃CR1⁺ donor-derived cells retained high CX₃CR1 and T-bet expression upon recall, further supporting that the CX₃CR1⁺ subset has progressed to a terminal state of differentiation.

In an independent study, Hudson et al. (ref) have shown that CD101⁻Tim3⁺ CD8⁺ T cells displayed a striking similarity to the CX3CR1⁺ subset identified in our study, as exemplified by their increased surface expression of CX3CR1, KLRG1, and the transcription factors T-bet, Zeb-2 and Klf-2. Consistent with their enhanced effector-like profile, CD101⁻Tim3⁺ cells also displayed augmented IFN-γ production and granzyme B expression relative to the exhausted CD101⁺Tim3⁺ subset. Importantly, Hudson et al. similarly observed a critical role for CX3CR1⁺ (CD101⁻Tim3⁺) CD8⁺ T cells in mediating antiviral immunity during chronic viral infection, thereby underscoring the importance of maintaining this cytolytic effector subset. Of note, although some small differences in the expression pattern of certain molecules (such as Tim3 and Eomes) and in the responsiveness to PD-1 blockade were observed when comparing effector versus exhausted CD8⁺ T cell subsets in these two studies respectively, this is likely a result of the stark differences in viral load that occurs following chronic infection under CD4⁺ T cell-sufficient versus CD4⁺ T cell “un-helped” conditions. Additionally, while our study demonstrates that CX3CR1⁺ CD8⁺ T cells (isolated at the late phase of chronic infection when viral titers decline) appear terminally differentiated upon recall with LCMV Cl13 in naïve recipient mice,, Hudson et al demonstrate that “un-helped” CD101⁻Tim3⁺ cells adopt a transitory differentiation state and maintain their proliferative potential when transferred into infection-matched mice (where viral load remains high). In response to continuous antigen stimulation, CD101⁻Tim3⁺ cells eventually progress into CD101⁺Tim3⁺ exhausted cells. Future work will be important for further dissecting the overall similarities and differences between CD101⁻Tim3⁺ and CX3CR1⁺Ly108⁻ effector subsets and determining how viral burden and CD4⁺ T cell help impacts their differentiation potential and function.

Although previous studies have identified an essential role for IL-21 in sustaining virus-specific CD8⁺ T cell responses and viral control (Yi, Du, and Zajac 2009a; Tian et al. 2016; Elsaesser, Sauer, and Brooks 2009a; Fröhlich et al. 2009), the mechanisms by which IL-21 confers this protection has thus far remained elusive. Importantly, our study identified that the generation of cytotoxic CX₃CR1⁺ CD8⁺ T cells, which are essential for viral control, was critically dependent on CD4⁼ T cell help and CD8⁺ T cell intrinsic IL-21R signaling. Our study further indicated that CD4⁺ T cells comprise approximately 90% of all IL-21-producing lymphocytes responding to LCMV Cl13 infection, suggesting that T helper cells are likely the major source of IL-21 that drives the differentiation of CX₃CR1⁺ CD8⁺ T cells. Importantly, our findings further identified that the CD4-IL-21-CX₃CR1 developmental pathway was also operational during tumorigenesis and may be harnessed to enhance immune-mediated control over tumor progression. Despite these findings, although we could readily detect donor-derived IL-21⁺CD4⁺ T cells infiltrating the tumor, it remains unclear whether CD4⁺ T cell-mediated CX₃CR1⁺CD8⁺ T cell differentiation can occur directly within the immunosuppressive tumor microenvironment or in secondary lymphoid tissues (where the Ly108⁺ progenitor population preferentially resides). A determination of whether CD4⁺ T cell help is essential in particular anatomical locations for driving the formation of CX₃CR1⁺ CD8⁺ T cells will have important implications for optimizing immunotherapy.

Notably, our data identify that the differentiation defect of “un-helped” CD8⁺ T cells was associated with diminished T-bet and upregulated Eomes expression. This finding is consistent with the Cx3cr1⁺ cell subset displaying uniformly high expression of Tbx21, whereas Pdcd1⁺ exhausted cells exhibited low Tbx21 expression, but high Eomes expression (Figure S2E-F). However, whether differential expression of these two transcription factors is responsible for the differentiation trajectory of virus-specific CD8⁺ T cells along a PD-1^hi exhausted or CX₃CR1⁺ developmental pathway remains to be further investigated. Moreover, although it is well appreciated that targeting the PD-1-PD-L1 axis can re-invigorate T cell responses during persistent infection and cancer (Barber et al. 2006; Wherry and Kurachi 2015), our data indicate that PD-L1 blockade is insufficient to rescue the defect in CX₃CR1⁺ CD8⁺ T cell differentiation that occurs in “un-helped” CD8⁺ T cells. These findings may be of particular relevance when considering immunotherapy to treat diseases wherein low CD4⁺ T cell numbers are often a common occurrence, such as during HIV infection (Klatzmann et al. 1984; Masur et al. 1989). Additionally, although several recent studies have identified that the TCF-1^hi progenitor population is the subset that provides a proliferative burst in response to PD-L1 blockade (Leong et al. 2016; He et al. 2016; Im et al. 2016; Utzschneider et al. 2016), we speculate that maintaining a pool of cytotoxic CX₃CR1⁺ CD8⁺ T cells in conjunction with the TCF-1^hi (Slamf6⁺) progenitor population is likely required for optimizing the therapeutic efficacy of blocking the PD-1-PD-L1 inhibitory axis.

Taken together, our study uncovers broad transcriptional heterogeneity amongst CD8⁺ T cells responding to persistent infection and supports a previously unrecognized paradigm of CD8⁺ T cell differentiation during chronic viral infection. Moreover, our data further indicate that CD4⁺ T cell help in the form of IL-21 may potentially be harnessed to bolster the formation of protective cytolytic CX₃CR1⁺ CD8⁺ T cells and improve control over chronic viral infection and tumor progression.

STAR METHODS

Lead Contact and Materials Availability

Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Weiguo Cui (weiguo.cui@bcw.edu).

Experimental Models and Subject Details

Mice and LCMV Cl13 infection

Six to eight-week-old female C57BL/6 and CD45.1 congenic mice were obtained through the National Cancer Institute grantees program (Frederick, MD). IL-21-tRFP mice were obtained from Dr. Joseph Craft, Yale School of Medicine, CT. Mice were bred and maintained in a closed breeding facility, and mouse handling conformed to the requirements of the Institutional Animal Care and Use Guidelines of Medical College of Wisconsin. Cx3cr1^DTR mice (stock #:025629) and dLck-cre-recombinase mice (stock #:012837) were purchased from Jackson and crossed to one another to generate either Cx3cr1^DTRdLck-cre− or Cx3cr1^DTRdLck-cre⁺ mice. Mice were infected with LCMV clone 13 (Cl13), which was intravenously injected into mice (2 × 10⁶ PFU/mouse) to establish chronic infection. LCMV Cl13 was prepared by a single passage on BHK21 cells and viral titers were determined by plaque formation assay on Vero cells.

Tumor cell lines and Tumor Induction

B16-F10 cells were obtained from ATCC and cultured in high-glucose DMEM (Cellgro) supplemented with 10% (vol/vol) FBS. Melanoma tumors were established by injecting 2 × 10⁵ B16-F10 cells s.c. on one flank of the C57BL/6 mice. Tumor growth was monitored by measuring with calipers every other day, and tumor volume was calculated as length * (width)²/2

Method Details

Generation of Polyclonal Tumor Reactive CD4⁺ T Cells

Bone marrow cells were isolated from C57BL/6 mice and cultured in RPMI (Cellgro) medium with 10% (vol/vol) FBS and 200 ng/mL Flt3L for 1 wk. On day 7, DCs were harvested and incubated with freeze-thawed tumor lysates at a ratio of one tumor cell equivalent to one DC (i.e., 1:1) as described (Liang et al. 2014). After 18 hours of incubation, DCs were harvested and maturated with LPS for 4 hours. The mature DCs and purified IL-21-tRFP CD4⁺ T cells were mixed in 1:2 ratio and cultured under Th17 cell conditions using IL-6 (50 ng/mL), TGF-β (10ng/mL), IL-23 (20ng/mL), and IL-21 (10ng/mL) and also with anti-IL-4 and anti-IFN-γ antibodies (10ug/ML). Then, activated CD4 T cells were sort-purified based on IL-21-tRFP expression and adoptively transferred into recipient mice that were inoculated with B16-F10 7-10 days prior.

Immune Cell isolation from solid tumors

Dissected tumor tissues were cut into small pieces and digested with 0.7 mg/mL collagenase XI Sigma-Aldrich) and 30 mg/mL type IV bovine pancreatic DNase (Sigma-Aldrich) for 45 min at 37 °C. Immune cells were then isolated by centrifugati on with Lymphocyte Cell Separation Medium (Cedarlane Labs)

Flow cytometry

All flow cytometry data were acquired on an LSRII (BD Biosciences, CA) and analyzed by FlowJo (Treestar, OR). Lymphocytes were isolated from tissues including spleen, blood, liver, bone marrow, and inguinal lymph nodes as described previously. Cells were then stained with GP_33-41 tetramer and antibodies against cell surface antigens for 30-60 minutes at 4 degrees. Transcription factor staining was performed using the True Nuclear transcription factor buffer set (Biolegend). All antibodies used in this study are listed in Supplemental Table 1. BrdU was administered to mice i.p. and measured in LCMV-specific cells according to the manufactures protocol (BD Bioscience).

IncuCyte cytotoxicity assay

Target EL4 cells were pulsed with either GP33 or irrelevant ovalbumin SIINFEKL (OT-I) peptide for 2 hours and then incubated with sorted GP_33-41⁺ CD8+ T cells (5:1 E: T) in the presence of IncuCyte Annexin V Red Reagent (Essen BioScience Cat #4641), which enables real-time detection of cells undergoing apoptosis. Images were taken every 2 hours and analyzed by incuCyte S3 Live-Cell Analysis System.

Administration of biologics

To deplete CD4 T cells, mice received i.p. injections of 500 μg anti-CD4 antibody (clone GK1.5 from BioXCell, NH) one day before LCMV Cl13 infection. InVivoMab anti-mouse PD-L1 (clone 10F.9G2) was purchased from BioXCell and administered in 200 ug doses on days 16, 19, and 22 p.i.

Cell transfer experiments

For recall experiments, CD8⁺CD44^hi CX₃CR1⁺, Ly108⁺ or CX₃CR1⁻Ly108⁻ subsets were sorted from LCMV-Cl13-infected mice on day 45 p.i. using an Aria IIIu cell sorter. Next we adaptively transferred 60k cells of each respective subset into separate groups of congenically marked naïve Ly5.1 mice that were subsequently challenged with LCMV-Cl13 one day later. In another experiment, 80k cells of each respective subset were transferred in experimental mice and identical results were observed. For P14 transfer experiments, congenically marked WT or Il21r^−/− P14 cells were co-transferred at a 1:1 ratio (5000 cells each) into naïve Ly5.1 mice that were then infected one day later.

Mixed bone marrow (MBM) chimeras

For MBM chimera experiments, recipient mice were irradiated with 6.5 and 5.5 Gy separated by 8 h. Bone marrow from various donor mice (as depicted in Figures) were mixed at the indicated ratios, and a total of ~ 6 × 10⁶ cells were transferred i.v. Mice were maintained on oral sulfamethoxazole for 2 weeks. Experimental MBM chimera mice were either infected with LCMV Cl13, or injected with B16-F10 melanoma cells at 8 weeks post reconstitution.

Single-cell RNA sequencing

LCMV-specific CD8 T cells were FACS-sorted from LCMV Cl13-infected mice on days 8 and 30 p.i. and were loaded on the Chromium Controller (10x Genomics). Single-cell RNA-seq libraries were prepared using the Chromium Single Cell 3’ v2 Reagent Kit (10x Genomics) according to manufacturer’s protocol. Libraries were loaded onto an Illumina NextSeq with the NextSeq 500/550 High Output Kit v2 (150 cycles) (FC-404-2002, Illumina) with the following conditions: 26 cycles for read 1, 98 cycles for read 2, and 8 cycles for i7 index. Python Run Downloader (Illumina) was used to download raw sequencing data. Cell Ranger (10x Genomics) functions mkfastq and count were used to demultiplex the sequencing data and generate gene-barcode matrices, respectively. All scRNA-seq analyses were performed in R (version 3.4.0) using the package Seurat (version 2.2.0) (Satija et al. 2015) and Monocle (Trapnell et al. 2014). Number of genes detected per cell, number of UMIs, and percent mitochondrial genes were plotted, and outliers were removed (number of genes over 2,500, number of UMIs over 8,000, and percent mitochondrial genes over 0.08) to filter out doublets and dead cells. Principal component analysis was performed, and the top 9 most statistically significant principal components were used for t-SNE analysis, with 2,000 iterations and a perplexity parameter of 30.

RNA isolation and quantitative real-time PCR

Fractions of spleens were snap-frozen on dry ice and subsequently stored in a −80°C freezer. Samples were stabilized in RNAlater-ICE (ThermoFisher) before proceeding with tissue homogenization and RNA isolation using the RNAqueous-Micro Kit (ThermoFisher) according to the manufacturer’s instructions. For sera samples, RNA was extracted from 10ul of sera using QiAamp MinElute Virus Spin Kit (Qiagen). Determination of viral load by qPCR was performed as has been previously described (McCausland and Crotty 2008) and the following primers were used during the qPCR reaction: GP-R (S pos. 970-991), GCAACTGCTGTGTTCCCGAAAC. GP-F (S pos. 877-901), CATTCACCTGGACTTTGTCAGACTC. Amplification was done for 40 cycles with each cycle consisting of two steps: 95°C, 15 s; and 60°C, 30 s . Standard curves were generated using serial dilutions of a gene fragment (gblocks IDT) derived from Lymphocytic choriomeningitis virus clone 13 segment S, with the following sequence: 5'- AGA GAA GAC TAA GTT CCT CAC TAG GAG ACT AGC GGG CAC ATT CAC CTG GAC TTT GTC AGA CTC TTC AGG GGT GGA GAA TCC AGG TGG TTA TTG CCT GAC CAA ATG GAT GAT TCT TGC TGC AGA GCT TAA GTG TTT CGG GAA CAC AGC AGT TGC GAA ATG CAA TGT AAA TCA TGA TGA AGA ATT CTG TG -3'.

Quantification and Statistical Analyses

Statistical tests were performed using Graphpad Prism 7. P-values were calculated using either two-tailed unpaired student’s t tests or one-way ANOVA while correcting for multiple comparisons via the Tukey method.

Data and Code Availability

The single-cell RNA-seq data will be deposited in the GEO database with the accession code GSE129139. All other raw data are available from the Lead Contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Mouse strains
C57BL/6	Charles River	N/A
Cx3cr1DTR mice	The Jackson Lab	Stock # 025629
dLck-cre-recombinase mice	The Jackson Lab	Stock # 012837
IL-21-tRFP reporter mice	(Shulman et al. 2014)	N/A
Flow Cytometry Reagents
Anti-mouse PD.1 (RMP1-30)	Biolegend	Cat #109110; RRID:AB_572017
Anti-mouse Thy1.1 (OX-7)	Biolegend	Cat #202506; RRID:AB_492882
Anti-mouse CXCR3 (CXCR3-173)	Biolegend	Cat #126512; RRID:AB_1088993
Anti-mouse CXCR5 (L138D7)	Biolegend	Cat #145526; RRID:AB_2566799
Anti-mouse CXCR6 (SA051D1)	Biolegend	Cat #151117; RRID:AB_2721700
Anti-mouse CX3CR1 (SA011F11)	Biolegend	Cat #149014; RRID:AB_2565698
Anti-mouse Ly108 (330-AJ)	Biolegend	Cat #134608; RRID:AB_2188093
Anti-mouse Tim3	BD Biosciences	Cat #747625; RRID:AB_2744191
Anti-mouse 2B4 (M2B4 (B6)458.1)	Biolegend	Cat #133503; RRID:AB_1595624
Anti-mouse KLRG1 (2F1)	Biolegend	Cat #138416; RRID:AB_2561736
Anti-mouse KLRa9 (Ly-49I; YLI-90)	eBioscience	Cat #12-5895-82; RRID:AB_466021
Anti-mouse CD127 (A7R34)	Biolegend	Cat #135014; RRID:AB_1937265
Anti-mouse CD107a (1D4B)	Biolegend	Cat#121616; RRID:AB_10643268
Anti-mouse TNF-α (MP6-XT22)	Biolegend	Cat#506306; RRID:AB_315427
Anti-mouse IFNγ (XMG1.2)	Biolegend	Cat#505826; RRID:AB_2295770
Anti-human/mouse Granzyme B (GB11)	Invitrogen	Cat#GRB04; RRID:AB_2536538
Anti-TCF-1 (C63D9)	Cell Signaling	Cat#2203S; RRID:AB_2199302
Anti-mouse BATF (D7C5)	Cell Signaling	Cat#8638S; RRID:AB_11141425
Anti-mouse Eomes (Dan11mag)	Invitrogen	Cat#25-4875-82; RRID:AB_2573454
Anti-mouse T-bet (4B10)	Biolegend	Cat#644806; RRID:AB_1595488
LCMV DbGP33 tetramer	Made in house	N/A
Brefeldin A Solution (1,000X)	Biolegend	Cat#420601
Fixation Buffer	Biolegend	Cat#420801
True Nuclear Transcription Factor Buffer Set	Biolegend	Cat#424401
Experimental Models: LCMV
LCMV Clone (Cl13) virus strain	Rafi Ahmed	Grew up in house
Experimental Models: Tumor Cell Lines
B16/F10 tumor cell line	ATCC	Cat# CRL-6475
Chemicals, Peptides and Recombinant Proteins
RNAlater-ICE	Invitrogen	Cat#AM7030
KAVYNFATM (GP33-41) peptide	GenScript	RP20257
Oligonucleotides
GP-F: CATTCACCTGGACTTTGTCAGACTC	(McCausland and Crotty 2008) IDT	N/A
GP-R: GCAACTGCTGTGTTCCCGAAAC	(McCausland and Crotty 2008) IDT	N/A
Critical Commercial Assays
Incucyte Annexin V Red Reagent	Essen Bioscience	Cat#4641
QiAmp MinElute Virus Spin Kit	Qiagen	Cat#57704
RNAqueous-Micro Kit	Thermofisher	Cat#AM1931
Mouse Treatment reagents
Anti-mouse CD4 Mab (GK1.5)	Bioxcell	Cat#BE0003-1
Anti-PD-L1 (10F.9G2)	Bioxcell	Cat#BE0101
Diptheria Toxin	Sigma-Aldrich	Cat#D0564
Single cell RNA sequencing
Chromium Single Cell 3’ Library & Gel Bead Kit v2	10X Genomics	Cat#PN-120267
Chromium Single Cell A Chip Kit	10X Genomics	Cat#PN-1000009
Chromium i7 Multiplex Kit	10X Genomics	Cat# PN-120262
Dynabeads™ MyOne™ Silane	Thermofisher	Cat#37002D
SPRIselect Reagent Kit	Beckman Coulter	Cat#B23318
Kappa NGS quantification kit	KAPABiosystems	Cat#KK4824
NextSeq 500/550 High Output Kit v2.5 (150 cycles)	Illumina	Cat#20024907
Computational analysis
Cell Ranger	10X Genomics	https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/installation
Seurat	(Satija et al. 2015)	https://satijalab.org/seurat/
Monocle-2	(Qiu et al. 2017) (Trapnell et al. 2014)	http://cole-trapnell-lab.github.io/monocle-release/docs/
Datasets
Day 8 and Day 30 single cell RNA sequencing on DbGP33+ CD8 T cells	In this paper	GSE 129139
Statistical or Data Analysis
Flowjo Version 10.5.3	Tree Star	N/A
Prism 8	Graphpad Software	N/A