CD4 T Cell Depletion Substantially Augments the Rescue Potential of PD-L1 Blockade for Deeply Exhausted CD8 T Cells

T cell exhaustion is a hallmark of chronic infection and is characterized by progressive downregulation of T cell function (1–6). In particular, the immunoinhibitory programmed cell death-1 (PD-1) pathway is critical in regulating T cell function during chronic infections and cancers (5, 7–9). PD-1 is upregulated on exhausted T cells (9) and ligation with programmed death-ligand 1 (PD-L1) results in reduced signal transduction after TCR triggering (10). In different models of chronic infection, blockade of the PD-1/PD-L1 pathway results in significant rescue of exhausted CD8 T cell responses (9, 11–16).

Until now, all studies with the chronic lymphocytic choriomeningitis virus (LCMV) infection model have assessed T cell exhaustion at early time points after the establishment of persistent infection (9, 17–20). These reports have shown that PD-L1 blockade within the first two months of chronic infection results in substantial rescue of exhausted CD8 T cell responses, but a detailed analysis of the impact of PD-L1 blockade during the later stages of chronic infection is lacking. In this study, we corroborated that PD-L1 blockade during the early stage of a chronic LCMV infection (about day 60) results in robust functional rescue of exhausted CD8 T cell responses. However, we observed reduced efficacy of PD-L1 blockade at rescuing exhausted CD8 T cell responses during the late stages of chronic infection (>150 d). Strikingly, the reduction in the efficacy of PD-L1 blockade in nonresponding mice (at late times postinfection) was associated with accumulation of PD-1⁺ regulatory T cells (Tregs). We also show that treatment with CD4 T cell–depleting Abs partially re-establishes responsiveness to PD-L1 blockade therapy at the late stage of chronic infection. These findings demonstrate an effective strategy for improving the efficacy of PD-L1 blockade in the context of advanced chronic diseases and highlight an inverse association between the levels of PD-1⁺ Tregs and response to PD-L1 blockade.

Four- to 8-wk-old C57BL6J mice (Jackson Laboratories) were infected with LCMV Armstrong or Cl-13. Memory T cell responses were generated by i.p. injection with 2 × 10⁵ PFU LCMV Armstrong (21), which results in an acute infection that is cleared within 8 d, resulting in the generation of memory immune responses. Lifelong chronic infections with exhausted CD8 T cell responses were generated by CD4 T cell depletion followed by i.v. injection with 2 × 10⁶ PFU LCMV Cl-13 as described previously (22). Transient systemic LCMV Cl-13 infections were induced by i.v. injection with 2 × 10⁶ PFU LCMV Cl-13 without prior CD4 T cell depletion. All animal experiments were performed with approval of the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee.

Titration of LCMV was performed on Vero cell monolayers as previously described (23). In brief, serial 10-fold dilutions from serum or homogenized tissues were distributed on Vero cell monolayers in six-well plates. Plates were then incubated for 1 h rocking every 15 min. A 1:1 solution of 1% agarose in 2× 199 media was overlaid on top of the monolayers. After 4 d, a 1:1 solution of 1% agarose in 2× 199 media with 1:50 neutral red was aliquoted on each well. PFUs were counted at day 5 with the aid of a transluminator. Adenoviral immunizations with various replication incompetent adenoviral vaccine vectors expressing LCMV glycoprotein (GP) were given i.m. at 10¹⁰ viral particles per mouse as described previously (24).

CD4 T cell depletions were performed by injection of 500 μg GK1.5 Ab (BioXCell) 2 d and again 1 d before PD-1/PD-L1 blockade. PD-L1 blockade was achieved by injection of 200 μg 10F.9G2 (BioXCell) at different times throughout the course of lifelong infection, and the regimen consisted of five doses every 3 d as previously described (9). All Ab treatments were given i.p. (diluted in DPBS). Rat IgG2b (BioXCell) was used as negative control. CD8 T cell rescue and antiviral control were assessed at day 15 posttreatment.

Single-cell suspensions were obtained from blood and tissues as previously described (25). Single-cell suspensions were stained with anti-CD8α (53-6.7), -CD4 (RM4-5), and -Foxp3 (FJK-16s) from eBioscience. Intracellular staining of Foxp3 was performed according to manufacturer’s instructions (eBioscience). Dead cells were excluded by gating out cells positive for Live/Dead fixable dead cell stain (Invitrogen). LCMV MHC class I tetramers were obtained from the National Institutes of Health tetramer facility (Emory University). LCMV-specific CD8 T cell responses were assessed by incubating splenocytes with 0.1 μg/ml LCMV peptides in the presence of brefeldin and monensin for 5 h at 37°C. Intracellular staining for IFN-γ was performed with Cytofix/Cytoperm kit (BD Biosciences). Annexin V staining was performed with the Annexin V Apoptosis Detection Kit (eBioscience), following the manufacturer’s instructions (cells were not fixed and were acquired within an hour of staining). Samples were acquired with a Becton Dickinson LSRII and analyzed using FlowJo (Tree Star).

Mann–Whitney U test was used to determine statistical significance. Statistical analysis was performed using Prism (GraphPad).

Functional exhaustion of virus-specific T cells occurs after chronic viral infection (1, 4, 5, 9). Earlier studies have assessed the early phase of T cell exhaustion post chronic LCMV infection in mice typically within the first two months of persistent infection (9, 17–20). However, analysis of exhausted CD8 T cell responses at later time points has not been previously reported.

To analyze the progression of CD8 T cell exhaustion throughout the course of chronic viral infection, we used the murine model of chronic uncontrolled LCMV Cl-13 infection, in which CD4 T cells are first depleted before i.v. challenge with LCMV Cl-13, resulting in a multiorgan and lifelong infection (9, 22). We then performed longitudinal phenotypic analyses of exhausted virus-specific CD8 T cells for 600 d (Fig. 1A), which is close to the life span of these mice. Although the frequencies of DbGP276-specific CD8 T cells were only modestly reduced during the course of 600 d (Fig. 1B), the total magnitudes of DbGP276-286–specific CD8 T cells declined over time (p ≤ 0.01, Mann–Whitney U test; Fig. 1C).

Importantly, this reduction in the number of LCMV-specific CD8 T cells was associated with a gradual increase in apoptosis as evidenced by increased Annexin V staining (mean 6.7% at day 60; mean 54.5% at day 600; Fig. 1D). Conversely, memory virus-specific CD8 T cell responses generated by an acutely controlled LCMV Armstrong infection showed negligible frequencies of apoptotic cells even at day 600 postinfection (Fig. 1D). These data indicate a time-dependent increase in the level of apoptosis of virus-specific CD8 T cells throughout the course of chronic viral infection. Throughout the course of chronic infection, exhausted CD8 T cells exhibited a modest upregulation of Eomes (Fig. 1E, p = 0.04) and PD-1 (Fig. 1F, p = 0.03), but we did not observe any difference in the expression of Tim-3 (data not shown). Moreover, there was a pattern of progressive downregulation of T-bet (Fig. 1G, p = 0.05). Overall, these continuous phenotypic changes (especially the inversely correlated expression of T-bet and PD-1) on exhausted LCMV-specific CD8 T cells suggested a gradual differentiation throughout chronic LCMV infection.

We next performed PD-L1 blockade at different times postinfection and compared its effects on CD8 T cell rescue and antiviral control (Fig. 2A). Consistent with previous reports (9, 19, 20, 26), PD-L1 blockade around day 60 resulted in robust expansion (Fig. 2B) and functional rescue (Fig. 2C) of exhausted virus-specific CD8 T cell responses. PD-L1 blockade on day 60 postinfection resulted in a 9.6-fold increase in CD8 T cell responses (DbGP276-286–specific; p = 0.02) as compared with controls. However, PD-L1 blockade proved less efficient at later stages of chronic infection (Fig. 2B–D). PD-L1 blockade on day 600 postinfection resulted in only a marginal 1.6-fold increase (p = 0.8) in CD8 T cell responses (Fig. 2D). As expected, memory CD8 T cells elicited by acutely controlled LCMV Armstrong or various adenovirus vector-based vaccines did not expand after PD-L1 blockade at day 60 postinfection (Fig. 2E).

We then assessed viral loads in both tissues and blood. Improved antiviral control in tissues (Fig. 3A, 4-fold viral reduction in spleen, p = 0.02) and sera (Fig. 3B, 2.4-fold viral reduction, p = 0.04) was observed when PD-L1 blocking Abs were administered on day 60 postinfection. However, antiviral control was not observed when PD-L1 blocking Abs were administered on day 600 postinfection (Fig. 3A, 3B). Taken together, these results suggest that there is a limited window of therapeutic efficacy for PD-L1 blockade and that its rescue effect wanes with progressive infection.

In addition, we also performed PD-L1 blockade at different times postinfection in the CD4-helped model of LCMV Cl-13 infection, in which systemic infection lasts for only 1–2 mo (1). In this model, enhancement in CD8 T cell responses and antiviral control (Supplemental Fig. 1) after PD-L1 blockade similarly declined when treatment was initiated after 1 mo of infection. However, this may have been due to the fact that in this less stringent model of LCMV Cl-13 infection, systemic viral infection is gradually cleared (1) and CD8 T cell rescue by PD-1 inhibition requires TCR recognition of Ag (27).

We investigated whether the reduced efficacy of PD-L1 blockade in mice infected for >150 d could simply be because of age rather than stage of infection. We infected naive 40-d-old or naive 400-d-old mice with chronic LCMV Cl-13 and started PD-L1 blocking treatment at day 60 postinfection (Supplemental Fig. 2A). Although there was a slight pattern of reduced responsiveness to PD-L1 blockade in older mice relative to young mice, there was still robust CD8 T cell rescue (Supplemental Fig. 2B, 2C) and viral control (Supplemental Fig. 2D). Therefore, age alone does not fully explain the lack of responsiveness to PD-L1 blockade in chronically infected mice at the late stages of infection.

The previous experiments suggest biased modulation of exhausted CD8 T cell responses by alternate immunoregulatory pathways in addition to the PD-1/PD-L1 pathway, especially at the later stages of chronic infection. Surprisingly, the frequencies of CD4⁺ T cells that were Foxp3⁺ (Tregs) increased 2.8-fold from days 60 to 600 postinfection (p = 0.0001, two-tailed, paired t test; Fig. 4A, 4B). However, there was no significant change in the overall number of Tregs in spleen (p = 0.1; Fig. 4C), and this was due to reduction in overall lymphocyte counts observed during the late stage of chronic LCMV infection (data not shown). The progressive increase in Treg frequencies suggested a possible role to the process of CD8 T cell exhaustion. Of note, PD-L1 blockade also resulted in a significant activation of Treg responses, evidenced by increased levels of inhibitory CTLA-4 and PD-1 molecules (Fig. 5A), with a marked increase in the numbers of PD-1⁺ Tregs (Fig. 5B, 5C). Also notable, PD-L1 blockade resulted in an increase in the PD-1⁺ non-Foxp3 subset (Fig. 5B), but all CD4 T cell subsets were not LCMV-specific by MHC class II (I-A^b GP66⁺) tetramer staining (Fig. 5B). Note that in this model of lifelong infection, in which CD4 T cells are first depleted before Cl-13 infection, there is no emergence of LCMV-specific CD4 T cells even after CD4 T cell recovery. In the absence of pre-existent virus-specific CD4 T cells, PD-L1 blockade does not induce any generation of LCMV-specific CD4 T cell responses, corroborating that PD-L1 blockade rescues pre-existing T cell responses but does not induce de novo responses, consistent with previous reports (28–30).

Strikingly, PD-L1 blockade during the late stage of chronic viral infection resulted in only an expansion of PD-1⁺ Tregs (Fig. 5D) with essentially no rescue of CD8 T cell responses (Figs. 2A–D, 3A, 3B). These results suggested a dual effect of PD-L1 blockade at rescuing not only PD-1⁺ CD8 T cell responses, but also PD-1⁺ Treg responses.

Continuous depletion of CD4⁺ Tregs has been shown to be lethal (31), and combining CD4⁺ Treg depletion with PD-L1 blockade results in an accelerated onset of autoimmunity (18). The lethal immune activation after specific Treg depletion is mediated by CD4 T cells. This highlights the fact that CD4 T cells can sometimes help the immune response, but in other situations, they could impair the immune response and become lethal (32–34). We noticed that CD4 T cell depletion in chronically infected mice did not result in any evident emaciation or signs of autoimmunity despite the absence of Tregs (data not shown). We thus assessed whether transient depletion of CD4 T cells could augment the efficacy of PD-L1 blockade at the late stages of infection (Fig. 6A). Consistent with our previous results, exhausted CD8 T cells appeared to resist functional rescue by PD-L1 blockade during the late stage of chronic infection (day 200; Fig. 6B). However, CD4 T cell depletion with PD-L1 blockade resulted in enhanced cellular proliferation (Supplemental Fig. 3), increased frequencies of virus-specific CD8 T cells in spleen (Fig. 6B), and also augmented cytokine coexpression (Supplemental Fig. 4; mean of IFN-γ and TNF coexpression for PD-L1 blockade alone was 1.4%, compared with 13.2% after combined CD4 T cell depletion and PD-L1 blockade at day 200 postinfection; p = 0.02; Fig. 6C). Combined CD4 T cell depletion and PD-L1 blockade also resulted in a statistically significant increase in the total numbers of splenic virus-specific CD8 T cells at the later times of chronic infection (p = 0.05; Fig. 6D). Similar augmentation of virus-specific CD8 T cell rescue was observed in nonlymphoid tissues, such as in liver (Fig. 6E, 6F) and lungs (Fig. 6G, 6H).

Consistent with the improved CD8 T cell rescue, combined CD4 T cell depletion and PD-L1 blockade during the late phase of chronic LCMV infection also improved antiviral control in tissues (Fig. 7A) and sera (Fig. 7B) (p ≤ 0.05, Mann–Whitney U test). During the early stage of chronic infection (day 60), the reduction in viremia was only 3.1-fold with PD-L1 blockade alone, but strikingly, it was 10.7-fold with combined CD4 T cell depletion and PD-L1 blockade. Importantly, during the late phase of chronic infection (day 200), PD-L1 blockade alone resulted in no detectable reduction in viremia, but the reduction in viremia was 2.1-fold with combined CD4 T cell depletion and PD-L1 blockade. Because the synergistic effect of combined CD4 T cell depletion and PD-L1 blockade was more modest when treatment started during the late stage of infection (compared with the early stage of infection), it is possible that alternative inhibitory pathways may be concertedly induced during advanced exhaustion. Note that <1% of CD4 T cells express LCMV (35–37), and thus the reduction in viral loads after CD4 T cell depletion and PD-L1 blockade was likely not due to Ab-mediated ablation of a few infected CD4 T cells. Consistent with this model, CD4 T cell depletion alone did not induce any change in viral loads (Fig. 7A, 7B). Taken together, our data demonstrate that transient depletion of CD4 T cells not only enhanced the rescue potential of PD-L1 blockade, but also significantly extended the window of therapeutic opportunity after PD-L1 blockade.

The inhibitory PD-1/PD-L1 pathway is involved in the regulation of diverse immune processes such as tolerance, T cell exhaustion, and normal immune responses to microorganisms (7, 15, 38, 39). In vivo blockade of the PD-1/PD-L1 pathway has been shown to be well tolerated and effective at rescuing CD8 T cell function in various animal models of chronic infections (9, 12, 13, 40). However, there is substantial heterogeneity in objective responses in patients with advanced melanoma, colorectal cancer, and other malignancies (41). Although expression of PD-L1 on tumors and intratumoral immune cells may correlate with objective responses after PD-1 blockade (41, 42), additional factors may also contribute to therapeutic responsiveness. In this report, we compared the effect of in vivo PD-L1 blockade during early versus late stages of chronic LCMV infection. We observed a marked decline in the efficacy of PD-L1 blockade to rescue exhausted CD8 T cells during advanced infection. Early after the establishment of chronic LCMV infection (day 60), PD-L1 blockade resulted in a 9.6-fold rescue of exhausted virus-specific CD8 T cells and a 2.4-fold decrease in viremia. However, during the late phase of a chronic LCMV infection (day 600), PD-L1 blockade resulted in negligible rescue of exhausted virus-specific CD8 T cell responses and no detectable reduction of viral loads. This lack of immune rescue during the late stages of infection was associated with a profound expansion of CD4⁺ PD-1⁺ Tregs, which expanded substantially after PD-L1 blockade. Interestingly, dual CD4 T cell depletion together with PD-L1 blockade was able to induce immune rescue at the late stages of chronic infection.

Several factors could explain the reduction in the therapeutic efficacy of PD-L1 blockade at the later stages of infection (CD8 T cell–intrinsic and CD8 T cell–extrinsic reasons). During chronic viral infection, there is a progressive differentiation of exhausted CD8 T cells (43). The Tbet^hi population terminally differentiates into the more exhausted Eomes^hi subset, which expresses higher levels of inhibitory PD-1. Importantly, PD-1^high CD8 T cells cannot be rescued after PD-1/PD-L1 blockade (44). Therefore, our phenotypic data in Fig. 1 suggest that the reduced CD8 T cell rescue by PD-L1 blockade during the late stage of chronic infection may be because of the more terminal differentiation of virus-specific CD8 T cells. These data are consistent with a model in which CD8 T cell rescue by PD-L1 blockade is dependent upon the pool size of a responsive T-bet^high Eomes^lo subset that expresses lower levels of inhibitory receptors. Consistent with previous observations (45), immune rescue by PD-L1 blockade seemed to accelerate the terminal differentiation of exhausted CD8 T cells (mean fluorescence intensity for T-bet after PD-L1 blockade was 639 versus 778 for control, p = 0.05; mean fluorescence intensity for PD-1 after PD-L1 blockade was 2408 versus 1648 for control, p = 0.0034, two-tailed test).

The reduced immune rescue after PD-L1 blockade during the late stage of chronic infection could also be attributed to the progressive increase in the frequencies of CD4 T cells that were Tregs over time, although in this study we do not directly assess the specific contribution of Tregs to CD8 T cell exhaustion during the late stage of infection. Throughout chronic LCMV Cl-13 infection, there is an increase in the frequencies of Tregs, and this has also been demonstrated in various types of chronic infections in mice and humans (46–52), and has been associated with immunosuppression leading to chronic infections (48, 53–56). Surprisingly, PD-L1 blockade also increased the expression of CTLA-4 on Tregs. CTLA-4 is an inhibitory molecule that is critical for Treg-mediated suppression (57, 58). Thus, our data are consistent with a model in which PD-L1 blockade, by inducing the expression of CTLA-4 on Tregs, may indirectly increase their suppressive capacity. However, we cannot exclude the possibility that CD4 T cell lymphopenia after CD4 T cell ablation may result in increased availability of homeostatic cytokines, which may improve immune rescue after PD-L1 blockade. Previous reports have shown that combining anti–CTLA-4 with anti–PD-1 blocking Abs resulted in >40% objective responses in advanced melanoma patients (59, 60). Taken together, we hypothesize that the synergy between CD4 T cell depletion and PD-L1 blockade was due to the removal of activated CTLA-4⁺ Tregs that were induced after PD-L1 blockade.

A prior report showed that transfer of LCMV-specific (SMARTA) CD4 T cells into Cl-13–infected mice results in rescue of CD8 T cell responses and reduction of viral loads (61). This beneficial effect was likely due to the fact that LCMV-specific CD4 T cells do not differentiate into Tregs (expanded Treg subsets after Cl-13 infection are specific for endogenous retroviruses, but not LCMV) (50). Other studies have suggested synergy between CD4 T cell depletion and PD-L1 blockade at enhancing immune control in a murine model of retroviral infection and renal cell carcinoma (62, 63), but it remains unknown whether this dual regimen can improve immune function during advanced disease. In this study, we show for the first time, to our knowledge, that the combination of PD-L1 blockade and CD4 T cell depletion can partially rescue severely exhausted CD8 T cell responses that are otherwise refractory to rescue by PD-L1 blockade alone. The immune rescue by dual CD4 T cell depletion and PD-L1 blockade therapy was well tolerated with 100% survival.

In summary, our data demonstrate that as a chronic viral infection progresses over time, there is a reduction in the therapeutic efficacy of PD-L1 blockade. This may be because of the progressive differentiation of CD8 T cells into more exhausted subsets, whose function may be regulated by multiple inhibitory pathways that override PD-1/PD-L1 inhibition. Some of these potent inhibitory pathways that impinge on deeply exhausted CD8 T cells may be induced in part by expanded Treg responses, which may be activated by PD-L1 blockade. Intriguingly, PD-L1 blockade at the later stages of infection appears to only “rescue” Treg responses, which may possibly counteract the desired CD8 T cell rescue. Altogether, these data suggest that CD4 T cell depletion together with PD-L1 blockade is able to rescue deeply exhausted CD8 T cells. These findings may be relevant for the treatment of advanced chronic diseases such as those caused by persistent viruses and cancers.

We thank Drs. Rafi Ahmed, John Wherry, Alice Kamphorst, Rafael Larocca, and Kathryn Stephenson for discussions and assistance.

The authors have no financial conflicts of interest.