Platinum-Based Chemotherapy Attenuates the Effector Response of CD8 T Cells to Concomitant PD-1 Blockade

Cancer immunotherapy with immune checkpoint blockade has revolutionized the clinical management of non–small cell lung cancer (NSCLC), which still represents the leading cause for cancer mortality worldwide (1–2). Treatment with monoclonal antibodies blocking the interaction of the coinhibitory receptor programmed cell death 1 (PD-1) with its main ligand programmed death-ligand 1 (PD-L1) has been shown to improve antitumor T-cell immunity (2–5). To date, blockade of the PD-1 pathway, with anti–PD-1 or anti–PD-L1 antibodies, represents the backbone of the systemic treatment for non-oncogene addicted NSCLC (3–4). For metastatic or recurrent disease, the choice to use PD-1 blockade alone or in combination with chemotherapy and/or ipilimumab is based on PD-L1 tumor proportion score on biopsy specimen (3). In association with PD-1 blockade, the chemotherapy regimen consists of a platinum derivative (cisplatin or carboplatin) plus pemetrexed or paclitaxel, whether the histological subtype is non-squamous or squamous (3, 5).

Studies using syngeneic tumor models have already investigated the mechanisms involved in the antitumor activity of chemo-immunotherapy, showing that chemotherapy has immune-modulating properties and can synergize with PD-1 pathway blockade (6–9). In addition to directly reducing the tumor burden, chemotherapy has also been shown to stimulate tumor antigen release and presentation, resulting in increased activation of tumor-infiltrating lymphocytes (8–9). The magnitude of these phenomena seems to depend on the type and dose of chemotherapy, and recently the anti-folate agent pemetrexed has been shown to share these immunogenic features (10–11). On the other hand, the alkylating agent cisplatin has been reported to be a weak inducer of immunogenic cell death (9); however, it was also shown that in vitro cisplatin increased CD4 T cell activation through the release of IFNβ by monocyte-derived dendritic cells (12).

Despite the immune-modulating properties proposed above, how chemotherapy affects the proliferation and differentiation of CD8 T cells in the context of persistent antigen stimulation has yet to be addressed. For this purpose, we used the well-established murine model of T-cell exhaustion with chronic lymphocytic choriomeningitis virus (LCMV) infection. This prototypical model presents the advantage to study the impact of chemotherapy on CD8 T cells that proliferate solely in response to PD-1 blockade, in the absence of the immunogenic effects of chemotherapy in terms of tumor antigen release and presentation. Using this same LCMV model, previous studies have shown how persistent antigen stimulation—as in chronic infection and cancer—upregulates PD-1 expression on T cells and leads to CD8 T-cell exhaustion, a process characterized by suboptimal proliferative capacity, cytokine production, and cytotoxicity (13–15). In both mice and humans, it has now been documented that antigen-specific PD-1+ CD8 T cells are heterogeneous and composed of at least 3 subpopulations during chronic infections and cancer, recapitulating the progressive stages in the differentiation and maintenance of exhaustion (16–21). We and others have found that the transcription factor T cell factor 1 (TCF-1) identifies a functional subset of PD-1+ CD8 T cells that is endowed with stem-like features and resides in lymphoid tissues (17, 21–23). Indeed, these resource cells, also referred to as precursor of exhausted CD8 T cells, sustain the effector immune response through slow self-renewal and generation of more differentiated cells that upregulates the co-inhibitory receptor TIM-3. Within this effector population, the expression of the chemokine receptor CX3CR1 defines a circulating transitory effector subset that is highly proliferating and represents the early progeny of the stem-like CD8 T cells (18, 24). Eventually, after persistent antigen exposure, virus- or tumor-specific PD-1+ CD8 T cells will further evolve into a terminally differentiated resident subset, characterized by the expression of the CD101 marker (18, 23, 25).

Here, our aim was to assess whether and how chemotherapy affects the CD8 T-cell response to PD-1 blockade under persistent antigenic stimulation. In particular, due to the cytotoxicity of chemotherapy on highly proliferating cells, we assessed if chemotherapy affects the proliferative burst of CD8 T cells needed to generate an effective immune response upon blockade of the PD-1 pathway.

Female C57BL/6J were obtained from The Jackson Laboratory. All mouse experiments were performed with approval of the Emory University Institutional Animal Care and use Committee. For chronic LCMV, mice were infected at 6 to 8 weeks of age through intravenous injection with 2 × 10⁶ pfu LCMV clone 13, according to procedures previously described (17, 26). Two days before, and the same day of infection, mice were injected intraperitoneally with 300 ug of CD4 T cell-depleting antibody GK1.5 (Bio X Cell), as temporarily CD4 T-cell depletion is required to maintain life-long systemic infection and to induce the CD8 T-cell exhaustion (27). LCMV clone 13-infected CD4-temporarily depleted mice were analyzed after day 45 with blood sampling, to check for effective infection. For acute infection, C57BL/6J mice were injected intraperitoneally with 2 × 10⁵ pfu LCMV Armstrong (13).

In LCMV chronically infected mice, treatment interventions started at least 45 days post-infection. Chemotherapy consisted of cisplatin and pemetrexed, the first platinum-doublet approved for combination with PD-1 blockade, and still widely used for the treatment of non-squamous NSCLC (3, 28). All drugs were administered intraperitoneally. PD-1 blockade was obtained injecting 10 mg/kg of aPD-L1 (anti-mouse PD-L1 antibody; clone 10F.9G2, prepared in house) dissolved in PBS. Chemotherapy was based on cisplatin and pemetrexed, purchased from Sigma-Aldrich and LC Laboratories, respectively. Cisplatin (CAS 15663­27­1) was dissolved in saline solution 0.9% and injected at the dose of 2.5 mg/kg, as in previous reports (6, 29). Pemetrexed (CAS 357166–29–1) was dissolved in PBS and injected at the dose of 300 mg/kg, similar to previous reports considering treatment intensity per week (10–11, 30). Cisplatin and pemetrexed were administered as separate injections, 3 to 6 hours following aPD-L1 injection. Conventionally, mice dosages were calculated considering 20 grams as average mice weigh.

For the chronic LCMV infection experiments, at day 15 after treatment initiation, mice were euthanized and blood, spleen, and lungs were collected and processed for lymphocyte isolation and staining for flow cytometry analyses. Spleen was also processed and analyzed for intracellular cytokine staining and plaque assay. For the acute LCMV infection experiment, mice sacrifice, spleen collection and lymphocyte isolation were performed at day 8 post-infection. Tissues from LCMV infected mice were processed as previously reported (13, 31). Briefly, for isolation of lung ymphocytes, lungs were first perfused by injecting in the right ventricle with 10 mL of ice-cold PBS, then removed and trimmed with scissors. Following incubation in 0.25 mg of collagenase B (Boehringer Mannheim)/mL and 1 U/mL of DNase (Sigma) at 37°C for 60 minutes, the digested lungs were passed through a cell strainer and centrifuged, and the pellet was resuspended in 5 to 10 mL of 44% Percoll (Sigma). This suspension was underlaid with 67% Percoll and spun at 850 × g for 20 minutes at room temperature, the lung lymphocyte population was harvested from the interface, and red blood cells were lysed using 0.83% ammonium chloride.

2 × 10⁶ splenocytes isolated from LCMV infected mice were incubated in RPMI with 10% FBS with 0.2 mg/mL of LCMV peptides (Gp33–41, Gp276–286) for 5 hours at 37°C in the presence of BD GolgiStop and BD GolgiPlug, according to the manufacturer's protocols. After incubation, cells were stained for flow cytometry (see below).

The following antibodies were used in flow cytometry experiments: live/dead (Invitrogen, Fixable Viability Dye), CD8a (BD Biosciences, 53–6.7), CD4 (BD Biosciences, GK1.5) CD44 (BD Biosciences, IM7), TIM-3 (BioLegend, RMT3–23), PD-1 (BioLegend, 29F.1A12), CD101 (Invitrogen, Moushi101), CX3CR1 (BioLegend, SA011F11), and LCMV-specific tetramers (DbGp33–41; DbGp276–286) as surface markers, and TCF-1 (Cell Signaling Technology, C63D9, RRID:AB_2199302), Ki-67 (BD Biosciences, B56) as intracellular markers. IFNγ (BioLegend, XMG1.2) and TNFα (BioLegend, MP6-XT22) were also used for intracellular cytokine staining. MHC class I tetramers were prepared in-house and used as previously described (31). Single cell suspensions were stained with antibodies and/or tetramer in PBS with 2% FBS and 2 mmol/L EDTA for extracellular targets. For intranuclear staining, the eBioscience Foxp3 Transcription Factor Staining Buffer Set (Thermo Fisher Scientific) was used for fixation, permeabilization, and intracellular staining. Samples were acquired using a 5 L Aurora-Cytek spectral flow cytometer or BD LSR II, and analyzed using FlowJo v.10 (BD Biosciences, RRID:SCR_008520).

Splenic viral titers were determined by plaque assay on Vero E6 cells (RRID:CVCL_XD71) as described previously (32). Briefly, 7.5×10⁵ Vero cells were plated in 35-mm wells in 6-well dishes (Costar, Cambridge, MA). The plates were incubated at 37°C and used for the assay when the cell monolayers were confluent. The medium was removed and the samples to be titrated were added to the cells (0.2 mL vol). After adsorption for 60 minutes at 37°C, the cells were overlaid with 3 mL of 0.5% Seakem agarose (FMC Corporation) in Medium 199 (Gibco Laboratories) supplemented with 10% heat-inactivated FCS, antibiotics, and l-glutamine. The plates were incubated for 3 days at 37°C and then overlaid with 2.0 mL of 0.5% agarose in Medium 199 containing 0.01% neutral red (Gibco Laboratories). Plaques were scored the following day.

All experiments were done in duplicate or triplicate, and the quantified data shown in dot plots represent combined data, unless otherwise noted. Summary graphs and statistics were performed in GraphPad Prism v9.3 (RRID:SCR_002798). Statistical significance was set at a p value lower than 0.05. To test differences across the treatment groups, t test, one-way or two-way ANOVA with multiple comparison test were used, as indicated in the figure legends.

The raw data generated in this study are available upon reasonable request to the corresponding author.

To directly assess the effects of chemo-immunotherapy on the effector response of CD8 T cell, we used the well-established model of T-cell exhaustion with the chronic LCMV infection. Groups of chronically infected mice (≥45 days post-infection) were either left untreated, given chemotherapy alone, treated with aPD-L1, or given combination therapy with chemotherapy and aPD-L1 (Fig. 1A). To recapitulate the treatment schedule used in patients with cancer, chemotherapy, based on cisplatin and pemetrexed, and aPD-L1 were administered on the same day, every 3 days (Fig. 1B). After five doses of treatment, mice were sacrificed for T-cell analysis in a lymphoid (spleen) and nonlymphoid tissue (lungs), and for the quantification of splenic viral titers.

Compared with the untreated group, treatment with aPD-L1 did not change the overall number of lymphocytes in the spleen, while chemotherapy either alone or in combination with aPD-L1 induced a modest reduction of the total lymphocytes (about 1.7-fold decrease for both chemotherapy and chemotherapy + aPD-L1 groups). A similar trend was observed also for the total number of CD8 and CD4 T cells (Supplementary Fig. S1A).

In the spleen, treatment with aPD-L1 alone mediated a significant expansion of PD-1+ and LCMV-specific (PD-1+) CD8 T cells, the latter of which were detected by using DbGp33 and DbGp276 tetramers (Fig. 1C and D; Supplementary Fig. S1B). The number of PD-1+ CD8 T cells more than doubled after aPD-L1, whereas such expansion was not observed after combination treatment (2.2-fold lower compared with PD-1 blockade; P < 0.0001). Compared with the untreated mice, PD-1+ CD8 T cells were significantly reduced after chemotherapy alone (P = 0.0003; Fig. 1D). We observed the same trend for LCMV-specific CD8 T cells, where the number of Gp33+ and Gp276+ CD8 T cells in the combination group was nonsignificantly higher than in the untreated mice, and significantly lower compared with the aPD-L1 group (Fig. 1D). Chemotherapy alone significantly reduced the number of both Gp33+ and Gp276+ CD8 T cells (Fig. 1D).

The response to PD-1 blockade in the spleen was noted also in terms of relative frequency of PD-1+ CD8 T cells, which was over 3-fold higher than the untreated group (P < 0.0001). The frequency of Gp33- and Gp276-specific CD8 T cells was 2.8- and 4.3-fold higher, respectively, after aPD-L1 monotherapy (Supplementary Fig. S1C) compared with the untreated mice. However, the frequency of PD-1+ and LCMV-specific CD8 T cells was not significantly different between the chemotherapy and the untreated groups, or between the aPD-L1 group and the combination groups.

The analysis of PD-1+ and LCMV-specific CD8 T cells in the lung showed similar results to what was observed in the spleen across the treatment groups (Fig. 1E; Supplementary Fig. S1D). Overall, these results indicate that the co-administration of chemotherapy and PD-1 blockade largely abrogates the therapeutic effects of PD-1 blockade on expanding LCMV-specific CD8 T cells.

Hypothesizing that the reduced number of LCMV-specific cells in chemotherapy + aPD-L1 group compared with aPD-L1 alone group was a consequence of the detrimental effect of chemotherapy on dividing cells, we quantified proliferating PD-1+ and LCMV-specific CD8 T cells by staining them with Ki-67 (Fig. 2A and B). After single-agent aPD-L1 treatment, the number of proliferating (Ki-67+) PD-1+ CD8 T cells was almost 4-fold higher compared with the untreated mice (P = 0.03) in the spleen (Fig. 2A). In contrast, in mice receiving the chemotherapy + aPD-L1 combination there was no increase in the number of proliferating PD-1+ CD8 T cells (1.3-fold higher compared with the untreated mice; P = 0.97). Notably, chemotherapy affected the proliferation of PD-1+ CD8 T cells even in steady-state conditions, when administered alone. At this time point of chronic LCMV infection, the proportion of Ki-67+ PD-1+ CD8 T cells was similarly low in the untreated and chemotherapy groups (mean of 6.7% in the untreated mice). After PD-1 blockade, the proportion of proliferating PD-1+ CD8 T cells increased to an average of 15% (P = 0.0001 compared with untreated mice; Fig. 2A). The restrained proliferative burst of PD-1+ CD8 T cells in the chemotherapy + aPD-L1 combination group, was confirmed also by the lower relative frequency of Ki-67+ PD-1+ CD8 T cells (P = 0.04 compared with the aPD-L1 group). Number and frequency of proliferating LCMV-specific CD8 T cells recapitulated the same trend as observed for overall PD-1+ CD8 T cells across the treatment groups (Fig. 2A). In the lung, the difference in proliferating LCMV-specific cells between the aPD-L1 and the combination group was more pronounced, especially in terms of relative proportions (Fig. 2B).

To further assess the cytotoxic effect of chemotherapy on proliferating CD8 T cells, we tested the same platinum-doublet in another condition characterized by an even greater proliferation, during the early effector phase of acute LCMV infection. Upon antigen encounter, newly activated CD8 T cells undergo a rapid cell division, mounting a robust effector response, with viral clearance by day 8 post-infection (13). Chemotherapy was given 6 hours before inducing acute LCMV infection, and mice received two more doses at day 3 and 6 post-infection, with animal sacrifice and T-cell analyses performed at day 8 (Supplementary Fig. S2A). We found reduced number of both total (P = 0.0004) and activated CD8 T cells (2.8-fold change, P = 0.002; Supplementary Fig. S2B). On the other hand, the number of naïve CD8 T cells (CD44^low) was not reduced after chemotherapy. Consistently, the number of LCMV-specific CD8 T cells was also significantly lower in acutely infected mice receiving chemotherapy [3-fold change in the number of both Gp33 (P = 0.005) and Gp276 (P = 0.003) cells; Supplementary Fig. S2C]. On the basis of these findings, chemotherapy attenuates the proliferative burst of effector CD8 T cells, both in response to PD-1 blockade in the context of persistent antigen stimulation, and in response to antigen encounter in the context of the expansion phase of acute LCMV infection.

Once observed that chemotherapy attenuates the proliferation of CD8 T cells responding to concomitant PD-1 blockade, we assessed if any specific CD8 T-cell population among LCMV-specific CD8 T cells was more vulnerable to the cytotoxicity of chemotherapy. Across treatment groups, we examined the number and frequency of the stem-like cells (PD-1+TCF-1+TIM-3-) and the more differentiated progeny of transitory effector (PD-1+CX3CR1+TIM-3+ or PD-1+CD101-TIM-3+), and terminally differentiated (PD-1+CD101+TIM-3+) LCMV-specific CD8 T cells in the spleen. Likely due to the slow self-renewal, the stem-like subset was mostly spared by the cytotoxicity of chemotherapy alone (Fig. 3A). The number of stem-like cells did not change after chemotherapy compared with the untreated mice, and increased in both the aPD-L1 and the combination groups. However, the magnitude of expansion of stem-like cells tended to be greater after PD-1 blockade compared with the combination therapy, in particular for the Gp276+ CD8 T cells (1.6-fold change; P = 0.048; Fig. 3A). Of note, the relative proportion of stem-like cells was higher in the chemotherapy group for both the Gp33+ and Gp276+ CD8 T cells compared with the groups of untreated, aPD-L1 treatment, and chemotherapy + aPD-L1 combination therapy (Supplementary Fig. S3A). These data suggest the resistance of stem-like cells to cytotoxicity, at least in their resting phase.

The transitory effector PD-1+CX3CR1+TIM-3+ CD8 T-cell subset expanded after both PD-1 blockade and chemotherapy + aPD-L1 combination (Fig. 3B; Supplementary Fig. S3B). After aPD-L1 treatment alone, the number of transitory effector cells was 2.4-fold higher for the Gp33+ CD8 T cells (P = 0.03) and 8-fold higher for the Gp276+ CD8 T cells (P < 0.0001) compared with those in untreated mice. It is worth to notice that among the three subsets, the transitory effector is the only subset that increases in terms of relative frequency after treatment with aPD-L1 (from an average of 10% in the untreated mice to an average 18%–20% in the aPD-L1 group). Even though the proportion of transitory effector cells did not change by adding chemotherapy to PD-1 blockade (Supplementary Fig. S3B), their absolute number was slightly lower in mice treated with chemotherapy + aPD-L1 combination compared with aPD-L1 for Gp33+ CD8 T cells (1.5-fold change; P = 0.1) and significantly lower for Gp276+ CD8 T cells (2.2-fold change; P = 0.02; Fig. 3B).

Terminally differentiated CD8 T cells represent the most prevalent subset at this late time-point of chronic infection, accounting for over 70% of the LCMV-specific CD8 T cells in untreated mice (18). While chemotherapy alone resulted in a decreased number of terminally differentiated LCMV-specific CD8 T cells, after PD-1 blockade these cells were significantly augmented compared with the untreated group (Fig. 3C). In contrast, adding chemotherapy to PD-1 blockade abrogated the expansion of terminally differentiated CD8 T cells seen after PD-1 blockade alone [1.6-fold decrease for Gp33+ CD8 T cells (P = 0.01) and two-fold decrease for Gp276+ CD8 T cells; P = 0.004; Fig. 3C]. The relative distribution of terminally differentiated CD8 T cells was comparable across the 4 groups for the Gp33+ CD8 T cells, while for the Gp276+ CD8 T cells, the proportion was higher in the untreated and chemotherapy groups, compared with both the aPD-L1 and the combination group (Supplementary Fig. S3C).

We then assessed the extent of proliferation, defined by the Ki-67+ population, of stem-like (PD-1+TCF-1+TIM-3-), transitory effector (PD-1+CD101-TIM3+), and terminally differentiated (PD-1+CD101+TIM-3+) CD8 T-cell subsets among LCMV-specific Gp276+ CD8 T cells across the treatment groups. Similar to what was shown by Gill and colleagues (33), the proliferation of the stem-like subset was low at this time-point, but with numerically higher after PD-1 blockade alone (mean 4.3% in aPD-L1 vs. 1.4% in untreated and 2.4% in the combination groups; Fig. 4A and B). As expected, the transitory effector subset presented the most intense proliferation, which further increased after aPD-L1 therapy (from 17% in the untreated to 41% in the aPD-L1 groups; P < 0.0001; Fig. 4B). It should be noted that compared with single-agent aPD-L1, in the chemotherapy + aPD-L1 group, both the intensity of Ki-67 expression (Fig. 4A) and the proportion of Ki-67+ cells within the PD-1+CD101-TIM-3+ subset were significantly reduced (mean value 26.7% vs. 42% in the combination and the aPD-L1 group; P = 0.0002; Fig. 4B). The frequency of Ki-67+ cells in the transitory effector Gp276+ CD8 T cells was lower also after chemotherapy compared with the untreated control (Fig. 4B). In the terminally differentiated subset, Ki-67 expression slightly increased in both the groups receiving aPD-L1 (Fig. 4A), and the proportion of Ki-67+ cells in terminally differentiated CD8 T cells was not different between mice treated with aPD-L1 and chemotherapy + aPD-L1. Overall, these findings indicate that upon concomitant PD-1 blockade, chemotherapy preferentially affects proliferating PD-1+CD101-TIM-3+ transitory effector cells among the three CD8 T-cell subsets during chronic LCMV infection.

The reduced burst of effector CD8 T cells seen after chemotherapy + aPD-L1 combination treatment, compared with aPD-L1 alone, had negative effects on the IFNγ production of LCMV-specific CD8 T cells. Upon ex vivo stimulation with LCMV peptides, the number of IFNγ producing CD8 T cells in the combination group was comparable to that in the untreated control (P = 0.3), and lower compared with the aPD-L1 group (mean fold change 1.9; P = 0.07; Fig. 5A and B). Consistent with the exhausted state of LCMV-specific CD8 T cells, only a small portion of CD8 T cells secreted IFNγ in untreated mice (mean frequency 1%). In the combination therapy group, the proportion of IFNγ producing CD8 T cells was higher than that in untreated mice, but lower compared with that in the aPD-L1 group.

Viral titers were measured as the in vivo readout of the effector functions of CD8 T cells. The splenic viral load was higher in mice treated with chemotherapy + aPD-L1 combination therapy than in those treated with PD-1 blockade (2.8 mean fold change, P = 0.03), while comparable to the untreated groups (1.7 fold lower compared with the untreated; P = 0.2; Fig. 5C). These results demonstrate that concomitant chemotherapy + aPD-L1 has negative effects on the therapeutic efficacy of PD-1 blockade in the chronic LCMV infection.

To assess if an alternative timing of administration of chemotherapy could preserve the proliferative burst of effector CD8 T cell upon PD-1 blockade, we tested a sequential approach of chemotherapy, followed by aPD-L1. As shown in Fig. 6A, chronically infected mice were treated with five doses of chemotherapy (cisplatin and pemetrexed), followed by five doses of aPD-L1 (sequential group). Both treatments were given every three days, and aPD-L1 therapy started two days after the completion of chemotherapy. This schedule was compared with the conventional concomitant combination of five doses of chemotherapy + aPD-L1 (concomitant group). During the chemotherapy phase in the sequential group, mice in the concomitant group received mock (PBS) intraperitoneally. After treatment completion, the numbers of total lymphocytes, CD8+, PD-1+ and LCMV-specific CD8 T cell in the spleen were significantly higher in the sequential group compared with the concomitant [3- and 3.8-fold change for PD-1+ (P < 0.0001) and Gp33+ (P = 0.01) CD8 T cells, respectively; Supplementary Fig. S4A; Fig. 6B]. The number of Ki67+TIM-3+PD-1+ and of Ki67+TIM-3+Gp33+ CD8 T cells was also superior after sequential combination (for Gp33 CD8 T cells, 5- fold change; P = 0.02; Fig. 6C). The number of stem-like Gp33+ CD8 T cells was significantly lower in the concomitant group (two-fold change; P = 0.04; Supplementary Fig. S4B and S4C). As for the effector compartment, both the effector transitory and the terminally differentiated LMCV-specific CD8 T cells were more abundant after sequential chemotherapy and aPD-L1 [4.4-fold change for transitory effector (P = 0.04); 4-fold change for the terminally differentiated Gp33 (P = 0.01) CD8 T cells; Supplementary Fig. S4B and S4C]. In line with these findings, CD8 T cells producing IFNγ upon ex vivo stimulation with LCMV peptides significantly expanded after sequential exposure to chemotherapy and PD-1 blockade, compared with the concomitant approach (5.3-fold change; P = 0.01; Fig. 6D).

As a results of the preserved proliferation and differentiation of LCMV-specific CD8 T cells, viral control was also improved with the sequential compared with the concomitant combination strategy, in which the viral load in the spleen was 2.8-fold higher (P = 0.0001; Fig. 6E).

In this study, we investigated the in vivo effects of the concomitant and the sequential administration of chemotherapy and PD-1 pathway blockade on T-cell exhaustion. The synergistic antitumor activity of chemotherapy with platinum + pemetrexed and PD-1 blockade has already been proven in tumor models (10–11). Here, we explored how chemotherapy affects the T cell proliferative response to PD-1 directed therapy, and the maintenance and differentiation of CD8 T-cell subsets using the chronic LCMV infection model of T-cell exhaustion. Using this model, we were able to assess the cytotoxicity of chemotherapy in a more controlled environment, where CD8 T cells proliferate only in response to PD-1 directed therapy. We have found that chemotherapy has a detrimental effect on the proliferative response of CD8 T cells to concomitant PD-1 blockade, and accordingly, on the ability of CD8 T cells to produce cytotoxic cytokines, which are both necessary for a successful immune response (14). Indeed, the restrained effector functions after combination treatment eventually resulted in impaired viral control, compared with treatment with aPD-L1 alone.

Even though the differentiated effector subset was the preferential target of the cytotoxicity of chemotherapy, the stem-like population was more resilient in the 2-week treatment window of our experiment. So far, only memory CD8 T cells have been shown to be chemo-resistant, in both animal and human studies. It seems likely that stem-like cells survive chemotherapy due to their quiescence at the steady-state condition. However, it is possible that this is only one of the factors at play. It has been shown in progenitor cells the existence of a cytoprotective machinery regulated at multiple levels to withstand genomic insult, thus allowing long-term survival and preventing the propagation of misrepaired DNA lesions to the downstream progeny (34–39). These mechanisms have been shown to protect epithelial stem cells and memory CD8 T cells from genotoxins, while more differentiated CD8 T cells seem to be affected (34, 35, 39, 40). On the basis of this evidence, we can hypothesize that the contraction of the terminally differentiated subset observed in our study could be due to more than one reason. On one hand, it could be simply the result of the reduction of the upstream population of proliferating transitory effector cells; on the other hand, chemotherapy could directly induce cell death of terminally differentiated cells, potentially more vulnerable to DNA damage, compared with the less differentiated counterpart of early generated effector and stem-like cells. Future studies would be needed to confirm and thoroughly investigate the program to maintain genome integrity in exhausted CD8 T-cell subsets.

The antitumor activity of chemotherapy mostly relies on the ability to deplete rapidly proliferating cells, a shared characteristic of chemotherapeutics, irrespective of the specific agent. Therefore, it seems plausible that the detrimental effects observed after platinum + pemetrexed on the proliferative response of CD8 T cells do not strictly depend on the type of chemotherapy. However, dedicated studies would be needed to confirm this hypothesis. In this regard, it would be clinically relevant to assess if our findings apply to the platinum-based doublet with paclitaxel, that is also widely used for treatment of patients with NSCLC, in combination with PD-1 blockade, with or without anti-cytotoxic T lymphocyte antigen-4 (CTLA-4; ref. 5).

While the high antitumor efficacy of chemo-immunotherapy has been widely confirmed in clinical trials, our findings provide a proof-of-concept that current combinations could be further improved in clinical practice. Our data from the timing experiment suggest that the detrimental effect of chemotherapy (alone) on the proliferation of CD8 T cell could be transient and reversible after chemotherapy discontinuation, at least in the relatively short treatment window of our experiment. The underlying reason could be that PD-1+ stem-like CD8 T cells are spared when chemotherapy is administered during their quiescent phase, thus preserving their differentiation and self-renewal capacity when later exposed to PD-1 blockade, as observed in the sequential combination regimen. However, stem-like CD8 T cells appear to be more vulnerable to the cytotoxicity of chemotherapy in case of concomitant combination, in light of the accelerated self-renewal they undergo when stimulated by PD-1 blockade. As we recently described (33), this cellular mechanism confers long-term durability to the stem-like subset, counteracting the simultaneous increased differentiation into effector cells that occurs under PD-1 blockade. With this regard, it would be valuable to explore the consequences of prolonged concomitant chemo-immunotherapy on PD-1+ TCF-1+ stem-like cells. Upon the sustained differentiation into effector cells mediated by concomitant PD-1 directed therapy, an impairment in the self-renewal of stem-like cells could lead to a progressive loss of this population, compromising their long-term persistence. Investigating the long-term maintenance of the stem-like exhausted CD8 T cells after chemotherapy ± PD-1 blockade has key translational value, especially in the context of lung cancer, where the use of chemo-immunotherapy has been recently expanded to the adjuvant and neoadjuvant setting of the early stage disease, for curative intent (41–43).

Compared with the concomitant combination, a sequential strategy may spare the self-renewal and differentiation of stem-like cells boosted by PD-1 blockade. Another potential advantage of the sequential combination regimen is that the proliferative response of CD8 T cells to PD-1 blockade could be enhanced upon conditions of homeostatic proliferation, as in the case of chemotherapy-induced lymphopenia. In support of this hypothesis, in our study, the attenuation of the proliferative response in concomitant combination group was greater compared with the sequential combination than compared with PD-1 blockade alone.

Translational studies should confirm the findings we obtained in the LCMV model, with potential clinical implications on optimized timing and dosing of current and future chemo-immunotherapy strategies. At present, patients with NSCLC with evidence of clinico-radiological response after 4 cycles of platinum-based chemo-immunotherapy are eligible to maintenance immunotherapy with PD-1 blockade, that is usually combined with pemetrexed in case of non-squamous histology (3, 5). In this context, our findings support the choice for a maintenance treatment based on PD-1 pathway blockade alone, and in general for a shorter chemotherapy course. Overall, the optimal duration of chemo-immunotherapy still needs to be established and it is a matter of debate in clinical practice. It should be mentioned that alternative chemo-immunotherapy combination regimens are underway or have already been approved across different tumor types (44–47).

We and others have shown that in the peripheral blood of patients with cancer the expansion of PD-1+ CD8 T cells early after initiation of PD-1 blockade is predictive of clinical benefit (48–49). Our findings provide a solid rationale for the translational investigation of how the association of chemotherapy with PD-1/PD-L1 blockade can modify the magnitude or the duration of this proliferative CD8 T-cell response in patients with cancer. Furthermore, the recent approval of chemo-immunotherapy in the neoadjuvant setting would represent an opportunity to assess more in depth its impact on the self-renewal and differentiation of stem-like CD8 T cells, that are crucial for sustained antitumor immunity.

V. Nicolini reports employment at Roche. D. Sangiolo reports personal fees from GSK, Roche, and AstraZeneca outside the submitted work. G.V. Scagliotti reports personal fees from AstraZeneca, Eli Lilly, MSD, Pfizer, Roche, Johnson & Johnson, Takeda, and Verastem and grants from Tesaro and MSD during the conduct of the study. S.S. Ramalingam reports grants from Bristol-Myers Squibb, AstraZeneca, Merck, and Pfizer outside the submitted work. No disclosures were reported by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

A. Mariniello: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft. T.H. Nasti: Conceptualization, data curation, supervision, investigation, visualization, methodology, writing–review and editing. D.Y. Chang: Conceptualization, investigation, visualization, methodology, writing–review and editing. M. Hashimoto: Conceptualization, supervision, visualization, methodology, writing–review and editing. S. Malik: Investigation, visualization, writing–review and editing. D.T. McManus: Investigation, visualization, writing–review and editing. J. Lee: Investigation, visualization, writing–review and editing. D.J. McGuire: Investigation, visualization, writing–review and editing. M.A. Cardenas: Investigation, visualization, writing–review and editing. P. Umana: Visualization, methodology, writing–review and editing. V. Nicolini: Visualization, methodology, writing–review and editing. R. Antia: Visualization, methodology, writing–review and editing. A. Saha: Visualization, methodology, writing–review and editing. Z. Buchwald: Visualization, writing–review and editing. H. Kissick: Visualization, writing–review and editing. E. Ghorani: Visualization, writing–review and editing. S. Novello: Resources, visualization, writing–review and editing. D. Sangiolo: Resources, supervision, validation, writing–review and editing. G.V. Scagliotti: Resources, supervision, visualization, writing–review and editing. S.S. Ramalingam: Conceptualization, resources, supervision, funding acquisition, visualization, methodology, writing–review and editing. R. Ahmed: Conceptualization, resources, supervision, funding acquisition, visualization, writing–original draft, project administration, writing–review and editing.

This project was supported by the NCI of the NIH grants P50CA217691 (to S.S. Ramalingam and R. Ahmed) and R01AI030048 (to R. Ahmed).