The Folate Pathway Inhibitor Pemetrexed Pleiotropically Enhances Effects of Cancer Immunotherapy

PD(L)1 inhibitors have markedly changed the therapeutic landscape in many tumor types including non–small cell lung cancer (NSCLC), and these agents are becoming standard of care across an increasing number of tumor types (1, 2). However, clinical benefit from these therapies is limited, and tumor recurrences are common (3, 4). One strategy to improve the efficacy of PD(L)1 inhibitors is to combine these agents with tumor-targeting therapies that have the potential for cooperative mechanistic interactions with immune agents (4, 5). Indeed, numerous clinical trials are underway to evaluate the potential to combine immune checkpoint inhibitors (ICIs) and chemotherapies (6).

Pemetrexed is an established chemotherapeutic that disrupts the folate pathway and is part of the standard of care for nonsquamous NSCLC and mesothelioma (7). The front-line treatment with pemetrexed, carboplatin, and anti-PD-1 (pembrolizumab) has been evaluated in patients with NSCLC in the randomized KEYNOTE-021G and KEYNOTE-189 trials (8, 9), leading to the accelerated approval of this regimen based on substantial increase in progression-free survival and overall response rate in the KEYNOTE-021G study. These improvements represented the first approval of chemo-immunotherapy combination (10). The rationale to combine chemotherapy with ICIs is based at least in part on the concept of immunogenic cell death (ICD) that can be a consequence of the cytotoxic effects of chemotherapeutic agents on tumor cells (11). ICD involves the release of immune-stimulating factors from dying tumor cells that drive antigen cross-presentation, T-cell priming, and adaptive immune response against tumors. Cytotoxic agents are not equipotent in their ability to induce ICD; only a few cytotoxic agents (e.g., anthracyclines, oxaliplatin) have been demonstrated to induce ICD, whereas most chemotherapeutics induce non-ICD (11). Cytotoxic agents can also be deleterious to the immune compartment by cytotoxic targeting of immune cells (11). The positive interaction between pemetrexed-based chemotherapy and immune checkpoint blockade in KEYNOTE-021 and KEYNOTE-189 trials may seem counterintuitive given that antifolate agents (e.g., methotrexate) have been used as immunosuppressive agents to treat patients with inflammatory conditions, and part of their immunosuppressive activity appears to involve the inhibition of T cells (12–15). Furthermore, recent work has identified one-carbon metabolism, which includes the folate and methionine cycles, as a top ranked metabolic pathway engaged during T-cell activation and survival (16, 17).

The main objective of this work was to obtain mechanistic insights into the immunostimulatory activity of pemetrexed ±PD1 blockade rather than justify clinical development of pemetrexed/anti-PD(L)1 combinations in NSCLC and other tumor types. We demonstrate that pemetrexed therapy exerts previously unknown immunomodulatory effects that result in an immune-permissive tumor microenvironment and improves the antitumor efficacy of PD(L)1 blockade. These results provide fundamental insights into the mechanisms underlying the combinatorial activity of pemetrexed and anti-PD-1 therapy, and provide a strong rationale for further exploration of combinations of pemetrexed and other folate pathway modulators with immunotherapies.

Colon26 and MC38 cell lines were purchased from DTP and NCI DCTD Tumor Repository, respectively. Female BALB/c and C57BL/6 mice were purchased from Envigo. All experimental procedures were done in accordance with the NIH Guide for Care and Use of Animals and were approved by the Institutional Animal Care and Use Committee.

Mouse splenic T cells stimulated with CD3/CD28 were cultured in the presence of pemetrexed as indicated. Oxygen consumption rate (OCR) was analyzed using Seahorse XF Cell Mito Stress Test Kit and Seahorse XFe96 instrument (Agilent). Cells were sequentially stimulated with oligomycin (1 μmol/L), FCCP (1.5 μmol/L), and rotenone/antimycin A (0.5 μmol/L each) and the spare respiratory capacity (SRC) was measured as the difference between basal OCR values and maximal OCR values obtained after FCCP uncoupling. To assess T-cell ability to metabolize fatty acids, XF Palmitate:BSA FAO substrate (Agilent) was incorporated into XF Cell Mito Stress Test assay. Wave 2.4 software (Agilent) was used for data acquisition and analysis of Seahorse data.

Splenocytes from ovalbumin-specific T-cell receptor transgenic OT-1 mice were incubated in the presence of 0.1 nmol/L of SIINFEKL peptide and IL2 for 5 days. CD8⁺T cells were then isolated and cultured with B16 tumor cells that had been previously labeled with cell tracer BV421 and pulsed with 100 nmol/L of SIINFEKL peptide for 2 hours, at a 10:1 effector-to-target ratio. Tumor cell death was analyzed by 7AAD incorporation by flow cytometry after 4 hours of coculture.

Colon26 and MC38 tumor cell lines were treated with pemetrexed, carboplatin, paclitaxel, gemcitabine, or doxorubicin. for 96 hours followed by analysis of high-mobility group B1 (HMGB1) protein and calreticulin (CRT) in culture supernatants using commercially available kits (IBL International, Hamburg, Germany, and Cloud Clone Corporation, Katy, TX, respectively). The viability of remaining cells was measured by Cell Titer-Glo assay (Promega) according to manufacturer's protocol.

QuantiGene Plex and nCounter gene expression assays were done as reported previously with slight modifications (18).

Group-wise statistical comparisons were performed as indicated in each figure using standard paired T tests, 1-way ANOVA, or 2-way ANOVA models with Tukey adjustment per time point, comparing treatment/dose and time point. Additional details are provided in the Supplementary Materials and Methods.

To characterize the effects of pemetrexed on intratumor immune response, initial experiments were performed in immunocompetent syngeneic mouse tumor models. To meet our research objective, the models had to meet 2 prerequisites: (i) demonstrate sensitivity to pemetrexed and (ii) responsiveness to PD(L)1 blockade. We found that MC38 and Colon26 colorectal tumor cell lines were sensitive to pemetrexed, whereas Lewis lung carcinoma (LLC), a commonly used lung cancer model, was pemetrexed-refractory (Supplementary Fig. S1A; ref. 20). Furthermore, although genetically engineered mouse models (GEMM) of lung carcinoma may sound like a logical choice, they are poorly fit for studying effects of immunotherapies, largely because GEMMs are driven by specific oncogenic events and do not exhibit high tumor mutational and neoantigen burden which have emerged as important molecular hallmarks underlying responsiveness of lung tumors to immunotherapy in humans (21, 22). Because single-agent treatment with pemetrexed induced tumor responses across multiple tumor types in early clinical trials (7, 23), we rationalized that it was appropriate to use tumor models with the right biological context irrespective of histology rather than using lung tumor models with no sensitivity to either pemetrexed or anti-PD(L)1.

MC38 tumors are modestly responsive to both PD-1 blockade (20) and pemetrexed (Fig. 1A). The effect of pemetrexed in MC38 model was consistent with thymidylate synthase inhibition (increased deoxyuridine, dUMP and decreased thymidine, dTMP in the tumor and plasma; Supplementary Fig. S1), assessed by metabolomics analysis (19) and the highest dose used in these studies was determined to be the maximum tolerated dose (7, 24, 25). MC38 tumors were responsive to pemetrexed at 50 and 100 mg/kg (% tumor growth inhibition of 30% and 52%, respectively; Fig. 1A). Tumors collected after 14 days of pemetrexed therapy were analyzed for changes in immune cell frequencies using flow cytometry. These analyses revealed that pemetrexed increased the frequency of total intratumoral leukocytes (live CD45⁺cells) at both doses, with a trend towards an increased percentage of total CD3⁺and cycling (Ki67⁺) CD8⁺cells, particularly at 50 mg/kg (Fig. 1B–E). This appeared to be driven mainly by an increase in Ki67⁺CD8⁺T cells, without any other significant differences in myeloid cell subsets (Fig. 1F). Molecular analysis of tumor samples using a custom-made immune profiling QuantiGene Plex (QGP) gene expression panel revealed that treatment with pemetrexed at 50 and 100 mg/kg promoted a T-cell inflamed phenotype, exemplified by upregulation of T-cell activation-associated genes including Pdcd1, Cd8b, Prf1, and Gzma (Fig. 1G) (26). The QGP data also suggest activated vascular endothelium (↑ Icam1, Vcam1, and chemokine Cx3cl1 also known as fractalkine that is induced in endothelium as result of immune activation) and enhanced interferon (IFN) response (↑ Irf7) and antigen presentation [upregulated Itgax and Zbtb1 associated with dendritic cells (DC)]. Beyond these changes, one of the genes most significantly modulated by pemetrexed treatment was Vegfc, which encodes vascular endothelial growth factor C (VEGF-C), a key regulator of lymphangiogenesis. VEGF-C is known to be regulated through the NF-κB pathway and is believed to promote T-cell infiltration rather than inhibit antitumor immune response (27, 28). Nos2, which encodes inducible nitric oxide synthase, is produced by myeloid-derived suppressor cells (MDSC) and DCs, was downregulated at both dose levels, suggesting that pemetrexed could potentially negatively impact myeloid cell subsets (Fig. 1G). Although Nos2 displayed downregulation, we did not observe a significant reduction in CD11b⁺cells by flow cytometric analysis (Fig. 1F). Collectively, these results suggest that pemetrexed influences the functionality rather than frequency of myeloid cells.

Because pemetrexed is administered in combination with platinum agents such as carboplatin and cisplatin in front-line treatment of patients with metastatic NSCLC, we next asked if carboplatin had immunomodulatory effects on the tumor immune microenvironment, and whether the immunomodulatory effects of pemetrexed were affected by the addition of carboplatin. In these experiments, we also evaluated the immunomodulatory effects of the chemotherapy doublet of carboplatin and paclitaxel, a commonly used treatment option in NSCLC, as well as paclitaxel monotherapy. Mice bearing MC38 tumors were treated with pemetrexed, paclitaxel, carboplatin, or combination of pemetrexed with carboplatin, or paclitaxel with carboplatin, at doses designed to model clinical exposures for these agents. Tumors were harvested 14 days after treatment initiation, and immune-related gene expression changes were evaluated by QGP analysis (Fig. 2A). Pemetrexed monotherapy resulted in upregulation of multiple immune-related genes and induced an immune activation signature indicative of IFNγ pathway activation (increased Cd274, Cxcl10, Cxcl11, Psmb8), cytolytic activity (increased Gzma, Prf1), IFN type I response (increased Irf7, Oas3), and activated vascular endothelium (increased Icam1, Vcam1). Paclitaxel monotherapy had a more modest effect on the expression of the gene sets tested, with the immunomodulatory effect mainly associated with moderate upregulation of myeloid cell-related genes (increased Il6, Cxcl1, Ccl2, Ccl3, Ccl4, Timd4). Although carboplatin monotherapy had a weak effect, addition of carboplatin to the pemetrexed regimen appeared to reduce the immunomodulatory effects of pemetrexed and to a lesser extent paclitaxel. Cisplatin had a similar effect (Supplementary Fig. S2), suggesting that platinum agents in general can attenuate immunomodulatory effects of pemetrexed.

To investigate the breadth of pathways modulated by pemetrexed ± carboplatin, we performed NanoString analysis of tumor tissues using the nCounter panels spanning key molecular pathways and cellular compartments of innate and adaptive immunity. Consistent with the QGP data, single-agent pemetrexed treatment significantly modulated the expression of a large number of genes associated with immune response [136 and 133 differentially expressed genes (DEG) for immune profiling and myeloid panel, respectively; Fig. 2B]. Paclitaxel had a quantitatively weaker effect (31 and 39 DEGs for immune profiling and myeloid/innate immunity panel, respectively) with a few DEGs shared between pemetrexed and paclitaxel monotherapy groups (Fig. 2B and C). The combination of pemetrexed and carboplatin yielded less prominent gene expression changes compared with pemetrexed monotherapy (87 and 98 DEGs for immune profiling and myeloid/innate immunity panel, respectively; Fig. 2B and C). Paclitaxel monotherapy and paclitaxel/carboplatin combination induced somewhat different immunomodulatory effects, with a limited number of DEGs overlapped between the 2 treatment groups (Fig. 2B and C).

Ingenuity pathway analysis (IPA) was used to further explore the immune-related molecular and/or cellular pathways modulated by pemetrexed. These analyses revealed macrophage, DC/NK cell, Th1/Th2 enrichment, and evidence of enhanced inflammatory response; innate immune activation and increased IFN signaling in MC38 tumors (Supplementary Table S1). Evaluation of the individual genes associated with these pathways suggested macrophage enrichment/reprograming and DC maturation (upregulation of Cd86, Tlr3, Tlr9, Tnf, Il1b, Il1rl1), increased IFN response and JAK/STAT signaling (upregulation of Ifnar1, Ifngr1, Jak3, Stat1, Stat2, Ifit3, Ifi35, Isg15, Psmb8, Tap1) and enhanced T-cell signaling mainly driven by increased expression of T-cell–specific transcripts (Ifngr1, Il2ra, Il2rb, Il12rb1, Il21r; Fig. 3). The same pathway enrichment was identified in pemetrexed monotherapy and pemetrexed/carboplatin groups; however, combination with carboplatin appeared to qualitatively weaken the effect compared with pemetrexed monotherapy (Supplementary Table S1; Fig. 3). As mentioned earlier, paclitaxel-based treatments exerted a less prominent immunomodulatory effect, and the IPA results showed a similar trend (Supplementary Table S1; Fig. 3). Collectively these data indicate that in MC38 tumors, treatment with pemetrexed or paclitaxel induced both qualitatively and quantitatively different immunomodulatory effects, and addition of carboplatin appeared to attenuate rather than enhance these changes.

The observed immunomodulatory effects of pemetrexed prompted us to evaluate pemetrexed in combination with PD(L)1 blockade. To this end, we performed in vivo combination studies with anti-PD-L1 antibody in MC38 and Colon26 tumor models on 2 distinct genetic backgrounds, C57BL/6 and BALB/c, with distinct immunologic Th1 and Th2 profiles, respectively (29, 30). In MC38 model, combining pemetrexed with anti-PD-L1 resulted in a modest but statistically significant tumor growth delay (Supplementary Fig. S3A). However, Colon26 model displayed greater sensitivity to the combination therapy; the combination of pemetrexed and anti-PD-L1 resulted in more substantial tumor growth delay accompanied by durable responses in some animals (Fig. 4A). No combination benefit was observed in LLC model (Supplementary Fig. S3B).

QGP analysis revealed transient immune-related changes in Colon26 tumors after treatment with monotherapies, whereas the combination effect was most pronounced at a later time point (D14 posttreatment, D24 postimplantation; Supplementary Fig. S3C). To further characterize the effects of the pemetrexed/anti-PD-L1combination in Colon26 model, we performed nCounter analysis of tumor samples collected at D14 posttreatment, where the differences between groups were most apparent. Pemetrexed affected the expression of a limited number of genes (n = 13), with anti-PD-L1 modulating a broader set of genes (n = 57). Combination treatment altered the expression of a large number of genes (n = 198), with the majority (n = 152) uniquely modulated by the combination treatment (Fig. 4B). Although pemetrexed treatment predominantly resulted in the downmodulation of genes in this model (10/13 genes), the combination therapy of pemetrexed and anti-PD-L1 resulted in the upregulation of a substantial number of genes (173/198), including a large set of genes not significantly upregulated by anti-PD-L1 monotherapy.

IPA was used to understand these extensive changes and further explore the immune-related molecular and/or cellular pathways modulated by the combination therapy. The pathways most significantly modulated by the combination involved CD4⁺T cell-mediated immunity (Th1/Th2 pathway) and a pathway referred to as “Granulocyte/Agranulocyte Adhesion and Diapedesis” (Supplementary Table S2). The latter was largely driven by genes encoding cell adhesion molecules (↑ Icam1, Icam2, Pecam, Sell), DC maturation (↑ H2-Ab1, Cd40, Tlr4, Tlr8), and CXC family chemokines and their receptors (↑ Cxcl10, Cxcl12, Cxcl13, Cxcl14,Cxcl16, Cxcr4), which have been associated with the activated vascular endothelium, leukocyte trafficking, and formation of tertiary lymphoid organs (Fig. 4C; ref. 31).

To confirm gene expression changes described above, we subjected Colon26 tumor samples after pemetrexed and/or anti-PD-L1 treatment to flow cytometry analysis (Supplementary Fig. S4A; Fig. 5). Consistent with the IPA results, we detected increased frequency of CD8⁺T cells, CD8⁺/CD4⁺and CD8⁺/Treg ratios along with enhanced activation of effector T cells (Ki67⁺, CD4+ Foxp3^NEG; Fig. 5A). Because the aforementioned gene expression data suggested that modulation of the myeloid cell compartment upon treatment with pemetrexed, we also evaluated the effects of pemetrexed ± anti-PD-L1 on myeloid cells. The combination treatment resulted in a decreased frequency of granulocytic MDSCs (Ly6G+) population and a trend towards greater DC infiltration along with increased activation phenotype of macrophages and Ly6C^highmonocytes which displayed higher expression of MHC class I and II. In addition, the combination treatment also resulted in marked upregulation of MHC class II on tumor cells (Fig. 5B). These data indicate that the combination treatment promotes antigen-presenting properties of myeloid cells, thus supporting the conclusion that pemetrexed induces T-cell-permissive changes in the myeloid cell compartment, leading to an activated and “T-cell priming-competent” phenotype. Given that the percentage of myeloid cells did not change, the data suggest that the effect of pemetrexed on myeloid cells is rather indirect, and may be mediated by immunogenic effects (e.g., ICD) on tumor cells. Of note, although the M1 and M2 phenotypes are well established in mouse macrophage biology, the emerging translational data generated on human breast and lung tumors suggest that the phenotypes of human tumor-associated macrophages are much more complex and cannot be dichotomized into binary M1/M2 states (32, 33). Given the lack of clinical relevance of M1/M2 macrophage polarization in human tumor biology, the bona fide markers of M1 and M2 macrophages were not pursued in our studies.

To understand if the changes in myeloid cells and subsequent T-cell priming in lymph nodes are required for the antitumor effects of pemetrexed and anti-PD-L1, we treated Colon26-bearing mice with pemetrexed and/or anti-PD-L1 together with the well-characterized sphingosine-1-phosphate receptor 1 (S1P1R) antagonist (FTY720) to block T-cell egress from lymph nodes (Fig. 5C). Although FTY720 treatment did not have an obvious impact on either monotherapy, the combination therapy benefit was lost after FTY720 treatment. To understand these observations in the context of an antigen-specific immune response, we examined the effect of the combination therapy on the frequencies of tumor antigen-specific T cells compared with monotherapies in Colon26 model. After 14-day treatment with pemetrexed and/or anti-PD-L1, we evaluated the activation status and frequency of tumor-specific T cells in the tumor, tumor-draining lymph node, and spleen using ELISpot and MHC tetramer assays. The results of these experiments indicate that although a trend towards increased tumor-specific CD8⁺T-cell responses was observed in the periphery during the combination treatment (as exemplified by IFN-gamma ELISpot and gp70 tetramer assay), no appreciable difference in the frequency of gp70 tetramer positive CD8⁺T cells was detected in the tumor. However, a small but statistically significant increase in the frequency of TNFα⁺CD8⁺T cells was observed in Colon26 tumors after treatment with pemetrexed and anti-PD-L1 suggesting that that the increased priming during the combination treatment increases the functionality rather than the quantity of tumor-reactive CD8⁺T cells (Supplementary Fig. S4B).

Collectively, these data demonstrate the development of an integrated antitumor immune response mediated by pemetrexed/anti-PD-L1 combination, and suggest that the underlying mechanism involves perpetuation of T-cell priming in lymph nodes, presumably through the enhanced antigen presentation function of myeloid cells.

Increased antigen presentation and DC maturation gene signatures suggest that pemetrexed treatment may lead to ICD of tumors, activating innate pathways leading to enhanced immune activation. To investigate the ability of pemetrexed to induce ICD, we evaluated the extracellular levels of CRT and HMGB1, both of which are specifically released from cells during ICD. Binding of HMGB1 to Toll-like receptor 4 and CRT to CD91/LRP1 leads to DC migration and maturation and enhanced antigen presentation and T-cell priming (34). Colon 26 and MC38 tumor cells were treated with pemetrexed or other chemotherapeutics (carboplatin, paclitaxel, doxorubicin, or gemcitabine), followed by measurement of the extracellular HMGB1 and CRT release (Fig. 5D). Although all agents tested appeared to induce some degree of ICD, as exemplified by increased CRT and HMGB1 release, pemetrexed was the most potent inducer of ICD in both Colon26 and MC38 cells, particularly across lower concentrations that reflect clinical exposure (0.02–0.05 μmol/L). These results suggest that the immunomodulatory effects of pemetrexed are mediated, at least in part, by tumor cell-intrinsic mechanisms involving ICD.

Purine nucleotide synthesis in general and the folate pathway in particular depend on metabolic intermediates (e.g., 3-phosphoglycerophosphate) supplied through glycolysis, and one-carbon metabolism plays a critical role during T-cell activation because T cells require high levels of glycolysis and mitochondrial respiration during the activation and effector phase (17, 35). To evaluate the impact of pemetrexed on T-cell glycolysis and mitochondrial respiration, we used the Seahorse mitochondria stress test, using primary T cells activated with anti-CD3/CD28 antibodies and IL2 in the presence of pemetrexed over a broad pharmacologic range spanning clinical exposure (0.004–0.1 μmol/L), and determined extracellular acidification rates (reflective of glycolysis), as well as basal and maximal OCR and SRC (reflective of mitochondrial respiration; Fig. 6A). Pemetrexed increased both basal and maximal OCR in an inverse concentration-dependent manner, with the maximum effect at 45 hours of stimulus with the lowest concentration tested (0.004 μmol/L; ref. 36). SRC (representing the difference between basal and maximal OCR values) was markedly increased by pemetrexed in a concentration-dependent manner, and this effect was particularly evident at 45 and 70 hours (Fig. 6A). Pemetrexed also enhanced OCR when activated T cells were supplemented with fatty acid (palmitate), an effect that was abrogated by etoxomir, an inhibitor of carnitine palmitoyltransferase-1 which blocks fatty acid oxidation (Fig. 6B). Because β-oxidation of fatty acids is directly linked with the tricarboxylic acid (TCA) cycle, these results suggest that pemetrexed may increase metabolic fitness of T cells through the enhancement of mitochondrial function or biogenesis that increases the bioenergetic reserve of T cells that might be critical for their survival.

Since chemotherapeutic agents including antifolates are known to have inhibitory effects on T-cell activation and survival (16), we next evaluated the direct impact of pemetrexed compared with paclitaxel on T-cell function and activation. Primary human T cells were activated with CD3 and CD28 antibodies and exposed to fixed, clinically relevant concentrations of pemetrexed (0.05 μmol/L) or paclitaxel (0.2 μmol/L) for various time intervals [days (D) 0–3, 3–9, 0–9] to mimic treatment during different stages of T-cell activation. In vitro activated T cells showed modest but significant attenuation of total proliferation in the presence of pemetrexed, yet this effect was reversible, and most pronounced when pemetrexed was present in the culture medium for the duration of the study, resulting in approximately half the number of total cells compared with untreated cells (Fig. 6C). In contrast, the cytotoxic effect of paclitaxel was detrimental to T-cell proliferation and viability, and paclitaxel-treated T cells did not survive beyond D6 regardless of the duration and timing of exposure (Fig. 6C). Flow cytometry analysis during T-cell expansion revealed that exposure to pemetrexed enhanced the activation state of T cells, as reflected by significantly increased and sustained surface expression of CD137 and GITR on CD8⁺and CD4⁺T cells with continuous pemetrexed exposure (Fig. 6D). The enhanced T-cell activation state was also accompanied by significantly increased mitochondrial content in CD8⁺and CD4⁺T cells (Fig. 6E). Finally, QGP gene expression analysis revealed pemetrexed-dependent upregulation of IFNγ-dependent transcripts (IFNG, CXCL9, CXCL10, CXCL11, IDO1, HLA-DRA), cytolytic genes (GZMB, PRF1) as well as transcripts encoding costimulatory receptors CD137, GITR, OX40 (TNFRSF9, TNFRSF18, TNFRSF4; Fig. 6F), and these results were further confirmed using nCounter analysis (Supplementary Fig. S5).

To assess if the increased metabolic fitness and activation state of T cells translate to enhanced effector function, we measured antigen-dependent tumor cell killing using ovalbumin (OVA)-specific OT-1 T cells. OT-1 T cells were primed with OVA peptide (SIINFEKL) in the presence or absence of pemetrexed, and their ability to kill tumor targets loaded with OVA peptide was evaluated in vitro. These results demonstrate that priming in the presence of pemetrexed resulted in ∼50% increase in tumor cell killing (∼50% and ∼30% dead tumor cells with pemetrexed and control, respectively). These data therefore suggest that the enhanced activation state induced by pemetrexed translates into increased effector T-cell function exemplified by increased cytotoxicity of antigen-specific T cells (Fig. 6G).

Taken together, these data reveal that pemetrexed exerts pleiotropic immunomodulatory effects by inducing ICD in tumor cells, enhancing the metabolic state of T cells by increasing their oxidative respiration and the mitochondrial content, leading to increased activation and effector function.

Chemotherapeutic agents are part of standard of care treatment in many tumor types and across lines of therapy. A fundamental premise for combining chemo- and immunotherapies is that the chemotherapeutic agents preferentially target tumor cells and do not incapacitate relevant immune functions. The potential for enhanced combinatorial activity of chemotherapy and immunotherapy is based on the principle that in some cases chemotherapeutic agents may cause immunogenic tumor cell death, resulting in immune enhancing activities through immune cells, and/or the tumor microenvironment without incapacitating relevant immune cell function (4, 5, 11).

It is generally believed that folate pathway inhibitors such as methotrexate are immunosuppressive (12, 13). Although the cytotoxic activity of pemetrexed has been attributed to inhibition of 4 enzymes in the folate cycle (24), very little is known about how pemetrexed modulates antitumor immunity. Our data indicate that pemetrexed exerts immunomodulatory effects across multiple pathways and immune cell subsets. These effects were not observed with other chemotherapeutic agents such as carboplatin or paclitaxel. Furthermore, our data indicate that carboplatin and cisplatin attenuate, rather than enhance the immunomodulatory effects of pemetrexed, and suggest that pemetrexed can potentially be combined with ICIs without platinum agents. This finding might be critical, and additional mechanistic, translational, and clinical studies are needed to further understand the effects of various platinum doublets on tumor immune microenvironment as well as their combinatorial potential with ICIs.

One hypothesis supporting platinum-based chemotherapy in combination with pemetrexed and immune checkpoint blockade is that platinum agents could induce somatic mutations in the tumor cells and, consequently, induce new immunogenic neoantigens. It is therefore possible that adding a platinum agent before pemetrexed/anti-PD(L)1 treatment could potentially enhance the immunomodulatory and antitumor effects through increased priming against these de novo induced neoantigens. The accumulation of somatic mutations in tumor cell DNA requires some time, and syngeneic mouse tumor models have very limited time window that makes them poorly fit for studying effects of cytotoxic agents on tumor mutational burden. It would be worthwhile to test this hypothesis in the clinical setting given that patients with advanced/metastatic NSCLC typically receive front-line chemotherapy every 3 weeks up to 6 cycles.

Gene expression profiling indicated that treatment with pemetrexed ± anti-PD-L1 induced macrophage reprograming, DC/NK cell enrichment, and enhanced inflammatory response, activated innate immune mechanisms (PRR/TLR signaling), granulocyte/agranulocyte diapedesis, and IFN signaling that play an important role in priming and establishing an efficient T-cell immunity. Additionally, the benefit from the combination treatment was abrogated when T-cell egress from the lymph nodes was blocked by the S1P1R antagonist. These results strongly support the hypothesis that pemetrexed induces an integrated antitumor immune response by enhancing antigen presentation and T-cell priming in tumor-draining lymph nodes.

The gene expression analyses also revealed that pemetrexed monotherapy was accompanied by activation of vascular endothelium and genes associated with tertiary lymphoid structure formation; this effect was even more evident when pemetrexed was combined with anti-PD-L1 therapy. It is plausible that the T-cell inflamed phenotype observed in pemetrexed-treated tumors might be attributable, at least in part, to enhancement of these pathways, which have the potential to promote T-cell trafficking and infiltration.

The high-content analyses suggested and in vitro data revealed that the immunomodulatory effects of pemetrexed also included direct effects on tumor cells via induction of ICD exemplified by the extracellular release of HMGB1 and CRT, in a manner superior to other ICD-inducing chemotherapies. Our data suggest that the immunogenic effects of pemetrexed on tumor cells exemplified by CRT and HMGB1 release may require lower drug concentrations compared with the cytotoxic effects. These results may explain the more robust gene expression changes observed in MC38 tumors after treatment with lower doses (50 mg/kg) of pemetrexed, and suggest that part of the mechanism of action for pemetrexed might involve increased tumor immunogenicity followed by the priming and establishment of an antitumor T-cell response.

It is worth noting that although carboplatin also demonstrated evidence of ICD in mouse tumor cell lines in vitro, these results did not translate in vivo as exemplified by the gene expression data in MC38 tumors. A potential explanation of this discrepancy could be due to different effects of carboplatin on tumor versus immune cells. It is also important to highlight that in MC38 tumors, treatment with all chemotherapeutic agents tested was accompanied by PD-L1 (encoded by Cd274) upregulation highlighting the need for PD1 pathway blockade to overcome adaptive resistance in T-cell compartment induced by chemotherapy.

To our knowledge, a positive effect of anti-folates in general and pemetrexed in particular on T-cell biology has not been described previously. Although the molecular mechanisms for the T-cell-intrinsic effects have not been fully elucidated, because the folate pathway (which regulates purine nucleotide synthesis and the TCA cycle) depends on 3-phosphoglycerophosphate (3-PG) generated through glycolysis (37), it is possible that inhibition of the folate cycle may increase the abundance of 3-PG and downstream metabolic intermediates required for optimal T-cell activation (37, 38). The enhanced metabolic fitness of T cells exposed to pemetrexed is notable as an immune-enhancing mechanism because activated T cells require adequate mitochondrial mass to support bioenergetic needs required for cytokine production and development of cytotoxic effector function. Indeed, multiple lines of evidence link T-cell metabolic state with activation, survival, and intratumoral exhaustion (35), and the association between metabolic fitness, activation phenotype, and effector function of tumor-reactive T cells has also been demonstrated in the present study. The T-cell-intrinsic effects described here can also potentially be attributable to the unique ability of pemetrexed, relative to other anti-folates, to also inhibit 5-aminoimidazole-4-carboxamide ribonucleotide formyltransferase (AICARFT), because blockade of this enzyme results in the elevated intracellular levels of ZMP, a metabolite structurally related to AMP that is capable of promoting mitochondrial biogenesis and respiration function via AMP kinase-mediated mechanisms (39–41). Emerging data indicate that epigenetic and/or metabolic mechanisms rather than immunosuppressive tumor microenvironment play a dominant role in driving intratumoral T-cell dysfunction (42–45). The results of this study, particularly with regard to the ability of pemetrexed to enhance metabolic fitness and effector function of T cells have an important biological and translational significance given limited therapeutic options to revert metabolically exhausted T cells in the tumor.

Collectively, the data from these studies suggest that pemetrexed therapy has the potential to induce an integrated antitumor immune response in tumors. These observations provide mechanistic rationale for the clinically observed combination activity between pemetrexed and anti-PD-1 therapy, identify pathways and mechanisms to be explored in translational studies and highlight the potential for pemetrexed as an important therapeutic modality to be investigated further in the context of combination immunotherapies. Finally, these studies provide context and direction for the exploration of the immunotherapeutic potential for other tumor-targeting agents currently being used or contemplated for use in the clinic.

D.A. Schaer, S. Geeganage, E.R. Rasmussen, K. Chodavarapu, J.R. Manro, G.P. Donoho, and M. Kalos hold ownership interest (including patents) in Eli Lilly and Company. R. Novosiadly holds ownership interest (including patents) in Eli Lilly and Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D.A. Schaer, S. Geeganage, F.C. Dorsey, D. Surguladze, M. Kalos, R.D. Novosiadly

Development of methodology: D.A. Schaer, N. Amaladas, Z.H. Lu, E.R. Rasmussen, Y. Li, S. Luo, C. Carpenito, G.E. Hall, M. Kalos, R.D. Novosiadly

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Amaladas, Z.H. Lu, A. Sonyi, D. Chin, A. Capen, Y. Li, C.M. Meyer, B.D. Jones, S. Luo, A. Nikolayev, B. Tan, F.C. Dorsey, G.P. Donoho, D. Surguladze, G.E. Hall

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.A. Schaer, S. Geeganage, N. Amaladas, Z.H. Lu, E.R. Rasmussen, A. Sonyi, D. Chin, A. Capen, Y. Li, C. Carpenito, K.D. Roth, A. Nikolayev, K. Chodavarapu, J.R. Manro, T.N. Doman, D. Surguladze, G.E. Hall, M. Kalos, R.D. Novosiadly

Writing, review, and/or revision of the manuscript: D.A. Schaer, S. Geeganage, N. Amaladas, E.R. Rasmussen, D. Chin, C. Carpenito, M. Brahmachary, F.C. Dorsey, J.R. Manro, G.P. Donoho, M. Kalos, R.D. Novosiadly

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Amaladas, X. Huang, J.R. Manro, T.N. Doman, D. Surguladze, G.E. Hall, M. Kalos

Study supervision: D.A. Schaer, S. Geeganage, A. Capen, G.P. Donoho, D. Surguladze, M. Kalos, R.D. Novosiadly

We thank Gregory D. Plowman, Levi Garraway, Ana Oton, and Jong Seok Kim (Eli Lilly) for review and helpful discussions during preparation of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.