Pemetrexed Enhances Membrane PD-L1 Expression and Potentiates T Cell-Mediated Cytotoxicity by Anti-PD-L1 Antibody Therapy in Non-Small-Cell Lung Cancer

Immunotherapy has significantly changed the treatment landscape for advanced non-small-cell lung cancer (NSCLC) with the introduction of drugs targeting programmed cell death protein-1 (PD-1) and programmed cell death ligand-1 (PD-L1). In particular, the addition of the anti-PD-1 antibody pembrolizumab to platinum-pemetrexed chemotherapy resulted in a significantly improved overall survival in patients with non-squamous NSCLC, regardless of PD-L1 expression. In this preclinical study, we investigated whether chemotherapy can modulate PD-L1 expression in non-squamous NSCLC cell lines, thus potentially affecting immunotherapy efficacy. Among different chemotherapeutic agents tested, only pemetrexed increased PD-L1 levels by activating both mTOR/P70S6K and STAT3 pathways. Moreover, it also induced the secretion of cytokines, such as IFN-γ and IL-2, by activated peripheral blood mononuclear cells PBMCs that further stimulated the expression of PD-L1 on tumor cells, as demonstrated in a co-culture system. The anti-PD-1/PD-L1 therapy enhanced T cell-mediated cytotoxicity of NSCLC cells treated with pemetrexed and expressing high levels of PD-L1 in comparison with untreated cells. These data may explain the positive results obtained with pemetrexed-based chemotherapy combined with pembrolizumab in PD-L1-negative NSCLC and can support pemetrexed as one of the preferable chemotherapy partners for immunochemotherapy combination regimens.

The surface expression of PD-L1 in A549 and H322 cells ( Figure S1) was then evaluated upon treatment with IFN-γ, a key cytokine known to upregulate PD-L1 at the membrane level [19]. As expected, the addition of IFN-γ significantly increased PD-L1 in both cell lines, indicating the presence of a functional Jak/STAT signaling pathway [20], known to control PD-L1 expression. Then, we tested the effect of different chemotherapeutic drugs (the antimetabolites, gemcitabine and pemetrexed; the microtubule-targeting agents, paclitaxel and vinorelbine; and the DNA-damaging agent, cisplatin) on mPD-L1 expression. A549, Calu-6, H292, and H322 cells were treated with each compound at the corresponding IC 50 value indicated in Figure 1B. As reported in Figure 1C, pemetrexed caused a marked increase in mPD-L1 expression in all the tested cell lines. In H292 cells, cisplatin increased mPD-L1 expression. Pemetrexed-mediated induction of mPD-L1 expression was confirmed by confocal microscopy in H322 cells, as shown in Figure 1D.
We then evaluated the effect of pemetrexed at different exposure times and concentrations on PD-L1 expression in A549 cells. Pemetrexed at 100 nM increased PD-L1 mRNA level ( Figure 2A) and protein expression ( Figure 2B) in a time-dependent manner with the highest levels of PD-L1 protein detected at 72 h. At this time, we evaluated the effect of increasing concentrations of the drug on PD-L1 induction, demonstrating that PD-L1 level started to increase at 100 nM with the maximum expression observed at 500-1000 nM ( Figure 2C). Pemetrexed at 500 nM enhanced PD-L1 level after 24 h ( Figure 2D and Figure S2).
With the aim to evaluate whether the induced PD-L1 expression was also maintained after pemetrexed removal, A549 cells were treated for 24 h with 500 nM of the drug and then incubated in drug-free medium for up to 48 h. Total ( Figure 2D) and membrane ( Figure 2E) PD-L1 expression were unchanged for 48 h after pemetrexed removal, maintaining levels comparable to those of cells continuously exposed to pemetrexed for 48 h, despite the significant decrease in PD-L1 mRNA expression ( Figure 2F). These results suggest that, upon induction, PD-L1 is a stable protein.
With the aim to evaluate whether the induced PD-L1 expression was also maintained after pemetrexed removal, A549 cells were treated for 24 h with 500 nM of the drug and then incubated in drug-free medium for up to 48 h. Total ( Figure 2D) and membrane ( Figure 2E) PD-L1 expression were unchanged for 48 h after pemetrexed removal, maintaining levels comparable to those of cells continuously exposed to pemetrexed for 48 h, despite the significant decrease in PD-L1 mRNA expression ( Figure 2F). These results suggest that, upon induction, PD-L1 is a stable protein. cells were continuously exposed to 500 nM pemetrexed for the indicated period of time or treated for 24 h and, after drug removal, the cells were incubated with fresh medium for 24 h or 48 h. At the indicated times, total PD-L1 protein, membrane PD-L1 protein, and PD-L1 mRNA were quantified by western blotting (D), flow cytometry (E), and RT-PCR (F), respectively. * p < 0.05; ** p < 0.01; *** p < 0.001. Data in (A), (E), and (F) are mean values ± SD of three independent experiments. Results in (B-D) are representative of three independent experiments.

Pemetrexed-Induced PD-L1 Expression Is Mediated by mTOR and STAT3 Signaling
The regulation of PD-L1 expression is very complex, varying among different tumor types and including both transcriptional and post-transcriptional mechanisms of control [8]. To unravel the mechanisms of pemetrexed-mediated upregulation of PD-L1, we compared the effects of IFN-γ and pemetrexed on intracellular signaling pathways. Results showed that IFN-γ or pemetrexed caused an increase in the phosphorylation of mTOR, P70S6K, and STAT3 proteins in A549 and H322 cells ( Figure 3A,B).
To define the timing of PD-L1 induction with respect to cell signaling modulation, we performed a time-course experiment demonstrating that both mTOR and STAT3 signaling were activated before the induction of PD-L1, at 6 h and 16 h of treatment for mTOR and STAT3 activation, respectively ( Figure S2).
To clarify the role of these signaling cascades on PD-L1 upregulation, treatment with specific inhibitors was performed ( Figure 3C). Pre-incubation of A549 and H322 cells with the mTOR inhibitor RAD001 reduced PD-L1 induction by pemetrexed; a more evident reduction was observed after pre-incubation with the STAT3 inhibitor C188-9. Interestingly, when the cells were pre-treated with both compounds, PD-L1 expression was further downregulated, suggesting that both pathways control PD-L1 expression induced by pemetrexed. To clarify the role of these signaling cascades on PD-L1 upregulation, treatment with specific inhibitors was performed ( Figure 3C). Pre-incubation of A549 and H322 cells with the mTOR inhibitor RAD001 reduced PD-L1 induction by pemetrexed; a more evident reduction was observed after pre-incubation with the STAT3 inhibitor C188-9. Interestingly, when the cells were pre-treated with both compounds, PD-L1 expression was further downregulated, suggesting that both pathways control PD-L1 expression induced by pemetrexed.

Pemetrexed Induces PD-L1 Expression through Stimulation of IFN-γ Production by T Lymphocytes
Since chemotherapeutic drugs interact not only with cancer cells, but also with immune cells in the tumor microenvironment, we tested the effects of pemetrexed on cancer cells co-cultured with activated PBMCs in a non-contacting transwell system ( Figure S3A).
A549 and H322 cells were co-cultured with primary human T-cells stimulated with CD3/CD28, in the absence or presence of pemetrexed. As reported in Figure 4A, a significant increase of mPD-L1 was observed when activated lymphocytes were added to the upper chamber of the transwell. The addition of pemetrexed to the co-culture system promoted a further increase in mPD-L1 level.
We then sought to elucidate the mechanisms underneath the effects of pemetrexed treatment on T-cell activation. Pemetrexed strongly enhanced the mRNA expression of IFN-γ and IL-2, and downregulated TGF-β mRNA levels ( Figure 4B) in activated T-cells.
We further investigated pemetrexed-induced cytokine production by quiescent and anti-CD3/CD28-activated T cells in vitro. Treatment with pemetrexed enhanced IFN-γ, TNF-α, and IL-2 production in both quiescent and activated cells after 24 h of treatment ( Figure 5). Of note, the same significant trend was evident for CD3-, CD4-, and CD8-positive T cells ( Figure 5A). However, after 48 h of pemetrexed treatment, a significant increase in cytokine production was observed only in anti-CD3/CD28-stimulated T cells ( Figure 5B). In addition, the cytotoxic potential was augmented after 24 h of pemetrexed treatment as compared to 48 h ( Figure 5C), predominantly for CD8-positive T cells. Then, we investigated the effect of paclitaxel and cisplatin on T-cell responses in quiescent and stimulated cells, and demonstrated that neither of these drugs affected cytokine production and the cytotoxic potential ( Figures S5A,B and S6A,B). Of note, all the chemotherapeutic treatments induced only a slight increase in cell death in lymphocytes ( Figure S4). in the absence or presence of pemetrexed. As reported in Figure 4A, a significant increase of mPD-L1 was observed when activated lymphocytes were added to the upper chamber of the transwell. The addition of pemetrexed to the co-culture system promoted a further increase in mPD-L1 level.
We then sought to elucidate the mechanisms underneath the effects of pemetrexed treatment on T-cell activation. Pemetrexed strongly enhanced the mRNA expression of IFN-γ and IL-2, and downregulated TGF-β mRNA levels ( Figure 4B) in activated T-cells. We further investigated pemetrexed-induced cytokine production by quiescent and anti-CD3/CD28-activated T cells in vitro. Treatment with pemetrexed enhanced IFN-γ, TNF-α, and IL-2 production in both quiescent and activated cells after 24 h of treatment ( Figure 5). Of note, the same significant trend was evident for CD3-, CD4-, and CD8-positive T cells ( Figure 5A). However, after 48 h of pemetrexed treatment, a significant increase in cytokine production was observed only in anti-CD3/CD28-stimulated T cells ( Figure 5B). In addition, the cytotoxic potential was augmented after 24 h of pemetrexed treatment as compared to 48 h ( Figure 5C), predominantly for CD8-positive T cells. Then, we investigated the effect of paclitaxel and cisplatin on T-cell responses in quiescent and stimulated cells, and demonstrated that neither of these drugs affected cytokine production and the cytotoxic potential ( Figures S5A,B and S6A,B). Of note, all the chemotherapeutic treatments induced only a slight increase in cell death in lymphocytes ( Figure S4). In conclusion, these findings suggest that pemetrexed can affect the expression of PD-L1 on cancer cells either directly, by activating signaling pathways controlling PD-L1 transcription, or indirectly, through the release of IFN-γ by immune cells in the tumor microenvironment.   In conclusion, these findings suggest that pemetrexed can affect the expression of PD-L1 on cancer cells either directly, by activating signaling pathways controlling PD-L1 transcription, or indirectly, through the release of IFN-γ by immune cells in the tumor microenvironment.

Effect of Pemetrexed Treatment on the Release of Soluble Form of PD-L1
Besides the membrane-bound form, PD-L1 has a soluble form (sPD-L1) that is released in the extracellular milieu. Previous data indicated that cancer cell lines with high mPD-L1 present elevated PD-L1 protein levels in the supernatant [21]. It is also well known that cytokines, such as IFN-γ, can increase the release of sPD-L1 [22]. We then evaluated if sPD-L1 can be released in response to pemetrexed treatment. Pemetrexed, similar to IFN-γ, not only increased mPD-L1, but also induced the release and accumulation of sPD-L1 in the medium from both A549 and H322 cells ( Figure 6A).
increase the release of sPD-L1 [22]. We then evaluated if sPD-L1 can be released in response to pemetrexed treatment. Pemetrexed, similar to IFN-γ, not only increased mPD-L1, but also induced the release and accumulation of sPD-L1 in the medium from both A549 and H322 cells ( Figure 6A).
To evaluate whether our preclinical data could be translated into a clinical setting, we assessed sPD-L1 levels in plasma samples from eight NSCLC patients, collected 4 days after pemetrexed treatment. As reported in Figure 6B, this drug significantly increased sPD-L1 levels in patients.

The Addition of the Anti PD-L1 Antibody Atezolizumab to Pemetrexed Significantly Potentiates T cell-Mediated Cytotoxicity in NSCLC Cells
Finally, we investigated in a co-culture assay, the anti-tumor immune activity of atezolizumab (anti PD-L1) with or without pemetrexed in the presence of activated T-cells.
For this purpose, A549 cells were co-cultured in a transwell system (non-contacting co-culture) with activated PBMCs in the presence or absence of 100 nM pemetrexed for 3 days ( Figure S3A). Then, activated PBMCs were transferred from the insert to the corresponding well ( Figure S3B) to be in contact with A549 cells ( Figure S3C). The contacting co-culture was then treated for 24 h with or without 200 μg/mL atezolizumab. Phase contrast microscopy images ( Figure 7A) and crystal violet assay ( Figure 7B) showed that atezolizumab significantly promoted T cell-mediated cytotoxicity of To evaluate whether our preclinical data could be translated into a clinical setting, we assessed sPD-L1 levels in plasma samples from eight NSCLC patients, collected 4 days after pemetrexed treatment. As reported in Figure 6B, this drug significantly increased sPD-L1 levels in patients.

The Addition of the Anti PD-L1 Antibody Atezolizumab to Pemetrexed Significantly Potentiates T cell-Mediated Cytotoxicity in NSCLC Cells
Finally, we investigated in a co-culture assay, the anti-tumor immune activity of atezolizumab (anti PD-L1) with or without pemetrexed in the presence of activated T-cells.
For this purpose, A549 cells were co-cultured in a transwell system (non-contacting co-culture) with activated PBMCs in the presence or absence of 100 nM pemetrexed for 3 days ( Figure S3A). Then, activated PBMCs were transferred from the insert to the corresponding well ( Figure S3B) to be in contact with A549 cells ( Figure S3C). The contacting co-culture was then treated for 24 h with or without 200 µg/mL atezolizumab. Phase contrast microscopy images ( Figure 7A) and crystal violet assay ( Figure 7B) showed that atezolizumab significantly promoted T cell-mediated cytotoxicity of A549 cells treated with pemetrexed. These data suggest that PD-L1 expression induced by pemetrexed is important for atezolizumab to enhance T cell-mediated cytotoxicity in NSCLC cells.

Discussion
In this work, we demonstrated that pemetrexed has the ability to stimulate PD-L1 expression SCLC cells, either as a membrane-bound protein or as a released soluble form. Moreove metrexed caused a marked increase in the production of pro-inflammatory cytokines, includin N-γ, by T cells, which further enhanced PD-L1 expression on cancer cells. This increased lev rongly enhanced T cell-mediated cytotoxicity induced by anti-PD-L1 therapy, as evaluated in c lture systems.
Since the approval of the anti-PD-1 monoclonal antibodies nivolumab and pembrolizumab an e anti-PD-L1 inhibitor atezolizumab for the treatment of advanced NSCLC, new strategies hav en developed in order to further improve the efficacy of these therapies. Recently, the phase I inical trial Keynote-189 [4] showed that pembrolizumab combined with the pemetrexed-platinu andard chemotherapy regimen significantly increased overall survival in advanced non-squamou SCLC patients with different levels of PD-L1 expression. Interestingly, the results of this clinic ial showed that immunotherapy was able to improve chemotherapy outcome regardless of baselin -L1 expression. In particular, a subgroup analysis showed no correlation between PD-L pression and the amount of benefit derived from the addition of pembrolizumab to platinum metrexed chemotherapy and, surprisingly, even patients with null PD-L1 expression received atistically and clinically relevant benefit from the combined treatment as those with different leve PD-L1 expression. One of the likely explanations for this unexpected clinical finding may be relate ith the possible modulation of PD-L1 during chemotherapy treatment. Very recently, it has been reported [7] that the PD-L1 tumor proportion score was much high re-biopsy tissue samples when compared with the initial biopsy in some NSCLC patients aft

Discussion
In this work, we demonstrated that pemetrexed has the ability to stimulate PD-L1 expression in NSCLC cells, either as a membrane-bound protein or as a released soluble form. Moreover, pemetrexed caused a marked increase in the production of pro-inflammatory cytokines, including IFN-γ, by T cells, which further enhanced PD-L1 expression on cancer cells. This increased level strongly enhanced T cell-mediated cytotoxicity induced by anti-PD-L1 therapy, as evaluated in co-culture systems.
Since the approval of the anti-PD-1 monoclonal antibodies nivolumab and pembrolizumab and the anti-PD-L1 inhibitor atezolizumab for the treatment of advanced NSCLC, new strategies have been developed in order to further improve the efficacy of these therapies. Recently, the phase III clinical trial Keynote-189 [4] showed that pembrolizumab combined with the pemetrexed-platinum standard chemotherapy regimen significantly increased overall survival in advanced non-squamous NSCLC patients with different levels of PD-L1 expression. Interestingly, the results of this clinical trial showed that immunotherapy was able to improve chemotherapy outcome regardless of baseline PD-L1 expression. In particular, a subgroup analysis showed no correlation between PD-L1 expression and the amount of benefit derived from the addition of pembrolizumab to platinum-pemetrexed chemotherapy and, surprisingly, even patients with null PD-L1 expression received a statistically and clinically relevant benefit from the combined treatment as those with different levels of PD-L1 expression. One of the likely explanations for this unexpected clinical finding may be related with the possible modulation of PD-L1 during chemotherapy treatment.
Very recently, it has been reported [7] that the PD-L1 tumor proportion score was much higher in re-biopsy tissue samples when compared with the initial biopsy in some NSCLC patients after chemotherapeutic treatment. A clinical trial (NCT03701607) aiming to evaluate whether PD-L1 may change after platinum-based chemotherapy in advanced NSCLC is presently active for patient recruitment.
In our study performed in a panel of NSCLC adenocarcinoma cell lines, wild-type for EGFR and ALK, we demonstrated that PD-L1 levels were enhanced after treatment with pemetrexed, but not with other chemotherapeutic agents, as a consequence of the activation of mTOR/P70S6K and STAT3 intracellular signaling pathways. Although the ability of pemetrexed to induce mTOR activation has already been reported [23], here we demonstrated that pemetrexed also activated STAT3, the main transcription factor responsible for PD-L1 activation mediated by IFN-γ [24]. According to this observation, pharmacological inhibition of both the signaling pathways with RAD001 and C188-9 reduced PD-L1 levels in cancer cells. Pemetrexed exposure also caused an increase in the soluble form of PD-L1. Recent data described discordant results on the predictive and prognostic significance of sPD-L1 in lung cancer patients [25], and the ability of sPD-L1 to interact with its ligand PD-1 on T cells is still on debate. Nevertheless, PD-L1 plasma levels were significantly enhanced in NSCLC patients receiving standard first-line chemotherapy [8] and our data indicated that pemetrexed, as a single regimen, can increase sPD-L1 in plasma patients.
In activated PBMCs, pemetrexed induced a marked increase in the expression and release of the cytokines IFN-γ and IL-2, and a decrease of the anti-inflammatory cytokine TGF-β, a sign of the acquisition of a pro-inflamed phenotype. Similar results were observed in patients with adenocarcinoma of the pancreas treated with the anti-folate compound; the authors reported an increase of IFN-γ and IL-2 secretion by natural killer (NK) cells during the treatment, enforcing the role of pemetrexed as a drug that can alter cytokine production in cancer patients [26].
The secretion of IFN-γ by T lymphocytes increased the expression of PD-L1 on cancer cells, thereby enhancing the efficacy of the immune checkpoint inhibitor atezolizumab, as demonstrated in the contacting co-culture assay, thus suggesting that a high PD-L1 level can be considered as a "driver" for T cell-mediated anti-tumor activity of immune checkpoint inhibitors.
These results are in agreement with recent findings showing that pemetrexed treatment in murine syngeneic colon tumor models (MC38 and Colon 26) activated T cells, inducing an inflamed phenotype, and that the addiction of an anti PD-L1 antibody improved the anti-tumor properties of the anti-folate compound [27].
Most chemotherapeutic drugs have been shown to have a detrimental effect on CD8+ T-cells in lymph nodes and blood; however, the same effect has not been detected on tumor-infiltrating lymphocytes [28]. Conventional chemotherapy can instead increase anti-tumor responses through the induction of immunogenic cell death by depleting immunosuppressive immune cell subsets, such as myeloid-derived suppressor cells or regulatory T-cells, or by exerting direct stimulatory effects on immune cells by cytokine production [29,30]. For example, cisplatin and paclitaxel treatment has been reported to generate a marked CD8+ T-cell response with high secretion of IFN-γ and IL-2 in patients with ovarian cancer [31]; 5-FU or gemcitabine selectively killed myeloid-derived suppressor cells [32] and in patients with esophageal squamous cell carcinoma, adjuvant chemotherapy with cisplatin and 5-FU increased the trafficking of CD4+ and CD8+ cells in the tumor microenvironment [33].
In conclusion, we propose that pemetrexed, by promoting PD-L1 upregulation in tumor cells and by inducing the secretion of cytokines by T cells, may generate, in the tumor microenvironment, a favorable condition for the efficacy of immune checkpoint inhibitors, independently of PD-L1 tumor cell expression at baseline.

Cells and Cell Culture
Human NSCLC A549, Calu-6, H292, and H322 cell lines were purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA) and cultured in Roswell Park Memorial Institute medium (RPMI 1640) supplemented with 2 mM glutamine, 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 µg/mL streptomycin. The cell lines were incubated at 37 • C in a humidified atmosphere containing 5% CO 2 .

Flow Cytometry
For the determination of membrane levels of PD-L1, cells were treated for 3 days with chemotherapeutic drugs, collected, and incubated with phycoerythrin (PE) isotype control mouse IgG1 κ or PE anti-human PD-L1 (BD Biosciences, San Jose, CA, USA). After incubation, quantification was performed with a Beckman FC500 flow cytometer as previously described [34] and analyzed with FCS express software (De Novo software, Pasadena, CA, USA). The values of mean fluorescence intensity (MFI) were converted into units of equivalent fluorochrome (MEF) using the FluoroSpheres 6-Peak kit (Dako, Santa Clara, CA, USA).

Cell Proliferation Assay
Cell proliferation was evaluated by cell counting, the MTT assay, and crystal violet staining as previously described [35].

Confocal Microscopy
Cells were grown on poly-L-lysine-coated glass slides and treated with 100 nM pemetrexed. After 72 h, the cells were fixed with 4% paraformaldehyde and permeabilized with 0.2% Triton X-100, and unspecific epitopes were blocked with 2% bovine serum albumin (BSA). Then, for PD-L1 staining, the cells were incubated overnight at 4 • C with anti-PD-L1 (Abcam, ab205921) and Alexa Fluor 546-conjugated donkey anti-rabbit IgG (Invitrogen, Carlsband, CA, USA) was used as the secondary antibody. Nuclei were stained with Draq5 (Biostatus, Shepshed, Leicestershire, UK). Samples were observed using a confocal system (LSM 510 Meta, scan head integrated with the Axiovert 200 M inverted microscope; Carl Zeiss, Jena, Germany) through a 40×, NA 1.3, oil immersion objective. Alexa Fluor 546 and Draq5 were excited with 543 nm and 633 nm He-Ne laser lines, respectively. Image acquisition was carried out in a multitrack mode (through consecutive and independent optical pathways) with relevant beam splitters; barriers filters were 560-615 band pass and 650 long pass filters for the above signals, respectively.

Soluble PD-L1 Quantification
To assay soluble PD-L1 (sPD-L1), cells were seeded in a cell-well culture plate. After 24 h, pemetrexed or IFN-γ was added for 3 days; at the end, the media were collected and tested for soluble PD-L1 expression by ELISA assay (R&D System, Minneapolis, MN, USA).

Co-Culture System
A non-contacting co-culture system of cancer cells with PBMCs was established using a transwell suspension culture chamber with a polyethylene terephthalate film (PET) (Corning, NY, USA). Briefly, 1 × 10 5 cells were seeded in a six-well plate. At the same time, freshly isolated PBMCs were activated with CD3/CD28. The day after, 1 × 10 6 activated PBMCs were added to the transwell in the presence or absence of pemetrexed and after 24 h, PBMCs were removed and PD-L1 membrane levels on cancer cells were analyzed by a flow cytometer.
For T cell-mediated cytotoxicity [39,40], after 3 days of incubation in the previous reported transwell system, PBMCs were transferred from the insert to the corresponding well in a contacting co-culture with A549 cells and maintained for 24 h in the absence or presence of 200 µg/mL atezolizumab. Finally, PBMCs and dead tumor cells were washed away and the crystal violet assay was performed. Images of viable cells were obtained with a Nikon Eclipse E400 microscope with a digital net camera.

Patients
Plasma samples were obtained from NSCLC patients enrolled within a phase Ib trial of pemetrexed-enzastaurin combination therapy, as described previously [41]. All patients received standard pemetrexed as a 10 min intravenous dose of 500 mg/m 2 . All these patients also received oral folic acid (350-1000 µg daily) and vitamin B12 (1000 µg by intramuscular injection every 9 weeks), beginning 5-7 days before the first dose of pemetrexed. An analysis of soluble PD-L1 expression was performed by ELISA assay (R&D System, Minneapolis, MN, USA).

Statistical Analysis
Statistical analyses were carried out using the GraphPad Prism 6.00 software. The normality distribution of data was tested by the Kolmogorov-Smirnov test. Comparisons were performed by the two-tailed Wilcoxon matched-pairs test, two-tailed Student's t-test, or one-way ANOVA followed by Bonferroni's post-test, depending on the assay; and p-values are indicated where appropriate (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001).

Conclusions
Our findings provide a biological rationale to explain the superior efficacy of anti-PD1/PD-L1 therapy added to platinum-pemetrexed chemotherapy in patients with PD-L1-negative NSCLC and support pemetrexed as an ideal drug partner for chemoimmunotherapy combination regimens in advanced non-squamous NSCLC treatment. The optimal pemetrexed scheduling along with immune checkpoint inhibitors, concurrent vs sequential, in order to take maximum advantage of the interaction between chemotherapy and PD-L1 expression, remains to be elucidated and deserves further pre-clinical and clinical research.