Cytotoxic T Cells in PD-L1–Positive Malignant Pleural Mesotheliomas Are Counterbalanced by Distinct Immunosuppressive Factors

The use of immune checkpoint inhibitors to engage the immune system in the fight against cancer has revolutionized the management of an increasing number of tumor types in recent years and offers the opportunity of achieving durable responses for patients (1–5). In certain cancer types, such as non–small cell lung cancer, renal cell carcinoma, and bladder cancer, only a minority of patients respond to immune checkpoint inhibitors, and because these drugs can be associated with significant immune-related toxicities, identification of reliable predictive biomarkers for these treatments is imperative. In some cancers, expression of PD-L1 appears to correlate, in part, with response to PD-1 or PD-L1 inhibitors; however, some PD-L1–positive tumors do not respond to these agents, and, in contrast, some PD-L1–negative tumors do respond to these drugs (6–8).

PD-L1 expression is not likely the sole determinant of tumor immune evasion, and the efficacy of PD-1 pathway blockade may be impacted by the expression of other inhibitory regulators of the immune response, such as cytotoxic T lymphocyte–associated antigen 4 (CTLA-4), T-cell immunoglobulin and mucin domain 3 (TIM-3), and lymphocyte activation gene 3 (LAG-3), among others (9–13). To identify additional biomarkers of immunotherapy response besides PD-L1 immunohistochemical expression, several techniques are currently under development to more comprehensively and quantitatively characterize the tumor immune microenvironment, at the RNA, protein, and cellular level (14–18).

Herein, we present a method for comprehensive immune profiling of solid tumors using flow cytometry, and we describe our initial experience in a cohort of 43 patients with malignant pleural mesothelioma, a cancer arising from the mesothelial surfaces of the pleural cavity. The majority of mesothelioma patients are not candidates for surgical resection, and first-line chemotherapy with platinum and pemetrexed has remained the standard of care for more than a decade (19). Beyond the first-line setting, no treatments have been associated with a survival benefit, and response rates to single-agent drugs like gemcitabine or vinorelbine are low (20, 21).

Recent data have emerged to suggest that immune checkpoint inhibition may be a promising strategy for treating mesothelioma. PD-L1 expression occurs in 20% to 40% of mesotheliomas, particularly in the sarcomatoid subtype, and may be an adverse prognostic marker for overall survival (22, 23). In a study of 29 patients with mesothelioma, treatment with the CTLA-4 inhibitor tremelimumab resulted in durable partial responses in two patients (7%; ref. 24). In the KEYNOTE-028 phase Ib trial, 6 of 25 (24%) patients with PD-L1–positive mesothelioma achieved an objective response (25), consistent with response rates in other solid tumors in which anti-PD-1 therapies have been approved by the FDA (6–8, 26–28).

We hypothesized that flow cytometric analysis of tumor immune infiltrates, when coupled with PD-L1 IHC, would more accurately reveal distinct and heterogenous immunologic phenotypes across mesothelioma tumor samples. A more detailed understanding of the immunosuppressive factors within mesothelioma tumors may uncover alternative or combination immunotherapeutic strategies for the treatment of this disease.

All patients in this study were consented to a tissue collection protocol, which was approved by the Institutional Review Board at Brigham and Women's Hospital and the Dana-Farber Cancer Institute (Boston, MA).

For patients in this study who had consented to our institutional tumor genomics protocol, next-generation sequencing (results available upon request) was performed on DNA isolated from formalin-fixed paraffin-embedded tissue using previously described methods (1, 2).

Fresh patient tumor samples, which were surgically resected, were placed in RPMI containing 10% FBS. Approximately 50 mg of each sample was fixed in 10% buffered formalin phosphate for immunohistochemical analysis. The remaining tissue was mechanically minced using scissors and subsequently incubated for 45 minutes at 37°C in dissociation buffer: RPMI containing 10% FBS, 100 U/mL collagenase type IV (Life Technologies), and 50 μg/mL DNase I (Roche). The resulting cell suspension was strained through a 70-μm nylon mesh to remove larger aggregates of tissue. Cells were treated with RBC Lysis Buffer (BioLegend) for 5 minutes at 37°C to remove red blood cells and then washed in FACS buffer: DPBS containing 2% FBS.

The nonfixed cell suspension was stained using LIVE/DEAD Fixable Yellow Dead Cell Stain Kit (Life Technologies) diluted 1:500 in FACS buffer and incubated for 10 minutes in the dark at room temperature. After two washes, the cell pellet was incubated in a 1:100 dilution of FcR blocking reagent (Miltenyi Biotec) in FACS buffer for 15 minutes in the dark on ice. After a single wash, 100 μL of cell suspension was aliquoted into multiple wells in a 96-well round-bottom plate for subsequent tagging using fluorescently conjugated antibodies. Pre-made antibody cocktails, each containing five to seven different surface-staining antibodies, were added to their respective wells and incubated for 20 minutes in the dark on ice. For wells containing a complete set of seven surface-staining antibodies, cells were fixed using a 1% buffered formalin phosphate solution in FACS buffer. Samples to be permeated to tag intracellular antigens were fixed in Fix/Perm buffer following the FOXP3 Staining Kit protocol (eBioscience). Antibodies targeting intracellular antigens were diluted 1:20 in Permeabilization buffer (eBioscience) and incubated for 30 minutes in the dark on ice. After two washes in the Permeabilization buffer, cells were resuspended in FACS buffer and stored at 4°C until flow analysis.

Single-cell suspensions were stained using mouse anti-human antibodies. Surface antibodies against CD3 (HIT3a; UCHT1), CD8 (RPA-T8), CD14 (M5E2; MphiP9), CD45 (HI30), CD56 (B159), CD279 (EH12.1) and its isotype control (MOPC-21), and HLA-DR (G46-6) were purchased from BD Biosciences. Surface antibodies against CD4 (RPA-T4), CD16 (3G8), CD19 (HIB19), CD33 (WM53), CD66b (G10F5), CD123 (6H6), and TIM-3 (F38-2E2) and its isotype control (MOPC-21) were purchased from BioLegend. The surface antibody against CD45 (2D1) was purchased from eBioscience, and the surface antibody against LAG-3 (polyclonal) and its isotype control (polyclonal) were purchased from R&D Systems. The intracellular antibody FOXP3 (236A/E7) was purchased from eBioscience. For CD3 (HIT3a), LAG-3, and its isotype control, and HLA-DR, 8 μL of antibody was added to each well. For all other surface antibodies, 2.5 μL was added to each well. Sample volume was brought up to a total of 100 μL using FACS buffer, resulting in a 1:12.5 or a 1:40 dilution, respectively. For all intracellular antibodies, 2.5 μL of antibody was added to 50 μL of FACS buffer for a 1:20 dilution.

Cells were analyzed using flow cytometry within 72 hours of fixation. Prior to analysis, cells were filtered through 35-μm nylon mesh. Samples were acquired on a BD FACSCanto II HTS cell analyzer with FACSDiva software v8.0.1 (BD Biosciences) and analyzed using FlowJo software v10.

For IHC, 4-μm thick paraffin-embedded sections were baked in a 60°C oven for one hour. Heat-induced antigen retrieval was performed using ER2 solution (pH8; Leica Biosystems) for 20 to 30 minutes.

Immunohistochemical staining of CD3 (polyclonal rabbit anti-human, Dako) and PD-L1 (monoclonal rabbit anti-human, E1L3N, Cell Signaling Technology) was performed using an automated staining system (Bond III, Leica Biosystems) following the manufacturer's protocols. CD3 immunostaining was performed with 1:250 dilution using Bond Primary Antibody Diluent (Leica) for 30 minutes at room temperature. PD-L1 immunostaining was performed with 1:200 dilution using Bond Primary Antibody Diluent (Leica) for 2 hours at room temperature. Detection and development of the primary antibody was performed using the Bond Polymer Refine Detection Kit (Leica Biosystems). Slides were counterstained with hematoxylin, dehydrated, and mounted. Positive and negative controls were included in each panel of staining for both markers.

PD-L1 expression in tumor cells was considered positive if ≥1 % of tumor cells had distinct membranous staining. The intensity (0, negative; 1, weak; 2, moderate; and 3, intense) and the percentage of positively stained tumor cells were recorded. For CD3, three representative hotspots on each slide were chosen to count the positive cells under 400× high-power field (hpf) view, and the average number was recorded. All the slides were evaluated and scored blinded to clinical data.

The unsupervised nonlinear dimension reduction method t-distribution–based stochastic nonlinear embedding (t-SNE; ref. 3) was applied to the flow cytometry data to investigate in reduced dimension space how samples from all mesothelioma tumor samples and 7 paired normal controls were located in relation to each other. t-SNE minimizes the divergence of neighborhood closeness moving from high dimensions to low dimensions. Data were first Z-score normalized and the perplexity parameter input to t-SNE was 3. The following flow cytometry parameters were used for embedding: percent CD45⁺ cells, percent CD3⁺ cells, percent CD19⁺ cells, percent CD56⁺ cells, percent CD3⁺CD56⁺ cells, percent CD33⁺ cells, percent CD66b⁺ cells, percent CD123⁺ cells, percent CD16⁺ cells, percent CD4⁺ cells, percent CD8⁺ cells, CD8:CD4 ratio, percent TIM3⁻PD1⁺CD4⁺ cells, percent TIM3⁺PD1⁺CD4⁺ cells, percent TIM3⁺ PD1⁻CD4⁺ cells, percent TIM3⁻PD1⁻CD4⁺ cells, percent LAG3-PD1⁺CD4⁺ cells, percent LAG3⁺PD1⁺CD4⁺ cells, percent LAG3⁺PD1⁻CD4⁺ cells, percent LAG3⁻PD1⁻CD4⁺ cells, percent TIM3⁺CD4⁺ cells, percent PD1⁺CD4⁺ cells, percent LAG3⁺CD4⁺ cells, percent FOXP3⁺CD4⁺ cells, percent TIM3-PD1⁺CD8⁺ cells, percent TIM3⁺PD1⁺CD8⁺ cells, percent TIM3⁺PD1⁻CD8⁺ cells, percent TIM3⁻PD1⁻CD8⁺ cells, percent LAG3⁻PD1⁺CD8⁺ cells, percent LAG3⁺PD1⁺CD8⁺ cells, percent LAG3⁺PD1-CD8⁺ cells, percent LAG3⁻PD1⁻CD8⁺ cells, percent TIM3⁺CD8⁺ cells, percent PD1⁺CD8⁺ cells, percent LAG3⁺CD8⁺ cells, percent CD14⁺CD16⁻ cells, percent CD14⁺CD16⁺ cells, and percent CD14⁺HLADR⁻ cells.

The clinical data, such as histology and immunohistochemical PD-L1 staining, are mapped to the tSNE1∼tSNE2 scatter plot to visualize the distribution of patient cases. Subsequently, k-means clustering was applied to identify the two major clusters, and Fisher exact tests and Wilcoxon rank sum tests were applied for clusters-associated statistics. Linear regression was used to identify the association between a tumor's immune marker expression and the distance from the normal cluster. Overall survival was determined using the Kaplan–Meier method.

Patients with malignant pleural mesothelioma who were undergoing surgical resection during routine clinical care at our institution were offered participation in a correlative tissue collection protocol. Clinical, pathologic, and genomic characteristics of the first 43 consecutive mesothelioma samples analyzed in this study are shown in Table 1. The histologic subtypes among the 43 cases were: 29 (67%) epithelioid, 4 (9%) sarcomatoid, and 10 (23%) biphasic. Three quarters of the samples came from male patients, and nearly all 43 of the patients were Caucasian. Of the 39 mesothelioma cases for which next-generation sequencing was completed, 20 (57%) had a BAP1 mutation and 13 (37%) had an NF2 mutation. Prior to surgical resection, 6 (14%) patients received neoadjuvant chemotherapy, and the remaining 86% of patients did not receive chemotherapy before surgery.

Sufficient tissue was available for PD-L1 IHC for 39 (91%) of the cases. We defined PD-L1 positivity using the criteria for the mesothelioma cohort on the KEYNOTE-028 pembrolizumab study: any intensity of PD-L1 membranous expression in at least 1% of tumor cells (25). Among the 39 samples, 18 (46%) cases were PD-L1 positive and 21 (54%) were PD-L1 negative. By histologic subtype, nonepithelioid (sarcomatoid and biphasic) tumors were significantly more likely (P = 0.01) to be PD-L1 positive (10/13 cases, 77%) compared with epithelioid tumors (8/26 cases, 31%).

Tumors from each of the 43 mesothelioma cases were dissociated into single-cell suspensions and analyzed by flow cytometry. A sufficient number of viable cells was available for comprehensive analysis by flow cytometry in 38 of the 43 cases (88%). Across cases, there was considerable variability in the percentage of live cells that consisted of CD45⁺ leukocytes, from 17.6% to 99.8%, and we noted that the percentage of infiltrating immune cells appeared to correlate with PD-L1 status (Fig. 1A). Of all the live cells isolated from each tumor, PD-L1–positive tumors had a significantly higher percentage of CD45⁺ immune cells (median 87.7%) than PD-L1–negative tumors (median 68.2%, P = 0.05, Fig. 1B).

When analyzed by histologic subtype, nonepithelioid (sarcomatoid or biphasic) tumors were also significantly more likely to have a higher fraction of infiltrating immune cells (median 91.4%) than epithelioid tumors (median 64.1%, P < 0.0001, Supplementary Fig. S1B). Among CD45⁺ immune cells, nonepithelioid tumors were also significantly more likely to have a higher fraction of CD3⁺ T cells than epithelioid tumors (P = 0.004, Supplementary Fig. S1). There was no significant difference in the proportion of CD45⁺ live cells in tumors from patients who received neoadjuvant chemotherapy compared with tumors from patients who received no treatment before surgery (Supplementary Fig. S2).

To determine which immune cell subtype contributed to the increase in CD45⁺ cells in PD-L1–positive tumors, we next determined the relative fraction of T cells, B cells, monocytes, granulocytes, dendritic cells (DC), and natural killer (NK) cells within each sample (Fig. 1C). Again, there was considerable variability in the immune cell composition across tumors, and we noted that the percentage of T-cell infiltration also appeared to correlate with PD-L1 status. CD3⁺ T cells were commonly identified in tumor samples (range 5.2%–81.2% of CD45⁺ cells), and there were significantly more CD3⁺ T cells in PD-L1–positive tumors than in PD-L1–negative tumors (median 30.4% vs 19.3%, P < 0.05, Fig. 1D). There was no significant difference in the number of CD19⁺ B cells in PD-L1–negative versus PD-L1–positive tumors (median ∼3%, P = 0.50), although some samples showed markedly high levels of B-cell infiltrates (up to 51.8% of CD45⁺ cells). Comparing PD-L1–positive tumors with PD-L1–negative tumors, there were no significant differences in the fraction of CD66b⁺ neutrophils (median ∼6%, P = 0.05), CD123⁺ DCs (median ∼0.4%, P = 0.82), CD3⁻ CD56⁺ NK cells (median ∼4%, P = 0.88), CD3⁺ CD56⁺ natural killer T (NKT) cells (median ∼0.6%, P = 0.27) as shown in Fig. 1D, or CD16⁺ cytotoxic NK cells (median ∼37% of CD56⁺ NK cells, P = 0.17, not shown). CD33⁺ monocytes were commonly found in most tumor samples, representing approximately 42% of CD45⁺ cells (range, 5.7%–86.1%), but there was no significant difference between PD-L1–positive and PD-L1–negative tumors (P = 0.65). Among CD33⁺ cells, there was no significant difference in the proportion of CD14^high/HLA-DR^low monocytic myeloid-derived suppressor cells (MDSC) in PD-L1–positive and -negative tumors (range 0.6%–31% of CD33⁺ cells, P = 0.47) and no differences in CD14⁺/CD16⁻ or CD14⁺/CD16⁺ monocytes by PD-L1 status (Supplementary Fig. S3).

To validate flow cytometry as an accurate tool for quantifying immune cell subsets, we also determined the number of infiltrating CD3⁺ T cells in each tumor using IHC. Similar to the flow cytometry results, PD-L1–positive mesothelioma specimens had a significantly higher number of infiltrating T cells compared with PD-L1–negative tumors (median 176 vs. 59 cells/hpf, P = 0.006, Fig. 2A). As with flow cytometric analysis, some PD-L1–negative and PD-L1–positive tumors showed a relative paucity of infiltrating CD3⁺ T cells, whereas others showed an abundance of infiltrating CD3⁺ cells (range, 3–549 cells/hpf, Fig. 2A). Representative CD3 IHC images with paired flow cytometry results from four cases are shown in Fig. 2B.

Given that both flow cytometry and IHC demonstrated that T cells were more abundant in PD-L1–positive tumors than in PD-L1–negative tumors, we next analyzed the T-cell subsets within mesothelioma samples using flow cytometry. We found no significant differences in the proportion of CD4⁺ (P = 0.26) or CD8⁺ (P = 0.11) live cells in PD-L1–negative tumors, as compared with PD-L1–positive tumors, (Fig. 3A) and no significant difference in the CD8:CD4 ratio (Fig. 3B).

We then analyzed CD4⁺ and CD8⁺ T-cell lineage subpopulations by flow cytometry within each tumor sample. For CD4⁺ T cells, we observed no significant differences between PD-L1–negative and PD-L1–positive tumors in the proportion of CD45RA⁺CCR7⁺-naïve (P = 0.88), CD45RO⁺CCR7⁺ central memory (P = 0.16), CD45RO⁺CCR7⁻ effector memory (P = 0.88), or CD45RO⁺ total memory (P = 0.54) T cells, but there were significantly fewer CD45RA⁺CCR7⁻ effector CD4⁺ cells in PD-L1–positive tumors compared with PD-L1–negative tumors (P = 0.01; Fig. 3C). Among CD8⁺ T cells, however, there was a significantly higher proportion of CD8⁺ memory T cells (P = 0.007) with an increase in CD8⁺ effector memory T cells (P = 0.03) and a lower proportion of CD8⁺ effector T cells (P = 0.001) in PD-L1–positive tumors compared with PD-L1–negative tumors. We detected no differences between PD-L1–negative and PD-L1–positive tumors in the proportion of CD8⁺-naïve (P = 0.39) or central memory (P = 0.90) T cells (Fig. 3D).

We also used flow cytometry to interrogate the functional status of infiltrating T cells with respect to markers of T-cell activation, proliferation, and inhibition. CD3⁺ T cells in PD-L1–positive tumors were significantly more likely to display the activated HLA-DR⁺ CD38⁺ phenotype than T cells in PD-L1–negative tumors (P = 0.001, Fig. 4A). Although there was no significant difference in the proportion of proliferating Ki67⁺ CD4⁺ T cells in PD-L1–positive versus PD-L1–negative tumors (P = 0.15), CD4⁺ T cells in PD-L1–positive tumors were more likely to express the regulatory T-cell (Treg) marker FOXP3⁺ (P = 0.005). In terms of T-cell–inhibitory markers, CD4⁺ T cells in PD-L1–positive tumors were also significantly more likely to be TIM-3⁺ (P = 0.002) as well as PD-1⁺ (P = 0.01) than CD4⁺ T cells in PD-L1–negative tumors. LAG-3 was infrequently detected on CD4⁺ T cells, and there was no significant difference based on PD-L1 status (P = 0.58, Fig. 4B).

In PD-L1–positive tumors, CD8⁺ T cells were significantly more likely to be proliferating than those in PD-L1–negative tumors (P = 0.02, Fig. 4C). As with the CD4⁺ T-cell analysis, CD8⁺ T cells in PD-L1–positive tumors were also more likely to express the inhibitory markers TIM-3 (P = 0.003) and PD-1 (P = 0.008) than CD8⁺ T cells isolated from PD-L1–negative tumors (Fig. 4C). LAG-3 was also rarely seen on CD8⁺ T cells, with no difference between PD-L1–positive and -negative samples (P = 0.43, Fig. 4C).

We next used flow cytometry to examine the coexpression patterns of TIM-3 and PD-1–inhibitory receptors on T cells. PD-L1–positive tumors had higher levels of TIM-3⁺/PD-1⁺ “double positive” CD4⁺ (P = 0.002) and CD8⁺ (P = 0.005) T cells and lower levels of TIM-3⁻/PD-1⁻ “double negative” CD4⁺ (P = 0.01) and CD8⁺ (P = 0.005) T cells as shown in Fig. 5A and B. About 50% of CD4⁺ T cells and 30% of CD8⁺ T cells expressed PD-1 in the absence of TIM-3 (“PD-1 single positives”), but expression of TIM-3 in the absence of PD-1 was rarely observed (“TIM-3 single positives”, Fig. 5A and B).

In addition to comparing mesothelioma samples according to PD-L1 status and histologic subtype, we also compared immune infiltrates by tumor genotype. We found no significant differences in the degree of immune cell infiltration in BAP1 wild-type versus BAP1-mutant mesothelioma (Supplementary Fig. S4A), and no difference in NF2 wild-type versus NF2-mutant samples (Supplementary Fig. S4B). Most PD-L1–negative tumors harbored BAP1 mutations (11/12 samples), whereas roughly half of PD-L1–positive tumors were BAP1 mutant (7/15, P = 0.02). NF2 mutation status did not correlate with PD-L1 expression (not shown). Germline DNA from patients was not available in this study to discriminate germline versus somatic tumor mutations.

To better visualize immunologic differences among tumor samples, we used t-stochastic neighbor embedding (t-SNE) to incorporate 38 flow cytometry parameters into a two-dimensional scatter plot, where close proximity between any two samples on the t-SNE map represents their immunophenotypic similarity (29). In addition to the mesothelioma tumor samples, flow cytometric analysis was also performed on adjacent pieces of normal tissue in 7 cases.

Two main immunologic groups were identified through t-SNE analysis, termed “Normal Cluster,” which contained all the paired normal tissue samples, and “Tumor Cluster,” which was significantly enriched for tumor samples (P < 0.0001, Fig. 6A). Sarcomatoid and biphasic tumors were more immunologically distant from the Normal Cluster than epithelioid tumors (P = 0.03, Fig. 6A; Supplementary Fig. S5A). A greater distance from the Normal Cluster was associated with an increasing fraction of infiltrating CD3⁺ T cells (P = 0.0001, Fig. 6B; Supplementary Fig. S5B), along with a decreasing fraction of CD66b⁺ granulocytes (P < 0.0001, Fig. 6C; Supplementary Fig. S5C). Distance from the Normal Cluster was also associated with PD-L1 positivity (P = 0.001, Fig. 6D; Supplementary Fig. S5D), increasing proportions of TIM-3⁺/PD-1⁺/CD8⁺ T cells (P < 0.0001, Fig. 6E; Supplementary Fig. S5E), as well as an increasing expression of FOXP3 on CD4⁺ T cells (P < 0.0001, Fig. 6F; Supplementary Fig. S5F).

Overall survival data were collected on all patients in this study. We found that patients with nonepithelioid mesothelioma had a significantly worse prognosis compared with patients with epithelioid tumors (P = 0.02, Supplementary Fig. S6A). There was no significant difference in overall survival among patients with PD-L1–positive tumors versus PD-L1–negative tumors (P = 0.15, Supplementary Fig. S6B). There was also no significant difference in survival among tumors with high versus low CD3 infiltrates, whether assessed by flow cytometry (P = 0.90, Supplementary Fig. S6C) or IHC (P = 0.89, Supplementary Fig. S6D).

Chronic inflammation in response to inhaled asbestos fibers has long been recognized as a contributing factor for the development of malignant pleural mesothelioma (30, 31). Dissecting the properties of these inflammatory cells within tumors will provide greater insights into the immunologic mechanisms of response and resistance to immunotherapy in this disease. Here, we used flow cytometry to characterize 43 resected malignant pleural mesothelioma specimens and uncovered distinct immunologic phenotypes in PD-L1–positive tumors as compared with PD-L1–negative tumors, and in sarcomatoid/biphasic tumors as compared with epithelioid tumors. We found that PD-L1–positive and sarcomatoid/biphasic tumors have a significantly greater proportion of infiltrating T cells than PD-L1–negative and epithelioid tumors, respectively. PD-L1–positive tumors also show significant increases in T-cell proliferation and activation, along with significant increases in Tregs and expression of T-cell–inhibitory markers.

Among PD-L1–positive tumors, we observed considerable immunophenotypic variability across samples, which may explain why only a minority of PD-L1–positive mesotheliomas responded to the PD-1 inhibitor pembrolizumab in the KEYNOTE-028 study (25). Factors other than PD-L1 expression that may modulate the efficacy of PD-1 inhibitors include (i) the relative abundance of infiltrating lymphocytes; (ii) coexpression of multiple inhibitory receptors on T cells; and (iii) the influence of MDSCs and tumor-associated macrophages (32, 33). In the current study, we demonstrate that tumor immune profiling with flow cytometry enables deeper investigation into these factors.

We found that some PD-L1–positive samples contained only a small proportion of infiltrating lymphocytes, whereas others were associated with dense immune infiltrates. Although the presence of tumor-infiltrating lymphocytes (TIL) may be a favorable prognostic factor in some cancers (34, 35), it is not clear whether TILs are predictive of a response to immune checkpoint blockade. One study of 41 patients with different tumor types showed no association of TILs with response to nivolumab, but the vast majority of samples analyzed in that study were not obtained immediately prior to initiating treatment with nivolumab (36). We also observed considerable variation in the proportion of infiltrating monocytic MDSCs across tumor samples, which in some cases may have a substantial role in creating an immunosuppressive microenvironment and promoting tumor growth and metastasis (32). There also appeared to be an inverse correlation between T-cell and neutrophil infiltrates in our study. The interaction between T cells and tumor-associated neutrophils is becoming increasingly recognized as an important modulator of cancer control and progression (37–39) and warrants further study in mesothelioma.

In animal cancer models, coexpression of additional inhibitory receptors on immune cells may also blunt responses to PD-1 inhibition. For example, TIM-3, a negative regulator of T cells, is frequently coexpressed with PD-1 on TILs in mice with solid tumors or leukemia, and targeting both PD-1 and TIM-3 can be more effective at controlling tumor growth than targeting each pathway alone (40, 41). We also found frequent coexpression of PD-1 and TIM-3 on CD8⁺ T cells, suggesting that combined inhibition of the PD-1 and TIM-3 pathways might be an effective therapeutic strategy in mesothelioma. High expression of PD-1 and TIM-3 on CD8⁺ T cells has previously been shown to identify clonally expanded, mutation-specific, tumor-reactive immune cells in melanoma (13). The high proportion of effector memory CD8⁺ cells that we observed in PD-L1–positive tumors compared with PD-L1–negative tumors may also reflect chronically stimulated T cells with high-affinity T-cell receptors that recognize tumor antigen (42). In our tumors with the highest proportion of TIM-3⁺/PD-1⁺/CD8⁺ T cells, we also saw high PD-L1 expression on tumor cells as well as increased FOXP3 expression on CD4⁺ T cells. This finding is in keeping with the concept of adaptive immune resistance (43), and recent work in melanoma has similarly shown that CD8⁺ T cells within a tumor can both upregulate expression of PD-L1 and indoleamine-2,3-dioxygenase through IFNγ signaling and also recruit FOXP3⁺ Tregs by cytokine-mediated CCR4 signaling in the tumor microenvironment (44, 45).

Immune profiling by flow cytometry in solid tumors has several potential advantages over traditional IHC. Analysis of immune infiltrates by flow cytometry can be performed rapidly and is quantitative, objective, and largely automated, whereas IHC tends to be semiquantitative and relies on subjective interpretation by a pathologist. Flow cytometry allows for multiplexed analysis of several markers simultaneously as well as phenotyping of specific immune cell types. In contrast, IHC has historically been restricted to looking at one marker per slide, making it difficult to analyze subpopulations of immune cells and challenging to determine the coexpression of markers at single-cell resolution, although newer multiplex techniques are under investigation for examining multiple markers per slide (46, 47). One disadvantage of flow cytometry compared with IHC is the loss of spatial relationships between tumor cells and immune cells, which may be an important determinant of the response to immune checkpoint blockade (48). However, this limitation may be overcome by combining data from IHC and flow cytometry on clinical samples.

Along with PD-1, we found frequent coexpression of TIM-3 in several samples. This may partially explain why only a minority of PD-L1–positive mesothelioma patients responded to the PD-1 inhibitor pembrolizumab (25). Comprehensive immunoprofiling by flow cytometry will hopefully improve our understanding of the factors that determine response and resistance to immunotherapy and lead to rationally designed immunotherapy combination trials.

J. English is the vice president (Head of Discovery IONC TIP) at EMD Serono. S.J. Rodig reports receiving a commercial research grant from Bristol-Myers Squibb and is a consultant advisory board member for Perkin-Elmer Inc. F.S. Hodi reports receiving a commercial research grant from Bristol-Myers Squibb (to the institution) and is a consultant/advisory board member for EMD Serono, Genentech, Merck, and Novartis. P.A. Janne reports receiving commercial research grants from AstraZeneca and Astellas, has ownership interest (including patents) in Gatekeeper Pharmaceuticals, and is a consultant/advisory board member for AstraZeneca, Boehringer Ingelheim, Chugai Pharmaceuticals, Merrimack Pharmaceuticals, Pfizer, and Roche/Genentech. L.M. Sholl is a consultant/advisory board member for Genentech. R. Bueno reports receiving commercial research grants from Genentech and Verasten. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.M. Awad, P.H. Lizotte, M.A. Bittinger, J.M. English, W.G. Richards, F.S. Hodi, K.-K. Wong, R. Bueno

Development of methodology: R.E. Jones, P.H. Lizotte, M. Kulkarni, G.S. Herter-Sprie, X. Liao, S. Koyama, D.A. Barbie, S.J. Rodig, K.-K. Wong

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.M. Awad, R.E. Jones, P.H. Lizotte, E.V. Ivanova, M. Kulkarni, G.S. Herter-Sprie, X. Liao, A.A. Santos, L. Keogh, C. Almonte, J. Barlow, W.G. Richards, S.J. Rodig, K.-K. Wong, R. Bueno

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.M. Awad, R.E. Jones, H. Liu, P.H. Lizotte, E.V. Ivanova, M. Kulkarni, A.A. Santos, L. Keogh, D.A. Barbie, A.J. Bass, S.J. Rodig, F.S. Hodi, K.W. Wucherpfennig, P.A. Jänne, L.M. Sholl, K.-K. Wong

Writing, review, and/or revision of the manuscript: M.M. Awad, H. Liu, P.H. Lizotte, E.V. Ivanova, X. Liao, L. Keogh, C. Almonte, J.M. English, W.G. Richards, F.S. Hodi, K.W. Wucherpfennig, P.A. Jänne, P.S. Hammerman, K.-K. Wong, R. Bueno

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.M. Awad, R.E. Jones, A.A. Santos, L. Keogh, C. Almonte, P.S. Hammerman, K.-K. Wong

Study supervision: M.A. Bittinger, J.M. English, P.S. Hammerman, K.-K. Wong, R. Bueno

M.M. Awad received funding from the American Society of Clinical Oncology (ASCO) Conquer Cancer Foundation and from the International Association for the Study of Lung Cancer (IASLC). P.S. Hammerman received funding from the Starr Consortium for Cancer Research and from the Damon Runyon Cancer Research Foundation. K.-K. Wong received funding from the Expect Miracles Foundation. This research was also supported by Stand Up To Cancer – American Cancer Society Lung Cancer Dream Team Translational Research Grant (grant number: SU2C-AACR-DT17-15). Stand Up To Cancer is a program of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.