A Phase I Trial of Regional Mesothelin-Targeted CAR T-cell Therapy in Patients with Malignant Pleural Disease, in Combination with the Anti–PD-1 Agent Pembrolizumab

Malignant pleural diseases, comprising metastatic lung and breast cancers and mesothelioma, are aggressive solid tumors. Malignant pleural mesothelioma (MPM) is an aggressive cancer characterized by resistance to treatment and poor survival (1). Median overall survival (OS) for patients with MPM following first-line treatment comprising cisplatin and pemetrexed is 13 to 16 months; the addition of bevacizumab prolongs OS to 18.8 months, albeit at the cost of increased toxicity (2). We and others have shown that immune responses are prognostic in patients with MPM (3–6). Immune checkpoint inhibitor (ICI) therapy has been investigated in MPM, a solid tumor with a low tumor mutational burden (TMB). PD-L1 expression is very low in patients with an epithelioid histologic profile, the most common type of MPM (7). This is consistent with the lack of improved survival in patients with MPM with single-agent checkpoint blockade inhibitors (Supplementary Table S1; ref. 8). The National Comprehensive Cancer Network guidelines for MPM include the use of ICIs as second-line treatment. A recent phase III trial of first-line dual nivolumab and ipilimumab therapy in unresectable MPM reported a median OS of 18.1 months, compared with 14.1 months for standard-of-care chemotherapy (9). This dual-ICI-agent drug regimen was approved by the FDA in October 2020 (https://www.fda.gov/news-events/press-announcements/fda-approves-drug-combination-treating-mesothelioma), 16 years after the previous approval of any systemic therapy for MPM.

Chimeric antigen receptor (CAR) T-cell therapy induces durable and curative responses in patients with hematologic malignancies, but attempts to treat solid tumors with this approach have so far met with limited success (10, 11). The obstacles to effectively treating solid tumors with T-cell therapy include heterogeneous antigen expression, an inability to achieve T-cell infiltration of the tumor, and inhibition of CAR T-cell function by an immunosuppressive microenvironment (12, 13). To address these obstacles, we developed a regionally delivered mesothelin-targeted CAR T-cell therapy (14, 15) and translated it for use in clinical trials. We selected mesothelin as a target cell-surface antigen because of its overexpression (14, 16) and association with aggressiveness—malignant transformation, cancer cell invasion, proliferation, and metastases (17–21)—in a majority of MPM and other solid tumors, together with its low levels of expression in normal tissues (22). All of the components in our CAR construct are fully human, including the single-chain variable fragment (scFv), an intentional choice to avoid human anti-mouse antibody reactions observed in some trials using CARs that comprised an scFv of murine origin.

As MPM is locoregionally aggressive (23, 24), we investigated regional versus systemic CAR T-cell administration in an orthotopic MPM mouse model: regional CAR T-cell therapy achieved superior efficacy at lower doses by avoiding sequestration of CAR T cells in lungs and by augmenting CD4 helper function (15). Regionally activated CAR T cells were able to circulate and establish systemic immunity.

In our clinically relevant mouse model, we observed that even costimulated CAR T cells became functionally exhausted when faced with large tumor burdens, in part because of inhibitory PD-1/PD-L1 signaling. We demonstrated that the administration of anti–PD-1 agents rescues function in exhausted CAR T cells and enhances antitumor efficacy (25, 26).

We therefore initiated an open-label, dose-escalating, single-center, first-in-human phase I trial of intrapleural delivery of mesothelin-targeted CAR T cells in patients with previously treated histologically proven pleural cancer from MPM, metastatic lung cancer, or metastatic breast cancer. Patients with mesothelin expression in at least 10% of tumor cells on IHC analysis and/or serum soluble mesothelin-related peptide (SMRP) >1.0 nmol/L were eligible (as detailed in the protocol; see Supplementary Material). We demonstrate the feasibility and safety of this approach and report our experience in a subcohort of patients who received pembrolizumab after CAR T cells.

From November 2015 to April 2019, 71 patients with malignant pleural disease were screened: 36 patients were not eligible, and 35 patients underwent leukapheresis (Fig. 1A). Two patients did not undergo cell production; 6 patients who had cells produced did not undergo infusion (Fig. 1A). In total, 27 patients were infused: 25 with MPM, 1 with metastatic lung cancer, and 1 with metastatic breast cancer (Fig. 1B). All patients had received ≥1 prior line(s) of therapy (median, 2; range, 1–13); 33% of patients had received ≥3 prior lines.

Patients in cohort 1 (2 MPM, 1 metastatic lung cancer) did not receive cyclophosphamide preconditioning. Of the 24 patients who received cyclophosphamide, 23 presented with MPM, and 1 presented with metastatic breast cancer (Table 1). Of the 23 patients with MPM, 17 (74%) had clinical stage 2 to 4 disease at enrollment, and 6 (26%) had relapsed disease (Table 2). Before CAR T-cell infusion, 11 patients with MPM (48%) had progressive disease (PD), and 12 (52%) had stable disease (SD) from prior lines of therapy (Table 2). Median mesothelin expression in the tumor was 100% (range, 25%–100%), median serum SMRP level at the time of infusion was 2.7 nmol/L (range, 0.9–17.5 nmol/L), median TMB was 2.6 (range, 0–4.9), and median PD-L1 percentage was 0 (range, 0%–80%; Supplementary Table S1). Median time from diagnosis to T-cell infusion was 6.1 months (range, 2.9–73.3 months; Table 2).

The mesothelin-specific CAR was engineered as a fusion protein encoding a fully human scFv (m912) and the CD28/CD3ζ sequences, inserted in the SFG gamma-retroviral vector. The CAR was linked to iCaspase-9 via an EMCV internal ribosomal entry site (15, 27–30). Details of the CAR T-cell manufacturing process (31) and analyses are described in the protocol (provided in Supplementary Material). The median time from apheresis to infusion of the product was 3.2 months (range, 1.1–20.2 months). There were no production failures. Median CD3⁺ CAR T-cell transduction was 41% (range, 10%–62%; Fig. 1C), without notable differences in CAR T-cell phenotype across cohorts (Supplementary Fig. S1A–S1D). Patients with clinical benefit and without significant adverse events following the first dose of CAR T cells were considered for reinfusion with a second dose of CAR T cells.

Following a single dose of preconditioning cyclophosphamide (1,500 mg/m²), mesothelin-targeted CAR T cells (0.3M–60M CAR T cells/kg) with the iCaspase-9 safety gene (27) were administered intrapleurally either through a pleural catheter after complete evacuation of pleural effusion in 11 patients (41%) or via interventional radiology–guided imaging in 16 patients (59%; Fig. 1D; Table 1). Leveraging our interventional radiology expertise, we were able to infuse even in patients with fused pleural cavities, including patients who had undergone thoracotomy or hemithoracic radiation, with no interventional failures (intervention was required at two different sites in three patients). The site of administration was chosen on the basis of safety and ease of accessibility by either CT- or ultrasound-guided access or a combination of both methods.

Dose escalation is detailed in Table 1. Three patients in cohort 1 were treated without cyclophosphamide preconditioning as a precautionary safety measure. From cohort 2 onward, three patients were treated at each dose level, and dose escalation continued to dose level 8 (three additional patients were treated at dose level 6; these patients underwent apheresis and cell manufacturing at a lower dose while on a prior treatment and were treated upon disease progression). There were no dose-limiting toxicities. There were no grade 5 adverse events. Supplementary Table S2 lists the treatment-emergent adverse events that occurred in ≥15% of the cohort (n = 27 patients); all adverse events one month postinfusion are detailed in Supplementary Table S3. All grade 4 adverse events were reversible laboratory abnormalities associated with lymphodepleting chemotherapy. Six patients experienced grade 3 clinical adverse events: constipation (two patients) and dysphagia, dyspnea following reinfusion, thromboembolic event, and febrile neutropenia (one patient each). No patients experienced cytokine release syndrome grade >2, neurotoxicity grade >2, or on-target, off-tumor toxicity. The maximum tolerated dose was not reached. Given the absence of toxicities, the iCaspase-9 safety switch was not activated in any patient. Four patients (15%) received a second intrapleural infusion at the time of disease progression, 7 to 24 months after demonstration of safety and clinical benefit from the first dose.

Peak levels of C-reactive protein were observed within 72 hours of infusion and during episodes of fever (n = 14 patients; Supplementary Fig. S2A–S2C). One patient was admitted to the intensive care unit as a precaution for monitoring following reinfusion. Two patients required IL6 blockade (tocilizumab, maximum of two doses), 1 each after infusion and reinfusion. Eleven patients (41%) were readmitted for monitoring, mostly within two weeks for fever, fatigue, and malaise. Serum IL6 levels >50 pg/mL during the first week of treatment were associated with readmission for fever (Supplementary Fig. S2D).

Two patients had grade 1 pleuritic pain, three had pleural effusion (grade 1, two patients; grade 2, one patient), and one had pericardial effusion (grade 2), following reinfusion (Supplementary Fig. S3A–S3H). In two patients with pleural metastatic non–small cell lung cancer (patient #1) or breast cancer (patient #5), no adverse events were noted. These patients received multiple lines of therapy after CAR T cells and survived 12 and 11 months, respectively.

The outcomes-analysis population included 23 patients with MPM who received cyclophosphamide preconditioning plus CAR T cells (Table 2). Results are presented across dose levels, as no dose-limiting toxicities were observed. At treatment-assessment cutoff (June 2020), median follow-up was 20.3 months (interquartile range, 19.2–26.1 months; Fig. 1B), median OS since T-cell infusion was 17.7 months [95% confidence interval (CI), 13.2 months to not estimable (NE)], and one-year OS was 74% (95% CI, 58%–94%; Fig. 2A–C).

Twenty-two of the 23 patients with MPM had a decrease in serum SMRP following CAR T-cell infusion (Supplementary Figs. S4, S5A, and S5B); CAR T cells were detectable in peripheral blood within four days in 87% of patients, for >100 days in 39% of patients, and for >200 days in 17% of patients (Fig. 3A). In two patients, CAR T cells were detected in pleural tumor at weeks 25 and 32 by PCR.

The combination immunotherapy outcomes analysis population included patients with MPM who received cyclophosphamide plus CAR T cells and subsequent pembrolizumab for a minimum three doses, with at least three months of follow-up after the third dose, representing 18 of the 23 patients with MPM. Seven of these patients (39%) had received ≥2 lines of therapy before CAR T-cell infusion (three of seven received ≥3 prior lines; Tables 1 and 2). The median time to initiation of pembrolizumab after CAR T-cell administration was six weeks (range, 4–17 weeks; Supplementary Fig. S6; Table 1). The adverse events experienced by these 18 patients up to 6 months after T-cell infusion are detailed in Supplementary Table S4.

Median OS after CAR T-cell treatment was 23.9 months (95% CI, 14.7 months to NE; Supplementary Fig. S7A). One-year OS was 83% (95% CI, 68%–100%); time to next treatment was not reached (Supplementary Fig. S7B). Reduction in serum SMRP level was observed after administration of both CAR T cells and pembrolizumab (Supplementary Figs. S4, S8A, and S8B). In peripheral blood, peak levels of CAR T cells, quantified on the basis of vector copy number (VCN), were higher following intrapleural CAR T-cell infusion (median, 1,660 VCN/mL; range, 0–77,300 VCN/mL), relative to peak levels after pembrolizumab administration (median, 397 VCN/mL; range, 0–48200 VCN/mL]; Fig. 3A–C; Supplementary Fig. S9A–S9G).

Radiologic evaluation by computed tomography (CT) scan using modified Response Evaluation Criteria in Solid Tumors (mRECIST) was performed at four to six weeks (not predefined in the protocol). Among combination immunotherapy patients with measurable disease by mRECIST (N = 16), the best overall response was partial response (PR) in 2 of 16 (12.5%), SD in 9 of 16 (56.3%), and PD in 5 of 16 (31.3%; Fig. 3D and E). In 8 of 16 patients, SD or better was sustained for ≥6 months (Figs. 2A, 3E; Supplementary Fig. S6). Radiologic measurements following CAR T-cell infusion before and after pembrolizumab administration are shown in Fig. 3E. Most importantly, patients with SD remained functionally well and did not require a next treatment for a prolonged duration (Figs. 2A, 3D; Supplementary Figs. S3A–S3H, S6, and S10).

A 76-year-old patient with epithelioid MPM had a 28% reduction in the target lesion with CAR T cells alone (Fig. 4A), with subsequent complete metabolic response on PET scan and PR on CT scan, with 78% reduction in the target lesion following pembrolizumab treatment (Fig. 4A). This response lasted for 26 months. The patient's serum SMRP level remained at baseline, and the patient remained functionally well (Fig. 4B). CAR T cells were detected in peripheral blood up to 12 weeks, with an increase in T-cell Simpson clonality, T-cell clonal expansion (Supplementary Fig. S11), and detection of new and sustained IgG responses (Fig. 4C).

In a 72-year-old patient with biphasic MPM, a similar complete metabolic response on PET scan and PR on CT scan (mRECIST; 40% reduction in target lesion) was observed (Fig. 4D). This response persisted, with no other treatments, for 16 months, and the patient remained functionally normal, gained and maintained weight, and maintained baseline serum SMRP level (Fig. 4E). Peripheral blood analyses showed new IgG responses following both CAR T cells and pembrolizumab administration (Fig. 4F). Tumor biopsy at 32 weeks showed reduction in tumor cell burden, and CAR T cells were detected by PCR (Fig. 4G). A patient who received first dose of pembrolizumab 121 days after CAR T-cell infusion had a 50% reduction in the target lesion by mRECIST, with redetection of CAR T cells in the peripheral blood following pembrolizumab (Fig. 4H and I). Among three patients with MPM, CAR T cells were redetected in peripheral blood up to 100 days after palliative radiation in one patient (Fig. 4J and K) and after reinfusion in two patients (Figs. 4L and M; Supplementary Fig. S3A–S3H). One patient with no measurable disease by mRECIST remained well, with no treatments administered other than pembrolizumab and reinfusion, for 32 months.

Analysis of peripheral blood T-cell receptor beta clonality (n = 10) revealed T-cell clonal expansion at weeks 2 to 5, relative to baseline (Fig. 5A), with greater clonal expansion in patients with PR or SD than in patients with PD (Fig. 5B). Simpson clonality was lower in patients with PR than in patients with SD or PD (Fig. 5C). On ProtoArray analysis of patient serum samples (n = 9; MPM subjects without disease served as controls), new IgG responses were observed after administration of both CAR T cells and pembrolizumab, compared with baseline (pre–CAR T cells; Fig. 5D). New IgG responses accounted for 15% and 20% of total IgG responses at the post-CAR and post-pembrolizumab time points, respectively (Fig. 5E). Prevalent IgG responses (>3.5-fold over baseline) and the percentage of unique IgG responses among prevalent responses are shown in Fig. 5F and G, respectively.

In patients with MPM with pleural effusion available for analysis, CAR T cells and elevated cytokine levels were detected in pleural effusions even in the absence of detection or low levels of these in the peripheral blood (Fig. 5H). Luminex analyses of pleural fluid and serum for 68 effector response proteins showed that peak protein levels were seen in pleural fluid at two weeks, and much lower levels were seen in serum (Fig. 5I and J).

T-cell clonal expansion was evident in the pleural fluid by week 2 (Fig. 5A; Supplementary Fig. S11). In pleural fluid, there was an 84% to 100% overlap of expanded clones and a 66% to 82% overlap of all clones found in the peripheral blood (Fig. 5K). The peripheral clonality of each patient was determined at several time points and was compared with the infused product clonality to dynamically track changes in the identified clones. Peripheral clonality overlap with infusion product was seen from week 1 to week 52 (Fig. 5L).

The trial reported herein reflects the clinical translation of preclinical testing and validation performed to overcome obstacles posed by solid tumors (26) by incorporating the following novel features: (i) intrapleural administration of CAR T cells, which is safe and feasible and has the potential to treat solid tumors and establish long-lasting systemic circulation (15); (ii) regional administration of CARs adjacent to tumors with high antigen expression without on-target, off-tumor toxicities observed in regional or systemic normal tissues (14, 16, 17, 19–21); (iii) the combination of T cells and pembrolizumab to further enhance CAR T-cell persistence and function (25); and (iv) manufacturing of CAR T cells (32), with the ability to produce up to 60M T cells/kg per patient.

Intrapleural administration of mesothelin-targeted CAR T cells in patients with pleural cancer was well tolerated, without overt toxicity to normal tissues known to express mesothelin. Mesothelin CAR T-cell administration followed by PD-1 blockade has the potential to durably treat solid tumors. Our findings of antitumor efficacy, with no on-target, off-tumor toxicity as observed in other solid tumor T-cell therapy trials (33), in patients with MPM and low TMB and PD-L1 expression (ref. 7; Table 1) are noteworthy. The scFv in our CAR construct was designed with a unique combination of IgM and IgG components, with relatively low affinity (34). The fully human elements of our CAR, including the scFv (34), possibly contributed to the ability of the CAR T cells to evade immune rejection and to be safely administered a second time, with efficacy (Fig. 4L and M; Supplementary S3A–S3H) and without adverse reactions, in a small number of patients (n = 4). Lack of intense cytokine release syndrome in our trial, compared with patients with hematologic malignancy following CAR T-cell infusion, may be indicative of insufficient tumor infiltration or the solid tumor microenvironment limiting CAR T-cell activation.

CAR T-cell engraftment, as measured in blood, required cyclophosphamide conditioning and overall increased with infusion dose (Fig. 2A). In our preclinical model, we observed that regionally delivered CAR T cells achieved rapid and efficient tumor infiltration throughout the pleural tumor, followed by systemic circulation within a short period (15, 25). These findings are consistent with our observations in this clinical trial (Figs. 2, 3, 4; Supplementary Figs. S9, S11, S12A–S12C). We did not observe any difference in adverse events, circulating CAR T cells (as demonstrated by VCN in peripheral blood), or outcomes between patients who were administered CAR T cells through a pleural catheter or via interventional radiology–guided imaging. The detection of circulating CAR T cells, 100 days after a single infusion, underscores the potential of achieving long-lasting T-cell persistence following locoregional infusion (26).

Evidence of CAR T-cell activity following infusion is reflected in T-cell and CAR T-cell clonal expansion (Figs. 3A, 5A and H) and new IgG responses (Figs. 4C, F, 5D–G) before administration of pembrolizumab. The clonal expansion of endogenous T-cell clones not overlapping with the CAR T-cell product (Fig. 5) demonstrates that PD-1 blockade expanded endogenous T cells as well. The finding of responses in patients with biphasic MPM with mesothelin expression as low as 25% (Fig. 4D–G) is consistent with the recruitment of endogenous T cells against mesothelin-negative tumor cells. The expansion of preexisting IgG responses and the development of new IgG responses further suggest that CAR T-cell infusion followed by PD-1 blockade promotes endogenous immunity and antigen spreading (Figs. 4 and 5; Supplementary Figs. S3A–S3H and S11). In murine models, we found that PD-1 blockade sustained mesothelin CAR T-cell therapy, which we attributed to a direct effect on CAR T cells, given the human specificity of pembrolizumab and the absence of endogenous T cells in NSG mice (26). In our patients with MPM, baseline PD-L1 expression was low but may have increased following infusion of CAR T cells. Our study was not funded to support iterative biopsies to investigate this question, but in a single instance where tissue was available before and after CAR T-cell infusion (patient #7), we found 30% PD-L1 expression before CAR T-cell infusion and 70% eight weeks thereafter. It is thus possible that PD-1 blockade rescues both CAR T cells and endogenous T cells that are recruited to the pleura following CAR T-cell infusion. The overlap of peripheral T-cell clones and expanded clones with infused CAR T cells (Fig. 5L) before the initiation of pembrolizumab raises an intriguing possibility that potent preexisting T-cell immunity may have a predisposing effect, resulting in better efficacy of CAR T-cell therapy. Together, these observations suggest the potential importance of combined regional CAR T cells and PD-1 blockade to recruit polyclonal endogenous immunity to overcome tumor antigen heterogeneity and decrease the likelihood of antigen escape.

Radiographic assessment of response in patients with MPM is challenging for multiple reasons: tumor distribution as a pleural rind, invasion into the chest wall and organs that cannot be assessed, interobserver variability in quantitating diffuse tumor, presence of pleural effusion masking underlying disease as well as fissural disease, false-positive PET avidity due to talc, and pseudoprogression, which is seen with immunotherapies (35–37). A consensus report from the NCI Thoracic Malignancy Steering Committee and the International Association for the Study of Lung Cancer attests to these challenges in this uniquely pleural-based cancer and notes that assessment using mRECIST, with the same measurement parameters and the same observer, can be useful (37). The mRECIST responses reported here were measured by an independent radiologist without knowledge of clinical data, and sustained complete metabolic responses were noted on PET scan. The best responses (PR by mRECIST and complete metabolic response on PET) were observed in two patients who received CAR T cells and pembrolizumab, lasting for 18 to 26 months, and a third patient had CAR T cells detected in peripheral blood for 18 months and did not require further treatment for 25 months. Four patients received reinfusion of a second dose of CAR T cells at 25, 12, 8, and 8 months after the first infusion and were alive at 32, 24, 20, and 19 months. No adverse events grade >3 were observed in these patients; CAR T cells, as monitored by VCN in the peripheral blood, were detected 90, 146, 7, and 91 days after the second infusion, suggesting they were not immunologically rejected by the recipient.

Although we provide mRECIST measurements as well as outcomes analyses, these results should be interpreted with caution, as they were not a prespecified part of the phase I study design, and pembrolizumab treatment was not initiated in a uniform fashion. With the known difficulties in assessing response by imaging, OS is the only reliable parameter to date to assess response to therapy in patients with MPM (37); OS was used as an endpoint in the approval of pemetrexed as well as recent dual-agent checkpoint blockade for MPM where there are no improvements in progression-free survivals (Supplementary Table S5; refs. 1, 9). Median OS among patients who received combination immunotherapy in our study (39% of patients received ≥2 prior lines of therapy; Supplementary Table S6) was 23.9 months after CAR T-cell infusion. Several additional measurements of antitumor activity were presented (Figs. 2, 3, 4, 5; Supplementary Figs. S3A–S3H, S9, S10, and S12A–S12C). However, as the patients in this phase I study designed to assess safety were selected for reasonable life expectancy and functional status to be able to await CAR T-cell manufacturing and treatment, the survival results cannot be directly compared with outcomes of other immunotherapy trials.

Our data strongly support the investigation of combination immunotherapy with CAR T cells and PD-1 blockade agents in solid tumors. Owing to safety concerns following regional administration of CAR T cells at a higher dose, we initially administered pembrolizumab several weeks after administration of CAR T cells. With confirmation of safety, we moved pembrolizumab closer to and more consistently following CAR T-cell administration. On the basis of the results of this trial, we are now conducting a phase II study with a fixed dose of mesothelin-targeted CAR T cells (6 × 10⁷ CAR T cells/kg) followed by initiation of pembrolizumab four weeks after CAR T-cell administration.

This is an open-label, dose-escalating, single-center, phase I study of mesothelin-targeted CAR T cells in patients with previously treated histologically proven pleural cancer from MPM, metastatic lung cancer, or metastatic breast cancer (trial registration number: NCT02414269). Patients with mesothelin expression of at least 10% of tumor cells on IHC analysis (Supplementary Fig. S13A–S13D) and/or patients with epithelioid mesothelioma with serum SMRP >1.0 nm/L were eligible (as detailed in the study protocol; see Supplementary Material). Following a single dose of preconditioning cyclophosphamide (1,500 mg/m²), mesothelin-targeted CAR T cells (0.3M–60M CAR T cells/kg) with the iCaspase-9 safety gene (27) were administered intrapleurally either through a pleural catheter or via interventional radiology–guided imaging.

The study protocol and amendments were approved by our institutional review board. All patients provided written informed consent. Response and toxicity outcomes were reviewed by an independent committee established by the institutional Clinical Research Oversight Committee to manage potential conflicts of interest in the interpretation of responses. No one who is not an author contributed to the writing of the manuscript.

Details of the CAR T-cell manufacturing process (31) and analyses are described in the protocol (provided in the Supplementary Material).

Our primary objective was to assess the safety, dose requirement, and targeting efficiency of CAR T cells. Secondary objectives included evaluation of changes in serum SMRP level and assessment of CAR T-cell persistence in peripheral blood. All patients, including those who received a second dose of CAR T cells and/or pembrolizumab (detailed in the amended protocol; see Supplementary Material), were included in the reporting of adverse events.

The outcomes-analysis population included patients with MPM who received cyclophosphamide preconditioning plus T cells and those who received cyclophosphamide plus combination immunotherapy (CAR T cells and subsequent pembrolizumab for a minimum three doses, with at least three months of follow-up after the third dose).

The sample size was based on a standard dose-escalation design. Descriptive statistics are presented as medians, with minimums and maximums (or interquartile range where specified) for continuous variables and with counts and percentages for categorical variables. Safety data are described as the number and proportion of patients who had treatment-related adverse events. Exact methods (Clopper–Pearson 95% CIs) were used for categorical variables. In the preliminary outcomes analysis, the time of the first CAR T-cell infusion was used as the time of origin in all time-to-event analyses. The analysis of OS used death as the event and was summarized using the Kaplan–Meier method. Time to next treatment after combination immunotherapy was summarized and displayed graphically using the cumulative incidence function. The administrative cutoff date was June 2020; no patient was lost to follow-up before this date. Analyses were performed using R software (3.6.1, R Foundation for Statistical Computing).

Cytokine release syndrome was graded in accordance with the Memorial Sloan Kettering cytokine release syndrome grading system (available on request). All patients were evaluated by the neurology team before and after CAR T-cell infusion. Neurologic and other toxicities were assessed in accordance with version 4.0 of the National Cancer Institute Common Terminology Criteria for Adverse Events.

All CT scans were reviewed by two radiologists after the period of data collection had ended. Measurable disease and response were defined per mRECIST (36); details are provided in the protocol (see Supplementary Material). Per mRECIST, measurable disease was defined as disease with a pleural thickness of 1.0 cm. Tumor thickness perpendicular to the chest wall or mediastinum was measured in two positions at three separate levels on transverse cuts of a CT scan. The sum of the six measurements defined a pleural unidimensional measure. Transverse cuts at least 1 cm apart were selected, when possible, in the upper, middle, and lower chest. At the time of reassessment, pleural thickness was measured at the same position and level. Nodal, subcutaneous, and other bidimensionally measurable lesions were measured unidimensionally. Unidimensional measurements were added to obtain the total tumor measurement. Complete response was defined as the disappearance of all target lesions, with no evidence of tumor elsewhere. PR was defined as a reduction in the total tumor measurement of at least 30%. PD was defined as an increase in the total tumor measurement of at least 20%, compared with the nadir measurement, or the appearance of one or more new lesions. SD was defined as disease that did not fulfill the criteria for PR or PD.

Measurements of serum SMRP, C-reactive protein, and IL6 were performed in a Memorial Sloan Kettering chemistry laboratory. The apheresis and infusion product obtained from each patient was characterized by flow cytometric analysis to determine its cellular composition. Flow cytometric quantification of the immune cell markers CD45, CD14, CD3, CD4, CD8, CD28, and CD62L and CAR T-cell expression by protein L assay were used to identify different cellular subsets. A quantitative real-time PCR assay was used to identify and quantify the presence of CAR T cells in the peripheral blood (vector copies per milliliter) and pleural fluid (vector copies per microgram of genomic DNA). Peripheral blood mononuclear cells and pleural effusion cells collected from each patient at different time points were subjected to genomic DNA isolation to study T-cell receptor beta clonality by use of ImmunoSEQ analysis (Adaptive Biotechnologies). Serum samples collected from each patient at baseline, three to four weeks after CAR T-cell therapy, and after anti–PD-1 treatment were assessed using the HuProt ProtoArray assay (CDI Laboratories) to identify anti-IgG responses against 24,000 proteins at each time point. Pleural fluid and serum profiles of 68 proteins were assessed using Luminex-based assays (65-plex Human ProcartaPlex panel, Thermo Fisher Scientific) and the 3-plex Millipore panel (Millipore Sigma).

P.S. Adusumilli has received research funding from ATARA Biotherapeutics, has served on the Scientific Advisory Board or as consultant to ATARA Biotherapeutics, Bayer, Carisma Therapeutics, Imugene, ImmPACT Bio, and Takeda Therapeutics, and has patents, royalties, and intellectual property on mesothelin-targeted CARs and T-cell therapies, which has been licensed to ATARA Biotherapeutics, as well as method for detection of cancer cells using virus, and pending patent applications on T-cell therapies. M.G. Zauderer reports grants and personal fees from Atara and grants from Department of Defense during the conduct of the study; grants and personal fees from Takeda and GSK, personal fees from Aldeyra and Novocure, grants from Department of Defense, NIH, Epizyme, Polaris, Sellas Life Sciences, BMS, and Curis, other support from Mesothelioma Applied Research Foundation, personal fees from Research to Practice, Medical Learning Institute, and OncLive outside the submitted work. I. Rivière reports other support from Atara Biotherapeutics during the conduct of the study; other support from Fate Therapeutics, Takeda Pharmaceutical, and Juno Therapeutics outside the submitted work. S.B. Solomon reports grants from Johnson & Johnson, personal fees from Varian, Olympus, Advantagene, and XACT Robotics and grants from GE Healthcare outside the submitted work. V.W. Rusch reports grants from NIH during the conduct of the study; Genelux Inc and Genentech, and other support from Intuitive Surgical outside the submitted work. R.E. O'Cearbhaill reports grants from Department of Defense and NIH during the conduct of the study; grants and personal fees from GlaxoSmithKline/Tesaro, Regeneron, and Seagen/Genmab, personal fees from Fresenius, grants from Bayer/Celgene/Juno, Ludwig Cancer Institute/CRI, TCR2 Therapeutics, AbbVie/StemCentrx, Syndax Pharmaceuticals, Marker Therapeutics, Sellas Therapeutics, Atara Biotherapeutics, Kite Pharmaceuticals, Genentech, Gynecologic Oncology Foundation, and AstraZeneca Pharmaceutical outside the submitted work; and AstraZeneca Pharmaceuticals meal 6/2/2019. B. Daly reports other support from Eli Lilly & Company and Roche outside the submitted work. J.L. Sauter reports stock ownership with Merck. S. Modi reports personal fees from Genentech, Daiichi Sankyo, AstraZeneca, Macrogenics, and Seattle Genetics outside the submitted work. C.M. Rudin reports personal fees from AbbVie, Amgen, AstraZeneca, Celgene, Epizyme, Genentech/Roche, Ipsen, Jazz, Lilly, Syros, Bridge Medicines, Harpoon Therapeutics, and Earli outside the submitted work. R.J. Brentjens reports personal fees from BMS and other support from Juno Therapeutics outside the submitted work. D.R. Jones serves as a consultant for AstraZeneca and Merck. M. Sadelain has patents, royalties, and intellectual property on mesothelin-targeted CARs and T-cell therapies, which has been licensed to ATARA Biotherapeutics. Memorial Sloan Kettering Cancer Center (MSK) has licensed intellectual property related to mesothelin-targeted CARs and T-cell therapies to ATARA Biotherapeutics, and has associated financial interests. No disclosures were reported by the other authors.

P.S. Adusumilli: Conceptualization, resources, data curation, software, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M.G. Zauderer: Methodology, writing–review and editing. I. Riviere: Resources, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. S.B. Solomon: Data curation, investigation, methodology, writing–review and editing. V.W. Rusch: Resources, writing–review and editing. R.E. O'Cearbhaill: Investigation, writing–review and editing. A. Zhu: Data curation, visualization, writing–review and editing. W. Cheema: Data curation, visualization, writing–review and editing. N.K. Chintala: Data curation, software, visualization, writing–review and editing. E. Halton: Data curation, writing–review and editing. J. Pineda: Data curation, writing–review and editing. R. Perez-Johnston: Data curation, visualization, methodology, writing–review and editing. K.S. Tan: Data curation, formal analysis, investigation, writing–review and editing. B. Daly: Resources, writing–review and editing. J.A. Araujo Filho: Data curation, investigation, visualization, writing–review and editing.D. Ngai: Data curation, writing–review and editing. E. McGee: Data curation, writing–review and editing. A. Vincent: Data curation, writing–review and editing. C. Diamonte: Data curation, writing–review and editing. J.L. Sauter: Resources, investigation, writing–review and editing. S. Modi: Resources, writing–review and editing.D. Sikder: Investigation, writing–review and editing. B. Senechal: Data curation, investigation, writing–review and editing. X. Wang: Data curation, investigation, writing–review and editing. W.D. Travis: Investigation, writing–review and editing. M. Gonen: Data curation, formal analysis, investigation, writing–review and editing. C.M. Rudin: Resources, writing–review and editing. R.J. Brentjens: Investigation, methodology, writing–review and editing. D.R. Jones: Resources, writing–review and editing. M. Sadelain: Conceptualization, supervision, validation, methodology, writing–original draft, writing–review and editing.

David B. Sewell of the Department of Surgery, Memorial Sloan Kettering Cancer Center, provided editorial assistance. P.S. Adusumilli's laboratory work is supported by grants from the NIH (P30 CA008748, R01 CA236615–01, and R01 CA235667), the U.S. Department of Defense (BC132124, LC160212, CA170630, and CA180889), the Baker Street Foundation, the Batishwa Fellowship, the Comedy vs. Cancer Award, the Derfner Foundation, the Dalle Pezze Foundation, the Esophageal Cancer Education Fund, the Geoffrey Beene Foundation, the Memorial Sloan Kettering Technology Development Fund, the Miner Fund for Mesothelioma Research, Mr. William H. Goodwin and Alice Goodwin, the Commonwealth Foundation for Cancer Research, and the Experimental Therapeutics Center of Memorial Sloan Kettering Cancer Center.

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.