BAP1-Altered Malignant Pleural Mesothelioma: Outcomes With Chemotherapy, Immune Check-Point Inhibitors and Poly(ADP-Ribose) Polymerase Inhibitors

Objectives Little is known regarding the outcomes of systemic treatments in BAP1-altered malignant pleural mesothelioma (MPM). Materials and Methods Forty five patients with MPM [group A: eight MPM patients with BAP1 inactivating mutation/copy number loss (FoundationOne® CDx/TEMPUSxT), selected from the electronic databases of four Israeli cancer centers (ICC); group B: 37 consecutive (years 2016–2018) MPM patients selected from the electronic databases of two ICC—of those six patients without a BAP1 alteration (group B1) and 31 patients not tested for BAP1 (group B2)] were analyzed for ORR, PFS (mRECIST), and OS with 1st-line platinum/pemetrexed+/−antiangiogenic drug (CT, n-28), immune check-point inhibitors (ICPi, n-16) and poly (ADP-ribose) polymerase inhibitors (PARPi, n-4). OS since diagnosis (OSDx) was assessed. Results There were no differences in ORR or mPFS with CT between the groups: ORR-50% vs. 47% vs. 50% vs. 47% (p>0.9), mPFS-9.1mo (95% CI, 1.2–16.1) vs. 9.2mo (95% CI, 2.9–13.3) vs. 7.2mo (95% CI, 2.3-NR) vs. 10.9mo (95% CI, 2.9–20.3) (p>0.8) in groups A, B, B1, and B2, respectively. There were no differences in ORR or mPFS with ICPi between the groups: ORR-0% vs. 27% vs. 33% vs. 25% (p>0.2), mPFS-2.5mo (95% CI, 1.4–3.7) vs. 3.0mo (95% CI, 1.3–10.5) vs. 2.0mo (95% CI, 1.9-NR) vs. 4.5mo (95% CI, 0.3–10.5) (p>0.3) in groups A, B, B1, and B2, respectively. In group A, no responses were seen with PARPi; mPFS with PARPi was 1.8mo (95% CI, 1.8-NR). OSDx was 98.3mo (95% CI, 9.7–98.3) vs. 19.4mo (95% CI, 9.7–47.3) vs. 18.8mo (95% CI, 8.5-NR) vs. 19.5mo (95% CI, 8.3–82.2) in groups A, B, B1, and B2, respectively (p>0.3). Conclusions BAP1-altered MPM, as compared to non-selected MPM, is characterized by similar efficacy of CT and ICPi. Numerically longer OS in BAP1-altered MPM may reflect favorable tumor biology. No responses were observed with PARPi.


INTRODUCTION
Mesothelioma is a rare malignancy that originates from mesothelial cells, mostly from the pleura [malignant pleural mesothelioma (MPM)], with an incidence rate of 1.4 cases per 100,000 population in the United States (1,2). The median OS (mOS) of patients with MPM is in the range of 5.9-12.6 months, 1-year survival is 21-55%, and 5-year survival is about 10% (2)(3)(4)(5).
The recommended 1 st -line regimen for advanced MPM is a combination of cisplatin and pemetrexed (6). The addition of bevacizumab to cisplatin and pemetrexed results in a moderate survival improvement (7), and it is recommended for use in selected patients (8). Carboplatin-pemetrexed is also appropriate when cisplatin is contraindicated (8)(9)(10)(11)(12). This therapy is associated with an objective response rate (ORR) of 34-41%, median progression-free survival (mPFS) of 5.7-9.2 months, and mOS of 12.1-18.8 months (6,7).
Little is known about the value of BAP1 alterations as potential predictive biomarkers and as targets for various systemic treatments in MPM. It has been hypothesized that BAP1-altered MPM are characterized by extremely high sensitivity to platinum-based chemotherapy-similarly to BRCA2-mutant ovarian cancer (41). This hypothesis was based on the assumption that BAP1-altered MPM cells cannot efficiently repair platinum-induced DNA cross-links. It has also been suggested that BAP1-altered MPM might be susceptible to poly (ADP-ribose) polymerase inhibitors (PARPi) (22,42,43). Mesotheliomas harboring BAP1 aberrations are characterized by elevated immune signaling and inflammatory tumor microenvironment (36,44), and, therefore, may predict longterm responses with ICPi.
However, llittle is known regarding the clinical outcomes of the above-mentioned systemic treatments in BAP1-mutant MPM. We conducted this retrospective analysis, aiming to compare the clinical outcomes with platinum/pemetrexed+/ −antiangiogenic drugs, ICPi and PARPi in patients with BAP1altered MPM to outcomes in BAP1-wild type or non-selected MPM patients.  (49,50). PD-L1 tumor proportion score (TPS), which is the percentage of tumor cells showing partial or complete membrane staining, was determined and classified as negative, intermediate, or high (TPS of <1%, 1-49%, and >=50%, respectively) (49). TMB was calculated according to either the FoundationOne ® CDx algorithm (45,47), TEMPUSxT algorithm (46), or GPS Cancer ™ algorithm (48)-according to the next-generation sequencing platform used in each case. MSI status was determined using either the FoundationOne ® CDx algorithm (45,47), or TEMPUSxT algorithm (46), and reported as "MSIhigh" and "MSI-stable." MMR status was assessed by IHC staining for MLH-1 (mouse, clone M1 Ventana), MSH-2 (mouse, G-219-1129 Ventana), MSH-6 (mouse, clone 44 Ventana or SP93 Cell Marque), and PMS-2 (rabbit, EPR 3947 Ventana) proteins on Ventana's BenchMark XT platform (Ventana Medical Systems, Tucson, AZ), and reported as "MMR deficient" or "MMR proficient." ORR, PFS, and OS were assessed and compared for each different systemic treatment modality administered (platinum/ pemetrexed chemotherapy, ICPi). For that purpose, only therapies administered for the treatment of advanced-stage disease were considered (neo-adjuvant/adjuvant systemic therapies were excluded from the analysis); for platinum-based combinations only 1 st line treatments were analyzed (rechallenge with platinum-based combination were not included in the comparison). Additionally, OS since disease diagnosis (OSDx) was compared between the groups. In group A, ORR, PFS, and safety with PARPi were assessed. Univariate analyses of PFS and OS with platinum/pemetrexed+/−antiangiogenic drug, ICPi, and OSDx were performed using Cox proportionalhazards regression model.

Patient Selection and Group Assignment
ORR and PFS were assessed using Modified Revised Response Evaluation Criteria in Solid Tumors (mRECIST) used for response assessment in MPM (51); radiological assessment was done by the investigators. PFS was calculated from treatment initiation until disease progression or death; the outcome was censored if a patient was alive without known progression of the disease at the time of last follow-up. OS was calculated from the day of treatment initiation until death; the outcome was censored if a patient was alive at the time of last follow-up. OSDx was calculated similarly, since MPM diagnosis. Adverse events were graded using Common Terminology Criteria for Adverse Events, version 4.03 (CTCAE, v. 4.03) (52).
The study was conducted in accordance with the principles of good clinical practice, and institutional review board approval was obtained before the study initiation.

Statistical Analysis
The sample size was determined by the available patients meeting the inclusion criteria. The statistical analysis was generated using SAS Software, version 9.4 (53).
Categorical variables were presented by numbers and percentiles, medians and ranges were reported for continuous variables. Fisher's exact test and T-test were used to compare the baseline demographic, clinical, and pathologic characteristics. PFS and OS were assessed by the Kaplan-Meier method, with the log-rank test for the comparison. The Cox proportional-hazards regression model was used for univariate and multivariate OS analysis. All reported p-values are based on two-sided hypothesis tests.

Patient and Tumor Characteristics
Baseline demographic and clinical characteristics of patients are presented in Table 1. The vast majority of patients were midaged males, tumors with epithelioid histology predominated; the majority of tumors were "good-prognosis" by EORTC and CALGB prognostic scoring systems. Patients in group A were younger as compared to patients in group B1 (p-0.02). Patents in group A were more likely to have a family history of malignancy (including breast cancer, gastric cancer, cancer of ovary, lung cancer, colorectal cancer, and CNS tumors), and less likely to have an asbestos exposure-these differences were not statistically significant. Interestingly, 1 patient appearing to have no BAP1 alteration in his tumor (using GPS Cancer ™ ) had a family history of MPM in two brothers. More patients in group A, as opposed to group B, received platinum-based ECOG PS at diagnosis, n (%) chemotherapy as a 1 st -line treatment for advanced disease (p-0.02); as expected, only patients in group A received PARPi. No other imbalances in terms of clinical baseline and treatment characteristics between the groups were observed. From the molecular perspective, the proportion of tumors with TPS ≥50% was higher in group A as opposed to group B (25 vs. 0%, p-0.03); no significant differences in terms of tumor PD-L1 expression between groups A and B1 were observed. None of the 14 tumors tested (8 in group A, and 6 in group B) had an MSI-high or MMR-deficient status. Median TMB was low in either group and comprised three mutations per megabase (mut/ Mb) (range, 0.8-5), 1.5 mut/Mb (range, 0-2), and 1.5 mut/Mb (range, 0-2) in 5, 4, and 4 tumors tested in groups A, B, and B1, respectively, without significant differences between the groups ( Table 1). The BAP1 alterations diagnosed in patients in group A (all predicted to be associated with loss of BAP1 function(were as follows: BAP1 Q684 mutation, BAP1 F660fs*32 mutation, BAP1 loss exons 9-17 mutation, BAP1 E198fs*45 mutation, BAP1 splice site 375 + 2T>C mutation, BAP1 loss exons 13-17, BAP1 loss, BAP1 copy number loss. The co-existing alterations observed in BAP1-altered tumors are listed in Table 2. No matched normal DNA was available for the analysis. Upon our request, in seven out of eight cases included in group A, an additional review of the molecular data has been done by Foundation Medicine, however, no differentiation could be made with regards to somatic or germline nature of the BAP1 alterations ( Table 2) In group A, four patients (50%) had a partial response (PR), three patients (38%) had stable disease (SD), and one patient (12%) had disease progression (PD). In group B, one patient (5%) developed complete response (CR), eight patients (42%) developed PR, seven patients (37%) experienced SD, three patients (16%) had PD, and one patient was not evaluable for response assessment (the treatment was stopped before radiological assessment was done). In group B1, two patients  Figure 1). Median PFS comprised 9.1 months (95% CI, 1.2-16.1), 7.2 months [95% CI, 2.3-not reached (NR)], and 10.9 months (95% CI, 2.9-20.3) in groups A, B1, and B2, respectively (p>0.8 for each comparison; Figure 1).
In the univariate analysis, the only variables which correlated with PFS and OS with 1 st -line platinum/pemetrexed chemotherapy+/− antiangiogenic agents were tumor histology and EORTC risk score ( Table 3). Presence or absence of BAP1 alteration did not affect PFS or OS with 1 st -line platinum/ pemetrexed chemotherapy+/− antiangiogenic agents in a significant manner. In group A, two patients (100%) had PD, and one patient was not evaluable for response assessment. In group B, three patients (27%) achieved PR, four patients (36.5%) experienced SD, four patients (36.5%) had PD, and two patients were not evaluable for response assessment. In group B1, one patient (33%) achieved PR, two patients (67%) had PD, and one patient was not evaluable for response assessment. In group B2, two patients (25%) achieved PR, four patients (50%) experienced SD, two patients (25%) had PD, and one patient was not evaluable for response assessment. ORR with ICPi comprised 0 and 27% in groups A and B, respectively (p-0.28), and 0, 33, and 25% in groups A, B1, and B2, respectively (p-0.35).

ORR, PFS With PARPi
Four patients in group A were treated with PARPi: two patients received veliparib, two patients received olaparib, and one patient was also treated with the combination of carboplatin and olaparib after olaparib failure. Olaparib was administered at a dose of 300 mg #2/d; veliparib was administered at a dose of 200 mg#2/d; the combination included carboplatin AUC-2 weekly and olaparib 300 mg #2/d. No objective responses with PARPi occurred; three patients had PD, and one patient achieved SD for 3.4+ months (SD is ongoing at the time of the analysis). PFS with PARPi comprised 3.4+, 1.9, 1.8, and 1.8 months; median PFS was 1.8 months (95% CI, 1.8-NR). The patient treated with the combination of carboplatin and olaparib after olaparib failure demonstrated PD 1.5 months after the initiation of the combined treatment.
One patient developed grade 3 thrombocytopenia during olaparib treatment; one patient developed grade 3 fatigue and grade 3 anorexia during veliparib treatment. The combined treatment was not associated with any treatment-related adverse events.
In the univariate analysis, the variables which significantly correlated with OSDx were sex, tumor histology, Eastern Cooperative Oncology Group performance status (ECOG PS) at diagnosis, and EORTC prognostic risk score ( Table 3). Presence or absence of BAP1 alteration did not demonstrate a statistically significant correlation with OSDx.

DISCUSSION
To the best of our knowledge, our series represents one of the first one reporting on different systemic treatment outcomes in BAP1-altered MPM patients. According to our observation, 1 st - . Furthermore, the results of univariate analysis of outcomes with platinum/pemetrexed did not indicate that BAP1 alterations have any predictive ability in association with this type of treatment. These observations, overall, do not support the assumption regarding the extreme sensitivity of BAP1-mutant MPM to platinum-based chemotherapy. Importantly, Kumar et al. did not observe an association between the loss of nuclear BAP1 expression and outcomes with platinum-based chemotherapy in MPM either (54). In another retrospective analysis assessing the predictive value of different genomic aberrations in MPM performed by Lo Iacono et al., PIK3CA and TP53 mutations, but not BAP1 mutations, predicted time-to-tumor progression and OS with platinum/ pemetrexed (38). With regards to predictive value of BAP1 aberrations with another chemotherapy regimens, a nonsignificant trend toward improved OS with vinorelbine in MPM with loss of nuclear BAP1 expression was observed by Kumar et al., suggesting a potential modulatory effect of BAP1 on microtubule organization and, as a result, response to vinorelbine (54). In addition, non-functional BAP1 has been associated with resistance to gemcitabine in cell lines (55,56). In our series, the presence of BAP1 aberrations did not seem to modify response to ICPi. Indeed, treatment with ICPi, mainly PD-1/PD-L1 inhibitors, resulted in similar ORR and mPFS of 0 and 27% (p-0.28), and 2.5 months (95% CI, 1.4-3.7) and 3.0 months (95% CI, 1.3-10.5) (p-0.39) in BAP1-altered and nonselected MPM patients, respectively. Again, similar ORR (33%, p-0.35) and mPFS [2.0 months (95% CI, 1.9-NR), p-0.96] with ICPi were seen in wild-type BAP1 tumors. The outcomes in both groups were in line with the reported in the literature with PD-1/ PD-L1 inhibitors in similar clinical scenario (ORR of 9-29%, mPFS of 2.6-4.1 months) (13)(14)(15)(16)(17). Furthermore, univariate analysis of outcomes with ICPi did not demonstrate a correlation between the presence of BAP1 alteration and ICPi efficacy.
The presence of BAP1 aberration in our cohort was associated with numerically longer OS since MPM diagnosis of 98.3 months (95% CI, 9.7-98.3), as compared to 19.4 months (95% CI, 9.7-47.3) in non-selected MPM and 18.8 months (95% CI, 8.5-NR) in wild-type BAP1 MPM-despite the lack of striking intergroup differences in other baseline patient and tumor characteristics. The inter-group differences in proportion of patients receiving platinum/pemetrexed and PARPi are less likely to explain such a big numerical difference in OS. Similar ORR and PFS with platinum/pemetrexed and ICPi, but numerically longer OS since MPM diagnosis in the BAP1altered group, in our opinion, reflected favorable natural history and indolent character of the disease, and not necessarily better responsiveness to systemic treatments. Giving the higher frequency of family history of malignancy in the BAP1-mutant cohort (24), we can hypothesize, that in some patients in our cohort BAP1 mutation reflected germline abnormalities, which are known to be associated with longer OS (39,40). Every attempt to elucidate the nature of the BAP1 genomic alterations in our cohort, unfortunately, was unsuccessful. However, considering the low prevalence of germline BAP1 mutations in MPM [1-7% (25-28)], such an explanation for the longer OS in the BAP1-altered group seems less likely.
In our series, none of the four patients with BAP1-altered MPM demonstrated an objective response with PARPi; three patients demonstrated progressive disease at the first radiological response evaluation. This represents an early lack-of-activity signal of PARPi in the BAP1-altered MPM, however, the response evaluation in our series has been done retrospectively, and no central radiological revision has been performed. PARPi are currently being evaluated in two ongoing phase 2 trials: one trial assessing niraparib in BAP1-altered malignant mesothelioma and other DNA damage response-deficient neoplasms (NCT03207347), and another trial evaluating olaparib in separate cohorts of patients with malignant mesothelioma in accordance with the BAP1 status (NCT03531840); no results have been presented so far. One of the promising biological agents in BAP1-altered MPM is enhancer of zeste homolog 2 inhibitor (EZH2i) tazemetostat. This is based on its selective in vivo activity in BAP1-mutant MPM (57), and positive results obtained in a phase 2 trial conducted in BAP1-altered MPM (the study met its primary end point demonstrating a disease-control rate of 51% at 12 weeks) (58). The role of histone deacetylase inhibitors (HDACi) in MPM with BAP1 alterations warrants further exploration as well. BAP1 downregulation increases the sensitivity to HDACi in mesothelioma cell lines (59), and therefore, it would be important to see whether BAP1 aberrations modulate response to vorinostat in VANTAGE 014 study (60). One of the major limitations of our series, in addition to its retrospective nature, small sample size, and lack of central radiological assessment, is the absence of routine molecular profiling for all MPM patients resulting in a significant chance of contamination of one of the comparator groups (group B, representing the "non-selected" MPM) by the BAP1-altered tumors. Whereas the prevalence of BAP1 somatic alterations (including point mutations, deletions, splice alterations, chromosomal alterations, and copy number loss) in MPM ranges between 20 and 64%-depending on the technology used (22,26,(29)(30)(31)(32)(33)(34)(35)(36)(37)(38), most large series implementing nextgeneration sequencing or Sanger sequencing report on the prevalence of 20-25% (30,34,37). Based on the latter estimation and considering performance of genomic testing for some of the patients in the comparator group, the proportion of BAP1-altered tumors in the comparator group (group B) in our cohort is expected to be around 20-25%. Moreover, the estimation of prevalence of BAP1 somatic alterations of 50% results in the proportion of BAP1-altered tumors in the comparator group (group B) of 43%. Overall, the significant chance of contamination of the "non-selected" comparator group by the BAP1-altered tumors represents an additional important limitation of our analysis. This limitation, at least partially, was addressed by the comparative analysis of BAP1altered and wild-type BAP1 MPM which demonstrated similar outcomes in both groups. Inability to differentiate between the germline and somatic aberrations in the BAP1 gene represents an additional weakness of the analysis. Finally, although patients in the selected cohort were consecutive patients, these appeared to be mainly "good prognosis" by EORTC and CALGB prognostic scoring systems, reflecting the typical patient population treated at tertiary cancer centers. As a result of this selection bias, there is an uncertainty regarding the representability of the selected cohort. Noteworthy, many of the recent clinical trials assessing novel treatment strategies in MPM have selection bias, which misleads the understanding of these novel treatments' efficacy in the real-world setting.
Since loss of BAP1 nuclear staining correlates with BAP1 lossof-function mutations with a sensitivity and specificity of 88 and 97% respectively (61), it would be interesting to assess its predictive value on the clinical outcomes of different systemic treatments. Considering the involvement of several oncological centers, and the absence of formal guidelines for IHC BAP1 assessment, a centralized pathological testing was essential for such analysis. Such a centralized pathological testing, however, was not feasible-which represents an additional important limitation of our analysis.

CONCLUSIONS
According to our retrospective analysis, the presence of BAP1 genomic aberration in MPM does not seem to modulate responses to platinum/pemetrexed or ICPi. Numerically longer OS since diagnosis in BAP1-altered MPM has been observed probably reflecting favorable natural history of this disease subset. In four BAP1-altered MPM patients treated with PARPi no responses have been seen. The clinical research to identify effective biological agents in BAP1-mutant MPM is ongoing.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by 0391-14-RMC. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

ACKNOWLEDGMENTS
The paper was presented as a poster during the WCLC 2019 congress (62). Since this publication, the manuscript was updated, and additional analyses have been done.