Characterizing soluble immune checkpoint molecules and TGF-β1,2,3 in pleural effusion of malignant pleural mesothelioma

The clinical impact of soluble molecules in pleural effusion (PE) is unclear in patients with malignant pleural mesothelioma (MPM). In this single-center, retrospective, observational study, we assessed soluble forms of cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1), and PD-1 ligand 1 (PD-L1) using enzyme-linked immunosorbent assays; three TGF-β isoforms were measured via multiplex assay in PE of patients with fibrinous pleuritis (FP) or MPM, to assess relationships between the levels of six molecules, clinicopathological characteristics, and efficacy of immune checkpoint inhibitors. Soluble forms of CTLA-4, PD-L1, PD-1, TGF-β1, TGF-β2, and TGF-β3 were variably produced in PE of FP (n = 34) and MPM (n = 79); we found significant relationships between the six molecules and clinicopathological features. Although none of the three soluble immune checkpoint molecules showed diagnostic or prognostic effects in patients with MPM, TGF-β2 level in PE is a useful differential diagnostic marker between FP and MPM. Both TGF-β1 and TGF-β3 levels are promising prognostic markers for MPM. Moreover, we found that higher baseline levels of PD-1 soluble forms predicted the response to anti-PD1 monotherapy. Our findings identify novel diagnostic, prognostic, and predictive biomarkers for anti-PD1 therapy in patients with MPM.


Clinical variables
The following patient characteristics were obtained from medical records for analysis: age, sex, histological subtype, clinical stage (cStage), and treatment (surgery, radiotherapy, chemotherapy, and ICIs).The follow-up period was set at a maximum of five years (60 months).Histological subtype and cStage were determined according to the criteria from the World Health Organization 2015 22 and the eighth edition of the TNM stage classification system [23][24][25] , respectively.The last follow-up date was August 31, 2023, and the median length of follow-up for censored cases was 13.4 months (range, 4.7-33.4months).

BAP1 status and tumor infiltrating CD8 T cell (TIL) using IHC
Formalin-fixed and paraffin-embedded blocks of MPM were obtained from the institution archives.IHC for BRCA1 associated protein 1 (BAP1) was performed on surgically removed tissues using an automated immunostainer (Ventana BenchMark GX; Roche Diagnostics, Basel, Switzerland).Briefly, slides were deparaffinized using the EZ buffer (Roche Diagnostics) for 1 min at 75 °C and antigen was retrieved using Cell Conditioning 1 buffer (Roche Diagnostics) at 100 °C for 60 min.In addition, a mouse monoclonal anti-BAP1 antibody (1:50, clone C-4, Santa Cruz Biotechnology, Dallas, TX) was used as the primary antibody at 37 °C for 32 min, followed by secondary antibody and detection using the Ventana UltraView Universal DAB detection kit (Roche Diagnostics) as per manufacturer's instructions.Loss of BAP1 was defined as the absence of BAP1 staining in the MPM cell nuclei.Immunoreactivity for BAP1 (loss or retention) was assessed by experienced pathologists (T.M. or E.I.).Nucleus staining for BAP1 in lymphocytes and vascular endothelial cells was used as an internal positive control for BAP1 expression.IHC for CD8 TIL was performed as double staining with rabbit anti-CD8α IgG antibody (clone D8A8Y, Cell Signaling Technology, Danvers, MA) and mouse anti-EGFR IgG2a antibody (clone A-10, Santa Cruz Biotechnology); the status of CD8 TIL was judged by two scorers (R.O. and T.S.) following previously described protocol 14 .

Statistical analysis
The cutoff values of each parameter were individually determined according to the receiver operating characteristic (ROC) curves for differential diagnosis between FP and MPM or for predicting 2-year survival or response to ICI in cases of MPM.A t-test was performed to compare the levels of the six PE parameters between the FP and MPM groups or the historical subtypes of MPM.A chi-square test was performed to evaluate the relationship between patient characteristics and TIME parameters.Spearman's rank correlation test was performed for correlation analysis between the levels of the six PE parameters.Kaplan-Meier survival analysis was used to determine the association between the status of each parameter and the 5-year OS until death or last follow-up.The significant differences in the 5-year OS between groups were assessed using the log-rank test.Statistical analyses were performed using GraphPad Prism 6.01 (GraphPad Software, La Jolla, CA).Univariate and multivariate analyses were performed with the Cox proportional hazards model to identify independent prognostic factors using the SPSS statistical package (version 17.0; SPSS, Chicago, IL).In all cases, statistical significance was set at p < 0.05.

Ethics approval and consent to participate
This research was approved by the Yamaguchi Ube Medical Center Ethics Committee (No. 28-13, 30-5 and 2020-17), followed the principles of the Declaration of Helsinki, and complied with the relevant guidelines and regulations.The requirement for written informed consent for the use of medical records was waived and an opt-out method on the website was used to obtain patient consent.

Patient cohort
Overall, 113 patients were enrolled and our cohort comprised 34 and 79 patients with FP and MPM, respectively.A definitive pathological diagnosis was performed via pleural biopsy in all patients with MPM at our hospital (72) or elsewhere (7).Patient backgrounds are shown in Table 1.

Relationships between the six PE parameters and BAP1 and CD8 TIL statuses in MPM
Proteomic data from MPM showed that BAP1-deficient tumors were characterized by inflammatory TIME and activation of immune checkpoint mechanisms 26 .Therefore, the relationship between the six PE parameters and the statuses of BAP1 and CD8 TIL in MPM was analyzed.Tissue blocks from 72 patients with MPM were obtained from the institution archives, whereas blocks from the other seven cases could not be assessed because the biopsy was performed outside our hospital.Representative images of BAP1 and CD8 + TILs are shown in Supplementary Figure S3.The BAP1 status was considered a "loss" in 46 cases and "retained" in 26 cases of MPM, whereas the CD8 TILs were negative in 36 cases and positive in 36 cases of MPM.TGF-β 1 in PE of MPM was significantly lower in cases with BAP1 loss tumors (9218 ± 6776 pg/ml) than in cases with BAP1-retained tumors (15,474 ± 15,944 pg/ml, p = 0.023).The sPD-L1 level in PE of MPM was lower in BAP1 loss tumors (208.0 ± 167.4 pg/ml) than in BAP1-retained tumors (338.7 ± 457.4 pg/ml, p = 0.085), although the difference was not significant (Supplementary Fig. S4a).The sCTLA-4 level in PE of MPM was lower in cases without CD8 TILs (10.41 ± 8.390 pg/ml) than in those with CD8 TILs (19.70 ± 25.35 pg/ml.p = 0.041), whereas TGFβ 1 in PE of MPM was higher in cases without CD8 TILs (14,105 ± 14,598 pg/ml) than in those with CD8 TILs (8849 ± 5646 pg/ml, p = 0.048) (Supplementary Fig. S4b).Optimal cutoff values for each parameter in predicting 2-year survival determined using ROC analyses were as follows: 9.56 pg/ml for sCTLA-4, 165.73 pg/ml for sPD-L1, 601.50 pg/ml for sPD-1, 8100.75 pg/ml for TGF-β 1 , 198.28 pg/ml for TGF-β 2 , and 28.56 pg/ml for TGF-β 3 .Both TGF-β 1 (p = 0.008) and TGF-β 3 (p = 0.004) significantly predicted the 2-year survival in patients with MPM, whereas the other four PE parameters did not (Supplementary Fig. S5).Kaplan-Meier survival analysis demonstrated that higher TGF-β 1 (p < 0.0001) and TGF-β 3 (p = 0.028) levels were associated with poorer OS than were lower levels of these molecules, whereas none of the other four PE parameters and statuses of BAP1 and CD8 TIL showed any impact on the 5-year OS (Fig. 2).

Univariate or multivariate analysis of the six PE parameters for 5-year OS
Cox regression analysis was performed to determine the predictive value of clinical variables for the 5-year OS.Prognostic impacts of the three soluble ICMs and three TGF-β isoforms in PE of MPM were assessed independently via univariate and multivariate analyses, with the ICMs and isoforms as confounding factors.The parameter number was set to six, as the multivariate analysis for soluble ICM and TGF-β cohorts included 72 cases with 61 events and 70 cases with 60 events, respectively.The other five parameters were sex 27 , histological subtype 27,28 , and TNM Stage 25 , which were broadly recognized as prognostic factors for MPM, including BAP1 and CD8 TIL statuses.In comparison, both BAP1 and CD8 TIL statuses were undecidable in seven cases, as biopsy was performed outside our hospital, and the TGF-β 2 or TGF-β 3 levels in PE were undetected in the other two cases.Univariate analysis of the soluble ICM cohort showed that histological subtype, cStage, and BAP1 status were significantly associated with 5-year OS, and multivariate analysis showed that histology and cStage were independent poor prognostic factors for 5-year OS in soluble ICM cohorts, whereas none of the soluble ICMs had any impact on 5-year OS (Supplemental Table S3).Conversely, univariate analysis for the TGF-β cohort showed histological subtype, cStage, BAP1 status, TGF-β 1 , and TGF-β 3 to be significantly associated with 5-year OS, and multivariate analysis showed that histology and cStage were independent poor prognostic factors for 5-year OS in TGF-β cohort; both TGF-β 1 and TGF-β 3 were also significantly associated with 5-year OS, while TGF-β 2 did not affect OS (Table 2).

Relationships between statuses of six PE parameters and ICI efficacy in patients with MPM
We determined the predictive impact of these six parameters on ICI efficacy in patients with MPM.In our cohort, seven patients received anti-PD-1 monotherapy, two patients were administered a combination of anti-PD-1 and -CTLA4 therapies, and one patient was treated with a combination of anti-PD-1 therapy and cytotoxic chemotherapy for MPM.The six PE parameters were divided into low or high levels using the ROC curve following the cutoff values for response to anti-PD-1 monotherapy: responder (partial response or stable disease for > 6 months) or non-responder (progressive disease or stable disease for < 6 months) as per modified RECIST criteria 29 (Supplementary Fig. S6).We found that higher sPD-1 in PE was significantly associated with responders in patients treated with anti-PD-1 monotherapy via chi-squared test, although no significant relationship was observed between the other five PE parameters and response to anti-PD-1 monotherapy or all six PE parameters and response to patients treated with ICI (Table 3).

Discussion
PE, which could reflect the TIME of MPM, is easy to collect under local anesthesia.The most popular diagnostic marker for PE in MPM is hyaluronic acid; however, its accuracy is insufficient to provide a definitive diagnosis.
To assess diagnostic markers of MPM in PE, we simultaneously assessed three ICMs and three TGF-β isoforms in PE from FP and MPM.We found that PE in MPM had higher TGF-β 2 , suggesting it is a diagnostic biomarker to distinguish between FP and MPM.A detailed understanding of the TIME in PE of MPM is important.Positive correlations were observed between levels of sCTLA-4 and sPD-L1, sCTLA-4 and sPD-1, and sPD-L1 and sPD-1 in PE of MPM.The Rs was particularly high for the relationship between sCTLA4 and sPD-1.Both CTLA-4 and PD-1 are expressed in exhausted T cells, suggesting that soluble ICMs originate from tumor-infiltrating exhausted T cells.These findings suggest that not only single ICM but multiplex ICMs help tumors to escape host immunity, which is a potential reason for the limited efficacy of ICI therapy in patients with MPM.We assessed only three soluble ICMs; however, the three soluble ICM levels positively correlated with each other, suggesting unexamined ICMs may also be rich in the PE of MPM.If CD8 + TILs are exhausted with higher multiplex soluble forms of ICM in the PE of MPM, ICI monotherapy-mediated activation of antitumor immunity may be difficult with high probability.We found that sCTLA-4 was higher in cases with CD8 + TIL than in those without CD8 + TIL, supporting the hypothesis that CD8 + TILs in the TIME of MPM were exhausted.
We assessed the relationship between clinicopathological features and six parameters in the PE of patients with MPM.We found that BAP1 status was significantly associated with TGF-β 1 levels in PE of MPM and was associated with sPD-L1 via t-test, although not significantly.BAP1 is not only a tumor suppressor gene 30 but also a regulator of proteins involved in DNA damage repair, cell differentiation, cell proliferation, and immune response 31,32 .BAP1 loss reflects BAP1 mutation 33 .Proteomic data from malignant peritoneal mesothelioma showed BAP1-deficient tumors characterized by an inflammatory TIME and immune checkpoint activation 26 .In PE of MPM with BAP1 loss, sPD-L1 was slightly lower than that in PE of MPM with BAP1 expression, suggesting that BAP1 function loss because of gene mutation directly decreased sPD-L1 secreting or indirectly inhibited www.nature.com/scientificreports/membrane binding PD-L1 shedding in PE.TGF-β 1 levels in PE were significantly lower in MPM with BAP1 loss than in MPM with BAP1 expression.This indicates that BAP1 function loss inhibits TGF-β 1 secretion in PE, thus implying that BAP1 alterations affect the EMT and tumor immunoescape 34 , although its clinical significance remains unknown.Another potential reason for the observed significant difference in TGF-β 1 levels between cases with BAP1 loss and retention might be that it is independent of BAP1 function but rather reflects variations in histological subtypes.Notably, BAP1 loss is strongly associated with the epithelioid subtype of MPM, which is significantly correlated with lower TGF-β 1 levels.We also found that TGF-β 1 in PE was higher in cases without CD8 TIL than in those with CD8 TIL.TGF-β 1 is an immune suppressor that causes CD8 T cell exclusion from tumors in TIME 35 .The effects of BAP1 loss on EMT and tumor immune escape should be further investigated.The potential of the six PE parameters in predicting MPM prognosis is worth considering.Although none of the three ICMs affected prognosis in patients with MPM, high TGF-β 1 and TGF-β 3 levels predicted poor prognoses.Our findings suggest that TGF-β may represent a promising target for MPM treatment, as it appears to fulfill dual roles in EMT and immune suppression 19,20 , which contribute to the worsening of MPM prognoses.Functional differences among three TGF-β isoforms should be investigated.TGF-β 1 knockout mice develop a spontaneous fatal inflammatory syndrome until 3-4 weeks of age 36 , whereas both TGF-β 2 and TGF-β 3 knockout mice die as fetuses.TGF-β 2 knockout mice showed developmental defects in multiple organs including cardiovascular, respiratory, and skeletal organs 36 , whereas TGF-β 3 knockout mice exhibited retarded growth in lungs with normal cardiovascular development 37 .Theoretically, TGF-β is a promising target for anti-cancer therapy; however, the pan-TGF-β inhibitor has not been used clinically because of cardiotoxicity in animal models 38 .We showed that the three TGF-β isoforms indicated different consequences in MPM; a selective inhibition strategy for one or two of the three TGF-β isoforms may be feasible for MPM treatment.An antisense oligonucleotide targeting TGF-β 2 inhibits lung metastasis in a preclinical in vivo mouse model 39 , which supports this strategy.The relationship between the six PE parameters and ICI efficacy in patients with MPM was of interest.Although our findings support that a higher baseline status of sPD-1 in PE predicts the response to anti-PD-1 monotherapy in patients with MPM, these findings should be interpreted with caution because of the small sample size.Therefore, further re-evaluation is warranted after accumulating a larger number of cases.
Our study has several strengths.First, this is the first study to evaluate the multiplex soluble forms of ICM and three TGF-β isoforms in PE of MPM simultaneously.Second, to the best of our knowledge, this is the largest study to evaluate sCTLA4, sPD-L1, sPD-1, and TGF-β in PE of MPM.PE sampling can be performed under local anesthesia without hospitalization and provides rapid results compared to pleural biopsy.This method may facilitate earlier diagnosis and prognosis prediction in patients with MPM.However, this study has some limitations.First, the sample size was small compared with that in similar studies on common cancer types; however, obtaining data for many patients with MPM was difficult because of its rarity.Another limitation of our study was the assessment of only six PE molecules despite the presence of numerous soluble factors in PE.Future studies should consider comprehensive PE analysis to identify diagnostic, prognostic, and predictive markers.
In conclusion, we found positive correlations among all combinations of the three soluble ICMs.In addition, positivity for CD8 + TIL in the tumor tissue was associated with higher levels of sCTLA-4 in the PE of MPM.Although none of the three soluble forms of ICMs showed a prognostic impact in patients with MPM, the baseline sPD-1 levels in PE predicted the response to anti-PD-1 monotherapy.Conversely, TGF-β 2 in PE is a differential diagnostic marker between FP and MPM, and TGF-β 1 and TGF-β 3 levels are promising prognostic biomarkers for MPM.

Table 3 .
Relationship between efficacy of ICI and statuses of BAP1, CD8 TIL, and the 6 PE parameters.ICI immuno checkpoint inhibitor, PR partial response, SD stable disease, PD progressive disease.*p < 0.05.