Excess of blood eosinophils prior to therapy correlates with worse prognosis in mesothelioma

Background Only a fraction of patients with malignant pleural mesothelioma (MPM) will respond to chemo- or immunotherapy. For the majority, the condition will irremediably relapse after 13 to 18 months. In this study, we hypothesized that patients’ outcome could be correlated to their immune cell profile. Focus was given to peripheral blood eosinophils that, paradoxically, can both promote or inhibit tumor growth depending on the cancer type. Methods The characteristics of 242 patients with histologically proven MPM were retrospectively collected in three centers. Characteristics included overall survival (OS), progression-free survival (PFS), overall response rate (ORR) and disease control rate (DCR). The mean absolute eosinophil counts (AEC) were determined by averaging AEC data sets of the last month preceding the administration of chemo- or immunotherapy. Results An optimal cutoff of 220 eosinophils/µL of blood segregated the cohort into two groups with significantly different median OS after chemotherapy (14 and 29 months above and below the threshold, p = 0.0001). The corresponding two-year OS rates were 28% and 55% in the AEC ≥ 220/µL and AEC < 220/µL groups, respectively. Based on shorter median PFS (8 vs 17 months, p < 0.0001) and reduced DCR (55.9% vs 35.2% at 6 months), the response to standard chemotherapy was significantly affected in the AEC ≥ 220/µL subset. Similar conclusions were also drawn from data sets of patients receiving immune checkpoint-based immunotherapy. Conclusion In conclusion, baseline AEC ≥ 220/µL preceding therapy is associated with worse outcome and quicker relapse in MPM.


Introduction
Malignant pleural mesothelioma (MPM) is a cancer associated with very poor prognosis mainly induced by occupational exposure to asbestos fibers (1). Despite the ban or limitation of asbestos use (2), incidence of MPM is still increasing worldwide (3) due to the long latency time between exposure and neoplasm development. There are 3 main histological subtypes of MPM: epithelioid (60-80% of cases), sarcomatoid (< 10%) and biphasic/mixed (10-15%) (4,5). Therapeutic standard options include conventional treatments (surgery, radiotherapy, chemotherapy) and, more recently, immunotherapy (6)(7)(8). Thus, since 2003 the first-line standard-of-care for unresectable MPM has been chemotherapy based on the combination of a DNA cross-linking agent (cisplatin or carboplatin) and an antifolate (pemetrexed) (6). The median overall survival (mOS) obtained with this regimen ranges between 13 and 16 months (6,9). Addition of an anti-VEGF antibody (bevacizumab) to cisplatin/pemetrexed improved mOS up to 18.8 months compared to 16.0 months in the control arm (9). As many MPM patients have a weakened immune system, chemotherapy initially seemed to be a better option than immunotherapy (10). However, the recent first-line dual immunotherapy by immune checkpoint inhibitors (ICIs) (nivolumab and ipilimumab, targeting PD-1 and CTLA-4, respectively) extended mOS from 14.1 months with standard chemotherapy to 18.1 months (11). Immunotherapy has only a limited benefit for the epithelioid subtype but is particularly effective for non-epithelioid MPM (12). Compared with chemotherapy, ICIs clearly provide much better OS rates at 4 years in non-epithelioid MPM (i.e., 14% vs 1%, respectively).
Despite these recent improvements, the prognosis of MPM remains globally poor. The biological mechanisms that drive the effectiveness of available therapies are still not well understood. However, the recent breakthroughs of ICIs indicate that the tumor microenvironment (TME) is a major parameter in cancer development and response to therapy. Even though mesothelioma was initially considered as a "cold" tumor (i.e., absence of T cells within or at the edges of the tumor), the paradigm has recently been revisited (10). In the mesothelioma TME, tumor-associated macrophages (TAMs) are the most abundant immune infiltrating cells (13)(14)(15)(16)(17)(18). The phenotype of these TAMs is shaped by mediators expressed by tumor cells. Therefore, the ability of TAMs to orchestrate the innate immune response and to modulate activation of effector T-cells is impaired in MPM. Among immune cells that regulate macrophage polarization, eosinophils favor the M1 phenotype through the production of IFN-g and TNFa. However, eosinophil-derived IL-4 and IL-13 can also promote suppressive TAMs and shape the TME (19,20). The balance between Th1-and Th2-related cytokines modulates the migration and activation of CD8 + T-cells and affects the local anti-tumor response. Among their pleiotropic activities, eosinophils also promote angiogenesis and tissue healing via VEGF, FGF and PDGF production. Besides their ability to shape the TME through the expression of cytokines, eosinophils display cytotoxic effects by secreting granule proteins and granzyme A.
Altogether, this evidence thus indicates that eosinophils exert both pro-and anti-tumorigenic activities. The final outcome will depend on a variety of parameters that include the cytokine balance, the interaction of eosinophils with other immune cells and the resulting cytotoxicity against the tumor. In this context, we investigated the correlation of blood eosinophil counts with mOS, progression-free survival (PFS) and duration of response in patients undergoing chemo-or immunotherapy.

Materials and methods
Patients' selection and data collection Two hundred and forty-two eligible MPM patients were included in this study. Between January 2009 and December 2021, these patients were given chemo-or immunotherapy in 3 hospitals: 68 at the University Hospital of Liege (Belgium), 61 at the University Hospital of Antwerp (Belgium) and 101 at the University Hospital of Lille (France). According to standard guidelines, 230 patients received cisplatin or carboplatin and pemetrexed as firstline chemotherapy (4,21). Among these, 32 patients also received 2 nd or 3 rd line immunotherapy with nivolumab and ipilimumab. Twelve patients were given ICIs in first-line therapy.
Exclusion criteria included autoimmune disease, congenital or acquired immunodeficiency including HIV, asthma, and active parasitic infection at diagnosis, requiring systemic treatment. Patients diagnosed less than a year before the study was initiated or who did not complete a full treatment plan were also excluded as the follow-up period was too short.
All data were collected for medical purposes and obtained retrospectively. The following data were collected from hospital databases: date of birth; date of diagnosis; sex; histological subtype; BAP-1 deletion; date and type of treatment; response to treatment at 3 months, 6 months and 1 year; hematological lab tests before, during and after treatment; smoking status; diabetes status; asbestos exposure information; comorbidity information; date of death if applicable. Clinical staging was not available for most patients.
This study was performed in compliance with the Helsinki Declaration and was approved by the local Ethics Committee with the reference 2020/45 (University Hospital of Liege) and 2022/1844 (University Hospital of Antwerp) and declared to the local Data Protection Officer (DPO), per General Data Protection Regulation (University Hospital of Lille). As this was a retrospective and noninterventional study, informed consent was not required. Medical records were analyzed pseudonymously.

Outcomes and statistical analysis
Absolute eosinophil counts (AEC) are routinely determined from hemograms collected at presentation. They were retrieved from the available medical records. Optimal AEC cutoff was determined with the X-tile 3.6.1 software (Yale University, New Heaven, CT) and validated by the receiver operating characteristics (ROC) curve. The analysis was based on the mean AEC, averaged during the last month preceding the administration of chemoor immunotherapy.
The primary studied endpoint was mOS, defined as the time from the diagnosis to the date of death due to any cause. Secondary endpoints included PFS, response rate, duration of response and disease control rate. The response was assessed with radiographic tumor assessment according to the modified Response Criteria (mRECIST) [version 1.1] (4,22). PFS was defined as the time between diagnosis and first-documented tumor progression or death due to any cause, whichever came first. Response rate was defined as the best overall response of complete response (CR) or partial response (PR). Duration of response was defined as the time from the first response to the first documented tumor progression or death due to any cause, whichever occurred first. Disease control rate was defined as the best overall response of CR, PR, or stable disease (SD).
Hazard ratios (HRs) and confidence intervals (CIs) of 95% were assessed using an unstratified Cox proportional hazards regression model. Survival curves and rates were estimated with the Kaplan-Meier method and Log-Rank test. Patients with missing values were excluded from the analysis. For statistical purposes, age was categorized as less than 65 years and more or equal to 65 years, whereas subtype was classified as epithelioid and non-epithelioid (i.e., sarcomatoid, biphasic or desmoplastic).
Statistical analysis and graphs were performed by using Prism GraphPad 8 or RStudio 2022.07.1 + 554.

Results
A threshold of AEC at 220/µL splits the cohort into two groups with different overall survival The X-tile software was used to identify the optimal AEC cutoff associated with survival in the cohort of 230 MPM patients receiving first-line chemotherapy. This bioinformatic tool is a graphical method for biomarker assessment and outcome-based cut-point optimization (23). The program provides the optimal division of the data by selecting significant uncorrected p-value and the highest Chi-square. An average AEC was calculated for each patient using the counts of the last month preceding the first administration of chemotherapy. The optimal AEC cutoff determined with the X-tile software was 220 eosinophils/µL of blood (Chi-square = 10.5992, uncorrected p = 0.00113, Figure 1A). This threshold divided the cohort into two groups of 169 (72.40%) and 61 (27.60%) subjects with AEC < 220/µL (in grey) and AEC ≥ 220/µL (in blue), respectively ( Figure 1B). These settings optimally segregated the Kaplan-Meier survival curves of the two subsets ( Figure 1C). The relative risk was estimated by dividing the death incidence corresponding to each AEC by that of the population ( Figure 1D). The AEC 220/µL cutoff classified the patients into two populations with highly significant different distributions (p = 0.0005, Figure 1E). The ROC curve illustrating the true (sensitivity) and false (1-specificity) positive rates validated the cutoff of 220 eosinophils/µL of blood (AUC = 0.6475, p = 0.0006, Figure 1F).
To verify that the measured AEC levels did not result from an increase of all white blood cells, the average absolute counts of other leukocytes were calculated. In populations with AEC < 220/µL and AEC ≥ 220/µL, the absolute counts of lymphocytes, monocytes and neutrophils were similar ( Figure 1G). Since the absolute counts of eosinophils differed significantly (p < 0.0001), it was concluded that high levels of AEC did not result from a general increase of all leukocyte subsets. Furthermore, X-tiles analysis of neutrophils, lymphocytes, monocytes and neutrophil-to-lymphocyte ratio (NLR) did not highlight any threshold or correlation with mOS.
This cut-point selection analysis thus indicated that a threshold of AEC ≥ 220/µL within the last month preceding the first administration of chemotherapy optimally divided the total population into two subsets displaying statistically significant different overall survival times.

Study population
Among the 230 eligible patients treated by chemotherapy, 53 males and 8 females' cases were above the threshold of AEC ≥ 220/ µL ( Table 1). The median age at the time of diagnosis of the patients with AEC < 220/µL and AEC ≥ 220/µL was similar (67 +/-10.4 vs 67 +/-10.9 years, respectively). In both categories, most patients were male (74.0% and 86.9%) and presented an epithelioid subtype of MPM (87.0% and 77.0%). These characteristics were thus representative of typical gender and histologic distributions of MPM (4).
Due to limitations of a retrospective study, only partial information was available for asbestos exposure, Eastern Cooperative Oncology Group (ECOG) performance status prior to chemotherapy, smoking status, diabetes, and BAP-1 expression (Table 1). Prior exposure to asbestos was confirmed in 28.4% and 34.4% of patients with AEC < and ≥ 220/µL, respectively. The proportions of patients presenting different ECOG performance status were similar. OS and AEC/mL were not statistically different in patients with ECOG status 0, 1 and 2 (Supplementary Figure 1). Both active tobacco consumption and diabetes affected a minority of patients. Loss of BAP-1 expression determined by immunohistochemistry was validated in 24.9% of AEC < 220/µL and 18.0% of AEC ≥ 220/µL subsets.
It thus appeared that the two populations split by the AEC 220/µL cutoff shared similar characteristics of age, gender and histological subtype.
AEC ≥ 220/µL is correlated with shorter overall survival Kaplan-Meier analysis showed that patients characterized by AEC ≥ 220/µL during the month preceding their chemotherapy had a highly significant shorter OS compared to subjects with AEC < 220/µL (Figure 2A). The mOS of the 230 individuals enrolled in this study were 14 months and 29 months for AEC above or equal to and below 220/µL, respectively (p = 0.0001, HR of 2.063 [95% CI corresponding to a 2.0-fold improvement in mOS when AEC < 220/ µL. The lower mOS in the AEC ≥ 220/µL subset was observed independently of the histologic subtype ( Figures 2B, C).
Although the proportion of patients with AEC ≥ 220/µL differed in the 3 hospitals (i.e., 17 Figure 3). Altogether, this retrospective observational study thus indicated that MPM patients with AEC ≥ 220/mL had a shorter mOS. AEC superior or equal to 220/mL is associated with earlier relapse The median PFS after chemotherapy was significantly lower in the AEC ≥ 220/mL group compared to the AEC < 220/mL subset (8 months vs 17 months; p < 0.0001, HR 2.589 [1.606 -4.173]) ( Figure 3A). Notably, PFS at 2 years was 13% [4.6 -25.4] vs 42% [33.5 -51.1] in patients with AEC ≥ 220/mL and AEC < 220/mL, respectively. Furthermore, the median time until progression or relapse differed significantly (7 months when AEC ≥ 220/mL vs 16 months when AEC < 220/mL; p = 0.0011, HR 1.950 [1.307 -2.908]) ( Figure 3B). Analysis of this retrospective dataset thus indicated that relapse after chemotherapy occurred more rapidly when AEC ≥ 220/mL. Partial information on response to treatment was available in the retrospective data set (145 and 45 patients in the AEC < 220/mL and AEC ≥ 220/µL groups, respectively (Table 2). Information on response to treatment was missing in 31 patients. A single CR was observed in each category, consistently with other MPM trials (11,24,25). The objective response rate (ORR) combining CR and PR was similar in the 2 subsets (15.9% vs 20.4% at 3 months and 9.0% vs 9.3% at 6 months). In contrast, SD was significantly more common in patients with AEC < 220/mL (56.6%) than in those with AEC ≥ 220/µL (38.9%). This difference in SD was due to a higher proportion of patients with progressive disease (PD) in the AEC ≥ 220/µL subgroup (33.3%, vs 17.2%). Only 33% [21.4 -45.6] of patients with AEC ≥ 220/mL displayed a disease control, including CR, PR and SD of at least 1 year, compared to 53% [43.9 -61.0] in subjects with AEC < 220/mL. This difference was still observed after 2 years (17% in AEC ≥ 220/µL vs 41% in AEC < 220/mL).
Together, these data showed that the AEC cutoff of 220/µL identified groups of patients with different mOS (Figure 2A) and response to chemotherapy ( Figure 3C). The same conclusion was drawn when the study was extended to patients who received immunotherapy (Supplementary Figures 4, 5). Indeed, Kaplan-Meier analysis highlighted that, patients with AEC ≥ 220/µL prior to immunotherapy had a shorter OS (p = 0.0022) and was characterized by a higher proportion of PD (42.9% vs 18.9%) compared with the AEC < 220/µL group.

Discussion
In this report, we showed that patients with an AEC ≥ 220/µL prior to their therapy appear to have a worse outcome and relapse more rapidly. Importantly, we have considered the mean AEC value measured during the month preceding administration of chemo-or immunotherapy. In particular, the disease control rate was improved in chemotherapy-treated patients with AEC < 220/µL and, consistently, the proportion of subjects with a response at two years was increased by 2.4-fold (i.e., 41% vs 17%, Table 2). While the proportion of patients with objective response rate (CR + PR) was similar above and below the threshold of AEC 220/µL, there was a statistically significant difference of SD (Table 2; Supplementary Figures 4C and 5C).
It should be mentioned that, in this study, we excluded patients with hypereosinophilia induced by asthma, allergy, parasitic infection, autoimmune disease, and medication (26, 27). Indeed, these conditions require systemic treatments that would have affected the immune system. It should also be noted that, within the "normal" range (0-450 eosinophils/µL of blood), there is no clear mechanism that explains the fluctuations of eosinophil levels.
In this retrospective study, successive CT evaluations and over time distinguishable tumor margins were often missing. It should however be mentioned that multiple radiographic assessments are particularly challenging in MPM (28). Therefore, OS is preferred and considered to be a more objective and reliable endpoint compared to PFS, response rate and duration of response (11). In this perspective, we showed that the AEC 220/mL threshold predicted a significant difference in mOS (14 vs 29 months in patients treated with chemotherapy and 25 vs 48 months with immunotherapy, Figure 2 and Supplementary Figure 4). The significant association between AEC and OS does not preclude that eosinophilic MPM patients could still respond to chemotherapy or ICIs (29). Consistently, MPM case reports of poor response and fast deterioration have been described in eosinophilic patients (29)(30)(31). If validated by prospective and interventional studies, this conclusion could thus be of particular interest for MPM management.
In fact, the association of AEC and OS has been investigated in other cancers, yielding to opposite conclusions. Indeed, excess of eosinophils in the peripheral blood has been correlated with either a better or a worse prognosis depending on the cancer type (20,32,33). For example, in non-small cell lung cancer (NSCLC) and melanoma, an AEC equal or superior to 300/µL measured before therapy was associated with a better outcome (34)(35)(36)(37)(38)(39)(40)(41)(42)(43). By contrast, the level of peripheral blood eosinophils is an independent prognostic factor for disease progression and disease-specific death in Hodgkin's lymphoma and primary cutaneous T-cell lymphoma (40,(44)(45)(46).
Due to the more recent advent of immunotherapy in MPM, the number of first-line immunotherapy-treated patients included in this study was limited. However, the difference of OS in the AEC ≥ 220/µL and AEC < 220/µL groups was nevertheless statistically significant (L-R test p = 0.0022; Supplementary Figure 4). This conclusion was valid providing that AECs were determined before, but not during or after, the initiation of therapy. In contrast, increase of peripheral blood eosinophils during treatment with ICIs is associated with better response and clinical outcome in NSCLC, indicating that the correlation could be dependent on the tumor type (47,48). Although the biological mechanisms underlying this difference are still not well understood, it is likely that the TME is a central parameter of this cancer specificity. The TME most likely shapes the phenotype of eosinophils into diverse subpopulations with opposite functions, as illustrated in asthma (49-51). In MPM, the interaction of eosinophils with other immune cells such as macrophages, monocytes and neutrophils may direct pro-or anti-tumor functions as well as response to therapy (13)(14)(15)(16)(17)(18). Consistently, inflammation markers such as lymphocyte predominance, NLR and absolute monocyte count (AMC) have been correlated with poor survival (52-57). Analysis of the data set of our cohort did not reveal any association of OS with NLR, AMC and monocyte-to-lymphocyte ratios. Although a causal link still needs to be demonstrated, the correlation between AEC and OS possibly opens direct prospects for therapeutic intervention. Indeed, our report suggests that there might be a benefit to decrease the AEC below the 220/mL threshold before initiating the chemo-or immunotherapy. For example, glucocorticoids (e.g., methylprednisolone) used to prevent pemetrexed-associated rash, emesis and inflammation (58-60) are able to induce apoptosis of eosinophils (61). In our study, a single dose of methylprednisolone at 48mg effectively reduced inflammation but did not reduce myeloid cell counts as numbers remained approximately constant before and after administration. More specific approaches targeting eosinophils have recently been developed in the treatment of asthma (62). Monoclonal antibodies interacting with cytokines associated with eosinophilia (e.g., IL-5, IL-33) are currently evaluated in clinical trials to treat eosinophilic COPD patients: Mepolizumab (anti-IL-5; NCT04075331), MEDI3506 (anti-IL-33; NCT04570657), REGN3500 (anti-IL-33; NCT04701983 and NCT04751487) and Astegolimab (anti-ST2; NCT03615040). Whether these targeted approaches are effective as add-on therapy in MPM could thus merit further evaluation.

Conclusion
In summary, this retrospective study shows that an AEC threshold of 220/µL measured prior to therapy identifies populations with distinct outcomes in mesothelioma, supporting further prospective analysis and possibly interventional trials.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Ethics statement
The studies involving human participants were reviewed and approved by University Hospital of Liège Ethical Committee, reference 2020/45 University Hospital of Antwerp Ethical Committee, reference 2022/1844. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Author contributions
MW, AS and EW collected the dataset in Lille CHU. JR collected the dataset in Antwerp CHU. MW, AF and MH collected the dataset in Liege CHU. AF, AS, EW, HB, JR, MG, LH, MJ, VH, RL participated in data interpretation and manuscript reviewing. MW, LW and MH designed the study. MW and LW drafted the manuscript. All authors contributed to the article and approved the submitted version.

Funding
This work was supported by the Belgian Foundation against Cancer, the Fonds National de la Recherche Scientifique (FNRS), the Teĺevie, and the Fondation León Fredericq.

Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.