Introduction

Endometrial cancer (EC) is the most frequently diagnosed gynecologic cancer (GC) [1] with peculiar increasing in incidence and mortality [2, 3]. While the prognosis is favorable for early stage, advanced disease displays 5-year survival of 15% (13.2–17.3%) [1]. EC first-line recurrent/advanced standard treatment recommends carboplatin/paclitaxel or hormonal therapy based on histological and clinical features [4, 5]. Nearly 50% of patients progress within a year [6] and there is no efficient second-line chemotherapy or hormone therapy [7]. Recently, four TCGA molecular categories were identified: (1) DNA polymerase epsilon (POLE)/ultramutated, (2) microsatellite unstable (MSI)/hypermutated, (3) low degree of somatic copy-number alteration (SCNA), and (4) high degree SCNA [8, 9] with a gain of ≥ 5 copies or homozygous deletion (< 10 Mb) [10]. POLE/ultramutated and MSI/hypermutated tumors make ECs a reasonable candidate for immunotherapy [11]. The TCGA molecular risk classifier [12, 13] identifies groups with divergent prognoses [14], with (i) POLE mutated and mismatch repair deficient (MMRd), uncommon and with excellent prognosis; (ii) the p53 abnormal group with the poorest prognosis; (iii) no specific molecular profile (NSMP), namely, no MSI or POLE mutations and low SCNA/no p53 mutations, largest group [15], lackings univocal prognostic significance [16, 17]. Several clinical trials evaluated the addition of immunocheck point inhibitors (ICIs) to standard first-line chemotherapy in advanced/recurrent EC [18]. NRG-GY018 study enrolled 816 patients to receive pembrolizumab with paclitaxel plus carboplatin. Consistent reduction in risk of progression (PFS) or death (OS) was detected in pembrolizumab treated, deficient mismatch repair patients (dMMR) (70%) and in the proficient mismatch repair patients pMMR (46%) [19]. The RUBY study evaluated the anti-PD-1, dostarlimab, as maintenance in combination with niraparib, in 607 primary advanced or recurrent ECs. A statistically significant PFS benefit was detected in treated patients with a particular benefit in dMMR subgroups [20, 21]. The AtTEnd study evaluated carboplatin, paclitaxel, and the anti-PD-L1 atezolizumab in 551 advanced or recurrent ECs [22] showing significant survival benefit in the dMMR atezolizumab treated patients [22]. The DUO-E trial [19] evaluated the anti-PD-L1 durvalumab, and olaparib in 718 advanced or recurrent ECs. Similarly to NRG-GY018, in DUO-E improvement in PFS was observed for the durvalumab arm irrespective of MMR status, although the greatest benefit was detected in the dMMR subgroups [19]. In MITO-END3 trial, anti-PD-L1, avelumab was beneficial in MSI-high but detrimental in MSS/TP53-mutated [23].

In EC, the tumor microenvironment (TME) orients toward the suppression as Tregs/CD8+ or Treg cell/CD4+ ratios were significantly higher [24]. Tumors with high mutational load engender stronger immune responses which in turn promote prolonged patient survival. POLE-Ultramutated and MMRd subtypes are enriched for tumor-infiltrating lymphocytes (TIL), although 22% of tumors belonging to these categories display TIL-low. Conversely, p53abn and p53wt are generally TIL-low tumors, yet also contained significant proportions of TIL-high tumors [25]. Additionally, T cell exhaustion markers (PD-1, CTLA4, LAG3, TIM3, TIGIT) might be expressed in “hot” tumors [26,27,28]. Thus, all molecular subtypes can present with immunologically “hot” and “cold” tumors, suggesting that the molecular subtype may not be exhaustive to stratify patients for immunotherapy [25]. Myeloid-derived suppressor cells (MDSC) represent heterogeneous cell population with potent immune suppressive activity [29]. MDSC are immature or early myeloid cells (eMDSC)(Lin-HLA-DR-CD11b+CD33+), monocytic (M-MDSC) (CD11b+CD14+CD15−) and polymorphonuclear (PMN-MDSC), (CD11b+CD14−CD15+) [29, 30] that inhibits T cells activity through arginase-1 (Arg-1), nitric oxide (NO), and reactive oxygen species (ROS). In EC, monocytic-MDSC (M-MDSC) are more represented and highly suppressive. M-MDSC attracted to tumors via CCL2 rapidly differentiate to tumor associated macrophages (TAM) [31,32,33]. In ipilimumab-treated melanoma patients, M-MDSC variation correlates with overall survival [34], and in ECs, MDSCs infiltration associated with unfavorable clinic pathological parameters, bone marrow, and lymph node metastases [35, 36] as well as lymph node premetastatic niche formation [37].

Tregs, CD4+CD25+ lymphocytes [38], comprise three functionally and phenotypically distinct subpopulations: naïve Tregs (CD45RA+  FoxP3lo), activated/effector-type Tregs (CD45RA−FoxP3hi), sharing in vitro suppressive function and non-suppressive T cells, cytokine-secreting (CD45RA−FoxP3lo) (non-Tregs) [39,40,41]. In ECs, Tregs play a controversial role as FoxP3 significantly associated with poorer outcome [42] or with prognostic surrogate of good prognosis [43]. Grading, deep invasion, and FoxP3 or FoxP3/vessel density correlated with [44,45,46] prognosis and EC-MMRd exhibit a significantly higher number of FoxP3+ Treg cells [43, 47].

The recruitment of Tregs cells [48, 49], as for MDSCs from the peripheral blood, has been linked to immunotherapy failure [50, 51]. Thus, with the intent to identify prognostic/predictive biomarkers that could integrate the molecular categories, peripheral MDSC and Tregs were longitudinally evaluated on peripheral blood from a subgroup of patients enrolled in the MITO-END3 trial.

Material and methods

Study design and participants

This study investigated peripheral MDSCs and Tregs in advanced/recurrent EC patients treated with CP/CPA [52]. All participants were Caucasian woman with median age 65.4 years and interquartile range (IQR) from 59 to 71.6 years. Thirty-three MITO-END3 patients (15 received CP and 18 received CPA) were enrolled at Istituto Nazionale Tumori IRCCS “Fondazione G. Pascale” and peripheral blood was collected and analyzed from 29 patients, as single institute analysis. All patients provided informed consent before the initiation of trial procedures. Healthy donors (HD) (n = 23) with median 51 years IQR 43.2–54.7 were analyzed as easy implementable reference value of peripheral MDSC and Tregs (CE 36/22 oss n.1177).

Flow cytometry

EDTA peripheral blood (6 mL) samples were collected in K3 vacutainer tubes. Intracellular staining for FoxP3 was performed with a commercially available kit (BD Transcriptor Factor Buffer Set; BD Pharmingen; RRID:AB_2869424). Tregs were characterized as follows: Horizon-V450 anti-FoxP3 (clone 259D/C7; BD Biosciences Cat# 560044, RRID:AB_1645589), Pe anti-CD25 (clone 2A3; BD Biosciences Cat# 341011, RRID:AB_2783790), Pe-Cy7 anti-CD127 (clone HIL-7R-M21; clone HIL-7R-M21) BD Biosciences Cat# 560822, RRID:AB_2033938), and APC-Cy7 anti-CD4 (clone RPA-T4; BD Biosciences Cat# 557871, RRID:AB_396913) anti-CD45RA BB515 (clone HI100; BD Biosciences Cat# 564552, RRID:AB_2738841) (gate strategy in Supplementary Fig. 1). Tregs could be further differentiated in: activated/effector Tregs (a-Tregs) identified based on FoxP3HI CD45RA−, (CD4+CD25+CD127low FoxP3HICD45RA−); resting or naïve Tregs (n-Tregs) based on FoxP3intCD45RA+ (CD4+CD25+CD127low FoxP3intCD45RA+) and not true or non-Tregs based on FoxP3intCD45RA− (CD4+CD25+CD127low FoxP3intCD45RA−).

MDSCs subsets were characterized as follows: FITC Lineage Cocktail 1 (CD3, CD14, CD16, CD19, CD20, CD56; BD Biosciences Cat# 340546, RRID:AB_400053), PE anti-CD11b (clone Mac-1; BD Biosciences Cat# 557321, RRID:AB_396636), PER-CP anti-CD33 (clone P67.6; BD Biosciences Cat# 341650, RRID:AB_400242), Pe-Cy7 anti-HLA-DR (clone G46-6; BD Biosciences Cat# 555813, RRID:AB_396147), APC anti-CD15 (clone HI98; BD Biosciences Cat# 551376, RRID:AB_398501), and APC-Cy7 anti-CD14 (clone MφP9; (BD Biosciences Cat# 333945, RRID:AB_399960) (gate strategy in Supplementary Fig. 2). CD4-lymphocyte and Tregs were defined as percentages of whole blood, while MDSCs subsets were defined as percentages of CD45 expressing PBMCs. Samples were acquired by BD FACS ARIAIII (BD Biosciences, San Jose, CA, USA). Viability was verified with LIVE/DEAD cell stain (Invitrogen by ThermoFisher, Cat. #L34966). Data were analyzed using BD FACS Diva Software 8.0 cat. 23-14523-00 (BD Bioscience; RRID:SCR_001456) [53].

Fluorescent multiplex immunofluorescence

Primary tumors were evaluated for M-MDSC (CD11b+CD14+) infiltration by multiplex immunofluorescence (mIF). Sections were selected for tumor cellularity (> 70%), absence of large areas of central necrosis or fibrosis. Moreover, when possible, nonneoplastic tissue was included in the sample. M-MDSCs were quantified on the high-resolution and high-contrast images captured by THUNDER Technology (Leica Microsystems), as double-positive, CD14+, and CD11b+ cell membrane expression were considered monocytic (M-MDSC). The Opal mIF protocol (PerkinElmer—Akoya Biosciences Cat# NEL811001KT, RRID:AB_3665660) was applied. Formalin-fixed paraffin-embedded (FFPE) human endometrial cancer tissue blocks were cut at 5 μm. Tissue sections were submerged in citrate buffer (BOND Epitope Retrieval Solution 2, pH 8), and antigen retrieval was performed in a decloaking chamber for 20 min 95 °C (Biocare Medical). Tissue sections were then blocked for 30 min at room temperature in Serum-Free Protein Block (DAKO) and stained with Rabbit recombinant monoclonal antibody. Anti-CD11b antibody [EPR1344] ab133357 (Abcam Cat# ab133357, RRID:AB_2650514) was diluted in antibody diluent (DAKO) 2 h at 25 °C, followed by secondary: HRP-conjugated AKOYA biosciences Opal Polymer HRP Ms + Rb (Akoya Biosciences Cat# ARH1001EA, RRID:AB_2890927) and Opal™ 520 (1:250, PerkinElmer—Akoya Biosciences Cat# NEL811001KT, RRID:AB_3665660). After labeling was complete, antibodies are removed in a manner that does not disrupt the Opal fluorescence signal in a decloaking chamber for 20 min 95 °C with AR6 Buffer AKOYA biosciences (AR600250ML); a second run was conducted with Recombinant CD14 (EPR3653) Rabbit Monoclonal (Cell Marque Cat# 114R, RRID:AB_2827391) for 2 h at 25 °C, followed by secondary: HRP-conjugated IgG Opal Polymer HRP Ms + Rb (Akoya Biosciences Cat# ARH1001EA, RRID:AB_2890927). Substrate: Opal™ 570 (1:250 (PerkinElmer—Akoya Biosciences Cat# NEL811001KT, RRID:AB_3665660). Counterstain: DAPI (blue). Confocal images were acquired on an Leica THUNDER Imager Tissue microscope (RRID:SCR_026034). Multispectral images allow analysis of double-positive cells (positive cells/mm2). M-MDSCs were counted in at least three regions of interest (ROI)/slide, through peritumoral (along the stromal–tumor interface at tumor edge) to intratumoral region (cells totally surrounded the neoplastic cells). M-MDSC were manually enumerated in five consecutive, not-overlapping high-power fields (400× magnification—0.237 mm2/field) for each ROI and expressed in cell/mm2 for slide/patient. Immunostainings were evaluated by three independent observers (C.D., G.S., and C.M.) blinded to patients’ clinical characteristics.

Statistical analyses

Plasma samples were collected at two time points: before starting treatments (B1) and at the end of cycle 6 (B2). The variation in the concentration of each immune cell population between the two time points was calculated as follows: delta, Δ = B1–B2. Continuous variables were described using the median and interquartile range (IQR), while qualitative variables were presented in terms of absolute numbers and proportions. The distribution of all peripheral CD4, MDSCs, and Tregs immune populations was represented using histograms. The normality of the data was assessed using the Shapiro–Wilk test. In case of a violation of the normality assumption, nonparametric tests were employed. For continuous variables, differences between two independent groups were tested using the T test or Mann–Whitney U test, as appropriate, depending on the distribution of the data. To test the association between treatment and the variation of each cell population (Δ), a linear regression model was applied. Immune populations were stratified in four categories: POLE mutated (POLEmut) (category 1); MSI-high (category 2); MSS/TP53 wt; (category 3); MSS/TP53 mut (category 4) according Cancer Genome Atlas (TCGA) and microsatellite instability (MSI) molecular data. Two multivariable regression models were conducted with TCGA and MSI status serving as covariates. In the post hoc analysis, the Mann–Whitney U test was used to compare differences between baseline levels of peripheral cells between EC patients and healthy volunteers. Data were analyzed using R software version 4.2.2 (R Project for Statistical Computing; RRID: SCR_001905). The MITO-END3 trial was based on the hypothesis (tested with 80% power) that adding avelumab to carboplatin/paclitaxel would prolong PFS as the primary endpoint. All the planned analyses related to the translational part of MITO-END3 trial were conducted for descriptive purposes to generate hypotheses and did not include a statistical sample size calculation.

Results

Sample collection

MITO-END3 is a multicenter phase 2 trial that showed improving in PFS in the dMMR population following the addition of avelumab to carboplatin and paclitaxel (CP) in first-line treatment of ECs [52]. To identify possible predictive peripheral biomarkers, a sub-cohort of 29/33 MITO-END3 patients (15 received CP and 18 received CPA) enrolled at Istituto Nazionale Tumori IRCCS “Fondazione G. Pascale” were evaluated. Baseline patient characteristics are shown in Table 1. Peripheral blood samples were collected before starting treatments (B1) and at the end of cycle 6 (B2). B1 and B2 paired samples analysis were available for 27 patients. Distribution of immune cell populations at B1 and B2 was described in Supplementary Fig. 3. Twenty-three/27 patients were characterized for TCGA and MSI status (Fig. 1). Longitudinal variations in peripheral cell population were analyzed (delta, Δ = B1–B2).

Table 1 Baseline patient characteristics
Full size table
Fig. 1
figure 1

Study profile. Blood Tregs and MDSCs were evaluated at B1 (pretreatment) and at B2 (post 6 cycles of CP/ CPA). Flow cytometry identified Monocytic CD14+HLA-DR−/low-(M-MDSC), early stage Lin−HLA-DR−CD33+-(E-MDSC) and polymorphonuclear CD14−CD15+CD11b+ (PMN-MDSC); activated/effector Tregs (a-Tregs) identified based on FoxP3HI CD45RA−, (CD4+CD25+CD127low FoxP3HICD45RA−); resting or naïve Tregs (n-Tregs) based on FoxP3intCD45RA+ (CD4+CD25+CD127low FoxP3intCD45RA+) and not true or non-Tregs based on FoxP3intCD45RA− (CD4+CD25+CD127low FoxP3intCD45RA−) from MITO-END3-EC patients

Full size image

M-MDSCs frequency decreases over treatment in TCGA-category 2 (MSI-high) patients in comparison with TCGA-category 4 (MSS)

ECs patients were stratified according to the TCGA classification [12, 13] (supplementary Tables 1 and 2). The association between TCGA and variations in circulating immune cell populations (Δ) was studied in 23 patients for which B1 and B2 samples were available (Table 2). Over treatment, either CP or CPA, a significant M-MDSCs decrease Δ of 5.41% (95% CI 2.34, 8.50; p = 0.004) was detected in TCGA-category 2 (MSI-high) patients in comparison with TCGA-category 4 (MSS). Thus, TCGA-category 4 ECs displayed a significant increase in M-MDSC Δ of 5.41% (95% CI 2.34, 8.50; p = 0.004) as compared to TCGA-category 2 (Table 2). Accordingly, when patients were stratified based on MSI status, significantly higher baseline frequency (%) of M-MDSCs was detected in 8 MSI-high EC patients (5.40% [3.94;8.70]) as compared to 18 MSS patients (1.19% [0.31; 2.64]) (p = 0.005) (supplementary Table 3). In Table 3, MSS patients had a significant increase Δ of 5.34% (95%CI − 8.11, − 2.56; p = 0.001) in M-MDSC at B2 as compared to MSI-high (Table 3). Tregs frequency significantly increased in CPA as compared to CP patients (0.67% [0.21; 0.87] vs 0.30% [0.18; 0.43], respectively), p = 0.038 (supplementary Table 4). Moreover, Treg subsets [naïve CD45RA+ FoxP3lo–(n-Tregs), activated/effector-CD45RA−FoxP3hi (a-Tregs), and not true Tregs CD45RA− FoxP3lo) (non-Tregs)] were evaluated. A significant increase in naïve-Tregs (n-Tregs) was detected in CPA compared to CP treated patients at B2 (0.07% [0.029; 0.140] vs 0.0150% [0.0007; 0.0300], respectively), p = 0.0044 (supplementary Table 5). Analysis of variations in circulating Treg subsets (Δ) showed modest but significant higher Δ a-Tregs (B1–B2 value) of 0.27% (95%CI − 0.55; − 0.001); p = 0.049) in MSS patients as compared to MSI-high, regardless of treatment type (Table 3).

Table 2 Multivariate regression model for relationships between variation in CD4, Tregs, e-MDSCs, PMN-MDSC, and M-MDSCs and Cancer Genome Atlas (TCGA)-based molecular endometrial cancer subgroups
Full size table
Table 3 Multivariate regression model for variation in CD4, MDSCs (e-MDSCs, PMN-MDSC, and M-MDSCs), and Tregs (resting or naïve Tregs, not true Tregs and a-Tregs, (non-Tregs) MSI status)
Full size table

Higher frequency of tumor infiltrated M-MDSC in MSS versus MSI-high ECs

To evaluate tumoral MDSC infiltration, 19/23 primary EC tissues, subgroup of peripheral analyzed patients molecularly characterized, were subjected to multiplex immunofluorescence for M-MDSC (CD11b+CD14+). Interestingly, although not significant, human CD11b+ and CD14+ stained cells were highly represented in MSS (n = 14) tumors as compared to MSI-high (n = 5). The median cell/mm2 [IQR] was 45.00[33.00, 69.00] in MSS vs 29 [15.00, 45.00] in MSI-high (Fig. 2).

Fig. 2
figure 2

EC patients displayed high tumor-infiltrating M-MDSCs in MSS compared to MSI-high. A Upper panel: Multiplex immunofluorescence (mIF) for M-MDSC (CD11b+CD14+) in primary tumor FFPE sections from EC-MITO-END3 patients. Box and whisker plot. Limits of boxes are 25th and 75th percentiles, with the thick line across the boxes being the median value. Whiskers indicate either the minimum and maximum values or a distance of 1.5 times the interquartile range from the edge of the box (whichever distance is smaller). Lower panel: microphotographs representing two cases of MSS and MSI-high primary tumors (original magnification 400× field); B peripheral pretreatment M-MDSC, (from supplementary Table 3) Box and whisker plot of peripheral

Full size image

Lower CD4+ and higher M-MDSC cells were detected at baseline in peripheral blood of EC patients. Post hoc analysis.

CD4, Tregs, and MDSC were evaluated at baseline in 29 ECs versus 23 HDs. Lower frequency (%) of CD4+ cells (p < 0.005) and higher (%) of M-MDSCs (CD14+HLA-DR−/low-) (p = 0.020) were detected in EC patients as compared to HDs (Supplementary Fig. 4). Tregs or Tregs subset showed no significant variation in distribution between ECs and HDs (Supplementary Fig. 4).

Discussion

In the present manuscript, 29 EC patients participating to the MITO-END3 study were evaluated for the peripheral frequency of MDSCs and Tregs at pretreatment (B1) and after 6 cycle (B2) of assigned treatment, either CP (carboplatin/paclitaxel) or CPA (carboplatin/paclitaxel/avelumab).

At B2, peripheral effector Tregs frequency was significantly higher in patients treated with CPA while both treatments induced significant decrease in peripheral M-MDSC (− 5.41%) in TCGA 2-MSI-high as compared to TCGA-category 4 tumors (p = 0.004). In accordance, both treatments induced M-MDSCs (+ 5.34%) in MSS patients as compared to MSI-high patients (p = 0.001). Moreover, in 19/23 primary tumors higher frequency of M-MDSC was associated with MSS as compared to MSI-high ECs patients. Accordingly, previous data showed MDSC increase in advanced EC correlating with poor response to cytotoxic treatments [37].

In EC patients, MDSC were reported to correlate with peripheral blood MDSC and few CD8+ cells/abundant MDSCs represent independent prognostic factors [54]. MSS tumors are characterized by few TILS as compared to POLE and MSI tumors [55], although tumor/peritumor-infiltrating CD3+ and CD8+ lymphocytes did not significantly associate with objective response, mutational status, TMB, MSI in MITO-END3 study [23]. In the RUBY trial, though pMMR-MSS-ECs are considered poorly responder, dostarlimab benefit was also demonstrated in the pMMR/MSS patients, suggesting that the molecular subtype may be insufficient to stratify patients for immunotherapy [20]. Insights into TME can reveal “hot” tumors in dMMR or MSI-high and “cold” tumors inMMRp or MSS with limited benefit or resistance to ICIs [20].

High frequency of peripheral Tregs, in particular of naïve Treg subset, was detected in avelumab treated EC patients (CPA) at the end of cycle 6 of therapy. This evidence can be interpreted as a general biological effect of avelumab or as prognostic biomarker. Although previous data reported Tregs expansion in peripheral blood after ICI treatment [49], further investigation is needed to understand the interplay between avelumab, Tregs, and treatment outcomes. In preclinical models, tumoral Tregs increased after anti-PD-1(L1) [56,57,58] modifying the cyto/chemokines that regulate Tregs tumor recruitment [59,60,61] and associate with primary or secondary ICIs resistance [48, 50, 62]. Herein, regardless of treatment, increase in activated/effector-type Tregs associated with MSS suggesting multiple immune inhibitory networks in the MSS ECs-TME. Anti PD-(L)1 promoted tumor progression and concomitantly higher frequency of Tregs was reported [63,64,65]. Tiragolumab, an anti-TIGIT antibody Treg inhibitor, combined with atezolizumab demonstrated improved outcomes in NSCLC) [66, 67]. Interestingly, tazemetostat, an EZH2 inhibitor that impairs differentiation of n-Tregs to eTregs [68], is actually evaluated in ARID1A mutated ovarian/endometrial cancer.

Conclusion

Despite limited by patient number and lack of direct correlation between peripheral MDSCs/Tregs and clinical response, this is the first report showing association between longitudinal MDSC/Tregs and molecular characterization. With additional validation, these findings could enrich the selection of patients that may benefit from new treatment combinations.