Immune cell composition in the endometrium of patients with a complete molar pregnancy: Effects on outcome.

OBJECTIVE
In 15% of patients with complete hydatidiform mole (CHM), disease progresses to post-molar gestational trophoblastic neoplasia (GTN) after curettage. Tumor infiltrating lymphocytes (TILs) are essential in overcoming disease in many tumors. Infiltrating lymphocyte composition and density may influence trophoblast regression and development of post-molar GTN. We analyzed immune cell composition and density in curettaged endometrium of patients with CHM which spontaneously regressed, and of patients with CHM which progressed to post-molar GTN.


METHODS
Sixteen patients with CHM and spontaneous regression, and 16 patients with CHM which progressed to post-molar GTN were selected. Immune cell composition and density of natural killer (NK) cells, natural killer T (NKT)-like cells, Cytotoxic T cells, T-Regulatory and T-Helper cells, were determined by multiplex immunohistochemistry (mIHC).


RESULTS
Curettaged endometrium of patients with CHM and spontaneous regression contained a slightly higher number of immune cells compared to patients with CHM which progressed to post-molar GTN. NKT-like cell density was significantly higher in patients with spontaneous regression compared to patients with CHM which progressed to post-molar GTN (483 ± 296 vs.295 ± 143 (mean ± SD), p = 0.03) respectively. NKT-like cell density in the spontaneous regression group was split in 'high' and 'low' (i.e. above and below the median number of NKT-like cells). In patients with high NKT-like cell density, hCG normalized earlier than in patients with low NKT-like cell density (9.5 weeks, (range 3.7-14) vs. 12.9 weeks, (range 8.6-17.9), p = 0.05).


CONCLUSION
A high number of NKT-like cells in the endometrium of CHMs may contribute to spontaneous regression of molar trophoblast cells.


Introduction
Gestational trophoblastic disease represents a heterogeneous group of pregnancy related disorders originating from placental tissue [1]. The most common form of gestational trophoblastic disease is the hydatidiform mole, which is subdivided in complete (CHM) and partial hydatidiform moles [1,2]. CHMs are diploid and androgenetic, resulting from duplication of the haploid genome of a single sperm, or dispermic fertilization of an ovum. In both cases, maternal DNA is lost, and nuclear DNA is entirely paternal [1]. In approximately 15% of all CHMs, trophoblastic activity after uterine suction curettage persists, a condition referred to as post-molar gestational trophoblastic neoplasia (GTN) [3]. Although much effort has been put into understanding the development from hydatidiform moles to GTN, the underlying mechanism is still unclear [4][5][6].
Immune tolerance towards the trophoblast cells still present after suction curettage could be a contributing factor to the development of post-molar GTN, since studies show that in normal pregnancies immune tolerance is essential [7][8][9]. In normal pregnancies, trophoblast cells expressing both maternal and paternal antigens are tolerated by the surrounding maternal immune system to sustain foetal development [7,8,10]. Trophoblast cells do not express the classical major histocompatibility (MHC) class I molecules human leukocyte antigen (HLA) -A, and \ \B, or MHC class II molecules, which are involved in T cell mediated recognition [9][10][11]. The immune checkpoint programmed death ligand 1 (PD-L1), is highly expressed on the syncitiotrophoblast and functions as an immune suppressant when bound to its receptor programmed death 1 (PD-1) [12]. In mice, a blockade or deficiency of PD-L1 resulted in decreased foetal survival rates [8]. Trophoblast cells of both hydatidiform moles and GTNs highly express PD-L1, suggesting that the PD-1/PD-L1 pathway might be involved in uncontrollable growth of these cells, leading to malignant tissue formation [13][14][15][16][17]. A recent study showed that anti PD-1 therapy in patients with chemo-resistant GTN was successful in three out of four treated patients [16].
In many tumor types, a high number of tumor infiltrating lymphocytes (TILs) is beneficial for clinical outcome of cancer patients. Absence or low density of TILs often results in a poor response to treatment with immune checkpoint inhibitors [16,[18][19][20][21]. Trophoblast cells of CHMs display many features of malignant cells and have the potential to become malignant. Therefore, an inadequate immune response towards trophoblast cells from CHMs, may contribute to progression into postmolar GTN. Since knowledge of immune cell presence in CHMs is limited, the aim of this study was to get insight in immune cell composition and density in the curettaged endometrial tissue of patients with a CHM which spontaneously regressed after suction curettage and of patients with a CHM which progressed to post-molar GTN.

Patients
We selected 16 patients with complete hydatidiform moles which spontaneously regressed after suction curettage and 16 patients with complete moles which progressed to post-molar GTN. Selection was based on the availability of formalin-fixed and paraffin-embedded specimens (FFPEs) of the first curettage at the Radboud university medical center pathology archive. Sixteen samples of patients with a CHM and progression to post-molar GTN were available for further analysis. To achieve an equal amount of cases and controls, 16 samples of patients with a CHM followed by spontaneous regression were included. Follow-up data was retrieved from the Dutch Central Registry for Hydatidiform Moles (DCRHM). Patient data are collected in this voluntary registry since 1977. For patients with spontaneous regression, the minimum follow-up period after normalization was six months. For patients diagnosed with post-molar GTN, follow-up was one year after normalization of hCG levels. No patients of the spontaneous regression or post-molar GTN group showed signs of intra-uterine retention of trophoblast tissue. Patients with a history of auto-immune diseases, recurrent hydatidiform moles or with no treatment data available were excluded. Post-molar GTN was defined according to the FIGO 2000 guideline (i.e., mola hydatidosa with serum hCG plateauing for three consecutive weeks or rising over a period of two consecutive weeks) [22]. In line with the Dutch guidelines, the following criterion was added to this definition: at least one of the values should exceed the 95th percentile of an hCG normogram of uneventful hCG as constructed by Yedema et al. [23]. The local ethical committee of the Radboud University Medical Center approved this study (reference number 2018-4132).
The slides were scanned using the Vectra Automated Quantitative Pathology Imaging System (Version 3.0.4, PerkinElmer). In each sample, the endometrial tissue was selected using Phenochart (Version 1.0.9, PerkinElmer Inc.), for multispectral imaging at 20× magnification, based on HE slides annotated by the gyneco-pathologist. Only endometrial regions were analyzed, since no immune cell infiltration was detected in trophoblast. Spectral unmixing of Opal fluorophores, DAPI, autofluorescence, and downstream image analysis was performed with inForm software (Version 2.2.1, PerkinElmer). The pathologist and researchers were blinded to patient follow-up details during the staining, scanning and analysis period.
A selection of 30-35 representative original multispectral images were used to train the inForm software to distinguish hemorrhagic endometrial tissue from non-hemorrhagic endometrial tissue, trophoblast tissue, and background, based on DAPI and autofluorescence. Settings for adaptive cell segmentation were based on DAPI and membrane signals. All the settings applied to the training images were saved in an algorithm to allow batch analysis of multiple original multispectral images of the same CHM slides.

Image cytometry
Segmented cell data was converted into Flow Cytometry Standard (FCS) files, and analyzed using FlowJo software (v10; Tree Star). Phenotype of the immune cells in non-hemorrhagic endometrial tissue regions were determined by gating cells and divided by surface area of the tissue region (mm 2 ).

Statistical analysis
As data were not normally distributed, we used the Mann-Whitney U test to compare baseline characteristics between the spontaneous regression and post-molar GTN group. Immune cell density of the immune phenotypes was calculated per mm 2 non-hemorrhagic endometrial tissue, and means were compared with the two-sided Students t-test when data was normally distributed after log-transformation. When after log-transformation data was skewed we used the Mann-Whitney U test. A p-value of ≤0.05 was considered significant. In this hypothesis-generating study, we did not correct for multiple testing.

Results
Baseline characteristics of the included patients are displayed in Table 1. There was no difference between both groups in terms of patients' age, and gestational age at diagnosis. Pre-evacuation serum hCG levels tended to be higher in the post-molar GTN group, although not significant (p = 0.06). Samples of the post-molar GTN group contained more hemorrhagic tissue within the imaged tissue areas compared to samples of the spontaneous regression group (Table 1). We selected endometrial tissue regions for multispectral imaging at a 20× magnification ( Fig. 1A and \ \B). Different immune cell populations were detected within samples (Fig. 1C). Non-hemorrhagic tissue could adequately be separated from hemorrhagic tissue and background by tissue segmentation with inForm (Fig. 1D). Only non-hemorrhagic endometrial tissue was included in the analysis. Single cell information was transformed into FCS files and further analyzed using image cytometry (Fig. 2). We calculated immune cell density per mm 2 non-hemorrhagic endometrial tissue in both groups for natural killer (NK) cells (CD56 + ), natural killer T (NKT)-like cells (CD56 + , CD3 + ), total T
In general, the curettaged non-hemorrhagic endometrial tissue of patients with a CHM and spontaneous regression contained a slightly higher number of total CD3 + immune cells compared to the curettaged non-hemorrhagic endometrial tissue of patients with CHM which progressed to post-molar GTN (Fig. 3). Median density of cytotoxic T cells, T helper cells, and T regulatory cells were similar in both groups (Fig. 3). Although not significant, the number of NK cells observed in the spontaneous regression group, tended to be higher than the number of NK cells in the post-molar GTN group. Non-hemorrhagic curettaged endometrial tissue of patients with CHM and spontaneous regression after curettage had a significantly higher number of NKT-like cells per mm 2 compared to endometrial tissue of patients which progressed to post-molar GTN (483 ± 296 versus 295 ± 143 (mean ± SD) NKT-like cells per mm 2 , respectively, p = 0.03), (Fig. 3).
To investigate the relevance of a higher number of NKT-like cells in the curettaged endometrial tissue of CHM patients for clinical outcome, each of the two patient populations (i.e. patients with a CHM which spontaneously regressed after suction curettage, and patients with a CHM which progressed to post-molar GTN), were divided in NKT-like cell density above and below the median density. There was no correlation between pre-evacuation hCG level and NKT-like counts. (Pearson correlation −0.57, p = 0.11). In patients with CHM and spontaneous regression after curettage, who had an NKT-like cell infiltrate in the endometrial tissue above the median, mean time until hCG normalization took 9.5 weeks (±3.0) compared to 12.9 weeks (±3.3) in patients with an NKT-like cell density below the median (p = 0.05), (Fig. 4). In patients with a CHM which progressed to post-molar GTN, and therefore treated with methotrexate (MTX) chemotherapy, mean number of MTX courses needed until hCG normalization was 6.3 (±3.9) in patients with an NKT-like cell density above the median, and 7.4 (±4.6) in patients with an NKT-like cell density below the median, (p = 0.77).

Main findings
Significantly more CD3 + CD56 + NKT-like cells were present in the endometrium of patients with a CHM and spontaneous regression, compared to patients with a CHM which progressed to post-molar GTN (Mean ± SD 483 ± 296 versus 295 ± 143 cells per mm 2 , respectively, p = 0.03). Also, patients in the spontaneous regression group tended to normalize earlier when NKT-like cell infiltrate in the endometrium was above the group median (mean 9.5 weeks, ±3.0) compared to patients with a NKT-like cell infiltrate below the group median (mean 12.9 weeks, ±3.3), p = 0.05. These observations may indicate that NKT-like cells influence spontaneous regression of CHMs after curettage.

Strengths and limitations
This study is the first to use seven-color mICH to simultaneously quantify different immune cells within the curettaged endometrial tissue of patients with a CHM. With this technique, we were able to analyze presence and density of five immune cell phenotypes within one sample, and link trophoblast regression to density of NKT-like cells. Another advantage of this technique is that it offers detailed information regarding distribution of immune cell phenotypes, while maintaining the morphological context of the tissue. Although selection of the analyzed samples was based on presence of endometrial and implantation trophoblastic tissue, retrieval of these samples was through uterine curettage. Therefore, we were unable to apprehend the exact approximity of the endometrial tissue to the trophoblast cells. Although a large part of the analyzed endometrial tissue will be representative of the decidua basalis and capsularis located at the implantation site, a part of the measured immune cell populations might be representative of the decidua parietalis. This however is true for all samples used in this study and may have had an equal influence on both the spontaneous regression, and the post-molar GTN group. Additionally, uterine suction curettage is the least invasive manner to remove a molar pregnancy, and all retrospective and future samples most probably will be collected in this manner. Because post-molar GTN is rare, we were limited by a retrospective design and a relatively small sample size. Differences in absolute NK cell numbers may exist between the spontaneous regression and postmolar GTN group, but our study may have been underpowered to detect this difference. This also may apply for the role of NKT-like cells in the clearance of CHMs in combination with chemotherapy. Unlike most malignant diseases, diagnosis of post-molar GTN is currently based on clinical parameters, i.e. level and course of serum hCG, instead of confirmation by histology [1,22]. To ensure diagnosis of post-molar GTN is not erroneously based on a missed uterine retention, macroscopic complete removal of the molar pregnancy was confirmed during or after the initial uterine curettage.

Interpretation
The most prevalent cell type within the endometrial tissue of CHM patients were the CD56 + NK cells. This cell type is also the most abundant in the decidua of normal pregnancies, representing 70% of the leukocytes present in utero 26 . Of the two subpopulations of NK cells present in normal pregnancies: CD56 dim CD16 + , and CD56 bright CD16 − , the majority of decidual NK cells (dNK cells) consists of the less cytotoxic CD56 brigh t Fig. 1. Multiplex immunohistochemistry of curettaged endometrial tissue of a patient with a CHM. Slides were stained with a lymphocyte panel containing CD3, FOXP3, CD8, CD56, CD45ro, pan-cytokeratin (CK) and DAPI, and scanned with the Vectra 3. A and B) Regions of endometrial tissue were selected for multispectral images (MSIs) at 20× magnification (light-blue rectangles) from the prescans (using DAPI, FITC and Cy5 filters) within Phenochart. From these overviews DAPI (blue), CD3 (green), CK (red) and autofluorescence coming from hemorrhagic tissue areas (green) are mainly visible. Scale bars are 3 and 1 mm, respectively. C) Composite image from unmixed MSIs using inForm Cell Analysis software, visualizing CD3 (red), CD8 (cyan), FOXP3 (green), CD45ro (orange), CD56 (yellow), pan-CK (white) and DAPI (blue) at 20× magnification. Scale bar is 0.1 mm. D) Overview of batch-analyzed tissue segmentation dividing the tissue into "non-hemorrhagic endometrial tissue" (grey) and "hemorrhagic endometrial tissue" (black) regions. Only the areas of non-hemorrhagic endometrial tissue were analyzed. Scale bar is 1 mm. CD16 − phenotype [10]. The origin of these dNK cells is still unclear. Possibly, NK cells are recruited from other organs or from peripheral blood to the decidua, where they undergo tissue-specific differentiation. Another theory is that dNK cells originate from hematopoietic progenitor cells or endometrial NK cells residing in the endometrium, which proliferate and differentiate into dNK cells during early pregnancy [27]. Regardless, dNK cells are thought to play an important role in pregnancies, promoting uterine vascular change, necessary for maternal blood flow through the placenta [26]. In the pre-malignant CHM, the role of dNK cells might be different. Biologically, NK cells respond to tumor formation by killing MHC class I negative cells. Indeed, trophoblast cells lack the classical MHC class I molecules HLA-A and \ \B, they however do express HLA\ \C, HLA-E and HLA-G. HLA-C is the most dominant ligand for killer-cell immunoglobulin-like receptors (KIR), expressed by NK cells. Interaction between these receptors and ligands results in inhibition of cytotoxic functions of NK cells [9,11,28]. NK cell function is furthermore inhibited following interactions with HLA-E and HLA-G [26,29]. Although these interactions might result in NK cell tolerance towards proliferating trophoblast cells, conversely, reduced HLA-G expression was recently linked to monochemotherapy resistance of gestational choriocarcinomas [30]. Therefore, the exact role of NK cells in endometrial tissue of CHM patients remains to be elucidated. CD3 + CD56 + NKT-like cells were the second most prevalent cell type within the endometrial tissue of CHM patients. They represent a heterogeneous group of T cells, consisting of a number of diverse T cell subsets, including classical NKT cells [31][32][33]. In normal pregnancies,~0.5% of the decidual CD3 + T cell population are classical NKT, a frequency 10 times higher than observed in peripheral blood [31]. Classical NKT cells are characterized by the expression of both T and NK cell markers. They differ from conventional T cells in the recognition of antigens [34]. Antigen recognition in NKT cells occurs via lipids presented by CD1d, a nonclassical MHC class-I-like molecule, whereas T cells detect antigens presented by conventional MHC molecules [35]. As CD1d is expressed by villous and extravillous trophoblast cells, NKT cells might play a role in trophoblast regression [31,36]. In mice, activated NKT cells have a dramatic impact on pregnancy, resulting in early or mid-gestation pregnancy loss and pre-term birth [37]. Additionally, NKT cells may be able to modulate uterine NK cell function through transactivation [34]. A study in pregnant mice showed, when NKT cells were activated through administration of αGalCer, that dNK cells became activated and produced IFN-gamma within hours of NKT cell activation. Whether this mechanism is crucial in mediating NKT-cell induced pregnancy loss is unknown. However, it demonstrates that NKT-cell function is capable of modulating dNK cell function, and that NKT cells are important in coordinating functional interactions among decidual leukocytes at the maternal-fetal interface [33,34]. Although due to the design of this study we mainly identified NKT-like cells, from which only a fraction are classical CD1d restricted NKT cells, theoretically, a higher number of NKT-like cells means a higher number of classical NKT cells. However, there are two types of NKT cells which have distinct and opposing roles. Type I NKT cells mainly promote immune reactions against tumor cells and produce cytokines to activate NK and cytotoxic T cells, whereas type II NKT cells are known to inhibit tumor directed immune reactions [33,34]. The immune function of NKT cells, either pro-inflammatory or tolerogenic, depends on tissue type and presence of many other immune cells [33,35,38,39]. Thus, NKT cells present in endometrium of CHMs might be activated or suppressed due to factors derived from trophoblastic cells and surrounding endometrium. Nevertheless, in our study, increased density of NKT-like cells was associated with a greater occurrence of spontaneous regression. We might therefore suggest that a predominant fraction of the NKT-like cells excerted the classical immune stimulating NKT cell function. To further elucidate the underlying biology of NKT cells present in the decidua of patients with trophoblastic disease, future studies should focus on the identification of classical and non-classical NKT cells through the use of αGalcer loaded CD1d multimers. Density of the present classical and non-classical NKT cells may be compared between different types of trophoblastic disease and their relation to pre-evacuation hCG levels, time to normalization, and courses of chemotherapy needed.
Similar to the cellular composition of the decidua during normal pregnancies, approximately 20% of leukocytes present in the curettaged endometrium in this study are CD3 + T cells, and of these CD3 + cells, 70% are CD8 + cytotoxic T cells, and~30% are CD4 + T helper cells [26]. We observed no significant differences in density of these immune cells between the CHM groups. In normal pregnancies, about 5% of the CD4 + are FOXP3 + regulatory T cells [26]. We observed very low numbers of FOXP3 + T regulatory cells within the endometrial tissue of patients in both CHM groups. As chromosomes of CHMs are entirely of paternal origin, a strong maternal immune response against these trophoblast cells would be expected [1]. Nevertheless, trophoblast cells of CHMs are able to proliferate and invade maternal tissue, suggesting a strong local immune tolerance towards these cells. Immune tolerance induced by FOXP3 + T regulatory cells could be a mechanism. However, in line with our results, other studies showed low numbers of FOXP3 + T regulatory cells in the endometrium of CHMs, or observed FOXP3 + T regulatory cells only in association with extremely high numbers of T effector cells [20,40]. Additionally, at the implantation site of normal pregnancies, FOXP3 + T regulatory cells are nearly absent [40]. Therefore, FOXP3 + T regulatory cells may have no primary role in local protection of trophoblast cells in normal-and molar pregnancies.

Conclusion
The expression of immune checkpoint molecules, lack of classical MHC class I and MHC class II complexes, and low density of immune stimulating tumor infiltrating lymphocytes may all contribute to a tolerogenic environment within molar pregnancies and GTN. Conversely, presence of a high number of NKT-like cells may prevent malignant cell formation and contribute to tumor regression. This hypothesis should be tested in additional research with the ultimate aim to identify patients at risk of progression to post-molar GTN at an early stage. The immune microenvironment is an extensive web where the balance between immune stimulating and suppressive factors will probably decide whether a CHM wil regress or progress to post-molar GTN. Although we observed differences in immune cell composition and density between the spontaneous regression and post-molar GTN groups, prediction of persistance based on the number of the analyzed immune cells present solely, is currently not possible.

Contribution to authorship
All authors have been involved in the study design, including the conceptualization and methodology of this study. YMH, MG, and JB collected the data, YMH, MG, and JB, anlyzed and investigated/interpreted the data. FCGJS, PBO, LFAGM and IJMV supervised this process and took care of the funding acquisition. YMH wrote the original draft of the manuscriptorig orand all authors were equally involved in reviewing and editing the manuscript.

Details of ethical approval
This study was approved by the local ethical committee of the Radboud University Medical Center (reference number 2018-4132).

Funding
This study did not receive any specific grant from funding agencies in the public, commercial, or non-for-profit sectors. MG was supported by NWO-Vici 916.14.655 awarded to IJMV.

Declaration of Competing Interest
The authors have nothing to disclose.