Human Reproduction, Vol.32, No.4 pp. 794–810, 2017 Advanced Access publication on February 9, 2017 doi:10.1093/humrep/dex014

ORIGINAL ARTICLE Gynaecology Platelets impair natural killer cell reactivity and function in endometriosis through multiple mechanisms

Yanbo Du, Xishi Liu, and Sun-Wei Guo* Shanghai Obstetrics and Gynecology Hospital, Fudan University Shanghai College of Medicine, 419 Fangxie Road, Shanghai 200011, China

*Correspondence Address. Shanghai Obstetrics and Gynecology Hospital, Fudan University Shanghai College of Medicine, Shanghai 200011, China. Fax: 86-21-6345-5090; E-mail: [email protected]

Submitted on October 28, 2016; resubmitted on December 27, 2016; accepted on January 13, 2017

STUDY QUESTION: Do platelets have any role in the reduced cytotoxicity of natural killer (NK) cells in endometriosis? SUMMARY ANSWER: Platelets impair NK cell reactivity and function in endometriosis through multiple mechanisms. WHAT IS KNOWN ALREADY: Platelets play an important role in the development of endometriosis, and platelet-derived transforming growth factor-β1 (TGF-β1) suppresses the expression of NK Group 2, Member D (NKG2D) on NK cells, resulting in reduced cytotoxicity in women with endometriosis. STUDY DESIGN SIZE, DURATION: Experiments on mice with induced endometriosis in which either platelets, NK cells or both were depleted and controls (none depleted). In vitro experiments with NK cells, platelets and, as target cells, endometriotic epithelial cell and endo- metrial stromal cell lines. PARTICIPANTS/MATERIALS SETTING METHODS: Immunohistochemistry analysis of ectopic endometrial tissues from mice with induced endometriosis receiving either platelet depletion (PD), NK cell depletion, or both or none. Immunofluorescence, flow cytometry and gene expression analysis for major histocompatibility complex class I (MHC-I) expression in target cells. Cytotoxicity and degranulation assays and the measurement of interferon (IFN)-γ secretion for the evaluation of NK cytotoxicity. Flow cytometry and gene expression for the expression of NK cell receptors. MAIN RESULTS AND THE ROLE OF CHANCE: PD resulted in significantly reduced lesion weight in mice with induced endometriosis, but NK cell depletion as well as concomitant platelet and NK cell depletion increased the weight, suggesting that the anti-endometriosis effect of PD is mediated, at least in part, by increased NK cell cytotoxicity against endometriotic cells. Co-incubation of target cells with platelets resulted in rapid platelet coating as well as increased MHC-I expression in these cells, effectively providing a cloak of ‘pseudo-self’ to these cells to shield against NK cell lysis. It also reduced the expression of NKG2D ligands MICA and MICB and reduced the NK cell cytotoxicity. In addition, co-incubation of NK cells with platelets impaired the NK cell cytotoxicity as well. This impaired NK cell cytotoxicity was not due to the increased NK cell apoptosis, but, rather, through reduced NK cell degranulation and IFN-γ production, and reduced expression of activat- ing receptors NKG2D and NKp46 and increased expression of inhibitory receptor KIR2DL1 in NK cells. Inhibition of TGF-β1 signaling par- tially restored the aberrant expression of NKG2D, NKp46 and KIR2DL1, and partially restored the impaired NK cell cytotoxicity induced by activated platelets and their releasate. LARGE SCALE DATA: Not applicable. LIMITATIONS REASONS FOR CAUTION: This study is confined by the limitation of animal and in vitro experimentation and the lack of direct human data. WIDER IMPLICATIONS OF THE FINDINGS: Anti-platelet treatment holds promise in treating endometriosis. STUDY FUNDING/COMPETING INTERESTS: The National Natural Science Foundation of China (81471434 to S.W.G., 81270676 to S.W.G., 81370695 to X.S.L. and 81671436 to X.S.L). None of the authors has anything to disclose.

© The Author 2017. Published by Oxford University Press on behalf of the European Society of Human Reproduction and Embryology. All rights reserved. For permissions, please e-mail: [email protected] Platelets impair natural killer cell function 795

Key words: cytotoxicity / endometriosis / IFN-γ / MHC class I / NK cells / NKG2D / platelet / TGF-β1

also expresses higher MHC-I than that of healthy women (Vernet- Introduction Tomas Mdel et al., 2006). Thus, the aberrant expression of NK cell Characterized by lesions of endometrial-like tissue outside of the uter- KIRs and activating receptors may be responsible for decreased NK ine cavity, endometriosis is an estrogen-dependent and inflammatory cell cytotoxicity in women with endometriosis. Conceivably, the disease associated with pelvic pain and infertility (Johnson et al., 2013). reduced cytotoxicity would diminish the removal of ectopic endomet- Despite extensive research, our understanding of the pathogenesis rial tissues, allowing their survival, implantation and proliferation. As and pathophysiology of endometriosis is still fragmentary (Johnson such, impaired NK cell function may play an essential role in the patho- et al., 2013). It has long been suspected that a deficient immune sys- genesis of the disease (Lebovic et al., 2001). Despite all these findings, tem may be involved (Dmowski et al., 1981; 1989). however, the molecular mechanisms underlying the reported dysregu- As a component of innate immunity, natural killer (NK) cells are a lated NK cell surface receptors are poorly understood, and it is subset of lymphocytes that provide the first-line defense against unclear what the factors are in the peritoneal or sera of women with infected or transformed cells by release of cytotoxic granules and cyto- endometriosis that impair the NK cell cytotoxicity (Thiruchelvam kine secretion without prior sensitization (Smyth et al., 2002; Caligiuri, et al., 2015; Jeung et al., 2016). It is even unclear whether depletion of 2008). The function of NK cells is tightly regulated by a combination of NK cells would facilitate the development of endometriosis. functionally opposing cell surface receptors, inhibitory killer We have recently shown that platelets play important roles in the immunoglobulin-like receptors (KIRs) that bind major histocompatibil- development of endometriosis (Ding et al., 2015; Guo et al., 2015; ity complex class I (MHC-I) molecules and protect ‘self’ and activating Zhang et al., 2015, 2016a,b) and found that endometriotic stromal receptors that bind ligands on transformed cells and eliminate them cells release platelet-activating molecules (Guo et al., 2016b). We (Lanier, 2003). Cells with low or absent expression of MHC-I (‘missing have also reported that platelet-derived transforming growth factor-β1 self’), such as tumor cells, will be identified as ‘non-self’ and targeted (TGF-β1) suppresses the expression of NKG2D on NK cells, resulting by NK cells while harmful or transformed cells expressing stress- in reduced cytotoxicity in women with endometriosis, but neutraliza- induced activating ligands (‘induced self’ or ‘altered self’) will turn on tion of TGF-β1 reverses the reduction (Guo et al., 2016c). However, it and energize NK cell activating receptors, resulting in lysis by NK cells is unclear whether the three entities, i.e. endometriotic cells, NK cells (Vivier et al., 2008; Orr and Lanier, 2010). Both activating and inhibi- and platelets, engage in any cross-talk, and, if so, what the molecular tory receptors can transduce, respectively, positive (‘missing self’ or mechanisms and the consequences of the cross-talk between platelets ‘altered self’) or negative (‘self’) signals to regulate NK cell cytotoxicity and NK cells and between platelets and endometriotic cells might be. and cytokine release that kills the cells deemed to be harmful (Cheent Given the well-documented roles of platelets in modulation of NK cell and Khakoo, 2009). Two checkpoints control NK cell activation: the function in tumorigenesis and metastasis (Nieswandt et al., 1999; first checkpoint is represented by ligands for activating receptors that Palumbo et al., 2007; Kopp et al., 2009; Placke et al., 2012a,b,c), it is are expressed de novo by potentially dangerous cells while the second likely that platelets may also modulate NK cell functions in the devel- involves the MHC-I-specific KIRs that sense the loss of ‘self’ labels on opment of endometriosis. these cells (Moretta et al., 2002). We hypothesized that through interaction with either NK cells or In endometriosis, the impaired NK cell cytotoxicity as well as the target cells, platelets and their secreted soluble factors impair NK decreased NK cells in peritoneal fluid (PF) and sera were noted in early cell cytotoxicity by regulating NK cell surface receptors, reducing inter- 1990s (Oosterlynck et al., 1991, 1992; Tanaka et al., 1992; Garzetti feron (IFN)-γ production and attenuation of NK cell degranulation. In et al., 1993; Kikuchi et al., 1993; Somigliana et al., 1996a, b). Peritoneal addition, since activated platelets express large amount of MHC-I fluid (Oosterlynck et al., 1993) and sera (Kanzaki et al., 1992) from (Chapman et al., 2012), the close proximity of NK cells with endome- patients with endometriosis as well as endometriosis-secreted factors triotic cells following adhesion and aggregation of platelets to ectopic (Hirata et al., 1994; Somigliana et al., 1996a, b) have been reported to implants may transfer MHC-I to endometriotic cells, conferring a reduce NK cell cytotoxicity. In addition, the increased expression of ‘pseudo-self’ phenotype and thus subverting NK cell immunosurveil- some KIRs, such as KIR, two Ig domains and long cytoplasmic tail 1 lance. This study was undertaken to test these hypotheses. (KIR2DL1) (Maeda et al., 2002, 2004; Matsuoka et al., 2005), KIR3DL1 (Wu et al., 2000) and NK Group 2, Member A (NKG2A) Materials and Methods (Galandrini et al., 2008), has also been reported. Moreover, reduced expression of activating receptors, such as NK cell p46-related protein Cells and reagents (NKp46) (Funamizu et al., 2014), has been reported. Consistently, An immortalized endometrial stromal cell line (ESC), established by Krikun HLA-G, a non-classical MHC-I molecule and a ligand for KIR2DL4, has et al. (2004), was kindly provided by Dr Asgi Fazleabas. Following Krikun been reported to be expressed in endometriotic cells (Barrier et al., et al., (2004), cells were maintained in phenol red-free DMEM/F12 2006; Wang et al., 2008), and the role of MHC-I in the impairment of (Gibco, CA, USA) medium supplemented with 10% (v/v) fetal bovine ser- NK cell functions in endometriosis has long been suspected (Semino um (FBS) (Gibco Laboratories, Life Technologies, Grand island, NY, USA), et al., 1995). The eutopic endometrium of women with endometriosis 1 µl/ml sodium pyruvate (Gibco, 11360070) and 100 U/ml Pen/Strep/ 796 Du et al.

Fungizone (Gibco), at 37°C with 5% CO2 in air in a humidified incubator. group and also received i.v. injection of anti-platelet antibody just like the A human endometriotic epithelial cell line (HEEC), 11Z, established by PD group. The choice of timing and dosage of injection was based on the Dr Anna Strazinski-Powitz (Gaetje et al., 1997), was kindly provided by manufactures’ information sheets and our pilot experimentation. Before Dr Jung-Hye Choi of Kyung Hee University, Seoul,Republic of Korea and any procedure was performed, the hotplate test was administrated and cultured in RPMI 1640 medium (Gibco Laboratories) supplemented with the bodyweight was measured for all mice. 5% (v/v) FBS (Gibco Laboratories),100 IU/ml penicillin G, 100 mg/ml Twelve days after induction, the hotplate test was administrated again streptomycin and 2.5 mg/ml Amphotericin B (Hyclone, Utah, USA). Non- and the bodyweight was also measured for all mice. The abdominal cavity immune (NI) IgG (C301, which is isotype-matched with R300) and rat was opened up, and all lesions were excised, carefully weighed and then anti-mouse GPIbα (CD42b) antibody (R300) were purchased from Emfret fixed for immunohistochemistry analysis (see below). The lesion dry Analytics (Eibelstadt, Germany). The anti-ganglio-N-tetraosylceramide weight was determined following Bacci et al., 2009. The schema for the (GM) antibody was from Wako (Duesseldorf, Germany). Anti-mouse pri- design of this protocol is shown in Supplementary Fig. S1. mary antibodies against TGF-β1, proliferating cell nuclear antigen (PCNA), vascular endothelial growth factor (VEGF), CD31 (for the measurement of microvessel density (MVD)), anti-human KIR2DL1-PerCP and vimentin Histology and immunohistochemistry antibodies were purchased from Abcam (Cambridge, MA, USA). The NK For the mouse experiments, serial 4-μm sections of lesions were obtained from Cell Isolation Kit II was purchased from Miltenyi Biotech (Bergisch each formalin-fixed, paraffin-embedded tissue block, with the first resultant slide Gladbach, Germany). Anti-human CD56-FITC, NKp44-PE, KIR3DL1-APC being stained for H&E to confirm pathologic diagnosis and the subsequent slides and MHC-I (HLA-A, -B and -C)-PE antibodies were from Biolegend (San stained for TGF-β1, PCNA, VEGF and MVD. The primary antibodies used for Diego, CA, USA). Anti-human CD3-PE, NKG2D-APC, NKp46-APC, the immunohistochemistry analysis, along with vendor names, catalog numbers CD61-APC and CD62P-FITC were purchased from eBioscience (San and the concentrations, are listed in the Supplementary Table S1.Theproce- Diego, CA, USA). Anti-human pan-cytokeratin and MHC-I antibodies dures and quantification of immunohistochemistry are fully described in the were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Supplementary Information, along with the representative figures for positive Dylight 649 (red) and Dylight 488 (green) were purchased from EarthOx and negative controls (Supplementary Fig. S2). (San Francisco, CA, USA), and Alexa Fluor 555 (yellow) was purchased from Cell Signaling Technology (Boston, MA, USA). The IFN-γ ELISA kit was purchased from Abcam. A83-01, an inhibitor of TGF-β1 Type I recep- Hematoxylin-eosin (H&E) staining and tor (Tojo et al., 2005), was purchased TOCRIS (Missouri, MO, USA). calculation of proportion of epithelium and stroma components in ectopic endometrium Animals and the mouse experiment protocol All lesion samples were evaluated by H&E staining using the H&E staining Sixty-eight 8-week-old female Balb/C mice, 18–20 g in weight, were pur- kit (Sun Biotec, Shanghai, China). Slides were dehydrated, mounted and × fi chased from the SLAC Laboratory Animals (Shanghai, China) and used for observed under 200 magni cation using an Olympus microscope this study. All mice were maintained under controlled conditions with a (Olympus, Tokyo, Japan) and the area of epithelium or stroma of lesions light/dark cycle of 12/12 h and had access to food and water ad libitum in were calculated by the Image-Pro Plus 6.0 following Hull et al., 2012. the Animal Facility at Fudan University Shanghai College of Medicine. All experiments were performed following the guidelines of the National Preparation of platelets and NK cells from Research Council’s Guide for the Care and Use of Laboratory Animals (Council, 1996) and approved by the institutional experimental animals peripheral blood samples review board of Shanghai OB/GYN Hospital, Fudan University. Blood samples were donated by healthy male volunteers who gave written Twenty mice were used for pilot studies of platelet depletion (PD) and informed consent and denied taking any medications for at least 3 weeks NK cell depletion to ensure that the depletion scheduling and doses used prior to the donation, as described previously (Ding et al. 2015). Twenty in the mouse experiment protocol were correct. The procedures and milliliters of citrated blood was first centrifuged at 150g for 10 min, the results are detailed in Supplementary Information. supernatant platelet rich plasma (PRP) was then centrifuged at 300g for After 1 week of acclimatization, but 3 days before the induction of endo- 5 min and the supernatant PRP was harvested again and then centrifuged metriosis (for more details on the procedures of endometriosis induction, at 1000g for 14 min following Merolla et al., 2012. Finally, the depositing see Supplementary Information), all mice, except 16 that were selected platelets were suspended in Dulbecco’s Modified Eagle Media: Nutrient randomly as donors that contributed endometrial tissue fragments, were Mixture F-12 (DMEM/F12) culture medium with 10% (v/v) FBS, for fur- administrated the baseline hotplate test as reported previously and ther use, yielding ~2 × 108 platelets from 20 ml of blood samples. weighted (Lu et al., 2010). The remaining 32 mice were randomly divided Activated platelets (PLT+THB) were obtained by adding 0.075 IU/ml into four groups of equal sizes: (i) control group (NI); (ii) PD group; (iii) thrombin (T8885, Sigma, St. Louis, MO, USA) into the platelets. Then NK cell depletion group (ND) and (iv) concomitant platelet and NK cell platelet suspensions were stirred for 10 min at 200 g and subsequently depletion group (CD). centrifuged at 900g for 15 min. The supernatant was used as platelet relea- Designating Day 0 as the day when the endometriosis-inducing proced- sate (PLT-RLS), and the precipitation was used as platelet pellet (PLT-PT) ure was performed, the PD mice were injected i.v. with 1.5 μg/g body- (Kopp et al., 2009). weight rat anti-mouse GPIbα (CD42b) antibody (R300) in 150 μl sterile Human peripheral blood mononuclear cells (PBMCs) were derived phosphate buffer saline (PBS) to deplete platelets at Days −2, 3, and 8, i.e. from blood samples donated by healthy male volunteers. PBMCs were iso- every 5 days. The NI mice were injected i.v. with 1.5 μg/g bodyweight NI lated by Ficoll-Hypaque density gradient centrifugation (STEMCELL isotype-matched IgG (C301) also in 150 μl sterile PBS at the identical time Technologies, Vancouver, Canada), from which the NK cells were col- points as the PD mice. The ND mice received an i.p. injection of 30 μg/g lected by the NK Cell Isolation Kit II. The purity of the freshly isolated NK − + bodyweight anti-asialo-GM Ab in 150 μl sterile PBS to depletion NK cells cells was ~90% as indicated by CD3 CD56 determined by fluorescence- at Days −3, 0, 3, 6 and 9, respectively, i.e. every 3 days. The CD mice activated cell sorting (FACS) (Becton, Dickinson and Company, Franklin received both i.p. injection of anti-asialo-GM antibody just like the ND Lakes, NJ, USA). Platelets impair natural killer cell function 797

Immunofluorescence (AV+) and late (AV+ and PI+) apoptotic and also viable (AV−/PI−)NK cells were evaluated. The cell culture procedure is described fully in the Supplementary Information. ESCs and HEECs were treated with PBS, platelets (PLT), thrombin (THB) alone or platelets activated by THB (PLT+THB) over- RNA isolation and real-time RT-PCR night, then washed with PBS twice and fixed with 4% formaldehyde. For Target cells (ESCs/HEECs) or NK cells were treated with PBS, PLT, blocking nonspecific binding sites, 5% (v/v) goat serum was used. The indi- PLT+THB, PLT-RLS, PLT-PT or thrombin overnight, then washed with cated primary antibodies were incubated overnight at 4°C. After washing PBS twice. The mRNA abundance of NKG2D, NKp44, NKp46, KIR2DL1 with PBS thrice, appropriate secondary antibodies were added for 60 min and KIR3DL1 on NK cells and of HLA-C, MICA and MICB on target cells at 37°C. After labeling, cells were mounted on slides using VECTASHIELD were evaluated by real-time RT-PCR using SYBR Premix Ex Taq (Takara) with DAPI (Vector Laboratories, Burlingame, CA, USA) and examined using primers listed in Table I. These expression levels of these genes were ΔΔ under a confocal laser scanning microscope (Leica TCS SP5 Confocal determined using the -CT method with GAPDH as the housekeeping Microscope, Solms, Germany) at room temperature. gene (Livak and Schmittgen, 2001).

The effect of suppressing TGF-β1 signaling Cytotoxicity assay on the expression of select NK cell surface NK cell cytotoxicity was determined using the standard lactate dehydro- receptors genase (LDH) release assay (Broussas et al., 2013). The concentration of LDH in the co-culture media was measured by LDH assay kit (Biovision, To investigate the effect of suppressing TGF-β1 signaling on NK cell func- Milpitas, CA, USA). The percentage of lysis was calculated as: (experimen- tion, the expression of the select NK receptors and NK cytotoxicity were tal LDH release − spontaneous target cells LDH release − spontaneous measured after NK cells were treated with the activated platelets or PLT- effector cells LDH release)/(maximum LDH release − spontaneous target RLS with additional 10 μΜ A83-01 or vehicle at 37°C for 8 h, then NK cells LDH release) × 100%, as reported previously (Broussas, Broyer and cytotoxicity was measured as above. The gene expression levels of Goetsch, 2013). NKG2D, NKp44, NKp46, KIR2DL1 and KIR3DL1 were determined by real-time RT-PCR using the same method as described above.

Detection of apoptosis in NK cells Fluorescence-activated cell sorting NK cells (5 × 105) isolated from healthy donors were treated with 20 μM NK cells were treated with PBS, PLT, PLT+THB, PLT-RLS, PLT-PT or of either cyclosprin A (as a positive control), 1640 + 10% (v/v) FBS, plate- THB alone overnight, PBS washed and then re-suspended in 100 μl PBS. let, PLT+THB, PLT-RLS, PLT-PT or THB overnight, and then stained with The cell suspension incubated with either FITC anti-human CD56, PE anti- FITC-labeled annexin V (AV) and propidium iodide (PI; BD Biosciences, human CD3, APC anti-human NKG2D, APC anti-human NKp46, PerCP San Jose, CA, USA) as suggested by the manufacturer and the proportion anti-human KIR2DL1, PE anti-human NKp44 or APC anti-human KIR3DL1 of apoptosis of NK cells was analyzed by FACS. The percentages of early for 30 min in the darkness at 4°C. Then the cells were washed with cold

Table I List of primers used in this study.

Gene name Direction Sequence ...... GAPDH Forward GGG AAG GTG AAG GTC GGA Reverse GGG GTC ATT GAT GGC AAC A NK group 2, member D (NKG2D), also called Forward TCC CTT GAC CGA AAG TTA C Reverse GGC TGG CAT TTT GAG ACA T Natural cytotoxicity triggering receptor 2 (NCR2), coding for NK cell p44-related protein (NKp44) Forward GTG GTA TCT CCA GCC TCT GC Reverse GAG ACT CAG GGG CTT GTC TG Natural cytotoxicity triggering receptor 1 (NCR1), coding for NK cell p46-related protein (NKp46) Forward ACAACCATGCCTGGTCTTTC Reverse AAAAGGTAGGTGCCCCAAGT KIR, two Ig domains and long cytoplasmic tail 1 (KIR2DL1) Forward TCC CTT GAC CGA AAG TTA C Reverse GGC TGG CAT TTT GAG ACA T KIR, three Ig domains and long cytoplasmic tail 1 (KIR3DL1) Forward TTC CAA ACA TAC AAG AGG C Reverse AGT CCG AAG AAA GGT GAA A HLA-C Forward GGACTCACATTCTCCCCAGA Reverse GCGGTGTCGAAATACCTCAT MHC class I polypeptide-related sequence A (MICA) Forward AAGACCAAGACACACTATCACGC Reverse GGTGTCGTGGCTCAAAGATAC MHC class I polypeptide-related sequence B (MICB) Forward CTGATGGGAATGGAACCTACC Reverse GTCTGTCCGTTGACTCTGAAGC

NK, natural killer; MHC, major histocompatibility complex class I; KIR, killer immunoglobulin-like receptors. 798 Du et al.

PBS twice and the percentage of NKG2D, NKp44, NKp46, KIR2DL1 and platelets and NK cells resulted in significantly increased lesion weight − + KIR3DL1 expression was measured on CD3 CD56 gated cells using (P = 0.016; Fig. 1A). In fact, a multiple linear regression analysis with FACS. bodyweight, PD status, NK cell depletion status and the interaction The target cells (ESCs or HEECs) were treated with platelets activated between PD and ND as covariables indicated that while PD reduces by 0.075 IU/ml thrombin for 30 min, then platelet alone, ESCs or HEESs but ND increases the lesion weight (P < 0.001 and P < 0.01, respect- alone and platelet treated target cells were incubated with CD61-APC, ively), concomitant platelet and NK cell depletion increases the lesion CD62P-FITC and MHC-I (HLA-A, -B, -C)-PE 30 min in dark at 4°C, weight (P < 0.001, R2 = 0.77). Of note, PD resulted in an average of respectively, then the cells and platelets were wash with cold PBS twice and the fluorescence intensity of CD61, CD62P and MHC-I were mea- 68% reduction in lesion weight, while ND and CD yielded an increase sured by FACS. of average lesion weight by 67% and 69%, respectively (Fig. 1A). As expected, there was no difference in hotplate latency prior to Degranulation assay the induction of endometriosis as well as prior to any intervention (P > 0.05; Fig. 1B). Twelve days after the induction procedure, how- We performed degranulation assays using CD107a expression as surro- ever, there was a significant reduction in latency (P < 0.01, Fig. 1B). gate marker for granule mobilization. NK cells were incubated with PBS, PLT, PLT + THB, PLT-RLS, PLT-PT or THB alone for 4 h, and then co- A multiple linear regression analysis on the change in latency using cultured with ESCs/HEECs for additional 4 h. The CD107a expression on baseline latency, bodyweight, lesion weight, PD status, NK cell deple- NK cells was determined by FACS. tion status and the interaction between PD and ND as covariables indicated that both the baseline latency and NK cell depletion were sig- ELISA nificantly associated with decrease in latency (both P-values < 0.001; R2 = 0.58). PD also significantly reduced but ND significantly increased The quantification of IFN-γ was done using an ELISA kit (Abcam) following the manufacture’s instruction. After NK cells co-cultured with platelets the stromal versus epithelial ratio in ectopic endometrium (both < overnight, the co-culture media with target cells were collected, and cen- P-values 0.001; Fig. 1C), and joint depletion of platelets and NK cells trifuged at 300g for 10 min, and then the IFN-γ concentration was mea- resulted in significantly increased the ratio (P < 0.01; Fig. 1C). In particular, sured using the kit. The assay was performed in triplicate. The detection the stroma-versus-epithelium ratio was positively correlated with the lesion limit of the IFN-γ kit was 5 pg/ml. The intra-assay coefficient of variation weight (r = 0.72, P < 0.001, both variables were log-transformed; Fig. 1D). was 7.8%. We further performed an immunohistochemistry analysis of PCNA (as a marker for proliferation), VEGF (angiogenesis), CD31 (for the Statistical analysis MVD) and TGF-β1 (Fig. 2). As expected, CD31 staining was seen The comparison of distributions of continuous variables between or mostly in vascular endothelial cells. PCNA staining was seen in cellular among two or more groups was made using the Wilcoxon’s and Kruskal’s nuclei of both the stromal and glandular epithelial cells of the ectopic test, respectively, and the paired Wilcoxon test was used when the lesion. And VEGF staining was seen in vascular endothelial cells and before–after comparison was made for the same group of subjects. mostly glandular epithelial cells of the ectopic lesion, while TGF-β1 Pearson’s or Spearman’s rank correlation coefficient was used when evalu- was seen both the stromal and glandular epithelial cells of the ectopic ating correlations between two variables when both variables were con- lesion, especially in the glandular epithelial cells. tinuous or when at least one variable was ordinal. To evaluate the effect of We found that, for all measurements, PD resulted in significant platelet and/or NK cell depletion and other factors on lesion size and reduction while ND yielded an increase. While non-parametric statis- immunoreactivity measurements, multiple linear regression models were tical analysis indicated that for all measures there was no difference used. For gene expression levels, the relative fold-change of gene expres- – sion was determined by normalizing to GAPDH and calculating the between groups NI and CD (Fig. 2B E), multiple linear regression ana- −ΔΔ fi 2 CT(Livak and Schmittgen, 2001). lyses, which were more ef cient in using all data, indicated that, with P values of <0.05 were considered statistically significant. All computa- the only exception for VEGF, concomitant PD and ND also increased tions were made with R 3.3.1 (R Core Team, 2013)(www.r-project.org). immunoreactivity to PCNA and TGF-β1, and MVD (Fig. 2B–E).

Results Platelet coating causes enhanced MHC-I expression by ESC/HEEC Effect of NK and/or PD on ectopic We co-incubated ESCs or HEEC with either PBS, freshly isolated pla- endometrium telets (PLT), platelets activated by thrombin (PLT + THB), or throm- PD resulted in an average reduction in platelet count by 89%, 71%, bin alone (THB) and then evaluated the interaction, if any, between 63% and 28%, respectively, on Days 1, 2, 3 and 4 from the basal plate- the two by immunofluorescence analysis. We found that platelets, let counts of 299 × 109/l at Day 0 before anti-platelet antibody was especially the activated ones, adhered to or surrounded the ESCs/ administrated to the mice (Supplementary Fig. S3). NK cell depletion HEECs, effectively providing a physical shield (Fig. 3A and B). None of resulted in an average reduction in NK cell count by 61%, 32% and ESCs/HEECs expressed integrin αIIb (CD41), nor did they express 11%, respectively, on Days 1, 2 and 3 (Supplementary Fig. S4). These MHC-I but platelets, especially activated ones, expressed both results provided a justification for the injection scheduling that was (Fig. 3A and B). The coating by activated platelets resulted in ESCs/ used in this experiment. HEEC displaying ‘pseudo-expression’ of CD41 and high-level MHC We found that PD significantly reduced but ND significantly Class I (MHC-I) on the cell surface in all cases (Fig. 3A and B). This increased the lesion weight (both P-values < 0.01; Fig. 1A; indicated that ECS/HEEC cells coated with platelets resulted in sub- Supplementary Fig. S5). Just as suspected, concomitant depletion of stantial MHC-I ‘pseudo-expression’. Platelets impair natural killer cell function 799

Figure 1 (A) Boxplot of lesion weight (g) among different treatment groups. The level of statistical significance of the difference between the group indicated and the NI group is shown in the figure. The P-value of the difference among the four groups is shown in the figure. (B) Boxplots of changes in hotplate latency (in seconds) in different treatment groups evaluated before the induction of endometriosis and before sacrifice. The ‘NS’ and ‘**’ are the statistical significance levels for the difference among the four groups of mice. (C) Boxplot of the ratio of the numbers of stromal cells versus epithelial cells among different treatment groups. The P-value of the difference among the four groups is shown in the figure. The dashed line indicates the median level of all mice. The symbols of statistical significance levels refer to the comparison between the control (i.e. NI) group and the group indi- cated. (D) Scatter plot of the (log−) stroma-versus-epithelium ratio and (log−) lesion weight. The Pearson’s correlation coefficient of the two variables and its statistical significance level are shown in the figure. The dashed line represents the regression line. Group labels: M or NI, Non-immune IgG (mock antibody); P or PD, platelet depletion; N or ND, NK cell depletion; C or CD, concomitant depletion of both platelets and NK cells. Labels of statistical significance: ‘*’, ‘**’ and ‘***’ mean that the P-value is <0.05, 0.01 and 0.001, respectively. NS: statistically not significant (P > 0.05).

We also measured the expression of CD61 (β3 integrin), P-selectin that shields the target cells) and found that in both cases the HLA-C (CD62P) and MHC-I (HLA-A, -B and -C) on platelets, ESCs/HEECs expression levels were significantly increased as compared with con- alone or ESCs/HEECs after co-incubation with activated platelets trols (Fig. 3D). respectively using FACS. As expected, platelets displayed high levels of Taken together, these data demonstrate that endometriotic stromal CD61 and MHC-I while ESCs and HEEC, in contrast, were negative and epithelial cells are rapidly coated by platelets in vitro, resulting in for CD61 and had negligible levels of P-selectin and low level of MHC-I expression of platelet markers and also MHC-I by endometriotic cells. (Fig. 3C). However, after co-incubation with platelets, platelets adhered onto the target cells (ESCs/HEECs) and resulted in the trans- Co-incubation of endometriotic cells with fer of platelet marker expression to the target cells or the ‘pseudo- expression’ of CD61 as well as MHC-I on the cell surface (Fig. 3C). platelets reduces the expression of NKG2D We also measured expression levels of HLA-C on target cells using ligands MICA and MICB real-time PCR. We found that ESCs/HEECs co-incubated with plate- Since the reduced NK cell cytotoxicity in endometriosis is attributable, lets and their releasate/pellets displayed significantly elevated HLA-C at least in part, to the reduced NK cell activating receptor NKG2D expression (Fig. 3D). In addition, since platelet-derived soluble factors, (Guo et al., 2016c), we wondered whether platelets have any effect on such as MHC-I or TGF-β1, may be involved in modulating the NK the expression of its ligands MICA and MICB in the target cells. We cytotoxicity (Guo et al., 2016c), we also evaluated the expression found that the expression levels of MICA/B in both endometriotic levels of HLA-C in ESCs/HEECs co-incubated with platelet releasate stromal and epithelial cells were significantly reduced after co- (PLT-RLS, which subsumes all platelet-derived soluble factors) as well incubation with PLT+THB and, in some cases, PLT, PLT-RLS and PLT- as platelet pellets (PLT-PT, which resembles more a physical cloak PT but not THB (Fig. 4A). 800 Du et al.

Figure 2 (A) Representative immunoreactivity staining of proliferating cell nuclear antigen (PCNA), vascular endothelial growth factor (VEGF), MVD (microvessel density, CD31-labeled endothelial cells) and transforming growth factor-β1(TGF-β1) in ectopic endometrium harvested from different groups of mice. The four different columns represent, respectively, ectopic endometrium from mice receiving a mock antibody (NI), platelet depletion antibody (PD), NK cell depletion antibody (ND), and both PD and NK cell depletion antibodies (CD). Magnification in all figures: ×400. The scale bar represents 125 μm. (B) Boxplots of the staining levels of PCNA, VEGF, CD31-labeled MVD and TGF-β1 among the four groups of mice. The p-value of the difference among the four groups is shown in each figure. The statistical significance levels of the difference between the indicated group and the NI group are also shown in the figure in symbols. The dashed line in each figure represents the overall median of all mice. The symbols of statistical significance levels: ‘*’, ‘**’ and ‘***’ indicate that the P-value is <0.05, 0.01, and 0.001, respectively. NS: statistically not significant (P > 0.05), all by Wilcoxon’stest. Platelets impair natural killer cell function 801

Figure 3 The activation of platelets and the subsequent granule release resulting from the coating of endometrial stromal cells (ESCs) (A) or endometriotic epithelial cells (11Z, human endometriotic epithelial cell line (HEEC)) (B) with activated platelets, leading to pseudo-expression of MHC-I in ESCs or HEECs. (A) Confocal microscopy analysis of ESCs after co-incubation with different conditions, via immunfluorescent staining of ESCs by the vimentin-Alexa488 (green) and of platelets by CD41-Alexa647 (red), as well as MHC-I molecules by MHC-I-Alexa555 (yellow) antibodies. Cell nuclei were counterstained with DAPI (blue). Similarly in (B), but the vimentin antibody was replaced by a pan-cytokeratin antibody conjugated with Alexa488 (green). (C) Flow cytometric analysis of platelets, ESCs/HEECs and both after co-incubation using antibodies against platelet marker (CD61, conjugated with FITC), activated platelet marker (CD62P/P-selectin, conjugated with PECy-5), and MHC-I molecules (conjugated with PE). The curves in red indicate a shift in fluorescent staining intensity after co-incubation with activated platelets. (D) The abundance of HLA-C mRNA in target cells (ESC and HEEC) after different co-incubation condi- tions. Platelets adhered to the surface of target cells, ESCs or HEECs, resulted in increased HLA-C gene expression levels in these target cells. In allexperi- ments, n = 8. *: the difference between the cells indicated and the control cells reached a statistical significant level, i.e. P < 0.05. NS: P > 0.05.

Co-incubation of endometriotic cells resulted in significantly reduced NK cytotoxicity as compared with those with platelets results in reduced NK untreated cells (all P-values < 0.01; Fig. 4C). These results suggest that cell cytotoxicity platelets impair the NK cell cytotoxicity in vitro, likely through reduced NKG2D ligands MICA/B expression in target cells, the pseudo- Given the above results, we evaluated the NK cytotoxicity after co- expression of MHC-I posing as ‘pseudo-self,’ or because of physical incubation of endometriotic cells under different conditions. We shield provided by activated platelets (Fig. 3AandB).Asexpected, found that the co-incubation of HEECs/ESCs with PLT and PLT NK cytotoxicity was decreased by reduced effector:target cell (E:T) + fi THB, but not THB alone, resulted in signi cantly reduced cyto- ratio (Fig. 4C). toxicity in NK cells as compared with those treated with PBS alone (all P-values < 0.01), especially those treated with PLT + THB (all P-values < 0.05; Fig. 4C). In fact, the cytotoxicity against ESCs/ Co-incubation of NK cells with platelets HEECs incubated with PLT + THB was significantly lower than that impairs NK cytotoxicity of those treated with PLT alone (both P-values < 0.001; Fig. 4C). Given impaired NK cell cytotoxicity for endometriotic cells co- Incubation of ESCs/HEECs with either PLT-RLS or PLT-PT also incubated with platelets, we wondered whether the cytotoxicity also 802 Du et al.

Figure 4 The abundance of major histocompatibility complex class I (MHC-I)-related chain (MICA/B) mRNAs in ESCs (A) and HEECs (B) after co- incubation under conditions as indicated. The statistical significance levels for the difference between the co-incubation condition and the controls are indicated by symbols. Natural killer (NK) cell cytotoxicity at different effector-to-target (E:T) ratios (15:1 and 10:1) after ESCs (C) or HEECs (D) were co-incubated with PLT, platelets; PLT+THB, platelets activated by thrombin; THB, thrombin only; PLT-RLS, platelet releasate; PLT-PT, platelet pellets. Symbols of significance levels: *P < 0.05; **P < 0.01; NS: P > 0.05. In both (C) and (D), the comparison was made between the cells co-incubated with phosphate buffer saline (PBS) at the same E:T ratio. In all experiments, n = 8. Platelets impair natural killer cell function 803

Figure 5 (A) NK cell cytotoxicity against different target cells at different effector-to-target (E:T) ratios (10:1 and 15:1) after the NK cells were co- incubated with PLT, platelets; PLT+THB, platelets activated by thrombin; THB, thrombin only; PLT-RLS, platelet releasate; PLT-PT, platelet pellets. The statistical significance levels for the difference between the indicated treatment condition and the control are indicated by symbols. (B) The per- centages of early apoptotic, late apoptotic and viable NK cells after co-incubation with different co-incubation conditions. For all three measurements, the difference between the NK cells in all indicated co-incubation conditions was significantly different (all P-values < 0.05) from those co-incubated with cyclosprin A (as a positive control). Data are presented in mean ± standard deviation. Symbols of significance levels: *P < 0.05; NS: P > 0.05.

is impaired if the effector cells themselves are co-incubated with plate- respectively. Similarly, we found that the NK cytotoxicity against lets. To see this, we co-incubated NK cells with PBS, PLT, PLT+THB, ESCs/HEECs was significantly reduced after co-incubation of NK cells PLT-RLS, PLT-PT or THB overnight, and evaluated the NK cytotox- with PLT, PLT+THB, PLT-RLS, PLT-PT as compared with those trea- icity against ESCs/HEECs at the E:T ratio = 10:1 and 15:1, ted with either PBS or thrombin (all P-value < 0.05, Fig. 5A). 804 Du et al.

Figure 6 (A) Results of the degranulation assays. The percentage of CD107a-positive NK cells after co-incubation under different conditions and then exposed to different target cells ((ESCs) or (HEECs)) are shown. Untreated indicates NK cells co-cultured in the absence of ESCs/HEECs. The statistical comparison was made between NK cells co-incubated with the condition indicated and that with PBS. (B) The concentration of interferon (IFN)-γ in the culture supernatants of NK cells exposed to different target cells (ESCs or HEECs). Untreated: NK cells co-cultured in the absence of ESCs/HEECs; PLT, platelets; PLT+THB, platelets activated by thrombin; THB, thrombin only; PLT-RLS, platelet releasate; PLT-PT, platelet pellets. In all experiments, n = 8. *P < 0.05; NS: P > 0.05.

Co-incubation with platelets does not induce Co-incubation with platelets inhibits NK cell apoptosis in NK cells degranulation Given impaired NK cytotoxicity after co-incubation of either NK cells Degranulation is the first step toward NK cell killing. We carried out with platelets, we asked the question as whether such an impairment was degranulation assays for NK cells that were co-cultured with ESCs/ due to the direct effect of platelets on NK cells since activated platelets HEECs and used counterstaining for CD3 and CD56 to see the dif- are reported to exert cytotoxic effect on red blood cells (Okada et al., ferent cytotoxic lymphocyte subpopulations. The presence of ESCs/ 1992). We thus evaluated whether plateletsortheirreleasateorpellet HEECs substantially induced CD107a expression, a marker for gran- induce apoptosis in NK cells. After NK cells were incubated with PBS, ule mobilization on CD3-CD56+NK cells (Fig. 6A). However, PLT, PLT+THB, PLT-RLS, PLT-PT or THB overnight, we determined the thepresenceofPLT+THB, PLT-RLS and PLT-PT, but not THB percentages of apoptotic NK cells by FACS. We found that none of these (P > 0.05), significantly reduced the CD107a expression levels conditions induced NK cell apoptosis, whereas high percentages of apop- (all P-values < 0.05; Fig. 6A). This indicates that platelets and their totic NK cells were detected in the positive control (Fig. 5B). Thus, the releasate and pellets, but not thrombin, inhibited the degranulation impaired NK cell cytotoxicity modulated by platelets, their releasate or of NK cells. pellets is not attributable to increased apoptosis in NK cells. Platelets impair natural killer cell function 805

Figure 7 (A) Platelets down-regulate the expression of NKG2D and NKp46 but up-regulate the expression of KIR2DL1 in NK cells. NK cells were incubated with PLT, platelets; PLT+THB, platelets activated by thrombin; THB, thrombin only; PLT-RLS, platelet releasate; PLT-PT, platelet pellets overnight, and the expression of the select surface receptors were evaluated by real-time RT-PCR (A) and flow cytometry analysis (B, C). In all experi- ments, n = 8. *P < 0.05; NS: P > 0.05. All comparison was made between the condition of interest and the control group (PBS).

Co-incubation with platelets reduces IFN-γ releasate, platelet pellets or their respective controls. The ESCs or secretion by NK cells HEECs were then added and the culture was carried out for an add- itional 24 h. While NK cells produced very low levels of IFN-γ in the Since the release of IFN-γ is one of the major mechanisms through absence of target cells (Fig. 6B), the presence of either ESCs or HEECs which NK cells exert their cytotoxicity, we evaluated whether platelets significantly increased the IFN-γ production by NK cells, which was and their releasate also reduce NK cell IFN-γ production. NK cells were significantly reduced by the presence of PLT+THB or PLT-RLS (both cultured overnight in the presence or absence of platelets, platelet P-values < 0.05), but not PLT or PLT-PT (both P-values > 0.05, Fig. 6B). 806 Du et al.

Figure 8 (A) The effect of the inhibition of TGF-β1 signaling on the expression of select NK cell surface receptors under different conditions. NK cells were cultured overnight in the presence or absence of the indicated conditions. PLT+THB, platelets activated by thrombin; PLT-RLS, platelet releasate. (B) NK cell cytotoxicity against different target cells at the effector-to-target (E:T) ratio of 15:1 after the NK cells were co-incubation with indicated conditions. N = 8 in all experiments. *P < 0.05.

Platelets impair NK cell surface receptor obtained using FACS analysis. Compared with NK cells incubated with expression controls, the protein expression levels of both NKG2D and NKp46 in NK cells were significantly reduced while that of KIR2DL1 were signifi- To further understand the molecular mechanisms underlying platelet- cantly elevated after incubation with PLT+THB and, to a lesser induced NK cytotoxicity impairment, we investigated whether pre- degree, also with PLT, PLT-RLS and PLT-PT (all P-values < 0.05; incubation of NK cells with PLT, PLT+THB, PLT-RLS and PLT-PT Fig. 7B and C). results in aberrant expression of NK cell surface receptors that have been reported to be involved in endometriosis. NK cells were incu- β bated overnight with PBS, PLT, PLT+THB, PLT-RLS and PLT-PT or Effect of inhibition of TGF- 1 signaling on NK THB, and the expression of NK cell receptors, such as NKG2D, cell receptors expression NKp44, NKp46, KIR2DL1 and KIR3DL1, was determined by real-time Platelet-derived TGF-β1 reduces the expression of NKG2D in NK RT-PCR. We found that, compared with NK cells incubated with PBS, cells in PF from endometriosis patients and is responsible, at least in the gene expression levels of NKG2D, NKp46 (both NK cell activating part, for the reduced NK cell cytotoxicity in endometriosis (Okada receptors) in NK cells were significantly reduced after incubation with et al., 1992). To determine whether the signaling of TGF-β1, which can PLT+THB, PLT-RLS, and, to a lesser degree, with PLT and PLT-PT (all be released copiously by active platelets (Smyth et al., 2009) and is P-values < 0.05; Fig. 6A) while that of KIR2DL1 (an inhibitory recep- activated in endometriosis (Komiyama et al., 2007) and high in concen- tor) were significantly elevated for NK cells incubated with activated tration in serum and PF from women with endometriosis (Pizzo et al., platelets and, to a lesser degree, with PLT-RLS and PLT-PT (all 2002), is the major culprit responsible for aberrant expression of P-values < 0.05; Fig. 7A). No significant difference in expression levels NKG2D, NKp46 and KIR2DL1 in NK cells after incubation with acti- of NKp44 and KIR3DL1 among the different treatment groups was vated platelets and PLT-RLS, we further evaluated the expression found (all P-values > 0.05; data not shown). Similar results were of these receptors after pre-treatment with A83-01, an inhibitor of Platelets impair natural killer cell function 807

activating receptors NKG2D and NKp46 and increased expression of inhibitory receptor KIR2DL1 in NK cells. TGF-β1 neutralization abolishes the aberrant expression of NKG2D, NKp46 and KIR2DL1, and partially restores the impaired NK cell cytotoxicity induced by acti- vated platelets and their releasate. Taken together, these data provide a strong piece of evidence that platelets impair NK cell cytotoxicity in endometriosis through multiple mechanisms and both soluble and membrane-bound factors are required for NK cell evasion of endome- triotic cells. Our data are consistent with the previous report that platelet- derived soluble factors do not induce NK cell death (Kopp et al., 2009). They are also consistent with our previous report that platelet- derived TGF-β1 suppresses the expression of NKG2D on NK cells, Figure 9 Schematic illustration of multiple mechanisms underlying resulting in reduced cytotoxicity in women with endometriosis, but platelet-impaired NK cell function and cytotoxicity in endometriosis. inhibition of TGF-β1 signaling reverses the reduction (Guo et al., Platelet coating provides endometriotic cells a physical shield against 2016c). NK cells as well as increased MHC-I expression, effectively providing Our study demonstrates that depletion of NK cells and platelets a cloak of ‘pseudo-self’ to endometriotic cells to protect against NK resulted in an average of 67% increase and 68% decrease in lesion cell lysis. Activated platelets or their secreted soluble factors such as weight, respectively, in a mouse model of endometriosis. In addition, TGF-β1 induce the expression of NKG2D ligands MICA and MICB and reduce the NK cell cytotoxicity. Platelets and their secreted sol- concomitant depletion of NK cells and platelets also resulted in an uble factors such as TGF-β1 can also have direct effect on NK cells, average of 69% increase in lesion weight. Thus, it highlights the point impairing the cytotoxicity. This impaired NK cell cytotoxicity is not that NK cells provide an effective impedance to the development of due to the increased NK cell apoptosis, but, rather, through reduced endometriosis, but platelets protect endometriotic cells from NK- NK cell degranulation and IFN-γ production, and reduced expression mediated immunosurveillance and removal. This is very similar to the of activating receptors such as NKG2D and NKp46 and increased case of cancer cells (Nieswandt et al., 1999). expression of inhibitory receptor KIR2DL1 in NK cells. Our data provide evidence that this platelet-mediated impairment of NK cell activity is through multiple mechanisms, i.e. reduced expression of NKG2D ligands MICA/B, reduced NK cell degranulation and IFN-γ TGF-β Type I receptor. We found that the inhibition of TGF-β1 signaling production, the provision of a physical cloak, thus yielding a steric hin- with A83-01 significantly attenuated the reduction in gene expression drance of the interaction between NK cells and endometriotic cells, and, of NKG2D and of NKp46 in NK cells co-incubated with PLT-RLS and more importantly, through modulation of the NK cell surface receptors PLT+THB, and also significantly abrogated the increase in KIR2DL1 and hence interfering with the identification and lysis of endometriotic expression when co-incubated with PLT-RLS (all P-values < 0.05; cells by NK cells. By physically cloaking endometriotic cells, platelets may Fig. 8A). While NK cell KIR2DL1 expression was significantly elevated transfer MHC-I antigen to the cells, effectively conferring an immune after co-incubation with activated platelets and TGF-β1 neutralization phenotype of ‘pseudo self’ or decoy to endometriotic cells. In this way, attenuated the increase in gene expression, the reduction did not endometriotic cells evade ‘missing self’ recognition, undermining the NK reach statistical significance (P > 0.05; Fig. 8A). These data indicate cell immunosurveillance. Our data provide evidence that platelets expres- that platelet-derived TGF-β1 is responsible for reduced NKG2D and sing MHC-I coat target cells and induce enhanced MHC-I expression in NKp46 expression and increased KIR2DL1 expression in NK cells. As target cells, effectively providing a decoy cover that allows immune- expected, the inhibition of TGF-β1 signaling partially restored the escaping. Alternatively, platelets may transfer immunosuppressive pro- impairment of NK cell cytotoxicity induced by activated platelets and teins to endometriotic cells as in cancer cells (Placke et al., 2012a,b,c). their releasate (all P-values < 0.05; Fig. 8B). Termed as trogocytosis (Joly and Hudrisier, 2003), such a transfer of immunoregulatory molecules between lymphocytes with other cells has previously been documented (Millet et al.,2008; Qu et al., 2009). Discussion Our data also provide evidence that the impaired NK cell cytotoxicity In this study, we have provided firstly in vivo evidence that anti- in endometriosis is also attributable to platelet-derived TGF-β1, which is endometriosis effect of PD is mediated, at least in part, by increased released copiously upon activation (Smyth et al.,2009). TGF-β1down- NK cell cytotoxicity against endometriotic cells. We also have shown regulates the activating NK receptor NKG2D as shown here and also that platelet coating provides target cells a physical shield against NK reported previously (Guo et al.,2016c) as well as NKp46, which was cells as well as increased MHC-I expression, effectively providing a reported to be down-regulated in endometriosis (Funamizu et al., cloak of ‘pseudo-self’ to target cells to shield against NK cell lysis. Co- 2014). It also up-regulates KIR2DL1 but not KIR3DL1 (Wu et al., 2000) incubation of target cells with platelets reduces the expression of or NKG2A (Galandrini et al.,2008), again as we found also. Hence, pla- NKG2D ligands MICA and MICB and reduces the NK cell cytotoxicity. telets and their interaction with both NK cells and endometriotic cells In addition, co-incubation of NK cells with platelets also impairs the can explain several aberrations in NK cell activity that have been NK cell cytotoxicity. This impaired NK cell cytotoxicity is not due to reported, namely, the increased expression of KIR2DL1 (Maeda et al., the increased NK cell apoptosis, but, rather, through reduced NK cell 2002, 2004; Matsuoka et al.,2005) and reduced expression of NKp46 degranulation and IFN-γ production, and reduced expression of (Funamizu et al.,2014) and of NKG2D (Guo et al.,2016c). In addition, 808 Du et al. both platelet-derived and endometriosis-derived TGF-β1dueto be highly desirable. Second, we used an endometriotic epithelial cell TGF-β1/Smad3 activation by platelets (Zhang et al., 2016)can line and an ESC line as target cells, both of which may not behave exert negatively on NK cell function and should be responsible for exactly the same way as the primary endometriotic epithelial and stro- reported reduced NK cell cytotoxicity resulting from exposure to mal cells. Future studies using primary endometriotic epithelial and PF (Oosterlynck et al., 1993)andsera(Kanzaki et al.,1992)from stromal cells, preferably from women with endometriosis of different patients with endometriosis as well as endometriosis-secreted fac- stages, as target cells may be desirable. In addition, the use of platelets tors (Hirata et al., 1994; Somigliana et al., 1996a,b). Of course, and NK cells derived from the same women would also be ideal. endometriosis-derived ICAM-1 (Vigano et al., 2001)orIL-15(Yu However, since currently the most widely used staging system, i.e. et al.,2016)mightalsoinfluence NK cell reactivity in endometriosis. rASRM classification, only measures the extensiveness of endometri- We summarize our findings in a schematic diagram depicting osis and is by no means synonymous with the stages of the endometri- platelet-mediated impairment of NK cell activity in endometriosis osis development, extra care needs to be taken to determine as what (Fig. 9). constitutes the ‘stage’ of endometriosis. Furthermore, as implicated in Once again, this study underscores the importance of platelets in our data, the NK cells in women with endometriosis may already have the pathophysiology of endometriosis. Platelets and endometriotic impaired functionality and/or cytotoxicity due to the hypercoagulable cells constantly engage cross-talk, ultimately resulting in fibrosis state in women with endometriosis (Wu et al., 2015; Munros et al., (Ding et al.,2015; Guo et al., 2015; Zhang et al.,2015; Guo et al., 2016), as shown in our study. Even platelets in women with endo- 2016a,b,c, Zhang et al., 2016a,b). While endometriosis is widely metriosis might also be hyper-reactive or have an aberrant lipid pro- viewed as an hormonal disease (Bulun et al., 2005)andaninflamma- file, for which little data are available as of now. This reveals just how tory condition (Burney and Giudice), emerging data indicate that little we know about the roles of coagulation signaling pathways in endometriosis also is a hypercoagulable condition (Wu et al.,2015; endometriosis and also highlights the challenges we face in disentan- Munros et al.,2016). As shown previously (Guo et al., 2016c)andin gling the relationship between platelets and NK cells in the develop- this study, platelets also are responsible for impaired NK cell cyto- ment of endometriosis. Further research is warranted to elucidate toxicity in endometriosis. Apart from these, other immunoregula- these issues. tory molecules known to be expressed on platelets upon activation In summary, this study not only elucidates mechanisms by which may also impact negatively on NK cell function and/or reactivity. endometriotic lesions interact with platelets to impair NK cell function For example, glucocorticoid-induced TNF-related ligand (GITRL) is and activity to evade cell immunosurveillance but also demonstrate expressed by platelets upon platelet activation. Like MHC-I, GITRL that platelets, as a newly indicted culprit in the development of endo- might also be rapidly mobilized to the platelet surface and trans- metriosis, are an important player in this estrogen-dependent inflam- ferred to endometriotic cells as in tumor cells, resulting in reduced matory disease and, as such, should be taken into account in strategies NK cell cytotoxicity (Placke et al., 2012a,b,c). In addition, NK cells when devising interventional measures. co-incubated with platelet ectosomes, i.e. vesicles budding from platelet surface, results in reduced expression of several activating surface receptors (NKG2D, NKp30 and DNAM-1) and decreased Supplementary data NK cell function, as measured by CD107a expression and IFN-γ pro- Supplementary data are available at Human Reproduction online. duction (Sadallah et al., 2016). Consequently, anti-platelet therapy may hold promises for treating endometriosis as shown in animal models (Guo et al., 2015, 2016a,b,c). Acknowledgement Our finding that the co-incubation of target cells with platelets We thank the two anonymous reviewers for their constructive com- reduces the expression of NKG2D ligands MICA and MICB may ments on an earlier version of our manuscript. appear to be at odds with a recent report that the concentrations of soluble NKG2DLs MICA/B, and ULBP-2 are elevated in the PF of women with deep infiltrating endometriosis (Gonzalez-Foruria et al., Authors’ roles 2015). Difference in materials aside (ESCs or endometriotic epithelial S.W.G. conceived and designed the study, performed data analysis cells versus PF harvested from women with deep infiltrating endomet- and data interpretation, and drafted the manuscript. Y.D. carried out riosis), we point out that the shed soluble MICA/B (sMICA/B) may all the experimentations, and X.S.L. recruited patients and secured actually reduce the expression of the ligands on target cells and cause their biological samples. All participated in writing the manuscript. the degradation of NKG2D on NK cells, resulting in suppression of NK cell function (Groh et al., 2002). Alternatively, prolonged chronic exposure of NKG2DLs may render NK cells desensitized (Pradeu Funding et al., 2013), again yielding suppressed NK cell function. While we believe that NK cell functions and cytotoxicity are The National Natural Science Foundation of China (81471434 to impaired in women with endometriosis just as the way as we depicted S.W.G., 81270676 to S.W.G., 81370695 to X.S.L. and 81671436 here, caution should be exercised due to several limitations in this to X.S.L). study. First, most data that we presented are in vitro in nature. Due to ethical constraints, the best in vivo data we had are from mouse experi- Conflict of interest mentation, in which the disease may not completely recapitulate its human counterpart. Hence more direct evidence from humans would None declared. Platelets impair natural killer cell function 809

References Guo SW, Du Y, Liu X. Platelet-derived TGF-beta1 mediates the down- modulation of NKG2D expression and may be responsible for impaired Bacci M, Capobianco A, Monno A, Cottone L, Di Puppo F, Camisa B, natural killer (NK) cytotoxicity in women with endometriosis. Hum Mariani M, Brignole C, Ponzoni M, Ferrari S et al. Macrophages are alter- Reprod 2016c;31:1462–1474. natively activated in patients with endometriosis and required for growth Hirata J, Kikuchi Y, Imaizumi E, Tode T, Nagata I. Endometriotic tissues and vascularization of lesions in a mouse model of disease. Am J Pathol produce immunosuppressive factors. Gynecol Obstet Invest 1994;37: 2009;175:547–556. 43–47. Barrier BF, Kendall BS, Ryan CE, Sharpe-Timms KL. HLA-G is expressed Hull ML, Johan MZ, Hodge WL, Robertson SA, Ingman WV. Host-derived by the glandular epithelium of peritoneal endometriosis but not in euto- TGFB1 deficiency suppresses lesion development in a mouse model of pic endometrium. Hum Reprod 2006;21:864–869. endometriosis. Am J Pathol 2012;180:880–887. Broussas M, Broyer L, Goetsch L. Evaluation of antibody-dependent cell Jeung I, Cheon K, Kim MR. Decreased Cytotoxicity of Peripheral and cytotoxicity using lactate dehydrogenase (LDH) measurement. Methods Peritoneal Natural Killer Cell in Endometriosis. Biomed Res Int 2016; Mol Biol 2013;988:305–317. 2016:2916070. Bulun SE, Lin Z, Imir G, Amin S, Demura M, Yilmaz B, Martin R, Johnson NP, Hummelshoj L, World Endometriosis Society Montpellier C. Utsunomiya H, Thung S, Gurates B et al. Regulation of aromatase Consensus on current management of endometriosis. Hum Reprod expression in estrogen-responsive breast and uterine disease: from 2013;28:1552–1568. bench to treatment. Pharmacol Rev 2005;57:359–383. Joly E, Hudrisier D. What is trogocytosis and what is its purpose? Nat Burney RO, Giudice LC. Pathogenesis and pathophysiology of endometri- Immunol 2003;4:815. osis. Fertil Steril, 98:511–519. Kanzaki H, Wang HS, Kariya M, Mori T. Suppression of natural killer cell Caligiuri MA. Human natural killer cells. Blood 2008;112:461–469. activity by sera from patients with endometriosis. Am J Obstet Gynecol Chapman LM, Aggrey AA, Field DJ, Srivastava K, Ture S, Yui K, Topham 1992;167:257–261. DJ, Baldwin WMIII, Morrell CN. Platelets present antigen in the context Kikuchi Y, Ishikawa N, Hirata J, Imaizumi E, Sasa H, Nagata I. Changes of of MHC class I. J Immunol 2012;189:916–923. peripheral blood lymphocyte subsets before and after operation of Cheent K, Khakoo SI. Natural killer cells: integrating diversity with function. patients with endometriosis. Acta Obstet Gynecol Scand 1993;72: Immunology 2009;126:449–457. 157–161. Ding D, Liu X, Duan J, Guo SW. Platelets are an unindicted culprit in the Komiyama S, Aoki D, Komiyama M, Nozawa S. Local activation of TGF- development of endometriosis: clinical and experimental evidence. Hum beta1 at endometriosis sites. J Reprod Med 2007;52:306–312. Reprod 2015;30:812–832. Kopp HG, Placke T, Salih HR. Platelet-derived transforming growth factor- Dmowski WP, Gebel HM, Rawlins RG. Immunologic aspects of endomet- beta down-regulates NKG2D thereby inhibiting natural killer cell antitu- riosis. Obstet Gynecol Clin North Am 1989;16:93–103. mor reactivity. Cancer Res 2009;69:7775–7783. Dmowski WP, Steele RW, Baker GF. Deficient cellular immunity in endo- Krikun G, Mor G, Alvero A, Guller S, Schatz F, Sapi E, Rahman M, Caze R, metriosis. Am J Obstet Gynecol 1981;141:377–383. Qumsiyeh M, Lockwood CJ. A novel immortalized human endometrial Funamizu A, Fukui A, Kamoi M, Fuchinoue K, Yokota M, Fukuhara R, stromal cell line with normal progestational response. Endocrinology Mizunuma H. Expression of natural cytotoxicity receptors on peritoneal 2004;145:2291–2296. fluid natural killer cell and cytokine production by peritoneal fluid natural Lanier LL. Natural killer cell receptor signaling. Curr Opin Immunol 2003;15: killer cell in women with endometriosis. Am J Reprod Immunol 2014;71: 308–314. 359–367. Lebovic DI, Mueller MD, Taylor RN. Immunobiology of endometriosis. Gaetje R, Kotzian S, Herrmann G, Baumann R, Starzinski-Powitz A. Fertil Steril 2001;75:1–10. Nonmalignant epithelial cells, potentially invasive in human endometri- Livak KJ, Schmittgen TD. Analysis of relative gene expression data using osis, lack the tumor suppressor molecule E-cadherin. Am J Pathol 1997; real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. 150:461–467. Methods 2001;25:402–408. Galandrini R, Porpora MG, Stoppacciaro A, Micucci F, Capuano C, Tassi I, Lu Y, Nie J, Liu X, Zheng Y, Guo SW. Trichostatin A, a histone deacetylase Di Felice A, Benedetti-Panici P, Santoni A. Increased frequency of human inhibitor, reduces lesion growth and hyperalgesia in experimentally leukocyte antigen-E inhibitory receptor CD94/NKG2A-expressing peri- induced endometriosis in mice. Hum Reprod 2010;25:1014–1025. toneal natural killer cells in patients with endometriosis. Fertil Steril 2008; Maeda N, Izumiya C, Kusum T, Masumoto T, Yamashita C, Yamamoto Y, 89:1490–1496. Oguri H, Fukaya T. Killer inhibitory receptor CD158a overexpression Garzetti GG, Ciavattini A, Provinciali M, Fabris N, Cignitti M, Romanini C. among natural killer cells in women with endometriosis is undiminished Natural killer cell activity in endometriosis: correlation between serum by laparoscopic surgery and gonadotropin releasing hormone agonist estradiol levels and cytotoxicity. Obstet Gynecol 1993;81:665–668. treatment. Am J Reprod Immunol 2004;51:364–372. Gonzalez-Foruria I, Santulli P, Chouzenoux S, Carmona F, Batteux F, Maeda N, Izumiya C, Oguri H, Kusume T, Yamamoto Y, Fukaya T. Chapron C. Soluble ligands for the NKG2D receptor are released during Aberrant expression of intercellular adhesion molecule-1 and killer endometriosis and correlate with disease severity. PLoS One 2015;10: inhibitory receptors induces immune tolerance in women with pelvic e0119961. endometriosis. Fertil Steril 2002;77:679–683. Groh V, Wu J, Yee C, Spies T. Tumour-derived soluble MIC ligands impair Matsuoka S, Maeda N, Izumiya C, Yamashita C, Nishimori Y, Fukaya T. expression of NKG2D and T-cell activation. Nature 2002;419:734–738. Expression of inhibitory-motif killer immunoglobulin-like receptor, Guo SW, Ding D, Geng JG, Wang L, Liu X. P-selectin as a potential thera- KIR2DL1, is increased in natural killer cells from women with pelvic peutic target for endometriosis. Fertil Steril 2015;103:990–1000 e1008. endometriosis. Am J Reprod Immunol 2005;53:249–254. Guo SW, Ding D, Liu X. Anti-platelet therapy is efficacious in treating endo- Merolla M, Nardi MA, Berger JS. Centrifugation speed affects light trans- metriosis induced in mouse. Reprod Biomed Online 2016a;33:484–499. mission aggregometry. Int J Lab Hematol 2012;34:81–85. Guo SW, Du Y, Liu X. Endometriosis-derived stromal cells secrete Millet V, Naquet P, Guinamard RR. Intercellular MHC transfer Thrombin and Thromboxane A2, inducing platelet activation. Reprod Sci between thymic epithelial and dendritic cells. Eur J Immunol 2008;38: 2016b;23:1044–1052. 1257–1263. 810 Du et al.

Moretta A, Bottino C, Mingari MC, Biassoni R, Moretta L. What is a nat- Smyth SS, McEver RP, Weyrich AS, Morrell CN, Hoffman MR, Arepally ural killer cell? Nat Immunol 2002;3:6–8. GM, French PA, Dauerman HL, Becker RC, Platelet Colloquium P. Munros J, Martinez-Zamora MA, Tassies D, Coloma JL, Torrente MA, Platelet functions beyond hemostasis. J Thromb Haemost 2009;7: Reverter JC, Carmona F, Balasch J. Total circulating microparticle levels 1759–1766. are increased in patients with deep infiltrating endometriosis. Hum Somigliana E, Vigano P, Gaffuri B, Candiani M, Busacca M, Di Blasio AM, Reprod 2016; Dec 17. [Epub ahead of print]. Vignali M. Modulation of NK cell lytic function by endometrial secretory Nieswandt B, Hafner M, Echtenacher B, Mannel DN. Lysis of tumor cells factors: potential role in endometriosis. Am J Reprod Immunol 1996a;36: by natural killer cells in mice is impeded by platelets. Cancer Res 1999; 295–300. 59:1295–1300. Somigliana E, Vigano P, Gaffuri B, Guarneri D, Busacca M, Vignali M. Okada M, Kodama T, Tominaga A, Kon K, Sagawa T, Utsumi S. Human endometrial stromal cells as a source of soluble intercellular Cytotoxicity of activated platelets to autologous red blood cells. Br J adhesion molecule (ICAM)-1 molecules. Hum Reprod 1996b;11: Haematol 1992;82:142–150. 1190–1194. Oosterlynck DJ, Cornillie FJ, Waer M, Vandeputte M, Koninckx PR. Tanaka E, Sendo F, Kawagoe S, Hiroi M. Decreased natural killer cell activ- Women with endometriosis show a defect in natural killer activity ity in women with endometriosis. Gynecol Obstet Invest 1992;34:27–30. resulting in a decreased cytotoxicity to autologous endometrium. Fertil R Core Team. R: A Language and Environment for Statistical Computing. Steril 1991;56:45–51. Vienna, Austria: R Foundation for Statistical Computing, 2013. Oosterlynck DJ, Meuleman C, Sobis H, Vandeputte M, Koninckx PR. Thiruchelvam U, Wingfield M, O’Farrelly C. Natural killer cells: key players Angiogenic activity of peritoneal fluid from women with endometriosis. in endometriosis. Am J Reprod Immunol 2015;74:291–301. Fertil Steril 1993;59:778–782. Tojo M, Hamashima Y, Hanyu A, Kajimoto T, Saitoh M, Miyazono K, Oosterlynck DJ, Meuleman C, Waer M, Vandeputte M, Koninckx PR. The Node M, Imamura T. The ALK-5 inhibitor A-83-01 inhibits Smad signal- natural killer activity of peritoneal fluid lymphocytes is decreased in ing and epithelial-to-mesenchymal transition by transforming growth women with endometriosis. Fertil Steril 1992;58:290–295. factor-beta. Cancer Sci 2005;96:791–800. Orr MT, Lanier LL. Natural killer cell education and tolerance. Cell 2010; Vernet-Tomas Mdel M, Perez-Ares CT, Verdu N, Molinero JL, Fernandez- 142:847–856. Figueras MT, Carreras R. The endometria of patients with endometri- Palumbo JS, Talmage KE, Massari JV, La Jeunesse CM, Flick MJ, Kombrinck osis show higher expression of class I human leukocyte antigen than the KW, Hu Z, Barney KA, Degen JL. Tumor cell-associated tissue factor endometria of healthy women. Fertil Steril 2006;85:78–83. and circulating hemostatic factors cooperate to increase metastatic Vigano P, Somigliana E, Di Blasio AM, Cozzolino S, Candiani M, Vignali M. potential through natural killer cell-dependent and-independent Suppression of natural killer cell function and production of soluble mechanisms. Blood 2007;110:133–141. ICAM-1: endometrial stroma versus melanoma. Am J Reprod Immunol Pizzo A, Salmeri FM, Ardita FV, Sofo V, Tripepi M, Marsico S. Behaviour of 2001;46:342–348. cytokine levels in serum and peritoneal fluid of women with endometri- Vivier E, Tomasello E, Baratin M, Walzer T, Ugolini S. Functions of natural osis. Gynecol Obstet Invest 2002;54:82–87. killer cells. Nat Immunol 2008;9:503–510. Placke T, Kopp HG, Salih HR. The wolf in sheep’s clothing: Platelet- WangF,WenZ,LiH,YangZ,ZhaoX,YaoX.Humanleukocyte derived “pseudo self” impairs cancer cell “missing self” recognition by antigen-G is expressed by the eutopic and ectopic endometrium of NK cells. Oncoimmunology 2012a;1:557–559. adenomyosis. Fertil Steril 2008;90:1599–1604. Placke T, Orgel M, Schaller M, Jung G, Rammensee HG, Kopp HG, Salih Wu MY, Yang JH, Chao KH, Hwang JL, Yang YS, Ho HN. Increase in the HR. Platelet-derived MHC class I confers a pseudonormal phenotype to expression of killer cell inhibitory receptors on peritoneal natural killer cancer cells that subverts the antitumor reactivity of natural killer cells in women with endometriosis. Fertil Steril 2000;74:1187–1191. immune cells. Cancer Res 2012b;72:440–448. Wu Q, Ding D, Liu X, Guo SW. Evidence for a hypercoagulable state in Placke T, Salih HR, Kopp HG. GITR ligand provided by thrombopoietic women with ovarian endometriomas. Reprod Sci 2015;22:1107–1114. cells inhibits NK cell antitumor activity. J Immunol 2012c;189:154–160. Yu JJ, Sun HT, Zhang ZF, Shi RX, Liu LB, Shang WQ, Wei CY, Chang KK, Pradeu T, Jaeger S, Vivier E. The speed of change: towards a discontinuity Shao J, Wang MY et al. IL15 promotes growth and invasion of endomet- theory of immunity? Nat Rev Immunol 2013;13:764–769. rial stromal cells and inhibits killing activity of NK cells in endometriosis. Qu C, Nguyen VA, Merad M, Randolph GJ. MHC class I/peptide transfer Reproduction 2016;152:151–160. between dendritic cells overcomes poor cross-presentation by monocyte- Zhang Q, Ding D, Liu X, Guo SW. Activated platelets induce estrogen derived APCs that engulf dying cells. J Immunol 2009;182:3650–3659. receptor beta expression in endometriotic stromal cells. Gynecol Obstet Sadallah S, Schmied L, Eken C, Charoudeh HN, Amicarella F, Schifferli JA. Invest 2015;80:187–192. Platelet-derived ectosomes reduce NK cell function. J Immunol 2016; Zhang Q, Duan J, Liu X, Guo SW. Platelets drive smooth muscle metapla- 197:1663–1671. sia and fibrogenesis in endometriosis through epithelial-mesenchymal Semino C, Semino A, Pietra G, Mingari MC, Barocci S, Venturini PL, transition and fibroblast-to-myofibroblast transdifferentiation. Mol Cell Ragni N, Melioli G. Role of major histocompatibility complex class I Endocrinol 2016a;428:1–16. expression and natural killer-like T cells in the genetic control of endo- Zhang Q, Duan J, Olson M, Fazleabas A, Guo SW. Cellular changes con- metriosis. Fertil Steril 1995;64:909–916. sistent with epithelial-mesenchymal transition and fibroblast-to- Smyth MJ, Hayakawa Y, Takeda K, Yagita H. New aspects of natural-killer- myofibroblast transdifferentiation in the progression of experimental cell surveillance and therapy of cancer. Nat Rev Cancer 2002;2:850–861. endometriosis in baboons. Reprod Sci 2016b;23:1409–1421.