Human HLA-G+ extravillous trophoblasts:Immune-activating cells that interact withdecidual leukocytesTamara Tilburgsa,1, Ângela C. Crespoa,b, Anita van der Zwana, Basya Rybalova, Towfique Rajc, Barbara Strangerd,Lucy Gardnere, Ashley Moffette, and Jack L. Stromingera,1

aDepartment of Stem Cell and Regenerative Biology, Harvard University, Cambridge, MA 02138; bPhD Program in Experimental Biology and Biomedicine,Center for Neurosciences and Cell Biology, University of Coimbra, 3004-504 Coimbra, Portugal; cDepartment of Neurology, Brigham and Women’s Hospital,Boston, MA 02115; dInstitute for Genomics and Systems Biology, University of Chicago, Chicago, IL 60637; and eDepartment of Pathology, University ofCambridge, Cambridge CB2 1QP, United Kingdom

Contributed by Jack L. Strominger, April 24, 2015 (sent for review February 27, 2015)

Invading human leukocyte antigen-G+ (HLA‐G+) extravillous tro-phoblasts (EVT) are rare cells that are believed to play a key role inthe prevention of a maternal immune attack on foreign fetal tis-sues. Here highly purified HLA‐G+ EVT and HLA‐G− villous tropho-blasts (VT) were isolated. Culture on fibronectin that EVT encounteron invading the uterus increased HLA‐G, EGF-Receptor-2, and LIF-Receptor expression on EVT, presumably representing a further dif-ferentiation state. Microarray and functional gene set enrich-ment analysis revealed a striking immune-activating potential forEVT that was absent in VT. Cocultures of HLA‐G+ EVT with samplematched decidual natural killer cells (dNK), macrophages, and CD4+and CD8+ T cells were established. Interaction of EVT with CD4+T cells resulted in increased numbers of CD4+CD25HIFOXP3+CD45RA+resting regulatory T cells (Treg) and increased the expression level ofthe Treg-specific transcription factor FOXP3 in these cells. However,EVT did not enhance cytokine secretion in dNK, whereas stimulationof dNK with mitogens or classical natural killer targets confirmed thedistinct cytokine secretion profiles of dNK and peripheral blood NKcells (pNK). EVT are specialized cells involved in maternal–fetal tol-erance, the properties of which are not imitated by HLA‐G–express-ing surrogate cell lines.

Treg | FOXP3 | NK cell | human | pregnancy

During pregnancy invading fetal human leukocyte antigen-G+(HLA-G+) extravillous trophoblasts (EVT) play a key role

in the process of placental anchoring, opening of the uterinespiral arteries, and prevention of a maternal immune attack onforeign placental and fetal tissues. Failures of these processes havebeen postulated to play a role in miscarriages, preterm birth, andpreeclampsia although little is understood about the mechanismsinvolved (1). Upon blastocyst implantation, the extraembryoniccells differentiate into several types of trophoblasts that have dis-tinct functions and unique ways of interacting with maternal im-mune cells (1). The main types of trophoblasts include the MHCnegative villous trophoblasts (VT) and the invading HLA-G+EVT. VT, including cytotrophoblasts and syncytiotrophoblasts,form the placental floating villi and are the barrier between fetalcord blood and the maternal blood in the intervillous space.HLA-G+ EVT, found at the tips of anchoring villi (columnar

EVT), detach from these villous columns and adhere to extra-cellular matrix proteins, migrate through decidual tissue (inter-stitial EVT), and invade the uterine spiral arteries where theyreplace the endothelial cell layer (endovascular EVT). The fac-tors required to promote differentiation and proliferation of EVTare unknown but may include interaction with extracellular ma-trix proteins and growth factors such as epidermal growth factor(EGF), leukemia inhibitory factor (LIF), bone morphogeneticfactor-4 (BMP4), and fibroblast growth factor-4 (FGF4) (2–5).EVT invade decidual tissue and interact with maternal decidualnatural killer cells (dNK), decidual macrophages (dMϕ), and de-cidual T cells (dT) (1, 6). EVT are difficult to study due to the low

EVT numbers that can be obtained from tissue and the lack ofproliferative capacity that precludes their expansion in vitro (7–9).EVT express HLA-C, HLA-E, and HLA-G (but not HLA-A

and HLA-B) and avoid immune rejection by maternal leukocytes(1). HLA-E and HLA-G have been shown to limit NK cytotox-icity (10, 11). HLA-G–expressing cells also have been shown toincrease the secretion of cytokines such as interleukin-6 (IL-6)and IL-8 (6, 12, 13), to induce regulatory T cells (Treg) (14), andto modulate antigen-presenting cells by binding to the HLA-Greceptor Leukocyte Ig-like receptor subfamily B member 1(LILBR1, also known as ILT2) (13, 15). At term, pregnancy al-logeneic paternal HLA-C expressed on EVT was correlated withmaternal dT activation and induction of fetus-specific Treg (16).Furthermore, genetic association studies have shown that motherswho have the activating Killer Ig-like Receptor-2DS1 (KIR2DS1)are protected from pregnancy complication when its ligand, theC2 allo-type of HLA-C, is expressed by the fetus (17). From thesedata it may be postulated that immune activation by EVT is re-quired to activate dNK to facilitate trophoblast invasion whilepreventing immune rejection by active induction of Treg.Thus, far the potential of dNK, dT, and dMϕ to secrete cy-

tokines or kill target cells has relied on the use of trophoblast-like cell lines (e.g., JEG3), target cell lines (e.g., 721.221 or K562),the transfectant 721.221/HLA-G, or stimulation with mitogens

Significance

Fetal extravillous trophoblasts (EVT) invade uterine tissue andinteract with maternal immune cells during pregnancy. EVTexpress human leukocyte antigen-C (HLA-C) and -G (HLA-G).Although polymorphic HLA-C can elicit a maternal immuneresponse, HLA-G has been associated with induction of im-mune tolerance. We have succeeded in isolating all maternalimmune cell types as well as EVT from human placental tissue.These methods were used to elucidate the unique charateristicsof EVT as well as their interaction with maternal immune cells.We demonstrate that EVT are specialized cells whose propertiesare not imitated by HLA‐G–expressing surrogate cell lines.Studies using primary EVT are crucial for understandingmaternal–fetal tolerance and development of pregnancy complicationssuch as preeclampsia and miscarriages.

Author contributions: T.T. and J.L.S. designed research; T.T., Â.C.C., A.v.d.Z., and B.R.performed research; T.R., B.S., L.G., and A.M. contributed new reagents/analytic tools;T.T. and Â.C.C. analyzed data; and T.T. and J.L.S. wrote the paper.

Conflict of interest statement: J.L.S. is a consultant for King Abdulaziz University, Jeddah,Saudi Arabia.

Data deposition: Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession no. E-MTAB-3217.1To whom correspondence may be addressed. Email: jlstrom@fas.harvard.edu ortilburgs@fas.harvard.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1507977112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1507977112 PNAS | June 9, 2015 | vol. 112 | no. 23 | 7219–7224

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0

(e.g., phorbol myristate acetate, or PMA) or antibodies that tar-get specific receptors (e.g., KIR2DS1 or CD3). These protocolsunderlined the numerous differences that exist between peripheralblood and decidual leukocytes (6, 13, 18, 19). dNK were shown tohave decreased cytotoxicity and IFNγ secretion and increasedproduction of cytokines such as IL-8, vascular endothelial growthfactor (VEGF), granulocyte monocyte-colony stimulating factor(GM-CSF), and galectin-1 (1LGALS), compared with peripheralblood NK cells (pNK) (6, 13, 20–23). dMϕ, of which there are atleast two subtypes, produce high levels of pro- as well as anti-in-flammatory cytokines (19). dT are a heterogeneous populationwith increased levels of CD4+ Treg and CD8+ Effector-MemoryT cells compared with peripheral blood (18, 24). However, thus farno study has systematically examined the immune response ofdNK, dT, and dMϕ directly to purified VT or HLA-G+ EVT.In the present study, a procedure was developed to obtain

robust VT and EVT preparations as well as all major maternaldecidual leukocyte types (dNK, dMϕ, CD4+, and CD8+ dT)from the same pregnancy. These preparations have been used tostudy the properties of EVT by flow cytometry, microarray, andgene set enrichment analysis. In addition, cocultures of EVT anddecidual leukocytes were established to investigate their contri-butions to the establishment and/or maintenance of maternal–fetal tolerance in human pregnancy.

ResultsIsolation of VT and EVT. Cell preparations of human first-trimestervillous tissue (Materials and Methods), were directly stained forEGF Receptor 1 (EGFR1, also known as ERBB1), HLA-G, andthe common leukocyte antigen (CD45) and analyzed by flowcytometry (Fig. 1A). FACS analysis demonstrated a median of∼9% CD45-HLA-G+EGFR1dim EVT (with some samples ashigh as 20%), ∼77% CD45-HLA-G-EGFR1+ VT, and ∼7%CD45+ leukocytes (Fig. 1B). The fraction of EVT did not varywith gestational age (SI Appendix, Fig. S1). However, the per-centage of CD45-HLA-G-EGFR1+ VT decreased whereasCD45+ cells increased with gestational age (SI Appendix, Fig. S1).The yield was 0.3–7.4 × 105 EVT and 0.4–3.3 × 106 VT perpreparation (SI Appendix, Table S1); the range is due to gesta-tional age and size of the tissue. Both EVT and VT expressed thetrophoblast marker cytokeratin-7 but not the stromal cell markervimentin (SI Appendix, Fig. S2A). FACS analysis confirmed HLA-Cexpression on EVT, whereas HLA-E was relatively low. VT wereMHC-negative (SI Appendix, Fig. S2B).

Up-Regulation of HLA-G and Growth Factor Receptors on EVT byCulture on Fibronectin. When trophoblast isolates were culturedon fibronectin for 2 days and afterward washed and harvested, amarked increase in the percentage of HLA-G+ cells as well asthe intensity of HLA-G on the cells was observed (Fig. 1C).After FACS sorting for CD45-HLA-G+EGFR1dim EVT (Fig. 1A),98% of cells in the EVT fraction were HLA-G+, whereas theCD45-HLA-G-EGFR1+ VT and the remaining CD45-HLA-G-EGFR1− cells were also at 99% purity (Fig. 1D). For the CD45-HLA-G-EGFR1+ VT fraction, very few adherent cells werefound, and neither the adherent nor the nonadherent fractionsshowed HLA-G expression after culture on fibronectin. Furtherstudies are needed to establish whether ligands present in fi-brinoid material in decidua and Matrigel that can engage ad-ditional integrins have distinct effects on VT and EVT phenotypeand capacity to proliferate or differentiate as found previously(25). EVT have been shown to express many receptors for growthfactors that can play a role in trophoblast proliferation and/ordifferentiation. The expression of EGFR2 (also known as CD340or ERBB2), LIFR, and, at low level, EGFR1 on EVT is confirmed,and, moreover, a marked increase in expression of EGFR2 andLIFR, but not EGFR1 was observed after culture on fibronectin(Fig. 1E) (2). VT expressed higher levels of EGFR1 and did notexpress EGFR2 or LIFR. Culture on fibronectin did not change theexpression on VT (SI Appendix, Fig. S3A). In addition, trophoblastisolates were cultured on fibronectin for 1 or 2 days and stained for

the proliferation marker Ki67, but no proliferation indicated byKi67+ cells was observed (SI Appendix, Fig. S3B). Total trophoblastpreparations were also labeled with the fluorescent dye carboxy-fluorescein succinimidyl ester (CFSE) and analyzed at day 3, but noproliferation was observed (SI Appendix, Fig. S3C).

VT and EVT Have Very Distinct Transcriptional Profiles. RNA wasobtained from seven patient-paired and freshly isolated VT andEVT from the HLA-G–expressing cell line JEG3 on two dif-ferent occasions and from four different decidual stromal cell(DSC) cultures. The 20 RNA samples were hybridized indepen-dently to Human Genome Affymetrix U133 plus 2.0 chips. Un-supervised hierarchical cluster analysis resulted in a dendrogramwhere each cell type formed a distinct cluster (Fig. 2A). Asexpected, DSC formed a separate branch from the other threecell types, whereas EVT were closer to JEG3 than to VT.To evaluate the genomic differences that distinguish VT and

EVT, a volcano plot was generated based on the mean expres-sion value of an individual probe’s fold change and the P valueassociated with reproducibility of these changes between the VTand EVT (Fig. 2B). This identified 2,252 probes up-regulated

A

HLA

-G

98% 1%

2% 99%

1%

CD45-HLA-G+EGFRdim CD45-HLA-G-EGFR1+ CD45-HLA-G-EGFR1-

HLA

-GEGFR1

B

0%

C

VT

EVT

R2

CD45+

FSC

SSC

CD45 EGFR1

D

EGFR1 LIFREGFR2

− IgG− fresh− day 2

E

− IgG− fresh− day 2

Fig. 1. Trophoblast preparations. (A) Representative FACS plots of tropho-blast preparations. A live gate (R1) was set in the Forward-Side Scatter plotand CD45+ leukocytes were excluded with gate R2. Subsequently, EGFR1–FITC and HLA-G–PE were plotted, and separation of EVT (CD45-EGFR1dimHLA-G+) and VT (CD45-HLA-G-EGFR1+) is depicted. (B) Percentage of CD45+leukocytes (within live gate), CD45-EGFR1dimHLA-G+ EVT, and CD45-HLA-G-EGFR1+ VT (both within live gate and CD45− fraction) of the trophoblastpreparations. (C) HLA-G expression on freshly isolated trophoblast prepara-tions and preparations cultured for 2 d on fibronectin. (D) RepresentativeFACS plots of HLA-G and EGFR1 expression on FACS-sorted CD45-HLA-G+EVT, CD45-HLA-G-EGFR1+ VT, and the remaining CD45-HLA-G-EGFR1− cellsthat were cultured on fibronectin for 2 d. (E) EGFR1, EGFR2, and LIFR ex-pression on freshly isolated EVT and EVT cultured for 2 d on fibronectin.

7220 | www.pnas.org/cgi/doi/10.1073/pnas.1507977112 Tilburgs et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0

in VT and 1,953 probes up-regulated in EVT with a twofolddifference and indicated that VT and EVT are very distinct celltypes. To generate more stringent gene signatures for VT andEVT, genes were selected based on fourfold differential expres-sion present in all seven patient pairs. This identified 340 probes(encoding 221 unique genes after exclusion of duplicate genes anduncharacterized loci) that were up-regulated in VT (Fig. 2B, bluedots) whereas 328 probes (encoding 210 unique genes) were up-regulated in EVT (Fig. 2B, red dots). The annotated probe lists

are included in Datasets S1 and S2. The gene signature of EVTwas overlaid on volcano plots that compare JEG3 vs. EVT andDSC vs. EVT and showed that the majority of the genes identifiedfor EVT were absent in JEG3 and DSC (Fig. 2C and SI Appendix,Fig. S4A). The gene signatures for HLA-G+ EVT generated bytwo previous studies were overlaid on the volcano plots anddemonstrated the similarity of the findings in both studies al-though distinct probes were also identified (SI Appendix, Fig. S4 Band C) (26, 27). Interestingly, the gene list for EGFR1+ VT (26)demonstrated a clear match whereas the signature for tumor-associated calcium signal transducer-2 (TACSTD2)–expressingVT (27) did not match. The trophoblast isolation procedure usedto isolate TACSTD2+ VT was different in trypsin digestion time(8 vs. 30 min), density gradient (ficoll vs. percoll), and selectionmarker (EGFR1+ vs. TACSTD2+). This may indicate thatEGFR1+ VT and TACSTD2+ VT are distinct subtypes of VT.Detailed analysis of transcriptional differences between EVT

and VT with respect to growth factors and cytokines, receptorsfor growth factors and cytokines, cell adhesion molecules, andECM components are presented as heat maps in SI Appendix,Fig. S5. Here, focus is placed on the integrin family of cell ad-hesion molecules because of their possible role in up-regulationof HLA-G on EVT after culture on fibronectin (Fig. 1C) andbecause of an extensive literature on integrins on the de-velopment of EVT (28). In addition to the integrin subunits thatare differentially expressed (ITGA5 on EVT and ITGB5 andITGB8 on VT), the α- and β-subunits of a number of integrinheterodimers were expressed on both trophoblast cells (SIAppendix, Fig. S6). Integrin α5β1, a fibronectin receptor, wasuniquely expressed by EVT and may contribute to the up-reg-ulation of HLA-G. Integrin αVβ5 and αVβ8, both vitronectinreceptors, are uniquely expressed on VT. Three integrins areexpressed by both EVT and VT: α6β4 and α6β1 (both lamininreceptors) and αVβ1 (a vitronectin receptor). FACS and micro-array data demonstrated ITGA6 and ITGB4 expression on VTand EVT. However, the expression level on VT was increasedand may explain the difference found in a previous study (29).The ligands for integrins are present in Matrigel as well as in thefibrinoid material in the maternal decidua (25).

Gene Set Enrichment Analysis. To identify the functional differ-ences that reflect the transcriptional differences between EVTand VT, a Gene Set Enrichment Analysis (GSEA) was per-formed. GSEA is a computational method that uses an a priori-defined set of genes and determines which “functional” sets ofgenes are up-regulated. GSEA software was used to generate alist of functional gene sets that are specifically overrepresentedin VT or EVT. Interestingly, in EVT, 14 functional gene sets ofthe 20 most significantly enriched gene sets were associated withdirect immune and lymphocyte activation (SI Appendix, TableS2). The core genes of these immune-activation–related pathwaysare cytokines such as Epstein–Barr virus-induced gene-3 (EBI3),tumor growth factor beta-1 (TGF-β1), TGF-β2, IL-8, and the cell-surface molecule CD276 (B7-H3) and cytotoxic and regulatoryT-cell molecule (CRTAM), which can all directly influence lym-phocyte binding and are potent inhibitors of lymphocyte activa-tion (SI Appendix, Fig. S7). EBI3 is known to dimerize with twoα-chains, IL-12(p28) and IL-12(p35), to form the immune-sup-pressive cytokines IL-27 and IL-35 (30). However, neither of theseIL-12 α-chains are expressed by EVT. How EIB3 functions inEVT and whether EBI3 can form homo-dimers or pairs with anunknown α-chain remain to be determined. The functional genesets enriched in VT were less significant (increased false discoveryrate) and more diverse in function (SI Appendix, Table S3).

Coculture of EVT with Sample Matched Maternal Leukocytes. To eval-uate the implications of the striking immune-activating and -reg-ulating potential of EVT demonstrated in the GSEA, VT- andEVT-enriched cocultures with all major maternal leukocyte subsets(dNK, dMɸ, CD4+, and CD8+ dT cells) were established. For thissample, matched trophoblast and decidual leukocytes or trophoblasts

Fig. 2. Expression profiling of EVT and VT. (A) Unsupervised hierarchical clusteranalysis results in a dendrogramwhere each distinct cell type (VT, EVT, JEG-3, andDSC) forms a separate cluster. Volcano plots were generated based on the meanexpression value of an individual probe’s fold change and the P value associatedwith reproducibility of these changes between (B) VT and EVT and (C) EVT andJEG-3. The unique gene signatures based on a more than fourfold difference forVT (340 probes, blue dots) and EVT (328 probes, red dots) are highlighted.

Tilburgs et al. PNAS | June 9, 2015 | vol. 112 | no. 23 | 7221

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0

and peripheral blood leukocytes from unrelated nonpregnant donorswere used. The FACS sorting strategy for all cell types can be foundin SI Appendix, Fig. S8. Microscopy images revealed that EVT adhereto fibronectin with a distinct, large, spread-out morphology and in-teract with multiple other EVT. Furthermore, HLA-G+ EVT withvarious sizes and morphology are found in the cell cultures whereasVT are homogenous, small, rounded cells that do not adhere tofibronectin (SI Appendix, Fig. S9A). Light microscopy also revealedclustering of NK, CD4+, and CD8+ T cells around the large EVT.Multiple NK or T cells can interact with one EVT, either at the cellbody or at filopodia-like structures that spread out from the EVTcell body (SI Appendix, Fig. S9B). No signs of cytolysis of EVT wereobserved in any of the trophoblast–leukocyte cocultures.

EVT Increase the Fraction of CD4+CD25HIFOXP3+CD45RA+ Treg. Tregare major regulators of immune responses at the maternal–fetalinterface (18, 31). To assess the influence of EVT on Treg,CD4+ dT and peripheral blood CD4+ T cells (CD4+ pT) werecultured alone and together with VT and EVT for 3 days. T cellswere harvested and analyzed for the presence of CD4+CD25HI

FOXP3+ Treg. Coculture of CD4+ dT and CD4+ pT with EVTsignificantly increased the percentage of CD4+CD25HIFOXP3+Treg. In addition, EVT significantly increased the expressionlevel (mean fluorescence intensity, or MFI) of FOXP3 within theCD4+CD25HIFOXP3+ Treg after coculture with CD4+ pT(Fig. 3 A–C). The FOXP3 level in CD4+CD25HI dT was al-ready very high and did not further increase by coculture withEVT. When CD4+CD25HI-depleted CD4+ pT (containing bothnaive CD4+CD25− and activated CD4+CD25DIM T cells but

not CD4+CD25HI Treg) were added to EVT, no induction ofCD4+CD25HIFOXP3+ Treg was observed (SI Appendix, Fig.S10B). In contrast, when both total CD4+ pT and purified CD4+CD25HI pT were cocultured with EVT, a marked increase in CD4+CD25HIFOXP3+ Treg compared with CD4+ or CD4+CD25HI pTcultured alone was seen (SI Appendix, Fig. S10 A and C). Thus,EVT specifically increased FOXP3 expression in CD4+CD25HI

cells as opposed to a de novo induction of FOXP3 in CD4+CD25− T cells. CD4+ pT or CD4+CD25HI pT were labeled withCFSE and incubated with or without EVT, VT, or anti-CD3/CD28 beads. Only the pT cells incubated with anti-CD3/CD28beads showed proliferation at day 3, demonstrating that theincrease in FOXP3 expression is not due to increased pro-liferation of these cells (SI Appendix, Fig. S10D). Furthermore,when purified pCD4+CD25HI cells were cocultured with EVT,a significant increase of CD4+CD25HIFOXP3+ CD45RA+Treg, previously described to be resting Treg (32), was observedin the EVT but not in the VT cultures (Fig. 3 D and E). Thus,EVT but not VT directly influence the generation of CD4+CD25HIFOXP3+CD45RA+ resting Treg.

EVT Do Not Elicit Specific Cytokine Responses by Maternal dNK, dMϕ,or dT Cells. Cell culture supernatants were harvested from EVT-enriched and leukocyte cocultures at day 1 (NK) and day 3 (Mɸand T cells), and cytokines were analyzed by using a multiplexcytokine assay. As a negative control, all cell types were culturedalone and as positive controls with 721.221 (221) MHC class Inegative target cells, with 721.221/HLA-G (221.HLA-G) or withPMA to activate NK or anti-CD3/CD28 to activate T cells.Natural killer cells. Upon stimulation with PMA (but not with 221cells), pNK secreted significantly higher levels of the proin-flammatory cytokines granulocyte monocyte–colony-stimulatingfactor (GM-CSF), IFN-gamma (IFNγ), and tumor necrosis fac-tor-alpha (TNFα) compared with dNK (Fig. 4 A–C). Recip-rocally, upon stimulation with 221 cells (but not with PMA),dNK secreted significantly higher levels of vascular endothelialgrowth factor (VEGF) and interleukin-6 (IL-6) than pNK (Fig. 4D and E), confirming previous studies (6, 13, 22). Interestingly,expression of HLA-G on 221 cells reduced IL-6 secretion indNK. Thus, HLA-G expression may play a direct role in sup-pression of at least one proinflammatory factor. However, HLA-Gexpression on 221 cells did not alter secretion of VEGF, animportant factor that contributes to vascular remodeling. Mostsurprisingly, when dNK were cocultured with VT- or EVT-enriched preparations, little or no effect on cytokine secretionwas observed. IL-6 was slightly elevated but no secretion ofGM-CSF, IFNγ, TNFα, VEGF, or IL-17 was induced in dNK(or in pNK) (Fig. 4 A–F). Thus, PMA, 221 cells, and 221/HLA-Gcells used as stimulators revealed differences in pNK anddNK cytokine responses, but this did not reflect VT or EVTresponses. Thus, to understand the unique properties of dNKand EVT–dNK interactions, primary EVT need to be included infurther studies. Secretion of IL-8 (neutrophil chemotactic factor)and galectin-1 (involved in suppression of activated T cells) wassignificantly higher in dNK compared with pNK (Fig. 4 G andH). However, the secretion of IL-8 and galectin-1 by dNK wasnot amplified by PMA, VT, or EVT stimulation. Thus, dNK (butnot pNK) constitutively secreted high levels of IL-8 and galectin-1.Prior activation or differentiation of dNK in vivo may be re-sponsible for inducing IL-8 and galectin-1. Supernatants werealso analyzed for the anti-inflammatory cytokines IL-10 andIL-22, but their secretion was not observed. Furthermore, all dNKand pNK were analyzed for the expression of KIR2DS1. Nodifferences were found in cytokine responses between dNK orpNK from KIR2DS1+ and KIR2DS1− individuals toward any ofthe target cells, including EVT.Macrophages. As demonstrated previously, dMϕ constitutively se-crete both the pro- and anti-inflammatory cytokines IL-10, IL-1β,TNFα, and IL-6 (19). However, in contrast to stimulation withlipopolysaccharide (LPS) or LPS/IFN-γ (19), stimulation of dMϕor monocytes VT- or EVT-enriched preparations did not change

CD4+ pT CD4+ pT +VTCD4+ pT +EVT

FOXP

3

CD25

3.5% 2.5%2.0%A

FOXP

3

CD45RA

E F

D18.7% 5.6% 15.1% 4.1%15.2% 3.0%

CD25HI pT CD25HI pT +VTCD25HI pT +EVT

B C

IgG

0% 0%

CD25HI pT

CD4+ pT

IgG

0.1%

Fig. 3. Coculture with EVT increases the percentage of CD4+FOXP3+ Treg.(A) Representative FACS plots of CD25 and FOXP3 expression of CD4+ pT inthe absence of stimulation or after coculture with VT- or EVT-enrichedpreparations. (B) The percentage of FOXP3+ cells within total CD4+ pT anddT and (C) MFI of FOXP3 expression within CD4+CD25HIFOXP3+ pT and dTcocultured with or without VT- or EVT-enriched preparations. (D) Repre-sentative FACS plots of CD45RA and FOXP3 within CD4+CD25hi pT cocul-tured with or without VT- or EVT-enriched preparations. (E) The percentageof FOXP3+CD45RA− and (F) FOXP3+CD45RA+ cells within each group.

7222 | www.pnas.org/cgi/doi/10.1073/pnas.1507977112 Tilburgs et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0

the secretion of any of these cytokines (SI Appendix, Fig. S11). IL-12(p70), a potent T-cell–stimulating cytokine, and IL-33, a mem-ber of the IL-1 family recently implicated in enhancing proliferationof primary trophoblasts (33) and promotion of Treg survival (34),were not detected in any of the dMϕ or monocyte cultures.T cells.CD4+ and CD8+ dT and pT were incubated with or withoutanti-CD3/CD28 beads and VT- or EVT-enriched preparations for3 days. Although both CD4+ and CD8+ dT and pT had a profoundIFNγ response to anti-CD3/CD28 beads, none of the T cells se-creted IFNγ in response to VT or EVT (SI Appendix, Fig. S12).

DiscussionEVT are found in at least four different locations in the pregnantuterus, i.e., as the cells at the tips of anchoring villi in cell col-umns (columnar EVT), as interstitial EVT within the decidua incontact with maternal leukocytes, as endovascular EVT withinthe uterine spiral arteries in contact with the vascular endothe-lium, and as cells in chorionic tissue directly adjacent to the maternaldecidual or parietal surface (chorionic EVT). In interpreting thepresent data, it is important to note the origin of the EVT studiedhere. These cells were obtained from first-trimester villous tissueseparated from the decidua. They largely represent cells obtainedfrom cell columns at the tips of anchoring villi. Interstitial EVT andendovascular EVT would represent a very small population—toosmall to be isolated under the present circumstances. Some datasuggest that these EVT are all distinct in expression of different

molecules related to distinct functions (28, 35). In that context, thepresent preparation may represent EVT at an intermediate level ofdifferentiation. Culture on fibronectin-enhanced HLA-G expressionas well as expression of EGFR2 and LIFR, presumably representinga further differentiation state. The up-regulation of HLA-G on EVTby culture on fibronectin during the preparations is likely to involvethe interaction of fibronectin with an integrin on EVT, probably α5β1that is prominently expressed by these cells (25, 28).Importantly, coculture of CD4+ T cells with EVT increased the

proportion of FOXP3 as well as the expression level of FOXP3within Treg, suggesting that EVT may directly enhance Tregfunction. Initial experiments demonstrate that the increase ofFOXP3+ is not dependent on proliferation but involves conver-sion of CD4+CD25HI FOXP3− cells into FOXP3+ Treg. In ad-dition, EVT, but not VT, significantly increased the proportion ofCD4+CD25HIFOXP3+CD45RA+ resting Treg. The increase inresting Treg after coculture with EVT demonstrates that EVTdirectly contributes to augmentation of the maternal Treg pool atthe maternal–fetal interface. Although the mechanisms responsi-ble remain unknown, the increase in FOXP3 and proportion ofTreg may contribute to suppression of maternal immune responsesdirected to polymorphic fetal HLA-C molecules expressed onEVT (16). However, the antigen specificity of the Treg at thefetal–maternal interface as well as whether cell contact and/orsoluble factors are required to promote Treg are questions of keyimportance to understanding maternal–fetal tolerance.Gene chip analysis on RNA obtained from seven patient-

paired VT and EVT preparations was performed. In contrast totwo recent studies (26, 36), the VT and EVT preparations werehighly purified and used directly after isolation without in vitroculture to minimize transcriptional changes that may occur withcell culture. Interestingly, the signature for TACSTD2+ VT (27)did not match the transcriptional signature for EGFR1+ VT andmay indicate that these are distinct subtypes of VT. Similarly,confocal imaging suggests that multiple HLA-G+ EVT subtypeswith various sizes and morphology are found in the preparations.Previous studies using different isolation protocols may haveselected trophoblasts with distinct characteristics (37). Furthereffort needs to be made to characterize distinct VT and EVT typespresent in placental tissue. The gene set enrichment analysisrevealed a striking immune-activating potential for EVT andidentified genes able to directly modulate trophoblast–lymphocyteinteractions. The core genes identified by the GSEA include knowncytokines expressed by EVT such as EBI3, TGF-β1, TGF-β2,Inhibin beta A (INHB), and IL-8 and molecules such as CD276(B7-H3) and CRTAM that can directly influence NK and T-cellbinding to EVT. Furthermore, the EVT gene signature revealedthat EVT express a variety of cytokine receptors that mayrender them responsive to cytokines and chemokines secretedby leukocytes such as IL-10, IL-2, IL-1, MIP-1α, RANTES,MCP-3, GM-CSF, M-CSF, and G-CSF.Previous studies have suggested that EVT are derived from

VT at the end tips of the villous tree. In this study, 4,205 probesthat are differentially expressed between EVT and VT were iden-tified (2,252 probes are overrepresented in EVT and 1,953 in VTwith a twofold difference). Furthermore, the GSEA and the uniqueVT and EVT gene signatures revealed major differences in cellularfunctions and expression of transcription factors. These large dif-ferences may indicate that VT and EVT are derived from a distinctcellular origin. Alternatively, EVT precursor cells or trophoblaststem cells at the end tips of the villous tissue may give rise to fullydifferentiated EVT. The proliferation and/or outgrowth of EVTobserved by BrDU incorporation in placental explant cultures (7)may represent a growth burst derived from specialized precursorcells that differentiate into EVT or proliferation by HLA-G+cells at the villous tip. Alternatively, VT may include distinctcell types (e.g., EGFR1+ and TACSTDC2+) of which some canand others cannot differentiate into HLA-G–expressing cells.Furthermore, cell culture conditions used here may lack additionalgrowth factors, integrin ligands, or oxygen/CO2 levels required fortrophoblast differentiation.

-pNKdNK

A B

DC

E F

HG

Fig. 4. Cytokine secretion profiles of pNK and dNK during coculture with VTand EVT. pNK (blue) and dNK (red) were incubated alone or with PMA, 221,221/HLA-G, and VT- or EVT-enriched preparations for 18 h. Cell culture su-pernatants were analyzed for (A) GM-CSF, (B) IFNγ, (C) TNFα, (D) VEGF, (E) IL-6,(F) IL-17A, (G) IL-8, and (H) galectin-1. VT- or EVT-enriched preparations cul-tured alone secreted IL-6 and IL-8 (E and G, depicted by gray dots), but did notsecrete any of the other cytokines.

Tilburgs et al. PNAS | June 9, 2015 | vol. 112 | no. 23 | 7223

IMMUNOLO

GYAND

INFLAMMATION

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0

Because of the difficulties in obtaining human HLA-G+ EVT,many studies have attempted to generate human HLA-G+ EVTlines. However, careful assessment of the cell lines and theirMHC expression demonstrates that these cell lines do not expressa similar MHC profile compared with EVT (38). The closestmodel cell line for EVT and its atypical HLA-C, HLA-E, andHLA-G expression remains the choriocarcinoma cell line JEG-3.The transcriptional profiles of JEG-3 differed from EVT by 1,475probe sets, many representing key functions of EVT, e.g., cyto-kine production, growth factor receptors, ECM components andenzymes, and MHC regulatory molecules. In this study, robustand highly purified VT and EVT preparations were obtained andused to establish cocultures with decidual leukocytes. Most in-terestingly, in contrast to previous studies that used surrogateHLA-G+ cell lines, HLA-G+ EVT did not elicit a profound cy-tokine response by dNK, pNK, dMϕ, monocytes, and CD4+ orCD8+ dT or pT, even though the distinct cytokine secretionprofiles of dNK and pNK in response to stimulation either by 221cells or 221/HLA-G target cells or by PMA was confirmed.Thus, careful and systematic validation of EVT–leukocyte in-

teractions needs to be carried out by using primary HLA-G+EVT to understand the unique contribution of EVT to the de-cidual immune response in human pregnancy. The surrogate celllines presently available are not adequate for this purpose.Studies on the role of primary EVT in maternal–fetal tolerancewill be crucial to understanding the development of pregnancycomplications such as preterm delivery and preeclampsia.

Materials and MethodsDiscarded human placental and decidual material was obtained fromwomenundergoing elective pregnancy termination at 6–12 wk at a local reproductivehealth clinic. Peripheral blood leukocytes were isolated from discarded leuko-packs from healthy volunteer blood donors from the Massachusetts General

Hospital (Boston, MA). All of the human tissue used for this research was de-identified, discarded clinical material. The Committee on the Use of HumanSubjects (the Harvard Institutional Review Board) determined that this use ofplacental and decidual material is not human subjects research. Decidual andvillous tissues were macroscopically identified and separated. Trophoblasts wereisolated as described previously (26, 39). In short, villous tissue was scraped fromthe basal membrane and digested for 8 min at 37 °C with trypsin (2 g/L) andEDTA (0.2 g/L). Trypsin was quenched with F12 medium with 1.0 mL/1 mLnewborn calf serum and penicillin/streptomycin (Gibco) and filtered over agauze mesh. Filtrate was layered on Ficoll (GE Healthcare) for density gradientcentrifugation (20 min at 800 × g) and thereafter incubated for 20 min at 37 °Cin a culture dish for removal of macrophages. Nonadherent cells were collectedand directly stained for flow cytometric analysis or FACS sorting.

Decidual tissue was washed, minced, and digested with 0.1 mg/mL col-lagenase IV and 0.01mg/mL DNase I (Sigma) shaking in a water bath for 1 h at37 °C. Released lymphocytes were filtered through 100-, 70-, and 40-μm sieves(BD Labware). Lymphocytes were dissolved in 20 mL 1.023g/mL Percoll (GEHealthcare) and layered on 10 mL 1.080g/mL and 12.5 mL 1.053g/mL Percollfor density gradient centrifugation (30 min at 800 × g). Lymphocytes wereisolated from the 1.080 to 1.053 g/mL interface and dMϕ from the 1.053 to1.023 g/mL interface. Cells were washed and directly stained for flow cyto-metric analysis or FACS sort (SI Appendix, Fig. S8). pNK, pT, and monocytesfrom leucopacks were isolated with RosetteSep (StemCell Technologies). Forall leukocyte types, >95% purity was obtained.

Cocultures, RNA isolation, microarray hybridization and analysis, flowcytometry, cytokine analysis, and imaging and statistics used are described inSI Appendix.

ACKNOWLEDGMENTS. We thank Patricia Rogers and Joyce Lavecchio forhelp with cell sorting; Jennifer Couget for help with microarray experiments;and all past and current laboratory members for their helpful discussions.This work was supported by National Institutes of Health Grant AI053330.Â.C.C. was supported by the Portuguese Foundation for Science and Technology(FCT) (SFRH/BD/33885/2009).

1. Moffett A, Loke C (2006) Immunology of placentation in eutherian mammals. Nat RevImmunol 6(8):584–594.

2. Jokhi PP, King A, Loke YW (1994) Reciprocal expression of epidermal growth factorreceptor (EGF-R) and c-erbB2 by non-invasive and invasive human trophoblast pop-ulations. Cytokine 6(4):433–442.

3. Kimber SJ (2005) Leukaemia inhibitory factor in implantation and uterine biology.Reproduction 130(2):131–145.

4. Das P, et al. (2007) Effects of fgf2 and oxygen in the bmp4-driven differentiation oftrophoblast from human embryonic stem cells. Stem Cell Res (Amst) 1(1):61–74.

5. Roberts RM, Fisher SJ (2011) Trophoblast stem cells. Biol Reprod 84(3):412–421.6. Hanna J, et al. (2006) Decidual NK cells regulate key developmental processes at the

human fetal-maternal interface. Nat Med 12(9):1065–1074.7. Aplin JD, Haigh T, Jones CJ, Church HJ, Vi�covac L (1999) Development of cyto-

trophoblast columns from explanted first-trimester human placental villi: Role of fi-bronectin and integrin alpha5beta1. Biol Reprod 60(4):828–838.

8. Xu G, Guimond MJ, Chakraborty C, Lala PK (2002) Control of proliferation, migration,and invasiveness of human extravillous trophoblast by decorin, a decidual product.Biol Reprod 67(2):681–689.

9. Weier JF, et al. (2005) Human cytotrophoblasts acquire aneuploidies as they differ-entiate to an invasive phenotype. Dev Biol 279(2):420–432.

10. Pazmany L, et al. (1996) Protection from natural killer cell-mediated lysis by HLA-Gexpression on target cells. Science 274(5288):792–795.

11. Long EO (1999) Regulation of immune responses through inhibitory receptors. AnnuRev Immunol 17:875–904.

12. Rajagopalan S, et al. (2006) Activation of NK cells by an endocytosed receptor forsoluble HLA-G. PLoS Biol 4(1):e9.

13. Li C, Houser BL, Nicotra ML, Strominger JL (2009) HLA-G homodimer-induced cytokinesecretion through HLA-G receptors on human decidual macrophages and naturalkiller cells. Proc Natl Acad Sci USA 106(14):5767–5772.

14. Brown R, et al. (2012) CD86+ or HLA-G+ can be transferred via trogocytosis from my-eloma cells to T cells and are associated with poor prognosis. Blood 120(10):2055–2063.

15. Shiroishi M, et al. (2006) Structural basis for recognition of the nonclassical MHCmolecule HLA-G by the leukocyte Ig-like receptor B2 (LILRB2/LIR2/ILT4/CD85d). ProcNatl Acad Sci USA 103(44):16412–16417.

16. Tilburgs T, et al. (2009) Fetal-maternal HLA-C mismatch is associated with decidual T cellactivation and induction of functional T regulatory cells. J Reprod Immunol 82(2):148–157.

17. Hiby SE, et al. (2004) Combinations of maternal KIR and fetal HLA-C genes influencethe risk of preeclampsia and reproductive success. J Exp Med 200(8):957–965.

18. Tilburgs T, et al. (2008) Evidence for a selective migration of fetus-specific CD4+CD25bright regulatory T cells from the peripheral blood to the decidua in humanpregnancy. J Immunol 180(8):5737–5745.

19. Houser BL, Tilburgs T, Hill J, Nicotra ML, Strominger JL (2011) Two unique humandecidual macrophage populations. J Immunol 186(4):2633–2642.

20. Koopman LA, et al. (2003) Human decidual natural killer cells are a unique NK cellsubset with immunomodulatory potential. J Exp Med 198(8):1201–1212.

21. Kopcow HD, et al. (2005) Human decidual NK cells form immature activating synapsesand are not cytotoxic. Proc Natl Acad Sci USA 102(43):15563–15568.

22. Apps R, et al. (2011) Ex vivo functional responses to HLA-G differ between blood anddecidual NK cells. Mol Hum Reprod 17(9):577–586.

23. Xiong S, et al. (2013) Maternal uterine NK cell-activating receptor KIR2DS1 enhancesplacentation. J Clin Invest 123(10):4264–4272.

24. Tilburgs T, et al. (2010) Human decidual tissue contains differentiated CD8+ effector-memory T cells with unique properties. J Immunol 185(7):4470–4477.

25. Damsky CH, et al. (1994) Integrin switching regulates normal trophoblast invasion.Development 120(12):3657–3666.

26. Apps R, et al. (2011) Genome-wide expression profile of first trimester villous andextravillous human trophoblast cells. Placenta 32(1):33–43.

27. Telugu BP, et al. (2013) Comparison of extravillous trophoblast cells derived fromhuman embryonic stem cells and from first trimester human placentas. Placenta 34(7):536–543.

28. Harris LK, Jones CJ, Aplin JD (2009) Adhesion molecules in human trophoblast: Areview. II. Extravillous trophoblast. Placenta 30(4):299–304.

29. Burrows TD, King A, Smith SK, Loke YW (1995) Human trophoblast adhesion to matrixproteins: Inhibition and signal transduction. Hum Reprod 10(9):2489–2500.

30. Vignali DA, Kuchroo VK (2012) IL-12 family cytokines: Immunological playmakers. NatImmunol 13(8):722–728.

31. Aluvihare VR, Kallikourdis M, Betz AG (2004) Regulatory T cells mediate maternaltolerance to the fetus. Nat Immunol 5(3):266–271.

32. Miyara M, et al. (2009) Functional delineation and differentiation dynamics of humanCD4+ T cells expressing the FoxP3 transcription factor. Immunity 30(6):899–911.

33. Fock V, et al. (2013) Macrophage-derived IL-33 is a critical factor for placental growth.J Immunol 191(7):3734–3743.

34. Schiering C, et al. (2014) The alarmin IL-33 promotes regulatory T-cell function in theintestine. Nature 513(7519):564–568.

35. Red-Horse K, et al. (2004) Trophoblast differentiation during embryo implantationand formation of the maternal-fetal interface. J Clin Invest 114(6):744–754.

36. Bilban M, et al. (2009) Identification of novel trophoblast invasion-related genes:Heme oxygenase-1 controls motility via peroxisome proliferator-activated receptorgamma. Endocrinology 150(2):1000–1013.

37. Hunkapiller NM, Fisher SJ (2008) Chapter 12. Placental remodeling of the uterinevasculature. Methods Enzymol 445:281–302.

38. Apps R, et al. (2009) Human leucocyte antigen (HLA) expression of primary tropho-blast cells and placental cell lines, determined using single antigen beads to charac-terize allotype specificities of anti-HLA antibodies. Immunology 127(1):26–39.

39. Trundley A, Gardner L, Northfield J, Chang C, Moffett A (2006) Methods for isolationof cells from the human fetal-maternal interface. Methods Mol Med 122:109–122.

7224 | www.pnas.org/cgi/doi/10.1073/pnas.1507977112 Tilburgs et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 9,

202

0