Human and mouse hemochorial placentas share many features, but differ in ways that affect immunity. The human placenta is structured as chorionic villous trees that are bathed in maternal blood. A single layer of multinucleated syncytiotrophoblast (SynT) surfaces human placental villous trees and lines the intervillous space perfused by maternal blood. Mononucleated cytotrophoblasts are undifferentiated progenitor cells, subjacent to the SynT, that differentiate and fuse to replenish the SynT. Other cytotrophoblasts differentiate into EVTs, which are located at tips of the anchoring villi and invade the decidua basalis and maternal decidual vessels. During first and into early second trimester, human placental villi are hemodichorial, covered with a continuous bilayer of SynT and cytotrophoblasts. The villous cytotrophoblast layer becomes discontinuous during the second trimester yielding a hemomonochorial structure with only a SynT cell barrier between fetal and maternal blood.

In mouse, the definitive chorioallantoic placenta is labyrinthine with two separate, maze-like vascular systems. Branches of the central arteries traverse the junctional zone then perfuse the fetal labyrinth with maternal blood. The irregularly-shaped junctional zone lies between the decidua basalis and labyrinth. The labyrinth is surfaced by two syncytial layers of trophoblast (SynT I and II) and a mononuclear trophoblast layer adjacent to the maternal blood. The labyrinthine structure comprises the interhemal unit, which continuously thins during gestation. Scanning electron microscopy indicates that midgestational mouse placenta is hemodichorial, with direct maternal blood contact of SynT-I as the cellular trophoblast layer becomes incomplete, exposing areas of SynT-II (18). These differences in cellular composition and placental structure may alter immunity at the maternal-placental interface. Thus, mechanisms that protect fetuses from an activated maternal immune system, and block access of maternally circulating pathogens to the fetus may differ spatiotemporally in human and mouse gestations.

Typically, placental membranes separate the semi-allogenic fetus from the maternal immune system throughout gestation, an arrangement that protects against maternal, immune-mediated elimination of the fetus. Multiple immunological interfaces occur between the placenta and maternal immune system. One such interface is between the trophoblast cells that line the chorionic villi (human) and labyrinth structure (mouse) that bathe in maternal blood (Figure 1-II). In human placentas, the SynT cells exposed to maternal blood do not express MHC-I or MHC-II, and do not elicit maternal T cell responses. In mouse placentas, the MHC expression on labyrinthine trophoblast is less clear. A second immunological interface occurs in the decidualized endometrium. In a process called interstitial trophoblast invasion the EVTs in human and extra labyrinthine trophoblasts in mice are intimately positioned to interact with the maternal immune cells. For example, EVTs express the class I molecule HLA-C and non-classical class I molecules HLA-E and HLA-G. HLA-C, HLA-E, and HLA-G interact with KIR and CD94/NKG2 receptors expressed in NK cells. HLA-G binds, members of the immunoglobulin-like transcript (ILT) ILT2 and ILT4, a family of receptors expressed in NK cells to induce growth factors important for fetal development (19). Invading giant trophoblast cells (TGCs) in mouse do not express non-classical MHC-I molecules (20, 21); however, TGCs that potentially contact immune cells in the decidua basalis of the C57BL/6J B6 mice do express the classical MHC-I, H-2K (Figure 1-I). Finally, in human placentas during endovascular trophoblast invasion, the EVTs hijack and transform the maternal circulatory system to yield high capacitance, low resistance arterial flow in the placenta and interact with circulating immune cells (Figure 1-III). Hence, the different immunological interfaces, with differential expression of paternally-inherited MHC receptors, may elicit distinct maternal immune responses such as systemic responses to SynT, or local tissue-specific responses to invading trophoblasts, or both.

NK cells are the originating members of an assortment of innate lymphoid cells (ILCs). NK cells are a heterogeneous population in the spleen, circulating blood, and many tissues (27, 32–34). Historically, most studies investigated conventional NK (cNK) in mouse spleen and human peripheral blood. NK cells do not express antigen-specific T-cell receptors (TCRs) or B-cell receptors (BCRs), distinguishing them from the adaptive immune system (35, 36). Developmental studies indicate that all ILC lineages arise from common lymphoid progenitors (CLPs) (37, 38). CLPs differentiate into NK cells, ILC1s, ILC2s, ILC3s and lymphoid tissue inducer (LTi) cells (39, 40). The common helper ILC precursor (CHILP), the progenitor to all helper-like ILC lineages, gives rise to ILC1s, ILC2s, and ILC3s, but not LTi cells or NK cells (41). This separates cNK cell development from other ILCs and ensures that cytotoxic NK cells are distinct from ILC1s (38).

We identified tissue-resident NK (trNK) cell populations in mouse liver, skin, and virgin uterus that are distinct from cNK cells (32). The trNK cells do not circulate in parabiotic mice, whereas cNK cells circulate (Figure 2). Phenotypic and RNA-seq analyses revealed that trNK and cNK cells express CD49a and DX5, respectively (32, 33). Liver trNK cells lack Eomesodermin (Eomes), a transcription factor expressed in cNK cells. Nfil3^−/− mice have trNK cells in liver, skin, and virgin uterus, but lack cNK cells. Tbx21^−/− mice lack trNK cells in liver and skin, whereas cNK cell distribution is relatively unaffected at these tissues. In virgin uteri, trNK cells dominate over cNK cells and are unaffected in Nfil3^−/− and Tbx21^−/− mice. Ly49 receptor expression repertoires differ between trNK and cNK cells; liver trNK cells do not express Ly49D or Ly49H. Although trNK cells have an immature phenotype, they have more diverse cytokine production than cNK cells. Thus, trNK and cNK cells represent divergent NK cell lineages, with a distinct trNK cell lineage in virgin uterus differing not only from cNK cell but also from liver/skin trNK cells (32).

Although trNK and cNK cells represent different lineages, recent studies suggest cNK cell conversion to NK cells phenotypically similar to liver ILC1 cells (42). For example, mouse CD49a⁻ DX5⁺ Eomes⁺ cNK cells can change their phenotype in a tumor microenvironment to become CD49a⁺ DX5⁻ Eomes^int, while, cNK cells convert into ILC1-like cells during Toxoplasma gondii infection (43). Human peripheral blood (pb) NK cells and hematopoietic progenitor cells isolated from decidual tissue also convert in vitro to a decidual NK cell phenotype (44, 45). It is currently unknown whether cNK cell conversion occurs in pregnancy.

Despite differences in gestational lengths (9 months in humans vs. 19.5 days in C57BL/6J mice) and placental structure, early pregnancy consistently triggers lymphocyte accumulation within successful implantation sites (23, 46). Although uNK cells are ultimately derived from bone marrow (26, 47), studies using immunocompetent donor and alymphoid recipient mice report uNK cell accumulation during pregnancy using uterine segment transplants (48, 49) or adoptive splenocyte transfer (48). These results indicate uterine homing of peripheral NK cells when the uterus lacked endogenous uNK cells. These findings were challenged when adoptively transferred splenic NK cells failed to home to a syngeneic, immunocompetent pregnant uterus, indicating expansion of resident progenitor NK cells (50). We recently used the parabiosis model together with experimentally-induced decidualization to study early pregnancy events and demonstrated that trNK cells do not circulate. Rather, trNK cells proliferate locally and expand the uNK cell pool with minimal cNK cells contributions (13, 27). These data support a two-wave hypothesis for uNK cell accumulation during typical pregnancies (13, 27, 51). The first wave involves local proliferation of trNK cells during decidualization; the second wave postulates the accumulation of cNK cells recruited into the decidua basalis during placentation. Whether or not trNK cells contribute to the NK cell pool of early human pregnancy remains unknown (discussed below).

In first trimester human pregnancy, NK cells represent ~70% of the lymphocytes present in decidua basalis and are phenotypically and functionally distinguished from pbNK cells. Human uNK (often referred to as decidual NK cells, dNK) cells are CD56^bright CD16⁻KIR⁺CD9⁺CD49a⁺ while pbNK cells are CD56^dimCD16⁺KIR⁺ (52). Compared to pbNK cells, uNK cells are not cytotoxic, produce lower amounts of IFN-γ and, as in mouse, secrete VEGF that promotes angiogenesis (28, 53, 54). Single-cell RNA sequencing comparing cells isolated from first trimester decidual basalis and matched peripheral blood demonstrated three distinct NK cell subsets in decidua basalis, dNK1-3, distinct from pbNK cells (55). High dimensional CyTOF analysis found dNK1-3 mixed with populations of proliferating dNK, dILC3 and pbNK cells suggestive of mixed origins of decidual lineages (31). The dNK1 cells resembled a recently identified “pregnancy trained dNK cell” subset found in repeated pregnancy (56) similarly to expanded ILC1s of mouse second pregnancy (57). Mouse models should be instrumental in clarifying the functional contributions of each uNK cell subset at the maternal-fetal interface.

Understanding of how human uNK cells affect pregnancy and the placental vasculature has been challenging; however, studies of mouse uNK cells have been informative. Several mouse genetic models deficient in NK cells display aberrant modification of the uterine gestational vasculature, as their strongest reproductive phenotype. Rag2^−/−gc^−/− alymphoid mice, display defective spiral artery remodeling during pregnancy (58) but no defects in implantation (59). The arterial defects were rescued in bone marrow (BM) chimeric mice when Rag2^−/− gc^−/− recipients received wild-type BM but not BM from interferon-γ-deficient mice. Similarly, Nfil3^−/− mice that lack cNK cells (32), have a smaller MLAp, aberrant uterine arterial modifications, and smaller pups (60, 61). Roles for interferon-γ in dNK cell-mediated vessel modification have not been confirmed in human studies. In the first trimester, uterine artery Doppler ultrasound screening can be used to identify women with poor spiral arterial remodeling predicting greater risk for preeclampsia and fetal growth restriction (62). Using in vitro analysis of EVT lines and primary explant cultures, dNK cells from the typical pregnancies promoted trophoblast motility and invasion while dNK cells from women identified as high risk for preeclampsia limited trophoblast function (63). Taken together, human and murine placental pathology supports that vascular remodeling defects are associated with preeclampsia and fetal growth restriction and linked to deficits in uNK cells numbers or functions (64, 65).

Establishing hemochorial placentation requires aggressive invasion by trophoblast cells. Human trophoblasts show interstitial invasion, penetrating the uterine epithelium and expanding in the subjacent decidua. Rodent trophoblast invasion is significant but not as deep as human (66). Modifications of maternal blood flow are established by endovascular trophoblast invasion of implantation site decidual vessels. In humans and rats, interstitial and endovascular trophoblast invasion both extend past the decidua basalis and penetrate the inner myometrium. In mice, trophoblast invasion is interstitial and limited to the junctional zone (67). Uncontrolled trophoblast invasion that penetrates beyond the decidua into the myometrial muscle creates the life-threatening pregnancy complication called placenta accreta. A recently reported dystocia mouse model resembles human placenta accreta (60). The phenotype includes stillbirths, hemorrhage, and undelivered placentas. Histopathology revealed excessive trophoblast invasion and reduced numbers of trNK cells (68). In this model, a deficiency in the signaling adaptor molecule Grb2-associated binding protein 3 (Gab3) was important for trNK cell proliferation in response to local cytokine (IL15) stimulation. A similar outcome of dysregulated trophoblast invasion occurs in a rat model of IL15 deficient NK cells (69). These results suggest that defects in trNK cell proliferation lead to interstitial trophoblast hyperinvasion and provide insights into placenta accreta.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.