Sir Peter Medawar, a pioneer in transplantation biology, was the first to identify the immunological paradox of pregnancy. By formulating the research question 60 years ago “how does the pregnant mother contrive to nourish within itself, for many weeks or months, a fetus that is an antigenically foreign body?’’, he set the ground for the field of reproductive immunology. In pregnancy, the fetus remains protected from maternal immune responses as the multilayered placental villi acts as a semi-permeable barrier for the bi-directional migration of immune cells from fetal and maternal blood (1). However, the semi-allogenic placental trophoblast cells are in direct contact with the leucocyte-rich uterine mucosa, referred to as decidua. Medawar’s hypotheses to solve this paradox included the anatomical separation between the mother and the fetus, fetal antigenic immaturity and the immunological indolence of the mother (2, 3). Although these hypotheses have been tested across the years, this question remains incompletely answered, and investigation on this complex topic is still ongoing. As yet, a fine tuning of uterine leucocyte function has become evident. Systemic changes synergize with the intra-uterine microenvironment to shift the differentiation of antigen presenting cells (e.g. macrophages and dendritic cells) towards tolerogenic responses that promote the accumulation of regulatory T cells while restricting local cell activation and cytotoxicity. Of particular abundance in the early pregnant uterus, NK cells are critical for vascular changes and trophoblast invasion in early pregnancy (4). Less well understood to date is the role of CD8+ T cells. Due to their cytotoxic nature towards allogenic cells, the presence of uterine CD8+ T cells is of high interest in the context of fetal immune tolerance. In organ transplantation, CD8+ T cells are important effectors mediating allograft tissue damage. By contrast, in the pregnant uterus, CD8+ T cells despite being the most abundant T cell subset do not trigger lytic responses towards the allogenic placental trophoblast. Within the nonpregnant human uterus, CD8+ T cells have cytotoxic capabilities with fluctuating activity through the menstrual cycle (5). In first trimester pregnancy, T cells constitute approximately 5-20% of total decidual leucocytes of which CD8+ T cells are approximately 45% (6, 7). Although decidual NK cells have received great attention, their frequencies remain stable throughout pregnancy, whereas T cells actively increase (8–10). In the past few years, studies on decidual CD8+ T cells have kept pace as new findings on CD8+ T cell properties are emerging, which characterize CD8+ T cells by e.g. expression of functional receptors, secretion profile and antigen specificity. Due to the heterogeneity of CD8+ T cells and current methodological approaches, classification of CD8+ T cell subsets remains a dilemma as controversy colors the scientific community within CD8+ T cell biology (11–13). With that in mind, in this review, we aim to provide an overview of the phenotypes and function of CD8+ T cells in the pregnant uterus. In particular, we seek to highlight recent findings of human decidual CD8+ T cells and their role in pregnancy, including their contribution to the immunological paradox of pregnancy. When required, we will refer to translational aspects, for example in mouse models. We will focus on conventional CD8+ T cells, which express the ɑ and β chains of the T cell receptor (TCR), although the decidua contains also unconventional CD8+ T cells, such as those expressing the γ and δ chains of the TCR (14). We will discuss the most relevant issues limiting our understanding of CD8+ T cell biology in the context of pregnancy and highlight the gaps in knowledge that are required to be bridged, in order to advance this field. In general, investigations of the decidual CD8+ T cell compartment have largely complied with a selection of methods and markers ( Supplementary Table 1 , Supplementary Figure 1 ) that allows for a relatively unifying interpretation of their differentiation profile. In order to compare this information about decidual CD8+ T cell subsets, we re-analysed data on first trimester CD8+ T cell single cell transcriptome, published by Chen et al. (15) and report the findings about equivalent populations that are described in the literature.

T cells are atypical somatic cells as after enduring quiescence for years, they can initiate massive proliferation and differentiation, resulting in multitude of heterogeneous cell subpopulations with diverse properties (16). Supplementary Table 1 and Supplementary Figure 1 contain information on the different CD8+ T cell subsets with a breakdown of CD8+ T cell subsets, their abbreviation, and cell-specific markers listed throughout the manuscript. The differentiation profile of CD8+ T cells populating the decidua ( Figure 3 ) differs largely from that in peripheral blood ( Table 1 ), indicating a highly tuned homeostasis of decidual CD8+ T cells.

According to the “circular” or “effector first” model, T cells undergo a “naïve->effector->memory” differentiation hierarchy. This model indicates that cells activated by a primary antigen exposure differentiate towards cytotoxic effector cells. Upon antigen clearance, a large subset of effector cells undergo apoptosis/clonal deletion (short-lived effector cells, SLEC), whereas a small fraction (memory precursors effector cell, MPEC) survive, give rise to long lived memory cells (17) ( Supplementary Figure 1 ). This proposed lineage relationships is currently under debate, and the so-called “linear model”, or “memory first model” places memory cells as an intermediate stage in T cell development, proposing that naïve cells can directly develop into memory cells, some of which later develop effector function (18–22). In first and third trimester decidua the frequency of naïve cells within CD8+ T cells is generally lower than in peripheral blood (6, 7, 23–27) as expected in a peripheral tissue ( Supplementary Table 2 ). Limited information on effector subsets is available due to the scarce use of the markers CD127 and KLRG1 that could serve to discriminate MPEC from SLEC in the immune phenotyping of decidual T cells (28). Our analysis of single cell sequencing data pinpoints that the small cluster (3%) number 7 includes SLEC ( Figure 1 ), as indicated by the negativity for CD127, and the expression of KLRG1, granzyme H, B, perforin 1, CD160 and additional killer cell lectin-like receptors. Whether these cells are recently activated in the context of the current pregnancy or recruited to the uterus after previous antigen exposure i.e. in secondary lymphoid organs remains unclear. Bridging these gaps in knowledge could further elucidate important steps in the immune responses towards fetal/placental antigens in pregnancy.

More data is available on the heterogeneous memory CD8+ T cells subsets within human decidua, as summarized in Table 1 . After responding to their cognate antigen, memory CD8+ T cells acquire distinct phenotypes and migration potential and endure long-term. They can quickly respond to re-stimulation by vigorous proliferation, cytotoxicity and secretion of effector cytokines (11). Compared to other memory CD8+ T cells, the stem cell memory (SCM) subset represents a naïve-like memory cell population with enhanced capacity to self-renew that serves as a precursor of other memory cell subsets (29). Due to the massive expansion of memory T cells across pregnancy and the long-term maintenance of robust T cell-memory to former pregnancies (30), it is tempting to hypothesize that fetal specific SCM are induced in pregnancy. To date it remains uninvestigated whether SCM cells are present in the decidua or more likely in uterus draining lymph nodes, as SCM are mainly found in lymphoid organs (31). A further elucidation of the transcriptional signature of this cell subset may allow its identification in the future, which is currently challenging e.g. in our single cell sequencing data analysis due to the limited number of cells analyzed and the low expression of key transcription factors. In this regard, within cluster 1, now defined as CM/TPM, we observed an enhanced expression of the mRNA encoding TCF-1 and CCR7, suggesting that, if present in the uterus, SCM could also cluster here ( Figures 1 and  2 ).

The relative presence of two long-standing and divergent memory populations namely central memory (CM) and effector memory (EM) cells is well described in the uterus (6, 7, 23–27, 33). Due to expression of lymphoid homing receptors, CM cells are highly prevalent in secondary lymphoid organs and circulate through blood and lymph. They only account for a small fraction in decidua that ranges from 3-15% of CD8+ T cells in first and third trimesters as shown by flow cytometry data (6, 7, 23–26, 34, 35). Recently, an additional memory population capable of circulating in blood and lymph nodes but also of trafficking within peripheral tissues was identified and referred to as peripheral memory (Tpm) T cells (11, 36). Tpm cells display intermediate CX3CR1 expression, which has not yet been studied in the decidua and appears as a novel and important marker for distinguishing the distinct migratory patterns that define EM, CM and Tpm CD8+ T cells (11). Analysis of the existing scRNA sequencing data for decidual immune cells reveals a rather large cluster (21%) characterized by high levels of CD127 and CCR7, which could be indicative of CM-like and Tpm subsets, advocating a more prevalent population than shown by flow cytometry data in the pregnant uterus ( Figure 1 ). As intermediate expression range of CX3CR1 was not detectable on mRNA level, we cannot separate CM and Tpm subsets. It remains to be elucidated whether these observations are due to a potential contamination with PBMCs or due to particular migration patterns of peripheral cells to the uterus during pregnancy.

EM cells, expressing receptors necessary for entering local inflammation sites, circulate through non-lymphoid tissues and exhibit higher cytolytic activity. It is not surprising that EM cells accumulate in the decidua (6, 7, 23–27) to account as the most prevalent CD8+ T cell population in first and third trimester, ranging from 48-83% of CD8+T cells (6, 7, 23–27, 33, 34, 37, 38). Decidual CD8+ EM T cells possess elevated expression of inhibitory receptors such as Programmed cell death protein 1 (PD1), T-cell immunoglobulin mucin family member 3 (Tim3), Cytotoxic T-lymphocyte-associated protein 4 (CTLA4), Lymphocyte activation gene 3 (Lag3) and CD39 compared to peripheral CD8+ T cells (6, 24, 25, 27, 39). Among the EM subset, the less differentiated EM1 population (CD27+CD28+) is of the highest frequency in the first trimester decidua (25, 26). In contrast, in the third trimester decidua, EM1 are outnumbered by more differentiated EM2 and EM3 ( Table 1 ), whose frequency is significantly increased compared to first trimester (25). An additional terminally differentiated EM cell population defined by the re-expression of CD45RA are the so-called effector memory re-expressing CD45RA (EMRA) cells. EMRA cells are found in decidua, ranging from 11-45% (6, 7, 25, 26, 34, 38) among which the more differentiated phenotype (CD27-CD28-) is the most prevalent (26). Similarly to EM, cell surface membrane expression of inhibitory receptors such as PD1 and Tim3 have been detected on EMRA cells (39). The accumulation of particular subpopulations of EM and EMRA subsets may mirror a dynamic differentiation throughout pregnancy e.g. in response to increasing antigen levels. Notably, defining markers for EMRA (CD45RA+CCR7-CD127-), would also include effectors, which should be kept in mind when relying on such gating strategy (11).

Remarkably, this notion of a dominating EM CD8+T cell profile in the decidua is challenged by the recognition of tissue resident memory (TRM) subsets which share some markers with EM. TRM migrate to non-lymphoid tissues and enter specific differentiation programs under the influence of tissue specific signals. TRM cells infrequently re-enter the circulation and account therefore for the first line of defense within tissues upon re-encounter with antigens, by subsequent secretion of proinflammatory cytokines or lysis of infected cells (40). TRM can have overlapping properties with other CD8+T cells including effector functions and/or high levels of inhibitory receptors (12, 40) and it is still debated whether TRM cells are terminally differentiated or not (16). TRM CD8+ T cells have been recently identified in mucosa of the whole human reproductive tract (41, 42) and in the decidua (6, 24–26, 33, 34) by the single or co-expression of the mucosal/tissue retention markers CD103 and CD69 (11). Among decidual CD8+ T cells ~26% are CD69+CD103+, which likely include the true tissue resident memory cell population, whereas ~50% are CD69+CD103-. CD69+CD103+ TRM cells expressed significantly higher levels of the inhibitory molecules PD1 and CD39 (34). These observations invite to reevaluate the relative abundance of EM and TRM subsets in the decidua, their similarities as well as differences, as reports before the identification of TRM may have mistakenly referred to them as EM.

The single cell transcriptome analysis scattered four further clusters of CD8+ memory T cells. Among them the most abundant cluster 0 (22%) represents TRM cells ( Figure 1 ), further supporting recent flow cytometry data on the high prevalence of TRM cells in the uterus. The TRM profile was evident by the high expression of mRNA coding for TRM markers CD69, CXCR6 and Hobit, although the markers CD103 and Blimp1 were not particularly co-expressed in this or other clusters. Markers such as Hobit, CD69, and IFNy are also abundant in Cluster 2, which defines a population of stressed and/or dying cells. Hence, Cluster 2 may also contain TRM cells, which are known to be extremely susceptible to ATP-induced cell death (43). In contrast, the clusters 3 and 4 presented mixed gene signatures of EM and TRM subsets ( Figure 1 and Supplementary Figure 2 ). Here the prevalence of EMRA cells could not be described since CD45RA cannot be determined at the mRNA level. Memory cells scattered in cluster 3 presented a pattern of regulatory/inhibitory and pro-angiogenic genes together with Hobit, whereas cluster 4 was enriched in markers for effector function and low in Hobit. These observations suggest that under the complex stimuli at the decidual microenvironment TRM and EM may gain similar functional signatures that complicate their discrimination.

Collectively, the continuous progress in research on differentiation, phenotypical and functional features of CD8+ T cell subsets (16) opens new questions on the CD8+ T cell compartment in the pregnant uterus. The current findings indicate that decidual CD8+ T cells are largely antigen-experienced cells, on a path of enhanced differentiation (24). The progress of pregnancy and likely the chronic exposure to allo-antigens favor the accumulation of EM cells with a highly differentiated phenotype, with high expression of inhibitory receptors, which however remain as EM3 rather than transitioning to EMRA. Many of these previously called EM and EMRA cells express markers for tissue residency, indicating that they are retained in the tissue, instead of recirculating to blood and that previous observations about decidual EM cells may apply to more recently identified TRM cells.

The accumulation of antigen experienced CD8+ T cells in the uterus opens questions about their antigen specificity. The clonal repertoire of the TCR of decidual CD8+ T cells indicated that clonally expanded EM CD8+ T cells were more abundant in the decidua than in blood (27, 44) although it could not be tested if an expansion occurred in response to fetal-derived antigens. Along with these findings, CD8+ EM T cells expanding clonally in decidua during healthy pregnancy expressed high levels of PD1 underpinning the importance of the modulation of CD8+ T cell responses to promote immune tolerance (27). In line with these observations, the detection of CD8+ T cells specific for peptides derived from the HY chromosome allowed to identify a local expansion of fetal (HY) specific CD8+ T cells particularly in the decidua of women carrying male but not female offspring (24). HY-specific T cells were enriched in EM and EMRA subsets expressing high levels of PD1 and CD69 indicating a highly differentiated and tissue resident-like profile, likely in response to local antigen stimulation (24). Also compared to blood, an enrichment of CD8+ T cell clones specific for the frequent human cytomegalovirus (HCMV) and Epstein Barr (EBV) viruses (38) has been reported, pinpointing an increased capacity of decidual CD8+ T cell to fight these maternal infections.

In conclusion, while particular clones of CD8+ T cells, e.g. specific for fetal and viral antigens, are enriched at the decidua, their function may be modulated by the expression of PD1 and other inhibitory receptors to promote fetal immune tolerance. Intriguingly, it is to date unclear whether these selected clones are specifically recruited to the uterus, or if they undergo clonal proliferation locally in an antigen specific or unspecific fashion (45).

During T cell development in the thymus, the incomplete deletion of self-reactive T cells makes further peripheral tolerance checkpoints crucial (46). Those include the generation of distinct responsive states to a given stimulus, that is, its functional state which can confer tolerance in the case of quiescence, ignorance, anergy, exhaustion, dysfunction and senscence. Hence, CD8+ T cell fate could be considered to follow two parallel and overlapping developmental courses/trajectories, one being the differentiation of a naïve cell to a variety of effectors and memory cells, the other being the functional state of the cell. These functional states may hold relevance at the feto-maternal interface to support a successful pregnancy, as Sir Peter Medawar’s hypothesis about the anatomical separation between maternal immune cells and fetal cells has been refuted. Tolerance states can be recognized in diverse CD8+ T cell subsets, thus rather than discussing how these states are attained, we will revisit current information about their presence and importance in the decidua.

Decades before the discovery of the CD4+ Treg cells, CD8+ T cells embarked as the first T cell subset with a suppressive potential (62). However, difficulty in confidently distinguishing them phenotypically from other CD8+T cell subsets halted the progress of this research topic while research on CD4+ Treg cells bloomed following the identification of the transcription factor Foxp3 (63). In recent years, findings on suppressive CD8+ T cells, also called CD8+ Treg cells have underscored their importance in biological contexts such as in tumor tolerance, organ transplantation, autoimmune diseases in mice and humans (64–70). Yet, due to the overlapping characteristics with other effectors, memory and exhausted T cells populations, there is no general consensus about markers to unambiguously identify CD8+Treg cells ( Figure 2 ). To date, CD8+ Treg cells have been identified based on the expression of FOXP3, the CD8aa receptor, the absence of CD28, and the specificity for hemoxygenase (HMOX)-1 derived peptides in humans (64, 71–75) as well as by CD122 positivity and restriction to Qa1 in mice. Inhibitory molecules, such as PD1, Tim3, CTLA4, Ly49, Lag3 and Tigit have also served to identify CD8+ Treg cells (63, 64). The expression of such inhibitory receptors was also described in CD8+HLA-DR+ cells with regulatory properties, whose enhanced positivity for IFNy and TNFα and degranulation ability appear related to induction of tolerance rather than exhaustion (76). Under the influence of environmental cues and of the particular immune phenomenon, diverse regulatory pathways were described for CD8+ Treg subsets (67) including suppression of proliferation of cytotoxic/autoreactive T cells, e.g. by triggering apoptosis of target cells, and by inhibition of their secretion of inflammatory cytokines. These actions may be mediated via cell-cell contact, suppression of costimulatory molecules in dendritic cells, Fas/FasL pathway, secretion of inflammatory (perforin 1, granzyme B (GrzB), IFNy) and/or immunosuppressive molecules (IL-10, TGFβ) (71). Due to the heterogeneity of the findings available, the great phenotypical and functional overlap of CD8+ Treg cells and conventional subsets ( Figure 2 ) their classification and nomenclature as “regulatory” has been brought into question (71). However, the immune regulatory properties of CD8+ T cells in important biological phenomena cannot be denied.

To date, studies on decidual CD8+ T regulatory cells, as such, are scarce. In humans, the regulatory subsets CD8+HLA-DR+CD8+ T cells (76) and CD8+CD28- T cells (77–79) were enriched in decidua, compared to peripheral blood. Additionally, decidual CD8+ T cells expressing inhibitory receptors have been referred to as having a regulatory function in the decidua in mice and humans (39, 80) colliding with those studies that defined exhausted cells by the expression of the same inhibitory markers (e.g. PD1+ Tim3). In mouse pregnancy, a regulatory function for CD8+CD122+ T cells was evident as their adoptive transfer prevented inflammation-induced intrauterine growth restriction (IUGR) restoring placenta and fetal growth (72).

Taken together, the possibility of distinguishing CD8+ Treg from exhausted cells and other subsets is urgently needed for unifying the scientific community and the progress of the field of CD8+ T cell biology. Coupling an in-depth analysis of markers with functional assessment of candidate subpopulations may serve to dissect unambiguously the potential functional states and suppressive capabilities (25).

Rather than from a widespread suppression of the immune responses, immune tolerance to fetal antigens results from a fine-tuned immune regulation. CD8+ T cell differentiation and responses are primarily influenced by the intra-uterine microenvironment. The local expression of regulatory ligands, the particular cytokine milieu (9, 24, 26, 33, 34, 37, 93), and the high levels of steroid hormones (133, 134) are critical for the establishment of immune tolerance towards the fetus ( Figure 3 ). Further, the allogenic placental trophoblasts in direct contact with maternal tissues (135) may interact with maternal CD8+ T cells to upregulate their PD1 and Tim3 expression (39), whereas decidual stromal cells can promote a differentiation of peripheral CD8+CD69+ T cell into decidual TRM (34), as shown in vitro studies.

The independent studies here reported underscore the relevance of CD8+ T cell modulation in immune tolerance towards the fetus. The abundance of CD8+ T cells at the decidua together with their particular tissue-specific attributes portraits a heterogeneous cell compartment with varied functional features ( Figure 3 ). Decidual CD8+ T cells in healthy pregnancies are cytolytic to a low extent due to low basal secretion of perforin and GrzB, despite their activated phenotype with effector and/or memory potential. Also, the potent secretory profile of decidual CD8+ T cell e.g. with regards to IFNy suggests an active response to pregnancy and a potential involvement e.g. in vascular processes in the decidua. These profiles along with the enriched expression of co-inhibitory molecules in highly differentiated cell subsets, mainly EM and TRM, appear as general features of decidual CD8+ T cells. Clearly, not only the trajectory but the functional states of decidual CD8+ T cells are tightly modulated in the intrauterine compartment, which may account as local mechanisms of tolerance towards the allogenic trophoblast. It is argued that the decidual CD8+ T cell compartment include dysfunctional, exhausted or regulatory T cells, however, a definitive characterization remains still missing. With recent advances in transcriptome analysis of single cells and in CD8+ T cell biology, overlaps of what previously was commonly accepted cellular trajectories are coming to light. This diversification is likely due to the small number of markers that can be simultaneously detected in different biological settings, such as different stages of pregnancy and pathologies, resulting in arbitrary new findings or markers and cytokines, not completely grasping the whole picture. The publicly available single cell RNA sequencing datasets on decidual immune cells include insufficient CD8+ T cells for an in-depth analysis (15). Therefore, single cell RNA sequencing tailored to CD8+ T cells, e.g. by improving cell isolation methods, including higher amounts of cells, antibody tags to identify relevant carbohydrate epitopes (CD45RA, CD57) may overcome current limitations for a better comparison to existing flow cytometry data.

Unifying their characterization and shedding light on CD8+ T cells mechanisms of pregnancy success and which components of these processes are at fault in pregnancy failure or complications, will allow to introduce new tools into current prevention strategies for pathologies such as IUGR, preeclampsia, miscarriage and preterm birth.

LH and MS wrote the manuscript. MB and ET wrote sections of the manuscript. LGl and ET performed the single cell RNA sequencing analysis. LGa, ET, and AK critically edited the manuscript. All authors contributed to the article and approved the submitted version.

This research has been funded by Deutsche Forschungsgemeinschaft (SO1413/3). MS position and MB position are financed by Deutsche Forschungsgemeinschaft (SO1413/2) and Alexander von Humboldt-Stiftung respectively. ET is financed by Deutsche Forschungsgemeinschaft (TO235/7-2) and KFO296.

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

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.