Maternal natural killer cells at the intersection between reproduction and mucosal immunity

Many maternal immune cells populate the decidua, which is the mucosal lining of the uterus transformed during pregnancy. Here, abundant natural killer (NK) cells and macrophages help the uterine vasculature adapt to fetal demands for gas and nutrients, thereby supporting fetal growth. Fetal trophoblast cells budding off the forming placenta and invading deep into maternal tissues come into contact with these and other immune cells. Besides their homeostatic functions, decidual NK cells can respond to pathogens during infection, but in doing so, they may become conflicted between destroying the invader and sustaining fetoplacental growth. We review how maternal NK cells balance their double duty both in the local microenvironment of the uterus and systemically, during toxoplasmosis, influenza, cytomegalovirus, malaria and other infections that threat pregnancy. We also discuss recent developments in the understanding of NK-cell responses to SARS-Cov-2 infection and the possible dangers of COVID-19 during pregnancy.


INTRODUCTION
The vulnerability of pregnant women to infectious disease has been explained historically by assuming that the immune system weakens during pregnancy so that it can tolerate the fetus. Indeed, the maternal immune system does not reject the fetus or its placenta despite being aware of fetal antigens and responding to them, as illustrated by antibodies to paternal Rhesus or HLA antigens, 1 or by male antigen-specific T cells found in the mothers of sons. 2 Recent research is painting a more detailed picture of the immunological changes that occur during pregnancy, challenging the view of immunosuppression during pregnancy and revealing that, surprisingly, some aspects of immune responses are even stronger in pregnancy. 3,4 The acronym TORCH refers to a set of pathogens (Toxoplasma, Other agents, Rubella, Cytomegalovirus and Herpes Simplex) that can cause abortion, intrauterine fetal growth restriction (FGR) and congenital infection. HIV, Malaria, and Influenza also cause severe adverse pregnancy outcomes. Complications of pregnancy due to infectious disease can be due to systemic or local factors, or to vertical transmission (Fig. 1). Immune responses to the pathogen may actually cause pregnancy complications. Complications may also occur without the pathogen crossing the placenta, as demonstrated by placental malaria when infected erythrocytes become sequestered in the placenta, triggering a syndrome similar to pre-eclampsia, the hypertensive disorder of pregnancy. With a focus on NK cells, we review the literature on pregnancy complications caused by major infections, including TORCH, malaria, influenza, ZIKA, and, possibly, SARS-Cov-2. NK and other immune cells may affect reproduction during implantation, gestation or parturition. We limit our discussion mainly to gestation, as NK cells are not key players during other stages of pregnancy (Fig. 2).

DOUBLE-EDGED SWORD: NK-CELL ACTIVATION DURING INFECTIONS IN PREGNANCY NK cells
Peripheral blood NK (pbNK) cells play a substantial role against viral infections. Viruses have evolved numerous mechanisms to escape NK immunity by targeting their ligand-receptor pairs, signaling pathways, chemokines, and cytokines affecting NK-cell activation and recruitment. 5 In return, NK cells respond by: (1) the recognition of "missing self" that viruses cause by downregulating MHC class I to escape T-cell responses; (2) activating receptors that sense host-and pathogen-derived ligands such as KIRs in humans, Ly49s in mice and, in both species, NKG2C, NCRs, and NKG2D; (3) antibody-dependent recognition via FcγRIII (CD16); (4) responding to IL-12, IL-15, and IL-18 and other cytokines produced by other cells. 6 We review mechanisms of both direct and cytokine-induced NK-cell activation.
Direct NK-cell activation KIR are a set of variable genes (each individual inherits 4-20 genes from each parent) that code for inhibitory or activating receptors expressed in a variegated manner on subsets of NK cells and some T cells. Inhibitory and some activating KIR bind to HLA class I molecules, while other activating KIR bind to pathogen components within HLA class I molecules. 7,8 Generally, inhibitory KIR suppress, while activating KIR enhance NK-cell functions such as cytotoxicity and cytokine production. However, inhibitory receptors can also promote NK-cell activation though the process of NKcell education, which, broadly speaking, is the acquisition of functional competence through the interaction of inhibitory receptors with self MHC molecules. 9 Each KIR gene can have dozens of allelic variants. Certain combinations of KIR and HLA genes associate with the outcome of HCV, HIV, and HCMV infections. [10][11][12][13][14][15] Trophoblast expresses a unique combination of HLA class I molecules, including non-classical HLA-E, and -G, but does not express class II molecules. While HLA-A and -B are not expressed, the only polymorphic class I on trophoblast is HLA-C. Many studies link increased ability to control infection with either weak inhibitory or activating KIR variants, 16 suggesting that enhanced activation of pbNK cells is beneficial for the host. Similarly, decidual NK (dNK) cells may require sufficient activation. 17 Pre-eclampsia is positively associated with combinations of KIR and HLA that favor inhibition and is negatively associated with combinations that favor dNK-cell activation. 18 Women pregnant with a HLA-C2 fetus have a lower chance of developing preeclampsia, 19 presumably because KIR2DS1 + dNK cells are activated upon ligating with fetal HLA-C2, and facilitate placental development. 20 KIR and HLA variants may also influence responses to infection in pregnancy, and dNK cells can be induced to destroy pathogen-infected cells. 21 HCMV infection of human decidual stromal cells (DSC) results in significant temporal downregulation of HLA-C expression; however, the expression of HLA-C is restored during the late stage of infection, which enables the activation of NK cells through KIR2DS1. 22 Maternal KIR2DS1 + dNK cells, as well as pbNK cells, exhibit higher cytotoxicity to HCMV-infected DSC, especially to HLA-C2 + cells. Interestingly, dNK, but not pbNK cells, failed to secrete pro-inflammatory cytokines and degranulate in response to HCMV-infected HLA-G + EVT. 22 KIR2DS1-mediated NK-cell activation might require modification of HLA-C by HCMV, 23 so KIR2DS1-mediated dNK-cell degranulation may occur only against HCMV-infected cells. HLA-G engages KIR2DL4 on dNK cells and LILRB1 on myeloid cells and some dNK cells, and may have the tolerogenic function. 24 HLA-G may also help infected and cancer cells to escape immunosurveillance. 25,26 Soluble HLA-G (sHLA-G) levels in the amniotic fluid of Toxoplasma gondii-infected pregnant women were greater in the group with congenital infection than in the group without vertical transmission. 27 This illustrates the conflict of the maternal immune system between Fig. 1 Influence on pregnancy outcomes by immune response to pathogens. Several aspects of the pathogenesis of a given infection can influence pregnancy outcomes: (i) tissue immunity to pathogen invasion at the fetal-maternal interface, (ii) in utero vertical transmission, and (iii) systemic immune response to infection. Both systemic and local tissue immune responses can be triggered by microbial invasion and may lead to pathogen elimination, and thus protect both mother and fetus. However, the activation of the immune system and the infection per se can lead to placental tissue damage, which might facilitate in utero vertical transmission and/or placental failure leading in turn to adverse pregnancy outcomes.
sustaining the growth of the fetus and protecting it from pathogens. Higher sHLA-G levels might modulate the immune response and prevent fetal loss but can also lead to congenital infection. HLA-G may supress NK cells, 28,29 although it can induce cytokine and chemokine production through interaction with KIR2DL4 30-32 or LILRB1. 33 HLA-E is expressed by trophoblast, maternal leukocytes, and stromal cells, where it interacts with activating CD94/NKG2C, expressed by a minority of pbNK and dNK cells and, preferentially, with inhibitory CD94/NKG2A, expressed by roughly 50% of pbNK cells and >90% dNK cells. The expression of HLA-E depends on peptides derived from other MHC class I molecules. While it may suppress dNK functions, 34 NKG2A may also promote dNK-cell functions by enhancing NK-cell education. 35 Natural Cytotoxicity Receptors (NCRs) on NK cells interact with host-as well as pathogen-derived ligands such as B7-H6, viral hemagglutinins, hemagglutinin neuraminidases, and other pathogen components. 36,37 Among NCRs, NKp46 activates NK cells, while splicing variants of NKp44 and NKp30 mediate either activation or inhibition. While dNK cells predominantly express inhibitory isoforms, pbNK express activating isoforms. 38 The switch from activating to inhibitory NKp44 and NKp30 is associated with decreased cytotoxicity and is induced by exposing pbNK cells to cytokines present in the decidual microenvironment, such as IL-15, IL-18, and TGF-β. 38 Degranulation, target cell lysis or the production of pro-inflammatory cytokines and chemokines by dNK cells can be induced via NKp46 or NKp30 stimulation and abrogated by co-engagement of NKG2A. 39 The only NCR in mice is NKp46; mice lacking NKp46 do not survive influenza A viral (IAV) infection. 40 Interestingly, NKp46-deficient mice show defective uterine vasculature adaptations to pregnancy. 41 NKG2D is an activating receptor expressed on the majority of pbNK cells and dNK cells in both humans and mice. It binds ligands normally absent on healthy cells and induced on stressed cells, therefore alerting immune cells. While these ligands are not found on trophoblast, 42 they can be induced on uterine stromal cells and/or produced by trophoblast in soluble forms to interact with both dNK cells 43 and pbNK cells. 44 In a mouse model, trophoblast-expressed NKG2D ligand engaged NKG2D on NK cells, leading to the production of IFN-γ, which, in mice, is essential for uterine vascular remodeling. 45 NKG2D is also involved in responses to infectious pathogens, as discussed below. NK cells can shape memory immune responses. 46 Human cytomegalovirus (HCMV)-seropositive individuals have an expanded CD94/NKG2C + pbNK cell population 47 and, in mice, Ly49H + spleen NK cells amplify in response to mouse cytomegalovirus (MCMV) and protect against re-infection. 48 Interestingly, during pregnancy, NKG2C + dNK cells with enhanced IFN-ɣ and VEGFα production expand and may play functional roles in subsequent pregnancies. 33 Cytokine-induced NK-cell activation Cytokines are essential for normal pregnancy but can also induce miscarriage, preterm labor, stillbirth, infertility, congenital fetal diseases and even behavioral changes in offspring. 49,50 In mouse models, the administration of IFN-ɣ or poly(I:C) (a synthetic dsRNA which mimics viral infection and stimulates innate immune responses) to dams leads to fetal resorption, 51 with the latter also causing increased serum TNF-α and upregulation of NKG2D expression on dNK cells and its ligand Rae-1 on F4/80 + decidual macrophages. 52 The fetal resorption caused by poly(I:C) is prevented by depletion of dNK cells or by either neutralizing TNF-α or blocking NKG2D. 52 Lipopolysaccharide (LPS), a component of the outer membrane of Gram-negative bacteria, also leads to fetal resorption and preterm labor. Interestingly, while Ifng −/− dams are more resistant to LPS-induced resorption, 53 Il10 −/− dams undergo preterm labor with low LPS doses that do not induce preterm labor in WT dams. LPS-induced preterm labor in Il10 −/− dams is associated with increased cytotoxicity of dNK cells and is prevented by the depletion of NK cells, neutralization of TNF-α, or recombinant mIL-10, which also restores low dNK cytotoxicity. 54 Thus, dNK cells and macrophages, TNF-α, IFN-ɣ, and the NKG2D pathway are potential components of cytokine-induced pathogenesis of pregnancy complications that may be regulated by IL-10. These studies also illustrate how dNK cells, key players in the physiology of pregnancy, can mediate fetal resorption if the decidual milieu is perturbed. Similarly, in a mouse model of fetal/ neonatal alloimmune thrombocytopenia, antibody-dependent cellular cytotoxicity (ADCC) toward anti-β3 integrin resulted in decidual infiltration of NK cells and trophoblast apoptosis, leading to miscarriage and intrauterine FGR, which could be prevented by NK-cell depletion or by blocking NKp46 or FcɣRIIIa receptors. 55 While these are clear examples of dNK cells causing pregnancy complications in pathological situations, there is no clear evidence that pbNK and dNK are implicated in recurrent miscarriage or implantation failure in humans. 56 Table 1 summarizes the known roles of dNK and pbNK cells in human pregnancy complications and mouse models. In the following sections we review these roles in individual infectious diseases.  During infection, the maternal immune system has to balance between sustaining the growth of the fetus and protecting both mother and fetus from pathogens.  59 In response to T. gondii infection, NK cells help to recruit and differentiate monocytes, macrophages, T lymphocytes, and dendritic cells (DCs) via NKG2D and IFN-ɣ, 52,60-63 a key cytokine during toxoplasmosis. 64 Indeed, animal experiments demonstrated that T. gondii infection leads to high expression of genes related to type I and especially type II interferon pathways in blood and lungs. 65,66 Activation of dNK cells in response to T. gondii may contribute to tissue pathology during pregnancy: coincubation of T. gondii-infected human dNK cells with trophoblast resulted in increased caspase 3 and 8 mRNA and trophoblasts apoptosis, which were reduced by IFN-ɣ neutralizing antibodies 67 or IL-10, which also upregulated the apoptosis inhibitor c-FLIP in infected primary human trophoblasts. 68 Animal experiments and in vitro studies with human dNK cells showed that IL-12 and TNF-α are essential for NK-cell stimulation and defense against T. gondii. 69,70 Human and murine DC and DCderived IL-12 facilitated IFN-ɣ production by NK cells 61,69,71 and decidual DC-derived IL-12 enhanced NKG2D-mediated dNK cytotoxicity toward T. gondii-infected cells. 70 Cytotoxicity of human dNK was increased when incubated with T. gondii-infected trophoblast cells. 72 T. gondi infection enhanced the expression of NKG2D, KIR2DL4, and LILRB1 mRNA in dNK cells and HLAG in trophoblast cells. 72 In vivo experiments on pregnant mice also demonstrated increased expression of NKG2D as well as NKG2A on dNK after T. gondi infection. 72 Most information about immunopathogenesis of acute toxoplasmosis during pregnancy comes from mouse models. However, acute toxoplasmosis in immunocompetent humans is generally asymptomatic, while experimental infection in mice may be severe or even fatal. [73][74][75] Generally, T. gondi-infected dams abort 72 and most experiments suggest a critical role for NK-cell derived IFN-ɣ. 61,[76][77][78] Indeed, T. gondii infection of Ifnɣ −/− mice or NK-celldepleted WT mice resulted in increased parasite load. 65,79 Type I IFNs also participated in controlling T. gondii infection and, in its absence, blood and liver parasite burden increased, 66 while recombinant IFN-β protected animals from lethal infection. 80 Also, enhanced NK-cell-mediated and perforin-independent 81 cytotoxicity was demonstrated in response to T. gondii compounds. 82,83 Despite higher parasite load in Ifnɣ −/− dams, all mice delivered live pups, whereas only a fraction of infected WT dams gave birth to live pups, varying from 5-10% to 40-60% depending on whether infection occurred during the first or second half of pregnancy, respectively. 78 Nonetheless, in comparison with WT dams, Ifnɣ −/− mothers died and when their pups were fed by infected WT foster mothers, Ifnɣ −/− pups had significantly lower body weight and died by the age of 6 weeks. 78 Interestingly, after Toxoplasma infection, dNK cells declined in Ifnɣr −/− but not in WT dams. 84 These experiments demonstrate that IFN-γ signaling in mice is key to controlling parasite loads in blood and uterus, but not conducive to progression of pregnancy and live birth. 84 High fetal loss and the reduction of fetoplacental weight were also found in dams infected with a less lethal strain of T. gondii (mutant Wh3Δrop16). The escalation of infection with this mutant strain was associated with the production of pro-inflammatory cytokines such as IL-12, IL-17, IFN-ɣ and reduced secretion of IL-10, IL-4, and TGF-β in the placenta. 85 Treatment of T. gondii-infected dams with TGF-β decreased resorption rates, improved fetal weights and diminished inflammation. 86,87 This improved pregnancy outcome was linked to a TGF-β-mediated reduction of dNKcell cytotoxicity and IFN-ɣ production. 86 Blood and liver NK cells are also the main source of IL-10 in acute toxoplasmosis: IL-10 production is induced by systemic IL-12 during infection dissemination. Animal experiments demonstrated that NK cells participate in a negative feedback loop by IL-10 secretion, which leads to reduced IL-12 production. 88 NK cells can restrict pathogen clearance via IL-10 secretion: T. gondii burden was significantly decreased in mice with impaired NK-cell IL-10 production. 89 Interestingly, Il15 −/− mice survive T. gondii infection that is lethal to WT mice. The successful survival of Il15 −/− mice was not due to better ability to clear the pathogen, but related to less severe intestinal inflammation, which was in turn associated with impairment of NKp46 + NK1.1 + CD127 − group 1 Innate Lymphoid Cell (ILC1)-mediated recruitment of inflammatory monocytes via secretion of CCL3 and the inability of inflammatory monocytes to induce damage of the small intestine. In this model, ɣδ and CD8 + T cells did not influence the course of infection. 79 In another study, ILC1s were potent IFN-ɣ and TNF producers and participated in the control of T. gondii infection. 90 Mice lacking the ILC1-master transcription factor T-bet or alymphoid Rag2 −/− Il2rɣ −/− mice had higher T. gondii loads than WT animals and recruited less inflammatory monocytes into the small intestine. 90 Adoptive transfer of ILC1s into Rag2 −/− Il2rɣ −/− mice decreased parasite burden and significantly enhanced monocyte infiltration. RNAseq analysis of blood and lungs obtained from mice at 7 days post infection with T. gondii showed signatures of increased NK cells. 66 However, other studies demonstrated a decrease of NK cells in spleen of mice after T. gondii infection, 65,91 while splenic ILC1s increased due to a conversion of NK cells into ILC1-like cells. 65 After T. gondii infection of mice, NK and ILC1-like cells had increased expressions of KLRG1, DNAM-1, Neuropilin-1, CXCR8, and CCR8. 65 Taken together, in vitro experiments with human cells and in vivo mouse models suggest that type I INF, IFN-γ and TNF-α are key to controlling T. gondii infection, however, activation of DC, IL-12 and IL-15 may lead dNK cells to damage trophoblast, and this may be counteracted by IL-10 and TGF-β.
Other infectious agents Influenza. Pregnancy is a risk factor for severe influenza. [92][93][94][95] Influenza virus does not infect or cross the placenta, however, the elevation of some systemic cytokines, especially type I and II IFNs, TNF-α, IL-1β, IL-6, and IL-15 may cause collateral damage. [96][97][98][99] There is evidence that either a lack of NK-cell function or NK-cell hyperactivity contribute to influenza immunopathology during pregnancy. Three patients with severe H1N1/09 influenza had low pbNK cell counts, with one lethal case of a pregnant women with IAV viremia, scarce infiltration of lung tissues with immune cells and a total absence of NK cells in lung samples. 100 IAV could also directly infect human pbNK cells, induce their apoptosis, and downregulate NKp46-and NKp30-mediated cytotoxicity. 101,102 Conversely, upon co-incubation of pbNK with H1N1-infected autologous monocytes, pbNK cells isolated from pregnant individuals produced higher levels of IFN-ɣ and degranulated more than pbNK cells isolated from control women. Moreover, the expression of NKp46 and CD38 were also increased during pregnancy. 103 In addition, pbNK, CD4, and CD8 T cells obtained from pregnant individuals at day 7 post vaccination with the seasonal vaccine had more robust responses to IAV infection than cells from non-pregnant vaccinated women. 104 Mouse models support clinical observations and demonstrate more severe influenza infection in pregnant than in non-pregnant animals. [105][106][107] In comparison with non-pregnant infected mice, dams had similar counts of lung-infiltrating NK cells 108 but higher IL-1β, IL-6, CCL3, and CXCL2, while IFN-ɣ was lower in lungs. 105 Also, BALs of IAV-infected dams had higher levels of proinflammatory cytokines/chemokines and contained greater counts of macrophages and neutrophils. 106 LPS causes preterm birth in a dose-dependent manner, and IAV primes dams to preterm birth with lower LPS doses. 109 Type I IFN signaling was key for the priming effect of IAV, which was also observed for LCMV and listeriosis. WT dams injected with IL-6 neutralizing antibodies or Ifnar −/− dams did not exhibit preterm birth. 109 The "double-hit" hypothesis suggests that viral infections predispose the organism to subsequent bacterial infection, leading to more severe disease. [110][111][112] Similar to humans, mouse NK cells may also be protective or contribute to the immunopathology of severe influenza. While NKp46-deficient mice were more susceptible to influenza, 40 Il15 −/− mice, as well as NK-cell-depleted WT mice, were resistant to IAV infection in comparison with WT mice. The resistance was not related to viral clearance, as IAV titers in lungs were similar in all groups. [113][114][115] Interestingly, leukocyte infiltration and productions of IL-6 and IL-12 were decreased, but IL-10 increased in bronchoalveolar lavage of Il15 −/− mice, and in NK-cell-depleted mice. 113 In addition, TNF-α and IL-6 serum levels, as well as lung NK-and CD8 T-cell infiltration were decreased in Il15 −/− mice. 114 Malaria. Women without pre-existing immunity to Plasmodium falciparum are at a greater risk of maternal and perinatal mortality. Common complications of pregnancy during malaria are low birth weight, miscarriage, preterm labor and stillbirth. [116][117][118] Pathogenesis of malaria during pregnancy is a complicated process which includes erythrocyte lysis, hormonal changes (e.g., cortisol release), red blood cell (RBC) accumulation, intravascular thrombosis, and immune responses. 119,120 NK cells contribute significantly to immune responses to malaria. 121,122 Human pbNK cells form conjugates with infected RBC and lyse them via ADCC. 123,124 Co-incubation of human PBMCs with live P. falciparum erythrocytic schizonts induced IFN-ɣ production and more than 50% of IFN-ɣ + cells were pbNK cells. 125,126 Interestingly, the association of placental malaria with certain KIR genes is modified by coinfection with HIV and CD4 T-cell counts, suggesting co-operation between NK and T cells in response to P. falciparum. 127 INF-ɣ plays dual roles in malaria pathology, being crucial for parasite clearance, but also responsible for disease severity and cerebral malaria. 128 Expression of IL-1β, IL8, and TNFA mRNA were significantly increased while IL-6 and TNFB were lower in placentas of women with malaria. TNFA correlated with placental hemozoin, a marker of malaria infection; and TNFA and IL8 correlated with intrauterine FGR. 129 Women with placental malaria had higher peripheral parasitaemia and displayed lower IFN-ɣ production by NK as well as by CD4 and CD8 T cells, in response to in vitro stimulation with PMA/ionomycin/IL-2. 130 Animal models have been widely used to study malaria infection during pregnancy. Infected pregnant rhesus macaques had detectable serum TNF-α and lost their fetuses, however, animals with no detectable serum TNF-α delivered live babies. 131 Similar to humans and monkeys, pregnant mice experienced more severe malaria and pregnancy complications than non-pregnant infected counterparts. 132,133 Upon infection with Plasmodium chabaudi, Ifnɣ −/− or Ifnɣr −/− animals had greater mortality rates, along with significantly higher parasite loads, lower serum concentrations of IL-12 and nitric oxide, diminished numbers of splenic macrophages and NK cells, as well as higher serum concentrations of IL-10. 134,135 IFN-ɣ and TNF are responsible for fetal loss in plasmodium-infected mice. Despite Ifnɣ −/− dams experiencing higher parasitaemia and more pronounced anemia, fetal loss was delayed in comparison with WT dams. 136 Plasma TNF was increased in infected dams and anti-TNF treatment decreased fetal loss, despite having no effect on parasite load. 136 Listeriosis. Pregnant women are more susceptible to infection with Listeria monocytogenes, which is more common in the third trimester and can lead to pregnancy loss, preterm birth, stillbirth, vertical transmission, and congenital diseases. 116 Transmitted via oral route, L. monocytogenes crosses the intestinal barrier, disseminates to placenta via blood and then passes the placental barrier which results in fetal infection in at least 25% of cases. 137 L. monocytogenes infects extravillous trophoblasts (EVT) and syncytiotrophoblast where two bacterial internalins, InlA (a ligand for Ecadherin) and InlB (a ligand for c-Met), independently play critical roles in fetoplacental listeriosis. 137,138 While neutrophils, monocytes, and macrophages play important roles in the innate immunity to listeria, [139][140][141] NK cells are also involved. Most of our knowledge comes from mouse models. Similar to other acute infection diseases after L. monocytogenes infection, Ifnɣ −/− or WT mice treated with anti-IFN-ɣ blocking antibodies have a greater bacterial burden than control groups. 142,143 NK cells are the main source of IFN-ɣ during listeriosis. 144,145 Also, NKp46 + RORɣt -ILC1s and NKp46 + RORɣt + ILC3s secreted IFN-ɣ and IL-22, respectively, in the small intestine and mesenteric lymph nodes of orally infected mice. 143 Interestingly, splenic and hepatic bacterial loads were lower in Il15 −/− or NK-depleted mice, which also survived at a greater rate than WT or sham-treated mice, 144,146,147 and no difference was seen between Il15 −/− and WT-infected dams, suggesting that NK cells are not required for protection against L. monocytogenes in either pregnant or non-pregnant mice, and can even be detrimental for pathogen clearance. 146 However, very recently, dNK cells were shown to kill L. monocytogenes from within EVT, by injecting anti-microbial Granulysin through nanotubes. 148 NK cells may influence the response of myeloid cells to L. monocytogenes by secreting IL-10. Indeed, L. monocytogenes stimulated IL-10 production by mouse splenic, liver, and pbNK cells at 72 and 96 h post infection, but not at 24 h, which coincides with the peak of IFN-ɣ production by NK cells. Bacterial burden was lower in infected Il10 −/− and NK-depleted mice; moreover, both Il10 −/− and anti-NK1.1 treated animals had increased splenic neutrophil and monocyte numbers at 2-4 days post infection. 144 Similarly, NK-cell IL-10 inhibited host resistance to Leishmania donovani. 149 NK cells were the main IL-10 producer among splenocytes of mice infected with L. donovani; nonetheless, the early NK-cell response to leishmania was IFN-ɣ secretion, and NK cells switched to IL-10 production only at later time points of infection. 149 Hepatitis C virus (HCV). Acute HCV infection evolves into chronic infection in 75-80% of cases, with 10-20% of HCV-positive individuals resulting in cirrhosis and hepatocellular carcinoma after 20-30 years of virus persistence 150,151 and the immune responses may both clear the virus and contribute to liver injury. 152,153 Seventy-one million people globally are estimated to live with chronic hepatitis C. 154 Although the rate of mother-to-fetus transmission is estimated to occur in 5-10% of cases, 155 and the current etiotropic treatment is effective, HCV vertical transmission is a serious healthcare problem. Transmission is associated with the use of injection drugs, high viremia, PBMCs infection, HIV coinfection, the rupturing of membranes, and amniocentesis. [156][157][158][159][160] HCV transmission may occur during delivery, while an estimated one third of infected babies may acquire HCV in utero. 161,162 Although HCV may significantly impact pregnancy outcomes, the associations between the virus and low-birth weight, intrauterine FGR or preterm birth are still controversial [163][164][165][166][167][168][169] because chronic hepatitis C can be accompanied by extrahepatic HCV-associated disorders, such as autoimmune diseases and low-grade inflammation. [170][171][172][173][174] The role and activation status of pbNK and hepatic NK cells varies throughout the course of HCV infection and differs during early and later stages of liver fibrosis or cirrhosis, 153,[175][176][177] and both defensive and pathogenic roles are considered. 177 As mentioned earlier, certain KIR and HLA-C variants have been associated with the resolution of hepatitis C. 16 It is difficult to model chronic HCV disease and the course of pregnancy during HCV infection in small animal models. 178 Therefore, it is unknown how HCV vertical transmission occurs and if immune cells at the fetal-maternal interface participate in the prevention of transmission. A small study with only five cases of vertical HCV transmission did not reveal signs of placenta pathology. 179 Human villous cytotrophoblast express receptors for HCV uptake and therefore, can be infected by HCV, 180,181 which results in upregulation of type I and III IFNs, CXCL11, and CXCL12, as well as potential dNK activation by HVC-infected trophoblast. 180 Human immunodeficiency virus (HIV). Around 38 million people worldwide are currently HIV positive. 182 HIV has two major types: HIV-1 and HIV-2, where HIV-1 is the most common type and responsible for the pandemic. 183 HIV infects CD4 T cells, macrophages and DC. 184 Before the era of highly active antiretroviral therapy (HAART), HIV vertical transmission occurred in 15-40% of cases 185,186 but HAART has reduced the perinatal transmission rate to <1%. [187][188][189] The virus can be detected in maternal blood, vaginal fluid, and breast milk. 190,191 Greater risk of mother-to-child transmission is associated with high maternal viral load in blood, low blood CD4 T-cell count, long duration of membrane rupture prior to delivery and breastfeeding. [192][193][194][195] How HIV crosses the placenta is unclear. HIV can infect trophoblast cells in vitro, however, cell-associated HIV can also cross the placenta presumably by cell-to-cell contact and transcytosis. 196 Although no specific histopathological abnormalities are demonstrated, there are indications of chorionitis 197,198 and placental membrane inflammatory lesions in HIV-positive women, though these do not correlate with increased risk of vertical transmission. 199 The above changes may be due to HIV-associated opportunistic infections. Elective cesarean section significantly reduces the risk of perinatal transmission to 2-10 % in women without antiviral therapy, 192,[200][201][202] suggesting that the majority of HIV perinatal transmission occurs intrapartum through contact with vaginal fluid or maternal blood. 196 Infants born from HIV-positive women are at higher risk of intrauterine FGR and low-birth weight 203,204 ; however, it is difficult to discern the effects directly caused by HIV from those caused by opportunistic infections and antiretroviral therapy. [205][206][207][208] Most research on NK-cell responses to HIV infection is done using human pbNK cells. The role of NK cells in controlling HIV infection and disease has been reviewed, and is related to natural cytotoxicity, ADCC, cytokine/chemokine production, and their influence on adaptive immune responses. [209][210][211] Low percentage of pbNK cells is associated with increased risk of HIV vertical transmission. 194,195 There is evidence that dNK cells may participate in the protection of the fetus against HIV transmission. Indeed, uterine-derived NK (uNK) cells (but not pbNK cells) activated in vitro with IL-12 and IL-15, inhibited infection of cell lines, PBMCs, and primary human endometrial cells with strains of HIV that use CXCR4 for cell entry. 212 The inhibitory activity was associated with uNK-produced CXCL12, a CXCR4 ligand, and supressed by CXCL12 neutralising antibodies. 212 Moreover, dNK cells significantly delayed and reduced infection of decidual macrophages with HIV strains that use CCR5 for entry. and they did so via cell-cell contact and IFN-ɣ release. 213 Zika virus (ZIKV). The recent epidemic brought ZIKV to attention due to its involvement in the development of fetal/infant central nervous system abnormalities. 214 Although ZIKV infections may result in fetal/infant abnormalities during any stage of pregnancy, the first trimester has the highest risk for adverse outcomes. 215,216 ZIKV can infect a range of cells in decidua and placenta explants such as decidual fibroblasts and macrophages, trophoblasts, fetal macrophages known as Hofbauer cells, and umbilical cord mesenchymal stem cells. [217][218][219][220][221] While trophoblast at early-and mid-gestation may be more susceptible to ZIKV, 220,222 trophoblast from full-term placenta is relatively resistant to ZIKV infection, which may be due to its constant IFN-λ production. 223 However, rather than in trophoblast, the virus predominantly replicates in Hofbauer cells, [224][225][226] which respond by secreting IFN-α, IL-6, and CXCL10. 219 Chemokines and cytokines increase modestly in the blood of patients with acute ZIKV infection. [227][228][229] Amniotic fluid from pregnant women with microcephalic babies have significantly increased levels of IL-6, IL-15, IL-17, IFN-ɣ, and TNF-α, which have the potential to contribute to fetal neurological disorders. 50,230,231 Ex vivo experiments on ZIKV-infected human decidual tissues have shown upregulation of type I and III IFN signaling pathways. 232 Studies on the role of dNK cells during ZIKV infection have not yet been conducted. Co-incubation of human pbNK cells with ZIKV-infected cells did not activate NK-cell ligand expression on epithelium cells. 233 In contrast, ZIKV led to IFNβ-mediated upregulation of MHC-I expression, which resulted in the inhibition of NK-cell killing; however, NK cells were still able to produce IFNɣ, but not TNF-α when co-incubated with ZIKV-infected cells. 233 While both myeloid and lymphoid cell populations, including NK cells, were decreased in blood of patients with viremia in comparison with healthy donors and nonviremic patients, 228 the percentages of blood NK and T cells expressing the Ki67 proliferative marker significantly increased in rhesus macaques after ZIKV infection of pregnant and non-pregnant animals. 234 Vertical transmission and fetal pathology also occur in animal models of ZIKV infection. 216,[235][236][237] Increased IFN-β production and activation of IFN-stimulated genes were found in placentas of pregnant mice after intrauterine ZIKV infection. 238 The importance of type I IFNs for protection against Zika virus was demonstrated in animal models, where Ifnar1 −/− mice are broadly used to model ZIKV infection. 216 Placentas of mice infected with ZIKV during the first half of pregnancy had signs of trophoblast destruction and vascular damage; also, FISH analysis of placental sections revealed ZIKV RNA in different trophoblast cells. 236 Rubella virus (RV) Rubella in pregnancy is associated with 85% risk of birth defects, especially if the maternal infection occurs in the first trimester. 239,240 Vaccination has substantially reduced cases, but some are still reported and globally, around 100,000-200,000 babies are born with congenital rubella every year. 241 RV infection is associated with major pregnancy complications including miscarriage, stillbirth, congenital disease, and a wide spectrum of fetal abnormalities such as cataract, hearing loss, cardiovascular and central neural system defects. 239,242 RV can be found in any fetal/infant organ. 239 RV-infected placentas display non-specific pathological changes similar to other viral infections, such as necrosis of trophoblast and endothelial cells, focal mononuclear and lymphocyte infiltration, fibrin deposition, interruptions of the syncytium and vasculitis. 243,244 Studies on immune responses to RV ex vivo are lacking. RV infection of human fibroblast cell lines in vitro resulted in significant upregulation of interferon-stimulated genes. 245 Pretreatment of cell cultures with anti-type I IFN serum before RV infection led to increased viral titers and numbers of infected cells. 245 Cytomegalovirus (CMV) NK cells play a key role in the control of herpesvirus infections. Patients with primary NK-cell deficiencies are susceptible to severe infectious diseases caused by varicella zoster virus, herpes simplex virus, cytomegalovirus, or Epstein-Barr virus. 246,247 Among all herpesviruses, HCMV plays the most prominent role in pregnancy complications, with 30-50% of primary and 0.5-2% recurrent HCMV maternal infections vertically transmittable in utero, leading to a variety of disorders including hearing and vision loss, intracranial calcifications, and intellectual disability. [248][249][250][251][252] Although infection during early pregnancy only rarely leads to congenital infection, it can lead to severe adverse pregnancy outcomes. 253 The cellular tropism of HCMV is very broad, including epithelial and endothelial cells, smooth muscle cells, neurons, dermal fibroblasts, and myeloid cells. 74,254 Examination of HCMVpositive placentas as well as ex vivo-infected decidual and villous explants revealed that HCMV is present in DCs, macrophages, endothelial cells, fibroblasts, cytotrophoblast, and rarely in syncytiotrophoblast, where it does not replicate. [255][256][257][258] Immunohistochemical analysis of HCMV-positive placentas showed leukocyte infiltration and the presence of CD8 + T and CD56 + NK cells. 256,259,260 Co-incubation of human dNK cells with HCMV-infected fibroblasts changed dNK receptor repertoire as well as their secretory profile, thus restoring cytotoxicity. Activating NKG2D and CD94/NKG2C/E receptors, but not NKp30 or NKp46, played crucial roles in killing HCMV-infected autologous decidual fibroblasts. 260 HLA-E and MICA/B expressions on HCMVinfected DSC were not impaired, while HLA-C expression was downregulated. Because of this HLA-C downregulation, inhibitory KIRs on NK cells could no longer be engaged, and therefore killing of virus-infected DSC might occur via missing self-recognition. 261 Accordingly, HLA-G expression on trophoblast is also impaired during HCMV infection, 255 which might activate NK-cell cytotoxicity because the inhibitory LILRB1 receptor on NK cells is no longer engaged. 262 However, the downregulation of HLA-G expression by EVT was not observed in a more recent study. 22 It is important to note that protein and gene expression profiles of human dNK cells, as well as their ability to respond to infection, significantly differ through pregnancy. During incubation with HCMV-infected DSC, first trimester dNK cells degranulates similar to pbNK cells, while the degranulation capacity of term pregnancy dNK cells is significantly declined. 261 Systemic immune responses to cytomegalovirus have also been described, where patients with acute symptomatic HCMV were found to have increased numbers of NK, T, and B cells in their blood. 263 Although the expression of activating NKG2C, NKG2D, and CD38 as well as inhibitory NKG2A receptors were increased on pbNK, degranulation and IFN-ɣ production by pbNK was similar to control groups. 263 Pregnant women with confirmed HCMV fetal infection had elevated serum CXCL10 and elevated amniotic fluid levels of TNF-α, IL-1β, IL-10, IL-12, IL-15, IL-17 and CCL2, CCL4, and CXCL10. 264 HCMV infection increased IFN-ɣ and CXCL10 mRNA in human decidual tissues and CXCL10 in placental villi. 265 Anti-TNF-α antibodies added to trophoblast cultures at the time of HCMV challenge significantly reduced trophoblast apoptosis. 266 Acute murine cytomegalovirus (MCMV) infection of pregnant mice results in immune cell infiltration and focal necrosis of placental tissues as well as reductions of placental and fetal brain weights; however, vertical transmission of MCMV in mice does not occur. [267][268][269] Herpes simplex and other herpesviruses NK cells have numerous tools to combat herpesviruses, which have evolved escape mechanisms to establish latent infection. 5,270,271 Blood CD56 dim NK cells decrease in the second and third trimesters of pregnancy, 4 and this may predispose pregnant women to herpesviruses. Despite two studies reporting pregnancy loss associated with HSV-infection in the first trimester of pregnancy, 272,273 there is not enough evidence to conclude that HSV causes adverse pregnancy outcomes. The importance of NK cells for the control of HSV-2 infection was demonstrated in mouse models, using alymphoid Rag2 −/− Il2rg −/− and NK-deficient Il15 −/− animals, which did not survive intravaginal HSV-2 infection. 274 The evidence for the involvement of HHV-6 and HHV-7 in pregnancy complications is clearer. HHV-6A DNA was expressed in endometrial samples of 43% of patients with idiopathic infertility, while no virus was detected in control women. 275 Interestingly, endometrial NK cells from HHV-6A-positive infertile women displayed higher cytotoxity toward HHV-6A-infected endometrial epithelial cells and increased levels of surface CXCR3, CX3CR1, CCR2. 276,277 Patients with spontaneous abortion had four times higher titers of anti-HHV-6 IgG and IgM in comparison with control pregnant women. 278 A higher percentage of miscarriages and premature deliveries were observed in pregnant patients with pityriasis rosea, which may be related to the reactivation of endogenous HHV-6 or HHV-7. HHV-6/7 DNA was detected in plasma, placenta and/or fetal tissues of 36% of patients with adverse pregnancy outcomes. 279,280 Shedding of HHV-6 /7 during pregnancy and vertical transmission of HHV-6 have also been reported. [281][282][283][284] However, it is not clear whether HHV-6 actually crosses the placenta, as chromosomal integration in the fetus can also occur. Inherited chromosomally integrated (ici) HHV-6 can indeed reactivate and cause harm to the fetus and newborn. 285,286 Since HHV-6/7 shedding was associated with a variety of diseases such as multiple sclerosis, chronic fatigue syndrome, AIDS, organ transplant complications, autoimmune diseases, cancer and many others, 287,288 a substantial challenge is to establish whether HHV-6/7 is actually an etiological factor for these pathologies and not just a harmless satellite finding of virus replication due to loosened immunosurveillance. Nonetheless, recently published data demonstrate that infection of fetuses with iciHHV-6-positive predisposes their mothers to pre-eclampsia independently of the parental origin of iciHHV-6. 289 Interestingly, HHV-6 infection suppressed angiogenesis and lymphangiogenesis in vitro, 290,291 and this might affect placenta development. The role of dNK cells in HHV-6 infection during pregnancy is yet to be examined.

EMERGING INFECTIONS SARS-CoV-2
The COVID-19 pandemic had affected more than 38 million people worldwide by the middle of October 2020. 292 SARS-CoV-2 may affect a variety of physiological and pathological processes, including gestation and pregnancy outcomes. Both SARS-CoV and particularly MERS-CoV, the agents of previous coronavirus pandemics, induce serious pregnancy complications, including intrauterine death, FGR, and maternal death. [293][294][295][296][297] Although there are some reported cases of SARS-CoV-2 vertical transmission, a thorough investigation is required to confirm this. [298][299][300][301][302] Other characteristics of COVID-19 pathology may be relevant to possible unfavorable pregnancy courses/outcomes. For example, COVID-19 associates with hypercoagulability and endotheliopathy, 303,304 which can potentially impact placental blood flow. Although histopathological changes in the placenta of SARS-CoV-2 infected women have been described, their causal association with SARS-CoV-2 infection is unclear. 299,305,306 The cytokine storm characteristic of some cases of COVID-19, 307,308 might indirectly mediate pregnancy complications and fetal developmental pathology. 49,50 A recent meta-analysis and a systematic review suggests good outcomes and no immediate concern for SARS-CoV-2 infections in pregnancy, although a high degree of statistical heterogeneity of the data makes it difficult to draw definitive conclusions 309 and cases with pregnancy complications are likely underreported. Studies on large cohorts of SARS-CoV-2 positive pregnant women with long-term follow-up of the babies are required.

CONCLUDING REMARKS AND FUTURE DIRECTIONS
There are still gaps in our knowledge of both systemic and local immune responses to pathogens during pregnancies. For example, while the changes in symptoms of rheumatoid arthritis and multiple sclerosis during pregnancy clearly illustrate the immunological alterations occurring during gestation, how exactly these changes affect the outcome of pregnancy during infections is unknown. However, new technologies such as mass cytometry and single-cell RNA sequencing, allow the generation of algorithms that register the chronology of immunological changes during pregnancy. Deviations from the norm can then be measured during infections or other pregnancy complications to capture the pathogenesis of certain infectious diseases in pregnancy. Physiological changes during gestation may affect the course of an infectious disease even if the immune response is not altered. For example, respiratory viruses may cause more harm in pregnant women due to changes in pulmonary functions in gestation. Placental malaria discussed above is also a case in point. Mass cytometry and single-cell RNA sequencing should help to generate a map of the complexity of dNK-cell heterogeneity at the maternal-fetal interface. The challenge is to determine the role of each cellular subset. In mice, for example, it appears that tissueresident NK cells may be responsible for fetal growth, while bloodlike conventional NK cells may regulate vascular changes by producing IFN-γ. 315 In humans, dNK1 may engage with trophoblast through KIR. A subset of NKG2C + dNK1 may provide benefits in secondary pregnancy through immunological training, 33 while dNK2 and dNK3 cells may engage with other cell types in the decidua through the production of other factors such as XCL-1, which can attract both DCs and trophoblast. 316 As inflammation and infections are associated with abnormal placental vascular development 317,318 and dNK cells play a significant role in placental vascularization, one of the remaining important research questions is how placental and systemic infection can influence the vascular remodeling function of dNK cells.
The relative roles of NK and CD8 + T cells are also difficult to discern, as these cell types share many features. However, temporal changes in the frequencies of the two types of cells in the decidua suggest that tissue-resident T cells may be more important in late gestation, while NK cells may specialize in certain defense mechanisms, like the recently reported nanotube-mediated transfer of granulysin into Listeria-infected trophoblast. 148 Decidual CD8 + T cells are able to degranulate and produce pro-inflammatory cytokines in response to stimulation and are therefore competent in their response to pathogen invasion. 319 How decidual NK and T cells operate during infections and which cells they interact with can be inferred by looking at the sets of MHC class I molecules expressed on cells in the decidua. Stromal cells express HLA-A, -B -C and -E, and so are capable of both presenting antigens to CD8 + T cells and engaging with certain receptors on NK cells, like KIRs, NKG2A, or NKG2C. Trophoblast instead expresses only selected MHC molecules, if any. So, while most trophoblast express no HLA molecules at all, invasive trophoblast cells express HLA-C, a known ligand for KIRs that can also present antigens to CD8 + T cells. Invasive trophoblast cells also express HLA-E and HLA-G, which respectively engage NK receptors NKG2A and NKG2C or LILRB1 and KIR2DL4.
Another open question is the presence of a microbiome in the placenta. This is a controversial area and, while the placenta is indeed sterile, 320 colonization of the fetal intestine by microbial products may be a physiological process that could be affected by pathogens during pregnancy. Finally, another fascinating question is how endogenous retroviruses, with many of their genes expressed in the placenta and some having impacted its evolution, 321 may engage with immune cells in the decidua and thereby influence immune responses to pathogens during pregnancy.
In this review, we have discussed the ambivalent role of NK-cell immunity to different pathogens in pregnancy outcomes. A comprehensive investigation and understanding of which immune responses are detrimental or, conversely, beneficial for mother and offspring health are critical for further therapeutic development and the improvement of clinical management of pregnant women with infectious diseases.