REVIEW ARTICLE: Uterine NK Cells, Spiral Artery Modification and the Regulation of Blood Pressure During Mouse Pregnancy


  • Invited review for 30th Anniversary Issue of the American Society for Reproductive Immunology.

Anne Croy, Department of Anatomy and Cell Biology, Queen’s University, Room 915, Botterell Hall, 18 Stuart Street, Kingston, ON, Canada K7L 3N6. E-mail:


Citation Burke SD, Barrette VF, Gravel J, Carter ALI, Hatta K, Zhang J, Chen Z, Leno-Durán E, Bianco J, Leonard S, Murrant C, Adams MA, Anne Croy B. Uterine NK cells, spiral artery modification and the regulation of blood pressure during mouse pregnancy. Am J Reprod Immunol 2010

Reproductive success in mammals involves coordinated changes in the immune and cardiovascular as well as in the neuroendocrine and reproductive systems. This review addresses studies that identify potential links for NK cells and T cells with the local and systemic cardiovascular adaptations of pregnancy. The studies reviewed have utilized immunohistochemisty and in vivo analyses of vascular parameters by ultrasound, chronic monitoring of hemodynamics via radiotelemetric recording and intravital microscopy. At the uterine level, functional subsets of uterine natural killer cells were identified. These included subsets expressing molecules important for vasoregulation, in addition to those previously identified for angiogenesis. Spiral arteries showed conducted responses that could account for conceptus control of vasoactivity and mouse gestational blood pressure 5-phase pattern. Vascular immunology is an emerging transdisciplinary field, critical for both reproductive immunology and cardiovascular disease.


Mammalian gestational success is a dynamic, exquisitely coordinated process involving systemic and local changes in the maternal neuroendocrine, reproductive, cardiovascular and immune systems. The mothers’ roles during pregnancy are provision of an optimal environment for implantation and of adequate nutrition for the conceptuses’ changing demands until parturition, including removal of waste products. Maintenance of maternal cardiovascular homeostasis is critical during the interval of gestation. Studies of cardiovascular disease have defined roles for both adaptive and innate immunity in onset and progression of pathology. Indeed, immune-modulating therapies are being assessed for their clinical value in arrest of cardiovascular disease progression. Studies that address potential links between the immune system and maternal cardiovascular adaptations of normal pregnancy are the focus of this review, particularly as instructive to our understanding of uterine natural killer (uNK) cell-based functions in pregnancy and in maternal vascular remodeling.

Uterine natural killer cells

uNK cells are a transient, differentiated population of NK lymphocytes, defined by both location and specialized functions as being distinct from circulating NK cells. uNK cells have been especially well studied in women and in mice.1–3 uNK cells differentiate within implantation sites during endometrial decidualization. This has been linked with the onset of interleukin (Il)-15 synthesis by endometrial stromal cells.4–7 In mice, uNK cells appear post-implantation while in humans their appearance is pre-implantation and begins during the late secretory phase of each menstrual cycle. In both humans and mice, uNK cells have angiogenic functions8–11 and the cell population declines from mid pregnancy.10,12 uNK cells secrete vascular endothelial growth factor, placental growth factor and other angiogenic molecules.8–11 Co-transplantation of human CD56bright uNK cells and choriocarcinoma cells subcutaneously into immune deficient nude mice showed that human uNK cells promote more vessel formation and trophoblast invasion than CD56bright cells from peripheral blood of the same woman. In mice, absence of uNK cells results in the failure of spiral arterial (SA) modification between gestation day (gd)9–10. A similar function has been more difficult to address in human implantation sites because of the more extensive invasion of these vessels by fetal trophoblastic cells.13 However, a number of reports are appearing confirming that as in mice,14 changes in human SA occur in advance of the presence of intravascular trophoblast in that region.2,15 Studies from mice suggest that it is the secretion of interferon gamma (Ifng) by uNK cells that provides the signaling to transiently change SA from constricted, muscular, vasoactive arterioles into dilated, thin-walled vein-like structures. Because of the absence of a similar arterial change in women with pre-eclampsia, who clinically have hypertension and renal protein excretion as their presenting pathologies, it is widely held that SA modification within the uterus is essential for maintenance of normal blood pressure (i.e. mean arterial pressure; MAP) during pregnancy. This thinking logically leads to the question of whether uNK cells contribute to regulation of MAP.16–18

The principal regulator of hemodynamic function is the renal renin-angiotensin-aldosterone-system (RAS; Fig. 1). RAS molecules work in concert with the nervous system, specialized baro- and chemo-receptors within the vasculature and endocrine mediators to rapidly detect changes in electrolytes, blood pressure or blood volume. They directly or indirectly return the cardiovascular system to homeostasis by inducing vasoconstriction, vasodilation, retention or excretion of water and/or release of vasoactive compounds (angiotensin II, aldosterone). While cardiovascular homeostasis is maintained during pregnancy, the normal regulatory mechanisms are altered. Because of significant blood volume expansion, there is activation of RAS; however, the vasculature becomes insensitive to RAS products. The cause of this insensitivity is unknown, but it may contribute to the transient fall in MAP during early gestation, as the RAS is a potent vasopressor.

Figure 1.

 Diagram of Renin-Angiotensin System, illustrating major functions of angiotensin II upon interaction with the AT1 and AT2 receptors.

Human NK cells and T cells from blood were recently shown to respond to Ang II by proliferating, shifting cytokine production toward IFNG production and migration.19 Further study showed that NK cells (defined as CD56+) and T cells (defined as CD3+) expressed not only the key receptors for Ang II (AT1 and AT2), but all RAS components (renin, renin receptor, angiotensinogen, angiotensionogen converting enzyme).

Expression and in vivo function of T cell-based RAS has been shown in mice. Not only was expression of RAS confirmed but mice genetically deficient in T and B cells themselves (Rag-1 null) or transplanted by marrow from donors deleted in AT1 had, like pregnant women, blunted responses to Ang II infusion.20 We asked whether mouse uNK cells express AT1 and AT2 (Hatta, MSc Thesis, 2009, Queen’s University). Using immunohistochemistry on samples from C57BL/6 mice, we found that at gd6, ∼10% of uNK cells were AT1 reactive and ∼20% AT2 reactive. By mid-gestation, these values were ∼20 and 40% respectively. Atrial natriuretic peptide, a molecule that inhibits renin and aldosterone production and counteracts Ang II-induced effects was not found in uNK cells at gd8 but was expressed by all uNK cells studied at gd10, once SA modification had occurred. These data indicate that mouse uNK cells may have vasoactive functions in addition to the functions for the promotion and regulation of angiogenesis.

The data on expression of AT1 and AT2 as well as other markers by mouse uNK cells strongly suggest uNK cells are functionally heterogenous and should now be classified by as functionally distinct subsets. Mouse uNK cells are commonly recognized histologically by one of two stains, periodic acid Schiff’s (PAS) reagent that reacts with glycoproteins in the cytoplasmic granules of uNK cells or Dolichos biflorus (DBA) lectin that recognizes terminal N-acetylgalactosamine expressed by the cell membrane as well as by the cytoplasmic granules of uNK cells.21,22 Cells combining both DBA lectin reactive surfaces and cytoplasmic granules are not found outside of the uterus in virgin mice or mice pregnant between gd 0.5–7.5,23 indicating that there is no peripheral source for mature mouse uNK cells. PAS and DBA lectin stains are not fully coincident but only two uNK cell subpopulations can be defined in normal pregnant mice by dual staining for these markers, PAS+DBA− or PAS+DBA+.24 Only the latter cell phenotype develops at implantation sites in pan-lymphocyte deficient mice of genotype Rag2−/−/Il2rg−/− (formerly Rag2−/−γc−/−) grafted by normal bone marrow at either gd6 (Zhang and Croy, unpublished data) or at gd12.5. Thus, uNK cells with the phenotype PAS+DBA+, which are ∼50% of the uNK cells at gd6 and >80% or more by mid-gestation (gd10 and 12) are thought to represent cells differentiated from uterine homed precursors while PAS+DBA− cells may arise from in situ progenitors. A number of angiogenic functions differ between these two uNK cell subsets. The PAS+DBA− cell subset appears to be regulatory and encompasses Ifng and Il-22 producing cells (Chen, Zhang, Colucci, Croy, MS in preparation).

Spiral arteries in pregnant mice

The histological appearance of SA in pregnant normal mice is typical of arteries for at least 8 days. Rapid changes are found between gd9 and 10. The resulting arteries have reduced muscular coats and greater lumen areas. EFNB2 and EPHB4 are typical molecules used to discriminate between arterial and venous vessels. When we applied antibodies against these molecular markers to normal mouse implantation sites, SA observed at gd6.5 and 8.5 were EFNB2+EPHB4−, the phenotype expected for arteries. At gd10.5 however, mixed expression of these molecules was seen even in single vessel cross-sections, and by gd12.5 the vessels had strong EPHB4+ (venous) expression. It was of considerable interest that uNK cells expressed both markers. At gd6.5, uNK cells were EFNB2+EPHB4− but, from gd8.5, had EPHB4+ expression, which was strong at gd10.5–12.5. Indeed, at gd10.5, both vascular phenotyping markers were more strongly expressed by uNK cells than by blood vessels. The functional significance for uNK cells of this expression remains undefined. In alymphoid mice of genotype Rag2−/−Il2rg−/− that do not experience gestational SA modification, SA remained EFNB2+EPHB4− from gd6.5 to 12.5, suggesting retention of arterial function and myometrium became unusual with very strong expression of EFNB2 and for EPHB4.25

Clinical literature suggests that absence of SA modification is associated with vasospasm in pregnancy and with placental and possibly fetal hypoxia.26–28 The retention of arterial phenotype by SA in alymphoid mice led us to examine spiral and various upstream arteries across pregnancy of normal and Rag2−/−Il2rg−/− mice by ultrasound using an RMV704 40 MHz probe (Vero770; VisualSonics, Toronto, ON, Canada). There were no detectable differences in blood velocity, pulsatility or resistance (Fig. 2). This is inconsistent with our measured, histological differences in SA at gd12 that provided a calculated structural resistance indicating there would be a 40% increased resistance in SA in alymphoid mice.18 Further, we infused mated, normal and Rag2−/−Il2rg−/− mice on BALB/c backgrounds with Hypoxia Probe™-1 to evaluate conceptus and maternal tissue hypoxia at killing. Unexpectedly, when a full time course of study (gd6–18) was reviewed, there were no significant differences in reactivity with the probe between normal or Rag2−/−Il2rg−/− placentas or fetal tissues (Leno-Duran et al., submitted). This suggests that retention of arterial phenotype does not limit perfusion of implantation sites through mid- to late-gestation. It is of considerable interest that relative hypoxia was found in only one of five maternal tissues studied: the kidney (Fig. 3). This relative hypoxia was seen in the normal, but not in Rag2−/−Il2rg−/− mice. We postulate that, in comparison to controls, Rag2−/−Il2rg−/− mice will have increased blood flow in the uterine and renal vessels to compensate for the lack of SA remodeling. Analyses of our serial ultrasound data to address this question is ongoing and it will be correlated with the systemic hemodynamic data we have obtained using radiotelemetry.

Figure 2.

 Doppler waveforms of spiral arteries of normal (upper panel) or alymphoid Rag2−/−Il2rg−/− (low panel) mice at mid-gestation (gd12). Peak systolic velocity (S) and end-diastolic velocity (D) are indicated. Hemodynamics of spiral arteries are equivocal between normal and alymphoid mice when examining blood velocity, resistance index (RI) and pulsatility index (PI), or vascular pattern. Ultrasound imaging was performed with Vevo 770 (Visual Sonic, Toronto, ON, Canada) equipped with a 40-MHz probe.

Figure 3.

 Immunohistochemical staining of kidney from gestation day 12 mice injected with Hydroxy-Probe™. (a) Immune competent BALB/c mouse and (b) alymphoid Rag2−/−Il2rg−/− mouse. Images taken at 400× times magnification.

Systemic effects of the immune system on maternal hemodynamics

Radiotelemetry is an extremely powerful research tool for continuous hemodynamic monitoring of animals using surgically implanted probes with inter-arterial solid-state catheters. This technology has been miniaturized for mice but its cost, comparative to less-accurate measurement of MAP, is high. David Merrill and Davisson et al. pioneered use of this method in pregnant mice.29 Radiotelemetric recording identified BPH/5 mice as developing late gestational hypertension. In comparison to controls, pregnant BPH/5 mice had equivalent uNK cell numbers but poor SA remodeling.17 Their placentas were significantly smaller at mid-gestation (i.e. placental insufficiency) but normalized by term. Despite their unusual late placental growth, BPH/5 fetuses were growth restricted throughout gestation and litters were smaller than in controls because of mid- and late-gestation resorptions.

We have undertaken conception to term radiotelemetric analysis of the hemodynamic changes of mouse pregnancy for normal (C57BL/6, BALB/c and outbred CD-1) and Rag2−/−Il2rg−/− females. Our study (Burke et al.) differs from most published studies, which have focused on mid-pregnancy or post-implantation and beyond using radiotelemetry or measured MAP daily or less frequently over the study time course. For normal mice with SA modification and minimal fetal loss, a five-phase pattern of MAP was found that coincided with defined stages of placental development or growth (Fig. 4). During pre-implantation development, ΔMAP was stable but heart rate increased, which is consistent with blood volume expansion and was supported by small gains in body weight. After implantation, MAP declined to a nadir at gd9. This time point is significant as it is the time when SA begins to remodel and when fusion of the allantois is completed in mice, initiating placental blood flow. Following the mouse gd9 nadir, MAP rises until gd12, where, depending on strain, MAP becomes variable but hovers about its pre-pregnancy baseline level. This pattern mimics the hemodynamic changes observed in pregnant women with a first trimester drop in MAP and a second trimester rise.30 Surprisingly, despite absent SA remodeling, gestational hemodynamics were similar in Rag2−/−Il2rg−/− mice and there was no evidence of hypertension compared with BALB/c control pregnancies. Placental weights were collected from mid-gestation to term and were equivalent until gd14, then diverged with peripartum Rag2−/−Il2rg−/− placentas being larger. These data are most simply interpreted as fetal and/or placental signals regulating the patterns of change in maternal MAP.

Figure 4.

 Mean arterial pressure patterns of mice during normal gestation, pseudopregnancy and pregnancy failure.

This interpretation is supported by hemodynamic profiles from hormonally pseudopregnant females who mated (i.e. confirmed copulation plugs) but failed to gain weight and had no palpable fetuses at gd12. These females had stable MAP between mating and post-mating day 12 when their study was terminated and they were confirmed to be non-pregnant (Fig. 4). This excludes ovarian hormones as the regulators of the pattern of change in early gestational MAP.

A further indication that conceptus-based signals might exclusively regulate maternal, gestational MAP came from two Rag2−/−Il2rg−/− dams that were unusual because they delivered small litters of predominately dead pups. From mid-gestation, their MAP profiles deviated because of the rapid onset hypertension.

Because Rag2−/−Il2rg−/− lack both NK cells and T cells, it is not possible to state from our study whether NK cells or T cells have roles in regulation of MAP. It can only be concluded that in the absence of both of these lymphocyte subsets, a normal gestational pattern of MAP is maintained. To examine the role of T cells, (we have not addressed B cells because Guzik et al. showed by adoptive transfer that B cells were ineffective modulators of vascular responses to Ang II infusion20) gestational radiotelemetric studies have been completed in some syngeneically mated T and B cell deficient Rag2−/− and in NOD.CB17-Prkdcscid (NOD.scid) females. A pattern similar to that shown in Fig. 4 was observed for Rag2−/−, indicating that loss of T cells with maintenance of NK cells and SA modification does not modulate the normal pattern of MAP in mice. This is in contrast to a challenge with Ang II in non-pregnant mice in which MAP responses were statistically blunted. Thus, pregnancy may mask the importance of T cell regulatory contributions. In contrast, gestational MAP in NOD.scid had a unique pattern that was unusually stable with little variability beyond pre-pregnancy baseline MAP. It should be noted that because of background strain, NOD.scid mice lack normal NK cell function in addition to their T cell deficiency. Gestational telemetric study of the newly described E4BP4 mice that lack NK cells but are T cell sufficient may resolve this question.31

It seems difficult to conceive how an early conceptus could regulate its mother’s MAP, and we are addressing the alternate hypothesis that endometrial decidualization is the regulatory process, using radiotelemetric study of pseudopregnant mice with induced deciduomata.32 However, one method for regulation of upstream vessels by downstream cells is known as reverse conductance. Conducted responses are vasomotor signals initiated at discrete locations and transmitted along the vascular wall to coordinate vasomotor responses at sites distant from the site of local stimulation (for review see Schmidt et al., 2008).33 They are often considered a means by which tissues can induce changes in blood flow to match metabolic demand via signaling to the upstream (against the flow of blood) feed vasculature. The vasoactive signals travel through gap junctions between adjacent endothelial cells, vascular smooth muscle cells and/or myo-endothelial junctions. While research has favored the investigation of upstream conducted responses because of their importance in regulating blood flow into the tissue,34–43 evidence indicates that vascular signals are conducted in both upstream and downstream directions.34, 35, 40, 44, 45 The fetus may take advantage of this inherent machinery to induce upstream maternal vasomotor responses in accordance to its own developmental needs. The heavy invasion of fetal tissues in rodents and humans places fetal cells in very close proximity to the maternal vasculature, including intramural positions and replacement of maternal vascular endothelial cells with fetal trophoblast. Therefore, all necessary components for fetal regulation of maternal blood flow via upstream conducted responses are present. To address whether the endometrial vasculature was capable of transmitting conducted responses, we performed preliminary work using an intravital microscopy technique developed in our lab.46 Briefly, the technique allows visualization of the intact, blood-perfused endometrial vasculature of an anesthetized gd12 mouse. We locally pipetted 10−4M phenylephrine (PHE) onto the mesenteric vessel (baseline diameter = 61.4 ± 3.6 μm) feeding the implantation site and then measured resultant vessel diameters in the downstream vasculature (>1000 μm from local site) (Fig. 5a). A dye tracer (FITC) was added to the micropipette contents to help ensure that the drug did not flow over the downstream sites of interest. Indeed, our data indicate that signals are conducted downstream along the endometrial vasculature and that the signal strength eventually diminishes as the vasculature bifurcates into daughter vessels. While the uterine and arcuate arteries (baseline diameters = 71.0 ± 5.3 μm and 68.9 ± 8.7 μm, respectively) showed similar vasoconstrictions as measured locally in the mesenteric feed (Fig. 5b,c), constrictions decreased in magnitude in the more distal radial (baseline diameter = 71.7 ± 7.1 μm), spiral (baseline diameter = 98.6 ± 8.8 μm) and basal (baseline diameter = 39.8 ± 5.4 μm) arteries (Fig. 5d–f). These trends were mirrored when vessel diameters were normalized to their branch order maximum diameters. Additionally, we directly exposed each level of the endometrial vasculature to 10−4M PHE to determine their capacity to constrict to direct application of the drug for comparison to the conducted response capacity (Fig. 5b–f). These data show that the endometrial vasculature is capable of transmitting conducted responses. Given that conducted responses can be bi-directional, our data indicate that the machinery is in place for the fetus to directly regulate maternal blood flow without relying on hormonal delivery of signals via the systemic circulation. Therefore, it is entirely plausible that the fetus can mediate maternal blood flow by initiating conducted responses at the maternal–fetal interface that travel in a retrograde manner to effect hemodynamic changes in the maternal feed vessels.

Figure 5.

 The endometrial vasculature is capable of conducting vasoactive signals along the vascular wall. (a) Schematic representation of the endometrial vascular sites of interest. Indicated are the site of drug application (mesenteric artery), and the endometrial vasculature located >1000 μm downstream. (b–f) Changes in vessel diameters at the site of drug application (Local), the induced change in vessel diameter at the branch order of interest (Downstream) as well as the change in vessel diameter induced by direct application of 10−4M PHE to the downstream site (Direct). *Indicates significantly different compared to ‘Local’, # indicates significantly different compared to ‘Direct’, both at P < 0.05.


This review summarizes recent work from our laboratory initiated to explore the relationships between lymphocytes and early gestational changes in the cardiovascular system. We find this an exciting and important emerging area with significant clinical relevance. We have only very preliminary insights into these potential relationships, but have shown that NK cells from lymphoid organs do not show the dynamic changes in or in some cases even the expression of the vascularly related molecules that we have addressed. These studies challenge long-established ideas concerning outcomes from incomplete SA modification and they indicate that pregnancy is a state that alters the immunoregulatory effects of T cells, particularly Th17 cells, on MAP that have been reported in mice.47 It is important that cross disciplinary studies continue to develop in reproductive immunology because formal demonstration of immunological involvement in cardiovascular hemodynamics provides new approaches for therapeutic interventions such as inhibitors to block specific cytokine signaling pathways through selective inhibition of tyrosine kinases or other signaling.48 The incidence of pre-eclampsia remains stable at ∼3–5% of all human pregnancies, making hypertensive disease important to pregnancy. Understanding and developing in vivo models that will effectively address the importance of relationships of lymphocytes in cardiovascular responsiveness will also have the potential to be important for men and women who are not pregnant.


These studies have been supported by awards from the Natural Sciences and Engineering Council Canada, the Canadian Institutes of Health Research, the Canada Research Chairs Program, the Canadian Foundation for Innovation. Contributing research trainees have been supported by CIHR (SDB), NSERC (AC), Programa de Formación del Profesorado Universitario (ELD), Queen’s University (KH), and by CAPES and CNQp (JB).

Declaration of contributions

BAC, MAA and CM designed and supervised these studies and contributed to writing of the manuscript. SDB contributed to writing of the manuscript and provided Fig. 1. SDB, VFB and JG conducted and analyzed the radiotelemetric studies and provided Fig. 4. AC, KH and JZ conducted the immunohistochemical studies. KH, ELD and JB conducted the hypoxia probe studies and provided Fig. 3. JZ conducted the ultrasound.