Angiogenic responses in the pregnant mouse uterus under uNK cell deletion and hypoxia

Inadequate uterine vascular remodeling is associated with pregnancy complication. Pregnancy specific uterine Natural Killer (uNK) cells produce angiogenic factors such as vascular endothelial growth factor (VEGF), nitric oxide (NO) and interferon-γ (IFNγ), but hypoxic effects of uNK cell functions are unknown. Here, we investigated the effects of hypoxia on murine uNK cell driven vascular remodeling and on pregnancy outcome. At gestational day (gd) 7.5, normal CD1 mice and geneticallymodified IL-15-/mice that lack uNK cells were exposed to 48 h of 420 Torr hypoxia. Pregnancies were then assessed at gestation day (gd) 9.5 for decidual gene and protein expression, histopathology or litter size at birth. Hypoxia delayed CD1 fetal development, but term deliveries occurred, with the exception in IL-15-/uNK deficient mice. VEGF protein and gene expressions were down-regulated, but not abrogate in the uNK deficient mice, suggesting an alternative source of this angiogenic factor independent of uNK cell. High eNOS/iNOS expression seems to be uterine compensatory response in the uNK deficient mice and it is probably by increasing NO release to stimulate local angiogenesis. The reduced successful pregnancy outcome is not related to uNK cell cytotoxic activity. Otherwise, the high incidence of pregnancy failure of IL-15-/animals confirmed the benefit of uNK cells in the homeostasis of pregnant uteri.


Introduction
Hormone and cytokine driven morphological and physiological changes occur in the uterus during interstitial embryo implantation and development of the hemochorial placenta.Stromal decidualization and vascular remodeling, including spiral arterial distention at implantation sites, play key roles supporting adequate trophoblast invasion and placental organogenesis (Burton et al., 2009;Liu et al., 2017).In humans, failure of spiral arterial adaptation is associated with complications that include recurrent miscarriages, intra-uterine growth restriction (IUGR) and preeclampsia (Burton et al., 2009;Kalkunte et al., 2010;Avagliano et al., 2011;Alecsandru and Garcia-Velasco, 2017;Franasiak and Scott, 2017).Spiral arterial modification is thought to be initiated in humans and mice by unique lymphocytes, the uterine Natural Killer (uNK) cells.
A shared feature of human and mouse pregnancy specific uNK cells is their ability to produce angiogenic and Lippe Braz.J. Biol.Sci., 2018, v. 5, No. 9, p. 13-24.
These studies underlie the current paradigm that suggests uNK cells have critical angiogenic and vascular remodeling functions that promote successful pregnancy.Importantly, pregnancy with litters of normal size progress to term in uNK cell deficient mice, a finding attributed to maternal cardiovascular adaptations.
The present work assessed effects of 48 h hypoxia during the postimplantation, pre-placental stages between gd 7.5 to 9.5 on angiogenic and vasoactive molecules on normal CD1 and NK cell deficient IL15 mice pregnancy outcomes.420 Torr hypoxia provoked developmental delay in conceptuses that was reversed and not fatal when the dams were returned to normoxia.However, in mice with uNK cell deficiency, recovery was not observed.The uNK deficient IL-15 -/-mice downregulated uterine VEGF expression and increased the Nos2 and Nos3 expression in comparison with CD1 mice.This aberrant local compensatory response may adequate local vascular remodeling to support at least 60% of successful pregnancies.

Animals
Randombred +/+ (CD1) and IL15 -/-mice on the CD1 background were purchased from Charles River (USA) and syngeneically mated.The morning of vaginal plug presentation was considered gestational day (gd) 0.5.All procedures for animal handling and experimental protocols were approved by the University of Kansas Medical Center Animal Care and Use Committee (Protocol No. 2007-1647) and by Committee of Experimental Animal Care and Use, Institute of Biology, University of Campinas (Protocol No. 2111-2).

Normoxic and hypoxic conditions
The first study manipulated pregnant CD1 and Il15 -/-mice (n = 10/genotype) on gd 7.5.Mated females pregnant were placed in hypobaric chambers (Biospherix Instruments, USA) and exposed to 420Torr (11% oxygen) for 48 h as reported by Ho- Chen et al. (2006).Pair-fed control gd 7.5 mice (at least five/group) were held in the same chamber at normoxia (760 Torr, 21% oxygen) for 48 h.These animals were immediately sacrificed by cervical dislocation for uterine tissue collection.The second group of pregnant CD1 and IL-15 -/-mice received the same treatments but were allowed to give birth to quantify litter size and neonatal viability.

Tissue sampling
Implantation sites were enumerated and classified as normal (N), atrophic (Atr) or hyperemic (Hym).Atrophic sites were that 20%-40% smaller in size than normal embryo implantation sites and of normal or pale color while hyperemic sites were dark red.Implantation sites with normal appearance were dissected to isolate fetal tissue-free mesometrial endometrium for RNA and protein assays or for freezing (-8 °C) in dry ice-cooled heptane (Fisher Scientifics, USA) for subsequent cryosectioning.
For uNK cell identification and quantification Dolichos biflorus (DBA) lectin cytochemistry (Paffaro Jr. et al., 2003) and perforin (PRF) immunocytochemical staining (rabbit polyclonal anti-perforin; Torrey Pines Biolabs, Houston, TX) (Ain et al., 2003) were used.Endoglin (EDG) (rat polyclonal anti-endoglin; Developmental Hybridoma, Iowa, USA) immunostaining was used to identify proliferating endothelial cells and TROMA-I immunostaining (mouse monoclonal anti-TROMA-I, Developmental Hybridoma, Iowa, USA) was used to identify trophoblast cells (Gomes et al., 2012) In sections placed with the mesometrial side of the uterus to the top, the placental area delimited at the bottom by the amniotic cavity and at the top by TROMA-I + trophoblast giant cells facing the decidualized endometrium was measured using Image J software (NIH, USA).Taking CD1 placental areas as reference, ratios of IL-15 -/-placental areas were calculated and used as trophoblast invasive indices.The statistical analysis from 3 independent assays of each experimental group was determined by ANOVA.

Enzyme-Linked-Immuno Sorbent-Assay (ELISA)
VEGFA and IFNγ concentrations were measured in homogenates of fetal tissue-free mesometrial tissue using mouse VEGFA-ELISA KIT and IFNγ-ELISA KIT (cat.KMG0112 and KMC4021, Invitrogen, Carlsbad, CA, respectively), according to the manufacturer's instructions.The statistical significance of data obtained from 5 independent assays of each experimental group was determined by ANOVA.
The Real-Time PCR (ABI7500/ 7500 software) transcriptome amplifications were performed using SYBR Green PCR Kit (Qiagen, USA) and reactions according to manufacturer's instructions for Nos2, Nos3, Tnfα, Tnfr1, Tnfr2, Vegfa, Vegfc and Perforin (Prf1) genes using 18S RNA as the endogenous housekeeping gene.Sense and anti-sense primer sequences were designed using the Primer 3 software program (Table 1).Quantitative RT-PCR data from four independent experiments of each group were used to assess statistical analysis by ANOVA.

Gross postmortem assessments of hypoxic gd 9.5 implantation sites does not coincide with litter sizes at birth in deficient in Il15 mice
The ratios between normal (N) and abnormal (either Atr or Hym) gd 9.5 implantation sites in CD1 and Il15 -/- pregnancies are given in Table 2. Using CD1 in normoxia as reference, the ratio of Atr-sites doubled in CD1 by hypoxia.
No Hym-sites were present in CD1 suggesting onset of conceptus delay was recent.In contrast, no Atr sites but only Hym sites were present in Il15 -/-mice.For uNK cell deficient strain, the proportion of failed sites was 5 to 6 fold greater than in CD1 under normoxia and hypoxia.These data suggest that conceptus delay or failure was more advanced when uNK cells were absent from decidua whether or not hypoxic insult was additionally present.
Pregnancy success evaluated as litter size at term compared to CD1 was reduced significantly under normoxic conditions in the gene-deleted strain (Table 3).Hypoxia did not alter CD1 litter size, but significantly reduced in the absence of uNK cells.The reduced number of pups in the IL-15 -/-animals coincides with high proportion of Hymsites found at gd 9.5.Similarly, the same litter size between normoxia and hypoxia conditioned CD1 animals suggest the embryo from Atr-site affected by 48 h hypoxia at gd 9.5, survive and growth to the term.Table 3. Litter size of CD1 and IL15 -/ mice at term under normoxia (N) and hypoxia (H).

Influence of hypoxia and IL-15 deficiency on implantation site histopathology
Both DBA lectin and perforin immunocytochemistry independently identified uNK cells distributed in the MLAP and DB of uterine embryo developing sites of CD1 mice.As expected, uNK cells were completely absent in the IL-15 -/-pregnant uterus at gd 9.5 (Figure 1).
Analyses of DBA + or PRF + uNK cells (Table 4) revealed greater numbers of DBA + or PRF + uNK cells in the DB under hypoxia than normoxia and lower numbers in the MLAP.This shift suggests recruitment of premature uNK cells into decidual regions to supplement the function of the decidual uNK cell subset.The percentage and standard deviation represents quantification from five independent experiments.
Based on endoglin+ endothelial cell distributions the endometrial vascular networks in CD1 mice did not show marked differences between normoxia and hypoxia.Endoglin-defined vascular networks were poorly remodeled in normoxic IL-15 -/-endometrium.The fetal blood vessels of labyrinthine placental layer were also endoglin-reactive.
Trophoblast giant cells (TGC) are the outermost layer TROMA-I positive cells of placenta in contact with decidua in all strains.The placental area of TROMA-I positive cells were measured and the trophoblast invasiveness index was established using CD1 in normoxia as a reference.All mutant placentas were significantly larger than CD1 placenta and those from IL-15 deficient mice being 4 fold larger than CD1 placenta under normoxic conditions.TROMA-I + TGC intruded deeply into the mesometrial endometrium of IL-15 -/- animals both in normoxia and hypoxia, and corresponding trophoblast invasiveness index was 3 and 5 folds higher, respectively.Hypoxia did not statistically alter placental size within a strain.Taking into account the limited development of decidua and poor endoglin expression in the endometrium of uNK deficient IL-15 -/-mice, these abnormal conditions at the maternalfetal interface seems to increase Lippe Braz.J. Biol.Sci., 2018, v. 5, No. 9, p. 13-24.
trophoblast invasion and placenta hypertrophy.

Influence of hypoxia and IL-15 deficiency on levels of IFNγ and VEGF in implantation sites
VEGF and IFNγ are molecules related to angiogenesis and produced by uNK cells in normal pregnancy (Ashkar et al., 2003;Lima et al., 2012).Protein quantification of VEGF showed the absence of uNK cells reduced the VEGF at the maternal-fetal interface to 23% if compared to CD1 mice in normoxia.The hypoxia did not change significantly the IFNγ concentration in the maternal-fetal interface of CD1 mice.Unexpectedly, the uNK cells deficiency increased significantly the amount of IFNγ in the uterus both in normoxia and hypoxia, suggesting the uNK cell is not the main source of this cytokine in the pregnant uterus at gd10th (Figure 2).

Influence of hypoxia and IL-15 deficiency on gene expression in implantation sites
The effect of hypoxia on the expression of selected genes expressed by uNK cells (Vegfa, Vegfc, Nos2 (iNOS), Nos3 (eNOS), Prf1 (perforin), Tnfa, Tnfr1, Tnfr2 was evaluated in gd 9.5 decidual homogenates from normal mice and The Nos2 isoform gene expression showed down regulation under hypoxia in all groups.The expression of the Nos3 isoform gene was higher in the absence of uNK cells while under hypoxic conditions it was downregulated.These data clearly show the dual influence of oxygen tension and seems mobilize alternative cells.
The Tnf-α transcript expression of CD1 was similar to uNK cells deficient mice in normoxia.Under hypoxia the transcript level of this pro-inflammatory cytokine was down-regulated in the uNK cells deficient mice, suggesting no involvement in the increasing pregnancy loss of these experimental groups.However, the expression of the TNF-α receptor Tnfr1 that induces apoptosis of target cell was similar to CD1 or was down-regulated in the uNK cells deficient uterus.The expression pattern of receptor Tnfr2 looks like to seek undo the TNF-α and TNFR-1 activity in the pregnant uterus of uNK deficient and hypoxia experimental groups.

Discussion
The present work demonstrated experimentally the absence of uNK cell and 48h continuous threshold hypoxia during pre-placentation stage deeply affects vascular remodeling and increase trophoblast invasiveness at mid gestation, but compromise only partially the pregnancy outcome.The reduced deleterious effects of isolated or combined adverse conditions can be attributed to the finely orchestrated compensatory capability of the uteroplacental tissues response to regulate angiogenic molecules at the maternalfetal interface.
The 420Torr oxygen concentration corresponds to hypoxic conditions of altitude higher than 4500 meters (m) and doubled the incidence of abnormal hypotrophy of embryo implantation sites in the CD1 pregnant uterus.Hypoxia has been associated with preeclampsia in high altitude human pregnancy in which increases nitrative stress (Zamudio et al., 2007;Deng et al., 2018;Schoots et al., 2018) and increases uterine vascular eNOS and NO production in the ovines and ewes (Hempstock et al., 2003;Fujita et al., 2010).
Uterine-NK cells, uterine endothelial cells, macrophages, stromal and epithelial cells and TGC among other cells, produce NO in the mouse pregnant uterus (Chen et al., 2017;Fortis et al., 2018).The up-regulation of Nos2 (iNOS) and Nos3 (eNOS) found in the CD1 endometrium under hypoxia could be either due to increasing numbers of DBA + /PRF + uNK cells in the DB and/or activation of nitrative stress pathway of uNK cells to increase NO concentration available in the uterus.On gd 9.5 the MLAP region usually contains small agranular DBA+uNK cells and the mature granulated forms dominates in the DB (Croy et al., 2010;Lima et al., 2012).Conversely, the expression of Prf, Tnf-α and Ifn-γ genes of molecules related to cytotoxicity of uNK cell (Croy et al., 2010;Chen et al., 2012;Lima et al., 2012), were not up-regulated under hypoxia in the CD1 mice.Therefore, these cytolytic and pro-inflammatory proteins are not the main cause of increasing of abnormal Atr-implantation sites.Intriguingly, despite the high incidence of Atrimplantation sites at gd 9.5 under hypoxia, this does not affect the litter sizes at term in CD1 animals and presumes the embryo in the Atr-sites at gd 9.5 by 48h hypoxia can be rescued to normalcy at term by restoring the normal oxygen supply.
Expectedly, IL15 gene deletion abrogate DBA + and PRF + uNK cell incidence as has been reported by Ahskar et al (2003).The high proportion (5-6 folds) of abnormal Hym-embryo developing sites at gd 9.5 seems to be irreversible, different from those Atrsites of CD1, and results in reduced litter size of these animals at term independently of oxygen tension influence.These data suggest the impact of uNK cell deficiency deeply affecting normal utero-placental onset.In fact, those apparently normal IL-15 -/-embryo developing sites showed poor endoglin positive endothelial cells distribution in the mesometrial endometrium and hypertrophic placenta at gd 9.5.Consistent to this, uNK cell deficiency including the IL15 -/-mice encompass hypocellular, edematous and incomplete uterine spiral artery wall remodeling of mesometrial endometrium (Croy et al., 2003;Wu et al., 2006).Endometrial vascular remodeling is essential for normal pregnancy and dependent of angio-modulatory factors like NO, angiotensin-I and -II, VEGF and IFNγ produced by human and mouse uNK cells (Chen et al., 2012;Lima et al., 2012).
VEGF, IFNγ and, eNOS and iNOS to produce nitric oxide (NO) are all attributed as key molecule of uNK cells related angiogenesis in the pregnant uterus (Fortis et al., 2018) were not abrogated in the uNK cell deficient IL15 -/- animals.Otherwise, low VEGF protein concentration in the IL15 -/-pregnant uteri confirms the uNK cells as putative main source of this angiogenic factor, but not exclusive (Ain et al., 2003) is not the only mechanism regulating VEGF expression in the pregnant mouse uterus.This alternative uNK independent pathway ensures constitutive VEGF expression as compensatory mechanism driving the utero-placental vascular remodeling.
Endoglin is consistently present in the proliferating endothelial cell and it is stimulated by VEGF.The reduced endoglin positive endothelial cells in the IL15 -/-mice endometrium matches to the low VEGF concentration.It means poor endometrial vascular remodeling and consequently low blood and oxygen supply to the placenta that stimulate increasing of trophoblast invasiveness (Zhang et al., 2011), as were unveiled by deeply intruded TROMA-I positive TGC into the mesometrial endometrium and enlarged placenta in the IL-15 -/-pregnant uteri.This hypertrophy of placenta could mechanistically ensure the necessary blood exchange area to the embryo in the poorly remodeled endometrial vascular bed.High incidence of abnormal Hymembryo implantation detected at gd 9.5 of uNK deficient uteri looks like consequence of increased trophoblast invasiveness with disruption of uteroplacental unit (data not shown) and impaired gestation.
Furthermore, more than 40% of the pregnancy failures in the IFNγ null mice reported by Ashkar et al (2003) demonstrated the importance of this cytokine in the pregnancy.One of the potential route of IFNγ mediated uNK cells gene expression regulation is the iNOS expression to produce NO (Frérat et al., 2009).Interestingly, the uterine eNOS expression is sensitive to hypoxia and is not dependent of uNK cells, while iNOS expression is only partially dependent of uNK cells affected by low oxygen tension.
The high concentration of IFNγ in the IL-15 -/-mice is an unexpected, contradictory result to the established paradigm attributing the uNK cell as main source of this cytokine in the pregnant uterus (Ashkar et al., 2003).This discrepancy could be due to the high reproductive performance of CD1 mice strain used as background for IL15 -/-mice instead of C57/Bl6.Previous reports using C57/Bl6 background IL15 -/-mice conclude no difference of litter size if compared to B6 wild type (González et al., 2012) while our data detected Hymuterine sites during mid-gestation and significant lowering of litter size in the CD1 background IL15 -/-mice.Further investigations regarding influence of mouse strains in the reproductive performance should be carried out since most of the inbred strains have small litter sizes if compared to CD1 outbred Lippe Braz.J. Biol.Sci., 2018, v. 5, No. 9, p. 13-24.
strains and probably influences uterine response capability to support pregnancy.

Conclusions
In conclusion, this study confirms the benefit of uNK cell present at the maternal-fetal interface as key modulatory cell working for adequate uterine tissue remodeling and supporting the successful pregnancy (Burkle et al., 2010;Lima et al., 2012).Secondly, despite the combined severe adverse conditions of uNK deletion and 48h hypoxia used in the present experiment to compromise angiogenesis and vascular remodeling in the pregnant uterus, at least 60% of implanted embryo were preserved or rescued to reach the successful gestation.This robust uterine support arise from the capability of uterine tissue compensatory response, namely the molecules essential for vascular remodeling to provide the blood supply for growing placenta.It also means the essential regulatory pathway necessary for successful pregnancy is not restricted or dependent of one single molecule, nor these are dependent of single sources in the pregnant uterus.

Table 4 .
DBA positive-and perforin (PRF) positive-uNK cells percentage in the MLAP and DB of pregnant mice uterus at gd9,5 in normoxia (N) and hypoxia (H).

Figure 1 .
Figure 1.Influence of hypoxia and IL-15 deficiency on implantation site histopathology.

Figure 2 .
Figure 2. Influence of hypoxia and IL-15 deficiency on levels of IFNγ and VEGF in implantation sites.

Figure 3 .
Figure 3. Influence of hypoxia and IL-15 deficiency on gene expression in implantation sites.

Table 1 .
Primers sequences us ed in real-time PCR and respective amplicon length.