Trophoblast-Specific Expression of Hif-1α Results in Preeclampsia-Like Symptoms and Fetal Growth Restriction

The placenta is an essential organ that is formed during pregnancy and its proper development is critical for embryonic survival. While several animal models have been shown to exhibit some of the pathological effects present in human preeclampsia, these models often do not represent the physiological aspects that have been identified. Hypoxia-inducible factor 1 alpha (Hif-1α) is a necessary component of the cellular oxygen-sensing machinery and has been implicated as a major regulator of trophoblast differentiation. Elevated levels of Hif-1α in the human placenta have been linked to the development of pregnancy-associated disorders, such as preeclampsia and fetal growth restriction. As oxygen regulation is a critical determinant for placentogenesis, we determined the effects of constitutively active Hif-1α, specifically in trophoblasts, on mouse placental development in vivo. Our research indicates that prolonged expression of trophoblast-specific Hif-1α leads to a significant decrease in fetal birth weight. In addition, we noted significant physiological alterations in placental differentiation that included reduced branching morphogenesis, alterations in maternal and fetal blood spaces, and failure to remodel the maternal spiral arteries. These placental alterations resulted in subsequent maternal hypertension with parturitional resolution and maternal kidney glomeruloendotheliosis with accompanying proteinuria, classic hallmarks of preeclampsia. Our findings identify Hif-1α as a critical molecular mediator of placental development and indicate that prolonged expression of Hif-1α, explicitly in placental trophoblasts causes maternal pathology and establishes a mouse model that significantly recapitulates the physiological and pathophysiological characteristics of preeclampsia with fetal growth restriction.

The junctional zone (JZ) was also evaluated by measuring the relative area of Tpbpa expression, which is present in spongiotrophoblasts and glycogen trophoblasts. Identification of the JZ, by Tpbpa expression, allowed for the measurement of the areas of the decidua, JZ and labyrinth (Fig. S8). At E14.5, there was a reduction in the relative area of the JZ in CA-Hif-1α placentas, compared to GFP controls (Figs 4a,b and S8). As gestation proceeds, the relative area of the JZ decreases as the labyrinth expands in size. Control GFP placentas at birth displayed the expected reduction of Tpbpa expression in the JZ area (Fig. 4c). In contrast, analysis of CA-Hif-1α placentas at In luciferase assays using the Pgk1-luciferase reporter, both pc3-Hif-1α3XSDM and the lentiviral pLB2V5-CA-Hif-1α overexpression constructs promoted increased luciferase expression compared to the pLB2V5 control, p < 0.0001 by ANOVA with Tukey's multiple comparisons test (*). (c,d) Expression of Hif-1α mRNA was detectable by in situ hybridization in E14.5 placentas derived from blastocysts infected with lentiviral particles containing GFP (c) or CA-Hif-1α (d) expression constructs. Expression was more intense and broadly distributed in CA-Hif-1α placentas. (e) CA-Hif-1α placentas collected at E19.5 showed increased expression of Hif-1α protein and its downstream target, Pdk1 compared to GFP controls, but an absence of Hif-2α protein, as detected by Western blot. CoCl 2 -treated COS7 cells were used as a positive control for Hif-1α and Hif-2α protein expression. Pan-actin was used as a loading control (a,e), (raw data, Figs S3-6) Scale bar = 100 μm.
www.nature.com/scientificreports www.nature.com/scientificreports/ birth indicated that they did not undergo the normal decrease in the relative area of Tpbpa expression and thus maintained an area similar to that of CA-Hif-1α placentas at E14.5 (Figs 4d and S9). To further assess trophoblast differentiation in the junctional zone, we examined the expression of Ascl2 (Mash2) to identify JZ progenitors 51 .

Figure 2.
Expression of CA-Hif-1α in trophoblast cells results in reduced fetal weight and altered placental development. (a) Placental weights at E14.5 or birth were not significantly altered by CA-Hif-1α expression, compared to GFP controls. (b) At E14.5, fetal weights from blastocysts infected with CA-Hif-1α were not significantly different, compared to GFP-infected controls (E14.5 GFP pups: n = 7 pups from 3 pregnancies; E14.5 CA-Hif-1α pups: n = 8 from 4 pregnancies). At birth, however, CA-Hif-1α pup weights (n = 21 pups from 6 pregnancies, avg. 1.20 g) were significantly reduced by 12.10%, compared to GFP control pup weights (n = 23 pups from 8 pregnancies, avg. 1.37 g), (*p = 0.03; by mixed model analysis). (c) Assessment of the different layers of the placenta showed that at E14.5 and at birth, the labyrinth layers of CA-Hif-1α and GFP placentas were comparable to each other and that both underwent a statistically significant increase in size between E14.5 and birth (*p = 0.0002; two way ANOVA with Sidak's Multiple Comparison test). (d,e) Alkaline phosphatase staining in GFP and CA-Hif-1α placentas at E14.5, respectively, is expressed in sinusoidal TGCs and identifies maternal blood spaces in the labyrinth layer (arrowheads). Pseudo-colouring of fetal (green) and maternal (red) blood spaces in GFP (f) and CA-Hif-1α (g) images, which when quantified, identified changes in the makeup of maternal and fetal blood spaces in the labyrinth layer, suggesting reduced branching and limited development of the fetal capillary network. This difference was evident at E14.5 but not at birth (h,i); *p = 0.0255 maternal, **p = 0.0246 fetal by two way ANOVA with Sidak's Multiple Comparison test. Scale bar = 50 μm. Expression of markers of syncytiotrophoblast demonstrated disorganized labyrinth development in CA-Hif-1α placentas. Gcm1, a marker of syncytiotrophoblast layer 2, was detectable in both GFP (a,c) and CA-Hif-1α (b,d) E14.5 placentas. In GFP controls, Gcm1 was associated with fetal blood spaces (c, arrowheads) and evenly distributed throughout the labyrinth layer. In CA-Hif-1α placentas, however, Gcm1 expression was observed in clusters of cells with limited association with blood spaces (d, arrowheads). Syncytin a (Syna), a marker of syncytiotrophoblast layer 1, was detectable at E14.5 in both GFP (e,g) and CA-Hif-1α (f,h) placentas; however, in CA-Hif-1α placentas, Syna expression was less evenly distributed throughout the labyrinth and was also observed in clusters of cells that were not associated with blood spaces (h, arrowheads), compared to the GFP controls (g, arrowheads). Boxed regions in a,b and e,f are shown at high magnification in c,d and g,h, respectively. Scale bar = 500 μm in (a,b,e,f) and = 100 μm in (c,d,g,h). . Junctional zone morphology is altered in CA-Hif-1α placentas at E14.5 and birth. Tpbpa, a marker of spongiotrophoblast and glycogen trophoblast was detectable in both GFP (a,c) and CA-Hif-1α (b,d) placentas and identified the junctional zone at both E14.5 (a,b) and birth (c,d). At E14.5, the junctional zone in CA-Hif-1α placentas appeared smaller and more compact than in GFP controls (brackets). While the GFP controls undergo a typical decrease in junctional zone size at birth (c), the CA-Hif-1α placentas did not exhibit the same rate of JZ reduction (d). Aldh1a3 identified glycogen trophoblasts (arrowheads) in the junctional zone of GFP (e,g) and CA-Hif-1α (f,h) placentas at E14.5 (e,f) and birth (g,h). At high magnification, it was evident that at E14.5 in CA-Hif-1α placentas (f), glycogen trophoblasts appeared smaller in size compared to GFP controls (e), with an increased density of nuclei. By birth, glycogen trophoblasts were mostly absent from the junctional zone of GFP control placentas (g); however, they were still present in CA-Hif-1α placentas (h). Scale bar = 500 μm in (a,b,c,d) and = 100 μm in (e,f,g,h). www.nature.com/scientificreports www.nature.com/scientificreports/ No significant differences were noted in JZ Ascl2 progenitors at E14.5 (Fig. S11a,b). In addition, in situ hybridization of Aldh1a3 and PAS staining were used to investigate glycogen trophoblasts (Figs 4e-h and S11c-f), while Prl3b1 (PL2) was used to demarcate spongiotrophoblasts ( Fig. S12a-d) [51][52][53] . Analysis of glycogen trophoblast cells at E14.5 indicated that, in individual pockets of glycogen cells, the density of nuclei were increased in CA-Hif-1α placentas, compared to GFP controls (Figs 4e,f, S11c-f and S12a-d), which suggested a decrease in cell size. This reduction in glycogen cell size may account for the smaller size of the JZ at E14.5 in CA-Hif-1α placentas. While the populations of glycogen trophoblast cells themselves were not noticeably different at birth, the spongiotrophoblasts were more compact in the GFP controls, indicative of normal placental development; however, in CA-Hif-1α placentas, spongiotrophoblasts still maintained a larger, spongy morphology. Taken together, these characteristics likely contributed to the overall junctional zone phenotype and the difference in JZ size between GFP control and CA-Hif-1α placentas at birth (Figs 4c,d,g,h and S12).
In a normal pregnant mouse, invasive trophoblast giant cells reach the maternal spiral arteries around E10.5 and become exposed to increased levels of oxygen (~12%), leading to the rapid and continuous degradation of Hif-1α protein and resultant loss of Hif-1α activity throughout the rest of gestation. Hif-1α inactivation at this developmental window is spatially and temporally coordinated and is required for trophoblast differentiation, giant cell invasion, and remodeling of the maternal spiral arteries; critical steps needed to facilitate increased blood flow and oxygen to the developing placenta and fetus [45][46][47]54 . Poorly regulated and reduced levels of oxygen at this invasive developmental time point; however, may prolong the expression of trophoblast Hif-1α and thereby alter critical steps involved in placental development; potentially leading to pregnancy-associated disorders, such as preeclampsia and/or fetal growth restriction.
To determine whether prolonged expression of Hif-1α affected trophoblast giant cell differentiation, a placental assessment was conducted to evaluate trophoblast cells that invade the maternal decidua and the extent of spiral artery remodeling. Healthy spiral artery remodeling includes the breakdown of maternal smooth muscle that lines the artery, as well as a reduction in maternal endothelial cells, which are replaced by invasive spiral artery-trophoblast giant cells (SA-TGCs) 47 . When we prolonged the expression of CA-Hif-1α in trophoblast cells beyond its' normal developmental window, we observed a continued presence of CD31 (Pecam1)-positive maternal endothelial cells (Fig. 5a-d), as well as a dramatic reduction in Prl2c2 (Plf)-positive invasive SA-TGCs (Figs 5e-h and S13a-d) associated with the maternal spiral arteries, both at E14.5 and at birth, compared to GFP controls. Additionally, fewer Prl2c2-positive trophoblast cells were associated with the maternal blood canals that deliver blood to the labyrinth in CA-Hif-1α placentas at E14.5 and at birth (Fig. S13a-d). These findings indicated that prolonged expression of trophoblast-specific CA-Hif-1α resulted in the lack of maternal spiral artery remodeling and also suggested that differentiation and/or invasion of Prl2c2-expressing TGCs was substantially decreased. Our results suggest that initial alterations in placental development are due to the lack of spiral artery remodeling in response to an inhibition of trophoblast differentiation in the presence of prolonged CA-Hif-1α. This is supported by our previous studies that show that Hif-1α inhibits trophoblast differentiation in rat trophoblast cells 19 . The lack of spiral artery remodeling would likely generate a subsequent hypoxic environment in the junctional zone and labyrinth and the end result would be the development of preeclampsia-like symptoms.
Based on the placental disorganization and reduced maternal spiral artery remodeling, we hypothesized that maternal blood pressure would be elevated, consistent with the pathophysiology that occurs in human preeclampsia. To examine the effects of trophoblast-specific CA-Hif-1α expression on maternal hypertension, blood pressure was analyzed via VPR tail cuff (Coda, Kent Scientific) in GFP control and CA-Hif-1α mothers throughout and following pregnancy. Although telemetry may be a preferred form of blood pressure measurement, analysis by non-invasive VPR tail cuff is also an accepted alternative and has been shown to produce comparable and similar blood pressure measurements in non-simultaneous recordings 55,56 , as this study was. Our analysis of blood pressure data indicate that CA-Hif-1α pregnant mice developed significantly elevated maternal hypertension, when compared to GFP controls, beginning at E17.5, that remained elevated until parturition (Fig. 6a). Maternal blood pressure in CA-Hif-1α pregnant mice decreased following delivery and returned to normal levels one day after birth (Fig. 6a), consistent with and characteristic of the preeclamptic phenotype.
Along with maternal high blood pressure, maternal glomeruloendotheliosis is a classic hallmark of preeclampsia in humans 57 . To determine if trophoblast-specific expression of CA-Hif-1α mice resulted in altered maternal renal morphology and function, maternal kidneys were evaluated. Maternal kidneys from GFP controls at E19.5 displayed normal architecture (Fig. 6b,d). Maternal kidneys from CA-Hif-1α mice at E19.5, however, had discernable collagen deposition in the glomeruli, as well as occlusion of capillary lumens, as indicated by Masson-Trichrome staining (Fig. 6b,c). PAS staining further confirmed these results and also indicated that CA-Hif-1α mothers have thinner glomerular basement membranes (Fig. 6d,e), indicative of glomeruloendotheliosis. To determine if the observed glomerular changes in CA-Hif-1α mice are accompanied by altered kidney function and proteinuria, glomerular barrier function was assessed by measuring urinary proteins. Consistent with the observed glomerular pathological damage, total urinary proteins were significantly increased in CA-Hif-1α mice (Fig. 6f). Additionally, the urinary albumin/creatinine ratio (ACR), indicative of proteinuria, was also are significantly elevated CA-Hif-1α mice compared to GFP control mice (Fig. 6g). Collectively, our data demonstrate that glomerular damage and dysfunction occurs in CA-Hif-1α mice and indicates that the prolonged presence CA-Hif-1α, specifically in placental trophoblasts, promotes kidney damage and results in proteinuria, consistent with a preeclamptic phenotype 57,58 .
While several animal models have been shown to exhibit some of the pathological effects present in human preeclampsia, these models often do not represent the physiological aspects that have been identified 57,58 . Reduced uterine perfusion pressure (RUPP), administration of autoimmune antibodies or infusion of pro-inflammatory cytokines to induce a pathological preeclamptic phenotype, are not pregnancy specific and result in systemic inflammation 59,60 . COMT and Elabela gene knockout mouse models have also been shown to display several characteristics of the preeclamptic phenotype; however, both the fetus and placenta have the cognate genes (2019) 9:2742 | https://doi.org/10.1038/s41598-019-39426-5 www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 5. Maternal spiral arteries are not remodelled in CA-Hif-1α placentas. CD31/Pecam, a marker of endothelial cells, was nearly undetectable in cells lining the maternal spiral arteries (arrowhead) in GFP E14.5 placentas, as expected (a); however, CD31/Pecam was clearly present (arrowhead) in CA-Hif-1α placentas at E14.5 (b). At birth, endothelial cells were undetectable in these vessels in GFP placentas (arrowhead) indicating normal trophoblast remodelling had occured (c). However, endothelial cells, which should have been remodelled and replaced by invading trophoblasts, were still CD31/Pecam positive and present at birth in the CA-Hif-1α placentas (arrowhead) (d). Prl2c2 is expressed in spiral-artery associated trophoblast giant cells that invade and identify maternal vessels that are undergoing or have undergone remodelling by invading trophoblast cells. At E14.5 (e) and birth (g), Prl2c2 expression was evident in GFP controls in cells lining maternal spiral arteries that have undergone normal trophoblast remodeling. In CA-Hif-1α placentas, while some Prl2c2 positive cells were present, they did not line maternal vessels in the decidua at either E14.5 (f) and birth (h). The staining pattern of CD31/Pecam, combined that of Prl2c2, in CA-Hif-1α placentas indicated a lack of appropriate trophoblast remodelling of the maternal spiral arteries. Scale bar = 100 μm.
www.nature.com/scientificreports www.nature.com/scientificreports/ ablated and these models can have fetal defects that are not seen in preeclampsia 61,62 . In addition, maternal administration of adenoviral sFLT1 or adenoviral Hif-1α have been shown to induce preeclampsia like symptoms as well. However, these models are not pregnancy specific, do not induce defects in spiral artery remodeling, and can cause hypertension in the absence of pregnancy 63,64 . Furthermore, maternal administration of adenoviral (a) Blood pressure in dams was read sequentially throughout gestation at the time points indicated. Dams pregnant with CA-Hif-1α placentas (n = 5-8) displayed elevated blood pressure beginning at E14.5, compared to dams carrying GFP placentas (n = 5-8). At E17.5-E19.5, the difference in blood pressure elevation was statistically significant in CA-Hif-1α mice by Sidak's multiple comparison test: E17.5, *p = 0.0270; E18.5, *p = 0.0005; E19.5, *p = 0.0006. Following birth, blood pressure of dams pregnant with CA-Hif-1α placentas resolved to levels comparable with dams pregnant with GFP placentas. Kidneys were dissected from pregnant dams carrying GFP or CA-Hif-1α placentas at E19.5 and sections were stained with Masson's Trichrome (b,c) or periodic acid Schiff (PAS) (d,e). Kidneys from dams with GFP placentas appeared histologically normal at E19.5 (b,d). Maternal kidneys from CA-Hif-1α mice at E19.5, however, showed glomerular damage and had discernable collagen deposition in the glomeruli, as well as occlusion of capillary lumens, as indicated by Masson-Trichrome staining (b,c). PAS staining further confirmed these results and also indicated that CA-Hif-1α mothers have thinner glomerular basement membranes (d,e), indicative of glomeruloendotheliosis. Scale bar = 100 μm. To assess kidney damage and function, spot urine samples from E18.5 and E19.5 control GFP (n = 3) and CA-Hif-1α (n = 4) dams were collected and analyzed. (f) To determine total urinary protein, the bicinchoninic acid (BCA) protein assay was used (Pierce). (g) To determine kidney dysfunction and the detect the presence of proteinuria, albumin and creatinine were measured by ELISA (EXOCELL). Maternal kidneys from GFP controls at E19.5 displayed normal architecture (b,d). In CA-Hif-1α mice; however, total urinary proteins were significantly increased in CA-Hif-1α mice (f). Additionally, the urinary albumin/creatinine ratio (ACR), indicative of proteinuria, was also are significantly elevated in CA-Hif-1α mice compared to GFP control mice (g). Collectively, our data demonstrate that glomerular damage and dysfunction occurs in CA-Hif-1α mice. ( www.nature.com/scientificreports www.nature.com/scientificreports/ Hif-1α suggests that Hif-1 overexpression is sufficient for the development of HELLP-like syndrome, a rare form of preeclampsia; however, this was independent of the presence of the placenta 64 . In addition, maternal adenoviral Hif-1α infects several organ systems and causes substantial liver damage, which is not due to Hif-1α expression, but rather due to nonspecific viral hepatitis 64 . Interestingly, it was suggested that the liver damage induced by maternal adenoviral Hif-1α administration is the likely source of elevated sFLT1 in this and potentially other maternal adenoviral models that produce pathological symptoms of preeclampsia, and not the placenta 64 . In CA-Hif-1α mice, endothelial dysfunction is present, as the maternal spiral arteries are not remodeled (Fig. 5); however, we did not observe an elevation in sFLT1 in the placenta by in situ hybridization or in maternal serum by ELISA, compared to GFP control mice (data not shown). This is consistent with studies that have reported that levels of sFLT1 can be variable or may originate from non-placental sources, and may be due to the trophoblast specificity of our model 65,66 .
In summary, our studies have prolonged the expression of CA-Hif-1α beyond its normal developmental window, as is observed in numerous studies of human preeclampsia, exclusively in trophoblast cells of the placenta. This prolonged expression of Hif-1α, specifically in trophoblasts, resulted in pregnancy-specific placental disorganization, altered placental lineage specification, and inhibition of trophoblast differentiation with endothelial dysfunction and failure to remodel the maternal spiral arteries. These physiological changes in CA-Hif-1α mice, that have also been shown to occur in human preeclampsia, lead to classic preeclamptic pathophysiology of maternal glomeruloendotheliosis, proteinuria, maternal hypertension with post birth resolution and fetal growth restriction 57 . Our findings identify trophoblast Hif-1α as a critical molecular mediator of placental development and indicate that prolonged expression of Hif-1α, explicitly in trophoblast cells of the placenta, induces maternal and fetal physiological and pathophysiological characteristics that significantly recapitulate that of preeclampsia with FGR and thereby establishes a mouse model that closely resembles the human condition.

Methods
Detailed materials and methods are described in Supporting Information.
All trophoblast-specific gene transfer experiments involving mice in this study were conducted under a Wright State University IACUC-approved protocol.
Blood Pressure Analysis. Blood pressures were measured and recorded via tail cuff plethysmography using the CODA Non-Invasive Blood Pressure System (Kent Scientific) 55,74,75 . Induction of Superovulation, Pseudopregnancy and Embryo Culture. C57BL/6 females were superovulated and mated as previously described 38 . Copulation plugs found the following morning were designated as E0.5. Superovulated C57BL/6 female embryos were collected at E1.5 (2-cell) in M2 media and cultured in pre-equilibrated 20 µl KSOM AA microdrops at 37.5 °C and 5.5% CO2 until E3.5 blastocysts 38 . ICR females were mated with vasectomized ICR males to generate pseudopregnancy, checked for copulation plug the next morning and designated as 0.5 d.p.c. 38 .
Placental Histological Staining and Analysis. Alkaline phosphatase (AP) staining to identify maternal blood spaces and isolectin staining to identify fetal blood spaces were conducted as previously described 44 . Periodic Acid Schiff (PAS) staining to identify glycogen accumulation was conducted according to the manufacturer's protocol (Sigma-Aldrich). Semi-quantitative assessment of placental layers and in situ hybridization staining in placenta sections was conducted using NIH ImageJ software. Different layers of the placenta and regions of staining for individual gene markers were outlined and measured manually for layers (in pixels) and using threshold adjustment and selection of threshold area by the software, using Image J, and assessed by two independent observers in a blinded fashion. Area was calculated as a proportion of the total measured area of each of the placentas and reported as % area. At least three independent pictures representing three independent placentas, were measured for each gene or layer quantified. Kidney Analysis. Kidneys from dams carrying GFP or CA-Hif-1α infected embryos were isolated at E19.5 and fixed in 4% PFA for embedding in paraffin. To assess kidney pathology, tissue processing, embedding, sectioning and histological staining with hematoxylin and eosin (H&E), periodic acid Schiff (PAS) without diastase and Masson's trichrome staining were conducted by Reveal Biosciences (San Diego, CA). To assess kidney damage and function, spot urine samples were collected and analyzed from E18.5 and E19.5. Total urinary protein was measured using a bicinchoninic acid (BCA) protein assay (Pierce). Protein concentrations were calculated by extrapolation of values from a protein standard curve. The p-value for total urinary protein was p = 0.0175**. To confirm kidney dysfunction and detect the presence of proteinuria, urinary albumin and creatinine were measured by ELISA (EXOCELL). Microalbuminuria was determined by calculation of the albumin/creatinine ratios (ACR). The p-value for ACR was p = 0.0326**.