Placental Vasculogenesis Is Regulated by Keratin-Mediated Hyperoxia in Murine Decidual Tissues

Abstract

The mammalian placenta represents the interface between maternal and embryonic tissues and provides nutrients and gas exchange during embryo growth. Recently, keratin intermediate filament proteins were found to regulate embryo growth upstream of the mammalian target of rapamycin pathway through glucose transporter relocalization and to contribute to yolk sac vasculogenesis through altered bone morphogenetic protein 4 signaling. Whether keratins have vital functions in extraembryonic tissues is not well understood. Here, we report that keratins are essential for placental function. In the absence of keratins, we find hyperoxia in the decidual tissue directly adjacent to the placenta, because of an increased maternal vasculature. Hyperoxia causes impaired vasculogenesis through defective hypoxia-inducible factor 1α and vascular endothelial growth factor signaling, resulting in invagination defects of fetal blood vessels into the chorion. In turn, the reduced labyrinth, together with impaired gas exchange between maternal and embryonic blood, led to increased hypoxia in keratin-deficient embryos. We provide evidence that keratin-positive trophoblast secretion of prolactin-like protein a (Prlpa) and placental growth factor (PlGF) during decidualization are altered in the absence of keratins, leading to increased infiltration of uterine natural killer cells into placental vicinity and increased vascularization of the maternal decidua. Our findings suggest that keratin mutations might mediate conditions leading to early pregnancy loss due to hyperoxia in the decidua.

Epithelial cells line the surface of internal organs and tissues. They provide mechanical support and protection from the external environment but are contemporaneously essential for the communication and the exchange of nutrients and oxygen from the environment, as in the gut and lung tissue, respectively.

The intermediate filament system of the epithelial cytoskeleton, formed by members of the keratin multiprotein family, is particularly suited to fulfill these functions. Keratins have been confirmed to provide mechanical stability, as diverse skin mutations account for.^1–4 In addition, keratins have been shown to exert important signaling functions in an isotype- and context-dependent manner in epithelial cells. Previous mutation and knockout (KO) studies showed their involvement in the regulation of cell cycle, in protein translation through 14-3-3 proteins and the mammalian target of rapamycin complex, modulation of apoptotic signals, organelle transport and protection against metabolic stress.^5–9 The functional analysis of individual keratins has been obscured by compensatory expression of other keratins expressed in the same epithelia. The recently reported deletion of all keratins in mice provided unexpected and novel insights into keratin function during mouse development.¹⁰ It showed that keratins play an essential role in embryogenesis, because mutant embryos died at embryonic day (E) 9.5 in midgestation because of severe growth retardation, caused by a drastic decrease in protein biosynthesis. This was due to a mislocalization of the glucose transporters GLUT1 and GLUT3 from the apical plasma membrane, which resulted in the inhibition of the mammalian target of rapamycin pathway through AMP kinase.¹⁰ In contrast to previous gene KO studies, in which the deletion of K8 or the combined deletion of K18/K19 and of K8/K19 caused fragility of trophoblast giant cells, followed by extensive hemorrhages,^11,12 we found that tissue integrity and apical cell polarity of the embryonic epithelia were largely maintained in keratin null embryos. The depletion of keratins rather caused an attachment defect of endodermal and mesodermal tissues in the yolk sac, resulting in decreased hematopoiesis and vasculogenesis through reduced Forkhead box protein F1 signaling and its downstream targets bone morphogenetic protein and P-p38 mitogen-activated protein kinase in the yolk sac mesoderm.⁹

The function of keratins in the placenta has not been fully elucidated. In a subset of K8-deficient embryos, increased sensitivity to tumor necrosis factor–mediated apoptosis was observed in a mouse strain-dependent manner.^13,14 In contrast, other studies reported trophoblast fragility after the combined deletion of K18/19 or K8/K19.^11,12 Survival and growth of the embryo critically depend on the placenta. The placenta forms the interface between maternal and fetal circulation, facilitating gas, nutrient, and waste exchange.¹⁵ It comprises the epithelial trophectoderm cells that on implantation of the embryo into the uterine wall at around E4.5 expand and differentiate. The first embryonic cells to interact with the maternal tissue are the trophoblast giant cells (TGCs) that invade and attach to the uterine wall and induce decidualization by altering specific gene expression among others for vascular remodeling and angiogenesis secreting various hormones such as prolactin-like protein a (Prlpa).¹⁶

The vascularization of the murine placenta starts at E8.5 with the fusion of the mesodermal allantois to the chorion and the invagination of the fetal blood vessels. This process relies on a highly coordinated cross talk between the epithelial trophoblast and the embryonic endothelium, modulated by transcription factors and signaling receptors on the epithelial^17–29 as well as on the endothelial side.^30–39 These factors are essential for the development of the proper vasculature in the placental labyrinth (Tables 1 and 2).

Table 1.

Mouse Mutants of Epithelial Trophoblast Cells with Similar Phenotypes

Gene	Gene product	Expression pattern in placental tissue and yolk sac	Mutant phenotype	Reference
ARNT (Hif1β)	bHLH/PAS transcription factor	Placental trophoblasts, trophoblast giant cells	Growth retardation at E9.5, defective placental and yolk sac vascularization	17, 24
Cx26	Connexin	Yolk sac epithelium, syncytiotrophoblast	Small labyrinth	21
FGFR2	Fibroblast growth factor receptor	Trophoblast cells	Small labyrinth layer, reduced vascularization, block of trophoblast cell proliferation	29
Fzd5	Wnt receptor	Labyrinth trophoblast, yolk sac	Reduced vascularization of yolk sac and placenta	22
Hif1α and Hif2α	Hypoxia-inducible factors	Placenta, yolk sac	Reduced placental vascularization, differentiation regulation of TGC and spongiotrophoblast	20
HSP90	Heat shock protein	Labyrinth trophoblast, yolk sac	Small labyrinth layer, reduced vascularization	23, 28
JunB	Transcription factor of the AP-1 family	Trophoblast derivatives	Small labyrinth layer, defective placental and yolk sac vascularization	26
p38	Mitogen-activated protein kinase	Diploid trophoblasts, Labyrinth trophoblast	Reduced vascularization of yolk sac and placenta	25
PPARγ	Nuclear hormone receptor, ligand activated transcription factor	Diploid trophoblast lineage	Reduced vascularization of placenta, impaired differentiation of syncytiotrophoblast	18, 19
Tfeb	Transcription factor of the bHLH zipper family	Labyrinth trophoblast	Reduced placental vascularization with embryonic blood vessels	27

Table 2.

Mouse Mutants of Endothelial Cells with Similar Phenotypes

Gene	Gene product	Expression pattern in placental tissue and yolk sac	Mutant phenotype	Reference
Ang1	Angiopoietin 1, Tie2 ligand	Arterial endothelial cells of yolk sac and placenta	Defects in angiogenic vascular remodeling in yolk sac and placenta	30
Ang2	Angiopoietin 2, Ang1 antagonist	Arterial endothelial cells of yolk sac and placenta	Overexpression leads to defects in angiogenic vascular remodeling in yolk sac and placenta	31
Dll4	Delta-like ligand	Arterial endothelial cells of yolk sac and placenta	Arterial atrophy	33, 35
Flk1	Vascular endothelia growth factor receptor 2	Yolk sac mesoderm, labyrinth	Lack of endothelia cells and hematopoietic system	32, 37
Flt1	Vascular endothelia growth factor receptor 1	Yolk sac mesoderm, spongiotrophoblast	Disorganized embryonic vasculature	32, 34
Notch1	Single-pass transmembrane receptor	Arterial endothelial cells of yolk sac and placenta	Defects in angiogenic vascular remodeling in yolk sac and placenta	36
Tie2/Tek	Receptor tyrosine kinase	Arterial endothelial cells of yolk sac and placenta	Reduced number of endothelia cell, disorganized embryonic vasculature	38, 39

Furthermore, a localized oxygen gradient has been shown to be critical in human decidualization and placentation by influencing vascular remodeling in the decidua and differentiation of trophoblast cells.⁴⁰ Increasing evidence suggests that the same holds true for mice.⁴¹ Both implantation and ensuing placentation start under hypoxic conditions until the vascular network between maternal and embryonic tissues is properly established and the placenta is flooded with maternal blood. Hypoxia is necessary for normal vascularization.⁴² The hypoxia-inducible factors (HIFs), HIF1α, HIF2α, and HIF1β (ARNT) directly promote vascular endothelial growth factor (VEGF) expression,⁴³ because their absence in various KO models showed a reduced vascularization and developmental retardation of the placenta.^17,20,24 In humans, increased hypoxia leads to preeclampsia, a syndrome causing substantial maternal and fetal mortality in pregnancies. Hallmarks of preeclampsia are reduced trophoblast invasion, reduced vasculature of the placenta, and increased serum levels of soluble fms-related tyrosine kinase 1 (sFlt1).^44,45 Conversely, untimely elevated oxygen levels in decidual and placental tissues cause regression of placental villi, spongiotrophoblast tissue, and intrauterine growth restriction through premature maternal blood flow to the placenta.^46,47 The underlying gene defects, however, are not well known, and molecular defects causing these alterations still have to be identified.

Here, we demonstrate a crucial role of keratins for vascular remodeling in the decidua and in the labyrinth through an altered localization of TGCs. Subsequently, TGC mislocalization altered proper hormone secretion and induced hypervascularization and increased oxygen levels in the decidua. Further, we find that keratins control GLUT1 localization also in extraembryonic tissues. Our findings show that keratins control embryo development in an isotype- and context-dependent manner and suggest that keratin mutations may contribute to early pregnancy losses in conditions of maternal hyperoxia.

Materials and Methods

Animals

All experimental animal procedures were conducted according to approved experimental animal licenses issued by the responsible animal welfare authority (Landesamt für Natur, Umwelt und Verbraucherschutz NRW). KtyII^−/− mice were bred from heterozygous animals, generated and genotyped as described previously,¹⁰ and maintained in a C57BL/6 background.

Transgenic Mice, Immunofluorescence Microscopy, and in Situ Hybridization

Mouse concepti at different gestational ages were prepared. For light microscopy and immunofluorescence analysis, tissues were snap-frozen in isopentane precooled at −80°C and stored at the same temperature. Immunofluorescence analysis was performed as follows: Frozen sections (8 to 12 μm thick) were fixed in acetone at −20°C for 10 minutes and dried for a few hours before further processing. All antibodies were diluted in Tris-buffered saline containing 1% bovine serum albumin and 5% normal goat serum, Alexa 488/594–conjugated secondary antibodies (Dianova, Hamburg, Germany) were used 1:800 diluted. Slides were mounted in ProLong Gold antifade reagent (Invitrogen, Carlsbad, CA).⁴⁸ GLUT1 (Millipore Corporation, Billerica, MA), K8/K18 (Progen Biotecknik GmbH, Heidelberg, Germany), and K19 (Magin Lab) staining were performed as described before.¹⁰

For routine histology, tissues were fixed overnight at 4°C in 4% formalin, sequentially incubated at 4°C overnight in 15% and 30% sucrose, embedded in paraffin, and sectioned (5 μm). Sections were placed on superfrost-plus slides (Menzel-Gläser, Braunschweig, Germany) and dried. After deparaffination, sections were either stained with H&E,⁴⁸ periodic acid-Schiff,⁴⁹ Dolichos bilforus agglutinin lectin (Vector Laboratories, Burlingame, CA), diluted 1:100 for 1 hour in Tris-buffered saline, 1% bovine serum albumin, 5% normal goat serum or used for immunohistochemistry. For CD31 (SZ31; Dianova), CD34 (MEC 14.7; Hycult Biotech, Uden, The Netherlands), Hif1α (Novus Biologicals Inc., Littleton, CO), VEGF (VG-1), and placental growth factor (PlGF; Abcam, Cambridge, MA) staining on paraffin sections, antigens were retrieved with citrate buffer (18 mmol/L citric acid, 82 mmol/L sodium citrate, pH 6.0), and the antibodies were diluted in Tris-buffered saline containing 1% bovine serum albumin and 5% normal goat serum 1:20, 1:50, 1:100, 1:50, and 1:20, respectively. For cleaved caspase 3 (Cell Signaling Technologies, Danvers, MA) staining of paraffin sections, antigens were retrieved with EDTA buffer (1 mmol/L EDTA, pH 8.0), peroxidases blocked 10 minutes in 3% hydrogen peroxide, and the antibodies were diluted in Tris-buffered saline containing 1% bovine serum albumin and 5% normal goat serum 1:200. The histochemistry was performed with Super Sensitive Link Label IHC Detection system (BioGenex, San Ramon, CA) and visualized with diaminobenzidine (DAB; Dako, Giostrup, Denmark).

For mRNA in situ hybridization experiments, concepti were processed as above and sectioned. Sections (10 μm) were hybridized with digoxigenin-labeled riboprobes for the various placental lineage markers,^50,51 K8, placental lactogen 1 (Pl1),⁵² trophoblast specific protein alpha (Tpbpa),⁵³ Prlpa,⁵⁴ glial cells missing (Gcm1),⁵⁵ and SyncytinA.⁵⁰ The probes were detected with alkaline phosphatase-conjugated anti-digoxigenin antibody and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT) as chromogenic substrate.⁵⁰

Immunofluorescence Microscopy and Data Processing

Images of stained paraffin sections were acquired with the use of a fluorescence microscope (Axioplan 2; Carl Zeiss Inc., Thornwood, NY) with a Plan-Neofluar 10 × 0.30 NA objective or a Plan-Apochromat 63 × 1.4 NA oil immersion objective at room temperature using an AxioCam MR camera (Carl Zeiss Inc.). Image analysis and processing were performed with AxioVision 4.6 (Carl Zeiss Inc.) and Photoshop 6.0 (Adobe Systems, Mountain View, CA) software. Lookup table (brightness and gamma) was adjusted with Photoshop.

Western Blotting

Western blotting was performed as described.¹⁰ In short, total proteins were extracted in SDS–polyacrylamide gel electrophoresis sample buffer under repeated heating (95°C) and sonication cycles. Total protein concentration was determined by the Bio-Rad protein quantification kit, and equal amounts of protein were loaded. Separation of total protein extracts was carried out by standard procedures (8% and 10% SDS–polyacrylamide gel electrophoresis). The immunostaining was performed as described earlier.¹⁰ Primary antibodies were diluted as follows: Hif1α at 1:500 (Novus Biologicals), VEGF at 1:500 (VG-1; Abcam), and tubulin at 1:5000 (B 5-1-2, Sigma-Aldrich, St. Louis, MO). Secondary antibodies (Dianova) were diluted at 1:20,000 and detected with Super Signal (Thermo Fisher Scientific Inc., Rockford, IL).⁴⁸

Hypoxia Assay and sFlt1 ELISA

Mice were injected with 60 mg/kg pimonidazole intraperitoneally (Hypoxyprobe-1; Hypoxyprobe, Inc., Burlington, MA) 2.5 hours before they were sacrificed. E9.5 mouse embryos were dissected and prepared for histology as described above. Immunohistochemical analysis with Hypoxyprobe-1 was performed as described by the manufacturer (Hypoxyprobe, Inc.) with the use of the hybridoma 4.3.11.3 (1:25). For the sFLT1 enzyme-linked immunosorbent assay (ELISA), nonpregnant and pregnant mice (E9.5) were sacrificed by cervical dislocation, and blood was extracted from the heart. Serum was collected, and the ELISA for sFLT1 was performed according to manufacturer's protocol (R&D Systems, Minneapolis, MN).

Results

Keratins Are Necessary for Placental Labyrinth and Decidual Development but Do Not Determine Cell Fate

To investigate to which extent formation of placental epithelia depends on keratins, we first analyzed placenta morphology in histologic sections of various developmental stages of keratin-deficient (KtyII^−/−) and wild-type (WT) embryos with their surrounding placental and decidual tissues (Figure 1 A, model). Until E8.5, the placenta of keratin null embryos developed normally and underwent a typical chorioallantoic attachment (Figure 1, H–M). By E9.5, however, a drastic difference in size became apparent (Figure 1, B and C). Mutant placentas were smaller and enriched in cells (Figure 1, C, F, and G) compared with WT controls (Figure 1, B, D, and E) and showed significantly impaired vasculature. H&E staining of these placentas showed that fetal blood vessels were present, but that they were fewer in number and appeared dilated (Figure 1, F and G) compared with WT placentas (Figure 1, D and E). In addition, embryonic vessels of the KO placenta showed a defect in embryonic hematopoiesis, containing less embryonic erythrocytes in comparison to WT placentas. KtyII^−/− embryos also had defects in the vascularization of the decidua, characterized by a discontinuous, decomposed loose structure of the peri-vitelline network (Figure 1, C and I; Figure 2D; Figure 3B; and Figure 4, C and D), the aggregation of maternal blood lacunae in between the trophoblast layers of the TGCs and the spongiotrophoblast with infiltrating lymphocytes (Figure 1, I and M; Figure 3, B and H; Figure 4D; and Figure 5 G) which became apparent already at E8.5. In contrast to WT embryo TGCs showed an irregular distribution. They were accumulated in some regions and underrepresented in others (Figure 1, C and I; Figure 4C; and Figure 5G). This was accompanied by the leakage and accumulation of maternal blood (Figure 1, I and M; Figure 3, B and H, Figure 4D; and Figure 5G) in >90% of the KO embryos.

Figure 1.

Severe reduction of labyrinth, spongiotrophoblast, and vascular defects were seen in the placenta of keratin-depleted embryos. A: Diagram of placental architecture. B–M: Semithin sections through WT (B, D, E, H, J, and K) and mutant (C, F, G, I, L, and M) embryos were stained with H&E. Note size difference of the labyrinth in keratin KO placenta (C) compared with WT placental control (B), starting at E8.5 in KO embryos (I). Increased vasculature in the decidua, surrounding the mutant embryo begins at E8.5 (arrow in C and I), resulting in maternal leakage of blood and lymphocytes into embryonic tissues (I and M). In higher magnifications, marked in the overview of E9.5 WT (D, E, magnified from B) and KO placenta (F, G, magnified from C), WT embryonic blood vessels are packed with embryonic erythrocytes compared with sparsely filled KO blood vessels. Further magnification of WT (E, magnified from D) and KO placentas (G, magnified from F) show that WT embryonic blood vessels are in close proximity to maternal blood sinuses in contrast to KtyII^−/− labyrinth, which contains few maternal blood sinuses. MV, maternal blood vessels; De, decidua; TGC, trophoblast giant cells; SpT, spongiotrophoblast; L labyrinth; ML, maternal blood lacuna; Ly, lymphocytes; Al, allantois; RM, Reichert's membrane; SyT, syncytiotrophoblast; EB, embryonic blood; MS, maternal blood sinus. Scale bars: 33 μm (B); 50 μm (C, H, and I); 10 μm (D–G and J–M).

Figure 2.

Spatial mislocalization of TGCs, but no differentiation defects in KtyII^−/− embryos. In situ hybridizations of labyrinth trophoblast markers K8, Pl1, Tpbpa, and Prlpa. A and D: Keratin 8 hybridization differentiates between WT (A) and KO (D) embryos. Note increased vasculature in the KO decidua (arrow in D). Pl1, a marker for TGCs, shows no differentiation defects of TGCs in KtyII^−/− embryos (E) compared with control TGC (B); however, Pl1 staining shows a discontinuous layer of TGCs in KO embryos, with areas of accumulation (black arrowheads) and areas of underrepresented TGCs (white arrowheads). In situ hybridizations of the spongiotrophoblast marker Tpbpa display no difference in KO (F) compared with WT (C) placentas. H–L: Higher magnifications of above-mentioned markers. G and J: Expression for Prlpa, a hormone secreted by the TGCs, is already at E8.5 notably reduced in the KO TGC (J) compared with WT TGC (G). Pl1 hybridization in KO placenta demarcates an area with reduced TGCs (K), compared with a continuous signal in WT TGC layers (H). At higher magnification no difference between WT (I) and mutant (L) placentas in Tpbpa signal was notable. Scale bars: 100 μm(A–F); 10 μm (G–L).

Figure 3.

Keratin KtyII^−/− embryos show a dysfunctional vasculature in the placental labyrinth. Immunohistochemical staining with anti-CD34 antibodies in WT (A, C, D, G, I, and J) and KO (B, E, F, H, K, and L) placentas. The WT labyrinth shows extensive staining with CD34 antibodies (A), displaying an extensive vasculature network. In contrast, embryonic vessels fail to invade the labyrinth properly. Further note accumulation of TGCs (B and E) adjacent to the extensive maternal vasculature (black arrowhead) and the presence of maternal blood lacunae in the KO placenta (B). In the higher magnifications (C–F), embryonic WT controls displayed small vessels filled with nucleated embryonic erythrocytes, in close vicinity to maternal blood sinuses (D, inlay of C) In contrast, KO vessels seemed dilated with a thickened syncytiotrophoblast layer, containing few nucleated embryonic erythrocytes (F, magnified from E). Vascular defects already became apparent at E8.5, including maternal leakage into the KO embryo (H, K and L, magnified from K) compared with the WT situation (G, I and J, magnified from I). De, decidua; TGC, trophoblast giant cells; SpT, spongiotrophoblast; L labyrinth; ML, maternal blood lacuna; Al, allantois; RM, Reichert's membrane; Ys, yolk sac; E, embryo; SyT, syncytiotrophoblast; EB, embryonic blood; MS, maternal blood sinus. Scale bars: 50 μm (A and B); 10 μm (C–F).

Figure 4.

Hyperoxia in decidual tissue surrounding keratin KO embryos leads to reduced induction of the downstream signal Hif1α. Localization of hypoxic regions in implantation sites of E9.5 WT (A) and KO (C) embryos using Hypoxyprobe-1–specific antibody. Strong hypoxic staining is present in WT decidual tissue enclosing the EPC, more explicit at higher magnification (B, magnified from A) in contrast to mutant decidua, where the hypoxic area around the EPC is reduced because of an increased vasculature of the maternal decidua, marked by black arrows (D, magnified from C). Controls in A and C are negative controls without antibody staining. Hif1α expression manifested predominantly in the placenta of WT (F, magnified from E) and to some extent in KO placentas (H, magnified from G), but less so in the decidua (E–H). P, placenta; De, decidua; TGC, trophoblast giant cells; E, embryo; Ys, yolk sac; EPC, ectoplacental cone; ML, maternal blood lacuna; Hyp, hypoxia. Scale bars: 100 μm (A–H).

Figure 5.

Hyperoxia in decidual tissue surrounding keratin KO embryos leads to reduced induction of downstream signals VEGF and PlGF. Hypoxia induces the angiogenic factors VEGF (A–D) and PlGF (E–H) through Hif1α, visualized by histochemical antibody staining. VEGF, as soluble factor, is expressed in decidual and placental tissues of WT (A and B) and KO embryos (C and D). In line with the hyperoxia in the KO, VEGF showed reduced staining in KtyII^−/− placenta and decidua (D) compared with the WT (B). Note also the absence of PlGF in the KO placenta (H), whereas decidual expression remained (G and H). WT tissues showed even distribution of PlGF through placental and decidual tissues (E and F). P, placenta; De, decidua; TGC, trophoblast giant cells; E, embryo; EB, embryonic blood; EPC, ectoplacental cone; ML, maternal blood lacuna; Al, allantois. Scale bars: 100 μm (A–H).

To examine whether the phenotype resulted from increased apoptosis, histochemical analysis for cleaved caspase 3 was performed. This clearly showed absence of trophoblast cell lysis and no increase in trophoblast cell apoptosis. In agreement with published data,¹⁰ the embryo proper displayed an increased number of apoptotic cells. Previously, trophoblast fragility was reported after the deletion of K18/19 or K8/K19.^11,12 In KtyII^−/−, serial sections indicated a slight decrease in TGC numbers but failed to detect any evidence for trophoblast cytolysis (see Supplemental Figure S1 at http://ajp.amjpathol.org). This indicated that aggregated keratins may contribute to cell lysis in the former setting. To further analyze the misdistribution of TGCs, epithelial differentiation was investigated with the use of several lineage markers for the major trophoblast cell types (Figure 2). In situ hybridization showed that the outermost trophoblast layer, the giant cells, expressed Pl1 (Figure 2, B, E, H, and K), and the spongiotrophoblast cells were positive for Tpbpa (also known as 4311; Figure 2, C, F, I, and L).^52,53 This showed that lack of keratins did not alter epithelial differentiation in the placenta per se. However, we noted an altered distribution of the TGC marker Pl1, indicating a mislocalization of TGCs (Figure 2E arrowheads). Moreover, the mRNA for the secreted Prlpa, a member of the prolactin family, was strongly reduced in KtyII^−/− TGCs compared with WT cells (Figure 2, G and J). Prlpa is produced by trophoblast cells and specifically interacts with uterine natural killer (uNK) cells.^56,57 During pregnancy, uNK cells accumulate around the mesometrial uterine vasculature^58,59 and stimulate vascular remodeling through the secretion of angiogenic factors⁶⁰ facilitating nutrient flow to the placenta and the fetus.^61,62 In line with these data, we found an increase in uNK cell numbers in the decidua of KO embryos (Figure 6E). Visualizing uNK cells with Dolichos bilforus agglutinin lectin and PAS staining⁶³ further showed a relocalization of uNK cells to the direct vicinity of the ectoplacental cone (EPC) of keratin KtyII^−/− trophoblast cells (Figure 6, C and D) in contrast to an even distribution in decidual tissues of WT embryos (Figure 6, A and B). Uterine NK cells were furthermore shown to secrete growth factors such as VEGF and the VEGF-related factor PlGF, adding to the decidual hypervascularization phenotype.

Figure 6.

Increased numbers of uNK cells concentrated in the direct vicinity of the KO placenta. uNK cells are visualized with Dolichos bilforus agglutinin lectin (A and C) and PAS (B and D) staining on semithin sections. Increased amounts of uNK cells (E) were localized closely along the embryonic trophoblast in the KO placenta (C and D), whereas the WT uNK cells are distributed evenly through the maternal decidua and sparsely around the WT trophoblast cells (A and B).

We conclude that the keratin-dependent altered spatial distribution of the TGCs and the compromised secretion of hormones¹⁰ trigger an increase in decidual uNK cell numbers and result in an extension of maternal vasculature. These changes contribute to the embryonic death of KtyII^−/− embryos.

In situ hybridization on implantation sites of E8.5 and E9.5 embryos with a Gcm1 probe, which demarcates the sites for the initiation of branching morphogenesis along the leading edge of the chorion at the interface with the allantois⁵¹ and for SyncytinA, a marker for the development of the syncytiotrophoblast,⁵⁰ did not show any alterations in expression levels in KtyII^−/− embryos (see Supplemental Figure S2 at http://ajp.amjpathol.org).

Vascular Defects in KtyII^−/− Placentas

We recently demonstrated that KtyII^−/− embryos at E9.5 could be easily discriminated from WT embryos by their pale yolk sacs. The depletion of keratins caused an attachment defect of endodermal and mesodermal tissues in the yolk sac, resulting in decreased hematopoiesis and vasculogenesis through reduced Forkhead box protein F1 signaling and its downstream targets bone morphogenetic protein and P-p38 mitogen-activated protein kinase in the mesoderm.⁹

The distribution of fetal capillaries in the labyrinth was investigated by immunohistochemistry with the use of antibodies that recognize CD34 (Figure 3) and CD31/platelet endothelial cell adhesion molecule 1 (see Supplemental Figure S3 at http://ajp.amjpathol.org), adhesion molecules expressed on the surface of vascular endothelial cells.^64,65 In the WT placenta of E9.5 embryos the fetal vessels were evenly distributed throughout the placental labyrinth, always in close contact to maternal blood sinuses to enable the fetal-maternal exchange of gas and nutrients (Figure 3, A, C, and D; also see Supplemental Figure S3 at http://ajp.amjpathol.org). However, in the KtyII^−/− placenta, fetal blood vessels mostly innervated the chorioallantoic plate and did not show extensive branching into the labyrinth trophoblast as the WT vessels did. They were fewer, appeared dilated, and contained less embryonic blood, and, in addition, very few maternal blood sinuses were identified near the embryonic vessels (Figure 3, B, E, and F; also see Supplemental Figure S3 at http://ajp.amjpathol.org).

KtyII^−/− Embryos Develop Decidual Hyperoxia with an Altered Expression of Vasculogenesis-Promoting Factors

In human embryonic development, vascularization of the placenta depends on oxygen tension in the surrounding decidual tissues.⁴⁰ Recent murine gene knockout studies of HIF proteins indicated that hypoxia is required for vasculogenesis, angiogenesis, and hematopoiesis in the mouse as well.^17,20,24,43

This prompted us to examine the state of hypoxia in KtyII^−/− embryos (Figure 4, A–D). Surprisingly, hypoxia was reduced in the KO decidua compared with the WT controls (Figure 4, A and B) because of increased maternal vasculature (Figure 4, C and D). However, Hypoxyprobe-1 detects only a severe decrease in oxygen levels of ≤10 mmHg (1.5% oxygen) and hence would not detect concentrations of oxygen consistent with milder hypoxia. Therefore, we assessed the localization of Hif1α, which is one of the earliest induced proteins in hypoxia, as soon as 8% oxygen. We found a strong placental staining in the WT placenta of E9.5 embryos and a weaker staining in the hypoxic areas of the decidua, because Hif1α is no longer responsive under continuous low oxygen levels and its accumulation levels drop⁶⁶ (Figure 4, E and F). Similar to the WT, the implantation side of the KO also showed an increase of Hif1α staining in the placental tissues (Figure 4, G and H). However, the level of Hif1α in the embryo proper seemed to be elevated in the KtyII^−/− embryo compared with the WT embryos. This was confirmed by Western blot analysis of protein lysates of the embryo proper and the yolk sac (Figure 7, A and B). Recent studies have shown that Hif1α induces VEGF expression to induce placental vasculogenesis⁴³ through a direct binding of the Hif1α dimer to the VEGF promoter region.⁶⁷ Therefore, we localized and measured VEGF expression in the WT decidua and placenta (Figure 5, A and B) and compared them with the keratin KtyII^−/− specimen (Figure 5, C and D). We found that there was a decrease of Hif1α and VEGF in KO placental tissues, which is consistent with the poor vascularization of the KO placenta (Figure 3). In the embryo proper, no differences in VEGF expression levels were observed (Figure 7, A and B).

Figure 7.

Increased hypoxia inducible factor in the embryo proper of KtyII^−/− embryos. A and B: Western blotting of total lysates of E9.5 embryo proper and yolk sac were analyzed against Hif1α and VEGF in both genotypes. Total protein levels of Hif1α and VEGF were quantified by densitometry and normalized to tubulin (n = 3; B), showing a drastic increase of Hif1α expression in KtyII^−/− embryos. VEGF levels remained unaltered. Results are shown as means and SEM; *P < 0.05 compared with WT controls (two-tailed t-test). C: sFLT1 levels in maternal blood serum of untreated versus pregnant WT and KtyII^+/− mice determined by ELISA were not increased in pregnant KtyII^+/− mice, ruling out preeclampsia as a possible cause for the prenatal death of KO embryos. Nonpregnant wild-type, n = 8; nonpregnant KtyII^+/−, n = 13; day E9.5 pregnant wild-type mice, n = 9; day E9.5 pregnant KtyII^+/−, n = 8. Results are shown as means and SEM; **P < 0.05 compared with nonpregnant WT controls (two-tailed t-test).

VEGF and PlGF are expressed simultaneously during mouse development⁶⁸ and target the same receptor Flt1.⁶⁹ We observed a similar distribution of PlGF protein compared with VEGF in WT embryos (Figure 5, E and F). Furthermore, we found that decidual hyperoxia caused no changes in decidual expression levels in KO embryos (Figure 5, G and H) but led to a drastic decrease in placental PlGF levels, corresponding to the observed decrease of placental vasculogenesis (Figure 5H).

Hypoxia was also shown to induce the expression of GLUT1 transporters in early embryonic development.⁴¹ In line, we found a twofold increase in GLUT1 expression in the embryo proper by quantitative PCR (data not shown), corresponding with our previous findings.¹⁰ Furthermore, GLUT1 transporters were mislocalized to cytosolic compartments in placental tissues (Figure 8) as previously seen in yolk sac tissues.¹⁰ Moreover, a reduction of glycogen storage in the KO trophoblast cell lineage was identified in the EPC region of the KO embryo compared with WT trophoblast cells (Figure 6, B and D). These findings suggest that keratins act through similar mechanisms in embryonic and extraembryonic tissues, affecting the vascularization and the localization of GLUT1 transporters. The subsequent lack of nutrition contributes to metabolic failure in KtyII^−/− embryos.

Figure 8.

Mislocalization of GLUT1 transporters in trophoblast cells of the labyrinth. Comparing immunofluorescence analysis of WT to KO placental staining with K8/K18 (A, D and G, H) and GLUT1 (B, E and H, K) antibodies shows a prominent location of GLUT1 transporters in WT syncytiotrophoblast cells, which regulate gas and waste exchange, as well as nutrient transport (C and F). In contrast, GLUT1 transporters are mislocalized in KO placental tissues (I and L) similar to yolk sac tissues.¹⁰ TGC, trophoblast giant cells; SpT, spongiotrophoblast; L, labyrinth; SyT, syncytiotrophoblast; EB, embryonic blood. Scale bars: 100 μm (A–C and G–I); 10 μm (D–F and J–L).

KtyII^−/− Mice Represent a Novel Model for Decidual Hyperoxia

Hypoxia-induced HIF1α protein accumulation in placentas is associated with shallow invasion of trophoblasts into the spiral arteries and the uterine wall (see Supplemental Figure S4 at http://ajp.amjpathol.org), probably resulting in vascular remodeling defects and further hypoxia.⁴⁴ In addition, secretion of sFlt1 into the maternal blood stream and proteinuria are phenotypic hallmarks of the severe pregnancy complication, preeclampsia, in humans and mice.^45,70 Therefore, we measured protein contents in urine of E9.5 pregnant mice (data not shown), as well as sFlt1 levels in nonpregnant versus pregnant WT and KtyII^+/− mice at E9.5. Both assays excluded keratins as a major contributor to preeclampsia, because there neither was an increase in protein levels in urine (data not shown) nor could a difference in sFlt1 levels in WT and mutant mice sera be detected (Figure 7C).

Discussion

The results reported here provide the first evidence that keratins coordinate vascular remodeling in the decidua as well as proper vascularization of the labyrinth in placental development. Their loss led to a reduced oxygen uptake and severe hypoxia. Furthermore, absence of keratins led to an extensive mislocalization of TGCs and a decreased signaling of the placental growth hormones Prlpa and PlGF in the extraembryonic trophoblast. In analogy to previously reported vascularization defects and mislocalization of GLUT1 and GLUT3 transporters in yolk sac tissues,^9,10 we also found altered GLUT1 distribution and vascular defects in the extraembryonic trophoblast. Collectively, our data show that the prenatal death of KtyII^−/− embryos at E9.5 is caused by a sequence of keratin-dependent mechanisms that culminate in metabolic failure in embryonic and extraembryonic tissues. This strongly supports the concept that keratins function in an isotype- and context-dependent manner.

We propose that this defect originates from a mislocalization of trophoblast giant cells, which were identified by Pl1 in situ hybridization. The primary giant cells are the first embryonic epithelial cells to interact with the maternal tissue and express keratins at high levels; they are responsible for the implantation of the embryo.¹⁶ KO embryos developed normally until E8.5 when growth retardation defects emerged; therefore, we propose a defect in secondary giant cells, which arise from the EPC.⁷¹ This became evident from their failure to form a continuous perivitelline network, because the decidua of KtyII^−/− embryos displayed hypervascularization compared with the WT placenta, leading to a very loose meshwork with wide maternal blood lacunae already apparent at E8.5. In humans, high oxygen concentration in the periphery of early placentas, induced by a premature onset of maternal placental circulation, is a key factor for early pregnancy loss. In fact, the intervillous space of the developing placenta is separated from the uterine circulation by plugs of trophoblast cells that occlude the tips of the uteroplacental arteries. In complicated pregnancies, invasion of the endometrium by the extravillous trophoblast is severely restricted compared with the normal situation. Plugging of the spiral arteries is therefore less complete and may predispose to the early onset of the maternal circulation.^40,47 In mice, the TGCs, which function equivalent to human extravillous trophoblasts, invade into the spiral decidual arteries and remodel the maternal blood vessels^71,72 (see Supplemental Figure S3 at http://ajp.amjpathol.org). The mislocalization of the TGCs could account for a defective plugging of maternal vessels caused by keratin deletion, because keratin-depleted cells were shown to exhibit severe migration and attachment defects.⁷³

In addition, TGCs are required for initiating decidualization and regulate its development by the secretion of paracrine hormones, such as proliferin and proliferin-related protein, which have a direct influence on promoting and inhibiting decidual vasculature, respectively.⁷⁴ In this fashion they stimulate angiogenic signals in the decidua-like angiopoietin 1 and 2 and secrete hormones that suppress the immune system of the mother.¹⁶ An involvement of keratins in the secretion of hormones is in line with their reported function in vesicular trafficking and in receptor and transporter localization at the plasma membrane.^10,75 One of these hormones is Prlpa, expressed in secondary TGCs,⁷⁶ which suppresses the maternal uNK cells. Here, we demonstrated a reduced number of TGCs, starting as early as E8.5 in KtyII^−/− embryos and an increased migration of uNK toward the KtyII^−/− placenta.⁷⁷

Hypoxia is a prerequisite for embryonic implantation and development in mammals^40,41 and is necessary for a normal development of the vasculature.⁴² The increased vascularization of the mutant decidua led to an increase in oxygen level of the placenta, and, in strong support, we detected hyperoxia. Furthermore, we found a drastic reduction in Hif1α in the KtyII^−/− decidua and placenta compared with WT. HIFs form heterodimers of Hif1β (ARNT) and Hif1α or Hif2α. Knocking out any of these 3 proteins, leads to a small labyrinth due to drastically impaired vasculogenesis of the placenta, similar to the phenotype we reported in our KO embryos.^17,20,24,43 As a consequence, reduction of placental vasculogenesis led to a drastic increase in hypoxic levels of the embryo proper itself.

Hypoxia triggers multiple responses to overcome decreased oxygen availability, including the expression of growth factors required for establishment of a functional circulatory system. HIF proteins can induce VEGF, a growth factor important for angioblast specification and vasculogenesis.⁴³ Entailing the reduced amount of Hif1α, we saw a reduction in the VEGF expression in KO decidua and placenta, which accounted for the vascularization defects. In line with these data, PlGF, a VEGF-related factor, was also drastically decreased in KO placenta but was not reduced in the surrounding decidua. Reduced amounts of PlGF in turn stimulate uNK cell proliferation. This could account for the increased amounts of uNK cells observed in the surrounding decidua of the KO placenta.⁷⁸ Uterine NK cells were further shown to secrete angiogenic factors such as VEGF and PlGF,⁶⁰ balancing the inhibitory effects of the decidual hyperoxia. Furthermore, VEGF is a known survival factor for embryonic stem cells during prolonged hypoxia as well as for hematopoietic stem cells,^79,80 which are generated among others in the placenta.⁸¹ In line with these findings, we reported that keratin KtyII^−/− placentas did not only exhibit reduced vasculature but also the vessels that were present were dilated and filled with fewer embryonic blood cells, corresponding to hematopoiesis defects in KtyII^−/− embryos in the yolk sac.⁹

The development of the labyrinth depends on an intensive cross talk between the epithelial trophoblast lineage and the endothelial cells of the allantois. Different receptor knockout studies, including fibroblast growth factor receptor 2 and Frizzeld 5,^22,29 as well as various transcription factors JunB, peroxisome proliferator-activated receptor γ, and transcription factor eb,^18,19,26,27 displayed very similar phenotypes to the KtyII^−/− embryos, which might be linked to keratin expression. JunB, for example, an activator protein 1 transcription factor, can induce K18 expression in differentiating embryonic stem cells by binding to an enhancer complex in the first intron.⁸² Various kinases were also reported to show vascularization defects in the placenta and the yolk sac when deleted, such as p38 mitogen-activated protein kinase.²⁵ We linked keratin deficiency to a reduction of phospho-specific activation of p38 in yolk sacs earlier,⁹ which would suggest that keratins influence these signaling functions in the trophoblast cells of the labyrinth as well.

Important signaling factors expressed in the allantoic endothelial side of the labyrinth contribute to the development of the labyrinth. Mutations in the Notch signaling pathway, for instance, block chorioallantoic branching and prevent proper invagination of the allantois into the placenta.^33,35,36 This is also the case for angiopoietin 1 and 2, the vascular remodelers and their receptor Tie2.^30,31,38,39 Defects in yolk sac and placental vascularization defects were also reported for the VEGF receptors, Flk1 and Flt1.^32,34,37

Increased secretion of sFlt1 into the maternal blood stream is a hallmark of preeclampsia. Keratin KO embryos seemed to express the common symptoms, defective trophoblast invasion, hypoxia of the embryo, growth retardation, and defective placental vasculature. However, we could not detect an increase in sFLt1 in the maternal serum. The notion that we did not discover preeclampsia may result from the fact that KO embryos die before the onset of this condition.^70,83

In summary, we propose that keratins are crucial for regulating the maternal-fetal vascularization through the correct localization of TGCs and the secretion of TGC hormones. Keratin-dependent failure of trophoblast localization and secretory function elicited multiple decidual defects (see Model in Figure 9). On the basis of our data, we suggest that mutations in K8 or K18 might be a plausible cause for early pregnancy loss due to decidual hyperoxia.

Figure 9.

Model for decidual hyperoxia caused by keratin deficiency of the TGCs. TGCs, on the one hand, mislocalize and, on the other, fail to secrete appropriate levels of hormones needed for proper decidualisation. This results in a lower degree of vascularization of the placental labyrinth (keratin positive cells in green and keratin KtyII^−/− cells in blue). MV, maternal blood vessels; De, decidua; TGC, trophoblast giant cells; SpT, spongiotrophoblast; L labyrinth; ML, maternal blood lacuna; Al, allantois; Hyp, hypoxia.

Supplementary Data

Supplemental Figure S1

Maternal blood leakage is not caused by TGC lysis. To examine whether the absence of keratins caused trophoblast cytolysis and/or apoptosis, cleaved caspase 3 (cC3) antibody was used to monitor apoptotic cells. Note equivalent levels of apoptosis in KO trophoblasts (C and D and inlay of C) compared with the WT cells (A and B and inlay of A). In line with previous results⁹ apoptosis is increased in the KO embryo itself (D). Furthermore, note relocalization and a reduced number of TGCs (E). P, placenta; De, decidua; TGC, trophoblast giant cells; E, embryo; Ys, yolk sac; EPC, ectoplacental cone. Scale bars: 100 μm (A–D).

Supplemental Figure S2

No differentiation defects in KtyII^−/− embryos. In situ hybridization for labyrinth trophoblast markers Gcm1 and SyncytinA. Gcm1 marks in the chorion the invagination points of the allantois into the labyrinth. Hybridization signals show Gcm1 unchanged gene expression in KtyII^−/− embryos (C) compared with control embryos (A). Note comparable expression of SyncytinA (SynA) in KO (D) compared with WT (B) placentas. E–H: Higher magnifications of above markers. Similar levels of Gcm1 expression in KO placentas (G), compared with a WT labyrinth (E). At higher magnification, no obvious differentiation defect in the KO (H) compared with the WT (F) placenta was notable. However, different localization pattern of SynA in the syncytiotrophoblast cells enveloping the dilated embryonic blood vessels in the KO labyrinth compared with the controls was observed (F and H). Scale bars: 100 μm (A–D); 10 μm(E–H).

Supplemental Figure S3

Dysfunctional vasculature in the placental labyrinth area of KtyII^−/− embryos. Immunohistochemical staining with anti-CD31 antibodies in WT (A, C, and D) and KO (B, E, and F) placentas showed extensive vasculature in WT labyrinth (A). In contrast, note the near absence of embryonic vessels in the KO placenta. Also note accumulation of TGC and the presence of maternal blood lacunae in the KO placenta (E). At higher magnifications (C and D), embryonic WT controls displayed small vessels filled with nucleated embryonic erythrocytes, in close vicinity to maternal blood sinuses (D and inlay of C). In contrast, KO vessels were dilated with a thickened syncytiotrophoblast layer, containing few nucleated embryonic erythrocytes (F and inlay of E). De, decidua; TGC, trophoblast giant cell; SpT, spongiotrophoblast; L labyrinth; ML, maternal blood lacuna; Al, allantois; RM, Reichert's membrane; Ys, yolk sac; E, embryo; SyT, syncytiotrophoblast; EB, embryonic blood; MS, maternal blood sinus. Scale bars: 50 μm (A and F); 10 μm (B–E and G–J).

Supplemental Figure S4

TGCs lining maternal spiral arteries in the WT. Immunofluorescence analysis of WT decidual tissue staining with K19 depicts TGCs lining the maternal spiral arteries in E9.5 embryos. De, decidua; TGC, trophoblast giant cell; SpA-TGC, spiral artery-associated trophoblast giant cells; MB, maternal blood (showing nonspecific cross-reactivity). Sale bars: 10 μm (A–D).