HIF2α-induced lysyl oxidase safeguards successful pregnancy by remodelling collagens at the feto-maternal interface

TL;DR

The study shows how hypoxia in the uterus helps pregnancy by triggering HIF2α and lysyl oxidase to reshape the tissue where the embryo implants.

Contribution

The study identifies HIF2α and lysyl oxidase as key players in hypoxia-driven extracellular matrix remodeling during early pregnancy.

Findings

HIF2α induces lysyl oxidase to stabilize collagen fibrils at the feto-maternal interface.

Disruption of HIF2α or lysyl oxidase in the uterus leads to infertility due to impaired trophoblast invasion.

Extracellular matrix remodeling in the decidua is critical for successful embryo implantation and placental formation.

Abstract

During the initial stage of gestation, precise constructions of the microenvironment at the feto-maternal interface are critical for successful embryonic development to term. It consists of the fetal placenta and the maternal decidua, contributing to favorable pregnancy outcomes and long-term health in both mother and child. Remarkably, although our previous work demonstrated that physiological uterine hypoxia promotes blastocyst implantation into the decidua basalis and onset of trophoblastic invasion via the hypoxia-inducible factor (Hif) signaling pathway, it remains unclear which key regulatory cascades are triggered underlying. Here, we harnessed recent advances of the spatial transcriptomic technology in combination with genetic mouse models to gain a more comprehensive understanding of physiological uterine hypoxia. We revealed that hypoxia-induced remodeling of the extracellular matrices (ECMs) allows the blastocyst-derived trophoblastic cells to invade into the uterine stroma upon implantation. Mechanistically, Hif2α drives the expression of lysyl oxidase (Lox) that crosslinks collagen I fibrils and their stabilization in the vicinity of the attached embryos. Uterine-specific knockout (uKO) of Hif2α and Lox compromised the process of embryo invasion, leading to female infertility. Collectively, our findings provide mechanistic insight into how hypoxia governs early embryonic invasion and contributes to successful pregnancy.

Hypoxia in the primary decidual zone induces Hif2α, which regulates lysyl oxidase (Lox) expression to remodel the extracellular matrix at the feto–maternal interface.Genetic disruption of Hif2α or Lox in the uterus impairs trophoblast invasion and compromises fertility, highlighting a critical hypoxia-driven signaling axis.Remodeling of the maternal decidual ECM network directly influences embryonic invasive capacity and subsequent placental formation, underscoring the importance of maternal tissue architecture in pregnancy outcomes.

Hypoxia in the primary decidual zone induces Hif2α, which regulates lysyl oxidase (Lox) expression to remodel the extracellular matrix at the feto–maternal interface.

Genetic disruption of Hif2α or Lox in the uterus impairs trophoblast invasion and compromises fertility, highlighting a critical hypoxia-driven signaling axis.

Remodeling of the maternal decidual ECM network directly influences embryonic invasive capacity and subsequent placental formation, underscoring the importance of maternal tissue architecture in pregnancy outcomes.

Figures

A; Supplementary Table 1). We analyzed further the stromal cells and identified a range of well-documented endometrial cell types, which evolved over the course of these three days of pregnancy: on Day 4, before the embryo attachment occurred, the stromal cell types comprised mainly non-proliferative and epithelial-mesenchymal transition (EMT)-like cell types; on Day 5, after attachment, the number of proliferative and the embryo-attached cells increased, and on Day 6, when the embryos invade the lining of the uterus, PDZ and SDZ cells appeared (Fig.[1D](#Fig1) and Supplementary Fig. 1B; Suppl

A; Supplementary Table 3). There were no significant alterations in the ratios of each cell cluster per total stromal cells (Supplementary Fig. 2B). We focused on the PDZ cells for further analysis because the PDZ region is avascular and hypoxic as well as these cells express high levels of *Hif2α* [[20](#CR20), [21](#CR21)](Supplementary Fig 3). By comparing gene expression in PDZ cells of *Hif2α* uKO and control uteri, we found 167 upregulated genes in the *Hif2α* uKO and 397 upregulated genes in the controls, of which 40 genes were targets of Hif proteins (Fig. [2C](#Fig2) and Supplementary

Abnormal collagen deposition in the decidua of *Hif2α* uKO uteri during trophoblast invasion.A Representative TEM images of embryo implantation sites in sections of control (*Hif2a*^*ff*^) and *Hif2α* uKO (*Hif2a*^*ff*^ *Pgr*^*Cre/+*^) uteri on the evening of Day 5. Trophoblasts and luminal epithelial cells are demarcated by dashed lines in blue or orange, respectively. Red arrowheads, remaining basement membrane; open, green arrowheads, collagen fibers; Tr, trophoblasts; Epi, luminal epithelial cell; Dec, decidual cell. Scale bar, 2 µm. *n* = 3 for each genotype. B Co-immunostaining of collag

Lox remodels the collagen-containing ECM around implantation sites.A Uniform manifold approximation and projection (UMAP) analysis of spatial transcriptomics data categorized by cell type, as indicated in the color key, from control (*Lox*^*ff*^) and *Lox* uKO (*Lox*^*ff*^*Pgr*^*Cre/+*^) uteri on Day 6 of pregnancy. uKO_Unique_Str, stromal cells found only in the *Lox* uKO. B Histology (top) and spatial transcriptomics (bottom) of implantation sites in control and *Lox* uKO uteri on Day 6. M, mesometrial pole; AM, anti-mesometrial pole. Arrowheads indicate blastocysts. Scale bar, 500 µm. C Ups

Lox is required for breakdown of the basement membrane and invasion of trophoblasts.A Representative TEM images of embryo implantation sites in control (*Lox*^*ff*^) and *Lox* uKO (*Lox*^*ff*^*Pgr*^*Cre/+*^) uteri on the evening of Day 5. The regions indicated by dashed lines are enlarged in the right-hand panels. Scale bars, 20 µm (left-hand images), 5 µm (right-hand images). Trophoblasts and luminal epithelial cells are demarcated by dashed lines in blue or orange, respectively. Red arrowheads, remaining basement membrane. Tr trophoblasts, Epi luminal epithelial cells, Dec decidua cells. *n*

Collagens I and IV are remodeled by Lox during embryo invasion.A Co-immunostaining of collagen I (Col I) and cytokeratin (CK-8) and DAPI staining of DNA in nuclei (Nu) in sections of control (*Lox*^*ff*^) and *Lox* uKO (*Lox*^*ff*^*Pgr*^*Cre/+*^) uteri on Day 6. The regions indicated by dashed lines are enlarged in right panels. M, mesometrial pole; AM, anti-mesometrial pole; le, luminal epithelium; ge, glandular epithelium; s, stroma. Asterisks (_*_) indicate blastocysts. Scale bars, 100 µm (left), and 20 µm (right). *n* = 6 for each genotype. B Immunostaining of collagen IV (Col IV) in secti

Full text

Introduction

Embryo implantation into the lining of the uterus is crucial for a successful pregnancy [1–3]. Implantation defects in mice, for example, can cause miscarriage and pre-term births [1, 2], and such defects may contribute to infertility in humans, which affects around 15% of couples globally [1].

The mouse blastocyst comprises an inner cell mass, which will go on to form the embryo proper, in a fluid-filled space surrounded by a layer of trophoblast cells called the trophectoderm, which will eventually give rise to the placenta. Implantation-competent blastocysts enter the uterine cavity on the morning of Day 4 after fertilization (Day 1 = the day of vaginal plug found). They space out, and by midnight of that day, they attach to the luminal epithelium lining the uterus. By the following morning, Day 5, an implantation crypt has formed around each blastocyst, and the neighboring stroma cells of the uterus begin to differentiate into decidua cells, a process known as decidualization.

The stroma cells nearest the blastocyst differentiate into epithelial-like primary decidual cells, which, by Day 6, form a primary decidual zone (PDZ) around the blastocyst in the implantation crypt. The cells of this layer adhere to each other tightly due to large amounts of the intercellular adhesion proteins E-cadherin and ZO-1 [4], and the PDZ is avascular, so it provides a permeability barrier that protects embryos from immune cells, pathogens and other harmful agents in the maternal circulation [5, 6]. Consequently, this layer is thought also to be hypoxic [7]. Physiological uterine hypoxia has been proposed to promote blastocyst attachment and trophoblast differentiation, in part through the hypoxia-inducible factor (Hif) signaling pathway, yet the downstream targets and mechanisms remain only partly understood.

Attachment of the blastocyst facilitates differentiation of PDZ cells in the afternoon of Day 5 [5, 8]. The luminal epithelial cells facing the blastocyst begin to detach from each other, creating gaps in the epithelium [8–10]. The basement membrane underlying the luminal epithelium collapses, and the number of epithelial cells decreases due to engulfment by trophoblasts or cell death [8, 9]. By the evening of Day 5, trophoblasts migrate out of the blastocyst, through the gaps in the epithelium and invade the PDZ. Subsequently, direct contacts with PDZ cells guide the trophoblasts to the secondary decidual zone (SDZ), which comprises polyploid cells differentiated from the outer stroma region, where trophoblasts eventually differentiate into the placenta lineage. Coincident with trophoblastic invasion, the extracellular matrix (ECM) network undergoes spatiotemporal remodeling: notably, the expression and distribution of collagen I (Col I) and collagen IV (Col IV) change upon embryo attachment [11–14]. Col I is a fibrillar collagen that forms dense, mesh-like structures in the tissue stroma [15, 16]. Col IV, by contrast, is a network-forming collagen and a major component of the basement membrane, which separates epithelia from their surrounding tissue compartments [16]. Previous studies have shown that ECM remodeling is tightly regulated during implantation and that aberrant ECM composition is associated with implantation failure and pregnancy disorders [17–19]. A recent in vitro study culturing embryos on collagen gels suggested that endometrial ECM facilitates embryo growth [17]. ECM components, including collagens, are degraded by metalloproteases (MMPs), which are inhibited by tissue inhibitors of MMPs (TIMPs) [15, 18]. In the mouse uterus, Mmp2 is constitutively expressed during pregnancy, and Mmp9 is strongly expressed in trophoblasts during invasion [19]. Timp-1 and -2 are highly induced after embryo attachment, which may inhibit Mmp2 function [19]. However, despite these advances, it remains unclear how maternal ECM networks are actively remodeled in vivo during implantation, and how their regulation by hypoxia contributes to the success or failure of pregnancy.

In this study, we aimed to elucidate how hypoxia at the feto-maternal interface, and specifically in the primary decidual zone (PDZ), contributes to the successful establishment of pregnancy. Building on our previous finding that the hypoxia-inducible transcription factor Hif2α is indispensable for implantation, we sought to identify the downstream pathways by which hypoxia-Hif2α regulates extracellular matrix remodeling around the implantation site. In particular, we focused on whether collagen organization and its enzymatic regulation by lysyl oxidase (Lox) play critical roles in enabling trophoblast invasion and placentation. To address this, we combined spatial transcriptomics with ultrastructural analyses and genetic mouse models, allowing us to dissect the mechanisms by which uterine hypoxia orchestrates tissue remodeling required for embryo implantation.

Results

Hif2α deletion compromises collagen remodeling in the decidua

To investigate the potential mechanisms by which the decidua, especially PDZ, regulates the invasion of the trophoblast cells into the endometria, we performed spatial transcriptome analysis of mouse uterine tissues on Days 4, 5 and 6 of pregnancy. To identify the implantation sites, we intravenously injected Chicago blue dye (Fig. 1A), stained sections of tissue for histology and processed them for spatial transcriptomics (Fig. 1B and Supplementary Fig. 1). Uniform manifold approximation and projection (UMAP) analysis of the spatial transcriptomics data of whole tissues identified stromal cells and myometrial cells as well as luminal and glandular epithelial cells (Fig. 1C and Supplementary Fig. 1A; Supplementary Table 1). We analyzed further the stromal cells and identified a range of well-documented endometrial cell types, which evolved over the course of these three days of pregnancy: on Day 4, before the embryo attachment occurred, the stromal cell types comprised mainly non-proliferative and epithelial-mesenchymal transition (EMT)-like cell types; on Day 5, after attachment, the number of proliferative and the embryo-attached cells increased, and on Day 6, when the embryos invade the lining of the uterus, PDZ and SDZ cells appeared (Fig.1D and Supplementary Fig. 1B; Supplementary Table 2). MSigDB Hallmark and GO term analyses of the data for the PDZ and SDZ cells revealed upregulation of hypoxia signaling (Fig. 1E) and collagen-containing extracellular matrix gene transcription (Fig. 1F) in both cell types. These findings are consistent with previous evidence that the PDZ is avascular [6] and therefore hypoxic, and that stromal Hif2α is indispensable for embryo invasion [10].Fig. 1. Spatial transcriptomics of peri-implantation uteri reveals enrichment of hypoxia- and ECM-related gene transcription in the decidua.A Schematic diagram of embryo implantation. Four days after fertilization, blastocysts arrive in the uterine horns. By the midnight of Day 4, they attach to the luminal epithelium (LE). From the morning of Day 5 onwards, implantation sites can be visualized by injecting Chicago blue dye intravenously, due to the increased vascular permeability of the stroma (Str) during the attachment reaction. By the morning of Day 6, the primary decidual zone (PDZ) and secondary decidual zone (SDZ) become evident, and trophoblasts (Tr) invade the decidua. ICM, inner cell mass; TE, trophectoderm. B Histology (top) and spatial transcriptomics analysis (using 10× Visium) of sections of whole tissue (middle) and stroma only (bottom) from pregnant uteri on Days 4–6. Arrowheads indicate blastocysts. Scale bar, 500 µm. C Uniform manifold approximation and projection (UMAP) analysis of spatial transcriptomics data from whole tissues, as in (B) and categorized by cell type, as indicated in the color key. LE, luminal epithelial cells; Str, stromal cells; GE, glandular epithelial cells; Myo, myometrial cells. D UMAP analysis of spatial transcriptomics data from stroma as in (B) and categorized by stromal cell types, as indicated in the color key. Prolif, proliferating cells; Non-prolif, non-proliferating cells; EMT, epithelial-mesenchymal transitional cells; Str-M, mesometrial stromal cells; Str-AM, anti-mesometrial stromal cells; PDZ, PDZ cells; SDZ, SDZ cells. E MSigDB Hallmark prediction of upstream transcription factors and F gene ontology (GO) analysis in PDZ (top) and SDZ cells (bottom), determined using Enrichr. In both E and F, bubble size reflects statistical significance (–log_10_ adjusted P-value) of enrichment, while color scale represents the average expression level of the associated genes. “Gene Ratio” indicates the proportion of genes in each term relative to the total number of genes.

Spatial transcriptome analysis of uterus tissues from the implantation sites of Hif2α uKO mice and control mice on Day 6 of pregnancy identified five cell types in both, including PDZ and SDZ cells (Fig. 2A, B and Supplementary Fig. 2A; Supplementary Table 3). There were no significant alterations in the ratios of each cell cluster per total stromal cells (Supplementary Fig. 2B). We focused on the PDZ cells for further analysis because the PDZ region is avascular and hypoxic as well as these cells express high levels of Hif2α [20, 21](Supplementary Fig 3). By comparing gene expression in PDZ cells of Hif2α uKO and control uteri, we found 167 upregulated genes in the Hif2α uKO and 397 upregulated genes in the controls, of which 40 genes were targets of Hif proteins (Fig. 2C and Supplementary Table 4). We predicted the molecules upstream of these upregulated genes in the control by using the enrichment analysis tool Enrichr [22] and the RNA-seq data resource ARCHS^4^, which showed that genes upregulated in the control were targets of HIF1α and HIF2α (Fig. 2D). Moreover, reactome analysis indicated that the upregulated genes specifically in either control or Hif2α uKO PDZ cells were involved in collagen assembly, biosynthesis and modification trimerization and fibril crosslinking (Fig. 2E), suggesting that Hif2α deletion in the PDZ compromises collagen remodeling in the extracellular matrix.Fig. 2. Spatial transcriptomics of Hif2α uKO uteri reveals upregulation of ECM-related genes in the PDZ.A Uniform manifold approximation and projection (UMAP) analysis of spatial transcriptomics data categorized by cell type, as indicated in the color key, in control (Hif2a^ff^) and Hif2α uKO (Hif2a^ff^ Pgr^Cre/+^) uteri on Day 6 of pregnancy. PDZ, PDZ cells; SDZ, SDZ cells; Str-AM, anti-mesometrial stromal cells; Str-M, mesometrial stromal cells; Myo, myometrial cells. B Histology (top) and spatial transcriptomics analysis (bottom) of implantation sites in the control and Hif2α uKO uteri on Day 6. Arrowheads indicate blastocysts. M, mesometrial pole; AM, anti-mesometrial pole. Scale bar, 500 µm. C Volcano plot depicting differentially expressed genes (DEGs) in control and Hif2α uKO PDZ cells. Upregulated genes in control cells are shown in magenta and those in Hif2α uKO PDZ cells are shown in blue. DEGs targeted by Hif proteins are indicated by open black circles. D Upstream transcription factors of the upregulated genes in control cells predicted using ARCHS^4^ at Enrichr. E Analysis with Reactome at Enrichr of genes upregulated in PDZ cells from controls (left) and in those from Hif2α uKO cells. In (D) and (E) “Gene Ratio” indicates the ratio number of GO term per the total number of genes.

To validate our conclusions from spatial transcriptomics, we used transmission electron microscopy (TEM) to analyze embryo implantation sites in Hif2α uKO and control uteri on Day 5 of pregnancy. We found control epithelial cells were detached from the basement and trophoblasts invaded decidua, while there was poor collagen deposition in Hif2α uKO decidua, with a substantial epithelium on the basal membrane, and flawed trophoblast invasion (Fig. 3A). Immunostaining of the implantation sites on Day 6 showed that Col l fibrils were abnormal in the absence of Hif2α: Col I fiber network was evident in the controls while it was fragmented in Hif2α uKO decidua (Fig. 3B).Fig. 3. Abnormal collagen deposition in the decidua of Hif2α uKO uteri during trophoblast invasion.A Representative TEM images of embryo implantation sites in sections of control (Hif2a^ff^) and Hif2α uKO (Hif2a^ff^ Pgr^Cre/+^) uteri on the evening of Day 5. Trophoblasts and luminal epithelial cells are demarcated by dashed lines in blue or orange, respectively. Red arrowheads, remaining basement membrane; open, green arrowheads, collagen fibers; Tr, trophoblasts; Epi, luminal epithelial cell; Dec, decidual cell. Scale bar, 2 µm. n = 3 for each genotype. B Co-immunostaining of collagen I (Col I) and cytokeratin 8 (CK-8), and DAPI staining of DNA in nuclei (Nu) in sections of control and Hif2α uKO uteri on Day 6. CK-8 staining indicates luminal epithelia and trophoblasts. The regions indicated by dashed lines are enlarged in the middle and right panels. M, mesometrial pole; AM, anti-mesometrial pole; le, luminal epithelium; ge, glandular epithelium; s, stroma. Asterisks (*) indicate blastocysts. Scale bars, 200 µm (left), 50 µm (middle), and 20 µm (right). n = 3 for each genotype. C In situ hybridization of Lox mRNA (black) in sections of the PDZs of control and Hif2α uKO uteri on Day 6. Nuclei were counter stained by Nuclear Fast Red (Pink). Scale bar, 200 µm. n = 3 for each genotype.

We hypothesized that the abnormal Col I network in Hif2α uKO decidua might be due to low expression of Lox as the Lox promoter contains a Hif-response element, and expression of this gene is induced by Hif proteins under hypoxic conditions [23]. Moreover, spatial transcriptomics showed Lox was highly expressed at PDZ implantation sites in the control on Day 6 (Supplementary Fig. 3). To investigate this hypothesis, we used in situ hybridization to analyze Lox expression at implantation sites in Hif2α uKO and control uteri on Day 6 and found much reduced expression in the absence of Hif2α when compared with the control (Fig. 3C and Supplementary Fig. 4). These data indicate that Lox-induced changes in collagen architecture in the decidual stroma may be important for Hif2α-mediated embryo implantation.

Lox deletion in the uterus compromises fertility

To investigate whether Lox plays a similar role in Hif2α-mediated trophoblast invasion, we used in situ hybridization to examine Lox expression in the stroma surrounding blastocysts during early pregnancy and found high Lox mRNA levels on Day 5 and Day 6 (Supplementary Fig. 5A), consistent with a previous report [24]. Lox expression was localized to the stroma adjacent to the blastocysts (Supplementary Fig. 5A), similar to Hif2α expression [21]. These findings suggest that Lox functions after embryo attachment.

To assess the biological significance of Lox in pregnancy, we established mice with a conditional deletion of Lox in the uterus (Lox uKO mice) by mating Lox-floxed mice with Pgr-Cre (uterine-specific Cre) mice (Supplementary Fig. 5B). Exons 3 and 4 of the Lox gene were flanked by loxP sites to facilitate Cre-mediated deletion. We validated Pgr*-Cre*-mediated genomic deletion of the floxed region in uterine tissues by PCR (Supplementary Fig. 5C) and the absence of Lox mRNA from implantation sites by in situ hybridization (Supplementary Fig. 5D).

When Lox uKO female mice were mated with fertile male mice, the pregnancy success rate was significantly lower than that when control females were mated (Fig. 4A), and those Lox uKO females that successfully delivered had litter sizes only half of the average litter size in controls (Fig. 4B). On Day 12 of pregnancy, we observed embryo resorption and hemorrhaging between the implantation sites in the uteri of Lox uKO females, which was not observed in the controls (Fig. 4C), and the lower weights of the uteri indicated impaired embryo growth (Fig. 4D). Indeed, the ratio of abnormal embryos per total litters was significantly increased in uKO uteri (Supplementary Fig. 6).Fig. 4Lox deletion in the uterus compromises fertility.Pregnancy rates (A) and litter sizes (B) of control (Lox^ff^) and Lox uKO (Lox^ff^Pgr^Cre/+^) mice are shown. n = 8 and 13 in (A), n = 7 and n = 5 in (B) respectively. The numbers on the bars in (A) indicate dams with successful term delivery per total plug-positive ones. C Representative images of pregnant uteri from control and Lox uKO mice on Day 12. Open-red arrowheads indicate bleeding, and black arrowheads indicate embryo resorption. Scale bar, 5 mm. D Average weight per implantation site in C; n = 4 and 6, respectively. Representative images of pregnant uteri from control and Lox uKO mice on Day 6 (E) and Day 5 (F). n = 12 and 7 in (E) n = 4 and n = 3 in (F) respectively. Scale bar, 5 mm. G Representative images showing in situ hybridization of mRNAs encoding the decidua markers Bmp2 (top) and Ptgs2 (bottom) at implantation sites in control and Lox uKO mice on Day 6. E-cadherin (E-Cad; a marker of epithelial cells) was co-immunostained and DNA in nuclei (Nu) was stained with DAPI. M, mesometrial pole; AM, anti-mesometrial pole; le, luminal epithelium; s, stroma. Asterisks (*) indicate blastocysts. Scale bar, 200 µm. n = 3 for each genotype. H Three-dimensional imaging of luminal and glandular epithelia at implantation sites in control and Lox uKO mice on Day 6. Asterisks (*) indicate blastocysts. Scale bar, 200 µm. Data are means ± SEM and statistical significance was determined by using Student’s t test. n = 3 for each genotype.

Embryo resorption can be caused by defects in implantation [1, 2, 25]. To investigate whether implantation defects might explain the resorbed embryos we saw in the uteri of Lox uKO females, we examined the uteri of pregnant Lox uKO and control females on Days 5 and 6 of pregnancy, when implantation typically occurs, by staining with Chicago blue dye, as before, to visualize the embryo attachment sites. Surprisingly, we saw similar numbers of blue bands in the uteri of Lox uKO and control females on Day 6 (Fig. 4E) and even on Day 5 in Lox-deficient uteri (Fig. 4F), indicating apparently normal embryo attachment despite the absence of Lox. Moreover, the receptivity-related molecules Ki67 (a marker of proliferation, which should be expressed in the receptive stroma), Pgr and Esr1 (female hormone receptors), and FoxA2 (a marker of glandular epithelium) were expressed similarly in the uteri of Lox uKO and control females on Day 4 of pregnancy (Supplementary Fig. 7), indicating normal uterine receptivity in Lox uKO.

To assess decidualization at the implantation sites on Day 6, we used in situ hybridization to analyze expression of Bmp2 and Ptgs2, which are markers of decidualization following embryo attachment [1]. Consistent with our observation of normal attachment sites and uterine receptivity in Lox uKO females, we observed that these markers of decidualization were expressed similarly in both Lox uKO and control females (Fig. 4G and Supplementary Fig. 8) Ptgs2 expression was seen in mesometrial stroma, and Bmp2 broadly in decidualizing stromal cells.

Another important feature of pregnant uteri on Day 6 is the epithelial crypt that surrounds each attached embryo [26, 27]. After embryo attachment, the surrounding luminal epithelium layer initiates invagination into the anti-mesometrial pole, accompanied by glandular epithelium, and anomalies in this crypt-shaping, as seen by three-dimensional (3D) imaging, lead to defective pregnancy outcomes [21, 26, 27]. We performed 3D imaging of the luminal and glandular epithelia at implantation sites in Lox uKO and control females on Day 6 and found no abnormalities in crypt shapes (Fig. 4H). Likewise, in Hif2α-deleted uteri, we found deeply invaginated implantation crypts (Supplementary Fig. 9), indicating that neither loss of Hif2a nor loss of Lox influences crypt shape. In addition, embryo attachment and crypt formation on the morning of Day 5 (Fig. 4F and Supplementary Fig. 10A), and PDZ formation with epithelial and avascular characters depicted by ZO-1 and Flk1 staining, which starts on the afternoon of Day 5 (Supplementary Fig. 10B and C) [5], were all normal in Lox uKO mice. These data indicate that in the uterus, Lox contributes to healthy pregnancy outcomes by regulating early events through an unknown molecular mechanism.

Lox remodels the collagen-containing ECM around implantation sites

To gain insights into the possible mechanisms by which the absence of Lox might impact pregnancy, we performed spatial transcriptomics on sections of uterus tissue from Lox uKO and control female mice on Day 6 of pregnancy and used UMAP analysis to categorize seven cell types (Fig. 5A, B and Supplementary Fig. 11; Supplementary Table 5). Most strikingly, one category of stromal cells was found only in the Lox uKO mice (uKO_Unique_Str: Fig. 5A and Supplementary Fig. 11). These cells were seen specifically around the PDZ region in tissue sections (Fig. 5B), suggesting abnormal stromal cell differentiation. Marker gene analysis revealed lower proliferative activity compared with SDZ cells, as indicated by decreased Mki67 expression (Supplementary Fig.11 ). Notably, uKO_Unique_Str cells expressed Lox at much lower levels than other stromal cell types (Supplementary Fig. 11). To further characterize these uKO_Unique_Str cells, we analyzed the uniquely downregulated genes in uKO_Unique_Str cells compared to other cell types by using ARCHS^4^ in Enrichr to predict upstream factors (Fig. 5C). These data predicted that gene expression in uKO_Unique_Str cells is regulated by HIF2α, among other transcription factors, and by fibrillin 1 (Fbn1), a component of the ECM (Fig. 5C). Gene ontology (GO) analysis of the uniquely downregulated genes in uKO_Unique_Str cells further revealed that they were related to collagen-containing ECM pathways and cytoskeleton-related pathways (Fig. 5D). These data indicate that deletion of the Lox gene in uKO_Unique_Str cells acts downstream of Hif2a and affects the ECM and cytoskeleton.Fig. 5. Lox remodels the collagen-containing ECM around implantation sites.A Uniform manifold approximation and projection (UMAP) analysis of spatial transcriptomics data categorized by cell type, as indicated in the color key, from control (Lox^ff^) and Lox uKO (Lox^ff^Pgr^Cre/+^) uteri on Day 6 of pregnancy. uKO_Unique_Str, stromal cells found only in the Lox uKO. B Histology (top) and spatial transcriptomics (bottom) of implantation sites in control and Lox uKO uteri on Day 6. M, mesometrial pole; AM, anti-mesometrial pole. Arrowheads indicate blastocysts. Scale bar, 500 µm. C Upstream transcription factors of uniquely downregulated genes in uKO_Unique_Str cells were predicted by using ARCHS^4^ in Enrichr. D Gene ontology analysis of uniquely downregulated genes in uKO_Unique_Str cells performed using Enrichr. E Volcano plot depicting differentially expressed genes (DEGs) in control and Lox uKO PDZ cells. Genes upregulated in the control cells are shown in magenta and those in Lox uKO cells in blue. F Upstream transcription factors of highly upregulated genes in control PDZ cells predicted using ARCHS^4^ in Enrichr. G GO analysis of highly upregulated genes in control PDZ cells performed using Enrichr. H–J as for E–G for genes upregulated in SDZ cells. In (C), (D), (F), (G), (I), and (J), ”Gene Ratio” indicates the ratio number of GO term per the total number of genes.

In addition to this unique stromal cell population in the Lox uKO uteri, we saw differential gene expression in the PDZ and SDZ cell types when compared to the control. In the PDZ cells, we identified 705 differentially expressed genes: 554 of which were upregulated in Lox uKO and 151 in control pregnant females (Fig. 5E and Supplementary Table 6). Prediction of upstream factors targeting genes upregulated in the control included HIF2α (Fig. 5F). GO analysis revealed that genes involved in extracellular structure and ECM organization were both up- and down-regulated in Lox uKO PDZ cells (Fig. 5G). Similarly, the SDZ cells in Lox uKO females showed differential gene expressions compared with controls (Fig. 5H and Supplementary Table7). HIF2α was predicted to be a highly enriched upstream factor in the controls (Fig. 5I), and, as in PDZ cells, genes involved in extracellular structure and ECM organization were highly expressed in Lox uKO cells and genes involved in collagen-containing ECM processes were highly expressed in control cells (Fig. 5J).

Taken together with our evidence implicate signaling through HIF2α and Lox in PDZ cells plays a crucial role in remodeling the collagen-containing ECM around implantation sites in early pregnancy.

Lox is required for breakdown of the basement membrane and invasion of trophoblasts

To examine whether the altered ECM-related gene expression we saw in the PDZ and SDZ cell types of Lox uKO pregnant females when compared to the controls was reflected in the morphology of the tissue around implantation sites, we used TEM to image these sites on Day 5 and Day 6 of pregnancy. By the evening of Day 5, the luminal epithelium of the uterine cavity adjacent to the pre-implantation embryo had been removed in the controls (Fig. 6A). However, in the Lox uKO females, the trophoblasts remained near the epithelial layer and within the basement membrane underlying the luminal epithelial cells, whereas in the controls, they were in direct contact with decidual cells (Fig. 6A). This indicates that in the absence of Lox, the basement membrane of the epithelium fails to break down, so the trophoblasts cannot invade the decidual layer.Fig. 6. Lox is required for breakdown of the basement membrane and invasion of trophoblasts.A Representative TEM images of embryo implantation sites in control (Lox^ff^) and Lox uKO (Lox^ff^Pgr^Cre/+^) uteri on the evening of Day 5. The regions indicated by dashed lines are enlarged in the right-hand panels. Scale bars, 20 µm (left-hand images), 5 µm (right-hand images). Trophoblasts and luminal epithelial cells are demarcated by dashed lines in blue or orange, respectively. Red arrowheads, remaining basement membrane. Tr trophoblasts, Epi luminal epithelial cells, Dec decidua cells. n = 3 for each genotype. B As in A), but the regions indicated by pink dashed lines are enlarged in the middle panels, and those indicated by green dashed lines are enlarged in the right panels. Scale bars, 5 µm (left-hand images), 2 µm (middle and right-hand images). Trophoblasts and luminal epithelia are demarcated by dashed lines in blue or orange, respectively. Open, green arrowheads, collagen fibers; green arrowheads, spotty collagen; blue dashed lines, trophoblasts. n = 3 for each genotype. C Representative images showing the immunostaining of tissue sections at implantation sites for cytokeratin 8 (CK-8; a marker of trophoblasts and epithelial cells) and E-cadherin (E-Cad; a marker of epithelial cells), and DAPI staining of DNA in nuclei (Nu) in control and Lox uKO uteri on Day 6. White arrowheads, abnormally invading trophoblasts from the ecto-placental cone. M mesometrial pole, AM anti-mesometrial pole. Asterisks (*) indicate blastocysts. Scale bar, 100 µm. n = 6 for each genotype.

Anomalies in trophoblast invasion were also evident on the morning of Day 6. Although epithelial cells engulfed by trophoblasts were seen in both Lox uKO and control uteri (Fig. 6B, middle panels), in Lox uKO uteri the trophoblasts remained within the epithelial layer (Fig. 6B, right bottom panel), whereas in the controls they infiltrated further into the decidual regions, making contact with ECM fibers (Fig. 6B, right top panel). Moreover, ECM fibers were sparse in the stroma of the following the Lox uKO uteri (Fig. 6B, right bottom panel), indicating that the anomalies due to loss of Lox affect the stroma rather than the epithelia.

We confirmed the anomalous trophoblast invasion on Day 6 by immunostaining sections of tissue at the implantation sites for CK-8, a marker of trophoblast and epithelial cells and for E-Cad, a marker of epithelial cells. In Lox uKO uteri, we saw poor trophoblast invasion of the decidua whereas the control uteri exhibited prominent trophoblast invasion into the primary decidua at the lateral sides (Fig. 6C). Together, these data indicate that Lox is required for breakdown of the basement membrane, invasion of trophoblasts from the blastocyst into the decidual layer of the uterus at embryo attachment sites, and for maintenance of ECM fibers in the stroma.

Collagens I and IV are remodeled by Lox during embryo invasion

Consistent with our evidence that Lox is required for breakdown of the basement membrane, and for maintenance of ECM fibers in the stroma, the composition of the ECM is known to undergo significant changes during embryo implantation in rodents [11]. To investigate whether the requirement for Lox for breakdown of the basement membrane and invasion of trophoblasts during embryo implantation might be due to effects on these collagen structures in the wall of the uterus, we immunostained sections of Lox uKO and control uteri from mice on Day 6 of pregnancy. In control uteri, we observed prominent Col I fibers surrounding decidual cells and invading trophoblasts, whereas in Lox uKO uteri Col I signals appeared fragmented and trophoblast invasion was impaired (Fig. 7A), similar to those seen in Hif2α-deficient uteri (Fig. 3). These data suggest that Lox supports trophoblast invasion by crosslinking Col I fibers. We also detected abnormal localization of the Col IV basement membrane in Lox uKO uteri (Fig. 7B). In control decidua, the basement membrane was degraded around invading cells with enlarged nuclei—a characteristic of trophoblasts—whereas in uKO samples the membrane remained intact (Fig. 7B). To further investigate the relationship between these two types of collagens, we performed double immunostaining for Col I and Col IV in uterine sections collected at multiple time points from 10 am on Day 5 to 10 am on Day 6. In control uteri, Col I fibers progressively assembled in the stroma after embryo attachment, while the Col IV-containing basement membrane near the blastocyst degraded (Fig. 7C). In Lox uKO uteri, by contrast, Col I staining was weak at early time points and later appeared stronger; however, as shown in Fig. 7A, Col I remained fragmented and failed to form normal fibrillar structures. At the same time, Col IV-containing basement membranes persisted adjacent to the blastocysts in uKO samples (Fig. 7C), as further confirmed by higher-magnification images (Fig. 7B). This abnormal network of Col I and IV seen in Lox uKO mice may explain the failure of the basement membrane to break down and the failure of trophoblasts to invade the decidual layer in the absence of Lox, as shown above (Fig. 6).Fig. 7. Collagens I and IV are remodeled by Lox during embryo invasion.A Co-immunostaining of collagen I (Col I) and cytokeratin (CK-8) and DAPI staining of DNA in nuclei (Nu) in sections of control (Lox^ff^) and Lox uKO (Lox^ff^Pgr^Cre/+^) uteri on Day 6. The regions indicated by dashed lines are enlarged in right panels. M, mesometrial pole; AM, anti-mesometrial pole; le, luminal epithelium; ge, glandular epithelium; s, stroma. Asterisks (*) indicate blastocysts. Scale bars, 100 µm (left), and 20 µm (right). n = 6 for each genotype. B Immunostaining of collagen IV (Col IV) in sections of control (Lox^ff^) and Lox uKO (Lox^ff^Pgr^Cre/+^) uteri on Day 6. DNA in nuclei (Nu) was stained with DAPI. Labels as in (A). n = 6 for each genotype. C Co-immunostaining of Col I, Col IV and CK-8 around attached embryos in control (Lox^ff^) and Lox uKO tissue sections during embryo attachment and invasion (Days 5–6). DNA in nuclei (Nu) was stained with DAPI. Labels as in (A); bv, blood vessel. Scale bar, 100 µm. n = 3 for each genotype and time point. D In situ hybridization of Mmp9 mRNA in trophoblasts (green, arrowheads) in control and Lox uKO uterus tissue on Day 6. Sections were co-immunostained for Col IV and CK-8 and DNA in nuclei (Nu) was stained with DAPI. Top panels show Col IV, CK-8, Mmp9 staining with DAPI, while bottom ones show only Col IV and DAPI to clarify the area with breakdown of basement membrane. Labels as in (A). Scale bar, 100 µm. n = 6 for each genotype. E Representative images of hematoxylin and eosin (H&E)-stained tissue sections at implantation sites on Day 8 in control and Lox uKO uteri. Labels as in (C) except asterisks (*) indicate embryos; epc, ectoplacental cone; dec, decidua. Scale bar, 100 µm. n = 3 for each genotype. F Representative images of H&E-stained tissue sections at implantation sites on Day 12. The regions indicated by dashed lines are enlarged in the bottom panels. Red arrowheads, trophoblast giant cells. Black arrowhead, spiral artery. dec decidua, sp spongio layer, lab labyrinthine layer. Scale bar, 200 µm. n = 3 for each genotype.

We hypothesized that Mmp9 is a key factor in degrading the Col IV-containing basement membrane during implantation, as it is highly expressed by invading trophoblasts [19, 28] and the embryonic deletion of this gene affects trophoblast differentiation and invasion [29]. To test this hypothesis, we used fluorescent in situ hybridization to look for Mmp9 expression in trophoblasts at implantation sites in Lox uKO and control uteri on Day 6 of pregnancy; we found no Mmp9 expression in Lox uKO, whereas there was significant expression in the control (Fig. 7D). Histological analysis of implantation sites on Day 8 of pregnancy showed that the embryos in Lox uKO uteri lacked ectoplacental cones, resulting in abnormal embryo shapes (Fig. 7E; similar effects were observed in Mmp9-deleted embryos [29]). Moreover, similar analyses on Day 12 revealed abnormal placentation in Lox uKO uteri (Fig. 7F; like that observed in Mmp9-deleted embryos [29]). The spongio- and labyrinthin layers observed in placentas from control uteri were absent from Lox uKO, but trophoblast giant cells were seen in their place (Fig. 7F). This abnormal placentation explains the lesser weight of the uteri from Lox uKO pregnant females on Day 12 when compared to the controls, as well as the lower pregnancy rates and smaller litter sizes in Lox uKO (Fig. 4).

These data indicate that Lox facilitates trophoblast invasion during placentation by remodeling Col I in the stroma and Col IV in the basement membrane of the decidua surrounding the blastocyst.

Discussion

In this study, we aimed to understand the mechanism whereby Hif2α expressed in the hypoxic feto-maternal interface ensures successful pregnancy. Spatial transcriptomics of uterine tissue around implantation sites in Hif2α uKO mice pointed to the collagen-containing extracellular matrix as a potential factor, and we subsequently demonstrated defective Col I fibers in tissue at the implantation sites. We hypothesized the involvement of Lox, which oxidizes and cross-links Col I fibers and has an HIF-response element in its gene promoter. Indeed, Lox expression in the uterus tissue lacking Hif2α was much reduced. Moreover, Lox was highly expressed at implantation sites. Subsequently, we created a uterus-specific knockout of Lox in female mice and found they had fewer pregnancies and smaller litter sizes than normal. We characterized the implantation stages of pregnancy in these animals and discovered that Lox safeguards fertility by maintaining Col I fibril structure around implantation sites. It also provokes degradation of Col IV in the basement membrane of the luminal epithelium by inducing the metalloprotease Mmp9 in trophoblasts, promoting invasion of trophoblasts into the decidual zone, where they will eventually differentiate into a placental lineage. Our findings thus explain how hypoxia in the PDZ, by inducing Lox expression, promotes a uterine environment conducive to implantation and successful establishment of pregnancy.

We previously showed that pregnancy defects in Hif2α uKO could not be mitigated by administering either the P_4_ or leukemia inhibitory factor [10], two key molecules involved in early pregnancy [1], suggesting a previously unexplored mechanism, which we have uncovered in this study. The Lox uKO did not affect receptivity, attachment, or crypt formation—essential for early pregnancy [1, 2, 26]—but resulted in abnormal embryo invasion by failing to maintain Col I fibers in the stroma and to provoke Col IV degradation in the basement membrane. This aligns with previous reports showing that proper ECM remodeling is indispensable for blastocyst invasion and placental development [28, 29], and highlights Lox as a critical downstream effector of Hif2α in orchestrating this process.

In cancer tissues, extensive cell proliferation causes hypoxia, triggering the transcriptional activities of Hif proteins [30]. Lox, a transcriptional target of Hif proteins, is required to crosslink Col I fibers that aid tumor migration [16]. Similar parallels have been noted in invasive placentation, where mechanisms of ECM remodeling share features with tumor invasion [31]. Recent in vitro research demonstrated that human placental trophoblast cells express LOX under low-oxygen conditions to promote extravillous trophoblast lineage commitment and column formation [31]. Considering this evidence together with our observations from this study, Lox appears to be crucial for trophoblast invasion at the feto–maternal interface, and this function is potentially conserved across species. LOX is expressed in trophoblasts in humans [32], however, it is unclear whether LOX is found in the decidua as in mice. This might explain the differences in trophoblast invasiveness in the two species. Trophoblast invasion is more profound in primates than in rodents, as they penetrate deeply into maternal tissues to modify maternal spiral arteries in the decidua and myometria [33]. Clinically, shallow embryo invasion is a cause of preeclampsia and intrauterine growth restriction [33]. Thus, our findings provide a mechanistic link between hypoxia-driven LOX activity and ECM remodeling, which could inform translational approaches to diagnose or treat human pregnancy disorders characterized by impaired trophoblast invasion. Assessing LOX expression or activity may serve as a biomarker for implantation competence, while therapeutic modulation of LOX could represent a novel strategy to improve reproductive outcomes.

Evidence about the oxygen requirements for trophoblast growth is inconsistent. Hypoxia enhances trophoblast invasion and proliferation in vitro [31, 34], whereas in vivo, trophoblasts migrate toward higher oxygen levels [35]. In this study, we confirmed that Hif2α-induced Lox expression promotes trophoblast invasion into SDZs, which have more developed blood vessels than PDZs [5, 36]. This observation reconciles previous contradictory findings by suggesting that hypoxia-induced Lox provides an initial scaffold for trophoblast entry, after which trophoblasts migrate toward oxygen-rich environments to support growth and differentiation. We observed persistent hypoxia in *Lox-*deficient deciduae around embryos on day 8 (Supplementary Fig. 12). As embryo growth was similarly impaired, failure to invade high-oxygen environments may hinder proliferation and differentiation of embryonic cells.

Given the omnidirectional expression of Hif2α and Lox around the embryo, it may be surprising that mouse embryos generally implant toward the lateral side of the implantation crypt. As embryogenesis is reported to require intrauterine forces [37, 38], we hypothesize that embryos sense mechanical forces generated by uterus tissues. The nuclear concentration of Yap, a mechanosensing protein that translocates to the nucleus in response to mechanical stress [39], was lower in the cells of Lox uKO deciduae when compared with that in controls (Supplementary Fig. 13). Consistent with our hypothesis, spatial transcriptome analysis predicted high expression of Tead1 in normal PDZ cells, suggesting that Yap–Tead signaling may contribute to mechanosensing in this context. We propose that the Hif2α–Lox–ECM axis could influence intrauterine mechanical forces that facilitate embryo development. Future studies employing genetic models, such as uterine- or trophoblast-specific knockouts of Yap or Tead, will be required to test this possibility and to integrate mechanical signaling into the broader framework of implantation biology.

Despite the notable effect of Lox deletion on fertility (Supplementary Fig. 14), some females could still carry pregnancies to term, which is in contrast to the fact that Hif2α uKO females are completely infertile [10]. We speculate that this may reflect differential expression patterns of Hif2α and Lox during pregnancy: Hif2α is robustly expressed throughout the uterine lining by Day 8, whereas Lox expression becomes restricted to certain decidual regions. Hypoxia and Hif1α expression, in contrast, are not evident at this stage. Consistent with this, uterine-specific deletion of Hif1α decreases fertility but does not abolish it, suggesting that Hif2α may contribute to later stages of embryo invasion in a manner partially independent of hypoxia. However, our current Cre-driver system ablates genes before or during implantation, making it difficult to directly assess roles of Hif2αin decidua after trophoblast invasion. The development of stage- and tissue-specific, inducible Cre lines will be essential to address these questions and to explore how Hif2α may regulate uterine remodeling and pregnancy maintenance beyond the implantation phase.

Materials and methods

Mice

Hif2α (Epas1) uKO mice were established, as previously reported by our group [10]. Lox-floxed mice were prepared using the CRISPR-Cas9 system as previously reported [40]. The gRNAs to insert loxP sites into introns 2 and 4 (Supplementary Fig. 5B) were injected into fertilized eggs. After germline transmission, Lox-loxP/+ mice were intercrossed to generate Lox-loxP/loxP mice containing homozygous recombinant alleles. Lox-loxP/loxP females were then crossed with Pgr-Cre [41] males to generate mice with Lox deletions throughout the uterus (Lox uKO mice).

Evaluation of pregnancy outcomes

To examine pregnancy outcomes, Lox uKO or Lox^f/f^ (control) female mice were mated with C57BL/6 N fertile male mice as previously reported [42, 43]. The day when the vaginal plug was detected was considered Day 1 of pregnancy. These female mice were euthanized on the designated days of pregnancy to evaluate pregnancy phenotypes and collect samples. On Day 4, a uterine horn was flushed with saline to confirm the presence of a blastocyst. The Chicago blue dye solution was injected intravenously on Days 5 and 6 [44], enabling visualization of embryo attachment sites as blue bands. When no attachment sites were visible, the uterine horns were dissected and flushed to collect embryos. Uterine tissues were weighed on Days 8 and 12 to evaluate embryo growth. All mice were housed in the University of Tokyo Animal Care Facility according to the institutional guidelines for the use of laboratory animals. All animal experiments were approved by the Institutional Animal Experiment Committee of the University of Tokyo Graduate School of Medicine (Approval numbers P16-066, P20-076, and A2023M165).

PCR to check the genomic deletion of Lox

DNA collected from the uterine tissues was used as a template. One hundred nanograms of DNA was amplified using the primers 5′-AATTACTGCTGAAGCCCACC-3′ and 5′-GAATCTCCTAGAACTGACTGG-3′ and rTaq (Takara), with 35 cycles of 98 °C for 30 s, 58 °C for 30 s, and 72 °C for 30 s. The PCR products were electrophoresed on a 2% agarose gel to detect a band shift.

In situ hybridization

We performed fluorescence- or colorimetric ISH as previously reported [5, 45], with some modifications. Frozen sections (12 µm thick) were attached to Superfrost™ Plus Microscope Slides (Fisher Scientific). After fixation and acetylation, the sections were hybridized overnight at 55 °C with a DIG-labeled Lox, Bmp2, or Ptgs2 antisense probe. Following probe washing, the sections were treated with an anti-DIG antibody-POD (11207733910, Roche, 1:200) followed by Alexa 488-conjugated tyramide (B40953, Invitrogen) to detect fluorescence signals. For colorimetric ISH, anti-DIG-AP (11093274910, Roche, 1:2000) and NBT/BCIP (S3771, Promega) were used. Nuclei were counterstained with 4ʹ,6-diamidino-2-phenylindole (DAPI; Dojindo, 1:500) or Nuclear fast red (40871, Muto kagaku). Fluorescence or colorimetric images were acquired using an AXR confocal microscope (Nikon) or DM2500 (Leica). Quantification of in situ hybridization was performed using ImageJ (NIH).

Spatial transcriptome analysis

Spatial transcriptome analysis using 10× Visium (10× Genomics) was performed according to the manufacturer’s protocol. Uterine tissues collected on day 6 of pregnancy from Hif2α uKO, Lox uKO, and control mice were sectioned into 10 μm slices, mounted on gene expression slides, and sent to KOTAI Bio. Inc (Osaka, Japan). Following a 30 min Proteinase K reaction, the sections were hybridized with spatial tags on the slides and reverse-transcribed in situ. cDNAs were analyzed via RNA sequencing using DNBseq (MGI) with 300 million reads per sample.

Raw FASTQ files and microscope slide images for each sample were processed using the Space Ranger software (version 1.1, 10× Genomics) with the “spaceranger count” pipeline, which used STAR with default parameters for aligning reads against the mouse reference genome mm10 “refdata-gex-mm10-2020-A.” This pipeline used the Visium spatial barcodes to generate a feature-spot matrix of unique molecular identifier (UMI) counts. Clustering analysis was performed via Seurat (version 4.0.3), and the clusters were visualized using UMAP. DEGs between genotypes were identified with an adjusted P-value < 0.05 and a fold change >1.5. Enrichr was used to analyze gene ontologies and upstream transcription factors within each cluster, for the data from Days 4–6 of pregnant uteri were previously deposited to the NCBI Gene Expression Omnibus (GEO accession no. GSE253520).

Hematoxylin and eosin (H&E) staining and immunohistochemistry

H&E staining was performed as previously described [5]. Frozen sections (12 μm) or 10% formalin-fixed paraffin-embedded sections (6 μm) were used for day 8 and 12 tissues, respectively. Images were acquired using a DM2500 (Leica) or an Eclipse Si (Nikon).

Immunofluorescence

Frozen sections (12 μm thick) were prepared for immunofluorescence. After fixing in 4% PFA-PBS, the sections were washed with 0.1% (v/v) Tween20-PBS (PBST), and blocked by 5% (w/v) BSA-PBST for 1 h. Sections were then incubated with primary antibodies at 4 °C, O/N. Following primary antibodies were used: Ki67 (Thermo Fisher Scientific, SP6, 1:300), Pgr (Cell Signaling Technology, 8757, 1:300), Esr1 (Abcam, ab32063, 1:300), E-Cadherin (Cell Signaling Technology, 3195, 1:300), CK8 (DSHB, TROMA-I, 1:300), Collagen I (Cell Signaling Technology, 72026, 1:300), Collagen IV (Merck, AB769, 1:300), Flk1 (BD Pharmingen, 550549, 1:300), ZO-1 (Santa Cruz, sc-8147, 1:300), ZO-1 (eBioscience, 14-9776-82, 1:300) and Yap (Cell Signaling Technology, 14074, 1:300). Signals were detected using secondary antibodies conjugated with Alexa Fluor 594 (anti-rabbit immunoglobulin G; Thermo Fisher Scientific, A21428, 1:500), Alexa Fluor 488 (anti-rat immunoglobulin G; Thermo Fisher Scientific, A11006, 1:500), Alexa Fluor 594 (AffiniPure Donkey Anti-Rabbit IgG; Jackson, 711-585-152, 1:500), Alexa Fluor 488 (AffiniPure Donkey Anti-Goat IgG; Jackson, 705-545-003, 1:500), and Alexa Fluor 647 (AffiniPure Donkey Anti-Rat IgG; Jackson, 712-605-153, 1:500). Sections were incubated with secondary antibodies and DAPI (Dojindo, 2 µg/mL final) for 1 hr, RT. Hypoxic regions were identified by treating mice with Hypoxyprobe-1^TM^ (Hypoxyprobe, Inc.) following slight modifications to a previously described protocol [7]. Briefly, pregnant females were intraperitoneally administered 60 mg/kg of Hypoxyprobe-1 in saline on day 6 or 8 and euthanized 45 min later. Snap-frozen uteri were sectioned and incubated with an FITC-conjugated anti-hypoxyprobe antibody. Images were captured using an AXR microscope (Nikon).

Three-dimensional visualization of the implantation sites

Three-dimensional visualization of Day 5 and 6 implantation sites was performed according to a previously reported protocol [26]. On Days 5 and 6, tissues were incubated with an anti-E-cadherin antibody (Cell Signaling Technology, 24E10, 1:500) to label luminal and glandular epithelial cells and subsequently with an Alexa 555-conjugated anti-rabbit antibody (Thermo Fisher Scientific, A21428, 1:500). Images were captured using an AXR microscope (Nikon). The 3D structures were constructed using the surface tool in Imaris (v9.8, Oxford Instruments). Luminal and glandular epithelial structures were segmented from E-cadherin-stained 3D images using Imaris, according to previous studies [26, 27].

Transmission electron microscopy

TEM was performed as described previously [10]. Briefly, uterine tissues were collected after sequential perfusion with saline and TEM fixative (2% glutaraldehyde/2% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4). Tissues were then sliced into individual implantation sites and fixed with TEM fixative. The fixed tissues were sent to the Hanaichi UltraStructure Research Institute (Okazaki, Japan) for embedding in resin, ultrathin sectioning, and TEM analysis.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 10. Data normality was assessed using the Shapiro–Wilk test in GraphPad Prism. The equality of variances was tested using F-test before performing Student’s t tests. A two-tailed Student’s t-test or Fisher’s exact test was used to determine significance at P < 0.05.

Ethics approval and consent to participate

All methods were performed in accordance with the relevant guidelines and regulations. All animal experiments were approved by the Institutional Animal Experiment Committee of the University of Tokyo Graduate School of Medicine (Approval numbers P16-066, P20-076, and A2023M165).

Supplementary information

Supplementary Figures and Legends Original data Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Table 4 Supplementary Table 5 Supplementary Table 6 Supplementary Table 7