Introduction

Infertility is a social concern that affects 17.5% of the adult population worldwide¹. Some patients who undergo in vitro fertilization and embryo transfer (IVF_-ET) repeatedly fail to become pregnant, even after ET using high-quality embryos (2). Embryo implantation, which is the starting point of pregnancy, can be divided into several processes, including blastocyst spacing, apposition, attachment to the uterine luminal epithelium, and invasion into the endometrial stroma^{2, 3}. Establishing implantation requires an exquisite interaction between the endometrium and embryo; however, the detailed underlying mechanism remains unclear.

Based on the similarities in the influence of female sex hormones on the endometrium during pregnancy, rodents have been utilized as an in vivo model of human pregnancy. In mice, day 1 of pregnancy is defined by the observation of a vaginal plug. After coitus, high serum levels of estrogen (E_₂) induce epithelial proliferation and P_₄ production from the ovary increases on day 3. Embryos in the oviduct reach the uterus on day 4 in a P_₄-dominant hormonal environment. In the uterus, proliferation-differentiation switching (PDS), indicating the inhibition of epithelial proliferation with induction of stromal proliferation the endometrium, is evident on day 4 under the continuous influence of P_₄, resulting in the endometrium acquiring implantation potential^{4, 5}. With this dynamic change, the morphology of the endometrial luminal epithelium reveals a slit-like narrowing, known as the formation of a slit-like luminal structure^{4, 6, 7} (Fig. 1a). Late on the morning of day 4, blastocysts are activated by a small estrogen surge that occurs as the starting signal for implantation^2,8. The blastocyst finally attaches to the luminal epithelium on day 4 midnight. The luminal epithelium initiates the formation of a uterine crypt after attachment and the embryo can be observed at the bottom of the crypt⁹. Stimulation from the embryo attachment seems to be transmitted from the endometrial epithelium to the stroma, where vascular permeability is increased (attachment reaction), and the surrounding stromal cells initiate differentiation into decidual cells (decidualization). By the evening of day 5, the endometrial luminal epithelium facing the attached blastocysts disappears, and trophoblast invasion is initiated.

Fig. 1

In recent studies, our group and others have demonstrated that a series of luminal changes, that is PDS, formation of a slit-like luminal structure, and crypt formation, are crucial for embryo implantation^{4, 5, 6, 7, 9, 10, 11}. For PDS and formation of the slit-like luminal structure, the P -progesterone receptor (PGR) pathway is involved^{4, 5, 6}. We previously reported that mice with epithelial deletion of Pgr (Pgr^fl/fl;Ltf^Cre/+ mice; Pgr eKO) showed sustained epithelial cell differentiation, resulting in defective uterine receptivity⁶. As for embryo attachment and subsequent crypt formation, the Lif-Stat3 pathway is crucial, in addition to P₄-Pgr signaling. Leukemia inhibitory factor (Lif) activates uterine Lif receptors (Lifrs) to evoke Stat3-mediated gene transcription, thus initiating the formation of implantation chambers (crypts)^{9, 11, 12, 13, 14}. We have also reported that uterine-specific retinoblastoma knockout mice (Rb uKO) and enhancer zeste homolog 2 knockout mice (Ezh2 uKO) show impaired PDS because of cell cycle abnormalities in the endometrium^{15, 16}. Notably, in these uKO models, although epithelial proliferation was continuously observed even on day 4, embryo attachment occurred but subsequent embryo invasion was flawed. These differential phenotypes of uterine-specific gene deletions suggest that essential mechanisms other than PDS are involved in P₄- and Lif-induced embryo attachment and subsequent pregnancy maintenance.

Accumulating evidence has demonstrated that PDS, slit-like formation, and crypt formation are important changes in the luminal epithelium for embryo attachment. However, how the overall changes in the luminal epithelia are dynamically regulated during the short period of embryo implantation remains unclear. It appears that PDS, slit formation, and crypt shaping occur sequentially; however, how each step influences the others remains uncertain. In particular, crypt formation occurs only in the presence of embryos¹⁰, indicating that epithelial changes require reciprocal interaction with embryos. To understand the molecular mechanism underlying these epithelial changes, we especially examined p38α, a MAP kinase activated by phosphorylation in response to various environmental stresses and inflammatory cytokines, which regulates fundamental cellular processes such as proliferation, apoptosis, cell differentiation^{17, 18, 19}. Notably, p38α plays an important role in development and tissue differentiation by modulating the localization of E-cadherin, which is expressed in tissue epithelial cells, thus altering epithelial morphology^{20, 21}. Embryo implantation requires cell differentiation and death as well as glandular epithelial development and secretion, which may be influenced by p38α because it regulates various cell differentiation processes including mammary gland duct formation^{22, 23, 24}. Recently, deletion of p38α was reported to result in complete pregnancy failure in mice²⁵, supporting our notion of a p38α-regulated mechanism shaping the endometrial epithelium during the peri-implantation period.

In this study, we first investigated the spatiotemporal changes of luminal shapes in 3D spanning the period from just after the coitus to the peri-attachment stage using a time course analysis. Our analyses revealed that the surface of the endometrial lumen not only became flattened but also showed creases, resulting in the even zoning of embryos before attachment. The embryos were then attached to the flattened area and crypts were formed. This epithelial shaping was impaired in the mice with uterine-specific knockout (uKO) of p38α, giving rise to failed embryo attachment. While treatment with P₄ and Lif restored embryo attachment in p38α uKO mice, embryo invasion and subsequent pregnancy maintenance remained disturbed because of abnormal epithelial shaping, which was not rescued. In summary, we discovered that dynamic morphological changes in the endometrial lumen prior to implantation may influence the process of pregnancy establishment and maintenance, which is regulated by previously unknown mechanisms independent of P₄-Pgr and Lif-Stat3.

Results

The morphology of the endometrial luminal surface dynamically changes before embryo attachment

To identify the spatiotemporal changes in the luminal epithelium prior to embryo attachment, we adopted 3D visualization, which recently revealed the detailed topography of the endometrial epithelium and the mechanism of embryo implantation¹⁰. Blood vessels enter the uterus from the mesometrium, situating the uterus along the mesometrial–anti-mesometrial (M–AM) axis. Once blastocysts attach to the surrounding luminal cells, an implantation chamber (crypt) is formed by luminal epithelial (LE) evaginations toward the AM pole^{10, 26, 27} (Fig. 1a, b). Although the morphological changes in the endometrial lumen after embryo attachment have been reported^{10, 28}, little is known regarding the luminal morphology prior to attachment. How the epithelial dynamics prior to embryo attachment influence subsequent pregnancy processes also remains unclear.

Therefore, we analyzed the luminal morphology in wild-type mice on the mornings of days 1–4 of pregnancy. On day 1, the endometrial lumen extended in a disorderly manner against the M-AM plane. During days 2–4, the lumen became flattened with some folding in the M-AM axis that remained even (Fig. 1c). A cross-sectional view of the tissues revealed that the lumen became more slit-like with epithelial mass decreasing daily prior to embryo attachment (Fig. 1c, d), which was consistent with the results of luminal changes previously depicted using conventional histology^{4, 29}.

Next, we investigated the relationship between luminal folding and blastocyst movement by observing day 4 endometria from the morning to midnight, when blastocysts arrived and attached to the uterine lumen, respectively. We observed the 3D morphology of the uterine lumen and positions of the blastocysts on day 4 morning (day 4 10:00), day 4 evening (day 4 16:00), day 4 evening (day 4 20:00), day 4 midnight (day 5 0:00), and day 5 morning (day 5 10:00). Once embryo attachment occurred, vascular permeability increased in the surrounding endometrium, which could be observed by the injection of blue dye into mice (blue dye reaction)³⁰. The blue dye reactions were 0% (0/5) at 16:00 on day 4, 18.8% (3/16) at 20:00 on day 4, 66.7% (4/6) at 0:00 on day 5, and 100% (7/7) at 10:00 on day 5 (Fig. 2a-c). Embryo attachment was completed in more than half of the individuals at midnight on day 4; therefore, we defined the evaluation time immediately before embryo attachment as day 4 20:00. As previously described, the surface of the endometrial lumen was flattened at day 4 10:00. On Day 4, from 16:00 to 20:00, the flat lumen exhibited alternating shrunken and stretched areas. Shrunken areas with multiple folds were observed at regular intervals. No regularity in the position of the embryos was observed until day 4 16:00; however, at day 4 20:00, the embryos entered the shrunken areas and gradually moved to the stretched areas over the time course. At day 4 midnight, the embryos were attached to the AM pole of the lumen in the stretched areas, and crypts were then formed on day 5 morning (day 5 10:00) (Fig. 2d). These results demonstrate that the uterus just prior to embryo attachment showed dynamic changes in luminal morphology once the embryos arrived.

Fig. 2

Uterine p38α activation is crucial for luminal morphology and the subsequent embryo attachment

We then examined how luminal dynamics before attachment were regulated. We performed single-cell RNA-seq (scRNA-seq) analysis for days 4 and 5 uteri (Fig. 3). Uterine tissues contain multiple cell types, including epithelial, stromal, vascular endothelial (VE), and immune cells (Fig. 3a, Supplementary Fig. 1a, Table S1 and S2). Among them, luminal cell clusters (LE) can be divided into two types: conventional luminal cells (LE) which are highly observed on day 4, and activated LE (LE_activated), whose population increases on day 5 (Fig. 3a). To determine the characteristics of the LE_activated cells, we performed enrichment analyses using Enrichr³¹ (Fig. 3b, Table S2). Transcription factor protein-protein interactions (PPIs) revealed enrichment of ATF2 and JUNB, which are related to stress signalling³². In agreement with this result, a pathway analysis using MSigDB HallMark also revealed TNFα-related and hypoxia signaling. Furthermore, gene ontology analysis using Metascape³³ identified pathways related to oxidative stress and cell motility (Fig. 3c, Table S2). We then investigated stromal cell types in the same manner (Fig. 3d, Supplementary Fig. 1b, and Table S3). These were clustered into five types: non-proliferative (Non-pro), proliferative (Pro), sub-epithelial (Sub-epi), sub-endothelial (Sub-endo), and attached (Attached). Because the day 5 stroma exclusively contained Attached cluster, we analyzed the enriched pathways in this cell type. Similar to LE_ activated, this cluster was enriched in stress-related transcriptional factors, signals and Gene Ontology (GO) terms (Fig. 3e, f, and Table S4).

Fig. 3

As a possible regulator of day 5-specific LE and stromal cell types, we focused on p38α, a Map kinase protein. p38α is phosphorylated for activation in response to various kinds of cellular stimuli¹⁸. Notably, phosphorylated p38α (pp38α) translocates into nuclei to activate ATF2-JUN-induced transcription of cytokines, including TNFα³². Our immunostaining revealed that p38α was highly expressed and activated by phosphorylation in the pre-attachment luminal epithelia (Fig. 3g), indicating the role of p38α in the luminal epithelium before embryo attachment. After attachment, pp38α was expressed both in the endometrial epithelium and stroma, especially around the attached embryos on days 6 and 8.

To investigate the roles of uterine p38α, we established mice with uterine-specific deletion of p38α (p38α uKO) by crossing p38α-floxed mice with those carrying a Pgr-Cre driver. Efficient deletion and inactivation of p38α protein in p38α uKO was confirmed by immunostaining for p38α and pp38α (Fig. 4a). To examine the reproductive phenotypes of p38α uKO and littermate p38α-floxed females (p38α Ctrl), we mated them with fertile wild-type (WT) male mice. p38α uKO mice showed complete infertility (Fig. 4b). As we detected corresponding numbers of hatched blastocysts by flushing the uterine cavity with saline solution on day 4 morning in each genotype (Fig. 4c), uterine p38α did not influence embryo development before attachment. We then intravenously injected Chicago blue dye solution to examine the implantation sites on day 5 of pregnancy. However, p38α uKO uteri showed no attachment sites on day 5 of pregnancy, and blastocysts were recovered by saline flushing of the uterine cavities (Fig. 4d), which indicated flawed embryo attachment. On day 6 of pregnancy, the number of embryo attachment sites was significantly reduced in p38α uKO compared with those in the p38α uCtrl (Fig. 4e). Further, our 3D imaging on day 5 morning revealed failed crypt formation in p38α uKO, suggesting that p38α uKO mice show infertility because of embryo attachment failure (Fig. 4f).

Fig. 4

Our observations indicating luminal activation of p38α in the pre-attachment period (Fig. 3g) as well as infertility with flawed embryo attachment in p38α uKO females (Fig. 4) motivated us to investigate the involvement of this molecule in luminal dynamics. We then compared the morphological changes in the lumen of p38α uCtrl and p38α uKO from days 1 to 4 using 3D imaging (Fig. 4g). Similar to the observation in the wild type mice, p38α uCtrl showed that the folding of the endometrial lumen in the M-AM axis gradually disappeared but some folding remained evenly. Accordingly, the endometrial lumen flattened as the pregnancy progressed from days 1 to 4. In contrast, in p38α uKO, folding in the both cross-sectional longitudinal axes remained even on day 4, which appeared as serrated luminal shapes in the 2D view. Considering that embryo attachment failed in p38α uKO, the luminal changes in the pre-attachment phase could influence successful embryo attachment.

Supplementation with P₄ and Lif, two major factors supporting embryo implantation, rescues the flawed embryo attachment, but not the subsequent pregnancy maintenance in p38α uKO

Our group and others have previously revealed that PDS, wherein epithelial cell proliferation is terminated before implantation, is an indicator of endometrial embryo receptivity⁵. P is an inducer of PDS as well as slit formation in the endometrial lumen^{4, 6}. Immunostaining for Ki67, a cell proliferation marker, showed an increased number of proliferating cells in the luminal epithelium in p38α uKO on day 4 morning, suggesting that PDS was impaired in this milieu (Fig. 5a, b). This result motivated us to examine whether P_₄ supplementation could rescue the phenotype of p38α uKO. We thus treated p38α uKO with P_₄ in the preimplantation period and examined the resulting morphological changes in the endometrial lumen using 3D analysis. On days 3 and 4, P₄ treatment suppressed folding in both the cross-sectional and longitudinal axes, thus flattening the lumen (Fig. 5c-e). Further, PDS was also rescued by P₄ injection to p38α uKO (Fig. 5f, g). Notably, p38α uKO showed a normal hormone-producing capacity of the ovary, as serum estradiol-17β (E_₂) and progesterone (P_₄) levels were comparable (Supplementary Fig. 2a). We also confirmed that the expression of estrogen receptor (ERα) and progesterone receptor (PGR) in the uterus were not influenced by p38α deletion (Supplementary Fig. 2b). Although luminal morphology was improved by P₄ treatment, embryo attachment was still failed in the mutant (Fig. 5h), suggesting that some other factor is required for p38α-dependent embryo attachment.

Fig. 5

Besides the P_₄-PGR pathway, leukemia inhibitory factor (Lif) is a critical factor for embryo attachment^{9, 11, 34}. Lif is an interleukin-6 family member secreted by the endometrial gland on the day 4 morning^{34, 35, 36}. In situ hybridization showed significant decreases of Lif in the p38α uKO uterus on day 4 morning even after P₄ treatment (Fig. 6a). Further, activation of Stat3, a transcription factor downstream of Lif, was also downregulated as determined by immunostaining for phosphorylated Stat3 in p38α uKO uterus on day 4 morning (Fig. 6b). Based on these results, we examined whether activation of Lif-Stat3 can restore the flawed embryo attachment in p38α uKO.

Fig. 6

These results prompted us to determine whether supplementation of Lif along with P_₄ can ameliorate abnormal implantation in p38α uKO. Five groups were established: vehicle administration to p38α uCtrl and p38α uKO, P_₄ alone to p38α uKO (at 10:00 on day 2–day 4), Lif alone to p38α uKO (at 9:00 and 18:00 on day 4), and both P_₄ and Lif to p38α uKO (Fig. 6c). The number of embryo attachment sites in the single treatment of either P_₄ or Lif did not differ compared with that in the vehicle group. In contrast, simultaneous supplementation with P_₄ and Lif increased the number of embryo attachment sites in p38α uKO, which was comparable to that in the p38α uCtrl group as shown by the clear blue reactions on day 5 morning, indicating that both P_₄ and Lif are required for p38α-dependent embryo attachment (Fig. 6d, e).

COX2 (Fig. 6f) and phosphorylated Stat3 (Supplementary Fig. 3), which are induced at embryo attachment sites¹¹, were found to be expressed in p38α uKO upon P_₄ and Lif treatment, also supporting our notion. However, despite embryo attachment, P₄ and Lif treatment could not recover full term pregnancy with litters in p38α uKO (Fig. 6g). On day 8, although the number of implantation sites was comparable between the two groups, none of the implantation sites underwent normal embryogenesis and formed hematomas in p38α uKO treated with P_₄ and Lif (Supplementary Fig. 4), indicating that pregnancy maintenance did not occur normally. Eventually, P₄ and rLif-treated p38α uKO females failed to give birth, accompanied by severe embryo resorption (Fig. 6g). This suggests that p38α is required for P₄- and Lif-induced embryo attachment as well as healthy pregnancy maintenance.

p38α plays an important role in establishing the luminal epithelial morphology for embryo attachment, and changes by day 4 morning are an important scaffold for embryo attachment site formation

Flawed pregnancy maintenance in the mutant even after treatment led us to examine how pregnancy events after the embryo attachment were disturbed in p38α-deleted uteri. We thus investigated the luminal morphologies on days 5 and 6 when embryo attachment and invasion were evident (Fig. 7a). We found that folding was evident in the longitudinal axis in p38α uKO regardless of any treatment.

Fig. 7

Simultaneous treatment with P₄ and Lif created a crypt on day 6, but still longitudinal folding was observed around the crypt. Further, crypt formation was poorly initiated on day 5 morning in P₄ and Lif-treated p38α uKO showing an obvious longitudinal folding. Considering that this folding should be eliminated by day 4 night in normal pregnancy (Fig. 2), we investigated the luminal morphology at day 4 20:00, just before embryo attachment (Fig. 7b). Similar to the preparatory phase of embryo attachment on day 4 morning (10:00 am), the surface of the stretched lumen, i.e., the site of embryo attachment, was flat in p38α uCtrl. In contrast, p38α uKO and p38α uKO+Lif (P_₄ non-treated group) showed remarkable persistence of luminal folding in the longitudinal axis with a poorly established stretched area. In p38α uKO+P_₄ and p38α uKO+Lif+P_₄ (P_₄ treated group,) the flattening and stretching of the lumen was partially rescued but longitudinal folding remained (Fig. 7b, left). We also observed the luminal shapes in cross-sectional views to examine the relationship between the shape of the slit-like lumen and embryo location (Fig. 7b, right, c, and d). In the p38α uCtrl, the stretched luminal areas serving as embryo attachment sites, were flat and slit-like, predicting smooth guidance of the embryo to the AM pole (Fig. 7b, right). In contrast, regardless of the treatment, flattening of the lumen failed in the p38α uKO, accompanied by strayed positioning of embryos probably because of failure of M-AM axis formation. These abnormal morphologies of the luminal epithelia appeared as an increased epithelial mass (Fig. 7c) and increased epithelial branching in the uKO (Fig. 7d). Overall, these results suggest that p38α is responsible for morphological changes in the lumen before embryo attachment, which P_₄, but not Lif, partially assists in. Considering that P₄ could not solely restore embryo attachment in p38α uKO, Lif seems to act as an inducer of attachment under the influence of P₄.

p38α is an important signal transducer between luminal epithelia and stroma for embryo attachment

We then examined how p38α influences feto-maternal interactions to accomplish healthy embryo implantation and luminal dynamics. Notably, we found that epithelial-specific deletion of p38α did not critically alter female fertility (Supplementary Fig. S5), indicating that p38α regulates epithelial-stromal crosstalk. To obtain this information, we performed scRNA-seq in Ctrl uteri on day 4 2000 h (just before attachment) and 2400 h (at attachment), as well as in uKO mice with or without P₄ + rLif treatment on day 4 2400 h (Fig. 8a, Table S5). Similar to Fig. 3a, we found multiple cell types in the uteri, including epithelial, stromal, myometrial, and immune cells. Among these, we focused on stromal cells as their population increased during the embryo attachment process in the uCtrl (Fig. 8a). The stromal cells were further clustered into non-proliferative (Non-prolif), proliferative (Pro), decidualizing (Dec), sub-epithelial (Sub-epi), sub-endothelial (Sub-endo), and uKO-specific clusters (Fig. 8b, Table S6). The uKO-specific cluster was highly enriched in uKO uteri. We then traced stromal differentiation using pseudo-time analysis (Fig. 8c and d). Intriguingly, the uKO-specific cluster was poorly differentiated compared with other stromal cell types, except for the Non-prolif cluster (Fig. 8d). We then examined the signal transduction between the epithelial (LE and GE) and stromal clusters, focusing on secretory molecules that can bridge the epithelial and stromal compartments (Fig. 8e). In uCtrl, non-canonical Wnt (ncWnt) from the LE and stroma and IGF from the stroma affected the decidualizing and luminal cells, which were enhanced upon embryo attachment (Fig. 8e, upper). In contrast, in p38α-deficient uteri, only epithelial cells strongly sent and received signals (Fig. 8e, lower left), which remained even with P₄ and rLif treatment (Fig. 8e, lower right). These results demonstrate flawed epithelial-stromal communication in the p38α uKO milieu.

Fig. 8

Among ncWNTs, Wnt5a plays a critical role in early pregnancy^{27, 28}. Wnt5a activates receptor tyrosine kinases Ror1 and Ror2 in the uterus Both overexpression and deletion of Wnt5a compromises pregnancy outcomes owing to sustained apicobasal polarity in the luminal epithelia²⁷. Similar to Wnt5a, Igf1 is also involved in epithelial depolarity and is highly expressed in day 4 uterine stroma, activating Igf1r and downstream Stat3 in the luminal epithelia³⁷. Uterine-specific deletion of Igf1r results in flawed embryo attachment because of sustained epithelial polarity³⁷. These contexts prompted us to examine these two molecules in the p38α uKO milieu. We first compared Wnt5a and Igf1 expression using scRNA-seq data and found significantly upregulated Wnt5a and downregulated Igf1 in both the luminal epithelia (LE) and decidualizing stroma (Dec) (Fig. 9a and b). In agreement with these results, we found stronger staining of epithelial markers, β-catenin and E-cadherin, in day 4 uKO uteri (Fig. 9c). We then asked whether embryo invasion was influenced by p38α deletion. Our group has previously shown that epithelial removal after the loss of epithelial polarity facilitates embryo invasion and the subsequent pregnancy processes^{15, 38}. To assess embryo invasion, the implantation sites were stained for cytokeratin 8 (CK8), a marker of epithelial and trophoblastic cells. In p38α uCtrl, the endometrial luminal epithelium disappeared and trophoblasts invaded the uterine stroma on the morning of day 6, whereas the endometrial luminal epithelium around the embryo remained and embryo invasion was incomplete in p38α uKO even after treatment with P₄ and Lif (Fig. 9d).

Fig. 9

We also examined the spatial transcriptome of the embryo attachment site on the morning of day 5 from Ctrl- and P₄ + rLif-treated uKO mice to observe epithelial-stromal interactions (Fig. 9e-h). Cross-sectional tissues of each implantation site were clustered into six types – LE, embryo attached stroma (Str_attached), proliferative stroma (Str_prolif), uKO-specific Str (Str_uKO_specific), GE, and myometria (Myo) (Fig. 9e, f, Table S7). Notably, uKO tissues solely contained Str_uKO_specific as the stromal cluster. To understand the characteristics of this cell type, pathway analyses were performed using Metascape. We found that the upregulated genes in Str_uKO_specific were enriched in pathways such as “regulation of inflammatory response,” “negative regulation of cell differentiation,” and “negative regulation of cell population proliferation” (Fig. 9g, Table S8)

Down-regulated genes were related to “Extracellular matrix organization,” “negative regulation of canonical Wnt signaling pathway,”, and “regulation of cellular response to growth factor stimuli” (Fig. 9h, Table S8), which are known to be associated with healthy pregnancy outcomes^{36, 39}. These results indicate that embryo attachment to the uKO luminal epithelia induces inflammatory signals rather than physiological reactions in the stroma, compromising subsequent pregnancy maintenance. In summary, our data demonstrate the previously unappreciated role of uterine p38α as crucial for appropriate embryo attachment independently of the P₄ and Lif pathways.

Discussion

Communication between embryos and endometria is crucial for the successful establishment of pregnancy⁴⁰. In preparation for blastocyst attachment, the endometrium undergoes massive morphological changes, especially in the epithelial layers: (1) PDS before embryo attachment, (2) slit formation of the endometrial lumen before embryo attachment, and (3) crypt formation of the endometrial luminal epithelium after embryo attachment. In this study, we conducted 3D histological analysis as well as single cell and spatial transcriptome analyses to elucidate the details of dynamic morphological changes in the endometrial lumen. Two molecular pathways, the P_₄-PGR and Lif-Stat3 pathways, have been considered important for embryo attachment so far. Further, we demonstrated a previously unappreciated role of p38α in addition to the above two major pathways (Fig. 10).

Fig. 10

The major finding of this study was that luminal shapes were altered throughout the uterine horns during early pregnancy. In particular, changes were evident from the morning of day 4 to midnight. This may explain the mechanism of embryo attachment as well as embryo spacing: a previous study demonstrated that embryo spacing occurs during day 4 noon to evening, which was compromised in mice with systemic KO of Lpar3, a lipid receptor expressed in the luminal epithelia⁴¹. Indeed, we observed that epithelial folding in the M-AM axis gathered evenly, corresponding to the position of the embryos from morning to evening on day 4, indicating that luminal layer movement contributes to embryo spacing in addition to myometrial, as previously reported⁴². This 3D imaging experiment had some limitations. Because the data were obtained from euthanized mice, following changes over time in the same individual was impossible, and the influence of phenotypic time differences among individuals could not be completely ruled out. To examine the epithelial morphology and phenotype over time in the same individual, experimental system using techniques such as live imaging in vivo needs to be established.

Considering the functional feature of p38α as a Map kinase, the outer- or inner-cellular stimuli evoking p38α-dependent epithelial shaping remain unclear. Map kinases can be activated downstream of various receptors, including tyrosine receptor kinases and G protein-coupled receptors¹⁸. One candidate is the Ror1/Ror2 tyrosine receptor kinase, which can be activated by Wnt5a²⁷. Uterine-specific deletion of Wnt5a-Ror1/Ror2 axis resulted in defective embryo implantation because of abnormal luminal morphology, similar to our observation in p38α uKO. Similarly, deletion of Igf1-Igfr signaling causes poor embryo attachment, accompanied by abnormal luminal integrity³⁷. These contexts may explain why Wnt5a and Igf1 were upregulated in p38α uKO uteri.

New insights into the morphological changes in the endometrial lumen may provide an innovative approach for treating embryo attachment failure, focusing on the endometrial lumen morphology. In humans, healthy implantation occurs at the fundus of the uterus⁴³, which is different from that in rodents with turbinal uterine structures. However, epithelial integrity is a common feature that regulates appropriate embryo implantation across species; for instance, after ovulation in humans, epithelial-mesenchymal transition with reduced epithelial polarity occurs during menstruation and implantation, thus influencing implantation outcomes^{44, 45, 46}. A previous study using human endometrial epithelial cell lines demonstrated that deletion of p38α influenced the cellular transcriptome and metabolome, contributing to cancer cell-like characters⁴⁷, suggesting the role of p38α in maintaining luminal epithelial integrity in humans. In addition to specific molecular mechanisms, implantation failure may also be overcome from the mechanical aspect of tissue mobility for pathological conditions such as uterine myoma and uterine adenomyosis, wherein abnormal morphological changes in the endometrial lumen are presumed⁴⁸. These findings are expected to have broad applications in diagnosing and treating human implantation failure, including the search for biomarkers of implantation ability and supplementation with relevant molecules. As reported previously, p38α also plays critical roles in mammary gland lumen formation²²; therefore, the mechanism discovered in this study could be applicable in other epithelial systems as well.

Methods

Mice

WT (C57BL/6N, SLC), p38α-floxed (Mapk14-floxed; kindly provided by Dr. Kinya Otsu, University of Osaka)¹⁹, Pgr-Cre⁴⁹, were used in this study. Pgr-Cre is expressed throughout the uterine layers⁴⁹. Mice with p38α deletion in all uterine layers (Mapk14^flox/flox Pgr^Cre/+; p38α uKO) were generated by crossing Pgr-Cre with p38α-floxed mice. Cre-negative littermates (Mapk14^flox/flox) served as controls. All mice used in this study were housed at the University of Tokyo Animal Care Facility, following the institutional guidelines for using laboratory animals.

Evaluation of pregnancy outcomes

To examine the pregnancy outcomes, p38α-uKO, or p38α-floxed (control) female mice were mated with C57BL/6N fertile male mice, as reported in a previous study^{15, 38, 50}. The day of vaginal plug detection was considered day 1 of pregnancy. Pregnant mice were euthanized by cervical dislocation on the designated day of pregnancy to evaluate pregnancy phenotypes and for sample collection. On days 2 and 3, both sides of the oviducts were flushed with saline to confirm the presence of 2-cell embryos on day 2 and 8-cell embryos or morula embryos on day 3. On day four, one uterine horn was flushed with saline to confirm the presence of blastocysts. Embryo attachment sites were observed as blue bands soon after intravenous injection of a 1% solution of Chicago blue dye (Sigma-Aldrich) in saline on days 5 and 6. When no embryo attachment sites were observed as of day 5, both uterine horns were cut and flushed with saline to collect the embryos.

To analyze implantation failure, pregnant mice were sacrificed on day 8 at 1000 h, and implantation sites were histologically assessed. If obvious hematopoietic cell infiltration was observed, implantation failure was diagnosed. Parturition events were monitored daily from days 19 to 22; all mice were dissected, and the abdominal cavity was observed for findings of miscarriage.

To evaluate the phenotype of embryo attachment in wild-type mice over time, we observed the uteri on day 4 morning (day 4 10:00), day 4 evening (day 4 16:00), day 4 night (day 4 20:00), day 4 midnight (day 5 0:00), and day 5 morning (day 5 10:00). For day 5 night, the observation time was defined as 18:00–22:00. In preliminary experiments, we confirmed that the phenotypes were equivalent, with no significant changes in the morphology of tissues collected during these 4 h. As described above, samples with visible embryo attachment sites were evaluated for their numbers, and those with no visible embryo attachment sites were evaluated as pregnant specimens by flushing the contralateral uterus with saline and observing the blastocysts. Daily subcutaneous injections of P_₄ (2 mg/mouse/day) to p38α uKO mice were performed from day 2 of pregnancy or from the criterion day of pregnancy at 10:00 as previously described¹⁵.

rLif injections were performed as previously reported¹¹; female mice received rLif (20 µg/head, i.p.) at 9:00 and 18:00 on day 4 of pregnancy. The rLif expression vector was a kind gift from Prof. Eichi Hondo¹³.

Transmission electron microscopy (TEM)

TEM was performed on mouse uterine specimens collected on day 4 at 10:00 h. The fixation solution was 2% glutaraldehyde-2% paraformaldehyde dissolved in 0.1 M phosphate buffer (pH 7.4). The mice underwent the following perfusion procedures: deep anesthesia was administered, the mouse was fixed in place, and the abdomen to chest was incised; the diaphragm was quickly incised with tweezers and the heart was exposed; next, a 26G needle was inserted into left ventricle, and saline solution was injected; then, the saline was switched with the aforementioned fixative solution, which was injected to ensure fixative solution spreading to the tissues. Finally, the specimens were kept refrigerated in fixative solution. Specimen processing and imaging were performed at the Hanaichi Institute of Chemical Microscopy after embedding.

RNA extraction and real time-quantitative PCR (RT-qPCR)

RNA was prepared from homogenized frozen tissues, as previously described¹⁶. qRT-PCR was performed using THUNDERBIRD SYBR qPCR Mix (TOYOBO). The housekeeping gene Actb was used for internal standardization of mRNA expression. Relative expression levels were determined using the ΔΔCt method⁵¹. The following primers were used.

H&E staining and immunostaining

H&E staining and immunostaining of uterine tissues were performed using paraffin-embedded sections (6 μm) or frozen sections (12 μm) as previously described¹⁵. For immunohistochemistry, the sections were incubated overnight with primary antibodies including p38 (8690, Cell Signaling Technology, 1:800), pp38 (4511, Cell Signaling Technology, 1:800), Esr1 (ab32063, Abcam, 1:200), Pgr (ab63605, Abcam, 1:100), pStat3 (ab76315, Abcam, 1:100), FOXA2 (8186, Cell Signaling Technology, 1:200), and COX2 (AA570-598, Cayman,1:200). For immunohistochemistry, signals were detected using a DAB substrate kit (#425011, Nichirei) after incubation with horseradish peroxidase-conjugated secondary antibodies (K4003, Dako). The images were captured using the Leica DM5000 B light microscope.

For immunofluorescence analysis of the paraffin sections, the sections were incubated overnight with primary antibodies, including CK8 (DSHB, 1:500), and signals were detected using Alexa Fluor 488-conjugated anti-rat immunoglobulin G (Thermo Fisher Scientific, A11006,1:500); nuclei were stained with 6-diamidino-2-phenylindole (DAPI) (Dojindo, 1:500). For immunofluorescence analysis of frozen sections, sections were incubated overnight with primary antibodies, including Ki67 (20701, Cell Signaling Technology, 1:200, Alexa Fluor® 555 Conjugate), Ecad (3199, Cell Signaling Technology,1:200, Alexa Fluor® 488 Conjugate), and β catenin (83539, Cell Signaling Technology, 1:200, Alexa Fluor® 555 Conjugate). Nuclei were detected using 6-diamidino-2-phenylindole DAPI (1:500). Images were captured using an AXR microscope (Nikon).

Automated western blots with simple western (WES)

Proteins were extracted from cryopreserved and homogenized day 4 uterine tissues using RIPA buffer (Sigma) supplemented with a proteinase inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail (Sigma).

Equal amounts of protein (2 µg/µl) were loaded into 12–230 kDa separation module kit and analyzed using the Protein Simple Wes® System (Protein Simple, San Jose, CA, USA) following the manufacturer’s instructions. The antibodies used included Actin (C-11, sc-1615, Santa Cruz, 1:2000), Stat3 (4904, Cell Signaling Technology, 1:2000), and pStat3 (9145, Cell Signaling Technology, 1:2000). Anti-goat IgG and anti-rabbit IgG antibodies were used as secondary antibodies. Actin served as the loading control.

3D visualization of uterine endometrial luminal epithelium

3D visualization of the day 1–4 uteri or day 5 and 6 implantation sites was performed as previously reported¹⁰. To stain luminal and glandular epithelial cells, day 1–6 tissues were incubated with anti-E-cadherin antibodies (Cell Signaling Technology, 24E10, 1:500), followed by incubation with an anti-rabbit antibody conjugated with Alexa 555 (A21428, Thermo Fisher Scientific, 1:500). 3D images were acquired using the LSM 880 (Zeiss) and AXR (Nikon) microscopes. The surface tool in Imaris (version 9.8; Oxford Instruments) was used to construct a 3D structure from the images.

Measurement of serum E₂ and P₄ levels

Blood samples were collected from mice on the indicated day of pregnancy. Serum P₄ levels were measured as described previously³⁸, using a progesterone enzyme-linked immunosorbent assay (ELISA) kit (582601, Cayman). Serum E_₂ levels were measured using an estradiol ELISA kit (501890, Cayman).

Spatial transcriptomics

Spatial transcriptomes were analyzed using 10x Visium (10x Genomics) following the manufacturer’s protocol. Day 6 uteri from control and Taz-uKO females were collected. Frozen sections (10 μm) were mounted on gene expression slides and sent to KOTAI Bio Inc. (Osaka, Japan) for processing. Following a 30-minute Proteinase K reaction, the sections were hybridized with spatial tags on the slides and reverse-transcribed in situ. The cDNAs were analyzed by RNA sequencing using DNBseq (MGI) with 300 million reads per sample. Raw FASTQ files and microscope slide images for each sample were processed with Space Ranger software (version 1.1, 10× Genomics) using the “spaceranger count” pipeline, involving STAR with the default parameters for aligning reads against the mouse reference genome mm10 “refdata-gex-mm10-2020-A.” This pipeline uses Visium spatial barcodes to generate a feature spot matrix with unique molecular identifier counts. Clustering analysis was performed using Seurat (version 5.0.0)⁵² and clusters were visualized using UMAP. Differentially expressed genes between genotypes were identified using an adjusted p-value < 0.05 and a fold change > 1.5. Metascape³³ and Enrichr⁵³ were used to analyze the GO terms and upstream transcription factors within each cluster, respectively. The Mouse Visium data were deposited to the GEO database (Accession No. GSE305995).

scRNA-seq and data analysis

The 10x Genomics Chronium FRP protocol was followed for scRNA-seq analysis. On days 4 and 5 for WT, or day 4 evening and midnight for p38α floxed and uKO, uterine horns were excised, snap-frozen, and sent to Takara Bio Co.(Osaka, Japan). After fixing the cells with formaldehyde, a single-cell suspension was prepared using a GentleMACS (Miletenyi Biotec). The cells were used for RNA sequencing library preparation using the Chromium Next GEM Single Cell Fixed RNA Sample Preparation Kit, Chromium Fixed RNA Kit, Mouse Transcriptome, Chromium Mouse Transcriptome Probe Set v1.0.1, Chromium Next GEM Chip Q Single Cell Kit, Dual Index Kit TS Set A, and Chromium X (10x genomics, USA). Paired-end sequencing was performed on an Illumina next-generation sequencer (NovaSeq 6000; Illumina). Raw FASTQ files were processed using Cell Ranger software (10x Genomics, USA), and Seurat (http://www.satijalab.org/seurat) v5.0.0⁵² was used to process read counts. Cell trajectory was determined using Monocle3⁵⁴. CellChat was used for the cell-cell interaction assay. The mouse scRNA-seq data were deposited in the GEO database (Accession No. GSE296581 and GSE305994).

Statistical analyses

Statistical analyses were performed using a two-tailed Student’s t-test or one-way analysis of variance (ANOVA), followed by Bonferroni post-hoc tests, in GraphPad Prism10. Statistical significance was set at P < 0.05.

Study approval

All animal experiments were approved by the Institutional Animal Experiment Committee of the University of Tokyo Graduate School of Medicine (approval numbers P20-076 and A2023M165).

Data availability

The RNA-seq experimental data will be made publicly available upon publication (GSE305994 and GSE305995). This study did not involve the development of custom code or algorithms. All the software used in this study is publicly available and is cited in the main text and Methods sections.

Acknowledgements

We thank Ms. Atsumi Miura for providing technical assistance. We are grateful to Francesco J. DeMayo (National Institute of Environmental Health Sciences) for providing Pgr-Cre mice, and to Kinya Otsuki (Osaka University) for providing Mapk14-floxed mice.

Additional information

Funding

This work was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI (grant nos. 23K08278, 23K27176, 24K22157, 24K21911, 25K02779, 25H01065), Japan Agency for Medical Research and Development (AMED) (grant no. JP25gn0110085, JP24gn0110069, JP25gk0210039, JP24lk0310083, JP25gn0110097, JP25gk0210042 and JP25gk0210045), Children and Families Agency (Grant Number JPMH23DB0101), Japan Science and Technology Agency (JST) Fusion Oriented Research for Disruptive Science and Technology (FOREST) (grant no. JPMJFR210H), Mochida Memorial Foundation for Medical and Pharmaceutical Research, Uehara Memorial Foundation, Inoue Foundation for Science, Astellas Foundation for Research on Metabolic Disorders, The Naito Foundation and the fund of joint research with NIPRO corporation.

Author contributions

Conceptualization: S.A. and Y.H.; Funding acquisition: S.A. and Y.H.; Investigation:C.I., S.A., Y.F., X.H., R.S.H., D.H., T.H., M.M.; Data analysis: C.I., S.A.; Data interpretation: C.I., S.A., Y.H.; Project administration and supervision: Y.H.; Writing - original draft preparation: C.I., S.A.; Writing – review and editing: S.A.

Funding

MEXT | Japan Society for the Promotion of Science (JSPS) (23K08278)

• Ryoko Shimizu-Hirota

MEXT | Japan Society for the Promotion of Science (JSPS) (23K27176)

• Shizu Aikawa

MEXT | Japan Society for the Promotion of Science (JSPS) (24K22157)

• Yasushi Hirota

MEXT | Japan Society for the Promotion of Science (JSPS) (24K21911)

• Yasushi Hirota

MEXT | Japan Society for the Promotion of Science (JSPS) (25K02779)

• Yasushi Hirota

MEXT | Japan Society for the Promotion of Science (JSPS) (25H01065)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP25gn0110085)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP24gn0110069)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP25gk0210039)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP24lk0310083)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP25gn0110097)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP25gk0210042)

• Yasushi Hirota

Japan Agency for Medical Research and Development (AMED) (JP25gk0210045)

• Yasushi Hirota

Children and Families Agency (JPMH23DB0101)

• Yasushi Hirota

MEXT | Japan Science and Technology Agency (JST)

https://doi.org/10.52926/jpmjfr210h

• Shizu Aikawa

Mochida Memorial Foundation for Medical and Pharmaceutical Research (公益財団法人 持田記念医学薬学振興財団)

• Shizu Aikawa

Uehara Memorial Foundation (UMF)

• Shizu Aikawa

Astellas Foundation for Research on Metabolic Disorders

• Shizu Aikawa

Inoue Foundation for Science

• Shizu Aikawa

Naito Foundation (内藤記念科学振興財団)

• Shizu Aikawa

Nipro (ニプロの)

• Yasushi Hirota

Additional files

Supplementary figures

Table S1

Table S2

Table S3

Table S4

Table S5

Table S6

Table S7

Table S8