|Home | About | Journals | Submit | Contact Us | Français|
An effective bidirectional communication between an implantation-competent blastocyst and the receptive uterus is a prerequisite for mammalian reproduction. The blastocyst will implant only when this molecular cross-talk is established. Here we show that the muscle segment homeobox gene (Msh) family members Msx1 and Msx2, which are two highly conserved genes critical for epithelial-mesenchymal interactions during development, also play crucial roles in embryo implantation. Loss of Msx1/Msx2 expression correlates with altered uterine luminal epithelial cell polarity and affects E-cadherin/β-catenin complex formation through the control of Wnt5a expression. Application of Wnt5a in vitro compromised blastocyst invasion and trophoblast outgrowth on cultured uterine epithelial cells. The finding that Msx1/Msx2 genes are critical for conferring uterine receptivity and readiness to implantation could have clinical significance, because compromised uterine receptivity is a major cause of pregnancy failure in IVF programs.
Normal implantation is the gateway to pregnancy success and is realized by a reciprocal molecular dialogue between the blastocyst and uterus (Cross et al., 1994;Dey et al., 2004; Rinkenberger et al., 1997) The uterus proceeds through several phases with respect to implantation - prereceptive, receptive and refractory (nonreceptive) (Wang and Dey, 2006). Ovarian progesterone (P4) and estrogen direct these phases and their coordinated actions regulate proliferation and differentiation of various uterine cell types in a spatiotemporal manner to determine the window of uterine receptivity for implantation. In mice, uterine epithelial cells undergo proliferation under the influence of preovulatory estrogen on the first day of pregnancy. Rising P4 levels from newly formed corpora lutea from day 3 onward initiate stromal cell proliferation which is further stimulated by ovarian estrogen on day 4 morning. In contrast, epithelial cells cease proliferation and undergo differentiation under this condition, making the uterus conducive to blastocyst attachment to the luminal epithelium (LE) on the day 4 evening (Dey et al., 2004).
Blastocysts implant only when the uterus achieves a short window of receptivity (Wang and Dey, 2006). In mice, the uterus is prereceptive on days 1-3, but becomes receptive on day 4 of pregnancy or pseudopregnancy. By day 5, the uterus becomes refractory to implantation. Increased endometrial vascular permeability at the site of blastocyst apposition coincides with blastocyst attachment to the LE. This process can be visualized as discrete blue bands (sites of increased vascular permeability) after an injection of a blue dye solution prior to sacrifice (Paria et al., 1993; Psychoyos, 1973). The blastocyst attachment normally occurs on the evening of day 4 (2000-2400h) and becomes more prominent on day 5 when stromal cells begin extensive proliferation and differentiation to decidual cells (decidualization).
Various uterine phases can be replicated in ovariectomized delayed implanting mice given exogenous ovarian hormones (McLaren, 1968; Paria et al., 1993; Yoshinaga and Adams, 1966). The uterus becomes non-responsive to implantation with blastocysts undergoing dormancy when exposed to P4 alone. However, the delayed uterus will respond to the presence of blastocysts for implantation if exposed to estrogen 24-48 h after P4 priming. Even so, the induced window of receptivity lasts only for a limited period (~24 h). The uterus then spontaneously proceeds to the refractory phase. During the delayed state, the uterine responsiveness to implantation can be extended with continued P4 treatment (neutral phase), but readily responds to implantation with an estrogen injection (Paria et al., 1993). While few signaling pathways are critical for uterine receptivity in mice, the underlying mechanism by which a uterus transits from the prereceptive to the receptive to the nonreceptive phase is far from clear. Therefore, learning the mechanism by which this transition is achieved could have serious implications for IVF clinics, allowing clinicians to potentially extend the window of receptivity and grant transferred embryos more time to implant.
Homeobox transcription factors have critical roles during embryogenesis, but their functions are limited in most adult tissues with reduced plasticity. One exception is the female reproductive tract which undergoes morphological, cellular, and molecular changes during pregnancy. Because of this plasticity, genes encoding specific members of the homeobox, Wnt, BMP and hedgehog families confer critical functions to the uterus during pregnancy. Msx genes that encode homeodomain transcription factors are one of the oldest and most conserved families of homeobox genes in animals (Finnerty et al., 2009). They are related to the Drosophila msh gene (Cornell and Ohlen, 2000) and implicated as downstream targets of BMPs during development (Bei and Maas, 1998; Timmer et al., 2002). The mouse Msx family is comprised of three genes, Msx1, Msx2 and Msx3 (Davidson, 1995) and show overlapping expression patterns during development. While loss of Msx1 adversely affects many developmental processes and leads to perinatal lethality (Bach et al., 2003; Chen et al., 1996; Satokata and Maas, 1994), mice with loss of Msx2 show defects in hair follicle, calvarial bone, tooth, skin, heart, and mammary gland development, and are prone to developing seizures (Satokata et al., 2000). Evidence suggests that Msx1 acts as a negative regulator of differentiation (Hu et al., 2001; Woloshin et al., 1995). Studies also show that Msx1 and Msx2 can function as transcriptional repressors (Catron et al., 1996;Newberry et al., 1997; Zhang et al., 1996). Less is known about Msx3, and gene targeting of Msx3 in mice has not yet been reported (Mehra-Chaudhary et al., 2001).
Our observation of persistent Msx1 expression in pregnant or pseudopregnant uteri of Lif−/− mice which show implantation failure (Daikoku et al., 2004; Stewart et al., 1992) suggested that uterine Msx1 is responsive to LIF signaling and critical for implantation. Previously, two studies examined Msx1 expression in the adult mouse uterus; one showed Msx1 expression in non-pregnant uteri (Pavlova et al., 1994), while our study showed transient cell-specific expression around the time of implantation (Daikoku et al., 2004), suggesting its role in implantation. Defining how Msx1 establishes and/or maintains the window of receptivity may allow us to potentially manipulate this window to increase the success of IVF programs.
Here we define the roles of Msx1 and Msx2 in pregnancy by conditionally deleting Msx1 and/or Msx2 in the uterus. While Msx1 is robustly expressed in the receptive uterine epithelium on day 4 of pregnancy, Msx2 expression is barely detectable. However, Msx2 is expressed in a similar fashion in Msx1-deleted uteri, suggesting Msx2’s compensatory role. Indeed, partial retention of fertility in Msx1-deleted females was totally abolished by superimposition of Msx2 deletion. More intriguingly, loss of Msx expression alters luminal epithelial cell integrity and apical-basal polarity. These changes are correlated with altered E-cadherin/β-catenin complex formation at adherens junctions via Wnt signaling. These results illustrate an unexpected role for Msx genes in altering the epithelial cell polarity required for blastocyst attachment to the LE and its homing into the stroma.
Msx1 is expressed in the LE and glandular epithelium on the morning of day 4 of pregnancy (Daikoku et al., 2004). To examine its expression in more detail, we used a reporter mouse line where nuclear LacZ (nLacZ) is expressed as a readout for Msx1 expression (Houzelstein et al., 1997). We found that uterine Msx1 expression is very low to undetectable on days 1 and 2 of pregnancy, but begins to appear from day 3, becoming robust on day 4 morning before decreasing by day 4 evening where its expression is remarkably downregulated with blastocyst attachment (Figure 1A). Uterine Msx1 expression does not depend on the presence of blastocysts, since it is also expressed on day 4 in pseudopregnant mice (Daikoku et al., 2004) and becomes undetectable past the anticipated time of implantation (Figure 1A), suggesting that Msx1 is important for the transitions of the uterus from prereceptive to receptive to nonreceptive phases regardless of the embryo’s presence.
The delayed implantation model serves to study various phases of uterine sensitivity to implantation. To induce delayed implantation, pregnant mice ovariectomized on day 4 morning were maintained by daily P4 injections from day 5 until sacrifice. To initiate implantation, P4-primed females received an injection of estradiol-17β (E2) on day 7. Implantation sites were recorded 24 h post-injection. We found that Msx1 expression persists in P4-treated delayed uteri, but disappears with blastocyst implantation after E2 injection (Figure 1B).
Maternal LIF is critical for implantation in mice. Msx1 is expressed in a similar fashion in WT and Lif−/− uteri on day 4 of pregnancy. While Msx1 expression is downregulated with the onset of implantation and thereafter in WT uteri, its expression persists in Lif−/− uteri beyond the normal window of implantation (Daikoku et al., 2004). We asked whether estrogen acting via LIF is responsible for downregulating Msx1 expression. Pseudopregnant WT or Lif−/− mice were ovariectomized and treated daily from days 5-7 with P4. On day 8, WT and Lif−/− mice received P4, P4 + E2, or P4 + LIF (20 μg/mouse, ip) and were sacrificed 24 h post-injection. While E2 downregulated Msx1 expression in WT uteri, it failed to do so in Lif−/− uteri. However, Msx1 expression was downregulated if P4 treatment was combined with LIF (Figure 1C). We also found that the rapid Lif induction by E2 in ovariectomized P4-primed WT uterus is coincident with downregulation of Msx1 (Figure 1D & E).
Msx1 downregulation by LIF suggests a role for Msx1 in implantation, since LIF injection in Lif−/− females on day 4 rescues implantation (Song and Lim, 2006). LIF binds the LIF receptor and its partner gp130 to execute its function (Hudson et al., 1996). We found that conditional uterine deletion of gp130 (gp130d/d) also results in sustained Msx1 expression (Figure 1F), suggesting that Msx1 is downstream of LIF signaling. Notably, blastocysts failed to implant in gp130d/d mice; unimplanted blastocysts were recovered from these mice (n=3 females, 7.0 ± 2.0 blastocysts/mouse, mean ± SEM).
We conditionally inactivated Msx1 in the uterus by crossing Msx1loxP/loxP mice with progesterone receptor-Cre (PgrCre/+) transgenic mice. Breeding Msx1loxP/loxP males with Msx1loxP/loxP females yielded normal litter sizes (n=3 females; 7.0 ± 1.0 pups/litter, mean ± SEM), confirming their normal reproductive functions. In situ hybridization showed normal uterine expression of Msx1 in WT and Msx1loxP/loxP mice on day 4 morning and evening and day 5 morning of pregnancy (Figure. S1A & B). Breeding Msx1loxP/loxP females with PgrCre/+ males generated Msx1loxP/loxP/Pgr+/+ (Msx1f/f) and Msx1loxP/loxP/PgrCre/+ (Msx1d/d) mice. To ensure that Msx1 was deleted in Msx1d/d uteri, Msx1f/f and Msx1d/d females were mated with WT males and sacrificed on day 4 morning. Deletion of uterine Msx1 was confirmed by detecting the deleted allele in genomic DNA. Msx1 expression was undetectable both at the mRNA and protein levels in Msx1d/d uteri (Figures S1C-E).
Msx1d/d mice have significantly reduced litter sizes (n=11; 3.3 ± 0.8 pups/litter, mean ± SEM) compared with those in Msx1f/f littermates (n=26; 8.5 ± 0.4 pups/litter, mean ± SEM). In addition, 7 Msx1d/d plug-positive females did not produce any pups (Figure 2A). These results show that Msx1d/d females have markedly compromised fertility, leading us to explore the site- and stage-specific effects of Msx1 during pregnancy.
To compare stage-specific pregnancy phenotypes, Msx1f/f and Msx1d/d mice were examined on day 5. Msx1d/d females showed defective implantation; 7 of 18 (~39%) plug-positive Msx1d/d mice failed to show any blue bands, while the remaining 11 females showed faint bands as compared to distinct bands in all Msx1f/f females (n=24) (Figure 2B, Table S1A). Blastocysts were recovered from uteri with no or very weak blue bands, suggesting defective implantation in Msx1d/d mice. Notably, Msx1d/d females have normal ovulation, fertilization and preimplantation development (Figure S2A).
To address if the uterine milieu in Msx1d/d females was comparable to that in Msx1f/f females, we examined the expression of genes known to be critical for uterine receptivity (Lee et al., 2006; Lim et al., 1999; Matsumoto et al., 2002). Similar expression patterns of Indian hedgehog (Ihh) and Hoxa10 were noted in Msx1f/f and Msx1d/d mice on day 4, suggesting that Msx1 has little influence on these P4-responsive genes (Figure 2C). We also examined the uterine cell proliferation and differentiation status by Ki67 immunostaining on day 4 when epithelial cells become differentiated and stromal cells undergo extensive proliferation in preparation for implantation and decidualization. In Msx1d/d uteri, numerous epithelial cells were still Ki67-positive, but Ki67-positive cells were sparse in the stroma, indicating impaired epithelial differentiation and stromal cell proliferation (Figure 2C). These results imply that Msx1 deficiency leads to compromised epithelial-mesenchymal interactions required for implantation. Cox2 and Bmp2 are critical for implantation and decidualization (Lee et al., 2007; Lim et al., 1997; Paria et al., 2001). The expression of both Cox2 and Bmp2 was reduced surrounding the blastocyst in Msx1d/d uteri showing weak blue bands (Figure 2D). These results led us to examine uterine histology and cellular architecture in Msx1d/d mice.
Blastocysts normally attach and implant at the end of invaginated slit-like lumens in crypts (nidus) at the antimesometrial pole of the uterus. Histological analysis of implantation sites showed notable differences between Msx1f/f and Msx1d/d mice on day 5. In Msx1d/d females, blastocysts attached laterally and/or in the middle of lumens rather than at the end of lumens with typical crypts (Figure 2D, top and bottom panels). In fact, uterine lumens in Msx1d/d females often did not assume slit-like architecture to encourage apposition of the blastocyst with the LE, and lacked well-defined crypts for blastocyst attachment and homing. Moreover, the stromal bed surrounding the blastocyst showed poor vascular permeability demarcated by more compact stroma resulting from reduced edema (demarcated area in Figure 2D, bottom panel).
We and others have shown that defective implantation leads to adverse ripple effects throughout the course of pregnancy, including abnormal embryo spacing, retarded fetoplacental growth and higher rate of resorptions (Song et al., 2002; Ye et al., 2005). Indeed, implantation sites were smaller in Msx1d/d females with abnormal embryo spacing when examined on days 6 and 8 (Figure 3A & B). By day 12, Msx1d/d females showed an increased number of resorptions (Figure 3C & D, Table S1A). We attribute these defects to defective implantation since Msx1 is not expressed in the uterus at later days of pregnancy (Figure S2B).
The blastocyst is the normal stimulus for decidualization; however, decidualization can also be induced experimentally by intraluminal oil infusion in pseudopregnant mice. Mice were sacrificed on day 8 (day of maximal decidualization) after intraluminal oil infusion on day 4 to assess the extent of decidualization by recording fold increases in uterine weights of infused over non-infused horns. Msx1d/d females had severely impaired decidual response (Figure 3E & F). Since the induction of decidualization entails the presence of a functional LE (Lejeune et al., 1981), compromised decidualization in Msx1d/d females perhaps resulted from an aberrant epithelial cell function due to loss of Msx1. These results suggest that poor implantation and decidualization arising from loss of Msx1 is due defective LE function.
Defective implantation and subsequent adverse processes could be secondary to reduced levels of serum P4 and E2 and/or their nuclear receptors ERα and PR in the uterus. This is a possibility since PgrCre/+ can delete floxed genes in the ovary. We found that Msx1 is not detectable in the WT ovary (Figure S2B) and serum levels of P4 and E2 are comparable between Msx1f/f and Msx1d/d mice on critical days of pregnancy (Figure S2C-D). ERα and PR expression was also comparable in day 4 uteri of both genotypes (Figure S2E). These results suggest that defective implantation in Msx1d/d females is not due to altered levels of P4 and/or E2 or their receptors.
During delayed implantation, the P4-primed uterus is unresponsive to implantation, but implantation is readily initiated by an E2 injection. When examined 24 h post-injection, implantation sites were remarkably scarce in Msx1d/d mice with the recovery of unimplanted blastocysts (Table S2A). These results show that implantation failure is not due to aberrant E2 and/or P4 function, and more importantly, Msx1 is necessary to sustain uterine readiness to implantation.
While Msx1 expression on day 4 is crucial for normal pregnancy, we sought to uncover the cause of partial retention of fertility in Msx1d/d females. We reasoned that Msx2 plays a compensatory role in the absence of Msx1, since Msx2−/− mice have normal fertility (Satokata et al., 2000). First, we conditionally deleted uterine Msx2 and found that these females have normal implantation examined on day 5 (n=4; 11.3 ± 0.6 implantation sites/mouse, mean ± SEM). Uterine Msx2 expression is very low to undetectable in WT and Msx1f/f mice on day 4 (Figure 4A). However, Msx2 expression was upregulated in Msx1d/d uteri, notably in a similar pattern as Msx1 in WT uteri (Figure 4B & C), prompting us to examine fertility in females with uterine deletion of both Msx1 and Msx2. We found total infertility in Msx1loxP/loxP/Msx2loxP/loxP/PgrCre/+ (Msx1/Msx2d/d) females as opposed to normal pregnancy in Msx1loxP/loxP/Msx2loxP/loxP/Pgr+/+ (Msx1/Msx2f/f) littermate females (Figure 4D, Table S1B); this infertility resulted from implantation failure (Figure 4E), but not due to defective ovulation, fertilization, or preimplantation embryo development (Figure S2F). Serum P4 and E2 levels in Msx1/Msx2f/f and Msx1/Msx2d/d females were within physiological ranges and uterine expression of ERα and PR was comparable between the two groups (Figure S2G & H). These data suggest that implantation failure was not due to altered ovarian hormone levels or their receptors. Furthermore, comparable expression of P4-responsive genes, Hoxa10 and Ihh (Figure S3A), and downregulation of E2-responsive genes, complement C3 (C3) and Lactoferrin (Ltf) (Figure S3B), in Msx1/Msx2f/f and Msx1/Msx2d/d uteri on day 4 suggest that compromised implantation did not stem from altered uterine sensitivity to hormonal changes. This is consistent with the failure of implantation in delayed implanting Msx1/Msx2d/d females after E2 and P4 injections (Table S2B). Implantation failure in Msx1/Msx2d/d females was associated with the loss of Bmp2 expression at the site of blastocyst apposition on day 5 and aberrant expression of Cox2, which was detected only in the LE surrounding the blastocyst as opposed to its expression in both the LE and stroma in Msx1/Msx2f/f mice (Figure 4F). These results led us to further explore the cause of implantation failure in Msx1/Msx2d/d mice.
LIF is critical for blastocyst implantation and is expressed in uterine glands on day 4 and in the stroma surrounding the blastocyst at the time of its attachment on day 4 night and persists through day 5 morning (Song et al., 2000; Stewart et al., 1992). We reasoned that defective implantation in Msx1d/d females is due to aberrant Lif expression. Indeed, Lif expression was remarkably downregulated in Msx1d/d uteri on days 4 and 5 (Figure 5A). Similar downregulation was also seen in Msx1/Msx2d/d uteri on day 4 (Figure 5B). These results suggest an interaction between Msx1 and LIF in regulating uterine receptivity and implantation. If Msx1 mediates its effects through LIF, then LIF administration should rescue implantation in females missing both Lifand uterine Msx1. We generated mice with a null mutation in Lif and uterine deletion of Msx1 (Lif−/−/Msx1d/d). Day 4 WT blastocysts were transferred to day 4 pseudopregnant Lif−/−/Msx1d/d or Lif−/− /Msx1f/f recipients. Recipients then received an injection of LIF (20 μg/mouse) and implantation sites were examined on day 6. Blastocysts failed to implant in Lif−/−/Msx1d/d recipients even after LIF injection with recovery of unimplanted blastocysts. In contrast, a similar LIF injection induced implantation in Lif−/−/Msx1f/f recipients (Figure 5C & D, Table S3A). We also injected LIF to Lif−/− and Msx1/Msx2d/d females on day 4. While LIF rescued implantation in Lif−/− females, it failed to do so in Msx1/Msx2d/d females (Figure 5E, Table S3B). These results suggest that uterine responsiveness to LIF is profoundly disturbed with loss of uterine Msx genes. These results led to the quest for the cause of implantation failure in these mice.
Normally, uterine lumens assume slit-like structures and epithelial cells transit from a columnar to cuboidal configuration approaching implantation (Finn, 1975). These features were often missing in Msx1d/d and Msx1/Msx2d/d mice, suggesting alteration in epithelial cell integrity and polarity. Adhesion of epithelial cells is primarily regulated by cadherins. E-cadherin, a Ca++-dependent transmembrane adhesion molecule, connects adjacent epithelial cells by linking their cytoskeletons via catenins to maintain cell-cell adhesion and apical-basal polarity (Witze et al., 2008). The loss of cell-cell adhesion and polarity resulting from downregulation of E-cadherin and/or β-catenin is seen in invasive tumors (Jeanes et al., 2008; Medrek et al., 2009), and loosening of cell-cell adhesion in the mouse uterine epithelium prior to blastocyst attachment is associated with the cleavage of E-cadherin extracellular domain (Potter et al., 1996). There is evidence that trophoblast adhesion to the epithelium involves transition in epithelial apical-basal cell polarity from a high to less polar state (Thie et al., 1996; Thie et al., 1998).
We examined the status of E-cadherin and β-catenin in uteri of Msx1/Msx2f/f and Msx1/Msx2d/d mice on days 4 and 6 of pregnancy by Western blotting. While levels of β-catenin remained similar or somewhat higher in Msx deleted uteri compared to floxed uteri, E-cadherin levels decreased on day 6 in Msx1/Msx2f/f uteri with implantation in progress (Figure 6A & B). Notably, E-cadherin was more sharply co-localized with β-catenin at the apicolateral border of the LE in Msx1/Msx2d/d than Msx1/Msx2f/f uteri, suggesting greater apical-basal polarity in deleted uteri prior to blastocyst attachment (Figure S4A). Histological analysis showed that LE cells are more columnar and polarized in deleted mice with increased cell height (Figure S4B).
This high polar state with co-localization of β-catenin and E-cadherin in the LE persisted in Msx1/Msx2d/d uteri on day 6 showing failure of implantation; no sign of penetration by the trophectoderm through the LE was noted in deleted uteri, as opposed to invading trophoblasts with loss of epithelial barrier and E-cadherin in Msx1/Msx2f/f uteri (Figure 6B); E-cadherin loss is due to LE breakdown at the site of implantation in Msx1/Msx2f/f uteri. However, co-localization of β-catenin with E-cadherin in the intact LE away from implantation sites in these females was also less organized as opposed to their highly organized localization in Msx1/Msx2d/d females (Figure 6C). These results are consistent with Co-IP data showing that more E-cadherin is physically associated with β-catenin in Msx1/Msx2d/d endometrial samples compared to those from floxed females on day 5 morning when the LE is still intact (Figure 6D).
We propose that one potential cause of implantation failure with uterine deletion of Msx genes is a lack of transition of the LE from a high to a less apico-basal polar state. The number of microvilli in the LE markedly diminishes with implantation (Murphy, 1993; Schlafke and Enders, 1975). Consistent with this, Msx1/Msx2d/d LE exhibited retention of apical microvilli and cilia, signatures of polarized cells, as assessed by staining of acetylated α-tubulin, a marker of cilia (Piperno and Fuller, 1985), and ezrin, a marker of stereocilia and microvilli (Yonemura and Tsukita, 1999). Relative to Msx1/Msx2f/f uteri, Msx1/Msx2d/d uteri exhibited more distinct staining for both markers at the apical region of the LE (Figure 6E & F). In addition, the localization of ZO-1 and αPKC, involved in tight junctions, appeared to be more organized in the apical region of the Msx1/Msx2d/d LE than in Msx1/Msx2f/f LE, although these proteins were present in tight junctions in epithelia of both genotypes (Figure S4C & D).
Wnt5a was shown to induce cell polarity by promoting E-cadherin/β-catenin complex formation via non-canonical pathway and adhesion receptors (Medrek et al., 2009; Witze et al., 2008). Wnt5a has multiple Msx1 binding sites, and Msx1 can bind to the human Wnt5a gene (Iler and Abate-Shen, 1996). We explored whether Msx1/Msx2 affects uterine receptivity and implantation by influencing Wnt5a signaling. Wnt5a is normally expressed in the uterine subepithelial stroma and epithelium, albeit at lower levels in the epithelium (Hayashi et al., 2009; Hou et al., 2004; Mericskay et al., 2004). In Msx1d/d and Msx1/Msx2d/d uteri, Wnt5a was upregulated in both the epithelium and stroma on days 4 and 6 at the protein and mRNA levels (Figure 6A, Figure S4E & F). After estrogen treatment, Wnt5a expression is upregulated in neonatal uterine epithelia of Msx2−/− mice, suggesting a role for Msx in regulating Wnt5a expression (Huang et al., 2005). By ChIP assay, we found that Msx1 binds to the Wnt5a promoter in a human uterine epithelial cell line HEC50B stably expressing Msx1 (Figure 6G). Of 10 primer sets, nine encompassed regions upstream and one within the coding region of the Wnt5a gene; each set contained one to three putative Msx1 binding sites. The region within primer set eight (P8), with two putative Msx1 binding sites, showed enrichment compared to other primer sets (Figure S4G).
To further show that Wnt5a influences E-cadherin/β-catenin complex formation, we cultured day 4 WT LE cells with more than 95% epithelial cell purity as assessed by cytokeratin-8 staining (Figure S5A). We found increased E-cadherin and β-catenin co-localization upon addition of Wnt5a to primary cultures (Figure 7A). Co-IP assays using a human uterine epithelial cell line (Ishikawa cells) also showed increased E-cadherin/β-catenin complex formation in the presence of Wnt5a (Figure S5B). We next examined whether Wnt5a affects blastocyst attachment and trophoblast outgrowth. Day 4 WT blastocysts were seeded onto attached primary epithelial cells in culture with or without Wnt5a. In these experiments, epithelial cells isolated from WT day 4 pseudopregnant females were cultured atop stromal cells (Figure 7B) or in chambers separated from stromal cells by an insert (Figure 7C). In both cases, blastocysts in the presence of Wnt5a showed initial adhesion, but trophoblast outgrowth was severely compromised as compared to those in the absence of Wnt5a (Figure 7B & C). However, Wnt5a did not inhibit trophoblast outgrowth in culture in the absence of epithelial cells, suggesting that the effects of Wnt5a requires epithelial cells (Figure S5C).
Canonical Wnt/β-catenin signaling in the uterus is also implicated in implantation (Mohamed et al., 2005). To see whether Msx1 affects this pathway, TOPGAL reporter mice were crossed with Msx1f/f and Msx1d/d mice, and β-gal staining was performed to see regional changes in Wnt/β-catenin signaling. No significant alteration in β-gal staining was noted in day 4 uteri of TOPGAL/Msx1f/f and TOPGAL/Msx1d/d mice (Figure S5D). Further, uterine levels of active β-catenin (unphosphorylated) and expression of Axin2, a known target of Wnt/β-catenin signaling (Jho et al., 2002), were comparable between floxed and deleted uteri on day 4 (Figure S5E & F). These results suggest that Wnt/β-catenin signaling is apparently not affected in the absence of Msx. Since LE cells transit from a more to a less apico-basal polar state for blastocyst implantation, we propose that E-cadherin/β-catenin association modulated by Wnt5a is a potential downstream target of Msx genes to influence implantation (Figure 7D).
Effects of uterine loss of Msx genes on epithelial morphology and polarity with heightened Wnt5a levels with no apparent changes in Wnt/β-catenin signaling suggest that an alternative Wnt pathway is active. Wnt5a is traditionally considered a non-canonical ligand and participates in Wnt/intracellular Ca++ or Wnt/PCP/JNK pathways (Kohn and Moon, 2005; Seifert and Mlodzik, 2007; Veeman et al., 2003). However, specific execution of these pathways depends on the receptor status and cellular context (Grumolato et al., 2010; van Amerongen et al., 2008).
We reasoned that if Wnt5a is to mediate its effects via a non-canonical pathway, a receptor and an effector system should be in place in the LE. Frizzled2 (Fzd2) and co-receptor ROR2 are known to bind Wnt5a to modulate JNK signaling (Oishi et al., 2003). Indeed, these putative Wnt5a receptors and a downstream kinase casein kinase 1ε (CK1ε) (Kani et al., 2004; Minami et al., 2010) are present in Msx1/Msx2f/f and Msx1/Msx2d/d uteri (Figure S5G-J). The levels of both receptors were apparently higher in Msx1/Msx2d/d uteri with Fzd2 being expressed in the stroma and epithelium, and ROR2 in the epithelium. CK1ε was also expressed in primary epithelial cells (Figure S5K). It is thus possible that E-cadherin/β-catenin complex formation in the LE is modulated by this downstream pathway.
Lessons learned from developmental biology show a remarkable degree of conservation of pathways/signaling molecules to program developmental processes across species. Alongside these roles, eutherian mammals have further utilized distinct developmental strategies to home developing embryos in the womb, and to nurture their growth through the processes of implantation, decidualization and placentation. This study presents two major unexpected findings. First, we show that Msx genes, known for their critical roles in development, particularly in craniofacial and limb development, have profound effects on implantation. Second, we present evidence suggesting that this role for Msx genes may be mediated by altering epithelial cell integrity through Wnt5a and E-cadherin/β-catenin complex formation.
Delayed implantation (diapause), a condition of suspended animation, is widespread in mammals. Nearly 100 mammals in seven orders exhibit this condition under various physiological and environmental conditions (Lopes et al., 2004; Renfree and Shaw, 2000), but the underlying mechanism by which the uterus and the embryo temporarily achieve quiescence and then resume implantation under favorable conditions remains largely unknown. The failure to initiate implantation in delayed implanting uteri lacking Msx1/2 strongly suggests that Msx genes are crucial for conferring readiness (neutral phase) to respond to an implantation cue. A comparative study on Msx genes in mammals which undergo delayed implantation may provide valuable information.
The relationship between Msx1 and LIF is also intriguing. Our present results suggest a positive-negative feedback loop between Msx1 and LIF in regulating each other’s activity. We propose that increasing Msx1 expression with rising P4 levels superimposed by preimplantation estrogen on day 4 facilitates Lif expression, and increased levels of LIF then downregulate Msx1 prior to implantation. The fact that Msx1 expression persists in Lif−/− mice and that Lif is downregulated with loss of Msx genes suggests that the regulation of one is influenced by the other, but their effects on implantation are mediated by different mechanisms, since implantation fails to occur in females deleted of Msx genes even after LIF administration.
The absence of Bmp2 and aberrant Cox2 expression at the site of blastocyst apposition in mice deleted for Msx genes suggests that the epithelial-mesenchymal interactions prerequisite for implantation are disturbed. Defective implantation in these mice suggests that Msx genes are required for optimal uterine receptivity with appropriate epithelial cell integrity for normal blastocyst attachment and invasion. Upregulation of Wnt5a in uteri deleted of Msx genes and binding of Msx1 to Wnt5a suggest that this morphogen is a potential downstream target of Msx genes and is important in regulating cell polarity. While Wnt5a is involved in uterine gland formation (Mericskay et al., 2004), the present results suggest that Wnt5a may also orchestrate epithelial apical-basal polarity by regulating E-cadherin/β-catenin complex formation. Our present in vivo and in vitro findings are consistent with previous work in cell culture systems that demonstrate roles for Wnt5a in directing cell polarity and cytoskeleton reorganization by influencing E-cadherin/β-catenin complex formation and adhesion receptors (Medrek et al., 2009; Mericskay et al., 2004; Witze et al., 2008) (Figure 7D).
Increased association of E-cadherin and β-catenin in Msx-deleted uteri with heightened Wnt5a levels suggests a non-canonical Wnt pathway without apparent changes in Wnt/β-catenin signaling. However, a role for Wnt/β-catenin cannot completely be ruled out, since there is evidence that Wnt5a and Wnt11 can form complexes and activate both non-canonical and canonical pathways in Xenopus axis formation (Cha et al., 2008). This suggests that nuclear β-catenin dependent and independent pathways are not always mutually exclusive.
Wnt5a can stabilize E-cadherin/β-catenin complex formation by increasing β-catenin phosphorylation by casein kinase 1 in cultured mammary epithelial cells (Medrek et al., 2009), and ROR2 can mediate non-canonical Wnt signaling in the presence or absence of Fzd (Green et al., 2008). Our results showing the presence of putative Wnt5a receptors Fzd2 and ROR2 and the downstream kinase CK1ε in the LE suggests that this signaling pathway is a potential mediator of Wnt5a’s effects in the uterus in the context of Msx genes. It is interesting that ROR2 mediates the transcriptional activity of Msx2 and Dlx5 (Matsuda et al., 2003) and that expression patterns of ROR2, Wnt5a and Msx genes show many similarities during development (Green et al., 2008; Satokata et al., 2000; Satokata and Maas, 1994; Yamaguchi et al., 1999). This suggests that Msx-Wnt5a-ROR2 pathway is also involved in uterine biology. Whether other effector pathways of Wnt5a are active in implantation will require further investigation.
The aberrant uterine luminal configuration, in the absence of Msx genes, notably its failure to assume a slit-like architecture and form a crypt, lend support to the importance of LE integrity in uterine biology and implantation. Since the expression of Msx genes is undetectable in stromal cells before, during and after implantation, the compromised decidualization in mice deleted of uterine Msx genes suggests that the defective LE function leads to poor stromal cell proliferation and differentiation to decidualization, ultimately compromising pregnancy outcome. The role of Msx genes in epithelial-mesenchymal interactions during development recapitulates their similar functions in epithelial-stromal interaction during implantation. Upregulation of Wnt5a in both the epithelium and stroma in uteri missing Msx genes is consistent with a role of these homeotic proteins in epithelial-stromal interaction.
Our findings have led us to believe that Msx genes play crucial roles in transitioning the uterus from the prereceptive to the receptive to the nonreceptive state. The importance of Msx1/Msx2 in human implantation is underscored by microarray gene expression analyses showing that both genes are downregulated during the window of implantation in women (Kao et al., 2002; Mirkin et al., 2005; Riesewijk et al., 2003), similar to which occurs preceding implantation in mice. The findings that Msx genes are critical to uterine receptivity and maintaining uterine readiness to implantation without altering the ovarian hormone levels or uterine sensitivity to these hormones are of high relevance to female fertility. These findings raise the possibility that clinicians may be able to develop new strategies to improve implantation rates in IVF programs by temporarily increasing uterine levels of Msx to extend the uterine responsiveness to implantation prior to embryo transfer. In the same vein, further uncovering the role of Msx may aid in developing non-hormonal contraceptives.
For detailed methods, see the expanded Experimental Procedures
PgrCre/+, Msx1loxP/loxP and/or Msx1loxPloxP/Msx2loxP/loxP mice were generated as described (Fu et al., 2007; Soyal et al., 2005). Lif−/− and Gp130loxP/loxP mice were generated by Philippe Brulet (Escary et al., 1993) and Werner Muller (Betz et al., 1998), respectively. Msx1-LacZ mice were provided by Benoit Robert (Houzelstein et al., 1997). All mice used in this investigation were housed in the Cincinnati Children’s Animal Care Facility according to NIH and institutional guidelines for the use of laboratory animals. All protocols were approved by the Institutional Animal Care and Use Committee.
Ovulation, fertilization, preimplantation embryo development and implantation were assessed as described (Hirota et al., 2010). The morning of finding the vaginal plug was considered day 1 of pregnancy.
Mice were ovariectomized on day 4 of pregnancy (0800-0900 h) and received daily P4 injections (2 mg/mouse, sc) from days 5-7. To induce blastocyst activation and implantation, P4-primed delayed mice were given an E2 injection (10 or 25 ng/mouse, sc) on day 7. Mice were killed 24 or 48 h post-E2 injection (Das et al., 1994). All steroids were dissolved in sesame oil.
Decidualization was experimentally induced as described (Lim et al., 1997). Sesame oil (25 μl) was infused intraluminally into one uterine horn on day 4 of pseudopregnancy; the contralateral horn served as a control. Fold increases in uterine weights of oil-infused horns over non-infused horns were used as an index of decidualization.
Paraformaldehyde-fixed frozen sections from control and experimental groups were processed onto the same slide and hybridized with 35S-labeled cRNA probes (Lim et al., 1997).
Total RNA (6 μg) was denatured, separated by formaldehyde-agarose gel electrophoresis, and transferred onto nylon membranes. Cross-linked blots were prehybridized, hybridized, and washed. 32P-labeled cRNA probes were used for hybridization and hybrids were detected by autoradiography (Das et al., 1994). rPL7 served as a housekeeping gene.
Protein extraction and Western blotting were performed as described (Hirota et al., 2010). The same blots were used for quantitative analyses of each protein. Bands were visualized by using an ECL kit (GE Healthcare). Actin served as a loading control.
Tissue sections from control and experimental groups were processed onto the same slide for immunohistochemistry and immunofluorescence (Hirota et al., 2010).
LacZ staining was performed as described (Ma et al., 2001).
RT-PCR was performed as described (Daikoku et al., 2008).
Serum E2 and P4 levels were measured by EIA kits (Cayman) (Hirota et al., 2010).
Epithelial and stromal cells were isolated from WT day 4 pseudopregnant females (Daikoku et al., 2005). Recombinant Wnt5a (R&D Systems) dissolved in PBS was used at a concentration of 1.2 μg/mL (Medrek et al., 2009).
The binding of Msx1 to Wnt5a was assessed by ChIP assays using a human uterine epithelial cell line (Hec50B) stably expressing Msx1-HA or HA (Ray and Das, 2006). See Supplemental Methods for primer sequences.
Co-IP was performed as described (Paria et al., 1999).
We thank Erin L. Adams for editing the manuscript, Bliss Magella for immunostaining and Sangwook Cha for helpful comments. This work was supported in part by NIH grants (HD12304 and DA06668 to S.K.D.), CCHMC Perinatal Institute Pilot/Feasibility Grant to T.D. J.C. is supported by an NICHD Training Grant, X.S. is supported by a Lalor Foundation Fellowship, and Y.H. was supported by a Japan Society for the Promotion of Science Fellowship for Research Abroad.
I, on behalf of all co-authors, declare that we do not have any financial conflict of interest with this submission.
Supplemental information includes Expanded Experimental Procedures and a total of 5 Supplemental Figures and 3 Tables.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errorsmaybe discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.