CDX2 is essential for cell proliferation and polarity in porcine blastocysts

After several cleavage divisions, mammalian zygotes develop into blastocysts comprising two distinct cell groups: the trophectoderm (TE) and inner cell mass (ICM). The TE is a single layer of polarized epithelial cells that forms the outer wall of the blastocyst and eventually develops into the fetal placenta, whereas the ICM comprises pluripotent cells and eventually develops into the embryo proper. As the earliest-born epithelial cells of mammalian embryos, the TE displays tissue characteristics that are typical of adult epithelia, including a concerted action of cell adhesion molecules, intercellular junctions and an extensive system of intermediate filaments that ensure epithelial integrity and polarized cytoplasmic organization (Fleming et al., 1992; Wiley, 1988). In turn, TE integrity is essential for blastocyst expansion, embryonic patterning, nutrition and implantation.

In mice, TE identity is maintained through a gene regulatory network (GRN) that is orchestrated by the transcription factor, caudal-related homeobox protein 2 (CDX2) (Johnson, 2009). CDX2 is first detected in some nuclei of the 8- to 16-cell stage mouse embryo, and its expression later expands to all blastocystic TE cells (Strumpf et al., 2005; Ralston and Rossant, 2008). CDX2 is one of the earliest transcription factors exhibiting differential expression on the outside and inside of blastomeres. Before the blastocyst stage, CDX2 becomes spatially restricted to the outer blastomeres immediately before the downregulation of OCT4 (POU5F1), NANOG and SOX2 (Cockburn and Rossant, 2010). Overexpression of Cdx2 alone is sufficient to commandeer the whole TE GRN and override mouse embryonic stem (ES) cell pluripotency to induce morphologically and functionally normal trophoblast stem cells (Niwa et al., 2005; Strumpf et al., 2005). Cdx2^−/− mouse embryos lose TE epithelial integrity and die around the time of implantation (Strumpf et al., 2005), and exogenous Cdx2 titration into early blastomeres changes the number of cells allocated to forming the TE and ICM by altering cell polarity (Jedrusik et al., 2008). Collectively, these data suggest that Cdx2 expression before blastocyst formation plays a pivotal role in TE identity and function in mice.

However, reports on CDX2 function before blastocyst formation are contradictory. Several studies have shown that CDX2 is not required for development of the mouse cleavage stage or for apical polarization and viability of mouse TE cells (Blij et al., 2012; Ralston and Rossant, 2008; Strumpf et al., 2005; Wu et al., 2010). These studies demonstrate that mouse embryos depleted of Cdx2 through either gene knockout or RNA interference (RNAi) are able to develop to normal blastocysts in terms of overall cell number, ICM:TE ratio, and TE apical polarity, despite abnormal cell junctions. Furthermore, Cdx2-deficient blastomeres allocate a normal number of cells to the TE in chimeric embryos (Ralston and Rossant, 2008). Despite using the same gene knockout and RNAi approaches, these results are directly contradictory to those reported in other studies, which have shown that Cdx2 depletion compromises cell polarity and induces developmental arrest before the blastocyst stage, as well as significantly reducing blastocyst rate (the ability of zygotes to normally develop into blastocysts) (Jedrusik et al., 2010, 2015). One plausible explanation for this discrepancy could be the different genetic backgrounds among strains. Nevertheless, given the contradictions and ambiguities, the molecular mechanisms associated with CDX2 function in mouse embryos remain elusive.

CDX2 is also expressed during the early embryonic development of other mammals; however, unlike in mouse, CDX2 protein is detectable after blastocyst formation in humans (Niakan and Eggan, 2013) and at the morula stage in monkeys (Sritanaudomchai et al., 2009). In pigs and cows, CDX2 is specifically expressed in blastocyst TE cells (Kuijk et al., 2008). Functional analysis of CDX2 in preimplantation embryos using RNAi has revealed that monkey blastocysts fail to hatch (Sritanaudomchai et al., 2009), similar to the mouse, whereas bovine blastocysts hatched normally (Berg et al., 2011). Taken together, these results indicate that although CDX2 is important for early embryonic development, it exhibits functional diversity among species. As such, it is necessary to study CDX2 in multiple species to fully elucidate its function, and these studies will undoubtedly shed light on the molecular regulation of preimplantation development across mammalian species.

In the present study, porcine CDX2 expression was assessed at both mRNA and protein levels in successive stages of preimplantation embryos. In addition, the CDX2 function in porcine embryonic development was systematically explored using siRNA-mediated gene knockdown and lentivirus-mediated TE-specific gene regulation.

To date, CDX2 expression has only been shown in porcine blastocysts. To gain better insights into the role of CDX2 in early embryonic development, the spatial and temporal dynamics of CDX2 were assessed in early porcine embryos. CDX2 was detected in a subset of TE nuclei in cavitated day (D)5 blastocysts (∼20 cells), and in all TE cells in expanded D6.5 and hatched D7.5 blastocysts (Fig. 1A). No CDX2 signal was observed in the nuclei of cleavage-stage embryos and the ICM. Interestingly, OCT4 (Fig. 1A) was present in all cells of the embryos, even in D7.5 hatched blastocysts. In vivo staining of CDX2 and OCT4 in porcine embryos at these stages was identical to that of in vitro counterparts (Fig. S1A,B versus Fig. 1A). Quantitative real-time PCR (qPCR) results indicated that CDX2 mRNA expression was remarkably increased at the 16-cell stage just before blastocyst formation, but OCT4 mRNA was steadily expressed during all developmental stages (Fig. 1B). Consistent with the previous RNA sequencing data (Cao et al., 2014), RT-PCR results confirmed the absence of two other CDX family members, CDX1 and CDX4 (three technical replicates, Fig. S1C). Moreover, western blot (WB) analysis revealed that meiosis II (MII)-stage oocytes expressed CDX2, but only ∼1/400 of that present in D6.5 blastocysts (Fig. 1C). Fluorescence in situ hybridization of RNA (RNA-FISH) was performed to further confirm the CDX2 mRNA expression in early porcine embryos, which clearly showed a perinuclear localization (Fig. 1D). The high fluorescence intensity of RNA-FISH was also seen at the 16-cell stage, consistent with our qPCR results (Fig. 1B).

To address CDX2 function in early porcine embryonic development, modified Stealth siRNA against CDX2 mRNA (siCDX2) (Fig. S1D) was injected into porcine zygotes, which were then cultured in vitro for 7 days. Stealth siRNA effectively decreased target gene expression for 3-4 days, during which time mouse embryos develop from zygotes to blastocysts (Wu et al., 2010). Porcine embryos require ∼7 days to develop from zygotes into blastocysts in vitro. As such, CDX2 expression was evaluated by performing qPCR at different developmental stages after injection of siRNA into zygotes to monitor the efficacy of the injected siRNA. Embryos that had been injected with a scrambled siRNA duplex were used as the control group (siControl embryos), and we set CDX2 expression in siControl embryos as 100%. As expected, CDX2 expression was effectively downregulated at least 80% in siCDX2 embryos at D7 and earlier (Fig. 2A). The effective CDX2 knockdown was also verified by performing RNA-FISH at the four-cell and blastocyst stages (Fig. 2B), and by performing immunofluorescent staining of blastocysts (Fig. 2C).

The cleavage rate and D7.5 blastocyst rate were consistent between siCDX2 and siControl embryos; however, the hatching rate was significantly lower in D7.5 siCDX2 blastocysts compared with the siControl group [0% (0/47) with siCDX2 versus 32% (16/50) with siControl] (Fig. 2D; Fig. S1G,H). In addition, the cell number was significantly lower in D7.5 siCDX2 blastocysts (Fig. 2E), which were smaller in size compared with the siControl group (Fig. 2F). Cytokeratin 8 (CK8) is a TE-specific intermediate filament and serves as a marker of epithelization. Notably, positive CK8 immunofluorescence signals were observed in the TE of siCDX2 blastocysts, indicative of normal epithelization (Fig. 2G), while the comparable cell numbers between D5 siCDX2 and siControl blastocysts showed that cell proliferation was compromised after blastocyst formation in siCDX2 embryos (Fig. 2E). These results imply that CDX2 knockdown affects total cell number and the hatching process in blastocysts, independent of previous embryonic development.

Polarization and adherens junctions have appeared by the eight-cell stage in both murine (Eckert and Fleming, 2008; Watson, 1992) and porcine (Reima et al., 1993) embryos, but the role of CDX2 in these events is still under debate. Thus, we investigated CDX2 function in blastomere polarization given that CDX2 mRNA expression was observed during cleavage (Fig. 1B; Fig. S1C). F-actin has been widely used as a mouse blastomere polarization marker (Johnson, 2009; Stephenson et al., 2010); therefore, we assessed the effect of CDX2 knockdown on eight-cell porcine embryo polarization by F-actin staining (Fig. 3A). These results showed that F-actin underwent normal apical polarization at the eight-cell stage in both siCDX2 and siControl embryos. Moreover, staining of E-cadherin (CDH1) demonstrated that adherens junction assembly was also intact in eight-cell siCDX2 embryos (Fig. 3B). To confirm these findings, the expression of CDH1 and other genes associated with polarization (PAR1 and PAR3, F2R and PARD3, respectively), cell junctions and compaction (CTNNB1, encodes β-catenin), and cavitation (ATP1B1, encodes a Na⁺/K⁺-ATPase subunit) were examined by performing qPCR, which showed equivalent levels of expression of these genes in eight-cell siCDX2 and siControl embryos (P>0.05; Fig. 3C). Collectively, these results suggest that CDX2 is not required to establish cell polarity and junctions before blastocyst formation.

CDX2 depletion did not affect the initial formation of polarity and junctions; however, whether CDX2 depletion affects later blastocyst stages has not been determined. As such, TE apical polarity was examined by F-actin immunostaining and scanning electron microscopy (SEM). Our results showed that F-actin localized apically on siControl blastocysts, but was more evenly distributed in cytoplasm of siCDX2 blastocysts (Fig. 4A). Given that apical F-actin is crucial to establish microvilli, SEM also showed hindered apical microvilli development on TE cells of siCDX2 embryos as compared to siControl counterparts (Fig. 4B). Due to the important role of F-actin in stress fibers (Tojkander et al., 2012), improper F-actin organization might be a potential reason for the hatching failure of siCDX2 blastocysts. Moreover, SEM images clearly showed that the lack of apical microvilli in siCDX2 blastocysts also influenced formation of the superficial ridge structure between cell-cell contacts, which is generally composed of microvilli from two adjacent epithelial cells (Gorelik et al., 2003). To determine the key factors mediating the deficient TE polarization and apical microvilli formation in siCDX2 blastocysts, qPCR was performed to analyze genes involved in these processes, such as ACTB, EZRIN, ARP2, CAPZA1 and PRKCA. The results showed that CDX2 depletion did not disturb ACTB, EZRIN, ARP2 or CAPZA1 mRNA expression, whereas PRKCA was significantly decreased (P<0.05) (Fig. 4C).

PRKCA is an important signal transducer that participates in cytoskeleton organization and function (Hong et al., 2011; Ng et al., 2001). As such, PRKCA expression was analyzed using qPCR during early porcine embryo development to determine whether its suppression was responsible for the abnormal apical polarity observed in siCDX2 blastocysts. Interestingly, PRKCA was maternally expressed in MII oocytes (Fig. 4D), which might be related to its roles in oocyte meiotic maturation and fertilization (Capo-Chichi et al., 2005; Fan et al., 2002). Thereafter, PRKCA expression decreased and remained at a low level until it significantly increased during morula-blastocyst transition (Fig. 4D), which corresponds to the period of CDX2 accumulation in the TE. To evaluate whether PRKCA expression is directly regulated by CDX2, we analyzed its expression using qPCR at 36 h after TE-specific lentivirus-mediated CDX2 overexpression or knockdown in D5 blastocysts (Fig. S2A,B). Notably, CDX2-knockdown (CDX-miR) embryos showed significantly lower PRKCA expression, whereas those with TE-specific red fluorescent protein-tagged CDX2 (CDX-RED) showed a marked increase when compared with that in controls that had been infected with lentivirus harboring negative control miRNA (Con-miR) or nucleus-localized red fluorescent protein (NLS-RED), or uninfected intact embryos. After a positive correlation between CDX2 and PRKCA expression was found, the effect of PRKCA on TE polarity was examined using siRNA knockdown in porcine embryos. These results demonstrated that PRKCA disruption produced the same partial phenotypes as CDX2 knockdown, including the failure to hatch (Fig. 4F,G), decreased blastocyst cell number (P<0.05, Fig. 4H) and loss of TE apical polarity (Fig. 4I).

Previous studies have indicated that phosphorylated EZRIN (p-EZRIN) acts as a bridge between F-actin and plasma membrane proteins during apical polarization, and that PKCα directly phosphorylates EZRIN (Baiocchi et al., 2008; Hong et al., 2011; Ng et al., 2001; Ren et al., 2009). Thus, EZRIN, p-EZRIN, CDX2 and PKCα levels were analyzed by western blotting of blastocysts that had been injected with siCDX2 or an siRNA against PRKCA (siPKCα). As expected, p-EZRIN was significantly lower in siCDX2 and siPKCα blastocysts when compared with that in siControl blastocysts (Fig. 4J); however, total EZRIN protein levels did not change. Moreover, CDX2 depletion reduced PKCα expression, but PKCα depletion did not affect CDX2 levels (Fig. 4J), indicating that PKCα is downstream of CDX2. These results were confirmed by immunostaining of p-EZRIN, which showed that both siCDX2 and siPKCα blastocysts lacked apically localized p-EZRIN (Fig. 4K). Collectively, these results strongly suggest that CDX2 is essential for maintaining cell polarity in the TE through induction of PKCα expression.

Since epithelial cell junctions and apical polarity are mutually dependent (Shin et al., 2006; Tepass, 2012), TE cell junctions were examined in siCDX2 blastocysts. Tight junctions are formed as a rather late molecular event that relies on both cell polarization and intact adherens junctions (Eckert and Fleming, 2008; Sheth et al., 2006). Immunostaining of ZO-1 (TJP1) at different developmental stages confirmed that tight junctions were established at the morula stage in porcine embryos (Fig. S3A). Moreover, we found that ZO-1 mRNA expression (Fig. 4L) and protein assembly was normal in siCDX2 embryos at D6.5 (Fig. S3B), but the mRNA was significantly upregulated at D7.5 (Fig. 4L, P<0.05). Additionally, although CDH1 mRNA levels were comparable in siCDX2 and siControl embryos (Fig. 4L), E-cadherin localization was more diffuse at the TE cell border in siCDX2 blastocysts (Fig. 4M), implying that adherens junctions in porcine TE are disrupted in the absence of CDX2 expression. Therefore, disrupted lateral junction formation, which is important for cavity sealing, could be another factor that aggravates the hatching failure of siCDX2 blastocysts, and this has been observed in mice (Strumpf et al., 2005; Wu et al., 2010).

To gain further insight into the decreased cell number in CDX2-deficient blastocysts, apoptosis was assessed via cleaved caspase 3 (Fig. 5A) and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining (Fig. 5B). Consistent with our observations, quantification of TUNEL analysis indicated a high-rate of apoptosis in siCDX2 blastocysts (Fig. 5C).

Subsequently, we examined the expression of select genes that are related to blastocyst development to evaluate the further development of siCDX2 embryos (Fig. 5D,E). The results showed that genes responsible for TE development – EOMES, ELF5 and GCM1 – were downregulated in D6.5 siCDX2 blastocysts; CDH3 (placental cadherin) and HAND1 (a gene important to placental development) were downregulated in D7.5 siCDX2 blastocysts; and FGFR2 (a signaling receptor important for elongation) was downregulated in both D6.5 and D7.5 siCDX2 blastocysts (Fig. 5D). However, key transcription factors for ICM development, SOX2 and NANOG, were both significantly upregulated in siCDX2 blastocysts, although mRNA encoding OCT4 was not (Fig. 5E). Immunofluorescence staining indicated that SOX2 expression was retained in the TE of siCDX2 blastocysts, whereas it was repressed in that of siControl blastocysts (Fig. 5F). Moreover, qPCR analysis of genes related to paracrine pro-proliferative factors (Fig. 5G) showed that BMP4 expression was slightly upregulated (P>0.05) in D6.5 siCDX2 blastocysts when compared to that in siControl embryos but that BMP4 levels subsequently diminished (P<0.05) at D7.5 only in siCDX2 blastocysts. The expression of both LIF and its receptor LIFR were also substantially increased in D7.5 siCDX2 blastocysts, while bFGF (FGF2) expression was comparable between siCDX2 and siControl embryos. However, expression of FGF4 was decreased, while that of its receptor, FGFR1, was increased in D6.5 siCDX2 blastocysts. In mouse blastocysts, CDX2 transcriptionally promotes Bmp4 expression in TE cells, which is subsequently secreted and binds to the ICM to promote proliferation and Fgf4 expression (Murohashi et al., 2010). In the current study, BMP4 expression was increased at the blastocyst stage during normal porcine development, consistent with CDX2 upregulation (Fig. S4B); however, both BMP4 and FGF4 were downregulated in siCDX2 blastocysts. In addition, we found that the BMP4 receptor, BMPR2, was expressed specifically in the porcine ICM (Fig. 6A). Together with previous reports of FGF4 expression occurring specifically in the porcine ICM (Fujii et al., 2013), we speculated that the CDX2-BMP4-FGF4 circuit observed in mice (Murohashi et al., 2010) might also be present in porcine embryos.

Based on this rationale, the effect of CDX2-BMP4 regulation on cell proliferation was examined in blastocysts that had been infected with TE-specific CDX-miR- or CDX-RED-encoding lentivirus. Notably, BMP4 expression was significantly higher in CDX2-overexpressing blastocysts at D6.5 than that in knockdown or control embryos (Fig. 6B). However, CDX2-overexpressing blastocysts severely degenerated at D7.5, most likely resulting from a harmful CDX2 overdose (Bou et al., 2016) (Fig. S2B). In TE-specific CDX2-knockdown embryos, BMP4 failed to increase at D7.5 and was significantly lower than that observed in the three control groups (Fig. 6B). Meanwhile, the mean number of ICM cells in D7.5 blastocysts was significantly increased (P<0.05) after TE-specific CDX2 overexpression, although the mean total of cells in the blastocyst and number of TE cells were decreased, most likely because of TE degeneration (Fig. 6C). Furthermore, both the blastocyst total cell number and TE cell number were significantly reduced in TE-specific CDX2-knockdown embryos compared to those in the three control groups (Fig. 6C).

To confirm the functional significance of BMP4 on porcine blastocyst proliferation, cell numbers were determined in D7.5 blastocysts that had been treated with BMP4 or the BMP4 antagonist Noggin, starting on D5. As expected, BMP4 addition significantly increased cell number in both siControl and siCDX2 blastocysts (Fig. 6D; Fig. S4A), whereas Noggin application inhibited cell proliferation specifically in siControl blastocysts (Fig. 6E), indicating that BMP4 is essential for porcine blastocyst proliferation.

Collectively, these results support our conjecture and suggest that CDX2 is important to blastocyst cell viability and proliferation, possibly by maintaining the BMP4-mediated blastocyst niche in pigs.

Mouse studies have shown that CDX2 is heterogeneously expressed in early blastomeres at the cleavage stage (Dietrich and Hiiragi, 2007), which is believed to contribute to lineage specification (Niwa et al., 2005). However, data from several studies show contradictory effects following the regulation of Cdx2 levels in early blastomeres (Jedrusik et al., 2008; Ralston and Rossant, 2008). Thus, we examined whether regulating CDX2 levels in porcine blastomeres during cleavage influences their developmental fate. For this, siCDX2 or CDX2 mRNA, along with an episomal plasmid expressing EGFP (pS/MAR-EGFP), were injected into one blastomere of a four-cell porcine embryo, and the distribution of EGFP-positive cells was traced in blastocysts (Fig. 7A). Results showed that the blastocysts segregated into three groups: ‘ICM only’ (all EGFP-positive cells contributed to the ICM), ‘TE only’ (all EGFP-positive cells contributed to the TE) or ‘ICM&TE’ (positive cells were found in both the ICM and TE) (Fig. 7B). Interestingly, the proportion of TE-only embryos was significantly increased in the CDX2-overexpressing group (Fig. 7C; P<0.05), whereas no significant difference was found between the CDX2-knockdown and control groups. Thus, these data suggest that CDX2 plays an important role in the GRN controlling the molecular features of the TE in porcine embryos but that CDX2 is insufficient to determine TE fate.

Genes are known to show species-specific expression patterns or functions in mammals. For example, Cdx2 expression in early mouse embryos resembles neither that in humans (Niakan and Eggan, 2013) nor monkeys (Sritanaudomchai et al., 2009). CDX2 is a major regulator of initial lineage segregation and trophoblast stem cell properties. The current study indicates that the CDX2 expression pattern in porcine embryos is relatively similar to that of humans, in which CDX2 is upregulated in TE cells after blastocyst formation (Niakan and Eggan, 2013). As such, exploring CDX2 function in porcine embryos might enhance our knowledge of molecular regulation in embryology in general and shed light on the function of CDX2 during human development.

The spatiotemporal expression of CDX2 in porcine embryo differs from that of mouse embryo (Dietrich and Hiiragi, 2007) (Fig. 7D). In mice, CDX2 protein is first detected in some blastomeres of eight-cell stage embryos and its expression becomes restricted to the TE by E3.5. In contrast, porcine CDX2 protein is initially observed in some TE cells right after cavitation at E5 and remains exclusively in the TE thereafter. Interestingly, the time point for porcine CDX2 protein expression is very similar to that in bovine (Goissis and Cibelli, 2014; Madeja et al., 2013) and human (Niakan and Eggan, 2013) embryos, in which CDX2 protein becomes detectable after blastocyst formation. Similarly, prolonged OCT4 expression in the TE is also observed in human (Cauffman et al., 2005) and bovine (van Eijk et al., 1999) embryos. Since CDX2 accumulation is primed by Hippo signaling in response to cell contact and positional information in mice (Hirate and Sasaki, 2014; Nishioka et al., 2009), it would be interesting to explore whether this signaling pathway exists in porcine embryos, as this knowledge would expand our mechanistic understanding of the various CDX2 expression patterns observed in mammalian embryos.

The current data on CDX2 function in mouse embryos before blastocyst formation is contradictory (Bruce, 2011; Wu and Schöler, 2011). Our results, derived from CDX2-knockdown porcine embryos, indicate that CDX2 is not essential before blastocyst formation, since the siCDX2 blastocyst rate was no different from that of the control. This result is consistent with several studies in mice, which show that Cdx2-mutant (Strumpf et al., 2005) or knockdown (Wu et al., 2010) embryos develop normally to the blastocyst stage; however, others have found that Cdx2 depletion induces developmental arrest before blastocyst formation (Jedrusik et al., 2010, 2015). Different time points for CDX2 accumulation in mice and other mammals might account for the different phenotypes of CDX2-depleted mouse and pig embryos. Therefore, the results from porcine studies may be more pertinent to our understanding of CDX2 function in embryonic development in large animals, including cows and primates, which also show the same temporal pattern of CDX2 accumulation (Goissis and Cibelli, 2014; Madeja et al., 2013; Niakan and Eggan, 2013). However, more studies in various strains are needed to clarify the functional significance of CDX2 in mouse cleavage-stage embryos (Wu and Schöler, 2011).

The effect of CDX2 on mouse TE-fate commitment is still a matter of debate. Results from a study by Jedrusik et al. (2008) indicate that CDX2 is a key factor in TE-fate commitment, as increased Cdx2 expression in individual blastomeres promotes differentiation towards the TE lineage and hinders that of the ICM. In contrast, Ralston and Rossant (2008) indicate that CDX2 functions downstream of lineage allocation since Cdx2-depleted blastomeres do not preferentially contribute to the ICM. Our study found that cell polarization and junctions in siCDX2 porcine embryos before the blastocyst stage were similar to those of controls. Although CDX2-overexpressing blastomeres from four-cell stage embryos contributed to the TE at a higher frequency; blastomeres in which CDX2 had been downregulated did not contribute to the ICM in the same way. Consistent with our results, a bovine study (Goissis and Cibelli, 2014) has indicated that CDX2 depletion does not affect the blastocyst rate and TE, ICM or total cell number. Interestingly, although CDX2 is first detectable at the morula-stage in monkey embryos, CDX2 knockdown did not affect blastocyst formation (Sritanaudomchai et al., 2009). Collectively, these results suggest that CDX2 is not involved in TE-fate decisions in large animal embryos.

Data regarding the role of CDX2 on TE cells and further development of blastocysts are similar among different studies and different species. In mice, both Cdx2-knockout and -knockdown blastocysts exhibit a hatching failure, decreased cell number, abnormal TE cell junctions/polarity and a disturbed TE GRN (Blij et al., 2012; Jedrusik et al., 2010, 2015; Nishioka et al., 2008; Ralston and Rossant, 2008; Strumpf et al., 2005; Wu et al., 2010). In primates (Sritanaudomchai et al., 2009) and cattle (Goissis and Cibelli, 2014; Madeja et al., 2013), CDX2 knockdown results in a hatching failure and deficient cell proliferation in blastocysts. Consistently, our study also showed that CDX2 knockdown compromises TE cell polarization and junctions. Furthermore, our results indicate that PKCα is a key intermediary for the effect of CDX2 on TE cell polarity.

A CDX2-BMP4-FGF4 functional circuit has been shown in mouse embryos (Murohashi et al., 2010). Similarly, our study determined that BMP4 promotes blastocyst cell proliferation and is regulated by CDX2 in the TE of porcine blastocysts. This suggests that downregulation of BMP4 in siCDX2 porcine embryos is partially responsible for the deceased cell number, along with the increasing rate of apoptosis. Thus, it is also reasonable to conclude that the hatching-failure phenotype seen in siCDX2 porcine embryos could result from defects in cell proliferation, as well as in TE cell junction/ polarity. Despite the discrepancy between species in CDX2 function before blastocyst formation, CDX2 participation in TE maintenance is clearly necessary for blastocyst development among various mammals.

It is noteworthy that CDX2 plays species-specific roles in regulating trophoblast-related gene expression. Reports from mouse trophoblast stem cell studies show that CDX2, EOMES and ELF5 form a positive-feedback loop that maintains lineage fate (Ng et al., 2008). However, according to a previous study (Valdez Magaña et al., 2014), this loop might not exist in the porcine trophoblast lineage since EOMES and ELF5 are only highly expressed in tubular-stage porcine embryos, when CDX2 expression in TE cells has already been downregulated, and ELF5 is not co-expressed with CDX2 in the porcine TE. Thus, it is assumed that CDX2 does not transcriptionally upregulate EOMES and/or ELF5 expression in porcine TE, and it is more likely that CDX2 and EOMES and/or ELF5 control the TE GRN at the early and late stages, respectively. Therefore, higher EOMES and ELF5 levels in siCdx2 D7.5 porcine embryos might occur due to a compensative mechanism following CDX2 downregulation.

Along with CDX2 studies in other mammalian embryos, the current study of porcine embryos highlights the importance of CDX2 in early mammalian development. In Fig. 7D, we compare preimplantation embryonic development between mouse and pig, in vitro and in vivo, and include key developmental events such as zygotic genome activation (ZGA), compaction, cavitation and implantation. In addition, the key results of this study are summarized in Fig. 7E. Comparisons of embryonic development among multiple mammalian species will not only improve overall knowledge of mammalian embryonic development but also provide clues to find more suitable animal models for human embryonic development. Collectively, our findings provide new ideas on CDX2 function in human embryonic development, especially with regard to TE development, due to the similarities in CDX2 spatiotemporal expression patterns between pig and human embryos.

All animal procedures conformed to the guidelines and regulatory standards of the Animal Care and Use Committee of Northeast Agricultural University. Porcine oocytes were matured in vitro, as previously described (Liu et al., 2008). To generate in vitro fertilization embryos, denuded oocytes were washed and held in modified Tris-buffered medium before fertilization using standard procedures (Abeydeera and Day, 1997). Presumptive zygotes were then cultured in PZM-3 medium (Yoshioka et al., 2002) for 7 days at 39°C under a 5% CO₂ atmosphere. Cleavage rates were evaluated at day 2 using microscopy, whereas blastocyst and hatching rates were assessed at day 7.5. Cell numbers at days 5, 6.5 and 7.5 were quantified by DAPI nuclear staining.

In vivo embryos were collected from Yorkshire-Danish Landrace sows from days 1 to 6 post insemination. The sows were mated twice within 24 h after detection of estrus, at an interval of 12 h. The last mating time was taken to be day 0 (D0) of conception. Reproductive organs collected from the sows were transported to the laboratory in warm saline (37°C) for a maximum period of 1 h. Embryos were rinsed twice using 50 ml of PBS supplemented with 5% fetal bovine serum (FBS). All procedures were approved by the Animal Ethics Committee of the Northeast Agriculture University, Harbin, China.

Stealth siRNA against porcine CDX2 designed with BLOCK-iT RNAi Designer was injected into zygotes or one blastomere of a four-cell stage porcine embryo. The most effective duplex (sequence given here) was selected from three Stealth siRNAs: siCDX2sense: 5′-CGAAAGACAAAUACCGAGUCGUGUA-3′; siCDX2 antisense: 5′-UACACGACUCGGUAUUUGUCUUUCG-3′.

CDX2 siRNA efficiency and specificity was demonstrated in porcine intestinal epithelial cells (Fig. S1E). At the same time, a scrambled siRNA duplex with the same nucleotide composition but no specific target was used as the control. The sequence of the control siRNA was as follows: siControl sense: 5′-CGAACAGAUAAAGCCGCUGUAAGUA-3′; siControl antisense: 5′-UACUUACAGCGGCUUUAUCUGUUCG-3′.

CDX2 was overexpressed in one blastomere of a four-cell stage porcine embryo through mRNA injection. The CDX2 full-length open reading frame was cloned from cDNA of a porcine blastocyst and inserted into pMACS KkHA(N) plasmid (Miltenyi Biotec) and used for in vitro transcription (RIBOMAX Large Scale RNA Production System T7 kit, Promega). The PCR primers used were as follows: CDX2F: 5′-ATGTACGTGAGCTACCTCCTGGACAAGGAC-3′; CDX2R: 5′-CTGGGTGACGGTGGGGTTTAACACGC-3′.

Microinjection of siRNA or mRNA was performed using an Eppendorf FemtoJet microinjector and Narishige NT-88NE micromanipulators. A glass capillary femtotipII (Eppendorf) was loaded with 5 μl of RNA (20 nM siRNA, 50 ng/µl mRNA) using a microloader (Eppendorf), and about 30 pl of solution was injected into the zygote or blastomere cytoplasm. To trace the fate of the single blastomeres that had been injected, 20 ng/µl of episomal plasmid pS/MAR-EGFP was co-injected with the siRNA or mRNA, or alone as control. Microinjection efficiency in this study was near 100%, as proven by injecting GFP mRNA under the same conditions (Fig. S1F).

To achieve TE-specific CDX2 regulation, the zona pellucida was removed from D5 early cavitated embryos with pronase and then washed. Each of 15 blastocysts was then infected in 30 μl of equilibrated PZM3 containing 5×10⁶ cfu of lentivirus in mineral oil and incubated at 39°C under 5% CO₂ for 3 h. Infected embryos were then washed and cultured in 500 μl of fresh PZM3. The total numbers of ICM and TE cells were calculated based on intranuclear – CDX2 and OCT4 – and fluorescent signals (Fig. S2C). The procedures for obtaining different lentiviruses with a FUW backbone were described previously (Zhi et al., 2014).

Stealth siRNAs against porcine PRKCA and mRNA encoding ZO-1 were designed using BLOCK-iT RNAi Designer. Because PKCα functions before the first cleavage (Capo-Chichi et al., 2005; Fan et al., 2002), PRKCA knockdown was accomplished by injecting siRNA into all blastomeres of two-cell embryos. siRNA against ZO-1 was injected into zygotes for knockdown. Stealth siRNA against mRNAs encoding PKCα and ZO-1 were: siPRKCA sense: 5′-UGGUUCACAAGAGGUGCCAUGAGUU-3′; siPRKCA antisense: 5′-AACUCAUGGCACCUCUUGUGAACCA-3′; and siZO1 sense: 5′-GCGCUACAAGUGAUGACCUUGAUUU-3′; siZO1 antisense: 5′-AAAUCAAGGUCAUCACUUGUAGCGC-3′.

Negative control siRNA (Cat#12935300, Invitrogen) was used as control siRNA for the above knockdown experiments.

Embryos were fixed with 4% PFA overnight at 4°C, and stored in methanol at −20°C. In situ hybridization was performed as previously described (Chazaud et al., 2006) with slight modifications. RNA probes for porcine CDX2 were generated by in vitro transcription of 936-bp full-length cDNA cloned into the pSPT18 plasmid. The sense and antisense probes were synthesized in vitro under control of the T7 and SP2 promoters, respectively, and labeled with digoxigenin-UTP using a DIG RNA labeling kit (SP6/T7) (Roche). The slide was sealed with ProLong Gold antifade reagent (Invitrogen) and visualized on a Leica TCS SP2 confocal microscope with a 40× oil immersion objective.

For real-time analysis of gene expression, 100 oocytes and embryos were harvested in 10 μl RLT buffer and RNA was extracted with PureLink Micro-to-Midi total RNA purification kit with DNAse I treatment (Invitrogen). The cDNA synthesis was performed using High Capacity cDNA reverse transcription kits (ABI).

The cDNA (∼8 oocytes/embryos per reaction) was used in 50-μl real-time reactions (SYBR Premix Ex Taq for Perfect Real Time, Takara) using oligonucleotide primers (Table S1). All transcripts levels were normalized against 18S rRNA for oocyte and embryo samples (Kuijk et al., 2007). For porcine intestinal epithelial cells, the 60S ribosomal protein L4 RNA (Nygard et al., 2007) was used as reference gene. qPCR was performed with the ABI 7500 real-time PCR system. Three to six biological and three technical replicates were performed. The 2^−ΔΔCt method was used for relative quantification analysis. To determine knockdown efficacy, embryos that had been injected with a scrambled siRNA duplex were used as controls (siControl embryos), and we set CDX2 expression of siControl embryos as 100%.

WB and immunofluorescence experiments in porcine embryos were conducted as previously described (Bou et al., 2016) with the antibodies described in Table S2.

After removing the zona pellucida, embryos were processed as previously described (Xia et al., 2011) and imaged with a S-3400N (Hitachi) microscope.

Apoptosis was determined by using the TUNEL technique. DNA strand breaks were directly labeled with red fluorescence using tetra-methyl-rhodamine-dUTP (In situ Cell Death detection kit, TMR red, Roche), as described previously (Hao et al., 2004).

Each experiment was repeated at least three times, and representative images are shown in figures. All values are reported as the mean±s.d. Statistical calculations were performed in GraphPad Prism 6 (GraphPad software Inc.). To compare the gene expression levels and embryonic development rate, we used unpaired two-tailed Student's t-tests (parametric) or unpaired two-tailed Mann–Whitney U-tests (non-parametric) when analyzing two groups. The one-way ANOVA with Tukey's post-hoc testing (parametric) or Kruskal–Wallis test with Dunn's correction for multiple comparisons (non-parametric) was used to analyze multiple groups, mainly when comparing cell numbers. The F-test, Browne–Forsythe test or Bartlett's test was used to determine the distributional assumption (normality and homogeneity of variance) of the data. Non-parametric tests were used when data did not fit a normal distribution. Significant differences were defined as P<0.05. Differences are shown with *, or different letters a, b, c.

We thank Lin Cui and Xuedong Wang (Northeast Agricultural University) for S.E.M. analysis and LetPub (www.letpub.com) for assistance with editing.