PC7 and the related proteases Furin and Pace4 regulate E-cadherin function during blastocyst formation

Early development from fertilization until the blastocyst stage in eutherian mammals is devoted to segregating a pluripotent inner cell mass (ICM) from the trophectoderm (TE) that enables attachment and survival in the mother. Differentiation of the ICM and the TE is initiated during the late eight-cell stage when individual blastomeres enlarge their cell–cell contact areas in a process termed compaction and begin to assemble junctional complexes in a polarized epithelial layer. Two subsequent rounds of symmetric or asymmetric divisions generate two identical daughters or one that is polar and one that is apolar, respectively (Handyside, 1980; Ziomek and Johnson, 1980; Johnson and Ziomek, 1981). Cultured apolar cells can become engulfed by the basolateral membrane of polarized cells, indicating that apical surfaces are less adhesive (Johnson and Ziomek, 1983; Dietrich and Hiiragi, 2007). Only apical membranes accumulate complexes of polarity proteins and atypical PKC (aPKC), and cells depleted of aPKC assume an inside position (Pauken and Capco, 2000; Plusa et al., 2005). It is possible, therefore, that cell positions are specified by asymmetric membrane inheritance. It was also reported that fates correlate with the angle of cell division (Bischoff et al., 2008). Other investigators concluded that only the most extreme symmetric divisions reliably predict outer fate (McDole et al., 2011) and that aPKC in reality promotes symmetric rather than asymmetric divisions by alleviating cortical tension and flattening cell shapes along the embryo surface (Dard et al., 2009). Accordingly, ICM fate may depend on uniform cell–cell contacts to block cell polarization and flattening (Hillman et al., 1972; Johnson and Ziomek, 1983). However, the precise mechanism specifying lineage differentiation remains unclear.

Dividing blastomeres in compacted morulae can still change positions: Inner cells occasionally rise to the surface to either assume an outer fate or quickly return to ICM, and some cells on the outside sink inside as late as during cavitation (Fleming, 1987; Yamanaka et al., 2010; McDole et al., 2011). Only the most surface-exposed mother cells in transition to the 16-cell stage give rise exclusively to outer cells, and they do so even if one daughter initially resides inside after asymmetric division (Watanabe et al., 2014). Lineage allocation, therefore, may not correlate with momentary cell positioning or polarization, but with the overall history of relative changes in cell–cell contacts.

Molecular differences among individual blastomeres already emerge before compaction. Increased DNA binding and distinct kinetics of the pluripotency determinant Oct4 and differential histone 3 arginine methylation may predict the fate of inner cells (Torres-Padilla et al., 2007; Plachta et al., 2011; Burton et al., 2013). All blastomeres initially also coexpress variable amounts of the TE lineage marker Cdx2 (Dietrich and Hiiragi, 2007; Ralston and Rossant, 2008). Unlike Oct4 kinetics, Cdx2 levels do not predict cell fate (Dietrich and Hiiragi, 2007). However, up-regulation of Cdx2 in outer cells is required to switch off Oct4 expression in TE after compaction (Strumpf et al., 2005) and to assemble tight junctions and boost mitochondrial activity (Ralston and Rossant, 2008; Wu et al., 2010).

Morula compaction, normal lineage segregation of inner and outer cells, and the regulation of Cdx2 expression critically depend on the cell–cell adhesion molecule E-cadherin (Stephenson et al., 2012). Until the early eight-cell stage, E-cadherin localizes on all cell surfaces, but thereafter becomes restricted to cell contacts during compaction (Vestweber et al., 1987). Meanwhile, apical surfaces of outer cells lacking contacts accrue aPKC, which enables the formation of tight junctions during TE epithelialization (Eckert et al., 2004, 2005). Mutant morula devoid of E-cadherin fail to compact and show ectopic aPKC membrane staining and elevated Cdx2 protein levels at the expense of Oct4 expression even in inner cells (Stephenson et al., 2010). In keeping with a role in ICM formation, depletion of E-cadherin by RNAi in only a subset of cells directs their progeny to the outside. In contrast, the progeny of blastomeres injected with dominant-negative mutant aPKC are diverted to ICM (Plusa et al., 2005). These observations established that E-cadherin and aPKC promote inner and outer fates, respectively, and that aPKC activity is limited by E-cadherin. E-cadherin is stabilized at cell–cell contacts by calcium-dependent homotypic interactions of its extracellular domain (Pey et al., 1998; Brasch et al., 2012). In inner cells that are surrounded by neighbors, E-cadherin thus accumulates circumferentially in adherens junctions that recruit the actin-binding proteins Nf2/Merlin and Angiomotin, thereby inducing the phosphorylation and cytoplasmic retention of Yes-associated protein (YAP) by Lats1/2 kinases. In contrast, polarization of nonadherent surfaces by activated aPKC sequesters Angiomotin and Lats1/2 beneath apical membranes (Cockburn et al., 2013; Hirate et al., 2013; Leung and Zernicka-Goetz, 2013; Anani et al., 2014). Because of the resulting inhibition of Lats1/2 signaling, polarized outer cells can translocate YAP to the nucleus, where it binds the transcriptional activator Tead4 to induce Cdx2 (Nishioka et al., 2008, 2009).

E-cadherin is synthesized as a precursor comprising an N-terminal prodomain, five cadherin repeats, a transmembrane, and a cytoplasmic domain. Prodomain cleavage enables calcium-dependent homotypic interactions among E-cadherin extracellular domains (Ozawa and Kemler, 1990). Cleavage is enhanced in insect cells that overexpress the endoprotease Furin in the trans-Golgi/endosome network and in the extracellular space (Bosshart et al., 1994; Molloy et al., 1994; Posthaus et al., 1998; Denault et al., 2002). However, E-cadherin processing was not inhibited in FURIN-deficient LoVo colon carcinoma cells and only reduced by <50% in breast cancer cells treated with 200-µM decanoyl-Arg-Val-Lys-Arg-chloromethylketone (CMK; Posthaus et al., 1998; Geng et al., 2012), a potent and membrane-permeable inhibitor of all proprotein convertases (PCs; Garten et al., 1994; Jean et al., 1998). It therefore remained unclear which of these proteases, if any, mediate E-cadherin processing.

Furin, PCSK5, PACE4 (also known as PCKS6/SPC4), and PC7 (PCSK7) all cleave secreted precursor proteins that contain the minimal recognition motif RXXR (Seidah and Prat, 2012). Furin and Pace4 activate TGF-β signals mediated by Nodal to specify the fate of progenitor cells in the developing mouse epiblast (Beck et al., 2002; Mesnard et al., 2011). In Xenopus laevis embryos, knockdown of PC7 suppresses eye development (Senturker et al., 2012), and a PC7-like activity can cleave pro-BMP4 during gastrulation (Nelsen and Christian, 2009). However, a role for PC7 in the mouse embryo is elusive. To identify novel functions of PCs, we developed the cell-linked indicator of proteolysis (CLIP) as a transgenic reporter substrate (Mesnard and Constam, 2010). In CLIP, a PC-cleavable linker connects secreted enhanced CFP (eCFP) to citrine that is targeted to the cell surface by a glycosyl-inositol phosphate anchor. Live imaging of CFP/YFP fluorescence ratios of CLIP and a comparison with an uncleavable mutant control (CLIPm) revealed that combined zygotic deletion of Furin and Pace4 inhibited CLIP processing already at the blastocyst stage by >80%. Because development did not arrest until after embryonic day (E) 5.5 (Mesnard and Constam, 2010), we asked whether the lack of Furin and Pace4 during the earliest stages of development is compensated for by yet another PC family member. Our results unmask combinatorial roles of Furin, PC7, and Pace4 in jointly regulating E-cadherin precursor processing and functionality in vitro and during morula compaction. In addition, detailed spatiotemporal mapping of PC activities in developing blastocysts revealed striking differences in their relative distribution and sensitivity toward the widely used pharmacological PC inhibitor CMK.

To evaluate a role for PC7, exons 3–7 encoding the translational start site, the pro region, and the catalytic domain were deleted by homologous recombination in embryonic stem (ES) cells (Fig. S1). Progeny of germline chimeras obtained from two independently targeted ES cell clones gave rise to fertile homozygous mutants at Mendelian frequency. No overt abnormalities could be identified, suggesting that PC7 is not essential under the conditions examined. To uncover overlapping functions of Furin and PC7 during early embryogenesis, we intercrossed Furin^+/−;PC7^−/− mice in an outbred NMRI genetic background. Analysis of progeny at E7.5 revealed 18% of empty decidua and <3% Furin^−/−;PC7^−/− double knockout (DKO) embryos (Fig. 1 A). Moreover, despite normal Mendelian ratios at E3.5, preimplantation embryos failed to hatch from the zona pellucida and were morphologically abnormal. To determine why Furin^−/−;PC7^−/− mutants might arrest development, embryos were collected at E2.5 and cultured until blastocyst stages (E3.5–4.0), followed by immunofluorescence staining of the cell polarity markers aPKC-ξ and E-cadherin and of tight junctions (ZO-1). Quantification of aPKC-ξ confirmed that it is enriched at the apical membrane of TE cells, whereas E-cadherin normally marks the basolateral side of TE and cell–cell contacts between ICM cells (Fig. 1, B and C; and Fig. S2 A; Vestweber et al., 1987; Pauken and Capco, 2000; Plusa et al., 2005). In DKO mutant blastocysts, aPKC-ξ staining increased at contacts between ICM cells expressing E-cadherin and on the basolateral side of one to four TE cells lacking contact with neighbors as determined by z-stack imaging (Fig. 1, B and C). A similar defect was observed in freshly flushed embryos (Fig. S2 B). Furthermore, even though all DKO mutants initiated blastocoel formation, this process was severely impaired both at E3.5 (n = 4/6) and after extended culture until E4.0 (Fig. 1 B, Fig. S2 C, and Table S1; n = 4/4), likely as a result of mislocalization of tight junctions (Fig. S2 D; n = 5/5). We conclude that combined loss of Furin and PC7 perturbs cell polarization and compaction at the morula stage and the development of a fully expanded blastocoel cavity.

To determine whether partial rescue of blastocyst formation involves an additional PC, we analyzed triple knockout (TKO) embryos lacking Furin, PC7, and Pace4. Both Pace4^−/− embryos and Furin^−/−;Pace4^−/− double mutants develop normal blastocysts irrespective of genetic background (Beck et al., 2002; Mesnard and Constam, 2010). PC7^−/−;Furin^−/−;Pace4^−/− TKO mutants were also found at the expected Mendelian ratios (Fig. S2 E). However, they all consisted of only loosely attached cells and failed to hatch at E3.5 or to cavitate (Fig. 1 B and Table S1; n = 5/5). Furthermore, double labeling of aPKC and E-cadherin revealed that E-cadherin staining at cell–cell junctions was decreased in TKO compared with DKO or wild-type embryos and relocalized to intracellular puncta, whereas aPKC-ξ was no longer confined to the embryo surface and instead accumulated ectopically along the circumference of both outer and inner cells (Fig. 1, B and C; and Fig. S2 A). aPKC and E-cadherin similarly mislocalized in freshly flushed TKO embryos (Fig. S2 B; n = 3/3), confirming that cell polarization is defective. Overall, these results suggest that overlapping Furin and PC7 activities are essential to correctly initiate compaction and cell polarization but that these processes are partially rescued in Furin;PC7 DKO embryos by Pace4.

To address whether rescue of DKO mutants temporally coincides with up-regulation of Pace4, we compared the activity profiles of individual PCs in preimplantation embryos expressing the transgenic reporter substrate CLIP. Although lack of expression of the CLIP transgene before E2.5 precluded analysis at the earliest stages, CLIP imaging was possible starting at E3.0. Therefore, mutant embryos and control littermates carrying the CLIP transgene were cultured from E2.5 until the stages of interest, followed by ratiometric imaging of eCFP and citrine (YFP) fluorescence intensities. At the late morula stage (E3.0), all cells showed strong YFP signals at the plasma membrane, whereas eCFP fluorescence of CLIP was not above background either in wild-type embryos or in Furin or PC7 single mutants (Fig. 2 A and Fig. S3 A), suggesting that CLIP was efficiently processed. In sharp contrast, Furin;PC7 DKO embryos retained both YFP and CFP fluorescence at the plasma membrane. Quantification at E3.0 in randomly chosen inner and outer cells thus revealed a significantly elevated CFP/YFP ratio of 1.13 in DKO compared with 0.73 in wild-type embryos, confirming a net decrease in PC activity. Also at E3.5, DKO embryos showed elevated CFP/YFP ratios, although CLIP cleavage progressively increased at this and during later stages (Fig. 2, B and C; Fig. S3 B; and Table S2). To test whether processing is partially rescued by Pace4, we crossed the CLIP transgene into a TKO mutant background. TKO embryos retained significantly more CFP fluorescence in all cells than either wild-type or DKO embryos both at E3.0 and E3.5 (Fig. 2, A and B; and Table S2; n = 8/8), and even though they were not viable at E4.5, they did not decrease their CFP/YFP ratios until at least E3.5 (Fig. 2 C; P = 0.28). In good agreement, similarly elevated CFP/YFP ratios were observed until E4.5 in normal embryos expressing the cleavage mutant biosensor CLIPm (Fig. S3 C), supporting the idea that no other PCs are active at morula and blastocyst stages. Cleavage by maternal pools also seems improbable given that TKO embryos were unable to rescue CLIP processing or compaction at the morula stage. Thus, we conclude that a net increase in PC activity in DKO mutants is specifically mediated by Pace4 at E3.5.

To assess the spatial distribution of individual PC activities, we compared CLIP processing in embryos cultured with or without the inhibitor deaconyl-Arg-Val-Lys-Arg-CMK. Concentrations of 5–75-µM CMK inhibit intracellular cleavage of the HIV gp160 precursor in Jurkat cells by 50–75% (Garten et al., 1994), and 10-µM CMK abolishes CLIP processing in HEK293T cells (Mesnard and Constam, 2010). Although a higher dose (20 µM) was toxic for uncompacted E2.5 wild-type embryos, incubation with 10-µM CMK within 30 h increased the CFP/YFP ratio of CLIP only 1.6-fold and only at the surface of ICM but not TE cells. However, significantly higher ratios were observed in DKO embryos even without CMK treatment (2.2- and 2.0-fold higher than the wild type in inner and outer cells, respectively; Fig. 3, A and B; and Table S3; n = 8/8 in each condition), suggesting that CMK efficacy was modest in the wild type and limited to inner cells. To address why CMK was so impotent especially in the TE lineage, we asked whether its efficacy can be unmasked in a PC mutant. Indeed, PC7^−/− embryos that were treated with CMK accumulated more uncleaved CLIP than DKO mutants and often failed to cavitate (n = 3/5), indicating that CMK efficiently inhibited Furin activity in inner and outer cells (compare Fig. 3 [A and B] with Fig. 2 [B and C] and Table S3). In contrast, CMK-treated Furin^−/− mutants cavitated and cleaved CLIP similar to the wild type, at least in TE. Compared with CMK-treated DKO embryos, they also more efficiently processed CLIP in the ICM. These results suggest that PC7 is relatively insensitive to CMK compared with Furin. Surprisingly, CMK treatment and loss of Furin nonetheless synergized to inhibit CLIP cleavage in the ICM (Fig. 3 B and Table S3). Moreover, a comparison with TKOs revealed that this synergism persisted in DKO mutants independently of PC7. Thus, we conclude that loss of Furin leads to compensatory up-regulation of CMK-sensitive Pace4 activity in the ICM. In sharp contrast, CFP/YFP ratios in the outer cells of CMK-treated DKOs remained 1.5-fold lower than those in TKOs, indicating that Pace4 activity in the TE lineage is CMK resistant. To investigate whether CMK sensitivity in inner cells might reflect a more accessible distribution of Pace4 in postsecretory compartments, we directly monitored the expression of PCs by quantitative RT-PCR (qRT-PCR) analysis in ICM and TE cells that were immunoisolated from blastocysts. Whereas PC7 and Furin were ubiquitously expressed in both ICM and TE, Pace4 was specifically transcribed in TE (Fig. S4). Therefore, the CMK-sensitive fraction of Pace4 in the ICM likely corresponds to secreted activity after its release by the TE. Whether an additional PC activity is present earlier remains unknown because 10-µM CMK as well as empty vehicle (DMSO) treatments led to premature developmental arrest of TKO embryos before the appearance of CLIP expression (unpublished data). Overall, our findings demonstrate that Furin, PC7, and Pace4 are all active in inner and outer cells but that they are differentially inhibited in these compartments by 10-µM CMK, with Furin being blocked in both TE and ICM lineages, PC7 in none, and Pace4 only in the ICM.

Compaction and ICM formation depend on cell adhesion mediated by E-cadherin (Stephenson et al., 2012). Because compaction also required overlapping Furin, PC7, and Pace4 activities, we investigated whether E-cadherin–dependent fate decisions are impaired in DKO or TKO mutants. Triple labeling of late blastocysts (E4.0) with antibodies against the lineage markers Oct4 and Cdx2 and with the DNA stain Draq5 revealed similar proportions of ICM and TE cells in wild-type and DKO embryos (Fig. 4, A and B; and Table S4). Moreover, contrary to E-cadherin mutants (Stephenson et al., 2010), DKO embryos showed no ectopic Cdx2 staining in inner cells (n = 7/7). However, 10.0 ± 6.0% of their outer cells still coexpressed Cdx2 and Oct4, compared with only 0.6 ± 0.4% in the wild type, indicating that the TE is incorrectly specified. Interestingly, TKO embryos show a drastic increase of Cdx2-positive cells (Fig. 4, A and B; and Table S4), even though the total number of cells was indistinguishable from the wild type (Fig. 4 C). Furthermore, 49% of inner cells in TKO embryos (n = 6/6) showed ectopic Cdx2 staining at the expense of Oct4. In addition, similar to what we observed in DKO mutants, some Cdx2-positive cells in TKO embryos (7.0 ± 2.0%) failed to down-regulate Oct4 expression, confirming that PCs are required for the proper specification of TE cells. A net enrichment of TE relative to ICM cells also occurs in E-cadherin mutants, and this has been attributed to ectopic nuclear accumulation of YAP (Nishioka et al., 2009; Stephenson et al., 2010; Cockburn et al., 2013; Hirate et al., 2013). To evaluate whether this lineage decision requires functionally overlapping PC activities, we analyzed the distribution of YAP by immunostaining. Interestingly, TKO ectopically expressed nuclear YAP in almost all outer and inner cells (Fig. 4 D), indicating that ectopic outer cell fate correlates with ectopic nuclear YAP.

Because TKO embryos lose E-cadherin expression at the plasma membrane and phenocopy E-cadherin mutants, we hypothesized that overlapping activities of PC7, Furin, and Pace4 may be responsible for E-cadherin precursor processing. To test this, we assessed whether E-cadherin precursor cleavage can be inhibited by PC-specific siRNAs or by CMK treatment in cultured cells. Concurring with earlier studies, HepG2, HEK293T, and LoVo cells all express FURIN, PACE4, PC5/6, and PC7 mRNAs (Fig. 5 A; Hatsuzawa et al., 1990; Miranda et al., 1996; Seidah et al., 1996; Hallenberger et al., 1997). Furin is not functional, though, in LoVo cells (Takahashi et al., 1995). Western blot analysis of endogenous E-cadherin showed that whereas processing was diminished by up to 25% already at a low dose of 10-µM CMK, 40–80% remained uncleaved in all of the cell lines tested, even at a high dose of 100-µM CMK (Fig. 5 B). These data suggest that CMK only partially inhibits endogenous PC activity, or that various cell types activate E-cadherin partly by an unrelated protease. To distinguish between these possibilities, we chose HepG2 cells to test whether CMK treatment and depletion of FURIN or PC7 have additive effects. We found that previously validated siRNAs deplete FURIN or PC7 mRNAs or both by ∼60 and 40%, respectively, as described previously (Fig. S5; Scamuffa et al., 2008). Moreover, although knockdown of FURIN or PC7 alone had little effect on E-cadherin, combined knockdown reduced processing by >40% (Fig. 5 C). Quantitative PCR revealed no compensatory up-regulation of PACE4 mRNA expression (Fig. S3). Nevertheless, treatment of double knockdown cells with 10-µM CMK further reduced the residual processing to 10%. Also, when added with siPC7 alone, CMK reduced cleavage in an additive manner to <25%, whereas only a minimal additive effect of CMK was observed with siFURIN compared with CMK alone. These results confirm that Furin, but not PC7, is sensitive to inhibition by CMK in vivo. CMK treatment also inhibited cavitation and partially inhibited E-cadherin processing in PC7^−/−, but not in wild-type blastocysts (Figs. 5 D and 3 A). In addition, PC7 RNAi synergized with the genetic ablation of Furin and Pace4 to inhibit E-cadherin processing in mouse ES cells (Fig. 5, E and F). Overall, these results show that PC7 together with functionally overlapping activities mediates E-cadherin precursor cleavage both in vivo and in cultured mouse ES cells and transformed human cell lines.

PCs can process multiple secreted proteins in cell-based assays, but in vivo functions are ill defined. Here, a targeted deletion of PC7 unmasked functional overlap with Furin and Pace4 in regulating cell polarization and adhesion during morula compaction and in localizing YAP signals that mediate the segregation of the ICM from the TE. Live imaging using the transgenic biosensor CLIP revealed that all of these PCs were active both in inner and outer cells and that partial rescue of Furin;PC7 DKO embryos by Pace4 temporally coincided with an up-regulation of Pace4 activity after E3.0. Consistent with their new role in ICM specification, Furin and PC7 each also promoted the processing and cell surface localization of E-cadherin. Our findings suggest that the mechanism to segregate the pluripotent ICM from differentiating TE involves dynamic regulation of E-cadherin maturation by multiple PCs.

PC7^−/−;Furin^−/−;Pace4^−/− TKO embryos showed impaired segregation of ICM from TE, marked by ectopic nuclear YAP and Cdx2 expression in half of all inner cells. Interestingly, these ectopic Cdx2-expressing cells continued to accumulate peripheral to the remaining Oct4-positive ICM, possibly reflecting differential adhesiveness. Their shapes were less hexagonal and more round than in the wild type, at least on the side facing the outermost TE layer, pointing to loosened contacts especially with outer neighbors. In keeping with impaired cell adhesion, TKO mutants also failed to compact at the morula stage. When detached from the TE by immunosurgery, the ICM spontaneously regenerates outer fates (Handyside, 1978; Hogan and Tilly, 1978; Spindle, 1978). Even in cells that normally do not reside at the surface, severing adhesion might prompt cell flattening and TE differentiation by alleviating cortical tension (Dard et al., 2009). In addition, diminished adhesion is expected to influence cell polarization because surfaces that cannot engage E-cadherin in contacts with neighboring cells accumulate aPKC (Stephenson et al., 2010). Activation of aPKC in turn blocks the kinases Lats1/2 to enable the assembly of nuclear complexes of YAP with Tead4 and thereby induces TE differentiation and Cdx2 expression (Plusa et al., 2005; Yagi et al., 2007; Ralston and Rossant, 2008; Dard et al., 2009; Nishioka et al., 2009; Hirate et al., 2013). Thus, defective morula compaction, ectopic aPKC accumulation, and nuclear YAP and Cdx2 staining in inner cells of TKO mutants likely reflect impaired cell adhesion (Fig. 6).

The combination of impaired cell adhesion, delayed compaction, and disorganized apical-basal cell polarization seen in TKO mutants is reminiscent of embryos lacking E-cadherin. Adherens junctions are required for the assembly of tight junctions (Eckert and Fleming, 2008), and apical and basolateral membrane domains fail to separate in the absence of E-cadherin (Stephenson et al., 2010). We found that although DKO embryos retained basolateral E-cadherin localization in at least some outer cells, assembly of tight junctions marked by ZO-1 staining and cavity formation were clearly perturbed. Furthermore, TKO embryos mislocalized E-cadherin to intracellular vesicles, and both outer and inner cells devoid of E-cadherin accumulated nonapical expression of aPKC. These observations agree with the notion that E-cadherin is essential to maintain epithelial cell polarity (Fig. 6 A; Stephenson et al., 2012). E-cadherin also mediates adhesion of filopodia and their myosin-induced traction on neighboring blastomeres of eight-cell stage embryos to initiate morula compaction (Fierro-González et al., 2013). Possibly, compaction of DKO embryos is delayed because of impaired adhesion of such filopodia.

Proteolytic maturation enables the extracellular domain of E-cadherin to engage in homotypic interactions (Ozawa and Kemler, 1990). Here, we found that combined depletion of FURIN and PC7 by RNAi in human HepG2 hepatoma cells reduced E-cadherin processing by almost 60% and that the remaining activity dropped to 10% upon addition of the PC inhibitor CMK. E-cadherin cleavage also decreased in CMK-treated cells transfected with PC7 siRNA alone, whereas FURIN siRNA alone failed to synergize with CMK. A role of endogenous Furin in E-cadherin precursor processing thus is masked in part by a CMK-resistant pool of PC7. Our discovery that PC7 and Furin respond differently to CMK provides an important simple criterion to assess their relative contributions to the processing of shared substrates. We also transfected siRNAs against PACE4 or PCSK5, but these knockdowns were inefficient, precluding quantification of a potential synergism (unpublished data). As these PCs are present, one or both may contribute at least to the CMK-sensitive pool of E-cadherin processing activity in HepG2 cells. CMK treatment also reduced precursor processing in PC7^−/− embryos and further delayed compaction of Furin^−/−;PC7^−/− DKO mutants, even though it failed to do so in wild-type or Furin^−/− morulae. Thus, Furin and CMK-resistant PC7 activities acting together with at least one partially CMK-sensitive convertase process E-cadherin both in vitro and in vivo. In good agreement, inactivation of three PCs was necessary to abolish E-cadherin cell surface localization in vivo. Because proteolytic maturation is essential for E-cadherin cell surface localization (Geng et al., 2012), it is plausible that PC7, Furin, and Pace4 enable TE cell polarization and ICM specification by jointly regulating E-cadherin cleavage.

We also mapped the distribution of PC activities and their sensitivity to CMK in vivo by CLIP imaging. Cleavage of CLIP in mouse blastocysts is enhanced by redundant Furin and Pace4 activities, but the time of onset of cleavage and a potential role of PC7 had not been analyzed (Mesnard and Constam, 2010; Mesnard et al., 2011). We found that the compaction and cavitation defects of Furin^−/−;PC7^−/− mutants correlated with a severe inhibition of CLIP cleavage at the morula stage and that partial rescue of these processes coincided with a marked up-regulation of CLIP processing by Pace4 activity at E3.5. Moreover, CFP/YFP fluorescence ratios reached 1.6-fold higher values in inner than in outer cells only in DKO but not in wild-type embryos, and this difference was lost in TKO mutants, indicating that Pace4 was most active in outer cells. In keeping with this conclusion, zygotic Pace4 expression has only been observed in the TE lineage and is stimulated by the TE-specific transcription factor Elf5 (Donnison et al., 2005; Mesnard et al., 2006). Accordingly, Pace4 activity in the ICM may be cell nonautonomous (Mesnard et al., 2011). In keeping with this interpretation, only the ICM-associated pool of Pace4 activity was inhibited by CMK. In sharp contrast, Furin and PC7 were blocked by CMK either in all cells or in none, respectively (Fig. 6 B). Importantly, CMK inhibits all of these PCs in cell-free assays, with K_i values of 0.1, 2.0, and 3.6 nM for PC7, Furin, and Pace4, respectively (Jean et al., 1998). CMK treatment also blocks extracellular Pace4 and sheds Furin activities in conditioned medium of ES cells (Mesnard and Constam, 2010). Together, these observations suggest that CMK-resistant PC activities reside in intracellular compartments that are less accessible than the CMK-sensitive pools of Furin and Pace4. In transfected cells, intracellular PC7 can rapidly enter post-Golgi compartments without passing through the trans-Golgi network (Rousselet et al., 2011). However, the trafficking itineraries of endogenous PC7 or Pace4 are unknown because of the lack of suitable antibodies. Our live imaging suggests that the abodes of these and other PC activities may be determined in future studies using compartment-specific CLIP variants to elucidate exactly where specific substrates are processed.

Does the enrichment of Pace4 activity in outer cells differentially modulate adhesive properties of E-cadherin compared with inner cells? Cells that lost contact with their neighbors in Furin^−/−;PC7^−/− DKO embryos were always positioned at the periphery. These could be inner cells that normally do not express Pace4 and therefore failed to maintain their adherens junctions with ICM neighbors (Fig. 6 A). However,unlike TKO or CMK-treated DKO mutants, DKO mutants cultured without CMK largely maintained Oct4 expression and diminished the cleavage of CLIP in inner cells by only 50% compared to the wild type, suggesting that their CMK-sensitive pool of Pace4 activity substantially rescued adhesion even within the ICM. In contrast, in outer cells, Pace4 alone may fail to reliably maintain contacts because of their epithelialization. Polarized epithelial cells secrete Pace4 basolaterally, whereas a shed form of Furin is released on the apical side (Mesnard et al., 2011), indicating that progressive cell polarization and its maintenance by E-cadherin may eventually segregate these PCs into distinct compartments and thereby limit their functional redundancy. Moreover, because delivery of E-cadherin to adherens junctions in polarized cells involves Rab11 recycling endosomes (Golachowska et al., 2010), alternative PCs may ensure efficient cleavage along this unique transport route. Recent studies revealed that only outer cells also activate Notch-1, an unrelated PC substrate acting in parallel to Tead4 to reinforce Cdx2 expression (Logeat et al., 1998; Rayon et al., 2014). However, if cell adhesion or Notch-1 processing are differentially regulated in outer compared with inner cells because of Pace4, redundant mechanisms must exist that maintain both cell fates in Pace4^−/− and other PC single mutants.

The spatiotemporal regulation of PC activities revealed by quantification of CLIP imaging at the morula stage suggests a role in modulating dynamic changes in cell adhesiveness and in the specification or maintenance of distinct cell fates. Interestingly, down-regulation of Furin and concomitant up-regulation of a secreted soluble isoform of PC5 in cancer cell lines from diverse tissues favor alternative cleavage of the related N-cadherin at a more distal site in the extracellular domain, thereby inhibiting instead of activating adhesive properties (Maret et al., 2012). It will therefore be important to investigate how dynamic changes in activity levels or subcellular localization of PCs influence epithelial–mesenchymal transitions and their reversion during cancer metastasis. Our finding that multiple overlapping PC activities can moonlight for one another to process E-cadherin suggests that future therapies targeting oncogenic PC activities individually should spare the function of this essential tumor suppressor.

A PC7 gene-targeting vector was derived from a 129 SvJ genomic phage library by flanking an inverted MC1-neomycin resistance cassette with 5′ and 3′ homology arms comprising a 4-kb NdeI–SalI and a 3.6-kb XbaI fragment of the Pcsk7 locus. The linearized vector was electroporated into 129 Sv/Ev CCE ES cells. Deletion of a 6.2-kb SalI–XbaI fragment encoding the translational start site, the pro region, and the entire catalytic domain of PC7 by homologous recombination gave rise to the allele Pcsk7^tm1Rob. When injected into C57BL/6 blastocysts, two targeted ES clones gave rise to male germline chimeras that were backcrossed to C57BL/6 females. Intercrosses between heterozygous F1 progeny yielded fertile adult homozygotes without any overt abnormalities at a Mendelian frequency. To detect homologous recombination, genomic DNA samples were digested with XbaI and analyzed by Southern hybridization. A 0.9-kb KpnI–XbaI probe upstream of the integration site detected a 12-kb fragment in the wild type and a mutant 5.7-kb fragment. Homologous recombination was verified using a 0.5-kb XbaI fragment 3′ of the integration site, which detected 12-kb or 6.2-kb KpnI fragments corresponding to the wild-type or mutant alleles, respectively (Fig. S1 A). PCR genotyping was performed using the reverse primer 5′-AGCCATCATAGTTGCAGTTGTCATTGTG-3′ in conjunction with the forward primer 5′-CCTCTTCCCATCCCGCCTTCATTA-3′ and the neomycin-specific primer 5′-GGCTGCAGGGTCGCTCGGT-3′ at an annealing temperature of 59°C to simultaneously amplify a 652-bp and a 478-bp fragment of the wild-type and mutant alleles, respectively. Animals were maintained in individually ventilated cages at the specific pathogen-free animal facility of Ecole Polytechnique Fédérale de Lausanne (EPFL). Experiments were performed in accordance with Swiss guidelines for the care and use of laboratory animals.

PC7^+/− heterozygous males on a mixed C57BL/6 × 129 SvEv background were backcrossed for >10 generations to C57BL/6 females. PC7^+/− animals were backcrossed for at least four generations to Furin^tm1Ajmr (Roebroek et al., 1998) and Pcsk6^tm1Rob heterozygotes (Constam and Robertson, 2000) on an outbred NMRI background with the CLIP transgene (Mesnard and Constam, 2010). Furin^tm1Ajmr contains a 3-phosphoglycerate kinase promoter–hygromycin phosphotransferase and a stop codon that disrupts the Furin coding sequence in exon 4. In the Pcsk6^tm1Rob allele, an MCl-neomycin phosphotransferase cassette was inserted by homologous recombination to delete the coding region for the catalytic domain of Pace4. Compound mutants were intercrossed to obtain double or triple homozygous mutant embryos. PCR primers used for genotyping are described in Table S5, Table S6, and Table S7.

The indicated human cell lines were cultured in DMEM containing 10% FBS and 1% penicillin/streptomycin (Gibco). Wild-type and mutant mouse ES cells were derived from blastocyst outgrowths in DMEM medium, 10% FBS, 5% knockout serum replacement (Invitrogen), and 30% ES cell–conditioned medium (Beck et al., 2002) and maintained on neomycin- and hygromycin-resistant STO fibroblasts in DMEM containing 15% FBS, β-mercaptoethanol, and 10³ units of leukemia inhibitory factor. For stable transfection, mutant ES cells were electroporated with the mouse Furin coding sequence containing a FLAG epitope after the autocatalytic cleavage site and an EF1-α promoter followed by an IRESpac cassette for puromycin selection (Beck et al., 2002). Specific siRNAs against PCs (Scamuffa et al., 2008) were purchased from Ambion and transfected using Lipofectamine 2000 (Invitrogen). ES cells were transfected using Lipofectamine 2000 in feeder-free conditions. Embryos were flushed at E2.5–2.75 in PBS supplemented with 10% FBS, washed twice in equilibrated potassium-supplemented simplex optimized medium (KSOM; EMD Millipore), and cultured with intact zona pellucida in KSOM at 37°C and 5% CO₂ until the stage of interest. Where indicated, decanoyl-RVKR-CMK (Tocris Bioscience) or empty vehicle (DMSO) was added at the indicated concentrations and replaced every 12 h. All experiments were repeated on more than three littermates.

Live embryos were scanned in a 37°C/5% CO₂ chamber in individual microdrops of KSOM on glass-bottomed microwell dishes (MatTek Corporation) at high speed (700 Hz) on a confocal microscope (SP2 or SP5; Leica) with Leica software (LAS AF 2009). All embryos were scanned using the same objective (HC Plan Apochromat 20×/0.70 NA; immersion), with a pinhole of 1 airy unit, a voxel size of 75 nm, 25% laser intensity, and number of z section. The gain of the photomultiplier was constant at any given stage but adapted to increasing CLIP expression levels at late compared with early stages. Heat maps were imaged using a fluorescence resonance energy transfer RatioQuantification plugin developed by the EPFL bioimaging platform. The YFP channel served as the reference channel and to define signal thresholds. Corresponding images of fluorescence ratios were colored using the Rainbow RGB.lut and a median filter of 1.0 pixel. For CFP/YFP ratio quantification, 7–15 random membrane regions were selected per embryo and quantified using Fiji. Each region of interest (ROI) was drawn on a magnified YFP channel and applied on a CFP channel. The mean of YFP and CFP background fluorescence was measured on 50 ROIs from 10 nontransgenic embryos scanned with the same conditions as mentioned before. After subtraction of this background, CFP and YFP signals of CLIP in individual ROIs were each divided by the gain of photomultiplier for signal normalization. The resulting corrected CFP values were then divided by their corresponding corrected YFP values in every ROI to obtain individual CFP/YFP ratios. Mean CFP/YFP ratios of up to 100 ROIs are displayed in arbitrary units (Table S2). For quantification of inner versus outer cells, ROIs in outer cells were drawn at cell–cell contacts between two outer cells and at membranes at the embryo surface, whereas ROIs in inner cells were drawn at cell–cell contacts between two inner cells. Cell–cell contacts between outer and inner cells were not considered in these experiments. All experiments were repeated on more than three littermates.

Embryos were fixed in 4% PFA, washed three times in PBS containing 0.1% Triton X-100, and incubated in 10% FBS for 1 h. Primary antibodies were incubated overnight at 4°C. Antibodies included anti-Oct4 (sc-8628; Santa Cruz Biotechnology, Inc.), anti-Cdx2 (MU392A-UC; Biogenex), anti-Yap (4912S; Cell Signaling Technology), anti-aPKC (sc-216; Santa Cruz Biotechnology, Inc.), anti–E-cadherin (U3254; Sigma-Aldrich), and anti–ZO-1 (sc-10804; Santa Cruz Biotechnology, Inc.). Embryos were washed three times in PBS/0.1% Triton X-100, and secondary antibodies coupled with Alexa Fluor 488, 568, and 647 (Jackson ImmunoResearch Laboratories, Inc.) and Draq5 were incubated at room temperature for 2 h, washed three times with PBS/0.1% Triton X-100, and analyzed at room temperature in individual microdrops of PBS/0.1% Triton X-100 on glass-bottomed microwell dishes at room temperature (19–21°C) using the confocal microscopy procedure as described in the previous section (Bessonnard et al., 2014). Images were prepared using Fiji and Gimp 2.8 software. Quantification of fluorescence was done as previously described for CFP/YFP ratiometric analyses. All experiments were repeated on more than three littermates.

Total mRNA was isolated using TRIzol, reverse transcribed using Superscript III (Invitrogen), and analyzed using SYBR green PCR Master Mix (Applied Biosystems) coupled with 7900ht Fast RT-PCR system (Applied Biosystems). Human qRT-PCR primers were designed using Primer Blast (National Center for Biotechnology Information) and are described in Table S6. Proteins were extracted in PBS supplemented with 1% (vol/vol) Triton X-100, 1-mM EDTA, and protease inhibitor cocktail (Roche). E-cadherin protein was analyzed by Western blotting using rat anti–E-cadherin antibody (U3254; Sigma-Aldrich), anti–rat IgG-HRP secondary antibody (Jackson ImmunoResearch Laboratories, Inc.), and Western Lightning ECL kit (Thermo Fisher Scientific). E3.5 wild-type embryos were flushed in M2 medium (Sigma-Aldrich). Zona pellucida was removed with Tyrode’s acid (Sigma-Aldrich) and incubated for 10 min with anti–mouse antibody (Sigma-Aldrich) at 37°C and with guinea pig complement serum at 37°C for 30 min (Sigma-Aldrich). Three embryos or six resulting TE/ICMs were pooled together, and mRNA was isolated using TRIzol (Nishioka et al., 2009), reverse transcribed using Superscript VILO (Invitrogen), and analyzed using SYBR green PCR Master Mix. Mouse qRT-PCR primers (Table S7) were designed using Primer Blast.

Statistical tests were performed as described using Prism (GraphPad Software). The normal distribution of CFP/YFP ratio values and ICM/TE distribution were verified with the Shapiro-Wilk normality test, and results were analyzed by a Mann-Whitney or by a Student’s t test depending on previous results. Mendelian distribution was analyzed using a χ² test.

Fig. S1 shows the knockout strategy in the PC7 locus. Fig. S2 shows E-cadherin/aPKC expression profiles in cultured or fresh embryos. Fig. S3 shows CFP/YFP ratios of CLIPv0 and CLIPm transgenes in several mutant embryos. Fig. S4 represents results obtained with embryo immunosurgery combined with qRT-PCR. Fig. S5 shows knockdown efficiencies in HepG2 cells. Table S1 recapitulates the cavity formation during development in wild-type, DKO, and TKO embryos. Table S2 summarizes CFP/YFP ratios at E3.0, E3.5, and E4.5. Table S3 summarizes CFP/YFP ratios in ICM or TE cells in control conditions or after CMK treatment (10 µM). Table S4 recapitulates the percentage of ICM/TE cells observed at E4.0 in wild-type, DKO, and TKO embryos. Tables S5, S6, and S7 correspond to the list of primers used for genotyping and gene expression profiling. Additional data are available in the JCB DataViewer at http://dx.doi.org/10.1083/jcb.201503042.dv.

We would like to thank the animal core facility of the EPFL School of Life Sciences for help with the maintenance of transgenic mice. We also thank the EPFL bioimaging core facility for advice on image acquisition and analysis, Dr. Michel Cohen-Tannoudji and Sandrine Vandormael-Pournin (Paris) for advice on ES cell manipulations, and Dr. Claire Chazaud (Clermont-Ferrand) and members of the Constam laboratory for comments on this manuscript.

This project was supported by the Swiss National Science Foundation (grant 3100A0-118080/1) and by EU FP6 MEMORIES Project (grant no. 037831) to D.B. Constam.

The authors declare no competing financial interests.

Author contributions: D. Mesnard and D.B. Constam conceived and designed initial experiments. D. Mesnard performed experiments on DKO embryos and human-derived cells and initiated biosensor analysis. S. Bessonnard performed experiments on TKO embryos, ES cells, immunosurgery, and fluorescent quantification. S. Bessonnard, D. Mesnard, and D.B. Constam analyzed the data and wrote the manuscript.