Phenotypical Analysis of Atypical PKCs In Vivo Function Display a Compensatory System at Mouse Embryonic Day 7.5

Background The atypical protein kinases C (PKC) isoforms ι/λ and ζ play crucial roles in many cellular processes including development, cell proliferation, differentiation and cell survival. Possible redundancy between the two isoforms has always been an issue since most biochemical tools do not differentiate between the two proteins. Thus, much effort has been made during the last decades to characterize the functions of aPKCs using gene targeting approaches and depletion studies. However, little is known about the specific roles of each isoform in mouse development. Methodology/Principal Findings To evaluate the importance of PKCι in mouse development we designed PKCι deletion mutants using the gene targeting approach. We show that the deletion of PKCι, results in a reduced size of the amniotic cavity at E7.5 and impaired growth of the embryo at E8.5 with subsequent absorption of the embryo. Our data also indicate an impaired localization of ZO-1 and disorganized structure of the epithelial tissue in the embryo. Importantly, using electron microscopy, embryoid body formation and immunofluorescence analysis, we found, that in the absence of PKCι, tight junctions and apico-basal polarity were still established. Finally, our study points to a non-redundant PKCι function at E9.5, since expression of PKCζ is able to rescue the E7.5 phenotype, but could not prevent embryonic lethality at a later time-point (E9.5). Conclusion Our data show that PKCι is crucial for mouse embryogenesis but is dispensable for the establishment of polarity and tight junction formation. We present a compensatory function of PKCζ at E7.5, rescuing the phenotype. Furthermore, this study indicates at least one specific, yet unknown, PKCι function that cannot be compensated by the overexpression of PKCζ at E9.5.


Introduction
The protein kinase C (PKC) family of serine-threonine kinases consists of 9 different genes giving rise to at least 12 isoforms, subdivided into 3 subfamilies. The subdivision is based on sequence homology as well as dependency on cofactors during their activation process. Thus the classical PKCs (cPKC: a, b and c) are dependent on Ca 2+ ions and diacylglycerol for their activation whereas the novel PKCs (nPKC: d, e, g and h) based on the lack of a binding domain are Ca 2+ ion independent but still require diacylglycerol for their activation. In sharp contrast, the subfamily of atypical PKCs (aPKC: f and i/l) are independent of Ca 2+ ions and diacylglycerol for their activation, thereby, representing quite a distinguishable subgroup among the PKC family. This becomes even more pronounced by the fact that aPKCs are unresponsive to Tumor Promoting Agent 12-O-Tetradecanoylphorbol-13-Acetate (TPA) treatment, an early described feature of PKC family members (see review [1]). Nevertheless, all PKC family members (including aPKCs) have been described to participate in a multitude of signaling pathways involved in cell growth, differentiation and apoptosis.
In 1989 Nishizuka and coworkers isolated the PKCf isoform from a rat cDNA library [2]. Four years later a second aPKC isoform (PKCi) was isolated and characterized [3] followed by the identification of the mouse ortholog PKC? [4]. As true for all other PKCs, the aPKC proteins can be divided into a regulatory Nterminal region and a catalytic C-terminal region, separated by a hinge region. The exception is PKMf which represents only the C-terminal kinase domain generated by an internal promotor whose activity has been assigned to neurons [5]. In contrast to other PKCs the N-terminal domain contains a Phox/Bem1 (PB1) motif mediating interaction with p62 [6,7]and other signaling molecules like Mek5 [8] and Par6 [9] all of which are supposed to mediate aPKC signaling. In addition both aPKC also contain a cystein-rich zinc-finger like domain within the regulatory Nterminal domain defined as C1. Whereas all other PKCs possess a tandem repeat, aPKCs possess only one C1 domain. Interestingly, this domain accounts for the binding of diacylglycerol and TPA in classical and novel PKCs. The aPKC C1 domain has been reported to bind directly to phosphatidylinositol (3,4,5)-trisphosphat thereby inducing conformational changes in the protein leading to activation similar to diacylglycerol binding to other PKCs [10]. But also other interacting partners, inhibitory as well as activating, have been described [11,12]. The aPKC C-terminal part represent the catalytic kinase domain sharing 86% homology to each other but only 45-55% to other PKCs. Overall both aPKCs show a 72% homology on amino acid level [4]. Due to high degree of homology and the limited availability of isoformspecific tools the in vivo analysis of isoform-specific aPKC functions remained insufficient in mammals. Nonetheless, it has been shown that aPKCs are conserved in numbers of organisms, including C. elegans [13] and D. melanogaster [14] in which only one isoform was detected.
We have previously shown that both aPKCs are expressed during mouse embryonic development [15] as well as in distinct domains in the adult mouse brain [16]. As a conclusion of these studies PKCi was defined as being ubiquitously expressed whereas PKCf expression was pronounced in lung, kidney and brain. The spectrum of physiological processes linked to aPKCs function is huge and covers cell proliferation [17], cell polarity [18], carcinogenesis, neurogenesis [19] and many more. Attempts to investigate individual in vivo functions also made use of the gene targeting approach. We and other have generated aPKC deficient mouse lines which were subjected to various phenotypical investigations. Interestingly, the phenotype of the conventional PKCf knockout did not display the expected phenotype during mouse preimplantation development [20]. Early studies using the conventional PKCf knockout revealed a functional link to NFkB signaling [21,22]. Subsequent studies also identified PKCf to act as tumor suppressor due to its regulatory function on the IL-6 promotor [23]. In sharp contrast a conventional aPKCi knockout displayed an embryonic lethal phenotype [24] (personal communication with Shigeo Ohno and own data presented within this publication), clearly distinguishing both aPKC isoforms for the first time. Later studies using the conditional gene targeting approach showed inter alia specific in vivo functions for PKCi but not for PKCf in muscle and podocytes [25,26]. Atypical PKCs have been described to form complexes with the partition defective proteins Par6 and Par-3 [27]. Par genes have been cloned and characterized in 1995 [28] and were shown to be crucial for asymmetric cell division in C. elegans and other organisms [29,30]. The association of aPKCs to this ternary complex, also called polarity complex [31] integrates aPKC signaling into all aspects of polarity without any isoform specificity.
Here, we subjected several established mutant mouse lines of the PKCi gene to a thorough developmental investigation. We present a detailed description of the embryonic lethal phenotype caused by the PKCi deficiency and provide more insights into the redundancy within the aPKC subfamily.

Generation of 2 Different Prkci Knock Out Alleles in Mice
To investigate PKCi in vivo function, we decided to generate various mutant alleles following standard gene targeting approaches in mouse embryonic stem cells (ESCs). First we generated a conventional Prkci knock-out allele by inserting a neo cassette into the second exon of the gene causing a disruption of gene transcription (PKCiNeo, details see Figure S1A and material & methods). Since homozygozity of this allele caused embryonic lethality (see later), we decided to generate a conditional allele for Prkci using the Cre/loxP system in addition. For this approach the Prkci gene was targeted by inserting a single loxP site 59prime of the 2nd exon followed by a floxed neo cassette 39prime of the exon. This so-called floxed allele (PKCiflox/flox) has been described earlier and its functionality has previously been shown [25,26,32,33]. In this study we subsequently crossed PKCiflox/ flox mice with a transgenic mouse line [34] expressing the Cre recombinase in the germ line, in a ubiquitous fashion, starting at embryonic day (E) 13.5 and obtained a second null allele for Prkci named PKCiD(details see Figure S1B).
To investigate whether homozygozity of Prkci null alleles (PKCiNeo/Neo and PKCiD/D) are compatible with embryonic development, we analyzed more than 100 offspring from matings of heterozygous carries for each line at 3 weeks after birth. For both individual lines we did not observe any homozygous offspring whereas heterozygous (66%) and wt (33%) mice were detected (with an average of 8 pups per litter). These findings implied an embryonic lethal phenotype for both mutant alleles in agreement with earlier data published by Hedrick and coworkers [24]. In addition we also havent been able to observed any abnormalities in heterozygous animals within the first three month after birth. We next examined litters from heterozygous crosses at different developmental stages to identify the time point when first morphological abnormalities manifested in homozygous embryos. At embryonic day (E) 6.5, no morphological differences between litter-mates were apparent. At day E7.5, morphological inspection recognized a group of embryos which appeared ''compressed'' along the proximal-distal axis of the egg cylinder and harbored a small amniotic cavity ( Figure 1A). Histological analysis showed that the epiblast was present but was thickened compared to nonaffected litter-mates. Genotyping of individual embryos confirmed that this phenotype was restricted to homozygous embryos (25% of the investigated embryos of both transgenic lines showed this morphological abnormalities at this stage). At E8.5, mutant embryos were severely growth retarded. An anterior-posterior polarity of the body axis was not readily apparent since head folds and a linear heart tube distinguishing the anterior end of the embryo and the allantois representing the most posterior aspect were reduced or lacking ( Figure 1B). In contrast to control embryos that harbored 4-6 somites at this stage, somites were not discernible by histological analysis in the mutants. By E10.5, mutant embryos were reabsorbed by the mother. Since phenotypic changes were indistinguishable between the two targeted alleles, we decided to exclusively use the PKCig/g line for further analysis.  In wt embryos, Otx2 expression marks the forebrain [35], Pax2 the midbrain [36] and Fgf8 the midbrain/hindbrain boundary [37]. In PKCi-deficient embryos all three genes were expressed in the ''anterior'' embryonic region but the domains were compressed indicating that the (neuro-) ectoderm is patterned along the anterior-posterior axis but the expansion of the different primordial brain segments is compromised. Fgf8 is additionally expressed in the primitive streak at the posterior end of the embryo overlapping with Brachyury (T) that additionally marks the notochord, a thin line of cells extending from the forebrain/ midbrain border to the primitive streak underneath the neural tube [37,38]. Expression of both genes was established in the mutant but the primitive streak appeared much reduced and the notochord was disrupted. This shows that PKCi mutant embryos are polarized along the anterior-posterior body axis but that differentiation of the axial mesoderm is impaired. Tcf15 (also known as paraxis) marks the paraxial mesoderm [39] which was organized into 4-8 distinct somite pairs at this stage in the wt. In PKCi-deficient embryos a domain of Tcf15 expression was found in the middle region of the embryos but individual somites were not formed. Expression of Bmp4 in the lateral plate mesoderm, the extraembryonic mesoderm and the extraembryonic ectoderm was found in mutant embryos as in the wt. Tbx20 was expressed in the allantois, a derivative of the extraembryonic mesoderm, and in the linear heart tube, which derives from the cardiac subregion of the anterior lateral plate mesoderm in the wt [40]. Again, both expression domains were established at opposite poles of the mutant embryos but neither domain was appropriately developed into an allantois. Finally, Gja, that is expressed in the cardiac endoderm and dorsal aorta in wt embryos [41] was restricted to cell clusters at the anterior and posterior pole of the mutant embryos but clearly lacked an organization into vessel like structures. Together, this analysis suggests that PKCi-deficient embryos exhibit a normal anterior-posterior polarization of their main body axis. Mesoderm formation is initiated but comes to a premature halt, mesoderm differentiation into axial, paraxial, lateral and extraembryonic subtypes occurs but subsequent formation of tissues and organs completely fails. As a consequence PKCi-deficient embryos die due to lack of a cardiovascular system.

Embryoid Body Formation is Impaired by PKCi Deficiency
ESCs, when cultured as aggregates, form spherical structures which are defined as embryoid bodies (EBs). These structures are thought to recapitulate early steps of the pre-implantation development including endoderm formation, basement membrane (BM) assembly, epiblast polarization and subsequent cavity formation [42]. When cultured in suspension for 5 days, EBs form epithelial cysts consisting of an outer endoderm and an inner columnar epiblast epithelium (CEE), separated by a BM. As PKCi mutants displayed an abnormal amniotic cavity at E7.5 we decided to use the embryoid body formation assay to analyze cavity formation in more detail.
When wt and PKCi deficient ESCs [43]were subjected to this assay, we detected obvious differences among the two genotypes. Wt EBs displayed the expected appearance in a phase contrast representation, PKCi D/D EBs showed a disorganized structure (Figure 3 B). Most of the wt EBs formed a single large cavity enclosed by CEE after 5 days, whereas the mutant EBs did not form a single cavity at all or formed multiple small cavities adjacent to the polarized CEE, although the endoderm differentiated normally. When cultured for additional 2 days, most mutant EBs formed small cavities but failed to complete cavitation ( Figure 3A, B and C). Since caspase-dependent apoptosis has been described as a key mechanism involved in the early steps of cavity formation [42] we immunostained 5-day EBs for cleaved (activated) caspase-3. Wt EBs displayed massive central apoptosis as evidenced by cleaved caspase-3 staining (Figure 3 D). By contrast, only scattered apoptotic cells were detected in PKCi deficient EBs. DAPI staining of condensed and fragmented nuclei, another hallmark of apoptosis, also showed reduced apoptosis at the center of mutant EBs ( Figure 3D). To further analyze the apoptosis during EB cavitation, we cultured EBs of both genotypes for 2 to 5 days and performed immunoblotting for cleaved caspase-3. Indeed, PKCi deficient EBs expressed less apoptotic activity at all time points tested than the corresponding wt extracts ( Figure 3E). These results demonstrate that PKCi deficient EBs fail to form a single cavity possibly due to reduced central apoptosis.
Gene targeting experiments have shown that laminin-mediated BM formation is essential for primitive ectoderm epithelialization and cavitation in EBs [44,45]. The laminin a-1 chain is a key component of the embryonic BM expressed early on during development [46]. In the wt EB the laminin a-1 staining formed a continuous thin layer between endoderm and epiblast cells ( Figure 3A). We did not observe obvious changes in BM assembly in mutant EBs. These findings were confirmed by using perlecan as an alternative marker for the BM ( Figure 3F and G). However, ectopic BM formation was detected at the center of ,20% mutant EBs ( Figure 3F). This may be caused by failure in endodermal cell migration to the EB surface. Nonetheless, these results suggest that the cavitation defect in PKCi deficient EBs is unlikely caused to ectopic BM assembly.
Atypical PKCs have been functionally linked to the establishment and maintenance of epithelial polarity and tight junction formation due to their association with Par-3 and Par6 (forming the so-called polarity complex). To determine if epiblast polarity is disrupted in the absence of PKCi, we immunostained EBs for the apical marker MUPP1, Par-3, the adherens junction protein Ecadherin and the Golgi marker GM130. The apical actin belt was visualized using rhodamine-phalloidin. As shown in Figure 3A, Factin staining was detected as a continuous layer at the most inner cell membrane surrounding the single cavity in wt EBs. In mutant EBs, no such staining was detectable. The apical markers MUPP1 and Par-3 were expressed at the same location ( Figure 3F). The Golgi marker GM130 was positioned away from the BM and localized on the apical side. Similarly, E-cadherin was enriched at the apex of the epiblast where they formed AJs. These results provide evidence that polarity of the epiblast was intact.
Since the embryo and the EB express both PKCi and PKCf, we were wondering if the remaining PKCf could possibly compensate for some PKCi function. Therefore, we immunostained aPKC using a pan aPKC antibody in wt and mutant EBs. We were able to detect the protein at the apex of the epiblast of both genotypes ( Figure 3G). Whether the signal in the wt reflects a combination of both aPKC isoforms is unclear but for sure the observed signal in the mutants represents PKCf. In addition we also carried out a western blot analysis for aPKCs using protein extracts from EBs after 4 and 5 days of differentiation. The aPKC signal (ergo PKCf in the PKCi deficient background) appeared slightly up-regulated at 4 days in PKCi deficient extracts whereas after 5 days it became down-regulated when compared with wt EBs ( Figure 3H). Taken together these results suggest that PKCf may compensate for the loss of PKCi to confer an apical-basal polarity to epithelial cells in the epiblast.

Immunohistochemical and Electron Microscopical Investigation Revealed Subtle Changes in the Overall Organization of PKCi Deficient Embryos
Since PKCi deficient EBs lacked apoptosis, we wished to evaluate the contribution of this process in the mutant embryos as well. However, we failed to detect increased apoptosis by cleaved (activated) caspase-3 staining nor did we detect changes in cell proliferation using a phospho-specific histone 3 antibody between E7.5 embryos of either genotype ( Figure S2).
We next analyzed whether the embryonic ectoderm of E7.5 mutant embryos harbored epithelial cells with normal apical-basal polarity and junctional complexes. Again we used an antibody against E-cadherin as a marker for adherens junctions (AJs). As shown in the overview and the higher magnification ( Figure 4A and C) both genotypes displayed a clear signal at the apical side of the embryonic ectoderm. This indicated that polarity in the epithelial cell layer was established. However, in mutant embryos the signal did not form a sharp thin line but was interrupted at the apical side ( Figure 4A and C). We then further applied occludin and zonula occludens-1 protein (ZO-1) specific antibodies as marker proteins for tight junctions (TJ). In both cases the wt showed a punctuated pattern at the apical pole as expected ( Figure 4C, white arrows). In PKCi mutants we detected signals for occludin at the apical pole of the epithelial cell layer, but weaker ( Figure 4C, open arrows). In contrast we were not able to detect any signal for ZO-1 at the apical side in mutants. Instead we observed a weak cytoplasmic background signal. A semiquantitative RT-PCR analysis of E7.5 embryos did not amplify a significant ZO-1 specific transcript. Western blot analysis did not detect ZO-1 protein in at E8.5 PKCi mutants either ( Figure S3). Thus we concluded that TJ exists in mutant embryos but that the composition of tight junction protein complexes is altered.
Based on our observation in EBs, we analyzed the aPKC content and localization in wt and mutant embryos. As shown in Figure 4B a commonly used antibody against aPKCs (Santa Cruz # sc121 C-20) stained primarily the apical pole of the embryonic ectoderm cell layer with a minor fraction also detected in the cytoplasm in the wt. In the mutant embryo the signal (representing PKCf) is maintained at the apical side. When applying a PKCi specific antibody [47] the signal in the wt appeared broader, indicating that PKCi is more widely distributed within the cells than indicated by the pan aPKC antibody, but still showing an accumulation at the apical membrane. The PKCi mutant embryo did not show staining proving the specificity of the antibody. Hence, both aPKCs showed overlapping localization at the apical pole of embryonic ectodermal cells at E7.5 suggesting possible compensatory mechanisms.
To further elucidate the structural organization of the cell-cell contacts, we analyzed wt and mutant embryonic tissue by using electron microscopy. Both genotypes showed a clear electron dense area at juxtaposed membranes located at the apical side defined as tight junctions ( Figure 5A). AJs and desmosomes (DSs) were detected without identifying any difference among the genotypes (see Figure 5A). However, when focusing on the overall appearance, we observed a general tendency of mutants to possess a looser contact zone between individual cells. This was indicated by more open gaps within the extracellular space when compared with wt (exemplary shown in Figure 5B). At the same time we observed in sections of mutants that F-actin normally localized in tight bundles in proximity to the membrane appeared less tight organized. In addition, the distance between the bundles and the outer membrane seemed to be larger ( Figure 5C). In summary the electron microscopic analysis provides further evidences for changes in the integrity of cell-cell contacts without PKCi.

PKCf is Able to Rescue the Early PKCi in vivo Function
PKCi and PKCf are structurally highly related, suggesting shared in vivo functions and redundancy. We tested this assumption by establishing a PKCi fRes/+ mouse line which expresses a chimeric PKCi/f protein. For this purpose we established a targeted (knockin) allele haboring an inserted PKCf cDNA in the 2nd exon of the Prkci gene at codon position 29 gave rise to a fusion protein consisting of the first 28 amino acids of the PKCi protein followed by the remaining 564 amino acids of PKCf which will be transcripted under PKCi conditions using its endogenous promotor. In addition we modified the 39 UTR of the artificial transcript in such a way that the 39 untranslated regulatory RNA sequences of the PKCi messenger RNA will be included thereby providing the same mRNA stability as the wt PKCi transcript (see Figure S1C and materials & methods).
To investigate whether homozygozity of the PKCi fRes allele is able to compensate for PKCi function during embryonic development, we first analyzed pups from matings of heterozygous carries for the PKCi fRes allele at E7.5. At that stage we did not detect any morphological changes within 111 embryos obtained from heterozygous intercrosses of this line. In addition, we were able to identify 21 homozygous embryos (18.9%) for the rescue allele (PKCi fRes/fRes ) when we genotyped these embryos by PCR (Figure 6, table).
A Western Blot analysis of individual embryos revealed that wt and PKCi D/+ embryos showed an aPKC signal (using a pan aPKC antibody) whereas there was no detectable signal in the PKCi D/D . Interestingly in PKCi fRes/fRes embryos an aPKC specific band was detectable including a 50 kDa fragment which is supposed to be a proteolytic product of PKCf, proving the isoform specificity of the expressed protein ( Figure 6C). We concluded that PKCf expressed under endogenous PKCi condition is able to rescue the early PKCi phenotype. In contrast, we failed to identify any PKCi fRes/fRes animals in 150 offspring derived from heterozygous crosses at postnatal day 21 indicating that PKCf is not able to compensate for PKCi's function during later stages of development (see table in Figure 6). To define the lethality of PKCi fRes/ fRes embryos, we screened several embryonic stages for morphologically divergent embryos. At E9.5, PKCi fRes/fRes embryos were severely retarded in growth ( Figure 6A). Western blot analysis revealed that PKCf but not PKCi was expressed in the rescue embryos whereas PKCf was not expressed in the wt ( Figure 6D).  Thus, PKCf is able to compensate for PKCi in vivo function at E7.5 but not at later stages.

Discussion
The aim of this study was to shed more light on early embryonic in vivo functions of PKCi in mouse. Earlier genetic studies in C. elegans, Xenopus and Drosophila melanogaster indicated a pivotal role of aPKC in the establishment and maintenance of polarity of oocytes and epithelial cells [48][49][50][51]. These findings were often related to the functional and physical association of aPKC to Par-3 and par6, which together form the polarity complex [31]. This ternary complex is conserved among species suggesting a related phenotype upon depletion of PKCi in mouse embryos. A report by Soloff et. al, 2004 described an embryonic lethal phenotype when the PKCi gene was knocked out by conventional gene targeting without a detailed developmental analysis. Both of our generated knock out mouse lines for PKCi showed an embryonic lethal phenotype with first morphological manifestations around E7.5. Mutant embryos appeared compressed and are characterized by a severely reduced amniotic cavity. Nevertheless a histological examination revealed still the existence of epithelial cells which was rather unexpected since all studies from other species identified alterations in epithelial cell layers. So far publications describing gene knockouts in mice of polarity related proteins do not exactly match phenotypically with the PKCi knockout. For example a conventional knock out of Cdc42, an established activator of aPKCs in the context of the Par-3/aPKC/ Par6 complex, revealed a much earlier phenotype. Embryos deficient for Cdc42 showed first signs of alterations at E5.5 and were degraded at E6.5 [52]. Rac1, another small GTPase, which had been implicated many times as signal mediator of aPKC signaling also revealed an earlier phenotype in the knockout. In this case the main cause of lethality was linked to failures in the actin cytoskeletal reorganization combined with reduced motility and increased programmed cell death [53]. A complete knock out of Par-3 caused mid-gestational embryonic lethality (around E9.5) which was correlated to a defective cardiac development [54]. Unfortunately published data on a par6 knockout in mice are currently not available to our knowledge. One of the most similar knock outs published so far, when taking morphological criteria into account, is represented by the afadin knockout. Similar to the PKCi knockout morphological changes occurred at E7.5 and resembled what we saw in the PKCi deficiency (compressed embryo and small amniotic cavity) [55]. Nevertheless the described disorganized cell-cell junctions did not exactly match our observation. Interestingly afadin has been described as adapter protein connecting nectin to actin filaments [56] and to colocalize with Par-3/aPKC at AJs in the neuroepithelium of the developing mouse central nervous system [55]. Thus a similar function earlier during development is conceivable but further experiments are needed to underpin this hypothesis.
Our marker analysis revealed that PKCi-deficient embryos start to gastrulate and establish distinct anterior-posterior and dorsoventral body axes at around E7.5-E7.75. However, axial elongation and mesoderm formation arrests around this stage and elaboration of all mesodermal tissues and organs fails. Absence of a functional cardiovascular system is the cause of the subsequent death. Given the fact that all embryonic tissue are affected our data suggest that PKCi is not regionally but cellularly required.
Applying the embryoid body formation assay we showed that mutant EBs failed to establish a single cavity instead they formed multiple small cavities. Nevertheless a proper formation of the basal lamina and the apical domain by various marker proteins could be detected (see Figure 3). Particular the correct localization of Par-3 and Mupp1 implied an established apical pole which to some content did not surprise since studies performed in C. elegans showed that loss of PKC-3 did not affect Par-3 localization either [56]. Interestingly EBs deficient for Cdc42 not only fail to form a single lumen cavity but also showed a complete loss of aPKC and Par-3 localization at the apical domain [57]. One reason for this difference might be that Cdc42 is not able to distinguish between the two aPKC isoforms and we showed that PKCf is still present in PKCi deficient EBs. Thus we assume that depletion of Cdc42 resulted in a robust reduction of both aPKC activities causing a more severe phenotype. This finds support by the fact that overexpression of a dominant negative PKCf version in wt EBs mimics the Cdc42 phenotype [57]. The formation of a single cavity has been described as a two-step process: first, multiple small lumen are formed in the periphery of the EB and within a second step smaller lumen are fused and remaining inner cells undergo apoptosis to clear the central lumen [42]. Our data suggest that the loss of PKCi did not interfere with the first step but disrupted the second. Whether PKCi acts on the fusion of the small cavities or the subsequently induced apoptosis or both is not solved yet but from our data its clear that PKCi deficiency caused a decrease of activated caspase-3 and less apoptotic cells in EBs.
An earlier study on the role of Par6 and aPKCs in Caco-2 cysts formation showed that down-regulation of either one of the proteins is causing multiple lumen formation as well. In this it was correlated to an impaired spindle orientation during morphogenesis [58]. These findings might represent an alternative explanation for the observed phenotype. However, we were not able to identify any indication for an impaired spindle formation, either in PKCi deficient embryos or in PKCi deficient EBs. Again, one likely reason for this difference might be that the knock-down of Par6 and both aPKCs is more dramatic than the loss of one single aPKC isoform. In this case the data would indicate compensatory functions in the context of spindle orientation in Caco-2 cysts.
The immunohistochemical analysis of E7.5 embryo sections revealed by the E-cadherin staining that the general polarity of the embryonic ectoderm is preserved in mutants. But it became apparent that the fine structure of the apical side is changed (indicated by its fuzzy appearance) when compared to the wt at higher magnification (Figure 4 C). As a possible explanation we found ZO-1, a tight junction protein, to be down-regulated and dislocated in epithelial cells deficient for PKCi. ZO-1 is believed to function as a junctional organizer by direct binding to tight junctional proteins (like occludin and claudins) and the actin cytoskeleton. Thus it could well be that the absence of ZO-1 in tight junction complexes could lead to a lack of proper connections of tight junctions to the actin cytoskeleton, causing a less organized actin belt at the apical side which leads to the initially observed fuzzy appearance. This hypothesis finds support by the EM analysis, which clearly indicate that F-actin filaments are less compact organized and the distance between the outer membrane and the F-actin filaments appears to be increased in mutants. Consequently we checked the ZO-1 localization in E7.5 rescue embryos. The ZO-1 protein appears to be localized at the apical site of the embryo comparable to what we see in the wt embryos ( Figure S4). This result indicates that the low expression of PKCf in the PKCi deficient background might be responsible for the loss of ZO-1 localization. Nevertheless a conventional ZO-1 knockout displayed a later phenotype mainly associated with defect in angiogenesis and increased apoptosis [59]. One plausible explanation for this might be the existence of compensatory mechanisms in the ZO-1 knock out, which do not exist in the PKCi deficiency. As for example the remaining ZO-2/3 proteins might function redundant when ZO-1 was deleted whereas a knockout of a crucial downstream kinase cannot be compensated for in this system.
When analyzing aPKCs cellular localization ( Figure 4B) we mainly detected aPKC at the apical side, representing the expected pattern for aPKC in epithelial cell layer [60][61][62]. To our surprise a nearly unchanged pattern was detected in mutants as well, indicating that either aPKCs share overlapping expression domains or PKCf becomes up-regulated in mutants. We favor the first since neither a semi-quantitative PCR nor a Western Blot analysis of E7.5 PKCi D/+ and E8.5 PKCi D/D was able to identify an up-regulation of PKCf ( Figure S3). Thus, we assume that PKCf can potentially take part at the predicted in vivo function of aPKCs with regard on the establishment and maintenance of polarity due to its cellular localization, thereby might compensate for PKCi deficiency. In addition the increased cellular localization pattern of PKCi, identified by a specific antibody, implied that PKCi might fulfill other in vivo functions than the ones predicted by its association to the Par-3 and par6 proteins.
When PKCf was expressed under the same spatial and temporal conditions like PKCi (PKCi fRes/fRes allele) we were not able to detect embryos at E7.5 displaying the morphological changes described earlier. We concluded that PKCf is able to compensate for PKCi in vivo function at E7.5 when expressed at higher levels than in the wt. Thus the originally defined PKCi specific in vivo function at E7.5 represents an aPKC signaling pathway which became PKCi specific through a very tight transcriptional control. However, besides no visible difference between wt and rescue embryos at E7.5 and E8.5 the PKCi fRes allele is not able to compensate for PKCi functions at E9.5 since embryos homozygous for this allele showed a clear growth retardation resulting in lethality one to two days later. This points to the fact that later during development PKCi in vivo function exists which are not redundant among the aPKC subfamily and could therefore be described as isoform-specific. Nevertheless at this point we are not able to exclude that an artificial overexpression of PKCf has toxic effects for the embryo or that the observe proteolytic 50 kDa fragment ( Figure 6D) which had been linked to apoptotic process [63] causes the embryonic lethality. Worth mentioning at this point is the fact that within an earlier publication we have reported a week but detectable PKCf expression by western blotting in wt E9.5 embryos [15] whereas no PKCf protein was detected in this study. One possible explanation for this discrepancy might be that different amounts of protein were loaded since in this study we used extracts from single embryos whereas in the earlier publication pools of embryos served as starting material. Other reasons might be a different batch of the antibody as well as variations in the exposure time. In any case expression levels of PKCf could be stated as lower when compared to PKCi at the same embryonic stage.
In summary we showed that the depletion of PKCi caused embryonic lethality with the earliest time-point of visible morphological changes at E7.5. To our surprise we still detected a strong staining for aPKC at the apical membrane in PKCi deficient embryos indicating that a reasonable amount of PKCf still localize to this area. This per se might be a reason why the basal to apical cell architecture in PKCi deficient embryos was still conserved. In addition it also identified a possible PKCf function in this context by its very precise localization which by far was not expected based on expression data earlier published. As a result of PKCi deficiency we identified a severe down-regulation of ZO-1 which could explain the subtle alteration in the cellular structure. Nevertheless the cellular mechanism of PKCi in vivo function at E7.5 remained unsolved. However, when we expressed PKCf in PKCi deficient embryos the initial lethal phenotype at E7.5 could be rescued. This identifies the aPKCs as a very redundant kinase subfamily in which transcriptional as well as spatial cues regulate its isoform specificity.

Generation of Mutant Prkci Alleles
To clone the mouse Prkci locus a 129/Ola genomic cosmid library (obtained from the Resourcenzentrum, Berlin, Germany) was screened using a full length mouse cDNA as a probe. Several cosmid clones were identified and further purified. One of those containing the genomic 59prime part of the gene was selected for further cloning. All further cloning strategies followed standard procedures described in [64] and [65]. To generate the following targeting constructs for the PKCi gene a 10.9 kb genomic EcoRI fragment, including the 2nd exon (corresponding to nucleotides 110-233 of the published murine PKCi cDNA), was subcloned into a bluescript backbone. Using this genomic DNA fragment the conventional targeting vector was generated by inserting an independent neo-cassette (derived from pMC1neoPolyA from Stratagene) into a Sal I restriction site, which was introduced into the 2nd exon by site directed mutagenesis. As a consequence of this insertion the transcription of the PKCi gene is supposed to be abrogated. Based on the same genomic DNA backbone the conditional targeting vector was generated by introducing a single loxP site into a unique EcoRV restriction site 3.3 kb located upstream of the 2nd exon subsequently followed by an insertion of a loxP-flanked neo-cassette (derived from pL2-neo8, kindly provided originally by Klaus Rajewskys laboratory, Cologne, Germany) into a BglII restriction site 1.5 kb downstream of the 2nd exon. Upon incubation with the Cre recombinase this will lead to a deletion of a , 6.0 kb genomic DNA fragment including the 2nd exon, thereby causing a frame shift within the transcript, leading to a nonsense mRNA. The functionality of this allele has been proven by a sufficient number of tissue specific Cre mediated deletions in mice (see for example [25]).
In order to subclone the PKCi rescue targeting vector again the 10.9 kb genomic EcoRI fragment served as the starting material, including the introduced SalI restriction site within the 2nd exon. A mouse PKCf cDNA fragment (consisting of the cDNA sequence for codon 29 till the STOP codon followed by the complete 39 UTR sequence of the PKCi transcript) was inserted in frame into the SalI site which corresponds to codon 29 of the PKCi gene. In addition a loxP-flanked neo-cassette for later selection purposes was subcloned right after the PKCi 39 UTR. The resulting PKCi rescue allele expresses a fusion protein consisting of the first 28 amino acids of PKCi and the remaining 564 amino acids of PKCf under the control of the endogenous PKCi promotor and the PKCi 59 and 39UTRs.
All three targeting vectors were linearized before electroporation. Applied ES cells were from the substrain E14.1 background (129/Ola), kindly provided by Ralf Kühn, Institut for Genetics, Cologne, Germany. For each electroporation 2696 G418-resistant ES cell clones were screened for a homologous recombination event of the targeting vector by Southern blot analysis [66]. The observed targeting frequency was about 30-40%, respectively. For each targeting event at least 2 ES cell clones were further characterized by Southern Blotting using indicated 59and 39probes and the internal neo cassette (see Figure S1) to check for a correct integration of the corresponding targeting vector into the PKCi gene locus. Verified ES cell clones were further used for injections into NMRI blastocysts. Chimeric males were obtained for each construct and subsequently mated to NMRI females to test for germ-line transmission. Each targeted mutation gave rise to F1 heterozygous males which were then further used for establishing the corresponding mouse lines for phenotypic analysis.

Mice and Genotyping
Mice carrying a null allele of Prkci and the PKCi fRes allele were maintained on an outbred (NMRI) background. For timed pregnancies, vaginal plugs were checked in the morning after mating; noon was taken as embryonic day (E) 0.5. Pregnant females were sacrificed by cervical dislocation; embryos were harvested in phosphate-buffered saline, decapitated, fixed in 4% paraformaldehyde overnight, and stored in 100% methanol at 220uC before further use. Genomic DNA prepared from yolk sacs or tail biopsies was used for genotyping by polymerase chain reaction (PCR).

Ethics Statement
All animal work conducted for this study was approved by The Norwegian Institute of public health and performed according to Norwegian legislation.

Hematoxylin and Eosin Staining of Embryonic Sections
For histological analysis, wt or aPKC mutant embryos were fixed in 4% paraformaldehyde and embedded in paraffin wax. Sections of 4 mm were processed for cytohistochemistry after deparaffinization. After rehydration the section were incubated in 16 Gill's Hematoxylin (haematoxylin. 6.0 g, aluminum sulphate 42.0 g, ethylene glycol 269 ml, citric acid 1.4 g, sodium iodate 0.6 g, distilled water 680 ml) for 5 minutes. After washing the samples were incubated in acid alcohol (1% HCl in 70% EtOH) and washed again. After incubation in ammonia water the samples were rinsed and incubated in eosin Y (Sigma-Aldrich, HT110132-1L) for 1 minute. Before mounting with a xylene-based mounting medium cells were dehydrated. Embryos were documented using Zeiss Lumar V.12 with AxioCam HRc camera.

Whole Mount RNA in situ Analysis
Whole-mount in situ hybridization on E8.5 embryos was performed following a standard procedure with digoxigeninlabeled antisense riboprobes [67].

Embryoid Body Culture and Processing
The ES cell line used for this study were wt E14 and PKCi D/D ES cells [43]. Wt and PKCi D/D ES cells were cultured on mitomycin C-treated STO cells. EB formation was initiated from ES cell aggregates in suspension culture as described previously [68]. For immunofluorescence, EBs were collected into 15-ml conical tubes, allowed to sediment by gravity, washed once in phosphate-buffered saline (PBS), and fixed with 3% paraformaldehyde. EBs were then embedded in OCT compound, sectioned on a Leica cryostat and immunostained for antigens of interest [68]. F-actin was stained with rhodamine-phalloidin (Molecular Probes). Nuclei were counterstained with DAPI. Slides were examined with a Nikon inverted fluorescence microscope (Eclipse TE 2000) and digital images were acquired with a cooled CCD camera (Hamamatsu) controlled by IP Lab 4.0 software (Scanalytics). For immunoblotting, EBs were washed 3 times in PBS and lysed in SDS lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% SDS) containing protease and phosphatase inhibitor cocktails. Protein concentrations were determined using BCA reagents (Pierce). Proteins of equal loads were resolved by SDS-PAGE and then transferred onto PVDF membranes, which were blocked with 5% nonfat dry milk. After incubation with primary antibodies, specific signals were detected with HRP-conjugated secondary antibodies and ECL reagents. Antibodies used: PKCf rabbit polyclonal antibody (Cat# 07-264) and GAPDH mouse monoclonal antibody were purchased from Millipore. Perlecan rat monoclonal antibody was purchased from Santa Cruz Biotechnology, Inc. Cleaved caspase-3 rabbit polyclonal antibody was bought from Cell Signaling Technology. Actin polyclonal antibody was bought from Sigma.

Electron Microscopy
The embryos were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer. After washing, specimens were rinsed twice for 10 min in 0.1 M sodium cacodylate buffer and postfixed with 2% osmiumtetroxide (OsO4) containing 1.5% potassium ferric cyanide ((K3Fe(CN)6)) for 1 h. Following extensive rinsing with water, dehydration with an alcohol series was performed. Specimens were placed in 100% alcohol and Epon 1:1 overnight and then embedded in pure Epon. Ultrathin sections were cut with a Leica microtome for TEM microscopy and placed on carbon coated copper grids. Images were taken with a Philips CM 100 transmission electron microscope (TEM) at various magnifications.

Western Blot Analysis of Embryos
Embryos were collected isolated at either E7.5 or E9.5 from the mother animal. All extraembryonic tissue was removed before lysing the whole embryo in protein extraction buffer for 20 min. on ice (50 mM Tris/HCl, pH 7.5-8.0; 2 mM EDTA, pH7.0; 10 mM EGTA, pH 7.0; 0.1% TritonX-100, 3% b-mercaptoethanol) supplemented with Proteinase Inhibitor Cocktail (Sigma, #P-2714). A 31 gauge syringe was used to disrupt the tissue. The samples were centrifuged at 8.000 rpm for 8 min and the supernatant was transferred into a new tube before supplementing protein sample loading buffer and boiling for 5 min. The samples analyzed on a 10% SDS-polyacrylamide gel and blotted onto BioRad nitrocellulose membranes. Unless stated otherwise the protein amounts were adjusted based on the b-actin levels. The membranes were blocked in 5% nonfat dry milk in TBST for 1 h at room temperature. Incubation with the primary antibody overnight in TBST at 4uC. The following antibodies were used: anti-PKCf (Santa Cruz, C-20), anti-PKCi 1:10.000 [47], anti-PKCf 1:5000 [47] and anti-b-actin (Santa Cruz, sc-47778 HRP).

Semi-quantitative RT-PCR Analysis
Embryos were isolated at appropriate stage and directly lysed in TRIZOL reagent from InvitrogenH. The RNA isolation was performed according to the TRIZOL manual. The reverse transcription to obtain cDNA was performed using the iSCRIPT-cDNA synthesis kit from Bio-RAD. For the expression analysis the HotStar Taq PLus DNA Polymerase from Qiagen in combination with following primers was used: ZO-1 (forward-primer AGC TGT TTC CTC CAT TGC TG, reverse-primer: GAG ATG TTT ATG CGG ACG GT), PKC f (forward-primer: GCC TCC CTT CCA GCC CCA GA, reverse-primer: CAC GGA CT C CT C AGC AGA CAG CA) and PKCi (forward-primer: AGG AAC GAT TGG GTT GTC AC, reverse-primer: GGC AAG CAG AAT CAG ACA CA). All primers were designed with an optimal annealing temperature at 60uC and a amplicon size of around 300 bp. Figure S1 Targeting of the PKCi locus in mice. Schematic representation of the chosen targeting strategies is shown. Restriction maps indicating for each targeting approach the relevant restriction sites of the wild type and mutant locus are shown (B, BamHI; Bg, BglII; E, EcoRI; EV, EcoRV; H, HindIII; K, KpnI). The black box represents the second exon of the PKCi gene which was used for all targeted modifications. The white boxes below the mutant alleles indicate the 59HindIII-EcoRI RFLP probe (59probe) and the 39KpnI-EcoRI RFLP probe (39probe) used for later southern blot analysis. Generated targeted alleles are listed as following: A) conventional knockout -PKCi WT locus, the targeting construct (harboring an inserted neo cassette into the 2nd exon) and the mutant PKCi locus (PKCi Neo allele) after homologous recombination are shown. B) conditional knockout -PKCi WT locus, the targeting construct (containing a single LoxP site 59 and a floxed neo cassette 39 of the 2nd exon) and the mutant PKCi locus (PKCi floxed allele) after homologous recombination and subsequent deletion via Cre recombinase are shown. C) rescue knockin (PKCi fRes allele) -PKCi WT locus, the targeting construct (containing an inserted PKCf cDNA into the coding sequence of the 2nd exon followed by a floxed neo cassette) and the mutant PKCi locus (PKCi fRes allele) after homologous recombination and subsequent deletion of the neo cassette via Cre recombinase are shown. D) Southern blot analysis of embryonic stem cell clones after G418 selection (PKCi Neo allele; PKCi floxed allele) and heterozygous offspring from intercrosses of the PKCi D allele and PKCi fRes allele. Top left; BamHI RFLP using the 59probe identifies 2 clones of the conventional targeting by a 6.5 kb band whereas the 8.0 kb band represents the wt band, top right: HindIII RFLP using the 5 probe identifies one targeted ES cell clone for the floxed allele (3.0 kb band repesents the wt whereas the 2.0 kb band indicates a mutant), bottom left; the PKCi D allele was characterized by EcoRI using the 39probe resulting in an 11.0 kb wt and a 3.2 kb mutant. (EPS) Figure S2 Proliferation and apoptosis in PKCi mutant embryos. E7.5 embryos were embedded in paraffin and then sectioned into a 5 mm thick slices. Immunofluorescence analysis were performed using the indicated antibodies. Markers for either apoptosis (cleaved caspase-3, 1:500) or proliferation (phospho-histone-3, 1:500) were used for immunofluorescence analysis of PKCi D/D embryos or wt. All analysis were done using the Zeiss LSM 510 confocal microscope. Scale bar: 100 mm (TIFF)