Prkci is required for a non-autonomous signal that coordinates cell polarity during cavitation

Abstract

Polarized epithelia define boundaries, spaces, and cavities within organisms. Cavitation, a process by which multicellular hollow balls or tubes are produced, is typically associated with the formation of organized epithelia. In order for these epithelial layers to form, cells must ultimately establish a distinct apical-basal polarity. Atypical PKCs have been proposed to be required for apical-basal polarity in diverse species. Here we show that while cells null for the Prkci isozyme exhibit some polarity characteristics, they fail to properly segregate apical-basal proteins, form a coordinated ectodermal epithelium, or participate in normal cavitation. A failure to cavitate could be due to an overgrowth of interior cells or to an inability of interior cells to die. Null cells however, do not have a marked change in proliferation rate and are still capable of undergoing cell death, suggesting that alterations in these processes are not the predominant cause of the failed cavitation. Overexpression of BMP4 or EZRIN can partially rescue the phenotype possibly by promoting cell death, polarity, and differentiation. However, neither is sufficient to provide the required cues to generate a polarized epithelium and fully rescue cavitation. Interestingly, when wildtype and Prkci^−/− ES cells are mixed together, a polarized ectodermal epithelium forms and cavitation is rescued, likely due to the ability of wildtype cells to produce non-autonomous polarity cues. We conclude that Prkci is not required for cells to respond to these cues, though it is required to produce them. Together these findings indicate that environmental cues can facilitate the formation of polarized epithelia and that cavitation requires the proper coordination of multiple basic cellular processes including proliferation, differentiation, cell death, and apical-basal polarization.

INTRODUCTION

The proper coordination of apical-basal cell polarity is critical for basic cellular functions such as cell migration, asymmetric cell division, epithelia formation, and cell differentiation. These cellular processes are necessary for normal embryonic development and to maintain tissue architecture during homeostasis. The well-known steps of cell polarity include: a polarity cue, regulation of the cytoskeleton, distribution of polarity molecules, and the transduction of polarity to other components in the cell (St Johnston and Ahringer, 2010). Apical-basal polarity is observed in diverse species and has been proposed to be controlled by the evolutionarily conserved Par3-Par6-aPKC (atypical Protein Kinase C) trimeric complex (Kemphues et al., 1988; Suzuki and Ohno, 2006). The aPKC trimeric complex proteins play a central role in the proper organization of proteins linked to tight junctions (TJs) and adherens junctions (AJs) in many contexts including Drosophila embryos, mammalian epithelial cells, and pre-implantation mouse embryos (Alarcon, 2010; Martin-Belmonte and Perez-Moreno, 2012). In zebrafish, aPKCs and PAR6 are required for lumen formation in the neural tube and the intestine (Horne-Badovinac et al., 2001; Munson et al., 2008). Polarization allows junctional proteins to be localized apically and during zebrafish gut formation this results in lumen formation at multiple foci. These multiple lumens then fuse to resolve a single lumen via Smo-dependent signaling and Rab11 membrane trafficking (Alvers et al., 2014). Lumen resolution is likely somewhat distinct from lumen enlargement which in the case of zebrafish intestine, involves Claudin15 and Na+/K+ ATPase mediated ion transport and fluid influx (Bagnat et al., 2007).

The early steps in mammalian development exemplify a key time-period when the coordination of cell polarization is required. For example, establishing proper cell polarity is needed as the extra-embryonic tissues form and the inner cells mass is converted from a solid to a hollow ball of cells. This conversion of a solid structure into a hollow ball or tube shape is recapitulated during many stages of embryogenesis and although cell polarity is important in most cases, the molecular mechanisms driving this type of morphogenesis can vary (Iruela-Arispe and Beitel, 2013; Lubarsky and Krasnow, 2003; Martin-Belmonte and Mostov, 2008). In the case of cavitation of the mouse epiblast, initial studies first pointed to the role for BMP proteins generated by the outer endoderm layer in inducing programmed cell death to remove cells from the center. The cells in the ectodermal epithelium, although subjected to these signals, were proposed to be protected from cell death by their association with a basal lamina (Coucouvanis and Martin, 1995, 1999). By studying the process of cavitation in cells null for a protein important for polarity, PRKCI, here we determine the relative role of such proposed BMP signals vs. the impact of cell polarity during cavitation.

The precise pathways downstream of aPKCs that mediate epithelial polarity have not been identified in detail. EZRIN is typically found at the apical portion of epithelia and is involved in cellular functions including epithelial cell morphogenesis, polarity, and cell adhesion (Fehon et al., 2010). Previous studies have shown that the PRKCI isozyme interacts with and can phosphorylate EZRIN (Wald et al., 2008). Conversely, expression of a dominant negative form of aPKC, the treatment with aPKC inhibitors, or knockdown of Prkci with shRNA decreases EZRIN phosphorylation (Liu et al., 2013). Thus one possibility is that EZRIN is a key mediator of polarity downstream of aPKCs. Here we determine if EZRIN may mediate the polarity function of PRKCI during cavitation.

The inactivation of Prkci results in embryonic lethality at E9.5 (Mah et al., 2015; Seidl et al., 2013; Soloff et al., 2004) with cavitation problems in the inner cell mass evident at E5.5 (Fig S1A). In order to more easily study how Prkci functions in embryonic development, we have employed an in vitro system that avoids early embryonic lethality. Organ-like structures can be produced in culture by inducing embryonic stem (ES) cells to differentiate into embryoid bodies (EBs) (Doetschman et al., 1985; Kurosawa, 2007). Although much larger than the mouse embryo, these EBs can form three germ layers and mimic many aspects of cell differentiation during early steps of embryogenesis (Desbaillets et al., 2000). Because early embryonic stages are challenging to access for biochemical and molecular assays, generating EBs with genetically modified ES cells can be a useful approach to study early embryogenesis and organogenesis (Coucouvanis and Martin, 1995; Li and Yurchenco, 2006). Using this system, we find that PRKCI is required for normal cavitation. Although BMP signaling and EZRIN expression is altered/reduced in null EBs, stimulating the BMP pathway and/or EZRIN expression only moderately rescues the phenotype. However we see a robust rescue when null and wildtype cells are mixed together. We suggest that although EBs made from Prkci^−/− cells do not appear to have a vastly different RNA expression profile compared to Prkci^+/− EBs, a non-autonomous signal from neighboring cells can induce Prkci^−/− cells to become regularly polarized and that this coordination of polarity then leads to normal epithelial organization and normal cavitation. We thus uncover the importance of an environmental cue that coordinates cell polarity and morphogenesis.

MATERIALS AND METHODS

ES and EB culture

Mouse Prkci^+/− and Prkci^−/− ES cells were derived as described (Mah et al., 2015; Soloff et al., 2004). Prkci^+/− ES cells and two different Prkci^−/− ES cell lines were maintained on mitomycin C treated SNL (STO cell line transformed with neomycin resistance and murine leukemia inhibitory factor (LIF) genes) feeders at 37°C in humidified air with 5% CO₂. ES media consisted of Dulbecco’s Modification of Eagle’s Medium (DMEM) supplemented with 15% Fetal Bovine Serum (VWR, SH30071.03), 1% Minimal Essential Media Sodium Pyruvate (Millipore, TMS-005-C), 1% non-essential amino acids (Millipore, TMS-005-C), 1% penicillin and streptomycin (JR Scientific, 101447–068), 1% L-Glutamine (Millipore, TMS-002-C), 0.1% 2-mercaptoethanol (GIBCO by Life Technologies, 21985023), and 2% CHO-LIF conditioned media. EBs made with Prkci^+/− cells were indistinguishable from EBs made from several other wildtype lines. We chose to use Prkci^+/− cells as our controls as the line had undergone the same selection regimen as those that were Prkci^-/−.

To make EBs, sub-confluent ES cultures were depleted of feeders, dissociated with 0.25% trypsin-EDTA into 20–50 cell aggregates to make EBs in suspension culture or into single cells to make EBs in a hanging drop (250 cells in 20ul droplets). EBs were cultured in ES media minus LIF (EB media) either in bacteriological plates or for 3 days in hanging drops and then transferred into suspension culture.

To test the role of BMP4, day 3 EBs were cultured in the presence of 15 ug/ul recombinant human BMP4 (R&D Systems, 314-BP, reconstituted in 4mM HCl and 0.1% bovine serum albumin) for 4 days. Control cultures without added BMP were included in each experiment.

To assess mixtures of wildtype and null cells, R1 GPI-GFP (Nowotschin et al., 2009) and Prkci^−/− ES cells were mixed at 10:90, 25:75, and a 50:50 ratios before generating hanging drop EBs.

To assess cell proliferation and doubling time, 5×10³ Prkci^+/− and Prkci^−/− ES cells were plated. After 24, 48, 72, 96, and 144 hours, the cells were dissociated, washed twice in 1X PBS, and then counted. Cell doubling times were calculated with software available from http://www.doubling-time.com/compute.php.

Histology and Immunofluorescence

To assess apoptosis, plastic sections (7 microns) (Immuno-Bed kit, Polysciences, Inc., 17324) were incubated with TUNEL reaction mixture (Roche, 1684795) as per kit instructions.

Cell proliferation was examined by incubating EBs with 10uM BrdU (Sigma, B5002) for 24hrs. The EBs were then fixed with 4%PFA and treated with 3% H₂O₂ to block endogenous peroxidase activity, soaked with 2N HCl to denature DNA, permeabilized with PBS/0.5% Triton-X100 (EMD, 9002–93-1) (PBST), and blocked with 3% normal goat serum in PBST. Incorporated BrdU was then detected with anti-BrdU antibodies (Sigma, B2531, 1:200) and an Alexa Fluor® 488 secondary.

To quantify TUNEL, BrdU, phospho-Histone H3, or Oct4 in EBs, positive pixels were counted from images and compared to a measure of DAPI signal using a threshold method (Fogel et al., 2012). At least 6 images per condition were analyzed for each experiment.

To assess protein expression by immunofluorescence, EBs were fixed in 4% PFA (15 minutes at room temperature), washed three times with 1X PBS, permeabilized for 15 minutes with PBS/0.5% Triton-X100 (PBST) and blocked in 3% NGS in PBST (1 hr.). EBs were then incubated with primary antibodies (Table S1). The EBs were then washed and incubated with the secondary antibodies diluted in PBST (Table S2). EBs were then embedded in plastic and sectioned.

To detect F-actin, fixed EBs were rinsed in 1X PBS two times and then stained with 1:40 diluted rhodamine phalloidin (Invitrogen, R415).

Western blotting analysis

Harvested ES cells and EBs were lysed with RIPA lysis buffer (50mM Tris pH8.0, 150mM NaCl, 01% SDS, 0.5% Sodium deoxycholate, 1% NP40, complete protease inhibitor (Roche, 11697498001), and PhoSTOP phosphatase inhibitor (Roche, 04906845001)). A DC™ Protein assay reagent kit (BioRad, 500–0111) was used for measuring protein concentration. Equal protein amounts were separated by SDS/PAGE gel and transferred on PVDF membrane (BioRad, 1620177). The membranes were blocked and incubated overnight at 4^oC with 1:50 anti-Ezrin (Abcam, ab4069), 1:1000 anti-alpha-Tubulin (Abcam, ab15246), 1:1000 anti-E-cadherin, 1:1000 anti- ß-catenin, and 1:1000 anti-LC3 (MBL, PD014). Antibody binding was detected by ECL. Protein quantification was analyzed with Image J.

RT-PCR

Total RNA was extracted using Qiagen RNeasy extraction kit (Qiagen, 74104). DNA was removed with DNAse I (Roche, 04716728001). The sample was re-extracted with phenol-chloroform and precipitated. 1μg RNA was reverse transcribed using M-MuLV reverse transcriptase (New England Biolabs, M0253S) and 2mM random hexamers (Amersham Biosciences, 27216601). PCR were carried out with 1ul of cDNA using Gotaq^® green master mix (Promega, M71231). Primer sequences and the annealing temperature used in the RT-PCR are summarized in Table S3.

Transient expression

Mouse Ezrin full length cDNA was generated with PCR.KOD hot start DNA polymerase (Novagen, 71086–3). Primer sequences used in the PCR were as follows: EcoR1-Ezrin forward:AAAAAGAATTCATGCCCAAGCCAATCAACGTCCGGGTG and Ezrin-HindIII reverse: AAAAAAAGCTTCTACATGGCCTCGAACTCGTCAATGCGTTGCTTGG. Restriction enzyme digested PCR products were inserted into the CS107 plasmid vector (Baker et al., 1999) using the Quick ligation kit (New England Biolabs, M2200S) and selected clones were verified by sequencing.

Before transfection, 3×10⁵ feeder-depleted ES cells were seeded overnight on gelatin treated plates. Fugene HD (Promega, E2311) was used to transfect the plasmid. The transfection involved treating cells with a mixture of 7.2ug of Ezrin DNA diluted into DMEM media (343ul) and adding Fugene HD reagent (14ul). Transfected ES cells were then used for making EBs. As a control to assess the efficiency of transfection, a GFP-expressing plasmid (CS2nls-eGFP) was used. After 1 day of culture, about 50–60 % of the cells were GFP-positive.

Expression profiling

To create uniform EBs, 96-well round-bottom plates (BD, 268200) coated with Poly (2-hydroxyethyl methacrylate) (20mg/ml in 95% EtOH, Sigma, P3932), were seeded with 250 feeder-depleted ES cells per well and incubated at 37˚C, 5% CO₂ in a humidified incubator. Between 325–375 EBs were collected per day at precavitation stages (Days 3–5) and during cavitation (Days 7–10) and stored in RNAlater (Qiagen, 76104). An additional 25–35 EBs were collected for histological analysis to confirm developmental staging. Total RNA was extracted from collected EBs and undifferentiated ES cells (Day 0) using the High Pure RNA Isolation Kit (Roche, 11828665001), following the manufacturer’s protocol except for a more vigorous vortex in Lysis/Binding buffer (30 instead of 15 seconds) and eluting with 20μl (instead of 50–100μl), twice to concentrate the sample. Based on total RNA concentration (between 0.2 and 2μg/μl) an average of 60ng total RNA was recovered per EB. cDNA was made from 20μg of total RNA per sample with the cDNA Synthesis System (Roche 11117831001) and purified with the High Pure PCR Cleanup Micro Kit (Roche, 04983955001) using 20% binding enhancer according to the manufacturer’s protocol. An average of 1.2μg cDNA was synthesized per sample. The Functional Genomics Core of the Beckman Research Institute of City of Hope then carried out QC and NimbleGen Gene Expression Array analysis. The cDNA samples were labeled using the NimbleGen One-Color DNA Labeling Kit (Roche, 05223555001) before analysis on NimbleGen 12-plex MM9-build mouse expression arrays. NimbleGen pair files were loaded and analyzed in Partek® Genomics Suite 6.6 Copyright ©; 2014 (Partek Inc., St Louis, MO, USA). Differentially expressed genes were identified using two-way ANOVA with cutoffs of p<0.05 and fold-change >2.0 or <−2.0. GO analysis was undertaken by establishing a SQL database, mapping the ENTREZ gene identifiers to MGI identifiers, and graphing the ratio of the median expression/average signal of the array for different GO categories.

Statistical analysis

Statistical significance between experimental groups was performed with the unpaired student t-test using Microsoft Excel.

RESULTS

Altered polarity in Prkci^−/− EBs.

Previously we and others have shown that loss of Prkci leads to a failure of cavitation in both embryos and embryoid bodies (EBs) made from Prkci^−/− cells (Fig. 1A, A’, S1A). Loss of one copy of Prkci, however, results in normal EB development and viable, fertile offspring (Mah et al., 2015; Seidl et al., 2013). In addition, previous studies using MDCK cells have shown that inactivation of Prkci results in the formation of discontinuous adherens junction (AJ) and aberrant tight junction (TJ) formation (Suzuki et al., 2002). AJs between cells involve large protein complexes that link intracellular ACTIN to E-CADHERIN via α- and β-CATENIN (St Johnston and Ahringer, 2010; Yamada et al., 2005). Thus we first tested whether EBs made of Prkci^−/− cells would also lead to defects in cytoskeletal architecture. In control (Prkci^+/−) EBs, ACTIN is found highly enriched at the apical side of the ectodermal epithelium and within the surface endoderm, where ACTIN inscribes each cell in a regular pattern (Fig. 1B, inset). In Prkci^−/− EBs, ACTIN was still found enriched in the endoderm however deposition on the surface was disrupted and irregular (Fig. 1B’, inset). In addition, ACTIN could be seen prominently in cells found in the EB core at the apical side of many small micro-cavities (Fig. 1B’). By Western blotting we found that the overall expression levels of E-CADHERIN and β-CATENIN were similar between Prkci^+/− and Prkci^−/− EBs (Fig. S1B). However similar to ACTIN, in Prkci^−/− EBs, E-CADHERIN was strongly expressed by cells deep within the EB core (Fig. 1C’). β-CATENIN and N-CADHERIN are found both strongly enriched on the apical side of the columnar ectodermal epithelium in Prkc^+/− EBs, (Fig. 1D and E), while in Prkci^−/− EBs, these proteins were found evenly distributed in cells across the entire EB (Fig. 1D’ and E’). aPKCs proteins are typically localized to the apical side of the cell and can indeed be found enriched apically in the ectodermal epithelia of the EB (Seidl et al., 2013). We therefore also examined the expression of apical proteins to determine if their localization was now perturbed in the absence of Prkci. We found that ZO-1, a marker of TJs and apical polarity, was strongly detected at the apical side of the ectodermal epithelium in Prkci^+/− EBs (Fig. 1F), whereas in Prkci^−/− EBs, ZO-1 expression was not distributed in a polarized pattern (Fig. 1F’). Both GM130, a cis-Golgi protein (Debnath et al., 2002) and Podocalyxin (PODXL), an important early marker of apical polarity (Bryant et al., 2014) were also found in the apical domain of the ectodermal epithelium of Prkci^+/− EBs (Fig. 1G,H), whereas no specific GM130 or podocalyxin orientation was observed in Prkci^−/− EB cells (Fig.1G’, H’). Similar observations have been made for the apical markers MUPP1 and PAR3 (Seidl et al., 2013). Taken together, the failure to localize these proteins to the apical domain in Prkci^−/− cells indicates in a severe defect in cell polarity resulting in disorganized complexes at the AJs and TJs, failure to properly traffic proteins around the cell, and an irregular cytoskeletal architecture. Others have drawn similar conclusions based on loss of Prkci (Bedzhov and Zernicka-Goetz, 2014; Seidl et al., 2013). Interestingly the ability of some cells to have polarized features (for example, note the localization of apical proteins in micro-cavities of null EBs in Figures 1B’, 3N’, and 4B) is not completely abolished. This may in part be because Prkci^−/− EBs still express Prkcz, the other atypical PKC isozyme in mammals, and this protein may provide some redundancy in function (Mah et al., 2015).

Figure 1. Abnormal polarity in Prkci^−/− EBs.

A. By day 4 of culture during normal EB development (as represented by Prkci^+/− EBs) an outer endoderm layer is apparent. By day 6 the ectodermal epithelium has become polarized. Small cavities form and fuse together to make a single central cavity. Cells in the center die and are remove by non-professional phagocytosis. A’. Day 4 Prkci^−/− EBs look similar histologically to controls with a distinct ectodermal layer. No ectodermal epithelium forms however small cavities arise surrounded by a few polarized cells. These small cavities do not fuse and the EBs remain filled with cells.

B, B’. In Prkci^+/− EBs, ACTIN enriched at the apical side of the ectodermal epithelium (arrow). While in Prkci^−/− EBs, ACTIN is mainly found near cells arranged in micro-cavities. On the Prkci^+/− EB endodermal surface ACTIN is found in a regular ‘cobblestone-like’ arrangement while endodermal cells of Prkci^−/− EBs have irregularly arranged ACTIN (insets).

C, C’; D, D’. In Prkci^+/− EBs, E-CADHERIN and β-CATENIN protein expression is enriched in the ectodermal epithelium (C and D) while in Prkci^−/− EBs, both these proteins are ectopically expressed in cells found in the EB core (C’ and D’).

E, E’. N-CADHERIN is found enriched on the apical side of the ectodermal epithelium in Prkci^+/− EBs while in Prkci^−/− EBs, expression is broadly distributed in the core and endoderm.

F, F’. ZO-1 is normally expressed strongly at the apical side of the ectodermal epithelium (arrow) while Prkci^−/− EBs display nonpolarized and patchy ZO-1 expression.

G, G’; H, H’. GM130 and PODXL are found enriched at the apical side of the ectodermal epithelium in Prkci^+/− EBs (arrow) while the expression of both is not oriented in Prkci^−/− EBs. Sections are counterstained with rhodamine-phalloidin for F-ACTIN (G, G’) and DAPI for cell nuclei (C-H), respectively. B-H’. Matched EBs are shown at post-cavitation stages (day 7–14 of culture).Scale bars: B,B’ = 50 microns; C-E’ = 25 microns; F-H’=12.5 microns

Figure 3. BMP4 treatment does not rescue the null phenotype fully.

A-C’. At day 5, both Prkci^+/− and Prkci^−/− EBs have phospho-SMAD1/5 (pSMAD1/5) expression throughout the EB (A and B). At later stages (day 7–9), Prkci^+/− EBs continue to exhibit pSMAD1/5 expression. However in Prkci^−/− EBs, pSMAD 1/5 is lost from the center and found at low levels only near the endoderm layer (B’). The addition of BMP4 protein (15ng/ul) induces pSMAD1/5 expression in the core of day 7–9 Prkci^−/− EBs in a pattern similar to that seen Prkci^+/− EBs (C’). Experiment was repeated at least 3 times.

D-F’. In Prkci^+/− EBs, ENDO A is enriched in the outer endoderm layer at both stages (D and D’). In Prkci^−/− EBs, although endoderm localization is present, ENDO A is also found in the cells located in the EB core up to day 7–9 (E and E’). The addition of BMP4 partially inhibits the ENDO A expression found in the core (F and F’). Experiment was repeated at least 3 times.

G-J. The addition of BMP4 stimulates additional cell death in Prkci^−/− EBs as visualized by TUNEL analysis (H vs. I) and quantified at day 5 and 7 (J, three independent experiments; mean ± SEM, *p<0.05). Graph shows quantification with Prkci^+/− EBs represented with blue bars versus Prkci^−/− EBs represented with red bars. BMP4 addition to Prkci^−/− EBs is indicated (darker red bars).

K. Prkci^+/− (blue) and Prkci^−/− (red) EBs with and without BMP4 treatment were scored for their ability to fully cavitate. Without BMP4 treatment, Prkci^−/− EBs fail to cavitate (solid bars) (***p<0.001). BMP4 treatment induces more Prkci^−/− EBs to fully cavitate (outlined bars) (**p<0.01) but not to the extent seen in +/− EBs (three independent experiments, n=23 (+/−); 25(+/− plus BMP4); 40 (−/−); 54 (−/− plus BMP4); mean ± SEM). Partial cavitation is represented by hatched bars.

L, M. Histological analysis of BMP4 treated Prkci^+/− and Prkci^−/− EBs. Insets show the overall morphology. We found that treated Prkci^+/− EBs have a distinct columnar epithelium as typically seen without BMP4 treatment (L) while treated Prkci^−/− EBs that have cavitated do not form a normal columnar epithelium (M).

N-N”. In Prkci^+/− EBs, EZRIN is found in the endoderm and in the apical side of cells surrounding forming cavities at early (N, inset, day 4) and later stages (N, day 7). Whereas in Prkci^−/− EBs, EZRIN is only found in the endoderm region (N’). BMP4 treatment of Prkci^−/− EBs fails to induce the generation of a thick columnar epithelium with apical EZRIN expression (N”).

Scale bars: A-I and N-N” = 100 microns; L-M = 150 microns

Figure 4. Overexpressed Ezrin recovers the phenotype partially.

A-C. High EZRIN expression can be detected at the apical side of the ectodermal epithelium and in the endoderm layer of ~day 7 Prkci^+/− EBs as cavitation is finishing (A). However in Prkci^−/− EBs, EZRIN is only found in the endoderm layer and the apical side of a few small micro-cavities (B). Transient overexpression of Ezrin increases the number of micro-cavities in Prkci^−/− EBs (C).

D. Ezrin over-expression in both suspension and hanging drop EB assays results in more Prkci^−/− EBs undergoing full cavitation (outlined bars, **p<0.01). The percent of Prkci^−/− EBs undergoing partial cavitation (hatched) is also increased with Ezrin overexpression (*p<0.05). However, the extent of cavitation is less than is normally seen in Prkci^+/− EBs (*p<0.05). Percent of EBs that failed to cavitate is represented by a solid bar (five independent experiments, n=157(+/−); 180 (−/−); 223(−/− plus Ezrin); mean ± SEM).

E. Phenotype of fully cavitated Ezrin-expressing Prkci^−/− EB. The columnar ectodermal epithelium is not rescued.

F, F’. No significant difference could be detected in the number of TUNEL positive cells found in Prkci^−/− EBs vs. Prkci^−/− EBs overexpressing Ezrin, day 7.

G, G’. Overexpression of Ezrin induces normal expression of ENDO A in null EBs, day 7.

H, H’. pSMAD1/5 expression is enhanced in null EBs by Ezrin overexpression, day 7.

I, I’. The number of cells expressing OCT4 (nuclear) is diminished when Ezrin is overexpressed, day 10.

Scale bars: A-C and F-F’ = 125 microns; E= 150 microns

Cell death and proliferation rate are not strongly affected in Prkci^−/− cells

Previous studies have proposed a model for cavity formation in EBs, in which the endodermal layer produces cell death signals that induce apoptosis of cells within the EB center. The adjacent ectoderm is proposed to be protected from these cell death signals by an association with the basement membrane (Coucouvanis and Martin, 1995, 1999). This model suggests that a failure to undergo cavitation might occur if an ectopic basal lamina formed in the EB core that could protect cells from dying. We therefore assessed the expression pattern of a common LAMININ in Prkci^+/− and Prkci^−/− EBs by immunofluorescence. We found that in Prkci^−/− EBs, LAMININ was deposited normally below the endoderm layer similar to Prkci^+/− EBs (Fig. 2A,A’) and was not present within the core of the EB. We also examined the expression of the Laminin β-1 gene by RT-PCR but we found no difference in expression levels (data not shown). These results indicate that PRKCI is not required to form the basal lamina found between the endoderm and ectodermal layers and that some epithelial-like structures (those with small micro-cavities) could still form in the absence of a basal lamina. Finally, these results demonstrate that the formation of ectopic basal lamina is not the cause of the failed cavitation.

Figure 2. Cell death and proliferation in Prkci^−/− EBs is not grossly altered.

A, A’. LAMININ is enriched in the basal lamina located between the endodermal and ectodermal layers in both Prkci^+/− and Prkci ^−/− EBs.

B, B’. In Prkci^+/− EBs, TUNEL positive cells are primarily found near the forming cavity while in Prkci^−/− EBs, TUNEL staining was distributed across the entire EB.

C, C’. The number of cells expressing phosphorylated H3 (marking cells in mitosis) was similar in both Prkci^+/− and Prkci^−/− EBs.

D. Images from TUNEL experiments were used for quantification. There were no significant differences between the pixel count ratios in Prkci^+/− (blue bar) vs. Prkci^−/− EBs (red bar) (three independent experiments, n=18, 20; mean ± SEM)

E. Quantification of the pHH3 staining indicates a similar proliferation ratio based on pixel count in Prkci^+/− (blue bar) and Prkci^−/− EBs (red bar) (three independent experiments, n= 27 (day 7); 29 (day 10) for +/−; n= 32 (day 7); 29 (day 10) for −/−; mean ± SEM).

F. The relative BrdU positive ratio based on pixel count is similar in Prkci^+/− (blue bar) and Prkci^−/− (red bar) EBs (three independent experiments; mean ± SEM).

G, H. The proliferation of Prkci^−/− ES cells (cultured with LIF) is slightly slower (not statistically significant) than Prkci^+/− ES cells. (H) Cell doubling time is similar between Prkci^+/− and Prkci^−/− cells. Six biological replicate experiments were performed.

A-D: EBs were assessed just prior to the completion of cavitation (typically day 5–7 of culture). Scale bars: A-A’:100 microns, B-C’:50 microns

Because apoptosis-mediated cell death is thought to be an important step in the formation of EB cavities (Coucouvanis and Martin, 1995), we performed TUNEL assays on EBs to determine if a change in cell death levels could account for a failure to cavitate. As expected, TUNEL positive cells were found near forming cavities in Prkci^+/− EBs (Fig. 2B, 3G). However in Prkci^−/− EBs, cell death was not reduced (Fig. 2D) and TUNEL positive cells were distributed across the entire EB (Fig. 2B’, 3H). Thus failed cavitation by loss of Prkci does not appear to be due to the general inability of Prkci^−/− cells to undergo apoptosis. Genes required for autophagy are involved in removing dying cells by non-professional phagocytes during EB cavitation (Qu et al., 2007). One possibility could be that although levels of apoptosis are normal in Prkci^−/− EBs, they fail to cavitate because of a defect in autophagy or phagocytosis such as would occur by the loss of LC3 (cytosolic LC3I and membrane bound LC3II), a marker of mammalian autophagy (Klionsky et al., 2016). Western blot analysis showed that there was no dramatic difference in LC3 levels in Prkci^+/− versus Prkci^−/− EBs (Fig. S2A). Thus, LC3 expression levels are not dramatically altered in Prkci^−/− EBs and autophagy and/or phagocytosis is likely not defective.

Another way for cavitation to fail could involve an increase in cell proliferation rate that exceeds the cell death rate. To determine whether cell proliferation is increased in Prkci^−/− EBs, we analyzed phospho-histone H3 (pHH3) expression and BrdU incorporation. The number of pHH3 positive cells appeared similar between Prkci^+/− and Prkci^−/− EBs at day 7 and 10 (Fig. 2C, C’) and this was confirmed with a quantification assay (Fig. 2E). BrdU incorporation was also not appreciably different in Prkci^−/− EBs (Fig. 2F). To further study cell proliferation, we examined cell proliferation and doubling times in undifferentiated ES cells. We found that Prkci^−/− ES cells proliferated at a slightly slower rate, but the cell doubling times were similar (Fig. 2G–H). Taken together, no dramatic difference was detected in cell proliferation rate comparing Prkci^−/− to Prkc^+/− cells, thus, it seems unlikely that the defective cavitation seen in Prkci^−/− EBs is caused by a sustained increase in cell proliferation.

BMP4 treatment does not rescue the phenotype fully.

During EB cavitation, BMP2 and BMP4 have been proposed to mediate the cell death that occurs in the EB core since treatment of S2 EBs (which normally do not undergo cavitation) with BMP4 protein induces cell death and cavitation. In addition, inhibition of BMP signaling with a dominant negative BMP receptor 1B suppresses cell death and cavitation (Coucouvanis and Martin, 1999). Although cell death is not reduced in Prkci^−/− EBs, we did observe TUNEL positivity distributed throughout the entire EB (Fig. 2B, B’). Thus, one possibility is that altered BMP signaling might lead to a different pattern of cell death in Prkci^−/− EBs. First, by RT-PCR we found that the expression levels of Bmp2 and Bmp4 were similar in both Prkci^+/− and Prkci^−/− EBs (Fig. S3A). To determine if the proposed responding cells had an altered BMP response, we also assessed the expression and localization of phosphorylated SMAD1/5 (pSMAD1/5), a readout of canonical BMP signaling (Eom et al., 2011). We observed that at day 5, pSMAD1/5 activity was present throughout both Prkci^+/− and Prkci^−/− EBs (Fig. 3A, B). During differentiation, pSMAD1/5 expression was still found internally in Prkci^+/− EBs, day 7–9 (Fig. 3A’). However, Prkci^−/− EBs at day 7–9, pSMAD1/5 activity was vastly decreased (Fig. 3B’). The visceral endoderm is proposed to release BMP4 during EB cavitation (Coucouvanis and Martin, 1999) so a decrease in pSMAD1/5 internally may be due to a decrease in visceral endoderm differentiation (Coucouvanis and Martin, 1999; Duncan et al., 1994; Makover et al., 1989). RT-PCR results showed that both the visceral endoderm markers Hnf4 and Ttr were expressed in Prkci^−/− EBs compared to Prkci^+/− EBs although somewhat delayed (Fig. S3A). Thus defective BMP signaling by loss of Prkci may lead to delayed differentiation of the visceral endoderm and failed cavitation. We also examined the expression of cytokeratin ENDO A (ENDO A) to determine whether the differentiation of endoderm is altered. Immunofluorescence assays showed that ENDO A is expressed exclusively in the endoderm region of Prkci^+/− EBs (Fig. 3D, D’). However, in Prkci^−/− EBs, while ENDO A is found in the endoderm layer, it is ectopically expressed in the core until day 7–9 (Fig. 3E, E’). In summary these findings suggest that Prkci is required for maintaining the proper levels of pSMAD1/5 activity in the core, promoting visceral endoderm differentiation, and is required to prevent the ectopic expression of ENDO A in the EB center.

Although the expression levels of Bmp2 and Bmp4 are not altered, the expression patterns of pSMAD1/5 and ENDO A are altered suggesting that there could still be a problem with the proper local delivery of BMP2/4 proteins, especially in a context where cell polarity is disrupted. We therefore reasoned that exogenously added BMP protein might facilitate either directly or indirectly (by inducing endoderm) transmission of death signals into the EB core and eventually enhance cavitation (Coucouvanis and Martin, 1999). To test this hypothesis, we studied the affect of adding BMP4 protein during Prkci^−/− EB development. By immunofluorescence, BMP4 treatment resulted in enhanced pSMAD1/5 activity in the core of Prkci^−/− EBs at day 7–9 (Fig. 3C’) although not at the same levels seen in Prkci^+/− EBs. At day 5, BMP4 treated Prkci^−/− EBs had stronger ENDO A expression in the endoderm layer compared to untreated Prkci^−/− EBs (Fig. 3F). However, ectopic ENDO A positive cells were still present in the EB core up to day 7–9 as seen in untreated Prkci^−/− EBs (Fig.3F’). Thus BMP4 treatment can induce enhanced pSMAD1/5 activity in the core of null EBs, but that treatment does not fully prevent ectopic ENDO A expression in the EB center.

Because BMP4 is considered a cell death signal (Coucouvanis and Martin, 1999), enhanced BMP signaling might lead to increased cell death in the EBs. Indeed, TUNEL positivity was significantly increased in BMP4-treated day 5 Prkci^−/− EBs compared to Prkci^+/− and Prkci^−/− EBs (Fig.3G–J). Furthermore, we found that BMP treatment increased the percent of Prkci^−/− EBs with cavities. However, the extent of recovery was partial with rescue of only half the percent of cavities found in Prkci^+/− EBs suggesting that BMP4 treatment is not sufficient to fully rescue the phenotype (Fig. 3K). These results also suggest that although the regulation of cell death is important for clearing cells in the EB center, the failure of null EBs to cavitate cannot be fully explained by decreased BMP signaling or a defect in the ability of cells to respond to a BMP signal.

Although cavitation could be stimulated with BMP4 treatment, a normal pseudostratified columnar ectodermal epithelium did not form (Fig.3L, M). EZRIN, an apical maker (Saotome et al., 2004), was also examined to determine if BMP4 treatment could stimulate the recovery of cell polarity in the ectodermal epithelium. In Prkci^+/− EBs, EZRIN is highly expressed within the endoderm and then in the apical region of the ectodermal cells that surround developing cavities (Fig.3N). In Prkci^−/− EBs the endoderm expression pattern did not appear altered, however EZRIN expression was much reduced internally and then only found at the apical side of micro-cavities (Fig.3N’). In Prkci^−/− EBs cultured with BMP4, polarity was not rescued since treated EBs still had the Prkci^−/− EB EZRIN expression pattern (Fig.3N”). In summary, these results suggest that the addition of BMP4 was not sufficient to recover the normal expression pattern of ENDO A and EZRIN or restore the formation of a columnar epithelium and can only partially recover cavitation perhaps by enhancing cell death. Thus it is likely that other factors besides BMPs are required to restore cavitation in Prkci^−/− EBs, especially those that can influence cell polarity.

Overexpressed Ezrin partially recovers the phenotype.

EZRIN is a membrane-to-cytoskeletal linker that is involved in various cellular functions, including epithelial cell morphogenesis and adhesion (Bretscher et al., 2002). Recent studies have shown that PRKCI interacts directly with EZRIN in Caco-2 cells and knockdown of endogenous Prkci decreases EZRIN expression and phosphorylation at T567 in intestinal epithelial cells. Apical localization of PRKCI has also been shown to be essential for the apical localization of activated EZRIN (Wald et al., 2008). These studies suggest that EZRIN might be an important downstream effector of PRKCI, and the loss of Prkci might lead to changes in expression levels of Ezrin during EB development. RT-PCR results, however, showed no large difference in the levels of Ezrin gene expression in Prkci^+/− versus Prkci^−/− EBs (data not shown) though the apical-basal localization pattern of EZRIN was altered (Fig. 3N’, 4B). We therefore hypothesized that by overexpressing Ezrin in Prkci^−/− EBs, enough would reach the apical side to facilitate and coordinate other cell polarity proteins and thus rescue cavity formation. At day 7, we found that overexpression of Ezrin resulted in an increase in the number of micro-cavities when compared to un-transduced control Prkci^−/− EBs (Fig. 4C,D). Also endodermal cells were more likely to express EZRIN strongly in Ezrin-transfected Prkci^−/− EBs. We then scored transfected Prkci^−/− EBs along with Prkci^+/− and Prkci^−/− EBs for their ability to cavitate. Quantification clearly showed that both partial and full cavitation was increased in Ezrin-transfected Prkci^−/− EBs although the extent of full cavitation was less than normally seen in Prkci^+/− EBs (Fig. 4D). As seen with BMP treatment, the columnar ectodermal epithelium did not form properly, even in EBs that had fully cavitated (Fig. 4E). Taken together, these results suggest that overexpressed Ezrin can partially recover polarity in Prkci^−/− EBs, increase the number of micro-cavities, however the formation of the ectodermal epithelium or resolution of a single lumen is not fully rescued.

One explanation for the enhanced cavitation observed in Ezrin-transfected Prkci^−/− EBs is that cell death was somehow increased. Using TUNEL analysis, however, we did not see any increased cell death in Ezrin-transfected Prkci^−/− EBs compared to untransfected Prkci^−/− EBs (Fig. 4F, F’). Another possibility is that Ezrin induces increased BMP activity. We therefore examined pSMAD1/5 activity and found that pSMAD1/5 was significantly increased in the core of Ezrin-transfected EBs compared to control Prkci^−/− EBs (Fig. 4H, H’). We also examined ENDO A expression and found that the expression pattern was rescued by Ezrin overexpression since similar to Prkci^+/− EBs, ENDO A expression was no longer found in the core of Prkci^−/− EBs (Fig.4G’). Thus overexpressed Ezrin may lead to a partial recovery of polarity and cavitation by promoting the normal expression pattern of pSMAD1/5 and ENDO A. In addition by enhancing the ability of cells to be polarized, differentiation is more likely to proceed. In support of this idea, we found that the number of OCT4 positive cells in Ezrin-transfected Prkci^−/− EBs was decreased compared to untransfected null EBs (Fig. 4I, I’ and Fig. S4A).

BMP4 treatment or Ezrin transfection each partially rescued the Prkci null phenotype, so we determined if combined BMP4 treatment and Ezrin overexpression would be additive. However, EBs subjected to both treatments were not fully rescued. Cavitation was still only partially rescued, the columnar epithelium did not form properly, and the normal EZRIN expression pattern was not restored (Fig. S4B and data not shown). Thus either we were not able to achieve high enough levels of these two factors to rescue the phenotype or additional factors are required.

Mixing wildtype cells with Prkci^−/− cells fully recovers the phenotype.

We next asked whether wildtype polarity-competent cells could influence Prkci^−/− cells to gain coordinated polarity and participate in cavitation. Wildtype R1 ES cells that express a glyco-phosphatidyl-inositol green fluorescent fusion protein (GPI-GFP, which directs GFP to the plasma membrane) and that form normal cavitated EBs ((Nowotschin et al., 2009) and data not shown) were mixed with Prkci^−/− ES cells. EBs were then generated with 10%, 25%, or 50% GPI-GFP cells with the remainder as Prkci^−/− cells. GPI-GFP ES cells could be easily distinguished from null cells by fluorescence imaging and at day 3–5, wildtype GPI-GFP cells were present in scattered patches among GFP-negative null cells in 50% GPI-GFP EBs (Fig. 5A). At day 6–8, GPI-GFP cells were located mainly in the endoderm, with some contribution to the ectodermal epithelium, and the EB core (Fig. 5A’). By day 9–10, GPI-GFP cells were mostly found in the endoderm layer (Fig. 5A”). We then assessed EB morphology and found that about half of the EBs made with 10% GPI-GFP cells formed a pseudostratified columnar epithelium and could undergo cavitation (Fig. 5C,D). In EBs made with 25% GPI-GFP cells, 55% had formed a columnar epithelium and 20% cavitated (Fig. 5C–D). In 50% GPI-GFP EBs, 75% of EBs formed a pseudostratified columnar epithelium and 60% of EBs cavitated, a frequency similar to Prkci^+/− EBs (Fig. 5B”, C–D). Thus a 50:50 wildtype-to-Prkci^−/− ratio was able to recover the null EB phenotype. Similar results were achieved by using another line wildtype for Prkci (a C57BL/6 ES line for example, data not shown).

Figure 5. Mixing wildtype cells with Prkci^−/− cells fully recovers the phenotype.

A-A”. When EBs are made with a 50:50 mixture (GPI-GFP wildtype ES cells: Prkci^−/− ES cells) in a hanging drop EB assay, GFP positive cells are widely distributed in day 3–5 EBs (A). At day 6–8 however, GFP positive cells are enriched in the endoderm with some cells in the core (A’). At day 9–10, GFP positive cells are mostly found in the endoderm layer (A”).

B-B”. The histological phenotype of 50:50 mixed EBs at different time points. Mixed EBs develop similarly to Prkci^+/− EBs with a distinct endodermal layer evident by day 3–5 (B). Micro-cavities have formed and some have merged in day 6–8 EBs (B’). A regular columnar epithelium (white arrow) and full cavitation is observed by day 9–10 (B”).

C. The percent of EBs that form a regular columnar epithelium is highly recovered in 50:50 mixed EBs compared to Prkci^−/− EBs (*p<0.05) and similar in frequency to Prkci^+/− EBs. (red bar: −/−, purple bars: mixed, blue bar: +/− control) (three independent experiments, n=32(−/−); 28(10%); 21(25%); 20(50%); 21(+/−)); mean ± SEM). By ttest, the difference between Prkci^+/− and 50% mixed EBs was not significant.

D. The percent of EBs that fully cavitate (outlined bar) is highly rescued in 50% GPI-GFP: 50% Prkci^−/− mixed EBs compared to 100% Prkci^−/− EBs (***p<0.001) and the percentage was similar to Prkci^+/− EBs. Failure to cavitate: filled bar; partial cavitation: hatched bar (three independent experiments n=40(−/−); 25(10%); 20(25%); 19(50%); 23(+/−); mean ± SEM). By ttest, the difference between Prkci^+/− and 50% mixed EBs was not significant.

E. ENDO A expression is rescued in mixed EBs (ectopic ENDO A in the core is much decreased).

F. GM130 in 50:50 mixed EBs is found at the apical side of the ectodermal epithelium at day 12 (white arrow). GPI-GFP cells are concentrated in the endodermal layer (green arrow).

G. 100% GPI-GFP EB showing strong expression in all cells.

H-J. EZRIN, PODXL, and ZO-1 expression in mixed EBs is enriched in apical region of developing ectodermal epithelium in 50% mixed EBs (white arrow).

K-M. pERK1/2 is found highly expressed in the columnar epithelial cells in Prkci^+/− EBs. Prkci^−/− EBs lack pERK1/2 as expected. In mixed EBs, pERK1/2 is now expressed in the columnar epithelial cells.

Scale bars: A-A”: 100 microns, B-B”: 150 microns, E-M=50 microns

Because the columnar epithelium forms so nicely, endoderm differentiation and cell polarity might also be rescued in the 50:50 cell mixing experiments. Indeed we found that ENDO A was now expressed predominantly in the outer endodermal layer and not ectopically in the core (Fig. 5E). Apical markers GM130, EZRIN, PODXL, and ZO-1 were now found enriched in the apical region of Prkci^−/− (negative for GPI-GFP) ectodermal cells lining the developing cavities (Fig. 5F, H–J). In addition, the amount of cell death in mixed EBs as determined by TUNEL was unchanged compared to both Prkci^−/− and Prkc^+/− EBs (Fig. S5A, B). Taken together, these results suggest that polarity-competent cells can have a non-cell-autonomous influence to rescue the polarity of Prkci^−/− differentiating ES cells and that establishing/maintaining epithelial polarity may be one of the most important features required for normal EB cavitation.

We found that GFP-GPI wildtype cells were intermingled with null cells during early EB development (Fig. 5A’, A”). How might signals emitted from wildtype cells rescue the polarity of the Prkci^−/− cells? One possibility is that wildtype cells influence null cells by via direct cell-cell contact and/or the local production of ECM or signaling molecules (reviewed in (Roignot et al., 2013)). Another possibility is that wildtype cells act on Prkci^−/− cells at a distance via signaling cascades among neighboring cells. In order to investigate these possibilities, we used the expression of pErk1/2, which is strongly found in normal cavitating EBs and is activated and required during the differentiation of pluripotent cells (Kunath et al., 2007), as a read-out for a differentiating signal (Fig. 5K). Thus we determined whether in mixed EBs, pErk1/2 is elevated in Prkci^−/− cells located near or far from GPI-GFP-positive wildtype cells or even within GPI-GFP-positive cells themselves. We found that although EBs made with 100% Prkci^−/− cells had very little pErk1/2 activation (Fig. 5L), when null cells were mixed with wildtype cells pErk1/2 activity was increased and largely observed in Prkci^−/− cells located in the columnar ectodermal epithelium below the GPI-GFP-positive wildtype endodermal layer but also at sites within the EB core (Fig. 5M, and not shown). These observations suggest that the GPI-GFP-positive wildtype cells can transmit signals to stimulate pErk1/2 activation via both direct cell-cell interactions and long distance signaling. Together the combined effects of paracrine signaling between cells and signaling among neighboring cells might spread throughout a tissue to establish normal polarity and an organized ectodermal epithelium, thus facilitating cavitation and differentiation.

Expression profile of EBs prior to cavitation

Our previous studies indicated that under differentiation conditions Prkci^−/− cells are more likely to remain in a stem cell state, although they are fully capable of differentiation (Mah et al., 2015). We hypothesized that gene expression in Prkci^−/− cells would be very similar to control Prkci^+/− cells at least during germ layer specification and prior to cavitation. We therefore assessed the expression profile of Prkci^+/− and Prkci^−/− EBs at different stages of differentiation (ES cells, before cavitation: days 3–5, and after cavitation: days 7–10) by microarray analysis using a NimbleGen array representing ~44,000 genes. Principle component analysis (PCA) showed that EBs at pre-cavitation stages were very similar to each other, and distinct from EBs at later stages and thus the overwhelming driver of differential gene expression was differentiation stage (Fig. 6A, PC1, 60.8%); this differentiation effect greatly exceeded any effects of genotype which was also not separated along PC2 or PC3. EBs that had cavitated exhibited much more variability especially along PC2 but again without any trend with respect to genotype. We then graphically compared gene expression levels between Prkci^+/− and Prkci^−/− EBs at the pre-cavitation stage across several large gene ontology (GO) categories. As predicted from the PCA analysis, when comparing expression levels in each category, we found a much greater difference at different stages than for different genotype. For instance, for GO:0042147, (regulation of proliferation), we found that although some differences in expression were certainly detectable (difference between red and blue y values) in Prkci^−/− (red) versus Prkci^+/− (blue) when plotted against the rank order of Prkci^+/− (x axis) at pre-cavitation stages, the levels of expression after cavitation varied much more from the precavitated baseline in both genotypes (Fig. 6B). Large differences were also evident when comparing gene expression levels before and after cavitation for Prkci^+/− EBs (Fig. 6C). In general, however when comparing Prkci^+/− and Prkci^−/− EBs at the same pre-cavitation stages and across different large GO categories, expression levels were very similar (Fig. 6D) indicating that although Prkci^−/− EBs exhibited altered polarity and morphological abnormalities, relatively few genes could be responsible for these effects. Our observation that relatively few genes show altered expression levels even though these two EB types ultimately have distinctly different morphology, allowed us to potentially detect specific important genes related to the morphogenesis outcome. Thus, to determine which genes showed statistically discernible differences in expression levels, we performed a two-way ANOVA (see methods) at the pre-cavitation stage to generate a list of 551 differentially expressed genes (full list available in Table S4). Using cutoffs for p values (<0.05) and fold expression (>2.0; <−2.0) we subjected the gene lists to Ingenuity Pathway Analysis (IPA). Genes both up- and down-regulated fell into a few main categories including Cancer, Molecular transport, and Nervous System Biology, and other smaller categories (Fig 6E). Our cell mixing experiments indicated that Prkci^−/− cells could respond to a non-autonomous influence and coordinate their polarity, but that they are not able to produce the effect themselves. Although a non-autonomous influence could occur via several different mechanisms, we identified 26 extracellular proteins (as defined by IPA) out of the 551 genes that are strongly expressed in Prkci^+/− cells but not in Prkci^−/− cells at pre-cavitation stages. We hypothesize that these proteins could be released by wild-type cells and induce Prkci^−/− cells to become polarized in our cell-mixing experiments. For the top 7 of these genes we confirmed their lower expression in Prkci^−/− cells by RT-PCR (Fig. 6F, G).

Figure 6. Gene expression profile of Prkci^+/− versus Prkci^−/− EBs.

A. Principle component analysis (PCA) for the first 3 principle components showing the largest separation at different stages of differentiation. Prkci^+/− ES cells (n=2 experiments), Prkci^+/− and Prkci^−/− EBs at precavitation stages (day 3–5, n=2 experiments); Prkci^+/− and Prkci^−/− cavitated EBs (day 7–9, n=3 experiments). Prkci^+/−:blue; Prkci^−/−:red

B. Expression of genes in the large GO category (regulation of proliferation) present on the NimbleGen Array for Prkci^+/− EBs at precavitation stages ranked in comparison to the expression levels in Prkci^−/− EBs (log₂). After cavitation, overall gene expression in Prkci^−/− EBs was much different (x axis rank remains equal to the precavitated rank). Prkci^+/−:blue; Prkci^−/−:red

C.Log2 expression for genes on the array in the same GO category as (B), comparing expression levels in Prkci^+/− EBs at precavitation stages versus after cavitation.

D. Expression for genes on the array in the same GO category as (C) as well as some other large categories, comparing expression levels in Prkci^+/− versus Prkci^−/− EBs at precavitation stages. Comparatively fewer genes are differentially expressed in Prkci^+/− versus Prkci^−/− EBs than when comparing stage of differentiation.

E. A 2-way ANOVA with cutoffs of p<0.05 and fold-change >2.0 or <−2.0 was used to identify differentially expressed genes (551) at precavitation stages. A graphical representation of associated functions from IPA identifies the top 15 biological functions/diseases in Ingenuity’s Knowledge Base that are most significant to the data set. Fisher’s exact test was used to calculate the probability that each biological function and/or disease assigned to the data set is due to chance. Functions are listed from most significant to least and the red vertical line indicates the cutoff for significance.

F. Of the 551 genes, those in the extracellular category that are reduced in expression in Prkci^−/− versus Prkci^+/− EBs (26 genes).

G. Validation by RT-PCR for the top 7 of the 26 genes.

DISCUSSION

In this study we found that loss of Prkci resulted in defective cell polarity and the mis-localization of the adhesion molecules that associate with AJ and TJs while cell death and proliferation rates were largely normal. SMAD phosphorylation was reduced and mis-localized in Prcki^−/− EBs however treatment with BMP4 protein only partially recovered EB cavitation without promoting apical-basal polarity or the generation of organized epithelia; rescue of cavitation likely occurred due to increased apoptosis. Overexpression of Ezrin also resulted in some increased cavitation but probably by a different mechanism— by promoting apical-basal polarity and the formation of micro-cavities. However Ezrin overexpression was not sufficient for a robust rescue and additional or different signals may needed. Interestingly, a full rescue could be achieved including restoration of polarity, epithelial organization, and full cavitation when wildtype cells were mixed Prkci^−/− cells suggesting that wildtype cells can provide missing environmental cues. Thus, our observations suggest that Prkci is primarily required for establishing and maintaining strict apical-basal cell polarity in ectodermal EB cells which allows for the stable formation of micro-cavities. Once these are formed they are maintained and can coalesce and resolve into a single cavity. In the absence of Prkci, cells are not sufficiently polarized, micro-cavities do not form or are not stable or large enough, and never coalesce and resolve into a single cavity. Expression profile comparisons show that Prkci^+/− and Prkci^−/− EBs at the pre-cavitation stage are very similar to each other despite differences in the localization of key polarity indicator proteins. Analysis of differential gene expression at this stage may therefore provide important clues for understanding what signals coordinate polarity and cavitation and why these signals are not produced in Prkci^−/− cells.

Cell death is unlikely to be the main mechanism driving cavitation

Despite Prkci^−/− EBs having a solid center, we find that cell death still occurs (Fig. 2B’, D). In addition the rate of proliferation is similar between Prkci^+/− and Prkci^−/− EBs (Fig. 2C,C’ and E–H). Consistent with our EB results, apoptosis and cell proliferation are not altered in conditional Prkci knockout mouse embryos (Imai et al., 2006). In addition, EBs null for CDC42 (upstream of aPKC signaling) also do not cavitate and or display a large change in apoptosis markers (Wu et al., 2007). However, another study investigating the consequence of removing Prkci function on EB development suggested that cavitation in Prkci^−/− EBs fails in part because of an observed reduction in cell death (Seidl et al., 2013). It is unclear why this study differs from our reported results and a more in-depth analysis will be needed to determine if Prkci^−/− cells could be more resistant to cell death signals. If cell death were an important player, however, we would have expected to see a dramatic increase in cell death in the rescued mixed wild-type/null EBs but we did not see an appreciable difference (Fig S5A, B). Furthermore, recent studies have shown that apoptosis is not necessary to form columnar epithelia in the early mouse embryo using a cell death inhibitor and p53 mutant (Bedzhov and Zernicka-Goetz, 2014). Thus we conclude that loss of Prkci is unlikely to play a major role in the ability of cells to respond to cell death signals and although a reduced ability to undergo cell death can result in a failure to remove cells in the center, cell death itself is not the driving mechanism that causes EB cavitation.

Cell polarity and cavitation

We found that major cytoskeletal and polarity proteins, both apical and basal are not present or mis-localized in Prkci^−/− EBs (Fig. 1, 2). These results are consistent with previous studies showing widespread apical-basal polarity defects when a dominant-negative form of aPKC is expressed in MDCK cells (Suzuki et al., 2002) and in another study investigating polarity in Prkci^−/− EBs and embryos (Seidl et al., 2013). CDX2 and CDC42 act as regulators of aPKC signaling in different cell contexts and are essential for cell polarity (Jedrusik et al., 2008; Joberty et al., 2000). Cdc42 null EBs also have defects in cell polarity, the formation of AJs and TJs, and complete failure of cavitation (Wu et al., 2007). In addition, loss of Cdc42 leads to the reduction of aPKC phosphorylation and abnormal localization. Similarly, reducing Cdx2 with inhibitory shRNA in Caco-2 cysts results in impaired apical basal polarity, multiple small cavities and mis-localized and ectopic distribution of aPKCs (Gao and Kaestner, 2010). In addition, inhibition of Par6B or aPKCs during the formation of Caco-2 cysts results in the multiple lumen formation (Durgan et al., 2011) not unlike the micro-cavities we see in Prkci^−/− EBs and embryos. Thus data presented here in combination with previous studies using diverse assays suggest that impaired signaling due to loss of Cdc42, Cdx2, aPKC, or Par6 leads to defects in cell polarity, epithelial organization, and cavitation in mammalian cells. Thus, ultimately proper cell polarity mediated by aPKCs and associated proteins may be the essential feature that drives normal morphogenesis during EB development.

A recent study on the role of Prkci in mouse embryogenesis has shown that over-expression of the Prkcz isozyme partially recovers the Prkci^−/− embryo phenotype at embryonic day 7.5 (Seidl et al., 2013). The localization of polarity proteins was not investigated in this context, however they are most probably restored. Thus, although not required, Prkcz likely has a similar role in regulating polarity in the absence of Prkci. Indeed, the polarized micro-cavities we observe in Prkci^−/− EBs do not form when both iota and zeta aPKC isoforms are inhibited (Mah et al., 2015). Further studies that investigate the relative role and contribution of Prkci and Prkcz are needed to determine if both proteins have identical or unique functions.

BMP signaling and the aPKC complex

Although the expression of Bmp2 and Bmp4 RNA is not greatly affected in Prkci^−/− EBs (Fig. S3A), a diminished response to BMP signaling is apparent (as evidenced by a decrease in pSMAD1/5 expression, particularly in EB core cells). BMP signaling appears to be most affected in Prkci^−/− EBs prior to cavitation since pSMAD1/5/8 is strongly expressed in early Prkci^−/− EBs but then is much decreased later at day 7–9 (Fig. 3B’). Treatment with BMP4 protein restored some pSmad1/5 expression in null EBs and increased cell death as measured by TUNEL, however ENDO A expression, the formation of epithelia, and cavitation were not completely restored (Fig. 3F’, J, K, M). The failure of BMP4 protein to fully rescue may be explained because a proposed interaction between phosphorylated SMAD1/5/8 with the aPKC complex at the tight junctions fails to occur (Eom et al., 2011). We therefore hypothesize that in the absence of Prkci pSMAD1/5/8 does not associate well with the aPKC complex and this impaired interaction could inhibit BMP signaling. Another possibility is that even if pSMAD1/5/8 can associate with the aPKC complex, it is rapidly degraded since recent studies also show that the depletion of Prkci in esophageal squamous cell carcinomas can enhance the ubiquitin-proteasome pathway (Liu et al., 2011). Regardless of mechanism, adding BMP4 is not able to rescue cell polarity. The partial rescue in cavitation may simply be due to a mild stimulation of BMP signaling which results in an increase in the number of dying cells and has little impact on cell polarity and epithelial organization.

EZRIN as a downstream effector of PRKCI

Overexpression of Ezrin in null EBs increased the number of micro-cavities and resulted in some recovery of cavitation. An improved polarity in some cells might lead an increase in the number of micro-cavities or an increase in their stability such that they would be more likely to fuse and resolve into a single cavity. Ezrin overexpression also rescued the pattern of ENDO A expression and brought the phosphorylation of SMAD1/5 to normal levels (Fig.4G–H). Rescuing cell polarity could restore the local release and trafficking of BMP or could improve the ability of null cells to respond to BMP signaling (perhaps by restoring normal receptor localization). In addition, an improved ability to segregate cellular determinants could have a broader impact. Indeed, we also found that the number of OCT4 positive cells was reduced with Ezrin overexpression and this could also have occurred by enhancing the ability of cells to segregate determinants that promote differentiation (Fig.4I–I’). However, like with BMP4 treatment, the full restoration of an organized and polarized columnar epithelium failed. There could be a number of technical reasons for this (expression needs to be higher and less transient), however another possibility is that factors in addition to Ezrin are needed. Indeed Ezrin is unlikely to mediate all functions downstream of Prkci because Ezrin knockout mice survive to birth (Saotome et al., 2004) while Prkci^−/− embryos do not develop past E8.5. In summary, overexpression of Ezrin enhances the formation of the initial multiple small lumens found in Prkci^−/− EBs likely by enhancing cell polarity. In some cases cavitation occurs because multiple micro-cavities fuse however, a fully polarized ectodermal epithelium does not persist.

Rescuing the null phenotype

When Prkci^−/− cells are mixed with wildtype cells, we found that the expression pattern of polarity markers was restored, a normal pseudostratified columnar epithelium formed, and complete cavitation was recovered (Fig. 5). This experiment suggested that wildtype cells can influence Prkci^−/− cells in a non-cell-autonomous fashion perhaps by releasing an extracellular signal. We also considered that the influence of wildtype cells could be indirect. For example, wildtype cells could have sufficient levels of the Na+/K+ ATPase pump to increase fluid influx and to resolve multiple micro-cavities into a single lumen as is seen in blastocyst formation (Eckert et al., 2004; Krupinski and Beitel, 2009) and zebrafish gut development (Bagnat et al., 2007). However to date there is no evidence that EB or ICM cavitation involves fluid influx or non-transport functions for the Na+/K+ ATPase (Barcroft et al., 2004; Madan et al., 2007). In addition, Prkci^−/− cells become strongly polarized in the presence of wildtype cells suggesting the existence of a direct non-autonomous influence. Interestingly, our results also suggest that PRKCI is not required in the responding cells when this non-cell-autonomous signal is provided. However, we speculate that the likely signal is an ECM protein or a secreted growth factor (Bryant and Mostov, 2008) and have included a list of candidates expressed in control EBs but not in Prkci^−/− EBs prior to cavitation that will require in-depth follow-up (Fig. 6). The identification of these signal/s could be useful for applications where cell polarity is disrupted such as in epithelial cancers. Such a signal might help stimulate polarity and differentiation instead of continued growth into the lumen and metastasis.

Regarding the regulation of genes in different disease and cell function pathways, we were not surprised to find that the cancer term was identified as a major pathway in our IPA analysis. Aberrant over-expression of aPKCs in epithelial cancers has been observed in multiple contexts and drugs that antagonize aPKC function are in various stages of clinical trial (Fields and Regala, 2007; Jin et al., 2008; Murray et al., 2011) (Jin et al., 2008; Mansfield et al., 2013). In addition, genes in the molecular transport category were also affected. The Na⁺/K⁺ ATPase was not included in the list but a number of Ca⁺², Na⁺, and K⁺ transporters were both up- and down-regulated suggesting that the polarity status, as mediated by Prkci, can feedback on the expression of ion channels. Genes related to nervous system biology included 12 genes important for motor coordination which may reflect an increase in neurogenesis in Prcki^−/− as observed previously (Mah et al., 2015). While genes altered in the lipid metabolism category included 12 genes related to the release of fatty acids. Farese and colleagues have studied the role of aPKC function in relation to diabetes and obesity for many years in hepatocytes (Farese et al., 2014) and our results suggest that some basic functions of aPKC in these pathways could potentially be studied in a simple EB model.

In summary, we suggest that cavitation is largely a process mediated by cell polarity and the ability to form polarized epithelia rather than a BMP-mediated cell death process as proposed by Coucouvanis and Martin, 1999. Our study is in agreement with recent studies showing that a polarity cue is required for epithelia formation, leading to normal cavitation of the mouse embryo (Bedzhov and Zernicka-Goetz, 2014). Thus one possibility is that in our EBs made of mixed cells, wild-type cells provide important extra-cellular cues that can coordinate cell polarity even in the absence of Prkci. Studies to identify which signals are critical and how they work will be an important future direction.

Supplementary Material

Table S4

Table S4. Two-way ANOVA analysis of Prkci^−/− vs. Prkci^+/− EBs at the pre-cavitation stage (p<0.05 and fold expression <-1.5 and >1.5)

Highlights.

• Loss of Prkci results in abnormal cell polarity and failed cavitation.

• Cell death and proliferation are largely unaffected in Prkci null EBs.

• BMP4 treatment or overexpressing Ezrin can only partially rescue the phenotype.

• A non-autonomous influence from wild-type cells rescues the null phenotype.