Par6B and Atypical PKC Regulate Mitotic Spindle Orientation during Epithelial Morphogenesis*

Cdc42 plays an evolutionarily conserved role in promoting cell polarity and is indispensable during epithelial morphogenesis. To further investigate the role of Cdc42, we have used a three-dimensional matrigel model, in which single Caco-2 cells develop to form polarized cysts. Using this system, we previously reported that Cdc42 controls mitotic spindle orientation during cell division to correctly position the apical surface in a growing epithelial structure. In the present study, we have investigated the specific downstream effectors through which Cdc42 controls this process. Here, we report that Par6B and its binding partner, atypical protein kinase C (aPKC), are required to regulate Caco-2 morphogenesis. Depletion or inhibition of Par6B or aPKC phenocopies the loss of Cdc42, inducing misorientation of the mitotic spindle, mispositioning of the nascent apical surface, and ultimately, the formation of aberrant cysts with multiple lumens. Mechanistically, Par6B and aPKC function interdependently in this context. Par6B localizes to the apical surface of Caco-2 cysts and is required to recruit aPKC to this compartment. Conversely, aPKC protects Par6B from proteasomal degradation, in a kinase-independent manner. In addition, we report that depletion or inhibition of aPKC induces robust apoptotic cell death in Caco-2 cells, significantly reducing both cyst size and number. Cell survival and apical positioning depend upon different thresholds of aPKC expression, suggesting that they are controlled by distinct downstream pathways. We conclude that Par6B and aPKC control mitotic spindle orientation in polarized epithelia and, furthermore, that aPKC coordinately regulates multiple processes to promote morphogenesis.

Cdc42 is a small GTPase that plays a fundamental role in promoting cell polarity across evolution and in a wide range of biological contexts (1). For instance, Cdc42 is required for bud emergence in Saccharomyces cerevisiae (2), asymmetric cell division in Caenorhabditis elegans (3) and directed cell migration in astrocytes (4). In epithelial cells, Cdc42 is indispensable during morphogenesis, controlling tight junction formation (5)(6)(7), the delivery of basolateral proteins (8), and apical surface formation through the regulated trafficking of vacuolar apical components (9). To further investigate the role of Cdc42, we have exploited a three-dimensional model of epithelial morphogenesis, in which single Caco-2 cells are cultured in matrigel to form polarized cysts (Fig. 1A). Using this system, we previously reported that Cdc42 controls mitotic spindle orientation to correctly position the nascent apical surface in a growing cyst (10). These findings highlight an intriguing relationship between epithelial mitosis and morphogenesis, which couples cell proliferation to tissue architecture.
A major challenge now is to determine the specific upstream regulators and downstream effectors through which Cdc42 controls epithelial morphogenesis. Recent work has focused on the identification of specific guanine nucleotide exchange factors that activate Cdc42 in this context. RNAi-based screens undertaken in Madin-Darby canine kidney (MDCK) 3 cysts have identified two distinct guanine nucleotide exchange factors, Tuba (11) and Intersectin 2 (12), which control the recruitment and activation of Cdc42, at a subapical domain, and the centrosome, respectively, to facilitate mitotic spindle orientation and proper morphogenesis. The downstream effectors through which Cdc42 regulates this process have yet to be systematically investigated.
In non-polarized, adherent cells, Cdc42 orients the mitotic spindle, parallel to the substratum, through two distinct pathways (13). Cdc42 regulates the actin cytoskeleton, through PAK2 and ␤Pix, and activates PI3K during mitosis. Both pathways are required for the localized accumulation of dynein⅐dynactin complexes at the midcortex and thus for proper spindle alignment (13). However, in polarized epithelial cells, spindle orientation is PI3K-independent (14), indicating that a different mechanism is at play. Furthermore, an additional requirement in these cells is that the spindle is aligned perpendicular to the axis of apical-basal polarity (10,15,16). The Cdc42 effector(s) that mediates spindle orientation in polarized epithelia is currently unknown.
Par6 (partitioning-defective 6) is a known Cdc42 effector (17)(18)(19)(20) with a critical role in establishing cell polarity during asymmetric division (21,22), directed migration (4), and the apical-basal polarization of epithelial cells (7,17,22). Par6 interacts specifically with Cdc42 in its GTP-bound, active form. Cdc42 binding induces a conformational change in Par6 (23) and promotes its polarized recruitment to specific subcellular compartments (24 -27). In the present study, we reasoned that Par6 represents a strong candidate as the Cdc42 effector that controls polarized spindle orientation during mammalian epithelial morphogenesis, and we sought to test this using the Caco-2 cystogenesis model.
In the present study, we investigated the hypothesis that Par6 and aPKC act downstream of Cdc42 to promote epithelial morphogenesis. We report that Par6B and aPKC function interdependently to control mitotic spindle orientation and proper positioning of the apical surface. In addition, we find that aPKC activity is indispensable for epithelial cell survival, suggesting that this kinase coordinately regulates multiple processes during epithelial morphogenesis.

EXPERIMENTAL PROCEDURES
Plasmids and Cloning-All primers are listed in the supplemental material. All point mutations were introduced by QuikChange site-directed mutagenesis, using Pfu Turbo (Stratagene). Mouse Par6B was rendered resistant to siRNA duplex siPar6B.3 with two silent mutations and subcloned into pQCXIP (BamHI/EcoRI), with a C-terminal HA tag, by PCR. A ⌬P136 deletion mutant was engineered, which is Cdc42 binding-deficient. Rat aPKC cDNA was kindly provided by Professor Peter Parker (Cancer Research UK, London). Two silent mutations were introduced to render aPKC resistant to siRNA duplex aPKC.1. Full-length or ⌬PB1 RNAi-resistant aPKC was then subcloned into pQCXIP (AgeI/EcoRI). Additional point mutations were introduced as follows: D62A, A118E, and D375A. Cdc42 L61 in pRK5myc has been described previously (48). All constructs were fully sequenced.
Cell Culture-293FT cells were maintained as recommended by Invitrogen. Caco-2 (10) and 16HBEo-(7) cells were cultured as described previously; stable lines were selected with 6 g/ml puromycin (InVivoGen). Three-dimensional cysts were grown on top of or embedded in Matrigel (BD Biosciences, catalog no. 354230) as described previously (10), or in 2% matrigel on glassbottomed, 4-well Lab-Tek chamber slides (Nunc). Cysts were routinely stimulated with 0.1 g/ml cholera toxin for 16 h prior to fixation to induce synchronous lumen expansion.
RNAi-All siRNA duplexes were purchased from Dharmacon (see supplemental material). Transfections were performed as described previously (10), except that 10 5 cells were seeded/6-well dish. 1 day post-transfection, cells were reseeded as indicated, or the medium was changed. For titrations, siaPKC.1 was mixed with siLamin A/C in different proportions to maintain a final siRNA concentration of 50 nM.
Retrovirus Production and Infection-Retroviral particles were prepared for the delivery of pQCXIP-based plasmids (see supplemental material). mCherry-H2B retrovirus was kindly provided by Dr. Oliver Florey (Memorial Sloan-Kettering Cancer Center). 10 5 Caco-2 cells were seeded/6-well dish and infected with 1.5 ml of viral suspension plus 8 g/ml Polybrene by centrifugation at 2250 rpm for 30 min. The viral suspension was removed and replaced with fresh medium; where indicated, puromycin selection was initiated 2 days later. To avoid clonal effects, all experiments were performed with short term cultures of selected, stable pools.
FACS Analysis-4 days post-transfection, cells were fixed and stained with propidium iodide as described in the supplemental material. DNA content was measured using a FAC-SCalibur machine (BD Biosciences) to capture 20,000 events/ sample/experiment; analysis was performed using FlowJo software. The data presented are representative of three separate experiments.
Western Blotting-Total cell extracts or radioimmune precipitation assay lysates were prepared as described in the supplemental material. PAGE and Western blotting were performed as described previously (49); Par6B and Par6B-HA were resolved using 10% BisTris gels. Antibody sources and dilutions are provided in the supplemental material. Blots were scanned using a Bio-Rad GS-800 calibrated densitometer, and band intensities were quantified using ImageJ software (National Institutes of Health). All Western blots shown are representative of data obtained from at least three separate experiments.
Time Lapse Microscopy-For video microscopy, cells were seeded on glass-bottomed, 6-well dishes (MatTek). Time lapse movies were acquired using a Nikon Eclipse Ti-U microscope, equipped with an S Plan Fluor ϫ20 ELWD objective (Nikon) and a Cool Snap HQ2 CCD camera (Photometrics), controlled by NIS-Elements imaging software (Nikon). Movies were processed and annotated using ImageJ software (National Institutes of Health).
Line Intensity Scans-The distribution of aPKC staining in three-dimensional cysts was analyzed using ImageJ (National Institutes of Health); a straight line was drawn along the apicalbasal axis of one cell per cyst, along which signal intensities were measured. To determine the ratio between apical and cytosolic signals, intensities at the apical surface (defined by the peak in actin) were compared with the mean intensity across the rest of the cell. A total of 10 cysts were analyzed per condition, from three separate experiments.
Spindle Orientation Assay-Three days post-siRNA transfection, 7 ϫ 10 5 Caco-2 cells were embedded in matrigel and seeded in an 8-well chamber slide (BD Falcon, catalog no. 354108) for 24 h. To increase the number of mitotic events detected, cells were synchronized at the G 2 /M border using 10 M RO-3306 for 16 h, washed, and then released into normal growth medium for 45 min. Growing cysts were fixed and stained for tubulin, actin, and DNA and imaged by spinning disk confocal microscopy, as described in the supplementary material. Spindle orientation was analyzed specifically during the early stages of cyst development (2-5 cells) in mitotic cells with a single apical surface. A complete z-stack was obtained for each cyst of interest, using 0.2-m intervals, and deconvolved using AutoQuant X2.2.0 software (Media Cybernetics). To obtain measurements, three-dimensional reconstructions were analyzed in Imaris software (Bitplane Scientific Software). Each spindle pole was marked as a point in three-dimensional space to give two sets of x,y,z co-ordinates (p1, p2), which define the spindle axis (p2-p1). The axis of apical-basal polarity was defined as a straight line through the center of the mitotic cell, running perpendicular to its apical surface. To determine this axis, three points were marked along the apical surface (p3, p4, p5) to construct two vectors (v1 ϭ p4-p3; v2 ϭ p5-p3), and the cross-product of these vectors was calculated to give the perpendicular line. Using Matlab software (MathWorks), the angle () between the spindle axis (x) and the apical-basal axis (y) was calculated using the formula, cosine ϭ dot(x,y)/ (norm(x)*norm(y)); angle ϭ acos(cos)*180/.
Statistics-Statistical significance was evaluated using Prism software. Mann-Whitney or unpaired Student's t tests were performed, with two-tailed p values and 95% confidence intervals.

Par6B Regulates Caco-2 Three-dimensional Morphogenesis-
We recently reported that Cdc42 controls Caco-2 morphogenesis by orienting the mitotic spindle, perpendicular to the apical-basal axis, during cell division ( Fig. 1A) (10). Misorientation of the spindle has profound implications for epithelial architecture, promoting cell division in the wrong plane, mispositioning the newly formed apical surface between daughter cells, and ultimately forming defective cysts with multiple lumens. In the present study, we sought to identify the downstream effector through which Cdc42 regulates spindle orientation and epithelial morphogenesis, using the formation of Caco-2 cysts with multiple lumens as a read-out. Taking a candidate approach, we focused on the polarity protein Par6, a known target of Cdc42 (17)(18)(19)(20).
Mammalian Par6 comprises a family of four related isoforms, A, B, C, and D (17). Preliminary studies using isoform-specific pools of siRNA indicated that depletion of Par6B was sufficient to disrupt Caco-2 morphogenesis (supplemental Fig. 1). To validate this observation, four individual siPar6B duplexes were tested, which depleted protein levels to varying extents (Fig.  1B); aPKC provided a loading control. Cells treated with either control or Par6B-specific siRNA were cultured in matrigel, and the resulting cysts were visualized by phase microscopy (Fig.  1C). Depletion of Par6B induced a gross defect in morphogenesis, comparable with the loss of Cdc42. Although Ͼ80% of control structures possessed a single, central, fluid-filled lumen, only around half of siPar6B cysts exhibited this normal morphology (arrowheads indicate defective structures).
To examine this morphological defect in more detail, fixed cysts were analyzed by confocal microscopy, staining for DNA to mark individual cells and for actin to visualize the apical surface ( Fig. 1, D and E). Depletion of Par6B phenocopied the loss of Cdc42, inducing the formation of aberrant cysts with multiple lumens. Importantly, this phenotype was observed with all four duplexes tested, indicating that this is a specific effect. Moreover, the level of Par6B knockdown correlated closely with the severity of the phenotype (Fig. 1, compare B and D).
Although partial depletion of Par6B promoted the formation of aberrant lumen in Caco-2 cysts, it did not disrupt apicalbasal polarity per se. Each individual lumen was surrounded by apical markers, such as actin (Fig. 1E) and ezrin (supplemental Fig. 2A), and both tight junctions and adherens junctions appeared intact, as judged by occludin and E-cadherin staining, respectively (supplemental Fig. 2, B and C). Together, these data reveal that Par6B controls the proper positioning of the apical surface during Caco-2 morphogenesis.
Par6B Functions Downstream of Cdc42 to Promote Caco-2 Morphogenesis-To confirm that Par6B functions downstream of Cdc42 during cyst development, an RNAi rescue experiment was performed. Stable pools of Caco-2 cells were derived, expressing either an empty vector (pQCXIP) or a mouse Par6B construct, C-terminally HA-tagged, and rendered resistant to duplex siPar6B.3 ( Fig. 2A). The function of wild type Par6B was compared with a ⌬P136 mutant, which is Cdc42 binding-deficient ( Fig. 2B) but still able to interact with other key partners, such as Par3, aPKC, and Lgl1 (50). As shown in Fig. 2, C and D, expression of wild type Par6B-HA was able to rescue proper morphogenesis following depletion of the endogenous protein.
In contrast, expression of the ⌬P136 mutant was unable to rescue the multilumen defect, indicating that Cdc42 binding is indispensable for Par6B function in this context. These data are consistent with a model in which Par6B acts as a downstream effector for Cdc42 during Caco-2 morphogenesis.
aPKC Regulates Caco-2 Three-dimensional Morphogenesis-Par6 typically functions with its binding partner, aPKC, to mediate signaling downstream of Cdc42 (32,51,52). To test whether aPKC is also required during Caco-2 morphogenesis, two non-overlapping duplexes were designed, each of which efficiently co-depletes the two mammalian isoforms of aPKC FIGURE 1. Par6B regulates Caco-2 three-dimensional epithelial morphogenesis. A, Caco-2 three-dimensional model of morphogenesis. During normal Caco-2 cyst development, the mitotic spindle is precisely oriented, perpendicular to the apical-basal axis, so that cells divide in the plane of the monolayer. Abscission occurs asymmetrically, and the nascent apical surface is deposited between daughter cells around a single, central lumen. Depletion of Cdc42 leads to misorientation of the spindle, cell division in the wrong plane, and ultimately, the formation of aberrant cysts with multiple lumens (10). The apical surface (red), microtubules (green), and chromosomes (blue) are indicated. B-E, Caco-2 cells were transfected with siRNA and reseeded as follows. B, on 6-well dishes. Total cell extracts were prepared 8 days post-transfection for Western blotting. C and D, "on top" of matrigel. Seven days post-transfection, cysts were treated with CTX to induce synchronous lumen expansion. C, live cysts were imaged using phase-contrast microscopy to visualize gross morphology; arrowheads indicate abnormal structures. The mean percentage of cysts with a normal morphology (single, central lumen) is indicated Ϯ S.D. Scale bars, 100 m. D, cysts were fixed and stained for actin/DNA to score for the presence of multiple lumens. At least 100 cysts were counted per condition. Bars, mean Ϯ S.D. (error bars) from three separate experiments. Student's t tests were performed to determine statistical significance. *, p Ͻ 0.02; ***, p Ͻ 0.0005. E, embedded in matrigel. Seven days post-transfection, cysts were stimulated with CTX and fixed and stained for actin (red) and DNA (blue). Single confocal sections are shown. Scale bars, 25 m.
( Fig. 3A; antibody specificity was established using recombinant proteins (supplemental Fig. 3)). Depletion of aPKC with either duplex induced major defects in morphogenesis; only around one-third of siaPKC cysts possessed a single, central, fluid-filled lumen ( Fig. 3B; arrowheads indicate defective structures). Notably, aPKC-depleted cysts were also significantly smaller and fewer than control structures, suggesting an additional role for this protein in either cell proliferation or survival (see below). Confocal imaging confirmed that depletion of aPKC increases the formation of Caco-2 cysts with multiple lumens (Fig. 3, C and D). We conclude that aPKC functions with Cdc42 and Par6B to control apical surface positioning.
Surprisingly, no obvious defects in apical-basal polarity or junction formation were detected following aPKC depletion under these conditions; each individual lumen in an aberrant cyst was surrounded by apical markers (e.g. actin (Fig. 3D) or ezrin (data not shown)) and filled with fluid, suggesting intact barrier function. We believe that a stronger knockdown is likely to be required to disrupt polarity and junctions, but we did not explore this further in the present study (see "Discussion").
Par6B and aPKC Control Mitotic Spindle Orientation-Cdc42 promotes proper epithelial morphogenesis by orienting the mitotic spindle during cell division (Fig. 1A) (10). To test whether Par6B and aPKC act through the same mechanism, these proteins were depleted, and spindle orientation was measured during the early stages of cyst development (Fig. 4A). Under control conditions, the majority of spindles were oriented perpendicular to the axis of apical-basal polarity (Fig. 4, B and C; median angle ϭ 82.2°), consistent with previous reports (10). However, upon depletion of either Par6B or aPKC, spindle orientation was essentially randomized (Fig. 4, B and C). We conclude that, like Cdc42, the Par6B⅐aPKC complex controls mitotic spindle orientation during cell division to correctly position the nascent apical surface in a growing cyst.
Par6B Is Apically Localized-In many contexts, the ability of Par6 to promote polarization depends upon its own asymmet- rical distribution (24 -27). However, the Par6B isoform has been reported to localize uniformly throughout the cytosol in mammalian epithelia, at least when epitope-tagged (N-terminally) and exogenously expressed (17,53). To investigate this further, Caco-2 cysts were stained for Par6B (Fig. 5A). Surprisingly, the endogenous protein was found to concentrate along the apical surface, together with actin. Importantly, this apical signal was reduced in cysts depleted of Par6B, confirming the specificity of the antibody. Furthermore, exogenously expressed Par6B, bearing a C-terminal HA tag, localized to the same compartment (Fig. 5B).
To determine whether the apical localization of Par6B is induced specifically in a three-dimensional environment, we cultured Caco-2 cells on glass coverslips for immunofluorescence. Mature monolayers were stained for either the endogenous (Fig. 5C) or exogenous protein (Fig. 5D). In both cases, a clear signal was detected at epithelial junctions, indicating that Par6B is polarized in two-dimensional as well as three-dimensional cultures.
To address whether the apical enrichment of Par6B is a cell type-specific phenomenon, we stained monolayers of 16HBE, a human bronchial epithelial cell line. In this cell line, a cytosolic signal could be detected, but both endogenous (Fig. 5C) and exogenous ParB (data not shown) were also observed at apical junctions, consistent with our observations in Caco-2. Together, these data reveal that Par6B is an apical protein and suggest that its asymmetric distribution may be important during epithelial morphogenesis.
Par6B Localizes aPKC to the Apical Surface-During the directed migration of astrocytes, Par6 recruits aPKC to the leading edge of the cell (4). To determine whether Par6B is similarly required to localize aPKC in Caco-2 cysts, siRNA-treated structures were analyzed by immunofluorescence (Fig. 6, A-C). Under control conditions, aPKC was polarized to the apical surface, consistent with previous reports (10). Strikingly, depletion of Par6B attenuated the apical enrichment of aPKC, whereas actin remained concentrated at the apical surface of each individual lumen. The reduction in apical aPKC clearly resulted from a change in its localization because total protein levels did not change following Par6B depletion (Fig.  1B). These data suggest that Par6B localizes aPKC to the apical surface to control morphogenesis. Notably, depletion of Cdc42 did not mislocalize aPKC (Fig. 6, A-C). Because Cdc42 can activate the Par6⅐aPKC complex in vitro (19,28), we surmise that Par6B acts to localize aPKC for subsequent activation by Cdc42.
To determine whether aPKC kinase activity is required for morphogenesis, pharmacological inhibitors were employed. A myristoylated peptide, based on the pseudosubstrate sequence of aPKC, is an effective aPKC inhibitor (46,47). However, this peptide appeared not to enter matrigel efficiently and was not amenable to these experiments. As an alternative, two small molecule inhibitors with different, but overlapping, specificities for PKC family members were exploited (54). Gö6983 inhibits all PKC isoforms, whereas BIM1 inhibits classical PKC and novel PKC (nPKC) subtypes but not aPKCs; the comparative use of these compounds is an established means of testing the requirement for aPKC activity (55). As shown in Fig. 6D, Gö6983, but not BIM1, induced a significant increase in the formation of cysts with multiple lumens, consistent with a specific role for aPKC kinase activity during morphogenesis. We conclude that aPKC promotes proper spindle orientation through the phosphorylation of downstream substrates.
aPKC Stabilizes Par6B through a Kinase-independent Mechanism-During the course of these studies, we also uncovered an unexpected role for aPKC in the regulation of Par6B. Depletion of aPKC, with either of two distinct duplexes, induced a corresponding down-regulation of Par6B protein expression (Fig.  7A). Under these conditions, Par6B levels could be partially restored by treatment with the proteasome inhibitor MG132, suggesting that in the absence of aPKC, Par6B is subject to proteasome-mediated degradation. To investigate the molecu-lar basis for Par6B stabilization, we performed RNAi rescue experiments, using a panel of aPKC domain mutants (Fig. 7B); these constructs were resistant to duplex siaPKC.1 but sensitive FIGURE 4. Par6B and aPKC control mitotic spindle orientation during Caco-2 three-dimensional morphogenesis. Three days post-siRNA transfection, Caco-2 cells were embedded in matrigel for 24 h, synchronized at the G 2 /M border using 10 M RO-3306 for 16 h, and then released into normal growth medium for 45 min. Cysts were fixed and stained for tubulin (green), actin (red), and DNA (blue) and imaged by spinning disk confocal microscopy. Spindle orientation was analyzed specifically during the early stages of cyst development (2-5 cells) in mitotic cells with a single apical surface. A, schematic summarizing the procedure for measuring mitotic spindle orientation. A correctly oriented spindle is angled at 90°with respect to the axis of apical-basal polarity. B, representative three-dimensional projections are shown. Scale bars, 5 m. The calculated spindle angle is indicated. C, quantification of spindle angles. The scatter graph represents data acquired across three separate experiments, and the total number of cells analyzed for each condition is indicated (n). The median angle is represented by a red circle and black line. Error bars, interquartile range. Statistical significance was evaluated using a Mann-Whitney test. **, p Ͻ 0.002; ***, p Ͻ 0.001. APRIL 8, 2011 • VOLUME 286 • NUMBER 14 to siaPKC.2 (Fig. 7C, top). As expected, expression of RNAiresistant wild-type aPKC* rescued the degradation of Par6B following treatment with duplex 1 but not duplex 2 (Fig. 7C). The D375A mutant was equally efficient in restoring Par6B levels, indicating that aPKC kinase activity is dispensable for this function. In contrast, neither Par6 binding mutant (⌬PB1 or D62A) was able to rescue, suggesting that physical association with aPKC is essential. From these data, we predicted that Par6B should be bound stoichiometrically to aPKC in Caco-2 cells. In agreement with this, gel filtration chromatography revealed that Par6B co-eluted with aPKC in fractions containing high molecular weight complexes (Fig. 7D), whereas monomeric Par6B was undetectable. We conclude that aPKC stabilizes Par6B in a kinase-independent manner and that the physical interaction of the two proteins protects Par6B from proteasomal degradation. Together, these data reveal that Par6B and aPKC function interdependently during epithelial morphogenesis.

Par6B and aPKC Control Spindle Orientation during Morphogenesis
aPKC Controls Cell Survival-In addition to inducing multiple lumens, aPKC depletion also led to the formation of consistently smaller and fewer cysts (Fig. 3, B and D). This observation suggests that aPKC plays an additional role during morphogenesis in the regulation of cell number. To confirm this, Caco-2 cells were transfected with siaPKC.1/2 and counted over 8 days. Growth curves show that depletion of aPKC with either duplex led to a significant decrease in cell number (Fig. 8A). To validate this observation, a rescue experiment was performed using aPKC* (resistant to duplex 1 but not duplex 2). Stable pools of cells, expressing either an empty vector or aPKC*, were transfected with siRNA and counted 6 days later (Fig. 8B). Expression of aPKC* partially rescues the decrease associated with siaPKC.1, but not siaPKC.2, confirming that depletion of aPKC specifically reduces Caco-2 cell number.
To examine this phenotype in more detail, the DNA content of siRNA-treated cells was analyzed by FACS. There was no detectable effect of aPKC depletion on cell cycle progression (Fig. 8C). However, a significant increase in the sub-G 1 fraction was observed, indicative of DNA fragmentation and cell death (Fig. 8D). Using time lapse microscopy, Caco-2 morphology was analyzed following treatment with siaPKC; these experiments were performed using cells stably expressing mCherryhistone 2B, so nuclear changes could also be monitored (Fig. 8E and supplemental Movie 1). Depletion of aPKC induces substantial DNA fragmentation and membrane blebbing, characteristic of apoptotic cell death (56). Wide field movies (phase only), captured over a longer time course, reveal the penetrance of this phenotype and highlight the dramatic effect on endpoint cell number (supplemental Movie 2). Importantly, comparable morphological changes were observed in aPKC-depleted cysts (Fig. 8F), suggesting that cell death is similarly induced in both two-and three-dimensional cultures. Cysts were fixed and stained for Par6B (green), actin (red), and DNA (blue). Images were acquired with fixed laser settings, so signal intensities may be compared. B, Caco-2 cells stably expressing either empty vector (pQCXIP) or wild type Par6B-HA were transfected as indicated and cultured in matrigel. Cysts were fixed and stained for HA (green) and DNA (blue). C, Caco-2 or 16HBE cells were seeded on glass coverslips and cultured to confluence. Mature monolayers were fixed and stained for Par6B (green) and DNA (blue). D, Caco-2 cells stably expressing either empty vector (pQCXIP) or wild type Par6B-HA were seeded on glass coverslips and cultured to confluence. Mature monolayers were fixed and stained for HA (green) and DNA (blue). All images shown are representative, single confocal sections. Scale bars, 25 m.
Finally, we sought to determine whether aPKC kinase activity is required for cell survival. RNAi rescue experiments were precluded because cell lines expressing kinase-dead aPKC (D375A) could not be stably maintained. This observation suggested that the inactive construct may behave as a dominant negative with respect to cell survival. To explore this further, cells were incubated with the myristoylated, aPKC pseudosubstrate peptide inhibitor; as a control, the myristoylated, nPKC pseudosubstrate peptide was employed (Fig. 8G). Whereas cells treated with the nPKC inhibitor appear similar to controls, the aPKC inhibitor induces synchronous cell death. Cells fixed 4 h after treatment show a significant increase in nuclear condensation and cleaved caspase 3 staining, indicative of apoptosis. We conclude that aPKC promotes cell survival in a kinase-dependent manner.
Apical Positioning and Cell Survival Depend upon Different Thresholds of aPKC Expression-The data presented indicate that aPKC controls both apical surface positioning and cell survival during morphogenesis. Because cysts with multiple lumens are readily detectable upon aPKC depletion (Fig. 3), despite the significant increase in apoptotic cell death (Fig. 8), we reasoned that these may represent distinct functions that depend upon different levels of aPKC expression. To test this, different amounts of siaPKC.1 were transfected into Caco-2 to titrate aPKC protein levels; cells were then seeded in parallel to analyze either cell number or three-dimensional morphogenesis (Fig. 8H). Cells expressing intermediate levels of aPKC (after treatment with 10 nM siRNA) were significantly impaired in apical surface positioning, with ϳ25% of cysts possessing multiple lumens. However, cell number was unchanged under these conditions. Upon a more robust depletion of aPKC (using 50 nM siRNA), the frequency of morphological defects was further increased to around 40%; however, cell number was now also significantly reduced. We conclude that apical positioning and cell survival depend upon different thresholds of aPKC expression. Positioning of the apical surface is sensitive to even a modest decrease in aPKC, whereas a significantly stronger reduction is required to induce apoptosis.

DISCUSSION
We have used a three-dimensional tissue culture model to explore the downstream effectors though which Cdc42 con- . C, mean ratios between apical and cytosolic signals. 10 cysts were imaged per condition, from three separate experiments, and each cyst was analyzed as described in B. The signal intensity at the apical surface (defined by the peak actin value) was compared with the mean signal intensity in the cytosol. Bars, mean ratio Ϯ S.D. (error bars). Student's t tests were performed to determine statistical significance. ***, p Ͻ 0.001. D, inhibitor studies. Caco-2 cells were seeded "on top" of matrigel and incubated for 3 days. Growing cysts were treated with or without DMSO (carrier control), 1 M BIM1 (control inhibitor), or 6 M Gö 6983 (aPKC inhibitor); incubated for a further 3 days; and then fixed and stained for actin/DNA, to score for the presence of multiple lumens. At least 100 cysts were counted per condition. Bars, mean Ϯ S.D. (error bars) from three separate experiments. Student's t tests were performed to determine statistical significance. ***, p Ͻ 0.001. trols mitotic spindle orientation during epithelial morphogenesis. Here, we report that Par6B and aPKC function interdependently to control this process and, furthermore, that aPKC activity is also indispensable for epithelial cell survival under basal conditions. During Caco-2 cyst development, the position of the apical surface is determined by the orientation of the mitotic spindle during cell division (10), highlighting an intimate relationship between epithelial mitosis and morphogenesis. Cdc42 is essential for orienting the spindle perpendicular to the apical-basal axis, and accordingly, depletion of Cdc42 promotes mispositioning of the apical surface and the formation of aberrant cysts with multiple lumens. In the present study, we explored the hypothesis that the Par6⅐aPKC polarity complex may be required to mediate Cdc42-dependent spindle orientation.
Using an RNAi-based approach, we found that depletion of Par6 phenocopies the loss of Cdc42, inducing misorientation of the mitotic spindle and the development of defective cysts with multiple lumens. Through rescue experiments, we confirmed that Cdc42 binding is indispensable for Par6 function, consistent with it acting as a direct Cdc42 effector. Specifically, we find that the Par6B species is required for cystogenesis, providing insight into isoform specificity within the mammalian Par6 family. However, our data do not exclude possible roles for other Par6 family members; the knockdown levels of Par6A and/or Par6G achieved here may simply be insufficient to induce a robust phenotype. Related to this, we note that mutation of zebrafish Pard6␥b leads to misorientation of the mitotic spindle and formation of multiple lumens in the developing neural tube (57).
Consistent with work in other systems, we find that Par6 functions together with its binding partner aPKC during epithelial morphogenesis. Depletion of aPKC leads to misorientation of the mitotic spindle, mispositioning of the apical surface, and ultimately, to aberrant cytogenesis. Surprisingly, we did not observe an obvious defect in apical-basal polarity or junction formation arising from depletion of either Par6B or aPKC under these conditions. It is likely that a partial depletion of these proteins is insufficient to completely disrupt polarity. With regard to Par6B, either a stronger knockdown may be required or the co-depletion of other Par6 isoforms, which may act redundantly. With respect to aPKC, we show that a more robust knockdown induces apoptotic cell death, precluding the analysis of cell polarity/junctions under these conditions. Cdc42, Par6, and aPKC are well known to control spindle orientation during asymmetric cell division, for instance during the first cleavage of the C. elegans embryo (30) and in Drosophila neuroblasts (29). In this context, spindle orientation controls cell fate. It is intriguing now to discover that conserved machinery regulates the symmetrical divisions of epithelia, where spindle orientation couples cell proliferation to tissue architecture. This also adds to the wide range of biological contexts in which Cdc42 signals through the Par6⅐aPKC complex (51).
With respect to molecular mechanisms, we identify a mutual dependence between Par6B and aPKC (Fig. 9A). First, aPKC protects Par6B from proteasomal degradation. This mechanism ensures that Par6B is constitutively bound to and able to signal through aPKC. Interestingly, aPKC kinase activity is not necessary for Par6B stabilization. Importantly, this finding excludes the possibility that aPKC functions solely to protect Par6B because its kinase activity is also indispensable for morphogenesis. In addition, these data add to the emerging view that PKC family members can mediate a collection of kinase-  Fig. 3D. Images were acquired using an epifluorescent microscope and a ϫ40 objective. Scale bar, 25 m. G, Caco-2 cells seeded on glass coverslips were treated with or without 50 M myristoylated pseudosubstrate inhibitors targeting either nPKC or aPKC. After 4 h, cells were fixed and stained for cleaved caspase 3 (CC3; green) and DNA (blue). Images were acquired using an epifluorescent microscope and a ϫ20 objective. Scale bar, 50 m. H, distinct aPKC functions depend upon different thresholds of protein expression. Caco-2 cells were transfected with a dilution series of siaPKC.1 and reseeded for Western blotting (as in Fig. 3A), counting (as in Fig. 8B), or three-dimensional analysis (as in Fig. 3C). Bars, mean Ϯ S.D. (error bars) from three separate experiments. Student's t tests were performed to determine statistical significance. *, p Ͻ 0.02; **, p Ͻ 0.003; ***, p Ͻ 0.0003. independent functions (58). To our knowledge, this is the first report to show that aPKC controls Par6 stability, although we predict that this effect may be relevant in other biological contexts that require the Par6⅐aPKC complex. In particular, it would be pertinent to investigate whether Par6 stabilization is important in processes that require aPKC protein but not kinase activity, such as the control of follicle cell polarity in Drosophila (59).
Conversely, Par6B plays a reciprocal role in localizing aPKC to the apical surface. This observation was initially surprising to us because Par6B itself is reported to distribute throughout the cytosol when exogenously expressed in MDCK cells (17,53). However, we find that endogenous Par6B is resident at the apical surface in Caco-2 cysts, suggesting that it directly recruits its interacting partner to this compartment. In addition to controlling aPKC localization, Par6 can also regulate its catalytic activity (32). As such, we envisage a two-step mechanism for aPKC activation in epithelial cells (Fig. 9A). Par6B recruits aPKC to the apical surface but is expected to hold the kinase there, poised in an inactive conformation. Binding of active Cdc42 to Par6B induces a conformational change, which relieves this inhibition and permits substrate phosphorylation (23). This mechanism would provide a precise means of regulating aPKC activity both spatially, through Par6B, and temporally, through Cdc42 (52).
The exogenous expression of aPKC was found to perturb Caco-2 cyst morphology, confounding further analysis of molecular mechanisms through RNAi rescue experiments. However, small molecule kinase inhibitors support a specific requirement for aPKC kinase activity during morphogenesis. Consistent with this, an inhibitory aPKC pseudosubstrate peptide has also been shown to misorient the spindle in MDCK cells (11). Furthermore, the zebrafish heart and soul mutation, which renders aPKC kinase deficient, randomizes spindle orientation in retinal progenitors, which can divide either symmetrically or asymmetrically (40). A major goal of our ongoing work is thus to identify substrates through which aPKC controls spindle orientation and positioning of the apical surface.
We have identified an additional function for aPKC during morphogenesis, in the regulation of Caco-2 cell number. Although aPKC is well known as a mitogen (60), we find no evidence for a change in cell cycle progression under the conditions tested. Instead, loss of aPKC activity promotes robust apoptotic cell death. Consistent with this, loss of aPKC function increases TUNEL staining at certain developmental time points in Drosophila (37) and Xenopus (39), and apoptotic markers have been observed in MDCK cells following aPKC depletion or inhibition (61). In the latter case, there was no corresponding change in cell number, possibly due to compensatory increases in proliferation (61,62). Here, we demonstrate that loss of aPKC has profound consequences for cyst size and number, indicating that support of cell survival is a further, vital function for aPKC during morphogenesis.
Atypical PKC is well known to protect cells from stimulusinduced death. aPKC suppresses apoptosis in leukemic cells following treatment with TNF␣, Fas ligand, or chemotherapeutic drugs, generally through activation of the NFB pathway (63,64), and in fibroblasts, inhibition of aPKC is induced by UV treatment (65) or Par-4 expression (66) prior to cell death. In the present study, we show that depletion or inhibition of aPKC is sufficient to trigger apoptosis in Caco-2, even under basal conditions. We conclude that aPKC must constitutively transduce critical survival signals. Because epithelial cells are the source of the majority of human tumors, their sensitivity to aPKC inhibition may present a significant therapeutic opportunity. Our data support the emerging view that aPKC represents a potential drug target in certain cancers (67).
Together, our current findings reveal that aPKC performs multiple functions during epithelial morphogenesis (Fig. 9B). Proper spindle orientation and cell survival depend upon different thresholds of aPKC expression, suggesting that they are likely to be regulated by distinct downstream pathways. The multifunctional nature of aPKC could have important implications for its therapeutic targeting. Although robust inhibition induces apoptosis, a desirable outcome in treating a tumor, cells would be expected to survive a partial inhibition but to be impaired in spindle orientation and apical surface positioning. These defects could disrupt epithelial architecture in proliferating tissues, potentially inducing unwanted side effects or even contributing to transformation (68). These may be important FIGURE 9. Par6B and aPKC regulate spindle orientation and cell survival during epithelial morphogenesis. A, Par6B and aPKC function interdependently to control spindle orientation downstream of Cdc42. aPKC protects Par6B from proteasome-mediated degradation in a kinase-independent manner. Reciprocally, Par6B localizes aPKC to the apical surface, restraining the kinase in an inactive conformation. Active Cdc42 binds to Par6, provoking a conformational change, which relieves the inhibition of aPKC. This two-step mechanism ensures that aPKC activity is precisely regulated, both spatially, by Par6, and temporally, by Cdc42. B, aPKC performs multiple, distinct functions during mammalian epithelial morphogenesis. Shown is a schematic summarizing the multiple roles of aPKC in mammalian epithelial cells; the threshold of expression required to support apical positioning and cell survival is indicated.
considerations for designing a therapeutic strategy based on inhibition of aPKC.
In summary, we describe two distinct functions for aPKC during mammalian epithelial morphogenesis. A major goal of our ongoing work is to identify those substrates through which aPKC controls spindle orientation/apical surface positioning and cell survival.