The Par-PrkC Polarity Complex Is Required for Cilia Growth in Zebrafish Photoreceptors

Specification and development of the apical membrane in epithelial cells requires the function of polarity proteins, including Pard3 and an atypical protein kinase C (PrkC). Many epithelial cells possess microtubule-based organelles, known as cilia, that project from their apical surface and the membrane surrounding the cilium is contiguous with the apical cell membrane. Although cilia formation in cultured cells required Pard3, the in vivo requirement for Pard3 in cilia development remains unknown. The vertebrate photoreceptor outer segment represents a highly specialized cilia structure in which to identify factors necessary for apical and ciliary membrane formation. Pard3 and PrkC localized to distinct domains within vertebrate photoreceptors. Using partial morpholino knockdown, photo-morpholinos, and pharmacological approaches, the function of Pard3 and PrkC were found to be required for the formation of both the apical and ciliary membrane of vertebrate photoreceptors. Inhibition of Pard3 or PrkC activity significantly reduced the size of photoreceptor outer segments and resulted in mislocalization of rhodopsin. Suppression of Pard3 or PrkC also led to a reduction in cilia size and cilia number in Kupffer’s Vesicle, which resulted in left-right asymmetry defects. Thus, the Par-PrkC complex functions in cilia formation in vivo and this likely reflects a general role in specifying non-ciliary and ciliary compartments of the apical domain.


Introduction
Cilia are hair-like structures that protrude from the apical surface of almost all vertebrate cells, including polarized epithelial cells. These cilia serve diverse functions, including sensory reception, motility-driven fluid flow, and signaling [1][2][3]. Within the cilium, the microtubule-based axoneme projects from a basal body anchored at the apical surface. This structure is surrounded by a ciliary membrane that is contiguous with the plasma membrane. The formation and continued maintenance of cilia requires intraflagellar transport (IFT), which refers to the bidirectional movement of IFT particles along the axoneme. Anterograde movement up the cilium is controlled by two kinesin motors, the heterotrimeric kinesin-II and the homodimeric Kif17 [4][5][6][7], while retrograde movement is controlled by the cytoplasmic dynein-2 complex [8][9][10][11].
Ciliogenesis initiates with the positioning of the centrosome/ basal body at the apical surface of epithelial cells. While the precise mechanisms controlling this action have not been fully elucidated, components of the cytoskeleton as well as determinants of cell polarity have been implicated. In primary cultures of multiciliated epithelial cells, basal body docking coincided with the assembly of an actin web at the apical surface [12,13]. Pharmacological inhibition of actin assembly prevented the apical accumulation of basal bodies and blocked ciliogenesis [12]. Although disruption of microtubules did not affect basal body migration in cultured quail oviducts [14], a dynamic microtubule reorganization was required for apical positioning of centrosomes and ciliogenesis within the zebrafish neural tube [15]. This microtubule reorganization and centrosome migration was also dependent upon the polarity protein Pard3 [15].
Pard3 is one of three key polarity determinants, along with Par6, and protein kinase C iota (PrkCi; formerly known as aPKCl), that form a heterotrimeric complex responsible for regulating apico-basal polarity in animal cells [16]. In zebrafish, three pard3 paralogs exist: pard3a, pard3ba, and pard3bb. We have focused on pard3a, which has been the subject of previous studies [15,17], and will be referred to herein as pard3.
Pard3, Par6, and PrkCi function in the maturation of tight junctions and adherens junctions but also determine the apical membrane domain (for reviews see [18,19]). Par6 and PrkCi physically interact through their N-terminal domains, while PrkCi physically interacts with Pard3 in a dynamic fashion. Following cell-cell contact, Pard3 recruits the Par6-PrkCi complex to contact sites, upon which tight junctions eventually form [20][21][22]. Following tight junction formation, growing evidence suggests that the PrkCi-Par6 complex dissociates from Pard3 and moves to the apical surface, where it localizes with other apical factors like Cdc42 and the Crumbs (Crb) complex [21,[23][24][25][26]. Elimination of any of these proteins or mutations that disrupt their interactions reduces the size of the apical membrane. Thus, while Pard3, Par6, and PrkCi all regulate apical domain size, only the Par6-PrkCi complex migrates to the apical surface to interact with the Crb complex [18].
The relationship between apical membrane size and ciliogenesis has been extensively studied in vertebrate photoreceptors. Vertebrate photoreceptors exhibit strong differences, both molecularly and morphologically, along the apical-basal axis. The presynaptic machinery localizes at the basal end of the cell while the apical domain consists of the distal part of the inner segment and the large outer segment. The photoreceptor outer segment is considered a modified ''sensory cilium'' [27] and extends from the apical surface of the inner segment. Numerous studies have implicated the Crb complex as a regulator of apical identity and outer segment size in both Drosophila and vertebrates [26,[28][29][30][31]. Loss of the zebrafish ortholog crb2b results in a reduction of apical domain size and loss of outer segments [28], while mutation of Crumbs homolog 1 (CRB1) causes early vision loss and photoreceptor degeneration [32,33]. Overexpression of the intracellular domain of Crb2a in zebrafish photoreceptors leads to an expansion of the outer segments, supporting a role for the Crb complex in positively regulating the size of the apical domain, including cilia [29]. The role other apical determinants play in ciliogenesis remains less clear. Zebrafish lacking either PrkCi or the MAGUK protein Nagie Oko do not exhibit obvious cilia defects in the retina, pronephros, or otic vesicle [28]. In contrast, Pard3, Par6 and PrkCi localize to cilia in cultured Madin Darby canine kidney (MDCK) cells and suppression of Pard3 or PrkCi by siRNA blocked ciliogenesis in cell culture [34,35]. Cilia growth in the neural tube was reported to be delayed, but not blocked, following suppression of zebrafish pard3 by morpholinos [15] but a role for pard3 in photoreceptor ciliogenesis was not investigated in a previous report [17]. Thus, it remains poorly understood whether the role of Crb in ciliogenesis is independent of other factors in vivo, or if other apical proteins, such as Pard3 and PrkCi, also play a role.
To better understand the role of apical identity proteins in ciliogenesis, we examined zebrafish photoreceptors following suppression of Pard3 or PrkCi. Partial suppression of these proteins by multiple experimental approaches dramatically reduced the size of the apical domain in photoreceptors and reduced cilia growth, resulting in smaller outer segments and mislocalization of rhodopsin. Our analysis suggests a general role for apical identity proteins in cilia formation in vivo and that cilia growth depends, in part, on the size of the apical membrane.

Nomenclature
In this paper, Pard3 (also known as Par3, Baz, Pard3a, or ASIP) will be referred to as Pard3. PrkCi (also known as aPKC lambda/ iota or aPKCl) will be referred to as PrkCi. PrkCz (also known as aPKC zeta or aPKCf) will be referred to as PrkCz. The term PrkC will refer to the combined activity of PrkCi and PrkCz in photoreceptors.

Ethics Statement
Wild-type zebrafish of a mixed AB-Ekkwill strain were housed, bred and staged according to standard procedures [36]. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the

Morpholino microinjection and design
One to two cell stage zebrafish embryos were injected as described [37]. All morpholinos (MOs) and photo-morpholinos (photo-MOs) were synthesized by Gene Tools, LLC. (Philomath, OR).
Sense photo-morpholinos (S-photo-MOs [40]) for pard3 (59 ACACCAGAGCApGGGAGCC), prkci (59 ATAATGCC-CApGCTGCGGGA 39) and prkcz (59 TATAATGCCTGpCAG-TAACGGAT 39) incorporate a photo-sensitive subunit (p) near the middle of the oligonucleotide, which can be cleaved by UVlight. S-photo-MOs can hybridize in a complementary fashion to the standard unmodified MOs, thereby permitting normal gene expression. Upon illumination by UV light, the photo-MO is cleaved and the standard MO is then free to bind the target RNA and block translation. Photo-MOs were mixed with their complementary MOs in 1:1 or 0.9:1.0 molar ratios in Milli-Q water at concentrations used for the standard MOs alone. The solutions were then heated to 100uC for 10 minutes and then slowly cooled to room temperature to permit hybridization. These solutions were stored as stock solutions at room temperature in the dark. At the appropriate time points, embryos were placed in petri dish and illuminated for 10-15 minutes with a light from a broad spectrum LED light source that passed through a Zeiss AxioZoom. V16 macroscope equipped with a FL Filter Set 49 for DAPI (Thornwood, NY, USA) at a magnification of 15x.

Mosaic analysis
Mosaic retinas were produced by blastomere transplantation [41]. Clutches of embryos from wild-type matings were dechorionated and injected at the 1-to 4-cell stage with morpholinos and a lineage-tracing label (1:9 mix of lysine fixable rhodaminedextran (Molecular Probes) at a total concentration of 5% w/v). At the 1000-cell stage, 10-40 donor cells from morpholino-injected embryos or wild-type embryos were transplanted to the animal pole of the dechorionated wild-type hosts, the region fated for eye and forebrain [42]. Host embryos were fixed in 4% paraformaldehyde at 4.5 dpf. Donor cells in host embryos were assessed by microscopy as described below. At least four independent blastomere transplantation experiments were conducted, with n$8 surviving embryos from each genotype combination.

Histology and electron microscopy
Larvae were fixed in a primary fixation solution of 2.5% glutaraldehyde, 1.0% paraformaldehyde prepared in a 0.06 M phosphate rinse buffer for 1.5 hours at 4uC. Larvae were then postfixed in 1% osmium tetroxide (prepared in 0.06 M phosphate rinse buffer for 1.5 hours and then dehydrated through a series of ethanol-water solutions. Specimens were then infiltrated in araldite resin. Fixation and dehydration was facilitated by a Pelco Biowave cold microwave. Transverse serial sections (1.0 mm) were cut and stained with an aqueous solution of 1% methylene blue, 1% azure blue, 1% borax. For electron microscopy transverse sections (0.1 mm) were cut and stained with uranyl acetate and lead citrate and viewed on a JOEL 1200EX transmission electron microscope.

SDS-PAGE and Western blotting
Adult retinas were homogenized in Laemmli buffer containing protease inhibitor cocktail and centrifuged for two minutes at 10006g. Two dilutions of the soluble fraction were separated by electrophoresis on 5-15% gradient SDS-PAGE gels (Biorad, Hercules, CA), and transferred to PVDF membranes. Rabbit anti-Pard3 (1:200) was used for Western blotting using standard protocols.

Immunohistochemistry and immunocytochemisry
A rabbit polyclonal anti-Pard3 antibody was generated against the synthetic peptide SPYTQKQNGRNGHPSTSDRY, which corresponds to amino (amino acids 1093-1112 in the C-Terminus of zebrafish Pard3, and was affinity purified by Open Biosystems. This antibody was used at 1:200 dilution. Immunolabeling procedures followed standard protocols, with the following specifications. Whole mount immunostaining was performed as previously described [43] with an anti-acetylated tubulin antibody at 1:1000. For immunohistochemistry larvae were fixed in 4% paraformaldehyde in phosphate buffered saline with 0.1% Tween-20 (PBST) for 2-18 hours at 4uC. Specimens were then washed twice in PBST and infiltrated with 30% sucrose (in PBST). Larvae were embedded in Tissue Freezing Medium (Triangle Biomedical Sciences) and frozen. Cryosections (10 mm) were mounted on gelatin-coated glass slides and dried for at least two hours at room temperature. Sections were washed in PBST and treated with block solution (PBST, 5% normal goat serum, 0.5% BSA (w/v), 1% DMSO, 1% Tween-20) prior to immunolabeling using standard procedures. Specimens were imaged with a Zeiss AxioImager microscope fitted with an ApoTome and AxioCam using 63x or 100x Plan-Neo-Fluar objectives and utilizing AxioVision software (Zeiss, Thornwood, NY). Each experiment was performed at least three independent times and at least 10 larvae from each condition were sectioned and stained per experiment.

In situ hybridization
Antisense riboprobes corresponding to southpaw were synthesized and in situ hybridizations performed as described [46]. Images were obtained on a Stereo LUMAR stereomicroscope (Zeiss).

Measurements and statistics
Measurements were performed with AxioVision software (Zeiss) and analyzed using GraphPad Prism5. The inner segment lengths were measured from the most proximal edge of the outer (D1-D3) Ift88 immunoreactivity stains cilia. (E1-E3) Acetylated tubulin immunoreactivity was seen in cilia and neural processes throughout the inner nuclear layer and plexiform layers. (F1-F3) Pard3 staining (red) was seen at the OLM (arrowhead), but did not colocalize with acetylated tubulin signal in the cilia (arrows). All sections were counterstained with DAPI (blue). Scale bar = 10 mm. doi:10.1371/journal.pone.0104661.g001 plexiform layer to the most distal edge of Zpr-1 staining. To measure the apical domain, the distance from the outer limited membrane to the distal edge of Zpr-1 staining was quantified. All measurements were taken from the central part of the retina in transverse cryosections sections at or near the optic nerve (n = 8-15 larvae) and analyzed using a 1-way ANOVA and Tukey posttest.

PKC inhibitor treatments
The myristolated PKCf pseudosubstrate inhibitor (EMD4Biosciences; 539624) and the non-myristolated form (EMD4Biosciences; 539610) were dissolved in water to a concentration of 1 mM and then diluted to a concentration of 3 mM in embryo media containing 0.5% DMSO [36]. Embryos were dechorionated at 50 hours post fertilization (hpf) and added to wells of a 24well culture plate containing 500 mL of compound at a density of no more than 8 embryos/well. Treatment at earlier time points resulted in lethality within 2-6 hours. The plates were stored at 28.5uC. At 96 hpf, larvae were collected and processed for immunohistochemistry.

Localization of polarity proteins in zebrafish photoreceptors
To investigate the link between polarity proteins and ciliogenesis in vivo, we first examined the spatial distribution of Pard3 and PrkC in cryosections of larval zebrafish retinas. Previous studies found that Pard3 localized to the apical surface of the retina and brain neuroepithelia at 33 hpf [17]. We generated a rabbit polyclonal antibody against a peptide of Pard3 (see Materials and Methods) and confirmed the specificity of this antibody by Western blotting (Fig. S1). Although the antibody should recognize both splice isoforms, we only detected the 180 kDa protein on Western blots (Fig. S1). Photoreceptor outer segments, which are considered specialized sensory cilia, are well-developed by 5 dpf and we examined the localization of Pard3 and other junction and the monoclonal antibody zpr-1 (middle column). The right panels show the merged images. Pard3 immunoreactivity was seen in the outer plexiform layer (OPL) and at cell junctions (A, arrowhead and arrows, respectively) in wild-type and control MO retinas but was significantly reduced in morphants retinas. All sections were also counterstained with DAPI (blue). ONL = outer nuclear layer; OPL = outer plexiform layer; INL = inner nuclear layer. Scale bar = 10 mm. doi:10.1371/journal.pone.0104661.g002 markers relative to cilia proteins ( Fig. 1). Phalloidin was used to stain the actin belt at the adherens junctions of the outer limiting membrane (OLM). As previously observed [28], actin localized vertically along the sides of the photoreceptors, immediately apical to the OLM, in a pattern resembling longitudinal fibers within the ellipsoid [47]. As expected, immunostaining for the cell junction proteins protein kinase C iota/zeta (PrkCi), Pard3, and ZO-1 revealed colocalization with actin at the OLM, although each protein had a distinctive spatial distribution ( Fig. 1A-C). An antibody that recognizes an epitope shared by PrkCi and the close paralog, protein kinase C zeta (PrkCz), colocalized with actin at the OLM and along the sides of the photoreceptor in the domain apical to the OLM. Crumbs proteins and the FERM domain protein Mosaic eyes (Moe), which function as apical determinants, also localized on the apical side of the OLM [28,48]. Consistent with earlier studies [17], the strongest Pard3 staining colocalized with actin at the outer plexiform layer (Fig. 1B). Pard3 also localized in a punctate pattern near the OLM, although the signal intensity was less pronounced than at the outer plexiform layer. ZO-1 distribution was restricted to the OLM (Fig. 1C). Immunoreactivity to Ift88, a component of the IFT particle, was seen in connecting cilia as previously described, but did not colocalize with actin ( Fig. 1D; [45]). Ciliary axonemes were visualized by immunostaining for acetylated tubulin, which also did not colocalize with actin ( Fig. 1D, E). As Pard3 localized to primary cilia in MDCK cells [34,35], we then asked if Pard3 localized to the photoreceptor connecting cilia or outer segments. While Pard3 was seen with other apical determinant proteins at the OLM, we did not observe Pard3 localizing with acetylated tubulin in the connecting cilia of photoreceptors (Fig. 1F). These results show that while Pard3, PrkC, and Crumbs proteins appeared at cell junctions, Pard3 was found at the outer plexiform layer, PrkC and Crumbs were seen in the domain immediately apical to the OLM, and none could be detected in the cilia in vivo.

Genetic mosaic analysis of pard3-deficient cells
Pard3 physically interacts with PrkCi and this interaction is necessary for establishing apical identity cultured cells [23]. Testing the role of Pard3 or PrkCi in photoreceptors, however, poses a number of experimental challenges. Complete loss-offunction mutations in genes required for apico-basal polarity, including heart and soul (has/PrkCi), mosaic eyes (moe), the Crumbs ortholog oko meduzy (ome), and nagie oko (nok/Stardust), disrupt the laminar organization of the retina and impair photoreceptor cell differentiation [28,38,39,[49][50][51]. Additionally, PrkCi and PrkCz function redundantly and suppression of both proteins may be necessary to fully test the role of PrkC activity in photoreceptors. These phenotypes are, however, cell-non-autonomous and can be rescued in genetically mosaic animals [38,49,50]. Injection of 4-5 ng of a previously characterized pard3 morpholino (MO1) fully suppressed Pard3 function but also caused cyclopia and disrupted retinal lamination and patterning [15,17]. In an attempt to examine the role of Pard3 on photoreceptor structure, morpholino oligonucleotides (MOs) against pard3 were injected into 1-cell embryos and we conducted genetic mosaic experiments using rhodamine-dextran as a lineage dye. At 4 dpf, wild-type donor cells contributed to all retinal cell types in wild-type hosts (Fig. S2A). The pard3 morphant donor cells also migrated to the appropriate laminar position in the wildtype hosts, indicating that the retinal cell positioning defect is cellnon-autonomous. We noticed, however, that the pard3 morphant cells formed few, if any, photoreceptors and only individual photoreceptors were ever observed in mosaic retinas (Fig. S2B). These rare pard3-deficient photoreceptors formed normal ad- herens junctions with neighboring wild-type cells, as determined by actin and ZO-1 staining (data not shown). In agreement with previous results [38], we also noticed that few photoreceptors formed when cells from prkci morphants or prkci/prkcz double morphants were transplanted into wild-type hosts (Figs. S2C and S2D). These results suggest that Pard3 and PrkC may function to limit cell fate or differentiation. Indeed, the loss of Pard3 activity in progenitor cells within the subventricular zone of the mouse cerebral cortex resulted in early neurogenic differentiation and fewer progenitor cell divisions [52]. This may explain why pard3deficient cells largely failed to differentiate. Given these results, we were reluctant to draw conclusions regarding cilia formation from single pard3-deficient photoreceptors due to the earlier effect of pard3 loss on retinal progenitor cell proliferation/differentiation.

Partial suppression of pard3 by morpholinos
To avoid the effects on retinal patterning and photoreceptor differentiation, but still partially suppress Pard3, doses of MO1 ranging from 1.2-1.5 ng were injected into 1-cell embryos. To verify that Pard3 expression was reduced, cryosections of wild-type and pard3 MO1-injected larvae were immunolabled with Pard3 antisera and colabeled with zpr-1, a marker for red-green double cones (Fig. 2). Pard3 immunostaining was seen at the OPL ( Fig. 2A, arrowhead) and in two rows of cell junctions ( Fig. 2A, arrows). In morphant retinas, Pard3 immunoreactivity was significantly reduced at the OPL and almost entirely absent at the OLM (Fig. 2D). Importantly, the Zpr-1 staining showed that photoreceptor differentiation was not impeded following Pard3 suppression (Fig. 2F). Injection of a control morpholino, which contained a 6-basepair mismatch, did not affect Pard3 expression (Fig. 2G-I). From these results, we concluded that smaller doses of pard3 morpholino could reduce Pard3 proteins levels without impairing photoreceptor differentiation.
The formation of cellular junctions requires Pard3 and suppression of Pard3 by RNAi disrupted the localization of other polarity proteins to tight junctions in cultured cells [20]. Studies in zebrafish, however, found normal ZO-1 localization to the apical midline in 12-13 somite pard3 morphants [15]. We asked if the loss of Pard3 at the OLM affected the cellular distribution of other polarity proteins within photoreceptors. Morpholino knockdown of pard3 did not alter the OLM localization of actin but the staining pattern appeared rough and less organized (compare Fig. 3A to 3B, D, F). The localization of PrkC to the apical side of the OLM was unaffected in pard3 morphants, but the apical surface (Fig. 3B''; arrows) was reduced and most of the signal was aggregated at the OLM (Fig. 3B). Immunohistochemistry with a polyclonal antibody that recognizes multiple Crumbs polypeptides gave a staining pattern similar to that of PrkC (Fig. 3C). In pard3 morphants, Crumbs staining was aggregated at the OLM and the apical membrane was reduced (Fig. 3D). ZO-1 staining was not lost in pard3 morphants, although the signal at the OLM was less discrete (compare Figs. 3E and F). Taken together, these results suggest that partial suppression of Pard3 was not sufficient to block the normal localization of polarity proteins or the formation of adherens junctions.
Pard3 promotes formation of the apical membrane domain [23], and the loss of PrkCi and Crumbs from the region immediately apical to the OLM may reflect a reduction in the apical domain size. The OLM divides the photoreceptor membrane into apical and basolateral domains, with the inner segment corresponding to the area between the OLM and the base of the outer segment. When wild-type retinas are stained with Zpr-1 at 5 dpf, the inner segments of cone photoreceptors are partitioned almost equally between the apical and basolateral domains [28]. We stained photoreceptors with Zpr-1 and quantified the size of the apical inner segment domain ( Fig. 4A; ''n'') relative to total length of Zpr-1 staining (Fig. 4A; ''m''). Consistent with previous reports [28], the apical domain accounts for approximately 50% of the length of Zpr-1 staining in 5 dpf larvae (Fig. 4A, quantified in Fig. 4G). In pard3 morphants, the apical domain was considerably smaller and was less than 30% of the total length of Zpr-1 staining (Fig. 4B). We found similar results when we labeled the OLM with ZO-1 and measured the distance from the outer plexiform layer to the most distal point of DAPI staining, which corresponded to the nuclei of the red/green double cones (Fig. 4D). Control morpholinos had no effect on apical domain size (Fig. 4C, F). Overall, these results indicate that low doses of pard3 morpholino reduced the apical domain in photoreceptors. Photoreceptor outer segment formation is reduced in pard3 morphants As cilia formation occurs on the apical surface of the inner segment, we next asked photoreceptor outer segment formation was affected. Histological sections of 4 dpf larvae confirmed that retinal lamination and organization was largely unaffected by suppression of Pard3 (Fig. 5), although some degree of disorganization was occasionally observed at higher morpholino concentrations (data not shown). Even at lower doses of morpholino, however, the pard3 morphants exhibited varying degrees of cyclopia, with some larvae having partially fused eyes and other larvae exhibiting only a slight reduction in distance between the eye fields ( Fig. 5A-C). Morphants with more pronounced cyclopia also showed reduced pigmentation of the RPE when compared to morphants without fused eyes (Fig. 5, panels E and F, respectively). Photoreceptors were present in the outer nuclear layer of pard3 morphants, but outer segments were not observed (Fig. 5D-F). Transmission electron microscopy confirmed that outer segments were significantly reduced, and often missing, from photoreceptors in pard3 morphants (Fig. 5G-H). The truncated outer segments that did form on pard3 morphant photoreceptors contained regularly stacked disc membranes, indicating that loss of Pard3 does not affect disc membrane organization. The positioning of the mitochondria and nucleus were normal, suggesting that apicobasal positioning of organelles was not disrupted.
The formation and growth of photoreceptor outer segments requires the process of Intraflagellar Transport, or IFT [45,53,54], and the Par-PrkCi complex has been linked to IFT [34]. Pard3 physically associates with Kif3a, the anterograde motor for IFT, and PrkCi binds to the complex through its association with Pard3 [34,55]. Furthermore, the interaction between the Par-PrkCi complex and the Crb complex is required for photoreceptor differentiation in Drosophila [26]. The absence of outer segments in pard3 morphants may reflect a failure in apical assembly of IFT particles. To investigate whether loss of Pard3 affected the localization of ciliary proteins, we examined transverse sections stained with antibodies against acetylated tubulin to label cilia and with Ift88, a component of the IFT particle. Although acetylated Figure 6. Connecting cilia are reduced in pard3 morphants. (A-C) Transverse cryosections were stained for Ift88 (left, red) and acetylated tubulin (middle, green) to label cilia in 5 dpf retinas. Merged images are shown in right panels. Ift88 localizes to the connecting cilia emanating from the apical surface of photoreceptors (arrows). Staining is also seen in connecting cilia of UV cones (arrowheads), which are tiered below the other cones. Sections were also counterstained with DAPI (blue). ONL = outer nuclear layer; OPL = outer plexiform layer; INL = inner nuclear layer. Scale bar = 10 mm. doi:10.1371/journal.pone.0104661.g006 tubulin staining was concentrated at the apical surface of pard3 morphants, very few cilia could be seen projecting toward the RPE (Fig. 6, compare panels A'' and B''). Ift88 immunoreactivity was also reduced in pard3 morphants, although the remaining signal localized at the apical surface ( Fig. 6B and B'').
IFT facilitates the trafficking of proteins, such as rhodopsin, to the photoreceptor outer segment. While opsin molecules function primarily to detect photons of light for the visual response, they are also the most abundant proteins in the outer segment and provide structural support to the outer segment [56]. Mutations in components of the IFT particle or in Kif3a cause opsin to accumulate in the photoreceptor inner segment and even partial loss of IFT results in severe outer segment defects [45,53,54,57]. As pard3 morphants show reduced levels of Ift88 and loss of outer segments, we anticipated that rhodopsin would be mislocalized. Indeed, opsin immunoreactivity could be seen in the inner segments of pard3 morphants (Fig. 7E, bracket) and the outer segment staining was reduced (Fig. 7E, compare to 7B). Thus, the loss of Pard3 reduces outer segment size and results in mislocalization of opsin.

Temporal regulation of pard3 and prkc by photomorpholinos
In an attempt to more effectively block Pard3 and PrkC activity at later time points, we utilized sense photo-active morpholinos (Sphoto-MOs) to keep morpholinos temporarily inactive (See Materials and Methods and [40]). Standard MOs and the complementary S-photo-MOs specific for pard3, prkci or prkcz were hybridized to form dimers in vitro prior to injection into 1cell embryos. Larvae injected with MO/S-photo-MO hybrids specific for either pard3 or prkci did not exhibit any overt morphological phenotypes or disruption to retinal structure in the absence of UV irradiation (Figs. S3 and S4). Larvae injected with the MO/S-photo-MO hybrids and subsequently irradiated with UV at 5 hpf showed both morphological and retinal lamination phenotypes that resembled the phenotypes seen in pard3 morphants and prkci mutants (Figs. S3 and S4; [17,38,51]). No adverse phenotypes were observed in larvae injected of S-photo-MOs alone. These results suggest that embryos retain endogenous Pard3 or PrkCi activity until the S-photo-MO is cleaved by UV light and the unmodified MO can target the cognate mRNA to suppress function.
To block the activity of Pard3 or PrkC at time points corresponding to the beginning of photoreceptor ciliogenesis, embryos were injected with MO/S-photo-MO hybrids and irradiated with UV light at 30 hpf. Although photoreceptor differentiation begins at 52 hpf, we found that irradiation at 30 hpf was required to give robust photoreceptor phenotypes at 4 dpf consistent with the loss of Pard3/PrkC activity. This may reflect the time needed to for the S-photo-MO to fully degrade, the MO to bind and inhibit the cognate mRNA, and for protein turnover. Irradiation at later times (e.g. 42 hpf or 50 hpf) resulted in little to no phenotype at 4 dpf (data not shown). Similar to the previous experiments, morphant retinas were stained with Zpr-1 and we quantified the size of the apical domain. At 4 dpf, the apical domain constitutes approximately 40% of the wild-type photoreceptors (Figs. 8A, E). When larvae were UV irradiated at 30 hpf, the apical domain was significantly reduced compared to wild-type larvae ( Fig. 8; n/m value: 38% vs. 16%, p,0.0001; n = $34). When larvae were irradiated at 30 hpf following co-injection of MO/S-photo-MO hybrids simultaneously targeting prkci and prkcz, the apical domain was essentially lost (Fig. 8D). The outer limiting membrane could not be detected by phalloidin, and zpr-1 staining was lost, suggesting that cells in the outer nuclear layer failed to differentiate into photoreceptors. Larvae injected with MO/S-photo-MO hybrids also showed significant reduction in the number of 1D1+ rod photoreceptors and mislocalization of rhodopsin (Fig. 9). Whereas rhodopsin localized predominantly to the outer segment of wild-type photoreceptors, suppression of Pard3 resulted in small outer segments and inner segment localization of rhodopsin. Suppressing PrkCi alone or all PrkC activity had more pronounced phenotypes, with few rhodopsinpositive cells being observed and rhodopsin localizing throughout the inner segments and cell bodies (Fig. 9). We quantified the number of 1D1+ rod photoreceptors in 10-micron cryosections. Compared to wild-type larvae, the number of rods was reduced 45% in pard3 photo-morphants (36 vs. 20 per section; p#0.001; n $10). Rods were reduced by 77% in prkci photo morphants and 92% following in larvae injected with both prcki and prkcz photo-MOs (Fig. 9I). As photoreceptors were missing almost entirely from the prcki/prkcz photo-morphants, cilia lengths were not assessed.

Pharmacological inhibition of PrkC reduced apical membrane size
Finally, we used a pharmacological approach to test the role of PrkC activity in photoreceptors. Zebrafish embryos were treated with a cell-permeable PKC-specific inhibitor peptide (3 mM, myristoylated PKCf pseudosubstrate inhibitor), which inhibits the activity of both PrkCi and PrkCz, and can block ciliogenesis in MDCK cells [34,58]. As a control, we used a non-myristoylated form of the peptide. Zebrafish were treated with the PKC inhibitor at 50 hpf, a time at which photoreceptors have become postmitotic but have not yet initiated outer segment formation [59], and were examined at 4 dpf, after two days of exposure to the peptide. Treatment at 24 hpf or 36 hpf caused lethality within 2-6 hours and embryos effectively disintegrated, possibly due to disrupting the integrity of the ectodermal epithelia. Similar to the previous experiments, retinas were stained with Zpr-1 and we quantified the size of the apical domain (Fig. 10). Compared to wild-type, inhibition of PrkC significantly reduced the size of the apical domain (45% vs. 28%; p#0.001; n$12 larvae). Treatment with the non-myristolated form of the inhibitor did not have any effect on apical domain size (Fig. 10C, D). Rhodopsin was mislocalized in animals treated with the PKC inhibitor but not in control animals (Figs. 11A-C). Cilia length was also reduced following inhibition of PrkC (Fig. 11D-G) but the number of cilia was not statistically altered (data not shown).

Loss of pard3 or prkc affected cilia in Kupffer's Vesicle
Vertebrates exhibit an asymmetry across the left-right (LR) axis that is linked to the function of motile cilia in Kupffer's Vesicle (KV) in zebrafish, or the node of mouse (reviewed in [60]. Disruption of KV specification or of KV cilia function randomizes LR asymmetry in zebrafish [61][62][63]. The nodal-related gene southpaw (spaw) is one of the earliest markers of LR asymmetry and is normally expressed in the left lateral plate mesoderm of 18-20 somite stage embryos [64]. Approximately 87% of wild-type embryos express spaw on the left side (L) with small percentages (, 9%) showing right-sided (R) expression, bilateral (B) expression, or an absence (A) of expression (Fig. 12A). Even with modest suppression of pard3 (1.2 ng morpholino), 48% show left-sided expression of spaw, while the others show right-sided (9%), bilateral (15%), or an absence (27%) of spaw expression. Interestingly, injection of either prkci MO or prkcz MO alone did not cause LR asymmetry defects, whereas injection of both prkci and prkcz MOs together did alter LR asymmetry. This indicates PrkCi and PrkCz function redundantly. As Pard3 functions with PrkCi and the function of Pard3 was only partially suppressed, we asked if morpholino suppression of both pard3 and prkci would show an additive, or perhaps synergistic, effect on LR asymmetry. However, suppression of pard3 and prkci did not alter the percentage of embryos showing left-sided expression of spaw, compared to suppression of pard3 alone (Fig. 12B).
We performed immunocytochemistry on 8-10 somite stage embryos and labeled cilia with acetylated tubulin antibodies (Fig. 12C). While the size of the KV was similar between wild-type and pard3 morphant embryos, the number of KV cilia was reduced (Fig. 12E) and cilia length was noticeably shorter (Fig. 12F). Taken together, the results suggest that Pard3 controls cilia formation in tissues other than the retina.

Discussion
The purpose of this study was to investigate the role of the Par-PrkCi complex in cilia formation in vivo, with a specific emphasis on photoreceptor outer segment formation. The roles for the Par-PrkCi complex and Crb proteins as apical determinants has been well established [19]. From flies to humans the Crb proteins are required for photoreceptor morphogenesis, particularly in formation of the light-sensitive rhabdomeres or outer segments [28,31,48,65]. This requirement reflects a role for Crb proteins in ciliogenesis [34]. While the Par-PrkCi complex was implicated in ciliogenesis in tissue culture [34,35], it was unclear if such function occurred in vivo, particularly in photoreceptors. Bazooka (the Drosophila homolog of Pard3), PrkCi, and Par6 are required for targeting Crb to the rhabdomere stalk [26]. However, Crb also regulates rhodopsin trafficking through interactions with myosin V [66], thereby raising the possibility Crb role in ciliogenesis may be independent of Par-PrkCi. Consistent with this hypothesis, the loss of prkci or pard3 in zebrafish clearly disrupted apico-basal polarity, but cilia defects were not previously reported [17,38,39,51]. Nevertheless, a recent study reported that inhibition of the PrkCi ortholog in sea urchin embryos dramatically shortened cilia, suggesting a conserved role for this complex in ciliogenesis [67]. We report that inhibition of Pard3 or PrkCi/z activity impairs cilia formation.

Role of pard3 in apico-basal polarity
The loss of pard3 impacts apico-basal polarity in two distinct events within the retina. Wei and colleagues [17] previously showed that a strong morpholino suppression of pard3 disrupted retinal lamination and reduced photoreceptor numbers. The loss of prkci or crb produces similar phenotypes, suggesting that these genes function in a similar pathway [28,38]. Based on observations in Drosophila and tissue culture, it is believed that Pard3 recruits PrkCi to cellular junctions, while Crb maintains junction integrity [21,26]. Thus, the retinal lamination defect in strong pard3 morphants likely reflects an initial requirement for pard3 to establish apical cellular junctions within the developing neuroepithelium, similar to prkci [38]. The second role for pard3 occurs in photoreceptors. Once photoreceptor cells acquire the appropriate laminar position, Pard3 and PrkC regulate the size of the apical domain. Suppression of pard3 by sub-saturating doses of morpholinos or by the use of photo-morpholinos did not affect retinal lamination (Figs. 5, 8) but apical membrane size was reduced (Fig. 4, 8). Similarly, inhibition of PrkC activity by photomorpholinos or by pharmacological inhibition resulted in a reduction of the apical membrane (Fig. 8, 10).
The precise mechanism by which Pard3 and PrkCi regulate apical size is unclear, but it likely involves the function of Crb proteins. Two crb2 paralogs, crb2a and crb2b, are expressed in zebrafish photoreceptors. The loss of crb2b function in zebrafish photoreceptors reduced apical membrane size similar to our results with Pard3 and PrkC [28]. In contrast, overexpression of various domains of crb2a significantly increased the size of the apical membrane and outer segment [29]. Whether these protein complexes function together or in a sequential process remains poorly understood. Based on protein localization, it is unlikely that Pard3 participates in a complex with PrkCi or Crb to regulate apical membrane size. While Pard3 localizes to the OLM and the outer plexiform layer, PrkC and Crb localize immediately apical to the OLM and extend apically along the inner segment in a region that lacks Pard3. This is consistent with studies showing PrkC and Crb excluding Pard3 from the apical surface [23,68]. It should be noted, however, that the formation of a Par-PrkC complex was required for apical membrane growth [23]. This suggests a sequential model where a temporary Par3-PrkC complex must assemble prior to PrkC migration to the apical surface [23]. Although modest suppression of Pard3 did not impact localization of PrkC or Crb to the region immediately apical to the OLM, we cannot exclude the possibility that some Pard3 activity remained.

Role of pard3 in cilia formation
As the ciliary axoneme grows during ciliogenesis, the ciliary membrane extends from the apical surface. The ciliary membrane is distinct from the apical cell membrane [69] and some cells form elaborate ciliary structures, such as the vertebrate photoreceptor outer segment. It is clear, however, that formation of the ciliary membrane requires the same apical determinants used to form the apical cell membrane. Manipulating crb expression can increase or decrease both apical and ciliary membranes [28,29], while inhibition of Pard3 or PrkC similarly reduces apical size and leads to loss of cilia (this study). Previous work in zebrafish found that positioning of centrosomes and basal bodies within the early neural tube requires Pard3 [15]. We found normal basal body localization in photoreceptors following suppression of pard3 by photo-MOs (data not shown), but this may reflect the early polarization of the retinal neuroepithelium.
The mechanism by which these apical determinants control ciliogenesis remains unclear. One possibility is that these proteins facilitate some aspect of intraflagellar transport (IFT). Work in tissue culture found that Pard3, PrkCi, and Crb3 localized in cilia and Pard3 and PrkCi immunoprecipitated with acetylated tubulin Figure 11. Pharmacological inhibition of PrkC activity reduces cilia length and causes opsin mislocalization. (A-C) Transverse cryosections of 4 dpf retinas were stained with phalloidin to label actin (red) and 1D1 to label rhodopsin (green). (D-F) Immunohistochemistry on cryosections to label Ift88 in cilia (red) and acetylated tubulin to label ciliary microtubules (green). Arrows identify cilia. Sections were also counterstained with DAPI (blue). (G) Quantification of cilia length. Cilia length in 4 dpf wild-type photoreceptors (43.2 microns) and the nonmyristoylated control inhibitor (48.9 mm) were not statistically different from each other, but both were statistically different from larvae treated with the PrkC inhibitor (27.1 microns). Error bars show standard deviation. (***p#0.0001) Scale bar = 10 mm. doi:10.1371/journal.pone.0104661.g011 [34]. Pard3 can also directly bind to Kif3a, the anterograde motor for IFT [55]. In contrast, numerous studies in Drosophila and mammalian cells have found Pard3 localizing to a subapical domain in a pattern distinct from PrkCi and Par6 [19,23,26]. We failed to observe colocalization of Pard3 with acetylated tubulin in the connecting cilia of photoreceptors and Pard3 did not localize along the apical surface with PrkC or Crb. While we cannot exclude the possibility that ciliary localization of Pard3 escaped our detection, our results are consistent with previous studies where a different polyclonal antibody did not detect Pard3 in photoreceptor cilia [17], and those showing a Pard3-GFP fusion protein not localizing to cilia when expressed within the zebrafish neural tube [15]. Given the distinct localization patterns between Pard3 and other apical determinants, our results favor a mechanism whereby Pard3 regulates the size of the entire apical membrane. While the ciliary and non-ciliary compartments have distinct protein profiles [69], cilia formation typically requires prior specification of the apical membrane.
The apical determinant factors pard3, prkci, nok, moe, and ome all give similar retinal phenotypes when fully mutated or suppressed, yet cilia defects in other tissues (e.g. kidney, otic vesicle, Kupffer's Vesicle) are not always observed [17,28,39,49,50]. We observed defects in left-right asymmetry and KV cilia in pard3 morphants and hydrocephalus was previously noted [15], although kidney cysts were not observed. In many cases, however, genetic redundancy by paralogs can compensate for the loss of individual genes. With crb genes, for example, ome functions in neuroepithelia, crb2b functions in the retina and kidney, while crb3a functions in the otic vesicle [28]. Similarly, prkci and prkcz exhibit functional redundancy within the retina [38]. A pard3b paralog [70], may explain why pard3 morphants do not exhibit other ciliary defects.
We used three distinct experimental approaches to inhibit Pard3 and PrkC activity and all gave similar results. Sub-optimal concentrations of morpholinos permitted retinal lamination in pard3 morphants and partially abrogated cilia and outer segment formation. A well-established peptide inhibitor of PrkC was effective in reducing apical domain size, but the inhibitor was toxic at earlier time points, which likely limits its utility for studies in embryos. The use of photo-morpholinos gave the strongest effects, particularly when targeting the prkc genes. While these reagents did not completely inhibit the MO activity over several days, the ability to cleave the photo-MO provided temporal control over MO activity and permitted normal retina lamination with before suppression of either pard3 or prkc. It should be noted that photo-morpholino suppression of prkci or prkci/z resulted in much more severe retinal effects than suppression of pard3. Although a biological difference cannot be definitively ruled out, these observations may reflect a difference in the efficacy of photo-MOs at different gene targets. While Tallafuss and colleagues reported strong suppression with 5-10 minutes of UV irradiation at 5 hpf [40], at least 20 minutes of irradiation was required to suppression gene function in our studies at either 5 hpf or 30 hpf. Longer irradiation times (e.g. 30 minutes) did not enhance the phenotype, although it is possible that the pard3 photo-MO was still not completely cleaved. As with any antisense reagent, the efficacy of photo-MOs can vary from target to target and this may explain the differences in phenotype.
In summary, we show that cilia formation in vertebrate photoreceptors requires the Par-PrkC complex in a manner that likely reflects a role in apical membrane regulation. Mutations in CRB1 cause retinitis pigmentosa [65] and crb genes were known to play key roles in apical differentiation. We now show that cilia formation in vivo requires the involvement of the Par-PrkC complex as well.