|Home | About | Journals | Submit | Contact Us | Français|
The Par3/Par6/aPKC protein complex plays a key role in the establishment and maintenance of apicobasal polarity, a cellular characteristic essential for tissue and organ morphogenesis, differentiation and homeostasis. During a forward genetic screen for liver and pancreas mutants, we identified a pard6γb mutant, representing the first known pard6 mutant in a vertebrate organism. pard6γb mutants exhibit defects in epithelial tissue development as well as multiple lumens in the neural tube. Analyses of the cells lining the neural tube cavity, or neurocoel, in wildtype and pard6γb mutant embryos show that lack of Pard6γb function leads to defects in mitotic spindle orientation during neurulation. We also found that the PB1 (aPKC-binding) and CRIB (Cdc-42-binding) domains and the KPLG amino acid sequence within the PDZ domain (Pals1- and Crumbs binding) are not required for Pard6γb localization but are essential for its function in neurocoel morphogenesis. Apical membranes are reduced, but not completely absent, in mutants lacking the zygotic, or both the maternal and zygotic, function of pard6γb, leading us to examine the localization and function of the three additional zebrafish Pard6 proteins. We found that Pard6α, but not Pard6β or Pard6γa, could partially rescue the pard6γbs441 mutant phenotypes. Altogether, these data indicate a previously unappreciated functional diversity and complexity within the vertebrate pard6 gene family.
Cell polarization distributes cellular constituents into discrete domains and results in asymmetric distribution of cellular activity. Three main protein complexes (Par3/Par6/aPKC, Crumbs/Discs lost/Stardust and Lethal giant larvae/Discs large/Scribble) regulate apicobasal polarity by localizing or restricting tight junction proteins toward the apical end of lateral membranes and by organizing the cytoskeleton (reviewed by Gibson and Perrimon, 2003). Epithelial cells exhibit apicobasal polarity, which allows the establishment of physiological barriers (reviewed by Rodriguez-Boulan and Nelson, 1989). Formation of polarized epithelial sheets is also essential for cellular and tissue morphogenesis, movement, and differentiation during metazoan development (reviewed by Gibson and Perrimon, 2003).
Forward genetic screens in zebrafish (Danio rerio) have identified embryonic lethal mutations in several genes that regulate epithelial development including heart and soul (has/prkci, previously known as apkcλ) (Horne-Badovinac et al., 2001; Peterson et al., 2001), nagie oko (nok/pals1/mpp5) (Wei and Malicki, 2002), oko meduzy (ome/crb2) (Omori and Malicki, 2006), and mosaic eyes (moe/epb41l5) (Jensen and Westerfield, 2004). Each of these mutations causes defects in heart tube assembly, retinal pigmented epithelium development, inflation of the brain ventricles, and body curvature. Further analyses have revealed distinct roles for each of these genes in different aspects of developing organs derived from epithelia (Horne-Badovinac et al., 2003; Kramer-Zucker et al., 2005; Omori and Malicki, 2006; Rohr et al., 2006; Bit-Avragim, 2008b; Christensen and Jensen, 2008). For example, zygotic expression of has/prkci and nok/mpp5 appears to regulate distinct aspects of heart development (Rohr et al., 2006) and ome/crb2 regulates the elongation of cilia (Omori and Malicki, 2006).
The has/prkci mutation affects one member of the Par3/Par6/aPKC complex. Studies in the C. elegans zygote first identified the components of this complex as essential for asymmetric cell division (Etemad-Moghadam et al., 1995; Guo and Kemphues, 1995; Watts et al., 1996). Since then, the influence of this complex on polarization and development has expanded to include Drosophila embryos and mammalian cells (Suzuki and Ohno, 2006). Analysis of zebrafish has/prkci mutants (Horne-Badovinac et al., 2001; Peterson et al., 2001), as well as pard3 morpholino injected embryos, aka morphants, (Geldmacher-Voss et al., 2003; Wei et al., 2004) and Pard3 deficient mice (Hirose et al., 2006) has increased our understanding of how these proteins regulate the complex events of vertebrate organogenesis. However, the role of Par6 during vertebrate development has remained elusive.
The ability of Par6 to polarize epithelial cells depends on at least four protein-protein interactions mediated through three domains. Par6 interacts with aPKC/Prkci through its Phox and Bem1p (PB1) domain (Suzuki et al., 2001; Suzuki et al., 2003), with Cdc42 through its partial Cdc42/Rac interactive binding (CRIB) domain (Joberty et al., 2000; Lin et al., 2000; Qiu et al., 2000; Garrard et al., 2003; Hutterer et al., 2004), and with both PAR-3/Pard3 (Joberty et al., 2000; Lin et al., 2000) and Pals1/Mpp5 (Hurd et al., 2003; Wang et al., 2004) through its PDZ (PSD-95/Dlg/ZO-1) domain. Only one Par6 protein exists in C. elegans and Drosophila, but a family of three proteins is present in mammals: Pard6A, Pard6B, and Pard6G (Joberty et al., 2000). Studies in MDCK cells have suggested that these different family members exhibit different sub-cellular localization and have distinct functions (Gao et al., 2004). However, there has been no in vivo analysis of the different vertebrate Pard6 family members during development.
Through neurulation, the neuroepithelium gives rise to the eyes, the brain and its ventricles, and the spinal cord. Many of these structures are disrupted in zebrafish embryos carrying mutations affecting apicobasal polarity (Horne-Badovinac et al., 2001; Wei and Malicki, 2002; Jensen and Westerfield, 2004; Lowery and Sive, 2005; Omori and Malicki, 2006). In zebrafish, the neuroepithelium begins as an epithelial-like structure at the neural plate stage, although it lacks polarized localization of apicobasal polarity markers (Papan and Campos-Ortega, 1999; Lowery and Sive, 2004; Ciruna et al., 2006). This neural plate converges to the midline and extends during somitogenesis to form the neural keel, a process that has the hallmarks of primary neurulation (Geldmacher-Voss et al., 2003; Lowery and Sive, 2004; Hong and Brewster, 2006). During this morphogenetic process, cells divide perpendicular to, and across, the midline (Geldmacher-Voss et al., 2003; Ciruna et al., 2006; Tawk et al., 2007). After the 12-somite stage (13 hours post fertilization (hpf)), Pard3 forms foci of polarity along the midline of the neural keel (Geldmacher-Voss et al., 2003; Tawk et al., 2007). These foci continue to assemble (Tawk et al., 2007) and by approximately 21hpf delineate continuous apical membranes and a slit lumen (Geldmacher-Voss et al., 2003). Following formation of these continuous apical membranes, fluid is pumped into the neurocoel and the brain ventricles inflate (Lowery and Sive, 2005). Coincident with the formation of the apical membranes, the mitotic spindles in dividing neuroepithelial cells rotate such that divisions now occur parallel to the plane of the midline (Geldmacher-Voss et al., 2003).
Several molecules implicated in vertebrate neurulation are components of the planar cell polarity or apicobasal polarity pathways (Ueno and Greene, 2003; Gotz and Huttner, 2005). Analysis of these molecular pathways in vertebrates is beginning to shed light on how the neural tube forms. The zebrafish provides a vertebrate genetic model that is amenable to live imaging and, therefore, offers distinct advantages for studying neurulation in vivo.
Here we report the identification of a mutation in the zebrafish pard6γb gene and analyze the resulting phenotype. pard6γb mutants show defects in the development of several epithelial tissues similar to has/prkci mutants (Horne-Badovinac et al., 2001; Peterson et al., 2001; Horne-Badovinac et al., 2003). Furthermore, they exhibit multiple lumens in the neural tube, a previously undescribed phenotype (Belting and Affolter, 2007). The Par3/Par6/aPKC(Prkci) complex has been suggested to affect orientation of the mitotic spindle in the neural tube stage (Geldmacher-Voss et al., 2003; von Trotha et al., 2006), and we show that pard6γb mutants exhibit defects in mitotic spindle orientation in the forming neural tube. Additional studies indicate that Pard6α, but not Pard6β or Pard6γa, could partially rescue the pard6γbs441 mutant phenotypes, thereby revealing a previously unappreciated functional diversity and complexity within the vertebrate pard6 gene family that need to be further explored.
Adult fish and embryos were maintained as described (Westerfield, 1995). We used wildtype and the following transgenic and mutant lines: heart and soulm567, nagie okom520, Tg(gutGFP)s854. pard6γbs441 was identified in a large-scale mutagenesis screen (Ober et al., 2006).
Genotyping of pard6γbs441 and nok/mpp5m520 embryos was performed using PCR and restriction fragment length polymorphisms or a dCAPS (Neff et al., 2002). Genotyping of has/prkcim567 was performed as previously described (Horne-Badovinac et al., 2001). We amplified full-length cDNA clones of pard6γb (accession NM 212563) and pard6γa (accession XM 689469) using PCR. Based on available human and mouse protein sequences, zebrafish genomic sequences, predicted sequences, published EST clones, and 5’-RACE analysis, we amplified full-length cDNA clones for pard6β(accession EF550990) and pard6β (accession EF550991), and introduced them into pCS2+. Deletion constructs were generated by PCR as described in Bit-Avragim et al. (2008b). All constructs were sequenced for accuracy. All primer sequences are available upon request.
To generate fusion proteins, eGFP or mCherry (Shu et al., 2006) followed by a short amino acid sequence (S-G-G-G-G-S) were placed into the existing expression constructs at the 5’ end of the coding sequence creating fluorescent N-terminal fusion proteins.
We designed in situ probes that targeted the 3’end and included portions of the 3’-UTR of the zebrafish pard6 genes. These sequences were cloned into pCRII (Invitrogen) and antisense DIG-labeled RNA probes were made using the appropriate SP6 or T7 promoters. We also utilized an existing myl7 (formerly cmlc2) probe (Yelon et al., 1999). Whole mount in situ hybridization was performed as described (Alexander et al., 1998).
For morpholino injections, 2ng of a splice-blocking morpholino designed against pard6γb (5’-ACATTCAACTCACCTTGCTTTTCAC-3’), a splice-blocking morpholino designed against pard6α (5’-TATTTAATGTAGGGACTCACCTCTC-3'), or a pard3 ATG morpholino (Wei et al., 2004) was injected into the yolk of wildtype embryos. For mRNA injections, we used the different constructs synthesized for the zebrafish Pard6 family: pard6γa, pard6γb, pard6α, pard6β, pard6γa-gfp, pard6γb-mch, pard6α-gfp, pard6β-gfp, pard6γb(Δcrib), pard6γb(Δpb1), pard6γb(Δcrib)-gfp, and pard6γb(Δpb1)-gfp. The generation of the mutant pard6γb constructs was described in Bit-Avragim et al. (2008). Two additional constructs were kindly provided by other laboratories, pard3-gfp (Geldmacher-Voss et al., 2003), and h2b-mrfp (Megason and Fraser, 2003). Capped RNA was synthesized in vitro by transcription with SP6 polymerase from the constructs described above using the mMessage SP6 kit (Ambion). mRNAs were injected in 2.3nL aliquots (100–400pg) into the yolk of zygotes of either wildtype or pard6γbs441 heterozygous incrosses. Several mRNA concentrations were tried for each rescue construct and only embryos showing clear fluorescent protein expression were analyzed.
Embryos were fixed in 4% PFA for 2 hours at room temperature, mounted in 4% agarose, and sectioned to 200μm using a vibratome. Sections were blocked in PBS with 4% BSA and 0.3% Triton for 1 hour and treated with primary antibody overnight at 4°C as applicable. We used the following primary antibodies: rabbit anti-aPKC (C-20) (Santa Cruz Biotechnology, 1:1000), rabbit anti-Nok/Mpp5 (Wei and Malicki, 2002), mouse anti-ZO-1 (Zymed, 1:200), mouse anti-β-catenin (BD Biosciences, 1:200), mouse anti γ-tubulin (Sigma, 1:200). Sections were washed and treated with Alexa Fluor conjugated secondary antibodies, Alexa Fluor conjugated phalloidin, or TO-PRO3 (Invitrogen) for 2 hours at room temperature. Following washing, sections were mounted on slides in Vectashield and imaged on a Zeiss LSM Pascal confocal microscope. Images were further analyzed using Image J Software including application of median filters.
To image the brain ventricles, embryos were prepared as previously described (Lowery and Sive, 2005). For imaging fusion proteins and time series, embryos were treated with Tricaine and mounted in 1% agarose. Embryos were analyzed on a Zeiss LSM5 Pascal confocal microscope or selective plane illumination microscopy (SPIM) set up (Huisken et al., 2004) and analyzed using Image J software.
The s441 mutant was identified in a forward genetic screen following ENU-induced mutagenesis (Ober et al., 2006). Compared to wildtype embryos at 36hpf (Fig. 1A), s441 mutants show phenotypes similar to the apicobasal polarity mutants has/prkci (Horne-Badovinac et al., 2001; Peterson et al., 2001), nok/mpp5 (Wei and Malicki, 2002), ome/crb2 (Omori and Malicki, 2006), and moe/epb41l5 (Jensen and Westerfield, 2004) including pericardial edema, failure of the brain ventricles to inflate, and dorsal body curvature (Fig. 1B). Complementation analyses showed that s441 was different from these four apicobasal polarity mutants (data not shown).
Since s441 mutants share several phenotypes with mutants known to have defects in epithelial tissues, we investigated epithelial tissues and their derivatives in s441 mutants. We examined the morphology of the myocardium (Fig 1C–D), gut lumen formation (Fig. 1E–F), epithelial integrity and asymmetric migration of the lateral plate mesoderm (LPM) (Fig. 1G–H), and brain ventricle inflation (Fig 1 I–K’). We found developmental defects in all of these tissues in s441 mutants. Thus, the s441 mutation causes phenotypes indicative of apicobasal polarity defects and affects the development of organs derived from epithelial tissues.
While investigating the brain ventricles of s441 mutants, we observed a novel phenotype (Belting and Affolter, 2007). Prior to 24hpf in wildtype embryos, two continuous apical membranes form along the midline to surround a lumen, the neurocoel, which extends from the anterior brain ventricles to the posterior neural tube (Geldmacher-Voss et al., 2003). In wildtype embryos, F-actin (Fig. 2A), β-catenin (Fig. 2C), aPKC/Prkci (Fig. 2E), Pard3-GFP (Fig. 2G), Nok/Mpp5 (Fig. 2I) , and ZO-1 (Fig. 2K), all localize to these continuous apical membranes. In contrast, s441 mutant neural tubes lack continuous apical membranes. In regions where apical markers are present, small lumens form, while in regions where no apical markers are present, no lumens form (Fig. 2B, D, F, H, J, L). This ‘mosaicism’ is present along both the dorso-ventral (Fig. 2) and antero-posterior axes (see Fig. 6). In wildtype embryos, the nuclei align away from the midline and thus toward the basolateral membranes (Fig. 2M). In unpolarized regions of s441 mutant neural tubes, the nuclei appear disorganized (Fig. 2N). Often, nuclei lie directly across the midline. Because s441 mutant neural tubes also show discontinuous apical membranes at 48 (Fig. 2O–P) and 72hpf (data not shown), this phenotype, which is fully penetrant (n > 400 mutants examined), does not appear to be a consequence of developmental delay. These observations suggest that the s441 mutation alters the function of a protein involved in neurocoel morphogenesis.
To get a better understanding of the s441 phenotypes, we utilized a positional cloning approach to isolate the affected gene. The s441 lesion was initially mapped to Chromosome 19 using bulk segregant analysis (Johnson et al., 1996). Fine mapping with 720 diploid embryos localized the s441 lesion to a region flanked by the simple sequence length polymorphisms (SSLP) markers z6079 and z9059. This region was narrowed using newly designed SSLP markers targeting CA-repeats (Fig. 3A). Examination of the genomic region between the two closest markers revealed the existence of three ESTs, one of which being pard6γb (Fig.3A). Because pard6 genes are known to play a role in apico-basal polarity in multiple organisms (Watts et al., 1996; Joberty et al., 2000; Lin et al., 2000; Petronczki and Knoblich, 2001), we tested pard6γb as a candidate for the gene affected by the s441 mutation.
To test whether a mutation in pard6γb caused the s441 phenotype, we took several approaches including analyzing the pard6γb expression pattern. pard6γb mRNA is maternally deposited (Fig. 3B). At 10hpf, its expression is ubiquitous and extends along the axial region of the embryo (Fig. 3C). Later at 18 (Fig. 3D) and 24hpf (Fig. 3E), pard6γb maintains its ubiquitous expression but becomes heightened in regions of the embryo that form epithelial structures. This heightened expression is particularly noticeable in regions of the neural tube (Fig. 1C–E; ZFIN, Thisse et al., 2001). This expression pattern is consistent with the tissues affected in s441 mutants. Second, we compared cDNA sequences of the wildtype and s441 mutant alleles of pard6γb. The 1302 base pairs of wildtype sequence encode a 434 amino acid protein. The mutant allele showed a base change from T to A at position 192 creating a premature stop codon at amino acid 64 (Fig. 3F–H). This single base change was confirmed by comparing genomic DNA sequences of the wildtype and s441 mutant alleles of pard6γb. To complement the information gained through genetic mapping, we injected morpholino antisense oligonucleotides designed against pard6γb to test whether lack of wildtype Pard6γb was responsible for the s441 mutant phenotypes. Injection of 2ng of a morpholino that blocks splicing between the first and second exon of the pard6γb transcript (Fig. 3A) phenocopied the s441 mutant in 92% of the injected embryos (n=98/106) (Fig. 3I). Morpholino-injected embryos showed defects in heart morphogenesis, failure of the brain ventricles to inflate, dorsal body curvature (Fig. 3I), and a discontinuous neurocoel (data not shown). To further test the hypothesis that the s441 mutation affects pard6γ γb, we injected 200pg of full length, wildtype pard6γb mRNA into embryos from heterozygous incrosses. Injection of this mRNA resulted in 97% (n=114/117) of the offspring displaying no phenotype at 30hpf. In addition, injection of s441 mutant pard6γb mRNA into s441 heterozygous incrosses resulted in 23% (n=65/282) of the offspring displaying the mutant phenotype, similar to uninjected incrosses. In similar injection experiments where embryos from heterozygous incrosses were injected with wildtype pard6γb mRNA and genotyped, 22 of 25 mutants were rescued. There was no observed phenotype caused by overexpression of Pard6γb at the quantities required for rescue. Altogether, the results from the genetic mapping, expression, allele sequence, morpholino, and mRNA rescue analyses indicate that the s441 mutation is caused by a lesion in pard6γb. Therefore, s441 will be referred to as pard6γbs441.
To determine the sub-cellular localization of Pard6γb, we constructed N-terminal fusion proteins between mCherry and Pard6γb (Pard6γb-mCh), as well as GFP and Pard6γb (Pard6γb-GFP). The addition of a fluorescent tag did not disrupt the function of these fusion proteins as evidenced by the ability of injected pard6γb-mch and pard6γb-gfp mRNA to rescue the pard6γbs441 phenotypes. Both fusion proteins showed a stage-dependent pattern of localization. In contrast to the cytoplasmic localization of Pard3-GFP at 10hpf (Fig. S1A), the Pard6γb fusion proteins localized primarily to the nucleus (Fig. S1B–C). Following gastrulation, Pard6γb fusion proteins continued to localize to the nucleus and some protein also localized to the cytoplasm and cell membranes (Fig. S1D). During the neural keel stage, when cells divide across the midline (termed mirror symmetric cell divisions) and Pard3-GFP localizes to the cleavage furrow of cells during cytokinesis (Tawk et al., 2007), Pard6γb-GFP shows an analogous localization (Figure S2A–C). After the neural keel stage, Pard6γb-mCh began to localize to the midline of the neural tube (Fig. 4A–B). As the apical membranes of the neurocoel formed and became continuous, both Pard6γb-mCh and Pard3-GFP became primarily localized to this surface (Fig. 4C–D). Through our analyses, however, we could not discern whether Pard3-GFP or Pard6γb-mCh localized to the apical membranes prior to the other one. Similar to the localization pattern of Pard3-GFP (Geldmacher-Voss et al., 2003), a low amount of Pard6γb-mCh remained in the cytoplasm. Thus, in contrast to Pard3-GFP localization prior to gastrulation, Pard6γb-mCherry and Pard6γb-GFP localize primarily to the nucleus, and during the neural keel, rod, and tube stages, they localize to the apical membranes of the neural tube and behave like Pard3-GFP.
Previous research has shown that the placement of Pard3-GFP during mirror symmetric cell divisions at the neural keel stage is important for the location of the future neurocoel (Tawk et al., 2007). At later stages, there is evidence that the re-orientation of cell division in the neural tube of zebrafish embryos is dependent on the function of the Pard3/Pard6/Prkci complex (Geldmacher-Voss et al., 2003). The multiple lumen phenotype in pard6γbs441 mutants prompted us to re-examine the role of apicobasal polarity and specifically the function of the Pard3/Pard6/Prkci complex in orienting mitoses at the neural keel and neural tube stages.
First, we examined pard6γb morphants prior to 18hpf, when mitoses occur in an orientation perpendicular to the midline and daughter cells end up on opposite sides of the midline (Geldmacher-Voss et al., 2003; Ciruna et al., 2006; Tawk et al., 2007). We also injected these embryos with pard3-gfp and ras-mcherry mRNA to facilitate this analysis, and observed no difference in cell behavior, or Pard3-GPF localization, compared to control (n=12 morphants; data not shown). Cells underwent mirror symmetric cell divisions and Pard3-GFP localized to the cleavage furrow of dividing cells.
Second, we examined embryos after 21hpf, when, in wildtype, the apical membranes along the neurocoel have become continuous and the orientation of mitoses has changed by 90° (Geldmacher-Voss et al., 2003). We analyzed mitotic events between 22–25hpf in the neural tube of live wildtype and pard6γbs441 mutants that had been injected with pard3-gfp and h2b-mrfp mRNA. We examined mitoses in regions where Pard3-GFP localized continuously along the midline in wildtype and pard6γbs441 mutant embryos (Fig. 5A–E) as well as mitoses in regions lacking Pard3-GFP localization in pard6γbs441 mutants (Fig. 5A, C). [We only analyzed mitotic events where we could clearly measure the angle of division and daughter cells remained in close proximity to their mother cell.] By measuring the angle of the plane of division in reference to the midline of the neural tube, we found that in polarized regions in both wildtype and mutant embryos, the angle of mitoses was tightly clustered between 80° and 90° (n=40/50 cell divisions) (Fig. 5B), in agreement with previous reports (Geldmacher-Voss et al., 2003; Reugels et al., 2006; von Trotha et al., 2006). In regions lacking polarity, we found the angle of mitoses to cluster between 40° and 70° (n=29/50 cell divisions) (Fig. 5C), a more dramatic effect than that observed in other polarity mutants or morphants (Geldmacher-Voss et al., 2003). Looking at this phenotype in more detail, we observed that in unpolarized regions that were 1–3 cells long, cells appeared to use information from nearby or surrounding polarized regions to orient mitotic divisions. In the rare situations when cell divisions occurred in larger unpolarized regions, the orientation of division was similar to that of cells prior to 18hpf. These data show that the multiple lumen phenotype of pard6γbs441 mutants correlates with altered cell behavior at the neural tube stage.
Experiments in C. elegans (Gotta et al., 2001), Drosophila (Hutterer et al., 2004), and MDCK cells (Joberty et al., 2000; Lin et al., 2000; Qiu et al., 2000; Hurd et al., 2003) have provided evidence that Pard6 function is dependent on three main protein-protein interaction domains: the PB1, CRIB, and PDZ domains. The PB1 domain interacts with Prkci, the CRIB domain interacts with Cdc42, and the PDZ domain (specifically the KPLG sequence) interacts with both Pals1 (Nok/Mpp5) and Pard3. These data prompted us to ask whether a particular domain of the Pard6γb protein was required for Pard6γb function in forming a continuous neurocoel.
In order to examine specifically how the PB1 and CRIB domains as well as the PDZ domain, which contains a KPLG motif essential for protein binding, affect Pard6γb localization and function, we constructed deletion alleles that eliminated the PB1 (ΔPB1) or CRIB (ΔCRIB) domain, and altered the KPLG amino acid sequence to AAAA. Injection of pard6γb(Δpb1), pard6γb(Δcrib), or pard6γb(kplg) mRNA did not rescue any of the pard6γbs441 phenotypes (data not shown). In order to analyze the localization of these proteins, we created fusion constructs and injected h2b-mRFP (a nuclear marker) and pard6γb(Δpb1)-GFP, pard6γb(Δcrib)-GFP, or pard6γb(kplg)-GFP mRNA into heterozygous pard6γbs441 incrosses and analyzed GFP localization in the neural tube of live embryos at 24hpf. In 24hpf wildtype embryos, Pard6γb(ΔPB1)-GFP localized primarily to the apical membranes around the neurocoel as well as the cytoplasm (Fig. 6A). In 24hpf pard6γbs441 mutant embryos, Pard6γb(ΔPB1)-GFP localized primarily to the discontinuous apical membranes as well as the cytoplasm (Fig. 6D). In 24hpf wildtype embryos, Pard6γb(ΔCRIB)-GFP localized to the cytoplasm as well as the apical membranes (Fig. 6B). In pard6γbs441 mutant embryos, Pard6γb(ΔCRIB)-GFP was found in the cytoplasm and the discontinuous apical membranes (Fig. 6E). Pard6γb(KPLG)-GFP exhibited similar localization as Pard6γb(ΔPB1)-GFP and Pard6γb(ΔCRIB)-GFP in wildtype (Fig. 6C) and pard6γbs441 mutants (Fig. 6F).
These data indicate that the PB1, CRIB and KPLG interaction domains are all required for Pard6γb function during neurocoel formation but not for Pard6γb localization in vivo. Alternatively, the presence of maternally derived Pard6γb or of other members of the Pard6 family may mask the requirement of these domains for Pard6γb localization.
In order to further analyze the role of the Par3/Par6/aPKC complex in apical membrane formation in the neural tube, we generated double mutants for pard6γb and has/prkci, pard6γb and nok/mpp5, as well as embryos with reduced levels of Pard6γb and Pard3. All three of these gene depletion strategies caused a more severe phenotype in apical membrane formation in the neural tube (Fig. 7). Apical membrane formation was much reduced in pard6γb/prkci double mutants (n=5) (Fig. 7B) compared to pard6γb (Fig. 7A) or prkci (data not shown) single mutants. Wildtype embryos injected with a pard3 morpholino did not exhibit an obvious defect in apical membrane formation in the neural tube (data not shown), while embryos with reduced levels of Pard6γb and Pard3 (n=5) exhibited little apical membranes in their neural tube (Fig. 7C). In nok/mpp5 single mutants, as previously reported by Lowery and Sive (2005), apical membranes in the neural tube may form in a continuous manner although the lumen does not inflate (Fig. 7D). In pard6γb/mpp5 double mutants (n=5), apical membrane formation was much reduced (Fig. 7E). In addition, all three of these gene depletion strategies led to severe epithelial disorganization in the neural tube. Altogether these analyses indicate that these four proteins function redundantly to regulate apical membrane formation in the neural tube.
Since pard6γbs441 mutants can form some apical membranes in their neural tube, we tested two hypotheses: (1) the maternal contribution of Pard6γb provides sufficient function for some apical membranes to form, or (2) one or more of the additional Pard6 family members compensates for the loss of Pard6γb function in pard6γbs441 mutants.
In order to test the first hypothesis we examined the neural tube in Maternal Zygotic (MZ) pard6γbs441 mutants, which lack both the maternal and zygotic contribution of Pard6γb. Embryos from a pard6γbs441 heterozygous incross were injected with wildtype pard6γb mRNA and raised to adulthood. Several homozygous mutant adults were identified through genetic crosses and PCR genotyping. Crossing homozygous adults produced MZpard6γbs441 offspring. MZpard6γbs441 mutant embryos exhibit body phenotypes that are similar to, yet somewhat more severe than, zygotic mutants. In particular, the retinal pigmented epithelium of MZpard6γbs441 mutants showed a more severe phenotype than that of zygotic pard6γbs441 mutants (Fig. S3A), and more similar to prkci mutants. The severity of the neural tube phenotype in embryos lacking the maternal and zygotic component of pard6γb was similar to that of embryos that only lack the zygotic component (Fig. S3B), suggesting that the maternal contribution of Pard6γb does not explain the presence of apical membranes in zygotic pard6γbs441 mutants.
To investigate the hypothesis that one or more of the additional Pard6 family members could compensate for the loss of Pard6γb function in pard6γbs441 mutants, we examined EST databases and genome browsers and found evidence for three additional pard6 genes. We amplified full-length cDNA’s for each of the additional pard6 genes. All four zebrafish Pard6 proteins show high similarity to the C. elegans and mouse Pard6 proteins, especially in the PB1, CRIB, and PDZ domains (Fig. S4A). The gene names pard6γa (located on Chromosome 16) and pard6γb have already been assigned. Using information from the ClustalW analysis (Fig. S4B) and synteny considerations, the genes currently assigned to chromosomes 18 and 23 were named pard6αand pard6β, respectively.
We examined the expression patterns of each of the three additional pard6 genes and found that they were all maternally deposited (data not shown), and subsequently showed similar expression patterns as pard6γb. There were, however, several distinct temporal and spatial expression patterns to note. For example, none of the other three pard6 genes appeared to be expressed along the axial region at 10hpf (data not shown) and pard6β showed heightened expression in the otic placodes beginning at 18hpf and present at 24hpf (Fig. S4C).
In order to determine whether any of the additional Pard6 family members could compensate for Pard6γb function, we constructed fluorescent fusion proteins (Pard6γa-GFP, Pard6β-mCh, and Pard6α-GFP) using the same method used for generating the functional Pard6γb-mCh. We injected 100–300pg of mRNA encoding these fluorescent fusion proteins into heterozygous pard6γbs441 incrosses. Pard6γa-GFP (Fig. 8A–D) and Pard6β-mCh (Fig. 8E–H) localized to the apical membranes in the neural tube but did not rescue the pard6γbs441 mutant phenotype (Fig. 8C, D, G, H). Pard6α-GFP also localized to the apical membranes of the neural tube (Fig. 8I–L). When we examined Pard6α-GFP localization in pard6γbs441 heterozygous incrosses, we observed a lower percentage of affected embryos from these injected incrosses (n=11/100 embryos) compared to uninjected incrosses (24/100 embryos). In addition, some embryos showed wildtype-like brain ventricle inflation but maintained other pard6γbs441 mutant phenotypes including pericardial edema and body curvature (not shown).
To test whether the fluorescent tags were interfering with protein function, we also performed these same experiments by injecting untagged pard6γa, pard6α, and pard6β mRNA’s into pard6γbs441 heterozygous incrosses. The results of these experiments were identical to the experiments with the fluorescent fusion proteins, indicating that the fluorescent tags were not disrupting protein function (data not shown). These data suggest that the function of Pard6γb in neurocoel formation is not completely shared with the other members of the zebrafish Pard6 family.
In order to further investigate the partial rescue of pard6γbs441 mutants with pard6α mRNA, we genotyped all embryos from a pard6γbs441 heterozygous incross injected with pard6α-gfp. Pard6α rescued the brain ventricle inflation and neurocoel phenotypes in 40% of mutants (n=4/10 mutants) (data not shown). Similarly, after co-injecting pard6α-gfp mRNA along with the pard6γb morpholino into wildtype embryos, only 58% (n=35/60) of the injected embryos exhibited collapsed brain ventricles. If Pard6α and Pard6γb function together to regulate neurocoel formation, pard6α morphants might at least partially phenocopy pard6γbs441 mutants. Using a pard6α morpholino targeted to the first exon-intron junction, we found that pard6α morphants did not phenocopy pard6γbs441 mutants, although they did exhibit pericardial edema at 30 and 56hpf (data not shown). Analysis of the neural tube of pard6γbs441 mutants injected with the pard6α morpholino revealed that apical membranes still formed (data not shown), although the pericardial edema was more severe than in pard6γbs441 mutants or pard6α morphants. Thus, while overexpression of Pard6α can partially rescue the neurocoel phenotype of pard6γbs441 mutants, it appears that the zygotic expression of pard6α does not play a primary/direct role in neurocoel formation. Of course, it remains possible that maternal supplies of Pard6α protein and/or pard6α mRNA play a role in this process.
In this paper, we report the isolation of a pard6γb mutation in zebrafish and analyze the resulting phenotypes. In addition to the phenotypes previously observed in embryos defective for components of the polarity machinery, pard6γbs441 mutants exhibit a distinct neural tube phenotype. We used this neural tube phenotype to investigate the extent of functional redundancy between the four Pard6 proteins as well as the role of the protein-protein interaction domains in Pard6γb localization and function. Our analyses of the first vertebrate mutation in a pard6 gene highlight the functional diversity of the Par6 protein family and should facilitate future analyses of this family in vertebrate systems.
The multiple lumen phenotype we describe in pard6γbs441 mutant neural tubes has not been reported before and is not observed in 24hpf has/prkcim567 or moe/epb41l5 mutants (data not shown). A defect in apical membrane formation resulting from a reduction in Pard6 function is not surprising, but the ensuing phenotype observed in pard6γbs441 mutants, i.e. the formation of multiple small lumens, is maybe less intuitive. Indeed, this multiple lumen phenotype may be driven by the affinity of apical membranes for each other, thereby leading to small clusters of cells around a central lumen. Mpp5 might be playing additional roles required for this cellular organization, or the inflation of the lumens necessary for their easy detection. A multiple lumen phenotype has also been recently described in the gut of certain zebrafish mutants/morphants (Bagnat et al., 2007), and in that tissue, single lumen formation appears to result from the inflation and coalescence of multiple small lumens. That process may also be at play in the neural tube although the alignment of the cells on either side of the midline and the subsequent alignment of their apical membranes suggest that other mechanisms must also be involved (Belting and Affolter, 2007).
Pard6 proteins are known to interact with Prkci and Cdc42 through their PB1 and partial CRIB domains, respectively. By deleting these domains, we found that the PB1 and CRIB domains are required for Pard6γb function, but dispensable for determining its subcellular localization in the neural tube at 24hpf, indicative of redundant localization mechanisms as discussed below. The PB1 domain was first identified using human Par6 in MDCK cells (Suzuki et al., 2003). Similar to our results with zebrafish Pard6γb, deletion of this domain had no effect on subcellular localization (Suzuki et al., 2001; Suzuki et al., 2003). Also consistent with our observations of ΔPB1:GFP lacking function, overexpressing the mutant human form in MDCK cells caused a different effect than overexpressing the wildtype form (Suzuki et al., 2001; Suzuki et al., 2003).
Previous investigations using Drosophila embryos and MDCK cells have addressed how DmPar-6 or mammalian Pard6A function and localization are affected when these proteins cannot interact with Cdc42. Analyses in Drosophila suggested that without interaction with Cdc42, DmPar-6 could neither function nor localize to apical membranes (Hutterer et al., 2004). However, analysis of a similar mutant form of mammalian Pard6A in MDCK cells suggested that without Cdc42 binding, Pard6A could still localize to cell membranes just as the wildtype form (Gao et al., 2004). The functional results in Drosophila were similar to those we observed with zebrafish Pard6γb(ΔCRIB)-GFP: neither protein was functional. However, the subcellular localization of the Pard6γb(ΔCRIB)-GFP form of zebrafish Pard6γb was not altered, similar to the analysis of mammalian Pard6A. One explanation for this difference could be the presence of multiple Pard6 family members in vertebrates. Because there is only one DmPar-6 protein and Drosophila par-6 mutants do not establish epithelial polarity (Hutterer et al., 2004), lack of Par-6 function (through deletion of the CRIB domain) in Drosophila embryos lacking both maternal and zygotic Par-6 would inherently lead to mislocalization. In contrast, functional redundancy between the multiple Pard6 family members in vertebrates could lead to localization of the Pard6 proteins’ binding partners including Prkci, Pard3, and Nok/Mpp5. Apical localization of these proteins (through interaction with remaining Pard6 family members) could recruit the Pard6γb(ΔCRIB)-GFP form of zebrafish Pard6γb to the apical membrane. Thus, redundant mechanisms appear to regulate Pard6γb localization, but functionality requires each of the domains tested.
Our results provide evidence that Pard6γb represents one of four apically localized Pard6 proteins in zebrafish. All four members of the Pard6 family (Pard6α, Pard6β, Pard6γa, Pard6γb) can localize to the apical membranes in the neural tube. These data are generally consistent with the reported analysis of the three mammalian Pard6 genes using MDCK cells. In agreement with our findings, mammalian Pard6B and Pard6C were localized primarily to the membranes of MDCK cells, overlapping with ZO-1, but mammalian Pard6A localized primarily to the cytoplasm (Gao and Macara, 2004). One explanation for the Pard6A localization discrepancy could be that the cytoplasmic localization was a result of overexpression or that it is cell type specific.
Despite their ability to localize apically, Pard6α, but not Pard6γa nor Pard6β, could partially compensate for the loss of Pard6γb function. Most surprisingly, injection of the pard6γa gene, which likely represents a genetic duplication of pard6γb, could not rescue the pard6γbs441 phenotype. To further test this result, we injected mRNA’s from two different wildtype alleles of pard6γa that originated from two different wildtype strains. Neither of these alleles was able to rescue the pard6γbs441 mutant phenotypes. While the possibility of an undetectable partial rescue remains, it appears that Pard6γa and Pard6γb are not functionally redundant during neurocoel formation.
Most, if not all, proteins involved in apicobasal polarity in zebrafish appear to be maternally provided, through deposition of proteins or their corresponding mRNAs. The stability of these maternally deposited components will influence when phenotypes first appear in the zygotic mutants. In such a model, it is possible that pard6γbs441 mutants, but not has/prkcim567 mutants, exhibit a neural tube phenotype because neural tube cells run out of Pard6γb before they can make enough apical membranes. Alternatively, the existence of proteins with at least partially redundant functions, e.g. Pard6α or aPKCzeta, may mask or further delay the appearance of these phenotypes. Thus, maternally deposited Pard6α may function to form apical membranes in the neural tube in the absence of maternal and zygotic Pard6γb. Alternatively, Pard6 independent mechanisms might be used for apical membrane formation as in C.elegans (Totong et al., 2007).
Despite cross-species conservation, small variations in the PB1, CRIB, and PDZ domains may contribute to differences in function (Gao and Macara, 2004). The amino acids outside these domains most likely also contribute to unique functions or interactions. One such interaction has been found between Pard6 and TGFβ receptors (Ozdamar et al., 2005). Alternatively, each Pard6 protein could be promiscuous and protein-protein interactions could thus be modulated by the expression of tissue-specific binding partners (Bit-Avragim et al., 2008b). In combination with the disparate expression patterns of the four Pard6 family members, these data suggest that the vertebrate Pard6 proteins may regulate unique aspects of development, as previously proposed (Gao and Macara, 2004). In contrast, vertebrate genomes only encode one Pard3 protein (Wei et al., 2004; Hirose et al., 2006).
Prior to 18hpf, the localization of Pard6γb fusion proteins to nuclei may be indicative of nuclear sequestration. Nuclear localization has also been observed for Nok/Mpp5 (Bit-Avragim et al., 2008a). This observation further supports the idea of unique temporal and spatial protein-protein interactions between Pard6 proteins and their binding partners, and suggests the existence of unappreciated means of regulating Pard6 localization. In addition, Par proteins have recently been associated with progenitor cell proliferation (Costa et al., 2008) and this role might be associated with their nuclear localization.
In agreement with previous observations (Gao and Macara, 2004), the vertebrate Pard6 proteins do not overlap in function, probably due to unique protein-protein interactions and/or subcellular localization. Zebrafish could provide a model system to further examine the functions of vertebrate Pard6 proteins in vivo using morpholino knockdown analysis and/or by identifying mutations in the other pard6 genes. It will be important to analyze the unique functions of known binding partners of the Pard6 protein family in vivo, such as Cdc42 (Atwood et al., 2007; Martin-Belmonte et al., 2007), and identify the additional binding partners of the Pard6 protein family in vertebrates. We examined the role of one Pard6 protein in zebrafish, Pard6γb, and found that it regulates the development of epithelial tissues. More specifically, it directs morphogenesis of the neurocoel by regulating formation of a continuous lumen along the neural tube. Further analyses of these processes and the functions of the pard6 genes in zebrafish may help identify new mechanisms regulating apicobasal polarity and neurulation.
Fig S1. Localization of Pard6γb prior to 24hpf. (A–D) Images of live wildtype embryos. Embryos were injected with pard3-gfp, pard6γb-mch, pard6γb-gfp, and/or h2b-mrfp mRNA as indicated. (A) At 10hpf, Pard3-GFP localizes primarily to the cytoplasm and cell membranes. (B) At 10hpf, Pard6γb-mCh localizes to the nucleus, cytoplasm and cell membranes. [No nuclear localization signal was found in the Pard6γb-GFP fusion protein]. (C) At 10hpf, Pard6γb-GFP co-localizes with H2B-mRFP in the nucleus. (D) At 13hpf, Pard6γb-GFP begins to change its localization pattern. It localizes primarily to the cytoplasm and cell membranes.
Fig. S2. Distribution of Pard6γb during neurulation. (A–C) Mitotic event from an embryo injected with pard6γb-gfp and h2b-mrfp mRNA. (C) Pard6γb-GFP is initially all around the membrane and then appears to localize to the cleavage furrow during cytokinesis (arrow) at 14 hpf.
Fig S3. Maternal Zygotic (MZ) pard6γbs441 mutant embryos. (A, A’) 36hpf MZpard6γbs441 mutant embryo. (A) The overall body phenotype of maternal zygotic mutant embryos is similar to zygotic mutants. (A’) The retinal pigmented epithelium is more severely affected in maternal zygotic mutants. (B) A MZpard6γbs441 mutant embryo sectioned between the first and sixth somite and stained for ZO-1 (blue), aPKC (red), and filamentous actin (green) at 30hpf. MZpard6γbs441 mutant embryos show a multiple lumen phenotype similar to that seen in zygotic pard6γbs441 mutants.
Fig. S4. The Pard6 protein family. (A) Clustal W (1.83) Multiple Sequence Alignment of the four zebrafish (ZF) Pard6 proteins with the three mouse (MM) Pard6 proteins and the C.elegans (CE) PAR-6 protein. The red box outlines the conserved PB1 domain. The green box outlines the conserved CRIB domain. The yellow boxes outline the conserved PDZ domain. (B) Cladogram of the four zebrafish (ZF) Pard6 proteins, three mouse (MM) Pard6 proteins, and the C.elegans (CE) PAR-6 proteins. (C) in situ hybridization analysis shows pard6β expression at 24hpf in the otic placodes (arrow).
We would like to thank A. Ayala, S. Waldron, N. Zvenigorodsky for excellent fish care and maintenance, the Alexander M. Reugels laboratory for Pard3-GFP and the Scott Fraser laboratory for the H2B-mRFP construct. H. Field, C. Shin, and M. Bagnat for screening, F. Del Bene and the Geraldine Seydoux laboratory for assistance with fusion protein design, K. Thorn and the Nikon Imaging Center for microscope use and expertise, D. Apollon, M. Bagnat, I. Cheung, S. Curado, S. Horne, A. Wehman, C. Yin, D. Bilder and reviewers for discussions and/or critical comments on the manuscript. C.A.M. was an NSF predoctoral fellow. J.H. is, and H.V. was, a long-term fellow of the Human Frontier Science Program Organization (HFSPO). This work was supported in part by grants from the NIH (NIDDK) and the Packard Foundation to D.Y.R.S..
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.