Zebrafish ale oko, an essential determinant of sensory neuron survival and the polarity of retinal radial glia, encodes the p50 subunit of dynactin

Microtubules play many essential roles in the structure and function of most, if not all, cells. As a component of the cytoskeleton, they provide structural support for morphological features of cells, such as axons and dendrites of neurons or apical specializations of sensory cells (e.g. Ward et al., 1975; Wen et al., 1982). Perhaps even more importantly, microtubules provide a surface for the movement of molecular motors, and so are necessary to generate the mechanical forces that translocate a variety of cargos inside the cell, including chromosomes during cell division. Among the two classes of microtubule-dependent motors, kinesins and dyneins, cytoplasmic dynein provides the predominant mechanism of minus end-directed movement. This large multi-subunit protein functions in conjunction with a second multi-subunit protein, dynactin, which is believed to increase dynein processivity and to mediate at least some of its cargo interactions (reviewed by Schroer,2004). Dynein/dynactin complexes are involved in the transport of a variety of intracellular cargos, including chromosomes(Sharp et al., 2000),mitochondria (Ebneth et al.,1998), pre-Golgi structures(Presley et al., 1997),phagosomes (Blocker et al.,1997) and lysosomes (Jordens et al., 2001).

The dynactin complex contains at least 11 different polypeptides organized into two subunits, referred to as the rod and the projecting arm (reviewed by Schroer, 2004). The rod consists of eight different components assembled around an octamer of the Arp1 polypeptide. The remaining three proteins, p150/Glued, p50/Dynamitin and p24 contribute to the arm subunit (Vaughan and Vallee, 1995; Waterman-Storer et al., 1995). The rod and arm moieties are thought to connect to each other via p50/Dynamitin. When overexpressed, p50 causes their dissociation and abolishes dynactin activity (see references below).

When studied in the context of a whole organism, dynein and dynactin mutations produce severe embryonic phenotypes. The strongest mutant alleles of the fly dynein heavy chain gene, Dhc64C, appear to be cell lethal,and the survival of early fly embryos is most likely due to the presence of the maternal contribution (Gepner et al.,1996). A similar phenotype is produced by strong Glued(dynactin 1) alleles in the fly (Garen et al., 1984). Likewise, mouse embryos mutant for the dynein heavy chain die very early at the blastocyst stage(Harada et al., 1998). Weaker mutations in genes encoding dynein and dynactin components are known to affect several aspects of neuronal development and physiology. A common consequence of such mutations in flies, nematodes and mice is the accumulation of synaptic and/or axonal proteins in neuronal cell bodies or in axonal swellings(Hafezparast et al., 2003; Koushika et al., 2004; LaMonte et al., 2002; Martin et al., 1999). Another common function of genes encoding dynein, dynactin, and associated proteins is the determination of nuclear position in a variety of cells, including fungal hyphae, the fly oocyte and zebrafish photoreceptors(Duncan and Warrior, 2002; Efimov and Morris, 1998; Shu et al., 2004; Tsujikawa et al., 2007; Whited et al., 2004). Related to these findings is the observation that interkinetic nuclear migration is abnormal in the retinal neuroepithelium of a mutant strain that carries a defect in the zebrafish mikre oko gene (also known as dctn1a) (Del Bene et al.,2008). Interestingly, this abnormality is associated with cell fate changes during retinal neurogenesis. Finally, defects in two components of the dynactin complex, Arp-1 (also known as Arp87C) and p150, were shown to destabilize synapses at the neuromuscular junction (Eaton et al.,2002).

To advance the understanding of dynactin activity in specialized cells of the nervous system, we embarked on the study of defects in the zebrafish ale oko (ako) locus, which encodes p50/Dynamitin, a component of the dynactin complex. Our analysis reveals that akoplays a role in multiple processes, including the survival of sensory hair cells, the positioning of the mitotic spindle in neuroepithelial cells, and the maintenance of the retinotectal projection. The most striking defects are,however, observed in photoreceptor cells and in retinal radial glia: the ako^jj50 mutation produces an exceptionally rapid degeneration of photoreceptors, and a severe cell-autonomous loss of apicobasal polarity in Müller glia.

The ako^jj50 allele was generated in the Malicki laboratory using a standard ENU mutagenesis protocol(Solnica-Krezel et al., 1994). The maintenance and breeding of zebrafish strains were performed using standard procedures (Malicki et al.,2002). Staging of zebrafish embryonic development was performed as described previously (Kimmel et al.,1995). All animal protocols used in this study were reviewed by the MEEI Animal Care Committee.

The preparations of plastic and frozen sections, as well as electron microscopy, were performed as previously described(Avanesov et al., 2005; Avanesov and Malicki,2004).

To map the ako locus, we genotyped F2 embryos from a cross between heterozygous carriers of the ako^jj50 allele (AB genetic background) and wild-type WIK strain homozygotes. To clone the full-length dynactin 2 cDNA, zebrafish coding sequences were obtained from Ensemble and NCBI public databases and used to design amplification primers. Total RNA was isolated with Trizol (Invitrogen) reagent at 4 days post-fertilization (dpf). Mutation site was detected by direct sequencing of PCR products amplified from both genomic DNA (primers:5′-GTTCAGGGCTCCAGTCTGACG-3′,5′-GACGCTGCTACACTGGACCA-3′) and cDNA (primers:5′-ATCTGGACTCTCTGCTCGGA-3′,5′-GTCGCCTTGTGTTTGGCAA-3′).

Morpholinos were injected into the yolk of embryos at the one-cell stage as described previously (Malicki et al.,2002). The following two morpholinos were used: ako-ATG,5′-GGTTCGCGTACTTCGGGTCGGCCAT-3′; and ako-SP,5′-GCTCTAAAGTCTCCTATGGACACAG-3′. For phenocopy experiments, we used 0.2 μg/μl of ATG and 2.0 μg/μl of SP morpholino. For maternal contribution tests, 2.0 μg/μl of ATG MO were used. To rescue the ako phenotype, the full-length coding sequence of dynactin 2was cloned into the pXT7 vector and transcribed as described previously(Malicki et al., 2002). Approximately 30 pg of mRNA was injected into embryos at the one-cell stage. mRNA-treated embryos were fixed at 3.5 dpf and cut in half to determine their genotype. The caudal halves were used to extract DNA for PCR amplification of the mutation site and sequencing, and the rostral halves of embryos were sectioned and stained to determine retinal architecture.

Fixation, infiltration and sectioning of embryos, and other steps of the antibody staining procedures have been described previously(Avanesov et al., 2005). The following primary antibodies and dilutions were used: mouse Zpr-1 (1:250,Zebrafish International Resource Center), rabbit anti-carbonic anhydrase(1:250, gift from Dr P. Linser, The Whitney Laboratory), mouse Zn8 (1:250,Oregon Monoclonal Bank), mouse anti-α-tubulin (1:1500, Sigma), mouse anti-acetylated-α-tubulin (1:1000, Sigma), rabbit anti-GABA (1:100,Sigma), mouse anti-parvalbumin (1:250, Chemicon), rabbit anti-neuropeptide Y(1:250, Immunostar), mouse anti-serotonin (1:500, Sigma), mouse HCS1 (1:250,gift from Dr J. Corwin, University of Virginia), mouse anti-ZO-1 (1:250,Invitrogen), rabbit-anti-Crb (1:250)(Omori and Malicki, 2006),rabbit anti-aPKC (1:500, Santa Cruz Biotechnology), rabbit anti-β-catenin(1:250, gift from Dr Steinbeisser, University of Heidelberg); and rabbit anti-GFP (1:1500, Clontech).

Blastomere transplantations were performed as previously described(Avanesov and Malicki, 2004),except that transgene expression was used to detect donor-derived cells. The Tg(gfap:GFP^mi2001; pax6-DF4:mCFP^q01) double transgenic ako line was used to simultaneously visualize all donor-derived cells based on CFP expression, and donor-derived Müller glia, based on GFP expression. CFP expression in donor-derived bipolar and photoreceptor cells allowed us to recognize these cell classes based on characteristic morphology. To enhance the detection of transgene expression,sections were stained with an anti-GFP antibody. This approach enhances a relatively weak CFP signal in the red channel without compromising the endogenous glial GFP signal in the green channel. Tg(brn3c:mGFP) and Tg(pax6-DF4:mGFP^s220) transgenes in ako^jj50 mutant background lines were used to visualize donor-derived ganglion and amacrine cells, respectively. During the analysis of the retinotectal projection, only sections that contained ganglion cell axons extending all the way from the optic chiasm to the tectum were taken into account.

The Tg(gfap:GFP)^mi2001 transgenic line was used for live imaging of Müller glia using the method described previously(Kay et al., 2004). To monitor the movements of Müller cells and their processes, z-series of images through selected cells were acquired every 5-10 minutes for a period of 8-16 hours. Each series consisted of 30-50 images and spanned 20-40 μm in the z dimension. Images from each z-series were automatically placed in a separate folder. ImageJ software was used to generate single plane projections from each z-series and assemble them into movies.

The ale oko (ako^jj50) allele was recovered in a chemical mutagenesis screen as a mutation that affects eye size of zebrafish larvae (Fig. 1A′,B′, compare with the wild type in A,B). Very few elongated photoreceptors were seen on histological sections of ale oko mutants at 3 dpf (not shown), and no morphologically normal photoreceptors were observed by 4 dpf (Fig. 1C′,D′, compare with C,D). Consistent with that,immunostaining with the Zpr-1 antibody revealed photoreceptor loss already at 2.5 dpf (red signal in Fig. 1F′, compare with F). By 3 dpf, all photoreceptors displayed rounded morphology (not shown), and by 4 dpf ∼90% of photoreceptors were absent (Fig. 1G′, compare with G; see also Fig. S1 in the supplementary material). By 7 dpf, the photoreceptor cell layer (PRCL) was absent (not shown). Mutant fish degenerated and died shortly afterwards. Ultrastructural analysis revealed that outer segments were missing in most regions of the ale okomutant retina. The few outer segments that did differentiate were frequently directed sidewise or even basally (Fig. 1E′, asterisk, compare with E).

To evaluate whether other retinal cell classes are also affected in ako mutants, transverse sections collected from the vicinity of the optic nerve were immunostained with several antibodies, including anti-carbonic anhydrase antibody for Müller glia(Fig. 1G,G′), and four antibodies for amacrine cells: anti-serotonin(Fig. 1H,H′),anti-parvalbumin (Fig. 1I,I′, green), anti-neuropeptide Y(Fig. 1I,I′, red) and anti-GABA (Fig. 1J,J′). In parallel, we also studied the ako mutant phenotype in two transgenic lines, Tg(brn3c:mGFP) and Tg(gfap:GFP), which express GFP in ganglion cells (Fig. 1K,K′) and Müller glia(Fig. 1F,F′),respectively. The most striking deficiency revealed by these studies was a severe reduction in the number of Müller cells(Fig. 1F′, compare with F). This decrease is consistent with the phenotype reported for mutations in the mok^s309 mutant strain, which harbors a mutation in the dynactin 1 (p150) gene(Del Bene et al., 2008). We did not observe striking differences in amacrine or ganglion cell populations at 3 dpf.

In addition to photoreceptors, mechanosensory hair cells are also affected in ako mutants. To visualize hair cells, we stained embryos with either anti-acetylated tubulin antibody, which highlights their apical surface and the kinocilia (Bang et al.,2001; Omori and Malicki,2006), or HCS-1 antibody, a hair cell marker(Gale et al., 2002). Sharply lower numbers of hair cells were observed in the posterior macula of ako animals at 5 dpf (Fig. 2A′, compare with A; quantitative data is shown in Fig. 2E). ako mutant cristae and the anterior macula were, however, relatively normal(Fig. 2B′, compare with B; data not shown). On electron micrographs, we observed an abundance of cellular debris in the posterior macula of ako mutants, a likely indication of cell death (arrowheads in Fig. 2C′, compare with C). The hair cell defect was most pronounced in the lateral line: the majority of hair cells in this organ degenerated between 4 and 5 dpf(Fig. 2D′, compare with the wild type in D; quantitative data is shown in Fig. 2F). These results indicate that ako is required for the survival of mechanosensory cells.

By comparing mapping results with data in public databases, we found that zebrafish dynactin 2 localizes to the vicinity of the akolocus in linkage group 6 (Fig. 3A). The cloning of dynactin 2 cDNA revealed that it contains an open-reading frame encoding 405 amino acids, and is composed of at least 15 exons. The analysis of the dynactin 2 sequence from ako^jj50 animals revealed a 9-bp insertion in the mutant transcript. This insertion introduces an in-frame stop codon just upstream of exon 11 (Fig. 3C, compare with B) and leads to the truncation of 143 C-terminal amino acids. To investigate whether the insertion was specific to the mutant gene, primers targeted to the insertion sequence were used for RT-PCR amplification. As expected, these primers generated an amplification product in the mutant only, supporting the claim that the insertion was associated with the ako mutant phenotype(Fig. 3D, lanes 2 and 3). The identity of this amplification product was confirmed by BfaI digestion (Fig. 3D, lane 4). Importantly, in the genomic sequence we also uncovered a CTG→CAG transition upstream of exon 11 (red asterisk in Fig. 3C, compare with B). This sequence change is likely to generate an ectopic AG splice donor site and is most likely responsible for the 9-bp insertion in the transcript. The injection of a combination of ATG- and SP- (splice site)-directed anti-dynactin 2 morpholinos produced a small eye phenotype that resembled that of ako^jj50 mutants(Fig. 3F, compare with E; see also Table S1 in the supplementary material). Consistent with that, dynactin 2 mRNA injections rescued both small eye size and photoreceptor phenotypes (Fig. 3H, compare with G; see also Table S2 in the supplementary material). The overexpression of dynactin 2 from a Tol2 transposon-based vector using a heat-shock-inducible promoter(Kawakami et al., 2004) also rescued photoreceptor phenotype (data not shown). These results indicate that the ako mutant phenotype is due to a defect in the dynactin 2 gene.

To test whether ako transcript is provided maternally, we performed a knockdown using an anti-ATG morpholino. This treatment resulted in a delayed epiboly so that in many embryos yolk plug closure was not completed even at 12 hours post-fertilization (hpf). By 24 hpf, the yolk was abnormally elongated, the tail region was severely stunted, and cell death appeared to be abundant along the entire embryonic axis. These defects were rescued by ako mRNA injections (see Table S2 and Fig. S3 in the supplementary material). These observations suggest that ako function is supplied maternally in the zebrafish embryo.

Several studies revealed that the dynactin complex is involved in axonal outgrowth, synapse stabilization, and/or axonal transport(Eaton et al., 2002; Grabham et al., 2007; LaMonte et al., 2002). To investigate whether this is the case in retinal neurons, we inspected the morphology of ale oko mutant ganglion and amacrine cells. To distinguish cell-autonomous and non-autonomous aspects of the akophenotype and to circumvent lethality of ako^jj50 mutants,we performed blastomere transplantation. To do that, we crossed the ako^jj50 mutation into Tg(brn3c:mGFP) and Tg(pax6-DF4:mGFP^s220) transgenic lines, which express GFP in ganglion and amacrine cells, respectively(Kay et al., 2004; Xiao et al., 2005). The use of transgenic strains can potentially allow one to detect donor-derived clones throughout the lifetime of mosaic animals, and also provides a better resolution for visualizing neuronal morphology. Mosaic analysis revealed that at 4 dpf homozygous ako^jj50 mutant cells contributed to the ganglion cell layer (GCL) and formed retino-tectal projections in the wild-type environment. ako^jj50 projections did not obviously differ from those produced by wild-type cells in the wild-type environment at the same stage (Fig. 4A′,B′, compare with wild-type donor-derived cells in A,B). In the wild-type host environment, ako^jj50 mutant cells were also able to produce extensive dendritic trees at 4 dpf(Fig. 4C′,D′,compare with wild-type donor cells in C,D). Similarly, the dendrites of ako^jj50 amacrine cells contributed to the inner plexiform layer and were grossly indistinguishable from those of wild-type cells(Fig. 4E′, compare with E). Finally, some ako^jj50 donor-derived bipolar cells produced normal apical and basal processes in the wild-type environment at 4 and 6 dpf (Fig. 4F′,compare with F; data not shown). These results suggest that in contrast to photoreceptors, retinal interneurons as well as ganglion cells do not require dynactin 2 cell-autonomously for the initial differentiation of their gross morphological features. However, we cannot exclude the possibility that subtle changes might exist in the stratification of dendrites, the distribution of synaptic termini or in the innervation of the optic tectum by ako^jj50 cells.

To evaluate whether dynactin 2 is required for the maintenance of ganglion cell morphology, we repeated mosaic analysis at 10 dpf. In most cases(12/17), wild-type donor-derived ganglion cells transplanted into wild-type retinae differentiated densely branched processes in the optic tectum(Fig. 4G,H). In contrast to wild-type cells, we have not observed densely branched axonal processes in any transplants that involved ako^jj50 mutant ganglion cells at this stage (0/11; Fig. 4G′,H′). Instead, mutant projections differentiated few branches or an irregular branching pattern. In some cases, we observed swellings near the termini of ako^jj50 mutant axons(Fig. 4H′, arrows). These results indicate that dynactin 2 is essential for the maintenance of the retino-tectal projection.

To determine whether ako functions cell-autonomously in photoreceptor cells, we generated genetically mosaic animals. Donor cells were distinguished based on expression of the Tg(pax6-DF4:mCFP^q01) transgene, and detected by staining with an anti-GFP antibody. Both wild-type and mutant-derived donor cells contributed to all layers of wild-type retinae, including the PRCL at 3.5, 4 and 6 dpf. At 4 dpf, the ratio of donor-derived photoreceptors to INL cells was 29% for mutant cells in the wild-type environment, compared with 35%for wild-type cells in the wild-type environment (see Table S3 in the supplementary material). This survival rate is dramatically better than that of ako^jj50 photoreceptors in ako^jj50mutant retinae, in which ∼90% of photoreceptor cells died by 4 dpf (see Fig. S1 legend in the supplementary material). In contrast to akomutant retinae, ako mutant cells in the wild-type host environment formed morphologically distinguishable, elongated photoreceptors even at 6 dpf(see Fig. S2 in the supplementary material). These results indicate that,similar to mikre oko (Tsujikawa et al., 2007), the ale oko photoreceptor phenotype contains a non-cell-autonomous component.

Our initial analysis of the ako retina revealed abnormal morphology of Müller glia (Fig. 1F′, compare with the wild type in F). To monitor gliogenesis, we crossed the ako^jj50 mutant allele into the Tg(gfap:GFP)^mi2001 transgenic line, which expresses GFP in Müller glia (Bernardos and Raymond,2006). In wild-type larvae, GFP-expressing Müller glia were well differentiated by 60 hpf (Fig. 5A). Their processes spanned the entire retinal thickness, and perikarya localized to the INL. By contrast, in the ako^jj50 retina, the number of GFP-positive glial cells was reduced by ∼40% (19±6, n=5 sections) compared with the wild type (33±6, n=5), and 40% (39/96, n=5) of mutant Müller cell perikarya were mislocalized, so that they were found basal to the INL, compared with ∼7% (11/168, n=5) in the wild type(Fig. 5A; data not shown). In addition, we found that only 18% (17/96, n=5 sections) of these cells had distinguishable apical processes at 60 hpf, compared with 82% (137/168, n=5) in the wild type. By 4 dpf, 74% (240/324, n=13) of ako Müller cell perikarya were basally mispositioned, and apical processes were absent (Fig. 5). In contrast to these observations, brain radial glia appeared to be grossly normal at least until 8 dpf (not shown). These observations indicate that dynactin 2 is a key determinant of Müller glia differentiation.

To investigate how the dynactin complex is involved in the positioning of Müller cell perikarya, we performed live imaging of these cells in the ako^jj50 mutant strain carrying the Tg(gfap:GFP)^mi2001 transgene. This transgene expresses GFP in Müller glia making it possible to continuously monitor the morphology of these cells in living animals. Imaging was performed for 8-16 hours,starting at 72-76 hpf. While using this imaging approach, respectively 94%(73/78, n=3) and 95% (74/78, n=3) of wild-type glial cells featured the apical and the basal process during the recording period. In all wild-type cells imaged, apical and basal processes, once differentiated, were continuously maintained, and we did not observe a displacement of cell perikarya (see Movie 1 in the supplementary material). By contrast, in ako mutants only 46% (32/69, n=6) of cells differentiated the apical process, whereas 84% (58/69, n=6) featured the basal process. All apical processes of mutant cells were retracted during the recording period. Interestingly, this was frequently followed by a rapid basal displacement of perikarya, which we observed in 38% (12/32) of cells imaged(Fig. 5B; see also Movies 2 and 3 in the supplementary material). This displacement frequently occurred in less than 1 hour (Fig. 5B,C),and in the most extreme cases its maximum velocity exceeded 1 μm/minute. In wild-type cells, the highest displacement velocity reached ∼0.2μm/minute, and was associated with small saltatory apicobasal movements of nuclei (Fig. 5D). It is worth pointing it out that we have not observed any cases of apical displacement of Müller cell perikarya (0/32), and that all cells that displayed basal perikaryal displacement maintained their basal process throughout the course of their translocation. These results indicate that dynactin 2 plays a key role in the maintenance of the apical process, and functions in localizing Müller cell bodies to the inner plexiform layer.

To investigate whether the Müller glia phenotype results from defects intrinsic to these cells or is a consequence of abnormal interactions with surrounding cell classes, we used genetic mosaics. Blastomere transplantation was performed using the Tg(gfap:GFP^mi2001;pax6-DF4:mCFP^q01) double transgenic ako line as the donor, and a non-transgenic line as the host. Embryos were stained with an anti-GFP antibody to visualize all donor-derived cells based on CFP expression(Fig. 6, red signal), and GFP endogenous transgene expression was used to distinguish donor-derived glial cells (Fig. 6, green signal,appears yellow owing to the overlap with the red signal). Wild-type Müller glia in wild-type hosts displayed several distinct morphological features: their cell bodies localized to the INL, and their processes extended both apically and basally, contacting the outer (OLM) and inner (ILM) limiting membranes, respectively. Moreover, the apical terminus of each glial cell was richly branched, and contacted numerous photoreceptors(Fig. 6G). All of these features were already well formed in wild-type cells at 3.5 dpf(Fig. 6A), and did not display significant changes at least until 8 dpf(Fig. 6D). Similarly, ako^jj50 mutant glia in the wild-type host environment frequently differentiated apical processes at 3.5 dpf (24/50, n=8 retinae; Fig. 6A′;quantitative data in Fig. 6F). By 4 dpf, however, the majority of ako^jj50 Müller glia that differentiated in the wild-type environment lost apical and basal processes (39/40, n=6; Fig. 6B′, compare with B; quantitated in Fig. 6F). This stunted morphology persisted for several days, until at least 8 dpf(Fig. 6D′).

The loss of the apical process in Müller cells could result from a photoreceptor defect. This possibility may appear likely as retinal cell clones tend to contribute to all layers of the retina simultaneously(Holt et al., 1988; Turner et al., 1990; Wetts and Fraser, 1988). For this hypothesis to be correct, mutant photoreceptor cells should be present apical to defective Müller glia at least during early stages of neurogenesis. Contrary to this prediction, we observed aberrant glial cells(9/19) in ako^jj50 clones that did not feature photoreceptors at 4 dpf and later (e.g. Fig. 6D′,E′). This observation eliminates the possibility that mutant photoreceptors survive in mosaic retinae but are unable to interact with Müller glia and thus cause glial defects. In an alternative scenario, mutant photoreceptors could be present earlier in development but then die and, consequently, cause glia defects. This scenario appears unlikely for several reasons. First, one has to note that a single glial cell (even mutant) contacts many photoreceptors (see Fig. 6A′,G), and thus many photoreceptor cells would have to die to result in the loss of apical junctions of Müller glia. Second, in the retinae of several zebrafish photoreceptor mutants, Müller glia persist largely intact following a complete photoreceptor loss (Doerre and Malicki, 2002). Third, to detect photoreceptor defects, we stained mosaic retinae with Zpr-1 antibody, which recognizes double cones (example shown in Fig. 6E,E′). We have not seen any aberrations suggestive of cell death in the photoreceptor cell layer. Finally, if photoreceptor cells died rapidly, then the ratio of photoreceptors to other cell classes would be drastically changed in ako^jj50 mutant clones at 3.5-4 dpf as compared with control wild-type donor-derived clones. This is not the case. Given these considerations, our data support conclusion that the Müller glia defect is cell autonomous.

Müller glia are highly polarized, their surface is subdivided into apical and basolateral domains by a belt of cell junctions, the outer limiting membrane (OLM) (Rodieck,1973). We have previously shown that apical surface determinants,Nagie oko (Nok, also known as Mpp5a) and Crumbs, localize to the vicinity of the OLM (Omori and Malicki,2006; Wei and Malicki,2002). To investigate the fate of apical surface proteins in ako mutants, we stained ako^jj50 retinae with antibodies to ZO-1 (also known as Tjp1), aPKC and Nok at 3 dpf. As expected,in the wild-type retina, ZO-1, Nok and aPKC staining was found almost exclusively at the OLM (Fig. 7A,C,E). Very few ectopically localized staining areas were observed near the inner surface of the retina (less than one per section for Nok and aPKC, and four per section for ZO-1; n=5 sections). By contrast, all three proteins frequently accumulated in ectopic aggregates in displaced Müller cells of the ako mutant (9, 10 and 16/section for Nok, aPKC and ZO-1, respectively; n=5; Fig. 7A-F″). Respectively, ∼80% and ∼90% of ectopic Nok and aPKC spots colocalized with ZO-1 (Fig. 7B-F″). These results indicate that ako is required for the proper localization of apical polarity proteins in Müller glia. As the centrosome of Müller glia localizes to the apical surface, far away from the bulk of the cytoplasm (Fig. 7G-H′), abnormal localization of these polypeptides is likely to have resulted from a defect in minus end-directed microtubule-dependent transport (schematically illustrated in Fig. 7I).

Previous studies demonstrated in several different contexts that dynactin is involved in the positioning of the mitotic spindle(Eshel et al., 1993; Gonczy et al., 1999; Li et al., 1993; Muhua et al., 1994; Toyoshima et al., 2007). To investigate whether ako is required for proper spindle orientation in neural progenitors of the retina, embryos were genotyped by sequencing at 30 hpf, sectioned, and stained with anti-α-tubulin antibody to visualize mitotic spindles. The orientation of mitotic spindle axes were evaluated relative to the apical surface of the neuroepithelium (see Fig. S4A in the supplementary material). As reported previously(Das et al., 2003), in wild-type embryos mitotic spindles were largely parallel to the apical surface of the neuroepithelium (see Fig. S4B in the supplementary material). Only 7%of cell divisions (8/121, in 39 embryos) occurred at orientations greater than 30° (see Fig. S4C in the supplementary material). In ako mutant embryos, however, mitotic spindles were frequently positioned at wider angles relative to the apical surface. Twelve percent of ako^jj50mitotic spindles (25/208, in 41 embryos) were oriented at an angle greater than 60° and 25% (53/208) of mitotic spindles were oriented at an angle greater than 30° relative to the apical surface (see Fig. S4D in the supplementary material). An extreme example of this is shown in Fig. S4B′ in the supplementary material. These observations indicate that ako function is required for the proper orientation of the mitotic spindle.

We demonstrated that p50, a dynactin component, is involved in several aspects of nervous system development. Prior to these studies,knockout of dynein heavy chain in the mouse was shown to produce early embryonic lethality (Harada et al.,1998). This phenotype is much more severe than those of the zebrafish dynactin mutants mok and ako (this work)(Del Bene et al., 2008; Tsujikawa et al., 2007). A likely factor that accounts for these differences is the presence of the maternal contribution in the zebrafish embryo. ako^jj50 is the second mutation of a dynactin component to be found in zebrafish. Previously characterized mutant strains, mok^m632 and mok^s309, carry defects in the p150 (dynactin 1) gene (Del Bene et al.,2008; Doerre and Malicki,2001; Malicki et al.,1996; Tsujikawa et al.,2007). The ako^jj50 mutant phenotype is more severe than that of mok^m632 in both photoreceptors and Müller glia (see Fig. S1 in the supplementary material).

Why are sensory cells and Müller glia particularly sensitive to the loss of dynactin? The answer to this question might lie in the fact that these cell types feature an elongated and highly polarized morphology; similar to epithelia, their surfaces are subdivided by cell junctions into apical and basolateral domains, and their centrosomes are positioned at the apical cell surface, far away from biosynthetically active perinuclear regions(Dowling, 1987; Rodieck, 1973). As a consequence of this, apically directed cargo travels a long distance and relies on minus-end-directed microtubule-dependent motors(Troutt and Burnside, 1988). The importance of a robust apically directed cytoplasmic transport mechanism is particularly obvious in photoreceptors. In these cells, many phototransduction cascade components are likely to be transported towards the outer segment via a microtubule-dependent mechanism. One has to note, however,that opsin itself may not require dynactin for its transport(Tai et al., 1999; Tsujikawa et al., 2007).

A striking feature of cell polarity in Müller glia is the particularly large distance between their nuclei and apically localized centrosomes(Fig. 7G,G′,I). Each Müller cell extends a long apical process which branches around several photoreceptors and forms junctions at their surface(Fig. 6G). The centrosome,presumably the main microtubule organizing center, localizes to the very apical termini of Müller glia (Fig. 7G,G′,I), and thus the minus ends of cytoplasmic microtubules most likely also point apically in these cells. Given the exceptional length and extensive branching of the apical process in Müller glia, its formation and maintenance is likely to require a very active minus end-directed microtubule-dependent trafficking system. Consequently, compared with other cell types, the reliance of Müller glia on the minus end-directed motor dynein may be more pronounced, and so defects in this motor might lead to cell degeneration. The observation that the apical determinants Nok and Has/aPKC accumulate in the cell bodies of Müller glia supports this scenario. As transport mechanisms in radial glia are poorly investigated so far, the analysis of the ako mutant offers one of the first insights into this area.

Although the mechanism of ako involvement in the maintenance of the retinotectal projection remains unclear, defects that we observed in the ako mutant might be medically relevant, as ganglion cell degeneration, glaucoma, is a frequent cause of human blindness(Quigley, 1996). It has been,in fact, documented that dynein components accumulate at the optic nerve head following an increase of intraocular pressure, a major risk factor for this disease (Martin et al., 2006). Finally, we note that the mitotic spindle phenotype in ako^jj50 mutants suggests that dynactin might be required for the proper distribution of cell fate determinants during asymmetric cell divisions and thereby might affect cell fate. Although the role of mitotic spindle positioning in retinal cell fate decisions remains unclear(Cayouette and Raff, 2003; Das et al., 2003; Silva et al., 2002; Zigman et al., 2005), in other developmental contexts the contribution of mitotic spindle positioning to the outcome of asymmetric cell divisions is well documented (for a review, see Gonczy, 2008).