Dynein Regulates Epithelial Polarity and the Apical Localization of stardust A mRNA

Intense investigation has identified an elaborate protein network controlling epithelial polarity. Although precise subcellular targeting of apical and basolateral determinants is required for epithelial architecture, little is known about how the individual determinant proteins become localized within the cell. Through a genetic screen for epithelial defects in the Drosophila follicle cells, we have found that the cytoplasmic Dynein motor is an essential regulator of apico–basal polarity. Our data suggest that Dynein acts through the cytoplasmic scaffolding protein Stardust (Sdt) to localize the transmembrane protein Crumbs, in part through the apical targeting of specific sdt mRNA isoforms. We have mapped the sdt mRNA localization signal to an alternatively spliced coding exon. Intriguingly, the presence or absence of this exon corresponds to a developmental switch in sdt mRNA localization in which apical transcripts are only found during early stages of epithelial development, while unlocalized transcripts predominate in mature epithelia. This work represents the first demonstration that Dynein is required for epithelial polarity and suggests that mRNA localization may have a functional role in the regulation of apico–basal organization. Moreover, we introduce a unique mechanism in which alternative splicing of a coding exon is used to control mRNA localization during development.

Introduction A polarized architecture increases the level of sophistication at which cells can interact with their neighbors and the environment. Thus, it is not surprising that highly polarized epithelia have become the predominant tissue type among metazoans. Epithelial cells have distinct apical and basolateral membrane domains, each of which contains a unique complement of lipids and proteins to regionalize cellular functions. Intracellular junctions separate the apical from basolateral domains and mediate cell-cell adhesion and selective permeability within the monolayer. Although the mechanisms that specify apico-basal polarity have been a topic of great interest in recent years, there are still many open questions about how this cell architecture is established and maintained.
Studies in model organisms and mammalian tissue culture have identified an elaborate protein network regulating epithelial polarity across animal species [1][2][3]. At the core of this network are three protein complexes, which sequentially localize to and specify one of the two membrane domains and position the adherens junctions at their interface. The Baz complex (Bazooka/aPKC/Par6) acts first in the hierarchy to specify the apical domain [4][5][6][7]. The Scrib complex (Scribble/Discs Large) then functions as a basolateral determinant by repressing the apicalizing activity of the Baz complex. Finally the Crb complex (Crumbs/Stardust/DPATJ) is targeted to the apical membrane to antagonize the Scrib complex [6,7].
Crb is the only transmembrane protein among the known core polarity determinants and is a potent regulator of apical identity. Over-expression of this protein is sufficient to cause an expansion of the apical membrane at the expense of the baso-lateral domain in Drosophila [8]. Crb is stabilized in the plasma membrane by the cytoplasmic scaffolding protein, Stardust/MPP5 (Sdt), which binds Crb's cytoplasmic tail through its PDZ (PSD-95/Discs large/ZO-1) domain [9][10][11][12]. A third member of the complex, DPATJ, interacts with Sdt via shared L27 domains [12,13]. The Crb complex is highly conserved both at the structural and functional level between flies and vertebrates [12,14,15], and mutations in human CRB1 have been linked to two severe forms of retinal dystrophy [16].
The Crb complex must be precisely localized to the apical domain to exert its effect on epithelial polarity, yet little is known about how the individual proteins within this complex are targeted to this position. Extensive studies in cultured mammalian cells have implicated exocytic vesicle trafficking in polarity regulation [17,18]. Under this paradigm, vesicles carrying newly synthesized transmembrane proteins are specifically targeted to either the apical or basolateral domain. While direct vesicle trafficking is an appealing potential mechanism for the localization of Crb, it is likely to be of limited use in the apical targeting of Sdt and DPATJ, the cytoplasmic members of this complex. Moreover, the currently recognized members of the Baz and Scrib complexes are entirely comprised of cytoplasmic proteins, which must be specifically localized to either the apical or lateral domains to perform their polarity functions. Uncovering the mechanisms that govern the subcellular targeting of cyto-plasmic polarity determinants is critical, therefore, to our understanding of apico-basal polarity regulation.
One mechanism that many cell types employ to position proteins within the cytoplasm involves the subcellular localization of their corresponding mRNAs [19][20][21][22][23][24]. There are several strategies by which mRNAs can become enriched within a given cytoplasmic domain, but the most common strategy involves the assembly of transcripts into large RNAprotein particles that are transported along cytoskeletal elements by molecular motors. Subcellular transport, followed by localized translation, ensures that new proteins will only be produced in the cytoplasmic region where their function is required [25]. This mechanism, which is conserved across such diverse taxa as fungi, plants and animals, has been particularly well studied in the Drosophila oocyte where mRNAs encoding patterning regulators such as oskar, gurken and bicoid are precisely targeted to the three different regions of the oocyte where their protein products are required [23]. mRNA localization has also been observed in the Drosophila blastoderm, where transcripts such as wingless, and those encoding pair-rule genes, are targeted to the apical cytoplasm [26][27][28][29].
Here we show that loss of the minus end-directed microtubule motor Dynein in the Drosophila follicle cell epithelium leads to defects in apico-basal polarity. Our data indicate that Dynein is required for the localization of Sdt and suggest a mechanism whereby the Dynein-dependent apical targeting of specific sdt mRNA isoforms enriches Sdt protein in the apical cytoplasm. Surprisingly, the apical localization signal for the sdt mRNA maps to an alternatively spliced coding exon. Selective removal of this exon coincides with a developmental switch between localized and unlocalized isoforms of the transcript, in which apically localized sdt transcripts play an early role in the assembly of the Crb complex, while unlocalized transcripts are sufficient to maintain polarity in mature epithelia. These results suggest that the strategy for the apical targeting of the Crb complex changes during epithelial maturation and introduce a new class of gene products, mRNAs, to the existing paradigm for apico-basal polarity regulation

Cytoplasmic Dynein Function Is Required for Epithelial Cell Shape and Integrity
To identify new genes required for epithelial architecture and morphogenesis, we performed a genetic screen in the follicle cell (FC) epithelium of the Drosophila egg chamber (EC). The FCs initially form a uniform, cuboidal epithelium surrounding the germ cells ( Figure 1A), but undergo dramatic rearrangements during mid-oogenesis to create a columnar epithelium in the posterior of the EC and a squamous epithelium in the anterior [30]. We used EMS mutagenesis and the Flp/FRT system, under the control of the e22c Gal4 driver, to screen for epithelial defects in genetically mosaic FC epithelia. Mutant clones were identified by the lack of green fluorescent protein (GFP) and stained with rhodamine phalloidin to reveal cell shape. We screened approximately 5000 chromosomes and recovered 61 mutations that disrupt various aspects of epithelial morphology (S. Horne-Badovinac and D. Bilder, unpublished data).
Among the many phenotypes identified in the screen, we found two complementation groups, one with 3 alleles (D12-5, N22-1, Q43-4) and one with a single allele (K194), that showed identical phenotypes. Each of the four mutations causes rounding and multilayering of the FCs predominantly in the posterior of the EC as well as epithelial gaps, which likely arise from reduced cell-cell adhesion ( Figure 1B and 1D, and data not shown). This combination of phenotypes is similar to those seen when either crb or sdt are mutated in the FCs ( [31], unpublished data) suggesting that these new mutations might also disrupt apico-basal polarity by affecting the specification of the apical membrane domain.
We employed deficiency mapping, followed by candidate gene analysis, to identify the genes disrupted by these new mutations. All three alleles of the first complementation group failed to complement Df(3L)Exel6102. Complementation tests with known lethal mutations in this chromosomal region revealed that this group corresponds to Drosophila's lone cytoplasmic Dynein heavy chain gene, Dhc64C. Moreover, the second group failed to complement Glued 1 , a mutation that disrupts the p150 subunit of Dynactin, which functions as an accessory complex for Dynein transport [32]. Together these data show that both complementation groups disrupt genes required for cytoplasmic Dynein transport and indicate that this motor is required for cell architecture and epithelial continuity in the FC epithelium.
It is curious that our screen identified three alleles of Dhc64C, as it had been previously reported that this gene is required for cell viability in somatic clones [33]. To provide further evidence that this complementation group does indeed disrupt Dhc64C, we made FC clones that were homozygous for Dhc 4À19 , a previously characterized strong Dhc allele [33]. Dhc 4À19 mutant FC clones are not only viable, but contain rounded, multi-layered cells and epithelial gaps highly similar to mutant clones for our newly isolated alleles ( Figure 1C). Moreover, in both Dhc 4À19 and D12-5 mutant clones, Dhc protein levels are significantly reduced when compared to wild-type cells ( Figure 1E and 1F). Finally, introduction of a Dhc transgene was sufficient to rescue

Author Summary
Cells within epithelial sheets are highly polarized with distinct apical and basolateral membrane domains. This cellular organization is critical to both epithelial form and function, and a failure to maintain epithelial polarity is often linked to tumor progression. The protein network that establishes and maintains the two membrane domains relies on the precise subcellular localization of its molecular components, but little is known about how these proteins are targeted to their sites of action. We have shown that the localization of the apical determinant protein Stardust depends on the microtubule motor Dynein. While investigating the relationship between Dynein and Stardust, we also made two unexpected observations about stardust mRNA regulation. First, the mechanism by which Dynein localizes Stardust may depend, in part, on the apical targeting of the stardust mRNA. Second, some stardust mRNA is apically localized during early stages of epithelial development, but the selective removal of the apical localization signal leads to the sole production of uniformly localized transcripts in mature epithelial cells. Together, these results introduce roles for Dynein in apico-basal polarity regulation and raise important questions about the role of mRNA localization in the targeting of polarity determinant proteins and epithelial maturation.
lethality of D12-5 hemizygotes (data not shown). From these data we conclude that D12-5, N22-1 and Q43-4 are bona fide Dhc alleles and that our ability to make strong Dhc loss-offunction FC clones provides a unique opportunity to study cytoplasmic Dynein function in epithelia.

Loss of Dhc Disrupts the Apical Localization of Sdt and Crb
How does loss of Dhc lead to cell architecture defects in the FCs? Dhc is the force-generating subunit of the cytoplasmic Dynein motor, which transports diverse cargoes, such as vesicles, mRNAs and organelles, toward microtubule (MT) minus ends in eukaryotic cells. As is typical for polarized epithelia, FCs contain a dense population of MTs along their lateral cortex with their minus ends pointed toward the apical surface ( Figure 2A) [34,35]. This cytoskeletal geometry suggests that Dynein could affect cell architecture by targeting apical determinant proteins to their site of action at the apical cell surface. Crb stands out as a potential Dynein cargo because it is the only transmembrane protein among the known apical determinants. We initially hypothesized, therefore, that Dynein might affect apico-basal polarity by transporting vesicles containing newly synthesized Crb to the apical surface.
To test whether Dynein is required for the apical targeting of Crb, we performed immunohistochemistry on Dhc FC clones and focused our analysis on lateral clones, in which mutant cells retain cuboidal morphology. The results show that, even in Dhc cells with relatively normal shape, Crb is missing from the apical surface ( Figure 2B). Although Dhc cells occasionally display a weak, cytoplasmic staining for Crb (data not shown), overall protein levels are severely reduced when compared to wild-type cells. These experiments show that Dynein is indeed required for the apical targeting of Crb and suggest that Dynein transport is also necessary for the maintenance of wild-type Crb levels in the FCs.
The fact that Crb is missing from the apical surface in Dhc clones is consistent with our initial hypothesis that Dynein could be transporting Crb-containing exocytic vesicles, but we were struck by the strong general reduction in Crb protein levels. It has been previously shown that Crb levels are reduced in embryonic epithelia when its cytoplasmic binding partner stardust (sdt) is mutated [36], so we next investigated whether Dynein is required for Sdt localization. Sdt is strongly depleted from at the apical surface in Dhc cells and is instead found throughout the cytoplasm ( Figure 2C). These data indicate that Dynein is also required for the apical targeting of Sdt and suggest that the mechanism by which Dynein localizes Crb could be indirect.
Although it is well established that Crb and Sdt depend upon one another for their apical localization, the degree of this dependency can vary with the identity and developmental stage of the tissue examined [9,10,36]. To determine whether the disruption of just one of these proteins in Dhc FC clones could account for the mis-localization of the other, we investigated the relationship between Crb and Sdt in previtellogenic egg chambers. Crb is only partially required for the apical localization of Sdt in the FCs, as there is still a significant amount of Sdt at the apical surface and cell junctions of crb mutant cells prior to stage 8 ( Figures 2D and  S1A). The Sdt localization defect in crb clones is, therefore, much weaker than that of Dhc clones. However, Crb protein is missing from the apical surface of sdt clones from the very earliest oogenic stages and overall Crb protein levels are strongly reduced ( Figures 2E and S1B), a phenotype indistinguishable from that of Dhc clones. Together these data indicate that that the mislocalization of Sdt in Dhc clones can potentially account for the loss of Crb.
Is Dynein's role in Crb and Sdt localization specific or does loss of Dynein lead to a general disruption of protein polarization? This question is particularly important because a recent study showed that Dynein helps localize Baz to the apical surface in the cellularizing embryonic blastoderm [5]. In contrast to the strong reductions in Crb and Sdt, Dhc FC clones show only modest reductions of Baz, aPKC and Par6 at the apical surface ( Figures 2F and S2). Because Crb is missing in Dhc mutant cells we investigated whether Baz complex components were similarly reduced in crb 11A22 cells and found a nearly identical phenotype ( Figure S2). These results suggest that the apical reduction of Baz in Dhc clones may be secondary to the loss of Crb. It is likely that Dynein does play some role in the apical targeting of Baz in the FCs, however, as we did observe rare Dhc clones in which Baz and aPKC were relocalized from the apical to the lateral domain, a phenotype never seen in crb clones ( Figure S2). Finally, the apical localization of b H -Spectrin is indistinguishable from wildtype (unpublished data), as is the localization of Discs-large (Dlg) to the lateral domain ( Figure 2G). In sum, these protein localization experiments indicate that Dynein is required for the proper targeting of a subset of apical proteins in the FCs and that it is particularly important for the localization of Sdt.
sdt Genetically Interacts with Dhc64c The molecular epistasis experiments above suggested a clear relationship between Dynein and Sdt in FC apico-basal polarity regulation. Since Sdt is a central regulator of embryonic epithelial polarity, we next wanted to determine whether this association also exists in embryos. However, unlike the FCs, it is not possible to study Dynein under strong loss of function conditions in Drosophila embryos due to its large maternal contribution and essential function in oogenesis [33,37]. We therefore looked for a genetic interaction between sdt and Dhc64C in embryos by asking whether reducing Dynein levels in a hypomorphic sdt mutant background could further impair Sdt activity. sdt XP96 is a strong allele that severely reduces cuticle production, resulting in only small cuticle fragments in almost all embryos ( Figure  3D). By contrast, embryos hemizygous for the hypomorphic allele, sdt P10 , typically have two large cuticle patches ( Figure  3B). We impaired Dynein function in this background by crossing the strong Dhc 4À19 allele to sdt P10 (see Methods for experimental details). Embryos collected from the cross in which the sdt P10 and Dhc 4À19 alleles were both segregating showed reduced cuticle production compared to sdt P10 mutants alone. In 48% of these embryos, the cuticle was reduced to either a single patch ( Figure 3C) or small fragments, similar to sdt XP96 mutants ( Figure 3A). The clear genetic interaction between sdt and Dhc64C indicates that the apical localization of Sdt is likely to depend on Dynein in both FC and embryonic epithelia.

The Apical Localization of sdt mRNA Is Dynein-Dependent and Developmentally Regulated
How does Dynein localize Sdt to the apical surface? Although Dynein could in theory be trafficking Sdt protein directly, this motor can also localize proteins indirectly by transporting the transcripts that encode them [21,23,24]. A single, localized mRNA molecule can generate many proteins in the desired location, making mRNA transport a highly efficient mode of protein localization. Because sdt transcripts have been reported to be apically localized in embryonic epithelia [9], we investigated whether Dynein might localize the Sdt protein through the apical targeting and localized translation of its mRNA.
To determine whether sdt mRNA is apically localized in the FCs, we performed fluorescent in situ hybridization on ovaries. During oogenic stages 2-4, sdt transcripts appear as small puncta that are greatly enriched in the apical FC cytoplasm ( Figure 4A). This apical localization is particularly evident when a sdt transgene (sdt cMAGUK, [9]) is overexpressed ( Figure 4B). Surprisingly, however, sdt transcripts are uniformly distributed in the FC cytoplasm during later oogenic stages ( Figure 4C). These data indicate that sdt mRNA is apically localized in the FCs, but that transcript localization can be developmentally regulated. Interestingly, we also found a similar developmental switch in embryos: sdt transcripts are predominantly apical during the first half of embyogenesis ( Figure 4F), but become randomly localized at later stages ( Figure 4G). That sdt mRNA localization is similarly regulated in both FC and embryonic epithelia suggests that a developmental switch in sdt mRNA localization may correlate with epithelial maturation in Drosophila.
We then took advantage of our ability to make strong loss of functions clones in the FCs to investigate whether Dynein is required for the apical enrichment of sdt transcripts during early oogenic stages. When compared to the ring of apical sdt puncta observed in wild-type FCs ( Figure 4D), sdt puncta are randomly distributed in FC epithelia composed entirely of Dhc mutant cells ( Figure 4E). To quantify this effect, we counted sdt mRNA puncta in each of three compartments: the nucleus, the apical cytoplasm or the basal cytoplasm. Among cytoplasmic puncta in wild-type cells, 74% were apical (n ¼ 8 ECs). In Dhc FCs, however, only 44% of cytoplasmic puncta were apical (n ¼ 7 ECs). The t value for this data set is 13.49 indicating statistical significance ( Figure  S4; Table S1). These data show that Dynein is required to enrich sdt transcripts in the apical FC cytoplasm during the same period when epithelial polarity defects are first seen in Dhc FC clones. In the strong sdt XP96 allele, the epidermis only produces small cuticle fragments (D). When Dhc levels are reduced in sdt P10 mutants, the cuticle in many embryos is reduced to a single patch (C) or strongly reduced to fragments, like sdt XP96 (D). doi:10.1371/journal.pgen.0040008.g003 The sdt Apical Localization Signal Is Located in an Alternatively Spliced Coding Exon To better understand the Dynein-dependence and developmental regulation of sdt mRNA localization, we mapped the cis-acting signal that targets sdt transcripts to the apical cytoplasm using a series of UAS-transgenes ( Figure 5B). The sdt locus produces several mRNA species that arise from a combination of at least three transcriptional start sites and alternative splicing [9,10,38]. We based our initial constructs on the sdt A isoform (also known as Sdt cMAGUK), which contains all 10 coding exons in the final transcript. As expected, transcripts from the sdt A transgene showed dramatic apical localization in the FCs ( Figure 5C). Because the majority of mRNA localization signals reside within the 39UTR, we replaced the sdt 39UTR on the sdt A transgene with the SV40 39UTR. Surprisingly, this manipulation had no effect on the apical transcript localization ( Figure 5D). Similarly, when we appended the sdt 39UTR to an eGFP transgene, eGFP transcripts were uniformly distributed in the FC cytoplasm ( Figure 5F). These results demonstrate that the sdt 39UTR is neither necessary nor sufficient to act as an apical mRNA localization signal.
In the uncommon cases in which mRNA localization signals lie outside the 39UTR, they are often found within the coding region [39][40][41][42][43]. Since the majority of sdt transcripts are apical at early oogenic stages, but unlocalized at later stages, there must be at least one naturally occurring sdt mRNA isoform that lacks the apical signal. The sdt B mRNA isoform is identical to sdt A except that the third exon is spliced from the final transcript, thus removing 433 amino acids that lie between the two ECR domains ( Figure 5A, [10]). Intriguingly, no obvious functional motifs are encoded by exon 3 and this region is absent from vertebrate Sdt homologs ( Figure 5A). Given the uncertainty about exon 39s contribution to Sdt protein function we hypothesized that this exon might contain the apical localization signal for the sdt mRNA.
To test this hypothesis we constructed two additional UAStransgenes and assayed their mRNA localization patterns. The first transgene mimicked the sdt B isoform, in that it contained all sdt coding exons except exon 3. Transcripts from the sdt B transgene are uniformly distributed in the FC cytoplasm ( Figure 5E), indicating that exon 3 is required for the apical localization of sdt mRNA. Conversely, when exon 3 was appended to the eGFP coding sequence in a second transgene, eGFP transcripts showed dramatic apical localization in the FCs, demonstrating that exon 3 is capable of targeting a normally unlocalized transcript to the apical cytoplasm ( Figure 5G-5H). Analogous experiments in embryonic salivary glands gave identical results ( Figure S5). These data indicate that exon 3 is both necessary and sufficient for the apical localization of the sdt mRNA.
By mapping the sdt apical signal to exon 3, our data suggest a model in which the subcellular localization of sdt transcripts could be developmentally regulated through pre-mRNA The apical ring of sdt puncta normally observed in wild-type egg chambers (D) disappears when the entire epithelium is mutant for Dhc D12À5 (E). (F) At embryonic stage 11, sdt transcripts are strongly enriched in the apical cytoplasm, but appear to be uniformly distributed by stage 15 (G and G') doi:10.1371/journal.pgen.0040008.g004 splicing. If this model is correct, the unlocalized, endogenous sdt transcripts observed in the FCs at oogenic stage 10 should lack exon 3 and thus correspond to variants of sdt B. We designed two exon-specific in situ probes, one that exclusively recognizes exon 3 and a probe of the same size that covers portions of exons 4-6. The ex4-6 probe is predicted to recognize all sdt mRNA isoforms, whereas the ex3 probe will reveal all sdt isoforms except sdt B variants. Although both probes stain localized transcripts in the oocyte with equal intensity (Figure S5), only the ex4-6 probe stains the unlocalized transcripts in the FCs at stage 10 ( Figure 5I and 5J). These data suggest that sdt B isoforms predominate in the FCs later in oogenesis. Moreover, our results reveal an unusual biological mechanism in which mRNA localization is developmentally regulated through the alternative splicing of a coding exon.

sdt A Is Specifically Required for the Early Formation of Embryonic Epithelia
To investigate the functional significance of sdt transcript localization, we examined the phenotype of sdt EH681 , an allele caused by a premature stop codon within the alternatively spliced exon 3 [10]. In this mutant background proteins translated from apical sdt A transcripts should be severely truncated and non-functional [44], while proteins from unlocalized sdt B transcripts should be unaffected ( Figure  6A). Immunohistochemistry on sdt EH681 FC clones revealed that Sdt and Crb protein levels are reduced at the apical surface of mutant cells ( Figure 6B and 6C). These data indicate that Sdt A is required to attain wild-type levels of Sdt and Crb at the apical FCs surface. Interestingly, sdt EH681 cells were morphologically indistinguishable from wild-type, indicating that sdt mRNA isoforms lacking exon 3 must also be present in the FCs during early oogenic stages. This supposition is consistent with our previous observation that 26% of sdt mRNA puncta are found in the basal cytoplasm during early oogenic stages (Table S1). Together these findings suggest that sdt A contributes to the apical targeting of the Crb complex, but that this function is dispensable for FC polarity under normal conditions.
In contrast to the relatively mild phenotype of sdt EH681 in the FCs, this allele produces a very strong phenotype in embryos ( Figure 6D and 6E, [36,38]). Since RT-PCR experi- ments demonstrated that sdt A and sdt B are both present during early embryogenesis ( Figure 6F, [38]), this suggested that apically localized sdt transcripts play a significant polarizing role in embryonic epithelia. To test this hypothesis, we compared the ability of sdt A and sdt B transgenes to rescue the sdt EH681 embryonic phenotype. For this experiment a maternal Gal4 driver ensured early availability of the exogenous mRNAs and the two transgenes were targeted to the same genomic attP2 site to provide uniformity in transcript levels.
Although both transgenes provide some function, sdt A is much more efficient than sdt B in rescuing the sdt EH681 phenotype ( Figure 6G). Among the embryos collected from the cross in which the sdt EH681 and the sdt A transgene were segregating, 43% showed continuous cuticle over greater than 2/3 of their surface. In contrast, only 13% of embryos expressing sdt B fell into this class, as compared to 8% in the negative control group. These data suggest that sdt A plays a specific and early role in epithelial formation in the embryo.

Dynein, Apical Determinants, and Polarity
Dynein's role in MT-based apical transport has been studied in the epithelia of both mammals and flies [5,28,[45][46][47][48], but an explicit link between Dynein and apico-basal polarity has not been found. One cause for this deficiency may be that Dynein is required for a number of essential cellular processes, making it difficult to study this motor under strong loss-of-function conditions. For instance, previous studies of Dynein function in Drosophila embryos have relied on combinations of hypomorphic Dhc alleles or injection of a-Dhc antibodies which only partially block Dynein function [5,28,33]. Notably, these manipulations have failed to produce epithelial polarity phenotypes. We have shown that strong loss of Dynein function in the FCs disrupts both molecular and morphological aspects of apico-basal polarity, providing the first direct evidence for the role of this motor in this key cell biological process. It is currently unclear why the FCs tolerate strong loss of Dynein function better than other tissues, but this property provides a unique opportunity to begin dissecting the many roles that Dynein is likely to play in epithelial organization.
How does Dynein regulate apico-basal polarity? Since Dynein is a minus-end directed MT motor and minus ends are apically oriented in epithelia, we investigated whether Dynein might ferry components of the Baz and/or the Crb complexes to their sites of action at the apical surface. Two pieces of data indicate that the Baz complex is not the primary Dynein cargo contributing to FC polarity. Although Baz is occasionally relocalized to the lateral FC surface in Dhc clones, most mutant cells display significant amounts of apical Baz. Furthermore, ECs in which the entire epithelium is mutant for baz or aPKC display multilayering at both EC poles ( [49], unpublished data), a phenotype that is distinct from the posterior multilayering in Dhc mutant epithelia. These observations are consistent with recent studies in the embryonic blastoderm, where it was shown that Dynein does play a role in Baz localization, but that it is not the only means by which this protein is targeted to the apical domain [5]. Together, these data indicate that, while Dynein likely does play a role in the apical targeting of Baz in both the embryo and the FCs, Dynein's major contribution to FC apico-basal polarity corresponds to a different cargo.
Our results favor a model in which Dynein works primarily through the Crb complex to influence epithelial polarity. When Dynein function is reduced in the FCs, Crb and Sdt disappear from the apical surface, even in Dhc cells that remain cuboidal. Moreover, the morphology of an egg chamber in which the entire epithelium is mutant for Dhc is quite similar to that seen for crb and sdt mutant egg chambers, as all three display multilayering predominantly in the posterior ( [31], data not shown). A major challenge comes in deciphering which Crb complex components are specifically transported by Dynein. We have focused on the relationship between Dynein and Sdt because genetic interaction and molecular epistasis experiments indicate that loss of apical Sdt can account for many aspects of the Dhc polarity phenotype. Our finding that the apical localization of sdt transcripts is Dynein-dependent suggests a mechanism by which Dynein could localize Sdt, in part, through the apical targeting and localized translation of its mRNA. sdt tran-scripts cannot be the only Crb complex component transported by Dynein, however, as the phenotype of sdt EH681 mutant FCs does not recapitulate all aspects of the polarity phenotype of Dhc mutant clones. Interestingly, crb transcripts are also targeted to the apical FC cytoplasm in a Dyneindependent manner, although at a later stage than sdt ( Figure  S3). This finding raises the possibility that, in addition to Baz, crb mRNA and/or Crb complex proteins represent other Dynein cargoes required for full epithelial polarization. Future work will be required to more finely dissect Dynein's complex contributions to the apical targeting of the Crb complex.

Developmental Regulation of sdt mRNA Localization
While investigating whether sdt mRNA was likely to be a primary Dynein cargo contributing to apico-basal polarity, we made the surprising discovery that the apical targeting of sdt transcripts is regulated through the alternative splicing of a coding exon, exon 3. We are aware of only two other genes in which transcript localization is regulated in this way [50,51]. In both instances, however, the signal lies within the 39UTR, so the splicing event does not affect protein structure. It is curious that alternative splicing of the sdt mRNA localization signal also deletes 433 amino acids from the protein. The role of the amino acids encoded by exon 3 in Sdt function is not yet known. These amino acids are not conserved in vertebrate Sdt homologs. Moreover, Sdt A and Sdt B bind the intracellular domain of Crb with equal efficiency in vitro [10] and our work with the sdt EH681 allele, as well as over-expression studies with the sdt A and sdt B transgenes ( Figure S1) indicate that both protein isoforms stabilize Crb in vivo. Although we have not ruled out the possibility that exon 3 regulates both mRNA localization and protein function, together these observations suggest that the splicing of exon 3 may primarily regulate mRNA localization.
A potential role for Dynein-dependent apical targeting of sdt mRNA in epithelial polarity is supported by our analysis of the relative contributions of the sdt A and sdt B isoforms. Although the lack of Sdt A in sdt EH681 FC clones reduces Crb at the apical membrane, this deficit leads to relatively mild effects on other aspects of polarity. This apparent discrepancy can be explained by the extrinsic cue for apical identity that is provided to the FCs through their direct contact with the germline [30,31], which may compensate for reduced Crb complex function in this tissue. However, in the embryo, where no such cue is available, sdt EH681 has a nearly null phenotype, which is rescued much more efficiently by sdt A than sdt B. Overall, these data suggest that the apical targeting of sdt transcripts may represent an important mechanism contributing to apico-basal polarity.
Interestingly, the apical targeting of sdt mRNA is developmentally regulated in both embryonic and adult epithelia. Specifically, we have shown that apically localized sdt transcripts are found only during early stages, while unlocalized transcripts predominate at later stages. Why is sdt transcript localization regulated in this way? It is tempting to speculate that apical transcripts are required primarily for the establishment of apico-basal polarity but not its maintenance. In reality, however, the functional distinction between the two phases of sdt mRNA localization is almost certainly more subtle. When apical sdt A transcripts are present, the epithelia are relatively immature; they tend to be prolifer-ative, display incomplete junctional structures and have yet to adopt their final cell morphology. By contrast, when unlocalized sdt B transcripts predominate, the epithelia are more likely to be post-mitotic and highly differentiated. These observations raise the possibility that a concentrated pool of Sdt protein, generated by localized translation of apical transcripts, could function to stabilize Crb primarily during periods when cell polarity is labile, but that this dynamic regulatory mechanism would be dispensable in fully differentiated cells.
Genetic interaction between Dhc64C and sdt. Embryos were collected from the three crosses detailed in Figure 6 over the course of 24 hours. The plates were then aged at 25 8C for an additional 24 hours to allow unaffected embryos to hatch and crawl away. Cuticles were prepared from the remaining mutant embryos. Because Dhc 4À19 is larval lethal [33] and the two sdt alleles are embryonic lethal, mutant embryos collected were presumed to be sdt (À/À). Within the cross segregating both the sdt P10 and Dhc 4À19 alleles, the sdt (À/À) embryos contain between 0-2 copies of Dhc 4À19 .
In situ probe synthesis. Antisense RNA Probes were synthesized using a DIG RNA labeling kit (Roche). crb (LD13435) and sdt (RE05272) EST clones were obtained from the Drosophila Genome Resource Center (DGRC). The crb clone was linearized with NotI and transcribed with the T7 polymerase and the sdt clone was linearized with MluI and transcribed with the T3 polymerase. To improve probe penetration, both the crb and sdt antisense RNAs were fragmented with carbonate buffer. Because the RE05272 EST contains the entire sdt B mRNA sequence, sdt probe fragments will anneal along the full length of sdt transcripts with the exception of exon 3. The antisense eGFP probe was synthesized from the pBS-eGFPA vector, by linearizing with KpnI and transcribing with the T7 polymerase. The ex3 and ex4-6 antisense probes were synthesized by direct transcription from PCR products where the reverse primer included a T7 promoter. Primer sequences can be found in Table S2.
In situ hybridization. In situ hybridizations were performed as described in Wilkie (1999) with minor modifications (protocol available upon request). Probes were detected using a horseradish peroxidase conjugated sheep anti-dig antibody (1:1000, Roche) followed by staining with AlexaFluor 488-Tyramide (1:100, Invitrogen). To detect clones in mosaic egg chambers, in situ hybridization was followed by staining with a rabbit anti-GFP antibody and AlexaFluor 555-conjugated secondary antibody (both at 1:200, Invitrogen). The nuclear envelopes and basal lamina were illuminated with AlexaFluor647-conjugated wheat germ agglutinin (1:200, Invitrogen). Images were obtained using a Leica TCS SL confocal microscope and processed with Adobe Illustrator and Photoshop.
Construction of transgenic animals. The six transgenes used to map the sdt localization signal were cloned into the pUASBP vector (a generous gift from Barret Pfeiffer, Rubin Lab unpublished), which contains the BamHI cassette from pUAST [56] as well as the attB element for site-specific recombination using phiC31 integrase [57]. A diagram of the vector and a complete list of primers can be found in Figure S3 and Table S2. PfuUltra High-Fidelity DNA Polymerase (Stratagene) was used to amplify eGFP, the sdt 39UTR, and sdt exon3. All other amplifications used the Phusion High-Fidelity DNA Polymerase (Finnzymes). To create the constructs with the eGFP coding sequence, eGFP was amplified from the pBS-eGFPA vector. Directional cloning of this amplicon into pUASBP using 59 NotI and 39 XhoI yielded the eGFP-SV40 construct. eGFP-sdt39UTR was synthesized in two steps. First, the sdt 39UTR was amplified from sdt cMAGUK [9]. This amplicon along with pUASBP were digested with XbaI and FseI, gel extracted, and ligated together. Second, eGFP was extricated from the eGFP-SV40 construct using NotI and XhoI and subsequently ligated into pUASB containing the sdt 39UTR. For the eGFP-sdtexon3-SV40 construct, exon 3 was amplified from sdt cMAGUK and cloned 59 XhoI and 39 KpnI into the eGFP-SV40 plasmid. To create the sdtA-SV40 transgene, the sdt 59UTR and coding region were amplified from sdt cMAGUK and cloned 59 NotI 39 XhoI into pUASBP. The sdt A construct was then created by cutting the 59UTR and coding sequence from sdtA-SV40 using NotI/XhoI and ligating it into pUASBP containing the sdt 39UTR. The sdtB transgene was constructed by amplifying the sdt 59UTR and coding sequence from the RE05252 EST (DGRC) and cloning 59 NotI and 39 XhoI into pUASBP containing the sdt 39UTR. Transgenic animals were produced by Genetic Services, who injected DNA for each construct into the attP2 line [57] and isolated transformants.
Reverse transcriptase PCR. Total RNA was prepped from embryos using TRIzol Reagent (Invitrogen) and first strand cDNA synthesis was performed using the Omniscript reverse transcription kit (Qiagen). Primer sequences can be found in Table S2.
Rescue of the sdt EH681 mutant phenotype. Rescue of the sdt EH681 embryonic cuticle phenotype was performed using the maternal PfGAL4-nos.NGTg40 driver [58] and the UAS-sdt A and UAS-sdt B transgenes. The UAS-eGFP-sdt 39UTR transgene was used as a negative control. Embryos were collected as described for the genetic interaction experiment above.
Addendum. A complementary study examining the contributions of Dynein and crumbs mRNA localization to apico-basal polarity regulation can be found in the following reference: Li Z, Wang L, Hays TS, Cai Y (2007) Dynein-mediated apical localization of crumbs transcripts is required for epithelial polarity. J Cell Biol. In press. Supporting Information Figure S1. Sdt Stabilizes Crb in the Apical FC Membrane (A) When the entire FC epithelium is mutant for crb, Sdt still localizes to the apical surface and junctional region during early oogenic stages, indicating that Sdt is only partially dependent on Crb for its localization in the FCs. This apical localization is gradually lost, however, until Sdt is entirely cytoplasmic by stage 8 (not shown). The bracket denotes Sdt staining in the somatic inner-cyst region. (B) By contrast, when the entire FC epithelium is mutant for sdt, Crb is never seen at the apical membrane, indicating that Crb absolutely requires Sdt for localization.

Accession Numbers
Information on sdt alleles and sequences for sdt mRNA isoforms can be found on Flybase (http://flybase.org/). In this database, sdt-A is called sdt-RB, and sdt-B is called sdt-RF.