The Ski2-family helicase Obelus regulates Crumbs alternative splicing and cell polarity

Alternative splicing increases protein diversity by generating multiple transcripts and protein isoforms from a single gene. Over 90% of intron-containing genes in humans are alternatively spliced, but the functional consequences of the vast majority of alternative splicing events are unknown (Wang et al., 2008; Brown et al., 2014). Alternative splicing plays an important role in a wide range of biological processes, including sex determination, axon guidance, neuronal arborization, immunity, and muscle differentiation, and defects in alternative splicing have been implicated in human disease (Faustino and Cooper, 2003; Tazi et al., 2009; Kalsotra and Cooper, 2011). Different isoforms of a single gene can perform distinct functions (Kelemen et al., 2013) and are often expressed in distinct temporal and spatial patterns (Graveley et al., 2011; Spencer et al., 2011; Brown et al., 2014). Splice-site selection in pre-mRNA transcripts is regulated by RNA-binding proteins that inhibit or promote the assembly of the spliceosome complex at particular splice sites (Fu and Ares, 2014). The molecular composition of the spliceosome itself can also vary between tissues, as spliceosomal proteins are expressed in different tissues and at different times during development, indicating that the spliceosome machinery is subject to extensive regulation (Park et al., 2004; Grosso et al., 2008; Celniker et al., 2009). Despite the pervasive presence and complexity of alternative splicing in multicellular organisms, the functions and molecular targets of most splicing regulators are currently unknown.

Epithelial cells are essential determinants of tissue structure, and defects in epithelial polarity and adhesion lead to developmental diseases and cancer. Cells in epithelia are connected by adherens junctions, which mediate cell interactions, maintain the structural integrity of the tissue, and enable dynamic tissue remodeling. The size, distribution, and dynamics of adherens junction complexes are regulated by several mechanisms, including interactions with the actin and microtubule cytoskeleton (Brieher and Yap, 2013), trafficking of junctional complexes to and from the plasma membrane (Wirtz-Peitz and Zallen, 2009), and spatial cues provided by the apical–basal polarity machinery (Harris and Tepass, 2010). Junctional proteins are distributed in a characteristic clustered fashion at the membrane, ranging from discrete puncta at nascent adherens junctions to a more continuous distribution at the mature zonula adherens (Tepass and Hartenstein, 1994; Adams et al., 1998; Cavey et al., 2008; McGill et al., 2009; Truong Quang et al., 2013). These clusters have been proposed to be important for the strength and dynamics of cell adhesion by increasing the avidity of weak intermolecular interactions and enabling rapid junctional remodeling during development (Yap et al., 2015). However, the mechanisms by which adherens junction organization is modulated by changes in the expression and activity of junctional regulators are not well understood.

Proteins involved in apical–basal polarity play an essential role in regulating adherens junction localization. Antagonistic interactions between apical and basolateral proteins direct the positioning of adherens junctions at the boundary between apical and basolateral domains (Bilder et al., 2003; Tanentzapf and Tepass, 2003; Blankenship et al., 2007). A transient expansion of the apical epithelial domain displaces adherens junctions basolaterally during tissue folding (Wang et al., 2012), and the apical domain is stably expanded in mature cells such as photoreceptors, which produce an elaborate apical light-sensing rhabdomere (Pellikka et al., 2002; Walther and Pichaud, 2010). The EGF repeat transmembrane protein Crumbs is an essential determinant of the apical domain and plays an important role in adherens junction organization in many tissues. Crumbs is required to maintain epithelial apical–basal polarity during development and to prevent light-induced degeneration of photoreceptor cells in the adult (Bulgakova and Knust, 2009; Tepass, 2012; Thompson et al., 2013). Crumbs and its associated proteins, Par-6, atypical PKC (aPKC), and Stardust/Pals1, act to exclude the junctional stabilizing protein Par-3 from the apical cortex, restricting Par-3 and adherens junctions to the adjacent junctional domain (Krahn et al., 2010; Morais de Sá et al., 2010; Walther and Pichaud, 2010). Overexpression of Crumbs leads to an expansion of the apical domain, raising the possibility that the level of Crumbs activity could provide a mechanism that modulates apical domain size (Wodarz et al., 1995; Grawe et al., 1996; Pellikka et al., 2002; Letizia et al., 2013). The crumbs pre-mRNA undergoes alternative splicing to produce three experimentally verified isoforms, each of which has a unique alteration in the extracellular domain (Brown et al., 2014). However, the factors that regulate the generation of different Crumbs isoforms, and how they influence epithelial polarity and junctional organization, are unknown.

Here, we report the identification of Obelus, a conserved Ski2-family DExD/H-box helicase isolated in a screen for Drosophila melanogaster mutants with defects in epithelial polarity and junctional organization. Embryos mutant for obelus display an expanded apical epithelial domain and aberrant aggregation of adherens junction proteins at the cell membrane, which are associated with a disruption of epithelial remodeling. Through whole-genome transcriptome analysis, we found that Obelus regulates the alternative splicing of the crumbs pre-mRNA. In obelus mutants, use of an alternative exon results in a switch in isoform expression of the crumbs pre-mRNA, resulting in increased expression of an isoform that contains an additional EGF-like repeat in the extracellular domain. Overexpression of this alternative Crumbs isoform is sufficient to recapitulate the centrosome mispositioning and junctional aggregation defects of obelus mutants. These results demonstrate that regulation of Crumbs alternative splicing by the Ski2-family helicase Obelus modulates cell polarity and adherens junction organization during development.

To identify genes that regulate cell polarity and adhesion in epithelia, we screened for Drosophila mutants that have defects in body axis elongation, a conserved process of epithelial remodeling. We identified a spontaneous mutation in the obelus (obe) gene in the background of several pBac mutant lines (Schuldiner et al., 2008). The embryonic progeny of homozygous mutant obe females crossed to heterozygous males (referred to as obe mutants) displayed significantly reduced axis elongation compared with wild type (Fig. 1, C–E; and Fig. S1 A). The obe¹ mutation failed to complement two deficiencies and a second mutation, obe², in the 88F6–88F7 interval on chromosome 3R (Fig. 1 D and Fig. S1 A). These mutations were all viable in trans but were male sterile and produced substantial maternal effect embryonic lethality (Fig. S1 A). These results indicate that obe¹ and obe² are alleles of a single gene that is required for epithelial remodeling and embryonic viability.

To determine whether these axis elongation defects are caused by a disruption of cell adhesion, we analyzed adherens junction localization in obe mutant embryos. During axis elongation in wild type, adherens junctions transition from discrete puncta to a more continuous distribution (Tepass and Hartenstein, 1994; Cavey et al., 2008; Truong Quang et al., 2013) and concentrate at dorsal and ventral cell borders in a planar polarized fashion (Zallen and Wieschaus, 2004; Blankenship et al., 2006). In obe mutants, the core adherens junction proteins E-cadherin and β-catenin and the junctional regulator Par-3 were enriched at dorsal and ventral cell borders as in wild type, indicating that planar polarity is correctly established (Fig. 1, A, B, and H–K). In contrast, adherens junctions showed an atypical clustering defect in a majority of obe mutant embryos. Par-3 was present in single, large aggregates at cell interfaces (Fig. 1, A and B). These aggregates colocalized with E-cadherin and β-catenin and failed to resolve into a continuous distribution during axis elongation (Fig. 1, H–K). Similar defects were observed in multiple obe alleles (Fig. S1, B–D). Junctional aggregates in obe mutants were more pronounced during axis elongation (stages 6–8), and a more uniform junctional distribution was restored at later stages (Fig. S1, E and F). These results demonstrate that Obelus is required for adherens junction localization and body axis elongation.

We observed several phenotypes in obe mutants that are commonly associated with defects in microtubule or centrosomal organization, including uncoordinated movement in adults, male sterility, and immobile flagella in sperm motility assays (unpublished data). We therefore tested whether obe mutants had defects in centrosome localization. Epithelial formation in the Drosophila embryo coincides with the reorganization of microtubules from a radial array focused at centrosomes to a noncentrosomal cortical array aligned with the apical–basal axis (Warn and Warn, 1986). In stage 5, centrosomes are located near the apical cortex in close proximity to cortical microtubules, consistent with their role as microtubule-organizing centers (Fig. 2 A). In stage 6, centrosomes dissociate from cortical microtubules and move away from the cortex to occupy a more basal and centrally located position within the cell (Fig. 2, C and G–I). In contrast, although a majority of wild-type centrosomes had moved away from the apical cortex by stage 7 (80 ± 5%), a smaller fraction of centrosomes had successfully moved away from the apical cortex by this stage in obe mutants (17 ± 3%; P < 0.0001; Fig. 2, B and D–F). Most centrosomes failed to dissociate from microtubules in obe mutants and instead remained in close proximity to adherens junctions throughout axis elongation in stages 6 and 7 (Fig. 2, J–L). Mispositioned centrosomes were often associated with adherens junction aggregates, such that two centrosomes flanked a single aggregate (an arrangement that resembles the division symbol, also known as an obelus). These results show that obe mutants have defects in centrosome positioning and adherens junction localization, indicating that Obelus is required for multiple aspects of cell polarity.

Centrosome positioning and adherens junction organization both require the microtubule cytoskeleton (Harris and Peifer, 2007; Siegrist and Doe, 2007). We therefore hypothesized that the defects in obe mutants could result in part from disrupted microtubule localization or activity. To test this possibility, we treated permeabilized obe mutant embryos with a low concentration of the microtubule-depolymerizing drug nocodazole, which inhibits microtubule polymerization without substantially disrupting overall microtubule levels (Liao et al., 1995). A brief (7 min) incubation of wild-type embryos in nocodazole caused a slight reduction in the level of cytosolic microtubules associated with centrosomes (Fig. 3, A and B). This treatment did not affect centrosome positioning in wild type, but did cause a reduction of cortical Par-3, consistent with the role of dynamic microtubules in cell adhesion. Notably, nocodazole treatment fully rescued the centrosome positioning defects in obe mutants (Fig. 3, C–E). In addition, nocodazole treatment completely suppressed the Par-3 aggregation defects in obe mutants, eliminating Par-3 puncta and restoring a more uniform distribution of Par-3 at the dorsal and ventral cell borders, similar to wild type (Fig. 3, D and E). These results demonstrate that the adherens junction aggregation and centrosome positioning defects in obe mutants are microtubule dependent.

The aberrant association between adherens junctions and centrosomes in obe mutants and the restoration of wild-type junctional localization and centrosome positioning by nocodazole treatment suggest that these defects may be caused by aberrant microtubule-nucleating activity at centrosomes. To test this possibility, we analyzed the localization of γ-tubulin, which is required for the microtubule-nucleating activity of centrosomes (Stearns and Kirschner, 1994). In wild-type embryos, γ-tubulin is rapidly down-regulated at centrosomes in stage 6, indicating that centrosomes lose their microtubule-nucleating activity as they move away from the cortex (Fig. S2, A and C). In obe mutants, the down-regulation of centrosomal γ-tubulin occurred at the same stage and to the same extent as in wild type, even though centrosomes failed to move away from the cortex (Fig. S2, B and C). These results indicate that the defects in obe mutants are not caused by increased microtubule nucleating activity at centrosomes, but may instead result from an aberrant association between microtubules and junctional proteins at the cell cortex.

The adherens junction defects in obe mutants are reminiscent of embryos that lack Par-6 or aPKC, which are associated with the Crumbs apical protein (Blankenship et al., 2006; Harris and Peifer, 2007). To test whether the defects in obe mutants are caused by a reduction in the activity of apical complex proteins, we analyzed Crumbs and aPKC localization and activity. Autophosphorylation of aPKC at T567/T574 (corresponding to T555/563 in the human protein), an indicator of kinase activity (Hirai and Chida, 2003), was unchanged in obe mutants (Fig. S3 C). In addition, Par-3 phosphorylation at serine 980 (Par-3 phospho-S980), an aPKC target site (Morais-de-Sá et al., 2010), occurred normally (Fig. S3, A and B). Moreover, unlike aPKC mutants, in which Crumbs fails to localize to the apical membrane (Harris and Peifer, 2007), Crumbs and aPKC were readily detected at the apical domain in obe mutants (Fig. 4, C and D). Similarly, although embryos lacking Crumbs have reduced levels of Par-3 (Morais-de-Sá et al., 2010), Par-3 levels were not affected in obelus mutants (Fig. S3 D). These results suggest that the defects in obe mutants are not caused by a strong reduction in the activity of apical proteins.

To investigate whether Obelus affects a distinct aspect of apical domain organization or activity, we analyzed the localization of Crumbs and aPKC in obe mutants. Crumbs and aPKC are normally localized just apical to and partially overlapping with the adherens junctions in wild type (Fig. 4, A and B; Harris and Peifer, 2005). In contrast, the Crumbs and aPKC domains were expanded to more than twice their usual height along the apical–basal axis in obe mutants (Fig. 4, D–H). Together, these results indicate that obe mutants do not have a strong loss of aPKC activity or cortical localization. Instead, obe mutants display an expansion of the apical epithelial domain.

To identify the gene that is responsible for the defects in obe mutants, we mapped obe to the 88F6–88F7 region on chromosome 3R (Fig. 5 A). The axis elongation, male sterility, and Par-3 aggregation defects in the progeny of obe¹, obe¹/obe², and obe²/Df females were fully rescued by the CH322-15H13 bacterial artificial chromosome (BAC), which covers a 21.6-kb region containing five genes (Fig. 5, A and B; and not depicted). Whole-genome DNA sequencing revealed a 19-bp deletion in the second exon and a point mutation at nucleotide 173 of the CG5205 gene in obe² mutants. Together, these changes are predicted to cause a frameshift after amino acid 48 and introduce a stop codon at amino acid 52 of the predicted 2,183–amino acid protein (Fig. 5 C). The CG5205 transcript was expressed at 2% of its wild-type levels in the progeny of obe²/Df females by quantitative RT-PCR (qRT-PCR) and is unlikely to encode functional protein. We did not identify mutations in exons or splice sites of CG5205 in obe¹, but the full-length CG5205 transcript was expressed at 10% and 7% of wild-type levels in the progeny of obe¹ and obe¹/obe² females, respectively, by qRT-PCR. Moreover, expression of the CG5205 cDNA from a ubiquitous promoter rescued the axis elongation, male sterility, Par-3 aggregation, and centrosome positioning defects of obe¹ and obe¹/obe² mutants (Fig. 5, B and D–F; and not depicted). These results demonstrate that obe¹ and obe² are strong loss-of-function and null mutations, respectively, in the CG5205 gene.

The obelus/CG5205 gene encodes a Ski2-like DExD/H-box helicase, a member of a subclass of superfamily 2 helicases with two tandem helicase domains and two Sec63 domains (Fig. 5, G and H; Johnson and Jackson, 2013). Obelus is highly conserved from yeast to humans, with 55% amino acid identity to human ASCC3 and 45% identity to human SNRNP200. The yeast paralogue Brr2 and the human paralogue SNRNP200 are components of the spliceosome, a large ribonucleoprotein complex that catalyzes the splicing of pre-mRNA transcripts (Lauber et al., 1996; Noble and Guthrie, 1996; Xu et al., 1996). Studies in Saccharomyces cerevisiae and human cell lines demonstrate that Brr2 and SNRNP200 promote the ATP-dependent unwinding of RNA duplexes in the spliceosome (Laggerbauer et al., 1998; Raghunathan and Guthrie, 1998; Kim and Rossi, 1999). Although the structure and biochemical activity of Brr2 are well characterized (Zhang et al., 2009; Cordin and Beggs, 2013; Johnson and Jackson, 2013; Mozaffari-Jovin et al., 2013), the physiological functions of this helicase family have not been previously analyzed in a multicellular organism.

Based on its homology to the Brr2 and SNRNP200 RNA helicases, we speculated that obelus may be required for pre-mRNA splicing or mRNA stability. To test this, we used RNA sequencing to compare the transcriptional and splicing profiles of staged wild-type embryos with obe¹/obe² and obe²/Df mutants. We focused on transcripts that were differentially expressed or spliced compared with wild type and were equally or more severely affected in the stronger obe²/Df mutant. False positives that could be explained by variations in developmental staging were excluded (see Materials and methods section RNA sequencing). By these criteria, CG5205 was the only transcript that was significantly down-regulated in both obe alleles, and five transcripts were significantly up-regulated compared with wild type (Table S1). In addition, 14 transcripts were differentially spliced in both obe alleles (Fig. S4). For three genes, crumbs, CG4449, and CG32581, splicing changes in obe mutants resulted in alterations to the coding region. These results indicate that obe is required for the alternative splicing of a small number of transcripts in the early Drosophila embryo.

As Crumbs is required for apical–basal polarity and adherens junction organization (Bulgakova and Knust, 2009; Tepass, 2012; Thompson et al., 2013), we focused on Crumbs as a potential target that could mediate the effects of Obelus on epithelial polarity. Four isoforms of Crumbs have been reported in Drosophila (dos Santos et al., 2015). Three isoforms (crumbs-RA, crumbs-RB, and crumbs-RC) have been verified by RNA sequencing (Celniker et al., 2009), but only the crumbs-RA isoform has been functionally characterized. The predominant isoform in the early embryo, crumbs-RA, encodes a transmembrane protein with a large extracellular domain composed of 28 EGF-like repeats and 3 laminin–AG domains (Fig. 6 A; Tepass et al., 1990). The crumbs-RC isoform is generated by an alternative splicing event that leads to the inclusion of exon 4, which encodes an additional EGF-like repeat after repeat 7 in the Crumbs extracellular domain (Fig. 6, A and B). Total crumbs mRNA levels in obe mutants were similar to wild type (Fig. 6, C and D). However, the crumbs-RC isoform was dramatically up-regulated at the expense of crumbs-RA in obe mutants (Fig. 6, C and E). The crumbs-RC isoform comprised <20% of total crumbs transcripts in wild type, but was >60% of crumbs transcripts in obe¹/obe² embryos and >80% of crumbs transcripts in obe²/Df embryos (Fig. 6, C and E). These results demonstrate that Obelus is required to generate the crumbs-RA isoform that removes exon 4 of the crumbs transcript.

Crumbs overexpression causes an expansion of the apical domain and a basolateral displacement of adherens junctions in Drosophila embryonic and pupal epithelia (Wodarz et al., 1995; Grawe et al., 1996; Pellikka et al., 2002). To test whether the defects in obe mutants are caused by increased Crumbs activity, we analyzed the effects of overexpressing Crumbs-PA and Crumbs-PC on apical–basal polarity, junctional organization, and centrosome positioning in the early embryo. Crumbs-PA and Crumbs-PC were expressed at similar levels and localized to the apical membrane when overexpressed at moderate levels, indicating that these two isoforms do not differ in overall stability or localization (Fig. S5, A and B). Moreover, both Crumbs-PA and Crumbs-PC retain apical-promoting functions, as overexpressing either isoform at high levels produced an expansion of the apical domain and a basolateral displacement of Par-3 (Fig. S5 B).

Overexpression of the two Crumbs isoforms at moderate levels revealed that these isoforms had distinct effects on centrosome positioning and Par-3 localization. More than half of Crumbs-PC–expressing embryos had centrosomes that were apically mislocalized at stage 7 (63 ± 5%, compared with 2 ± 1% in wild type; P < 0.0001; Fig. 7, A, C, and F), and more than one third of centrosomes were in close proximity to adherens junctions (38 ± 5%, compared with 1 ± 0.5% in wild type; P < 0.0001; Fig. 7 G). In contrast, centrosome localization was only weakly defective in embryos overexpressing Crumbs-PA (10 ± 3% of Crumbs-PA–expressing embryos had apically mislocalized centrosomes, and 6 ± 2% had centrosomes near adherens junctions; Fig. 7, B, F, and G). In contrast, centrosome positioning occurred normally in crumbs mutants (Fig. 7 D). These observations, in addition to the expanded apical epithelial domain in obe mutants (Fig. 4, D–H), are consistent with the idea that obe mutants have increased Crumbs activity.

In addition to distinct effects on centrosome positioning, Crumbs-PA and Crumbs-PC had different effects on Par-3 localization. Moderate overexpression of Crumbs-PC, but not Crumbs-PA, produced a significant proportion of embryos with single, large Par-3 aggregates that resembled obe mutants (Fig. 7, C and H). These large Par-3 aggregates were rarely observed in crumbs mutants or in embryos overexpressing Crumbs-PA (Fig. 7, B, D, and H). In addition, overexpression of either Crumbs-PA or Crumbs-PC produced a fragmented distribution of Par-3 into many small aggregates at the cell cortex in a dispersed pattern, similar to the defects in crumbs mutants (Fig 7, D and I), consistent with the idea that these two isoforms possess partially overlapping activities. Collectively, these results demonstrate that although both isoforms of Crumbs retain some common functions, overexpression of Crumbs-PC, but not Crumbs-PA, causes centrosome positioning and junctional aggregation defects that closely resemble obe mutants.

Alternative splicing events can have profound consequences for protein function, but few factors have been identified that regulate critical alternative splicing events in vivo. Here, we identify a functional target of the conserved Ski2-family helicase Obelus and show that Obelus is required for junctional organization and centrosome positioning in the Drosophila embryo. The loss of Obelus is associated with a failure to exclude exon 4 from the crumbs transcript, resulting in a switch in isoform expression from the Crumbs-PA isoform to the Crumbs-PC isoform that contains an additional EGF-like repeat in the extracellular domain. The apical epithelial domain is expanded in obe mutants, and the junctional localization and centrosome positioning defects in obe mutants are recapitulated by overexpression of the Crumbs-PC but not the Crumbs-PA isoform. These results suggest that the conserved Obelus helicase regulates the activity of the Crumbs apical determinant by inhibiting the expression of a potentially more active form of the protein, providing a mechanism for tuning the size of the epithelial apical domain during development.

We propose that increased or altered activity of the Crumbs-PC isoform triggers a series of events that alter centrosome positioning and junctional organization in obe mutants. Crumbs regulates both junctional position along the apical–basal axis and the extent of junctional clustering, and either a strong loss or gain of Crumbs activity disrupts adherens junction organization (Wodarz et al., 1995; Tepass, 1996). Here, we provide evidence that a more subtle modulation of Crumbs activity through the inclusion of an additional EGF repeat in its extracellular domain provides an endogenous mechanism that modulates the effect of Crumbs on junctional regulation. In one model, increased activity of Crumbs-PC could result in excessive exclusion of Par-3 from the cortex. As Crumbs and its associated apical proteins displace Par-3 and adherens junctions from the cortex (Krahn et al., 2010; Morais-de-Sá et al., 2010; Walther and Pichaud, 2010), the apical expansion in obe mutants could enhance Par-3 exclusion from the apical cortex and corral the remaining Par-3 protein into a subset of its normal territory (Fig. 7 J). Alternatively, the two isoforms could have distinct functions, for example, if Crumbs-PA inhibits junctional clustering and Crumbs-PC promotes clustering. In both models, increased clustering is predicted to enhance Par-3 oligomerization, which exposes the Par-3 N-terminal domain that binds to and bundles microtubules (Chen et al., 2013). The association between Par-3 and microtubules could drive further junctional aggregation in obe mutants, as microtubules enhance adherens junction clustering (Brieher and Yap, 2013). Microtubules could in turn capture and translocate centrosomes to the cortex through the activity of the junction-associated minus end–directed motor dynein (Ligon et al., 2001; Schmoranzer et al., 2009; Kotak et al., 2012). Thus, increased or altered Crumbs activity in obe mutants could trigger a cascade of events that lead to Par-3 aggregation and centrosome mispositioning, ultimately disrupting epithelial polarity and junctional remodeling. Microtubule-dependent positive feedback between Par-3 and centrosomes may be important for many processes within cells, including the establishment of apical–basal polarity (Feldman and Priess, 2012; Jiang et al., 2015) and oriented cell division in the stem cell niche (Inaba et al., 2015).

In contrast to Brr2, which regulates the constitutive splicing of most intron-containing genes in S. cerevisiae (Pleiss et al., 2007), Obelus is specifically required for the alternative splicing of a relatively small number of transcripts in the early Drosophila embryo. This specificity may be the result of differences in the functions of Obelus and Brr2, or an expansion of the mechanisms that regulate alternative splicing in Drosophila compared with yeast. Alternative splicing is more commonly used in multicellular organisms, and studies in zebrafish and in human and Drosophila cell lines have shown that depleting conserved spliceosome complex proteins, including Brr2, selectively affects only a subset of alternative splicing events (Park et al., 2004; Rösel et al., 2011; Brooks et al., 2015; Wickramasinghe et al., 2015). Here, we combine genome-wide transcriptome analysis with molecular genetic approaches to identify the functional target of a conserved Ski2-family helicase in a multicellular organism in vivo. Collectively, these findings demonstrate that proteins with homology to core components of the splicing machinery in yeast can perform highly tissue- and transcript-specific functions in multicellular organisms. Although the biochemical activities of Obelus and its orthologues have not been extensively characterized, the S. cerevisiae Obelus paralogue Brr2 is an RNA helicase that promotes RNA duplex unwinding in the spliceosome, leading to conformational changes that are important for pre-mRNA splicing (Cordin and Beggs, 2013; Johnson and Jackson, 2013). The human Obelus paralogue SNRNP200 has also been shown to function as an RNA helicase, and mutations that disrupt its helicase activity are associated with retinal degeneration (Zhao et al., 2009; Santos et al., 2012). These results suggest that Obelus may function as an RNA helicase, although whether Obelus regulates crumbs splicing directly or indirectly by controlling other RNA processing or unwinding events that are important for epithelial polarity remains to be determined. In addition, the inability of Crumbs-PC overexpression to fully recapitulate the defects in obe mutants could indicate the presence of other RNA targets or other functions of Obelus.

Mutations in the human paralogues of Crumbs and Obelus, CRB1 and SNRNP200, are associated with retinitis pigmentosa, a disease that causes progressive degeneration of retinal photoreceptor cells in humans, leading to blindness (den Hollander et al., 1999; Zhao et al., 2009; Daiger et al., 2013). The identification of Crumbs as a functional target of the Obelus helicase suggests a functional link between these proteins in regulating epithelial organization and integrity. Although previous work on Crumbs has focused on the role of its cytoplasmic domain, our results indicate an important role for the extracellular domain of Crumbs in epithelial polarization and remodeling. In obelus mutants, early embryos express increased levels of the Crumbs-PC isoform that contains an additional EGF-like repeat in the extracellular domain. How these two Crumbs isoforms differ mechanistically remains to be resolved. EGF repeats are present in many proteins involved in cell signaling, polarity, and adhesion, including Notch, Delta, Celsr/Flamingo, and several cadherins. Individual EGF repeats can have profound effects on cell adhesion (Balzar et al., 2001) and ligand–receptor interactions (Rebay et al., 1991) and can influence protein localization and activity by targeting proteins for glycosylation (Acar et al., 2008). The inclusion of an additional EGF repeat in the Crumbs extracellular domain could alter its ability to mediate interactions between cells, as Crumbs homologues promote homophilic adhesion in culture (Zou et al., 2012; Letizia et al., 2013) and the Crumbs extracellular domain is required for its effects on cell polarity, sorting, and survival in vivo (Fletcher et al., 2012; Hafezi et al., 2012; Röper, 2012; Letizia et al., 2013). Crumbs isoforms that differ in the number of EGF repeats have been reported in mice (Mehalow et al., 2003), zebrafish (Zou et al., 2012), and humans (Pruitt et al., 2014), and mutations located within the EGF repeats of the human CRB1 protein are associated with retinitis pigmentosa (den Hollander et al., 1999; Bujakowska et al., 2012). An important goal for further studies will be to determine how inputs that control the structure or activity of the Crumbs extracellular domain influence its essential roles in epithelial organization during normal development and human disease.

Oregon-R was the wild-type control unless otherwise specified. Embryos were raised and scored at 25°C. The obe¹ mutation was isolated in the background of a subset of piggyBac insertions on 3R (Schuldiner et al., 2008), and the obe² mutation was identified as linked to the d04274 P-element insertion (Thibault et al., 2004). Precise excision of the transposable elements did not revert the mutant phenotype, which was mapped by deficiency mapping to a different location on chromosome 3R. Both obe¹ and obe² failed to complement Df(3R)Exel6174 (Parks et al., 2004) and Df(3R)BSC515 (Bloomington Drosophila Stock Center; Cook et al., 2012). In all figures except Fig. S1, obe¹/obe² mutants and obe²/Df mutant embryos were the progeny of obe¹/obe² and obe²/Df(3R)Exel6174 females, respectively, crossed to obe²/+ males. The obe¹/obe¹ and obe¹/Df mutant embryos were the progeny of obe¹/obe¹ and obe¹/Df(3R)Exel6174 females, respectively, crossed to obe¹/+ males.

Embryos overexpressing Crumbs-PA or Crumbs-PC at moderate levels were the F2 progeny of UASp–crumbs-PA or UASp–crumbs-PC males × matαtub15 Gal4 females. Embryos overexpressing Crumbs-PA or Crumbs-PC at high levels were the F2 progeny of UASp–crumbs-PA or UASp–crumbs-PC males × matαtub67;15 Gal4 females (gifts of D. St. Johnston, University of Cambridge, Cambridge, England, UK). Untagged crumbs transgenes were used in all experiments except Fig. S5 A, in which GFP-tagged crumbs transgenes were used. The crb^11A22 (Tepass et al., 1990) germline clones were generated with the FLP recombinase-dominant female sterile system using FRT82B ovo^D2 (Chou and Perrimon, 1996).

To generate crumbs mutant embryos, larvae of the genotype hs-flp/+;FRT82Bcrb^11A22/FRT82B ovo^D2 were heat shocked and crossed to FRT82Bcrb^11A22/TM3, twi-Gal4, upstream activator sequence (UAS)–GFP males. Homozygous mutant embryos were identified by the absence of GFP from twi-Gal4, UAS-GFP balancers in embryos sorted before fixation on a fluorescence stereomicroscope (MZFLII; Leica). All GFP-negative embryos were maternally and zygotically mutant for crb^11A22 (referred to as crb^m/z or crumbs^m/z in the figures).

Time-lapse confocal imaging was performed with Resille:GFP (a gift of A. Debec, Institut Jacques Monod, Paris, France). Embryos were dechorionated for 2 min in 50% bleach, washed in water, mounted on an oxygen-permeable membrane (YSI Incorporated) with halocarbon oil 27 (Sigma-Aldrich), and imaged on a spinning-disk confocal microscope (UltraView RS5; PerkinElmer) with a Plan Neofluor 40×, 1.3 NA objective (Carl Zeiss). Z stacks were acquired at 1-μm steps at 15-s intervals. Maximum intensity projections of 2–3 μm in the apical junctional domain were analyzed. Germband length was measured as the long axis of an ellipse fit to the group of cells tracked using custom software in MATLAB (MathWorks; Simões et al., 2010) and was normalized to the value at the onset of elongation at t = 0 in early stage 7. The p-values were calculated using the F-test followed by the appropriate t test (using the t = 30-min value as the test statistic).

For bright-field imaging, embryos were analyzed 30 min and 3 h after the cellular blastoderm stage under halocarbon oil 27 without dechorionation. Weakly defective obe mutants had reduced elongation at 30 min but were indistinguishable from wild type after 3 h. The germband reached 30–40% egg length by 30 min in moderately defective obe mutants and <20% egg length by 30 min in strongly defective obe mutants and did not recover by 3 h in either case.

To generate sqh-CG5205, the full-length (6,970 nt) CG5205 coding sequence was amplified from the CH322-15H13 BAC (BACPAC Resources; Venken et al., 2009) using the following primers: CG5205F, 5′-CAACATGTGGGAGCCGCCACGATTG-3′; and CG5205R, 5′-TTAGACTTCTTTTGCATCAATG-3′. The PCR product was cloned into the pENTR/D TOPO vector (Invitrogen), recombined into the pSqh-GFP-W-attB destination vector (a gift of F. Wirtz-Peitz, Harvard Medical School, Boston, MA) using the Gateway system (Invitrogen), and inserted into attP40 on chromosome 2 (Genetic Services). For pEntr-crumbs-PA and pEntr-crumbs-PC, the 5′ portion of the gene containing exons 1–3 of the crumbs gene was amplified from UASt-crumbs^WT (a gift of D. ter Stal and U. Tepass, University of Toronto, Toronto, Canada) using the primers 5′-CCGCGGCCGCCCCCTTCACCATGGCTAAAATCGCCAATGC-3′ (both), 5′-AGTTGTAACCCGATGAGCCTGGCTTAC-3′ (PA), and 5′-CCAATGGGGCCCGATGAGCCTGGCTTAC-3′ (PC). The 3′ portion of the gene containing exons 5–14 was amplified using the primers 5′-CTCATCGGGTTACAACTGTCAAACGAG-3′ (PA), 5′-GGATACACAGGTTACAACTGTCAAACGAG-3′ (PC), and 5′-GGGTCGGCGCGCCCACCCTTCTAAATTAGTCGCTCTTCCG-3′ (both). Exon 4 (unique to the PC isoform) was amplified with the primers 5′-GGCTCATCGGGCCCCATTGGTCACTGCC-3′ and 5′-CAGTTGTAACCTGTGTATCCACTAGCACAAATACAG-3′. The vector pEntr was amplified using the primers 5′-AAGGGTGGGCGCGCCGAC-3′ and 5′-GGTGAAGGGGGCGGCCGC-3′. For pEntr-GFP:crumbs-PA and pEntr-GFP:crumbs-PC, an internal GFP tag (starting at amino acid 89) plus exons 1–3 was amplified from UASt-GFP:crumbs (gift of D. ter Stal and U. Tepass) using the primers above. The exons 1–3, exons 5–14, and pEntr fragments were assembled into pEntr-crumbs-PA and pEntr-GFP:crumbs-PA, and the exons 1–3, exon 5, exons 5–14, and pEntr fragments were assembled into pEntr-crumbs-PC and pEntr-GFP:crumbs-PC using the Gibson Assembly Master Mix (New England Biolabs, Inc.) and grown in DH5α chemically competent Escherichia coli. The pEntr-GFP:crumbs-PA and pEntr-GFP:crumbs-PC constructs were assembled with the same method using the exon 1–3:GFP fragment. All crumbs constructs were recombined into the pUASp-W-attB destination vector (gift of M. Buszczak, University of Texas Southwestern Medical Center, Dallas, TX) to generate pUASp-crumbs-PA, pUASp-crumbs-PC, pUASp-GFP:crumbs-PA, and pUASp-GFP:crumbs-PC. All transgenes were inserted in VK37 on chromosome 2. The CH322-15H13 genomic BAC clone (BACPAC Resources; Venken et al., 2009) was prepped using the Large-Construct kit (QIAGEN), and the transgene was inserted into attP40 on chromosome 2 (Genetic Services).

Embryos were dechorionated for 2 min in 50% bleach and washed with water. Primary antibodies were rabbit aPKC (1:500; Santa Cruz Biotechnology, Inc.), rabbit β-catenin (1:200; Riggleman et al., 1990), rabbit centrosomin (1:500; Lucas and Raff, 2007), mouse Crumbs (1:2; Developmental Studies Hybridoma Bank [DSHB]), rat E-cadherin (1:25; DSHB), mouse neurotactin (1:200; DSHB), guinea pig Par-3/Bazooka (1:500; Blankenship et al., 2006), rabbit Par-3 phospho-S980 (1:250; Krahn et al., 2009), mouse α-tubulin (1:1,000; DM1α; Sigma-Aldrich), and mouse γ-tubulin (1:1,000; Martinez-Campos et al., 2004). For immunostaining with antibodies to β-catenin, Par-3, Crumbs, and neurotactin, embryos were boiled for 10 s in 0.03% Triton X-100/0.4% NaCl, cooled on ice for >30 min, and devitellinized in heptane/methanol. For antibodies to E-cadherin and aPKC, embryos were fixed for 1 h in 4% paraformaldehyde in 0.1-M sodium phosphate buffer and manually devitellinized. For α-tubulin, embryos were fixed for 10 min in 10:9:1 heptane, 37% paraformaldehyde, and 0.5-M EGTA and manually devitellinized. For all other antibodies, embryos were fixed for 20 min in 4% formaldehyde in PBS/heptane and devitellinized in heptane/methanol. Secondary antibodies conjugated to Alexa Fluor 488, 568, or 647 (Molecular Probes) were used at 1:500. Embryos were mounted in Prolong Gold (Invitrogen), imaged on a confocal microscope (LSM700; Carl Zeiss) with a PlanApo 40× 1.4 NA objective (1-μm optical sections acquired at 0.5-μm steps), and staged by posterior midgut morphology.

The expression levels of the pUASp-GFP:crumbs-PA and pUASp-GFP:crumbs-PC transgenes were compared by Western blot analysis. Embryos expressing the respective UASp transgenes under the control of matαtub15 Gal4 were subjected to Western blot analysis using standard procedures. Late stage 5/early stage 6 embryos were lysed in a sample buffer and boiled for 5 min. Equal amounts of embryonic lysate per lane were run on a 4–12% (vol/vol) Bis-Tris SDS/PAGE gel or 4% Tris-acetate SDS-PAGE gel (NuPAGE; Invitrogen). Protein was transferred to a 0.45-μm nitrocellulose membrane by standard protocols. Primary antibodies were used at the following concentrations: rabbit GFP (1:1,000; Torrey Pines), rabbit phospho-aPKC T555/T563 (these sites on the human protein correspond to T567/T574 in Drosophila aPKC isoform PA; 1:1,000; 5813; Abcam), mouse α-tubulin (DM1α; 1:1,000; Sigma-Aldrich), and guinea pig Par-3 (1:1,500; Blankenship et al., 2006). Secondary antibodies were mouse, rabbit, and guinea pig HRP-conjugated antibodies (1:10,000; Jackson ImmunoResearch Laboratories, Inc.) and were detected by chemiluminescence imaging (ECL Plus Western Blotting Detection Reagents; GE Healthcare). For phospho-aPKC (T555/T563) and total Par-3 level analysis, Oregon-R was the wild-type control, and obe¹/obe² mutant embryos were the progeny of obe¹/obe² females crossed to obe²/+ males.

Oregon-R and obe¹/obe² embryos were shaken at 220 rpm for 7 min in 1:1 PBS/heptane in the presence or absence of 56-nM nocodazole. After nocodazole treatment, embryos were fixed immediately in 100 ml of 0.5-M EGTA, 900 ml of 37% formaldehyde solution, and 1 ml heptane, shaken at 220 rpm for 10 min, and manually devitellinized.

Genomic DNA was isolated from wild-type, obe¹/obe¹, and obe²/Df adult males and sequenced using paired-end SOLiD (Sequencing by Oligonucleotide Ligation and Detection) whole-genome sequencing technology (Applied Biosystems). Sequenced libraries were mapped to the Drosophila genome (DM3; UCSC) using a custom pipeline generated by N. Socci at the Memorial Sloan Kettering Cancer Center Bioinformatics core facility. The wild-type control was the starting chromosome on which the obe¹ mutation arose (Schuldiner et al., 2008). No such chromosome was available for obe², but the identified obe² mutations were not present in this control or in Oregon-R.

Total RNA was isolated from late stage 5/early stage 6 Oregon-R, obe¹/obe¹, obe¹/obe², and obe²/Df(3R)Exel6174 embryos as described in the next section. For each sample, 1 µg of total RNA was DNase treated and then reverse transcribed using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems). Then, 50 ng cDNA was amplified in each qRT-PCR reaction using predesigned TaqMan gene expression assays for exons 5 and 6 of CG5205 (Dm02140819_g1) and exons 2 and 3 of RpL32 (Dm02151827_s1; Applied Biosystems). Reactions were carried out using a 7900HT Fast Real-Time PCR system (Applied Biosystems). Relative expression levels were quantified using the 2^−ΔΔC(T) method (Schmittgen and Livak, 2008), and expression was normalized to RpL32 within each sample. The results are the means of three technical replicates.

Genotypes used for RNA sequencing were Oregon-R, obe¹/obe², and obe²/Df(3R)Exel6174, three biological replicates for each genotype. Total RNA was extracted from ~50 embryos hand selected at late stage 5/early stage 6 using the TRIzol (Invitrogen) extraction protocol. RNA sequencing was carried out by the New York Genome Center using the HiSeq2500 platform (Illumina) with 50-bp paired-end reads and >20 million reads per sample. Sequenced libraries were mapped to the Drosophila genome (DM3; UCSC Genome Browser) using the RNA-Seq aligner STAR (v_2.3.1z). Differential analysis of gene and transcript expression was performed using Cuffdiff 2 (v_cufflinks-2.1.1; Trapnell et al., 2013). Transcripts that were differentially expressed in both mutant samples compared with wild type, were similarly or more strongly affected in the stronger obe²/Df combination, and had ≥30 reads in all samples were selected for further analysis. Analysis of several maternally and zygotically expressed genes (De Renzis et al., 2007; Celniker et al., 2009) revealed that obe²/Df and obe¹/obe² samples were developmentally younger than controls. Transcripts whose differences could be explained by the fact that mutant embryos were slightly younger were excluded from analysis. The complete RNA sequencing data set has been deposited in the Gene Expression Omnibus (accession code GSE74545).

To quantify the localization of Crumbs, aPKC, and Par-3, mean intensity values along line scans from single confocal cross sections were measured using the ImageJ (National Institutes of Health) Plot Profile function with a line width of 0.8 microns. Crumbs, aPKC, and Par-3 intensity was the ratio of the mean pixel intensity along a line scan at the cell boundary relative to the mean intensity of three nonoverlapping cytoplasmic regions of 0.8 microns by 2.4 microns each. A single value was obtained for each image by averaging three to six cell interfaces per image; one to three images in four to six embryos were analyzed per genotype.

Par-3 planar polarity was measured in Scientific Image Segmentation and Analysis (SIESTA) software (Fernandez-Gonzalez and Zallen, 2011) and occurred normally in obe mutants. We measured the ratio of the mean Par-3 fluorescence intensity at the dorsal and ventral edges (oriented at 0–15° relative to the anterior–posterior axis) to the mean intensity at the anterior and posterior edges (oriented at 75–90°). Cytoplasmic intensity (the mean of all pixels >1 μm from a cell interface) was subtracted from both values before calculating the ratio. P-values were calculated using the F-test followed by the appropriate t test.

To quantify the level of γ-tubulin at centrosomes, centrosomes were automatically identified in maximum intensity image projections of a 3-μm-thick region along the apical–basal axis using Otsu thresholding (Otsu, 1979) in ImageJ. The fluorescence intensity of γ-tubulin was measured at each centrosome using the Analyze Particles function in ImageJ. The γ-tubulin intensity was the ratio of the mean intensity of γ-tubulin relative to the mean intensity of all pixels outside of centrosomes. A single value was obtained at each stage by averaging 100–200 centrosomes per image; one to three images in four to seven embryos were analyzed for each genotype. P-values were calculated using the F-test followed by the appropriate t test.

To quantify placement of centrosomes in embryos overexpressing Crumbs-PA or Crumbs-PC, centrosomes were identified in maximum intensity image projections of the adherens junctions domain (1.5–2.5 μm) along the apical–basal axis using Otsu thresholding (Otsu, 1979) in ImageJ. Centrosomes were classified as located near the cortex if the shortest distance from the centrosome to the cortex was <1 μm. A single percentage of apically or junctionally localized centrosomes was calculated for each embryo, and plots show the mean values across embryos. P-values were calculated using the F-test followed by the appropriate t test.

Fig. S1 shows the defects in multiple obe allelic combinations. Fig. S2 shows that normal down-regulation of pericentriolar material occurs in obe mutants. Fig. S3 shows Par-3 protein levels and Par-3 and aPKC phosphorylation in wild-type and obe mutants. Fig. S4 shows sashimi plots and isoform analysis for all transcripts with alternative splicing defects in obe mutants. Fig. S5 shows the level of expression and cross sections of embryos that overexpress Crumbs-PA or Crumbs-PC. Table S1 shows all transcripts that were up-regulated or down-regulated more than twofold in obe mutants.

We are grateful to Masako Tamada and Justin Sanny for help with the genetic screen that identified obelus, Nick Socci at Memorial Sloan Kettering Cancer Center and Nicolas Robine at the New York Genome Center for assistance with sequence analysis, Frederik Wirtz-Peitz for the pSqh-GFP-W-attB destination vector, David ter Stal and Ulrich Tepass for the pUASt-Crumbs^WT and pUASt-GFP:Crumbs plasmids, Mike Buszczak for the pUASp-W-attB plasmid, and Kathryn Anderson, Mary Baylies, Eric Brooks, Hiten Madhani, Adam Paré, German Sabio, Beate Schwer, and Naomi Stevens for helpful discussions and comments on the manuscript.

This work was supported by National Institutes of Health/National Institute of General Medical Sciences R01 grant GM079340 to J.A. Zallen and by National Institutes of Health T32 training grant GM008539 to A. Vichas. J.A. Zallen is an Investigator of the Howard Hughes Medical Institute.

The authors declare no competing financial interests.