Abstract
The adherens junction couples the actin cytoskeletons of neighboring cells to provide the foundation for multicellular organization. The core of the adherens junction is the cadherin-catenin complex that arose early in the evolution of multicellularity to link cortical actin to intercellular adhesions. Over time, evolutionary pressures have shaped the signaling and mechanical functions of the adherens junction to meet specific developmental and physiological demands. Evolutionary rate covariation (ERC) identifies genes with correlated fluctuations in evolutionary rate that can reflect shared selective pressures and functions. Here we use ERC to identify genes with evolutionary histories similar to shotgun (shg), which encodes the Drosophila E-cadherin (DE-Cad) ortholog. Core adherens junction components α-catenin and p120-catenin displayed strong ERC correlations with shg, indicating that they evolved under similar selective pressures during evolution between Drosophila species. Further analysis of the shg ERC profile revealed a collection of genes not previously associated with shg function or cadherin-mediated adhesion. We then analyzed the function of a subset of ERC-identified candidate genes by RNAi during border cell (BC) migration and identified novel genes that function to regulate DE-Cad. Among these, we found that the gene CG42684, which encodes a putative GTPase activating protein (GAP), regulates BC migration and adhesion. We named CG42684 raskol (“to split” in Russian) and show that it regulates DE-Cad levels and actin protrusions in BCs. We propose that Raskol functions with DE-Cad to restrict Ras/Rho signaling and help guide BC migration. Our results demonstrate that a coordinated selective pressure has shaped the adherens junction and this can be leveraged to identify novel components of the complexes and signaling pathways that regulate cadherin-mediated adhesion.
Author Summary The establishment of intercellular adhesions facilitated the genesis of multicellular organisms. The adherens junction, which links the actin cytoskeletons of neighboring cells, arose early in the evolution of multicellularity and selective pressures have shaped its function and molecular composition over time. In this study, we used evolutionary rate covariation (ERC) analysis to examine the evolutionary history of the adherens junction and to identify genes that coevolved with the adherens junction gene shotgun, which encodes the Drosophila E-cadherin (DE-Cad). ERC analysis of shotgun revealed a collection of genes with similar evolutionary histories. We then tested the role of these genes in border cell migration in the fly egg chamber, a process that requires the coordinated regulation of cell-cell adhesion and cell motility. Among these, we found that a previously uncharacterized gene CG42684, which encodes a putative GTPase activating protein (GAP), regulates the collective cell migration of border cells, stabilizes cell-cell adhesions and regulates the actin dynamics. Our results demonstrate that components of the adherens junction share an evolutionary history and that ERC analysis is a powerful method to identify novel components of cell adhesion complexes in Drosophila.
Introduction
The adherens junction (AJ) is a multiprotein complex that is essential for intercellular adhesion in metazoa. The core of the AJ is the cadherin-catenin complex. Classical cadherins are single-pass transmembrane proteins with an extracellular domain that mediates calcium-dependent homotypic interactions. The adhesive properties of classical cadherins are driven by the recruitment of cytosolic catenin proteins to the cadherin tail: p120-catenin binds to the juxta-membrane domain and β-catenin binds to the distal part of the tail. β-Catenin recruits α-catenin to the cadherin-catenin complex. α-Catenin is an actin-binding protein and the primary link between the AJ and the actin cytoskeleton [1–3].
The primary function of the AJ is to link actin to intercellular junctions. It is believed the AJ arose early in the evolution of multicellular metazoans to coordinate epithelial tissue formation and organization [4–7]. The AJ has since evolved to function in a range of physiological and developmental processes, including cell polarity, collective cell migration and cell division [8, 9]. AJ function in these diverse processes requires an array of ancillary regulatory proteins, including kinases, signaling molecules and adaptor proteins [10–14]. Defining the molecular networks that regulate AJ biology is critical to understanding cadherin-mediated adhesion in normal and disease states.
Evolutionary rate covariation (ERC) analysis is a comparative genomic approach that has been used successfully to identify genes with shared functions in canonical protein complexes and biological processes in prokaryotes, fungi, Drosophila and mammals [15–21]. ERC analysis takes advantage of the evolutionary rates shared between co-functional genes that have correlated rates of change due to common selective pressures. ERC represents the correlation coefficient of branch-specific evolutionary rates that are estimated from the phylogenetic trees of a pair of genes and their orthologs from multiple species[19]. ERC analysis permits the identification of genes within a given genome that evolved in a correlated manner and hence might function in the same pathway or molecular complex. These genes can then be screened by RNAi-based knockdown or similar genetic approaches to validate their role in a relevant biological process.
Border cell (BC) migration in the developing Drosophila egg chamber requires coordinated cell adhesion and migration. During BC migration, a group of 6–8 follicular cells delaminate from the anterior most tip of the epithelium and undergo haptotaxis and migrate collectively towards the developing oocyte [22, 23]. The BC cluster consists of migratory BCs and a centrally positioned pair of polar cells (PCs) that signal to BCs and contribute to cluster adherence [23]. BC migration is highly dependent on Drosophila E-Cadherin (DE-Cad, encoded by shotgun (shg)) [24–26]. Upregulation of DE-Cad is essential for the initial delamination and subsequent migration of BC since disruption of DE-Cad-mediated adhesion affects the ability of BC to detach from the follicular epithelium (FE) and collectively migrate [24, 25].
We performed ERC analysis of shg to identify genes that share a common evolutionary history, and therefore may share an overlapping function, with shg and assessed their role in BC migration. Genes encoding the primary components of AJ, including α-Cat and p120-catenin, display high ERC values relative to shg and one another, suggesting that these genes and their protein products are co-functional, which is well described in the literature [1, 27]. We show that genes showing high ERC values with shg are enriched for membrane-associated proteins and proteins that function in E-cadherin-dependent biological processes. We then conducted an RNAi-mediated genetic screen in BCs with 34 high-ranking ERC candidates and identified both novel and known genes that function to regulate DE-Cad at cell contacts. Among those, we characterized a GTPase activating protein (GAP) domain encoding gene, CG42684, which we have named “raskol” after the Russian term “to split”. We show that Raskol colocalizes with DE-Cad, regulates DE-Cad levels at the BC-BC interface and modulates actin-rich protrusions during BC migration. Our results demonstrate that components of the AJ share an evolutionary history and that ERC analysis is a powerful method to identify novel components of cell adhesion complexes in Drosophila.
Results
ERC analysis identifies genes that coevolved with shg
We used ERC analysis to identify novel genes that regulate DE-Cad-mediated adhesion in Drosophila. Since ERC signatures are often observed between proteins that function in a molecular complex [15, 16, 18, 19], we evaluated the ERC values of the fly AJ components – shotgun (shg, DE-Cad), armadillo (arm; β-catenin in vertebrates), α-catenin (α-Cat), p120-catenin (p120ctn), Vinculin (Vinc) and canoe (cno, afadin in vertebrates). Notably, shg, α-Cat and p120ctn displayed positive ERC values with each other (Fig 1A). arm, Vinc and cno did not show elevated ERC values relative to shg, α-Cat or p120ctn. Since Arm, Vinc and Cno are known to function independently of cadherin-mediated adhesion [28–31], we speculate that alternative selective pressures have influenced the evolution of these genes in flies, likely obscuring any ERC signature with purely AJ genes. Nonetheless, ERC analysis suggested that the AJ components DE-Cad, α-Cat and p120ctn coevolved to maintain their collective function in cell-cell adhesion. We postulated that other genes whose products regulate AJ biology would have similar evolutionary histories to maintain functionality.
We then used ERC analysis to identify genes with high ERC values relative to the core of the AJ complex, shg (S1 Table). We identified 137 genes with ERC values of 0.4 or greater, representing the top 1.3% of shg ERC values. Since α-Cat has an ERC value of 0.47 relative to shg (Fig 1A), placing their ERC value in the top 0.6 % of all gene pairs, we reasoned that genes with similar or higher ERC values would represent genes with similar evolutionary histories to shg. Accordingly, the thresholded shg ERC list contains genes with described roles in AJ regulation such as Hrb98DE [32, 33], PDZ-GEF [34, 35], babo [36, 37], CG16952 [38, 39] and Rab5 [40–42] (Table 1). Excitingly, the majority of identified genes have not been associated with the AJ and include transcription factors, kinases, GTPase regulatory proteins and calcium channel regulators (Table 1). A previous genomic RNAi screen conducted in Drosophila S2 cells [43] and E-Cadherin proximity biotinylation screens in epithelial cells [44, 45] identified multiple hubs of interactors and regulators. Cross referencing the shg ERC list with the hits from these screens revealed only a few common genes such as RhoGAPp190, Rab5, Appl and Stim. This suggests that ERC analysis is identifying additional shg regulatory components that were undetected in genetic or proteomic screens.
Using gene ontology (GO) based enrichment analysis, we found that the shg ERC genes are enriched for plasma membrane (PM) localized proteins and PM-associated protein complexes (Fig 1B). Additionally, genes that function in biological processes requiring E-cadherin-mediated adhesion, such as wing disc morphogenesis, imaginal disc morphogenesis, epithelial morphogenesis, cell migration and AJ organization, were significantly overrepresented in the shg ERC list (Fig 1C). Next, we analyzed the molecular functions of the human orthologs of the ERC identified genes. We found that genes involved in epithelial AJ remodeling and cancer molecular mechanism were overrepresented (Fig 1D; S2 and S3 Table). Also, genes that function in RhoA, CCR5, TGF-β, PTEN and AJ-mediated signalling pathways were enriched (Fig 1D, S2 and S3 Table). Thus, the shg ERC list contains genes with established roles in regulating AJs as well as novel candidate genes.
RNAi screen in BCs identifies genes that regulate cell-cell adhesion
To evaluate the function of genes in the shg ERC list, we conducted an in vivo RNAi-based genetic screen in the Drosophila egg chamber. We analyzed BC collective cell migration (CCM) because it is regulated by DE-Cad and is a powerful system to study the interplay between cell migration and cell-cell adhesion (Fig 2A) [23, 24]. BC detachment from the FE and concomitant CCM requires increased DE-Cad expression and loss of DE-Cad arrests BC migration [24, 25].
First, we downregulated levels of individual AJ genes by using the GAL4/UAS system to drive UAS-RNAi transgenes in the migrating BC cluster. We used a BC specific driver, slowborder-GAL4 (slbo-GAL4) [46, 47] to drive expression of a UAS-GFP reporter and a UAS-RNAi transgene targeted against the gene of interest. Stage 10 egg chambers expressing RNAi and GFP in BCs were fixed and scored for BC cluster position along the anterior-posterior migration axis (Fig 2H). In control egg chambers, nearly all BC clusters completed migration and were positioned adjacent to the oocyte (Fig 2B and 2I). In contrast, downregulation of shg caused a BC migration failure or delay in all egg chambers (Fig 2C and 2I). BC migratory defects were less severe in egg chambers with reduced expression of α-Cat compared to shg; however, the prevalence of defects was higher than in control egg chambers (Fig 2I). We could not assess the effect of arm downregulation since arm RNAi expressing flies did not survive to adulthood. The downregulation of CadN, Vinc, p120ctn or cno did not lead to BC migratory defects (Fig 2I).
Next, we screened 34 genes from the shg ERC list. We focused on genes that are expressed in the ovary [48] and for which an RNAi stock was readily available (S4 Table). The downregulation of target genes displayed a variable range of BC migration defects with 12 genes displaying defects in more than 15% of egg chambers assessed compared to 4% in control (Fig 2J). We also randomly selected and screened six genes that had either very low ERC values or did not appear in the shg ERC list for migration defects. As expected, we did not observe strong migration defects when these genes were knocked down (Fig 2K). Moreover, the shg ERC genes showed statistically lower average migration than the random negative control genes (Wilcoxon rank sum test, p=0.0162), supporting the hypothesis that genes with correlated evolutionary histories share functional characteristics. Pdk1 knockdown resulted in the most penetrant phenotype with 50% of egg chambers displaying either a failure or delay in BC migration (Fig 2D and 2J). Knockdown of babo and CG42684 caused migration delays similar to α-Cat (Fig 2E–J). Additionally, knockdown of multiple genes resulted in considerable migration delays relative to the control including CG16952 (Fig 2D), Hrb98DE, magu, InR, RhoGAPp190, PDZ-GEF and CG11593 (Fig 2J). Conversely, knockdown of genes such as enc, mppe and rfx did not affect BC migration (Fig 2J).
While screening α-Cat knockdown egg chambers, we noticed that about 20% of the egg chambers displayed a cluster disassociation phenotype where one or more BCs had separated from the cluster (Fig 3B and 3G). Since this phenotype is indicative of cell adhesion defects in BCs [24], we scored cluster disassociation for all genotypes. In control BCs expressing luciferase RNAi, disassociated clusters were rarely observed (Fig 3A and 3G). Likewise, downregulation of genes with low ERC values did not show a cluster disassociation phenotype (Fig 3I). Knockdown of CadN, vinc, p120ctn and cno also did not cause a penetrant cluster disassociation phenotype (Fig 3G). However, cluster disassociation phenotypes were observed with a number of ERC target genes. CG42684, PDZ-GEF, CG16952, CG11593, zormin, cac and Rab5 displayed similar or higher cluster disassociation phenotypes compared to α-Cat (Fig 3). Overall, these ERC target genes, chosen for their high shg ERC values, exhibited the cluster disassociation phenotype significantly more often than genes with low shg ERC values (Wilcoxon rank sum test, p= 0.00022). Together, these results demonstrate that highly ranked genes in the shg ERC list contain factors that may regulate cell adhesion during BC collective cell migration.
Top ERC candidates regulate DE-Cad levels in BCs
Since knockdown of a subset of shg ERC list genes disrupted BC migration, we hypothesized that these genes might regulate DE-Cad at BC contacts. To test this, we quantified DE-Cad levels along cell-cell contacts between BCs in RNAi-expressing clusters. We used a DE-Cad-GFP knock-in stock that express DE-Cad at endogenous levels [49] and drove RNAi constructs under the control of slbo-GAL4. In control egg chambers where either luciferase RNAi or RFP were expressed, high DE-Cad levels were observed at BC-BC contacts (Fig 4A, S2A Fig). As expected, shg RNAi expression severely reduced DE-Cad levels in the BC cluster (Fig 4B). Knockdown of CG42684 (Fig 4C), CG16952 (Fig 4D), CG11593 (Fig 4E), babo (Fig 4G) and Hrb98DE (S2B Fig) caused a significant reduction in DE-Cad levels at BC-BC contacts. Interestingly, knockdown of Pdk1, a kinase in the PI3K pathway [50], did not affect DE-Cad levels in BCs even though it displayed the most prominent migration defect (Fig 4F). As expected, expression of RNAi for genes capt and CG5872, which have very low ERC values relative to shg and were therefore not predicted to operate in this pathway, did not cause a reduction in DE-Cad levels at BC-BC contacts (S2C and S2D Fig). Collectively, these results indicate that a significant subset of genes with high ERC values relative to shg, many of which were previously not associated with cadherin function, regulate DE-Cad levels at BC-BC contacts.
Raskol colocalizes with DE-Cadherin
Knockdown of CG42684 displayed the most severe cell disassociation phenotype amongst all genes tested (Fig 3). CG42684 is reported to localize at the cell cortex and is enriched specifically at the apical surface of epithelial cells in Drosophila embryo [51] though almost nothing is known about its molecular function in flies. The mammalian orthologs of CG42684, Rasal2 and Dab2IP, also localize to the PM [52]. Interestingly, and similar to the impact we report here for CG42684, downregulation of Rasal2 disrupts E-Cadherin localization at the cell contacts [53], though the mechanism for this disruption remains poorly defined and its conservation across phyla has yet to be reported. Therefore, we wanted to determine whether CG42684, which we named “Raskol” (Russian for “to split”) associates with DE-Cad along the cell membrane. Consistent with earlier localization studies, a YFP-trap stock expressing Raskol-YFP localized to the cell periphery in embryonic epidermal cells (S3 Fig) [51]. We used this stock to assess Raskol colocalization with DE-Cad along the BC membrane. In Drosophila stage 8 embryos, before BCs have delaminated, Raskol localized to the cell membrane of BCs and PCs and was enriched at the PC apical membrane (Fig 5A, S2B Fig). During migratory stages, Raskol localization persisted at the PC apical membrane and at the cell-cell contacts of BCs and PCs (Fig 5B). Immunolabeling of egg chambers with DE-Cad antibody or imaging of endogenously RFP-tagged DE-Cad revealed strong colocalization between DE-Cad and Raskol (R=0.63, n=15) at the apical surface of PCs and at BC-BC contacts (Fig 5C and 5D). We also observed colocalization in the FE, particularly along the apical membrane (Fig 5A and 5C; S2A and S2B Fig). Raskol was not present at cell contacts between NCs (S2B Fig).
Next, we determined whether Raskol colocalizes with DE-Cad in other tissues that require AJ-mediated adhesion. Dorsal closure (DC) is an embryonic process in which the migrating ectoderm closes the dorsal hole [54–56]. The amnioserosa, an extra-embryonic tissue, covers the dorsal hole and contributes to ectodermal closure by providing contractile forces that pull the contralateral ectodermal sheets together [54, 57, 58]. DC requires DE-Cad-mediated adhesion for ectodermal migration and fusion [59, 60]. To analyze Raskol and DE-Cad dynamics, we conducted time-lapse live imaging of embryos expressing YFP-tagged Raskol and RFP-tagged DE-Cad during DC. Colocalization of Raskol and DE-Cad was observed both at the amnioserosa cell contacts as well as in the dorsal most ectodermal cells at the zippering interface (S3A-D Fig). Raskol and DE-Cad colocalize in multiple Drosophila tissues, suggesting that Raskol may be a fundamental regulator of DE-Cad.
Raskol regulates the distribution of polarized actin protrusions
Analysis of Raskol-YFP protein localization in BC clusters revealed that cytoplasmic levels of Raskol were ∼2x higher in PCs compared to BCs (S2B Fig and S4A Fig). Similarly, DE-Cad levels were up-regulated in PCs relative to BCs (Fig. 4A) [24] suggesting that Raskol protein expression pattern trends with DE-Cad. Accordingly, shg and raskol gene expression patterns overlap during embryonic development and both peak at 6–8hrs after egg laying (S4B Fig) [48]. To determine if Raskol is required for maintaining cell adhesions in PCs similar to BCs, we expressed raskol RNAi using unpaired-GAL4 (upd-GAL4) to drive expression specifically in the PCs [61]. Egg chambers were stained for F-actin (phalloidin) and nuclei (DAPI). Expression of control RNAi did not affect cluster adherence; however, shg RNAi expression caused cluster disassociation in ∼80% egg chambers (Fig 6A, 6B and 6D), similar to a previous report [24]. Expression of raskol RNAi in PCs caused BC disassociation in 63% of egg chambers (Fig 6C and 6D), suggesting that Raskol and DE-Cad might function together to promote BC cluster adhesion in both PCs and BCs.
Next, we focused on the Raskol-YFP signal at the leading edge of the BC cluster (Fig 5B and 5D). Here, Raskol colocalized with actin suggesting that it may play a role in actin dynamics. Since Raskol contains a GAP domain and colocalizes with the cortical actin cytoskeleton in BCs, the FE, the amnioserosa and the dorsal most ectodermal cells [55, 60], we sought to determine whether it functions to regulate actin organization in BCs. We stained egg chambers expressing raskol RNAi in the BCs under the control of slbo-GAL4 driver for F-actin (phalloidin) and nuclei (DAPI) to assess F-actin distribution in the migrating cluster. In control BCs, actin accumulated at the base of protrusions typically oriented in the direction of migration (Fig 6E). In contrast, downregulation of raskol resulted in dramatic formation of multiple ectopic actin patches around the BC cluster (Fig 6F and 6G).
We then performed time-lapse live imaging of BC clusters expressing UAS-lifeact-GFP under the control of slbo-GAL4 to analyze protrusion dynamics in more detail. In control egg chambers, we observed protrusions extending primarily from the front of the migrating cluster (Fig 6H, 6J-L; Movie 1). In contrast, when raskol was downregulated, protrusions extended indiscriminately around the cluster (Fig 6I-L; Movie 2), consistent with our previous observations. Interestingly, the number of front-oriented protrusions in control and raskol RNAi expressing BCs did not differ significantly (Fig 6K). These data suggest that Raskol acts to restrict actin protrusions to the front of the BC cluster, which is critical to regulate BC migration. In addition, raskol knockdown caused BC delamination defects (Movie 3) and cluster disassociation (Movie 4) thereby confirming its importance in controlling cell adhesion and providing initial mechanistic insight into its role in regulating actin dynamics.
Discussion
We combined evolution-guided bioinformatics with classical RNAi-based screening in Drosophila to identify regulators of DE-Cad-mediated cell adhesion. Our screen uncovered both established and novel regulators of DE-Cad function during BC migration. We demonstrated that one hit, the previously uncharacterized GAP domain containing protein Raskol, colocalizes with DE-Cad and regulates polarized actin dynamics in migrating BCs.
ERC analysis reveals an evolutionary relationship between core components of the AJ
The core AJ components shg, α-Cat and p120ctn display high ERC values relative to one another, suggesting that the AJ complex has coevolved under selective pressure. α-Catenin provides the mechanical link between the cadherin-catenin complex and the actin cytoskeleton. Actin linkage is believed to be the original, ancestral function of the adherens junction that provided the foundation for multicellularity [9], so it is not surprising that shg and α-Cat, given their functional roles, have coevolved. Interestingly, p120-catenin is not essential for cadherin-mediated adhesion in flies [62], though it plays an important role in cadherin endocytosis in flies [63], similar to its established role in vertebrates [64]. Our ERC analysis suggests that shared selective pressures guided the evolution of the p120-catenin and DE-cadherin complex, and we speculate that these pressures may have shaped the range of p120-catenin functions in higher vertebrates.
Notably, no significant ERC relationship was observed between shg and arm or shg and other secondary AJ complex genes. While Arm is a core component of the AJ, it also functions as a key transcription factor in the Wnt signaling pathway [28, 29, 65]. We speculate that Arm function in Wnt signalling placed additional evolutionary pressures and altered its ERC signature relative to the other AJ genes. Similarly, neither cno nor vinc showed a strong ERC relationship to AJ genes, possibly reflecting their individual roles in AJ-independent processes [31, 66].
A number of genes identified in the shg ERC analysis have been implicated in regulating cell adhesion (Fig 1, Table 1). However, many of ERC-identified genes have not been functionally associated with DE-Cad, the AJ or cell adhesion. A previous genomic RNAi screen conducted in Drosophila identified multiple regulators of DE-Cad [43]. Notably, there was little overlap between the two screens. This highlights the potential of ERC analysis as an alternative, unbiased approach to generate a target gene list based solely on the evolutionary rate comparison. However, it is important to conduct secondary screens to validate the function of the target genes in a relevant biological system to eliminate false positives [16, 18, 19]. Also, ERC analysis cannot predict where (e.g., tissue type, cellular component) or when (e.g., developmental stage) a putative interaction will occur. Refinement of the shg ERC list based on spatio-temporal expression data can further eliminate false positive hits [19]. Nonetheless, a major advantage of ERC analysis is that genes that would otherwise not arise in a functional or associative genetic screen can be identified.
DE-Cad and its regulators are required for BC migration
BC migration requires coordinated regulation of adhesion and motility [23, 67, 68] and is a good system for testing genes that regulate DE-Cad. BC migration is also an in vivo model for metastasis since many morphological characteristics of BCs resemble the invasive behaviour of metastatic cell clusters [68, 69]. However, in contrast to most models of epithelial-to-mesenchymal transition, the detachment of BC cluster requires upregulated levels of DE-Cad [23, 24]. DE-Cad-mediated adhesion between BCs and NCs is required for cluster polarization and directional migration, whereas adhesions between BCs and PCs are required for cluster adherence during migration [24, 25, 70]. Thus, BC migration is a useful system for genetic studies of cell adhesion and offers an opportunity to explore the role of adhesion genes in a relevant disease model [68, 69].
Our secondary genetic screen of shg ERC hits revealed potential roles for number of genes in BC migration, including kinases, GTPase regulators, transcription factors and cytoskeletal proteins. A small number of hits have putative or established roles in regulating DE-Cad and/or AJs. For example, PDZ-GEF was shown to colocalize with DE-Cad and function through GTPase Rap1 to regulate DE-Cad at the cell membrane [34, 35]. Hrb98DE is an RNA-binding protein that regulates DE-Cad mRNA processing [32, 33]. Mammalian orthologs of the small GTPase Rab5 regulate E-Cad trafficking [40, 42]. The mammalian homolog of transcription factor CG16952, Btbd7, regulates E-Cad expression [38, 39]. Additionally, genes such as RhoGAPpp190, Stim, Appl and Rab5 were also identified in functional and proteomics screens of E-Cad [43, 44]. We also identified numerous genes that have not been linked to DE-Cad, including raskol, CG11593, babo and zormin. Out of these genes, raskol, CG11593 and babo directly regulate DE-Cad levels at the BC-BC contacts (Fig 4). Using shg ERC analysis and BC migration as a genetic model we have identified multiple novel genes that function to regulate DE-Cad-mediated cell adhesion in Drosophila.
Raskol is a putative regulator of cell adhesion, polarity and actin dynamics
Downregulation of raskol caused severe BC cluster disassociation suggesting that Raskol is a critical regulator of BC adhesion. Consistent with this, Raskol colocalized with DE-Cad in multiple cell types and knockdown of raskol reduced DE-Cad levels at BC-BC cell contacts. Like shg, raskol is significantly upregulated in PCs relative to BCs and NCs. The encoded protein contains a highly conserved GAP domain that displays homology towards Ras- and Rho-GAPs, a plekstrin homology (PH) domain and a C2 domain that likely promote its membrane localization [71]. This suggests that, by colocalizing with DE-Cad, Raskol regulates adhesive strength between BCs to maintain cluster adhesion during detachment from the FE and subsequent migration. The mammalian orthologs of Raskol, Rasal2 and Dab2IP, were identified in a screen for RasGAP tumour suppressors [71] and are frequently downregulated in multiple types of cancer cells [53, 72–75]. Rasal2 and Dab2IP are capable of inactivating Ras through inducing GTP hydrolysis through their GAP domain and their downregulation leads to Ras overactivation [76–78]. Furthermore, inactivation of Rasal2 promotes invasive behaviour in a cell migration assay suggesting that Rasal2 has a conserved role in regulating cell adhesion and protrusive behaviour in mammals [71]. Dab2IP was identified in cadherin proximity biotinylation screens in mammalian epithelial cells [44] and mouse neonatal cardiomyocytes [79], further suggesting that the Rasal2/Dab2IP/Raskol family of proteins regulate AJ biology. Nonetheless, the mechanism of their function remains unclear.
Our study offers potential insight into Raskol function during collective migration. Epidermal growth factor receptor (EGFR) and PDGF- and VEGF-related receptor (PVR) localize to the leading edge of BC clusters and respond to a presumed gradient of guidance cues originating from the oocyte [67, 80–82]. The BC with the highest levels of EGFR/PVR activation becomes the leader cell and relays a signal to neighboring BCs through the DE-Cad adhesion complex to inhibit protrusion formation at the sides or rear of the cluster [24]. Interestingly, gurken, which encodes one of the four ligands for EGFR [81, 83, 84], also appeared on the shg ERC list (ERC value 0.62) as did its receptor Egfr (epidermal growth factor receptor, ERC value 0.36; S1 Table). The presence of both ligand and receptor suggests that EGF signaling has coevolved with DE-Cad to regulate cell adhesion. The primary GTPase that functions to regulate the directional migration of BC downstream of EGFR and PVR is Rac1, a member of the Rho GTPase family of proteins. We propose that Raskol, as a GAP, may function to suppress Rac1 signalling in non-leader BCs. Rac1 is expressed in all BCs, but the leader cells exhibit higher activity due to increased activation of EGFR and PVR [24, 80, 85]. Our results show that Raskol, like EGFR, PVR and Rac1 [80, 81, 85], restricts protrusions to the front of migrating BC cluster thus ensuring unidirectional migration. Downregulation of DE-Cad causes disruption in the polarized distribution of Rac1 in BC clusters, suggesting that DE-Cad regulates signaling downstream of EGFR and PVR [24, 86]. Therefore, Rac1 suppression might be achieved through Raskol GAP activity since knockdown of Rac1 or Raskol produce similar protrusion phenotypes [85]. Raskol may buffer the DE-Cad/Rac/actin mechanical feedback loop to regulate cell adhesion and promote collective cell migration. Whether Raskol directly interacts and regulates the GTPase domain of Rac1 remains to be explored.
Raskol localization is polarized with highest levels observed at the apical domain of ectodermal cells, FE cells and PCs. We also observe Raskol localization in the leading protrusion of BC cluster. This suggests that Raskol might regulate actin dynamics at the apical domain of polarized cells. However, Raskol does not directly regulate formation of protrusions since reducing Raskol levels does not affect the prevalence of leading protrusions. We predict that in leading BCs, Raskol limits active Rac1 to the tip of the protrusion to induce localized actin cytoskeletal remodelling. Accordingly, as reported by a Rac1-FRET sensor, high Rac1 activity is limited at the lamellopodial tip in the leading BC [24]. Overall, these data highlights two potential roles of Raskol function: 1) Raskol functions as a putative regulator of cell adhesion, and 2) Raskol regulates actin dynamics of the migrating cluster downstream of receptor tyrosine kinase signaling in BCs. Future studies dissecting the role of Raskol and other genes identified in our study are expected to offer insight in to the cooperative role of proteins that function along with the AJ to promote cell adhesion and cell migration.
Methods and Materials
Evolutionary Rate Covariance analysis
ERC values were calculated from protein coding sequences from 22 Drosophila species: D. ananassae, D. biarmpies, D. bipectinada, D. elegans, D. erecta, D. eugracilis, D. ficusphila, D. grimshawi, D. kikawaii, D. persimilis, D. pseudoobscura, D. melanogaster, D. miranda, D. mojavensis, D. rhopaloa, D. sechelia, D. simulans, D. suzukii, D. takahashii, D. virilis, D. willistoni, and D. yakuba. Protein coding sequences were downloaded from the Flybase website (http://www.flybase.org/) or the NCBI genome annotation website (https://www.ncbi.nlm.nih.gov/genome/annotation_euk/all/). Initially, coding sequences were evaluated for internal stop codons and the sequence was removed if found. For genes with multiple transcripts, the transcript with the longest sequence size was selected to represent the gene.
Orthology between genes across the multiple species were determined using the Orthofinder algorithm [87]. For each orthogroup, which are sets of genes that are orthologs and/or recent paralogs to each other, we omitted paralogous genes. Only orthogroups that had at least 6 species representation were analyzed further. Gene members of each orthogroup were aligned to each other using the PRANK aligner [88]. The multisequence alignment of each orthogroup was used by the PAML aaml program [89] to estimate the evolutionary rates on a single fixed species topology. A single species topology was estimated using a supertrees approach by combining individual orthogroup topologies that were estimated using RAxML [90]. Trees were combined using the matrix representation method implemented in phytools [91]
ERC was calculated using the branch lengths of each orthogroup. The overall species phylogenetic rates were normalized out for each orthogroup’s evolutionary rate, as described previously [18, 19]. Afterwards, ERC was measured as the Kendall’s τ correlation coefficients between two orthogroups and their species phylogeny normalized relative rates. ERC was then calculated for all pairwise orthogroup combinations.
Gene ontology analysis
The Drosophila gene lists were subjected to GO analysis using Flybase website (http://www.flybase.org/). Homologs of Drosophila genes in mammalian genomes were generated using Flybase website (http://www.flybase.org/). A mammalian gene was considered a homolog if the gene was reported by 45% or more algorithms. Mammalian homologs were analyzed for canonical pathway and disease & function enrichment using Ingenuity Pathway Analysis tools (https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis/).
Drosophila melanogaster strains
All GAL4, reporter TRAP and RNAi stocks were obtained from Bloomington Drosophila Stock Center (BDSC) (S4 Table). DE-Cad-GFP and DE-Cad-mCherry knock-in stocks were used as E-Cad reporters [49]. slbo-GAL4, UAS-lifeact-GFP, UAS-LacZ stock was generously provided by Jiong Chen (Nanjing University) [92]. Fly stocks were raised on standard yeast-based media at 20°C, unless otherwise noted.
BC migration screen
For BC migration analysis, RNAi-expressing female flies under the control of slbo-GAL4 were collected and transferred to vials containing fresh yeast paste and males. Flies were raised at 29ºC for 1–2 days. UAS-GFP was used as a reporter for RNAi expression. Dissected ovaries were fixed in 4% paraformaldehyde in PBS for 20 mins and washed 5 times with PBS. Ovaries were mounted on microscope slides in 70% glycerol and 20 μm z-stacks were acquired with a 20x objective on Nikon A1 scanning confocal microscope.
Immunostaining of egg chambers
1–2 day old females were incubated at 29ºC for 1–2 days in vials with fresh yeast paste and males. Ovaries were dissected, fixed in 4% paraformaldehyde for 20 mins in PBS with 0.1% Triton-X (PBST), washed 5 times with PBST and blocked in normal goat serum (NGS) for 30 mins. For primary antibody staining, ovaries were incubated with Dcad2 antibody (1:100, Developmental Studies Hybridoma Bank) overnight at 4ºC and washed 10 times the next day over 1 hour. Next, ovaries were incubated with Alexa Fluor-conguated secondary antibody (1:150, Thermo Fisher Scientific) and Alexa 647 phalloidin (1:150, Thermo Fisher Scientific) for 2 hours. Egg chambers were then incubated in DAPI for 10 mins. Ovaries were washed 5 times in PBST and washed overnight and washed again 5 times next morning. Ovaries were stored and mounted on microscope slide in 70% glycerol and then imaged as described previously.
Live imaging of BC clusters
Male flies containing slbo-GAL4, UAS-lifeact-GFP and UAS-LacZ were crossed to UAS-raskol-RNAi females. 1–2 day old F1 females were incubated at 29ºC for 1–2 days in vials with fresh yeast paste and slbo-GAL4, UAS-lifeact-GFP males. Ovaries were dissected for live imaging in imaging media (Schneiders’s medium, 15% fetal bovine serum (FBS) and 0.2mg/ml Insulin; Thermo Fisher Scientifc) according to published protocols [22, 93]. 100 μl of imaging media containing egg chambers was transferred to poly-D-lysine coated Mattek dishes for imaging. 20 μm Z-stacks (1 μm step size) covering the whole migrating border cell cluster were acquired every 2 minutes using a 40x objective on a Nikon A1 scanning confocal microscope.
Live imaging of Raskol-YFP in embryos
Raskol-YFP [51] homozygous female flies were crossed to DE-Cad-mCherry [49] homozygous males. Embryos were collected overnight on grape juice agar plates and transferred to microscope slides coated with double-sided tape. Embryos were manually dechorionated and immediately transferred to halocarbon oil on coverslips with the dorsal side facing down. Coverslips were then attached to imaging chambers using double sided tape and imaged using a 60x objective on a Nikon A1 scanning confocal microscope.
Quantification and statistics
Border cell migration defects were quantified as described previously [24]. Stage 10 egg chambers were analysed for each genotype. Border cell position along the migratory path was assigned into one of the following categories: 0–25% (no migration), 25–75% (delayed migration) and 75–100% (completed migration).
To quantify defects in border cell cluster adhesion, we determined the percentage of egg chambers where individual border cells had detached from the cluster.
To quantify DE-Cad levels, linescans across the BC-BC contacts were used to calculate the maximum pixel intensity at the contact in ImageJ. Peak values were then normalized to the peak intensity values of cell-cell contacts between NCs for each egg chamber. One-way ANOVA followed by Mann-Whitney tests were performed to determine significance. At least 22 border cell clusters (3 cell contacts per cluster) were imaged for each genotype.
To measure cytoplasmic levels of Raskol, an ROI was drawn in the cytoplasm of polar and border cells and average intensity determined. The ROI intensity was normalized to the average cytoplasmic intensity of Raskol-YFP in the nurse cells. To quantify colocalization of DE-Cad and Raskol, Pearson’s correlation coefficient (R) was calculated using the JACop plugin in ImageJ.
Ectopic actin patch number was quantified as described [92]. Actin patch present at the base of the leading edge protrusion was excluded from quantification since it is not ectopic. Welch’s t-test was performed to statistically compare number of ectopic actin patches between samples.
To quantify protrusion direction, we measured protrusions around the BC cluster in 45° increments at each frame of the movie (16 frames from 8 movies for each genotype). Protrusions between 315º and 45º angles were considered frontal protrusions; between 45º and 135º and 225º and 315º as middle protrusions; and between 225º and 135º as rear protrusions. Mann-Whitney test was performed to statistically compare number of protrusions between samples.
Acknowledgements
We would like to thank Dr. Jiong Chen for providing slbo-lifeact-GFP stock and Bloomington Drosophila Stock Center for providing GAL4 and RNAi stocks. We also thank members of the Kwiatkowski lab for critical feedback on the manuscript.
References
- 1.↵
- 2.
- 3.↵
- 4.↵
- 5.
- 6.
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.
- 12.
- 13.
- 14.↵
- 15.↵
- 16.↵
- 17.
- 18.↵
- 19.↵
- 20.
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.
- 74.
- 75.↵
- 76.↵
- 77.
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵