Mask, a component of the Hippo pathway, is required for Drosophila eye morphogenesis

Abstract Hippo signaling is an important regulator of tissue size, but it also has a lesser-known role in tissue morphogenesis. Here we use the Drosophila pupal eye to explore the role of the Hippo effector Yki and its cofactor Mask in morphogenesis. We found that Mask is required for the correct distribution and accumulation of adherens junctions and appropriate organization of the cytoskeleton. Accordingly, disrupting mask expression led to severe mis-patterning and similar defects were observed when yki was reduced or in response to ectopic wts. Further, the patterning defects generated by reducing mask expression were modified by Hippo pathway activity. RNA-sequencing revealed a requirement for Mask for appropriate expression of numerous genes during eye morphogenesis. These included genes implicated in cell adhesion and cytoskeletal organization, a comprehensive set of genes that promote cell survival, and numerous signal transduction genes. To validate our transcriptome analyses, we then considered two loci that were modified by Mask activity: FER and Vinc, which have established roles in regulating adhesion. Modulating the expression of either locus modified mask mis-patterning and adhesion phenotypes. Further, expression of FER and Vinc was modified by Yki. It is well-established that the Hippo pathway is responsive to changes in cell adhesion and the cytoskeleton, but our data indicate that Hippo signaling also regulates these structures.


INTRODUCTION
Since the final structure of any cell in a tissue is the outcome of the mechanical constraints and forces placed on that cell, adhesive junctions established with its neighbors, and the structure of the cell's internal cytoskeleton, understanding the mechanisms that regulate these elements in a developing or mature tissue is important. Recently, a number of studies in vertebrates have suggested that Hippo signaling can modify or respond to these aspects of cell anatomy and is therefore important in tissue and organ morphogenesis (Zheng and Pan, 2019). However, since the pre-dominant role for Hippo signaling is regulation of cell proliferation and survival (Boopathy and Hong, 2019;Misra and Irvine, 2018;Watt et al., 2017), clarifying the contribution of Hippo to tissue morphogenesis is challenging. To circumvent this issue we utilize the Drosophila pupal eye as a model since it is post-mitotic and, in addition, becomes refractive to apoptosis (Cagan and Ready, 1989b;Wolff and Ready, 1991a;Wolff and Ready, 1991b). Hence, in the fly eye we can examine more precisely the role of Hippo signaling in morphogenesis, independent of its function in tissue growth.
Mask (multiple ankyrin repeats single KH domain) is a 423 kDa protein that contains two ankyrin repeat domains (suggesting a scaffolding role) and a single K-homology domain (that may mediate interactions with nucleic acids) that was first identified during a screen for novel receptor tyrosine kinase (RTK) signaling components (Smith et al., 2002). Mask has two mammalian orthologues -Mask1 and Mask2 -and recent studies indicate that Mask family proteins regulate the import of Yki/YAP into the nucleus and are therefore required for their full transcriptional activity (Kwon et al., 2013;Li et al., 2017;Machado-Neto et al., 2014;Sansores-Garcia et al., 2013;Sidor et al., 2019;Sidor et al., 2013). Hence Mask is described as a Yki/YAP cofactor.
Studies that implicate Hippo signaling in tissue morphogenesis have mainly considered this role in vertebrates. For example, YAP is required for nephron development in the mammalian kidney (McNeill and Reginensi, 2017;Reginensi et al., 2016;Reginensi et al., 2013) and urinary tract morphogenesis (Reginensi et al., 2015). YAP also contributes to lung development where it promotes gene expression that is associated with myosin II activation and the generation of tensile forces necessary for branching morphogenesis (Lin et al., 2017). YAP also promotes transcription of genes associated with increased cellular tension in hepatocytes, which correlates with antagonism of adherens junction (AJ) formation (Bai et al., 2016). In contrast, YAP is required for VE-cadherin distribution and correct adhesion during angiogenesis in the mouse brain and retina (Kim et al., 2017a). Hence, mounting evidence indicates that cytoskeletal and junction structures are modified via YAP-mediated transcription in developing tissues and these effects are likely to extend to cancer as well. Indeed, in cancer-associated fibroblasts, ectopic YAP has been shown to modulate the expression of genes that modify actin and myosin structures to promote cell migration (Calvo et al., 2013). Mask1 has similarly been implicated as a regulator of cancer cell migration. For example, reduced expression of mask1 decreased the migration of multiple myeloma cells, hepatocellular carcinoma cells, and colorectal cancer cells in which YAP activity was associated with the transcription of genes that promote epithelial to mesenchymal transition (Dhyani et al., 2015;Yao et al., 2018;Zhou et al., 2019).
The effects of YAP and Mask1 in modifying junction/cytoskeletal structures are suggestive of a feedback loop since these structures are also well-documented modifiers of Hippo signaling. For example, activity of the core AJ components E-cadherin (E-cad) and αcatenin (α-cat) has been linked to activation of the core Hippo kinase LATS1/2 and inhibition of YAP (Kim et al., 2011;Schlegelmilch et al., 2011;Silvis et al., 2011), and in Drosophila the Ig-CAM protein Echinoid that localizes to AJs can promote Salvador/Hpo activity to inhibit Yki (Yue et al., 2012). A complex picture of cytoskeletal regulation of Hippo signaling, mainly via interactions with Wts/LATS or Hpo/MST, is emerging (Seo and Kim, 2018;Zheng and Pan, 2019). F-actin accumulation, elaboration of branched F-actin networks and activation of contractile actin-myosin networks have all been shown to increase nuclear Yki/YAP/TAZ activity (Aragona et al., 2013;Dupont et al., 2011;Fernández et al., 2011;Gaspar et al., 2015;Matsui and Lai, 2013;Rauskolb et al., 2014;Sansores-Garcia et al., 2011;Wada et al., 2011;Zhao et al., 2012). Conversely, activity of the spectrin-based membrane skeleton is considered to antagonize Yki/YAP (Deng et al., 2015;Fletcher et al., 2015).
The functions of Hippo pathway proteins, Yki, and its nuclear interactors have been extensively characterized in proliferative Drosophila tissues. Here, using the post-mitotic Drosophila pupal retina as a model, we found that correct activity of Mask, Yki and Wts is required for the appropriate distribution of AJs and hence eye patterning. In addition, we determined that Mask activity impacts the expression of numerous genes during eye patterning, including many associated with cell adhesion and cytoskeletal organization. Indeed two of these -FER tyrosine kinase (antagonized by Mask activity) and Vinculin (promoted by Mask) -contributed to the correct organization of AJs and eye patterning. We also found that Mask regulates a large number of genes associated with signal transduction and many genes that modify apoptosis to promote cell survival. These latter data emphasize that an entire gene program, rather than a select few genes, is modified by Hippo signaling to determine whether cells survive or die. Taken together, our data underscore a pivotal role for Hippo in epithelial morphogenesis.

Microscopy and image processing
Tissue was imaged with a Leica DM5500 B fluorescence microscope, Leica SP8 confocal microscope or Zeiss LSM510 confocal microscope and associated software. Adult eyes were imaged with a Leica M125 stereo-dissected microscope, Leica IC80HD camera and Leica Acquire version 3.3 software at 6.3X magnification. All adult animals shown are female with the exception of Fig. 1K. Confocal microscopy parameters were identical when imaging control and experimental tissues but for images gathered using standard fluorescence microscopy, imaging parameters were optimized for maximal E-cad detection. Image files were processed for publication using Adobe Photoshop. All images presented are of tissue in the center of retinas. Image tracings were drawn in Adobe Illustrator.

Analysis of density and distribution of E-cadherin at AJs, and retinal mis-patterning
Retinas of different genotypes were prepared and imaged in parallel, with identical conditions. Three independent replicates of each experiment were performed. Maximum projection images spanning the AJs were assembled from confocal Z-stacks and imported into ImageJ and pixel intensity of junctions between lattice and 1° cells located in the center of retinas was determined. Junctions were randomly selected for these analyses. Normalized junctional intensity was calculated as (average pixel intensity of cell junction) -(background pixel intensity), where the latter value was determined as an average of the pixel intensity of the apical cytoplasm in the center of both neighboring cells. For quantification of the distribution of E-cad along at AJs (coverage, %), maximum projection image files were imported into ImageJ and a) the length of randomly-selected AJs between lattice and 1° cells in the center of retinas measured, and b) gaps in E-cad distribution along the AJ measured. Gaps were identified as regions of the AJ with no detected fluorescence (pixel intensity = 0). E-cad coverage (%) was the defined as (1-(sum of all gap lengths/total AJ length))x100. To generate mean ommatidial mis-patterning scores (OMS) a hexagonal grid was superimposed over images of the central region of retinas, as previously described (Johnson and Cagan, 2009). Each hexagon was drawn so that the centers of 6 ommatidia surrounding a central ommatidium were connected. Ommatidia and the surrounding lattice cells and bristle groups within each drawn hexagon (or data point) were scored for defects to generate an OMS. Cone cells defects scored included changes in cell number, orientation, and failure to establish correct contacts between cells. For 1° pigment cells, defects scored included changes in 1° cell number, unequal 1° cell size, failure to establish the 1°-1° cell junction and resulting contacts between cone cells and lattice cells or bristle groups as a result. For 2° and 3° cells (lattice cells) defects scored included changes in lattice cell number and failure to correctly establish the 3° niche. For bristle groups, changes in the number and position of bristle groups were scored. For all analyses, statistical significance was determined using One-Way Analysis of Variance (ANOVA).
qRT-PCR and RNA-sequencing mRNA was prepared, in triplicate, from GMR>lacZ; GMR>GFP RNAi ; GMR>mask RNAi v29541 ; GMR>GFP RNAi , Dcr-2; and GMR>yki RNAi , Dcr-2 retinas using either standard Trizol extraction and reverse transcriptase (Invitrogen, #18090010) or the ReliaPrep RNA Tissue Miniprep System (Promega Corporation, # M3001), as previously described (DeAngelis and Johnson, 2019). Duplicate qRT-PCR analyses were performed using a Step One Plus Real-Time PCR System (Applied Biosystems, # 4376600) with primer sets listed in Table S2. For each gene assayed with qRT-PCR, expression was quantified by determining the threshold cycle for each reaction (C T ) and C T values compared to the housekeeping gene rp49 to generate estimates of relative expression (ΔC T ).
Replicate ΔC T values were then compared to generate estimates of differential expression (ΔΔC T ). Specificity of amplified products generated was confirmed with melt curve analysis and gel electrophoresis. Significant changes in gene expression were determined with twosample two-sided student's t-tests.
The UAS-mask RNAi-v29541 , UAS-GFP RNAi , UAS-GFP, mask EY01848 and GMR-Gal4 lines were isogenized by backcrossing to w 1118 for five generations and then rebalanced. mRNA was isolated in triplicate from 50-70 retinas of GMR>GFP RNAi and GMR>mask RNAi-v29541 ; and GMR>GFP and GMR>mask EY01848 as described (DeAngelis and Johnson, 2019). GMR>GFP RNAi and GMR>mask RNAi-v29541 replicates were dissected on different days than GMR>GFP and GMR>mask EY01848 replicates. Barcoded cDNA library preparation was performed using TruSeq library preparation kits, libraries were pooled and balanced pooling was confirmed using qPCR and paired-end 51bp RNAsequencing were all performed by the University of Michigan Advanced Genomics and Next Generation Sequencing Core. Sequencing reads were imported into Galaxy (https:// usegalaxy.org/) and their quality assessed with FASTQC (Afgan et al., 2018;Andrews, 2010). Bioinformatics processing of sequence data followed the approach of (Lanno et al., 2017). Briefly, sequence reads were aligned to the D. melanogaster reference genome and gene annotation files available at the time of submission (reference genome: Drosophila_melanogaster.BDGP6.dna.toplevel.fa, gene annotation: Drosophila_melanogaster.BDGP6.93.gff3) downloaded from Ensembl (Zerbino et al., 2018) with Bowtie2 (Langmead and Salzberg, 2012) using default parameters. The percentage of mapped reads was determined with Flagstat from SAMtools (Li et al., 2009). Gene expression quantification and differential gene expression statistical analyses were performed using Cuffdiff following geometric normalization and transcript length correction where bias correction was performed using the reference genome sequence (Trapnell et al., 2012). Sequencing reads mapping to the UAS-mask RNAi-v29541 transgene were quantified, to confirm mask reduction, using cufflinks (Trapnell et al., 2010). Statistical comparisons of mask long isoform expression to corresponding controls as well as comparison of reads mapping to the UAS-mask RNAi-v29541 transgene were performed with two-sample two-sided student's t-tests. Gene Ontology analyses was assessed utilizing The Gene Ontology Consortium resources (http://geneontology.org/). Scatterplots and volcano plots ( Figure 5) were generated using R-statistical software (CRAN, 2018). For further information on fly lines, software and other key materials, please see Table S1.

Mask is required for morphogenesis of the Drosophila retina
The fly eye is a neuroepithelium composed of approximately 750 ommatidia (Cagan and Ready, 1989a;Carthew, 2007;Kumar, 2012;Ready et al., 1976;Wolff and Ready, 1993). Each mature ommatidium contains eight photoreceptor neurons, four cone cells and two primary (1°) pigment cells ( Figure 1A). Secondary (2°) and tertiary (3°) pigment cells separate the ommatidia and are precisely positioned to generate an ordered honeycomb-like lattice that spans the eye field ( Figure 1A, B). In addition, each eye contains over 600 sensory bristle groups which are embedded within the interommatidial lattice ( Figure 1A). An antibody to all Mask isoforms detected the protein throughout the pupal eye in numerous cytoplasmic puncta and at AJs ( Figure S1, (Smith et al., 2002)). However, whilst antibodydetection of Mask was abrogated by expression of an RNAi transgene against mask ( Figure  S1C), AJ-localization of Mask, in particular, was not recapitulated in mask GFP-CC00924 nor mask sfGFP-TVPTBF retinas (transgenic lines in which the longer Mask isoforms, or all Mask isoforms are potentially GFP-tagged, respectively; Figure S1D, Figure S2A-C; (Sarov et al., 2016)). Yki was similarly observed in numerous apical and cortical puncta in yki YFP-VK37 and yki sf-GFP-TVPTBF retinas and occasionally also at AJs ( Figure S2D-E). In addition, a small number of Mask and Yki puncta were observed in the nuclei of cone and pigment cells, and photoreceptors ( Figure S2F-Q, and data not shown).
To assess the role of Mask in pupal eye development, we modified its expression using the Gal4/UAS system and Glass Multimer Reporter-Gal4 (GMR-Gal4), which is active in eye tissue after the passage of the morphogenetic furrow that establishes the eye field in the larva (Brand and Perrimon, 1993;Freeman, 1996). Patterning defects in retinas with reduced or increased mask expression were then examined and quantified at 40 h APF (ommatidial mispatterning score (OMS), Figure 1C, Table S3A and B, (Johnson and Cagan, 2009). UASmask RNAi-v29541 , which targets all predicted mask transcripts ( Figure S1D), generated severe mis-patterning phenotypes including errors in the stereotypical arrangement and size of cone cells, incorrect orientation of ommatidia along the dorsal-ventral axis, unequallysized 1° cell pairs, and angular rather than curved boundaries between 1°s and neighboring interommatidial cells ( Figure 1D). The interommatidial lattice was also disorganized with few correctly-shaped 2° and 3° cells: most were trapezoidal or even triangular, and the lengths of many lattice-lattice cell boundaries were reduced. Most bristle groups were also mis-positioned and some were located between 1° cell pairs. As discussed in more depth below, we also observed a marked reduction in the density of apical E-cad in all cells of GMR>mask RNAi-v29541 retinas at 40 h APF ( Figure 1D). When expressed together with UAS-Dcr-2, a second RNAi transgene, mask RNAi-v103411 , generated similar albeit more mild patterning defects including modest mis-orientation of ommatidia, straighter 1°-lattice cell boundaries, and mild disruption to the neat organization of the interommatidial cell lattice ( Figure 1E, Figure S1D). Ectopic mask, induced with mask EY01848 ( Figure S1D) or UASmask RA , also generated mild patterning defects including disruptions to the formation of correctly-shaped 3°s and grouping of interommatidial lattice cells in two or more rows between ommatidia ( Figure 1F-G). In addition, cone cells were occasionally observed in direct contact with lattice cells where 1°s had failed to adhere to each other to fully encircle the cone cell group ( Figure 1F', G'). The adults of each of these genotypes displayed "rough eye" phenotypes that corresponded with the degree of pupal eye mis-patterning or OMS scores ( Figure 1C, H-L; facet disruption in adult eyes was often more pronounced in the posterior eye where the period of transgene expression had been longest).

Mask promotes survival of retinal cells
Since Hippo signaling has a crucial role in regulating cell survival and mitosis, it was not surprising that the number of interommatidial cells in GMR>mask RNAi-v29541 retinas at 40 h APF was reduced from an average of 12.21 about an ommatidium in control GMR>lacZ retinas to 8.28 (p=1.9×10 −34 ) and ommatidia with one rather than two 1° pigment cells were often found in GMR>mask RNAi-v29541 retinas (hereafter simply referred to as GMR>mask RNAi ). These observations were also consistent with the initial description of Mask as a promoter of cell survival (Smith et al., 2002). These defects in cell number did not arise from errors in larval eye development: in GMR>mask RNAi larval retinas we observed no change in the final mitotic division of interommatidial cells, which occurs following the passage of the morphogenetic furrow ( Figure S5A,B) (Ready et al., 1976;Wolff and Ready, 1991a) and we also observed no increase in apoptosis in GMR>mask RNAi larval eye discs ( Figure S5C,D). Photoreceptor recruitment was also unperturbed (data not shown). Hence the ommatidial field is correctly established in GMR>mask RNAi retinas and the mispatterning and cell survival defects arise during pupal eye morphogenesis.
A range of signals contribute to the culling of excess interommatidial lattice cells from the eye, a process that begins at ~17-18 h APF and terminates at around 33 h APF. These include apoptosis-inducing signals (Notch, Wingless and Jun N-terminal kinase (JNK) signaling) (Bushnell et al., 2018;Cagan and Ready, 1989b;Cordero et al., 2004;Wolff and Ready, 1991b) and survival-promoting signals (e.g. Epidermal Growth Factor Receptor (EGFR) signaling) (Monserrate and Brachmann, 2007) which integrate to ensure the correct number of lattice cells remain about each ommatidium. Increased apoptosis was observed in GMR>mask RNAi retinas in comparison to control GMR>lacZ eyes at 18, 21 and 24 h APF ( Figure S6A-C) confirming that Mask promotes lattice cell survival. By 27 h APF, apoptosis in control and GMR>mask RNAi retinas was similar ( Figure S6A,D,E), suggesting that other survival signals (e.g. EGFR) protect the remaining lattice cells in GMR>mask RNAi eyes from apoptosis from this time on. Reducing expression of yki similarly reduced lattice cell number (discussed below) and we conclude that Mask and Yki contribute to survivalpromoting signals that counterbalance apoptosis to ensure appropriate lattice cell number.

Retinal patterning is independent of changes in lattice cell number
We next questioned whether mis-patterning of GMR>mask RNAi retinas was simply a consequence of ectopic apoptosis and, conversely, whether the additional interommatidial cells in GMR>mask EY01848 retinas disrupted lattice organization. Ectopic cell death was triggered by expression of reaper (rpr) (White et al., 1994), but as GMR>rpr was pupal lethal, we used Gal4-54C to express rpr only the lattice cells (Bao et al., 2010). At 40 h APF, 54C>rpr animals had an average of 7.53 lattice cells around an ommatidium (in comparison to 12.17 cells in 54C>lacZ retinas) yet, as long as at least 5 cells surrounded an ommatidium, the cells adopted contorted shapes to generate a honeycomb lattice and hexagonal ommatidia ( Figure 1M,N). Since driving mask RNAi transgenes with Gal4-54C did not sufficiently reduce mask even when co-expressed with Dcr-2, we had to compare lattice patterning in 54C>rpr retinas with that of GMR>mask RNAi eyes which had an average of 8.35 interommatidial cells. In this genotype the lattice was markedly distorted and many ommatidia shaped into pentagons ( Figure 1O). Many abutting ommatidia not separated by lattice cells were also observed ( Figure 1O), a phenotype not observed in 54>rpr retinas until the number of lattice cells about an ommatidium dropped to below 4 (data not shown). Blocking cell death in GMR>mask RNAi retinas via concurrent expression of the cell death inhibitor Diap1 (Hay et al., 1995) increased lattice cells to an average of 14.85 around an ommatidium, but the lattice was still disorganized with many grouped cells ( Figure 1P, green outlines). The ommatidia were seldom neatly hexagonal, often misoriented and unequally-sized 1°-cell pairs were frequent ( Figure 1P). None of these mispatterning phenotypes were observed in GMR>Diap1 retinas ( Figure 1R), which had an average of 23.61 lattice cells that nonetheless were organized into single rows around ommatidia and still generated the hexagonal lattice, although the 2° and 3° pigment cell niches were not always correctly patterned. This contrasted with distortions to the lattice in mask EY01848 retinas ( Figure 1S), where the additional interommatidial cells were arranged in groups rather than in single file, and few 3° cells were correctly established. Mis-oriented ommatidia were also observed in GMR>mask EY01848 retinas ( Figure 1S). Taken together, these data indicate that the patterning defects observed when mask expression is modified are independent of changes in the number of interommatidial cells. Our data also underscore that lattice patterning is a robust process that adapts to variations in the availability of lattice cells.

Yki and Wts are required for retinal morphogenesis, in concert with Mask
Since Mask interacts with Yki (Kwon et al., 2013;Li et al., 2017;Sansores-Garcia et al., 2013;Sidor et al., 2013), we hypothesized that Mask:Yki complexes contribute to pupal eye morphogenesis. When co-expressed with Dcr-2, UAS-yki RNAi-v104524 generated mispatterning phenotypes that were qualitatively similar to retinas with reduced mask expression (Figure 2A-D, Table S3D). Unfortunately yki RNAi-HMS00041 , a transgene commonly used to reduce yki in larval tissues, was pupal-lethal when driven with GMR-Gal4 and yki RNAi-JF03119 generated only very mild eye mis-patterning (not shown). However, in addition to reducing the number of interommatidial cells, yki RNAi-v104524 disrupted the lattice, caused misshapen ommatidia that were often not correctly aligned along the dorsal-ventral axis, 1° cell pairs that were unequally sized and mis-positioned bristles. The adult GMR>yki RNAi-v104524 , Dcr-2 eyes were accordingly 'rough' and similar to those of GMR>mask RNAi adults ( Figure 2E-H). Hence Yki, like Mask, is important for the organization of the Drosophila retina.
Ectopic yki also generated additional grouped lattice cells and other mild patterning defects similar to those in GMR>mask retinas ( Figure 2I). In addition, ectopic yki partially suppressed GMR>mask RNAi mis-patterning phenotypes ( Figure 2J, compare to Figure 2D). However whilst GMR>mask RNAi patterning defects were enhanced in yki heterozygous retinas ( Figure 2L), these effects were mild, as might be expected if little functional Mask, and hence few Mask:Yki complexes, remained in GMR>mask RNAi retinas such that further reducing yki expression had little effect. These changes in patterning defects were reflected in the disorder of the adult eyes and OMS values ( Figure 2H,M-Q, Table S3E). Taken together, our data suggest that Yki and Mask function together to promote eye morphogenesis, although independent roles for Yki are not precluded.
Since Wts is the major negative regulator of Yki, we hypothesized that ectopic wts would cause patterning defects similar to those observed when either mask or yki expression was reduced. Indeed, at 40 h APF, GMR-wts retinas were characterized by numerous mis-placed and incorrectly shaped lattice cells, unequal 1°-cell pairs, and mis-oriented ommatidia ( Figure 2R). Ectopic wts also significantly enhanced GMR>mask RNAi mis-patterning at 40 h APF ( Figure 2S, Table S3F), whilst a null allele of wts significantly reduced GMR>mask RNAi defects ( Figure 2T,U, Table S3F). As before, these genetic interactions were reflected in the disruptions to the adult eye and OMS values ( Figure 2V-Z). In addition, GMR>mask RNAi mis-patterning was partially suppressed in tissue heterozygous for mutant alleles for hippo (hpo), expanded (ex) or merlin (mer) ( Figure S7, retinas heterozygous for these alleles were correctly patterned). These data support that Hippo pathway activity regulates Mask during pupal eye morphogenesis.

Mask regulates AJ distribution and cytoskeletal structures in the retina
To better understand the cause of mis-patterning in retinas with reduced mask expression, we considered the requirement for Mask for correct AJ distribution and density. Loss of function mask clones failed to survive and mask RNAi-v29541 clones (generated using the standard weaker actin-Gal4 driver) had little phenotype. However the dorsal specific mirror-Gal4 driver ( Figure S8A,B) (McNeill et al., 1997) generated a gradient of mask  or mask RNAi-v29541 expression that mildly disrupted distribution of AJs, assessed at 24 h APF ( Figure S8C-H, patterning was also mildly disrupted). Specifically, AJs were not evenly distributed about the entire periphery of 1°s and lattice cells, leaving numerous 'gaps' in E-cad distribution ( Figure S8E,H). Since transgenes were only weakly expressed in mirr>mask RNAi pupae and these rarely survived beyond 24 h APF, we utilized GMR-Gal4 for all further analyses, although this restricted our comparisons to between retinas.
In wild type (or control) eyes, early patterning is characterized by local cell rearrangements and changes in cell shape and size, and coincident with this, AJs are not uniformly distributed ( Figure 3A-D) (Johnson, 2020). This likely reflects AJ remodeling or the formation of nascent junctions that have not yet been stabilized (Guillot and Lecuit, 2013). Reducing mask significantly increased the frequency and persistence of gaps in E-cad/AJ distribution, although this became less obvious at junctions between 1° cell pairs from 27 h APF ( Figure 3E-J). In addition, significantly less E-cad was detected at AJs in GMR>mask RNAi-v29541 retinas at 24 and 40 h APF ( Figure 1D, 3K). We conclude that Mask is critical for establishing, securing or maintaining AJs.
Similar defects in AJ organization were observed in GMR>yki RNAi-v104524 , Dcr-2 and GMR-wts retinas ( Figure 4A-D). In addition, whilst yki alleles failed to modify mask RNAi induced AJ disruption, ectopic yki significantly rescued these defects ( Figure 4E-J,O). A wts null allele similarly rescued AJ organization in GMR>mask RNAi retinas and AJ defects were severely augmented by ectopic wts ( Figure 4K-N,P). These data suggest that Hippo pathway activity must be correctly controlled for the appropriate regulation of adhesion during eye morphogenesis.
Accordingly, adhesion defects generated when mask, yki or wts expression were modified could account for retinal disorder since compromising AJ dynamics or stability would impair morphogenetic processes necessary to position and shape retinal cells (Figures 1, 2). Indeed, directly targeting AJs by reducing E-cad expression disrupted eye patterning ( Figure  S9), but phenotypic similarities between GMR>E-cad RNAi , GMR>mask RNAi , GMR>yki RNAi and GMR-wts retinas were limited. Specifically, reducing E-cad led to lattice cells that were poorly organized, but in GMR>E-cad RNAi retinas, 1° cell pairs were generally equal in size and most boundaries between 1° and lattice cells were curved rather than straight, as frequently observed in GMR>mask RNAi retinas. These data argue that morphogenetic defects in retinas with less Mask or Yki activity do not originate only from changes in AJ organization.
Given the functional importance of interactions between AJs and the cytoskeleton, we next examined actin and myosin structures in GMR>mask RNAi retinas. At both 24 and 40 h APF, the density of F-actin greatly increased ( Figure S10A-E), and the accumulation of nonmuscle myosin II (NMII) decreased (Figure S10F-J) when mask expression was reduced. In control 1° cells, the apical actin cytoskeleton is strikingly organized into numerous F-actin structures that appear to tile across the cells' width at right angles to the 1° cell-lattice cell interface by 40 h APF ( Figure S10C). The functional importance of these F-actin structures has not been explored but we note that in GMR>mask RNAi retinas they were entirely disrupted and 1° cells were seldom correctly sized or shaped, suggesting a role in determining or maintaining cell architecture ( Figure S10D). In control 1°s, NMII accumulated along the 'concave' surfaces at 1°-lattice cell interfaces at 40 h APF ( Figure  S10H), correlating with a model where myosin-mediated contractility contributes to the rounded shape of 1°s. Accordingly, in GMR>mask RNAi retinas approximately equal accumulation of NMII in abutting 1°-lattice cell neighbors could account for the straighter form of these cell interfaces ( Figure S10I). NMII puncta also accumulated through the cytoplasm of control lattice cells at 40 h APF ( Figure S10H), but not in GMR>mask RNAi lattice cells ( Figure S10I), which were also marked by a striking increase in F-actin ( Figure  S10D). We predict that these disruptions to the cytoskeleton contribute to the irregular cell shapes in GMR>mask RNAi retinas and propose that Mask is an important regulator of the cytoskeleton in the eye.

Mask regulates genes associated with adhesion and the cytoskeleton
To identify genes regulated by Mask that contribute to retinal morphogenesis, we used RNAsequencing to assess genome-wide gene expression at 24 h APF in GMR>GFP RNAi , GMR>mask RNAi , GMR>GFP and GMR>mask retinas ( Figure S11). This generated 5.22×10 8 sequence reads, with a range of 35,255,405 to 48,935,181 mapped reads per sample (with 92.12 to 93.29% reads mapping to the genome), suggesting that we had appropriate read-depth for confident quantification of most of the expressed genome (Table  S5). Reducing mask resulted in significant changes in the expression of 1674 genes (Table  S6, Figure 5A,B), and in tissue with ectopic mask, 255 genes were significantly differentially expressed (Table S7, Figure 5A,B). Expression of 129 loci was significantly modified in response to both reduced and increased mask. In addition, the expression of twelve known targets of Yki or Hippo pathway activity changed in GMR>mask RNAi , although these changes were not all statistically significant (Table S8).
Gene Ontology (GO) analyses revealed that Mask regulates genes involved in a variety of biological processes (Table S9) including genes associated with adhesion or with roles in the actin cytoskeleton (Tables S6 and S9), although expression of core AJ components (shotgun, which encodes Drosophila E-cad, and the Catenin proteins) was not significantly changed. Further, ectopic E-cad failed to rescue GMR>mask RNAi mis-patterning, which was also not modified in retinas heterozygous for E-cad (shg) ( Figure S12, Table S3G). Hence, we conclude that rather than regulating transcription of core AJ proteins, Mask instead contributes to mechanisms that influence AJ assembly or stability. Amongst the loci that could mediate this were FER tyrosine kinase (FER) and Vinculin (Vinc) which were repressed and promoted in the presence of Mask, respectively (Tables S6, S10). FER has been implicated in the phosphorylation and degradation of β-catenin (Murray et al., 2006;Piedra et al., 2003;Rosato et al., 1998) and Vinc contributes to AJ formation and is recruited to AJs in response to mechanical stress (Galbraith et al., 2002;Le Duc et al., 2010;Leerberg et al., 2014;Opazo Saez et al., 2004;Taguchi et al., 2011). We discuss our initial investigations into the roles of these loci during retinal morphogenesis in more detail below.
Amongst the transcriptional changes detected in GMR>mask RNAi retinas that could account for disruptions to F-actin structures were Abelson interacting protein (Abi) (log 2 fold change in expression = 0.42), which regulates actin dynamics through the WASP and WAVE complexes (Bogdan et al., 2005); washout (wash) (log 2 fold change = 0.63), which crosslinks F-actin and microtubules and is required for maintaining the actin cytoskeleton in the ovary (Liu et al., 2009); and RhoGEF3 (log 2 fold change = −0.40) and RhoGEF4 (log 2 fold change = 0.56), which have been implicated in activating Rac1 (Nakamura et al., 2017) and RhoA (Nahm et al., 2006). Changes in expression of Shroom (log 2 fold change = −0.41), which acts through Rho-kinase to promote NMII activity (Nishimura and Takeichi, 2008), may account for the reduced accumulation of NMII in GMR>mask RNAi retinas ( Figure S10G,I).
qRT-PCR confirmed that Vinc, FER, Abi and wash expression was similarly regulated by Yki and Mask in pupal retinas ( Figure 5C), suggesting that these loci are regulated by Mask:Yki complexes, although this regulation may be indirect. However, whilst Mask promoted Shroom expression, it was antagonized by Yki ( Figure 5C), suggesting independent roles for Mask and Yki in Shroom regulation.

Repression of FER downstream of Mask is essential for eye morphogenesis
FER expression significantly increased in GMR>mask RNAi retinas (log 2 fold change = 0.34, Table S6) and decreased, although not significantly, in GMR>mask retinas (log 2 fold change= −0.19, Table S7). Because FER has been implicated in β-catenin phosphorylation and consequent degradation (Murray et al., 2006;Piedra et al., 2003;Rosato et al., 1998), we hypothesized that elevated FER would contribute to AJ disruption when mask was reduced. Accordingly ectopic expression of FER generated discontinuous distribution of Ecad in retinal cells at 24 h APF similar to those observed in GMR>mask RNAi tissue ( Figure  6A-B). Ectopic FER also amplified errors in AJ distribution in GMR>mask RNAi retinas ( Figure 6C-E), increased the number of patterning errors observed by 40 h APF ( Figure 6F-J, Table 3H), and enhanced the consequent roughness of the adult eye ( Figure 6K-N, Table  3H). In addition, GMR>FER retinas were characterized by mild disorganization of the lattice and patterning errors commonly observed in retinas with reduced mask (Figure 6G), although additional lattice cells were also common, possibly due to ectopic Wg signaling consequent to reduced β-catenin (Chen et al., 2014). In the pupal retina, Wg activity contributes to apoptosis of excess lattice cells (Cordero et al., 2004;Lin et al., 2004) and, accordingly, reducing FER expression led to the occasional missing lattice cell and also generated mild patterning defects that qualitatively resembled phenotypes observed in GMR>mask retinas ( Figure S13, Table S3I). Further, disruptions to AJs in GMR>mask RNAi retinas were mainly suppressed in tissue also heterozygous for FER X21 ( Figure 6E,O-P) and mis-patterning was significantly suppressed and the adult eye relatively undisrupted ( Figure  6Q-T, Table S3H). Taken together, our data indicate that suppression of FER downstream of Mask activity is essential for correct AJ distribution and eye morphogenesis and ectopic FER contributes to patterning defects in GMR>mask RNAi retinas.

Vinculin is an effector of Mask during eye patterning
Expression of Vinc was significantly reduced in GMR>mask RNAi retinas (log 2 fold change = −0.73, Table S6). Given the role of Vinc in fortifying AJs (Huveneers et al., 2012;Taguchi et al., 2011;Yonemura et al., 2010), we then tested the hypothesis that Vinc was amongst the genes promoted by Mask that favored AJ stabilization during eye morphogenesis. Indeed, ectopic Vinc partially rescued defects in AJ distribution in GMR>mask RNAi retinas at 24 h APF ( Figure 7A-E) and this correlated with significantly fewer patterning defects at 40 h APF ( Figure 7F-J, Table S3J) and improved organization of the adult eye ( Figure 7K-N).
Conversely, in GMR>mask RNAi retinas also heterozygous for Vinc, gaps in AJs were wider and more frequent at 24 h APF ( Figure 7E,O-P), and mis-patterning defects were modestly enhanced at 40 h APF and in adults ( Figure 7J, Q-T, Table S3J). Patterning analyses (Table  S3J) also suggested a greater requirement for Mask-Vinc function in 1° cells than in lattice or bristle cell groups. Specifically, in GMR>mask RNAi retinas, errors in the number of 1°s (two per ommatidium) occurred with a frequency of 0.16 (SD=0.37), and ectopic Vinc reduced this to 0.04 (SD=0.20) whilst in Vinc heterozygotes this frequency increased to 0.23 (SD=0.43). Further, in GMR>mask RNAi retinas the junctions between 1° pairs were compromised at a frequency of 0.04 (SD=0.20) and 1° pairs remained 'open', leaving cone cells in contact with neighboring lattice or bristle cells, at a frequency of 0.04 (SD=0.20). These phenotypes were not modified by ectopic Vinc but in Vinc heterozygotes the frequency of shorter or disrupted 1°:1° junctions increased to 0.19 (SD=0.40) and 'open' 1° pairs were present at a frequency of 0.27 (SD=0.61). Taken together, our data allude to an important role for Vinculin in the formation of stable junctions, especially between neighboring 1°s, during eye patterning.

Mask regulates a set of genes that promote cell survival
Expression of the Yki target Diap1 is commonly used to assess Hippo pathway activity and considered central to Yki's role in limiting apoptosis (Huang et al., 2005). However, our transcriptome analyses detected only modest reduction in Diap1 expression (log 2 Fold change = −0.05) in GMR>mask RNAi retinas (Table S6, S8). This observation is consistent with a previous study where Diap1 expression was not reduced in larval eye discs despite increased Hpo, Sav or Wts activity (Verghese et al., 2012). Instead we identified changes in expression of a large number of other genes associated with apoptosis or cell survival in GMR>mask RNAi retinas (Table S11). These included significant increases in the expression of core components of the apoptosis machinery, including grim, rpr and Death regulator Nedd2-like caspase (Dronc, which conversely has previously been shown to decrease when Yki was activated (Verghese et al., 2012)). In addition, we detected gene expression changes that would modify signaling pathways associated with apoptosis or survival of lattice cells in the pupal eye. These changes included increased expression of wingless and its receptor frizzled and modified expression of multiple components of the Notch signaling pathway (Table S11). Both Notch and Wingless signaling promote apoptosis of lattice cells (Cagan and Ready, 1989b;Cordero et al., 2004;Lin et al., 2004;Miller and Cagan, 1998). Further, expression of Egfr, which promotes retinal cell survival (Domínguez et al., 1998;Freeman, 1996;Miller and Cagan, 1998), was decreased (Table S11). Hence, a comprehensive set of cell-death and survival factors are regulated downstream of Mask during pupal eye morphogenesis, underscoring a broad role for Mask in promoting cell survival.

Diverse signal transduction pathways are modified by Mask
Our RNA-sequencing data revealed that an array of signaling pathway components are modified by Mask activity in the retina (Table S12), although many of these changes may be indirect or reflect interactions between signaling networks. Nonetheless, it is striking that expression of multiple components of the Hedgehog, Notch, RTK, TGFβ, Toll and Wnt signaling pathways were modified in GMR>mask RNAi retinas, as well as numerous GPCRs (Table S12). It is plausible that the transcriptional changes in the EGF-Receptor, and other RTK components, identified in our analyses account for the initial description of Mask as a modifier of RTK signaling (Smith et al., 2002). Given the importance of RTK signaling in the Drosophila pupal eye (Malartre, 2016), these transcriptional changes would contribute, no doubt, to the complexity of the patterning defects observed when mask was reduced during eye morphogenesis. We also observed changes in the expression of several components of the planar cell polarity (PCP) system (diego (dgo), Van Gogh (Vang), frizzled (fz) and fat (ft)), which could account for the disrupted orientation of many ommatidia in GMR>mask RNAi retinas.
In addition to its role in PCP, Fat also functions to modify Hippo signaling, as does crumbs, which was also expressed at lower levels in GMR>mask RNAi retinas (Table S12). The expression of two transcription factors that complex with Yki, Mothers against dpp (Mad) and scalloped (sd), was also modified in GMR>mask RNAi tissue (log 2 fold change= 0.37, (Table S6 and S12); and −0.29, (Table S6)). Hence, multiple feedback loops appear to be triggered by Mask in the retina to transcriptionally modulate Hippo pathway activity.
Further studies are required to clarify whether the signaling pathways and networks modified by Mask function in specific retinal cell types or throughout the eye. For example, we note that several Semaphorins as well as roundabout 1 (robo1) were modified by Mask (Table S12). Given the role of these gene families in axon guidance, it is plausible that these function in the organization of axons projected by photoreceptor or bristle neurons (Hu and Zhu, 2018;Seiradake et al., 2016).

DISCUSSION
Drosophila epithelia have been used extensively to characterize the role of Hippo signaling in tissue growth (Irvine and Harvey, 2015;Snigdha et al., 2019), but here we describe Hippo as a major contributor to epithelial morphogenesis. In assessing the contribution of Yki and its cofactor Mask to tissue morphogenesis and cell architecture, we used an approach that modified their activity mainly after the eye field and photoreceptors were established and mitosis had ceased. Hence, we avoided modifying Hippo pathway activity early in eye development, which profoundly alters cell proliferation and also severely modifies activity of the retinal determination gene network to perturb early eye patterning, photoreceptor selection, and eye size (Wittkorn et al., 2015).
Using RNA-seq, we identified many genes that require Mask activity for their correct expression, although these expression changes were captured in whole retinas at 24 h APF and additional investigations are required to determine which expression changes are cellspecific (pigment cells, photoreceptors, neurons and support cells of the bristle groups; Table  S6 and S9). We modified mask for our transcriptional analyses rather than yki because, in our hands, the available RNAi transgenes that target mask were more effective. Indeed, we detected both Mask and Yki in most retinal cell nuclei (although sparsely, Figure S2) and Mask has previously been shown to promote transcription of Yki target genes, possibly via regulating nuclear localization of Yki (Sansores-Garcia et al., 2013;Sidor et al., 2019;Sidor et al., 2013). Accordingly, several loci already identified as Yki/YAP/TAZ targets were also modified by Mask (Table S8). That some of these Yki targets were not significantly modified in our experimental set-up may reflect differences in the transcriptional potential of post-mitotic versus mitotic tissues. Further, our qRT-PCR analyses confirmed that several loci we identified -FER, Vinc, Abi and wash -were similarly modified by Mask and Yki ( Figure 5C). Hence we predict that many transcriptional changes we identified downstream of Mask (Tables S6 and S7) were consequent to modified Yki activity. Of course some are likely to be Yki-independent. For example, we found that Shroom expression, which required Mask, was instead potentially suppressed by Yki ( Figure 5C). We also note that Mask has been identified as a modifier of splicing (Brooks et al., 2015) and expect that loss of this function contributed complexity to the transcriptional changes and patterning defects in GMR>mask RNAi retinas.
In particular, our in vivo data emphasize that Mask, Yki and Wts promote AJ assembly or stability in retinal cells (Figure 3, 4). We also found that Mask is essential for the organization of actin and NMII at both 24 and 40 h APF ( Figure S10). Specifically, Mask antagonizes F-actin and promotes NMII accumulation. These data are consistent with the work of Kim and colleagues, who found that YAP is required during angiogenesis in the murine retina and brain for maintaining VE-cadherin levels and distribution and actin/ myosin organization (Kim et al., 2017a). In contrast, several studies have shown that YAP or Mask1 antagonize cell adhesion and promote cell migration (Bai et al., 2016;Calvo et al., 2013;Dhyani et al., 2015;Yao et al., 2018;Zhou et al., 2019), but it is plausible that these inconsistencies reflect tissue or context-specific outcomes for Mask/Mask1 and Yki/YAP activities.
Of the many adhesion-related genes modified by Mask activity (Table S6, 10), we chose to focus on two for immediate validation. We found that Mask activity reduced FER expression, and since FER phosphorylates β-Catenin to promote its degradation (Murray et al., 2006;Piedra et al., 2003;Rosato et al., 1998), we predict that the excess FER present in GMR>mask RNAi retinas leads to rapid turnover of AJs, contributing to reduced AJ density and errors in AJ distribution ( Figure 6). In contrast, Mask promoted Vinc expression. Since Vinc has an established role in fortifying connections between the Catenins and the actin cytoskeleton when AJs are subject to mechanical stress (Bershadsky et al., 2003;Galbraith et al., 2002;Huveneers et al., 2012;Opazo Saez et al., 2004;Taguchi et al., 2011;Yonemura et al., 2010), we expect that reduced Vinc expression in GMR>mask RNAi retinas compromised this response (Figure 7). However, these hypotheses require validation.
Diverse studies have established that changes in adhesion and the actin-myosin cytoskeleton, can profoundly modify YAP/Yki activity (Aragona et al., 2013;Calvo et al., 2013;Dupont et al., 2011;Fernández et al., 2011;Gaspar et al., 2015;Kim et al., 2011;Matsui and Lai, 2013;Rauskolb et al., 2014;Sansores-Garcia et al., 2011;Schlegelmilch et al., 2011;Silvis et al., 2011;Zhao et al., 2012). However, our examination of Mask and Yki in the pupal eye, as well studies that examined YAP's role in cell invasion/migration and adhesion, identified many genes associated with actin, myosin and adhesion as transcriptional targets of Hippo activity (Bai et al., 2016;Calvo et al., 2013;Kim et al., 2017a;Lin et al., 2017;Yao et al., 2018). Hence, Hippo signaling appears to utilize a feedback mechanism to coordinate transcriptional and cytoskeletal/junction activities. Indeed, the expression of numerous Hippo pathway proteins was also modified in GMR>mask RNAi retinas, including crb, ft and sd (expression of these three loci was significantly decreased) and mad (expression significantly increased), hinting at multiple opportunities for feedback regulation of Hippo signaling by Mask. Crumbs has previously been identified as a transcriptional target of Yki (Genevet et al., 2009;Zhu et al., 2015b).
Not all feedback between Hippo pathway activity and the cytoskeleton or adhesion is mediated through changes in gene expression. Indeed, Wts directly impacts actin polarization in border cells of the fly ovary via phosphorylation of Enabled, to promote border cell migration (Lucas et al., 2013). In addition, cytoplasmic Yki has been shown to interact with Strn-Mlck to promote NMII accumulation and activation at the apical cortex of cells in the Drosophila larval wing disc, contributing to the generation of tensile forces in this tissue . It is therefore very plausible that the dense apical pool of Mask and Yki we identified in Drosophila retinas ( Figure S1 and S2) similarly contributes to apical myosin-structures and hence cortical tension in this tissue. Indeed, the disruption of NMII accumulation that we observed in GMR>mask RNAi retinas could reflect this role (Figure S10F-I). Although not recapitulated in mask GFP fly lines, we also detected Mask at AJs using a Mask antibody ( Figure S1). Similarly, we observed a subset of Yki at AJs ( Figure S2), consistent with the maintenance of inactive YAP at AJs via interactions with 14-3-3 and α-catenin (Schlegelmilch et al., 2011). It is plausible then that a subset of Mask is maintained in complexes with Yki at AJs, but this hypothesis, as well as the role of Mask at this location, remains to be tested.
Our RNA sequencing analyses also identified a large number of genes that regulate cell death that are modified downstream of Mask activity (Table S11). These included core components of the apoptosis machinery (Denton and Kumar, 2015) including Dronc, grim, Dark and rpr, which were expressed at higher levels in GMR>mask RNAi retinas. Expression of Diap1, which is an established target of Yki (Huang et al., 2005) was modestly reduced in these retinas (Table S8). We also found changes in transcription that would enhance signaling pathways that promote apoptosis in the fly eye (eg. Wg signaling) (Cordero et al., 2004) and impede those that protect cells from death (eg. EGFR signaling) (Miller and Cagan, 1998;Monserrate and Brachmann, 2007). Taken together, these transcriptional changes demonstrate and account for the powerful impact of Hippo pathway activity in regulating cell survival.
Inevitably, many of the genes identified in our RNA-sequencing analyses may not be direct targets of Mask or Mask:Yki transcriptional complexes but instead targets of the signaling pathways modified by Mask (including Hedgehog, Notch, RTK, TGFβ, Toll, and Wnt signaling pathways; Table S12). Indeed, Hippo signaling has also been shown in other systems to facilitate transcription of components of the Notch, EGFR, and JAK-STAT pathways (Ren et al., 2010a;Yu et al., 2008) and, perhaps not surprisingly, significant crosstalk between Hippo and other signaling pathways has been described (Kim et al., 2017b;Polesello and Tapon, 2007;Reddy and Irvine, 2013). These signaling networks surely add further complexity to the role of Hippo signaling in tissue morphogenesis.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.

ACKNOWLEGEMENTS
We thank our reviewers for helpful comments on our work, and the BDSC (NIH P400D018537), the VDRC (Dietzl et al., 2007), Iswar Hariharan, Mike Simon, Ken Irvine, Ulrich Tepass, Chunlai Wu, Richard Fehon, Cathie Pfleger and Nick Brown for fly lines or antibodies. We also thank Arielle Ashley, Redwan Bhuiyan and Kayla Jaikaran for technical assistance, and Cathie Pfleger, Michael Weir and members of the Johnson and Coolon Labs for helpful discussion. Lucas Coolon provided artistic assistance in matching colors in Figure 5. This work was supported by R15GM114729.

Manuscript Highlights:
• Mask functions within the context of Hippo signaling to regulate morphogenesis of the Drosophila pupal eye.
• Appropriate accumulation and distribution of Adherens Junctions requires Mask and Yki.
• Numerous genes associated with adhesion, the cytoskeleton, and tissue morphogenesis are responsive to Mask in the pupal eye.
• Correct expression of FER and Vinc, which are downstream of Mask, is essential for the accumulation of Adherens Junctions and hence pattern formation.  representative eyes of adults of these genotypes. See Table S3D  (Q) Mean OMS analyses at 40 h APF, for indicated genotypes. See Table S3E for further analyses of mis-patterning. (R) A retina heterozygous for GMR-Gal4 and a GMR-wts transgene, and (S) in addition with mask RNAi-v29541 expression. (T) A retina heterozygous for wts X1 and GMR-Gal4 or (U) in addition mask RNAi-v29541 . (V)-(Y) Representative eyes of adults of genotypes (R)-(U). (Z) Mean OMS analyses at 40 h APF, for indicated genotypes. See Table S3F for further analyses of mis-patterning. For panels (Q) and (Z), given the goal of testing modification of patterning in GMR>mask RNAi-v29541 retina when yki or wts expression was modified, significant changes in only these data are indicated. * denotes p-value < 0.1; ** denotes p-value < 0.01; ns = not significant. Abutting ommatidia are indicated with red lines; yellow * denote ommatidia missing 1°s; all other annotations as described in Figure 1. E-cad-imaging was optimized and images processed so that patterning defects could be scored.  Table S4B. (K) Quantification of amount of E-cad at AJs in GMR>lacZ and GMR> mask RNAi-v29541 retinas at 24 and 40 h APF. For N and p-values see Table S4C. In (J) and (K) all p-values were < 0.1 with the exception of the difference between E-cad coverage between 1° cells at 27 h APF (J). Error bars reflect standard error. Ommatidia at 24 h APF expressing (A) lacZ and Dcr-2, (B) yki RNAi and Dcr-2, or (C) ectopic wts and (D) quantification of AJ distribution; all p-values were < 0.1. For N and pvalues see Table S4D. Ommatidia at 24 h APF heterozygous for (E) GMR-Gal4 or expressing (F) mask RNAi-v29541 , (G) yki, and (H) yki and mask RNAi-v29541 , heterozygous for (I) GMR-Gal4 and yki b5 , or (J) in addition with mask RNAi-v29541 expression, heterozygous for (K) GMR-Gal4 and a GMR-wts transgene, and (L) in addition with mask RNAi-v29541 expression, heterozygous for (M) wts X1 and GMR-Gal4, and (N) in addition mask RNAi-v29541 . (O) Quantification of E-cad/AJ distribution for genotypes (E)-(J). For N and p-values see Table S4E. (P) Quantification of E-cad/AJ distribution for genotypes (E), (F),(K)-(N). For N and p-values see Table S4F. Given the goal of testing modification of AJ distribution by yki and wts in GMR>mask RNAi-v29541 retinas, significant changes in only these data are indicated in O and P. * denotes p-value < 0.1; ** denote p-value < 0.05; ns = not significant. Error bars reflect standard error. E-cad was enhanced in all images presented so that inconsistencies in AJ distribution can be observed. Orange arrows indicate examples of gaps in E-cad detection. (A) Scatterplots of gene expression (FKPM = fragments per kilobase per million reads) in retinas at 24 h APF in which mask expression was reduced (left) or increased (right), in comparison to control retinas. Yellow and green points indicate loci that were significantly differentially expressed when mask was modified (q < 0.05). Expression of 1674 genes was modified in GMR>mask RNAi-v29541 retinas and 255 in GMR>mask retinas, with 129 of these loci common to both data sets (see inset Venn diagram at right). (B) Volcano plots comparing significance (−log 10 (q-value)) with the magnitude of expression change when Ommatidia at 24 h APF (A) heterozygous for GMR-Gal4, (B) with ectopic FER P100 , (C) mask RNAi-v29541 or, (D) FER P100 and mask RNAi-v29541 . As before, tissue was imaged with identical confocal settings, but the images presented enhanced for better visualization of AJs. (E) Quantification of AJ (E-cad) distribution in retinas at 24 h APF. For N and p-values see Table S4G.  Table S3H for detailed analyses. Given the goal of testing modification of GMR>mask RNAi-v29541 by FER, significant changes in only these data are Ommatidia at 24 h APF in retinas (A) heterozygous for GMR-Gal4, (B) with ectopic Vinc, (C) mask RNAi-v29541 , and (D) with ectopic Vinc and mask RNAi-v29541 . As before, eyes were imaged with identical confocal settings, but image panels enhanced for better visualization of AJs distribution. (E) AJ distribution at 24 h APF. For N and p-values see Table S4H.  Table S3J for detailed analyses. Given the goal of testing modification of GMR>mask RNAi-v29541 by Vinc, significant changes in only these data are indicated in (E)