The ZO-1 protein Polychaetoid as an upstream regulator of the Hippo pathway in Drosophila

The generation of a diversity of photoreceptor (PR) subtypes with different spectral sensitivities is essential for color vision in animals. In the Drosophila eye, the Hippo pathway has been implicated in blue- and green-sensitive PR subtype fate specification. Specifically, Hippo pathway activation promotes green-sensitive PR fate at the expense of blue-sensitive PRs. Here, using a sensitized triple heterozygote-based genetic screening approach, we report the identification of the single Drosophila zonula occludens-1 (ZO-1) protein Polychaetoid (Pyd) as a new regulator of the Hippo pathway during the blue- and green-sensitive PR subtype binary fate choice. We demonstrate that Pyd acts upstream of the core components and the upstream regulator Pez in the Hippo pathway. Furthermore, We found that Pyd represses the activity of Su(dx), a E3 ligase that negatively regulates Pez and can physically interact with Pyd, during PR subtype fate specification. Together, our results identify a new mechanism underlying the Hippo signaling pathway in post-mitotic neuronal fate specification.


Introduction
Generating neuronal diversity during the development of a sensory organ is a prerequisite for the organ to perceive and discriminate various external stimuli. For example, the perception of color relies on comparing the outputs of multiple light-sensing photoreceptor (PR) subtypes with different spectral sensitivities [1][2][3]. During development, the fate of sensory neurons is progressively restricted toward terminal differentiation, finally generating diverse neuronal subtypes. Although the role of transcriptional regulations during neuronal terminal differentiation has been extensively studied [4,5], the details of how specific signaling pathways influence this process are not well understood. Here we use the blue-and green-sensitive PR binary fate decisions in the Drosophila eye as a model to understand the role of the Hippo pathway in post-mitotic neuronal terminal differentiation.
The Drosophila eye is a powerful model to understand the principles of neuronal development [1,[6][7][8]. The Drosophila compound eye contains~750 individual units, ommatidia, each of which consists of eight PRs: the outer PRs R1-R6 and the inner PRs R7-R8 (Fig 1A). There are two main subtypes of ommatidia: pale (p) and yellow (y) ommatidia, present in the adult Drosophila eye (Fig 1B). The outer photoreceptors R1-R6 in both p and y ommatidia express the broad spectrum light sensitive opsin Rhodopsin 1 (Rh1) and are responsible for dim light vision and motion detection. However, the inner R7 and R8 cells express Rhodopsins with different spectral sensitivities, making them capable of performing color vision [9]. In p ommatidia, R7s express UV-sensitive Rh3 and R8s express blue-sensitive Rh5, while in y ommatidia, R7s express UV sensitive Rh4 and R8s express green-sensitive Rh6 (Fig 1B). The p and y subtypes are randomly distributed throughout the retina in roughly a 35:65 p:y ratio [10]. The p vs. y fate decision is first made in R7s via the stochastic activation of the transcription factor Spineless in yR7s during mid-pupation [11]. R7s that do not express Spineless (i.e. the pR7s) instruct their underlying R8s to adopt pale R8 (pR8) fate through Activin and BMP signaling [12]. R8s that do not receive the pR7 signals default to yellow R8 (yR8) fate [13,14]. The effectors involved in p vs. y R8 fate in R8s involve two proteins-Melted (Melt) and Warts (Wts) [15] (Fig 1C). melt encodes a pleckstrin homology domain-containing protein [16], while wts encodes a serine/threonine kinase that is a core component in the Hippo signaling pathway [17][18][19]. melt expression is activated in pR8s by the pR7-driven Activin and BMP signals and its expression leads to the transcriptional repression of wts. Conversely, wts represses melt expression in yR8s by suppressing the activity of the transcriptional coactivator Yorkie (Yki), the downstream effector of the Hippo pathway. Yki is necessary for melt expression in pR8s. Therefore, wts, melt and Yki form a double negative regulatory loop to ensure pR8 vs. yR8 subtype fate decision (Fig 1C) [10]. Yki, together with its DNA-binding partner Scalloped (Sd), regulates the output of the regulatory loop to promote the expression of blue-sensitive Rh5 and prevent the expression of green-sensitive Rh6 [20].
The Hippo pathway was originally discovered in Drosophila for its pivotal roles in tissue growth and organ size control [21]. Its critical and conserved roles in mammals have been subsequently identified in a wide range of biological processes, including stem cell regeneration and homeostasis, innate immune biology, cell differentiation, as well as tumorigenesis [19,[22][23][24]. The components of the Hippo pathway can be classified into three categories: the core kinase complex, the downstream effectors and the upstream regulators. The core kinase complex contains the kinases Wts [17,18], Hippo (Hpo) [25][26][27][28][29], and Mob as tumor suppressor (Mats) [30], as well as the scaffold protein Salvador (Sav) [31,32]. Hpo phosphorylates Wts, affiliated by Sav and Mats, and Wts phosphorylates and inhibits the ability of Yki to enter the nucleus [33][34][35].
Taking advantage of the pR8 and yR8 binary fate assay, we generated and carried out a sensitive and efficient genome-wide screening to identify the new regulators of the Hippo pathway. We identified the Drosophila ZO-1 protein Pyd as a new upstream regulator of the Hippo pathway. Using loss-and gain-of-function studies, we show that Pyd is required and sufficient to promote green-sensitive yR8 fate and repress blue-sensitive pR8 fate. We additionally determined the roles in PR subtype fate specification for pez and suppressor of deltex (su(dx)), the upstream regulators of the Hippo pathway in the Drosophila midgut epithelium [51,52]. Using epistasis analyses, we revealed that Pyd acts upstream of the core components and the upstream regulator Pez in the Hippo pathway, while it may function in parallel to Kib to repress Su(dx)'s activity to specify R8 subtypes. Together, our study identifies a new upstream regulator of the Hippo pathway that functions in post-mitotic neuronal fate specification.

A triple heterozygote-based screening to identify new regulators of the Hippo pathway
A complementation test is generally used to determine whether or not two mutations define the same or different genes [53]. In most cases, if two recessive mutations that cause similar phenotypes are not in the same genes, the two mutations can be complemented by the corresponding wild type alleles in the F1 double heterozygotes and therefore manifest the wild type phenotype. However, when these mutations are in the genes of the same signaling pathway, they can fail to complement each other. Instead, the double heterozygote can exhibit a mutant phenotype [53,54]. To test whether this non-complementation between the genes in the same pathway can be used to screen the Drosophila deficiency collections [55,56] to identify new components of the Hippo signaling pathway, we performed complementation assays for pR8 and yR8 subtype fate specification between a mutation of wts, the core component of the Hippo pathway, and mutations in other genes in the pathway. Compared to the ratios of pR8 (Rh5+: 33.9±2.2%) and yR8 (Rh6+: 66.1±2.2%) in wild type flies (Fig 1D and 1H), flies heterozygous for a hypomorphic wts enhancer trap line wts Zn [15] (wts Zn /+) (Rh5+: 36.5±3.4%, Rh6+: 63.5±3.4%), a null kib allele kib 1 [43] (kib 1 /+) (Rh5+: 34.7±2.3%, Rh6+: 65.3±2.3%), or the sav deficiency mutation Df(3R)BSC803 (referred to as sav df hereafter) (sav df /+) (Rh5+: 33.0±3.7%, Rh6+: 67.0±3.7%) did not show any statistical difference in the ratios of pR8 and yR8 (Figs 1H and S1A-S1D), while the heterozygous mer 3 flies [48] (mer 3 /+) had a slight increase in pR8s at the expense of yR8s (Rh5+: 43.7±1.1%, Rh6+: 56.3±1.7%) (Figs 1H and S1E). We then analyzed pR8 and yR8 subtypes in the double heterozygous flies for these mutations. The ratios of pR8s had minor increases with corresponding decreases of yR8s in the double heterozygous flies for wts Zn  and S1G and S1H). There was a higher increase in pR8s at the expense of yR8s in wts Zn /mer 3 (Figs 1F and 1H and S1F) compared to heterozygous mer 3 (mer 3 /+) mutants and other double heterozygous flies (Figs 1H and S1E). These results showed the heterozygous mutations of the genes in the Hippo pathway can enhance each other's R8 subtype phenotype, but the effects are too subtle for large-scale screening.
We therefore generated a wts Zn -kib 1 chromosome by recombining wts Zn and kib 1 mutations and designed a triple heterozygote-based phenotype enhancement assay. We analyzed pR8 and yR8 subtypes in wts Zn -kib 1 /mer 3 and wts Zn -kib 1 /sav df triple heterozygous flies, respectively. The number of pR8s was significantly increased in all wts Zn -kib 1 /mer 3 (Rh5+: 89.0±3.7%) and wts Zn -kib 1 /sav df (Rh5+: 90.3±2.1%) flies. In contrast, yR8s were dramatically decreased in these flies (Rh6+: 11.0±3.7% and 9.7±2.1%, respectively) (Figs 1G and 1H and S1I). These results indicated that the wts Zn -kib 1 heterozygous flies provide a sensitive genetic background, into which introducing one copy of a loss-of-function mutation in a gene of the Hippo pathway to generate triple heterozygous flies can dramatically affect R8 subtypes. Therefore, the triple heterozygote-based assay can be used as an efficient and sensitive tool to screen the Drosophila chromosomal deficiencies to identify new regulators of the Hippo pathway.
The phenotype enhancement between pyd and wts Zn -kib 1 mutations suggests pyd is a regulator of the Hippo signaling pathway. To confirm that pyd regulates R8 subtype fate specification via regulating Hippo signaling, we analyzed R8 subtypes in yki knock-down retinas in pyd loss-of-function (LOF) background. Yki is the downstream effector of the Hippo pathway and

Pyd regulates wts and melt expression in R8 cells
Previous findings have shown that a key step in dictating pR8 (Rh5-positive) vs. yR8 (Rh6-positive) fate is through the transcriptional activation of melt and wts in pR8s and yR8s, respectively [15]. To determine whether pyd functions upstream of the melt-wts regulatory loop, we first analyzed wts and melt expression in pyd knock-down retinas by using an enhancer trap line for wts (wts-nLacZ, aka wts Zn ) [15] and an expression reporter for melt (melt450-nLacZ) [10]. wts Zn was expressed in~65% of R8s (yR8s, 65.2±3.6% of R8s) in wild type retinas (Fig 4A  and 4E). However, its expression was lost in most of R8s in pyd knock-down retinas (5.2±2.1% of R8s) (Fig 4B and 4E). In contrast, melt expression was expanded into most R8s in pyd knock-down retinas (91.9±3.2% of R8s) (Fig 4D and 4F), compared to its expression in~35% of R8s in wild type retinas (34.9±3.5%) (Fig 4C and 4F). Therefore, these results indicate Pyd is required for wts expression and melt repression in yR8 cells.
As pyd plays a similar role with wts to repress melt expression, we tested whether pyd is also expressed in an yR8-specific manner. We performed a Gal4 enhancer trapping by using the pGawB transposons which insertions are within or adjacent to the pyd gene.  NP4419 , which insertion is 8bp upstream of the pyd gene, was able to drive the expression of the reporter gene UAS-nuclear GFP (UAS-nGFP) in PR cells. The expression was found in all PR cells and also in non-neuronal cells (Fig 4G). We further tested its expression in melt GOF (otd>melt) or wts GOF (otd>wts) flies and found that its expression was not affected by melt or wts misexpression (Fig 4H). All these data suggest that pyd is transcriptionally expressed in both pR8 and yR8 subtypes, and its expression is not regulated by wts or melt.

Pyd functions upstream of the core components of the Hippo pathway for PR subtype fate specification
To further understand how Pyd regulates the Hippo pathway to specify PR subtypes, we performed epistasis assays for Pyd and the core Hippo components in pR8 and yR8 subtype fate specification. Misexpression of wts, hpo or sav in wild type retinas was sufficient to induce yR8 fate and repress pR8 fate in most or all R8 cells [15,49] (Rh5+: 7.1±2.5%, Rh6+: 92.9±2.5% in lGMR>wts retinas; Rh5+: 0%, Rh6+: 100% in lGMR>hpo retinas; Rh5+: 0%, Rh6+: 100% in GMR>sav retinas) (Fig 5B, 5E and 5I). We found that misexpression of these core genes wts, hpo or sav in pyd mutant retinas had the same abilities to promote yR8 and repress pR8 subtype fate specification with their misexpression in wild type retinas: with the misexpression of wts, hpo, or sav, all or most of pR8s in pyd mutant retinas adopted to yR8 subtype fate (Rh5+: 8.8±3.1%, Rh6+: 91.2±3.1% in lGMR>wts, pyd LOF retinas; Rh5+: 0%, Rh6+: 100% in lGMR>hpo, pyd LOF retinas; Rh5+: 0%, Rh6+: 100% in GMR>sav, pyd LOF retinas) (Fig 5C,  5D, 5F and 5I). To further confirm the epistasis of pyd and wts, we performed pyd and melt double LOF experiments. wts expression in pR8s is derepressed in melt mutant flies [15]. We  (Figs 5G-5I and S5A-S5F). Together, these results indicate that Pyd genetically acts upstream of or in parallel to the core components of the Hippo pathway. The expression of pyd in both pR8 and yR8 subtypes indicates that the Pyd protein at its endogenous level in pR8 cells might not be sufficient to induce yR8 fate. In order to test whether the core components of the Hippo pathway are necessary for Pyd-mediated yR8 fate specification, we tested whether overexpression of Pyd in all PR cells by using a strong GMR-GAL4 driver [60] (GMR-GAL4>UAS-pyd) can induce yR8 fate specification. We found that pyd overexpression induced Rh6 expression in most R8 cells (Rh6+: 88.9±3.1%). Rh5 expression was only observed in a small proportion of R8s in pyd overexpression retinas (Rh5 +: 11.1±3.1%) (Fig 5K and 5P). We then knocked down wts, hpo, sav or mats in pyd-overexpressing retinas, and found that knocking down any of these genes abolished the overexpressed Pyd's ability to promote yR8 and suppress pR8 fate specification ( (Fig 5L-5P), suggesting these Hippo core components are necessary for Pyd to promote yR8 and inhibit pR8 subtype fate specification. Together, these epistasis experiments suggested Pyd genetically functions upstream of the core component genes in the Hippo pathway for R8 subtype fate decisions.

Pyd functions upstream of pez for R8 subtype fate specification
Pyd was previously shown to directly interact with the E3 ubiquitin ligase Su(dx) in regulating the size of the Drosophila ovary stem cell niche [58]. Additionally, Su (dx) targets and degrades Pez during intestinal stem cell proliferation [52]. Pez is the Drosophila homolog of Protein tyrosine phosphatase non-receptor type 14 (PTPN14) and functions as a partner of Kib. Both Pez and Kib are required for the activity of the Hippo pathway to restrict intestinal stem cell proliferation in the Drosophila midgut epithelium [51]. However, the roles of both Pez and Su (dx) in post-mitotic PR subtype fate specification have not been explored. Here, we first analyzed pR8 and yR8 subtypes in heteroallelic pez 1 /pez 2 mutant flies. Almost all R8s expressed Rh5 (97.5±1.2%), the pR8 fate marker, in pez 1 /pez 2 mutant flies at the expense of Rh6 (2.5 ±1.2%) (Fig 6B and 6G). Furthermore, misexpression of pez (lGMR>pez) led to a significant increase in Rh6-expressing yR8s (94.6±2.5%) and a reduction in Rh5-expressing pR8s (5.4 ±2.5%) (Fig 6C and 6G). These results demonstrate that pez is necessary and sufficient to promote yR8 and repress pR8 subtype fate specification in the Drosophila eye. We then determined the genetic relationship between pyd and pez. Misexpression of pez was able to suppress the phenotype caused by pyd mutations (Rh5+: 6.1±2.9%, Rh6+: 93.9±2.9% in lGMR>pez, pyd LOF retinas) (Fig 6D and 6G), similar with misexpression of the core components of the Hippo pathway. Further, we knocked down pez in pyd-overexpressing eyes and found that loss of pez repressed the phenotype caused by overexpressed Pyd (Rh5+: 90.2±3.9%, Rh6+: 9.8 ±3.9% in GMR>pez RNAi +pyd retinas) (Fig 6E-6G), indicating Pez is required for Pyd to promote yR8 and repress pR8 subtype fate. Together, these results indicated that Pez acts downstream of Pyd in the Hippo pathway to specify R8 subtype fate specification. Su(dx) plays an opposite role to Pyd in regulating the size of the ovary stem cell niche [58]. In order to investigate the functional relationship of the two proteins in R8 subtype fate specification, we first tested whether Su(dx) plays any role in PR subtype fate specification by knocking down su(dx) in retinas. Knock-down of su(dx) (lGMR>su(dx) RNAi ) led to an opposite phenotype to loss of pez: Rh6-expressing yR8s were significantly increased (86.8±3.1%) at the expense of Rh5-expressing pR8s (13.2±3.1%) (Fig 6H and 6N). We additionally used su(dx) mutations su(dx) 2 and su(dx) 32 to generate heteroallelic su(dx) 2 /su(dx) 32 mutant flies and analyzed R8 subtypes in these flies. su(dx) 2 /su(dx) 32 retinas showed a similar R8 subtype phenotype (Rh5+: 15.2±4.3%, Rh6+: 84.8±4.3%) with su(dx) knock-down retinas (Fig 6I and 6N). Further, we found that misexpression of su(dx) (lGMR>su(dx)) caused an increase in pR8s (71.1±4.8%) and a reduction in yR8s (28.9± 4.8%) (S6A-S6C Fig). Therefore, Su(dx) and Pyd play opposite roles in R8 subtype fate decisions.

Pyd suppresses Su(dx) in R8 subtype fate specification
Previous yeast two-hybrid assays and co-immunoprecipitation (co-IP) tests in the Drosophila S2 cells have shown that Pyd and Su(dx) can interact with each other and form a complex. To explore the functional relationship of the two interacting proteins during R8 subtype fate

Discussion
In this study, we designed a sensitized genetic screen using a triple heterozygote-based PR subtype phenotype enhancement assay to identify novel regulators of the Hippo pathway in the Drosophila eye. Taking advantage of this genome-wide screening, we identified the Drosophila ZO-1 protein Pyd as a new PR subtype fate determinant. We demonstrated Pyd is an upstream regulator of the Hippo signaling pathway and is required for the pathway to promote yR8 and repress pR8 PR subtype fate specification. We also determined the roles of Pez and Su(dx) in R8 subtype fate specification and found they play opposite roles in this process (Fig 8), as they act in intestinal stem cell proliferation. Previous reports have shown that Pyd and Su(dx) can physically interact with each other. We found that Pyd and Su(dx) act antagonistically during R8 subtype fate specification (Fig 8). Further, our pyd and su(dx) double LOF and double GOF results have indicated that the R8 subtype phenotype in pyd LOF retinas depends on the presence of Su(dx), and on the other hand, the overexpressed Pyd represses Su(dx)'s activity to promote pR8 and inhibit yR8 fate specification. Considering that Su(dx) can induce Pez degradation [52], Pyd may be required for Hippo signaling by antagonizing Su(dx)'s activity and therefore stabilizing Pez (Fig 8). Interestingly, it is the WW domain of the Su(dx) protein that interacts with both Pez and Pyd. Therefore, it is possible that Pyd competes with Su(dx) to bind and stabilize Pez. Our data also showed that Kib suppresses Su(dx)'s activity during R8 subtype fate specification, consistent with the previous report that Kib can block Su(dx)induced Pez degradation [52]. However, Kib was not shown to interact with Su(dx) and it can't decrease the binding between Su(dx) and Pez [52]. Therefore, Pyd and Kib may use different mechanisms to stabilize Pez: Pyd competes with Su(dx) for Pez binding, while Kib directly binds to Pez. Since loss of pyd or kib lead to significant expansion of pR8s and reduction of yR8s, both of the two mechanisms is required in wild type retinas. However, overexpression of any one of pyd and kib can circumvent loss of another gene (Fig 7), suggesting the two mechanisms might function independently (Fig 8). It will be of interest and important to test this model using biochemical approaches in future studies and explore whether and how Pyd directly competes with Su(dx).
The Hippo pathway was originally discovered in Drosophila and its evolutionarily conserved roles in various biological processes have been subsequently found in mammals [19,34,[61][62][63][64][65][66][67][68]. However, the regulatory mechanisms upstream of the signaling pathway are less understood. Most of the core components of the Hippo pathway were first isolated as a result of their overgrowth phenotypes in mosaic mutant-based screens [69]. However, this strategy is not efficient to identify the upstream components of the pathway because the overgrowth defects caused by mutations of the upstream genes is much weaker compared to those induced by mutations of the core components due to the redundant roles of the upstream components in tissue growth control [42,43,51,70]. Interestingly, Fat, Expanded as well as Crumbs are not required for the activation of Hippo signaling during R8 subtype fate specification [49], making the upstream regulation of the Hippo pathway during R8 subtype fate decisions less complicated. Additionally, Hippo-dependent R8 subtype fate specification can be precisely quantified. Taking advantage of these features, we generated a sensitive genetic

PLOS GENETICS
background with double heterozygous wts and kib mutations that affect R8 subtypes modestly, but is able to significantly change pR8 and yR8 subtypes when coupled with one more mutation in a gene of the Hippo pathway. This sensitive genetic tool makes it possible to perform a genome-wide screening by testing the activities of the Drosophila deficiency lines to affect R8 subtypes. Notably, previous studies have demonstrated that Mer physically interacts with Wts and Kib [42,71]. Our results showed the R8 subtype phenotype was enhanced more in heterozygous wts Zn -mer 3 or kib 1 -mer 3 flies than in wts Zn -kib 1 , kib 1 -sav df or wts Zn -sav df flies. Therefore, the quantitative phenotype enhancement assays for R8 subtypes have a potential application to predict the physical interactions between the components of the Hippo pathway.
ZO-1 proteins are crucial for the formation and maintenance of tight junctions in vertebrate cells [72]. While in Drosophila cells, which lack tight junctions [73], Pyd is associated with both adherens and septate junctions [74,75]. ZO-1 proteins are members of the membrane-associated guanylate kinase (MAGUK) family and contain a GUK (guanylate kinase) domain, three PDZ domains, and a SH3 domain [58]. The cellular localization and the presence of multiple protein-protein interaction domains suggest the ZO-1 proteins may play important roles in coupling the extracellular signals to intracellular signaling pathways. A previous study in cultured cells have found that the transiently expressed ZO-1 protein can interact with the carboxy-terminal PDZ binding motif of TAZ, a downstream effector of the Hippo pathway in mammals, via its first PDZ domain [76]. Whether Pyd interacts with Yki, the Drosophila homolog of YAP/TAZ, hasn't been explored. However, our result that knock-down of wts is sufficient to suppress the phenotype in pyd overexpression retinas suggests the interaction between Yki and Pyd, if there is any, does not play a significant role for the cytoplasmic retention of Yki and thus inhibiting its activity as a transcription co-activator. Additionally, Pyd has been previously implicated in the regulation of the Notch pathway in context-dependent manners [59,77,78]. However, the Notch pathway has not been shown to cell-autonomously regulate R8 subtype fate specification in the Drosophila eye. Our results in this study provide evidence that Pyd is a regulator of the Hippo signaling pathway and functions as an upstream regulator of the pathway for PR subtype fate decisions. Considering its interactions with junctions and cytoskeleton proteins, Pyd might function as a scaffold to organize other components of the Hippo pathway at the plasma membrane to form functional complexes. Furthermore, genetic or direct interactions between Pyd and some transmembrane proteins have been previously reported [79,80]. Given that Pyd functions upstream of the Hippo pathway during R8 subtype fate decisions, it will be of great interest to test the role of these Pydinteracting transmembrane proteins for R8 subtype fate decisions and investigate whether any of them acts as a transmembrane receptor in Hippo signaling. Notably, Mer plays key roles to recruit the core Hippo components to apical membrane area [81]. Pyd and its transmembrane partner may be required for Mer membrane associations.
The R8 terminal differentiation into pR8 or yR8 subtype fate occurs in the late pupal stage and is dependent on the activation and deactivation of the Hippo signaling pathway [82]. As a negative regulator of the Hippo pathway, melt is expressed in a subset of R8s from 40% pupation [10] and is indispensable to transcriptionally repress wts expression and de-activate Hippo signaling [49], allowing these R8s to adopt the pR8 subtype fate. In this study, we determined that the E3 ligase Su(dx) as another negative mediator of the Hippo pathway for R8 subtype fate specification. Su(dx) was shown to degrade Pez and therefore inactivate Hippo signaling in midgut epithelium [52], indicating Su(dx) inactivates Hippo signaling by a different mechanism with Melt-mediated transcriptional repression of wts. It is possible that Su(dx) is present in a subset of R8s at 40-50% APF stage and its presence reduces the default Hippo signaling and thus results in elevated Yki activity which, together with the transcription factors Otd, Traffic jam and Scalloped [10,20], initiates the melt-wts bistable loop to activate melt and repress wts expression, and finally leads to the generation of pR8 subtype.

Immunohistochemistry
Fly head cryosections, dissection for whole mount retinas, and antibody staining were performed as previously described with modifications [82,86]. Adult fly heads were embedded and frozen in OCT and sectioned (12 μm) using the Cryostat CM1850 (Leica). The samples were then fixed in 4% paraformaldehyde/ PBS, and washed 3x10 min with PBX (PBS + 0.3% Triton X-100), incubated with primary antibodies overnight at 4˚C in antibody dilution buffer (PBX + 1% BSA), washed 4x10 min with PBX, and incubated 90 min at room temperature with secondary antibodies diluted in antibody dilution buffer. After 4x10min PBT washes, samples were mounted in anti-fade reagent, and imaged. Antibody dilutions were: rabbit Salm (1:150) [82]; mouse Rh5 (1:1000) [87]; rabbit Rh6 (1:100, this study); chicken LacZ (1:1000, Abcam). Alexa Fluor 488, 555 and 655-conjugated secondary antibodies (1:1500, Invitrogen) were used. Digital images were obtained with an Apotome deconvolution system (Zeiss), and processed with Axiovision 4.5 (Zeiss) and Adobe Photoshop software. Quantifications for the longitudinal sections were performed by counting at least 800 ommatidia from four or more individual flies per genotype, and only sections that include both R7 and R8 layers were counted. Quantifications for the tangential sections use one section for each retina to avoid repeatedly counting the same ommatidia. Retinas for quantifying the whole mount staining are from at least three flies per genotype.

Polyclonal antibody production
Polyclonal antiserum against Rh6 was generated against a KLH-conjugated peptide from the Rh6 deduced amino acid sequence (CLACGKDDLTSDSRTQAT corresponding to amino acids 344-361). Peptide synthesis, KLH-conjugation, rabbit immunizations and bleeds were performed by GenScript (Piscataway, NJ).