A screen for E3 ubiquitination ligases that genetically interact with the adaptor protein Cindr during Drosophila eye patterning

Ubiquitination is a crucial post-translational modification that can target proteins for degradation. The E3 ubiquitin ligases are responsible for recognizing substrate proteins for ubiquitination, hence providing specificity to the process of protein degradation. Here, we describe a genetic modifier screen that identified E3 ligases that modified the rough-eye phenotype generated by expression of cindrRNAi transgenes during Drosophila eye development. In total, we identified 36 E3 ligases, as well as 4 Cullins, that modified the mild cindrRNA mis-patterning phenotype. This indicates possible roles for these E3s/Cullins in processes that require Cindr function, including cytoskeletal regulation, cell adhesion, cell signaling and cell survival. Three E3 ligases identified in our screen had previously been linked to regulating JNK signaling.


Introduction
Covalent attachment of ubiquitin to a protein is a post-translational modification that can signal its degradation by the 26S proteasome (reviewed by [1][2][3][4]). This process is crucial for the clearance of proteins when no longer needed in a cell. Protein ubiquitination also serves important proteasome-independent roles and has been implicated in signal transduction [5], protein trafficking [6], endocytosis [7], DNA repair [8], transcriptional regulation [9] and histone modification [10]. Given these diverse and important roles, the ubiquitination system can profoundly influence the development and homeostasis of tissues.
Three core classes of enzyme complexes are required for ubiquitination [1][2][3][4]. The ubiquitin-activating enzymes (E1s) catalyze conversion of ubiquitin to ubiquitin-adenylate intermediates that are momentarily bound to E1s. The active ubiquitin-adenylate is then transferred to ubiquitin-conjugating enzymes (E2s). Finally, ubiquitin is transferred to target proteins in reactions catalyzed by ubiquitin ligases (E3s) that provide substrate specificity by dictating which target proteins are ubiquitinated.
E3 ligases are characterized according to their domains which catalyze transfer of ubiquitin to target proteins: the HECT (homologous to the E6AP carboxyl terminus) and RING (Really Interesting New Gene) domains [1,2]. A more elaborate RBR (RING-between-RING) domain characterizes a subclass of E3s. In addition, proteins containing Cullin, U-box, N-recognin, SKP1 and F-box domains contribute to the formation of functional E3 complexes. In recent annotations, 617 putative E3 ligases were identified in the human genome and 80 putative E3 ligases in the yeast Saccharomyces cerevisiae, accounting for 1-2% of the proteins encoded in the genomes of these species [11]. E3 ligases occupy a similar percentage of the Drosophila genome [12]. Identifying the substrates of these E3 ligases and the cell behaviors for which their functions are crucial will facilitate a comprehensive understanding of the importance of ubiquitination but is a considerable challenge. Determining which E3 ligases are required in specific tissues is an important first step in meeting this challenge.
The Drosophila eye neuro-epithelium has been extensively utilized to study cell behaviors and signals that integrate to generate functional epithelia. This tissue is patterned with high precision during development and disruptions to the epithelium are easily detected in the adult eye. Here we describe a genetic modifier screen that identified E3 ligases that genetically interact with Cindr, a conserved scaffold protein that is essential for eye development [13]. Expression of RNAi transgenes that targeted cindr generated a sensitized genetic background that could be modified by mutations in E3 ligase loci. The UAS-cindr RNAi.2.21 transgene was expressed by the driver line GMR-GAL4 (the genotype of these retinas is abbreviated to GMR>cindr RNAi2 throughout this manuscript). This modestly compromised multiple cell behaviors that require Cindr, including signal transduction, the correct localization of adhesion proteins, and regulated remodeling of the actin cytoskeleton [13][14][15][16]. The E3 ligases identified in our screen therefore have potential roles in regulating any of these conserved cell behaviors during the development of the eye epithelium.

Genetic modifier screen
Between six and eight young male flies of each stock screened were crossed to eight to ten virgin GMR-GAL4; UAS-cindr RNAi2.21A / SM5: TM6b females. For control crosses, males were crossed to GMR-GAL4 virgin females. Crosses were maintained at 25˚C. The parental flies were removed from vials on day seven. On day fourteen the F1 progeny that had emerged were scored blind and independently by two researchers. Scoring was repeated if their assessments differed. Adults were frozen rapidly at -70˚C and imaged using a Leica M125 stereo-dissecting microscope fitted with an LED5000HDI ring light and diffuser (data presented in

Pupal eye dissection, immunofluorescence and analyses
All crosses and pupae were maintained at 25˚C. Eye-brain complexes were dissected at 40 hours after puparium formation (APF) in PBS, fixed on ice in 4% formaldehyde in PBS for 35 minutes, and incubated overnight in rat anti-Drosophila Ecadherin (1:20, DSHB DCAD2) at 4˚C. Tissue was then incubated in goat anti-rat secondary antibodies conjugated to AlexaFluor 488 (Jackson ImmunoResearch). Retina were removed from the brain complexes and imaged using a Leica TCS SP5 DM fluorescent microscope and associated LAS AF Software (Leica Microsystem, Exton, PA). Images were processed using Adobe Photoshop CC (Adobe, San Jose, CA). Patterning errors were quantified as previously described [18].

Selection of E3 ligases
We explored the Drosophila melanogaster genome using the Gene Ontology search function hosted by FlyBase (the database for Drosophila Genes and Genomes) to identify loci annotated to have domains or properties associated with ubiquitin ligase activity. These searches generated an initial candidate list of 156 predicted or experimentally confirmed E3 ligases (S1 Table, summarized in Table 1), which included all ubiquitin ligases also identified by FlyBase curators [19][20][21]. Since Cullin proteins function as scaffolds to assemble E3 ligase complexes [22], we also included the six Drosophila Cullins into our candidate list. We did not include F-box and SKP1 proteins, which are components of the Cullin-based E3 complexes.
Our primary goal was to identify E3 ligases that function in the cytosol during epithelial tissue development, since Cindr resides in this compartment. Therefore, we removed from our candidate list all E3 ligases that, at the time, were known or predicted to function primarily in the nucleus, mitochondria, peroxisomes, the endoplasmic reticulum, lysosome or the endosome (see S1 Table). Next, we removed E3 ligases that (at the time) were known to modify Notch signaling (Bre1, Deltex, Mind Bomb 1, Suppressor of deltex, Nedd4, Neuralized) and Decapentaplegic/SMAD signaling (Smurf/Lack), because these signaling pathways are essential for processes that also require Cindr during Drosophila eye development [23,24]. However, we included Cbl, a proto-oncogene that modifies Receptor Tyrosine Kinase (RTK) signaling, including Epidermal Growth Factor Receptor Signaling (EGFR) which is extensively utilized during Drosophila eye development [25,26]. Since the vertebrate orthologs of Cbl and Cindr are known to interact [27], we predicted that alleles of cbl would be identified in our screen, which would verify the efficacy of our approach.
The Bloomington Drosophila Stock Center (BDSC) maintains stocks carrying classical mutations or transposable element insertions. Gene expression is modified in many of these Drosophila lines. Unfortunately, alleles were not available for 39 of the 110 loci that we wished to screen (S1 Table). Alleles for the remaining 71 loci were obtained.

The genetic modifier screen
The Drosophila eye is a simple neuro-epithelium composed of~750 hexagonally shaped ommatidia, each capped with a domed lens that is easily observed using a standard stereo-dissecting microscope (Fig 1A and 1B) [28]. Each ommatidium is composed of eight photoreceptor neurons  E3 ligases that modify eye mispatterning when Cindr is reduced encapsulated by four cone-cells and two primary pigment cells that secrete the material that becomes each lens. Secondary and tertiary pigment cells surround each ommatidium. The hexagonally-shaped ommatidia are neatly packed in rows, giving the adult eye its precisely-ordered appearance (Fig 1A and 1C). Even small disturbances during eye development can disrupt this simple pattern, alter the shape or dimensions of the ommatidial lenses, and give rise to mis-patterned 'rough' eyes. Expression of RNAi transgenes that targeted cindr generated a mild rougheye phenotype (Fig 1D). Cindr is required for the regulation of a multitude of cellular processes during eye development including the correct remodeling of the actin cytoskeleton, the appropriate spatio-temporal localization of adhesion proteins, cell-signaling, and the apoptotic removal of superfluous cells from the eye field [13][14][15][16]. Given these diverse cell behaviors, the GMR> cindr RNAi2 rough-eye provided a sensitized background for a modifier screen to identify E3 ligases and Cullins essential for eye patterning. Enhancement or suppression of the GMR>cindr RNAi2 eye phenotype could be easily scored.
Alleles of Cul1, Cul2, Cul3 and CG11261 modified the GMR>cindr RNAi2 eye. Diverse roles for Drosophila Cul1 and Cul3 have been suggested that may account for their interactions with GMR>cindr RNAi2 . Cul1 has been implicated in regulating Cyclin E to promote cell division [29] and may therefore modify mitosis during larval eye development. Cul1 and Cul3 are also regulators of Cubitus Interruptus [30][31][32], transcription factor activated by Hedgehog signaling, which is required during early eye patterning [23]. In addition, Cul3 has been described as a regulator of the actin cytoskeleton [33-36], though this role has not been explored in epithelia. Cul2 function has been mainly explored in Drosophila germline development [37,38].
Many of the E3 ligases that modified GMR>cindr RNAi2 have yet to be characterized and named (Table 2). In addition, we identified E3 ligases that have been linked to differentiation,   ; ari-1, which has been implicated in axon pruning and re-wiring and adult myogenesis and is especially important during metamorphosis as it targets the ecdysone receptor [54][55][56][57]; tn, which is crucial for the assembly and maintenance of myofibrils [58,59]; and sinah and sina, which have been implicated in photoreceptor and bristle differentiation [60]. In addition, several E3 ligases connected to Jun-N-terminal Kinase (JNK) signaling modified GMR>cindr RNAi2 phenotypes.

Phenotype: enhanced (E), suppressed (S) or no modification (NM) of GMR>cindr RNAi2
A cohort of JNK-associated E3 ligases were identified in the screen A set of sequentially-activated kinases comprise the core of the JNK signaling pathway, a developmentally regulated pathway that is also activated in response to stress signals [61][62][63]. JNK activity influences the establishment of planar polarity in the fly retina [64], but otherwise does not contribute significantly to eye development [65]. However, we found that a UAS-insertion allele of no poles (nopo) suppressed GMR>cindr RNAi2 mis-patterning (Fig 2C). Nopo promotes apoptosis in response to Eiger-TNF Receptor signaling [66]. Modification of GMR>cindr RNAi2 by alleles of parkin (park), was inconsistent: defects were suppressed by park 1 and mildly  repressed by park Δ21 but severely enhanced by park c00062 (Fig 2E-2G). It is possible that the park c00062 line contains additional mutations that contribute to eye disruption. Parkin has been implicated in inhibiting JNK activity to suppress apoptosis, possibly by indirectly reducing bsk expression [67,68]. Finally, two alleles of tumor necrosis factor receptor-associated factor 6 (Traf6) that potentially drive ectopic Traf6, Traf6 EP325 and Traf6 EP1516 suppressed cindr RNAi mis-patterning (Fig 2H and 2I). Traf6 promotes JNK activity downstream of the TNF Receptor [69]. Identifying alleles of nopo, parkin and Traf6 in our screen suggested that JNK activity is modified in the eye epithelium in response to expression of UAS-cindr RNAi2 transgenes, a relationship that we have observed in Drosophila wing epithelia [16]. To verify this, we tested whether alleles of other JNK signaling components modified the GMR>cindr RNAi2 adult eye phenotype. Over-expression of basket (bsk, which encodes the Drosophila JNK [65]) mildly enhanced cindr RNAi2 mis-patterning (Fig 3C). In addition, ectopic slpr, a JNKKK that functions upstream of Bsk, [70,71], also enhanced the cindr RNAi2 rough eye (Fig 3E) as did mutations in the Bsk inactivator, puckered (puc, [72], Fig 3G). As the potential for JNK activity would have been enhanced in these three genetic manipulations, we expected that manipulations that decreased JNK activity would suppress GMR>cindr RNAi2 adult eye phenotypes. However, mutations in bsk and over-expression of puc enhanced cindr RNAi2 mis-patterning (Fig 3H and  3I). These data could reflect cross-talk between JNK and other signals that are utilized during eye development, including Notch, Hedgehog and RTK networks [73]. In addition, ectopic Traf4, which promotes JNK signaling [74], also enhanced the cindr RNAi2 rough eye as did mutations in the ubiquitin-conjugating enzyme variant 1a (dUev1A) an E2 enzyme that promotes JNK activity [75] (Fig 3K and 3M). Interestingly, alleles of two Cullins tested in our screen-Cul1 and Cul3-also modified the GMR>cindr RNAi2 adult eye ( Table 2). Activity of these Cullins has not been linked to JNK signaling, but they have been implicated in dendrite morphogenesis and apoptosis, processes that require JNK activity. Surprisingly, Cul4, which has been shown to regulate JNK (as well as Wingless) activity [76], did not modify the GMR>cindr RNAi adult eye-it is possible that the Cul4 KG02900 allele used in the screen does not significantly perturb the locus.

Phenotype: enhanced (E), suppressed (S) or no modification (NM) of GMR>cindr RNAi2
To better understand how nopo, park and Traf6 modified the phenotype of GMR>cindr R-NAi2 adult eyes, we dissected retinas from pupae in which these E3 ligases were modified (Fig   4). At 40 hours after puparium formation (APF), the stereotyped arrangement of ommatidial cells and the interweaving cell lattice is established (Fig 4A and 4B). Specifically, four cone cells and two primary pigment cells encapsulate each group of photoreceptors, which are buried below the apical epithelium surface. Three bristle groups, three tertiary pigment cells and six secondary pigment cells are arranged about each ommatidium, thus generating the honeycomb lattice (Fig 4A and 4B). This precise cellular pattern is reflected in the arrangement and shapes of the lenses of adult eyes (Fig 1A and 1B). The neat arrangement of interommatidial cells was mildly disrupted in GMR>cindr RNAi2 retinas (Fig 4C and 4D) due to the introduction of a variety of patterning errors which are quantified in Table 3. Specifically, many cells failed to adopt correct positions and shapes and consequently the honeycomb interommatidial cell lattice was mildly distorted. Expression of nopo, park and Traf6 generated few defects in the arrangement of cells in the retina (Fig 4E, 4I and 4M, Table 3). However, ectopic nopo and park significantly restored patterning of interommatidial cells in GMR>cindr RNAi2 retina ( Fig  4F and 4J, compare to panel C, Table 3). To verify the contribution of Traf6 we obtained an additional UAS line (BL-58991), which mildly improved patterning of the pupal eye (Fig 4N, compare to 4C, Table 3). Mis-patterning of the lattice of GMR>cindr RNAi2 retinas was mildly enhanced by alleles of nopo (Fig 4G, 4H, compare to 4D) and improved by park 1 and park Δ21 (- Fig 4K and 4L, Table 3).

Discussion
Many signals and cell behaviors integrate to pattern complex epithelia. In this screen, we have identified 36 E3 ligases and 4 Cullins that interact with the adaptor protein Cindr, which is required for these processes ( Table 2). Few of these E3s/Cullins have been characterized and the roles of most of these in epithelia are unexplored. Deciphering the substrates of the E3 ligases and the conserved cell behaviors that they modify will be an important next step in understanding their contribution to epithelial patterning.
Amongst those loci identified in our screen were a set encoding E3 ligases that had previously been implicated in modifying JNK activity (Fig 2). These E3s modified cindr RNAiinduced patterning defects that were evident in the errant arrangement of cells in the pupal eye and reflected in the disordered arrangement of lens facets in the adult. Cindr is required to inhibit JNK activity in the developing Drosophila wing epithelium [16]. Hence identifying nopo, parkin and Traf6 in our screen likely reflects that Cindr-JNK interactions are important for the correct development of most epithelia. However, since many E3 ligases regulate multiple proteins, it is possible that Nopo, Parkin or Traf6 have targets besides components of the JNK cascade. Over-expression of these proteins did not disrupt the eye (Fig 4E, 4I and 4M) and investigations of loss-of-function phenotypes are required to clarify whether these E3 ligases contribute to JNK-independent processes that pattern epithelia.
It is intriguing that genetic manipulations that potentially increased JNK activity (ectopic bsk and slpr expression, mutations in puc) as well as those that perturbed JNK (mutations in bsk, expression of puc) enhanced mis-patterning of the adult GMR>cindr RNAi fly eye. Solving this anomaly will require investigation of the cell behaviors regulated by JNK during larval and pupal eye development. However, our data are less surprising if one considers the effect of specific cell behaviors that converge to organize the eye. For example, during pupal development, local cell movements rearrange interommatidial cells to generate the honeycomb lattice [77] and too much cell movement, as well as too little, can impede patterning to generate adult eyes that appear similarly disordered. Hence, whilst the adult Drosophila eye provides an excellent model for genetic screens such as the one described in this manuscript, further investigations may be essential to pinpoint the cell behaviors that generate adult eye phenotypes.
Our screen did not include all E3 ligases encoded in the Drosophila genome and some of the alleles used may not have disrupted gene expression sufficiently to modify the cell behaviors responsible for mis-patterning of the GMR>cindr RNAi2 eye. Nonetheless, we have identified a large number of E3 ligases and Cullins that potentially function with Cindr to modify the cytoskeleton, adhesion or signaling as the eye epithelium is organized [13][14][15][16]. Due to the high degree of conservation between Drosophila and vertebrates, the orthologs of these E3 ligases and Cullins are likely to modify processes regulated by CD2AP and CIN85, the vertebrate orthologs of Cindr [78][79][80][81][82][83].