JNK is antagonized to ensure the correct number of interommatidial cells pattern the Drosophila retina.

Apoptosis is crucial during the morphogenesis of most organs and tissues, and is utilized for tissues to achieve their proper size, shape and patterning. Many signaling pathways contribute to the precise regulation of apoptosis. Here we show that Jun N-terminal Kinase (JNK) activity contributes to the coordinated removal of interommatidial cells via apoptosis in the Drosophila pupal retina. This is consistent with previous findings that JNK activity promotes apoptosis in other epithelia. However, we found that JNK activity is repressed by Cindr (the CIN85 and CD2AP ortholog) in order to promote cell survival. Reducing the amount of Cindr resulted in ectopic cell death. Increased expression of the Drosophila JNK basket in the setting of reduced cindr expression was found to result in even more severe apoptosis, whilst ectopic death was found to be reduced if retinas were heterozygous for basket. Hence Cindr is required to properly restrict JNK-mediated apoptosis in the pupal eye, resulting in the correct number of interommatidial cells. A lack of precise control over developmental apoptosis can lead to improper tissue morphogenesis.


Introduction
The removal of cells by apoptosis is a fundamental feature of tissue and organ morphogenesis, homeostasis and pathogenesis (Fuchs and Steller, 2011). During development, apoptosis is utilized for the strategic removal of cells to sculpt organs and tissues. Further, the shrinking of apoptotic cells can exert pulling forces on their neighbors and contribute to overall tissue shape (Teng and Toyama, 2011). In addition, apoptosis removes normal cells that have been generated in excess and abnormal cells before these can compromise a tissue.
The Drosophila eye has been an indispensable tool for analyses of the apoptotic caspases and their regulators. It is an excellent model in which to examine the developmental signals that regulate apoptosis because death of surplus interommatidial cells (ICs) -epithelial cells that separate ommatidia -occurs within an 18-hour period of mid-pupal development that is accessible to genetic manipulation and imaging (Cagan and Ready, 1989b). Additionally, the pupal retina is post-mitotic and even experimental manipulations that significantly increase apoptosis do not trigger apoptosis-induced compensatory proliferation (reviewed by (Fuchs and Steller, 2015). Hence the consequences of experimental manipulations of apoptosis can be accurately quantified.
After apoptosis has ceased in the retina, a precise number of ICs surround each ommatidium, indicating the presence of mechanisms that ensure neither too many nor too few ICs are removed. The precise nature of this mechanism remains elusive, but it likely relies on a surge of 20-hydroxyecdysone that roughly coincides with the death of most ICs (Riddiford, 1993;Yin and Thummel, 2005) and the integration of developmental signals that promote or repress IC apoptosis. These signals include EGFR activity, which confers IC survival, and Notch and Wingless (Wg) which are required for IC death (Baker and Yu, 2001;Cagan and Ready, 1989b;Cordero et al., 2004;Miller and Cagan, 1998;Parks et al., 1995). Intriguingly, the position of an IC can determine whether it survives or not. ICs close to differentiating bristle groups are most likely to die, whilst ICs in contact with more than two primary pigment cells (1°s) are most likely to survive (Monserrate and Brachmann, 2007;Wolff and Ready, 1993). Hence positional information across the retina determines cell death/survival decisions, but how this information is encoded has not been clearly resolved. Differences in Notch or EGFR activities do not appear to correlate with cell death or survival (Monserrate and Brachmann, 2007), suggesting the integration of additional molecular mechanisms that confer susceptibility or resilience to apoptosis.
Here, we explore whether activity of the Jun-N-terminal Kinase pathway (JNK) is regulated to ensure correct elimination of ICs from the Drosophila pupal eye. JNK activity can promote apoptosis, although this context-dependent output has been mainly associated with a stress response (Craige et al., 2016;Hotamisligil and Davis, 2016). For instance, JNK is activated in response to DNA damage, leading to elevated expression of hid via JNKmediated activation of the AP-1 transcription factors (Luo et al., 2007). JNK is also activated in response to disruption of the Cdc42-Par6-atypical protein kinase C complex, which is essential for epithelial cell polarity, leading to epithelial cell death (Warner et al., 2010). Activation of JNK in dying epithelial cells can lead to morphogen expression that induces proliferation of other cells to maintain tissue size, or secretion of Eiger, the ligand of the Drosophila tumor necrosis factor receptor (TNFR) that triggers JNK signaling leading to additional cell death (Fuchs and Steller, 2015). In addition, JNK signaling can be activated downstream of Dronc, resulting in a positive feedback loop that amplifies the robustness of apoptosis (Shlevkov and Morata, 2012).
Here, we describe that Cindr, in addition to its other roles in the developing pupal eye epithelium, regulates the survival of epithelial cells in the Drosophila pupal retina by limiting JNK activity. When we reduced expression of Cindr, JNK activity was enhanced and triggered the removal of a large number of ICs, an effect that was reversed when JNK signaling was compromised. Cindr is richly expressed in the eye retina, which we hypothesize provides a mechanism for widespread JNK repression. However, Cindr neither limits Bsk concentrations nor its phosphorylation. In addition, our genetic data suggest that at least some JNK signaling is still activated during IC death, despite the presence of Cindr. Finally, modifying Cindr and JNK signaling potential introduced defects in the positioning of ICs, indicating that a low level of JNK activity in the retina contributes to IC patterning.

Dissection, immunofluorescence and microscopy
Retina were dissected from pupae collected at 0 h APF and maintained at 25°C until dissection, or wandering third instar larvae, in ice-cold PBS and fixed in 4% formaldehyde. Primary antibodies were rat anti-DE-Cad2 (1:20, DSHB), rat anti-Elav (1:20, DSHB), mouse anti-β-Galactosidase (1:20, DSHB), rabbit anti-cleaved Dcp-1 (1:100, Cell Signaling) and rabbit anti-phospho Histone3 (1:200, Upstate Biotechnology). Secondary antibodies were conjugated Alexafluor 488 or Cy3 (Jackson ImmunoResearch). Retinas were imaged with a Zeiss LSM 501 confocal and associated Zen software or Leica TCS SP5 DM fluorescence microscope and associated LAS AF software. Adult thoraces were imaged with a Leica M125 stereo-dissecting microscope, Leica IC80HD camera and Leica Acquire software. All images were prepared for publication using Adobe Photoshop: images were aligned and cropped; minimal and equal adjustments were applied to images of control and experimental retinas; pseudo-color was introduced to highlight all ICs.

Image analysis
To quantify the number of ICs, hexagonal data-points were drawn by joining six ommatidia surrounding a single ommatidium, as illustrated in Figure S1 and all ICs enclosed within these data-points were counted. Since 1° cells and bristle groups are recruited from the pool of ICs between ∼17 to 22 h APF, these were included in cell counts of 18, 21, 24 and 27 h APF retinas. Between 6 and 8 retinas were assessed for each genotype per age APF and the cells counted in 7 to 15 data points per eye. Only the central third of each retina was analyzed as the retina is characterized by a developmental gradient.
The severity of cell death was scored by assessing the amount of cleaved Dcp-1 observed in whole retinas dissected at 18, 21, 24 and 27 h APF. Each retina was scored to be exhibiting mild cell death (retinas with a low number of Dcp-1 positive cells, eg. Figure 2B), moderate death (retina characterized by moderate Dcp-1 activity across the eye field, eg. Figure 2E), severe death (a large number of cells that were Dcp-1 positive, eg. Figure S4D). Damaged regions of retinas were excluded from analyses. Between 13 and 39 retinas were assessed for each genotype per age APF.
Expression of the puc E69 enhancer trap, a proxy for JNK signaling activity, was assessed by detecting β-Galactosidase in GMR>GFP and GMR>cindr RNAi221+23 retinas dissected at 24 h APF. Retinas of both genotypes were dissected consecutively and processed using common solutions. Confocal imaging and analyses utilized identical parameters. To analyze fluorescence intensity of β-galactosidase expression, ImageJ 1.50i was used to measure the mean gray value of grayscale maximal projection images generated from the same number of serial confocal sections. Three or four retinas of each genotype were analyzed for each of the three independent data sets (analyses are presented in Figure 1H which plots mean fluorescence intensities). Regions of retinas that had been damaged during dissection and regions of GMR>cindr RNAi221+23 retinas that were marked by small 'holes' (probably due to severe apoptosis) were excluded from analyses as such damage could trigger stress-or repair-induced JNK activity. Examples of images analyzed are presented in Figure S2, with regions analyzed marked.
Patterning errors were assessed and quantified as previously described (Johnson and Cagan, 2009). Briefly, all patterning errors observed within hexagonal data points were counted and the mean ommatidial mis-patterning scores calculated from 75 data points per genotype.

Western blotting and analysis
Embryo lysates were prepared from embryos of genotypes da>GFP, da>cindr GFP and da>cindrR NAi2 . 100μL of embryos of each genotype were aged between 5 and 10 h after egg laying, dechorionated and crushed in lysis buffer: 20 mM HEPES at pH 7.5 with 125 mM NaCl, 1.5 mM MgCl 2 , 1mM EDTA, 1mM DTT, 1mM Na 3 VO 4 , 1mM β-glycerolphosphate, 25 mM NaF, cOmplete™ protease inhibitor cocktail (Roche) and 20% glycerol. The lysate was cleared with centrifugation for 1 min at 5,500 rpm and frozen. Three independent samples of embryo lysate were prepared for each of the three genotypes. Each lysate sample was analyzed via SDS-PAGE and Western Blotting at least three times. Wing discs from 24 wandering third larval instar larvae of genotypes c765>GFP, c765>cindr GFP and c765>cindr RNAi2 were dissected in ice-cold PBS supplemented with 1mM DTT, 1mM Na 3 VO 4 , 1mM β-glycerolphosphate, and cOmplete protease inhibitor cocktail (Roche). The wing discs were transferred to 25uL lysis buffer (as before). Each sample was analyzed by SDS-PAGE and Western Blotting three times.

Statistical analyses
Two-sample t-tests were used to assess differences between JNK activity (puc-lacZ expression, Figure 1E), cell number ( Figure 1H, Table 2, Table 3), ommatidial mispatterning scores (Table 4), total JNK and pJNK ( Figure 6B and D). To assess differences in cell number between genotypes at 18, 21, 24, 27, 30,33, 36 and 40 h APF (ages were considered separately) ANOVAs showed that genotype was significantly associated with cell counts at significance level 0.05 ( Figure S7A). Post-hoc analyses using the Tukey procedure revealed which genotypes varied significantly at the 0.05 level ( Figure S7B).

Cindr is required for the proper survival of ICs in the Drosophila pupal eye
The Drosophila eye is striking in its order. Each ommatidium, with its core of eight photoreceptor neurons encapsulated by four cone cells and a pair of primary pigment cells (1°s), is separated by pigment-producing interommatidial cells (ICs) arranged in a precise honeycomb lattice that spans the eye field ( Figure 1A and B) (Cagan, 1993;Cagan and Ready, 1989a;Wolff and Ready, 1993). Six secondary pigment cells (2°s) form the sides of each hexagonal unit. Bristle groups and tertiary pigment cells (3°s) occupy alternate corners of the hexagon ( Figure 1A and B). This distinctive cellular arrangement arises when the organism is a pupa and requires the integration of a variety of signals and cell behaviors.
The early pupal retina is characterized by a developmental gradient. At 20 hours after puparium formation (APF), ICs are untidily arranged between ommatidia in the younger, anterior part of the eye, but rearrange into single file and progressively adopt smaller, more ordered shapes ( Figure 1C). The early pupal retina is characterized by a large number of superfluous ICs. These are removed by apoptosis, leaving the precise number of cells that adopt characteristic 2° and 3° cell shapes by 40 h APF ( Figure 1A) (Cagan and Ready, 1989b).
Reducing cindr expression during pupal eye patterning compromised the arrangement and shape of many ICs, largely due to disruptions to adhesive junctions and the cytoskeleton during patterning (Johnson et al., 2012;Johnson et al., 2011;Johnson et al., 2008). These disruptions lead to classic rough-eye phenotypes in GMR>cindr RNAi adults (not shown, (Johnson et al., 2008). In addition, expression of UAS-cindr RNAi-2 transgenes that reduce expression of Cindr isoforms that contain SH3 domains (Johnson et al., 2008), also decreased the number of ICs within the eye field ( Figure 1D). To quantify this defect in IC survival, we compared the number of ICs within hexagonal 'data points' delineated by connecting the centers of six ommatidia (illustrated in Figure S1A and B). In control retinas expressing green fluorescent protein (GFP) or β-Galactosidase (encoded by lacZ), an average of 12.1 and 11.9 cells lay within each data point, respectively ( Figure 1E). Wild type Canton S retinas are similarly characterized by ∼12 ICs per data point (not shown). However, reducing cindr decreased this number to 10.8 cells ( Figure 1E) reflecting a significant change in the regulated removal of ICs. These data suggest that Cindr counteracts cell death to ensure a suitable number of ICs remain within the IC lattice.

JNK activity is repressed by Cindr in pupal retinas
Recently, we reported that Cindr interacted with the Drosophila JNK Basket (Bsk) in the wing epithelium (Yasin et al., 2016). This interaction was crucial to preserve epithelial integrity: loss of cindr increased JNK signaling that triggered extensive cell delamination and apoptosis (Yasin et al., 2016). Since Cindr is richly expressed in the pupal retina (Johnson et al., 2008), we posited that one role for Cindr would be to similarly oppose JNK signaling during eye patterning. Supporting this hypothesis, expression of puc-lacZ (puc E69 ), a read-out of JNK signaling activity (Ring and Martinez Arias, 1993), was increased in the retina when cindr was reduced ( Figure 1F-H, Figure S2).

Cindr inhibits JNK activity to promote survival of ICs
Apoptosis removes surplus ICs from the pupal retina as the lattice is patterned (Cagan and Ready, 1989b;Wolff and Ready, 1993). To directly assay the contributions of Cindr and JNK to cell death, we examined the activation of Death caspase-1 (Dcp-1) when Cindr and JNK activity were modified ( Figure 2). As a second measure, we counted the cells that lay within hexagonal data points (Figure 3). Since 1° cells are recruited from the pool of interommatidial cells from ∼18 through ∼23 h APF we included these in our cell counts (although their numbers did not vary significantly between the genotypes we examined at any time point, data not shown). Bristle groups were also included, as these originate from cells selected from the IC pool from ∼ 16 h APF. Our strategy for counting cells in these young retinas is illustrated in Figure S1C and D.
In control retinas expressing lacZ, we observed a mean decline of 7.36 cells, per data point, from 18 h APF to 40 h APF (Table 1, Figure 3A). Cell counts confirmed that this decline is not steady, but characterized by two periods of enhanced apoptosis: from 21 to 24 h APF and from 27 to 30 h APF (Table 1). Few cells were removed after 30 h APF. These data were similar to previous characterizations of the loss of ICs from pupal retinas (Cagan and Ready, 1989b;Monserrate and Brachmann, 2007), although note that the retinas analyzed by Cagan and Ready were raised at 20°C rather than 25°C in this study and hence developed at a slower pace. Dcp-1 activity was enhanced at the anterior periphery of the eye field at 18 h APF, as previously observed (Wolff and Ready, 1993), and apoptotic cells were also dispersed through the entire retina ( Figure S3A). Ranking Dcp-1 activity into one of three categories (mild, moderate or severe, reflecting the amount of Dcp-1 activity detected across entire retinas; see Methods) revealed that the amount and severity of apoptosis was qualitatively equivalent at 18 and 21 h APF in GMR>lacZ retina, declined at 24 h APF, and then increased modestly in intensity at 27 h APF (Figure 2A and B, Figure S3A, Figure  S4A, Figure S5A).
Reducing cindr resulted in fewer cells populating retinas even at 18 h APF ( Figure 3A and B), but apoptosis and cell proliferation were unchanged in third larval instar eye discs expressing cindr RNAi ( Figure S6), suggesting that apoptosis begins prematurely in GMR>cindr RNAi pupae. We began our analyses of cell death at 18 h APF, however, as dissecting pupae without damaging the retina prior to this age is technically very difficult, and because the major morphogenetic events that pattern the retina begin from around this age. At 18 h APF, the severity of apoptosis -indicated by Dcp-1 activity -was not markedly changed by cindr RNAi expression ( Figure 2A). However, by 21 h APF many GMR>cindr RNAi retinas were marked by severe apoptosis (Figure 2A, Figure S4). Death subsided modestly at 24 h APF and then increased again at 27 h APF (Figure 2A and E, Figure S5D, Figure 3). Few cells were removed after 30 h APF. Hence the pattern of cell death in GMR>cindr RNAi retinas mimicked that of GMR>lacZ retinas but with two important differences. First, cell death began early. Second, a larger number of cells were pruned per data point from 18 to 21 h APF (1.80 cells when cindr was reduced as opposed to 0.97 in control lacZ-expressing retinas, Table 1). Hence we conclude that Cindr is especially important to protect cells from death during this early developmental period. Interestingly, reducing Cindr did not modify the deceleration of apoptosis that began from ∼30 h APF, suggesting that a 'breaking mechanism' that halts cell death is independent of Cindr.
Ectopic bsk expression also reduced the number of cells populating retinas at 18 h APF ( Figure 3) but only modestly increased apoptosis observed at 18, 21, 24 and 27 h APF (Figure 2A and C, Figure S3B, Figure S4B and Figure S5B). However, after 27 h APF, cell death reduced and the number of ICs populating the lattice at 40 h APF was normal (Table1, Figure 3, Figure 4B). These data indicate that the 'breaking mechanism' limits loss of ICs after 30 h APF overcomes even ectopic bsk.
In contrast, ectopic expression of bsk in cindr RNAi -expressing retinas markedly enhanced apoptosis to reduce the number of cells in retinas (Figure 2A and F, Figure S3E, Figure S4E and Figure S5E). Further, cell death was reduced in bsk 1/+ , cindr RNAi retina (Figure 2A and G, Figure S3F, Figure S4F and Figure S5F), although the number of cells within these retinas was also reduced at 18 h APF (Figure 3; again cell death was unperturbed in the larval eye, Figure S6). Indeed the mean number of cells lost, per data point, was reduced to 0.72 cell from 18 and 21 h APF, and to 1.43 cells from 21 to 24 h APF in bsk 1/+ , cindr RNAi retina, in comparison to 1.80 and 1.58 cells in cindr RNAi retina during these same time intervals (Table 1). Taken together, these data support the hypothesis that Cindr protects ICs from JNK-mediated apoptosis.

Modifying JNK activity changes the final number of ICs in GMR>cindr RNAi retinas
To confirm that JNK signaling activity mediates death of ICs when cindr RNAi transgenes are expressed, we extended our analyses to include additional bsk alleles and alleles of additional components of the JNK pathway. This time we restricted our analyses to examining only the number of ICs within retinas at 40 h APF (Figure 4, Figure S8, Table 2). As before, ectopic bsk in GMR>cindr RNAi retinas severely reduced the number of ICs ( Figure 4C and D), whilst mutations in the kinases msn, slpr, hep, and bsk, as well as the transcription factors jun and/or fos increased the number of ICs in GMR>cindr RNAi retinas ( Figure 4E to P, Table 2). Partial suppression of the cindr RNAi -induced cell death indicates that Cindr interacts with additional molecular mechanisms that are independent of JNK signaling to protect ICs from apoptosis.

At least some JNK activity is required during apoptosis of ICs
Surprisingly, mutations in slpr and bsk increased the number of ICs in GMR>cindr RNAi to above the usual wild type number of 12. These observations led us to question whether JNK activity made at least a minor contribution to normal IC apoptosis -this activity might have been compromised in slpr and bsk heterozygotes. Indeed, although the number of cells removed from bsk 1/+ retinas between 18 and 40 h APF (an average loss of 7.18 cells per data point) was similar to the number of cells removed in control GMR>lacZ retinas (7.36 cells per data point, Table 1), the severity of Dcp-1 activity revealed a modest delay in significant pruning of ICs (Figure 2A). Reducing expression or function of the JNK kinases or Jun or Fos during pupal development lead to occasional extra ICs at 40 h APF ( Figure S9), although the average number of ICs observed across entire retinas was not significantly modified (Table 3). We conclude that at least some JNK activity is required for the normal progression of apoptosis in the eye field and that some JNK activity escapes repression to enhance efficient removal of ICs, although JNK is not a prominent driver of IC death. Our experimental approach, however, has not addressed whether other apoptosis-inducing mechanisms function redundantly with JNK to ensure efficient removal of ICs.

JNK activity contributes to IC intercalation
In addition to reducing the number of ICs within the lattice, reducing expression of cindr introduced defects in the arrangement and shape of ICs ( Figure 1D, Figure 4C). 1° cells and bristle groups were also frequently improperly positioned. These defects distorted the hexagonal lattice and can be quantified as an ommatidial mis-patterning score (OMS) ( Table  4) (Johnson and Cagan, 2009). Mutations in JNK pathway components modified the frequency of these patterning defects ( Figure 4E to M, Table 4). Patterning was modestly improved by mutations in hep, jun and fos but mutations in msn, slpr and bsk, which increased the number of ICs to above 12, increased cindr RNAi -induced patterning errors.
Many ICs, failed to adopt the correct positions and shapes: this was especially evident in genotypes characterized by additional cells.
To assess which cell behaviors were modified by interactions between cindr and JNK, we examined retinas at 24 h APF expressing cindr RNAi in the setting of reduced or compromised bsk expression ( Figure 5). In control GMR>lacZ retinas, most 1° cells were of comparative size and scalloped in shape, and the ICs had intercalated and arranged into single file between ommatidia by 24 h APF ( Figure 5A). These characteristics were not modified when bsk was over-expressed ( Figure 5B) nor in bsk heterozygotes ( Figure 5C). In GMR>cindrR NAi retinas, 1°s were variable in size and shape, scalloping often less pronounced and some 1° pairs failed to 'seal' properly around the four cone cells ( Figure  5D). In addition, ECad was not distributed about the entire periphery of cells in GMR>cindr RNAi retinas, indicating defects in the apical adherens junctions that suggest weakened adhesion between cells. ICs were also large and adopted random shapes in comparison to those in control GMR>lacZ retinas ( Figure 5D). These defects were not markedly modified when bsk levels were modified ( Figure 5E and F). However, reducing bsk in GMR>cindr RNAi retinas modified the arrangement of ICs, which remained grouped rather than positioned in single file in many places between ommatidia ( Figure 5F). These patterning defects were not resolved by 40 h APF ( Figure 4H and I). ICs are usually repositioned into single file via intercalation, a process that is mediated by actin and junction remodeling and changes in cell shape and which can occur even when too many ICs populate the lattice (Johnson et al., 2011;Larson et al., 2010). Importantly, errors in cell shape and arrangement were observed in retinas with reduced JNK activity, although these patterning defects were not widespread ( Figure S9). Taken together, these data indicate that JNK activity promotes lattice patterning by fine-tuning IC shape and position.

How does Cindr inhibit JNK signaling?
We have previously co-immunoprecipitated Cindr and Bsk from Drosophila embryos (Yasin et al., 2016), suggesting that these form a complex in vivo. Hence we hypothesized that Cindr could recruit ubiquitin-conjugating enzymes to direct Bsk degradation or recruit a phosphatase to trigger Bsk inactivation.
To test these hypotheses, we over-expressed Cindr in the embryo using daughterless-GAL4 or in the larval wing with c765-GAL4. Despite an eight-fold increase in Cindr in the wing (mean=7.95 fold increase, SD=2.13) and 13-fold increase in the embryo (mean=12.79 fold increase, SD=4.84), Bsk decreased by an average of only 7% in embryo and 12% in wing lysates ( Figure 6A to D). Phosphorylated Bsk similarly decreased by only 8% in wing discs ( Figure 6C to D). In addition, RNAi transgenes that reduced the availability of Cindr by up to 61% (SD=13%) did not significantly modify the amount of Bsk in embryos ( Figure 6A and B). Reducing cindr through the entire wing disc led to widespread cell death, precluding further analysis of this tissue. Given these modest changes in Bsk concentration and phosphorylation, it is unlikely that Cindr limits JNK signaling by regulating the amount or phosphorylation state of Bsk.

Conclusions
Our genetic data indicate that the conserved adaptor protein Cindr antagonizes JNK signaling in the pupal eye. This interaction is crucial to protect the interommatidial cell lattice from rampant apoptosis during development. Hence Cindr contributes to the mechanism that ensures an appropriate number of cells populate the honeycomb lattice to correctly pattern the eye field.
Our experimental strategies also uncovered a minor role for JNK in fine-tuning the shapes and positions of ICs. That JNK contributes to IC shape is consistent with JNK's wellcharacterized role in regulating cell shape in other developmental contexts. For example, during dorsal closure in the Drosophila embryo, JNK activity in the single row of lateral leading edge cells causes cells to elongate, contributing to the stretching of the lateral epithelial sheets to cover the amnioserosa and seal the embryo (Rios- Barrera and Riesgo-Escovar, 2013;Stronach, 2005). JNK similarly drives cell elongation during closure of the Drosophila thorax. In addition, genetic manipulations that activate JNK can profoundly alter the shapes of epithelial cells: for example, activating JNK in the peripodial membrane causes these hexagonal cells to elongate (Tripura et al., 2011). Additional experiments are required to determine whether JNK activity similarly contributes to shaping the epithelial cells of the developing fly pupal eye. In particular, live-imaging studies will be needed to ascertain whether JNK contributes to cell elongations that drive the intercalation of interommatidial cells from multiple to single rows (Hellerman et al., 2015). On the other hand, JNK activity has also been implicated in regulating adhesive junctions (Llense and Martin-Blanco, 2008;Wang et al., 2010) and it is possible that JNK activity reinforces adhesion during or after IC intercalation. A similar role has been ascribed to JNK during mammalian gut elongation (Dush and Nascone-Yoder, 2013).
The mechanism by which Cindr antagonizes JNK signaling remains to be resolved. Cindr and Bsk interact in vivo (these co-immunoprecipitate from Drosophila embryos) and interactions between Cindr and Bsk have been detected in two yeast 2-hybrid screens (Giot et al., 2003;Stanyon et al., 2004;Yasin et al., 2016). These data imply that Cindr complexes with Bsk in epithelia. Because Cindr lacks enzymatic activity, we hypothesized that appropriate enzymes are recruited to Cindr-Bsk complexes to regulate the Bsk kinase. However, our biochemical analyses do not support a model in which Cindr promotes JNK degradation via recruitment of the ubiquitin ligase machinery. Similarly, it is unlikely that Cindr promotes JNK inactivity via recruitment of a phosphatase. Instead, we propose that Cindr may modulate activity of the JNK pathway by sequestering Bsk away from its effector targets including the AP-1 transcription factors. Alternatively, Cindr may simply outcompete the AP-1 proteins in their quest to bind activated Bsk.
Cindr is richly expressed in Drosophila tissues and we therefore suggest that Cindr provides a general and effective mechanism to limit JNK signaling that might otherwise severely modify tissue morphology and function. Indeed, in our genetic analyses, ectopic expression of bsk was insufficient to induce JNK activity unless expression of cindr was reduced. Since we have observed similar control of JNK by Cindr in the developing wing epithelium (Yasin et al., 2016), we hypothesize that Cindr is a general regulator of JNK activity. How then do the JNK kinases overcome Cindr's repression in order to trigger signaling in the different developmental contexts that require JNK? In the Drosophila pupal retina, it is possible that the TNF receptor ligand eiger is spatially and temporally expressed and triggers high levels of JNK activity that momentarily overcome Cindr repression. Alternatively, Cindr is modified, displaced or sequestered, releasing JNK from its inhibitory grip.
Besides antagonizing JNK to promote IC survival, Cindr fulfills other important roles in tissues, including regulation of the cytoskeleton, junctions, vesicular trafficking and cytokinesis (Eikenes et al., 2013;Haglund et al., 2010;Johnson et al., 2012;Johnson et al., 2011;Johnson et al., 2008;Quinones et al., 2010). Indeed, Cindr contains a variety of interaction motifs to facilitate protein interactions but few of these interactions have been well-characterized. In this manuscript we describe that reducing expression of the Cindr isoforms containing the N-terminal SH3 domains impaired the survival of retinal ICs (although as discussed, additional patterning defects, independent of IC number, were also evident). Our data implies that Bsk is amongst the proteins that dock with these Cindr isoforms to influence cell death/survival decisions. However, our data also implies that other signals, independent of JNK, also interact to modify the survival or death of retinal cells.
The vertebrate ortholog of Cindr, CD2AP, is also important for cell survival, but this has been ascribed to PI3-K/Akt signaling and TGF-β activity Huber et al., 2003;Schiffer et al., 2004) Whether Cindr regulates the orthologous signals in the Drosophila pupal eye remains to be investigated. Similarly, whether CD2AP also limits JNK signaling in vertebrate cells requires investigation. Additionally, variants of CD2AP have been associated with susceptibility to Alzheimer disease (Chouraki and Seshadri, 2014;Karch and Goate, 2015). Experimental data indicates that CD2AP protects neurons from Tau-mediated toxicity (Shulman et al., 2013;Shulman et al., 2014). Whether this is because CD2AP protects neurons from JNK-mediated apoptosis deserves testing.

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

•
Cindr limits JNK signaling to promote the survival of epithelial cells in the Drosophila pupal retina: if unchecked, ectopic JNK activity triggers apoptosis of many cells in the retina.

•
Our data suggests that Cindr exerts widespread repression of JNK in the retina, but this is not via regulating Bsk levels nor phosphorylation.

•
Despite repression, at least some JNK is activated and contributes to efficient removal of superfluous cells.
• JNK activity also contributes to patterning of the interommatidial cell lattice, although this role is minor.    Figure S7 for further statistical analyses determining differences between each genotype at each age.    1 Interommatidial cells, 1° cells, and bristle groups were included in cell-counts (see Figure S1C).

2
For each genotype, mean values were calculated from analyses of 67 to 160 data points distributed across 10 to 16 independent retinas.
Dev Biol. Author manuscript; available in PMC 2019 January 01.