To determine the contribution of caspases to growth inhibition in emc mutant cells, we used the Flippase (FLP)/FLP Recombinase Target (FRT) system (Xu and Rubin, 1993) to generate clones of emc mutant cells that were unable to activate caspases normally. We achieved this by homozygously deleting the linked reaper (rpr), head involution defective (hid), and grim genes using Df(3L)H99. Rpr, Hid, and Grim proteins promote Diap1 degradation in response to apoptotic stimuli and allow activation of initiator caspases such as dronc to start apoptosis (Yoo et al., 2002). By removing all three pro-apoptotic proteins, Df(3L)H99 affects both apoptotic and non-apoptotic caspase functions in Drosophila (Tapadia and Gautam, 2011; White et al., 1994). Thus, clones of emc Df(3L)H99 mutant cells should be defective for caspase activation. In addition, we also made clones of emc mutant cells also mutated for dronc, which encodes the main initiator caspase in Drosophila that is necessary for most developmental apoptosis (Hawkins et al., 2000; Meier et al., 2000; Quinn et al., 2000). Dronc contributes to non-apoptotic caspase functions as well, so emc dronc mutant cells should also show reduced non-apoptotic and well as apoptotic caspase functions.

In comparison to neutral clones, which grew equivalently to their twin-spot controls, emc mutant clones showed greatly reduced growth in eye or wing imaginal discs, as described previously (Figure 1A–D, Andrade-Zapata and Baonza, 2014; Bhattacharya and Baker, 2011; Alonso and Garcia-Bellido, 1988). In contrast, emc mutant clones that were also mutant for dronc (droncⁱ²⁹; Xu et al., 2005), or emc Df(3L)H99 clones that lacked the rpr, grim, and hid genes, grew more like their twin-spot controls in both eye and wing (Figure 1E–H). Statistically, growth of emc clones also deleted for the rpr, hid, and grim genes was not distinguishable from the wild-type, and similarly for emc clones also mutated for dronc (see Figure 1N for quantification). Neither homozygosity for dronc, nor deletion of rpr, grim, and hid, significantly affected growth of otherwise wild-type imaginal disc clones (Figure 1I–L).

Previous studies of emc phenotypes have often used Minute genetic backgrounds (i.e. heterozygosity for mutations in Rp genes) to retard growth and enhance the size of the emc mutant clones (Bhattacharya and Baker, 2009). When emc Df(3L)H99 clones were induced in the M(3)67C background, the clones took over almost the entire disc, leaving only a few M(3)67C heterozygous cells remaining (Figure 1M). The growth advantage of emc dronc clones was not so marked.

These results indicate that the growth disadvantage of emc mutant imaginal disc clones is mostly attributable to cell death genes. Preventing caspase activation, either by mutating the main initiator caspase, or by preventing Diap1 turnover, partially or completed restored normal growth. Our data did not support a previous suggestion that dronc was required for normal wing disc growth (Verghese et al., 2012).

Emc mutations have multiple effects on the eye imaginal disc, although only a few aspects of retinal differentiation depend on proneural bHLH genes. The proneural gene atonal is required, along with da, for the specification of R8 photoreceptor precursors in the morphogenetic furrow that initiate each ommatidial cluster in the larval eye disc (Brown et al., 1996; Jarman et al., 1994). Later, during pupal development, proneural genes of the Achaete-Scute gene Complex (AS-C), along with da, are required for the specification of the interommatidial bristles (Cadigan et al., 2002). All the other cell types develop independently of proneural bHLH genes, and most of them develop independently of da (Brown et al., 1996; Jiménez and Campos-Ortega, 1987).

In emc mutant clones, retinal differentiation begins precociously, associated with more rapid transit of the morphogenetic furrow across the disc (Figure 2A, B, Bhattacharya and Baker, 2011; Bhattacharya and Baker, 2012; Brown et al., 1995). In contrast, we found that 75% of the time, the morphogenetic furrow progressed normally through emc dronc double mutant clones (Figure 2C). Eye discs containing emcDf(3L)H99 double mutant clones always appeared completely normal (Figure 2D). Control dronc and Df(3L)H99 mutant clones that lacked emc mutations also showed normal furrow progression (Figure 2—figure supplement 1A, B).

In addition, emc mutant clones also exhibit sporadic ectopic differentiation of neurons anterior to the morphogenetic furrow, which do not take photoreceptor cell fate (Figure 2A, B; Bhattacharya and Baker, 2011; Brown et al., 1995). This was likewise absent from emc dronc and emcDf(3L)H99 double mutant clones (Figure 2C, D).

The overall pattern of retinal differentiation revealed by labeling for Senseless, which is specific for R8 photoreceptor cells in the retina, and Elav, which labels all neuronal photoreceptor cells, appeared so normal in these genotypes that we deemed it necessary to confirm that the supposedly emc dronc mutant and emc Df(3L)H99 mutant clones were indeed mutated for emc. This was confirmed by lack of staining by an antibody against Emc protein, similar to plain emc mutant clones that did affect the morphogenetic furrow (Figure 2—figure supplement 1C–E). We also found that emc dronc and emc Df(3L)H99 clones upregulated Da protein expression to a comparable degree to plain emc mutant clones (Figure 2—figure supplement 1C–E, Bhattacharya and Baker, 2011). This further confirmed the absence of emc function from emc dronc mutant and emc Df(3L)H99 genotypes, and also showed that cell death pathways were not required for the regulation of Da protein levels by emc.

We have previously concluded that diap1 transcription is reduced in emc mutant clones, and in Da-overexpressing cells, due to Yki inhibition downstream of ex (Wang and Baker, 2015b). In eye imaginal discs, Diap1 protein was normally present uniformly (Figure 3A). Diap1 protein levels were cell-autonomously reduced in emc mutant clones posterior to the morphogenetic furrow, compared to emc/+ cells in the same eye discs, and shown and quantified in (Figure 3B, E). A comparable reduction was also seen in emc Df(3L)H99 clones, compared to emc/+ Df(3L)H99/+ cells in the same eye discs (Figure 3C, E). There was also no difference in Diap1 levels between Df(3L)H99 homozygous clones and Df(3L)H99/+ cells in the same eye discs (Figure 3D, E). These data suggest emc affects Diap1 protein levels, whereas rpr, grim, and hid affect Diap1 protein activity. A caveat is that Diap1 levels in mutant clones were not compared directly to wild-type cells, none of which were present in the same eye discs, and that Df(3L)H99 might affect Diap1 protein levels dominantly. Interestingly, we did not detect emc-dependent changes in Diap1 levels in wing discs (Figure 3—figure supplement 1B).

To test the role of Diap1 in the eye directly, diap1 was over-expressed in the posterior eye by transcription under GMR-Gal4 control. This rescued morphogenetic furrow speed to normal in emc clones (Figure 3—figure supplement 2B). We also saw restoration of morphogenetic furrow speed in emc clones that expressed Baculovirus P35 under GMR-Gal4 control (Figure 3—figure supplement 2A). Baculovirus P35 encodes a caspase pseudo-substrate that inhibits all Drosophila caspases except Dronc (Hawkins et al., 2000; Hay et al., 1994; Meier et al., 2000; Xue and Horvitz, 1995). The emc clones expressing p35 also lacked cell death (Figure 3—figure supplement 3B). The restoration of morphogenetic furrow speed by Diap1, a Dronc antagonist, as well as by the caspase inhibitor P35, suggest that the effect of morphogenetic furrow acceleration in emc mutant clones is due to the caspase cascade.

To test whether reduced Diap1 expression promoted apoptosis and thereby accelerated the morphogenetic furrow, we assessed apoptosis levels by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) of eye discs containing mutant clones. We found almost no cell death posterior to the furrow in emc clones (Figure 3—figure supplement 2D), and cell death in emc clones anterior to the furrow was comparable to controls and less than that of cells surrounding the clones (Figure 3—figure supplement 2C, D). Some cell death is expected outside emc or control clones, as these cells have the M background that is itself associated with an increase in apoptosis (Coelho et al., 2005; Kale et al., 2015; Li and Baker, 2007). Similar to TUNEL, we found no cell death in emc clones posterior the furrow with Dcp1 staining (Figure 3—figure supplement 3D). To test whether apoptosis in the eye disc would be sufficient to promote morphogenetic furrow progression, we generated mosaic clones for a pineapple eye (pie) mutation. These pie mutant cells have an elevated rate of apoptosis in imaginal discs, but not sufficient to prevent pie homozygous clones surviving late into larval and even adult life (Shi et al., 2003). The rate of morphogenetic furrow progression was unaffected in eye discs containing pie clones (Figure 3—figure supplement 3A). Because no excess cell death was detected in emc clones, and cell death was insufficient to accelerate the morphogenetic furrow in otherwise wild-type eye discs, emc clones might be affected by a non-apoptotic caspase activity.

To understand how caspases could affect the speed of the morphogenetic furrow, we analyzed pathways known to contribute. Hedgehog (Hh) and Decapentaplegic (Dpp) signaling drive this differentiation wave, along with a contribution from Notch signaling (Baonza and Freeman, 2001; Borod and Heberlein, 1998; Fu and Baker, 2003; Heberlein et al., 1993; Ma et al., 1993). A negative regulator of morphogenetic furrow progression is Wingless (Wg), which is expressed at the dorsal and ventral eye disc margins (Lee and Treisman, 2002; Maurel-Zaffran and Treisman, 2000).

Because we found that ex mutations affected thoracic bristle patterning through a caspase-dependent non-apoptotic effect on Wg signaling (Wang and Baker, 2019), we looked first to see whether emc mutations reduced Wg signaling in the eye. We used Frizzled-3 RFP (Fz3-RFP) as a reporter (Sato et al., 1999). In control eye discs, the Fz3-RFP recapitulates the pattern of endogenous Wg signaling activity at the wing margins (Figure 4—figure supplement 1, Treisman and Rubin, 1995). Frizzled-3 RFP expression was not changed in emc clones (Figure 4A). We then used a mutation in naked cuticle (nkd), encoding a negative feedback regulator of Wg signaling, to modulate Wg signaling (Chang et al., 2008; Zeng et al., 2000). If the morphogenetic furrow was accelerated in emc mutant clones due to reduced Wg signaling, more normal development should occur in emc nkd clones. The morphogenetic furrow was still accelerated in emc nkd clones, however (Figure 4B). These results provided no evidence that Wg signaling was the relevant emc target in the eye.

We also examined Dpp signaling, since ectopic Dpp signaling is sufficient to accelerate the morphogenetic furrow (Pignoni and Zipursky, 1997). The pattern of pMad, a readout of Dpp signaling, was identical in emc mutant and control clones spanning the morphogenetic furrow (Figure 4C, D). Thus, Dpp signaling did not seem altered by emc mutants either.

Hedgehog (Hh) signaling is a key mover of the morphogenetic furrow (Heberlein et al., 1993; Ma et al., 1993; Treisman, 2013). Elevated Hh signaling is sufficient to accelerate the morphogenetic furrow (Heberlein et al., 1995; Ma and Moses, 1995). Notably, it has been suggested previously that emc mutations affect the morphogenetic furrow by activating Hedgehog signaling, because emc mutant cells accumulate Ci protein (Spratford and Kumar, 2013). Full-length Ci protein (Ci155) is targeted to the proteosome by Cul1 for processing into a transcriptional repressor protein Ci75 (Aza-Blanc et al., 1997). By inhibiting this processing, Hh prevents repression of target genes by Ci75 and promotes transcriptional activation downstream of Ci155. Accordingly, Ci155 accumulation is a feature of cells receiving Hh signals (Motzny and Holmgren, 1995).

We confirmed that emc mutant cells contain higher levels of Ci155, as reported previously (Figure 5A, Spratford and Kumar, 2013). Ci155 was elevated in emc dronc clones (Figure 5B) but reduced to wild-type levels in emc Df(3L)H99 clones (Figure 5C, Figure 5—figure supplement 1). Thus, Ci155 levels did not correlate perfectly with behavior of the morphogenetic furrow.

We noticed that Ci155 levels were elevated in emc mutant clones in the posterior, differentiating eye disc, as well as in and ahead of the morphogenetic furrow (Figure 5A). This is significant, because Ci155 is not affected by Hh-dependent Cul1 processing posterior to the morphogenetic furrow (Baker et al., 2009; Ou et al., 2002). Ci155 accumulation posterior to the furrow suggests a Hh-independent mechanism.

To test whether Ci155 accumulation in emc clones indicates elevated Hh target signaling, we looked at the transmembrane receptor Patched (Ptc) which is a transcriptional target of Hh signaling, acting in a negative feedback loop (Hepker et al., 1997). We saw no changes in Ptc protein levels in either emc or emc Df(3L)H99 clones compared to controls, questioning the notion that Hh signaling was altered by emc (Figure 5—figure supplement 2).

To test functionally whether Ci155 is responsible for accelerating the furrow in emc clones, we generated emc ci double mutant clones. To achieve this, a genomic transgene that rescues ci⁹⁴ flies to adulthood (Little et al., 2020), was introduced into chromosome arm 3L where it is linked to the wild-type emc locus, so that mitotic recombination in the ci⁹⁴ null background leads to emc ci double mutant clones. For unknown reasons, emc ci double mutant clones were small and difficult to obtain, even in the Minute background. Eye differentiation was still accelerated in those emc ci double mutant clones we found that spanned the morphogenetic furrow (Figure 5D). The unexplained synergistic growth effects in particular could be consistent with interactions between emc and Hh signaling, but emc must regulate the speed of the morphogenetic furrow through at least one other target besides Ci in order to explain the furrow acceleration observed in emc ci double mutant clones.

The remaining signaling pathway that contributes to morphogenetic furrow movement is Notch. Specifically, only cells where Notch signaling is active are competent to initiate retinal differentiation in response to Dpp (Baonza and Freeman, 2001; Fu and Baker, 2003). Accordingly, ectopic expression of Delta, the transmembrane ligand for Notch, is sufficient to accelerate retinal differentiation anterior to the morphogenetic furrow by expanding the effective range of Dpp signaling (Baonza and Freeman, 2001; Li and Baker, 2001).

To test whether emc restrains Notch, we examined the bHLH proteins of the E(spl)-C, widely characterized targets of the Notch pathway (Bray, 2006). We used mAb323 to detect up to five E(spl) bHLH proteins with Notch-dependent expression (Jennings et al., 1994). E(spl) protein expression, and hence Notch signaling, was higher in the morphogenetic furrow in emc clones than in wild-type cells (Figure 6A). In contrast, E(spl) protein expression was normal in emc dronc clones (Figure 6B).

To test whether elevated Notch signaling was required to accelerate the morphogenetic furrow in emc mutants, we examined emc psn double mutant clones. Presenilin (Psn) is the enzymatic component of γ-secretase that releases the intracellular domain of Notch during active Notch signaling, and the psn gene is linked to emc. Loss of psn function leads to a Notch loss of function phenotype (Struhl and Greenwald, 1999; Ye et al., 1999). Accordingly, psn clones lead to a neurogenic phenotype in the eye, without affecting the progression of the morphogenetic furrow (Figure 6C, Li and Baker, 2001). The position of the morphogenetic furrow was also not affected in emc psn clones (Figure 6D). Thus, the furrow was not accelerated in emc mutant clones also defective for Notch signaling.

These two results together indicated that emc mutants promoted Notch signaling in the morphogenetic furrow, acting through caspase signaling on a step prior to γ-secretase cleavage of the intracellular domain of Notch. Accordingly, we decided to check Delta (Dl) protein levels. We found that Dl protein levels were significantly and consistently elevated cell-autonomously throughout emc clones, both posterior and anterior to the furrow (Figure 7B, E). This included the region just ahead of the morphogenetic furrow that lacks Delta expression in normal development (Baker and Yu, 1998; Parks et al., 1995, Figure 7B). In contrast, levels of Dl protein in emc Df(3L)H99 clones were similar to those of wild-type controls (Figure 7A, D, E). Levels in emc dronc clones were also similar to wild-type on average, although some clones seemed to show an increase, smaller than in emc clones (Figure 7A, C, E). This indicated that Dl protein is a target of caspases, directly or indirectly. The Delta protein sequence contains multiple predicted caspase target sites, as do many proteins (Wang et al., 2014). Only one candidate site lies in the intracellular domain, where it would potentially be accessible to caspases (Figure 7—figure supplement 1). Truncation of the Dl intracellular domain is usually associated with loss of Dl function, not stabilization and enhanced function, however (Daskalaki et al., 2011).

Because Dl activates Notch signaling cell non-autonomously, we wondered whether the effect of emc mutant clones was cell-autonomous. We note that, in all the experiments reported here, and in all the previous studies of emc mutant clones affecting morphogenetic furrow movement, the morphogenetic furrow is maintained as a continuous groove across the eye disc (Bhattacharya and Baker, 2011; Bhattacharya and Baker, 2012; Brown et al., 1995; Spratford and Kumar, 2013). That is, the advanced front of retinal differentiation within emc clones is always smoothly continuous with the normal morphogenetic furrow outside the clones, implying a progressive gradual increase in morphogenetic furrow speed near the lateral edges of emc mutant clones (e.g. Figures 2B, 4B, D—6A). Clones of Su(H) null mutants, which act cell-autonomously, provide a contrasting example. Complete loss of Su(H) accelerates the morphogenetic furrow due to loss of default Su(H) repression of Notch targets (Li and Baker, 2001). The boundaries of Su(H) null clones exhibit a clear discontinuity between the rates of differentiation within and outside the clones (Figure 6E). This difference between Su(H) and emc mutant clones is consistent with the idea that emc affects morphogenetic furrow progression differently from Su(H).

Because the general pattern of neurogenesis revealed by pan-neuronal anti-Elav staining appeared so normal in emc dronc and emc Df(3L)H99 clones (Figure 2C, D), we examined whether effects of emc on particular retinal cell fates was caspase dependent. Emc is also required for R7 differentiation and for timely onset of cone cell differentiation (Bhattacharya and Baker, 2009), two cell fate decisions that also depend on Notch signaling (Cooper and Bray, 2000; Flores et al., 2000; Tomlinson and Struhl, 2001; Treisman, 2013). These cell fates are normally independent of da, but like imaginal disc cell growth and morphogenetic furrow progression, ectopic da activity perturbs them in emc mutants (Brown et al., 1996; Reddy Onteddu et al., 2024).

Clones of emc mutant cells lack R7 photoreceptor cells (Figure 8A, B, Bhattacharya and Baker, 2009). R7 differentiation was restored to 90% of ommatidia in emc Df(3L)H99 clones (Figure 8C). As shown before, emc mutant clones delayed cone cell differentiation by two to three columns (corresponding to a delay of 3–6 hr) (Bhattacharya and Baker, 2009, Figure 8D, E). We found no delay in cone cell differentiation in most emc Df(3L)H99 clones (Figure 8F). These results indicated that caspase activity contributes to the R7 and cone cell differentiation defects that are also characteristic of emc mutants.

The following stocks were employed in this study and were maintained at 25°C unless otherwise stated – hsflp;emc^AP6 FRT80/TM6B, hsflp;droncⁱ²⁹FRT80/TM6B, hsflp;droncⁱ²⁹emc^AP6, hsflp;Df(3L)H99/TM3, hsflp;Df(3L)H99 emc ^AP6/TM3, Ubi-GFP M(3)67C FRT80, FRT42 M(2)56F Ubi-GFP, FRT82 M(3)95A Ubi-GFP, hsflp; Ubi-GFPFRT80, hsflp; Su(H)^Δ47FRT40, pie^eB3 FRT40, psn[v1]FRT80/TM6B, Fz3-RFP;D/TM6B (kind gift from Yu Kimata, University of Cambridge). We obtained genomic Ci construct from Kalderon lab and used BestGene Inc to target the transgene to the third chromosome and then recombined these flies to generate hsflp;ci+M(3)67Calz FRT80/TM6B;ci94/y+spa.

Mosaic clones were obtained using FLP/FRT-mediated mitotic recombination (Golic, 1991; Xu and Rubin, 1993). For non-Minute genotypes, larvae were subjected to heat shock for 30 minutes at 37°C, 60 ± 12 hr after egg laying. For Minute genotypes, heat shock was performed 84 ± 12 hr after egg laying for 50 minutes. Larvae were dissected 72 hr after heat shock. All flies were maintained at 25°C unless otherwise stated.

Clone and twin-spot areas were measured by tracing in ImageJ. To quantify the growth effects of various genotypes, the sum of clone areas per wing disc was divided by the sum of twin-spot areas in the same wing disc. This avoids any subjectivity in identifying individual clones and assigning them to individual twin spots. Clone/twin-spot ratios were log-transformed to ensure normality before statistical analysis.

Unless otherwise noted, preparation of eye and wing imaginal discs for immunostaining and confocal imaging were performed as described previously (Baker et al., 2014). Antibodies from Developmental Studies Hybridoma Bank (DSHB): anti-Ptc (mouse, 1:40), anti-Elav (mouse, 1:100), anti-Elav (rat, 1:50), anti-Cut (mouse, 1:50), anti-Delta C594.9B (mouse, 1:2000), anti-βGal (mouse, 1:100), mouse anti-βGal (1:100, DSHB 40-1a), and anti-Ci (rat, 1:10). Other antibodies: anti-phospho-Smad1/5 (rabbit, 1:100, Cell Signaling), anti-DIAP1 (rabbit, 1:50) (gift from Hyun Don Ryoo, NYU), anti-Dcp1 (rabbit, 1:100, Cell Signaling), anti-Sens (guinea pig, a gift from Hugo Bellen used at 1:500), anti-Da (mouse, 1:200), rabbit anti-Emc (1:8000), anti-GFP (rat, 1:50 from Nacalai Tesque # GF090R), rabbit anti-GFP (1:500), rabbit anti-β-Galactosidase (1:100, Cappel), E(spl)bHLH (1:50,mAb323), and guinea pig anti-runt (1:500). Secondary antibodies conjugated with Cy2, Cy3, and Cy5 dyes (1:200) were from Jackson ImmunoResearch Laboratories and Alexa 555 (1:500). Multi-labelling images were sequentially scanned with Leica SP8 confocal microscopes and were projected and processed with ImageJ. All images were assembled into figure format using Adobe Illustrator 2020.

Anti-DIAP1 and anti-Dl labeling were quantified within clones using the average density measurement in ImageJ and normalized to control regions in the same tissue. For anti-DIAP1 labeling, in particular, we lack any measurement of any non-specific background labeling that may occur. Differences in Diap1 levels between genotypes may be underestimated if a component of the labeling is non-specific.

For labeling dead cells with TUNEL assay, ApopTag Red In Situ Apoptosis Detection Kit (Cat #S7165) was used according to the manufacturer’s instruction. Briefly, dissected eye discs were fixed for 20 min at room temperature followed by three washes with 1× phosphate-buffered saline (PBS). Then the samples were incubated in equilibration buffer for 1 min followed by incubation in reaction buffer (TdT enzyme; ratio 7:3) at 37°C for 1 hr. TdT reaction mix was replaced with stop buffer (diluted 1:34 in dH₂O) and incubated for 10 min at room temperature. Samples were washed three times with 1× PBS, 5 min per wash; and incubated with anti-digoxigenin antibody solution (diluted 31:34 in blocking solution) for 30 min at room temperature. The samples were then washed three times in 1× PBS, 5 min per wash. For the subsequent antibody staining, the samples were blocked in PBST (1× PBS + 0.5% Triton-X) for 30 min, and incubated with primary antibodies in PBST overnight at 4°C. The samples were next washed with PBST and incubated for 2 hr with secondary antibodies in PBST, and then again washed with PBST, followed by PBS wash and samples were mounted in mounting media.

Statistical analysis was performed using GraphPad Prism 7. The statistical tests used are described in the figure legends. Statistical significance is shown as follows: n.s., p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Caspase cleavage sites were predicted for Delta using Cascleave 2.0 (Wang et al., 2014). Cascleave 2.0 was set to a medium stringency threshold for prediction of cleavage sites.

All new materials generated in this project are available from the corresponding author.