Different modes of APC/C activation control growth and neuron-glia interaction in the developing Drosophila eye

Development of multicellular organisms requires an intricate coordination of cell division and subsequent cell migration. A prominent model with which to study these processes is the developing compound eye of Drosophila. Here, several molecular mechanisms have been identified that control growth and tumorigenesis as well as neurogenesis, which in the eye disc epithelium occurs in a posterior-to-anterior fashion in the wake of the morphogenetic furrow (Dominguez, 2014; Freeman, 1997; Kumar, 2001, 2011). Differentiating photoreceptor neurons project their axons through the optic stalk towards the brain (Tayler and Garrity, 2003). In contrast, all glial cells that will eventually ensheath the photoreceptor axons are born in the central nervous system (CNS) and migrate onto the eye disc (Choi and Benzer, 1994; Franzdóttir et al., 2009; Rangarajan et al., 1999; Silies et al., 2007b). Therefore, the development of neurons and glial cells has to be coordinated to ensure the proper wrapping of the sequentially forming neuronal projections.

Migration of CNS-derived glial cells onto the eye disc is controlled in a two-stage process (Hummel et al., 2002). Early in development, eye disc cells prevent the ingrowth of glial cells by unknown signals, with ingrowth only triggered upon initiation of neurogenesis. To allow coordinated differentiation, glial cells move onto the eye disc on a glial sheath formed by the carpet glia that separates the migrating glial cells from differentiating axons. Thus, glial cells initially migrate independent of neuronal signals (Fig. 1A,B) (Silies et al., 2007b). Only when glial cells reach the edge of the glial sheath do they make contact with nascent axons. This contact stops their migration and triggers differentiation into wrapping glia along the photoreceptor axons back towards the brain in a fibroblast growth factor (FGF)-receptor dependent manner, ensuring the continuous ensheathment of the sequentially forming photoreceptor axons (Fig. 1B) (Franzdóttir et al., 2009; Sieglitz et al., 2013; Silies et al., 2007b).

Although there is a switch from glial-glial to glial-neuronal interactions in the eye imaginal disc, neuronal and glial development have to be coordinated in other contexts as well. During the development of the peripheral nervous system of the Drosophila embryo, glial cells are born in the CNS and migrate into the periphery along developing motoneurons. The coordination of axonal growth and glial cell migration is controlled by the anaphase-promoting complex/cyclosome (APC/C) (Silies and Klämbt, 2010).

The APC/C is a highly conserved multi-subunit E3 ubiquitin-protein ligase of the RING-domain family that regulates entry into anaphase and exit from mitosis (Peters, 2006; Sullivan and Morgan, 2007; Zachariae and Nasmyth, 1999). Most APC/C subunits are found in Drosophila, including the APC2 Cullin homolog Morula (Mr) and the APC11 RING-H2 finger protein homolog Lemming (Lmg) (Reed and Orr-Weaver, 1997; Silies and Klämbt, 2010). APC/C substrate specificity is conferred by its co-activator proteins Fizzy(Fzy)/Cdc20 and Fizzy-related(Fzr)/Cdh1. Fzr/Cdh1 functions as an adaptor, binding to the APC/C and to substrate proteins via its C-terminal WD40 repeats (Kraft et al., 2005). These substrates contain short amino acid sequence motifs, called destruction box or KEN box, that mediate the direct interaction with the activator subunits Fzy/Cdc20 or Fzr/Cdh1 (Burton et al., 2005; Glotzer et al., 1991; Peters, 2006; Pfleger and Kirschner, 2000). Fzr/Cdh1 function is regulated by a number of mechanisms, including phosphorylation and binding of the F-box protein Rca1/Emi1, which blocks substrate association of Fzr/Cdh1 (Blanco et al., 2000; Grosskortenhaus and Sprenger, 2002; Yamaguchi et al., 2000).

APC/C regulates mitosis but, together with Fzr/Cdh1, it is also expressed and required during neuronal differentiation (Gieffers et al., 1999; Harmey et al., 2009; Silies and Klämbt, 2010; Stegmüller et al., 2006). Expression of a dominant-negative Cdh1 variant suggests that APC/C^Cdh1 activity allows NGF-induced neurite outgrowth in a murine neuronal cell culture model (Harmey et al., 2009). However, RNA interference (RNAi)-mediated knockdown of murine cdh1 in primary neuronal cell cultures leads to increased axonal growth (Konishi et al., 2004). cdh1 knockout mice show early lethality, preventing an easy in vivo analysis (Garcí-Higuera et al., 2008). In the fruit fly, different postmitotic roles have been described for the APC/C. It controls synaptic size via the downstream effector Liprin-alpha at the larval neuromuscular junction, acts during the establishment of planar cell polarity and functions in neurons of the embryonic peripheral nervous system (PNS) to control glial migration (Silies and Klämbt, 2010; van Roessel et al., 2004; Weber and Mlodzik, 2017). As mentioned above, Drosophila peripheral glial cells migrate along the axons of the motoneurons, which are decorated with a gradient of the Ig-domain cell-adhesion molecule Fasciclin 2 (Fas2). Downregulation of Fas2 levels coincides with the onset of peripheral glial migration. In fzr/cdh1 mutants, the graded distribution of Fas2 is lost and glial cells remain in regions of high Fas2 expression (Silies and Klämbt, 2010; Silies et al., 2007a). Cell type-specific rescue experiments have demonstrated that fzr/cdh1 acts in motoneurons to control the expression of the adhesion protein Fas2, which in turn regulates glial migration (Silies and Klämbt, 2010). Given its independent roles in both cell cycle regulation and neuronal differentiation, the APC/C and its co-activators Fzr/Cdh1 and Fzy/Cdc20 are excellent candidates for coordinating different stages of development within one tissue.

Here, we have analyzed the role of fzy/cdc20 and fzr/cdh1 in the developing eye disc. Disruption of Fzy/Cdc20 or APC/C function blocks growth of the eye disc and allows a precocious entry of glial cells onto the disc. In contrast, loss of fzr/cdh1 affects photoreceptor axons and results in a glial-positioning phenotype specifically at the front of the developing eye field, reflecting defects in neuron-glia interaction. Thus, two distinct modes of APC/C-dependent protein degradation act during development. We then identified putative target proteins that control neuron-glia interaction downstream of fzr/cdh1. The function of two candidates, Dlc90F and CG15765, was verified by mutant analysis. In summary, we show that APC/C has distinct roles during eye development, linking cell division to axon-glia interaction and identify several genes acting downstream of Fzr/Cdh1 to fulfill its manifold functions during neural development.

In the embryonic PNS of Drosophila, postmitotic Fzr/Cdh1 regulates axonal growth and non-autonomously controls glial migration (Silies and Klämbt, 2010). To test whether fzr/cdh1 functions more broadly in regulating these processes, we determined its role during eye development. As fzr/cdh1 null mutants are embryonic lethal (Sigrist and Lehner, 1997), we performed mitotic recombination experiments. We generated eye discs containing almost entirely mutant fzr/cdh1 cells in an otherwise heterozygous background, using Flp-recombinase expressed under the control of the eyeless (ey) promoter (Fig. 1C,D). As this promoter is not expressed in the developing brain lobes (Fig. S1A,B), the organization of the brain is not affected in these animals. In fzr/cdh1 mutant eye discs, photoreceptor neurons are present but ommatidial units contain variable numbers of photoreceptor neurons (Fig. 1) (Karpilow et al., 1989; Pimentel and Venkatesh, 2005). When we followed the projection of the photoreceptor axons into the larval brain, a reduced number of axon fascicles were present that showed targeting defects in the developing lamina and medulla of the larval brain (Fig. 1C,D).

We have previously shown that glial migration from the brain onto the eye disc requires glial-glial interactions, whereas the subsequent interaction with nascent axons at the anterior-most part of the differentiated eye disc triggers differentiation (Silies et al., 2007b). Initially, glial cells show a collective migration behavior and move as a sheath along the surface of the carpet glia, which shields them from axonal contact (Fig. 1A,B). As soon as they reach a zone closely posterior to the morphogenetic furrow, only the anterior-most migrating glial cells come in contact with nascent photoreceptor axons, which stops their anterior-ward migration (Fig. 1A,B). Axonal contact triggers differentiation and glial cells move around to the other side of the carpet glial sheath. Subsequently, glial cells adhere to the photoreceptor axons, migrate in the opposite direction towards the brain and ensheath the forming photoreceptor bundles (Fig. 1B).

Given the non-autonomous requirement of fzr/cdh1 in the development of the embryonic nervous system (Silies and Klämbt, 2010), we tested a possible non-autonomous contribution of fzr/cdh1 function in neuron-glia interaction in the eye disc. In wild-type eye discs, glial cells migrate onto the eye disc from posterior to anterior and the anterior-most glial cells are usually located a few cell rows posterior to the anterior-most differentiating photoreceptors (Fig. 1E). We then silenced fzr/cdh1 expression using two independent fzr/cdh1^dsRNA constructs together with eyeless-Gal4 (ey-Gal4), which is not expressed in glial cells (Fig. S1A,B). Upon knockdown of fzr/cdh, glial migration terminated further anteriorly compared with wild type, reaching the front of the differentiating eye field (Fig. 1F), whereas the UAS-fzr/cdh1^dsRNA construct alone did not cause any phenotype.

When we generated fzr/cdh1 mutant cell clones using ey-Flp, glial cells migrated further anteriorly than their wild-type neighbors in control areas of the same disc (Fig. 1G,H, Fig. S1C,D). This phenotype correlated with the extent of the mutant tissue, suggesting that neuron-glia interaction is specifically disrupted at the position where neuronal signals normally trigger glial differentiation (Fig. 1A,B). In contrast to the phenotype noted upon fzr/cdh1 knockdown in the entire eye disc (Fig. 1F), glial positioning appeared less affected by small mutant clones, whereas large clones showed a more pronounced phenotype, suggesting that perdurance of the Fzr/Cdh1 protein in clonal tissue mitigated the glial positioning defects (Fig. 1G,H, Fig. S1C,D). Because with this clonal approach only the eye disc epithelium, not the glia, is mutant, fzr/cdh1 has a non-autonomous function in the regulation of neuron-glia interaction in the eye disc.

Knockdown of fzr/cdh1 function using ey-Gal4 also resulted in neuronal phenotypes. Whereas neuronal nuclei are regularly clustered at the apical side of the wild-type eye disc, knockdown of fzr/cdh1 resulted in irregularly spaced photoreceptor clusters and misplacement of neuronal nuclei into the basal layer of the eye disc (Fig. 1I,J). Together, these data show that Fzr/Cdh1 concomitantly controls neuronal patterning and neuron-glial interaction.

We next addressed whether Fzr/Cdh1 acts through the APC/C to control neuron-glia interaction in the eye disc. We analyzed mutants in the two core subunits of the APC/C: APC11 [or Lemming (Lmg)]; and APC2 [or Morula (Mr)] (Fig. S1E). Whereas homozygous mutant lmg⁰³⁴²⁴ animals died as L1 or L2 larvae, lmg^EY11317 mutant animals reached third instar larval stages. The eye discs of these mutants were formed but failed to grow (Fig. S1F,G). This phenotype is reminiscent of the phenotype caused by the absence of the APC/C subunit APC5, imaginal discs arrested (ida) (Bentley et al., 2002). An optic stalk with glial cells was identifiable, but the lack of a differentiated eye field precluded further analysis. In trans-heterozygous lmg⁰³⁴²⁴/lmg^EY11317 third instar larvae, the eye discs developed further. Although the eye discs were smaller than in wild-type larvae, photoreceptor cell differentiation clearly took place (Fig. S1H-J). Glial cells often migrated in large streams to ectopic, anterior positions (Fig. S1H-J). In addition, ey-Flp-induced mutant cell clones were generated that displayed similar phenotypes. Whereas a regular glial cell pattern was seen in wild-type control clones, glial cells entered the undifferentiated eye field over broad regions in lmg^EY11317 clones (Fig. 2A,B). In contrast to the fzr/cdh1 mutant eye discs, glial cells migrated further anteriorly and the phenotype did not correlate with the borders of the mutant clones, suggesting that this phenotype originated earlier in eye disc development. Consistent with this, similar glial overshoot phenotypes were seen in mutants of early eye development genes such as hedgehog and sine oculis (Hummel et al., 2002), indicating an early failure in the control of initial glial entry to the differentiating eye field.

We then tested whether these phenotypes were also observed upon eye-specific knockdown of the APC/C core component morula (Fig. 2C). Here, a glial migration phenotype similar to the one observed following loss of lmg could be observed. Streams of glial cells overshot their normal boundaries and were found in regions far ahead of the morphogenetic furrow, again suggesting an early glial migration phenotype (Fig. 2C). We therefore tested whether this phenotype had already developed in the second instar stage, before photoreceptor development took place (Fig. 2D,E). No glial cells were found in second instar wild-type eye discs (Silies et al., 2007b). However, upon knockdown of morula in the ey-Gal4 pattern, glial cells entered the eye disc proper before the onset of neurogenesis (Fig. 2E), confirming that the migration phenotype originated early in development. Complete loss of mr led to an undeveloped eye disc with only few glial cells with large nuclei (Fig. 2F). In summary, the APC/C appears to have additional roles during early eye disc development that are independent of Fzr/Cdh1.

To further analyze the function of Fzr/Cdh1 and APC/C, we tested the contribution of Rca1 (Regulator of cyclin A1), which has been shown to suppress Fzr/Cdh1-mediated APC/C activity (Grosskortenhaus and Sprenger, 2002). We first asked whether Rca1 overexpression in the eye disc phenocopies the fzr/cdh1 mutant phenotype. Upon ey-Gal4-driven overexpression of Rca1, third instar larval eye discs were severely reduced in size, no photoreceptor cells developed and glial cells migrated precociously on the immature eye field (Fig. 2G), resembling phenotypes of APC/C core components, but not Fzr/Cdh1. To test whether Rca1 overexpression phenocopied fzr/cdh1 phenotypes in later eye disc development, we used a temperature-sensitive repressor of Gal4 (McGuire et al., 2003) to modulate expression of Rca1 temporally. At the permissive temperature (18°C), tub-Gal80^ts ey-Gal4 UAS-Rca1 animals develop normally. When larvae were kept for 2 days after initiation of eye disc development at the restrictive temperature (29°C), the patterning of the eye disc appeared normal, but glial cells migrated too far anteriorly (Fig. 2H). This phenotype induced by Rca1 overexpression was again reminiscent of the ones caused by loss of APC subunits, but cannot be explained by Rca1-evoked reduction of Fzr/Cdh1 activity. We therefore explored the hypothesis that Rca1 might also interfere with Fzy/Cdc20 activity and that the qualitatively different phenotypes in eye disc growth and early glial entry, on the one hand, and neuronal patterning and glial positioning, on the other hand, might be caused by APC/C activation through either Fzr/Cdh1 or Fzy/Cdc20. We thus silenced fzy/cdc20 function in the eye disc by expressing a UAS-fzy^dsRNA construct. Indeed, similar to the Rca1 gain-of-function experiments or the APC/C subunit loss-of-function experiments, we observed a broad excessive migration phenotype of glial cells into the undifferentiated part of the developing eye disc (Fig. 2I). These data suggest that Rca1 also interacts with Fzy/Cdc20. Additionally, upon fzy/cdc20 knockdown, glial cells again already enter the undifferentiated eye field in L2 larvae (Fig. S1K,L). In summary, Fzy/Cdc20 and Fzr/Cdh1 activate APC/C at different stages during eye development. Whereas Fzy/Cdc20 acts during early stages and controls glial entry upon the differentiating eye field, Fzr/Cdh1 controls neuronal patterning and, concomitantly, neuron-glia interaction.

In the embryonic PNS, Fzr/Cdh1 functions in neurons to control glial cell migration by regulating the neuronal expression of Fas2 (Silies and Klämbt, 2010). Fas2 is also expressed in the developing eye disc where it is found in cells posterior to the morphogenetic furrow, including photoreceptor neurons (Mao and Freeman, 2009) (Fig. 3A,B). When we generated mutant cell clones of the amorphic fzr/cdh1^8F3 allele using ey-Flp, Fas2 expression was indistinguishable between the homozygous mutant fzr/cdh1 clones and the surrounding heterozygous tissue (Fig. 3A,B). To further test whether Fas2 participates in the control of glial migration in the eye disc, we generated fas2 fzr/cdh1 double mutant clones (Fig. 3C,D; Fig. S2). Eye discs containing single fas2^EB112 mutant clones did not show any glial phenotypes (data not shown). Although fas2 mutant clones were smaller than controls, both in a wild-type as well as in a fzr/cdh1 mutant background, fzr/cdh1 mutant phenotypes were not suppressed by the loss of fas2. In double mutant fas2 fzr/cdh1 clones, glial cells still migrated a few cell rows too far, as they did in fzr/cdh1 single mutants (Fig. 3C,D, Fig. 1G,H). Furthermore, the abnormal basal positioning of the mutant photoreceptor nuclei was still observed in fas2 fzr/cdh1 double mutant clones (Fig. S2).

At the end of development, the adult eye displays regular organization and ommatidia are arranged in hexagonal arrays. In adult flies carrying fzr/cdh1 mutant clones, this crystalline-like structure was disrupted and adult fly eyes showed a ‘rough eye’ phenotype, whereas loss of fas2 alone did not affect the regular pattern of the compound eye (Fig. 3E-G). fas2 fzr/cdh1 double mutant clones were dramatically reduced in size, but the ommatidia were still arranged in an irregular manner (Fig. 3H). In contrast to the embryonic PNS, loss of fas2 did not rescue any phenotypes caused by fzr/cdh1 mutants during eye development. Thus, unknown substrates are targeted by Fzr/Cdh1-activated APC/C to control neuron-glia interaction and patterning of the compound eye.

To identify Fzr/Cdh1 substrates controlling neuron-glia interaction in the eye disc, we developed a genetic suppressor screen employing the temperature dependence of the Gal4 system (Fig. 4). When we cultured flies expressing fzr/cdh1 dsRNA using the ubiquitously expressed da-Gal4 driver at 25°C, viable progeny were obtained. In contrast, culturing flies at 29°C resulted in lethality during second larval stage. When animals developed at 25°C for 2 days and were then transferred to 29°C, lethality was shifted to late pupal stages (Fig. 4). Late pupal lethality provided an easy read-out for a genetic suppressor screen. As reduction of fzr/cdh1 function is expected to lead to an accumulation of its downstream targets, we anticipated that concomitant knockdown of a Fzr/Cdh1 substrate might rescue the da-Gal4 UAS-fzr^dsRNA-induced pupal lethality to viability (Fig. 4).

We first selected proteins as substrate candidates that physically or genetically interact with Fzr/Cdh1 (Giot et al., 2003; Guruharsha et al., 2011; Stegmüller et al., 2006), including Drosophila homologs of potential Cdh1 target proteins expressed in nervous system tissue. In addition, we retrieved all D-box (“destruction box”) or KEN motif-containing proteins, two known targets for APC/C^Fzr/Cdh1 substrates expressed in the nervous system (Table S1). Of the 152 lines tested, 38 were able to suppress the pupal lethality of da>>fzr^dsRNA and flies eclosed (Table S1). Although this number is high and might include some false positives, the list of suppressor genes included well-known Fzr/Cdh1 interactors such as Cyclin B, Cyclin B3, Pimples (Drosophila Securin) and Drosophila SnoN (Peters, 2006; Stegmüller et al., 2006). As these substrates are associated with the function of Fzr/Cdh1 either in cell cycle control or in the control of axonal growth, it suggests that our assay was suitable to identify both mitotic as well as post-mitotic downstream targets of Fzr/Cdh1.

To find substrates affecting neuron-glia interaction in the developing eye disc, we tested which candidates could rescue the glial positioning phenotype evoked by silencing of fzr/cdh1 function. Knockdown of fzr/cdh1 function using ey-Gal4 resulted in an excessive migration of glial cells in about 65% of the eye discs (Fig. 5). We then tested whether concomitant silencing of candidates identified above suppressed this phenotype. The phenotypic rescue was determined according to the number of rows of photoreceptor neurons that were found anterior to the front of migrating glial cells. In wild type, glial nuclei are found about four rows behind the anterior-most differentiating photoreceptors. When we tested for suppression of the fzr/cdh1 knockdown phenotype, eye discs with three or four rows of photoreceptors ahead of a straight row of glial cells were considered as rescued (see Fig. S3 for examples). At least 17 eye discs were scored for the distribution of phenotypic classes for each genotype (Fig. 5).

Silencing of fzr/cdh1 in the eye disc caused an excessive glial migration phenotype in 65% of the eye discs (n=26). To exclude the possibility that the simultaneous expression of a second UAS-dsRNA construct could weaken the UAS-fzr^dsRNA-induced phenotype, we assayed 17 RNAi lines that did not suppress the da>>fzr^dsRNA-induced lethality. None of these control lines, e.g. Pak^dsRNA (phenotype in 64%, n=25), modified the fzr^dsRNA-induced glial migration phenotype (Fig. 5A,B,G).

Out of the candidate lines that were able to suppress da>>fzr^dsRNA-induced lethality, silencing of four candidate genes [Nrk (CG4007), Dlc90F (CG12363), CG15765 and KP78b (CG17216)] suppressed the ey>>fzr^dsRNA-induced glial migration phenotype using two RNAi lines each (Fig. 5, data not shown). Upon concomitant suppression of CG15765, we noted a 50% rescue and only 32% of the eye discs showed a glial migration phenotype (n=28, Fig. 5C,G). Upon suppression of Dlc90F, a 54% rescue was noted with 30% of the eye disc showing a glial migration phenotype (n=27, Fig. 5D,G). Suppression of KP78b caused the strongest rescue by 62%, leaving 25% of the eye discs with a migration phenotype (n=45, Fig. 5E,G) and upon simultaneous suppression of Nrk and fzr/cdh1 dsRNAs, 29% of the eye discs showed a glial migration phenotype (n=17, Fig. 5F,G). We next tested whether loss of any of these four candidates also suppresses the fzr/cdh1^dsRNA neuronal patterning phenotype. All four candidates also partially restored photoreceptor patterning, although neuronal nuclei could still be found in ectopic positions (Fig. 5A″-F″, data not shown). To test whether these four candidates are indeed expressed in the eye disc tissue, we referred to microarray data sets (FlyBase) and independently performed RT-PCR on isolated eye-brain complexes from third instar larvae. In all four cases, mRNA expression can be detected (FlyBase and Fig. S4).

Retinal patterning defects are often associated with a rough eye phenotype. We thus tested whether the knockdown of the four candidates also suppressed the adult eye phenotype caused by a reduction in fzr/cdh1 function using a scanning electron microscope (Fig. S5). Compared with controls, none of the four candidates knocked down suppressed the fzr/cdh1^dsRNA-induced adult rough eye phenotype (Fig. S5). Therefore, different genes act downstream of Fzr/Cdh1 to mediate neuronal and non-neuronal differentiation.

The above candidates were identified in an RNAi-based screen. If feasible, we tested whether mutants affecting these genes would lead to a similar genetic suppression of the glial migration phenotype in order to verify the above findings and exclude a rescue due to off-targets of the dsRNA constructs. KP78b encodes a calmodulin-dependent serine/threonine kinase that resides in a large intron of the prospero gene. Nrk encodes a tyrosine kinase. No KP78b- or Nrk-specific mutants or P-element insertions, that would allow one to easily generate a mutant, were available. However, the suppression of the ey>>fzr^dsRNA phenotype was verified using two independent RNAi lines for each of these two candidates. CG15765 encodes a large lectin-type protein and two FRT-bearing P-element insertions flanking the first coding exon could be used for Flp-mediated deletion generation (Parks et al., 2004) (Fig. S6A). Flies homozygous for CG15765^Δ1 are viable and fertile with no discernible phenotype. The glial positioning phenotype caused by suppression of fzr/cdh1 in the eye disc could be rescued by the CG15765^Δ1 allele. In control animals, where only fzr/cdh1 is silenced, 47 out of 69 imaginal discs showed a glial migration phenotype (68%, n=69, Fig. 6A,G) and all eye discs of the CG15765^Δ1 allele showed a normal neuronal and glial pattern (n=5, Fig. 6B,G). Upon loss of CG15765 in a fzr/cdh1^dsRNA background, only 23 out of 52 imaginal discs showed a phenotype (44%, n=52, Fig. 6C,G).

For the fourth candidate, Dlc90F, loss-of-function alleles were available. The Dlc90F^e155 allele is a null mutant (Li et al., 2004). Whereas heterozygous Dlc90F^e155 eye discs showed no phenotype (n=10, Fig. 6D,G), reducing the gene dose of Dlc90F^e155 suppresses the fzr/cdh1-induced glial migration phenotype (40% rescue, n=87, Fig. 6E,G). A similar phenotypic rescue was obtained using the transposon-induced Dlc90F^e00877 mutation (57% rescue, n=48, Fig. 6F,G). Whereas fzr/cdh1 knockdown phenotypes were suppressed by removing one copy of Dlc90F, the same was surprisingly no longer true if Dlc90F was completely missing, using either the Dlc90F^e155 or the Dlc90F^e00877 allele (Fig. 6G, Fig. S3D,E). This suggests that, upon complete loss of Dlc90F, other Fzr/Cdh1 targets, of which we have identified some in the present paper, become more relevant.

To further test whether Dlc90F is regulated by Fzr/Cdh1, we determined protein expression levels in a fzr/cdh1 loss-of-function background. As no anti-Dlc90F antibodies were available, we generated a 2.5 kb large mini-gene coding for an N-terminally HA-tagged Dlc90F protein (^HADlc90F) (Fig. S6B). In a wild-type background, we noted weak expression in the eye disc, including photoreceptor neurons and glial cells (Fig. 7A). A similar broad expression pattern was detected when analyzing Dlc90F P[lacZ] enhancer trap insertions or a reporter construct expressing stingerGFP under the control of the Dlc90F 5′ UTR (Fig. S7). When the expression of the ^HADlc90F mini-gene was analyzed in the background of a fzr/cdh1 knockdown (Fig. 7B), a significant upregulation in staining intensity was noted compared with eye discs expressing only the ^HADlc90F mini-gene (Fig. 7A-C). Dlc90F expression was increased 1.6-fold when comparing median staining intensity in neurons and 1.4-fold when comparing median staining intensity in undifferentiated epithelial cells (Fig. 7C). These data further support the finding that Dlc90F is a target of APC/C^Fzr/Cdh1 function in the eye disc, where it participates in the control of neuron-glia interaction.

Here we have identified two temporal phases during eye disc development in Drosophila that require two distinct modes of APC/C-dependent protein degradation to link cell division and neuronal differentiation to glial migration. Whereas APC/C^Fzy/Cdc20 is needed for eye disc growth and blocking of precocious entry of glial cells onto the eye disc, APC/C^Fzr/Cdh1 regulates photoreceptor differentiation and promotes neuron-glia interaction. The latter function is reminiscent to the embryonic PNS, where APC/C^Fzr/Cdh1 acts in postmitotic motoneurons to control neuron-glia interaction (Silies and Klämbt, 2010). In sensory neurons of the developing compound eye, APC/C^Fzr/Cdh1 uses different substrates, which we identified using a genetic suppressor screen.

The early loss of APC/C function efficiently blocks cell division in the developing eye and results in very small eye discs with only a few glial cells. When APC/C function is removed only in the eye disc, the eye disc remains small. In these animals, glial cells ectopically moved on the eye disc in broad streams. Similar glial phenotypes were noted for genes affecting early eye disc patterning (Hummel et al., 2002). This suggests that the eye disc epithelium actively prevents glial migration – possibly by expressing repulsive signals such as Slit or semaphorins as in other imaginal discs (Sasse and Klämbt, 2016). In the case of reduced proliferation, reduced amounts of repulsive guidance cues are produced and glial cells can ectopically move onto the eye disc. Interestingly, loss of APC/C^Fzy/Cdc20 function did not always lead to reduced eye disc sizes, but glial cells still migrated too far, suggesting that APC/C^Fzy/Cdc20 could also have a direct role in the control of onset of glial migration (Fig. 7D).

In normally developing eye discs, the repulsive guidance cues are ignored by the carpet glial cells, which slowly extend their anterior cell margins towards the differentiating neurons. All other retinal glial cells migrate along this thin glial sheath, which shields the glia from any signal coming from the eye disc proper (Silies et al., 2007b). Importantly, migrating glial cells do not make contact with any neuronal cell membranes until they reach the anterior margin of the carpet glial cells. Here, they contact nascent photoreceptor axons, triggering glial differentiation in an FGF-dependent manner (Bauke et al., 2015; Franzdóttir et al., 2009; Sieglitz et al., 2013; Silies et al., 2007b). The glial cells lose their affinity for the carpet glia, stop migrating anteriorly and instead move to the layer above the carpet glia where they adhere to the photoreceptor axons along which they start to differentiate (Fig. 7D). Within the differentiating photoreceptor neurons, Fzr/Cdh1 appears to be required to promote normal neuron-glia interaction. When fzr/cdh1 is removed from the eye disc, glial cells do not develop an affinity for photoreceptor axons or are missing specific signals presented by the axon and therefore might be pushed ectopically towards the anterior (Fig. 7D). This interpretation is specifically supported by the tight correlation of the glial phenotype with the mutant clone. Consequentially, some short-distance signaling could be affected.

Although we could not unambiguously separate glial migration from all neuronal phenotypes, we would like to point out that at least the misplacement of the neuronal nuclei was not rescued in any of the genetic interactions that rescued glial positioning. Furthermore, not every neuronal patterning or rough eye phenotype causes a glial phenotype. For example, other rough eye mutants (ro, rst) exhibit no abnormal glial migration phenotype onto the eye imaginal disc (data not shown). There are several possible scenarios that might explain the fzr/cdh1-induced mutant phenotype. On the one hand, missing adhesion or missing signaling systems could cause the failure of neuron-glia interaction; on the other hand, reduced or aberrant axonal differentiation could lead to disrupted neuron-glia interaction.

To dissect the role of APC/C activity further, we expressed Rca1, which in Drosophila has been shown to bind and block Fzr/Cdh1 but not Fzy/Cdc20 (Grosskortenhaus and Sprenger, 2002; Zielke et al., 2006). We found that in the eye disc, loss of function of APC/C subunits or fzy/cdc20, and Rca1 gain-of-function phenotypes were highly similar, whereas Rca1 overexpression did not recapitulate the fzr/cdh1 loss-of-function phenotype. Interestingly, in vertebrates, the Rca1 homolog Emi1 has been proposed to interact with both APC/C^Fzr/Cdh1 and APC/C^Fzy/Cdc20 (Miller et al., 2006; Pesin and Orr-Weaver, 2008; Peters, 2006; Reimann et al., 2001). This indicates that, depending on context, Rca1 can also be a regulator of APC/C^Fzy/Cdc20 function in Drosophila.

APC/C^Fzr/Cdh1 function is mediated by the degradation of specific target proteins. As Fas2 protein is not involved in neuron-glia interaction in the eye disc, we screened for novel candidates. Four genes affecting neuron-glia interaction downstream of fzr/cdh1 in the developing larval visual system were identified. Of these, we dissected the function of two more thoroughly by mutant analysis. CG15765 is thought to be specifically expressed in the nervous system and encodes a secreted protein carrying a C-type lectin. C-type lectins function as Ca²⁺-dependent carbohydrate-binding modules and are characterized by a CLECT domain. The CLECT domains of CG15765 and Brevican are related and the extracellular matrix (ECM) protein Brevican is known to promote glial cell migration (Lu et al., 2012). One possible scenario is that CG15765 is secreted by neurons to promote the attachment of glial cells to the photoreceptor axons.

Yeast two-hybrid data suggest a direct interaction of Dlc90F with Fzr/Cdh1 (https://thebiogrid.org/interaction/31998; Giot et al., 2003), indicating that Dlc90F may be a direct target. Although overexpression of Dlc90F does not lead to a mutant phenotype (data not shown), this could suggest a tight regulation of Dlc90F protein levels. Interestingly, the expression of the Dlc90F reporter construct is upregulated in the fzr/cdh1 knockdown background. The severity of the neuron-glia interaction phenotype caused by suppression of Fzr/Cdh1 is significantly reduced upon suppression of Dlc90F expression. Dynein light chains can associate with dynein heavy chains to orchestrate movement along microtubules. They are required during retrograde axonal transport as well as in the cytoplasm, where dyneins participate in the positioning of the Golgi complex. In addition, motor-independent functions were reported for the mammalian Dlc90F homolog Tctex-1 (Sachdev et al., 2007). Tctex-1 can sequester inactive RhoGEF, indicating a function of Tctex-1 in regulating actin dynamics (Meiri et al., 2012, 2014). Upon fzr/cdh1 knockdown, nascent photoreceptor axons may be delayed in their development due to upregulated Dlc90F function. Glial cells, in turn, are then missing their normal target structures and continue to migrate. When Dlc90F function is reduced, axonal growth and glial migration returns to normal, although neuronal patterning is not completely rescued. Interestingly, in the complete absence of Dlc90F, other Fzr/Cdh1 targets appear to be more relevant. Our genetic suppressor screen may have uncovered further potential APC/C substrates that could orchestrate mitotic and postmitotic APC/C function in different contexts.

In conclusion, we demonstrate a dual function of the APC/C during eye development. Initially APC/C^Fzy/Cdc20 is required for cell proliferation in the eye imaginal disc and controls glial entry onto the eye anlage. Subsequently, APC/C^Fzr/Cdh1 is needed for postmitotic differentiation and causes a neuronal patterning, as well as a glial positioning phenotype. The switch in co-activator dependence and substrate specificity nicely illustrates how the same molecular machinery can be used to control different developmental aspects, and link cell proliferation and tissue growth to differentiation and neuron-glia interaction.

Flies were cultured using standard conditions. Rescue and gain-of-function studies were carried out using the Gal4/UAS system (Brand and Perrimon, 1993). The following mutant alleles were used: fas2^EB112 (Grenningloh et al., 1991), fzr^8F3 and fzr^ie28 (Sigrist and Lehner, 1997; Silies and Klämbt, 2010), mr⁴ (Reed and Orr-Weaver, 1997), lmg⁰³⁴²⁴ and lmg^EY11317 (Bellen et al., 2011), and Dlc90F^e155 (Li et al., 2004). ey-FLP clones were generated using the following chromosomes: fzr^8F3 FRT19A, GMR-hid FRT19A; ubi::nlsGFP FRT19A; ubi::nlsGFP 40A FRT and lmg^EY11317 FRT40A. The GMR-hid transgene, which results in a loss of heterozygous and homozygous tissue, was employed to generate large mutant cell clones (Newsome et al., 2000; Stowers and Schwarz, 1999). ey-Gal4 flies were obtained from the Bloomington Stock Center. The ey-Gal4 element uses a 3.5 kb promoter fragment that is not expressed in glia (Fig. S1A,B).

UAS-dsRNA fly stocks were obtained through VDRC (Vienna, Austria), the Drosophila Genome Resource Center (DGRC, Kyoto, Japan) and the Bloomington Stock Center (Indiana University, Bloomington, IN, USA). UAS-dsRNA fly stocks used for knockdown were: fzy - w[1118]; P{GD11268}v40500 (no predicted off-targets); mr (van Roessel et al., 2004); and fzr - w[1118]; P{GD9960}v25550 (one predicted off-target). An additional UAS-fzr^dsRNA line was generated during this work. The UAS-fzr^dsRNA vector was generated with the pWiz plasmid. PCR was performed on the cDNA clone RE20929 and the fragments were subcloned into pWiz with XbaI+NheI for the first orientation and XbaI+BlnI for the second orientation. Primers used were: fzrRNAi for, CCCTTCTAGATAAATGTGGTGTACAT; and fzrRNAi rev, CTGGTTCTAGAAGAGACTCCGCTTTG. The piggyBac insertion lines were obtained from Exelixis (Harvard Medical School, MA, USA). For overexpression of Dlc90F we generated flies carrying a UAS-Myc::Dlc90F construct and used the GSV P-element insertion lines GS3099 and GS7372.

The fzr/cdh1 substrate candidate genes were identified using a FlyBase query containing ‘embryonic nervous system’. The corresponding sequences were analyzed for the presence of a D-Box using the Destruction Box Motif Finder (Liu et al., 2012) using an E-value of 0.2 as a cut-off, or for the presence of a KEN motif. Candidates were subsequently tested in the RNAi-based suppressor screen. To perform the lethality shift assay, we established the stock UAS-fzr^dsRNA/CyO^tub-Gal80ts; da-Gal4/ da-Gal4 in which the tub-Gal80^ts prevents constitutive expression of the fzr^dsRNA in the parental stock and crossed it to a selection of UAS-dsRNA strains. In addition, we screened genes encoding Drosophila homologs of recently described APC/C candidate target proteins (Merbl and Kirschner, 2009). Possible neural expression was checked with the help of the BDGP in situ database (http://www.fruitfly.org/cgi-bin/ex/insitu.pl). All stocks analyzed are listed in Table S1.

Fixation and preparation of tissues for immunohistochemistry was performed as described previously (Yuva-Aydemir et al., 2011). Antibodies used were mouse anti-Repo (1:5), mouse anti-Chaoptin (1:5), rat anti-Elav (1:10) (all DSHB), rabbit anti-β-Galactosidase (1:200; MP Biomedicals); rabbit anti-GFP (1:1000; Invitrogen); mouse anti-HA (1:1000; Covance); and goat anti-HRP-Alexa647 (1:500; Jackson ImmunoResearch Laboratories). Secondary antibodies used in this study were: goat anti-mouse IgG, goat anti-mouse IgG highly cross-absorbed, goat anti-rabbit IgG and goat anti-rat IgG with the fluorophore conjugates Alexa Fluor-488, -568 and -647 (1:1000; Molecular Probes). Specimen were analyzed using a Zeiss LSM710 and LSM880; orthogonal sections were displayed using the Zeiss LSM imaging software.

Scanning electron microscopy (SEM) experiments were carried out using a HITACHI Scanning Electron Microscope S 3000N. Whole flies were immersed in fixative for 2 h, dehydrated in an ethanol series at room temperature for 12 h each and dried overnight in a desiccator with silica gel. Flies were mounted on stick carbon conductive tabs placed on aluminum specimen mounts. Sputter coating with gold was carried out using an EMITECH K550X under vacuum for 4 min at 25 mA.

CG15765^Δ1 mutants were generated using Flp-mediated recombination between two FRT-containing transgenic insertions following standard procedures (Parks et al., 2004). Recombination between the FRT-elements carrying P-elements {XP}CG15766^d06145 and the PBac {WH}CG15765^f02896 deleted approximately 3.5 kb of the gene locus, including the 5′ UTR and the first coding exon of CG15765. The deletion was verified by loss of white⁺ expression and PCR (Parks et al., 2004). The following primers were used: XP5′ plus, AATGATTCGCAGTGGAAGGCT; WH5′ minus, GACGCATGATTATCTTTTACGTGAC; CG15765 for, CTGCTGAGGGAATTGGATGT; CG15765 rev, GCAGCAGATCGATGTCGTTA.

The putative regulatory sequence of Dlc90F (5′ UTR) was cloned upstream of a stgGFP into a pBPGUw vector (Addgene), resulting in a Dlc90F^{5′ UTR}-stgGFP construct. A 2.5 kb mini-gene was generated using standard procedures. Primers used for cloning of the mini-gene were: HindIII Dlc90F 2.2 kb for, CACCaagcttCGTGCAGAGTTACACAG (CACC added for TOPO cloning, HindIII in lowercase, genomic Dlc90F sequence in capitals); Dlc90F N HA tag rev, GGTCTCAtgatttgcttctaacggcag (BsaI site used for Golden Gate cloning in capitals, lowercase Dlc90F sequence); Dlc90F N HA tag for, accGGTCTCAatcaaagctgcgacgATGggaggaTACCCATACGATGTTCCAGATTACGCTggaggaatggatgact [added sequence in capitals (BsaI site, start codon, and spacer sequence to HA-encoding sequence), lowercase Dlc90F sequence]; and Dlc90F 3RNA rev, gCTAGcggtattatacacttttta (capitals – generates an NheI site). Owing to the close proximity of the insertion site of the mini-gene and the Dlc90F locus, we did not test if the mini-gene rescues Dlc90F mutant phenotypes.

We thank the Bloomington Stock Center, DGRC and VDRC for fly stocks. We are thankful to E. Naffin, P. Deing and K. Krukkert for excellent technical help, and to E. McMullen, and K. Adolphs for critical reading.