Zfh1 promotes survival of a peripheral glia subtype by antagonizing a Jun N-terminal kinase-dependent apoptotic pathway

Abstract

In Drosophila subperineurial glia (SPG) ensheath and insulate the nerve. SPG is under strict cell cycle and survival control because cell division or death of such a cell type would compromise the integrity of the blood–nerve barrier. The mechanisms underlying the survival of SPG remain unknown. Here, we show that the embryonic peripheral glia expresses the Zfh1 transcription factor, and in zfh1 mutants a particular SPG subtype, ePG10, undergoes apoptosis. Our findings show that in ePG10, Zfh1 represses the pro-apoptotic RHG-motif gene reaper in a cell-autonomous manner. Zfh1 also blocks the activation of the Jun N-terminal kinase (JNK) pathway, and reducing or enhancing JNK signalling in zfh1 mutants prevents or promotes ePG10 apoptosis. Our study shows a novel function for Zfh1 as an anti-apoptotic molecule and uncovers a cryptic JNK-dependent apoptotic programme in ePG10, which is normally blocked by Zfh1. We propose that, in cells such as SPG that do not undergo self-renewal and survive long periods, transcriptional control of RHG-motif gene expression together with fine tuning of JNK signalling is crucial for cell survival.

Introduction

In flies, as in vertebrates, the choice between cell survival and death is regulated by the balanced activities of the anti-apoptotic molecule Diap1 and the pro-apoptotic molecules of the RHG-motif family (Reaper, Head Involution Defective, Grim, Sickle and Jafrac2). Diap1 blocks activation of the intrinsic cell death machinery by binding to and neutralizing the activity of pro-apoptotic caspases (Meier et al, 2000; Muro et al, 2002; Wilson et al, 2002). When cells are challenged with death stimuli, the pro-apoptotic RHG proteins (White et al, 1994; Grether et al, 1995; Chen et al, 1996), bind to Diap1, thereby liberating and activating caspases (Wang et al, 1999; Goyal et al, 2000). During embryogenesis, RHG-motif genes are expressed in highly restricted patterns that seem to correlate with programmed cell death (PCD) (White et al, 1994; Grether et al, 1995; Chen et al, 1996). These observations suggest that fine transcriptional control of their expression is potentially one of the major mechanisms regulating apoptosis. This issue has just begun to be addressed in studies of PCD in which cells are committed to die or to survive during tissue remodelling and establishment of final cell numbers. Thus, the Drosophila Hox protein Deformed controls head morphogenesis by activating localized apoptosis through direct activation of rpr (Lohmann et al, 2002). In the central nervous system (CNS), the segment-specific PCD of dMP2 neurons is regulated by the differential expression of the Hox gene Abd-B, which in posterior segments prevents neuronal death by repressing reaper and grim (Miguel-Aliaga and Thor, 2004). This example illustrates how, in a cell type under death pressure, survival is actively triggered through transcriptional repression of RHG-motif genes. One could thus hypothesize that, more generally, in developing tissues normally devoid of PCD a cryptic pro-apoptotic programme is present but is actively held in check notably through RHG-motif gene repression. To date, the transcriptional cascades involved in such processes remain largely unknown. Moreover, whether transcriptional regulation of the three classical RHG-motif genes (reaper, hid and grim) has a central role in this context is a question that remains unanswered. Strikingly, in the embryonic peripheral nervous system (PNS) no PCD seems to occur in peripheral glia (PG), and their final number is already established by embryonic stage 12 at the time when these cells become postmitotic. Moreover, in the larvae neither cell death nor further production of PG occurs (Sepp et al, 2000). PG appear thus as a cell type whose numbers are tightly maintained from early embryonic to late larval stages and constitute a tractable system for understanding how a particular cell type avoids apoptosis.

The Drosophila zfh1 gene constitutes the prototype of an atypical family encoding transcription factors containing two separate krüppel-like zinc-finger clusters, and a homeodomain (Fortini et al, 1991). zfh1 displays pleiotropic functions, ranging from progenitor cell specification to postmitotic precursor differentiation. Indeed, it is required for the proper differentiation of mesodermal-derived tissues such as muscle, gonads or heart (Lai et al, 1993; Broihier et al, 1998; Su et al, 1999). In the CNS, zfh1 specifies the identity of motoneurons and interneurons issued from the NB7-3 lineage (Lee and Lundell, 2007). Moreover, it controls the exit of motor axons from the ventral nerve cord and has important functions in axon pathfinding of a subset of motoneurons as they grow towards their peripheral muscle targets (Garces and Thor, 2006; Layden et al, 2006). In addition, a recurrent feature observed is that dysregulation of zfh1 function often affects cell numbers in many tissues (Lai et al, 1993; Broihier et al, 1998; Su et al, 1999), although its involvement in the control of cell survival remains an open question.

In this study, we describe a cell-autonomous requirement for zfh1 in the survival of a subset of glial cells in the Drosophila embryonic PNS. We show that Zfh1 is expressed by all PG and that it is required for the survival of a gliotactin+ subperineurial glia (SPG) subtype named ePG10 (embryonic peripheral glia 10). Glia-specific expression of the pan-caspase inhibitor p35 in a zfh1 mutant context, efficiently rescues ePG10 survival, migration, morphology and molecular identity. We found that zfh1 prevents ePG10 apoptosis by repressing the expression of RHG-motif cell death activator reaper. Moreover, we show that in ePG10 Zfh1 blocks activation of the Jun N-terminal kinase (JNK) pathway, well known for regulating several essential developmental processes such as dorsal closure in the embryo, thorax closure, dorsal appendage formation as well as wound healing (Martin-Blanco et al, 1998; Zeitlinger and Bohmann, 1999; Suzanne et al, 2001; Ramet et al, 2002; Galko and Krasnow, 2004). Reduction in JNK signalling using a dominant-negative form of Basket in zfh1 mutants is sufficient to prevent ePG10 apoptosis. Altogether these results show that a cryptic JNK apoptotic programme is indeed held in check by Zfh1 in this glial population.

Results

Zfh1 is expressed by all PG

We first characterized the expression pattern of Zfh1 within the PNS and more specifically in glial cells. PG are named according to the nomenclature of von Hilchen et al (2008). During stages 14–16 these cells display a highly stereotyped organization (Figure 1A and A′; see also von Hilchen et al, 2008). PG can be easily identified at the molecular level by the expression of the transcription factor reversed polarity (Repo), which labels all postmitotic PG from stages 12–16 (Campbell et al, 1994; Xiong et al, 1994; Halter et al, 1995) (Figure 1B). In the dorsal PNS, ePG11 (also known as support cell of the dorsal bipolar dendrite cell) and ePG10 can be reliably distinguished by their locations and characteristic morphologies. At stage 15, Repo-Gal4∷UAS-Tau-EGFP clearly visualizes the dynamics of ePG10 migration and of its cytoplasmic processes that extend from stage 15 onward both dorsally and ventrally, along the anterior fascicle (Figure 1B).

Figure 1.

Zfh1 is expressed in all peripheral glia. (A, A′) Diagrams showing the location and morphology of PG nuclei from stages 14–16 and the molecular markers expressed by these cells; adapted from von Hilchen et al (2008). (B) In wild-type stage 15 flat-mounted embryos, Repo-Gal4, UAS-EGFP^F show the morphology of PG. Closeup view 1 shows Repo expression in the ePG12 neuron. In the dorsal region, ePG10 migrates from a dorsal to a ventral position (compare closeup views 2 and 3). (C, D) Double labelling for Zfh1 and Repo shows that all glial cells of the PNS express Zfh1. At stage 14, the nuclei of ePG10 and ePG11 display a characteristic rounded morphology, whereas at stage 16 these cells have an elongated nucleus (compare closeup views in C and D). (E) Double staining for Zfh1 and Elav shows that, with the exception of the ePG12 neuron (asterisk), PNS neurons do not express Zfh1. Closeup views show expression of Zfh1 in the ePG12 neuron. (F) Double staining for Zfh1 and Prospero shows that Zfh1 is not expressed in PNS accessory cells. In all panels two consecutive hemisegments are shown. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: 5 ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right. The dashed line indicates the lateral edge the CNS.

Double –immunostaining at embryonic stages 14 and 16 for Zfh1 and Repo showed that in the abdominal hemisegments, Zfh1 is detected in all PG (Figure 1C and D). We observed the expression of Zfh1 in a Repo+ cell at the position where ePG12 (also known as lateral bipolar dendrite) is located (asterisk in Figure 1C and D). Halter et al (1995) have earlier noted that this cell is positive for the neuron-specific antigen HRP and Repo. An anti-Elav antibody that labels all neuronal cells (O'Neill et al, 1994) confirmed Zfh1 expression in ePG12 and its exclusion from all other neurons (Figure 1E). Finally, we examined whether Zfh1 was expressed in sensory organ accessory cells using the specific marker Prospero (Campbell et al, 1994). No overlap between Zfh1 and Prospero could be detected in this cell type (Figure 1F). Altogether, these results show that in the embryonic PNS, Zfh1 is expressed in all PG and in ePG12, but is excluded from other neurons and accessory cells.

The ePG10 subperineurial glial cell is missing in zfh1 mutants

The glial-restricted expression pattern of Zfh1 suggested that it might have a function in glial development. We thus analysed glial cell numbers, position and morphology using combinations of two independent zfh1 loss-of-function alleles, zfh1⁵ and zfh1^75.26 (Lai et al, 1993; Broihier et al, 1998) and a hypomorphic allele, zfh1⁸⁶⁵, which is a P-element enhancer trap line with an insertion into the 5′ region of the zfh1 gene (Justice et al, 1995) and denoted zfh1^LacZ herein.

In the ventral region, the number and positions of PG in homozygous zfh1⁵, zfh1^75.26 or trans-heterozygote zfh1⁵/zfh1^75.26 mutants were similar to wild-type embryos, even if we occasionally noticed that some cells were missing and/or mislocated (Figure 2A and B). In contrast, in the dorsal region we consistently found, for all three alleles, that one glial cell was missing in the area where ePG10 and ePG11 are normally located (Figure 2A–C). To unambiguously determine the identity of this cell, we used the J29^LacZ enhancer trap for gliotactin (Auld et al, 1995) that distinguishes the subperineurial ePG10 cell (J29^LacZ+) from ePG11 (J29^LacZ−). We found that ePG11 was present in all hemisegments at all time points analysed; however, ePG10 was missing in 75% of hemisegments in zfh1⁵ stage 16 mutant embryos (arrowhead in Figure 2D, E and G). We also examined the status of ePG10 in zfh1^75.26 homozygote and in zfh1⁵/zfh1^75.26 trans-heterozygote mutant embryos at stage 16 and observed it to be absent in 85% (n=49) and 75% (n=105) of hemisegments, respectively (Figure 2G).

Figure 2.

The dorsal ePG10 glial cell is missing in zfh1 mutant. (A) In wild type, Zfh1 and Repo are co-expressed in all the glial cells of the PNS. (B) No Zfh1 protein was detected in homozygote zfh1⁵, indicating that this allele is a protein null. In the dorsal region of the PNS ePG10 is absent in 72% of the hemisegments whereas ePG11 is always present. Double arrowhead indicates shrunken and densely stained nucleus at the normal position of ePG10. (C) In zfh1^LacZ−/−, ePG10 is absent in two consecutive hemisegments. (D) In wild-type embryos, J29^LacZ shows ePG10 in all hemisegments. (E) In zfh1⁵ mutant embryos, ePG10 is rarely seen in the dorsal region of the PNS. (F) In first-stage larvae (stL1) ePG10 is frequently absent whereas other PG appear normal. Two consecutive hemisegments are shown. (G) Quantification of ePG10 cell survival at stages 15–16 in different mutant combinations compared with wild type. Of note, the rescue experiment using UAS-zfh1 was carried out at 29°C to increase Zfh1 dosage. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right. The dashed line indicates the lateral edge the CNS.

We next aimed to assess whether other PG were affected in zfh1 mutants at later stages of development. We used the hypomorphic zfh1^LacZ allele that allows analysis up to larval stages because zfh1⁵ and zfh1^75.26 mutants die soon after stage 16. In this mutant at stages 15–16 we noticed ePG10 being absent in 51% of hemisegments (n=376) (Figure 2C and G) whereas other PG appear unaffected. In the first instar larval stage (stL1) mutants we found no further apparent phenotype among PG other than the lack of ePG10 (Figure 2F). Altogether, these results show that the loss of the gliotactin+ subperineurial ePG10 in zfh1 mutants occurs primarily between stages 13 and 16 and shows that the majority of PG differentiate grossly normally.

Zfh1 prevents ePG10 apoptosis in cell-autonomous manner

The absence of ePG10 at stage 16 in zfh1 mutants prompted us to investigate whether this cell is never generated or progressively dies during development. In the late stage 12 of zfh1⁵ mutant embryos we found that ePG10 was present in 75% of hemisegments (Figure 3A, B and H) and in only 47% at stages 13–14 (data not shown; Figure 3H for quantification). This phenotype appeared even more severe by late stage 15, ePG10 being present in just 25% of the hemisegments as mentioned above (Figure 3H).

Figure 3.

zfh1 is postmitotically and cell autonomously required for ePG10 survival. (A) In late stage 12 Zfh1 is expressed in ePG10 but not in ePG11. (B) In late stage 12 zfh1⁵ mutants ePG10 is present or appears pyknotic (double arrowhead). (C) In wild type, no detectable Casp3^* staining is seen in Repo+ cells. (D) In zfh1⁵ mutant embryos, Casp3^* staining shows that, in the dorsal region of the most anterior hemisegment, ePG10 undergoes apoptosis. (E) In zfh1^LacZ+/− embryos, ePG10 (Repo+, βGal+) is found in close association with dorsally projecting inter segmental nerve (ISN) motor axons (Fas2+). (F) In zfh1^LacZ−/− embryos, ePG10 is absent in two hemisegments where ISN motor axons project normally. (G) Using UAS-zfh1; Repo^Gal4 in overall zfh1⁵ homozygote mutant embryos, ePG10 is present in 75% of the hemisegments and migrates towards its final ventral position. (H) Quantification of ePG10 cell survival in different mutant combinations compared with wild type. Note that the absence of ePG10 is more severe in stages 15–16 than in stages 13–14 and than late stage 12 indicating that in some hemisegments ePG10 cell death occurs mainly between late stage 13 and stage16. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right. The dashed line indicates the lateral edge the CNS.

The progressive decrease in ePG10 numbers between late stages 12 and 16 raised the possibility that in zfh1 mutants, ePG10 undergoes apoptosis. In line with this hypothesis, in zfh1 mutants we frequently found pyknotic cells with shrunken and densely stained nuclei at the site where ePG10 is normally found (double arrowhead in Figure 3B; see also Figure 2B). In late stage 12, pyknotic ePG10 nuclei were found in 15% of hemisegments (n=96) whereas they were never found in wild-type counterparts. To confirm that ePG10 undergoes apoptosis in zfh1 mutants we used an antibody raised against the activated form of the human caspase-3 protein (denoted Casp3^*). This antibody was shown earlier to recognize apoptotic cells in Drosophila tissues (Yu et al, 2002; Brennecke et al, 2003). Casp3^* staining showed no cell death in PG in stages 15–16 wild-type embryos (Figure 3C). In contrast, in zfh1⁵ mutants, Casp3^* was detected in all dorsally located Repo+ nuclear remnants, which presumably correspond to the dying ePG10 (Figure 3D). These results show that in the absence of zfh1, ePG10 is generated and degenerates by apoptosis.

We next aimed to determine whether zfh1 acts postmitotically and in a cell-autonomous manner to ensure ePG10 cell survival. The fact that motor axons bundles are affected in zfh1⁵ mutant embryos (Layden et al, 2006) could argue for an indirect effect of motor axons on ePG10 survival as this cell is found in close contact with the dorsally projecting inter segmental nerve (ISN). In contrast to zfh1⁵ allele we found that in zfh1^LacZ hypomorphic mutants ISN motoneurons develop and project normally towards their dorsal muscle fields. Nevertheless, ePG10 still dies in this context (Figure 3E and F; see Figure 2G for quantification) supporting the idea that motoneurons axons bundles do not have a key function in its survival. We then used, on a zfh1⁵ mutant background, the Repo-Gal4 driver to specifically restore zfh1 activity in postmitotic glial cells (and not in motoneurons) (Sepp and Auld, 2003). In such embryos, the death of ePG10 was efficiently (albeit partially) rescued, ePG10 being present in 75% of the hemisegments at stages 15–16 (n=130) (Figure 3G and H). Furthermore, in this context the position and morphology of ePG10 nuclei were similar to wild type (Figure 3G). Collectively, these results show that zfh1 acts postmitotically and cell autonomously to prevent ePG10 cell death.

Loss of ePG10 in zfh1 mutants is not due to aberrant differentiation

It could be argued that ePG10 cell death in zfh1 mutants resulted from impaired differentiation (i.e. aberrant molecular identity and/or migratory behaviour and/or cell morphology). To test this possibility we used the Repo-Gal4 line to express the baculoviral anti-apoptotic caspase inhibitor p35 (Hay et al, 1994) in glial cells, in homozygote zfh1⁵ mutant embryos. This resulted in the survival of ePG10 in 88% of hemisegments (n=110) (Figure 4A–F). Moreover, in this allelic combination we found that (i) rescued ePG10 expresses Repo, wingless, mirror (showed by the enhancer trap lines wg^LacZ and mirr^LacZ; see Figure 4A–D) and Gliotactin (data not shown); (ii) the migration and positioning of ePG10 is ventral to ePG11 as in wild-type embryos (Figure 4E and F) and (iii) its specific shape, i.e. elongated nucleus oriented perpendicular to ePG11, is similar to the wild type (Figure 4E and F). In addition, use of Mz97^Gal4, which is normally expressed in ePG10, further outlines the correct molecular identity and morphology of this cell (extended cytoplasmic processes lying perpendicular to ePG11) in zfh1 mutants (Figure 4G and H). Taken together these data show that zfh1 prevents apoptosis of the ePG10 glial cell but is not required for other aspects of its differentiation.

Figure 4.

zfh1 does not have an instructive function for ePG10 differentiation other than survival. (A–H) Using UAS-p35; Repo^Gal4, zfh1⁵ ePG10 survives in 88% of hemisegments. Rescued ePG10 expresses Repo (B, D, F, H), mirror (from mirr^LacZ) (B), wingless (from wg^LacZ) (D) and Mz97^Gal4 (H), similarly to wild-type counterparts. The morphology of rescued ePG10 (elongated cytoplasmic processes oriented perpendicular to ePG11) is similar to wild type (compare H and G). In stage 16 embryos, rescued ePG10 migrates ventrally as in wild type (compare E and F). In each case, only two consecutive hemisegments are shown. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right. The dashed line indicates the lateral edge the CNS.

Genetic interactions between zfh1, Diap1 and RHG-motif genes in ePG10

As Diap1 and RHG-motif genes are key regulators of cell survival/death, we examined their putative genetic interaction with zfh1 for the control of ePG10 survival. Using the th^j5C8 enhancer trap line (a P-element LacZ insertion upstream of Diap1; see (Hay et al, 1995) we showed that Diap1 is expressed in ePG10 (Figure 5A and A′). This line has been reported to be a hypomorphic allele of Diap1 and, in th^j5C8/th^j5C8 mutant embryos, we found that ePG10 undergoes apoptosis in 32% of the hemisegments (n=139) (Figure 5B, B′ and E). Moreover, over-expression of Diap1 in PG using the allelic combination UAS-Diap1; Repo-Gal4 in zfh1⁵ homozygous mutant embryos, led to survival of the ePG10 cell in 93% of hemisegments (n=176) (Figure 5C and E). In parallel we analysed the effect of the homozygous deletion Df(3L)H99 lacking rpr, hid and grim (denoted H99) on ePG10 cell death in a zfh1⁵ homozygous mutant background. In zfh1⁵,H99 double mutants embryos we found that apoptosis of ePG10 can be prevented since ePG10 (identified by the J29^LacZ) is present in 62% of hemisegments analysed (n=104) (Figure 5D and E). We also quantified ePG10 cell death in zfh1⁵ homozygous mutant embryos that are heterozygous for the H99 deletion and found ePG10 in 65% of hemisegments (n=204), showing that only one copy of the wild-type rpr, hid and grim genes is not sufficient to induce ePG10 cell death (Figure 5E). These results show that (i) the death of ePG10 observed in zfh1 mutants is apoptotic in nature, confirming our earlier experiments with p35; (ii) ePG10 survival is sensitive to the level of expression/activity of Diap1 and RHG-motif genes; and (iii) Diap1 activity and H99 deletion are epistatic to Zfh1 function.

Figure 5.

Diap1 and H99 are epistatic to zfh1 function. (A, A′) In diap1^LacZ+/− heterozygous embryos, ePG10 and ePG11 are present and express diap1^LacZ. (B, B′) In diap1^LacZ−/− homozygous embryos ePG10 is absent in 32% of the hemisegments (n=139) whereas ePG11 survival is not affected. The (A′, B′) panels are, respectively, identical to (A) and (B) but only the Repo channel is shown. (C) In overall zfh1⁵ mutant embryos, UAS-diap1;Repo^Gal4 rescues ePG10 survival. (D) In homozygote H99,zfh1⁵ mutant embryos, ePG10 (J29^LacZ+) is present in 62% of the hemisegments showing that ePG10 cell death in zfh1 mutants is RHG-motif genes dependant. (E) Quantification of ePG10 cell survival in different mutant combinations in stage 16 compared with wild type. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right. The dashed line indicates the lateral edge the CNS.

Analysis of Diap1 and RHG-motif genes expression in zfh1 mutants

As Zfh1 is a transcriptional regulator, these data raised the possibility that Zfh1 has an anti-apoptotic function in ePG10 by influencing Diap1 and/or RHG-motif gene expression. To test this idea, we first monitored Diap1 expression using the th^j5C8 insertion in zfh1⁵ mutants at late stage 12, before ePG10 cell death, and found normal Diap1 expression (Supplementary Figure S2F, G). In stage 16 homozygote zfh1⁵ mutant embryos in which ePG10 survival was rescued using UAS-p35, Repo-Gal4 we also found normal Diap1 expression (Figure 6A and B). Thus, despite the role of Diap1 in ePG10 survival this observation indicates that in zfh1 mutants, ePG10 cell death is not a consequence of downregulation of Diap1. We also confirmed this at the protein level by using an antibody against Diap1 (data not shown). We then studied the status of RHG-motif gene expression in the same genetic background. By double immunostaining we found that neither Hid nor Grim expression is upregulated in ePG10 (Supplementary Figure S1 A–D′). Since to date no Rpr antibody suitable for immunostaining procedures on fixed tissues has been reported, we thus monitored rpr expression in vivo using a Gal4 P-element insertion into rpr, rpr^NP0520 (Miguel-Aliaga and Thor, 2004) and denoted rpr^Gal4(520) herein. In wild-type embryos, rpr^Gal4(520) expression was not detected in ePG10 (n>82 hemisegments). In contrast, in zfh1 mutant embryos at stages 13–14 its expression was observed in this cell in 10% of hemisegments (n=130). Similar results were obtained using the rpr^Gal4(368) enhancer trap (Figure 6C and D; data not shown). This result most probably reflects that ePG10 apoptosis does not occur simultaneously in each hemisegment (see Figure 3H) and highlights that each analysed embryo should only be regarded as a unique picture taken at a given time. Altogether, these results show that while the expression of Diap1, Hid and Grim is not altered in ePG10 in zfh1 mutants, rpr is deregulated.

Figure 6.

Reaper is necessary for ePG10 cell death in zfh1 mutant. (A) In diap1^LacZ+/− embryos ePG10 and ePG11 express βGal. (B) In UAS-p35; Repo^Gal4, diap1^LacZ+/−, zfh1^5−/− embryos, we found normal βGal expression in ePG10 indicating that in this cell zfh1 does not regulate diap1 expression. (C) In UAS-nlsEGFP; rpr^{Gal4(520)+/−} embryos, GFP expression is not detected in the developing Repo+ cells. (D) In UAS-p35, UAS-nlsEGFP; rpr^{Gal4(520)+/−}, zfh1^5−/− embryos, GFP expression is seen in ePG10. (E₁) Schematic view of the rpr gene locus. The open reading frame for Rpr is indicated in red. Locations of the two P-Gal4 element insertions rpr^Gal4(368) and rpr^Gal4(520) used in this study as mapped by the NP Consortium. Small arrows denote the position of the oligonucleotides used for the genomic DNA PCR. (E₂) PCR on genomic DNA showing that the rpr^Gal4(368) P-element lays 335 base pairs upstream of the rpr start codon and 43 base pairs upstream of the rpr^Gal4(520) P-element. (E₃) western-blot analysis showing that the level of Rpr protein is decreased by 2.5-fold in homozygote rpr^Gal4(368)-pooled embryos compared with wild type. (F) Quantification of ePG10 cell survival in different mutant combinations in stage 16 compared with wild type. Note that ePG10 cell death is significantly rescued using heterozygous or homozygous combinations of the rpr^Gal4(368) or the rpr^Gal4(520) hypomorphic allele of rpr in overall zfh1^5−/− mutant embryos. As a control, zfh1^5−/− homozygotes mutants (denoted ‘reverted') obtained after removal of the rpr^Gal4(368) allele from the combined rpr^Gal4(368), zfh1⁵ chromosome leads to ePG10 survival in only 35% of hemisegments. These results strongly support the view that ePG10 cell death in zfh1 mutants is primarily rpr dependant. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right.

rpr as a major cell death executor in ePG10

The above observations indicate that zfh1 represses rpr and suggests that among the three classical RHG-motif genes rpr is the major cell death determinant of ePG10 in zfh1 mutant. To confirm this hypothesis we took advantage of the availability of a hypomophic allele of rpr, rpr^Gal4(520) (Miguel-Aliaga and Thor, 2004), and another uncharacterized rpr allele, rpr^Gal4(368). First, we characterized the Gal4 enhancer trap allele rpr^Gal4(368) by showing that this P-element lies 335 base pairs upstream of the rpr start codon and 43 base pairs upstream of the rpr^Gal4(520) P-element (Figure 6E₁ and E₂). Furthermore, by western-blot analysis we have found that the level of Rpr protein is decreased by 2.5-fold in homozygote rpr^Gal4(368)-pooled embryos compared with wild type (Figure 6E₃). We then analysed rpr^{Gal4(368)−/−},zfh1^5−/− and rpr^{Gal4(368)+/−},zfh1^5−/− embryos and found ePG10 survival in 75% (n=210) and 64% (n=150) of hemisegments, respectively (Figure 6F; Supplementary Figure S2H). Similar results were obtained using the rpr^Gal4(520) allele (Figure 6F). These figures are strikingly comparable with those found using the XR38 deficiency, which removes rpr and sickle (Peterson et al, 2002) (Figure 6F) and the H99 deficiency (see Figure 5E). Altogether these results highlight a key role for rpr in ePG10 cell death in zfh1 mutants.

ePG10 apoptosis in zfh1 mutants requires activation of the JNK pathway

We then aimed to determine whether the death of ePG10 in zfh1 mutants is the consequence of the activation of known pro-apoptotic signalling pathways. Among them, we focused our attention on the JNK pathway that represents one good candidate. Indeed, molecular determinants of this pathway are known to have a major function in triggering apoptosis during development in Drosophila (Adachi-Yamada et al, 1999a; McEwen and Peifer, 2005) and in vertebrates; it has been shown to influence the survival of PG (Parkinson et al, 2001). In Drosophila, JNK pathway activity can be monitored by transcriptional activation of puckered (puc), a dual-specificity phosphatase that acts in a negative-feedback loop to regulate the JNK pathway (Martin-Blanco et al, 1998). puc enhancer traps have been widely used to monitor at the cellular level resolution JNK activity in vivo (e.g. Adachi-Yamada et al, 1999a; Tateno et al, 2000; Igaki et al, 2002). Indeed, in response to apoptotic stimuli the pucE69 enhancer trap is induced in a JNK-dependent manner (Adachi-Yamada et al, 1999a). We thus first analysed the activation state of this pathway in ePG10 in wild type and zfh1 mutant backgrounds. In wild-type late stage 12 and stage 13 embryos we found that βGal expression from pucE69 is low or under detectable levels in the developing ePG10 (Figure 7A and C). In contrast, in equivalent stage zfh1 mutants, we found pucE69 expression to be significantly increased in ePG10 (Figure 7B and D) showing that JNK pathway is potentiated in zfh1 mutants.

Figure 7.

In zfh1 mutants JNK signalling is active in ePG10 and controls its apoptosis. (A, C) Double staining for Repo and βGal shows that in wild-type embryos Puc^LacZ expression is under detectable level (late stage 12) or low (stage 13) in ePG10. Note that higher levels of Puc^LacZ expression are detected in ePG11 cells. (B, D) In zfh1 mutants Puc^LacZ expression was increased in ePG10 both at late stage 12 and stage 13 embryos. (E) Forced expression of Bsk^DN in glial cells prevents ePG10 cell death. (F) Quantification of ePG10 cell survival in different mutant combinations compared with wild type in stage 15 embryos. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right.

We next decided to study the putative effect of this JNK pathway activation on the death of ePG10 in zfh1 mutant embryos. First, we used UAS-Puc driven by Repo-Gal4 to antagonize JNK signalling. This led to the significant survival of ePG10 in 58% of hemisegments (n=268) from stages 13–16 (Figure 7F; data not shown). These results support the idea that death of ePG10 in zfh1 mutants involves, at least in part, the recruitment of the JNK signalling. Second, we took advantage of the existence of a dominant-negative form of JNK (Basket), UAS-bsk^DN (Bsk^DN) (Adachi-Yamada et al, 1999b), which is the most potent JNK pathway inhibitor available so far (Galko and Krasnow, 2004). Specific over-expression of Bsk^DN in glial cells, resulted in the survival of ePG10 in 83% of hemisegments (n=367) compared with 35% (n=81) in zfh1 mutant embryos from stages 13 to 16 (Figure 7E and F; data not shown). Altogether these results show that in the absence of Zfh1 the JNK signalling pathway is activated in ePG10 and triggers its apoptosis. They also support the view that Zfh1 acts as a negative regulator of this signalling.

Zfh1 is capable of inhibiting apoptosis downstream of JNK activation

The earlier experiments showed that Zfh1 is epistatic to the JNK pathway. To investigate at which level(s) of the pathway (of which the core components are → Hep → Bsk → DJUN → Puc) Zfh1 may act, we took advantage of the availability of an activated form of the MAPKK Hemipterous (Hep) and denoted HepAct here. We thus over-expressed HepAct in all glial cells using RepoGal4∷UAS-HepAct in different conditions of zfh1 gene dosage. By monitoring pucE69 expression as a readout we noticed strong activation of the pathway in most PG, independently of the zfh1 gene dosage condition. It is notable that in the ventral region we consistently found a pool of three or four glial cells that never displayed robust pucE69 expression (Figure 8A–C) indicating they are not permissive to JNK pathway activation. In contrast, we found robust expression of pucE69 in ePG10 despite the presence of Zfh1. This result suggests that the earlier shown interference of Zfh1 with the JNK pathway occurs at the level of or upstream of Hep. Surprisingly, over-expression of HepAct in wild-type embryos did not trigger massive cell death among glial cells and in ePG10 in particular. However, careful quantification showed apoptosis of ePG10 in 4% of hemisegments at stage 16 (n=146) thus confirming the potential pro-apoptotic activity of this signalling (Figure 8A, B and D for quantification). This phenotype indicates that pro-survival determinants might act by blocking downstream cell death effectors triggered by this pathway. In light of the anti-apoptotic effect of Zfh1 in ePG10 we addressed whether this molecule could fulfil such a role. To test this hypothesis we over-expressed HepAct in zfh1⁵ heterozygotes. This led to the death of ePG10 in 16% of hemisegments (either absence or pyknotic nuclei of ePG10) showing a four-fold increase in cell death compared with the same experiment carried out in wild-type background (Figure 8B–D).

Figure 8.

Zfh1 is able to inhibit ePG10 apoptosis downstream of JNK activation. (A–C) When HepAct is over-expressed in glial cells ePG10, ePG11, ePG12 and a subset of PG located ventrally strongly express Puc^LacZ compared with wild type. Interestingly, in each hemisegment three or four ventrally located glial cells never displayed robust Puc^LacZ expression. (D) Quantification of ePG10 cell survival when JNK signalling is activated in different conditions of zfh1 gene dosage compared with wild type. (E) Proposed model for ePG10 survival. A DIAP1 and reaper-dependant cell death mechanism in ePG10 are controlled by zfh1 and JNK activation. Zfh1 may block transcription of one or more component(s) of the JNK signalling, including upstream actor(s), thus shutting down or maintaining a low-activated state of this pro-apoptotic pathway (grey line). Alternatively, Zfh1 may repress rpr independently of its control of the JNK pathway. The fact that rpr has been reported in other models to be able to activate its own expression by sequestering DIAP1 and thus recruiting the JNK pathway (grey line) could explain why in zfh1 mutants increased expression of puckered is observed. These two mechanisms are not exclusive as Zfh1 may be able to repress in parallel components of the JNK pathway and rpr. Red and green indicate pro-apoptotic and anti-apoptotic molecules, respectively. Arrowheads: ePG10; arrows: ePG11; asterisks: ePG12 neuron; brackets: five ligament cells of the chordotonal organ. Anterior is up and dorsal is to the right.

Altogether these experiments show that Zfh1, in addition to its role as an inhibitor of the JNK pathway, can also prevent apoptosis by blocking downstream pro-apoptotic effectors triggered by JNK signalling. On the basis of our genetic evidence we propose the model presented in Figure 8E to illustrate how Zfh1 may act to prevent cell death in ePG10.

Discussion

Control of cell survival by Zfh1 family members: a general feature?

In Drosophila, several studies have showed that loss of zfh1 function often leads to a reduction in cell numbers in mesodermal derivatives (Lai et al, 1993; Moore et al, 1998; Su et al, 1999). However, it is still not known whether these phenotypes are due to aberrant cell death and/or proliferation defects. In the light of our present work it will be interesting to see whether zfh1 also has an anti-apoptotic function in these structures. Recently, it was shown that specific inactivation of Zfhx1b in the mouse embryonic dorsal telencephalon leads to proliferation defects and to increased apoptosis in the developing hippocampus. In particular, apoptotic figures were found in the cortical plate in which differentiating neurons normally accumulate (Miquelajauregui et al, 2007). In humans, while mutations or deletions in ZFHX1B have been found to cause the Mowat–Wilson Syndrome, which is characterized by typical facial features, microcephaly and severe mental retardation (Mowat et al, 2003), nonsense and frameshift mutations in ZFHX1A (also named TCF8) have been associated with Fuchs corneal dystrophy (Krafchak et al, 2005). Patients with Fuchs' dystrophy display a progressive loss of cells in the corneal endothelium and it has been proposed that excessive apoptosis may be an important mechanism in the pathogenesis (Borderie et al, 2000; Li et al, 2001). In addition, ZFHX1A was shown to have a key function within a neuronal survival pathway rapidly induced in response to cortical ischemia. This pro-survival response, initiated through the rapid induction of p63, is mediated ultimately by the transcriptional repression of a pro-apoptotic isoform of p73 by ZFHX1A (Bui et al, 2009). These observations together with our results indicate that an anti-apoptotic role of Zfh1 family members may be a general feature and thus have important impacts on the understanding of the function of this family of transcriptional regulators in both physiological and pathological conditions.

During ePG10 differentiation zfh1 is specifically required for its survival

Several lines of evidence support the cell-autonomous action of zfh1 in the survival of the subperineurial cell ePG10. First, in the PNS, Zfh1 is specifically expressed by all embryonic PG. Second, in zfh1 mutants, the death of ePG10 can be prevented by targeted glial-specific expression of zfh1. Third, over-expression of the viral caspase inhibitor p35 or the anti-apoptotic molecule Diap1 in postmitotic glial cells in a zfh1 mutant background prevents ePG10 cell fragmentation and apoptosis. Of note, rescues with UAS-p35 and UAS-Diap1 seem more efficient than with UAS-zfh1. One possible explanation is that p35 and Diap1 interfere directly with the cell death machinery, whereas for Zfh1 intermediate transcriptional events are necessary to block apoptosis. This interpretation would suggest that in ePG10 pro-apoptotic signals need to be held in check soon after its birth, that is around stage 12.

Of particular importance, in zfh1 mutants in which ePG10 is rescued by p35, the molecular identity of this particular glial cell seems normal as it is Repo+, Gliotactin+, wingless+, mirror+ and Mz97+ and its shape and migratory behaviour is comparable to wild type. These observations indicate that zfh1 does not seem to have an instructive role for aspects of ePG10 differentiation other than survival, and that ePG10 cell death in zfh1 mutant embryos is likely not a consequence of impaired differentiation.

It is notable that intriguing links between factors that promote cell migration and those that control cell survival have been reported recently, as both events appear often to be induced or inhibited by the same molecules (Montell, 2006). During a given differentiation process, the cell migration and cell survival processes seem tightly linked as cells ‘live to migrate' and ‘migrate to live' (Montell, 2006). Interestingly, a recent report shows that a precise temporal and quantitative regulation of Numb and Notch activities is required for ePG10 migration (Edenfeld et al, 2007) without affecting its survival. Our observations, together with the results of Edenfeld et al (2007) indicate that, at least for ePG10, survival and migration mechanisms can be molecularly dissociated.

Cell survival requirement within different subtypes of PG

Despite the fact that Zfh1 is expressed by all PG, all three zfh1 mutant alleles used in this study (zfh1⁵, zfh1^75.26. and zfh1^LacZ) exhibit a highly penetrant phenotype only in ePG10. We cannot completely exclude, however, that other glial cells will be affected in zfh1⁵ or zfh1^75.26 mutants at later stages, but due to the severe disorganization of several tissues and to their embryonic lethality we could not test this hypothesis. However, use of the hypomorphic zfh1^LacZ allele allowed us to determine that even at larval stage L1 only ePG10 seem to be affected whereas other glial cells develop grossly normally. This observation underlines that ePG10 is more sensitive to loss of zfh1 function and might reflect the singularity of this cell.

Indeed, within the PG born at the periphery, ePG10 is the only subtype able to migrate over long distances as a single cell (Umesono et al, 2002; Edenfeld et al, 2007). It is also known that glia form a heterogeneous cell population in terms of shape, structure or molecular identity (von Hilchen et al, 2008). Interestingly, regarding its physiological role, ePG10 belongs to the SPG subtype (gliotactin+). SPG form a thin layer below the perineurial cells, ensheath the peripheral nerve and insulate it against the potassium-rich hemolymph (Auld et al, 1995; Stork et al, 2008; von Hilchen et al, 2008). As suggested by electron microscopy analyses, only few SPG populate the peripheral nerve, they are very large and can form autocellular septate junctions (Stork et al, 2008). These observations are in agreement with earlier studies that have shown that during larval life the SPG do not divide but grow enormously large in size (Sepp et al, 2000; Schulte et al, 2003, 2006). As cell division or apoptosis of such cell type would imply disintegration of septate junctions and thus result in the opening of the blood–brain barrier, one can consider that SPG are under strict cell cycle and cell survival control. There may thus exist intrinsic differences between glial cell types that make them more or less sensitive to the balance between death and pro-survival signals. Our data indicate that ePG10 is highly susceptible to the balance between pro- and anti-apoptotic signals. First, we have noted using a hypomorphic allele of Diap1 (th^j5C8) that ePG10 is the only PG to undergo apoptosis during stages 13–16 in th^j5C8−/− mutant embryos. Second, in zfh1 mutant embryos carrying only one copy of the H99 (or XR38) deficiency, ePG10 cell death is reverted. Third, ectopic activation of JNK pathway using HepAct in zfh1⁵ heterozygotes leads to ePG10 cell death whereas other PG appear unaffected.

Zfh1 as a new transcriptional regulator influencing RHG-motifs genes expression

It has been previously proposed that apoptosis may be controlled transcriptionally (Jacobson et al, 1997). The transcriptional regulation of the three classical cell death activators reaper, hid and grim has emerged as an important issue that has been addressed within different cellular systems in which apoptosis takes place. For example, during histolysis of larval salivary glands and midgut, signalling by steroid hormone activates reaper and hid expression (Jiang et al, 1997). In response to irradiation, the Drosophila p53 homologue has been shown to activate reaper (Brodsky et al, 2000). During development, the Hox gene deformed can directly activate reaper expression at the segmental boundaries in the maxillary segment (Lohmann et al, 2002).

Interestingly, in some other developing tissues such as the salivary gland, PCD does not occur (Myat and Andrew, 2000). In this model, it has been found that senseless (sens) together with Fork head (fkh) function as anti-apoptotic transcription factors by preventing the expression of reaper and hid. By blocking these pro-apoptotic genes, they allow survival of the salivary gland cells in the embryo (Myat and Andrew, 2000; Chandrasekaran and Beckendorf, 2003). In this study we have analysed SPG that are thought to be under strict cell cycle and survival control (Stork et al, 2008). By expression analyses and genetics we have found that, among the four earlier well-characterized contiguous RHG-motif genes (hid, grim, rpr and sickle), rpr is the major cell death executor in ePG10 in zfh1 mutants. At the expression level, using two different rpr^Gal4 enhancer trap lines we have found that in zfh1⁵ mutants, rpr is upregulated in this cell. Quantifications showed that this occurs in 10% of the hemisegments at stages 13–14. However, this percentage is likely to be an under-representation, and several arguments could account for such occasional rpr deregulation: (i) we have described that in zfh1⁵ mutants, death of ePG10 is a progressive process spanning from late stage 12 up to stage 16 (see Figure 3H), which might reflect a dynamic rpr expression. Indeed, this has been already reported by Miguel-Aliaga and Thor (2004) within the embryonic nervous system and in naturally dying dMP2 neurons; (ii) the time required for rpr^Gal4 levels to become high enough to be detected before the death of ePG10 may also represent an important limitation; (iii) finally, both rpr^Gal4 insertions represent hypomorphic rpr alleles, and in our system, reducing rpr gene dosage leads to a significant reversion of ePG10 apoptosis (from 75 to 35%), which might reduce the probability of detecting rpr upregulation in a dying ePG10 cell in a given embryo at a given stage. At the genetic level, using combinations of alleles affecting hid, grim, rpr and/or sickle, we concluded that rpr is the major determinant of ePG10 apoptosis. Indeed, zfh1 mutants carrying either a H99 deletion (which removes hid, grim and rpr) or a XR38 deletion (which removes rpr and sickle) or a hypomorphic rpr allele, ePG10 survive in a similar manner, that is in ∼70% of hemisegments. Intriguingly, this number reaches 93 and 88% using UAS-Diap1 or UAS-p35, respectively, raising the possibility that Jafrac2, the last known member of the RHG-motif gene family, could be implicated. As no loss of function allele has been so far reported for this gene, we were not able to test this hypothesis. Our results thus constitute a new example illustrating how, in cells devoid of PCD, the active transcriptional repression of RHG-motif genes is a defining factor enabling cell survival.

Fitting Zfh1 into the JNK signalling pathway for glial cell survival

In tissues normally devoid of PCD, what are the upstream pro-apoptotic signalling pathways that have to be actively held in check to ensure cell survival? How do these signalling interact with the transcriptional cascades involved in the regulation of RHG-motif genes expression?

In vertebrates, the JNK pathway regulates proliferation and/or death in several cell types (Leppa and Bohmann, 1999; Ham et al, 2000). Interestingly, in Schwann cells this pathway is required for TGFβ-induced apoptosis (Parkinson et al, 2001). Our study also points for the first time towards a key role for the JNK signalling in the regulation of PG cell survival/death in Drosophila. Indeed, the JNK pathway is activated in ePG10 in zfh1 mutants, and blocking this pathway (using Bsk^DN) reverts the zfh1 phenotype (ePG10 cell death). These results indicate that Zfh1 is epistatic to JNK signalling. It is well established that JNK pathway can induce RHG-motif gene expression (rpr, hid) leading to apoptosis (reviewed in Varfolomeev and Ashkenazi, 2004). As DIAP1 has been shown to be a negative regulator of the JNK pathway, Rpr by sequestering DIAP1, enhances its own expression by a feed-forward loop (Kuranaga et al, 2002; Ryoo et al, 2004). In our system, in light of our observation that pucE69 expression is increased in ePG10 in zfh1 mutants, several hypotheses can be proposed: (i) Zfh1 could block transcription of one or more component(s) of the JNK signalling, thus maintaining the pro-apoptotic activity of this pathway in a low or inactive state. Our gain of function experiments using HepAct showed that Zfh1 does not seem to influence the core components of the cascade but rather acts upstream or at the level of Hep. It also strongly supports the view that Zfh1 is able to prevent apoptosis by blocking downstream pro-apoptotic effectors triggered by JNK signalling; and (ii) Zfh1 could act downstream of the core JNK cascade by solely repressing rpr independently of its control of the JNK pathway. Indeed, the fact that rpr can enhance its own expression by recruiting the JNK pathway (as mentioned above) could explain why in zfh1 mutants we observe activation of this signalling. These two possibilities are non-exclusive as Zfh1 may be able to repress in parallel components of the JNK pathway and rpr (Figure 8E). The fact that Zfh1 orthologs have been mainly described as transcriptional repressors (Postigo and Dean, 2000) is consistent with these hypotheses. Further analyses are however needed to test whether these actions are direct or indirect and to delineate what are the precise target(s) of Zfh1 in this pathway. The genetic interaction of Zfh1 with the JNK pathway that we have reported in glial cell survival could be a more general feature in other processes and cell types. Indeed, during the course of this study we have noticed that in zfh1 mutant embryos JNK signalling is dramatically induced in fat body cells (in which Zfh1 is normally expressed, data not shown) thereby triggering massive cell death in this tissue (data not shown). In addition, it has been recently hypothesized that Zfh1 could act through the JNK pathway during neuromuscular junction growth (Vogler and Urban, 2008). Interestingly, a recent analysis on the role of Zfhx1b, one of the two vertebrate homologues of zfh1, has showed such an interaction during hippocampus formation. In this model, Zfhx1b seems to have an anti-apoptotic function in neuronal cells by acting as an indirect activator of a JNK-dependant pro-survival pathway (Miquelajauregui et al, 2007), whereas in our system we have shown that the JNK pathway displays a pro-apoptotic activity. However, this discrepancy can be explained by the fact that the JNK pathway functions in a cell type and stimulus-dependent manner, as its components can have opposing roles in apoptosis (reviewed in Lin, 2003; Varfolomeev and Ashkenazi, 2004). In conclusion, these data raise the possibility that the interaction between Zfh1 and the JNK pathway has been conserved during evolution (eventually through distinct mechanisms depending on the cellular context) with cell survival being the final outcome.

Materials and methods

Drosophila stocks and P-Gal4 lines analysis

All genetic crosses were performed at 25°C unless otherwise specified. The following fly stocks were used in this study: w¹¹¹⁸ was used as a wild-type stock; UAS-myc-EGFP^F (denoted UAS-EGFP^F) (Allan et al, 2003); UAS-nls-myc-EGFP (denoted UAS-nlsEGFP) (Callahan et al, 1998) (from D van Meyel); Repo^Gal4(Sepp et al, 2001) (from U Gaul); J29^LacZ enhancer trap insertion that reflects gliotactin expression (Auld et al, 1995) (from C Klämbt); zfh1⁵ (Lai et al, 1993) (obtained from JC Lai); zfh1^75.26 (Broihier et al, 1998) (from R Lehmann); zfh1⁸⁶⁵ (Justice et al, 1995); UAS-diap1 (Peterson et al, 2003) (from K McCall), UAS-puc (Martin-Blanco et al, 1998) (from S Noselli). The following lines th^j5C8 (Hay et al, 1995); H99 deficiency (White et al, 1994), UAS-p35 (Hay et al, 1994), UAS-bsk^DN (Adachi-Yamada et al, 1999b) and UAS-zfh1 were obtained from the Bloomington stock centre. Over-expression of each transgene UAS-p35, UAS-Diap1, UAS-Bsk^DN, UAS-zfh1, UAS-Puc with Repo-Gal4 in a wild-type situation is shown in Supplementary Figure S2A–E. The XR38 deficiency (Peterson et al, 2002), wingless^LacZ, mirror^LacZ and Mz97^Gal4 (von Hilchen et al, 2008) were provided by K White, O Vef and G Technau. The rpr^Gal4 lines (103634, NP: 520) and (112155, NP: 368) were generated and mapped by the NP Consortium (http://flymap.lab.nig.ac.jp/~dclust/getdb.html) are referred to as rpr^Gal4(520) and rpr^Gal4(368)_, respectively, in this study. For precise mapping of the rpr^Gal4 lines genomic DNA PCR was carried out using a sense primer located in the 5′end of the P-Gal4 vector (5′-gtttgggagagtagcgacac-3′) and an antisens primer (5′-caagcgaaggatctgctgct-3′) located in the 5′ end of the rpr gene. All mutants lines were kept over CyO,Act-GFP or TM3,Ser,Act-GFP balancer chromosomes.

Immunostaining of embryos and western-blot analysis

Immunolabelling was carried out as described earlier (Thor et al, 1999). For first instar larval (stL1) dissection embryos were preselected during the time when their main dorsal tracheae begin to fill with air, which represents 18 h after egg laying (AEL) and allowed to develop for further 3 h. First instar larvae (21 h AEL) were dissected as described in Baines and Bate (1998). The following antibodies were used: α-22C10 (1:40), α-Elav (1:50), α-Repo 8D12 (1:80), α-c-Myc 9E10 (1:50), α-Repo 8D12 (1:80), α-Prospero MR1A (1:40), Fas2 1D4 (1:40) and α- βGal 40-1a (1:10) (all from Developmental Studies Hybridoma Bank), rabbit α-Zfh1 (Van Doren et al, 2003) (1:5000), rabbit α-cleaved caspase-3 (Asp175, Cell Signaling; 1:200), rabbit α-Grim (Claveria et al, 1998) (1:1000), rabbit α-Hid (Bergmann et al, 2002) (1:500), rabbit α-Rpr (Freel et al, 2008) (1:1000 for western blot), mouse α-β tubulin (T5168 clone, Sigma; 1:10 000 for western blot), mouse α-GFP (A11120 clone, Molecular Probes, Eugene, OR; 1:250), rabbit α- βGal (Cappel; 1:5000), before use, the polyclonal α- βGal, α-GFP, α-Grim, α-Hid antibodies were pre-absorbed against early stage wild-type embryos. Secondary antibodies Alexafluor488 and Cy3 were used at 1:500 and 1:2000, respectively. Tissues were imaged on a Bio-Rad confocal microscope. Double-labelled images were false coloured using Photoshop for the benefit of colour-blind readers. For reaper western blot, preparations of proteins extracts were carried out as described in Freel et al (2008) and we used chemiluminescent HRP substrate for detection (Immobilon Western, from Millipore).

Supplementary Material

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure Legends

Review Process File