A molecular link A molecular link between Hairless and Pros26.4, a member of the AAA-ATPase subunits of the proteasome 19S regulatory particle in Drosophila

The proteasome is the major degradation machinery of the cell that regulates multiple cellular processes as diverse as cell cycle, signal transduction and gene expression. Recognition and unfolding of target proteins involves the regulatory cap whose base contains six AAA-ATPases that display reverse chaperone activity. One of them, Rpt2 (also known as S4), has an essential role in gating the degradative central core. We have isolated the orthologous gene Pros26.4 from Drosophila melanogaster as a molecular interaction partner of Hairless. Hairless plays a major role as antagonist of Notch signalling in Drosophila, prompting our interest in the Hairless-Pros26.4 interaction. We find that Pros26.4 negatively regulates Hairless at the genetic and molecular level. Depletion of Pros26.4 by using tissue-specific RNA interference (RNAi) resulted in a specific stabilization of the Hairless protein, but not in stabilization of the intracellular domain of Notch or the effector protein Suppressor of Hairless. Thus, the Hairless-Pros26.4 interaction provides a novel mechanism of positive regulation of Notch signalling.

Introduction Regulation of Hairless by Pros26.4 Pros26.4 gene, which encodes the Rpt2 orthologue in Drosophila (Hölzl et al., 2000). In accordance with its function in the proteasome, Pros26.4 encodes a general, essential factor during fly development. Knock-down of Pros26.4 activity by RNA interference (RNAi) caused lethality and cell degeneration in accordance with its essential cellular role. At the same time Pros26.4 depletion resulted in a specific stabilization of Hairless but neither of Su(H) nor Notch proteins. The Hairless-Pros26.4 interaction might thus reflect a novel mechanism of positive regulation of the Notch signalling pathway.

Proteasome subunit Pros26.4 interacts on a molecular level with the Notch antagonist Hairless
In a yeast two-hybrid screen for potential molecular partners of Hairless, clone pJG7-9 was the most frequently isolated and also corresponds to Pros26.4 (Marquart, 1999). Pros26.4 has a single transcript and associated protein (flybase). In comparison to the published sequence, clone pJG7-9 lacked 13 codons (K 47 -L 59 ) and had N-terminally an extension of seven codons that were derived from untranslated leader sequences. To exclude artificial interactions, the Pros26.4 coding sequence was recloned from a full-length cDNA clone to generate pJG-S4. In yeast two-hybrid assays, pJG-S4 behaved identical to pJG7-9. Using several deletion constructs of Hairless, the Pros26.4-interacting domain was mapped within the Cterminal third of the Hairless protein (CX; Fig. 1A,B). Moreover, the Hairless-interacting domain was mapped to the N-terminal half of Pros26.4, excluding an involvement of the AAA-ATPase domain (S4-I; Fig. 1C,D). The molecular interactions were confirmed in vivo: anti-Hairless immunoprecipitates from embryonic protein extracts indeed contained Pros26.4 protein (Fig. 1E).
Expression of Pros26.4 during Drosophila development Molecular interactions are relevant only when the respective proteins are expressed at the same time within the same tissues during development. We thus analysed the expression patterns of Pros26.4 mRNA and protein. Pros26.4 mRNA is uniformly distributed in early eggs owing to maternal product ( Fig. 2A), as has been described before for Hairless transcripts. Later in development, transcripts appeared enriched within the developing nervous system (Fig. 2B-H) thus overlapping the Hairless expression domain (Maier et al., 1992). In third instar larvae, Pros26.4 mRNA levels were higher in dividing cells,  Maier et al. (Maier et al., 1997)]. (B) Quantification of the interaction of Pros26.4 (pJG-S4) with Hairless and its deletion constructs. The quantitative yeast two-hybrid assay shows that the Hairless full-length construct (HFL) as well as all Hairless deletion constructs, with the exception of CX, retain interaction capacity. Control was empty vector (pEG). Values represent Miller units. (C) Drawing of Pros26.4 protein; the conserved AAA-ATPase domain (AAA) is located in the C-terminal half; ATP-bindings sites ('Walker'motifs A and B, and motif C) are highlighted (Confalonieri and Duguet, 1995). Below, the deletion constructs S4-I to S4-V in pJG are shown. Numbers refer to included codons. (D) Interactions of Pros26.4 deletion constructs and Hairless HFL in pEG were quantified: only S4-I, which contains the N-terminal half but lacks the ATPase domain, retains its binding activity, which is in contrast with any of the other constructs. Control was empty vector (pEG). (E) Coimmunoprecipitation of Pros26.4 and Hairless from embryonic extracts. Protein extracts from Drosophila embryos were taken for immunoprecipitation using anti-Hairless antiserum (IP H). Precipitates were probed with either Hairless antiserum (left) or Pros26.4 antiserum (anti-S4; right). For comparison, total embryonic extracts were loaded in the next lane (in, input about 12% of IP). As control (co), the precipitation was performed with unrelated antiserum. Molecular mass is given in kDa. e.g. in neuroblasts of the ventral chord or in eye imaginal discs anterior to the morphogenetic furrow ( Fig. 2I-L). This pattern is very similar to that of Pros28.1, which encodes an ␣-subunit of the central core (Fig. 2M-P;Udvardy, 1993). Pros26.4 protein is ubiquitously distributed in all tissues; it is cytoplasmic and also, albeit to a lesser degree, nuclear (Fig. 3C). During oogenesis the protein is enriched in the germinal vesicle (Fig. 3A,B) as was reported before for the ␣and ␤-subunits of the 20S central core (Adam et al., 2004). The general distribution of both Pros26.4 and Hairless proteins (Maier et al., 1999) indicates extensive overlap in many tissues.
Depletion of Pros26.4 during Drosophila development results in cellular degeneration and death Pros26.4 was overexpressed in a variety of different tissues using the Gal4/UAS system (Brand and Perrimon, 1993). Despite a strong expression, the flies developed completely normal without any obvious phenotypes (not shown). Based on the close molecular contacts between the six AAA-ATPases forming a ring at the base of the 19S regulatory cap (Wollenberg and Swafield, 2001;Lam et al., 2002), one might expect that overexpression of any of them disrupts the stoichiometry of the ring and, therefore, Journal of Cell Science 119 (2) (B) With the onset of neurogenesis, transcripts start to accumulate in presumptive neuronal cells (arrow; stage 11). (C) Nervous-system-specific accumulation of Pros26.4 mRNA is apparent with germ-band retraction (br, brain; vc, ventral chord). (D) At the time of dorsal closure (stage 14), expression is mainly detected in the ventral chord (vc) and the brain (br). (E) A close-up of the developing head of a stage-12 embryo shows a modulated expression in the brain (br). (F) A dorsal view on an extended germ band embryo shows the modulated expression (arrow) in the nervous system anlage (stage 10). (G) A ventral view on a stage-11 embryo shows accumulation of Pros26.4 mRNA in presumptive neuroblasts (arrow). (H) A dorsal view on a stage-14 embryo highlights the strong mRNA expression in the two brain lobes (br) and also enrichment in the posterior spiracles (ps). Embryos are oriented with anterior left and dorsal up unless otherwise noted. Stages are according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1997). (I) Pros26.4 mRNA transcription levels remain high in proliferating neuroblasts of the ventral chord (arrow) and also in the ring gland (rg) of third instar larvae. (J) In the eye disc (ed) expression is mainly observed in cells anterior to the morphogenetic furrow, whereas in the antennal disc (ad) mRNA distribution is more uniform. (K) In the wing disc, expression is observed all over in a modulated pattern; it is weaker in presumptive vein areas (arrow) and the zone of non-proliferation (asterisk). (L) A likewise modulated expression is observed in the leg disc. (M-P) Expression of Pros28.1, which encodes the ␣-subunit of the catalytic 20S core of the proteasome, is very similar when compared with that of Pros26.4. proteasome activity. However, this is not the case because Drosophila is insensitive to a large increase in the amount of Pros26.4.
Recently, it was shown in Drosophila S2 cells that reduction of Pros26.4 activity by RNAi affects cell growth rates and results in an increased number of apoptotic cells (Wojcik and DeMartino, 2002). Presumably, Drosophila cells that completely lack Pros26.4 cannot survive because respective mutations in yeast are lethal . To study loss of Pros26.4 activity in Drosophila, transgenic flies were constructed containing an inverted Pros26.4 construct (UAS-dsS4) thus allowing tissue-specific RNAi . In the animal, ubiquitous knock-down with the da-Gal4 driver caused death at about second larval instar. Likewise, larval or pupal death was observed upon tissue-specific RNAi when using a number of different driver lines (e.g. ptc-Gal4, en-Gal4, sca-Gal4). Apparently, depletion of Pros26.4 caused cell death not only in Drosophila cell lines but also during fly development (see Fig.  4). For example, specific expression of UAS-dsS4 within the eye anlagen using ey-Gal4 allowed us to recover pharate adults from the pupal case that lacked the eyes and most of the head capsule ( Fig. 4A,G). Closer inspection of the defects caused by loss of Pros26.4 was possible by using gmr-Gal4, which drives expression later in development within differentiating cells of the eye imaginal disc. In these cases, we found fusions of ommatidia in a posterior to anterior gradient in the adult eye (Fig. 4D, compare with 4A). Frequently, necrotic tissue was attached to the very posterior end of the eye. The underlying defect was revealed in sections, which showed a likewise graded loss of internal retinal tissue (Fig. 4E, compare with 4B), reflecting the respective differentiation wave. Only cells behind the morphogenetic furrow, which sweeps from posterior to anterior over the eye imaginal discs during larval life, express gmr-Gal4 (Wolff and Ready, 1993). Thus, cells furthest anterior were the last to be exposed to RNAi and survived, whereas those more posterior were destroyed. A strong increase in cell death was already observed in third larval instar eye discs (Fig. 4C,F), and could be rescued by simultaneous overexpression of the antiapoptotic baculoviral p35 protein (Fig. 4H). The defects seen in the adult, however, arose later in development, and only a marginal rescue was caused by the overexpression p35 (not shown). Apparently, interference with Pros26.4 activity occurs with a pronounced time-lag, presumably due to a stable pool of Pros26.4 protein. As a consequence, cells eventually die and the retina degenerates with a late onset.

Pros26.4 acts as a negative regulator of Hairless
The molecular interactions between Pros26.4 and Hairless prompted us to investigate possible genetic interactions. As Hairless functions as antagonist of Notch signalling, overexpression of Hairless results in typical Notch loss-offunction phenotypes throughout development (e.g. Maier et al., 1997). However, Hairless gain-of-function phenotypes were not altered by the concurrent overexpression of Pros26.4 (data not shown). Mutations in Hairless are haplo-insufficient and cause a dominant loss of bristles and wing venation defects. This phenotype was rescued to some extent by simultaneously removing one copy of Pros26.4 by using deficiency Df(3R)mbc-R1: flies heterozygous for H P8 in a wild-type background lacked 12.2 macrochaetae on average (see Maier et al., 2002), whereas in the background of Df(3R)mbc-R1, they lacked only 9.7 bristles on average. This slight amelioration of the Hairless lossof-function phenotype by concurrent loss of Pros26.4 argues for a negative regulation of Hairless by Pros26.4. One role of Pros26.4 might be, for example, to destabilize the Hairless protein. However, this deficiency removes further genes of unknown function that might influence the Hairless mutant phenotype as well. To investigate whether reduction of Pros26.4 activity by RNAi likewise restored bristle numbers, we overexpressed UAS-dsS4 within proneural clusters, including the bristle precursor cells using the sca-Gal4 driver line. Unfortunately, the animals died at late pupal stage and displayed a reduced number of bristles on their own. This is not unexpected, considering that Pros26.4 is required in cell survival as outlined above. However, it prevented us from investigating the genetic interactions. We therefore decided to study eye development instead.
Knock-down of Pros26.4 in the developing eye by driving UAS-dsS4 with gmr-Gal4 caused ommatidial fusions owing to the degeneration of the retina (Fig. 4D,E, Fig. 5A,B). Overexpression of Hairless results in rough and smaller eyes (Fig.  5C) as a consequence of interference with Notch signalling during differentiation and survival of cells in the retina (Artavanis-Tsakonas et al., 1999;Müller et al., 2005). This gain-of-function phenotype was markedly enhanced by a knockdown of Pros26.4: the eyes became even smaller with a very glossy surface, indicative of strong defects in retina development (Fig. 5D). This is in agreement with the hypothesis that Pros26.4 normally downregulates Hairless: reduction of Pros26.4 activity due to RNAi should therefore enhance phenotypes caused by a surplus of Hairless. By contrast, no effect was seen on the enlarged eyes caused by the overexpression of Su(H) or Notch (Fig. 5E-H). Overexpressing either Su(H) or Notch in combination with dsS4 gave the expected additive phenotype: eyes were still enlarged, but showed a glossy appearance and necrotic tissue caused by the depletion of Pros26.4 (Fig. 5F,H).

Pros26.4 depletion results in a stabilization of Hairless protein but not of Notch or Su(H)
To test the effect of Pros26.4 on protein stability more directly, we assayed levels of Hairless, Su(H) or Notch proteins while reducing Pros26.4 by RNAi. Cell death resulting from such a depletion was partially prevented by driving the Drosophila inhibitor of apoptosis protein DIAP1 (Lisi et al., 2000) simultaneously with the dsS4 construct. Notch, Su(H) or Hairless proteins were induced all over and, with respect to Pros26.4 knock-down, stability of each protein was assayed at specific time points after induction. For this experiment, either Notch, Su(H) or Hairless transgenes were induced from a heat-shock promoter by a 1-hour heatpulse in animals where dsS4 and/or DIAP1 had been activated along the antero-posterior border of imaginal discs (Fig. 6). After a given time at ambient temperature, wing discs were immuno-stained for the respective proteins. They were compared with controls that had not been subjected to the heatpulse and in which no effect on protein distribution was seen. As shown in Fig. 6A, Hairless protein was slightly stabilized by overexpression of DIAP1 and very strongly, when Pros26.4 activity was also reduced (Fig. 6B,B'). By contrast, the stability of either Notch or Su(H) proteins remained unchanged (Fig.

Journal of Cell Science 119 (2)
6C-F') arguing that Pros26.4 had no influence on these proteins. These data are in agreement with earlier findings that RNAi for Pros26.4 does not result in an immediate failure of proteasome activity (Wojcik and DeMartino, 2002;Boutros et al., 2004). Otherwise, we would have expected stabilization of Notch as well, because this effect was observed with a dominant negative mutation in the proteasomal ␤-subunit (Schweisguth, 1999). Rather, the particular stabilization of Hairless hints to a specific regulation of this protein by Pros26.4.

Pros26.4 is essential for cellular viability in Drosophila
Pros26.4 is one of six AAA-ATPases that form the base of the 19S regulatory particle of the 26S proteasome. It is an essential component not only in yeast but also in insect cells: a high throughput RNAi screen for defects in cell growth and viability identified, among a few others, Pros26.4 in Drosophila S2-and Kc-cells (Boutros et al., 2004). In accordance with our observation of a significant time-lag of the Pros26.4 RNAieffects, the cells initially grew normally before growth rate and cell number decreased significantly.
The 19S regulatory particle is indispensable to the recognition and unfolding of proteins destined for degradation by the 20S core particle. Thus, it is not surprising that interference with its activity results in cell death. In addition, it has been assigned a number of other, completely different, activities: in yeast it was shown that the 19S regulatory particle plays an independent role in nucleotide excision repair of damaged DNA and, moreover, in transcription activation and elongation (reviewed in Muratani and Tansey, 2003). To date it is unknown whether these multiple activities also apply to the fly. If this were the case RNAi with Pros26.4 might interfere with all these processes. Interestingly, impeding the activity of the central core subunits ␤6 or ␤2 causes superficially very similar phenotypes to depletion of Pros26.4 (Belote and Fortier, 2002). The authors mis-expressed respective dominant negative mutants DTS5 and DTS7 in the photoreceptor cells. They observed necrosis after a noticeable time-lag but did not analyse the underlying defects. The stunning resemblance of phenotypes strongly suggests that, above all, RNAi of Pros26.4 interferes with proteasome activity proper. Apparently, the function of Pros26.4 within the proteasome degradation machinery is the most dose-sensitive, such that reduction in activity causes instant and obvious defects. By means of the tissue-specific RNAi assay we can now identify and further analyse possible roles of Pros26.4 in Drosophila, e.g. in gene transcription or nucleotide-excision repair.
The ubiquitin-proteasome system and the regulation of Notch-dependent transcription In the last few years, an intimate relationship between the proteasome and the regulation of gene expression has been established (for a review, see Muratani and Tansey, 2003). In principle, the proteasome might control transcriptional regulators at the level of availability (i.e. their translocation into the nucleus or into nuclear bodies) or at the level of activity (i.e. by delimiting co-activators or alike) and finally at the level of abundance. Indeed, it has been reported that Notch-mediated gene activation is negatively regulated by targeting the intracellular domain of Notch for proteasomal destruction (Schweisguth, 1999;Gupta-Rossi et al., 2001). Transcription-coupled proteolysis seems common to unstable transcriptional activators, in which trans-activation domains and degradation signals often overlap. Thereby, activation of the transcription machinery is directly linked to the degradation of the trans-activator, a process coined the 'black widow' model (Tansey, 2001). In the process of activation, the transcriptional regulator is marked for degradation. Indeed, many transcriptional regulators are phosphorylated by components of the basal transcription machinery. This renders them susceptible to ubiquitylation and degradation by the proteasome (Muratani and Tansey, 2003). In case of Notch signalling the process has been analysed in detail (Fryer et al., 2004). The transcription-activation complex affecting Notch signals includes the intracellular domain of Notch, a CSL-type DNA-binding protein and also the coactivator Mastermind (Mam). Mam is essential for transcription initiation and promotes hyperphosphorylation of Notch by coupling to the CycC:Cdk8 kinase. This results in PEST-dependent degradation of active Notch (Fryer et al., 2004).
Besides coupling transcriptional activation to destruction, a non-proteolytic role for the 19S regulatory particle has been stated in transcriptional activation and elongation. Notably, the proteasomal AAA-ATPases Rpt6/Sug1 and Rpt4/Sug2 are located at promoters where they direct chromatin remodelling (Gonzalez et al., 2002;Muratani and Tansey, 2003;Ezhkova and Tansey, 2004). They are recruited by mono-ubiquitylated histone H2B, which participates in the activation of transcription and relief of telomeric silencing in yeast. H2B ubiquitylation is mediated by the Rad6 complex, which includes the E3 ubiquitin ligase Bre1. It is a first and mandatory step for subsequent histone H3 methylation at lysine residues 4 and 79, which also depends on the presence of Rpt6/Sug1 and Rpt4/Sug2 (Muratani and Tansey, 2003;Wood et al., 2003;Ezhkova and Tansey, 2004). It has not yet been studied in detail, whether similar mechanisms apply to Drosophila as Journal of Cell Science 119 (2) well. However, very recently it has been shown that a mutation in Drosophila Bre1 particularly affects Notch target gene expression , arguing that, in Drosophila, Notch target genes are particularly susceptible to the effects of histone H2B ubiquitylation. Albeit all six AAA-ATPases quickly associate with active promoters in yeast, there is little evidence for a specific role of Rpt2/S4 in transcriptional regulation (Gonzalez et al., 2002). In this case, knock-down of Pros26.4 by RNAi would result in a downregulation of Notch target genes, similar to mutations in Bre1.
Our results demonstrate that Pros26.4 positively influences Notch signalling by specifically destabilizing the Notch antagonist Hairless. Our data reveal a remarkable, Pros26.4dependent preference of the proteasome for Hairless compared with other potential substrates such as Notch. We propose that this is based on a direct molecular interaction of Pros26.4 and Hairless. It is conceivable that protein-protein interactions of proteasomal AAA-ATPase subunits with candidate substrates occur independently of ubiquitylation, which regulates proteasomal access. Hairless interferes with Notch activity in two ways. On one hand, it interferes with the assembly of the transcription-activator complex by competing for CSL/Su(H) and on the other hand, it forms, directly on Notch-responsive promoters, a repressor complex together with Su(H), and the co-repressors Groucho and CtBP (Morel et al., 2001;Barolo et al., 2002;Nagel et al., 2005). Thus, by decreasing the availability of Hairless protein, Pros26.4 promotes Notch target gene activation.
In conclusion, transcriptional regulation of Notch target genes is influenced in several ways by the ubiquitinproteasome system. First, activated Notch undergoes proteasomal destruction in a typical transcriptional activationdestruction process. Second, Notch target genes are particularly susceptible to ubiquitin-dependent chromatin activation. Third, by means of Hairless destabilization, the AAA-ATPase Pros26.4 promotes Notch signalling.

Isolation and verification of Pros26.4
A yeast two-hybrid screen was performed on approximately 1.7ϫ10 7 doubly transformant clones of the Drosophila melanogaster embryonic cDNA library RFLYI (Paroush et al., 1994) with the bait construct pEG-FLH (Maier et al., 1997) according to standard protocols (Fields and Song, 1989). After three rounds of verification, 57 clones were tested positive. Five different genes were isolated; the vast majority of clones (41%) represented Pros26.4 (pJG7-9). In comparison to the published DNA sequence, clone pJG7-9 carried a 39-nucleotide gap at position 137 and contained a 21-nucleotide leader sequence but retained an open reading frame and expressed a fusion protein of the expected size. A full-length Pros26.4 cDNA clone (ICRFp520D102) was obtained from RZPD (Berlin) and used to amplify the Pros26.4 coding sequence with forward primer CAA GCA CCA AGA ATT CAT GGG ACA AA and reverse primer TAT ATC TTC GCT CGA GTT AAT AAA TAA. To generate pJG-S4, the amplified sequence was digested with EcoRI and XhoI, and cloned into likewise linearized pJG4-5 vector (Gyuris et al., 1993). In subsequent qualitative and quantitative yeast two-hybrid assays, pJG-S4 was indistinguishable from pJG7-9. Clones were sequence verified and protein expression in yeast cells was tested in western blots with antibodies directed against HA (Roche Ltd) and LexA (Clontech).

Analysis of protein-protein interactions
Co-immunoprecipitations were performed on protein extracts of approximately 500 wild-type embryos. For immunoprecipitation, we used rabbit anti-Hairless A antiserum at 1:250 dilution and for detection rat anti-Hairless A (Maier et al., 1999) and rat anti-S4 antibodies at 1:500 dilution, respectively. Secondary AP-coupled antibodies were diluted 1:200 (Jackson Laboratories). Paired protein interactions were analysed as previously described (Matsuno et al., 1995). For bait constructs pEG-FLH, pEG-C1, pEG-C2, pEG-C3, pEG-CX and pEG-C6, see Maier et al. (Maier et al., 1997). To map the interacting site of Hairless within Pros26.4, a set of five fragments was generated by PCR using Pfu-polymerase (Stratagene) and Pros26.4 full-length cDNA as template. Forward primers carried an EcoRI and reverse primers a XhoI restriction site to allow cloning in respectively linearized pJG vector (Gyuris et al., 1993). Primer sequences are available upon request. Final constructs were sequence verified.

Manipulation of gene activity
Flies were raised under standard conditions; further information on fly lines can be obtained from http://flybase.bio.indiana.edu/. Tissue-specific expression of transgenes was controlled by use of the Gal4/UAS-system (Brand and Perrimon, 1993). Driver lines were da-Gal4, en-Gal4, ey-Gal4, gmr-Gal4, ptc-gal4 and sca-Gal4. The following UAS-lines were used: UAS-DIAP1 (gift of A. Müller, Heinrich-Heine-Universität, Düsseldorf, Germany), UAS-GFP, UAS-N ICD (Go et al., 1998), UAS-Su(H) and UAS-H (Nagel et al., 2000). Heat-inducible lines were hsH, hsSu(H) and hsN ICD (Lieber et al., 1993;Schweisguth and Posakony, 1994;Maier et al., 1997). Df(3R)mbc-R1 takes out 95A5/7 to 95D6/11; Pros26.4 is located at 95C and thus contained within the deletion. UAS-S4 transgenic flies were generated with a construct containing the complete open reading frame of Pros26.4 shuttled from pEG as an EcoRI-XhoI fragment into pUAST. Tissue-specific RNAi was induced by overexpression of UAS-dsS4, which was cloned as outlined before : a DNA segment covering codons 90 to 250 was PCR amplified with primers containing sites for BamHI (5Ј) CCA GGA TCC GAA GAA CGA GGA GGA GCG CTC TAA and Kpn I (3Ј) CTC GGT ACC TAC AAC GCG CAG GAA AGT GGC C, cloned into pHIBS and shuttled into pUAST, respectively. Transformant flies were generated according to standard protocols. For the genetic interaction studies, the following fly lines were constructed by standard recombination techniques: gmr-Gal4 UAS-H 2nd / CyO (Müller et al., 2005); UAS-DIAP1, gmr-Gal4; UAS-DIAP1, UAS-dsS4; UAS-N ICD UAS-dsS4; UAS-Su(H) UAS-dsS4 / TM3Sb. Crosses were carried out at 18°C or 25°C as specified and the offspring was compared with the siblings or with control crosses with UAS-GFP lines. Pictures of adult flies were taken with a Leica Wild M3C dissecting microscope equipped with a Pixera digital camera. Scanning-electron micrographs were taken with a Joel JSM35 microscope. For the protein stability experiments, fly stocks of the following genotypes were constructed: UAS-DIAP1, ptc-Gal4; UAS-dsS4 hsN ICD ; UAS-dsS4 hsH 3rd and UAS-dsS4 hsSu(H). Crosses were reared at 25°C. Then a single 1-hour heat-pulse at 37°C was given to wandering third instar larvae, which were aged for 1, 6 or 12 hours at room temperature, before histochemical analysis; they were compared with unshocked control animals.

Immuno-histochemistry
Polyclonal antibodies directed against Pros26.4 and DIAP1 were raised in rats and rabbits immunized with GST-and MBP-fusion proteins, respectively (PINEDA ABservice Berlin). Other primary antibodies were directed against Hairless (Maier et al., 1999), Su(H) (Schweisguth and Posakony, 1994) and Notch (Fehon et al., 1991). The latter was obtained from DSHB (University of Iowa under the auspices of the National Institute of Child Health and Human Development). Secondary antibodies (1:200) coupled to DTAF or Cy3 were purchased from Jackson Laboratory (Dianova). Samples were mounted in Vectashield (Vector Lab) and analysed on a Zeiss Axioskop or a Zeiss Axiophot linked to a Bio-Rad MRC1024 and a Zeiss LSM 510 confocal microscope, respectively. Pictures were assembled using CorelDraw software.