Role of the Drosophila Non-Visual ß-Arrestin Kurtz in Hedgehog Signalling

The non-visual ß-arrestins are cytosolic proteins highly conserved across species that participate in a variety of signalling events, including plasma membrane receptor degradation, recycling, and signalling, and that can also act as scaffolding for kinases such as MAPK and Akt/PI3K. In Drosophila melanogaster, there is only a single non-visual ß-arrestin, encoded by kurtz, whose function is essential for neuronal activity. We have addressed the participation of Kurtz in signalling during the development of the imaginal discs, epithelial tissues requiring the activity of the Hedgehog, Wingless, EGFR, Notch, Insulin, and TGFβ pathways. Surprisingly, we found that the complete elimination of kurtz by genetic techniques has no major consequences in imaginal cells. In contrast, the over-expression of Kurtz in the wing disc causes a phenotype identical to the loss of Hedgehog signalling and prevents the expression of Hedgehog targets in the corresponding wing discs. The mechanism by which Kurtz antagonises Hedgehog signalling is to promote Smoothened internalization and degradation in a clathrin- and proteosomal-dependent manner. Intriguingly, the effects of Kurtz on Smoothened are independent of Gprk2 activity and of the activation state of the receptor. Our results suggest fundamental differences in the molecular mechanisms regulating receptor turnover and signalling in vertebrates and invertebrates, and they could provide important insights into divergent evolution of Hedgehog signalling in these organisms.


Introduction
G-protein coupled receptors (GPCRs) are seven-transmembrane proteins that play critical roles during development and in the regulation of cellular physiology. GPCRs constitute the largest superfamily of cell membrane receptors [1]. The major GPCR regulatory pathway involves phosphorylation of agonist-activated receptors by G protein-coupled receptor kinases (GRKs), followed by binding of the cytosolic arrestin proteins [2]. This interaction prevents the receptor from activating additional G proteins in a process known as desensitization [3]. GRKs and ß-arrestins also participate in signal propagation by recruiting additional proteins to the receptor complex [4][5][6]. Thus, the GRK/ß-arrestin pathway facilitates receptor internalization from the cell surface through clathrin-coated pits, and this leads to numerous physiological outcomes, including receptor degradation, receptor recycling and the activation of distinct downstream signalling events [2,[7][8][9][10]. Finally, more recent evidence suggest a role for ß-arrestins in signalling by other families of cellular receptors, including receptor tyrosine kinase (RTKs), non-classical 7TMRs like Smoothened and Frizzled, Notch and TGFb receptors, and also by downstream kinases such as MAPK and Akt/PI3K [5,[11][12][13].
The arrestin family is divided in two classes: the visual arrestins (arrestin 1 and 4), which are located almost exclusively in photoreceptor cells, and the non-visual b-arrestins 1 and 2 (also named arrestin 2 and 3, respectively), which are ubiquitously distributed [4]. These proteins are closely related and their sequence is highly conserved across species [14]. In Drosophila melanogaster there is only a single non-visual b-arrestin, encoded by kurtz (krz), which function is essential for development, survival and neural function [15][16][17][18]. In addition, the gene CG32683 encodes a related protein that presents some homology with b-arrestins, but lacks the clathrin-binding domain (see Figure S1). The GRK family includes seven members in humans (GRK1-7) and two components in flies (Gprk1 and Gprk2). Gprk1 modulates the amplitude of the visual response, acting as a Rhodopsin kinase, whereas Gprk2 regulates the level of cAMP during Drosophila oogenesis [19]. In addition, Gprk2 and Gprk1 play a key role in the regulation of the Hedgehog (Hh) signal transduction pathway [20][21], where they seem to phosphorylate and activate the sevenpass transmembrane protein Smoothened (Smo) [22]. The ßarrestin Krz has also been involved in the regulation of Notch signalling, promoting the formation of a trimeric Notch-Deltex-Krz complex that mediates the degradation of the Notch receptor in an ubiquitination-dependent pathway [23], reminiscent of ßarrestin-mediated ubiquitination of other canonical GPCRs [8]. More recently, Krz has also been implicated in the regulation of Smo accumulation [21] and ERK phosphorylation [24]. Because Krz is the unique ß-arrestin present in Drosophila, it is likely that the protein has additional functions in the modulation of other signalling pathways.
To address the participation of Krz in signalling events, we have analyzed its function during the development of the imaginal discs, the epithelial layers that give rise to the adult structures of the fly. Imaginal discs are very convenient model systems to study the activity of signalling pathways in vivo, because their development is under the regulation of the Hh, Wingless, EGFR, Notch, Insulin and TGFb pathways [25]. In this manner, the response of these epithelia to the manipulation of Krz levels using genetic variants is a key diagnostic to identify the functional requirements of this protein in signalling during imaginal development. Surprisingly, considering the key roles identified for vertebrate non-visual arrestins, we find that the complete elimination of Krz in imaginal cells has no major consequences during imaginal development. Thus, and as claimed previously [15], krz mutant flies are morphologically normal. In contrast, the over-expression of Krz in the wing causes a phenotype identical to the loss of Hedgehog signalling. We find that excess of Krz inhibits Hh signalling by promoting Smo internalization and degradation in a clathrin-and proteosomal-dependent manner. Contrary to that observed in vertebrates, the effects of Krz on Smo are independent of Gprk2 activity and of the activation state of the receptor. We suggest that such differences in Hh signalling are based in the strict requirement of the primary cilia, a structure that is not present in fly epidermal cells, for Hh signalling in most vertebrates [8].

Expression of Krz
Krz is the Drosophila homologue of mammalian non-visual barrestins, and has all the molecular features of a canonical barrestin [15,23] ( Figure S1). The expression of krz occurs ubiquitously in all imaginal discs ( Figure 1A and data not shown).
To visualize the accumulation of the Krz protein, we generated an antibody against Krz. We found that the protein is localized in the cytoplasm of imaginal cells, being detected at higher levels close to the apical side of the epithelium ( Figure 1B-1C, and 1D-1E, red). krz mRNA is also expressed in early blastoderms and during embryogenesis, mostly in the central nervous system and gut during stages 12-17, as assessed both by mRNA ( Figure 1F-1H) and Krz protein ( Figure 1I-1K) expression. The specificity of the antibody was confirmed by analysing the expression of Krz in lossof-function conditions. Thus, Krz staining is lost in dorsal wing compartments expressing a krz interference RNA construct ( Figure 1M-1M9), and is also absent in clones of cells homozygous for of a krz genetic deficiency ( Figure 1L, 1N-1O9). The subcellular localization of the protein in wing discs over-expressing wild type Krz or different mutant forms (described below) is also in the cytoplasm, with higher levels at the apical side of the cells ( Figure 1P-1S, red).

Functional requirement of Krz during imaginal development
b-arrestin has widespread functions during mammalian development and cellular homeostasis [2,8]. To identify the functional requirements of Krz during the development of the fly wing, we constructed flies with wings homozygous for a krz deficiency ( Figure 1L). The elimination of Krz in the wing pouch (in 638-Gal4; FRT82 M(3)z/FRT82 Df(3R)krz; UAS-FLP/+ flies) produces a slightly folded wing of smaller than normal size but without any major defects in the pattern of veins or wing margin (Figure 2A-2B). This phenotype was also observed when the expression of krz was reduced in the entire wing blade (UAS-dicer/+; nub-Gal4/UASikrz; Figure 2C). In wing discs of a similar genotype (638-Gal4/ UAS-ikrz), we found a 70% reduction in krz mRNA levels ( Figure  S2). The very modest effects of krz elimination in the wing disc imply that the signalling pathways regulating wing patterning operate normally in the absence of Krz in imaginal cells. We also studied the consequences of krz elimination in the entire larva in Df(3R)krz homozygotes and in the Df(3R)krz/krz 1 combination. These two genotypes survive until the third larval instar, where they became immobile and flaccid and form melanotic tumors, as described for krz 1 homozygotes [16]. The imaginal discs of Df(3R)krz/Df(3R)krz and Df(3R)krz/krz 1 larvae are very reduced in size, and express high levels of activated Caspase 3, indicating massive cell death in all imaginal tissues ( Figure 2E). As expected, krz mRNA is absent in Df(3R)krz homozygous larva ( Figure S2). We rescued the viability of Df(3R)krz/krz 1 larvae by expressing Krz in the central nervous system in UAS-krz/wor-Gal4; Df(3R)krz/krz 1 flies. The surviving flies display slightly folded wings that were very similar to those of flies where Krz levels are eliminated only in the wing ( Figure 2D, compare with 2B and 2C). The corresponding wing discs have a normal appearance and the expression of signalling molecules occurs in the normal domains ( Figure 2F for Smo and data not shown). These data confirm the essential function of Krz in the CNS for larval development [15][16] and indicate that the degeneration of imaginal tissues observed in Df(3R)krz/krz 1 larvae is a consequence of the loss of krz in the CNS. Interestingly, the rescued flies, both males and females, albeit morphologically normal are sterile (data not shown), suggesting that krz is required during germ cell development.

Krz requirements for EGFR, Notch, and Smoothened signalling
The function of b-arrestin 2 [26][27][28] and Krz [23][24] has been related to the regulation of EGFR, Notch or Smo signalling in different experimental systems (reviewed in [8]). Therefore, and despite of the lack of a krz mutant phenotype suggestive of an alteration in any of these signalling pathways, we searched for genetic interactions between loss of krz and genetic variants affecting the efficiency of signalling by the Notch, EGFR or Hh

Author Summary
Non-visual b-arrestins are key proteins involved in plasma membrane receptor internalization, recycling, and signalling. The activity of b-arrestins is generally linked to seventransmembrane receptors, but in vertebrates they can also participate in many other signalling pathways. Consistently, b-arrestins play important roles during vertebrate development and are implicated in a variety of human pathologies. Here we take advantage of the fruit fly model to analyse the genetic requirements of the unique fly nonvisual b-arrestin (kurtz) in signalling during the development of imaginal epithelia. To our surprise, we find that the complete elimination of kurtz has no major consequences in imaginal cells. Our data suggest that insect epithelial cells have evolved arrestin-independent mechanisms to control receptor turnover and signalling, so arrestin function has become less critical. On the other hand, in contrast to previous reports in vertebrates, we find that the over-expression of Kurtz blocks Hedgehog signalling by promoting the internalization and degradation of the transductor Smoothened. We suggest that such differences are based on the specific requirement of the primary cilia for Hedgehog signalling in most vertebrates. These results could provide important insights into divergent modes of membrane receptor regulation and Hedgehog signalling in vertebrates and invertebrates.
pathways. We used the 638-Gal4/UAS-ikrz genotype as a background condition in which the levels of Notch, EGFR or Hh components were reduced. In these combinations, we only observed a genetic interaction between Notch and Krz, as the reduction of krz increases the phenotype of wing margin loss and thicker veins caused by loss of Notch ( Figure 2G-2H). No effects of loss of krz were detected upon a reduction of EGFR ( Figure 2I-2J) or Hh signalling ( Figure 2K-2L).
Since the activities of b-arrestins have been mostly linked to mechanisms of receptor turnover and activation, we next studied the expression and subcellular localization of the EGFR, Notch and Smoothened proteins in krz mutant cells, as a second approach to identify Krz roles during imaginal development. We first looked at the expression of EGFR, Notch and Smo proteins in wing imaginal discs where krz levels are reduced only in the dorsal compartment (ap-Gal4/+; UAS-ikrz/+), because in these discs we can compare the dorsal (mutant) with the ventral (control) compartments in the same disc ( Figure 3A-3D, 3F-3G and 3I-3I9). Only in the case of Notch did we observe a slight relative increase in the accumulation of Notch at the apical side of dorsal cells ( Figure 3A-3D). No differences were detected in the expression of EGFR or Smo in dorsal versus ventral cells in ap-Gal4/+; UAS-ikrz/+ wing discs ( Figure 3F-3G and 3I-3I9, respectively). To extend these results to krz null mutant cells, we generated clones of cells homozygous for the krz deficiency (Df(3R)krz) and for the krz 1 allele in the wing imaginal disc. In both cases the clones were grown at 25uC and at 29uC, to determinate whether there are temperature-dependent effects of the loss of Krz, as described for Gprk2 [21]. The expression of Smo and EGFR in krz null cells is normal, and the subcellular localization of these receptors remains as in wild type cells ( Figure 3H-3H9 and 3J-3J9). In the case of Notch, we could only observe a subtle increase of Notch accumulation in the apical  Figure 3E-3E9). We noticed that some of the clones contained cells shorter than wild type cells ( Figure 3E-3F). In these cases, the maximal expression of Notch is detected at a different focal plane of the epithelium because of the shortening of the cells in the apico-basal axis ( Figure 3E-3E9, Figure S3 and Figure S4). Otherwise no major changes in Notch accumulation were observed in transversal sections of the disc ( Figure 3E-3E9, Figure S3 and Figure S4). In summary, the analysis of krz loss-of-function conditions uncovered a modest requirement of this gene for Notch signalling, which was only observed upon a reduction of Notch levels of expression, and a variable effect of loss of krz on Notch protein levels. These results are in contrast to the key requirements of vertebrate Krz homologs in several signalling pathways described in cell cultures and in vivo systems (reviewed in [8]).
In Drosophila, besides the visual arrestins that are only expressed in the eye, there is another gene (CG32683) encoding a protein structurally related to Krz. It is unlikely that CG32683 is providing an arrestin-like function in the absence of Krz, because this protein lacks several conserved aminoacid motifs present in all members of the arrestin family ( Figure S1). Furthermore, the expression of CG32683 is only observed during embryonic development (data not shown), and no transcripts are detected in the wing imaginal disc by RT-qPCR or in situ hybridization ( Figure S2). As expected, the expression of interference RNA directed against CG32683, either alone or in combination with ikrz, does not cause any alteration in the wing (data not shown). Finally, the over-expression of CG32683-FLAG in the wing does not cause any mutant phenotype, although in this background (sal EPv -Gal4/UAS-CG32383-FLAG) the protein is present at high levels in a pattern similar to that of Krz ( Figure S2). In this manner, we conclude that CG32683 does not have any role during imaginal development, and that it cannot substitute for Krz in the absence of this gene.

Krz downregulates Smoothened signalling
The over-expression of b-arrestin is often sufficient to promote internalization of its agonist-activated GPCRs [28], [29]. In this manner, increasing the levels of Krz might reveal other activities of the protein not uncovered by the loss-of-function approach. To this purpose, we made several constructs to express under the UAS promoter the complete krz cDNA or different modified forms of the protein. When Krz is over-expressed in the wing blade (638-Gal4/+; UAS-krz/+), we obtained a variable phenotype of reduced wing size and changes in the pattern of veins ( Figure 4A-4D and Figure S5). The strength of the phenotype depended on the transgenic UAS-krz line used in these combinations, and in the most severe cases the wing was very reduced in size and all longitudinal veins failed to differentiate ( Figure 4D and Figure S5). These combinations were raised at 29uC, as we did not find any major pattern defects at 25uC. This is likely a consequence of insufficient levels of ectopic expression, because when we used two copies of the UAS-krz construct at 25uC we obtained similar results than with one copy of the UAS-krz at 29uC ( Figure S6B-S6D). The most obvious phenotype of gain of Krz expression is the reduction of the L3/L4 intervein territory. This phenotype is caused by Krz over-expression in the anterior compartment, because when Krz is over-expressed only in anterior cells located in the central domain of the wing blade (dpp-Gal4/+; UAS-krz/+) the fusion of the L3 and L4 veins is also observed ( Figure 4E). These veins and the L3/L4 intervein correspond to the territory specified by Hh signalling [30]. In fact, the observed phenotypes are very similar to those resulting from loss of Hh signalling, detected when, for example, Smo or hh expression is reduced or when Costal2 or Patched (Ptc)  To demonstrate that increasing the level of Krz diminishes Hh signalling, we analyzed the expression of several Hh-target genes, such as Ptc and Engrailed (En), in Krz over-expression conditions. We found that the expression of Ptc and En is strongly reduced or absent in anterior cells of wing discs over-expressing Krz in the entire wing blade (638-Gal4/+, UAS-krz/+; Figure 4J-4K and 4N-4O), the dorsal compartment (ap-Gal4/+; UAS-krz/+; Figure 4L and 4P) or in clones of cells (hsFLP1.22; act,FRT.Gal4; UAS-GFP/UAS-krz; Figure 4M and 4Q). All together, these results indicate that Krz has the potential to antagonize Hh signalling, although this antagonism is only observed upon its overexpression.  [31][32]. We confirmed that the expression of Smo in the membrane of posterior compartment cells and at the A/P boundary is very much reduced in wing discs over-expressing Krz in the entire wing (638-Gal4/+; UAS-krz/+; Figure 5A-5B, see also [21]). The loss of Smo is also observed in clones of cells over-expressing Krz (hsFLP1.22; act,FRT.Gal4; UAS-GFP/UAS-krz; Figure 5C-5C9) and in dorsal cells of ap-Gal4/+; UAS-krz/+ genotype ( Figure 5D). The resulting levels of Smo in Krz over-expressing cells are very similar to those of the anterior compartment, suggesting that Krz promotes Smo elimination. Interestingly, loss of Smo is also observed in wing discs raised at 25uC, even though the corresponding adult wings are almost normal ( Figure S6 and [21]). These results suggest that Krz mostly promotes Smo elimination, and that only above a certain  We identified in Krz several conserved aminoacids (Val94, Leu440/IsoLeu441/Leu443 and Ser427; see Figure S1) whose human counterparts are implicated in the targeting of GPCR to clathrin-coated pits without affecting receptor signalling (Val94) [33][34], in the binding of b-arrestin to clathrin (Leu440/ IsoLeu441/Leu443) [34][35], or that reduce its ability to promote internalization of the b2-adrenergic receptor (Ser427) [36][37]. To explore whether the function of these residues is conserved in Krz, we made several constructs with mutant forms of Krz fused to the Flag tag (UAS-krz V94D -Flag, UAS-krz S427D -Flag, and UAS-krz Leu -Flag). As a control construct we used wild type Krz fused to Flag  Figure 7J and 7N and 7Q-7R). These results indicate that Krz is able to eliminate Smo independently of its phosphorylation state by the kinases CK1 and PKA. In agreement, the Smo PKA form is also eliminated from the cell surface by the over-expression of Krz ( Figure S7).

Discussion
In this work we have analysed the requirement of krz during the development of the Drosophila wing disc. The wing disc is an epithelial tissue, and its patterning and growth depends on the activity of several conserved signalling pathways [25]. We therefore reasoned that any requirement of Krz in the regulation of these pathways should be uncovered by the phenotype of the complete genetic loss of krz in the disc. Surprisingly, we find that wing discs (and all other imaginal discs) can develop in an almost entirely normal manner in the total absence of Krz function (see also [15]). This finding implies that any role of Krz during normal development is dispensable for the regulation of the signalling pathways operating in the wing disc. We must emphasize that even small changes in the levels or domains of signalling by the Notch, EGFR and Hh/Smo pathways result in very characteristic and distinct phenotypes in the wing, and consequently we have to conclude that these pathways operate normally in the absence of Krz in the discs.
The function of Krz has been linked in imaginal discs with the regulation of Notch protein stability [23] and of MAPK phosphorylation [24]. These conclusions are base on sound biochemical data taken from cell culture experiments, and also on the analysis of genetic interactions evaluating the ability of krz mutations in heterozygosity to modify the phenotypes caused by Notch pathway components and MAPK alleles [23][24]. We also find that krz reduction enhances the phenotype of a Notch loss-offunction condition, but we never found any Notch-related phenotype in krz mutant wings. Furthermore, we only found changes in Notch accumulation in a small fraction of krz 1 and Df(3R)krz mutant clones, in contrast to [23]. In this context, it is interesting to note that we were able to detect a robust accumulation of Notch when krz mutant cells over-express the Notch ligand Delta (data not shown), suggesting that the function of krz becomes critical to promote Notch turnover upon Notch-Delta interactions. In this manner, the implication of our analysis and of previous works is that Krz might be required to optimise some aspects of Notch degradation or MAPK phosphorylation, but that these processes can occur normally in the absence of Krz. It might well be that only upon particular alterations of Notch levels, or in sensitized genetic backgrounds, such as overexpressing a non-dephosphorylable form of MAPK, these finetuning aspects of Krz are manifested in phenotypic modifications. It is unlikely that the paucity of krz requirements during imaginal development was due to functional redundancy with other arrestin proteins, because the only Drosophila candidate, CG32683, is not The lack of a krz mutant phenotype in the discs is also surprising considering the multitude of roles assigned to its vertebrate counterparts in the Wnt, IGF, Notch, Smo and TGFb signalling pathways and in ERK activation promoted by many GPCRs (reviewed in [8]). These roles rely both on the regulation by barrestins of receptor internalization and subcellular localization, and also on their functions as scaffold for a variety of proteins involved in cellular signalling. We have to postulate that insect epithelial cells have evolved arrestin-independent mechanisms to control receptor turnover and signalling, and consequently that arrestin function has become less relevant in these cells. This proposal is compatible with Krz retaining the capability to molecularly interact with similar proteins as its vertebrate counterparts, as Krz possesses both amino-and carboxy-terminal arrestin domains and is 72% similar to the mammalian ß-arrestin 2 and 74% similar to ß-arrestin 1 [15].

Implications of Krz in Smo biology and signalling
In contrast to the loss-of-function analysis of krz, the study of its over-expression offers clear-cut indications of its implication in The ability of Krz to interact with Smo in the Drosophila wing is very specific, as we did not observe any other alterations in the localization and activity of other receptors, such as Notch or EGFR. In this context, it is intriguing that the function of vertebrate b-arrestins has also been linked to Smo signalling in several experimental settings. First, b-arrestin 2 promotes Smo signalling by translocating this protein to the primary cilium in mouse NIH-3T3 cells [28,40]. Second, b-arrestin 2 promotes, upon GRK phosphorylation, the internalization of activated Smo in human embryonic kidney 293 cells [26]. Finally, b-arrestin 2 promotes Smo signalling in zebrafish embryos, and this seems to be a physiological function because it is detected in loss-of-function One of the main differences in the Smo signalling pathway between vertebrates and Drosophila is the localization in vertebrates of active Smo to the primary cilium, a structure that is only present in the fly in sensory neurons [41][42]. We can only speculate that the necessity to translocate Smo complexes associated with the type II kinesin motor Kif3A to the cilium [28], a structure not present in fly epidermal cells, imposes a requirement for b-arrestins that is not observed in the fly. Nonetheless, our results show that the capability of Krz to interact with Smo is retained in Drosophila, and this is revealed upon the over-expression of Krz. Once Krz is bound to Smo it would trigger the formation of clathrin-coated pits that targets Smo for degradation in the proteasome, leading to the insufficiency of Hh signalling we observe. In this way, we propose that Krz has retained some of the molecular targets typical of vertebrate barrestins, but that these interactions might not occur at physiological levels of expression, or being redundant with other mechanisms of receptor trafficking and signalling.

Generation of a krz deficiency (Df(3R)krz)
We used the Exelixis insertions e03507 and e00739, which are separated by 3.8 Kb of DNA including krz and the 59 untranslated end of modulo (mod). Flipase (FLP)-induced recombination was induced by a daily 1 hour heat shock at 37uC to the progeny of hsFLP1.22/+; e03507/e00739 females and w; TM2/TM6b males. Thirty putative w; e03507-e00739/TM2 offspring males were individually crossed to w; TM2/TM6b females and after 3 days were used to extract genomic DNA to determinate by PCR the existence of FLP recombination. The position of the Exelixis flanking insertions e03507 and e00739 and the extent of the krz deficiency are described in Figure 1L. Generation of krz and CG32683 constructs UAS-ikrz and UAS-iCG32683. The EST LD31082 was used as a template to amplify a 515 pb krz fragment using the following primers: 59GCGCTCTAGAGCAAATAATAAGGATAAA39 and 59GCGCTCTAGAGCATGCGCCGAAAATAATAGTAGT39. The EST RH70434 was used as a template to amplify a 697 pb CG32683 fragment using the following primers: 59GCG-CTCTAGAGCATGCAGCCAGTAAACCCACAGA39 and 59G-CGCTCTAGACCAAAATCGGAGAGAAAG39. In both cases, the amplified fragment was digested with the restriction enzyme XbaI (underlined sequence in the primers) and cloned into the pWIZ vector previously digested with AvrII. The resulting plasmid was digested with NheI to clone the krz PCR fragment digested with XbaI. The orientation of both XbaI fragments cloned into pWIZ was checked to confirm an inverted position.

Generation of FLIP recombination clones
UAS-krz. An EcoRV-XhoI fragment purified from the EST LD31082 was first cloned into pBluescriptKS, and a NotI-KpnI fragment from this construct was cloned in the pUASt vector.
UAS-CG32683-FLAG. The EST RH70434 was used as a template to amplify the coding sequence of CG32683 using the following primers: 59CACCATGTCGGACAAGCAGCAGGAA-AAGG39 (forward) and 59CACATTCGATGACTTGGGGA-CT39 (reverse). This PCR product was directionally subcloned into pENTR/D-TOPO (Invitrogen). We used the LR Clonase II reaction of CG32683-pENTR/D-TOPO and pTWF (3XFlag-tag at the C-terminal) to generate CG32683 C-terminal-Flag-tagged fusion protein following the instructions from Invitrogen.

RNA isolation and quantitative real-time RT-PCR
Total RNA was prepared from a pool of 50 wing imaginal discs (both wild type and 638-Gal4/UAS-ikrz) and a pool of 30 larvae (both wild type and homozygous Df(3R)krz) using the TRIzol reagent protocol following Life Technologies (Grand Island, NY) instructions. Total RNA (0.7 mg) was used for a first round of reverse transcription employing the Gene Amp RNA PCR kit (Applied Biosystems). Quantitative PCR analysis was performed in a APRI PRISM 7900HT SDS (Applied Biosystems) using the TaqMan probes from Universal Probe Library (Roche) for Krz and CG32683. To normalize the results of the qPCR in the ikrz and CG32683 experiments we used probes for the genes Act42A, Tub84A and RPL32 and to normalize the results of the qPCR in the Df(3R)krz experiment we used a RNApol-II probe. Three independent experiments were done and the quantification of cDNA reduction was performed using Student's t-test. A p-value#0.05 was considered to be statistically significant.

Generation of a Krz antiserum
Protein expression and purification. Fusion protein containing aminoacids 125-470 of Krz was generated using pBluescript-Krz as template and the following primer pair: 59GGGGATCCATTAAAAAGCTGGGGCCG 39 (foward primer) and 59CCCGAATTCTAGGCTC TGTTTCAG39 (reverse primer), containing BamHI and EcoRI restriction sites respectively (underlined sequence). The amplified fragment was digested with the restriction enzymes BamHI and EcoRI, cloned in the BamHI-EcoRI site of the gluthatione-S-transferase (GST) gene fusion vector pGEX-2T (promega) vector and transformed in E.coli BL21 DE3. Selected clones were verified by sequencing.
Antibody generation. After induction, The GST-Krz 125-470 protein was purified using the Profinia Protein Purification System (BioRad), and used for antibody generation in guinea pig following conventional procedures.

Inmunohistochemistry
We used rabbit anti-activated Cas3 (Cell signalling) and anti-panArrestin (BD transduction) and rat anti-EGFR (a gift from B. Shilo), we also utilized anti-En, anti-Ptc, anti-Smo anti-N intra and anti-FasIII mouse monoclonal antibodies from the Hybridoma Bank at University of Iowa (Iowa City, IA) and anti-FlagM2 mouse from SIGMA. Secondary antibodies (used at 1:200 dilution) were from Jackson ImmunoResearch (West Grove, PA). To stain the nuclei we used TOPRO (Invitrogen). Imaginal wing discs and embryos were dissected, fixed, and stained as described in [44]. Confocal images were taken in a LSM510 confocal microscope (Zeiss). In situ hybridization with krz and CG32683 RNA probes were carried out as described [44]. We used the ESTs LD31082 and RH70434 as templates to synthesize krz and CG32683 probes, respectively.

Cell culture and experimental treatments
Culture conditions. S2 cells were cultured at 25uC or lower temperature in Insect-XPress media (BioWhittaker) supplemented with 10% FCS, 100 units/ml penicillin, and 100 mg/ml streptomycin. Stocks were splitted every 3 days.
Stable cell lines generation. Transfection was carried out using the Cellfectin Reagent Kit (Invitrogen). The HhN inducible vector was provided by Stephen M. Cohen (European Molecular Biology Laboratory, Heidelberg, Germany). A Myc-Smo inducible vector was generated by cloning 6Myc-tagged smo (a gift from Jianhag Jia) in pRmHa-puro. Stable cell lines over-expressing either HhN or Myc-Smo were generated by puromycin treatment (25 mg/ml). Hh S2-conditioned medium was obtained by incubation with 0.7 mM CuSO 4 for 24-36 h as described [20].
Immunoprecipitation. 10 ml of cellular lysate was used to assess protein expression levels. The immunoprecipitation reactions were performed by incubating the cellular lysates with 1 mg/ml BSA and specific antibodies for Myc-Smo (15 ml of anti-Myc agarose, Santa Cruz Biotechnology) at 4uC for 4-16 h, followed by incubation with protein-A sepharose when the Krz antibody was used. A pre-immune serum (diluted 1:100) was used as a negative control.
Immunoblotting. Whole cell lysates or immunoprecipitated complexes were resolved by 6-7% SDS-PAGE and proteins transferred to nitrocellulose membranes using a wet-blotting apparatus (BioRad). Smo and Krz proteins were detected by incubating with anti-Myc monoclonal antibody (c-Myc 9E10, Santa Cruz Biotechnology) and anti-Krz guinea pig serum, respectively. Blots were also analyzed with anti-Tubulin (Sigma) as a loading control. Immunoblots were developed and quantified using IR680 and 800 labelled antibodies (Licor) with the Odyssey Infrared Imaging System (Li-Cor). When required, the amount of co-immunoprecipitated protein was normalized by the amount of the IgG protein, as assessed by specific antibodies. When the whole-cell lysates were used, the level of Myc-Smo protein was normalized with the amount of tubulin. Data are expressed as a mean value 6 SEM. Specific measurements were compared using Student's t-test. A p-value#0.05 was considered to be statistically significant. Figure S1 Amino acid sequence alignment of ß-arrestin 1 (human, accession number NP001032), ß-arrestin 2 (human, accession number NP001304), Krz (Drosophila, accession number AF221066) and CG32683 (Drosophila, accession number AAF46586). The putative Valine that prevents GPCR targeting to clathrin-coated pits (Fergurson et al 1996, science) is indicated by green asterisks. Those residues that could be phosphorylated by ERK (present in both ß-arrestin 1 and Drosophila CG32683 protein) are underlined in green. The residues that bind AP2 are boxed and the amino acid necessary for clathrin binding are underlined in red.