Auxilin is essential for Delta signaling

Notch signaling is used universally in animals and in nearly every developmental context (Bray,2006). In the signaling cells, transmembrane ligand endocytosis is necessary for signaling (Le Borgne et al.,2005; Le Borgne,2006; Chitnis,2006; Nichols et al.,2007a). In Drosophila, ligand endocytosis and signaling depend on Epsin (Overstreet et al.,2004; Wang and Struhl,2004; Wang and Struhl,2005), an endocytic factor first identified in vertebrates that functions in at the internalization step of endocytosis(Chen et al., 1998). Drosophila Epsin is encoded by the liquid facets(lqf) gene (Cadavid et al.,2000).

Recently, the endocytic protein Auxilin has been implicated in Notch signaling in Drosophila (Hagedorn et al., 2006; Eun et al.,2007). Auxilin is an adapter for the ATPase Hsc70 (Heat shock cognate 70); Auxilin binds Hsc70 with its J domain and also binds Clathrin cages, which are Hsc70 substrates. Auxilin catalyzes the binding of Hsc70 to Clathrin cages and stimulates its ATPase activity(Lemmon, 2001; Eisenberg and Greene, 2007). The evidence is most solid for two direct functions of Auxilin, one in internalization and a second in uncoating Clathrin-coated vesicles. At the internalization step, Hsc70/Auxilin exchanges Clathrin on cages forming at the plasma membrane, which might help to constrict vesicles prior to scission. After scission, Hsc70/Auxilin removes the Clathrin cages, allowing vesicle fusion with endosomes (Lemmon,2001; Eisenberg and Greene,2007).

Here, we show that Drosophila Auxilin is crucial in signaling cells for Delta internalization and signaling. In addition, we show that the requirement for Auxilin is, at least in part, to maintain the pool of free Clathrin. We infer that Clathrin has a key role in the signaling cells. We discuss how these observations impact current thinking on the roles of Epsin and endocytosis in the signaling cells.

From the Bloomington Stock Center: FRT42D (FBti0002072); ubi-nGFP FRT40A/CyO (FBti0015576, FBti0002071); CyO, YFP(FBti0058605); w, eyFLP (FBti0015982); Act5C-gal4/CyO(FBti0012293); Dl-lacZ (FBti0012268); and Chc¹(FBal0033511).

From our laboratory: w¹¹¹⁸; w; gaux+/CyO(Eun et al., 2007); aux⁷²⁷, aux^D136, aux^K47,aux^D128 (Eun et al.,2007); lqf^FDD9 (FBal0104483); UAS-aux+ (FBal0190739); UAS-GFP-Clc (FBtp0017428); and N^ts1 (FBal0012887).

Generated in this work: w; FRT42D, gaux+, ubi-GFP/CyO, YFP;aux⁷²⁷/TM6B; w; Tub-aux+, ubi-GFP, FRT40A/CyO, YFP;aux⁷²⁷/TM6B; w, eyFLP; FRT40A/CyO, YFP;aux^D128/TM6B; w, eyFLP; FRT42D/CyO, YFP;aux^D136/TM6B; w, eyFLP; FRT42D/CyO, YFP;aux^D136 Dl-lacZ/TM6B; w; ro-aux+/CyO;aux^K47/TM6B; w; Act5C-gal4/CyO;aux⁷²⁷/TM6B; w; Act5C-gal4/CyO;aux^K47/TM6B; w; UAS-aux^CBD+J/CyO;aux^D128/TM6B; w; UAS-aux^K+PTEN/CyO;aux^D128/TM6B; w;UAS-aux^ΔJ/CyO;aux^D128/TM6B; w; gaux^CBD+J/CyO;aux^D128/TM6B; w; gChc+/CyO; aux^K47/TM6B; w; gChc+/CyO; aux^D128/TM6B; and w; gChc+/CyO;lqf^FDD9/TM6B.

Molecular biology was performed using standard procedures. Cloning enzymes and standard oligonucleotides were from New England Biolabs and Roche, custom oligonucleotides were from IDT, and DNA purification kits were from Qiagen. The DNA sequences of all PCR products were verified.

An aux+ cDNA in pOT2 (GH26574 from the Drosophila Genomics Resource Center, Bloomington, IN) was purified as an XhoI-EcoRI fragment and ligated into those sites of pBSKSII+ (Stratagene) to generate pBS-aux+. The EcoRI site in pBS-aux+ was changed to AscI by ligating annealed oligonucleotides of the sequence 5′-AATTGGCGCGCC-3′ into the EcoRI site. Subsequently, the XhoI site in the resulting plasmid was changed to AscI by ligating annealed oligonucleotides of the sequence 5′-TCGAGGCGCGCC-3′ into the XhoI site. The 4.4 kb AscI fragment containing the aux+ cDNA was ligated into the pRo vector (Huang and Fischer-Vize, 1996) and the correct orientation determined by SalI digestion.

The aux+ cDNA coding region is 3498 nt long, including start and stop codons, and it encodes a protein of 1165 amino acids (aa), including the Met initiator codon.

This construct contains nt 2071-3498 of the aux+ cDNA coding region, corresponding to aa N691-A1165. A 1.4 kb region of the aux+coding region was amplified by PCR (Platinum PCR Supermix, Invitrogen) using pBS-aux+ as the template and the primer pair 5′-GGAATTCATGAACCAGGACTTGGATGATCTG-3′ and 5′-GTCTAGATTACGCATTAAACATATTTTGCTG-3′. The PCR product was subcloned into pGEM-T Easy (Promega). A 1.4 kb EcoRI-XhoI fragment containing aux^CBD+J was purified and ligated into pUAS_t (Brand and Perrimon, 1993) restricted with EcoRI and XhoI.

This construct contains nt 1-2103 of the aux+ cDNA coding region,corresponding to aa M1-E701. A 2.1 kb AscI fragment containing aux^K+PTEN was generated by PCR using pBS-aux+ as a template, the primers 5′-AAAGGCGCGCCATGGGCGAGTTCTTTAAGTCGCTC-3′and 5′-AAAGGCGCGCCTCACTCAGGGTCTGGCAGATCATCCA-3′, and Pfu Turbo Polymerase (Stratagene). The amplified product was gel-purified, restricted with AscI, and ligated into AscI-restricted pUAS_t.

This construct contains nt 1-3291 of the aux+ cDNA coding region,corresponding to aa M1-D1097. A 3.3 kb fragment containing aux^ΔJ was generated by PCR using pBS-aux+ as template, the primers 5′-AAAGGCGCGCCATGGGCGAGTTCTTTAAGTCGCTC-3′ and 5′-GGCTAGCTTAATGTCCCACAACACTGTGTGCATAG-3′, and Herculase DNA Polymerase (Stratagene). The product was ligated into pGEMT-Easy, re-isolated as an AscI-XbaI fragment, and ligated into pUAS_t restricted with those enzymes.

This construct is the 21 kb genomic DNA fragment in pgaux+(Eun et al., 2007), subcloned as an NheI fragment into the XbaI site of pCasper3(Thummel and Pirrotta, 1992)to generate pCasper3-gaux+, and with a deletion of 4101 bp that includes the kinase and PTEN domains, and an addition of two aa (Leu-Glu),corresponding to an XhoI site (5′-CTCGAG-3′) just downstream of the start codon. This was achieved by replacing an internal 10.7 kb KpnI fragment with a 6.6 kb KpnI fragment, which was generated by joining three PCR products (Herculase), a 5′ product, a middle product, and a 3′ product that generates the deletion. The 984 bp 3′ product was amplified using pCasper3-gaux+ as template and the primers 5′-ACTCGAGAACCAGGACTTGGATGATCTGCCAG-3′ and 5′-GGTACCGCCTGTGGCTGCG-3′, ligated into pGEM-T and re-isolated as an XhoI-KpnI fragment and ligated into pBSIIrestricted with the same enzymes. The middle fragment (2852 bp) was generated the same way using the primers 5′-AGATATCTCCGGATGGGCAGACACGAA-3′and 5′-ACTCGAGCATTTTGGTGGTGGCCAATGCTA-3′, isolated as an EcoRV-XhoI fragment and ligated into those two sites in pBSII containing the 3′ fragment. The 5′ fragment (2789 bp) was generated the same way using the primers 5′-AGCGGCCGCGGTACCGCCCGAGCCCG-3′ and 5′-TCCGGATGTTGCAAACTTTCCAA-3′, isolated as a NotI-BspEI fragment, and ligated into those two sites in pBSII containing the 3′ and middle fragments. Finally, a 6.6 kb KpnI fragment was isolated from the resulting pBSII plasmid and ligated into pCasper3-gaux+ restricted with KpnI.

A 14.3 kb AvrII-SacII fragment of BAC22H11 (BACPAC Resources, Oakland, CA) containing Chc+ genomic DNA was ligated into pCMC105 (a gift of C.-M. Chen and G. Struhl, Columbia University, New York, NY) restricted with AvrII and SacII. The resulting plasmid was restricted with AvrII, and a 4.5 kb NheI-XbaI fragment of pAT806 (a gift of K. Basler and G. Struhl) containing a w+ marker gene was ligated in.

An αTub84B (Tub) promoter fragment was excised from pKB700 (a gift of K. Basler and G. Struhl) as a 2.6 kb NotI-KpnI fragment and ligated into pBSKSIIrestricted with the same enzymes. The NotI site in the resulting plasmid was changed to AvrII by inserting the linker 5′-GGCCCCTAGG-3′. The Tub promoter was excised as an AvrII-KpnI(3′ overhang removed) fragment. A 4.4 kb EcoRI-XhoI fragment containing the aux+ cDNA was excised from clone GH26574 and ligated into pBSKSIIrestricted with the same enzymes. The XhoI site of the resulting plasmid was changed to NheI using the linker 5′-TCGAGCTAGC-3′, the EcoRI site was changed to EcoRV using the linker 5′-AATTCGATATCG-3′, and the NotI site was changed to AvrII using the linker 5′-GGCCCCTAGG-3′. The Tub promoter fragment was ligated into the resulting plasmid restricted with EcoRV and AvrII,and an AvrII-NheI fragment containing Tub-aux+ was excised. A 1.2 kb SpeI-XbaI fragment containing transcription termination signals was excised from pAT806 and ligated into pCasper4 (Thummel and Pirrotta, 1992) restricted with SpeI. A plasmid with the termination signal in the appropriate orientation was isolated, restricted with SpeI, and the Tub-aux+ ligated in.

P-element transformation of w¹¹¹⁸ flies was performed according to standard methods in our laboratory or by Genetic Services (Sudbury, MA).

Antibodies were generated to Auxilin amino acids Q702-A1165. The corresponding DNA sequence was amplified by PCR using primers 5′-GGATCCCAGGTGACACCTCGGTTCTGCG-3′ and 5′-GAATTCCGCATTAAACATATTTTGCTGCGTG-3′, with plasmid GH26574 as a template, and subcloned into pGEMT. A BamHI-EcoRI fragment containing the aux coding sequences was ligated into pRSET (Invitrogen) restricted with the same enzymes, which added six His codons at the beginning of the open reading frame. Protein expression from pRSET-aux was in BL21-CodonPlus-RIL cells(Invitrogen) induced with 0.1 mM IPTG. The 6xHis-Aux protein was purified from sonicated cells using HisBind (Novagen) and used to raise antibodies in rats(Pocono Rabbit Farm and Laboratory, Canadensis, PA). Prior to use for immunostaining, the antisera were preadsorbed with fixed w¹¹¹⁸Drosophila embryos.

Third instar larval eye discs were fixed in PEMS and antibody incubations and washes were in PBST (see Fischer-Vize et al., 1992). Primary antibodies used were: mAb323 (1:2) obtained from Sarah Bray (University of Cambridge, UK); mAb202 (1:10) from Developmental Studies Hybridoma Bank (DSHB), Iowa City, IA; rat anti-Elav(9:1) from DSHB; and anti-Aux (1:500). Secondary antibodies (Molecular Probes)were AlexaFluor 568 anti-mouse and AlexaFluor 633 anti-rat (1:300). AlexaFluor 568-phalloidin (Molecular Probes) was dried and resuspended in PBST at 0.1 unit/l. Eye discs were mounted in Vectashield (Vector) and viewed with a Leica TCSSP2 confocal microscope. Plastic sectioning of adult eyes was as described(Tomlinson and Ready, 1987),and wings were dehydrated in 70% ethanol and mounted in DPX (Fluka). Wing and eye sections were photographed with a Zeiss Axioplan and Axiocam HRc. Images were processed with Adobe Photoshop.

Throughout this work, we refer to aux⁷²⁷/aux^D128 or aux⁷²⁷/aux^D136 heterozygous cells or animals as aux-. [The aux alleles were analyzed previously(Eun et al., 2007).] The nonsense mutant aux^D136 is likely to be null because the truncated protein it may produce would lack the J domain(Fig. 1A), which we show below is necessary for Auxilin function. aux⁷²⁷ and aux^D128 are also likely null alleles because of their early nonsense mutations (Fig. 1A), and because their phenotypes in trans to each other or aux^D136 are indistinguishable. Also, we show below that the mutant eye phenotype of either aux- genotype is identical to lqf or Delta null mutants. Finally, in eye discs immunolabeled with a polyclonal antibody to the C-terminal 464 amino acids of Auxilin, whereas adjacent aux+ cells showed strong signals, clones of aux⁷²⁷/aux^D128(Fig. 1B,B′) cells showed no signal above background. Nevertheless, there could be some truncated and functional Auxilin protein produced by aux⁷²⁷ or aux^D128 that escaped detection. In aux⁷²⁷ or aux^D128 homozygous larvae,mutant aux mRNA accumulates to levels similar to those seen in wild type (data not shown), and thus through reinitiation of translation downstream of the nonsense mutation, aux⁷²⁷ or aux^D128 potentially could produce truncated Auxilin protein. As we show that the C-terminal region of Auxilin containing the Clathrin-binding and J domains is sufficient for function, truncated Auxilin protein so produced could be functional. We could not test whether or not these alleles produce truncated protein undetectable by immunolabeling because the Auxilin antibody we generated does not detect Auxilin protein on blots of Drosophila wild-type protein extracts.

aux- clones in third instar larval eye discs displayed phenotypes consistent with the failure of three Delta signaling events near the morphogenetic furrow (Fig. 1C). Cells anterior to the furrow require Notch activation to achieve neural potential (proneural enhancement); when Notch activation fails at this point,no neurons develop (Baker,2002). As cells exit the furrow, lateral inhibition results in evenly spaced ommatidial preclusters; failure of Notch activation at this point results in disordered photoreceptors (R-cells)(Baker, 2002). R-cell determination in aux- clones appeared identical to that in Delta or lqf null clones(Baker and Yu, 1996; Overstreet et al., 2004). No R-cells (Elav+) were present in the center of large aux- clones, and aux- R-cells present at clone borders were disorganized(Fig. 1D,D′). Another indicator of lateral inhibition failure is increased Deltatranscription (Baker and Yu,1998; Wang and Struhl,2004). aux- clones showed increased expression of a Delta-lacZ enhancer trap (Fig. 1E,E′) and also increased Delta protein(Fig. 1F,F′). It was reported previously that Delta-lacZ is not upregulated in eye discs(Hagedorn et al., 2006). Perhaps the discrepancy is due to differences in the mutant auxalleles used. A later Notch signaling event is R-cell restriction, in which R2/5 and R3/4 signal precluster cells in the developing ommatidium to limit the number of outer R-cells to six (Fig. 1C) (Overstreet et al.,2004). Like Delta or lqf hypomorphs(Parks et al., 1995; Cadavid et al., 2000), aux hypomorphic eyes contain ommatidia with extra outer R-cells(Hagedorn et al., 2006; Eun et al., 2007).

As is the case for Epsin, the requirement for Auxilin in eye patterning appears limited to Notch signaling. Several different signaling pathways are needed for morphogenetic furrow movement, and Epidermal growth factor receptor signaling is essential for R-cell determination(Lee and Treisman, 2002; Kumar, 2002; Nagaraj et al., 2002). These pathways are operating in aux- clones because furrow movement inside aux- clones resembles that in wild type(Fig. 1F,F′), and cells expressed Elav, a marker of R-cell determination(Fig. 1D,D′). In addition, neither Epsin (Overstreet et al., 2004; Wang and Struhl,2004) nor Auxilin is required for cell viability.

aux- cells become R-cells at clone borders, where they are adjacent to aux+ cells that signal and impart proneural potential(Fig. 1D,D′). This suggests that aux- cells can receive Notch activation signals. Indeed, aux- cells expressed Enhancer of split proteins [E(spl)], a direct response to Notch activation(Jennings et al., 1994), but only when adjacent to aux+ cells at clone borders(Fig. 1G,G′). In addition, in aux hypomorphic eyes, expression of aux+ in R2/5 and R3/4 by a transgene (ro-aux+) rescued the mutant phenotypes of ommatidia with ectopic R-cells: ∼25% of facets in aux^K47/aux^D128 were wild-type, and the addition of ro-aux+ increased the fraction to ∼95% (data not shown). The ro expression construct is remarkably specific for R2/5 and R3/4(Overstreet et al., 2004). Thus, R-cell restriction fails when R2/5 and R3/4 have little aux+activity because they cannot signal to extraneous precluster cells. We conclude that aux- cells in the eye can receive, but not send, Notch signals.

Membrane-localized Delta is almost undetectable in immunolabeled wild-type eye discs (Fig. 2A)(Parks et al., 1995; Baker and Yu, 1998; Overstreet et al., 2003; Overstreet et al., 2004). Rather, nearly all Delta detected is in endocytic vesicles that are probably irrelevant to signaling, as they are observed in lqf- cells that do not signal (Fig. 2C)(Overstreet et al., 2004; Wang and Struhl, 2004). In lqf- clones, lateral inhibition failure results in Delta overexpression (Baker and Yu,1998; Wang and Struhl,2004), which allows detection of Delta on the plasma membrane(Fig. 2C). This is observed also in eye discs that are hypomorphic for both lqf and heterozygous for a neuralized (neur) mutation(Fig. 2D)(Overstreet et al., 2004). neur encodes a Ubiquitin ligase required for Delta ubiquitination and signaling (Le Borgne et al.,2005; Le Borgne,2006). Delta eventually clears from the membrane and is observed again only in intracellular puncta several rows posterior to the furrow(Overstreet et al., 2003; Overstreet et al., 2004). A similar phenomenon is observed with Notch temperature sensitive(N^ts) mutant discs raised to the restrictive temperature for 6 hours (Fig. 2B)(Baker and Yu, 1998). However,in N^ts discs, where lateral inhibition fails owing to a failure in signal reception, Delta clears from the membrane quickly and only a tight band of membrane-associated Delta is apparent at the furrow(Fig. 2B)(Baker and Yu, 1998). By contrast, in lqf- cells, Delta remains on the membrane for a longer time (more rows of cells) posterior to the furrow(Fig. 2C,D)(Overstreet et al., 2003; Overstreet et al., 2004) and we interpret this as inefficient endocytosis. This observation is similar to that in the wing disc, where Delta was overexpressed using a transgene and clones of lqf-cells accumulated more Delta on the membrane than did adjacent wild-type cells (Wang and Struhl,2004). Delta also persisted on the membrane in aux mutant cells (Fig. 2E-E″,F-F″). Also, as in lqf- clones(Overstreet et al., 2004; Wang and Struhl, 2004), Delta puncta were present in aux- cells(Fig. 2F,G). We conclude that like Epsin, Auxilin is required for some, but not all, Delta internalization. Similarly, in yeast and vertebrate cells, aux mutations result in decreased endocytosis (Pishvaee et al.,2000; Umeda et al.,2000).

Drosophila Auxilin, like vertebrate auxilin 2 or GAK (cyclin G-associated kinase), has N-terminal kinase and PTEN (phosphatase and tensin-related) domains (Fig. 3), the functions of which are unclear(Umeda et al., 2000; Lemmon, 2001; Korolchuk and Banting, 2002; Lee et al., 2006; Massol et al., 2006; Eisenberg and Greene, 2007). The auxilins of yeast and C. elegans, which lack kinase and PTEN domains, uncoat clathrin-coated vesicles (CCVs) in vitro and in vivo(Gall et al., 2000; Pishvaee et al., 2000; Greener et al., 2001), and the Clathrin-binding and J domains of GAK are sufficient for uncoating in vivo(Holstein et al., 1996; Greener et al., 2000). The presence of the kinase and PTEN domains in Drosophila Auxilin suggests that the function of Auxilin in Notch signaling might not be limited to Clathrin uncoating or exchange.

To determine whether or not the kinase and PTEN domains or the J domain are required for Notch signaling, we generated transformants with the four different UAS transgenes shown in Fig. 3A. The transgenes were tested in single copy for their ability, when expressed ubiquitously with Act5C-gal4, to complement strong aux- mutants and hypomorphs. We found that Auxilin^FL or Auxilin^CBD+Jcomplemented all aux- mutant phenotypes, and that Auxilin^K+PTEN and Auxilin^ΔJ had no activity(Fig. 3A). Act5C>auxilin^CBD+J might overexpress Auxilin^CBD+J relative to endogenous Auxilin. Because genomic transgenes are likely to be expressed at physiological levels, we also generated a transgene (gaux^CBD+J) that expresses Auxilin^CBD+J in the context of a genomic DNA construct, gaux+ (Eun et al.,2007). A single copy of either gaux+ or gaux^CBD+J complemented all aux- phenotypes(Fig. 3A). These results suggest that the ability of Auxilin^CBD+J to complement aux- is unlikely to require its overexpression.

We conclude that the J domain is necessary and a C-terminal Auxilin fragment including the Clathrin-binding and J domains is sufficient for function in Notch signaling. Thus, the role of Auxilin is likely to depend on its function as a link between Hsc70 and Clathrin cages. Consistent with functions in endocytic vesicle internalization and uncoating, Auxilin was detected apically in cytoplasmic puncta near the plasma membrane(Fig. 3B).

In aux mutants, it would be expected that the levels of free Clathrin are limiting owing to inefficient uncoating of CCVs. In yeast and C. elegans, free clathrin is depleted in aux mutants(Gall et al., 2000; Pishvaee et al., 2000). Consistent with Clathrin depletion in Drosophila aux mutants, we found that a loss-of-function mutation in the Clathrin heavy chaingene, Chc¹, is a strong dominant enhancer of auxhypomorphs: Chc¹/+ flies appeared normal, and aux^K47/aux^D128 adult flies were viable with Notch-like phenotypes, but Chc¹/+;aux^K47/aux^D128 adults could not be obtained. It was reported previously that expression of GFP-Clathrin light chain (Clc) enhances aux phenotypes (Hagedorn et al.,2006), and this result was repeatable in our hands. We do not think, however, that this genetic interaction contradicts the results with Chc mutants. Rather, as GFP-Clc does not colocalize with Chc in the eye disc (data not shown), we propose that GFP-Clc acts as a dominant negative. There are no Clc mutants available to test for complementation by GFP-Clc.

We also tested whether an additional copy of the Chc+ gene(PChc+) suppresses the Notch-like phenotypes of aux- hypomorphs. We found that PChc+ suppresses the wing phenotype completely (Fig. 4A,B) and the eye phenotype significantly(Fig. 4C-F). Similarly, in C. elegans, aux(RNAi) mutant phenotypes are suppressed by overexpression of clathrin heavy chain(Greener et al., 2001). Taken together, these results suggest that Notch pathway phenotypes of auxmutants are due at least in part to Clathrin depletion. Clathrin is likely to be needed in some manner for Delta internalization. Indeed, we have shown that Delta endocytosis is inefficient in aux mutants. Thus, the requirement for Auxilin does not necessarily infer a requirement in the signaling cells for trafficking of Delta-containing endosomes downstream of internalization, as proposed previously(Hagedorn et al., 2006).

A role for Clathrin in Notch signaling cells was originally inferred from the observation that Chc mutants are strong dominant enhancers of lqf hypomorphs (Cadavid et al.,2000). As Epsin has both Ubiquitin- and Clathrin-binding motifs,and also binds the plasma membrane, the simplest scenario imaginable for Clathrin and Epsin function in Delta internalization is for Epsin to act as a Clathrin adapter that recognizes ubiquitinated Delta, and brings Clathrin to the membrane for CCV formation (Wendland,2002). However, in light of evidence that Epsin-dependent endocytosis of ubiquitinated transmembrane proteins such as Delta may not occur through formation of CCVs (Chen and De Camilli, 2005; Sigismund et al., 2005; Aguilar and Wendland, 2005; Hawryluk et al., 2006; Barriere et al.,2006; Madshus,2006), it became unclear how to interpret the Chc/lqfgenetic interaction. The results presented here point to a crucial role for Clathrin in Notch signaling cells. One intriguing possibility is that Delta internalization depends on Clathrin not because Delta is endocytosed in CCVs,but because Clathrin is a positive regulator of Epsin function. More experiments are required to test this idea.

Why do tissues that lack Epsin or Auxilin display Delta-like phenotypes, rather than phenotypes indicating failure of many signaling pathways or even cell death? One possibility is that the apparent specificity of both Epsin and Auxilin might simply reflect the usual redundancy of endocytic protein functions, and an unusual dependence of Notch signaling on efficient endocytosis. Alternatively, a special function of Epsin might be crucial to Notch signaling cells. Two kinds of models have been proposed to explain why Notch signaling requires ligand endocytosis by the signaling cells(Le Borgne et al., 2005; Le Borgne, 2006; Chitnis, 2006; Nichols et al., 2007a). One idea (the `pulling model') is that after receptor binding, ligand endocytosis generates mechanical forces that result in cleavage of the Notch intracellular domain (Notch activation), either by exposing the proteolytic cleavage site on the Notch extracellular domain, or by causing the heterodimeric Notch receptor to dissociate (Parks et al.,2000; Nichols et al.,2007b). Alternatively, ligand internalization prior to receptor binding might be required to process the ligand endosomally, and recycle it back to the plasma membrane in an activated form (the `recycling model')(Le Borgne and Schweisguth,2003; Wang and Struhl,2004; Wang and Struhl,2005; Emery et al.,2005; Jafar-Nejad et al.,2005). Epsin might generate an environment particularly conducive to either pulling or recycling, and Auxilin might be required specifically by Notch signaling cells because it activates Epsin, perhaps by providing free Clathrin. Alternatively, Auxilin might be needed to provide free Clathrin because Delta is internalized through CCVs. In this case, if Auxilin is required in Notch signaling solely to provide free Clathrin, the implication would be that efficient CCV uncoating is not important for generating uncoated Delta-containing vesicles per se, which are prerequisite for travel through an endosomal recycling pathway. Further understanding of the role of Auxilin in Notch signaling cells might be key to understanding the role of ligand endocytosis.

For flies, plasmids and antibodies we thank Nick Baker, Sarah Bray, Helmut Kramer, Ira Mellman, Marc Muskavitch, Gary Struhl and the Bloomington Drosophila Stock Center. We also thank Erin Overstreet for the N^ts image in Fig. 2, Xuanhua Xie for important discussions, and three anonymous reviewers who helped to improve the manuscript greatly. This work was supported by a grant to J.A.F. from the National Institutes of Health,Institute for Child Health and Human Development (RO1-HD30680).