The clathrin-binding motif and the J-domain of Drosophila Auxilin are essential for facilitating Notch ligand endocytosis

Background Ligand endocytosis plays a critical role in regulating the activity of the Notch pathway. The Drosophila homolog of auxilin (dAux), a J-domain-containing protein best known for its role in the disassembly of clathrin coats from clathrin-coated vesicles, has recently been implicated in Notch signaling, although its exact mechanism remains poorly understood. Results To understand the role of auxilin in Notch ligand endocytosis, we have analyzed several point mutations affecting specific domains of dAux. In agreement with previous work, analysis using these stronger dAux alleles shows that dAux is required for several Notch-dependent processes, and its function during Notch signaling is required in the signaling cells. In support of the genetic evidences, the level of Delta appears elevated in dAux deficient cells, suggesting that the endocytosis of Notch ligand is disrupted. Deletion analysis shows that the clathrin-binding motif and the J-domain, when over-expressed, are sufficient for rescuing dAux phenotypes, implying that the recruitment of Hsc70 to clathrin is a critical role for dAux. However, surface labeling experiment shows that, in dAux mutant cells, Delta accumulates at the cell surface. In dAux mutant cells, clathrin appears to form large aggregates, although Delta is not enriched in these aberrant clathrin-positive structures. Conclusion Our data suggest that dAux mutations inhibit Notch ligand internalization at an early step during clathrin-mediated endocytosis, before the disassembly of clathrin-coated vesicles. Further, the inhibition of ligand endocytosis in dAux mutant cells possibly occurs due to depletion of cytosolic pools of clathrin via the formation of clathrin aggregates. Together, our observations argue that ligand endocytosis is critical for Notch signaling and auxilin participates in Notch signaling by facilitating ligand internalization.


Background
The Notch pathway, a highly conserved signaling module, participates in diverse aspects of animal development, including cell proliferation, differentiation, and pattern formation [1,2]. Upon ligand binding, Notch undergoes proteolytic processing, resulting in the release and the nuclear translocation of NICD (Notch intra cellular domain) for transcriptional activation [3]. As the Notch receptor and its ligands are widely expressed, the activities of this important pathway need to be regulated at multiple levels to prevent inappropriate signaling output. Recent evidence from several systems has suggested that ligand endocytosis plays a key role in regulating the activ-ity of this cascade, although its exact function in Notch signaling remains unclear.
Genetic analysis using Drosophila has identified several factors required for ligand endocytosis during Notch signaling. In Drosophila, there are two known Notch ligands, Delta (Dl) and Serrate (Ser), both of which appear to utilize an ubiquitin-mediated endocytic pathway to activate Notch [4][5][6][7][8]. While polyubiquitination is known to facilitate protein degradation, covalent addition of a single ubiquitin moiety to the cytoplasmic tails of membrane proteins can serve as an internalization signal [9,10]. The ubiquitination of Dl and Ser for internalization is mediated by neuralized (neur) and mind bomb (mib1), two structurally unrelated but functionally similar E3 ubiquitin ligases [4,7,[11][12][13][14][15][16]. The subsequent recruitment of ubiquitinated Notch ligands into clathrin-coated vesicles is thought to depend on liquid facets (lqf, the Drosophila homolog of epsin), a cargo-specific clathrin adaptor [17][18][19][20]. Dynamin, a GTPase required for the pinching-off of clathrin-coated vesicles (CCVs) from the plasma membrane, is also required for this event [21,22]. Moreover, during the asymmetric cell division of sensory organ precursors, Rab11 (a GTPase associated with recycling endosomes) and Sec15 (a component of the exocyst complex) have been shown to promote Notch signaling and regulate Dl trafficking [23,24].
Mutations in Drosophila auxilin (dAux), a J-domain-containing regulator in clathrin-mediated transport, were recently shown to disrupt Notch-dependent processes [25][26][27]. Although originally identified as a factor promoting the assembly of clathrin cage from free triskelia [28], efforts investigating the role of auxilin during clathrinmediated endocytosis (CME) have mostly focused on its cooperation with Hsc70 in mediating the disassembly of clathrin coats form nascent CCVs [29]. In mammals, there are two different auxilin-like molecules (auxilin 1 and GAK/auxilin 2), differing in their tissue distributions and the presence of a N-terminal Ark family kinase domain [29][30][31]. In contrast, the Drosophila genome contains only one auxilin ortholog, which is structurally more similar to GAK, as it contains a N-terminal kinase domain, followed by a PTEN (phosphatase and tensin) homologous region, a clathrin-binding domain, and a C-terminal dnaJ domain ( Figure 1A) [32]. Biochemical data suggest that the CBM (clathrin-binding motif) and J-domain recruit ATP-bound Hsc70 to CCVs and stimulate the ATPase activity of Hsc70, which drive the disassembly of clathrin coats [30,33,34]. The PTEN-related region appears to be dAux mutations cause defects in photoreceptor specification (B-E) Fluorescent micrographs of larval eye discs from (B, C) Act5C>FLP/+; FRT 5-5Z3515 , dAux L78H /FRT 5-5Z3515 , ubi-GFP nls and (D, E) Act5C>FLP/+; FRT 5-5Z3515 , dAux F956 */FRT 5-5Z3515 , ubi-GFP nls animals. The mosaic discs are stained for Elav (red), which labels the nuclei of neuronal cells, and mutant tissues are marked by the absence of nuclear-GFP expression (green). (C, E) Boxed areas of dAux L78H and dAux F956 * clones from B and D are shown at a higher magnification. In all the panels, anterior is to the left. Scale bar, 50 µm.

D E
critical for the membrane recruitment of auxilin to clathrin-rich regions [35,36]. The N-terminal kinase is known to phosphorylate subunits of AP1 and AP2 adaptor complexes, possibly modulating the affinity of these adaptors for cargoes [31,[37][38][39][40]. However, as not all auxilin homologs (for example, yeast swa-2 and C. elegans auxilin) contain the kinase and the PTEN-related region [41][42][43], the physiological relevance of these domains remains to be determined.
Initially, the identification of auxilin as a component of the Drosophila Notch pathway suggests that, to activate Notch, ligand endocytosis needs to proceed past the step of clathrin cage disassembly. This implies that the relevance of ligand endocytosis is not the internalization per se, but a transit of ligand through subsequent endocytic compartments. However, this interpretation hinges on the assumption that auxilin participates solely in clathrin uncoating under physiological conditions. Recent in vitro investigations have suggested several new functions of auxilin during CME [32], including facilitating clathrin exchange during CCV formation [44,45], binding to adaptor complexes [46], binding to dynamin to facilitate the constriction of coated pits [47], and chaperoning dissociated clathrin for subsequent rounds of endocytosis [48]. Thus, to understand the significance of ligand endocytosis in Notch, it is critical to determine whether dAux is exclusively involved in clathrin uncoating or has additional roles during CME under physiological conditions.
Using strong dAux alleles generated from an FRT-containing chromosome, we have obtained compelling evidences that dAux mutations cause defects in multiple Notchdependent processes and dAux function is required in the signaling cells. Although the relevant cargo for dAux during Notch signaling seems to be Dl, the endocytic function of dAux is not limited to the Notch ligand, as the internalizations of other membrane proteins also appear disrupted. Deletion analysis shows that over-expressed CBM and J-domain are sufficient for dAux function, suggesting that the recruitment of Hsc70 to CCVs is a core function of dAux. Furthermore, while the disruption of dAux function alters the sub-cellular clathrin distribution, Dl appears to be trapped at the cell surface, away from the abnormal clathrin-positive structures. Together, our data support a model, in which dAux facilitates Notch ligand endocytosis by regulating clathrin dynamics. Moreover, the fact that Dl is trapped at the plasma membrane in dAux mutants suggests that the linking of dAux to the Notch pathway does not exclude the possibility that the Notch ligand internalization per se plays a pivotal role in activating Notch.

Mutations in dAux kinase-or J-domains disrupt photoreceptor specification
To further understand the role of dAux under physiological conditions, we have isolated additional dAux mutations from several rounds of F 2 non-complementation screens. We have nine mutations in dAux, and, for eight of these nine alleles, the molecular lesions are known (Figure 1A). Furthermore, three of these alleles, namely dAux L78H , dAux F956 *, and dAux G257E , were generated from an FRT-containing chromosome; thus permitting dAux phenotypic analysis in clones.
Previous studies have shown that mutations in the PTENrelated region cause defects in photoreceptor specification [25,26]. To test whether mutations in other dAux domains have similar effects, FLP-induced mutant clones of dAux L78H and dAux F956 *, mutations disrupting the kinase and the J-domain respectively, were stained with α-Elav, which labels the nuclei of neuronal cells [50]. In wild-type tissues, an organized set of eight Elav-positive cells is formed in each cluster. In contrast, as shown in Figure 1B-E, increased numbers of disorganized Elav-positive cells were seen in both dAux L78H and dAux F956 * mutant clones ( Figure 1B-E, indicated by the absence of the nuclear GFP). This suggests that the integrity of the kinase and the J-domain, in addition to the PTEN-related region, is important for the function of dAux during neuronal cell specification. In addition, this neural hypertrophy exhibited by dAux F956 * tissues seemed to be more severe than those of dAux L78H (in the sense that additional dAux L78H Elav-positive cells appeared to be grouped in clusters, whereas dAux F956 * cells did not), correlating well with the allelic strength defined by the lethal phase analysis.

Notch-dependent proneural enhancement and lateral inhibition are both disrupted in dAux mutant tissues
The Drosophila compound eye consists of regular arrays of 800 ommatidia, each containing eight stereotypically positioned photoreceptors (R1-R8) and other accessory cells. The formation of this elaborate pattern is initiated in the larval eye imaginal disc along the morphogenetic furrow (MF, an indentation in the eye disc epithelium), which sweeps from posterior to anterior across the eye disc [51]. Behind the furrow, undifferentiated cells are sequentially recruited to adopt distinct developmental fates. To further understand the roles of dAux in photoreceptor cell differentiation, we investigated the effect of dAux mutations on the specification of R8, the founder cell within each cluster. The specification of R8 cells is known to require at least two Notch-dependent events: 1) proneural enhancement: an up-regulation of atonal (ato, a proneural gene) [52] expression in cells near the morphogenetic furrow to confer them the capability to adopt a neuronal fate, and 2) lateral inhibition: a subsequent restriction of the broad Ato expression to one single R8 cell per cluster behind the MF [53,54].
To determine if dAux is required for these processes, mosaic eye discs containing dAux F956 * mutant clones were stained for Ato. As shown in Figure 2A, most dAux F956 * mutant cells located near the MF expressed less Ato than their wild-type counterpart, suggesting that dAux has a role in the proneural enhancement event. Noticeably, some dAux F956 * mutant cells at the clone border still expressed an elevated level of Ato ( Figure 2C and 2D, solid arrows). One plausible explanation is that dAux functions non-cell-autonomously during Notch signaling (see below); therefore, the Notch pathway in these dAux F956 * mutant cells could still be activated because they were juxtaposed to the wild-type cells.
Behind the MF in normal eye discs, the broad Ato expression is gradually restricted, by Notch-mediated lateral inhibition, to a single cell in each cluster. Unlike wild type, dAux F956 * mutant tissues contained clumps of multiple Ato-positive cells that are not restricted into single Ato-expressing cells (Figure 2A, open arrows), suggesting that the process of restricting Ato expression was dis-dAux mutations disrupt the proneural enhancement and the lateral inhibition during R8 specification  To further confirm that lateral inhibition during R8 specification was affected by dAux, FLP-induced eye disc clones were stained for Senseless (Sens), a Zn fingercontaining transcription factor expressed in the nuclei of R8 cells [55]. Unlike Ato, whose expression disappears in more mature R8 cells, sens expression persists in all R8 cells in the eye disc [55]. Similar to the Ato staining, clusters of extra Sens-positive nuclei were seen in dAux F956 * mutant tissues after the furrow ( Figure 2H). Together, these observations suggest that both proneural enhancement and lateral inhibition, two known Notch-dependent processes, require dAux.

Mutations in dAux disrupt the formation of the dorsalventral wing boundary
To ask if dAux regulates other Notch-dependent processes, we examined the effect of dAux mutations on the formation of the dorsal-ventral (DV) boundary in the developing wing discs. In normal wing discs, cells in both the dorsal and ventral compartments express Notch. However, because of the modification of Notch receptor by fringe in the dorsal compartment [56,57], these dorsal cells respond preferentially to Dl signaling from cells in the ventral compartment. Conversely, the cells in the ventral half respond preferentially to the other Notch ligand, Ser, which is expressed in the dorsal half. As a result, margin-specific genes like cut are expressed in a stripe of cells along the DV border, in a Notch-dependent manner [58].
To test if this process requires dAux, wild-type and dAux F956 * mosaic wing discs were stained with 2B10 α-Cut antibody [59]. In wild-type wing discs, a stripe of Cut-positive cells was seen along the DV border ( Figure 3A). In contrast, Cut staining was absent in dAux F956 * mutant clones located at the DV border ( Figure 3B-D), suggesting that dAux is required for the DV boundary formation in wing development. Again, at a higher magnification (Figure 3D, inset), some dAux F956 * mutant cells at the clone border still expressed Cut, supporting the possibility that dAux functions non-cell-autonomously during Notch signaling (see below).

dAux acts non-cell-autonomously during Notch signaling
To determine in which cell dAux is required during Notch signaling, eye discs containing FLP-induced dAux F956 * mutant clones were stained for Enhancer of split (E(spl)), a transcriptional target of Notch signaling [60]. We reasoned that, if dAux functions in the receiving cells (cell autonomous), all mutant cells, regardless of their locations within the clones, will be unable to activate Notch and will, therefore, not express E(spl). On the other hand, if dAux functions in the signaling cells (non-cell autonomous), mutant cells at the clone border can still receive signals from neighboring wild-type cells, and will express E(spl). Consistent with this reasoning, none of the cells mutant for Notch receptor (N 264-39 ) within the clones expressed E(spl) ( Figure 4A-C) [61,62]. Conversely, as Dl is one of the ligands of this signaling cascade, several Dl RevF10 mutant cells at the clone border expressed E(spl) ( Figure 4D-F) [19]. Similarly, some dAux F956 * mutant cells at the clone border clearly expressed E(spl) ( Figure  4G-I). This, along with the observations described in previous sections regarding Ato and Cut expression, suggests dAux mutations disrupt Cut expression at the DV boundary in developing wing discs

Mutations in dAux cause Dl accumulation at the cell surface
As our genetic data suggest that dAux functions in the signaling cells during Notch signaling, a relevant cargo of dAux-dependent endocytosis is likely to be Notch ligand.
To test this, FLP-induced mutant clones of dAux L78H and dAux F956 * were first stained with C594.9B α-Dl antibody, which recognizes the extracellular domain of Dl [63].
Compared to the wild-type cells, the intensity of Dl staining was increased in both dAux F956 * and dAux L78H mutant cells ( Figure 5C&F). This increase of Dl staining intensity was more pronounced when the clones were at or near the morphogenetic furrows, consistent with the phenotype shown by a mutation in the PTEN-related region [26].
Although this elevated level of Dl seemed to accumulate around dAux mutant cell periphery ( Figure 5C&F), it was not clear whether Dl proteins were trapped at the plasma membrane or in vesicular structures inside the cells. To distinguish between these possibilities, eye discs containing dAux F956 * mutant clones were stained with C594.9B under a non-permeabilized condition to label Dl at the cell surface. In wild-type tissues, a high level of surface Dl staining was first seen in cells behind the morphogenetic furrow ( Figure 5G&I). In more mature clusters located in the posterior region of the eye disc, less Dl was seen at the surface, suggesting that most of Dl was internalized [64].
In dAux F956 * mutant clones, the surface Dl staining appeared excessive, indicating that the previously observed peripheral Dl most likely represents Dl accumulated at the cell surface.
dAux mutants show elevated levels of EGF and Notch receptor expression

Internalization of EGFR and Notch appear disrupted in dAux mutant cells Knockdown of GAK function in mammalian cells was shown to greatly inhibit the internalization of EGFR [65].
To determine whether the endocytic function of dAux is specific to Notch ligand, FLP-induced dAux F956 * mutant eye disc clones were stained with a α-DER antibody [66] and a α-Notch antibody (C17.9C6) [67], respectively. In wild-type eye discs, DER is expressed in cells ahead of furrow, and its expression is reduced behind the furrow [68]. Compared to wild-type cells, DER staining is elevated in the mutant cells ( Figure 6A-C), suggesting that the internalization of DER was inhibited. Similarly, the intensity of Notch staining was increased in dAux F956 * mutant cells ( Figure 6D-F). Thus, although our genetic data suggest that dAux is required in the signaling cells during Notch signaling, the endocytic function of dAux is not limited to the Notch ligand.

Clathrin distribution was disrupted in dAux mutant clones
We have previously observed a genetic interaction between dAux and clathrin light chain (Clc) [26]. To test directly whether clathrin distribution was disrupted by dAux, the localization of a Clc-GFP fusion [69] was examined in FLP-induced dAux mutant clones in eye discs. In wild-type cells (marked by the presence of a membraneassociated, myr-mRFP), Clc-GFP appeared as vesicular structures near the cell periphery ( Figure 7). In both dAux L78H and dAux F956 * mutant cells (marked by the absence of myr-mRFP), vesicular Clc-GFP still appeared near cell periphery but its intensity was clearly elevated (Figure 7). Furthermore, large and bright spots of Clc-GFP staining could be seen (arrows), suggesting that Clc-GFP or Clc-positive structures may form aggregates in dAux mutant cells. These data suggest that normal clathrin distribution depends on both the kinase and the J-domain of dAux. Interestingly, although the level of Dl proteins appeared elevated, staining of Clc-GFP-expressing dAux F956 * mutant cells with C594.9B α-Dl antibody showed that Dl was not enriched in these large Clc-positive structures ( Figure 7G-N).

The CBM and J-domains are indispensable for dAux function in Notch signaling
To investigate the domains critical for dAux function, we generated a series of mRFP-tagged dAux derivatives, each with a particular domain or a subset of domains removed ( Figure 8A). These include UAS-dAux FL -mRFP (fulllength), UAS-dAux ∆K -mRFP (kinase domain deleted), UAS-dAux CJ -mRFP (kinase and PTEN-related domains deleted), UAS-dAux ∆C -mRFP (CBM deleted) and UAS-dAux ∆J -mRFP (J-domain deleted). Using a α-DsRed antibody (Clontech) and the Act5C-GAL4 driver, bands of expected sizes were detected in blots of extracts prepared from these transgenic flies ( Figure 8B), indicating that these truncated dAux proteins were expressed.
To investigate the functional relevance of various dAux domains during Notch signaling, the abilities of these deletions, expressed using the Act5C-GAL4 driver, to rescue the supernumerary photoreceptor phenotype exhibited by dAux mutant eye discs were determined. While Act5C-GAL4 or UAS-dAux-mRFP constructs (not shown) alone had no effect, expression of dAux FL -mRFP completely suppressed the disorganization and the extra Elavpositive cell defects in FLP-induced dAux F956 * clones (Figure 8C). Similarly, expression of dAux ∆K -mRFP or dAux CJ -mRFP in dAux F956 *clones displayed arrays of normal complement of Elav-positive cells, suggesting that over-expression of dAux without its kinase domain and the PTENrelated region can restore normal Notch signaling ( Figure  8D&E). In addition, it should be noted that expression of dAux FL , dAux ∆K or dAux CJ could rescue dAux mutants to adult viability (data not shown). In contrast, expression of dAux ∆C -mRFP and dAux ∆J -mRFP could not rescue the extra Elav-positive cell defects in dAux F956 * clones ( Figure  8F&G). Over-expression of dAux ∆J -mRFP appeared to have some dominant negative effects. For instance, animals mutant for dAux L78H /dAux W328 * died during the larval stage. However, these mutants died before the larval stage when the J-domain deletion was expressed under the control of Act5C-GAL4 (data not shown). The results of these rescue experiments were confirmed using two independent transgenic lines from each construct and held true for another dAux allele (dAux L78H ), indicating that it is not allele specific. Together, these results suggest that while the kinase and PTEN-related domains are less critical, the J-domain and the CBM region are essential for dAux function.

Discussion
From a F 2 non-complementation screen, we have isolated several new dAux alleles, some of which contain point mutations disrupting specific domains. Consistent with our previous analysis of a viable dAux allele, strong dAux mutations affect several Notch-mediated processes, including photoreceptor specification in the eye and DV boundary formation in the wing. These phenotypes are consistent with the genetic interactions exhibited between dAux and Notch [26] and between dAux and lqf [25]. Taken together, these genetic observations strengthen the notion that endocytosis plays a critical role in Notch signaling, and suggest that dAux functions in multiple Notchdependent events.
As the functional importance of endocytosis has been suggested for both the signaling and receiving cells during Notch signaling [22], it is critical to determine in which cell is dAux function required. Although we have previ-dAux mutation disrupts clathrin distribution ously concluded that dAux is needed in the signaling cells, the evidence, obtained from mitotic clones of a weak dAux allele, was less than convincing [26]. To adequately address this critical issue, we have monitored the expression of E(spl), a Notch target gene, in clones mutant for strong dAux alleles. Using these reagents, it is clear that dAux mutant cells at the clone border can still activate Notch (a similar result was seen with Cut and Ato staining), suggesting dAux acts non-cell autonomously. These genetic data imply that the relevant cargo is likely to be the Notch ligand. Indeed, as shown by the surface labeling experiment, Dl internalization is disrupted in dAux mutant cells.
Inhibition of auxilin function by mutations [25,26,42,43], RNAi [31,41,65,70], or injection of inhibitory peptides [71] is known to interfere with the endocytosis of many molecules. In mammalian cells, inhibition of GAK function causes a decrease in the internalization of EGFR and transferrin [31,65]. Our observation suggests that, similar to the mammalian cells, dAux participates in the endocytosis of EGFR, although we did not previously observe a genetic interaction between DER and dAux [26]. It is possible that this lack of interaction between dAux and DER reflects the low sensitivity of our genetic assay. Alternatively, it may be that a defect in DER internalization does not significantly impact its signaling during eye development. Consistent with this, we have been unable to detect a drastic increase in the phosphorylation of MAP kinase, a downstream event of DER activation, in dAux F956 * mutant clones (data not shown). Nevertheless, our data show that, although the developmental defects of dAux resemble those of Notch, Notch ligand is not the sole cargo of auxilin-mediated endocytosis. This apparent specificity of dAux's Notch-like phenotypes suggests that the Notch pathway, compared to other signaling cascades, may be more sensitive to disruptions in the clathrin-mediated endocytosis.
Sequencing analysis of our dAux alleles revealed that disruptions in the kinase, the PTEN-related region, and the Jdomain could all result in abnormal Notch signaling. Noticeably, our screen did not isolate any point mutation in CBM, although the deletion analysis suggests that the CBM is critical for dAux function. This apparent discrepancy is likely due to the fact that the CBM domain contains multiple redundant clathrin-binding motifs [72,73], thereby obscuring the effect of eliminating one single motif by a point mutation. Interestingly, the removal of the CBM from the yeast auxilin (swa-2) does not completely eliminate its function in vivo [73]. The reason for this difference is unclear but it is possible that swa-2 contains other protein domains capable of substituting for the CBM. Similar to a study of the mammalian GAK [31], our deletion analysis confirmed the importance of the J-domain, as over-expression of the dAux ∆J construct fails to restore the extra photoreceptor cell defect. The CBM and J domains are thought to facilitate the recruitment of Hsc70 to CCVs, and a fragment consisting of CBM and J domain alone has been shown to support clathrin uncoating in vitro [30,33,34]. In support of this notion that the recruitment of Hsc70 to CCVs is likely to be a critical step, overexpression of the CBM and J domain alone could restore the supernumerary Elav-positive cell phenotype.
Conversely, our observation also implies that the loss of the kinase and PTEN-related region could be compensated by the over-expression of the CBM and J-domain. The PTEN-related region is thought to participate in the membrane recruitment of auxilin during CME [35,36].
Thus it is imaginable that a defect in the subcellular localization is less deleterious when the fragment consisting of CBM and J-domain is over-expressed. It is unclear how the requirement of kinase domain can be compensated by the over-expression of the CBM and J-domain, as the relevant substrate for dAux kinase domain during Notch signaling is not known. It should be mentioned that elevated expression of dAux CJ rescued the extra Elav-positive cell phenotype in both dAux F956 * and dAux L78H (point mutations disrupting the J-domain and the kinase domain respectively), arguing against a scenario in which the kinase domain of endogenous dAux F956 * mutant proteins could complement the over-expressed dAux CJ in trans. It is possible that some functional redundancy exists between dAux and Numb-associated kinase (NAK, the Drosophila homolog of adaptin-associated kinase) [74], as the kinase domains from both factors are known to phosphorylate adaptor complexes [31,37,[75][76][77][78]. However, although mutations in subunits of Drosophila AP1 and AP2 complexes have been implicated in other Notch-dependent processes [79,80], it is not clear if these adaptor complexes have a role in the Notch processes we examined. Homozygous α -adaptin mutants do not appear to exhibit a neurogenic phenotype [69,80]. Furthermore, the removal of one copy of AP2 µ subunit (by a deletion) has no effect on the dAux I670K rough eye phenotype (data not shown). In any case, it should be stressed that the kinase and the PTEN-related region do play a role in Notch signaling, as point mutations disrupting these domains cause Notch-like defects, albeit to a weaker extent. Taken together, these results suggest the role of the kinase and the PTEN-related region during Notch ligand endocytosis is less than obligatory.
What is the role of ligand endocytosis in Notch signaling? It has been suggested that, after receptor-ligand binding, ligand endocytosis may provide a mechanical stress or other types of micro-environment (clustered ligand and receptor, etc.) to facilitate Notch cleavage or NECD shedding [21,81]. Alternatively, before binding to Notch, the ligands may have to enter a particular recycling pathway to render them active [20,23,24]. We initially viewed the linking of dAux to Notch as evidence favoring the latter model because it suggests that ligand endocytosis needs to proceed past clathrin uncoating. However, as an increased level of the Dl appeared to be trapped at the mutant cell surface, not inside CCVs, the linking of dAux to Notch certainly does not exclude the model that ligand internalization per se is critical for Notch signaling. Biochemical analysis has suggested several additional functions for auxilin during the CCV cycle besides uncoating [32]. Although abnormal clathrin distribution was observed in dAux cells, given the resolution of our analysis, it is unclear which particular step(s) were affected. It is possible that mutations in dAux directly inhibit Notch ligand endocytosis by disrupting one or more of these early steps during CCV formation. Alternatively, dAux mutations may indirectly inhibit Notch ligand internalization by causing an excessive formation of non-functional clathrin-dependent structures, thereby decreasing the cytosolic clathrin pool. Indeed, in dAux mutant cells, those large clathrin-positive structures did not appear to contain an elevated level of Dl. Consistent with this, it was recently shown that over-expression of Chc could restore the dAuxassociated defects [27].

Conclusion
Our genetic analysis of strong dAux alleles clearly strengthens the notion that ligand endocytosis plays a critical role in Notch signaling. Furthermore, the deletion analysis suggests that the recruitment of Hsc70 to clathrin is a key event for dAux to facilitate Notch signaling. More importantly, we showed that Dl accumulates at the cell surface in dAux mutant cells. This suggests that the linking of dAux to the Notch pathway does not exclude the model in which ligand endocytosis activates Notch by physically dissociating the receptor.

Drosophila genetics
All fly crosses were carried out at 25°C in standard laboratory conditions unless otherwise specified. To facilitate the analysis of dAux phenotypes in clones, 5-5Z3515, an FRT-containing P-element insertion from the DrosoDel project (Cambridge, UK), was used to isolate additional dAux alleles. This strategy was chosen because 1) the conventional FRT site for the third chromosome right arm [82] is more distal from the centromere then the dAux locus, therefore not applicable for generating dAux mutant clones, and 2) crossovers between FRT 5-5Z3515 and dAux would have been astronomically rare, as the P element insertion in 5-5Z3515 is only 20 kb closer to the centromere than the dAux locus.
The screens for additional dAux alleles on FRT 5-5Z3515 chromosomes were performed as previously described [26]. Briefly, w; FRT 5-5Z3515 males were mutagenized with 25 mM ethyl methane sulfonate (Sigma), and mass mated with w/w; TM3, Sb/TM6B, Hu virgins. Progeny were then individually mated with dAux I670K , p [w + ]/TM6B, Hu flies, and those that failed to complement dAux I670K were recovered and maintained over TM6B or TM3 balancers. To determine the mutations in dAux alleles, coding regions were amplified from genomic DNA extracted from homozygous mutant embryos by PCR. Multiple independent PCR products were analyzed by direct sequencing.

Molecular Biology
The mRFP-tagged dAux deletions were constructed using PCR and standard cloning techniques. The dAux deletions, dAux ∆K , dAux CJ and dAux ∆J , correspond to amino acids 339-1165, 762-1165 and 1-998 respectively. dAux ∆C lacks the amino acids 768-1049. In all dAux constructs, mRFP or EGFP was appended in-frame at the Cterminus. The sequence and construction detail will be provided upon request. All the constructs were verified by sequencing, and multiple transgenic flies carrying the constructs were generated by P-element-mediated transformation [83].
For surface Dl labeling, mosaic eye discs were dissected and fixed in 4% paraformaldehyde. The peripodial membranes were partially removed, and the discs were then stained with mouse α-Dl C594.9B and washed with PBS in the absence of any detergent. The stained eye discs were examined using a BioRad MRC1024 laser confocal microscope (Nikon OPTIPHOT-2) and the images were processed with Adobe Photoshop.

Western analysis and fly extract preparation
To prepare fly extracts, flies expressing mRFP-tagged dAux derivatives were homogenized in 2× SDS-loading buffer containing 0.2 mM DTT (10 flies/100µl), boiled immediately for 5 minutes and separated by SDS-PAGE using 10% acrylamide gel. The gel was transferred to nitrocellulose membrane and probed with α-DsRed rabbit polyclonal antibody (Clontech) at 1:1000 dilution, followed by HRP-conjugated goat anti-rabbit secondary antibody (Jackson Lab). The immunodetection was performed using ECL substrate (Amersham Biosciences).