Fat facets and Liquid facets promote Delta endocytosis and Delta signaling in the signaling cells

Endocytosis controls cell signaling through a variety of different mechanisms (Seto et al., 2002; Gonzalez-Gaitan and Stenmark,2003). For example, signaling by the epidermal growth factor receptor following ligand binding is attenuated by receptor endocytosis and lysosomal degradation. Endocytosis of epidermal growth factor receptor also enhances signaling by transporting activated receptor to its targets. In addition, endocytosis plays a variety of roles in establishing gradients of morphogens like Hedgehog, Decapentaplegic and Wingless. Moreover, several different aspects of Notch pathway function rely on endocytosis.

Two proteins required for pattern formation in the Drosophila eye,the deubiquitinating enzyme Fat facets (Faf) and its substrate Liquid facets(Lqf), are linked to both cell signaling and clathrin-mediated endocytosis(Fischer-Vize et al., 1992; Huang et al., 1995; Cadavid et al., 2000; Chen et al., 2002; Overstreet et al., 2003). Lqf protein levels in the Drosophila eye are controlled by the balance between ubiquitination, which targets the protein for proteasomal degradation,and deubiquitination by Faf, which prevents Lqf degradation(Huang et al., 1995; Wu et al., 1999; Chen et al., 2002). Faf and Lqf mediate a cell communication event that prevents overneuralization of the compound eye. Accordingly, faf or lqf mutant eyes contain more than the normal complement of eight photoreceptors in each facet (or ommatidium) of the eye. As mosaic experiments demonstrate that faf⁺ and lqf⁺ function outside of the ectopic photoreceptors, the extra photoreceptors must result from a failure of cell signaling (Fischer-Vize et al.,1992; Cadavid et al.,2000). Several observations suggest that Faf and Lqf facilitate endocytosis. First, Lqf is the Drosophila homolog of epsin, a multi-modular protein that binds phosphoinisitol lipids at the cell membrane,the adapter complex AP2, clathrin, ubiquitin and other endocytic accessory factors (Kay et al., 1998; De Camilli et al., 2001; Wendland, 2002). Epsin is required for endocytosis in yeast and in mammalian cells(Wendland et al., 1999; Itoh et al., 2001; Shih et al., 2002). In addition, faf and lqf mutations show dramatic genetic interactions with mutations in the clathrin heavy chain gene, which indicate that all three genes function in the same direction in a pathway(Cadavid et al., 2000). Finally, the Notch ligand Delta fails to be internalized normally in lqf mutant eye discs (Overstreet et al., 2003).

The overneuralization phenotype in faf and lqf mutants,and the altered Delta localization in lqf mutants suggest a role for Faf and Lqf in Notch/Delta signaling. The Notch pathway is highly conserved in metazoans and participates in a wide range of cell communication events that determine cell fate. Mutants in the Notch receptor and in other genes in the signaling pathway (`neurogenic' genes) were first isolated on the basis of their role in inhibiting neural cell fate determination in Drosophilaembryos (Lehmann et al.,1981). It is now apparent that Notch receptor activation, in different cellular contexts, can result in either inhibition or promotion of a variety of cell fates (Artavanis-Tsakonas et al., 1999). The mechanism of Notch signaling is unusual in that upon ligand binding, a fragment of the Notch intracellular domain is cleaved,travels into the nucleus, and acts a transcriptional regulator(Artavanis-Tsakonas et al.,1999). Although details of the events that lead to nuclear translocation of the Notch intracellular domain are contentious, there is a consensus model where binding of ligand to the Notch extracellular domain induces two cleavages of Notch. The first cleavage (called S2) detaches the extracellular domain from the remainder of the Notch protein, and is prerequisite for the second cleavage (S3) that releases the transcription factor domain (Baron,2003).

Endocytosis controls Notch signaling in both the signaling and receiving cells. The first evidence for this idea came from analysis of Drosophila shibire mutants. shibire encodes the Drosophila homolog of dynamin, a GTPase required for scission of endocytic vesicles(Chen et al., 1991). shibire mutant phenotypes resemble Notch loss-of-function phenotypes, and the results of mosaic experiments suggest that shibire is required in both the signaling and receiving cells(Poodry, 1990; Seugnet et al., 1997). A model for the dual function of shibire was formulated for Notch signaling during lateral inhibition, where both the signalers and receivers express both Notch and Delta. In this case, selective internalization of either Notch or Delta could bias cells to become either the signaler or the receiver. Recent experiments with Drosophila sensory organ precursors support the idea that Notch internalization may bias a cell to become the signaler. The Numb protein, which binds Notch and the endocytic protein α-adaptin, is asymmetrically distributed between two daughter cells and the Numb-containing cell becomes the signaler (Rhyu et al.,1994; Lu et al.,1998; Santolini et al.,2000; Berdnik et al.,2002; Le Borgne and Schweisguth, 2003). Thus, by stimulating Notch internalization,Numb may bias one sensory organ precursor cells to become the signaler.

In addition to preventing a cell from displaying either Notch or Delta at the cell membrane, endocytosis has also been proposed to play a positive role in Notch receptor activation (Parks et al., 2000). The idea is that the Notch extracellular domain, bound to Delta, is trans-endocytosed into the Delta-expressing (signaling) cell. This trans-endocytosis event is prerequisite for S2 cleavage, and therefore for S3 cleavage and activation of Notch in the receiving cell. Evidence for this model comes from experiments in the developing Drosophila eye using two non-neural cell types: cone cells and pigment cells(Parks et al., 1995; Parks et al., 2000). Delta is transcribed in cone cells, and Notch is transcribed in pigment cells. Yet, the extracellular domain of Notch (N^ECD) is detected with Delta in endosomes inside the cone cells. Moreover, in shibire mutants, Notch and Delta both accumulate at cone cell plasma membranes. In addition, in Delta mutants, there are significantly fewer N^ECD-containing vesicles in cone cells. In addition, in temperature-sensitive Delta loss-of-function mutants, Delta accumulates on cone cell membranes. Finally, in cell culture, cells expressing Delta alleles that encode endocytosis-defective ligands do not trans-endocytose N^ECD.

Consistent with the trans-endocytosis model, the ubiquitin-ligases Neuralized (in Drosophila and Xenopus) and Mindbomb (in zebrafish) modulate Delta endocytosis and Delta signaling. Neuralized (Neur)and Mindbomb ubiquitinate Delta thereby stimulating Delta internalization(Itoh et al., 2003; Yeh et al., 2001; Deblandre et al., 2001; Lai et al., 2001; Pavlopoulos et al., 2001). The results of several studies suggest that Neur and Mindbomb are required in the Delta signaling cells to promote Notch activation in the receiving cells(Pavlopoulos et al., 2001; Itoh et al., 2003; Le Borgne and Schweisguth,2003; Li and Baker,2004). However, the role of Neur is unclear, as other reports suggest that Neur is required for Delta internalization in the receiving cells, perhaps to bias those cells to become the receivers(Yeh et al., 2000; Lai et al., 2001; Lai and Rubin, 2001a; Lai and Rubin, 2001b).

Here, we report a unique mechanism for regulating Notch/Delta signaling. We show that the deubiquitinating enzyme Faf, through its substrate Lqf, promotes Delta internalization and Delta signaling by the signaling cells. The signaling cells, photoreceptor precursors R2/3/4/5, thus activate Notch in surrounding undifferentiated cells, preventing recruitment of ectopic photoreceptors (R-cells). We call this event R-cell restriction. In addition,we show that while Faf is required only for R-cell restriction, Lqf is needed also for two earlier events in the eye that require Notch/Delta signaling:proneural enhancement and lateral inhibition. We also provide evidence that Neur functions with Faf and Lqf in R-cell restriction. There are three main conclusions of this work. First, the results provide strong support for the model where Delta internalization by the signaling cell is required for Notch activation in the receiving cell. Second, the results support a model where Neur stimulates Delta internalization in the signaling cells rather than in the receiving cells. Finally, we demonstrate that deubiquitination by Faf of the endocytic factor Lqf is a novel mechanism for regulating Delta signaling. We propose that by elevating Lqf activity, Faf enhances the efficiency of Delta endocytosis and promotes Delta signaling.

Our laboratory maintains stocks of lqf^ARI FRT80B(Overstreet et al., 2003), lqf^FDD9 (Cadavid et al., 2000) and faf^FO8(Fischer-Vize et al., 1992; Chen and Fischer, 2000). FRT82B Dl^rev10 (Baker and Yu, 1996) was obtained from N. Baker. The following lines were obtained from the Bloomington Drosophila Stock Center: neur¹ and neur¹¹ (Lehmann et al.,1993); Ub-GFP FRT80B and FRT82B Ub-GFP(FlyBase, 2003; Xu and Rubin,1983); ey-FLP on X (Newsome et al., 2000); and EGUF; FRT82B GMR-hid l(3)CL-R(Stowers and Schwarz,1999).

Although neur¹ and neur¹¹ are reported to be null alleles, several results presented here suggest that neur¹¹ retains some neur⁺ activity. As described below, neur¹ enhances the lqf^FDD9 phenotypes much more strongly than does neur¹¹, and the eye disc patterning defects in neur¹ are more severe than in neur¹¹.

lqf^ARI eye disc clones were generated in larvae of the following genotypes: ey-FLP; lqf^ARI FRT80B/Ub-GFP FRT80B. Dl^rev10 eye disc clones were generated in larvae of the following genotype: ey-FLP; FRT82B Dl^rev10/FRT82B Ub-GFP. neur¹ eye disc clones were generated in larvae of the following genotype: ey-FLP; FRT82B neur/FRT82B Ub-GFP. neur¹¹ eye discs were generated in larvae of the following genotype: EGUF/RO-GFP; FRT82B neur¹¹/FRT82B GMR-hid l(3)CL-R.

Sectioning, light microscopy and photography of adult eyes was as described(Huang et al., 1995). Flies with neur¹¹ eyes were: EGUF/+; FRT82B neur¹¹/FRT82B GMR-hid l(3)CL-R. The faf^FO8/faf⁺ mosaic ommatidia are those described (Fischer-Vize et al.,1992) and they were reanalyzed here using different criteria. The faf^BX4/faf⁺ mosaic ommatidia were generated and prepared for microscopy exactly as described(Fischer-Vize et al.,1992).

Primary antibodies used were rabbit polyclonal anti-Ato at 1:2000(Jarman et al., 1994) from Y. N. Jan; anti-Boss mouse ascites at 1:2000(Kramer et al., 1991) from H. Kramer; anti-E(spl) mAb323 supernatant at 1:2(Jennings et al., 1994) from S. Bray; anti-Dl mAb202 supernatant at 1:10(Parks et al., 1995) from H. Kramer; and rat monoclonal anti-Elav supernatant at 1:9(O'Neill et al., 1994) from the Developmental Studies Hybridoma Bank. Secondary antibodies (Molecular Probes) were Alexa⁶³³-anti-mouse, Alexa⁵⁶⁸-anti-mouse,Alexa⁶³³-anti-rat and Alexa⁶³³-anti-rabbit, all used at 1:500. In addition, Alexa⁵⁶⁸- and Alexa⁶³³-phalloidin were used as described (Chen et al.,2002). Eye discs immunostaining and confocal microscopy were as described (Chen et al.,2002).

A DNA fragment containing GFP flanked by AscI sites was generated by PCR, using a GFP-containing plasmid(Siemering et al., 1996) as a template and the following primers:5′GGCGCGCCATGAGTAAAGGAGAAGAAC3′ and 5′GGCGCGCCTTATTTGTATAGTTCATCCC3′. The PCR product was ligated into pGEM-T-Easy (Promega) to generate pGEM-GFP. The GFP DNA sequence in pGEM-GFP was determined, and the AscI fragment containing GFP was isolated and ligated into the AscI site of pRO(Huang and Fischer-Vize,1996). A plasmid, pRO-GFP, with the AscI fragment in the appropriate orientation was isolated.

An AscI-NdeI DNA fragment containing GFP was generated by PCR using a GFP-containing plasmid(Siemering et al., 1996) as a template and the following primers:5′CAGATGGGCGCGCCATGAGTAAAGGAGAAC3′,5′CATATGTTTGTATAGTTCATCC3′. The PCR product was ligated into pGEM-T-Easy to generate pGEM-GFP-AN. The GFP DNA sequence in pGEM-GFP-AN was determined and the ∼700 bp AscI-NdeI GFP fragment was isolated and ligated into a plasmid containing the lqf cDNA called pMoPac-lqf-cDNA3. pMoPac-lqf-cDNA3 was constructed as follows: the lqf cDNA was generated in two parts by PCR using as a template a plasmid containing lqf cDNA-3(Cadavid et al., 2000). The 5′ part of lqf was generated as an NdeI-HpaI fragment using the primers 5′ATGCAGGTCAATGTCGCTGG3′ and 5′CGGTTTGATCAGATTGTCTAGG. The PCR product was ligated into pGEM-T-Easy to generate pGEM-Lqf5′ and the lqf DNA sequence in the plasmid was determined. The 3′ part of lqf was generated as an HpaI-AscI fragment using the primers 5′TTTCCTCGGCGAGAACTC3′ and 5′TTACGACAAAAACGGATTTGTTG3′. The PCR product was ligated into pGEM-T-Easy to generate pGEM-cDNA3-3′ and the lqf DNA sequence in the plasmid was determined. A ∼1650 bp NdeI-HpaI fragment of pGEM-Lqf5′ and ∼800 bp NdeI-AscI fragment of pGEM-cDNA3-3′ were isolated and ligated into pMoPac(Hayhurst et al., 2003)restricted with NdeI and AscI.

An SpeI-SalI fragment of pTM1 containing shi^K44A (Moline et al., 1999) (obtained from A. Bejsovec) was ligated into pBSKII(Stratagene) restricted with SpeI and SalI to generate pBSK-shi^DN. AscI sites flanking the shi^K44A gene were added as follows: pBSK-shi^DNwas restricted with SpeI, treated with Klenow fragment, and an AscI linker ligated in. A second AscI linker was ligated similarly into the SalI site. The resulting AscI fragment of shi^K44A was purified and ligated into pRO. A plasmid,pRO-shi^DN, with the AscI fragment in the appropriate orientation, was isolated.

A DNA fragment of Delta lacking the cytoplasmic domain and flanked by AscI sites was generated by PCR, using as a template pG1C(Fehon et al., 1990) (obtained from M. Muskavitch), which contains a complete Delta cDNA and the following primers and also inserted a stop codon:5′GGCGCGCCCACACACACACACAGCCCTG3′ and 5′GGCGCGCCTTACACCGCATTCTGTTC3′. The PCR product was ligated into pGEM-T-Easy to generate pGEM-Dl^DN. An AscI fragment containing the truncated Delta gene was purified from pGEM-Dl^DN and ligated into pRO. A plasmid, pRO-Dl^DN,with the AscI fragment in the appropriate orientation was isolated.

P-element transformants were generated by injection of w¹¹¹⁸ embryos using standard techniques.

Drosophila eye development is controlled by a complex network of cell signaling pathways, which includes many roles for Notch/Delta signaling(Mlodzik, 2002; Nagaraj et al., 2002). The Drosophila compound eye is composed of hundreds of identical ommatidia. The eye develops in larval and pupal stages from a cellular monolayer called the eye disc (Wolff and Ready, 1993). In third instar larvae, a wave of morphogenesis,initiated at the posterior of the disc by the morphogenetic furrow, moves anteriorly through the monolayer of undifferentiated cells. A column of organized preclusters emerges from the furrow (column 0)(Fig. 1G). A few cells are excluded from the initial preclusters and the remainder differentiate into five of the eight photoreceptors (R-cells; R8/2/3/4/5). These clusters then recruit R1/6/7, the four cone cells, and finally the pigment and bristle cells. As the furrow moves forward by one column approximately every 2 hours,each more posterior column is one step more mature and the sequence of ommatidial assembly can be observed in a single disc.

The pattern of Delta expression in wild-type eye discs has been well-characterized. Delta transcription is ubiquitous in the morphogenetic furrow, and then resolves to the R-cell preclusters as they emerge from the furrow (Parks et al.,1995). Delta protein is detected in a similar pattern of cells and its subcellular localization is intriguing. Although Delta is expected to function at the membrane, an antibody to the Delta extracellular domain detects most of the protein in endosomal vesicles posterior to the furrow(Fig. 1D)(Parks et al., 1995). Delta-containing vesicles first accumulate in preclusters emerging from the furrow, then in R-cells as they differentiate, and remain detectable in some R-cells until at least column 14 (Parks et al., 1995). Using unusual tissue preparation conditions (no detergent), low levels of membrane-bound Delta are observed in the same pattern as Dl transcripts (Baker and Yu,1998). These observations suggest that in some cells, most of the Delta at the cell surface is internalized and that endosomal Delta is not degraded rapidly.

In lqf^FDD9 eye discs, which produce low levels of wild-type Lqf protein, Delta accumulates on cell membranes in columns 0-3 posterior to the furrow (Fig. 1F) (Overstreet et al.,2003). Like lqf^FDD9, faf mutant discs have decreased levels of Lqf protein (Chen et al., 2002). In order to determine if Delta internalization is defective in faf mutant discs and in which cells, we double-labeled faf^FO8 third instar larval eye discs[faf^FO8 is a strong mutant allele(Fischer-Vize et al., 1992; Chen and Fischer, 2000)] with antibodies to the Delta extracellular domain and with phalloidin to outline the apical membranes of the ommatidial cluster cells. We find that Delta is present on the membranes of R2/3/4/5 and the ectopic R-cells in columns 0-3 of faf^FO8 discs (Fig. 1E,H-H″). Some vesicular Delta is also observed(Fig. 1H″). We conclude that both faf⁺ and lqf⁺ are required for Delta endocytosis in R-cell clusters in columns 0-3.

The observation that similar Delta internalization defects occur in faf and lqf mutant discs supports the idea that the faf mutant phenotype results from a decrease in the level of Lqf protein. However, more Delta-expressing cells emerge posterior to the furrow in lqf^FDD9 discs than in wild-type or faf discs. The difference in Delta expression between faf and lqf^FDD9 discs reflects a broader requirement for lqf⁺ in early developmental decisions (see below).

In faf mutants, the R2/3/4/5 precursors display Delta endocytosis defects. In order to determine whether faf⁺ and lqf⁺ function in these cells, we investigated the expression pattern of the vector pRO(Huang and Fischer-Vize,1996). pRO transgenes that drive expression of a faf cDNA(RO-faf) can substitute for the endogenous faf gene(Huang and Fischer-Vize,1996). Likewise, a RO-lqf transgene rescues to wild type the mutant eye phenotype of lqf^FDD9 or faf(Cadavid et al., 2000). We generated a RO-GFP transgene and observed the pattern of GFP expression in eye discs from three independent transformant lines. We find that GFP is expressed in R2/3/4/5 beginning in column1(Fig. 2A,B). The same results were obtained with a RO-GFP-lqf transgene which also complements the faf and lqf^FDD9 mutant phenotypes (data not shown). We conclude that expression of faf⁺ or lqf⁺ in R2/3/4/5 is sufficient to substitute for the endogenous faf gene or to compensate for the lower levels of Lqf protein in lqf^FDD9.

To investigate further the requirement for faf⁺ in R2/3/4/5, we analyzed adult ommatidia mosaic for marked faf⁺ and faf^- cells generated by mitotic recombination. Two types of genetically mosaic facets were observed and analyzed: phenotypically mutant ommatidia with more than six outer (R1-6)R-cells, and phenotypically wild-type ommatidia. The genotype of each outer R-cell (including ectopic cells) was scored in both types of mosaic facets(Fig. 2C-E). In assigning R-cell identities, we assumed that the ectopic R-cells arise between R3 and R4. If faf⁺ is required in all or a subset of R2/3/4/5,then we would expect to find no phenotypically mutant facets where R2/3/4/5 are all faf⁺. As expected, in not one of 86 mutant mosaic facets at the borders of 30 different faf^FO8 clones were R2/3/4/5 all faf⁺ (Fig. 2C,D). Moreover, in nearly half of the mutant mosaic ommatidia(42/86), none of the R2/3/4/5 group is faf⁺ and in only 2/88 mutant mosaics are three of the R2/3/4/5 group faf⁺(Fig. 2C,D,F). Conversely, we expected that R2/3/4/5 would not all be faf^- in phenotypically wild-type facets. For this analysis, we used faf^BX4, which is a null allele(Fischer-Vize et al., 1992). In only 1/51 phenotypically wild-type mosaic facets in 13 different clones were R2/3/4/5 all faf^-(Fig. 2E). Moreover, although no particular R-cells in the R2/3/4/5 cell group were always faf⁺, at least three of them were faf⁺in 36/51 mosaic facets, and at least two of them were faf⁺in 47/51 of the mosaic facets (Fig. 2E,F). The wild-type mosaic ommatidia where not one R-cell (1/51)or only one R-cell (3/51) of the R2/3/4/5 group is faf⁺can be explained by the observation that in faf^BX4homozygotes, ∼10% of the facets are phenotypically wild type. These results show that as more of the R-cells in the R2/3/4/5 group are faf⁺, there is an increasing tendency for the ectopic R-cells to be excluded.

faf⁺ and lqf⁺ activities are linked to endocytosis and Delta endocytosis fails in precluster cells with decreased lqf⁺ activity (faf^FO8 or lqf^FDD9). Is a failure of endocytosis the cause of the faf and lqf^FDD9 mutant eye phenotypes? If so,then disrupting endocytosis in R2/3/4/5 through a mechanism other than blocking faf⁺ or lqf⁺ gene activity should result in an eye phenotype similar to that of faf or lqf^FDD9. We interfered with endocytosis in R2/3/4/5 by expressing a dominant-negative form of Shibire(Moline et al., 1999) using the pRO vector (RO-shi^DN). We find that otherwise wild-type flies expressing RO-shi^DN display adult retinal defects similar to those in faf or lqf^FDD9mutants (Fig. 3A, Fig. 1A-C). The ectopic R-cells in RO-shi^DN join the clusters in columns 0-3 as in faf or lqf^FDD9 discs(Fig. 3B-D). Moreover, Delta internalization defects similar to those in faf or lqf^FDD9 are observed in RO-shi^DN eye discs (Fig. 3B-D, Fig. 1E,F). We conclude that R2/3/4/5 precursors require endocytosis to prevent inappropriate recruitment of neighboring precluster cells as R-cells.

Does the failure of Delta signaling in R2/3/4/5 cause the faf and lqf^FDD9 mutant phenotypes? If so, then specifically interfering with Delta endocytosis and signaling in R2/3/4/5 should phenocopy faf and lqf^FDD9 mutants. To test this, we used the pRO vector to express in R2/3/4/5 a dominant-negative form of Delta(Dl^DN) (Sun and Artavanis-Tsakonas, 1996). In RO-Dl^DNtransformant eye discs, ectopic R-cells join the clusters in columns 0-3(Fig. 3F-H) and are present in adult eyes (Fig. 3E). In addition, Delta protein accumulates on R-cell membranes near the furrow(Fig. 3F). The Dl^DNprotein has a truncated intracellular domain and if Delta endocytosis is required for Delta signaling, the dominant-negative activity of Dl^DN is probably due to its failure to be internalized. Thus, the membrane-associated Delta protein observed in RO-Dl^DNdiscs may be a mixture of Dl^DN protein and wild-type Delta that is prevented by Dl^DN from interacting with Notch. We conclude that specific disruption of Delta signaling and endocytosis in R2/3/4/5 results in the same developmental consequences as does interfering with faf or lqf function.

We have shown that in order to prevent recruitment of ectopic R-cells into the ommatidia, faf⁺ and lqf⁺ are required for Delta signaling by R-cell precursors just posterior to the furrow. faf⁺ appears to be essential only for this one Delta signaling event: in faf^FO8 (strong) mutants, Delta is on the membrane in R-cell preclusters, ectopic R-cells are recruited just posterior to the furrow and the adult eye phenotype (ectopic R-cells) reflects these events. By contrast, lqf⁺ appears to be necessary also for earlier patterning processes. In lqf mutant eye discs[lqf^FDD9 or discs with small lqf^ARI(null) clones], all cells emerging from the furrow express Delta(Fig. 1F)(Overstreet et al., 2003)(also see below), whereas in wild-type discs Delta is expressed in distinct clusters (Fig. 1D)(Parks et al., 1995). In addition, in the adult eye, the phenotype of lqf^ARI clones is much more severe than that of faf mutants(Fischer et al., 1997).

Prior to the faf⁺-dependent signaling event, two discrete Notch/Delta signaling processes are required for the evolution of expression of the proneural protein Atonal(Baker and Yu, 1996; Baker et al., 1996; Baker, 2002). First, Notch activation in groups of cells anterior to the furrow upregulates Atonal expression; this event is referred to as proneural enhancement. Elevated Atonal levels are necessary for neural determination of these cells. Second,Notch/Delta signaling is essential for lateral inhibitory interactions that resolve Atonal expression to one cell by column 0. The one Atonal-expressing cell becomes R8, the founder R-cell of each ommatidium(Baker and Yu, 1998).

In order to determine whether lqf⁺ is required for Delta signaling during proneural enhancement and/or lateral inhibition, we analyzed the phenotypes of large lqf^ARI (null) clones using a number of different antibodies and compared them with the phenotypes of large Dl^rev10 (null) clones. We find that the lqf^ARI clone phenotypes closely resemble those of Dl^rev10 clones described earlier(Baker and Yu, 1996). Upregulation of atonal (proneural enhancement) does not occur in the Dl^rev10 or lqf^ARI clone centers(Fig. 4); although the cells in the middle of the clone are Notch⁺, there are no Delta⁺ cells adjacent to them to activate Notch. As would be expected, Dl^rev10 or lqf^ARI mutant cells at the clone borders adjacent to Delta⁺ cells do upregulate atonal (Fig. 4). In the absence of proneural enhancement, no R-cells are expected to be determined posterior to the furrow. Consistent with this,R-cells are absent from the centers of Dl^rev10 or lqf^ARI clones (Fig. 5A,A′,C,C′). By contrast, at the clone borders where mutant cells undergo proneural enhancement, R-cells are present(Fig. 5A,A′,C,C′). Lateral inhibition also fails in Dl^rev10 and lqf^ARI clones. The R-cells at the Dl^rev10 or lqf^ARI clone borders are not organized into discrete ommatidia; instead, it appears that all of the mutant border cells are R-cells (Fig. 5A,A′,C,C′). As these cells cannot send Delta signals,lateral inhibition fails. Consistent with this idea, there are clusters of R8s at the borders of the clones (Fig. 5B,B′,D,D′). We conclude that lqf⁺is required in the Delta signaling cells for proneural enhancement and lateral inhibition.

The results so far suggest that faf⁺ and lqf⁺ are required for Delta internalization and Delta signaling. One prediction of this model is that faf⁺ and lqf⁺ should function non-autonomously; faf⁺ or lqf⁺ cells adjacent to mutant cells should fail to have their Notch pathways activated and should be misdetermined as R-cells. Ectopic R-cells present in faf⁺/faf^- mosaic ommatidia in adult eyes are often faf⁺(Fig. 2C,D)(Fischer-Vize et al., 1992). The same phenomenon was observed in lqf⁺/lqf^- mosaic facets(Cadavid et al., 2000). Thus, faf⁺ and lqf⁺ function non-autonomously. Conversely, if lqf⁺ functions in the Delta-signaling cells as opposed to the receiving cells, it should be possible to activate Notch in lqf-null mutant cells that are adjacent to lqf⁺ cells. To test this, we generated lqf^ARI clones and Dl^rev10 (null)clones as a control in eye discs and labeled them with mAb323, which recognizes several different Enhancer of split [E(spl)] proteins expressed in response to Notch activation (Jennings et al., 1994). There is little Notch activation in the middle of the Dl^rev10 clones (Fig. 6A,A′) or the lqf^ARI clones(Fig. 6B,B′) (see also legend). Thus, like Delta⁺, lqf⁺ is required for Notch activation in neighboring cells. At the borders of the Dl^rev10 clones near the furrow, Delta⁺Notch⁺ cells outside the clone can signal the Dl^rev10 Notch⁺ cells inside the clone. Thus,E(spl) protein is detected in many Dl^rev10 cells at the clone borders (Fig. 6A,A′). The same phenomenon is observed the borders of lqf^ARI clones (Fig. 6B,B′). Thus, the Notch signaling pathway may be activated in lqf^- cells. We conclude that cells lacking lqf⁺ activity can activate their own Notch pathway in response to signals from neighboring cells, but cannot signal to activate Notch in their neighbors.

If the effect of lqf⁺ on Delta endocytosis is direct,then when lqf⁺ and lqf^- cells are juxtaposed, Delta should accumulate only on the membranes of lqf^- mutant cells. In small lqf^ARI(null) clones in eye discs, Delta accumulates on the membranes of all cells emerging from the furrow (Fig. 7) (Overstreet et al.,2003). At the clone borders, high levels of membrane-bound Delta are observed only in the lqf^ARI mutant cells(Fig. 7). We conclude that the effect of Lqf on Delta internalization is cell autonomous.

Neur is required for Delta internalization in wing and eye discs(Lai et al., 2001; Pavlopoulos et al., 2001). However, the only specific functions demonstrated for neur⁺ in the eye are a weak requirement in proneural enhancement and lateral inhibition (Lai and Rubin, 2001a; Li and Baker, 2001; Li and Baker,2004). The observation that the neur adult eye mutant phenotype resembles that of faf and lqf^FDD9mutants (Fig. 8A)(Lai and Rubin, 2001a) and that neur⁺ is expressed specifically in R-cells that emerge from the furrow (Pavlopoulos et al., 2001; Lai and Rubin,2001b) led us to test whether neur⁺ is also required for faf⁺-dependent Delta signaling by R2/3/4/5 precursors.

In order to determine if neur⁺ is required in R2/3/4/5 precursors for Delta internalization and signaling, we performed three experiments. We first tested neur for genetic interactions with faf and lqf. We find that two strong mutant neuralleles (neur¹ and neur¹¹) are powerful dominant enhancers of lqf^FDD9. neur¹lqf^FDD9/lqf^FDD9 animals die as larvae. neur¹¹ lqf^FDD9/lqf^FDD9 are viable and their retinal defects are more severe than lqf^FDD9/lqf^FDD9 (compare Fig. 8B with Fig. 1C). In eye discs, neur enhances the lateral inhibition defects in lqf^FDD9; the clusters of Delta-expressing cells are larger in lqf^FDD9 neur¹/lqf^FDD9 discs(Fig. 8C,C′) than in lqf^FDD9 (Fig. 1F) and Delta is on the cell membrane. neur mutants enhance the faf mutant phenotype weakly (data not shown). The genetic interactions are consistent with the idea that neur⁺,lqf⁺ and faf⁺ function in the same direction in a pathway. Second, we monitored the distribution of Delta in neur eye discs. In neur¹¹ eye discs that express RO-GFP we find that, similar to faf mutants, Delta accumulates on membranes of the R-cell clusters(Fig. 8D,D′). The Delta mislocalization phenotype of neur¹ eye discs is stronger than neur¹¹ and similar to lqf^FDD9(Fig. 8E,E′). Finally, we asked what effect neur mutant cells have on Notch activation near the furrow. We find that neur^- cells behave similarly to lqf^- cells; Notch is activated in neur¹ cells at clone borders that are adjacent to neur⁺ cells, but not in neur¹ cells in the center of mutant clones (Fig. 8F,F′). These results suggest that an important function of neur⁺ in the eye is in R-cell restriction and that neur⁺ functions with faf⁺ and lqf⁺ in the Delta signaling cells.

Cells with decreased lqf⁺ activity accumulate Delta on apical membranes and fail to signal to neighboring cells. We examined three Notch/Delta signaling events in the eye: proneural enhancement, lateral inhibition and R-cell restriction (Fig. 9A). We find that loss of lqf⁺-dependent endocytosis during all three events has identical consequences to loss of Delta function in the signaling cells. We conclude that lqf⁺-dependent endocytosis of Delta is required for signaling, supporting the notion that endocytosis in the signaling cells activates Notch in the receiving cells. However, Lqf is not required absolutely for all Delta internalization in the eye. Even in lqf-null cells, which are incapable of Delta signaling, some vesicular Delta is present(see Fig. 7). Perhaps not all of the vesicular Delta present in wild-type discs results from signaling.

Genetic studies in Drosophila indicate clearly that deubiquitination of Lqf by Faf activates Lqf activity(Wu et al., 1999; Cadavid et al., 2000). Moreover, genetic and biochemical evidence in Drosophila suggests that Faf prevents proteasomal degradation of Lqf(Huang et al., 1995; Chen et al., 2002). In vertebrates, however, it is thought that epsin is mono-ubiquitinated to modulate its activity rather than poly-ubiquitinated to target it for degradation (Oldham et al.,2002; Polo et al.,2002). If Lqf regulation by ubiquitin also occurs this way in the Drosophila eye, the removal of mono-ubiquitin from Lqf by Faf would activate Lqf activity.

Whatever the precise mechanism, given that both Faf and Lqf are expressed ubiquitously in the eye (Fischer-Vize et al., 1992; Chen et al.,2002), two related questions arise. First, why is Lqf ubiquitinated at all if Faf simply deubiquitinates it everywhere? One possibility is that Faf is one of many deubiquitinating enzymes that regulate Lqf, and expression of the others is restricted spatially. This could also explain why Faf is required only for R-cell restriction (see below). Another possibility is that Faf activity is itself regulated in a spatial-specific manner in the eye disc. Alternatively, Lqf ubiquitination may be so efficient that Faf is needed to provide a pool of non-ubiquitinated, active Lqf. Similarly, Faf could be part of a subtle mechanism for timing Lqf activation. Second, why is Faf essential only for R-cell restriction? One possibility is that there is a graded requirement for Lqf in the eye disc, such that proneural enhancement requires the least Lqf, lateral inhibition somewhat more and neural inhibitory signaling by R2/3/4/5 the most. The mutant phenotype of homozygotes for the weak allele lqf^FDD9 supports this idea, as R-cell restriction is most severely affected. Alternatively, Lqf may be expressed or ubiquitinated with dissimilar efficiencies in different regions of the eye disc. More experiments are needed to understand the precise mechanism by which the Faf/Lqf interaction enhances Delta signaling.

In neur mutants, Delta accumulates on the membranes of signaling cells and Notch activation in neighboring cells is reduced. These results support a role for Neur in endocytosis of Delta in the signaling cells to achieve Notch activation in the neighboring receiving cells, rather than in downregulation of Delta in the receiving cells. Because neur shows strong genetic interactions with lqf and both function in R-cells,Neur and Lqf might work together to stimulate Delta endocytosis. Lqf has ubiquitin interaction motifs (UIMs) that bind ubiquitin(Polo et al., 2002; Oldham et al., 2002). One explanation for how Neur and Faf/Lqf could function together is that Lqf facilitates Delta endocytosis by binding to Delta after its ubiquitination by Neur (Fig. 9B). This is an attractive model that will stimulate further experiments.

One exciting observation is that the endocytic adapter Lqf may be essential specifically for Delta internalization. Although we have not examined these signaling pathways directly, hedgehog, decapentaplegic and wingless signaling appear to be functioning in the absence of Lqf. These three signaling pathways regulate movement of the morphogenetic furrow(Lee and Treisman, 2002) and are thought to require endocytosis (Seto et al., 2002). The furrow moves through lqf-null clones and at the same pace as the surrounding wild-type cells(Fig. 7)(Overstreet et al., 2003). Moreover, all mutant phenotypes of lqf-null clones can be accounted for by loss of Delta function. Further experiments will clarify whether this apparent specificity means that Lqf functions only in internalization of Delta, or if the process of Delta endocytosis is particularly sensitive to the levels of Lqf.

Lqf expands the small repertoire of endocytic proteins that are known targets for regulation of cell signaling. In addition to Lqf, the endocytic proteins Numb and Eps15 (EGFR phosphorylated substrate 15) are objects of regulation. In vertebrates,asymmetrical distribution into daughter cells of the α-adaptin binding protein Numb may be achieved through ubiquitination of Numb by the ubiquitin-ligase LNX (Ligand of Numb-protein X) and subsequent Numb degradation (Nie et al.,2002). In addition, in vertebrate cells, Eps15 is phosphorylated and recruited to the membrane in response to EGFR activation and is required for ligand-induced EGFR internalization(Confalonieri et al., 2000). Given that endocytosis is so widely used in cell signaling, endocytic proteins are likely to provide an abundance of targets for its regulation.

We are grateful to all of the individuals named in the Materials and methods for flies and reagents. We thank our colleagues Paul Macdonald and John Sisson for the use of their confocal microscopes. This work was supported by a grant to J.A.F. from the NIH (CHHDR01-30680). Support for E.O. also came from a University of Texas at Austin Continuing Fellowship. E.F. received support from a University of Texas at Austin Undergraduate Research Fellowship.