Genetic and biochemical analysis of the role of Egfr in the morphogenetic furrow of the developing Drosophila eye

The Drosophila Epidermal growth factor receptor homolog (Egfr) is similar to vertebrate ERBB1 and to Let-23 in C. elegans, and is a type 1 transmembrane protein with intrinsic tyrosine kinase activity(Bogdan and Klambt, 2001; Shilo, 2003). In the developing fly eye, the major positive ligand is Spitz(Freeman, 2002). Like vertebrate EGFR, upon ligand binding the fly protein dimerizes, is activated and trans-autophosphorylates. Inside the cell, the signal is propagated via Ras and a kinase cascade with the final step being the fly homolog of P42/44 mitogen activated protein kinase (MAPK): Rolled(Marshall, 1994; Freeman, 2002; Shilo, 2003). Activated MAPK(pMAPK) translocates to the nucleus, where it can modulate the activities of transcription factors such as Pointed and Anterior open (Yan) in the fly eye(Brunner et al., 1994; O'Neill et al., 1994; Treisman, 1996; Freeman, 2002; Voas and Rebay, 2004).

Patterning in the Drosophila eye imaginal disc begins in the third larval instar when the morphogenetic furrow begins to move from posterior to anterior (Ready et al., 1976). This furrow produces the spaced array of ommatidia, with a new column made roughly every two hours (Ready et al.,1976; Basler and Hafen,1989). The first photoreceptor cell to be specified and to differentiate is R8 (Ready et al.,1976; Tomlinson and Ready,1987). R8 later induces the remaining cells of the facet and is thus also known as the ommatidial founder cell(Tomlinson, 1988; Freeman, 1997).

Coincident with the process of R8/founder cell specification is the expression of the proneural basic helix-loop-helix transcription factor Atonal; in atonal loss-of-function mutants, R8 cells do not form(Jarman et al., 1993; Jarman et al., 1994; Jarman et al., 1995). The level of Atonal first rises in all cell nuclei, then, deep in the furrow, it is then lost from some cells to leave spaced groups (the `intermediate groups'). Finally, Atonal is restricted to the single founder cells, where it persists for a few columns (Baker et al.,1996; Dokucu et al.,1996). R8/founder cell specification and spacing involves the Notch and Hedgehog pathways, through the regulation of atonaltranscription (Cagan and Ready,1989; Baker and Zitron,1995; Baker et al.,1996; Dominguez,1999; Suzuki and Saigo,2000; White and Jarman,2000; Baonza and Freeman,2001; Baker,2004).

In the developing eye, Egfr signaling is required for late R8 cell maintenance, the induction of all cells following the R8/founder cell,proliferation, ommatidial rotation and cell survival(Freeman, 1994; Tio et al., 1994; Freeman, 1996; Freeman, 1997; Tio and Moses, 1997; Dominguez et al., 1998; Kumar et al., 1998; Kurada and White, 1998; Halfar et al., 2001; Brown and Freeman, 2003; Firth and Baker, 2003; Kumar et al., 2003; Strutt and Strutt, 2003; Yang and Baker, 2003). It has also been suggested that Egfr signaling has a primary function in the initial specification and spacing of the Atonal-positive intermediate groups and/or the R8/founder cells based on four observations: pMAPK is strongly expressed in the intermediate groups; the gain-of function mutation Egfr^Elp has reduced numbers of R8 cells; Ras pathway signals promote the expression of the Rough transcription factor in the cells that do not become R8s; and Ras pathway loss-of-function mosaic clones show reduced precision in the array of R8 cells(Baker and Rubin, 1989; Zak and Shilo, 1992; Dokucu et al., 1996; Kumar et al., 1998; Spencer et al., 1998; Baonza et al., 2001; Yang and Baker, 2001; Yang and Baker, 2003). Of these the most direct and elegant evidence is the last (Ras pathway loss-of-function mosaic clones).

Egfr null clones are very small because of an early proliferation defect (Xu and Rubin, 1993). Thus, we generated a conditional Egfr loss-of-function mutation(Egfr^tsla, a temperature-sensitive allele) and used it to remove Egfr function at specific times, by-passing the proliferation defect. We found that while this abolishes pMAPK in the intermediate groups,it does not affect the establishment or spacing of the Atonal-positive R8/founder cells (Kumar et al.,1998). We also observed that the high-level pMAPK antigen in the intermediate groups is predominantly cytoplasmic and proposed that there is a block to pMAPK nuclear translocation (`MAPK cytoplasmic hold')(Kumar et al., 1998; Kumar et al., 2003).

A second approach to overcome the proliferation defect was to use the Minute technique to give Egfr null cells a growth advantage(Baonza et al., 2001; Yang and Baker, 2001). In these clones, the single Atonal-positive cells are irregularly arranged rather than accurately spaced. This observation was made independently by two groups and led to the conclusion that Egfr signaling has a primary role in ommatidial spacing (Baonza et al., 2001; Yang and Baker, 2001).

How may these two views be reconciled? If Egfr signals do not regulate R8 spacing, then the pMAPK pattern may be explained by cytoplasmic hold, the Egfr^Elp phenotype may be an indirect consequence of a gain-of-function mutation, and the regulation of Rough expression in the non-R8 cells could be through a direct Egfr function in these cells, not due to any Egfr function in the R8s [indeed, a rough null has normal early ommatidial development(Tomlinson et al., 1988)]. However, if Egfr signals do regulate R8 cell spacing then the pMAPK pattern may be explained by a direct function, the Egfr^Elpphenotype may be due to over-inhibition of R8 cell formation and Rough expression in the non-R8 cells may be a direct result of Egfr signaling.

The two apparently direct loss-of-function experiments give different results: Egfr^tsla temperature shift versus Egfr-null Minute mosaics. We have now re-examined these two experiments and have tested them for four possible flaws. (1) The Egfr^tsla protein could be temperature-sensitive only during synthesis, so that after temperature shift the previously synthesized protein remains active and provides sufficient function for normal R8 patterning (a form of perdurance). (2) The Egfr^tsla mutation might not be null at the restrictive temperature, and residual function could support normal R8 patterning. (3) At the time of the temperature shift, the furrow in Egfr^tsla eyes might arrest, freezing the Atonal pattern so that it appears to be wild type. (4) Minute⁺ clones may have non-cell autonomous effects on Atonal patterning.

Here, we report the characterization of Egfr^tsla,including its mutant lesion, the effects of temperature on its activity and stability, and the relocalization of the protein from the cell surface into the cell. We show by biochemical and genetic means that Egfr^tsla is temperature-sensitive for activity and that it is functionally null at the restrictive temperature. We also report that, indeed, the Minutetechnique artifactually disturbs ommatidial spacing in the Egfr null mosaic experiments. We conclude that Egfr signaling does not normally function in the initial spacing of the R8/founder cells.

Egfr^tsla/Df(2R)Egfr⁵ and cn bw.

y¹ w¹¹¹⁸ ey:FLP;P{ry⁺=neoFRT}42D P{w^+mc=Ubi:GFP.nls}2R1P{Ubi:GFP.nls}2R2/P{ry⁺=neoFRT}42D Egfr^tsla.

y¹ w¹¹¹⁸ ey:FLP;P{ry⁺=neoFRT}42D P{w^+mc=Ubi:GFP.nls}2R1P{Ubi:GFP.nls}2R2/P{ry⁺=neoFRT}42D Egfr^f2.

P{ry^+t7.2=hsFLP}1 w¹¹¹⁸;P{ry⁺=neoFRT}42D Egfr^f2/P{ry^+t7.2= neoFRT}42D P{w^+mC=arm-lacZ.V}51D M(2)53.

P{ry^+t7.2=hsFLP}1 w¹¹¹⁸;P{ry⁺=neoFRT}42D Egfr^tsla/P{ry^+t7.2= neoFRT}42D P{w^+mC=arm-lacZ.V}51D M(2)53.

P{ry^+t7.2=hsFLP}1 w¹¹¹⁸;P{ry⁺=neoFRT}42D Egfr^top-18A/P{ry^+t7.2= neoFRT}42D P{w^+mC=arm-lacZ.V}51D M(2)56i.

P{ry^+t7.2=hsFLP}1 w¹¹¹⁸;P{ry⁺=neoFRT}42D Egfr^tsla/P{ry^+t7.2= neoFRT}42D P{w^+mC=arm-lacZ.V}51D M(2)56i.

P{ry^+t7.2=hsFLP}1 w¹¹¹⁸;P{ry⁺=neoFRT}42D/P{ry^+t7.2 =neoFRT}42D P{w^+mC=arm-lacZ.V}51D M(2)53.

PCR primer pairs were as follows.

Exon 1, type I: forward primer (fp), ATAGCTTGGAAGAGGGCTTGATTC; reverse primer (rp), TGGGCAGCCAGTCAGTCAGCCAGA.

Exon 1, type II: fp, TTGACTAGGCACCACAACGCCACC; rp,CAATATTTATGTGCCACATTTCAG;

Exon 2: fp1, GGGTCAACCGAAACAACAAAACCCATT; fp2, TGCATCGCCACTAAATCTCGG; rp1,GTACAGGGTATTAGGTCTAGTCCA; rp2, CTTGGGAAATACCACTTTCTT.

Exon 3: fp1, CTACAAATCATTCGCGGACGC; fp2, CCCAAGTGCCACGAGAGCTGC; fp3,CCGCCCATGCGCAAGTACAAT; fp4, GTCCTGCATGCCGGCAACATT; fp5,GGAACCCACCCGCAGTCCCGGAAT; fp6, TGGCCGGCCATTCAGAAGGAG; rp1,GCACTGATCCGAGCAAATGGT; rp2, TCTATTATGCTGGATGACCAC; rp3, GATCTCCTTCACCGTGGAGAA;rp4, GGGGCACGGTCCATTGCAGGG; rp5, CTCCTTGCATACGCCCTCGTC; rp6,ACACTCGCGCTCCGGAGCAGT.

Exon 4: fp1, GAGAAAAATGGAACCATTTGCTCG; fp2, GCACTGATCCGAGCAAATGGT.

Exon 5: fp1, GTCTGCCGAGAGTGCCACGCC; fp2, GAACAGGTGTGCTCCAAGTGC; fp3,GCCATTGGACCCTACTGTGCA; fp4, GCCAATCTATGCAAGTTGCGC; fp5, GTGCGAAATAACCGGGACAAG;fp6, GCACTGGAGTGCATTCGCAAT; fp7, TGGCACTTGGATGCCGCCATG; fp8,ACCGACTGCACGGATGAGATA; fp9, AATATGGCAGCTGTGGGCGTG; rp1,GACTCCCATCACCGGTATCCCGAT; rp2, CACGCCCACTTCCCGGGCACT; rp3,AATGGCCAGATAGCGACCCGG; rp4, CACACCAAAGGCCCAGACATC; rp5, ACCCTCCGGCACCCAAACGCC;rp6, CACCAGAACAGCACCAGTGATGTGATCGGCCGGACACTCGGT.

Sequencing was done at the University of Iowa facility.

Two rabbits were injected with an Egfr C-terminal peptide[(C)QRELQPLHRNRNTETR] (Lesokhin et al.,1999), coupled to keyhole limpet hemocyanin. Sera were affinity purified (by Zymed). For imaging, S2 cells were cultured on cover slips after transfection and prepared as previously described(Lee et al., 2001). Eye discs(except those used for the anti-Egfr stains) were prepared as previously described (Tomlinson and Ready,1987), modified as described by Tio and Moses(Tio and Moses, 1997), mounted in Vectashield (Vector Labs, H-1000) and imaged by confocal microscopy. For Egfr staining, the fix was 4% paraformaldehyde in PBS for 30 minutes at room temperature; eye discs were then washed and blocked as above, then incubated with primary antibody overnight at room temperature(Moberg et al., 2004). Primary antibodies: rabbit anti-Egfr [1:3000 for blots, gift of N. Baker(Lesokhin et al., 1999)],rabbit anti-Egfr (1:100 for immunoprecipitation, 1:500 for immunohistochemistry), rat anti-Spitz [1:20(Schweitzer et al., 1995)],mouse anti pTyr (PY20, 1:500, Santa Cruz Biotechnology SC-508), mouse anti-pMAPK [1:5000 for blots, 1:500 for immunohistochemistry, Sigma M-8159(Gabay et al., 1997)], rabbit anti-MAPK (anti-ERK, 1:1000, Cell Signaling Technologies 9102), rabbit anti-Ato [1:1000 (Jarman et al.,1993)], rat anti-Elav [1:500, 7E8A10 from DSHB(O'Neill et al., 1994)],guinea-pig anti-Senseless [1:1000, gift of G. Mardon(Nolo et al., 2000)], mouse anti-β-gal (1:1000, Promega 23783), guinea-pig anti-Hrs [1:100, gift from Ursula Weber (Lloyd et al.,2002)], mouse anti-Armadillo [1:10, DSHB N2 7A1(Riggleman et al., 1990)]. Secondary antibodies were mainly from Jackson ImmunoResearch: goat anti-mouse Cy5 (1:500, 115-175-003), goat anti-rabbit TRITC (1:250, 111-025-003), goat anti-rat Cy5 (1:200, 112-175-003), goat anti-mouse HRP (1:40, 115-035-003),goat anti-guinea pig TRITC, (1:150, 106-025-003). Goat anti-rabbit-HRP(1:8,000, 65-6120) was from Zymed. Syto-24 was used to stain DNA (1:10,000,Molecular Probes S-7559).

The Egfr⁺ plasmid pMtEgfrType I was a gift of N. Baker(Lesokhin et al., 1999). The Egfr^tsla plasmid pMtEgfr^tslaType I was derived from pMtEgfrType I using a Stratagene `QuickChange' Site-Directed Mutagenesis Kit. The secreted Spitz plasmid is pMTsSpitz, a gift of B.-Z. Shilo(Schweitzer et al., 1995). Schneider line 2 (S2) cells were maintained in Drosophila Schneider's medium (Invitrogen 11720-034) with 10% heat-inactivated Fetal Bovine Serum at 25°C. Transfections were performed with Cellfectin (Invitrogen 10362-010)in serum-free medium for 4 hours 25°C. DNA concentrations used for transfections were: 2-3 μg/ml for wild-type and mutant Egfrplasmids, or 12 μg/ml for pMtsSpitz. Egfr transfections were incubated at 18°C for 15-16 hours in serum + medium, then for 24 hours in serum-free medium with 100 μM CuSO₄. Fresh medium containing 10μg/ml cycloheximide (CHX) was added and the cells were then treated as described in the text. We tested cycloheximide between 0 and 100 μg/ml, as assayed by 35S-methionine incorporation (data not shown). Spitz transfections were incubated at 25°C, following the same protocol except for the use of 0.7 mM CuSO₄. Spitz supernatants were validated by immunoblot (data not shown). Cells were lysed with RIPA buffer as described previously(Schweitzer et al., 1995). Gel blots were performed by standard protocols (BioRad). Immunoprecipitation was performed as previously described(Schweitzer et al., 1995).

For surface biotinylation, the transfections and temperature shifts were as above, the cells were then incubated with 0.5 mg/ml EZ-Link™sulfo-NHS-Biotin (Pierce, 21217) at 4°C for 30 minutes and the extract was precipitated with streptavidin beads (Pierce 2349) as described previously(Salazar and Gonzalez, 2002). Quantification was by the following steps. (1) Densitometry from non-saturated films to measure total band intensities (integrated over a standard area) for Egfr antigen from the untransfected control cells and from those transfected for Egfr⁺ and Egfr^tsla, and for pMAPK.(2) Track background (measured below each band) was subtracted from each total band density. (3) The untransfected control value was subtracted from each experimental transfection (Egfr⁺ and Egfr^tsla). (4) The pMAPK value was divided by the Egfr antigen value (for each track) to give the activity per receptor. (5) Values were standardized to make the activity of Egfr⁺ at 30°C 100;all other values are percentages of wild type. (6) Three replicates were used for the mean and the standard error of the mean.

We sequenced Egfr^tsla and found a single coding change relative to the parent chromosome: the transition C1,754T [using the numbering system of Clifford and Schüpbach(Clifford and Schüpbach,1994)] produces a missense S511F mutation (in the type I isoform). The change lies in the extracellular, conserved L2 domain, which is a leucine-rich repeat that functions in ligand binding(Clifford and Schüpbach,1994; Burgess et al.,2003). S511 is not conserved in either vertebrate or nematode Egfr homologs, but lies in a variable region that functions in ligand specificity(Clifford and Schüpbach,1994; Burgess et al.,2003). Three other temperature-sensitive alleles of Egfrhave been reported, although, unlike Egfr^tsla, none have strong phenotypes at the restrictive temperature. Two of these lesions lie in the intracellular kinase domain, and one (Egfr^SH2) lies in the L1 domain [L206Q in the type I isoform(Clifford and Schüpbach,1994)].

To determine if the Egfr^tsla protein is temperature sensitive only during its synthesis or for its activity, we developed a tissue culture assay system using Drosophila Schneider line 2 cells (S2 cells)[Fig. 1A, assay adapted from Schweitzer et al. and Lesokhin et al.(Schweitzer et al., 1995; Lesokhin et al., 1999)]. These cells express low levels of endogenous Egfr, so all of the experiments were done in parallel with untransfected controls. We transfected cells to express additional Egfr⁺ or Egfr^tsla at 18°C (the permissive temperature). One day later, we shifted the cells to 30°C and added cycloheximide to inhibit new protein synthesis. One hour later, we added conditioned media containing Spitz and harvested the cells at different time points. The levels of Egfr⁺ antigen remained fairly constant after 60 minutes with cycloheximide, suggesting that the wild-type protein has a much longer half-life than one hour under these conditions;Egfr^tsla antigen was also detectable, although less so than wild type (Fig. 1B). In this assay,Egfr⁺ drives a ligand-dependent, high level of pMAPK within 5 minutes (asterisk in Fig. 1C),whereas Egfr^tsla does not.

As a second test for receptor activity we measured autophosphorylation. Again, we transfected S2 cells at 18°C for Egfr⁺ or Egfr^tsla expression. The cells were grown at either 18°C or 30°C for one hour and then treated with ligand(Fig. 1D). We assayed anti-Egfr immunoprecipitates for phospho-Tyrosine(Fig. 1E,F). We detected phosphorylated Egfr for both Egfr⁺ and Egfr^tsla at 18°C, but only for Egfr⁺ at 30°C (white asterisks). Egfr antigen was detected under all conditions(Fig. 1G,H); however, we detected reduced levels of Egfr^tsla at 30°C, perhaps through instability (see below).

We previously reported that Egfr^tsla mutants loose pathway activity within 30 minutes at 29°C in vivo(Kumar et al., 1998). We therefore determined whether Egfr^tsla becomes inactive in this system in a similar time frame after temperature shift. S2 cells were treated as described above, shifted to 30°C (for 0, 15, 30 or 60 minutes, Fig. 2A) and then treated with ligand. We detected reduced levels of Egfr^tsla antigen after the temperature shift. We always detected robust levels of MAPK phosphorylation driven by both Egfr⁺ and Egfr^tsla at 18°C. However,Egfr^tsla lost this activity within 15 minutes at 30°C in all experiments (Fig. 2B-D).

From these experiments, we conclude that Egfr^tsla protein is wild type at 18°C for ligand-dependent signaling activity. However, at 30°C, Egfr^tsla becomes inactive within 15 minutes. Furthermore,this is due to the temperature sensitivity of the protein after synthesis and this effect is rapid.

We quantified the experiments in Fig. 2, at the 0- and 60-minute time points. Each condition was repeated in triplicate and two such experiments were quantified per receptor at 30°C (see Materials and methods and Fig. 3). From the two experiments, we conclude that the activity of Egfr^tsla is wild type at 18°C but that at 30°C its activity is very low, and is indistinguishable from no activity at all.

Furthermore, we directly visualized the activity of Egfr in individual cells by immunofluorescence (Fig. 4). We examined cells at the 60-minute time point, as above(Fig. 2A). Untransfected cells show low background levels of Egfr antigen(Fig. 4A,D) and a low level of pMAPK antigen (Fig. 4B,D). Transfected cells express elevated levels of Egfr antigen(Fig. 4E-P). Cells that express additional Egfr⁺ at 30°C (white arrowheads in Fig. 4E-H) and 18°C (not shown), and Egfr^tsla at 18°C(Fig. 4I-L), drive increased levels of nuclear pMAPK antigen (arrows in Fig. 4F,H,J,L), but cells expressing Egfr^tsla at 30°C do not(Fig. 4N,P).

All of these biochemical and cell biological experiments show that Egfr^tsla is a mutation that is temperature sensitive for activity. The level of ligand-dependent Egfr^tsla activity at the permissive temperature is not distinguishable from that of the wild type, and the activity at the restrictive temperature is indistinguishable from no activity at all. In addition, these experiments show that the effect of the temperature shift is rapid, occurring within 15 minutes. These results are consistent with the in vivo study we published previously, which showed that Egfr^tsla mutants rapidly loose pMAPK antigen from the morphogenetic furrow (Kumar et al.,1998). We conclude that Egfr^tsla is indeed a rapidly acting mutation that is temperature sensitive for activity. Egfr^tsla is effectively wild type at 18°C and effectively null at 30°C.

We detected reduced levels of Egfr antigen in cells transfected with Egfr^tsla at 30°C (see above). It could be that after temperature shift, the Egfr^tsla protein is removed from the cell surface and targeted for degradation. To study this, we labeled transfected S2 cells with a non-membrane permeable biotinylation reagent to separate superficial from intracellular Egfr (Fig. 5A). We compared biotinylated Egfr antigen (see asterisk in Fig. 5B) to total Egfr antigen(see asterisk in Fig. 5C). We detected biotinylated Egfr at all time points from cells transfected with Egfr⁺. Biotinylated Egfr^tsla is also seen at 18°C, but is rapidly lost so that none is detectable after 60 minutes at 30°C (Fig. 5B), and a low level of total antigen remains (Fig. 5C). These results strongly suggest that Egfr^tslaprotein undergoes a ligand-independent conformational change at 30°C,leading to rapid internalization.

To investigate this relocalization in vivo, we raised a new anti-Egfr serum[to the same C-terminal peptide as previously reported(Lesokhin et al., 1999)]. We tested it for specificity by staining eye imaginal disc Egfr^f2 homozygous clones (data not shown). We stained wild-type eye discs and saw an Egfr expression pattern as has been previously described (data not shown) (Lesokhin et al., 1999). We see the wild-type pattern in Egfr^tsla homozygous retinal clones raised at 18°C(Fig. 6A,B). In the furrow, we observed antigen concentrated in a subset of the Armadillo-positive cell junctions and the first two columns of ommatidial preclusters (see white arrrowhead in Fig. 6B)(Takahashi et al., 1996; Ahmed et al., 1998). This junctional stain occurs around all sides of the cells deep in the furrow (see white arrrowhead in Fig. 6B),but, at the edges of the furrow, it becomes lost from the faces of the cells that lie away from the furrow. The junctions between the cells within the precluster are heavily stained (see white arrrowhead in Fig. 6B, inset); however, the junctions between the precluster cells and the surrounding cells are not (see black arrrowhead in Fig. 6B,inset). We also saw numerous cytoplasmic granules in the assembling ommatidia(see arrows in Fig. 6A,B).

Thus, we find that Egfr antigen is normally strongly localized in the furrow to a specific subset of cell junctions: essentially the cells held in G1 cell-cycle arrest (Ready et al.,1976; Wolff and Ready,1991; Thomas et al.,1994). This observation is intriguing, but the pattern does not correlate precisely with known Egfr signaling activity (from pMAPK staining). Egfr junctional staining is seen within and between the Atonal-positive intermediate groups, whereas pMAPK staining is seen only within them(Gabay et al., 1997; Kumar et al., 1998; Spencer et al., 1998). Thus,the biological significance of this pattern is presently unclear.

When we shift these discs to the non-permissive temperature, the junctional antigen in the furrow and early preclusters is rapidly converted to larger granules: partially after 15 minutes at 30°C (and then up to 15 minutes dissection time at room temperature, Fig. 6D and inset), and completely by 30 minutes (and then up to 15 minutes dissection time at room temperature, Fig. 6F). Under these conditions Armadillo staining is unaffected(Fig. 6G). Thus, although Egfr antigen may co-localize with Armadillo in some of the Armadillo-positive junctions (furrow and early preclusters), Armadillo antigen distribution is not dependent on Egfr function (at least not in the short term). In mammals,ligand bound EGFR has been shown to pass through an endocytic pathway(Carpenter, 2000), and the Drosophila Hepatocyte growth factor regulated tyrosine kinase substrate (Hrs) protein affects trafficking so that active Egfr is degraded(Lloyd et al., 2002; Weber et al., 2003). We co-stained Egfr^tsla homozygous retinal clones after 30 minutes at 30°C for Egfr and Hrs antigens(Fig. 6H) and found no significant co-localization. This suggests that the Egfr antigen is removed from the Armadillo-positive junctional domains following temperature shift,but that this is not via a pathway that involves Hrs.

These results are consistent with the biotinylation data described above. The simplest interpretation is that the Egfr^tsla protein undergoes a conformational change upon temperature shift, leading to ligand-independent internalization via a non-signaling pathway.

Egfr null mosaic clones are small(Xu and Rubin, 1993), so we compared the sizes of Egfr^tsla clones to Egfrnull clones. We used ey:Flp to induce Egfr^tsla clones and raised the animals continuously at 18°C (Fig. 7A), at 18°C and then at 30°C for their last day(Fig. 7B), or continuously at 30°C (Fig. 7C). The total area of the Egfr^tsla clone territory was roughly equal in size to the total area of wild-type twin-spots when raised continuously at 18°C and after one day at 30°C (although they are reduced towards the posterior side). However, the clones are very much smaller after continuous growth at 30°C. Likewise, Egfr homozygous null clones made without Minute mutations are very small and are similar to Egfr^tsla homozygous clones after continuous growth at 30°C (compare Fig. 7C and 7D).

Egfr is required for late ommatidial development, so we stained Egfr^tsla clones (raised at 30°C for 24 hours) for a late R8 marker, Senseless (Fig. 7E,F) (Frankfort et al.,2001), and a neural marker, Elav(Fig. 7G,H)(Robinow and White, 1991). As previously reported (Dominguez et al.,1998; Kumar et al.,1998; Spencer et al.,1998; Baonza et al.,2001; Yang and Baker,2001), we find that R8 cells do form, although at later stages they are sometimes abnormally close (Fig. 7E,F). Elav expression is lost from the non-R8 photoreceptors(Fig. 7G,H) and also from most of the R8s, because of a late maintenance requirement(Kumar et al., 1998).

Thus, by these proliferation and differentiation defects, Egfr^tsla (at the restrictive temperature) is indistinguishable from nulls.

Thus far, we have tested for the first two of the four possible artifacts defined above: Egfr^tsla is temperature sensitive for activity(there is no perdurance of activity), and it is biochemically and genetically indistinguishable from a null. The third possible artifact was that the normal Atonal expression pattern in temperature-shifted Egfr^tslaeye discs might result from a rapid arrest of furrow movement with no further change in the Atonal pattern. To test this, we stained Egfr^tsla homozygous clones that were raised at 18°C and held at 30°C for 24 hours for Atonal. In this experiment, the Egfr^tsla territories abut Egfr⁺ twin spots. If the loss of Egfr function causes the furrow to arrest with `frozen'Atonal expression, we would see the furrow about twelve columns more advanced in the wild-type tissue than in the mutant clones. As before(Kumar et al., 1998), we saw normal Atonal expression in this mutant tissue. We also saw that the position of the furrow was equally advanced in the mutant clones and in the wild-type twin spots (Fig. 8A-C). Thus,we conclude that Egfr does not regulate the rate of furrow progression. We also confirm again that Egfr^tsla has no effect on R8/founder cell initial spacing, neither with a 24-hour shift(Fig. 8A-C), nor when raised continuously at the non-permissive temperature(Fig. 8D-F).

Having investigated the possible artifacts of the Egfr^tsla studies, we turned to the fourth possibility:some artifact in the Egfr null Minute mosaic experiments. It could be that the Egfr mutations used in these experiments have some dominant effects on Atonal expression under these conditions (24 hours at 30°C). Therefore, we stained eye discs heterozygous for three Egfr alleles in trans to wild type (Egfr^tsla,Egfr^f2 and Egfr^top-18A, Fig. 9A-C), and we found that there was some variation in the Atonal antigen staining intensity in different specimens, but that this does not correlate with the three Egfrgenotypes. In all three cases, Atonal expression was normal.

It could be that Minute mutations have dominant and/or non-cell autonomous clone effects on Atonal expression and R8/founder cell patterning. Indeed, it has been reported that a Minute mutation affects gene expression and cell growth in competing, wild-type clones(de la Cova et al., 2004). To test for this and other possible artifacts, we repeated the Egfr null experiments, exactly as previously reported, using stocks kindly supplied by those investigators. We placed Egfr^f2 in trans to Minute(2)53 (Baonza et al.,2001), and we placed Egfr^top-18A in trans to Minute(2)56i (Yang and Baker,2001). We also placed Egfr^tsla in trans to both Minute(2)53 and Minute(2)56i. All four genotypes were raised at 18°C until the early second instar and then clones were induced with hs:Flp (90 minutes at 37°C). Thereafter, the animals were raised at 30°C and eye discs were stained for Atonal. In all four cases(Fig. 9D-O), we found that the pattern of Atonal expression was not wild type (compare with Fig. 9A-C). In these cases, the Egfr^tsla clones were indistinguishable from the clones of the two Egfr null alleles made with the same Minutemutations (compare Fig. 9D-F with 9G-I, and compare 9J-L with 9M-O). Thus, by this phenotype also, Egfr^tsla raised at 30°C is indistinguishable from the nulls.

Furthermore, we find that the pattern of Atonal expression is different to that of wild type, both within the Egfr homozygous mutant territories and outside of them (in the Egfr Minute heterozygous cells). These abnormalities include the occasional twinning of Atonal-positive cells(Fig. 9E,F), reduced differentiation of the intermediate groups (all four cases) and reduced Atonal expression in some cases (arrows in Fig. 9K,L,N,O). Taken together, these data suggest that some factor in both Minute experiments causes dominant and non-cell autonomous defects in the pattern of Atonal expression and R8/founder cell/ommatidial spacing.

How could the use of the Minute mutations produce these effects?The simplest possibility is that Minute(2)53 and Minute(2)56i have a dominant effect on Atonal patterning on their own. Indeed, as well as affecting body size, developmental time and bristle morphology, some Minute mutants are reported to have dominant rough eye phenotypes (Plough and Ives,1934; Dunn and Mossige,1937; Brehme,1939; Kalisch and Rasmuson,1974; Sinclair et al.,1981). However, neither Minute(2)53 nor Minute(2)56i show a dominant rough eye phenotype. Furthermore, we stained the heterozygous stocks for Atonal expression and saw no defects (data not shown). Thus, the Atonal defects are not due to a simple dominant effect of the Minute mutations.

It could be that the phenotype is due to a dominant synthetic interaction between the Minute mutations and Egfr. Again, this is not probable, as the Egfr mutant clones are Minute⁺after the somatic recombination. We stained animals that were Egfrmutant in trans to the two Minute mutations (without inducing clones)for Atonal expression and again saw no defects (data not shown). Thus, the Atonal defects are not due to a dominant synthetic interaction between the Minute mutations and Egfr.

It may be that cell competition effects between the Minute⁺ and Minute heterozygous territories produce the Atonal patterning defects. Such competition effects have been reported in other Minute experiments(de la Cova et al., 2004). Another possibility is that the dying Minute^– cells release developmental signals that affect retinal patterning in the surrounding tissues. Indeed, it has been reported that dying cells release both Wingless and Dpp (Perez-Garijo et al., 2004; Ryoo et al.,2004). If either of these possibilities is true, then we would expect to see defects in the Atonal expression pattern when clones are induced from Minute heterozygous flies, even without any Egfrmutation present in trans. In other words, fully wild-type clones adjacent to Minute heterozygous territories would have Atonal defects similar to those seen in the Egfr mutant clone-containing discs. We stained such wild-type clones for Atonal expression, and do indeed see just such defects,both within and outside of the clones (Fig. 9P-R). We observed a total of fifteen such clones in the Atonal expression domain, taken from five imaginal discs. Indeed, the Atonal pattern in these Minute (alone) clones is indistinguishable from that in clones for the same Minute with Egfr (compare Fig. 9Q to 9E,H).

Our data suggest that the Atonal patterning defects seen in mosaic clone experiments, when the Minute technique is used, may be due to a genetic effect of the Minute, and not of the Egfr,mutation.

We characterized the lesion in Egfr^tsla and found that it is a missense mutation in the conserved ligand-binding, extracellular L2 domain(Burgess et al., 2003). Our biochemical and localization data suggest that the Egfr^tsla protein functions normally at 18°C as a ligand-activated receptor. However, after shift to the non-permissive temperature, Egfr^tsla rapidly becomes inactive and is removed from the cell surface, probably via a non-signaling endocytic pathway (with or without ligand). It may be that the Egfr^tsla protein is conformationally unstable at 30°C and is degraded. Human EGFR is normally only internalized in response to ligand binding (Burke et al., 2001),but it has been reported that inhibition of PKA leads to internalization of unbound EGFR (Salazar and Gonzalez,2002). Our mutation in the extracellular L2 domain suggests that this domain may be involved in mediating the stability of Egfr in the membrane.

Atonal expression and R8/founder cell spacing are normal in Egfr^tsla eye discs incubated at the non-permissive temperature. We examined three possible artifacts that could have invalidated this observation. (1) Egfr^tsla could be temperature sensitive for synthesis but not activity. If so, then protein made before the shift to 30°C might continue to supply sufficient function at the non-permissive temperature to support normal R8/founder cell development. However, we have shown that Egfr^tsla is in fact temperature sensitive for activity,and that activity is lost within minutes of the temperature shift, while Atonal expression and R8/founder cell formation continues normally for 24 hours. (2) Egfr^tsla could be leaky (i.e. not null) at 30°C, and some residual activity might supply sufficient function at the non-permissive temperature to support normal R8/founder cell development. We have shown before that Egfr^tsla mutants (at the non-permissive temperature) are genetically indistinguishable from a null(Egfr^f24) for three phenotypes(Kumar et al., 1998). Here, we have shown in addition that Egfr^tsla mutants at 30°C are phenotypically indistinguishable from nulls (Egfr^f2and Egfr^top-18A), in Minute⁺ mosaic clones, as well as in their growth deficits in clones made without Minute mutations. Furthermore, we have undertaken quantitative biochemical experiments using S2 cells to show that Egfr^tsla at 30°C is indistinguishable from a null in its ability to drive both the phosphorylation of MAPK and its own autophosphorylation. (3) It could be that,at 30°C, the furrow just freezes in Egfr^tsla mutant eyes, leaving an arrested but apparently normal furrow. Here, we have shown that in adjacent Egfr^tsla and Egfr⁺territories, the furrow advances at the same rate after the temperature shift to 30°C. Thus, we conclude that our observation of normal Atonal and R8/founder cell patterning in Egfr^tsla mutant eyes is not invalidated by any of these three possible artifacts.

Next, we turned to the possible problems associated with the Minute⁺ mosaic method used to make the same observations with the Egfr nulls (Baonza et al., 2001; Yang and Baker,2001). We replicated the Minute⁺ Egfr null experiments exactly as previously reported, using the same Drosophilastocks, except to run them at 30°C. In parallel, we did the same experiments with Egfr^tsla. If the known Egfrnulls were to have a different phenotype to Egfr^tsla, we would be forced to conclude that Egfr^tsla is not behaving as a null in this assay. If, however, the two Egfr nulls have the same phenotype in this assay as Egfr^tsla does, then we could conclude that the temperature-sensitive allele is indistinguishable from the nulls at 30°C, and that the difference is due to the Minutetechnique and not to Egfr. Indeed, we did find that the Egfr^tsla and Egfr null phenotypes are indistinguishable, and thus we conclude that the Egfr^tslaphenotypes assayed without the Minute technique are valid and that the discrepancy stems from some aspect of the use of the Minutemutations. We went on to stain Minute clones made without any Egfr mutation present and obtained the very same Atonal expression defects. Taken together, these data suggest that the spacing defects previously reported by others, and replicated by us here, are genetically dependent on the presence of the Minute clones, and are not an affect of the Egfr mutations.

Therefore, we conclude that Egfr has no primary role in R8/founder cell spacing and also that Egfr^tsla is a rapidly acting temperature-sensitive mutation that is functionally null at the non-permissive temperature.

We thank A. D'Costa, V. Faundez, D. Marenda, K. Moberg, M. Powers, E. Rogers, A. Vrailas and D. Zarnescu for their helpful comments; D. Ritsick and G. Salazar for intellectual contributions; R. Boyd and J. Kumar for assistance with sequencing; and N. Baker, G. Mardon, B.-Z. Shilo and U. Weber for their gifts of reagents. This work was supported in the Moses laboratory by a grant from the National Eye Institute (EY12537).