Rough eye Is a Gain-of-Function Allele of amos That Disrupts Regulation of the Proneural Gene atonal During Drosophila Retinal Differentiation

Corresponding author: Françoise Chanut, S1334, Box 0452, University of California, 513 Parnassus Ave., San Francisco, CA 94143., chanut{at}itsa.ucsf.edu (E-mail)

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The regular organization of the ommatidial lattice in the Drosophila eye originates in the precise regulation of the proneural gene atonal (ato), which is responsible for the specification of the ommatidial founder cells R8. Here we show that Rough eye (Roi), a dominant mutation manifested by severe roughening of the adult eye surface, causes defects in ommatidial assembly and ommatidial spacing. The ommatidial spacing defect can be ascribed to the irregular distribution of R8 cells caused by a disruption of the patterning of ato expression. Disruptions in the recruitment of other photoreceptors and excess Hedgehog production in differentiating cells may further contribute to the defects in ommatidial assembly. Our molecular characterization of the Roi locus demonstrates that it is a gain-of-function mutation of the bHLH gene amos that results from a chromosomal inversion. We show that Roi can rescue the retinal developmental defect of ato¹ mutants and speculate that amos substitutes for some of ato's function in the eye or activates a residual function of the ato¹ allele.

THE compound eye of Drosophila melanogaster is a regular array of 800 ommatidia, each composed of 8 photoreceptor cells (R1–R8) and 12 accessory cells (WOLFF and READY 1993 ). The development of each ommatidium begins with the specification of an R8 photoreceptor precursor (TOMLINSON and READY 1987A , TOMLINSON and READY 1987B ). The R8 precursor acts as a founder cell around which all other ommatidial cell types are progressively recruited through a series of inductions mediated by Sevenless (Sev) and the epidermal growth factor receptor (EGFR; TOMLINSON et al. 1987 ; FREEMAN 1997 ). Retinal differentiation begins at the posterior margin of the eye-antennal disc in early third instar larvae and proceeds as a wave that reaches the anterior disc margin by the early pupal stage, 48 hr later (WOLFF and READY 1993 ). At the front of the differentiation wave, an indentation in the epithelium known as the morphogenetic furrow (MF) marks the transition between proliferating, undifferentiated cells in the anterior and differentiating cells in the posterior. In the MF, cells are arrested at the G1 phase of the cell cycle (READY et al. 1976 ; TOMLINSON 1985 ; WOLFF and READY 1993 ). Retinal neurogenesis begins with the broad expression of the proneural gene atonal (ato; JARMAN et al. 1993 ) in all cells of the furrow's anterior edge, followed by the progressive restriction of ato in the MF, eventually leading to the resolution of single ato-expressing R8 precursors at the furrow's posterior edge (JARMAN et al. 1994 ; DOKUCU et al. 1996 ; SUN et al. 1998 ). ato encodes a basic-helix-loop-helix (bHLH) protein with high sequence similarity to proneural proteins of the Achaete-Scute family (JARMAN et al. 1993 ). Like other members of this group, Ato is thought to exert its proneural function via its dimerization with the bHLH protein Daughterless (Da; JARMAN et al. 1993 ; BROWN et al. 1996 ). In spite of its broad expression pattern, ato is only strictly required for the specification of the R8 precursors (JARMAN et al. 1994 ) and for some aspects of their differentiation into R8 photoreceptors (WHITE and JARMAN 2000 ). Nevertheless, in the absence of ato function, all retinal cell types are missing because of the absence of ommatidial founders (JARMAN et al. 1994 ). Whether additional proneural genes specify the other photoreceptor fates is currently unknown.

Expression of ato is tightly controlled temporally and spatially, which reflects complex patterning mechanisms that ultimately ensure the regular spacing of nascent ommatidia behind the MF (HEBERLEIN and MOSES 1995 ; BRENNAN and MOSES 2000 ). ato is activated in the MF by the diffusing factor Hedgehog (Hh; DOMINGUEZ and HAFEN 1997 ; STRUTT and MLODZIK 1997 ; BOROD and HEBERLEIN 1998 ; DOMINGUEZ 1999 ), which is synthesized by photoreceptors differentiating behind the MF (HEBERLEIN et al. 1993 ; MA et al. 1993 ) and repressed anterior to the MF by the neuronal inhibitors hairy (h) and extramacrochaete (emc; BROWN et al. 1995 ). The transmembrane receptor Notch (N; WHARTON et al. 1985 ) and its ligands Delta (Dl; KOPCZYNSKI et al. 1988 ) and Scabrous (Sca; LEE et al. 1996 ; POWELL et al. 2001 ) then pattern ato's profile via a process of lateral inhibition: ato expression is first restricted to evenly spaced intermediate groups at the posterior edge of the MF and eventually to single R8 precursors that emerge at regular intervals behind the MF (BAKER and ZITRON 1995 ; BAKER et al. 1996 ; SUN et al. 1998 ; reviewed in BRENNAN and MOSES 2000 ). Resolution and spacing of single R8 precursors are in addition controlled by the homeodomain protein Rough (Ro; TOMLINSON et al. 1988 ; HEBERLEIN et al. 1991 ) that keeps ato transcription repressed in cells other than R8 (DOKUCU et al. 1996 ). ato also regulates its own expression and is required to maintain high levels of Ato protein in the intermediate groups and R8 precursors (SUN et al. 1998 ).

A number of mutations are known to disrupt the regular arrangement of the eye facets, causing roughening of the eye surface. While many affect the recruitment or specification of the various ommatidial cell types induced by the R8 founders (reviewed in ALBAGLI et al. 1997 ; KUMAR and MOSES 1997 ), several have been traced back to early patterning defects in the MF (BAKER et al. 1990 ; BAKER and RUBIN 1992 ; CAGAN 1993 ; THOMAS et al. 1994 ). Rough eye (Roi) has been known for many years as a dominant mutation that causes roughening of the eye surface (RENFRANZ and BENZER 1989 ; HEBERLEIN et al. 1993 ). On the basis of their observation of adult retinal sections and MF cell morphology, RENFRANZ and BENZER 1989 proposed that Roi disrupted early patterning in the MF. Consistent with this proposal, we previously reported that Roi acts as a strong suppressor of two mutations that cause a premature arrest of furrow progression, hh^bar3 and ro^Dom (HEBERLEIN et al. 1993 ). We showed that Roi restored the anterior progress of retinal differentiation and the expression of a furrow-specific reporter gene in both mutant backgrounds. hh^bar3 is a hypomorphic allele with a regulatory region mutation that abolishes hh expression in the eye (LEE et al. 1992 ; HUANG and KUNES 1996 ), leading to insufficient ato expression in the MF (F. CHANUT and U. HEBERLEIN, unpublished observation) and eventually to the arrest of the furrow (CHANUT and HEBERLEIN 1997 ). ro^Dom is a gain-of-function mutation that causes an anterior expansion of the domain of ro expression, leading to furrow arrest via the progressive repression of ato expression in the MF (CHANUT et al. 2000 ). To understand how Roi might restore furrow progression in both of these backgrounds, we decided to characterize the molecular nature of the Roi mutation.

Here we describe our phenotypic and molecular characterization of the Roi mutation. We find that Roi disrupts ato patterning and increases hh expression behind the MF. We show that Roi is linked to a genomic inversion between cytological positions 36A and 37A that causes misexpression of the proneural gene amos in the eye. We show that experimental overexpression of amos in eye discs mimics the Roi phenotype and that Roi rescues retinal differentiation in homozygous ato mutants. We discuss various mechanisms for the effects of Roi and amos on ato expression and retinal patterning.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks and culture:
Roi arose spontaneously in In(2L)t (22D3–E1; 34A8–9) but has also been introduced on the CyO balancer (LINDSLEY and ZIMM 1992 ). For most genetic interactions and mapping purposes, we used a Roi, CyO chromosome obtained from the Bloomington Stock Center. For the analysis of the Roi phenotype in somatic clones, we used a chromosome where Roi had been separated from other rearrangements by homologous recombination (K. HARSHMAN and D. BALLINGER, unpublished data). Briefly, a recombinant chromosome (al, b, Roi, pr, cn) was recovered in 1 out of 15,000 progeny from females of the genotype In(2L)Cy^L, t^R, Cy, Roi, pr/al, b, pr, cn. This chromosome had no large chromosomal aberration and was used to isolate a second recombinant (isoRoi), in which the portions of the second chromosome located outside the b-pr interval were replaced by an isogenic, homozygous viable chromosome. Recombination mapping placed Roi at position 2-52.5, between b (2-48.5; cytological position 34D1) and pr (2-54.5; cytological position 38B4).

The ato¹ mutation and UAS-amos construct have been described previously (JARMAN et al. 1993 ; GOULDING et al. 2000 ). da^UX136 (synonym, da¹⁰; BROWN et al. 1996 ) was obtained from N. BROWN; hh-lacZ (line P30; LEE et al. 1992 ), from P. BEACHY; dpp-lacZ (line H1-1; BLACKMAN et al. 1991 ), from R. BLACKMAN; hsFLP1 and FRT(40)P[w⁺]30C (XU and RUBIN 1993 ), from G. RUBIN; the h-Gal4 line (P{GAL4}h^H10; HUANG and FISCHER-VIZE 1996 ), from M. MLODZIK; the dpp-Gal4 line (P{GAL4-dpp.blk1}40C.6; STAEHLING-HAMPTON et al. 1994 ), from J. TREISMAN; the dac-Gal4 line (P{GawB}dac^p7d23; HEANUE et al. 1999 ), from G. MARDON; the GMR-Gal4 line (P{GMR-GAL4.12}; FREEMAN 1996 ), from M. FREEMAN; and hh^13c (synonym, hh⁸¹; JURGENS et al. 1984 ), dpp^blk (synonym, dpp^d-blk; BLACKMAN et al. 1987 ), the hs-GAL4 line (P{GAL4-Hsp70.PB}89-2-1; BRAND et al. 1994 ), Df(2L)r10 (ASHBURNER et al. 1990 ; SCHUPBACH and WIESCHAUS 1991 ), Df(2L)cact^255rv64 (TOWER et al. 1993 ), Df(2L)TW137, and Df(2L)TW50 (WRIGHT et al. 1976 ), from the Bloomington Stock Center. All crosses were carried out at standard temperatures on standard fly medium.

Induction of somatic clones:
Chromosomes where Roi was linked to FRT(40) [genotype Roi, P[w⁺], FRT(40) or Roi, FRT(40)] were recovered in the progeny of isoRoi/FRT(40)P[w⁺]30C females, after selection on neomycin (XU and RUBIN 1993 ). To generate homozygous Roi mutant clones in an Roi heterozygous background (Fig 1C), the following cross was performed: hsFLP1; FRT(40), P[w⁺]30C x FRT(40), Roi/CyO. To generate wild-type clones in a Roi mutant background (Fig 1D), Roi was recombined onto the FRT(40), P[w⁺]30C chromosome and the following cross performed: hsFLP1; FRT(40), Roi, P[w⁺]30C/CyO x FRT(40). Progeny were grown at 25° and subjected to a 1-hr heat shock at 38.5° once at the end of the first larval instar (48 hr after egg laying) and once at the end of the second larval instar (72 hr after egg laying). The presence of homozygous Roi or wild-type clones was inferred from unpigmented patches in the background of w⁺, Roi heterozygous eyes.

View larger version (171K): In this window In a new window Download PPT slide   Figure 1. The Roi phenotype in the adult retina. Tangential sections of the retina of Roi heterozygotes (A) and wild-type (B) adults are shown. Insets show scanning electron microscopy images of the adult eye surface. (C and D) Tangential sections through clones of homozygous Roi (C) or homozygous wild-type (D) tissue. The clones are marked with a mutation in the white (w) gene, which leads to the absence of pigmentation in the pigment cells that surround each photoreceptor cluster and along the photoreceptor rhabdomeres. The diagram at the bottom illustrates the strategy used to mark each clone. (C) In homozygous Roi mutant tissue, the retina is more severely disrupted than in heterozygous tissue, and the rhabdomere morphology is very abnormal, but ommatidial clustering is still evident. (D) Roi acts locally: Wild-type tissue surrounded by Roi/+ tissue differentiates a normal array of ommatidia. At the clone's boundaries, abnormal ommatidia often contain a wild-type (unpigmented) inner photoreceptor.

Mapping Roi against deficiencies:
Although Roi had been reported to be lethal over deficiencies that spanned the 36–37 region (LINDSLEY and ZIMM 1992 ), we were able to obtain viable trans-heterozygous escapers. The eye phenotype of the trans-heterozygotes proved difficult to interpret. For instance, Df(2L)r10 (35D1; 36A6–7) caused a slight suppression of the rough eye phenotype, suggesting that it might uncover the gene responsible for Roi, assuming that Roi was a hypermorphic allele. However, an overlapping deficiency, Df(2L)cact^255rv64 (35F6–12; 36D) had the opposite effect of slightly enhancing the rough eye phenotype, suggesting that it too might uncover the Roi gene, assuming that Roi was an antimorphic allele. Two more distal deficiencies, Df(2L)TW137 (36C2–4; 37B9–10) and Df(2L)TW50 (36E4–F1; 38A6–7) were also found to enhance the rough eye phenotype, while other deficiencies in the area had no detectable effect on Roi.

Interactions with amos also proved confusing: A recently generated loss-of-function allele (P. ZUR LAGE and A. P. JARMAN, unpublished results) causes a slight suppression of the rough eye phenotype, which is surprising since the wild-type amos gene is not expressed to detectable levels in the eye (GOULDING et al. 2000 ). In addition, if an amos loss-of-function allele acts as a suppressor of Roi, one would have expected Df(2L)TW137 and Df(2L)TW50, which remove the amos locus, to also act as suppressors, instead of enhancers. A simple interpretation of these contradictory observations is that the deficiency stocks, and perhaps the amos mutant stock as well, carry multiple lesions that can act as second-site modifiers of Roi. Consistent with this interpretation, we have found that Roi displays dominant interactions with many known and unknown loci (T. J. DONOHOE, S. PEREIRA and U. HEBERLEIN, unpublished observations).

Reversion of the Roi phenotype:
Since the deficiency mapping did not allow us to understand either the nature of the Roi allele or its precise location, we attempted to map Roi by generating revertants. We first mutagenized CyO, Roi/l(2) flies with X rays and recovered four CyO-linked potential revertants out of 30,000 flies screened. Chromosome squashes of the phenotypic revertants showed cytological abnormalities in the 36–37 region, confirming the original mapping and showing that Roi could be reverted. To obtain molecular access to the Roi gene, we reverted Roi by hybrid dysgenesis (ENGELS 1989 ). CyO, Roi/l(2) virgin females were crossed to males of the 2 P-element donor stock. Their dysgenic male progeny (F₁) were then crossed to ry⁵⁰⁶ virgin females en masse, and the CyO progeny (F₂) were examined for eye roughness. Five CyO-linked mutations that eliminated (or strongly reduced) the roughness of Roi eyes were recovered out of 100,000 F₂ screened. In situ hybridization to salivary gland chromosomes with P-element probes revealed that one of the five putative revertants carried a P element near 37 on the CyO chromosome, suggesting a revertant, rather than a second site suppressor of Roi. This mutant, referred to as Roi^Rev, also contained seven additional P elements. We were able to revert Roi^Rev to a rough-eye (presumably Roi) phenotype by remobilizing the P elements with the 2–3 transposase (ROBERTSON et al. 1988 ), confirming that the reversion was due to a P insertion. Molecular analysis of 16 germ-line revertants of Roi^Rev established that all had lost the P-element insertion at 37. This confirmed that the gene responsible for the rough-eye phenotype mapped to the 37 region.

Molecular analysis of the 36A–37A region:
Genomic DNA from Roi^Rev mutant flies was subjected to a partial Sau3A digest and cloned into a FIX BamHI vector (Stratagene, La Jolla, CA). Phages that hybridized to P-element probes were isolated and hybridized in pools of 10 to wild-type salivary gland polytene chromosomes. Phages from pools that hybridized to the 36–37 region were retested individually. Several of them yielded two signals, one at 36A and one at 37A. This suggested that the P element in Roi^Rev had inserted near a chromosomal inversion between 36A and 37A. Phage DNAs that spanned the inversion breakpoint were used to isolate wild-type cosmid clones from a library kindly provided by John Tamkun. Several cosmids yielded two signals when hybridized to Roi chromosomes, though they hybridized to either 36A or 37A in wild type, confirming the presence of a chromosomal inversion in Roi. DNA sequencing of one of the phage clones from Roi^Rev, 89, showed that the P element was inserted in DNA normally located at 36A, but translocated near 37A in Roi. Comparison with wild-type genomic sequences and the Drosophila genome sequence (ADAMS et al. 2000 ) identified the precise location of the Roi^Rev P-element insertion and the Roi breakpoint as shown in Fig 4F.

View larger version (75K): In this window In a new window Download PPT slide   Figure 2. The Roi phenotype in eye discs. Eye imaginal discs from third instar larvae carrying a dpp-lacZ (A and B) or hh-lacZ (C and D) reporter construct were stained for ß-galactosidase activity (blue) and for expression (brown) of the neuronal marker ELAV (A–D), the proneural protein Ato (E and F), and the R8-specific cell-surface marker Boss (G and H). In A–H, posterior is to the right, anterior to the left. (A and B) Expression of the MF marker dpp-lacZ (blue) is increased in Roi relative to wild type. (C and D) Expression of hh-LacZ is increased relative to wild type in differentiating Roi/+ clusters. The progressive growth of ommatidial clusters is highly disorganized in Roi compared to wild type. (E and F) Single Ato-expressing cells distribute unevenly behind the MF in Roi heterozygotes (F) and are often clustered (arrow). (G and H) Expression of Boss confirms the uneven distribution and clustering (arrows) of R8 precursors in the mutant relative to wild type.

View larger version (67K): In this window In a new window Download PPT slide   Figure 3. Genetic interactions between Roi and mutations that affect retinal differentiation. (A and B) Scanning electron microscopy images of the eye of a roDom heterozygote (A) and a Roi, roDom double heterozygote (B). (C–E) Tangential sections of adult retinae from flies heterozygous for Roi and strong alleles of hh (C, hh13c), da (D, daUX), and ato combined with da (E, ato1, daUX).

View larger version (76K): In this window In a new window Download PPT slide   Figure 4. A revertant of the Roi phenotype maps close to amos. (A) Tangential section through the retina of a w; RoiRev/+ adult. (B and C) Antibody staining of third instar eye imaginal discs to reveal Ato (B) and Boss (C) expression. (D) Scanning electron microscopy image of the eye of an adult heterozygous for RoiRev and roDom. (E) Genomic map of the 36A–37A region in wild type and Roi. The distance from the P element to the inversion breakpoint was inferred from sequence comparison with genomic P1 clone DS02780. The location of all genes indicated on the map is inferred from the BDGP/Celera genomic sequence release (ADAMS et al. 2000 ; RUBIN et al. 2000 ).

cDNA libraries from eye discs (gift from A. COWMAN) and embryos (ZINN et al. 1988 ) were screened with probes from 36A and 37A, and clones corresponding to gene BG:DS02780.1 and gene CG15160, respectively, were recovered and partially sequenced. The 36A inversion breakpoint was found to lie within the first intron of BG:DS02780.1, and the 37A breakpoint was 21 nucleotides within the last exon (exon 8) of CG15160. We attempted to detect a chimeric transcript consisting of the first exon of BG:DS02780.1 and the last exon of CG15160 by reverse transcription (RT)-PCR analysis. To detect transcription from BG:DS02780.1, we used a sense primer from the first exon (36A1: 5' CGCTCTCCTTTTTCATTTCGAATGCG 3') and an antisense primer from the second exon (36A2: 5' CCCCTGGCATCGAATATGCTACAGC 3'). To detect transcription from CG15160, we used a sense primer from the seventh exon (37A1: 5' CGTGGACGCCGCTGGACTCTAGCC 3') and an antisense primer from the eighth exon (37A2: 5' GCCCTGGTCCGGGTCCATCAAATCCCGG 3'). Each set of primers yielded the expected size fragments (375 bp for the 36A1–2 pair, 1000 bp for the 37A1–2 pair) when used against poly(A)⁺ RNA extracted from either wild-type or Roi/+ larvae. However, under the same conditions, the combination of primer 36A1 with primer 37A2 did not yield any product of the size expected from the chimeric transcript (500 bp). We concluded that the chimeric gene formed by the Roi inversion was not expressed.

Overexpression of amos:
We first attempted to drive the UAS-amos transgene using Gal4 driver lines with specific expression patterns in the eye. These included h^H10, which carries a P[Gal4] insertion at the hairy locus that is highly expressed anterior to the MF (HUANG and FISCHER-VIZE 1996 ), a dac-Gal4 line that reproduces the dac expresssion pattern around the MF (HEANUE et al. 1999 ), and a dpp-Gal4 construct expressed in the MF (STAEHLING-HAMPTON et al. 1994 ). All of them led to lethality prior to third larval instar when driving UAS-amos, which made it impossible to study their effect on eye patterning. When UAS-amos was driven by GMR-Gal4, a construct that is expressed behind the MF (FREEMAN 1996 ), the flies lived but had no eye defect. We next turned to hs-Gal4 to express amos ubiquitously in third instar larvae. Flies carrying the hs-GAL4 construct were crossed to flies carrying the UAS-amos construct. The progeny were raised at 25° and subjected to a 30-min 37° heat shock at the beginning of third larval instar. Eye discs were dissected out of larvae 24 hr later and stained with antibodies.

Histochemistry:
Antibody detection, ß-galactosidase activity staining, and retinal sections were performed as previously described (CHANUT et al. 2000 ). The rat-anti-ELAV antibody was a gift of G. RUBIN and was used at a 1:5 dilution. The mouse-anti-Boss antibody was a gift from L. ZIPURSKY and was used at a 1:1000 dilution. The rabbit-anti-Ato antibody was a gift from Y. N. JAN and was used at a 1:5000 dilution. The rabbit-anti-Amos antibody, which will be described elsewhere (P. ZUR LAGE and A. P. JARMAN, unpublished results), was used at a 1:5000 dilution.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Roi causes local disruption of the ommatidial lattice:
As previously described (RENFRANZ and BENZER 1989 ) tangential sections of adult retina show a severe disruption of the ommatidial lattice in Roi heterozygotes compared to wild type (Fig 1A and Fig B). Ommatidia with either more or less than the normal complement of photoreceptors are observed: For instance, the small rhabdomeres characteristic of inner photoreceptors R8 or R7 cells are missing in some ommatidia and clustered in others. Pigment cells are often missing, leading to large photoreceptor clusters that encompass the equivalent of two to three normal ommatidia. The homozygous Roi phenotype was assessed in somatic clones (Fig 1C). Homozygous mutant clones were smaller than their wild-type twin spots and contained very aberrant photoreceptors with fused or distorted rhabdomeres (Fig 1C).

To determine whether Roi disrupts ommatidial organization locally or at a distance, we generated marked homozygous wild-type clones in a Roi heterozygous background (Fig 1D). We found that the wild-type tissue organized into a regular array of normally structured ommatidia, while the surrounding mutant tissue developed into aberrant ommatidia. We conclude that the effect of Roi on ommatidial structure and organization is primarily local, though not necessarily cell autonomous. At the edges of the wild-type clones, ommatidia containing mutant and wild-type photoreceptors were usually abnormally structured. Ommatidial structure did not seem to correlate with the genotype of any given photoreceptor; in particular, the presence of a wild-type R8 did not guarantee the formation of normally patterned clusters.

Roi increases hh and dpp expression and disrupts the spacing of R8 precursors:
Because Roi suppresses furrow-stop mutations, we were interested in its potential effect on events occurring in the MF. Our marker of the MF, a reporter construct that places ß-galactosidase under the control of a disc-specific decapentaplegic (dpp) enhancer (dpp-lacZ; BLACKMAN et al. 1991 ), displayed a broadened expression domain in Roi relative to wild type, suggesting an expansion of MF cell fates (Fig 2A and Fig B). As Dpp signaling in the MF is known to regulate the cell cycle and, indirectly, the shape of cells (PENTON et al. 1997 ; HORSFIELD et al. 1998 ), the expansion of dpp expression may contribute in part to the cell shape anomalies that had previously been detected in the MF of Roi mutants using antibodies to cell surface markers (RENFRANZ and BENZER 1989 ).

In wild type, the expression of the dpp-lacZ reporter is activated by Hh, which is secreted by cells differentiating behind the MF (HEBERLEIN et al. 1993 ; MA et al. 1993 ). Expression of a hh-lacZ reporter construct that monitors faithfully hh transcription in discs (LEE et al. 1992 ; MA et al. 1993 ) was also greatly increased behind the MF in Roi mutants compared to wild type (Fig 2C and Fig D). We conclude that hh expression is increased in Roi mutants, leading to an expansion of MF cell fates as monitored by dpp-lacZ expression.

Expression of ato, another target of hh signaling in the MF (DOMINGUEZ and HAFEN 1997 ; STRUTT and MLODZIK 1997 ; BOROD and HEBERLEIN 1998 ), was not markedly increased in Roi heterozygotes relative to wild type (Fig 2E and Fig F). However, ato-expressing cells emerged from the MF of Roi mutant discs at irregular intervals and often remained in clusters of two or three cells (arrow in Fig 2F) instead of resolving to single evenly spaced cells as in wild type (Fig 2E). The irregular spacing and occasional twinning of the R8 precursors in Roi were maintained through later developmental stages, as shown by staining with an antibody against the R8-specific cell surface protein Bride of Sevenless (Boss; CAGAN et al. 1992 ). In wild type, the Boss antigen appears several rows behind the MF as a regular lattice of fine spots (Fig 2G), each of which corresponds to the constricted apical surface of a single R8 cell (KRAMER et al. 1991 ; CAGAN et al. 1992 ). In contrast, spacing between Boss-expressing cells was irregular in Roi heterozygotes (Fig 2H), and larger spots, consistent with clusters of two to three cells, were frequent. This suggests that Roi interferes with the early patterning events that allow the precise spacing of individual R8 precursors.

Irregular R8 selection was accompanied with the disorganized recruitment of other ommatidial cells, as shown in discs stained with an antibody against the pan-neural marker ELAV (ROBINOW and WHITE 1991 ). In wild type, the progressive induction of photoreceptor differentiation by the R8 founder follows a strict temporal and spatial sequence and creates a smooth gradient of ommatidial maturation, with clear polarity along the antero-posterior axis (TOMLINSON and READY 1987A , TOMLINSON and READY 1987B ). In Roi, while a maturation gradient was still evident behind the MF, the recruitment of photoreceptors did not appear to follow a stereotyped pattern (Fig 2B and Fig D). Whether this is a direct effect of the Roi mutation or an indirect consequence of abnormal R8 spacing is not known. In summary, our data demonstrate that Roi affects retinal patterning within the MF, at the level of R8 formation, and potentially behind the MF, at the level of further photoreceptor recruitment.

Mutations in da and hh act as strong suppressors of the Roi phenotype:
While Roi suppresses the stop-furrow phenotype of ro^Dom (Fig 3A and Fig B) and hh^bar3 (HEBERLEIN et al. 1993 ), it is itself modified by mutations in several genes known to affect eye differentiation or furrow progression. For instance, removing one copy of hh led to a noticeable suppression of eye roughness in Roi heterozygotes. In a tangential section (Fig 3C), the retina still appeared somewhat disorganized, but at least 50% of the ommatidia had regained a normal structure and orientation (Fig 3C, arrows).

The strongest suppression of the rough eye phenotype was achieved with the removal of one copy of da. In section, the retina appeared almost normal (compare Fig 3D with Fig 1B), with only an occasional abnormally structured ommatidium. da encodes a protein of the bHLH family (CAUDY et al. 1988 ) that is required for Ato's proneural function in the eye (BROWN et al. 1996 ). Interestingly, removal of one copy of ato had no detectable impact on the Roi phenotype (not shown), while simultaneous removal of ato and da only slightly improved the suppression of patterning defects relative to the removal of da alone (Fig 3E).

Molecular analysis of the 36A–37A region:
On the basis of its failure to complement the lethality of specific chromosomal deficiencies, Roi had previously been mapped to the 36F7–37B8 region (VOELKER and LANGLEY 1978 ). Deficiencies spanning this region do not lead to eye roughness, suggesting that Roi is a gain-of-function mutation. To identify the Roi gene, we obtained revertants of the rough eye phenotype by mutagenizing a Cyo, Roi chromosome using X-ray irradiation or by P-element-mediated dysgenesis (see MATERIALS AND METHODS). Several X-ray-induced revertants showed cytological abnormalities in the 36–37 region (not shown), confirming the initial mapping of Roi. One of the P-induced potential revertants, Roi^Rev, contained a P element at 37A and was retained for further analysis. In retinal sections, Roi^Rev ommatidial structure and patterning appeared indistinguishable from wild type (Fig 4A; compare with Fig 1B). In discs, staining for Ato and Boss expression confirmed that spacing and resolution of R8 precursors was normal (Fig 4B and Fig C; compare with Fig 2E and Fig G). In addition to reverting to wild-type eye morphology, Roi^Rev had also lost most of its ability to suppress the furrow-stop phenotype of ro^Dom (Fig 4D; compare with Fig 3A and Fig B) and hh^bar3 (not shown). Furthermore, excision of the P element at 37A restored the Roi phenotype (see MATERIALS AND METHODS). Taken together, these observations confirmed that Roi is a gain-of-function mutation that can be reverted by the inactivation of a gene located at 37A.

To identify this gene, genomic DNA flanking the insertion was obtained from a phage library of Roi^Rev genomic DNA (see MATERIALS AND METHODS). Phages that contained the P element at 37A were found to hybridize to two locations on wild-type polytene chromosomes: 37A and 36A. Two sites of hybridization were also seen when wild-type genomic DNA from the 37A region was hybridized to Roi or Roi^Rev polytenes. This indicated that Roi and Roi^Rev carried an inversion between 36A and 37A. This inversion may have gone unnoticed in previous cytological examinations of CyO, Roi chromosomes because of the abnormal conformation frequently adopted by polytene chromosomes in the 36–39 region (LINDSLEY and ZIMM 1992 ).

DNA sequence analysis of the Roi^Rev phage clones and of wild-type genomic clones obtained from the 36A and 37A regions identified the location of the inversion breakpoints in Roi and showed that the P element in Roi^Rev had inserted 8.4 kb distal to the 37A breakpoint, inside the inversion (Fig 4E). cDNA clones homologous to the 36A and 37A region were isolated from wild-type eye disc and embryonic libraries and shown by cytology and molecular analysis to span the Roi inversion breakpoint. Sequence comparisons revealed that the P element in Roi^Rev lay 14 bp upstream of the first exon of a gene of unknown function normally located at 36A, later identified as BG:DS02780.1 (ASHBURNER et al. 1999 ). The inversion breaks in the first intron of the 36A gene and within the last exon of a gene of unknown function normally located at 37A, later identified as CG15160 (ADAMS et al. 2000 ). We first hypothesized that a chimeric gene formed by the 3' portion of the 37A gene (CG15160) under the control of the 5' portion of the 36A gene (BG:DS02780.1) was responsible for the Roi phenotype. In Roi^Rev, expression of this gene would be abolished by the P insertion in the promoter region of BG:DS02780.1. However, we could not detect the expression of a chimeric RNA by RT-PCR in Roi heterozygotes (see MATERIALS AND METHODS), and this hypothesis was abandoned.

Later releases of the Drosophila genome sequence in the 36–37 area revealed other potential candidates for the Roi gene (Fig 4E). The BG:DS02780.1 gene was found to overlap a three-gene cluster encoding imaginal disc growth factors (IDGF1–3) related to Chitinase (KAWAMURA et al. 1999 ). In Roi, the 5' end of the Idgf gene cluster is located 5.3 kb away from the inversion breakpoint. The first exon of BG:DS02780.1 lies within the first intron of Idgf2, on the opposite strand. Consequently, the Roi^Rev P element is also inserted within Idgf2's first intron. The Idgf gene cluster, 8 kb long, lies 12.3 kb downstream of dachshund (dac), which encodes a transcription factor implicated in eye morphogenesis (MARDON et al. 1994 ). Outside the inversion, 2.6 kb proximal to the 37A breakpoint is amos, a proneural gene required for the development of olfactory organs (GOULDING et al. 2000 ; HUANG et al. 2000 ).

Roi causes misexpression of amos in the eye disc:
Of the genes mapping near the genomic breakpoint in Roi, dac and amos appeared as the most likely candidates to disrupt retinal development in Roi. dac belongs to a network of genes including eyeless, eyes absent, and sine oculis that imparts retinal fate to the cells of the eye epithelium (CHEN et al. 1997 ; PIGNONI et al. 1997 ). amos is a proneural gene most closely related to ato (JARMAN et al. 1993 )—the two genes share 74% sequence identity over their entire bHLH region—that has been reported to mimic ato in the induction of sense organs in embryos and adults (GOULDING et al. 2000 ; HUANG et al. 2000 ).

Staining with an antibody directed against the Dac protein failed to detect any difference among Roi, Roi^Rev, and wild-type eye discs (Fig 5, A–C). In addition, Roi complemented lethal dac alleles and the rough-eye phenotype was insensitive to a reduction in dac gene dosage (not shown); the revertant also complemented lethal and eye-specific dac alleles (not shown). We conclude that dac is not affected by the Roi inversion and not implicated in the resulting rough-eye phenotype. In contrast, amos expression was markedly different between Roi and wild type (Fig 5D and Fig E). In wild type, amos expression does not begin in the eye-antennal region until pupal stages and remains confined to the area giving rise to olfactory sensilla precursors in the antenna (GOULDING et al. 2000 ). In Roi mutant discs from third instar larvae, we found high levels of Amos protein in a broad area surrounding the MF (Fig 5E). Ectopic expression was sharply reduced in Roi^Rev discs (Fig 5F), although not completely abolished, which might explain that Roi^Rev retains some ability to suppress ro^Dom (Fig 4D). Together, these data suggested that the Roi phenotype might result from ectopic expression of the proneural gene amos in the eye disc.

View larger version (85K): In this window In a new window Download PPT slide   Figure 5. The Roi phenotype results from ectopic expression of amos in the eye disc. (A–F) Staining of wild-type (A and D), Roi/+ (B and E), and RoiRev/+ (C and F) third instar eye-antennal discs with an antibody against the retinal specification protein Dachshund (A–C) or the proneural protein Amos (D–F). (G–I) Overexpression of amos leads to R8 patterning defects. amos was ubiquitously expressed in third instar larvae using a UAS-amos construct under the control of a hs-GAL4 construct. Third instar eye discs were stained with antibodies against Ato (green) and Boss (red). (G and H) Ubiquitous amos causes an expansion of the front of ato expression, associated with frequent bulges (white arrows) and the prolonged expression of ato in isolated cells behind the MF (bracket). (I) Close-up of the area highlighted in (H). Staining for Boss expression reveals irregular spacing and clustering of R8 cells (arrows).

To test this hypothesis, we attempted to drive a UAS-amos transgene under the control of GAL4 constructs expressed in the MF or anterior to it. However, amos misexpression under all the drivers tested caused early larval lethality (see MATERIALS AND METHODS), making it impossible to study an effect in third instar eye discs. We therefore expressed amos ubiquitously in a short pulse during the third larval stage using a hs-Gal4 driver (see MATERIALS AND METHODS). This allowed survival to adult stages and led to a roughening of the adults' eye surface (not shown). In discs, amos overexpression led to an expansion of ato expression (Fig 5G and Fig H): Instead of the sharp band of Ato protein observed ahead of wild-type furrows (see Fig 2E), the front of differentiation was marked by an irregular and mottled zone of ato expression. Forward bulges of the Ato front (arrows in Fig 5G and Fig H) suggested regions of accelerated furrow progression, without proper patterning. Behind this expanded front of Ato protein, single Ato-expressing cells were found over a broader area than in wild type (bracket in Fig 5G), suggesting that ato expression persisted longer in the R8 precursors than in wild type. In addition, spacing of these cells was often irregular. Staining for Boss expression confirmed the presence of irregularly spaced R8 precursors and of occasional R8 clusters (Fig 5I).

In summary, overexpression of amos in eye discs under heat-shock control leads to defects in R8 patterning that are similar to those observed in Roi mutant discs, which is consistent with the proposal that the Roi phenotype is caused by misexpression of amos in the eye. On the other hand, heat-shock-driven misexpression of amos also causes a considerable expansion or stabilization of ato expression relative to wild type, an effect that was not observed in Roi. This discrepancy might simply reflect the different patterns of amos misexpression in the two situations: In Roi, amos expression is confined to a portion of the eye disc near the MF, whereas it is ubiquitous under heat-shock control. While we cannot eliminate the possibility that other genes in the vicinity of the Roi breakpoints participate in the Roi phenotype, we conclude that the effect of Roi on ato patterning is due mainly to ectopic expression of amos in the retinal portion of the eye-antennal disc.

Roi suppresses the differentiation defect of ato¹ homozygotes:
Because amos and ato encode related bHLH proteins with somewhat overlapping specificity in the differentiation of sense organs (GOULDING et al. 2000 ; HUANG et al. 2000 ), we were curious to see whether amos could assume some of ato's functions in retinal differentiation. We introduced the Roi mutation in the background of the ato¹ mutation, a viable, recessive loss-of-function ato allele that does not allow the development of R8 photoreceptors (JARMAN et al. 1994 ). While the eyes of ato¹ homozygotes are reduced to a slit of pigment cells (Fig 6A), the presence of one copy of Roi allows them to reach one-third to one-half of wild-type size (Fig 6B). Upon sectioning, ommatidial clusters appeared disorganized and composed of an abnormal number of photoreceptors whose rhabdomeres were often elongated and misshapen (Fig 6C). In many ommatidia, however, we were able to discern smaller rhabdomeres (Fig 6C, red circles), suggesting the presence of R8 photoreceptors. To confirm the identity of these cells, we stained imaginal discs from double mutant (Roi/+; ato¹) larvae with the anti-Boss antibody. The Boss protein was detectable as small spots in a disorganized array, but at distances comparable to those observed between R8 precursors in wild type (Fig 6D). This suggested that most, if not all, of the ommatidial clusters that developed in this double mutant background contained a Boss-expressing, presumed R8 cell.

View larger version (130K): In this window In a new window Download PPT slide   Figure 6. Roi restores retinal differentiation in ato1 homozygotes. (A and B) Photographs of adult heads from ato1/ato1 (A) or Roi/+; ato1/ato1 (B) adults. (C) Tangential retinal section of a Roi/+; ato1/ato1 double mutant. Ommatidia are disorganized and may contain an excess of photoreceptors (yellow circle), an excess of inner photoreceptors (red circles), or a lack of inner photoreceptors (green circle). (D–F) Third instar imaginal discs from ato1 homozygotes (E) or Roi/+; ato1/ato1 (D and F) stained to reveal expression of Amos (green, D), Ato (green, E and F) and Boss (red, D and F). (D) Roi causes Amos misexpression (green) and restores the formation of Boss-expressing (red) R8 cells in ato1 homozygotes. (E) ato1 homozygotes express a mutant Ato protein that fails to resolve to single cells. (F) Roi restores the resolution of Ato to isolated cells (arrows) behind the MF in ato1 homozygotes.

The ato¹ allele expresses a mutant protein that is recognized by the anti-Ato antibody and forms a continuous band near the posterior disc margin that fails to resolve to single cells in ato¹ homozygotes (Fig 6E; SUN et al. 1998 ). In the presence of Roi, however, resolution to single cells was clearly restored behind the continuous front of ato expression (Fig 6F, arrows). The distribution of these cells was uneven and they occasionally remained clustered, in a manner reminiscent of Roi's effect in an otherwise wild-type background (compare Fig 6F with Fig 2F). This shows that Roi can restore the resolution of the mutant Ato protein to isolated cells, as well as allowing the adoption of the R8 cell fate, as monitored by Boss expression. Whether the Boss-expressing cells observed in the Roi; ato¹ double mutant derive from the single ato-expressing cells that emerge from the MF could not be ascertained, because ato expression subsides several rows before the appearance of Boss (Fig 6F).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Roi is a gain-of-function allele of amos:
Several observations suggest that Roi is a gain-of-function allele of amos. First, the Amos protein is abundantly expressed in the eye imaginal discs of Roi heterozygotes, whereas it is never detected at this location in wild-type eye-antennal discs. Second, reversion of the Roi phenotype is achieved by a P-element insertion a few kilobases downstream of amos and is associated with the severe reduction of amos expression in the eye. Third, overexpression of amos in wild type, under heat-shock control, causes the disruption of R8 cells spacing in a manner very reminiscent of Roi. Fourth, the Roi phenotype is almost completely suppressed by mutations in daughterless, a gene that encodes a bHLH protein shown to form heterodimers in vitro with the Amos protein (HUANG et al. 2000 ). Together, these observations constitute strong evidence that the Roi phenotype is due to the ectopic expression of the bHLH protein Amos.

The basis of amos's misexpression in Roi is unclear. The 36A–37A chromosomal inversion that we mapped in Roi breaks 2.6 kb downstream of amos and brings it in the vicinity of the Idgf gene cluster (7.9 kb away) and of dac (29.1 kb away). The inversion appears to remove a 3' endogenous enhancer of amos, as Roi behaves as a loss-of-function amos allele in the development of antennal olfactory sense organs (P. ZUR LAGE and A. P. JARMAN, unpublished results). In the absence of endogenous regulatory sequences, amos may respond to other neighboring enhancers. Whether the genes brought closest to amos in Roi—the Idgf gene cluster and the 5'-most portion of BG:DS02780.1 gene—have enhancers that can direct gene expression in the eye is at present unknown. At this point, the most likely source of an eye-specific enhancer is dac, in spite of its distance from amos (29 kb), since its domain of expression in the eye coincides roughly with that of amos in Roi (MARDON et al. 1994 ; Fig 5A, Fig B, and Fig E).

Genes other than amos may be affected by the inversion and contribute to some extent to the Roi phenotype. Our experiments eliminated the possibility that dac, a likely candidate considering its normal involvement in eye differentiation (MARDON et al. 1994 ; CHEN et al. 1997 ), had a part in the Roi phenotype. We also do not believe that a chimeric gene straddling the 37A breakpoint participates in the Roi phenotype because this gene, made of a noncoding exon from the BG:DS07820.1 gene and a portion of the last exon of the CG15160 gene, is probably not coding. In addition, its expression was undetectable in Roi mutants. No mutation has been reported in the Idgf genes or in the putative genes immediately proximal to amos, and whether they contribute to the Roi phenotype cannot be ascertained. For the sake of simplicity, the rest of our discussion assumes that amos misexpression in the eye disc mediates all aspects of the Roi phenotype, unless otherwise mentioned.

Amos activates ato expression:
We find that Roi disrupts the patterning of the ato-expressing cells that emerge from the posterior edge of the furrow: Their distribution is irregular and they often subsist as small clusters of two or three cells, instead of resolving to evenly distributed single cells. This indicates that intermediate clusters distribute unevenly along the MF and fail to resolve properly to single R8 cell precursors. While the formation and distribution of intermediate groups depends mostly on Notch-mediated lateral inhibition (CAGAN and READY 1989 ; BAKER and ZITRON 1995 ; BAKER et al. 1996 ), the resolution of ato expression to single R8 precursors requires the repression of ato expression by the homeobox gene ro (DOKUCU et al. 1996 ). In the absence of ro function, ato expression is maintained in clusters of two to three cells known as the R8 equivalence group, which leads to the differentiation of additional R8 cells. Our data suggest that amos interferes both with lateral inhibition and with the inhibitory effect of ro on ato expression. The expression of a ro-lacZ reporter construct (HEBERLEIN et al. 1991 ) is not detectably repressed at the posterior edge of the MF in the background of the Roi mutation (F. CHANUT and U. HEBERLEIN, unpublished observation). We therefore do not believe that ectopic amos can prevent ro expression. It remains possible, however, that, in Roi, Amos somehow prevents the Rough protein from repressing ato. Antagonism between Amos and Ro could explain why Roi is a strong suppressor of the ro^Dom stop-furrow phenotype, which is caused by ectopic ro expression anterior to the MF (CHANUT et al. 2000 ).

Ubiquitous expression of amos under a heat-shock promoter gives rise to similar defects in the patterning of ato-expressing cells behind the MF as Roi, including the irregular distribution and frequent twinning of R8 precursors. In addition, there is considerable expansion of the domain of ato expression, both anterior and posterior to the MF, which suggests that Amos can, directly or indirectly, activate ato expression. Inhibitors of ato expression anterior to the MF include the HLH proteins Hairy and Emc (BROWN et al. 1995 ) and Hairless (CHANUT et al. 2000 ). It is possible that Amos interferes with their expression or with their activity (for instance, an excess of Amos protein might titrate out the repressor EMC). Alternatively, because of its high sequence similarity to Ato, Amos might directly activate ato expression, mimicking Ato's ability to autoregulate (SUN et al. 1998 ). This would also explain why ectopic expression of amos causes the prolonged expression of ato in the R8 precursors, where ato maintenance is due primarily to autoregulation (SUN et al. 1998 ). As precocious or excessive ato expression is known to lead to aberrant ato patterning (BROWN et al. 1995 ; DOKUCU et al. 1996 ), the patterning defects observed upon Amos overexpression could at least in part be explained by ato upregulation.

In conclusion, we propose that ectopic expression of Amos leads to inappropriate ato expression, either by directly activating ato or by interfering with repressors of ato expression. In Roi, the effect of Amos is milder—in particular, ato expression levels were not detectably elevated in the MF—presumably because amos misexpression is weaker and spatially more restricted than when amos is overexpressed ubiquitously. We note that a slight increase of ato expression in Roi could also explain the strong suppression of the ro^Dom stop-furrow phenotype, since ro^Dom is efficiently suppressed by heterozygous mutations in groucho and Hairless, which presumably lead only to slight elevations of ato levels (CHANUT et al. 2000 ).

Roi patterning defects are suppressed by mutations in hedgehog (hh) and daughterless (da):
Regularity and proper polarization of the ommatidial lattice are almost completely restored in Roi mutant eyes by a reduction of the da gene dosage and partially restored by a reduction of the hh gene dosage. Since Amos and Da have been shown to form heterodimers in vitro (HUANG et al. 2000 ), the strong sensitivity to da gene dosage suggests that all the effects of the Amos protein in Roi eyes are carried out by Amos-Da heterodimers. The sensitivity to hh gene dosage is intriguing because hh is upregulated behind the MF in Roi discs. This observation suggests that excess hh might contribute to eye roughness. Overexpression of hh behind the MF does not usually result in R8 patterning defects (WHITE and JARMAN 2000 ; F. CHANUT and U. HEBERLEIN, unpublished observations), although in a genetic background where R8 spacing is already compromised, it can further enhance the failure to resolve ato expression to single R8 cells (WHITE and JARMAN 2000 ). It is therefore possible that the excess Hh produced in Roi mutants exacerbates the effect of Amos on R8 patterning. Removing one copy of the hh gene would alleviate this effect and partially suppress eye roughness.

Why hh is overexpressed in Roi mutants is unclear. In wild type, expression of hh behind the MF is limited to the R2 and R5 precursors and requires the Da protein (BROWN et al. 1996 ). In Roi, the formation of ectopic Da-Amos heterodimers might activate hh expression in more cells than in wild type. In this case, da mutations would affect the Roi phenotype at two levels: in the MF, by reducing amos's interference with ato patterning, and behind the MF, by reducing hh expression. This could explain the strong suppression of Roi by halving the da gene dosage. At this point, however, it remains also possible that the increased transcription of hh in Roi results from the misexpression of other genes in the vicinity of the Roi inversion breakpoints. Regardless of its cause, increased hh expression behind the MF is likely to explain why Roi suppresses the furrow-stop phenotype of the hypomorphic allele hh^bar3, which is thought to cause an eye-specific transcriptional defect (LEE et al. 1992 ; HUANG and KUNES 1996 ). We propose that Roi overcomes the transcriptional block of hh^bar3, which in turn restores Hh production to sufficient levels for normal MF progression.

Can amos induce photoreceptor differentiation?
It has recently been shown that ectopic expression of the proneural gene scute (sc) in ato¹ homozygotes can lead to the differentiation of photoreceptors in the apparent absence of R8 founders (SUN et al. 2000 ). We similarly tested whether amos could induce photoreceptor differentiation in the absence of ato function by introducing Roi in the background of ato¹ homozygotes. We found that this restored retinal differentiation, but that, in contrast to sc overexpression, it was accompanied by the restoration of R8 cells. In addition, whereas in ato¹ mutants the Ato protein fails to become patterned, resolution of ato to single cells behind the MF was restored in Roi.

This experiment suggests that, in contrast to sc, amos does not induce the differentiation of photoreceptors independently of an R8 founder. Consistent with this proposal, Roi is not associated with an overall excess of outer photoreceptors relative to R8s in a wild-type background (RENFRANZ and BENZER 1989 ). In the Roi; ato¹ adult eyes, we do observe a number of photoreceptor clusters devoid of R8 cells (Fig 6C), which could indicate that they were induced directly by amos, in the absence of any R8 founder. However, R8 precursor cells are present at high density in double mutant discs (Fig 6D). We therefore find it more likely that all the photoreceptor clusters that form in Roi; ato¹ double mutants were seeded by an R8 founder; some R8 cells may later degenerate, for instance, because of incomplete fate specification.

As for the origin of the R8 cells in the Roi; ato¹ double mutants, we envision two scenarios. In the first one, all R8 cells derive from the isolated ato-expressing cells that are restored by Roi behind the MF of ato¹ mutant discs. This implies that ato¹ is able to support the differentiation of R8 cells, although it carries point mutations that are thought to abolish the Ato protein's DNA-binding activity (JARMAN et al. 1994 ). This would suggest that ato¹ retains some residual activity and becomes potentiated in the Roi background, either because its expression is elevated or via synergy with amos at the level of ato's transcriptional targets.

In the second and perhaps more likely scenario, amos, due to its high similarity with ato, directly induces the differentiation of cells with at least some R8 characteristics, including the ability to maintain ato expression, to express Boss, and to recruit other photoreceptor cell types. A potential difficulty with this scenario is that the R8 cells that develop in the Roi; ato¹ mutants form a lattice that, though imperfect, is reminiscent of wild type. How do R8 cells become patterned when they are induced by amos? We note first that amos expression in Roi, while continuous ahead of the MF, becomes restricted to groups of cells behind the MF (see Fig 5E). Second, experiments in which ato was expressed ubiquitously have shown that only the cells of the R8 equivalence group have the competence to adopt an R8 fate (DOKUCU et al. 1996 ). The combination of restricted cell competence and patterned amos expression behind the MF can probably account for the distribution of R8 cells in the Roi; ato¹ double mutant.

In conclusion, we propose that amos can promote photoreceptor differentiation in the eye and that its activity is biased toward the induction of the R8 fate, which it may achieve by substituting for ato or by increasing ato activity. The Roi mutation, a gain-of-function allele that misexpresses amos in the eye, provides a promising background in which to identify transcriptional targets of amos and ato, as well as genes that modulate their activity.

	FOOTNOTES

¹ These authors contributed equally to this work.
³ Present address: Department of Laboratory Medicine, University of California, San Francisco, CA 94143.
⁴ Present address: Boston University School of Medicine, Boston, MA 02118.
⁵ Present address: ITRI-BMEC, Hsinchu 310,Taiwan.

	ACKNOWLEDGMENTS

We thank John Tamkun for the use of his cosmid library, Gerry M. Rubin for his support in the initial stages of this project, Dennis Ballinger for sharing information on Roi and providing the isoRoi stock, the Bloomington Stock Center and various members of the fly community for providing fly stocks, and members of the Heberlein lab, past and present, for their technical and intellectual input. This work was supported by a grant from the National Science Foundation (IBN-9996214) and from the National Institutes of Health (EY-11410) to U.H.

Manuscript received September 7, 2001; Accepted for publication November 21, 2001.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ADAMS, M. D., S. E. CELNIKER, R. A. HOLT, C. A. EVANS, and J. D. GOCAYNE et al., 2000  The genome sequence of Drosophila melanogaster. Science 287:2185-2195[Abstract/Free Full Text].

ALBAGLI, O., M. P. LAGET, and F. CHANUT, 1997  Photoreceptor differentiation in Drosophila: transduction and interpretation of the RTK signaling pathway. Med. Sci. 13:184-191.

ASHBURNER, M., P. THOMPSON, J. ROOTE, P. F. LASKO, and Y. GRAU et al., 1990  The genetics of a small autosomal region of Drosophila melanogaster containing the structural gene for alcohol dehydrogenase. Genetics 126:679-694[Abstract].

ASHBURNER, M., S. MISRA, J. ROOTE, S. E. LEWIS, and R. BLAZEJ et al., 1999  An exploration of the sequence of a 2.9-Mb region of the genome of Drosophila melanogaster: the Adh region. Genetics