semang affects the development of a subset of cells in the Drosophila compound eye

Mechanisms of Development 95 (2000) 113±122

www.elsevier.com/locate/modo

semang affects the development of a subset of cells in the Drosophila compound eye Qian Zhang, Xiangyi Lu* Department of Molecular Biosciences, The University of Kansas, Lawrence, KS 66045, USA Received 18 November 1999; received in revised form 13 March 2000; accepted 11 April 2000

Abstract semang (sag), a mutation isolated as a suppressor of Drosophila Src42A, has previously been shown to affect some receptor tyrosine kinase mediated embryonic processes. Here we show that sag speci®cally affects the development of R1, R6 and R7 photoreceptor cells in a cellautonomous manner. These cells are absent in the mutant at the time when they normally appear in the ommatidial pre-clusters. Genetic analyses suggest that sag functions downstream of, or parallel to, Mapk and Yan in the photoreceptor differentiation pathway. The autonomous requirement of sag for R1/R6/R7 development could be explained by a selective impairment of the late, but not early, rounds of Egfr-induced precursor cell assembly by the sag mutations. Egfr signaling is highly regulated by autocrine or paracrine mechanisms in different cells. Knowing that the photoreceptor cluster formation is a complex process involving dynamic changes in cell±cell contact, our hypothesis is that the sag alleles affected certain special aspects of Egfr-signaling that are unique for the recruitment of R1/R6/R7 cells. q 2000 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Eye development; Ras1; Egfr; Src

1. Introduction The Drosophila compound eye is composed of 800 identical eye units, or ommatidia. Each ommatidium contains eight photoreceptor cells (R1±R8), four cone cells and several pigment cells. These different cell types that form an ommatidium are not related by cell lineage. Rather, these cells assemble into an ommatidial cluster via inductive recruitment of undifferentiated cells of the eye imaginal disc (reviewed by Wolff and Ready, 1993). Freeman (1997) has proposed that cellular recruitment during ommatidial assembly involves several rounds of cell-to-cell signaling triggered by the reiterative activation of Drosophila EGF receptor tyrosine kinase (Egfr RTK). This model is supported by the observation that a dominant negative form of Egfr, when expressed at the time of cellular recruitment, ®rst blocks the formation of photoreceptor cells R1± R7 (R8 was not tested), then the cone cells and ®nally the pigment cells (Freeman, 1996). Thus Egfr signaling is necessary for the recruitment of all cell types in the ommatidia. The differentiation of the R7 cells requires additionally the activation of the Sevenless (Sev) RTK in the R7 precursor cells. * Corresponding author. Tel.: 11-785-864-4167; fax: 11-785-864-5321. E-mail address: [email protected] (X. Lu).

The cells that are recruited into the developing cluster choose speci®c developmental fates. Mechanisms involved in determining photoreceptor subtypes as well as cone and pigment cell types are largely unknown. Since Egfr activation appears to recruit all cell types of the eye, the differentiation of a speci®c cell type is possibly determined by factors already present in the cell at the time when Egfr RTK is activated. In supporting this model, it has been shown that Egfr signaling induces different gene expression programs in different precursor cells. For example, Egfr signaling turns on the transcription of phyllopod (phyl) speci®cally in R1, R6 and R7 precursor cells, thereby allowing these cells to adopt the neuronal differentiation pathway (Li et al., 1997; Tang et al., 1997). If the phyl gene is absent, the R1, R6 and R7 precursors will all be transformed into extra cone cells. On the other hand, ectopic expression of phyl in the cone cell precursors is suf®cient to cause the transformation of these cells into R7-like cells (Chang et al., 1995; Dickson et al., 1995). phyl and most other cell type-speci®c genes identi®ed so far are nuclear factors (reviewed in Freeman 1997). For example, the homeobox gene rough is required for the formation of the R2/R5 photoreceptor pair (Tomlinson et al., 1988; Heberlein et al., 1991). When rough is expressed in the R7 precursors, these cells are transformed into R2/R5-like photoreceptor cells following the activation of the Sev RTK (Basler et al., 1990;

0925-4773/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0925-477 3(00)00340-3

114

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

Kimmel et al., 1990). These studies indicate that a combination of nuclear factors (pre-existing and/or induced) plays an important role in determining the outcome of cellular responses to an inductive signal. The signals from RTKs in many developmental systems appear to merely turn on the underlying cellular differentiation programs (reviewed in Freeman, 1997; Schweitzer and Shilo, 1997). In addition, speci®c cell±cell contact/signaling intrinsic to the ommatidial assembly could presumably contribute to cell fate decisions. Many components of the RTK pathways such as Drk, Sos and Dos are colocalized with the adherens junctions to the apical surface of pre-ommatidial cells (Olivier et al., 1993; Karlovich et al., 1995; Raabe et al., 1996). Thus speci®c cell±cell contacts that alter the membrane skeleton could possibly convey cell type speci®city through unique interactions with the ubiquitous RTK signaling components. The Drosophila Src family member, Src42A (also referred to as Dsrc41; Takahashi et al., 1996), is also found to colocalize with actin ®bers and putative adherens junctions in ommatidial pre-cluster cells. We have recently shown that Src42A negatively regulates Egfr and Torso RTK signaling (Lu and Li, 1999). Src42A appears to de®ne a branch pathway parallel to that of Ras1 since the reduction of maternal Src42A activity is able to enhance Torso signaling in embryos lacking the Ras1 protein (Lu and Li, 1999). In the eye, loss-of-function Src42A mutations enhance the rough eye phenotype caused by the hyperactivation of Ras or Raf (Lu and Li, 1999). To understand Src42A function, suppressor mutations of Src42A were isolated (Zhang et al., 1998). Here we show that a Src42A suppressor, semang (sag), speci®cally affects the recruitment of only a subset of photoreceptors, namely the R1, R6 and R7 cells. Different from phyl and other cell type-speci®c genes characterized so far, sag does not cause a transformation between different cell types and is thus unique among the currently known genes that de®ne subset of ommatidial cell types. Genetic analyses showed that Sag promotes RTK signaling in the eye. The autonomous requirement of sag for R1/R6/R7 development could be explained by a selective impairment of the late, but not early, rounds of Egfr-induced precursor cell assembly by the sag mutations. One may also speculate that sag transmits a signal from speci®c cell contact(s) during ommatidial assembly and this signal, upon converging with the Egfr signaling pathway, determines the recruitment of speci®c ommatidial precursors.

2. Results 2.1. sag is autonomously required for the development of R1, R6 and R7 Loss-of-function sag homozygotes (sag 13L and sag 32-3, see Section 4) die as late pupae with small rough eyes. All ommatidia of the mutant eye lack the normal complement of

photoreceptor cells. To ®nd out whether sag only affects speci®c photoreceptor cells in the eye, mitotic clones of sag homozygous cells were generated in a sag/1 genetic background. Inside the unpigmented patch of sag mutant clones, ommatidia lacked the inner R7, and two to three outer photoreceptors (Fig. 1A). At the clonal border, mosaic ommatidia with a normal complement of photoreceptors were composed of both the wild-type ( 1 /1 or sag/1) and mutant (sag/sag) cells (Fig. 1B). The sag/sag mutant cells lacked the pigment granules that are present in the wild-type cells at the base of rhabdomeres (Fig. 1B). In addition, the sag/sag photoreceptors formed smaller rhabdomeres compared to the wild-type photoreceptors (Fig.

Fig. 1. Wild-type sag is autonomously required for the formation of R1, R6 and R7 photoreceptor cells. Apical sections of sag 32-3 mosaic eyes. (A) Patches of pigmented wild-type ( 1 /1 or sag 32-3/1) and unpigmented sag 32-3 homozygous cells. Ommatidia in the center of the unpigmented patch were missing both the R7 and two to three outer photoreceptor cells. At the clonal border, mosaic ommatidia with the normal cellular complement were composed of both wild-type and mutant cells (B). Only the wild-type cells contain dark brown pigment granules at the base of rhabdomeres (B; arrowheads point to R1, R6 and R7 cells that contained pigment granules). Note that the mutant outer photoreceptor cells were smaller in diameter. The genotypes of photoreceptor cells observed in 111 mosaic ommatidia are shown below. Closed and open circles represent the wild-type and mutant cells, respectively. The total numbers of each ommatidial type scored in this analysis are indicated in the upper right hand corner of each panel (see also Table 1).

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

115

1B). Thus both the pigment granules and the size of the rhabdomere allowed us to score the genotype of each photoreceptor cell reliably. Fig. 1 shows a panel of 111 mosaic ommatidia that were examined through serial sections from the apical to the basal R8 cell layer; and the genotype of all photoreceptor cells were determined. All of the R7 cells contained pigments and were therefore wild-type cells. Most of the R1 (106/111) and R6 (109/111) cells were also wild-type cells. The remaining photoreceptors showed approximately equal numbers of the mutant and wild-type R4 and R5 cells and slightly higher numbers of the mutant R2 (62%) and R3 (62%) cells (Table 1). Similar results were obtained for both alleles (sag 13L and sag 32-3) in the mosaic analysis and all later analyses. These results suggest that wild-type Sag function is absolutely required for the development of the R7 cells and strongly required for that of the R1 and R6 photoreceptor cells. However, sag is not autonomously required for the speci®cation of R2/R5 and R3/R4 photoreceptor pairs although these cells developed smaller rhabdomeres in the absence of wild-type Sag activity. In some ommaditia composed of entirely sag/sag mutant cells, one additional outer photoreceptor besides R1 and R6 was occasionally missing. This is likely caused by a combined disruptive effect of lacking Sag activity in all ommatidial cells since approximately equal numbers of the wild-type and the mutant R2±5 photoreceptors formed in the normally constructed mosaic ommatidia. In ¯y eyes carrying constitutively activated Ras, while activated Ras leads to the formation of extra photoreceptor cells in many ommatidia, the disruptive effect of activated Ras during ommatidial assembly also causes the loss of photoreceptor cells in other ommatidia.

Table 1 Mosaic analyses of sag 32-3 and sag 13L alleles a Photoreceptor cell

Percentage of cells with genotype of sag 32-3/1 and 1/1

Percentage of cells with genotype of sag 13L/1 and 1/1

R1 R2 R3 R4 R5 R6 R7 R8

92 (n ˆ 100) 41 (n ˆ 100) 35 (n ˆ 100) 52 (n ˆ 100) 57 (n ˆ 100) 99 (n ˆ 100) 100 (n ˆ 99) b 53 (n ˆ 58)

96 (n ˆ 111) 37 (n ˆ 111) 37 (n ˆ 111) 47 (n ˆ 111) 48 (n ˆ 111) 99 (n ˆ 111) 100 (n ˆ 111) 48 (n ˆ 111)

a

Mitotic clones of sag 32-3 or sag 13L homozygous cells were induced in a sag/1 genetic background. Those mosaic ommatidia that contained the normal complement of photoreceptor cells were closely examined through serial sections and the genotype of each photoreceptor cell was determined unambiguously. The mutant photoreceptor cells lacked pigment granules and have smaller rhabdomeres. b The genotype of one R7 cell in this analysis could not be unambiguously determined.

Fig. 2. Ommatidial pre-clusters in sag mutants contain fewer cells than normal. Immunostaining of eye imaginal discs. (A,B) Anti-Elav antibody labels the nuclei of photoreceptor cells in the developing clusters. The clusters are progressively more mature posteriorly behind the morphogenic furrow (black arrowheads). In the wild-type larval disc (A), mature clusters contained eight cells (only ®ve cells were seen on the focal plane). In the sag 13L mutant larval disc (B), pre-clusters were irregular and missing three to four cells (arrow). (C±F) Anti-b -galactosidase antibody staining of eye discs derived from ¯ies that carried a copy of seven up-lacZ which is expressed in the R1, R6, R3 and R4 cells. The R1 and R6 (two weakly stained cells) were found in the wild-type larval disc (C), but absent in the sag 13L mutant larval disc (D). Wild-type (E) and mutant (F) pupal discs at 24 hours postpupariation. By this time of the eye development, ommatidial rotation and organization were severely disrupted and a small portion of R3 or R4 cells appears to be lost (compare E with F). (G,H) X-gal staining of third-instar larval eye discs derived from ¯ies that carried B38-lacZ with strong b -galactosidase expression in the R7 cell (weak expression is found in other cells, most notably in R1 and R6 cells; S. Butler, Y. Hiromi, unpublished). (G) Wild-type disc; (H) sag 13L mutant disc. Note that R7 cells are missing in the mutant.

116

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

2.2. R1, R6 and R7 cells are missing in the ommatidial preclusters in sag mutants To examine how sag affects the development of photoreceptor cells, anti-Elav antibody was used to visualize all photoreceptor neurons as they were recruited into the ommatidial pre-clusters. In third-instar sag mutant eye discs, the initial formation of the R8 cell appeared normal. Subsequently, more mature pre-clusters contained reduced numbers of Elav-positive cells (Fig. 2B). To ®nd out which cells were missing in the ommatidial pre-clusters, thirdinstar larval eye discs from sag mutant ¯ies that also carried a seven up-lacZ enhancer trap line were stained using antilacZ antibody. Because the seven up-lacZ line (H162; Mlodzik et al., 1990) expresses the lacZ reporter gene in the R1, R6, R3 and R4 cells, these four cells can be visualized by anti-lacZ antibody staining. The R1 and R6 cells (weakly stained cells) appeared in clusters starting at the ®fth row behind the morphogenic furrow in wild-type discs (Fig. 2C). However, these two cells were absent in the mutant disc (Fig. 2D). The positions of the cells in the pre-cluster indicate that the missing cells were R1 and R6 cells. Most of the R3 and R4 cells were present in sag mutant discs at the larval stage (Fig. 2D). At the pupal stage, the mutant ommatidia became dramatically disorganized, showing a loss of some R3 or R4 cells due to the disruptive effect of lacking Sag activity in all eye disc cells (compare Fig. 2E with F). These results suggest that R1 and R6 cells are not formed at the time when they normally appear in the ommatidial preclusters. Using the enhancer trap line B38-lacZ (an enhancer trap insertion in klingon as H214-lacZ) which express b galactosidase strongly in the R7 cell (S. Butler, Y. Hiromi, unpublished data), we showed that the R7 cells were missing in the mutant (Fig. 2G,H). These results suggest that R1, R6 and R7 cells are not formed at the time when they normally appear in the ommatidial pre-clusters.

1998). Thus unlike phyl, sag does not cause cell type transformation. sag not only affects R1, R6 and R7 photoreceptor cells but also the cone and pigment cells. However, the defects in cone and pigment cells could be indirectly caused by the loss of photoreceptor cells. 2.4. Constitutive activation of Ras1 was unable to suppress the formation of R1/R6/R7 cells in sag mutants We have shown previously that sag affects the terminal, tracheal and wing vein developmental pathways regulated by Torso, DFGF-R1 and Egfr RTKs, respectively (Zhang et al., 1998). The sag eye phenotype could be explained by reduced Egfr signaling during ommatidial cell recruitment. To test this, we determined whether sag alters the level of output response from Ras1, an essential component of the Egfr pathway. An activated form of Ras1, sev-Ras1 V12D CAAX (Therrien et al., 1995), causes the transformation of cone cell precursors to supernumerary R7-like cells when expressed in cells of the R7 equivalence group (the R7

2.3. Cone and pigment cells are partially missing in sag mutants In phyl mutants, the precursors of R1, R6 and R7 are transformed into cone cells. To examine whether sag causes a similar cell transformation phenotype, cone and pigment cells were visualized by cobalt sul®de staining of pupal discs. At 50 h postpupariation, four cone cells, two primary pigment cells and several secondary and tertiary pigment cells are all formed in the wild-type discs (Fig. 3A). In sag mutant discs, the numbers of cone and pigment cells were dramatically reduced but not eliminated (Fig. 3B). Many ommatidia contained only one cone cell wrapped around by just one primary pigment cell (Fig. 3B). The numbers of the secondary and tertiary pigment cells were also reduced. The remaining cone and pigment cells varied in numbers from one ommatidium to the next. However, scanning electron microscopy (SEM) showed that the density of eye bristles was not reduced (Zhang et al.,

Fig. 3. Cone and pigment cells are missing in sag mutants. Cobalt sul®de staining of pupal eye discs at 50 h postpupariation. Four cone cells (c), two primary pigment cells (p) and several secondary and tertiary pigment cells were visualized in the wild-type disc (A). All of these cell types are partially missing in the sag 13L mutant disc (B). Black arrow points to an ommatidium containing only one cone cell. Black arrowhead points to one primary pigment cell wrapping around two cone cells. White arrowhead points to the location of a missing secondary pigment cell.

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122 Table 2 Average numbers of R7-like cells per ommatidium a Genotype

Average number of R7-like cells per ommatidium

1/1 sev E4/Y; Sos JC21/11 sev E4/Y; Sos JC21/1 sag 13L

1.0 (n ˆ 698) 0.21 (n ˆ 1173) 0.007 (n ˆ 538)

1/1; sev-Ras1 V12D CAAX/1 sag 13L/1; sev-Ras1 V12D CAAX/1 sev-Ras1 N17/1 sev-Ras1 N17/sag 13L

2.395 (n ˆ 458) 1.483 (n ˆ 512) 0.757 (n ˆ 567) 0.412 (n ˆ 498)

yan 1/yan 1 yan 1 sag 13L/yan 1 sag 13L

2.0 (n ˆ 610) 0.9 (n ˆ 148)

sag 13L/sag X-10 sev-rl SEM/1 sag 13L/sag X-10; sev-rl SEM/1

0.8 (n ˆ 191) 1.6 (n ˆ 197) 0.99 (n ˆ 196)

a

The average numbers of R7-like cell (with small rhabdomeres) per ommatidium were determined from apical tangential sections. At least eight eyes of the various genotypes as shown were sectioned and the R7 cells were counted.

and cone cell precursors and the mystery cell). In sag/1; sev-Ras1 V12D CAAX/1 ¯ies, the formation of supernumerary R7-like cells was reduced (compare Fig. 4D with C; Table 2). Conversely, sag dominantly enhanced the loss of R7

Fig. 4. sag interacts with RTK signaling genes to affect the formation of R7 cells. Apical tangential sections of eyes of the following genotypes are shown. (A) sev E4/Y; Sos JC2/1. (B) sev E4/Y; Sos JC2 1/1 sag 13L. (C) sevRas1 V12D CAAX/1. (D) sag 13L/1; sev-Ras1 V12D CAAX/1. (E) sev-Ras1 N17/1. (F) sev-Ras1 N17/sag 13L. Arrows indicate the R7 cell or extra R7 cells. Note that sag suppressed the formation of R7 cells (also see Table 2).

117

cells caused by the expression of the dominant negative form of Ras1, sev- Ras1 N17 (compare Fig. 4F with E; Table 2). Since Ras1 mediates signaling from both Egfr and Sevenless (Sev) RTKs, the possible role of sag in Sev signaling was tested under a condition where Sev signaling was selectively impaired. sev E4 is a partial loss-of-function allele, but is strong enough to cause the transformation of all R7 precursor cells into cone cells. Sos JC2 is a partially activated form of Sos. Thus the sev E4/Y; Sos JC2/1 ¯y represents a condition where Sev (but not Egfr) signaling is reduced. In sev E4/Y; Sos JC2/1 ¯ies, all outer photoreceptor cells form normally whereas the R7 cells form in only 21% of the ommatidia (Fig. 4A; Rogge et al., 1991; Biggs et al., 1994). In contrast, R7 cells formed in only 0.7% of the ommatidia in ¯ies of the genotype sev E4/Y; Sos JC2 1/1 sag 13L (Fig. 4B; Table 2). This is a signi®cant result because sag 13L/1 ¯ies have normal ommatidia. This level of R7 cell reduction by one copy of sag is similar to what observed for rolled (rl) which encodes the Drosophila Mapk (Biggs et al., 1994). The above results suggest that wild-type sag activity promotes the cell differentiation pathways activated by probably both Egfr and Sev RTKs in the eye. If sag mutations prevented the formation of R1/R6/R7 cells by affecting the Egfr/Ras1 induced recruitment/differentiation pathways, why doesn't sag affect the formation of all photoreceptor cells? To show that the cell type-speci®c effect is not an artifact derived from the partial loss-of-function nature of the alleles used, we tested whether a constitutively activated Ras1 activity could rescue the sag ommatidial phenotype. Homozygous sag clones that also carried sev-Ras1 V12 were generated. The sev-Ras1 V12 transgene expresses activated Ras1 V12 mainly in cells of the R7 equivalence group, but there is also leaky ubiquitous expres-

Fig. 5. sag operates downstream or parallel to Mapk. SEM (A±C) and apical tangential sections (D±F) of ¯y eyes of the following genotypes are shown. (A,D) sag 13L/sag X-10. White arrow indicates an ommatidium with missing photoreceptor cells. (B,E) sag 13L/sag X-10; sev-rl SEM/1. (C,F) sev-rl SEM/1. The white arrowhead points to an ommatidium with extra R7 cells.

118

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

downstream effector pathways. The Ras effector Raf initiates the Mapk phosphorylation cascade which plays a major role in cell proliferation and differentiation. Two other Ras effectors, phosphatidylinositol-3-kinase (PI3-K) and Ral exchange factor, also regulate Ras-dependent activities such as cytoskeletal remodeling that does not involve the Mapk cascade (Katz and McCormick, 1997; Bos, 1998). To determine whether sag affects Mapk signaling, the effect caused by mutating both copies of sag in ¯ies carrying an activated form of Mapk (sev-rl SEM, Brunner et al., 1994) was analyzed. In the eyes of sev-rl SEM/1 ¯ies, 22% of the ommatidia contained two to three R7 cells (Fig. 5F). In the eyes of sag X10/sag13L mutant ¯ies, 19% of the ommatidia lacked the R7 cells (Fig. 5D). The ommatidia of double-mutant ¯ies of genotype sag X10/sag13L; sev-rl SEM/1 were examined and 99% of these ommatidia contained just one R7 cell (Fig. 5E). In other words, the double mutant showed an intermediate phenotype (also compare Fig. 5A with B,C; Table 2). Due to a technical dif®culty, sag 13L/ sag 13L; sev-rl SEM/1 ommatidia could not be recognized. However, we were able to generate and recognize yan 1 sag 13L double homozygous ommatidia. Yan is a substrate of Mapk and an Ets-domain transcriptional repressor (Lai and Rubin, 1992). The Ras1/Mapk cascade counteracts Yan-mediated repression by phosphorylating Yan followed by the rapid degradation of phosphorylated Yan (O'Neill et al., 1994). In the loss-ofFig. 6. Double-mutant clonal analysis. Apical tangential sections through sag 13Lclones (unpigmented area) induced in the genetic backgrounds of sevRas1 V12 (A) and yan 1 (B). The white arrows point to ommatidia with extra R7 photoreceptors. Note that sag 13L; sev-Ras1 V12 ommatidia (A) showed essentially the same phenotype as sag ommatidia (Fig. 1A). In contrast, yan 1 sag 13L ommatidia (B) showed an intermediate phenotype. Note that in both cases, none of the ommatidia inside sag clones contained extra R7 cells.

sion because these ¯ies have reduced viability and contained extra outer photoreceptors in the eye. If R1/R6/R7 cells were selectively affected because the formation of these cells normally requires a higher level of Egfr signaling, sev-Ras1 V12 would be likely to at least partially rescue the formation of these cells. sag 13L; sev-Ras1 V12 ommatidia, as recognized by the lack of pigments in all photoreceptors in the clusters (Fig. 6A; n ˆ 153), were examined. These ommatidia contained only three or four outer photoreceptors and none of these ommatidia contained R7 cells (Fig. 6A). Thus the sag 13L; sev-Ras1 V12 ommatidia showed essentially the same phenotype as sag 13L ommatidia (compared to Fig. 1A). This demonstrated that hyperactivation of Ras1 was not able to rescue the formation of R1/R6/R7 cells. Thus the cell type-speci®c effect of sag mutations probably re¯ects a genuine cell type-speci®c function of the gene. 2.5. sag affects Ras1-mediated Mapk signaling Ras has been shown to signal through at least three different

Fig. 7. The second mitotic wave is reduced in sag mutants. Mitotic cells in the third larval eye discs were detected by BrdU incorporation followed by anti-BrdU antibody staining. The second mitotic waves were indicated by the black arrowheads in the wild-type (A,B) and the sag 13L mutant (C,D) discs. (B,C) are enlargements of the boxed areas in (A,C), respectively. There was an obvious reduction of the number of mitotic cells in sag 13L mutant discs.

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

function yan 1 mutant, the reduction of Yan-mediated repression mimics the effect of Ras1 hyperactivation. Consequently, 90% of the ommatidia contain between one to four (average 2.0) supernumerary R7-like cells; 27% of the ommatidia contain one or two extra outer photoreceptor cells; while 6% of the ommatidia lack one or two outer photoreceptor cells (pigmented region in Fig. 6B; Lai and Rubin, 1992). Clones of yan 1 sag 13L double homozygous cells were induced in (yan 1 sag 13L/yan 1 1 ) ¯ies. The majority (90%) of yan 1 sag 13L double-mutant ommatidia contained the normal complement of photoreceptor cells (unpigmented region in Fig. 6B; Table 2). Thus the phenotype of the yan 1 sag 13L double mutant suggests that yan 1 suppressed the phenotype of sag 13L ommatidia. Since no photoreceptor develops in yan null clones (Rogge et al., 1995), we did not use yan null in our double-mutant analysis. These results suggest that sag affects Ras1-dependent Mapk signaling. The signal mediated by Sag antagonizes Yan to promote photoreceptor cell differentiation at a level downstream of, or parallel to, Yan. 2.6. Mitotic cells in the second mitotic wave are reduced in sag eye discs RTK signaling mediates cell proliferation of imaginal discs (Karim and Rubin, 1998). The eyes from sag 13L or sag 32-3 mutants contained only approximately 50% of the normal number of ommatidia. Third-instar larval eye discs dissected from these mutants were also smaller in size (compare Fig. 7A with C), suggesting that sag may have a role in cell proliferation of the eye disc during the ®rst and second instar larval stages. Two mitotic waves occur at the third-instar larval stage. The ®rst wave consists of asynchronous cell divisions that precede the onset of cluster formation at the morphogenic furrow. The second mitotic wave is a single synchronous division of unrecruited cells excluded from the ®ve-cell pre-clusters. Interestingly, the precursors for the cell types affected in sag mutants are all derived from the progeny of the second mitotic wave. To ®nd out whether the phenotype associated with sag results from a general mitotic defect, mitotic cells of the third-instar larval eye discs were detected by BrdU incorporation followed by anti-BrdU antibody staining. The results showed that the ®rst mitotic wave was not affected, but there was an approximately 50±60% reduction of the number of mitotic cells in the second mitotic wave (compare Fig. 7B with D). 3. Discussion The eye phenotype of a Drosophila mutation semang (sag) was characterized here. Mosaic analysis showed that sag is required for the development of R1/R6/R7 photoreceptors in a cell-autonomous manner. Approximately 50% of the remaining photoreceptor subtypes (R2, R3, R4, R5, R8) were homozygous for sag in the normally constructed mosaic ommatidia, suggesting that wild-type Sag function is

119

not autonomously required for their formation. However, these sag/sag outer photoreceptors are not entirely normal in that they have smaller rhabdomeres. Other cell types affected by sag in the eye are the cone and pigment cells which are reduced in numbers, but the number of the eye bristle cells appears normal. The loss of cone and pigment cells is less severe in weaker sag mutants. This reduction of cone and pigment cells could be an indirect consequence due to the loss of photoreceptor cells in the pre-clusters. The precursor cells for the cell types affected by sag arise from the second mitotic wave. Interestingly, the second mitotic division is reduced in the mutant. When the second mitotic division was completely blocked by the expression of the cyclin-dependent kinase inhibitor p21 CIP/WAF1, the formation of R1, R6, R7 and cone cells was not affected, but the pigment and bristle cells were severely reduced (de Nooij and Hariharan, 1995). This is presumably because there were not enough cells left to form the last two cell types, the pigment and bristle cells, when the second mitotic division was abrogated. Thus although the second mitotic wave is affected in sag mutant discs, this alone could not account for the near complete loss of R1, R6 and the complete loss of R7 cells. In summary, Sag plays a speci®c role in the development of R1, R6 and R7 photoreceptor subtypes although it may also have other functions in the eye (e.g. mitotic division). Using various cell-speci®c markers, we showed that the missing cell types in sag mutants were absent at the time when they normally appear in the ommatidial pre-clusters. This suggests that sag affects the recruitment of precursor cells that give rise to R1, R6, R7 cells. Since the reiterative activation of Egfr RTK triggers the recruitment of all ommatidial cells, sag appears to selectively affect some but not all rounds of Egfr-mediated cell recruitment. Even though the sag alleles analyzed may not be null mutations, it should be pointed out that partial loss-of-function mutations in known RTK signaling genes do not show a cell typespeci®c effect. In addition, constitutive activation of Ras1 was unable to rescue the formation of R1, R6 and R7 cells in sag mutants. This suggests that the loss of speci®c photoreceptors is not simply because these cells require quantitatively higher levels of Egfr signaling than other photoreceptor subtypes. The cell type-speci®c effect of sag mutations probably re¯ects a genuine cell type-speci®c function of the gene. Other evidence that sag affects Egfr-induced cell recruitment in the eye is provided by the extensive genetic interactions between sag and known RTK signaling genes. For example, even though sag/1 eyes do not have any phenotype, this heterozygosity enhanced the loss of R7 cells caused by dominant negative form of Ras1 N17, but suppressed the formation of supernumerary R7 cells caused by activated Mapk SEM. These results are consistent with that Sag activity is critically important in transmitting or responding to Ras-dependent Mapk activation. This is not surprising since Sag has been shown to affect several RTK

120

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

mediated embryonic processes. In embryos lacking both maternal and zygotic sag 1 activity, the posterior expression domains of the two genes (tailless and huckebein) induced by the Torso RTK pathway are reduced and some of the terminal structures deleted (Zhang et al., 1998). These mutant embryos also have reduced and fragmented tracheal branches, suggesting that sag affects DFGF-R1 RTK signaling. Interestingly, sag does not affect the Egfr-mediated patterning of the embryonic ventral ectoderm. However, sag mitotic clones in the wing resulted in the loss of wing vein which requires Egfr signaling for its development (Zhang et al., 1998). Overall, it appears that Sag activity promotes RTK signaling, but Sag does not make an equal impact on all Egfr pathways. If sag primarily affects Egfr signaling during ommatidial pre-cluster assembly, why doesn't sag affect the recruitment of R2, R3, R4, R5 cells? Egfr signaling is highly regulated by a number of mechanisms in different cells and tissues, including autocrine and paracrine feedback regulations (Wasserman and Freeman, 1998; reviewed in Perrimon and Perkins, 1997; Schweitzer and Shilo, 1997). It is conceivable that the sag alleles affected certain special aspects of Egfr-signaling that are unique for the recruitment of R1/R6/ R7 cells. The model for ommatidial assembly is that the Egfr RTK ligand Spitz is initially secreted by the ®rst formed R8 cell, resulting in the recruitment of R2 and R5 cells. Next, Spi is secreted by the three-cell pre-cluster (R8, R2 and R5), resulting in the recruitment of R3 and R4 cells (Freeman 1994). Later, Spi is presumably secreted by the ®ve-cell pre-cluster, resulting in the addition of R1 and R6 cells into the cluster (reviewed in Freeman, 1997). At the cellular level, however, photoreceptor cluster formation is a complex process involving dynamic changes in cell±cell contacts. It has been shown that proteins of the immunoglobulin superfamily and the spectrin membrane skeleton affect the formation of the R7 photoreceptor cell (Butler et al., 1997; Thomas et al., 1998). We could hypothesize that Sag mediates a cell-contact signal required for the development of R1/R6/R7 cells. The existence of such a signal is suggested by the retina aberrant in pattern (rap) gene which is required in the R8 cell for the development of R1/R6/R7 cells (Karpilow et al., 1996). The integration of Egfr signaling with the Sag-mediated signal could presumably result in the determination of R1/R6/R7. This idea would ®t well with the double-mutant analyses that suggest that sag operates genetically downstream of, or parallel to, Mapk/Yan. Further studies are required to address the detailed mechanism by which sag affects speci®c cell types in the eye.

4. Experimental procedures 4.1. Drosophila mutations sag 13L and sag 32-3 are the two strongest loss-of-function alleles among the 12 alleles isolated as suppressors of

Src42A Su(Raf)1 (Lu et al., 1994; Zhang et al., 1998). Src42A suppresses the lethality of a Drosophila Raf mutation, Raf C110 (Melnick et al., 1993). Flies of genotype Raf C110/Y die, whereas those of Raf C110/Y; Src42A Su(Raf)1/1 live. Flies of genotype Raf C110/Y; Src42A Su(Raf)1 1/1 sag, that carry one mutated sag copy, die despite having Src42A Su(Raf)1. sag 13L and sag 32-3 are pupal lethal mutations, one copy of which suppresses Src42A Su(Raf)1 fully (i.e. none of the Raf C110/Y; Src42A Su(Raf)1 1/1 sag 13L ¯ies survive). This suggests that sag 13L and sag 32-3 are strong loss-of-function alleles if not null alleles. The chromosome de®ciencies uncovering the sag region between l(2)k07110 and l(2)k14517 are not available to test if they are nulls at this time. sag X-10 is a weak allele. Flies of genotype sag X-10/ sag 13L or sag X-10/sag 32-3 are viable but have rough eyes due to disorganized ommatidial arrays and reduced numbers of photoreceptor cells. The seven up-lacZ line (H162; Mlodzik et al., 1990), sev-Ras V12D CAAX (Therrien et al., 1995), sevRas1 N17 (M. Simon, unpublished) and stocks used for making mitotic clones in the eye were kindly provided by G. Rubin's laboratory. sev E4; Sos JC2 stock was kindly provided by U. Banerjee's laboratory. B38-LacZ, an enhancer trap insertion in klingon (allelic to H214-lacZ; S. Butler and Y. Hiromi, unpublished), was kindly provided by Dr Y. Hiromi. From B38-LacZ, LacZ is strongly expressed in the R7 cell, and weakly in other cells (most notably R1 and R6 cells; Y. Hiromi, unpublished). Su(Raf)1

4.2. Mosaic analysis sag mutant clones in the eye were induced by the FLP technique (Xu and Rubin, 1993). The eye clones in wildtype background were induced in ¯ies of genotype y w P(ry 1; hs-FLP) 12; P(ry 1; hs-neo; FRT) 42D sag/P(ry 1; hsneo; FRT) 42D P(ry 1; w 1) 47A. Small patches of unpigmented sag homozygous cells were found to form disorganized ommatidial arrays. No clone in the eye was recovered in control experiments either without heat shock treatment or with heat shock treatment in the absence of P(ry 1; hsFLP) 12. The eye clones in yan 1 background were induced in yan 1 homozygous ¯ies using the same method as above. The eye clones in the activated Ras1 V12 background were induced in ¯ies carrying sev-Ras V12 (T2B) (Fortini et al., 1992; Karim et al., 1996). 4.3. Histology To obtain mutant eye imaginal discs, a sag/SM6-TM6B, Tb stock was used where SM6-TM6B is a balancer carrying Tubby (Tb). Tb causes shorter larvae and pupae, allowing sag/sag eye discs to be isolated from Tb 1 larvae or pupae. Antibody staining followed standard procedures. Cobalt sul®de staining and BrdU labeling followed the methods described in Melamed and Trujillo-Cenoz (1975) and Wolff and Ready (1991), respectively. Sectioning of eye tissues followed that described in Tomlinson and Ready

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

(1987). Scanning electron microscopy (SEM) followed the method described in Zhang et al. (1998).

Acknowledgements We are grateful to the following people for sending various ¯y stocks that are essential to this work: Y. Hiromi, T. Lebestky, U. Banerjee, Z.C. Lai, K. Matthew and T. Laverty. We also wish to thank B. Cutler for help with eye sectioning, and D. Ruden and J. Parr for helpful comments on the manuscript. This work was supported by the American Cancer Society (#RPG-96-13504-DDC), and partially by NIH (1 P50 DK57301-01) to X.L.

References Basler, K., Yen, D., Tomlinson, A., Hafen, E., 1990. Reprogramming cell fate in the developing Drosophila retina: transformation of R7 cells by ectopic expression of rough. Genes Dev. 4, 728±739. Biggs III, W.H., Zavitz, K.H., Dickson, B., van der Straten, A., Brunner, D., Hafen, E., Zipursky, S.L., 1994. The Drosophila rolled locus encodes a MAP kinase required in the sevenless signal transduction pathway. EMBO J. 13, 1628±1635. Bos, J.L., 1998. All in the family? New insights and questions regarding interconnectivity of Ras, Rap1 and Ral. EMBO J. 17, 6776±6782. Brunner, D., Oellers, N., Szabad, J., Biggs 3rd, W.H., Zipursky, S.L., Hafen, E., 1994. A gain-of-function mutation in Drosophila MAP kinase activates multiple receptor tyrosine kinase signaling pathways. Cell 76, 875±888. Butler, S.J., Ray, S., Hiromi, Y., 1997. klingon, a novel member of the Drosophila immunoglobulin superfamily, is required for the development of the R7 photoreceptor neuron. Development 124, 781±792. Chang, H.C., Solomon, N.M., Wassarman, D.A., Karim, F.D., Therrien, M., Rubin, G.M., Wolff, T., 1995. phyllopod functions in the fate determination of a subset of photoreceptors in Drosophila. Cell 80, 463± 472. de Nooij, J.C., Hariharan, I.K., 1995. Uncoupling cell fate determination from patterned cell division in the Drosophila eye. Science 270, 983± 985. Dickson, B.J., Dominguez, M., van der Straten, A., Hafen, E., 1995. Control of Drosophila photoreceptor cell fates by phyllopod, a novel nuclear protein acting downstream of the Raf kinase. Cell 80, 453±462. Fortini, M.E., Simon, M.A., Rubin, G.M., 1992. Signalling by the sevenless protein tyrosine kinase is mimicked by Ras1 activation. Nature 355, 559±561. Freeman, M., 1994. The spitz gene is required for photoreceptor determination in the Drosophila eye where it interacts with the EGF receptor. Mech. Dev. 48, 25±33. Freeman, M., 1996. Reiterative use of the EGF receptor triggers differentiation of all cell types in the Drosophila eye. Cell 87, 651±660. Freeman, M., 1997. Cell determination strategies in the Drosophila eye. Development 124, 261±270. Heberlein, U., Mlodzik, M., Rubin, G.M., 1991. Cell-fate determination in the developing Drosophila eye: role of the rough gene. Development 112, 703±712. Katz, M.E., McCormick, F., 1997. Signal transduction from multiple Ras effectors. Curr. Opin. Genet. Dev. 7, 75±79. Karim, F.D., Rubin, G.M., 1998. Ectopic expression of activated Ras1 induces hyperplastic growth and increased cell death in Drosophila imaginal tissues. Development 125, 1±9. Karim, F.D., Chang, H.C., Therrien, M., Wassarman, D.A., Laverty, T.,

121

Rubin, G.M., 1996. A screen for genes that function downstream of Ras1 during Drosophila eye development. Genetics 143, 315±329. Karlovich, C.A., Bon®ni, L., McCollam, L., Rogge, R.D., Daga, A., Czech, M.P., Banerjee, U., 1995. In vivo functional analysis of the Ras exchange factor son of sevenless. Science 268, 576±579. Karpilow, J.M., Pimentel, A.C., Shamloula, H.K., Venkatesh, T.R., 1996. Neuronal development in the Drosophila compound eye: photoreceptor cells R1, R6, and R7 fail to differentiate in the retina aberrant in pattern (rap) mutant. J. Neurobiol. 31, 149±165. Kimmel, B.E., Heberlein, U., Rubin, G.M., 1990. The homeo domain protein rough is expressed in a subset of cells in the developing Drosophila eye where it can specify photoreceptor cell subtype. Genes Dev. 4, 712±727. Lai, Z.C., Rubin, G.M., 1992. Negative control of photoreceptor development in Drosophila by the product of the yan gene, an ETS domain protein. Cell 70, 609±620. Li, S., Li, Y., Carthew, R., Lai, Z.C., 1997. Photoreceptor cell differentiation requires regulated proteolysis of the transcriptional repressor Tramtrack. Cell 90, 469±478. Lu, X., Li, Y., 1999. Drosophila Src42A is a negative regulator of RTK signaling. Dev. Biol. 208, 233±243. Lu, X., Melnick, B.M., Hsu, J.C., Perrimon, N., 1994. Genetic and molecular analyses of mutations involved in Drosophila raf signal transduction. EMBO J. 13, 2592±2599. Melamed, J., Trujillo-Cenoz, O., 1975. The ®ne structure of the eye imagina disks in muscoid ¯ies. J. Ultrastruct. Res. 51, 79±93. Melnick, M.B., Perkins, L.A., Lee, M., Ambrosio, L., Perrimon, N., 1993. Developmental and molecular characterization of mutations in the Drosophila-raf serine/threonine protein kinase. Development 118, 127±138. Mlodzik, M., Hiromi, Y., Webber, U., Goodman, C.S., Rubin, G.M., 1990. The Drosophila seven-up gene, a member of the steroid receptor gene superfamily, controls photoreceptor cell fates. Cell 60, 311±324. Olivier, J.P., Raabe, T., Henkemeyer, M., Dickson, B., Mbamalu, G., et al., 1993. A Drosophila SH2±SH3 adaptor protein implicated in coupling the sevenless tyrosine kinase to an activator of Ras guanine nucleotide exchange, Sos. Cell 73, 179±191. O'Neill, E.M., Rebay, I., Tjian, R., Rubin, G.M., 1994. The activities of two Ets-related transcription factors required for Drosophila eye development are modulated by the Ras/MAPK pathway. Cell 78, 137±146. Perrimon, N., Perkins, L.A., 1997. There must be 50 ways to rule the signal: the case of the Drosophila EGF receptor. Cell 89, 13±16. Raabe, T., Riesgo-Escovar, J., Liu, X., Bausenwein, B.S., Deak, P., 1996. DOS, a novel pleckstrin homology domain-containing protein required for signal transduction between sevenless and RAS1 in Drosophila. Cell 85, 911±920. Rogge, R., Green, P.J., Urano, J., Horn-Saban, S., Mlodzik, M., Shilo, B.Z., Hartenstein, V., Banerjee, U., 1995. The role of yan in mediating the choice between cell division and differentiation. Development 121, 3947±3958. Rogge, R.D., Karlovich, C.A., Banerjee, U., 1991. Genetic dissection of a neurodevelopmental pathway: Son of sevenless functions downstream of the sevenless and EGF receptor tyrosine kinases. Cell 64, 39±48. Schweitzer, R., Shilo, B.-Z., 1997. A thousand and one roles for the Drosophila EGF receptor. Trends Genet. 13, 191±196. Tang, A.H., Neufeld, T.P., Kwan, E., Rubin, G.M., 1997. PHYL acts to down-regulate TTK88, a transcriptional repressor of neuronal cell fates, by a SINA-dependent mechanism. Cell 90, 459±467. Takahashi, F., Endo, S., Kojima, T., Saigo, K., 1996. Regulation of cell±cell contacts in developing Drosophila eyes by Dsrc41, a new, close relative of vertebrate c-src. Genes Dev. 10, 1645±1656. Thomas, G.H., Zarnescu, D.C., Juedes, A.E., Bales, M.A., Londergan, A., Korte, C.C., Kiehart, D.P., 1998. Drosophila betaHeavy-spectrin is essential for development and contributes to speci®c cell fates in the eye. Development 125, 2125±2134. Therrien, M., Chang, H.C., Solomon, N.M., Karim, F.D., Wassarman, D.A.,

122

Q. Zhang, X. Lu / Mechanisms of Development 95 (2000) 113±122

Rubin, G.M., 1995. KSR, a novel protein kinase required for RAS signal transduction. Cell 83, 879±888. Tomlinson, A., Ready, D., 1987. Neuronal differentiation in the Drosophila ommatidium. Dev. Biol. 120, 366±376. Tomlinson, A., Kimmel, B.E., Rubin, G.M., 1988. rough, a Drosophila homeobox gene required in photoreceptors R2 and R5 for inductive interactions in the developing eye. Cell 55, 771±784. Wolff, T., Ready, D.F., 1991. Cell death in normal and rough eye mutants of Drosophila. Development 113, 825±839. Wolff, T., Ready, D.F., 1993. The Development of the Drosophila mela-

nogaster, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 1277±1325. Wasserman, J.D., Freeman, M., 1998. An autoregulatory cascade of EGF receptor signaling patterns the Drosophila egg. Cell 95, 355± 364. Xu, T., Rubin, G.M., 1993. Analysis of genetic mosaics in developing and adult Drosophila tissues. Development 117, 1223±1237. Zhang, Q., Zheng, Q., Lu, X., 1998. A genetic screen for modi®ers of Drosophila Src42A identi®es mutations in Egfr, rolled and a novel signaling gene. Genetics 151, 697±711.