Senseless represses nuclear transduction of Egfr pathway activation

Interactions between receptors and their ligands are frequently used during development to distinguish one cell type from another. While much is known about the consequences of ligand/receptor binding in the signal-receiving cell, the mechanisms by which autocrine effects are prevented in the signal-producing cell are less clear. Ligand/receptor interactions are of critical importance during Drosophila eye development, as recruitment and differentiation of almost all photoreceptors are dependent on high levels of activation of the Epidermal growth factor receptor (Egfr) pathway(Freeman, 1996; Kumar et al., 1998; Lesokhin et al., 1999; Yang and Baker, 2001). Binding of activating ligand to Egfr results in a cascade of membranous and cytoplasmic phosphorylation events, ultimately resulting in the translocation of phosphorylated ERK to the nucleus where it phosphorylates and activates the P2 isoform of Pointed, an ETS domain-containing transcription factor(Brunner et al., 1994; Kumar et al., 2003; O'Neill et al., 1994). Pointed-P2 is then thought to induce transcription of a second pointed isoform, P1, which encodes the final nuclear effector of the Egfr pathway (O'Neill et al., 1994). Egfr activation in the Drosophila eye is induced by the Spitz (Spi) ligand (Freeman,1994; Tio et al.,1994; Tio and Moses,1997). Interestingly, while Spi is produced initially and primarily by R8, the founding photoreceptor of each ommatidium, R8 itself is the only photoreceptor that does not require Egfr activation to differentiate(Fig. 1A)(Kumar et al., 1998; Yamada et al., 2003; Yang and Baker, 2001).

Although Egfr pathway signaling is not required for R8 differentiation,several lines of evidence indicate that the uppermost tiers of the Egfr pathway are activated at a high level in the differentiating R8 photoreceptor:(1) R8 is very probably exposed to high levels of secreted Spi, the Egfr ligand in the eye, as R8 itself is the major source of Spi during photoreceptor recruitment; (2) Egfr protein is ubiquitously expressed at high levels at the time when early photoreceptor fates are assumed; and (3) high levels of ERK activation, as assayed by an antibody to dpERK (dual phosphorylated ERK), are detected in R8 intermediate groups and R8 itself, but only in the cytoplasm(Chen and Chien, 1999; Freeman, 1994; Kumar et al., 2003; Kumar et al., 1998; Lesokhin et al., 1999; Rawlins et al., 2003; Spencer and Cagan, 2003; Tio et al., 1994; Zak and Shilo, 1992). However,despite this apparent activation of the pathway at the level of the receptor and within the cytoplasm, a number of nuclear indicators of Egfr activation are either not detected or decreased in R8. First, Rough (Ro), a homeodomain-containing protein that is an early target of Egfr activation, is not expressed in R8 (Dokucu et al.,1996; Frankfort et al.,2001; Kimmel et al.,1990). Second, transcription of argos (aos),which encodes a negative regulator of Egfr signaling, is decreased in R8 relative to other photoreceptors (Lesokhin et al., 1999). As aos expression is thought to occur proportionately to the degree of Egfr signaling in that cell, this implies that the Egfr pathway is activated at lower levels in R8 than in non-R8 photoreceptors (Golembo et al.,1996). It is probable that this reduction in Egfr signaling from cytoplasm to nucleus in R8 is developmentally important because high levels of activation of Egfr signaling induced by either ectopic expression or the Egfr^Elp mutation result in the development of very few R8 photoreceptors, suggesting that high levels of Egfr signaling are not compatible with R8 differentiation(Dominguez et al., 1998; Kumar et al., 1998). Thus, R8 serves as the signal-producing epicenter for Egfr-dependent recruitment in developing ommatidia, yet must remain refractory to the very signaling events that it initiates.

R8 development requires the actions of the proneural gene atonal(ato) and its downstream effector, senseless(sens), which encodes a conserved C2H2 zinc finger transcription factor (Frankfort et al., 2001; Jarman et al., 1994; Nolo et al., 2000). Several specific substages of R8 development have been identified (reviewed by Frankfort and Mardon, 2002). In particular, there is a distinction between R8 selection, the choosing of an R8 precursor from a group of developmentally equivalent cells, and R8 differentiation, a later process by which R8 fate is `locked' and the expression of neural markers is initiated. sens is required only in R8, and mutations in sens result in a total failure of R8 differentiation despite normal selection of R8 precursor cells (called presumptive R8s, or pre-R8s) (Frankfort et al., 2001; Nolo et al.,2000). Furthermore, in sens mutants, the pre-R8 cell instead consistently differentiates as a founder photoreceptor of the R2/R5 subtype (Fig. 1B)(Frankfort et al., 2001; Nolo et al., 2000). R2/R5 is normally the first subtype of paired non-R8 photoreceptors to be recruited by Egfr activation via Spi secretion from R8(Fig. 1A)(Tomlinson and Ready, 1987). Like wild-type R2/R5 cells, the R2/R5 cells that develop from the pre-R8 in sens mutants express Ro, which is required for R2/R5 differentiation(Frankfort and Mardon, 2002; Kimmel et al., 1990; Tomlinson et al., 1988).

R8 differentiation is restored when ro function is removed in sens mutant tissue, suggesting that Sens-mediated repression of ro is a critical event during R8 differentiation. However, complete loss of ro function does not rescue R8 differentiation in all sens mutant ommatidia. Therefore, it is probable that Sens also has a function in R8 which is distinct from its role as a repressor of ro(Frankfort et al., 2001). Since the pre-R8 cell in sens mutants consistently differentiates as a cell type that is normally Egfr dependent (R2/R5) and Egfr pathway activation appears incompatible with R8 differentiation, it is possible that this additional function of Sens may involve repression of Egfr signaling in R8(Fig. 1C).

We show that Sens prevents transduction of Egfr signaling to the nucleus of R8, despite both Egf receptor activation and ERK phosphorylation. This is accomplished via a novel regulatory mechanism – Sens causes the transcriptional repression of the P1 isoform of pointed. This ensures that the ommatidial signaling center is protected from the effects of autocrine stimulation by secreted Spi, and that R8 differentiation and normal ommatidial organization are preserved. Finally, analogous relationships that exist between Sens and Egfr pathway orthologs in T-lymphocytes may establish R8 development as a novel system with which to study lymphomagenesis,apoptosis and cancer.

All Drosophila crosses were carried out at 25°C on standard media. For clonal analyses in Figs 2 and 3, GMRp35;rho-3¹ rho-1^7M43 FRT80B/TM6B(Wasserman et al., 2000), w; sens^E2 FRT80B/TM6B(Frankfort et al., 2001), w; egfr^Elp px/CyO; sens^E2 FRT80B/TM6B (this work), or GMRp35; rho-3¹ rho-1^7M43 sens^E2FRT80B/TM6B (this work) were crossed to y w hsFLP122;P[w⁺=ubiGFP]61EF M(3)i(55) P[w⁺]70C FRT80B/TM6B(Frankfort et al., 2001); GMRp35; spi^SC1 FRT40A/CyO P[w⁺=hshid];sens^E1 FRT80B/TM6B (this work)(Tio et al., 1994) was crossed to y w hsFLP; P[w⁺=arm-lacZ] FRT40A; P[w⁺=arm-lacZ]FRT80B (this work); or GMRp35; rho-3¹ rho-1^7M43sens^E2 FRT80B ro^X63/TM6B (this work) was crossed to y w hsFLP122; P[w⁺=ubiGFP]61EF M(3)i(55) P[w⁺]70C FRT80B ro^X63/TM6B(Frankfort et al., 2001). Clones were induced as previously described(Frankfort et al., 2001). Misexpression of sens in the egfr^Elp background was accomplished through a cross between w; egfr^Elp px/CyO;ey-GAL4 and w; egfr^Elp px/CyO; UAS-sens. The flpout-GAL4 clones shown in Fig. 4 were generated as previously described(Dominguez et al., 1998) with crosses between y w P[w⁺=act<cd2<GAL4]; hsFLP MKRS/Tb and either UAS-Egfr^act UAS-lacZ(Dominguez et al., 1998), UAS-Egfr^act pnt¹²⁷⁷ (this work), or UAS-sens; UAS-Egfr^act pnt¹²⁷⁷ (this work). Clonal misexpression experiments in Fig. 5 were performed essentially as described, except with crosses between y w hsFLP; FRT40A sca-GAL4/CyO P[w⁺=hshid] females and w; M(2)24F P[w⁺=arm-lacZ] P[w⁺=tub-GAL80]FRT40A/+; UAS-gene/TM6B males, where `gene' represents Egfr^act, pnt-P1, or ro (this work)(Pappu et al., 2003).

pnt¹²⁷⁷ (pnt-lacZ) is an enhancer trap line in the pointed locus which is expressed in many cells posterior to the morphogenetic furrow (see Fig. 3). pnt-P1 transcript is normally detected in the morphogenetic furrow in intermediate groups and posteriorly in developing ommatidia (Rawlins et al.,2003). While the expression patterns posteriorly are very similar, pnt-lacZ cannot reflect the entire expression pattern of pnt-P1 because it is not expressed in the intermediate groups. It is also not clear whether pnt-lacZ expression is specific to the P1 or P2 isoform of pnt, or if it represents a combination of the expression patterns of both.

The sca-GAL4 line was a gift from Yash Hiromi. This line is expressed at high levels in R8, beginning with the first column of single R8 cells and at lower levels in both other cells in the morphogenetic furrow and some non-R8 cells posteriorly (our unpublished observations).

All antibodies were used and confocal microscopy performed as previously described (Frankfort et al.,2001). Adult eyes were fixed, embedded and sectioned as previously described (Frankfort et al.,2001). Whole adult eyes were examined using a Leica MZ16 stereomicroscope and processed with Image-Pro Plus image analysis software.

proc4-2 (Tomlinson et al.,1988), was digested with EcoRI to yield a 1.2 kb ro cDNA lacking the coding sequence for the first four amino acids. This fragment was subcloned into pBluescript containing an EcoRI site, which was modified with the following adapters:5′-AATTGCCTCAAACGAAATGCAG and 5′-AATTCTGCATTTCGTTTGAGGC. This created a modified ro cDNA that encodes a protein N terminus of MQNSSK instead of the wild-type MQRHK. Several protein and biochemistry prediction programs were used to assess the modified ro cDNA and no changes in behavior compared to wild-type were predicted. The ro cDNA was then excised from pBluescript with an XhoI/XbaI digestion and directionally cloned into pUAST. Vector DNA was injected into Drosophila embryos according to standard protocols. Ro protein was detected in wing and leg imaginal discs by antibody staining when UAS-ro was misexpressed with dpp-GAL4, and transgenic UAS-ro animals were sufficient to rescue the ro^X63 mutant phenotype when misexpressed with hsGAL4, suggesting that the encoded Ro protein is functional in vivo(Kimmel et al., 1990).

In our analysis of sens function in R8 differentiation, we found that the extra R2/R5 cell that develops from the pre-R8 in sensmutants expresses Ro, which is normally expressed in R2/R5 but not R8(Dokucu et al., 1996; Frankfort et al., 2001; Kimmel et al., 1990). Ro is expressed downstream of Egfr pathway activation, and both ro function and high levels of Egfr pathway activation are required for R2/R5 differentiation (Dominguez et al.,1998; Freeman,1996; Hayashi and Saigo,2001; Tomlinson et al.,1988; Yang and Baker,2001). Since the pre-R8 cell consistently expresses Ro and differentiates as an R2/R5 cell in sens mutants, we hypothesized that this transformation occurs as a consequence of high levels of Egfr activation in the pre-R8 cell.

We tested this hypothesis by simultaneously removing sens function and blocking Egfr activation in the developing Drosophila eye(Fig. 2A-C). We blocked Egfr activation by removing function of both rhomboid-1 (rho-1)and rhomboid-3 (rho-3; FlyBase: roughoid, ru). Loss of both rho-1 and rho-3 function prevents processing of secreted Egfr ligands, including Spi, and results in the loss of all ERK (MAP kinase) activation (Urban et al.,2002; Wasserman et al.,2000). Furthermore, loss of rho-1 and rho-3phenocopies Egfr loss-of-function in that only R8 cells differentiate(Fig. 2A)(Wasserman et al., 2000). Loss of sens function results in pre-R8 differentiation as a founder R2/R5 cell which is sufficient to recruit a reduced number of photoreceptors(Fig. 2B). However, the absence of rho-1, rho-3 and sens together causes total photoreceptor loss, except for a few photoreceptors near the clonal boundary that are rescued non-autonomously by neighboring wild-type cells that produce and process Spi appropriately (Fig. 2C) (Frankfort et al.,2001). A similar phenotype is detected in tissue mutant for both spi and sens (Fig. 2D). This loss of photoreceptors seen in rho-1 rho-3 sensand spi sens mutants is not due to cell death because apoptosis was prevented in these experiments by expression of GMR-p35 (see Materials and methods) (Hay et al.,1994). Furthermore, pre-R8 selection still occurs in both rho-1 rho-3 and rho-1 rho-3 sens mutant tissue, suggesting that a potential founding photoreceptor is present(Fig. 2E,F). Therefore, our interpretation of these results is that, in the absence of sensfunction, pre-R8 differentiation as a founder R2/R5 photoreceptor requires activation of the Egfr signaling pathway via the Spi ligand. In other words,in sens mutants, the pre-R8 switches from a Spi/Egfr-independent R8 differentiation pathway to a Spi/Egfr-dependent R2/R5 differentiation pathway.

When rho-1 and rho-3 function are removed in sens ro double mutants, R8 differentiation does not occur(Fig. 2G). This suggests that the requirement in the pre-R8 cell for Egfr activation remains even when ro function is removed, and that the ro-independent function of sens may involve a relationship with the Egfr pathway. Specifically, as the pre-R8 normally does not require Egfr activation but becomes completely dependent on Egfr activation when sens function is removed, we hypothesized that sens normally acts as a repressor not only of ro, but also of Egfr pathway activation in R8. This potential function of sens as a repressor of Egfr signaling is supported by genetic interactions between sens and the gain-of-function Egfr^Elp mutation. Egfr^Elp homozygotes have a greatly reduced number of ommatidia with large gaps of pigmented tissue between them (Fig. 3A)(Baker and Rubin, 1989). In contrast, sens mutant tissue is disrupted in appearance but does not contain undifferentiated gaps between ommatidia(Fig. 3B)(Frankfort et al., 2001). However, when clones of sens mutant tissue are induced in a background that is heterozygous for the Egfr^Elp mutation,gaps of undifferentiated tissue appear between ommatidia, a phenotype very similar to that of Egfr^Elp homozygotes(Fig. 3C). Thus, loss of sens function strongly enhances the Egfr^Elpheterozygous phenotype such that it closely approximates that of Egfr^Elp homozygotes. If this enhancement occurs by derepression of Egfr signaling by the loss of sens function, then misexpression of sens in an Egfr^Elp homozygote might have the opposite effect and suppress the phenotype of ommatidial loss and interommatidial gaps. Indeed, misexpression of UAS-sens with ey-GAL4 has precisely these effects on Egfr^Elphomozygotes (Fig. 3D). Together, these gain- and loss-of-function experiments suggest that sens functions as a powerful negative regulator of the Egfr pathway during Drosophila eye development. Since the expression of Sens is tightly restricted to R8 and the primary sens mutant phenotype occurs in the pre-R8 cell, it is most likely that this repression occurs specifically in the differentiating R8 photoreceptor. We therefore looked at expression of an enhancer trap in pointed (pnt-lacZ), which encodes the nuclear effector of the Egfr pathway. Consistent with our hypothesis, pnt-lacZ, while expressed in many non-R8 photoreceptors as they differentiate, is not expressed in Sens-expressing R8 cells(Fig. 3E,F; Materials and methods) (Scholz et al.,1993).

While the Egfr pathway is probably activated at a high level at the cell membrane and in the cytoplasm of R8, expression of nuclear outputs of the pathway is low (Fig. 3E,F, see Introduction). Moreover, whereas loss-of-function mutations in all major Egfr pathway members have no effect on R8 differentiation, high levels of activation of Egfr signaling as a result of either ectopic expression or Egfr^Elp mutations result in the development of very few R8 photoreceptors (Dominguez et al.,1998; Kumar et al.,1998; Lesokhin et al.,1999; Yamada et al.,2003; Yang and Baker,2001). Thus, the reduction in Egfr activation from high cytoplasmic levels to low nuclear levels in R8 may be of developmental importance. Since Sens acts as a negative regulator of the Egfr pathway, we hypothesized that Sens mediates this critical decrease in Egfr signaling from membrane/cytoplasm to nucleus in R8.

To test this hypothesis, we ubiquitously expressed an activated form of Egfr (Egfr^act) in small clones using flpout-GAL4. We first looked at Egfr^act clones positioned posterior to the morphogenetic furrow (MF). Neural differentiation occurs throughout such clones (Fig. 4A). Since ectopic Egfr activation is sufficient to induce photoreceptor differentiation prior to passage of the MF, these clones probably represent a field of cells that had already differentiated as neurons by the time the MF reached it(Dominguez et al., 1998). Consistent with the hypothesis that high levels of Egfr activation are not compatible with R8 differentiation, these clones show a cell-autonomous lack of Sens expression (Fig. 4A). Anterior to the MF, activation of Egfr signaling causes precocious neural development autonomously, and induces ectopic MFs non-autonomously. These ectopic MFs express Ato, Sens and dpERK appropriately(Fig. 4B,C)(Dominguez et al., 1998). Furthermore, the ectopic neurons generated with this system express pnt-lacZ at a high level, indicating that the nuclear target of Egfr activation is being induced and the canonical Egfr signaling pathway is probably the cause of neural differentiation(Fig. 4D). However, when sens is co-misexpressed along with Egfr^act,pnt-lacZ expression is prevented and neural differentiation is severely reduced in the cells that express sens, and ectopic MFs are not established (Fig. 4E,F). In contrast, dpERK expression still occurs when sens is co-misexpressed with Egfr^act, indicating that the Egfr pathway is being activated at the cell membrane and within the cytoplasm(Fig. 4F). This suggests that Sens cell-autonomously blocks transduction of the activated Egfr pathway to the nucleus and is sufficient to prevent the effects of cell membrane activation of the Egfr pathway. These results are also consistent with our proposed role for Sens as a repressor of high levels of Egfr signaling in R8.

If Sens acts to reduce Egfr signaling from the cytoplasm to the nucleus of R8, we hypothesized that activation of Egfr signaling downstream of the point at which Sens blocks the pathway could disrupt normal R8 differentiation,whereas activation of the pathway upstream of this point would have little or no effect. To test this hypothesis, we misexpressed members of the Egfr pathway in R8 using sca-GAL4 (Materials and methods). When Egfr^act or ras1^val12 (an activated form of Ras that functions in the cytoplasm upstream of ERK) is expressed in R8 with this system there is no appreciable effect on Sens, Boss, or Ro expression in third instar eye imaginal discs(Fig. 5A-C, not shown). This suggests that R8 differentiation proceeds normally when the Egfr pathway is activated at the cell membrane or within the cytoplasm of R8 and is consistent with our proposed role for Sens in R8. Since Sens acts as a repressor of pnt transcription, we also misexpressed both isoforms of pntin R8. Interestingly, misexpression of the P2 isoform of pnt(pnt-P2), or an activated form of pnt-P2, also has no effect on R8 differentiation (not shown) (Halfon et al., 2000). However, misexpression of pnt-P1 causes a disruption in Sens expression such that Sens-expressing nuclei are displaced apically in the imaginal disc (Fig. 5D,E). Since photoreceptor nuclei move basally during neuronal differentiation, this implies that misexpression of pnt-P1 in R8 may disrupt R8 differentiation. Consistent with this, many sca-GAL4× UAS-pnt-P1 adult ommatidia do not contain small rhabdomeres,suggesting an absence of R8 (Fig. 5F). Ommatidia also contain a variable number of photoreceptors and these adult phenotypes are very similar to the sensloss-of-function phenotype. These results imply that pnt-P1, but not pnt-P2, may be a target of sens repression. Misexpression of ro, an early target of Egfr signaling, has a more profound effect on R8 development as both Sens and Boss expression are absent(Fig. 5G, not shown). Adult ommatidia lack small rhabdomeres but are otherwise of relatively normal construction (Fig. 5H). These results are consistent with the known function of ro as a critical determinant of R2/R5 cell fate determination, as well as with our previous model in which Sens acts as a repressor of ro. Together, these data strongly suggest an important additional role for Sens as a novel nuclear repressor of pnt-P1 (Fig. 6).

Our work suggests that Sens acts to ensure that the organizing center of each ommatidium is refractory to the developmental signals it produces –the R8 cell can secrete Spi and even activate Egfr on its own cell membrane,yet remains protected from the deleterious effects of activation of Pnt and other Egfr targets, such as Ro, in R8.

The mechanism by which Sens regulates the discrepancy between levels of Egfr activation at the receptor/cytoplasmic and nuclear levels in R8 is probably through repression of pnt transcription. This is supported by the observation that pnt transcription is not induced by misexpression of an activated form of Egfr when sens is co-misexpressed (Fig. 4). Furthermore, expression of the pnt-P1 isoform in R8 disrupts R8 differentiation (Fig. 5D-F). As misexpression of pnt-P2 has no effect on R8 differentiation, this suggests that Sens negatively regulates transcription of pnt-P1, but not pnt-P2. This mode of regulation is consistent with established models for transduction of the Egfr signal to the nucleus. Specifically, ERK phosphorylates Pnt-P2, which is thought to be a transient positive regulator of pnt-P1 transcription (Brunner et al., 1994; O'Neill et al.,1994). In our model, transduction of Egfr activation occurs all the way into the nucleus of R8, but Sens represses the pathway at the final step – positive regulation of pnt-P1 by Pnt-P2(Fig. 6). When sensfunction is removed, the block on pnt-P1 transcription is relieved,and Pnt-P1 can exert its transcriptional effects on the nucleus, including ro induction.

There is evidence that pnt-P1 transcription can be regulated by Egfr signaling independently of pnt-P2 during Drosophilaembryogenesis (Gabay et al.,1996). If this is the case during eye development, our model would remain essentially the same – Sens would still act as a negative regulator of pnt-P1 in R8. However, this regulation would occur independently of pnt-P2 rather than downstream of pnt-P2.

Sens is also a potent negative regulator of ro and this relationship appears to specifically affect the cell fate decision between R8 and R2/R5 differentiation (Fig. 5G,H) (Frankfort et al.,2001). Several lines of evidence suggest that Sens-mediated repression of ro is distinct from other effects of Sens in R8. First,loss of ro function does not rescue R8 differentiation in all ommatidia in sens mutants(Frankfort et al., 2001). Second, even those R8 cells that do differentiate in sens ro double mutants require Spi/Egfr pathway activation(Fig. 2G). Third, misexpression of ro in R8 causes a different phenotype than misexpression of pnt-P1 in R8. Specifically, even though Egfr pathway activation is necessary and sufficient for Ro expression, misexpression of pnt-P1in R8 does not cause an obvious cell fate transformation from R8 to R2/R5,while misexpression of ro in R8 does(Fig. 5D-H)(Dominguez et al., 1998; Hayashi and Saigo, 2001). Indeed, R8 markers are still expressed when pnt-P1 is misexpressed in R8. However, aberrant nuclear movements and the absence of small rhabdomeres at the level of R8 in adults suggest that misexpression of pnt-P1does perturb R8 differentiation (Fig. 5D). Together, these results suggest that Sens repression of pnt-P1 occurs independently of Sens function as a repressor of ro, and that Sens-mediated repression of pnt-P1 is probably required for normal R8 differentiation upstream or independently of cell fate determination (Fig. 6).

Since Sens acts as a transcription factor and its mammalian homolog, Gfi-1,binds directly to enhancer regions of Ets1 and Ets3, two mammalian orthologs of pnt, is it possible that Sens repression of pnt-P1 expression occurs directly(Duan and Horwitz, 2003; Nolo et al., 2000; Zweidler-Mckay et al., 1996). Gfi-1 also interacts with nuclear matrix proteins to repress transcription(McGhee et al., 2003). Thus,it is possible that Sens represses transcription of Egfr nuclear effectors via a similar mechanism. Future experiments are required to determine which of these or other mechanisms are important during R8 differentiation. However, it is likely that Sens does not act as a positive regulator of Edl/Mae, a proposed cell-autonomous repressor of Egfr signaling, because edl/maefunction is not required for normal R8 differentiation(Yamada et al., 2003). Finally, it is also unlikely that sens functions as an activator of yan, which encodes a nuclear repressor of the Egfr pathway, because yan loss-of-function mutations also do not impact R8 differentiation(Lai and Rubin, 1992).

The positioning of Sens repression downstream of ERK activation may help explain interactions observed between sens and Egfr pathway homologs in T-lymphocytes. In Jurkat T-cells, activation induced cell death (AICD), a process that is required to prevent non-specific activation of T-cells, is dependent, in part, on ERK1/2 activation(van den Brink et al., 1999). Intriguingly, high levels of Gfi-1 have been shown to inhibit AICD despite high levels of ERK1/2 activation (Karsunky et al., 2002). The antagonistic relationship between Sens and the Egfr pathway in R8, in conjunction with the observation that Gfi-1 can bind to the enhancer regions of Ets1 and Ets3, suggest that this inhibition of AICD may occur via Gfi-1-mediated repression of ERK1/2 targets(such as Ets/pnt) in T-cells (Duan and Horwitz, 2003). Thus, our results may establish R8 development as a powerful and novel system with which to study mechanisms of lymphomagenesis, apoptosis and cancer.

We thank Matthew Freeman, Kevin Moses, Alan Michelson, Kwang Choi and the Bloomington Stock Center for stocks, Hugo Bellen and the Developmental Studies Hybridoma Bank for antibodies, and Cornelius Boerkoel for use of equipment. Special thanks go to the Mardon lab for support and to Kartik Pappu for critical reading of the manuscript. This work was supported by a fellowship from the McNair Foundation to B.J.F. and R01 EY11232 from the National Eye Institute to G.M.