Sequential activation of ETS proteins provides a sustained transcriptional response to EGFR signaling

Development of all multicellular organisms is dictated by a handful of signaling pathways that transmit information from the extracellular milieu and neighboring cells to the nuclei of the receiving cells. The final step in each of these pathways is the activation of transcription factors and induction of a set of target genes, the identity of which depends on the cellular context of the receiving cell. In order to make the transcriptional output sharp and resilient to noise, different pathways have adopted a common strategy that relies on a dual input. When a signaling pathway is silent, target genes are repressed by the recruitment of transcriptional repressors. Following signaling, proteins that activate transcription replace the inhibitory proteins, generating a bi-stable switch. For example, the Wnt pathway target genes bind TCF associated with the repressor Groucho/TLE, and, upon activation, Armadillo/β-Catenin replaces Groucho/TLE on TCF to trigger gene expression (Cavallo et al., 1998). Similar scenarios occur in the Hh, Notch and BMP pathways (Aza-Blanc et al., 1997; Ohlmeyer and Kalderon, 1998; Lai, 2002; Müller et al., 2003).

Following the initial burst of target-gene expression, sustained transcription requires continuous input from the signaling pathway. The inherent nature of the ‘gas and brake’ mechanism described above implies that, upon termination of activation, repression will follow by default. This poses a regulatory problem, as the required duration of transcription induction within the receiving cell may differ from the dynamic extracellular activation profile that triggers the signaling pathway. In such cases, it may be necessary to provide compensatory mechanisms that will propagate the initial and transient transcription-activation switch. We have examined this conundrum in the compound eye of Drosophila, which is patterned by repeated cycles of EGFR activation.

Development of the compound eye is a highly orchestrated process that is carried out in a sequential manner in the eye imaginal disc, to form the repeated ∼750 ommatidial units, each comprised of 20 distinct cell types (Kumar, 2012). Following the posterior-to-anterior progression of a morphogenetic furrow, a single R8 photoreceptor cell is specified, independently of EGFR (Yang and Baker, 2001; Hsiung and Moses, 2002). The subsequent recruitment events include successive activation rounds of the EGFR pathway, and will give rise to all cell types of the eye (Freeman, 1996). Although the same pathway is repeatedly induced, the altered intracellular ‘context’ of the recruited cells will give rise to the induction of different sets of photoreceptor and accessory cell fates at each round (Banerjee and Zipursky, 1990). The limited pool of non-differentiated cells necessitates restricted activation of the EGFR pathway, which occurs in discrete bursts. Thus, several mechanisms that limit the strength of the EGFR signal have been described. These mechanisms operate at the level of active ligand production (Yogev et al., 2008), ligand diffusion (Freeman et al., 1992b; Klein et al., 2004) and, in the signal receiving cell, downstream of the receptor (Casci et al., 1999; Kramer et al., 1999; Reich et al., 1999).

The final step of the EGFR pathway, following MAP kinase (MAPK) activation, is the induction of target-gene expression, mediated by the ETS-family protein Pointed (Pnt) (Klämbt, 1993; Scholz et al., 1993; Brunner et al., 1994; Klaes et al., 1994; O’Neill et al., 1994; Gabay et al., 1996; Hsu and Schulz, 2000). Similarly, ETS proteins such as C. elegans LIN-1 and mammalian Elk1 are important targets of Ras/MAPK regulation (Hart et al., 2000; Yordy and Muise-Helmericks, 2000), underscoring the widespread functions of ETS proteins in Ras/MAPK signaling. pnt is expressed from two alternative promoters, leading to the generation of two protein isoforms, PntP1 and PntP2 (Fig. 1A) (Klämbt, 1993; Scholz et al., 1993). MAPK activity regulates activation of both Pnt isoforms, but via distinct mechanisms. The PntP2 protein is directly activated by MAPK phosphorylation, such that only the phosphorylated form can induce transcription (Brunner et al., 1994). By contrast, PntP1 functions as a constitutively active transcription factor. However, expression of pntP1 is restricted, and is induced by unknown transcription factors, the activity of which depends on MAPK phosphorylation (O’Neill et al., 1994; Gabay et al., 1996). In addition, the ETS repressor Yan binds to target gene enhancers, and is removed by MAPK phosphorylation (Lai and Rubin, 1992; Rebay and Rubin, 1995).

The dynamic activation profile of EGFR in the eye, and the restricted amount of ligand that is released by the photoreceptor cells, pose a challenge. Is EGFR activation sufficiently sustained and/or prominent enough to reliably induce transcriptional activation that will define the required cell fates? The current work addresses the mechanism by which consistent induction of EGFR target genes takes place, despite these limitations. We have undertaken a detailed dissection of the transcriptional responses to EGFR signaling during Drosophila eye imaginal disc development. Full disruption of pnt function revealed a complete EGFR loss-of-function photoreceptor recruitment phenotype, in which only the R8 photoreceptor cell type is specified. Ommatidia in which the function of only one of the two Pnt isoforms was eliminated showed the same R8-cell-only phenotype, thus demonstrating that each isoform is essential for photoreceptor recruitment. Further analysis showed that the two Pnt forms are activated in a sequential manner: MAPK phosphorylates and activates PntP2, which is capable only of inducing transcription of pntP1 and not of other pathway target genes. Once expressed, PntP1 is sufficient to induce the target genes essential for photoreceptor development. The induction of EGFR-target genes may thus be sustained by PntP1 protein stability, beyond the time window where EGFR triggers local MAPK activity.

The following lines were used: GMR-Gal4, ey3.5-Gal4 and tub-Gal80^ts (obtained from the Bloomington Stock Center); pnt-HS20 [obtained from H. Scholz (University of Cologne, Germany) and C. Klaembt (University of Münster, Germany)]; pnt-1277 (obtained from the Bloomington Stock Center); rho-X81 (Freeman et al., 1992a); UAS-spi RNAi (VDRC 103817); UAS-pntP1 (obtained from C. Klaembt); UAS-pntP2 GFP; UAS-yan^act (obtained from I. Rebay, University of Chicago, IL, USA); FRT82B arm-lacZ (obtained from the Bloomington stock center); ey-flp; FRT82B Ubi-GFP, FRT82B pnt^Δ33, FRT82B pnt^Δ78, FRT82B pnt^Δ88 (obtained from H. McNeill, Samuel Lunenfeld Research Institute, Ontario, Canada); and ru^PLLb (obtained from M. Gallio, Northwestern University, Evanston, IL, USA). The information on mutations, genome organization and strains was obtained from Flybase (Tweedie et al., 2009).

pntP2 GFP was generated as follows: a pntP2 cDNA was amplified by PCR and cloned into the Drosophila Gateway vector pTGW (T. Murphy, Carnegie Institution of Washington), using the Gateway cloning system (Invitrogen). The resulting construct was injected into flies harboring the AttP40 site, in order to insert the transgene into a defined genomic site.

After dissection in PBS, fixation in 4% PFA was performed. Washes were performed using 0.1% Triton X-100 (or 0.3% Triton X-100 when staining involved anti-PntP1). BSA (0.1%) was used for 15-minute blocking (or 5% NGS when staining involved anti-PntP1). An additional 30-minute blocking step with Image-it FX signal enhancer (Invitrogen Molecular Probes) was used prior to BSA blocking, when staining involved anti-β-gal or anti-PntP1. Primary antibodies were added for overnight incubation at 4°C and secondary antibodies for 2 hours at room temperature. Embryos were stained with antibodies or X-Gal according to standard procedures.

Primary antibodies used were anti-GFP (chick 1:2000; Abcam), anti-β-gal (rabbit 1:5000 after pre-absorption; Cappel), anti-Senseless (guinea pig 1:2000, obtained from H. Bellen, Baylor College of Medicine, Houston, TX, USA), anti-Eve (rabbit, 1:1000, obtained from M. Frasch, University of Erlangen-Nuremberg, Erlangen, Germany) and anti-PntP1 (rabbit 1:500, obtained from J. Skeath, Washington University, St Louis, MO, USA). Anti-Elav (rat 1:2000) and anti-Prospero (mouse 1:100) were obtained from the Developmental Studies Hybridoma Bank, University of Iowa.

Secondary antibodies used were anti-chick DyLight 488 (1:800), anti-rat Cy3 (1:400) (obtained from Jackson ImmunoResearch), anti-rabbit Alexa405 and Alexa488, anti-mouse Alexa405 and Alexa488, and anti-guinea pig Alexa647 (1:800, obtained from Molecular Probes).

For each of the pnt forms, a digoxigenin (DIG)-DNA-labeled probe was prepared using a PCR-DIG probe synthesis kit (Roche). The following primers were used for amplifying probes from pnt cDNA.

For pntP1: 5′AATTCGCGCGGTGTGTG, 5′TGGACAGCGATAAGGAGCAG; 3′GGGCGCCATATCATTGAAGT5′.

For pntP2: 5′CGCGCGCAATATACAGCAAA; 3′AGGCGCACAATCTTGATGGA5′.

Fixation, hybridization and detection were carried out according to Kosman et al. (Kosman et al., 2004).

The simulation is presented graphically in Fig. 6A.

The following set of ordinary differential equations were used to generate the figure.

Boundary conditions are given by the following: [Spi]_t=0 = 0; [P2]_t=0 = P2_tot; [pP2]_t=0; [P1]_t=0 = 0.

[Spi] is the concentration of the EGFR ligand Spitz (Spi). Spi is transiently produced at a rate α_spi for a time 0 <t <t₀. Spi is then being degraded at some rate: β_Spi.

[P2] is the concentration of PntP2 protein. PntP2 is phosphorylated at a rate k, which depends linearly on the levels of Spi, with a time lag τ_p. This lag accounts for the time it takes Spi signal to propagate until phosphorylation produces phosphorylated PntP2 [pP2]. [pP2] is dephosphorylated at a constant rate: k_phos. Initially, all PntP2 is dephosphorylated and its levels are P2_tot. We assume total PntP2 levels are constant, such that [P2]+[pP2]=P2_tot at all times.

[P1] is the concentration of PntP1. It is produced at a rate α, which is linearly dependent on the levels of phosphorylated PntP2, τ_t seconds earlier. τ_t represents the time it takes to produce PntP1 protein following activation of the pntP1 gene by [pP2]. PntP1 is turned over at a rate β_P1. Parameter values are provided in supplementary material Table S1.

To begin our analysis of the manner by which ETS-domain transcription factors mediate EGFR signaling during Drosophila eye development, we sought to verify the functional requirements for Pnt in this setting. The pnt genomic locus gives rise to two partially overlapping transcripts, pntP1 and pntP2, which are generated by distinct transcriptional promoters, positioned over 50 kb apart. The two encoded protein isoforms, PntP1 and PntP2, share a common C-terminal region, that includes their ETS DNA-binding domain. (Klämbt, 1993; Scholz et al., 1993) (Fig. 1A). We used FLP/FRT recombination to generate eye-disc clones homozygous for mutant pnt alleles, which affect either or both Pnt forms (Fig. 1B-D′). The alleles used represent deletions that are predicted to be molecular nulls. Full disruption of Pnt function was achieved using the pnt^Δ88 allele, an intragenic deletion removing the 3′ end of the pnt gene (Brunner et al., 1994) (Fig. 1A). As was previously shown, only R8 photoreceptors, which express both the R8-specific factor Senseless and the pan-neuronal marker Elav are specified in the complete absence of Pnt activity (Rogers et al., 2005), implying an essential requirement in recruitment of additional cells to ommatidial units (Fig. 1B,B′).

We next examined the consequences of specifically removing each of the pnt forms alone. To achieve this, we used the pnt^Δ33 and pnt^Δ78 alleles, small intragenic deletions that specifically disrupt the pntP1 or pntP2 transcripts, respectively (O’Neill et al., 1994; Morimoto et al., 1996) (Fig. 1A; supplementary material Fig. S1C-K). The form-specific clones displayed an identical phenotype to the complete removal of pnt, i.e. absence of photoreceptor neurons, except for R8 (Fig. 1C-D′). This result, which confirms previous observations (Yang and Baker, 2003), indicates that there is no redundancy between the two forms of Pnt, such that both are required for the establishment of EGFR-induced photoreceptor cell fates.

Having verified an essential requirement for both forms of Pnt in mediating EGFR-based recruitment of photoreceptors, we examined the manner by which EGFR signaling influences pnt expression. The expression pattern of the two pnt forms during eye development was analyzed by following two enhancer trap lines, HS20 and 1277, integrated in the vicinity of the P1 or P2 promoters, respectively (Scholz et al., 1993) (Fig. 1A). Expression of the β-galactosidase (β-gal) reporter from these elements during embryogenesis indeed recapitulates the known distribution of the specific transcripts (Klämbt, 1993; Scholz et al., 1993) (supplementary material Fig. S1A,B).

In the eye disc, the two reporters show distinct patterns of expression. pntP1 is expressed in a restricted manner, as revealed by HS20 (Fig. 2A,B). Expression is observed only within the differentiated region of the disc, including and posterior to the morphogenetic furrow. HS20 colocalizes with Elav in all mature ommatidia, indicating that pntP1 expression is further confined to photoreceptor neurons. However, HS20 is also detected in cells immediately posterior to the morphogenetic furrow, only some of which express Elav. These observations imply a temporal expression pattern, consistent with the possibility that PntP1 induction leads to photoreceptor differentiation and the eventual expression of Elav (Fig. 2B). pntP2 expression, monitored by the 1277 reporter, is expressed in a broader pattern that is not restricted to the Elav-expressing cells (Fig. 2C,D), the only exception being R8 cells, which show lower expression (Fig. 2D).

To explore the basis for the different expression patterns of the two pnt forms, we examined their dependence on EGFR signaling. An RNAi construct targeting the EGFR ligand Spitz (Spi) was induced, using the ey3.5-Gal4 driver that is expressed in all eye disc cells, on both sides of the morphogenetic furrow (Fig. 2E-F′). This led to an effective block of EGFR signaling, as evidenced by the absence of Elav-positive photoreceptor cells, except for R8 cells, the specification of which does not depend on EGFR signaling (Yang and Baker, 2001; Hsiung and Moses, 2002). This phenotype is identical to the one observed in clones deficient for spi (Tio and Moses, 1997).

As was observed for Elav, pntP1 expression was retained only in R8 cells following knockdown of spi (Fig. 2E,E′). This indicates that, like Elav, expression of pntP1 depends on EGFR signaling. pntP2 expression, on the other hand, was retained in cells surrounding the R8 cells (Fig. 2F,F′). Expression was somewhat reduced in the most posterior cell rows, which we ascribe to secondary consequences of cell mis-specification and death. Therefore, pntP2 transcription appears to be EGFR independent, consistent with its broad distribution beyond the cells where EGFR signaling normally takes place.

Taken together, these observations of the differential expression patterns of the two Pnt forms are suggestive of their modes of regulation and activity. pntP2 is broadly expressed, independently of EGFR signaling, providing all cells with a competence to respond to MAPK activation in the restricted domains where triggering will take place. Expression of pntP1, however, is limited and overlapping with the domains of EGFR activation, consistent with a role for EGFR-induced MAPK activity in controlling its expression.

Having shown that both Pnt isoforms are required for photoreceptor specification, we next asked whether they are sufficient to induce EGFR target-gene expression on their own. To assess sufficiency of PntP1, we first used GMR-Gal4, which is broadly active posterior to the morphogenetic furrow, to drive expression of pntP1 in eyes bearing clones of the null pnt^Δ88 allele (Fig. 3A-A′). Such ectopic expression of PntP1 rescued the mutant phenotype of these clones, leading to the recruitment of multiple Elav-expressing cells around the R8 photoreceptors. The typical arrangement of photoreceptor cells in each ommatidia was lost, however, as the expression of PntP1 did not follow the orderly activation of EGFR (Fig. 3A-A′). This result demonstrates that once expressed, PntP1 can elicit photoreceptor cell fates in the absence of PntP2.

The sufficiency of PntP1 could be further demonstrated following overexpression in wild-type eye discs. As normal expression of PntP1 in the eye disc is restricted to the cells eventually expressing Elav, we used the ey3.5-Gal4 driver to express PntP1 throughout the eye disc epithelium. Under these circumstances, the Elav expression pattern lost its highly ordered pattern, and became broad and disorganized (Fig. 3B-C′), suggesting the recruitment of additional cells. Furthermore, a global alteration in Elav expression was observed, such that Elav-positive cells are detected closer to the morphogenetic furrow, indicating premature induction (Fig. 3B′,C′).

To monitor the excessive recruitment of cells more precisely, we used markers that are specific for distinct photoreceptor types. Rhomboid 1 (Rho1) is expressed in three photoreceptors, R8, R2 and R5 (Freeman et al., 1992a), whereas Prospero (Pros) is normally expressed only in the R7 cell (Kauffmann et al., 1996). Both of these fate markers normally display a highly ordered and stereotypic array (Fig. 3E,H). Following ectopic PntP1 expression, the regular arrangement broke down, and extra cells expressing Rho1 or Pros were detected in many ommatidia (Fig. 3F,I). On average, the number of Rho1- and Pros-expressing cells rose to 3.6 cells (n=460 ommatidia) and 1.2 cells (n=317), respectively.

Similar overexpression of PntP2-GFP did not result in detectable abnormalities in photoreceptor differentiation, despite the high levels of ectopic expression (Fig. 3D,G,J; supplementary material Fig. S2). We argue that as the activity of PntP2 depends on MAPK phosphorylation, ectopic expression in cells that are not immediately adjacent to the normal source of Spi, will have no consequences. Taken together, these experiments support a scenario of sequential activation of Pnt forms, as part of the EGFR signaling pathway. In this scheme, the broadly expressed PntP2 is first locally activated by MAPK, and subsequently triggers (directly or indirectly) the expression of PntP1, which then acts constitutively to induce pathway targets in the signal-receiving cells.

We next examined the role of PntP2 in the induction of pntP1 expression. The HS20 reporter for pntP1 expression is inserted in the locus, and thus could not be used in pntP2 mutant clones. We therefore followed expression of PntP1 protein, using an antibody that is specific to this form (Alvarez et al., 2003). PntP1 protein is most prominent in the proneural clusters ahead of the morphogenetic furrow, and shows a weaker expression posterior to the furrow. This pattern is similar to the distribution of activated MAPK, as monitored by dpERK antibodies (Gabay et al., 1997), and is therefore consistent with the induction of pntP1 expression by EGFR/MAPK signaling. As expected, PntP1 protein is not detected by the form-specific antibody in pntP1 mutant clones (Fig. 4A-A′). We then generated mutant clones for pntP2, and followed the expression of PntP1 protein in a similar manner. Both the prominent expression of PntP1 in the proneural clusters, as well as its expression posterior to the furrow were eliminated in these clones, demonstrating that PntP2 is required for induction of pntP1 expression (Fig. 4B-B′)

In view of the requirement for transcriptional activation of pntP1 by the PntP2 ETS protein, we asked whether PntP1 can also maintain its own expression. Ectopic expression of PntP1 leads to excessive photoreceptor differentiation (Fig. 4C,C′), but this is not accompanied by elevation in the level of the pntP1 transcriptional reporter HS20 (Fig. 4C′). Importantly, this result suggests that once PntP1 is expressed, the activity it provides is not self-sustaining. Rather, it depends on the stability of pntP1 RNA and protein that was produced initially, following the burst of activated MAPK and PntP2 activation. Similarly, ectopic expression of PntP1 did not elevate the expression levels of the pntP2 transcriptional-reporter 1277 (Fig. 4D-D′).

A sequential scenario, in which PntP1 expression is dependent on PntP2 activation, can readily explain the pnt-null phenotype observed when only the PntP2 form is inactivated. Less clear, however, is the insufficiency of PntP2, which appears to be incapable of inducing photoreceptor cell recruitment and differentiation on its own. This insufficiency may arise from distinct transcriptional activation capacities of the two Pnt forms, which may associate with different enhancer-binding proteins. An alternative possibility is that levels of MAPK-phosphorylated PntP2 within signal-receiving cells can support only pntP1 expression, but not other pathway targets.

Expression of PntP1 or PntP2 alone is commonly sufficient to trigger expression of EGFR and FGFR target genes (Gabay et al., 1996; Samakovlis et al., 1996; Alvarez et al., 2003; Zhu et al., 2011). The unusual reliance on sequential activation of both Pnt forms during eye disc development may reflect the low and transient levels of EGFR activation that are induced at each round of photoreceptor recruitment. If this were the case, endogenous PntP2 may suffice for target-gene induction under circumstances of elevated levels of EGFR signaling in the eye disc.

To address this issue, we generated and examined mutant pntP1 clones in eye discs also mutant for rho-3. EGFR signaling is elevated in this background, owing to more effective production of the active form of the EGFR ligand Spi (Yogev et al., 2008). Indeed, in these double-mutant clones, recruitment of Elav-expressing cells was observed in some of the ommatidia (Fig. 5A-A′). This observation indicates that PntP2 is mechanistically capable of inducing photoreceptor cell fates on its own. In addition, it supports the notion that the requirement for sequential recruitment of Pnt proteins stems from the low and transient nature of EGFR activation in the eye disc.

If PntP2 can be mechanistically sufficient for induction of target genes, raising the levels of PntP2 may also allow it to induce target genes in the absence of PntP1. To examine this possibility, the GMR-Gal4 driver was used to overexpress PntP2-GFP. We first determined that this construct was functional, by demonstrating restoration of Elav expression in mutant clones for pntP2 (Fig. 5B-B′). The same construct was then expressed in the background of pnt^Δ33 mutant clones, in which pntP1 is specifically impaired (Fig. 5C-C′).

Partial rescue was observed in the pntP1-mutant background, in that numerous ommatidia displayed recruitment of one or two cells expressing Elav, in addition to R8 (Fig. 5C′). High levels of PntP2 can therefore elicit photoreceptor recruitment and differentiation on their own, in cells where MAPK activation takes place. It is important to note that this experiment is conceptually distinct from ectopic PntP2 expression in a wild-type background, where no ectopic photoreceptor induction was observed (Fig. 3D,G,J). Here, we demonstrate that higher levels of PntP2 in cells that are normally exposed to the endogenous MAPK activation, can substitute for the absence of PntP1 and give rise to Elav expression.

Sequential activation of PntP2 and PntP1 implies that the duration of transcriptional activation shifts from a regimen that relies on PntP2 protein phosphorylation and dephosphorylation, to one that is dependent only on PntP1 RNA and protein stability. To examine the effect of such a shift on the time frame of transcriptional induction of target genes, the sequential induction of Pnt proteins was described by differential equations (Materials and methods), and parameters were chosen based on an estimate of biologically relevant values (supplementary material Table S1).

Signal transduction and phosphorylation are more rapid events than protein production and turnover rates. The simulation shows that kinetics of EGFR activation by Spi, and persistence of the phosphorylated active form of PntP2, are transient (less than 20 minutes). However, the induction of PntP1 expression sustains the biological response for 2-3 hours, as long as PntP1 RNA and protein are stable (Fig. 6A and supplementary material Table S1).

We sought to measure the stability of PntP1 protein in the context of the developing eye disc, as a means of demonstrating that the sequential signaling scenario is feasible. To achieve this, we used Gal80^ts, a temperature-sensitive variant of the Gal4 repressor Gal80 (McGuire et al., 2004). Expression of pntP1 was induced in the eye by GMR-Gal4 at 29°C, the restrictive temperature for Gal80^ts. Eye discs were fixed at different time intervals following a shift down to 18°C, and stained for PntP1 protein. The endogenous level of PntP1 protein posterior to the morphogenetic furrow is low, and represents only 8.3% of the total level that was observed after induction of pntP1. Even after a 10-hour chase, significant levels of ectopic PntP1 could be detected. The estimated half-life of PntP1 under these conditions is about 6 hours. Stability of PntP1 protein is therefore sufficient to promote the duration of response to EGFR signaling, irrespective of the transient and fluctuating level of the signal.

This study demonstrates that both forms of the Pnt ETS protein are required for induction of photoreceptor cell fates, and that the requirement is sequential. Our interpretation of the Pnt-based response in EGFR signal-receiving cells of the eye imaginal disc is as follows (see also scheme in Fig. 7A-C). PntP2 activity is dependent on EGFR/MAPK activation, and possesses a low capacity to trigger target genes, owing to insufficient duration of the activated phosphorylated form. This transient activation of PntP2 is sufficient, however, to trigger (directly or indirectly) expression of pntP1. Signaling within the cell is then propagated, irrespective of the activation of EGFR, as long as the PntP1 protein persists. When the expression of PntP1 was driven independently of EGFR signaling, it induced photoreceptor cell fates on its own, supporting the notion that PntP1 transcriptional activity constitutes a second tier of the cellular response. Sequential activation may also provide amplification of the signal, if each molecule of PntP2 is capable of triggering several rounds of pntP1 transcription.

Owing to the different modes of Pnt activation, the sequential use of the two forms impinges on the kinetics of signaling. Phosphorylation of PntP2 provides the first layer of response to phosphorylation by MAPK, but may be short-lived, owing to dephosphorylation and the transient nature of EGFR activation within the eye disc. It should also be noted that the activated form of MAPK may be transient, owing to the activity of MAPK phosphatases that were shown to attenuate EGFR signaling in the eye and in other tissues (Kim et al., 2004). By induction of pntP1 expression, the response is no longer dependent on the regimen of MAPK phosphorylation, and relies solely on the stability of PntP1 mRNA and protein, thus extending its duration. This scenario ensures that the required target genes will be induced appropriately. As PntP1 does not trigger its own expression, this response is eventually terminated.

The sequential Pnt activation mechanism may be especially crucial in the eye disc, where the levels of EGFR activation are low due to the attenuation of the amount of active ligand that is secreted. Thus, low levels of Spi maintain a short range of signaling. In parallel, negative-feedback loops, including Argos and Sprouty, are induced by EGFR signaling. The expression of PntP1 would make the cells that already responded to EGFR activation refractive to the negative signals they are now producing. Thus, PntP1 serves to propagate the EGFR transcriptional response, irrespective of the dynamic changes in EGFR/MAPK signaling. Although the cell is still dependent upon the perdurance of the PntP1 protein, photoreceptor cell fates will be appropriately induced, as PntP1 is sufficiently stable to trigger the crucial responses.

In various other developmental scenarios, each of the Pnt isoforms is expressed in a distinct set of tissues and functions on its own within that context. For example, PntP1 is expressed in the embryonic neuroectoderm, whereas PntP2 is expressed in the mesoderm (Klämbt, 1993; Gabay et al., 1996). It will be interesting to explore whether a sequential mechanism, similar to the one described in the eye, is operating in other contexts where the two Pnt forms are expressed within the same tissue, such as during EGFR-dependent recruitment of chordotonal cells in the leg disc (zur Lage and Jarman, 1999; zur Lage et al., 2004). As in the eye disc, EGFR activation in chordotonal precursors occurs in discrete bursts generated by a relay mechanism, in which a recruited cell becomes a source for new ligand. It will also be important to identify the transcription factor(s) that are phosphorylated by MAPK, and induce the expression of pntP1 in tissues where PntP2 is absent. It is interesting to note that sequential activation of ETS proteins was recently reported for the process of reprograming amniotic cells to endothelial cells (Ginsberg et al., 2012). However, it is not known whether the second cohort of ETS proteins that are recruited in this sequence are constitutively active in the same way as PntP1.

Mechanistically, PntP2 is capable of inducing the set of target genes that represent acquisition of photoreceptor cell fates. When boosting the signaling activity, either by elevating EGFR signaling in a rho-3 mutant background or upon overexpression of PntP2, this form alone is capable of partially inducing photoreceptor cell fates. We assume that, under normal conditions, activation of PntP2 is too transient to induce the necessary target genes. What makes the induction of pntP1 transcription more responsive to activated PntP2, and how it avoids perturbation by noise and fluctuations in PntP2 activation, are interesting questions for future study.

Although the two Pnt isoforms share the same DNA-binding domain, their distinct transcription-activation domains may impinge not only on their mode of activation, but also on different preferences for association with other transcription factors and thus on the choice of genes that are induced. We believe that the composition of other transcription factors, expressed by a given cell will determine whether PntP2 will activate the expression of pntP1. Indeed, there are tissues such as the embryonic mesoderm, where only pntP2 is expressed and does not induce expression of pntP1 (Klämbt, 1993).

The sequential activation of Pnt proteins presented here provides another solution to the perdurance of gene expression. The ability of a signal-induced transcription factor to maintain its own expression gives rise to prolonged expression. Alternatively, long-term inactivation of a transcriptional repressor can provide an effective mechanism. For example, transient MAPK activation was shown to induce long-term phosphorylation that inactivates the Groucho co-repressor (Helman et al., 2011).

Transcriptional responses by the canonical developmental signaling pathways rely on a ‘gas and brake’ model, where constitutive repression of transcription of the relevant target genes is overcome by triggering the activity of key transcription factors, to induce gene expression. This implies not only a direct link between the signaling cascade and the onset of transcription, but also a continuous obligatory coupling between the two. When the extracellular signaling regimen is transient, this coupling poses a problem for sustained and reliable transcription, and subsequent induction of cell fates.

In the case of ETS proteins, activation of transcription by RTKs requires simultaneous inactivation of an ETS repressor termed Yan (Lai and Rubin, 1992; Rebay and Rubin, 1995) and activation of ETS activators, which compete for the same binding sites on the DNA. The mechanism of sequential activation of Pnt protein forms presented in this work provides one solution for sustaining the transcriptional response, and is based on the distinctly different mode of activation of each Pnt form. Another advantage is that a constitutively active transcription factor represents a form that is refractive to negative-feedback responses to signaling that it may induce, assuring a sustained transcriptional response of the activated cells. All other Drosophila ETS proteins are similar to PntP2: they contain the Pnt domain that mediates MAPK association as well as putative phosphorylation sites, and are likely to be activated directly by MAPK phosphorylation (Tweedie et al., 2009). Acquisition of the PntP1-specific exons may have endowed the pnt locus with the unique capacity of sustained induction of target genes. With this paradigm in mind, it would be interesting to explore whether transcriptional responses of other signaling pathways may use such a two-tier mechanism to propagate signaling within the receiving cells under conditions where the extracellular signal is transient.

We thank H. Bellen, M. Frasch, M. Gallio, C. Klaembt, H. McNeil, I. Rebay, J. Skeath, the Development Studies Hybridoma Bank, and the Bloomington and VDRC Stock Centers for generously providing reagents and fly stocks. We are grateful to Danny Ben-Zvi for carrying out the numerical simulation and critical reading, and to all members of the Shilo lab for help and discussions.