Control of Drosophila wing growth by the vestigialquadrant enhancer

The Drosophila wing is a discrete organ of stereotyped pattern,size and shape specified by a selector gene, vestigial (vg)(Williams et al., 1991; Williams et al., 1993; Kim et al., 1996; Halder et al., 1998; Liu et al., 2000). Shortly after the wing primordium is first apparent as a cluster of ∼40 vg-expressing cells (Wu and Cohen, 2002), it is subdivided into dorsal (D) and ventral (V)compartments by the heritable activation of the selector gene apterous (ap) in the D compartment(Diaz-Benjumea and Cohen, 1993; Williams et al., 1993; Blair et al., 1994). Short-range Delta/Serrate/LAG-2 (DSL)-Notch signaling across the D-V boundary then initiates a dramatic 200-fold expansion of the wing primordium to a population of ∼8000 vg-expressing cells under the control of the long-range morphogens Wingless (Wg) and Decapentaplegic (Dpp)(Diaz-Benjumea and Cohen, 1995; Kim et al., 1995; Zecca et al., 1995; de Celis et al., 1996; Doherty et al., 1996; Lecuit et al., 1996; Nellen et al., 1996; Zecca et al., 1996; Neumann and Cohen, 1997). Here, we examine how morphogens and selector genes control organ growth, using the Wg-driven expansion of the population of vg-expressing cells -the wing primordium - as a paradigm.

In the accompanying study (Zecca and Struhl, 2007), we focused on how Wg signaling controls vgexpression and wing growth, taking advantage of ap mutant discs. Normally, short-range DSL-Notch signaling across the D-V boundary induces`border' cells flanking the boundary to express both wg and vg (Williams et al.,1994; Couso et al.,1995; Diaz-Benjumea and Cohen,1995; Kim et al.,1995; de Celis et al.,1996; Kim et al.,1996; Neumann and Cohen,1996; Rulifson et al.,1996). However, in ap mutant discs, border cells are not specified, the early expression of vg that normally precedes the D-V segregation dissipates, and the presumptive wing primordium fails to develop(Williams et al., 1993). By generating clones of cells that ectopically express Vg, Wg or both, we showed that cells within ap mutant discs could be recruited to express vg in response to Wg, but only if they were located near or next to cells that already express Vg. These results defined a previously unknown feed-forward mechanism of vg autoregulation, and led us to propose that D-V border cells normally control the expansion of the wing primordium by providing both a long-range morphogen, Wg, as well as the initial Vg-dependent feed-forward signal that entrains neighboring cells to express vg in response to Wg.

Here, we extend our results in ap mutant discs by testing whether this autoregulatory circuit is required for normal wing growth in wild-type discs. We first demonstrate that the previously identified quadrant enhancer(QE) of the vg gene mediates vg autoregulation in response to Wg, the feed-forward signal, and a newly defined third input:`priming' of the vg locus by pre-existing low levels of Vg. We then present evidence that QE-driven expression of vg is necessary and sufficient for the expansion of the wing primordium organized by D-V border cells. These findings support our hypothesis that wing growth normally depends on a non-autonomous autoregulatory circuit of vggene expression triggered by short-range DSL-Notch signaling and fueled by long-range Wg signaling.

Most mutant alleles, transgenes and protocols employed are described in the accompanying study (Zecca and Struhl,2007). Transgenes and protocols unique to this study are as follows.

5XQE>Tubα1-flu-GFP>vg and 5XQE>CD2,y2>vg transgenes were assembled using DNAs described in Zecca and Struhl (Zecca and Struhl,2007). 5XQE>vg derivatives of these transgenes generated by germ-line excision of the intervening Flp-out cassette resulted in dominant larval lethality. Similarly, induction of 5XQE>vgclones in first instar larvae carrying such transgenes resulted in late larval or pupal lethality. Hence, our analysis of 5XQE>vg transgene activity was restricted to the behavior of Flp-out clones in the wing disc.

The rp49>CD2,y2>vg transgene was assembled using the rp49 promoter, as well as the destabilizing Hsp70 3′UTR (Greenwood and Struhl,1997; Casali and Struhl,2004). rp49>vg derivatives (referred to subsequently as rp49-vg) of each of several rp49>CD2,y2>vgtransgene insertions were generated by germ-line excision of the >CD2,y2> cassette. All but one of these resulted in pupal lethality when present in one copy. However, the exceptional rp49-vgderivative was viable and fertile in one copy, albeit pupal lethal when homozygous, indicating that this transgene expresses a lower level of exogenous Vg than the others, and on this basis we selected it for use in all subsequent experiments.

Clones of vg or arrow (arr) mutant cells were generated by Flp-mediated mitotic recombination(Golic, 1991). The Minute technique (Morata and Ripoll, 1975) was used to generate vg83b27(vg^b) clones with a growth advantage(Fig. 2E). Clones expressing exogenous Vg were generated using the Flp-out technique(Struhl and Basler, 1993). In some cases, two types of clones were generated in the same disc to yield either (1) adjacent clones of different type (e.g. 5XQE>vg clones next to Tubα1>vg clones; e.g. Fig. 4); (2) coincident clones of different type [e.g. vg83b27R (vg⁰) 5XQE>vg clones; e.g. Fig. 3C]; or (3) `clones within clones' (e.g. arr<sup>0</sup>clones inside 5XQE>vg clones; Fig. 3D,E). Unless otherwise stated, clones were induced by heat shocking first instar larvae [24-48 hours after egg laying (AEL)] at 36°C for 30 minutes; for `clones within clones', larvae were heat shocked, as above, during the first instar, and then given a second heat shock of the same length and temperature at 60-84 hours AEL (late second to early third instar), or 48-72 hours AEL (second instar). In all of the experiments in this study, mature wing discs were dissected from late third instar larvae, and fixed and analyzed as previously described(Zecca and Struhl, 2002).

Larvae were heat shocked (35°C 10 minutes) at the times indicated in Fig. 2. Mutant clones were marked by loss of GFP expression, whereas their wild-type sibling (twin)clones were marked by strong GFP expression (owing to homozygosity of the Hsp70-flu-GFP transgene). All, and only those, wild-type clones that contributed to the presumptive wing pouch area (marked by 1XQE-lacZexpression) were scored for the presence and contribution of their associated mutant twins. For further details, see Fig. 2.

Genotypes are listed by figure panel; except where stated otherwise, the X chromosome was y w Hsp70-flp.

1B: 1XQE-lacZ vg^83b27/vg^83b27.

1C: 1XQE-lacZ vg^83b27R/vg^83b27R.

1D: 1XQE-lacZ vg^83b27/vg^83b27;rp49-vg/rp49-vg.

1E: 1XQE-lacZ vg^83b27R/vg^83b27R;rp49-vg/rp49-vg.

1F: 1XQE-lacZ vg^83b27/vg^83b27R;Tubα1>flu-GFP, y⁺>vg/+.

1G: 1XQE-lacZ vg^83b27R/vg^83b27R;Tubα1>flu-GFP, y⁺>vg/+.

2A,B,F: FRT42D vg^83b27/1XQE-lacZ FRT42D Hsp70-flu-GFP.

2C: y w 5XQE-DsRed/y w Hsp70-flp; FRT42D vg^83b27R/1XQE-lacZ FRT42D Hsp70-flu-GFP; rn-lacZ/+.

2D,F: FRT42D vg^83b27/1XQE-lacZ FRT42D Hsp70-flu-GFP;rp49-vg/rp49-vg.

2E: FRT42D Minute(2)IK Hsp70-flu-GFP/FRT42D vg^83b27;1XQE-lacZ/+.

2F: 1XQE-lacZ FRT42D Hsp70-flu-GFP/FRT42D Tubα1-DsRed (for vg⁺ clones).

3B,C: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; FRT42D vg^83b27R/1XQE-lacZ FRT42D Hsp70-flu-GFP.

3D: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; FRT42D arr²/FRT42D Tubα1-DsRed; 1XQE-lacZ/+.

3E: y w 5XQE-DsRed/y w Hsp70-flp; FRT42D arr²/FRT42D Hsp70-CD2; Tubα1>flu-GFP, y⁺>vg/+.

4A: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; FRT42D vg^83b27R/1XQE-lacZ vg^83b27R;Tubα1>flu-GFP, y⁺>vg/+.

4B,C: 1XQE-lacZ ap^56f vg^83b27R/1XQE-lacZ vg^83b27R; Tubα1>DsRed,y2>vg/5XQE>Tubα1-flu-GFP>vg.

5A: FRT42D vg^83b27R/1XQE-lacZ vg^83b27R;BE-vg^GFP/+.

5B,C: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; FRT42D vg^83b27R/1XQE-lacZ ap^56f vg^83b27R;BE-vg^GFP/+.

5D: FRT42D vg^83b27R/1XQE-lacZ vg^83b27R;BE-vg^GFP/5XQE>Tubα1-flu-GFP>vg.

5E: FRT42D vg^83b27R/1XQE-lacZ vg^83b27R;BE-vg^GFP rp49-vg/+.

5F,G: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; FRT42D vg^83b27R/1XQE-lacZ vg^83b27R; BE-vg^GFPrp49-vg/+.

5H: FRT42D vg^83b27R/1XQE-lacZ vg^83b27R;BE-vg^GFP rp49>vg/5XQE>Tubα1-flu-GFP>vg.

6A,C: y w 5XQE>CD2,y2>vg/y w Hsp70-flp.

6B: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; 1XQE-lacZ FRT42D Hsp70-flu-GFP/1XQE-lacZ ap^56f vg^83b27R.

6D: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; Dll-lacZ/+;C765-Gal4/+.

6E: y w 5XQE>CD2,y2>vg/y w Hsp70-flp; Dll-lacZ/5XQE-DsRed vg^83b27R; UAS-wg rp49-vg/C765-Gal4.

Following the segregation of the wing imaginal disc into D and V compartments, vg expression is induced in D-V border cells and then extends into the rapidly expanding `pouch' of the disc, defining the growing wing primordium (Williams et al.,1993; Williams et al.,1994; Kim et al.,1995; Kim et al.,1996). Border cell and pouch expression are associated,respectively, with the activity of distinct boundary enhancer (BE)and quadrant enhancer (QE) elements(Williams et al., 1994; Kim et al., 1996). Here, we have sought to determine whether QE elements are responsible for mediating the feed-forward propagation of vg in response to Wg, and if so, whether operation of this autoregulatory circuit is necessary and sufficient for the normal expansion of the wing primordium that occurs following the D-V segregation.

Our main approach has been to generate a vg transgene that is expressed under QE control, validate that it is activated in response to the combined inputs of Wg and the vg-dependent feed-forward signal, and then test whether it is necessary and sufficient to mediate vg expression and wing growth away from the D-V compartment boundary. Crucial to the success of this approach has been our discovery of a third input necessary for the vg autoregulatory response: priming of the vg gene by pre-existing low levels of Vg. We begin by describing our evidence for priming, which arose unexpectedly from experiments designed to test the capacity of a BE-deficient allele of vg to mediate feed-forward autoregulation.

vg^83b27 (henceforth vg^b) is an internal deletion of the intron 2 segment of vg that removes the previously defined BE as well as adjoining sequences, but leaves intact the rest of the gene, including the QE(Williams et al., 1993; Kim et al., 1996). Mature vg^b mutant discs, like vg-null (henceforth vg⁰) discs, lack the wing primordium(Fig. 1A-C)(Williams et al., 1993), as expected if D-V border cells require BE activity to express vg and to initiate propagation of vg expression into neighboring tissue. However, clones of vg^b cells should retain the capacity to propagate vg expression in response to wild-type border cells and, hence, to contribute normally to the developing wing. In testing this prediction, we obtained evidence that the vg^b mutation deletes a previously unknown `priming'enhancer (PE), in addition to the BE, and that cryptic, low levels of Vg, expressed under the control of this enhancer, are a prerequisite for feed-forward autoregulation.

Initially, we generated vg^b clones in otherwise wild-type discs carrying the standard 1XQE-lacZ reporter. As expected, vg^b clones induced after D-V segregation (mid-to late second instar) were able to contribute normally to the wing pouch and express 1XQE-lacZ (Fig. 2B,F; data not shown). However, vg^b clones induced at earlier times showed a progressive decrease in their ability to do so (Fig. 2A,F). Indeed, vg^b clones induced during the first instar and given a competitive growth advantage using the Minute technique(Morata and Ripoll, 1975)failed to express 1XQE-lacZ and appeared to be excluded from the wing pouch, forming non-wing tissue at its expense(Fig. 2E). Thus, early-induced vg^b clones appear compromised for the ability to activate the QE and develop as wing tissue, even when they abut wild-type Vg-expressing cells and are in a position to receive Wg as well as Dpp.

One explanation for this unexpected result is that vg^bcells lack an additional component of the proposed vg autoregulatory circuit. The QE contains binding sites for Scalloped (Sd), the DNA-binding protein that combines with Vg to form a composite transcriptional activator (Halder et al.,1998; Paumard-Rigal et al.,1998; Simmonds et al.,1998; Guss et al.,2001; Halder and Carroll,2001). Moreover, Sd and the presence of its binding sites are necessary for QE activity (Halder et al., 1998; Guss et al.,2001). Hence, the QE might need to be `primed' by cryptic, low-level Vg to mediate feed-forward autoregulation, and the presence of such pre-existing Vg might depend on a distinct `priming' enhancer(PE) deleted in the vg^b allele. According to this hypothesis, sufficient Vg would perdure in vg^b cells induced after the D-V segregation to supply the requisite priming function,but not in the descendents of vg^b cells induced before D-V segregation.

To test this, we generated an rp49-vg transgene in which the vg coding sequence is expressed at exceptionally low level, under the control of the uniformly active, but weak, ribosomal protein 49(rp49; also known as RpL32 - Flybase) promoter(Greenwood and Struhl, 1997; Casali and Struhl, 2004), and asked whether such low-level expression is sufficient to rescue normal wing development and endogenous vg expression in early-induced vg^b clones.

Wing discs homozygous for the rp49-vg transgene express so little Vg protein that we were unable to detect it by antibody staining in wild-type or vg⁰ discs. In addition, homozygosity for the transgene failed to rescue wing development in vg⁰ discs. Instead, vg⁰; rp49-vg discs formed abnormally small wing pouches composed of cells that appeared to correspond to the periphery of the normal pouch, where neither Vg nor 1XQE-lacZ expression were readily detectable (Fig. 1E). Nevertheless, homozygosity for the rp49-vg transgene almost completely rescued the capacity of early-induced vg^bclones to express endogenous Vg, as well as the 1XQE-lacZ reporter,and to contribute to the wing pouch (Fig. 2D,F). Indeed, it restored normal vg expression and wing development in the wing pouch of entirely mutant vg^bdiscs, including in D-V border cells, despite the absence of the well-defined and evolutionarily conserved BE(Fig. 1D).

Thus, the capacity of vg^b cells to express vgand to develop as normal wing tissue appears to depend on cryptic, low-level Vg activity, defining a third input, `priming', that is required together with Wg and the feed-forward signal, for upregulation of vg away from the D-V boundary. These findings also indicate that the vg<sup>b</sup>allele retains at least one additional BE, and that activity of this BE depends on priming. Hence, the primary cause of the vg<sup>b</sup> `no wing' phenotype appears to be the deletion of the PE, not the BE, in intron 2. In subsequent experiments, we used the rp49-vg transgene to satisfy the requirement for priming in the absence of the endogenous vg gene.

The QE is active in all cells in which we posit that feed-forward autoregulation occurs, consistent with the QE being responsible for mediating this autoregulatory circuit as well as normal wing growth. To test this, we sought to assay the behavior of cells in which the endogenous vg gene is replaced with a transgene that expresses Vg under the control of QE sequences.

To express Vg under the control of QE sequences, we generated transformants of a 5XQE>CD2>vg Flp-out transgene in which five copies of the QE drive the expression of either rat CD2 or Vg. In the absence of Flp recombinase, the 5XQE>CD2>vg transgene insertions behaved like the previously described 1XQE-lacZ and 5XQE-DsRed reporter transgenes, showing expression of CD2 that was tightly restricted to the wing pouch but excluded from the D-V border cells within the pouch (Fig. 3A)(Kim et al., 1996; Zecca and Struhl, 2007). Upon heat shock-induced expression of Flp, the >CD2> cassette is excised in single cells, generating clones of 5XQE>vg cells marked in the prospective wing pouch by the absence of CD2 and the expression of exogenous Vg. Different transgene insertions showed some variation in the level and extent of Vg expression following excision of the >CD2> cassette. For the most strongly active insertions, clones of 5XQE>vg cells that were also vg⁰ developed as wing tissue and expressed levels of exogenous Vg protein similar to those of endogenous Vg expressed in surrounding wild-type cells(Fig. 3B,C). By contrast,clones of 5XQE>vg vg⁰ cells generated using the less active insertions showed weaker, patchy expression of Vg and rescued wing development less well (data not shown). We therefore focused our analysis on one such strongly active 5XQE>CD2>vg insertion, and performed the experiments described below to validate that its activity depends on Wg and the Vg-dependent feed-forward signal.

To test whether expression of the chosen 5XQE>CD2>vgtransgene requires Wg input, we heat shocked first instar larvae to generate large clones of 5XQE>vg cells in otherwise 5XQE>CD2>vg wing discs, and then heat shocked them again during the late second to early third larval instar to generate smaller clones of cells mutant for arr (arr⁰), which encodes a co-receptor essential for transducing Wg(Wehrli et al., 2000). We observed that surviving arr⁰ clones showed greatly reduced or no expression of both CD2 and Vg, irrespective of whether the clones were located within the 5XQE>vg territories(Fig. 3D) or in surrounding 5XQE>CD2>vg tissue (data not shown). These results indicate that expression of both the excised and intact forms of the transgene require Wg input.

Because arr⁰ clones are associated with the loss of endogenous vg expression (data not shown), it is possible that activity of the 5XQE element depends only on the presence of Vg protein, and hence might `report' Wg input indirectly, via activation of other, as yet unidentified, Wg-responsive enhancers in the endogenous vg gene. To assess this, we replaced the 5XQE>CD2>vgtransgene with a Tubα1>GFP>vg transgene(Fig. 1G)(Tubα1 is also known as αTub84B -Flybase) (Zecca and Struhl,2007) to create clones of Tubα1>vgcells that continuously express moderate levels of exogenous Vg, irrespective of Wg input. We then generated clones of arr⁰ cells within such Tubα1>vg clones and asked whether activity of the 5XQE element still requires Wg input, using expression of a 5XQE-DsRed reporter to monitor 5XQE activity. Tubα1>vg clones make sufficient Vg protein to rescue expression of both the 5XQE-DsRed and 5XQE>CD2>vg transgenes, as well as wing development, in vg⁰ discs (Fig. 1G; data not shown). Nevertheless, arr⁰ clones generated within Tubα1>vg clones ceased to express the 5XQE-DsRed transgene (Fig. 3E). We conclude that 5XQE transgene activity does not merely reflect the presence of Vg protein, but instead depends on Wg input even when cells are supplied continuously with exogenous Vg. Significantly,such arr⁰ clones are subsequently lost from the wing pouch, within ∼12 hours after they cease to express the 5XQE-DsRed reporter, despite being independently and continuously supplied with exogenous Vg (data not shown). Hence, cells within the wing primordium still require continuous Wg input to survive and grow, even when they are provided with Vg protein by other means (see Discussion).

To determine whether activity of the 5XQE>CD2>vg transgene requires feed-forward input, we asked whether expression of either the intact or excised form of the transgene depends on the presence of neighboring Vg-expressing cells. vg⁰ wing discs carrying 5XQE>vg clones in a background of 5XQE>CD2>vg cells do not express either Vg or CD2 (data not shown). To test whether they fail to do so because both forms of the transgene require feed-forward input, we generated Tubα1>vg clones concomitantly in these same discs (by excision of a Tubα1>GFP>vgtransgene) and asked if such exogenous Vg-expressing clones could act non-autonomously to induce either 5XQE>vg or 5XQE>CD2>vg expression.

Activity of the intact 5XQE>CD2>vg transgene in this experiment was monitored by CD2 expression and that of the excised 5XQE>vg transgene was monitored by either Vg (data not shown) or 1XQE-lacZ expression (which is robustly expressed in all cells that express the 5XQE>vg transgene, e.g. Fig. 3B, Fig. 4A, Fig. 6B). Accordingly, the presence of 5XQE>vg clones can only be visualized if the experiment gives a positive result: namely, that Tubα1>vg clones (marked by the absence of GFP) can indeed induce 5XQE>vg expression in neighboring clones of 5XQE>vg cells (as monitored by Vg or 1XQE-lacZexpression). Nevertheless, we identified many such positively responding clones (lavender-colored clone in diagram in Fig. 4A; data not shown). Importantly, in all cases, these clones were adjacent to Tubα1>vg clones (green clone in diagram, Fig. 4A). Thus, it appears that clones of Tubα1>vg cells can induce 5XQE>vg expression in neighboring 5XQE>vg clones. Moreover, induction appears to depend on contact between the two clones.

Significantly, 5XQE>vg expression was not restricted to those 5XQE>vg cells that abut the neighboring Tubα1>vg clone. Instead, 5XQE>vgexpression appeared to spread many cell diameters into the 5XQE>vgclone, away from the abutting Tubα1>vg clone, and was associated with an expansion of the rescued wing primordium. In addition,such 5XQE>vg-expressing cells were also able to induce neighboring 5XQE>CD2>vg cells far from the abutting Tubα1>vg clone to express CD2 (yellow cells in Fig. 4A). Thus, the 5XQE>vg transgene appears to have the capacity not only to respond to the feed-forward signal, but also to propagate feed-forward signaling from one cell to the next. As evident in Fig. 4A, the range over which 5XQE>vg cells can induce 5XQE>CD2>vg expression across the clone border is tightly restricted to only a few cell diameters, consistent with the feed-forward signal being dependent on cell contact.

In principle, activation of the 5XQE>vg transgene by feed-forward signaling should require priming by low levels of Vg protein in the responding cells, and hence is unexpected in vg⁰discs. However, the 5XQE>vg cells in this experiment carry the 5XQE>vg as well as the Tubα1>GFP>vgtransgene, either one of which could provide cryptic Vg expression and hence the requisite priming activity. The same explanation also applies to activation of the 5XQE>CD2>vg transgene by adjacent 5XQE>vg-expressing cells, as even the intact 5XQE>CD2>vg transgene was able to provide a cryptic, priming activity in other experiments (Fig. 5A,B,E).

We note that the response of 5XQE>vg and 5XQE>CD2>vg cells appeared to depend on their capacity to express Vg, as expression of the 5XQE>vg transgene was strongly induced by adjacent Tubα1>vg cells(Fig. 4), whereas that of the 5XQE>CD2>vg transgene was not (data not shown). We infer that both the 5XQE>vg and 5XQE>CD2>vg transgenes initially respond only weakly to feed-forward input from abutting Tubα1>vg cells, but that the initial weak response of the 5XQE>vg transgene raises the level of Vg protein in these cells (and hence the strength of the priming input), thereby initiating an autoregulatory amplification of 5XQE>vg expression induced by the feed-forward signal. By contrast, 5XQE>CD2>vg cells would lack the capacity to autoregulate in this way, preventing them from mounting a robust response to Tubα1>vg cells. It is also notable that the response of 5XQE>CD2>vg cells depended on the level of Vg expressed in the inducing cells. Tubα1>vg cells express only moderate levels of Vg,well below peak endogenous levels (Fig. 1F′,G′), and were ineffective. However, 5XQE>vg-expressing cells make much higher levels(Fig. 3C) and were able to strongly activate 5XQE>CD2>vg expression in abutting cells(yellow cells in Fig. 4A). Hence, we infer that 5XQE>vg-expressing cells are more potent inducers of 5XQE>CD2>vg expression because they provide a correspondingly stronger feed-forward signal.

Although the experimental design of using the 5XQE>CD2>vgtransgene to generate 5XQE>vg clones has the virtue that it allows the 5XQE>CD2 response to be assayed in cells outside of the clone,it suffers from the fact that cells within such clones can only be identified if they respond positively, by expressing the 5XQE>vg transgene. We therefore repeated the experiment using an equivalent transgene, 5XQE>Tubα1-GFP>vg, which allows all of the 5XQE>vg cells to be scored independently by the loss of a Tubα1-GFP transgene within the Flp-out cassette (see Materials and methods). We again found that the resulting 5XQE>vgtransgene was only activated in 5XQE>vg clones that abut Tubα1>vg clones (the latter being marked independently in this experiment by excision of a >DsRed>cassette; Fig. 4B,C),confirming the requirement for the Vg-dependent feed-forward signal.

The experiments described above establish that activation of the 5XQE>vg transgene requires both Wg and the Vg-dependent feed-forward signal. In the following experiments, we use this transgene to test our hypothesis that feed-forward autoregulation mediated by the QE is necessary and sufficient for the dramatic expansion of the wing primordium organized by D-V border cells. To do so, we asked whether the presence of the 5XQE>vg transgene can rescue wing growth in discs in which Vg expression is otherwise driven only by the BE.

To generate such `BE-vg-only' wing discs, we used a BE-vg^GFP transgene that expresses a functional Vg-GFP chimeric protein under the control of a minimal form of the intron 2 BE (Zecca and Struhl,2007). In otherwise wild-type discs, this transgene behaves like the standard BE-lacZ reporter gene(Williams et al., 1994; Kim et al., 1996), being expressed in a thin stripe of border cells flanking the D-V compartment boundary within the wing pouch, and in a broader stripe in the surrounding hinge and notum primordia (Fig. 3A). In vg⁰ discs, the BE-vg^GFP transgene was expressed only weakly and sporadically in D-V border cells within the pouch, affording detectable, but very limited, rescue of wing development(Fig. 5A). This minimal response appeared to reflect a requirement for priming for efficient activation of the BE-vg^GFP transgene, as adding the rp49-vg transgene significantly enhanced border cell expression of Vg^GFP, as well as local rescue of wing development along the D-V boundary (Fig. 5E; data not shown).

Despite the limited response of the BE-vg^GFP transgene in vg⁰ discs, these discs still provide a context in which Vg expression in the pouch depends primarily on the BE. Hence, we asked whether clones of 5XQE>vg cells could rescue wing growth away from the D-V boundary in this context, and found that this was indeed the case. Clones of 5XQE>vg cells induced in vg⁰BE-vg^GFP discs were associated with activation of both the 5XQE>vg and 1XQE-lacZ transgenes, as well as with the expansion of wing tissue away from BE-vg^GFP-expressing border cells (Fig. 5B,C). In addition, they induced adjacent 5XQE>CD2>vg cells to express CD2, indicating propagation of the feed-forward signal across the clone border.

These results indicate that the 5XQE>vg transgene is both necessary and sufficient to restore wing growth in discs in which vgexpression is otherwise dependent only on BE and cryptic priming activity. However, it is apparent that growth is not fully rescued, as the expansion of wing tissue associated with 5XQE>vg clones was significantly less than the expansion that normally occurs following D-V segregation in wild-type discs. One explanation is that both BE- and QE-driven expression of Vg are compromised by inadequate Vg priming that derives from the BE-vg^GFP and 5XQE>vgtransgenes (the only possible sources of pre-existing Vg activity in the 5XQE>vg clones). We therefore repeated the experiment in the presence of the rp49-vg transgene, to ensure adequate priming in all cells.

In the absence of 5XQE>vg clones, vg⁰BE-vg^GFP discs carrying the rp49-vg transgene, as well as the 5XQE>CD2>vg and 1XQE-lacZ transgenes,showed robust expression of the BE-vg^GFP transgene in a narrow stripe of D-V border cells, and this was accompanied by weak expression of both the 5XQE>CD2>vg and 1XQE-lacZ transgenes in flanking cells (Fig. 5F). We note that in these discs, as well as in otherwise wild-type discs, both the BE-vg^GFP and standard BE-lacZ transgenes were expressed at a low level in cells up to several cell diameters away from the D-V boundary (data not shown). Hence, the weak 5XQE>CD2>vg and 1XQE-lacZ expression detected in cells flanking the D-V boundary might reflect a response to this low-level Vg^GFP activity.

Strikingly, when clones of 5XQE>vg cells were generated in this background, near or next to the D-V compartment boundary, they were associated with an autonomous upregulation of the 1XQE-lacZ transgene and a dramatic expansion of prospective wing tissue(Fig. 5G). Furthermore, they appeared to induce an equally dramatic, albeit short-range, induction of CD2 expression in neighboring 5XQE>CD2>vg cells (arrows in Fig. 5G). Corresponding experiments using the 5XQE>Tubα1-GFP>vgtransgene instead of 5XQE>CD2>vg confirmed the rescue of wing growth and also showed that it is an autonomous property of the 5XQE>vg clones (marked independently by the absence of GFP; Fig. 5D,H).

Thus, in the added presence of exogenous Vg priming activity provided by the rp49-vg transgene, the 5XQE>vg transgene appears both necessary and sufficient to greatly expand the domain of prospective wing tissue, mimicking the normal growth of wing tissue organized by D-V border cells.

If wing growth is governed by the capacity of QE sequences to mediate vg feed-forward autoregulation, the size of the wing primordium should depend on the strength of the QE response. To test this, we assayed the effects of 5XQE>vg clones on wing growth in otherwise wild-type discs, where they appear to generate a more sensitive and potent upregulation of Vg expression driven by the combined QE-dependent activities of the transgene and endogenous vg.

Early-induced 5XQE>vg clones were associated with an abnormal,cell-autonomous expansion of prospective wing tissue, extending beyond the normal limit of detectable Vg expression into the surrounding `rotund(rn)-only' territory of the wing pouch delimited by the inner ring of Wg expression (Fig. 6A-C) (see Zecca and Struhl, 2007). These clones also induced adjacent 5XQE>CD2>vg cells that abutted the abnormally expanded wing primordium to ectopically express CD2(Fig. 6A,B). Note that these CD2-expressing cells did not appear to express either Vg or 1XQE-lacZ, suggesting that the 5XQE transgene has a greater capacity to respond to one or more of its normal inputs than either endogenous vg or the 1XQE-lacZ transgene. Taken together, these results suggest that strengthening and/or sensitizing the QE response by introducing the 5XQE>vg transgene causes an enhanced expansion of wing tissue.

We note that even though 5XQE>vg clones formed abnormally enlarged domains of wing tissue, other elements of wing pattern were not similarly expanded within the clones. In particular, the domain of Distal-less (Dll) expression, which normally depends on Wg but is less broad than that of vg, remained unaltered in such clones(Fig. 6D). However, the Dll domain could be expanded in response to ectopic Wg(Fig. 6E). It follows that the effects of Wg on wing size (via control of vg expression) can be dissociated from its effect on wing pattern (via control of other target genes such as Dll).

The dramatic expansion of the Drosophila wing primordium following the D-V compartmental segregation provides a valuable paradigm of organ growth. Growth in this context is manifest as a rapid ∼200-fold expansion of the population of cells expressing the wing selector gene vg,under the control of the long-range morphogens Wg and Dpp. This system thus poses the fundamental question of how morphogens organize the increase in the mass and number of cells that express a given selector gene, to yield an adult appendage of appropriate size and shape.

In the accompanying paper, we defined a novel autoregulatory property of vg that appears crucial for this process. We presented evidence that vg-expressing cells send a short-range feed-forward signal that neighboring cells must receive in order to express vg in response to Wg. This led us to hypothesize that Wg controls wing development by fueling this non-autonomous autoregulatory mechanism. Here, we establish that the vg quadrant enhancer (QE) can mediate vgautoregulation in response to Wg and then use a transgene that expresses Vg under QE control to provide a proof-in-principle that wing growth normally depends on the operation of the autoregulatory circuit.

As illustrated in Fig. 7, we envisage wing growth following D-V segregation as an outcome of vgautoregulation, primed by cryptic, low-level Vg in all cells that is seeded by DSL-Notch-mediated induction of specialized D-V border cells that express high levels of vg and wg, and then propagated by the capacity of vg-expressing cells to induce and sustain vg expression in neighboring cells in response to Wg. In support, we have been able to restore wing growth in vg⁰ discs in a step-wise manner by the sequential addition of transgenes that provide, first priming(rp49-vg), then initiation (BE-vg^GFP), and finally feed-forward propagation (5XQE>vg). As we observe(Fig. 1E, Fig. 5), priming is necessary but not sufficient for wing development, initiation provides local rescue of wing tissue, and propagation is responsible for the dramatic expansion in the size of the prospective wing.

Importantly, priming and feed-forward signaling are linked in a self-reinforcing autoregulatory circuit in which a gain in either input leads to an amplification of both. We envisage that the QE normally integrates both the priming and feed-forward inputs together with Wg in a way that is sensitive to the initial strength of each input and subject to autoamplification. For example, in the `resting' state, cells have a low level of priming that falls beneath the minimal threshold necessary to specify the wing state or generate appreciable feed-forward signal. Upon receipt of sufficient Wg and feed-forward signal, the level of Vg expression rises,crossing the threshold defining the wing state and enhancing the capacity of the responding cell both to send and to receive the feed-forward signal. Amplification of this circuit then leads to the maximum output of Vg expression and feed-forward signaling that can be supported by the strength of the Wg signal received.

The self-reinforcing nature of this autoregulatory circuit, both between and within cells, helps explain how Wg spreading from D-V border cells normally fuels the expansion of the population of vg-expressing cells. It also helps account for the unexpected responses we observed in experiments using the rp49-vg, BE-vg^GFP and 5XQE>vg transgenes to mimic the normal priming, initiation and feed-forward inputs (Figs 4,5). All of these transgenes depend on heterologous promoters and potentially complex enhancer elements operating outside of their normal genomic contexts. Consequently, weak, inappropriate activities of any of these transgenes (e.g. cryptic priming by BE-vg^GFP and 5XQE>vgtransgenes, or faint QE activity of the BE-vg^GFPtransgene) could be amplified by the autoregulatory circuitry, yielding spatially inappropriate responses. Nevertheless, despite these experimental limitations, our results indicate that the major factor governing the expansion of the wing primordium is feed-forward autoregulation mediated by the QE.

As discussed in the accompanying paper(Zecca and Struhl, 2007), wing growth does not depend solely on the capacity of Wg to recruit and maintain cells in the wing primordium by fueling vg autoregulation. Instead,we show here that even when wing pouch cells are supplied constitutively with exogenous Vg (thus bypassing the requirement for vg autoregulation),they still depend on continuous Wg input to survive and grow within the context of the wing primordium (Fig. 3) (see also Johnston and Sanders, 2003). This is in contrast to cells in the more proximal hinge and notum primordia, which survive and grow without Wg input(Chen and Struhl, 1999; Giraldez and Cohen, 2003). Thus, Wg appears to promote wing growth via two distinct mechanisms: by continuously `selecting' which cells enter and remain within the wing primordium, and by allowing the survival and growth of cells so selected. We cannot, at present, distinguish the relative contributions of these two mechanisms. However, as we show here, both appear essential, as cells fail to enter, or stay, within the wing primordium when either one is eliminated.

Wing growth depends not only on Wg emanating from D-V border cells, but also on Dpp secreted by A compartment cells along the A-P compartment boundary(Zecca et al., 1995; Burke and Basler, 1996; Lecuit et al., 1996; Nellen et al., 1996; Zecca et al., 1996; Neumann and Cohen, 1997),suggesting that the QE might mediate feed-forward autoregulation in response to Dpp, as well as Wg. In support, the QE contains binding sites for the Dpp transducer Mad, and there is evidence that these sites, as well as Mad itself, contribute to QE activity(Kim et al., 1997; Halder et al., 1998; Guss et al., 2001). Moreover,clones of cells that cannot transduce Dpp behave like those that cannot transduce Wg: they cease to express Vg and are lost specifically from the wing primordium, in contrast to clones located in the more proximal hinge and notum primordia (Burke and Basler,1996; Martin-Castellanos and Edgar, 2002; Moreno et al.,2002; Gibson and Perrimon,2005; Shen and Dahmann,2005). Hence, we think it likely that Dpp and Wg act together to fuel the feed-forward autoregulatory circuit, and by so doing, regulate the size and shape of the developing wing.

The ability of Wg, and potentially Dpp, to promote wing growth by fueling a non-autonomous autoregulatory circuit of vg expression is, to our knowledge, novel, and has implications for the control of organ growth by morphogens. As epitomized by the developing wing, a long-standing enigma is that gradient morphogens drive relatively uniform growth and proliferation across a tissue at the same time that they function in a concentration-dependent manner to organize complex patterns of gene expression and cell differentiation (Garcia-Bellido and Merriam, 1971; Milan et al., 1996; Resino et al.,2002). We suggest that a minimum threshold level of morphogen might be sufficient to fuel both feed-forward autoregulation of organ selector genes and the growth and proliferation of cells so selected. Accordingly, as illustrated in Fig. 7, organ growth would be governed primarily by the progressive expansion in the range of morphogen (a process that might itself depend on the ability of morphogen to regulate expression of its receptors and other binding proteins) and by any boundary conditions that limit the availability and capacity of surrounding cells to respond.

We thank X.-J. Qiu, A. Adachi and C. Bonin for technical assistance, and K. Irvine, L. Johnston, P. A. Lawrence, R. Mann and A. Tomlinson for advice and comments on the manuscript. M.Z. is a Research Associate and G.S. an Investigator of the Howard Hughes Medical Institute.