Recruitment of cells into the Drosophila wing primordium by a feed-forward circuit of vestigial autoregulation

All animals are composed of distinct body parts and organs, many of which are specified by particular combinations of `selector' genes that control cell and tissue behavior by regulating downstream `realisator' genes(Garcia-Bellido, 1975). A central problem in animal development has been to determine how selector genes come to be active in the appropriate primordia, and how their activities govern the growth, size and pattern of the body parts they specify.

The first selector genes identified were those that specify developmental compartments in Drosophila(Garcia-Bellido, 1975; Blair, 1995). These include the Hox genes of the Bithorax complex, which control segment type, as well as the engrailed (en) and apterous (ap) genes,which govern the anterior-posterior (A-P) and dorsal-ventral (D-V)compartments within segments(Garcia-Bellido et al., 1973; Morata and Lawrence, 1975; Diaz-Benjumea and Cohen, 1993; Blair et al., 1994). All of these selector genes are initially activated in small, discrete groups of founder cells, and their states of expression, whether `on' or `off', are then stably inherited in descendent cells that constitute the compartment.

Despite the importance of compartment-specific selector genes, it is now apparent that they constitute a special class, and that most other selector genes are expressed in discrete `organ' domains that are defined by active signaling rather than by epigenetic inheritance(Mann and Morata, 2000). For example, much of the fuselage of the adult Drosophila thorax as well as the wing derives from a single `wing' imaginal disc. During development,this disc is subdivided into several sub-domains of organ-specific selector gene expression, each controlling a different portion of the adult structure: vestigial (vg) specifies the wing, homothorax(hth) specifies the hinge, and teashirt (tsh), pannier (pnr) and genes of the Iroquois (Iro-C) complex specify the body wall (notum), as well as distinct portions therein(Williams et al., 1991; Williams et al., 1993; Kim et al., 1996; Diez del Corral et al., 1999; Azpiazu and Morata, 2000; Calleja et al., 2000; Casares and Mann, 2000; Cavodeassi et al., 2000). For all these selector genes, the growth, size and shape of the organs they specify depend on the signaling mechanisms that control the expansion of the cell populations in which they are expressed. Here, we focus on how the morphogen Wingless (Wg) promotes the expansion of the population of vg-expressing cells that comprise the developing wing.

The vg gene is first activated in a small cluster of embryonic cells, defining the nascent wing imaginal disc, and is then expressed in a central portion of the disc during early larval life until the disc is segregated into dorsal (D) and ventral (V) compartments by the heritable activation of ap in D compartment cells(Williams et al., 1993; Wu and Cohen, 2002). Following the D-V segregation, cells in both compartments are programmed to send short-range Delta/Serrate/LAG-2 (DSL) signals across the compartment boundary(reviewed in Blair, 1995; Irvine and Rauskolb, 2001),activating the receptor Notch and inducing vg expression in a thin stripe of `border' cells flanking the boundary(Williams et al., 1994; Couso et al., 1995; Kim et al., 1995; Neumann and Cohen, 1996; Kim et al., 1997). DSL-Notch signaling also induces border cells to secrete Wg(Diaz-Benjumea and Cohen,1995; Rulifson and Blair,1995; de Celis et al.,1996), which acts at long range to drive vg expression in surrounding cells (Zecca et al.,1996; Neumann and Cohen,1997). Decapentaplegic (Dpp), secreted by A compartment cells along the A-P compartment boundary, also upregulates vg away from the D-V compartment boundary (Kim et al.,1996; Kim et al.,1997; Guss et al.,2001), centering a rapidly expanding population of vg-expressing cells on the intersection between the D-V and A-P boundaries.

Although generally accepted, the well-defined roles of DSL-Notch, Wg and Dpp signaling in wing development present a paradox: Wg and Dpp signaling coincide in many different contexts during Drosophila development,yet they only induce cells to express vg in the wing disc. Hence, it appears that Wg and Dpp can only recruit cells to express vg if they are already defined as prospective wing, a state that should itself depend on pre-existing Vg activity (Halder et al.,1998; Klein and Martinez-Arias, 1998; Guss et al., 2001; Curtiss et al.,2002). This paradoxical requirement suggests a crucial, but as yet unresolved, role for vg in controlling its own expression in response to Wg and Dpp.

Here, and in the accompanying paper(Zecca and Struhl, 2007), we provide evidence that Wg promotes the rapid expansion of the wing primordium following the D-V segregation by fueling a non-autonomous circuit of vg autoregulation. In this first paper, we show that vg-expressing cells send a short-range feed-forward signal that is required to entrain neighboring cells to upregulate vg in response to Wg. We also show that this process can reiterate from one cell to the next,propagating the recruitment of surrounding cells into the wing primordium. In the second paper, we show that the quadrant enhancer (QE) of the vg gene mediates this autoregulatory response and that activity of this enhancer is required for normal wing growth.

ap^56f, wg^CX4, vg^83b27R and vg^83b27 (see http://flybase.bio.indiana.edu/).

Tubα1-Gal4, UAS-GFPnls(Struhl and Greenwald, 2001); Tubα1-Gal80 (Lee and Luo, 2001); C765-Gal4,UAS>CD2,y⁺>Nrt-flu-wg,UAS>CD2,y⁺>flu-Δarm, UAS-flu-wg(Zecca et al., 1996); UAS>CD2,y+>N^intra(Struhl and Adachi, 1998), UAS-N^ECN (Struhl and Greenwald, 2001); Tubα1>flu-GFP,y⁺>Gal4,Tubα1>Gal80,y⁺>Gal4(Zecca and Struhl, 2002a; Zecca and Struhl, 2002b), UAS-wg (Struhl and Basler,1993); UAS-vg, 1XQE-lacZ(Kim et al., 1996), BE-lacZ (Williams et al.,1994), rn-lacZ (rn⁸⁹)(Couso and Bishop, 1998), wg-lacZ; dpp-Gal4(Wilder and Perrimon, 1995), Dll-lacZ (Dll⁰¹⁰⁹²)(Spradling et al., 1995).

The UAS>CD2,y⁺>vg,Tubα1>CD2,y⁺>Nrt-flu-wg,Tubα1>DsRed,y2>vg,Tubα1>flu-GFP,y⁺>vg, BE-vg^GFP,5XQE-DsRed and Tubα1-DsRed transgenes were assembled using the following DNAs: UAS promoter(Brand and Perrimon, 1993); Tubα1 promoter, Tubα1 3′UTR (Greenwood and Struhl,1997; Casali and Struhl,2004); vg QE enhancer/promoter(Kim et al., 1996), vg BE enhancer/promoter (Williams et al., 1994), DsRed coding sequence(Bevis and Glick, 2002), vg coding sequence (Kim et al.,1996); Nrt-flu-wg coding sequence(Zecca et al., 1996) and flu-GFP coding sequence (Zecca and Struhl, 2002b). The >flu-GFP,y+> and >DsRed,y2> Flp-out cassettes are derivatives of the >CD2,y+> cassette (Zecca et al., 1996) containing either the flu-GFP or DsRed coding sequences instead of CD2, respectively, and, in the case of the >DsRed,y2> cassette, a truncated form of the y+genomic fragment yielding a `y2' phenotype in y flies.

To generate the BE-vg^GFP transgene, the 750 bp EcoRI fragment defining the `minimal' vg BE(Williams et al., 1994) was inserted in (-750 to -1) orientation upstream of the minimal Hsp70-promotor (Lis et al.,1983) and the coding sequence for vg^GFP. The Vg^GFP chimera consists of the first 20 amino acids of Vg joined via one copy of the flu-tag to GFP, followed by two flu-tags, joined to the rest of Vg at amino acid 25. The N- and C-terminal joints to Vg are, respectively, PYLYGR/GSYPYDVPDYA and YAGPYDVPDYags/RSFYQYE (Vg sequences are underlined and slashes highlight the joints).

To generate the 5XQE promoter, the main portion of the 750 bp BE (-750 to -138) within the BE-vg^GFP transgene was replaced with five copies of the 806 bp QE fragment of the vg gene (Kim et al.,1996), arrayed each in (-806 to -1) orientation, while the vg^GFP coding sequence was replaced with that of DsRed. The remaining 137 bp fragment of the original BEsequence includes a Su(H)-binding site that is essential within the context of the intact 750 bp BE for its activity in D-V border cells. However,the presence of this binding site is not sufficient, within the context of the 5XQE-DsRed transgene [and related 5XQE transgenes reported in the accompanying study (Zecca and Struhl, 2007)], to drive detectable marker gene expression within boundary cells, whether within the wing pouch, or in the presumptive hinge or notum primordia (e.g. Fig. 1B,C). In addition, 5XQE reporter derivatives in which this Su(H)-binding site is inactivated as in Kim et al.(Kim et al., 1996) or deleted,generate a pattern of expression that is indistinguishable from 5XQEreporter transgenes in which the site remains intact (as in Fig. 1B,C).

Flp-out (Struhl and Basler,1993; Zecca et al.,1996), Gal4/UAS (Brand and Perrimon, 1993) and MARCM (Lee and Luo, 2001) techniques were used to manipulate gene activity in marked clones of cells or entire discs. To obtain clones ectopically expressing two coding sequences, we used the Tubα1>flu-GFP,y⁺>Gal4transgene to activate gene expression of two UAS- transgenes within the same cells (e.g. UAS-vg and UAS-wg; as in Fig. 7A,B) or, alternatively, a Flp-out transgene generating clones of cells expressing one gene under Tubα1-control in discs that express the other gene under C765-Gal4/UAS control (e.g. Tubα1>vgclones in C765-Gal4/UAS-Nrt-wg discs; as in Fig. 4B). To obtain discs containing clones ectopically expressing one coding sequence adjacent to clones or tissue ectopically expressing a second, we employed various permutations of the Flp-out and Gal4/UAS methods (e.g. Tubα1>vg Flp-out clones next to UAS>Nrt-wg Flp-out clones in C765-Gal4 discs; as in Fig. 5). Finally, the MARCM technique was used to generate wg^cx4 mutant clones ectopically expressing one or more coding sequences (e.g. UAS-vg plus UAS-Nrt-wg in Fig. 3E).

Clones were generated by heat shock-induced Flp recombinase as described previously (e.g. Zecca and Struhl,2002b). Unless otherwise stated, clones were induced during the first larval instar (24-48 hours after egg laying), prior to when the D-V compartmental segregation normally occurs (mid- to late second instar), and mature wing discs dissected, fixed and analyzed at the end of the third larval instar using standard protocols (e.g. Zecca and Struhl, 2002b). Antisera employed: anti-Wg (Brook and Cohen, 1996), anti-Vg(Williams et al., 1991),anti-Flu (Roche), anti-β-gal (Cappel) and anti-CD2 (OX34, Serotec).

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

1E: 1XQE-lacZ ap^56f/1XQE-lacZ ap^56f.

1F: y w Hsp70-flp UAS-GFPnls/y w Hsp70-flp; 5XQE-DsRed ap^56fvg^83b27R/ap^56f UAS-N^ECN;Tubα1>Gal80,y⁺>Gal4/rn-lacZ (no heat shock treatment).

1G: y w 5XQE-DsRed/y w Hsp70-flp; UAS-Nrt-flu-wg ap^56f/ap^56f; C765-Gal4/+.

1H: UAS-Nrt-flu-wg ap^56f/1XQE-lacZ ap^56f;C765-Gal4/BE-vg^GFP.

2A: 1XQE-lacZ ap^56f/ap^56f; Tubα1>Gal80,y⁺>Gal4/UAS>CD2,y⁺>N^intra.

2B: y w Hsp70-flp UAS-GFPnls/y w Hsp70-flp; 5XQE-DsRed ap^56fvg^83b27R/ap^56f UAS-N^ECN;Tubα1>Gal80,y⁺>Gal4/rn-lacZ.

2C,D: ap^56f UAS-N^ECN/1XQE-lacZ ap^56f;Tubα1>Gal80,y⁺>Gal4/BE-vg^GFP.

2E: 5XQE-DsRed ap^56f vg^83b27R/ap^56fUAS-N^ECN; Tubα1>flu-GFP,y⁺>Gal4/BE-lacZ.

3A: y w Hsp70-flp Tubα1-Gal4 UAS-GFPnls/y w Hsp70-flp;Tubα1-Gal80 FRT39 ap^56f/FRT39 ap^56f;UAS-vg/1XQE-lacZ.

3B: wg-lacZ ap^56f/ap^56f UAS-vg;Tubα1>flu-GFP,y⁺>Gal4/+.

3C: y w Hsp70-flp Tubα1-Gal4 UAS-GFPnls/y w Hsp70-flp;Tubα1-Gal80 FRT39 ap^56f/wg^cx4 FRT39 ap^56f; UAS-vg/1XQE-lacZ.

3D: y w Hsp70-flp Tubα1-Gal4 UAS-GFPnls/y w Hsp70-flp;Tubα1-Gal80 FRT39 ap^56f/wg^cx4 FRT39 ap^56f; UAS-wg UAS-vg/1XQE-lacZ.

3E: y w Hsp70-flp Tubα1-Gal4 UAS-GFPnls/y w Hsp70-flp;Tubα1-Gal80 FRT39 ap^56f/wg^cx4 FRT39 ap^56f; UAS-Nrt-flu-wg UAS-vg/1XQE-lacZ.

4A: 1XQE-lacZ ap^56f vg^83b27/1XQE-lacZ ap^56f; Tubα1>flu-GFP,y⁺>vg UAS-Nrt-flu-wg/+.

4B: 1XQE-lacZ ap^56f vg^83b27/ap^56f;Tubα1>flu-GFP,y⁺>vg UAS-Nrt-flu-wg/C765-Gal4.

5: UAS>CD2,y⁺>Nrt-flu-wg ap^56f/1XQE-lacZ ap^56f; Tubα1>DsRed,y²>vg/C765-Gal4.

6: 1XQE-lacZ ap^56f/1XQE-lacZ ap^56f;Tubα1>DsRed,y²>vg C765-Gal4/UAS>CD2,y⁺>flu-Δarm.

7A: UAS-vg/+; Tubα1>flu-GFP,y⁺>Gal4 UAS-wg/+.

7B: y w 5XQE-DsRed/y w Hsp70-flp; UAS-vg/vg^83b27R;Tubα1>flu-GFP,y⁺>Gal4 UAS-wg/rn-lacZ.

7D-F: Tubα1>CD2,y⁺>Nrt-flu-wg/1XQE-lacZ;dpp-Gal4/UAS-vg.

7G: UAS>CD2,y⁺>Nrt-flu-wg/+;UAS>CD2,y⁺>vg/C765-Gal4.

7H: UAS>CD2,y⁺>flu-Δarm/+;UAS>CD2,y⁺>vg/C765-Gal4.

Following segregation of the wing disc into D and V compartments, vg expression is driven by DSL-Notch signals received by border cells flanking the D-V boundary, and by Wg and Dpp received by neighboring cells in the wing pouch, the central portion of the disc that normally gives rise to the adult wing (reviewed by Blair,1995; Irvine and Rauskolb,2001). The `border' and `pouch' expression domains are mediated by distinct boundary and quadrant enhancers (BE and QE)(Williams et al., 1994; Kim et al., 1996). Here, and in the accompanying paper (Zecca and Struhl, 2007), we monitor the activities of these enhancers by assaying transgenes that express reporter proteins such as β-gal, rat CD2 and DsRed, or a functional Vg-GFP chimeric protein, under their control(Fig. 1A-D). In some experiments, we assay QE activity using transgenes containing a tandem array of five quadrant enhancers; these 5XQE reporters appear to provide a more sensitive and faithful indicator of Wg- and Vg-dependent pouch expression than the more standard 1XQE reporters(Fig. 1B,C)(Zecca and Struhl, 2007).

In the generally accepted model of wing development, Wg emanating from D-V border cells drives vg expression in the surrounding pouch cells via activation of the QE. Accordingly, if the D-V segregation is blocked,as in ap mutant (ap⁰) discs, border cells are not specified and the early expression of vg that normally precedes the D-V segregation and defines the nascent wing pouch rapidly dissipates(Williams et al., 1993)(Fig. 1E; data not shown). In mature, third instar discs, the wing pouch is normally surrounded by the inner of two rings of Wg expression (e.g. Fig. 1A,B) in the prospective wing hinge; in ap⁰discs, the pouch is absent and the inner ring reduced to a cluster of cells(e.g. Fig. 1E,F).

In agreement with the model, clones of cells that express Notch^intra (N^intra) or Notch^ECN(N^ECN) - constitutively active forms of Notch that are under Gal4/UAS control (Struhl and Greenwald, 2001) (henceforth UAS-N^* clones) -in ap⁰ discs behaved as ectopic border cells and rescued wing growth even when generated up to 48 hours after the time at which D-V segregation would normally occur (Fig. 2). This rescuing activity was associated with the autonomous activation of wg, vg and BE reporter gene expression within the clone and, more strikingly, with the long-range non-autonomous induction of vg and QE reporter gene expression in surrounding cells(Fig. 2). As in wild-type discs(Fig. 1B), the restored vg and QE expression did not extend all the way to the inner ring of Wg expression, but instead was separated from it by a zone of cells distinguished by expression of the transcription factor gene rotund(rn), without detectable vg (the `rn-only' domain; Fig. 1B, Fig. 2B).

In conflict with the model, supplying ectopic Wg to ap⁰discs, whether by making UAS-wg clones or by ubiquitously expressing UAS-wg throughout, failed to rescue the expression of vg, QEreporter genes, or other `pouch genes' such as Distal-less(Fig. 1G,H; data not shown)(see also Klein and Martinez-Arias,1998; Klein and Martinez-Arias, 1999). Hence, Wg is not sufficient to activate QE-dependent vg expression or to rescue wing development in the absence of D-V border cells.

The contrasting results obtained with UAS-N^* and UAS-wg clones suggests that, in ap⁰ discs, cells might need to express vg to be able to induce neighboring cells to upregulate vg expression. In support, we found that UAS-vgclones mimic the effects of UAS-N^* clones: when generated in ap⁰ discs, they were able to induce surrounding cells up to several cell diameters away to express endogenous vg as well as QE reporters and to develop as wing cells(Fig. 3A).

ap⁰ discs lack border cells and hence the normal source of Wg, posing the question of how UAS-vg clones induce surrounding cells within such discs to activate the QE. UAS-vg clones express levels of exogenous Vg that greatly exceed the peak levels normally expressed in the developing wing (Fig. 3A; data not shown): such physiologically abnormal levels might artifactually activate wg expression, providing an ectopic source of Wg required for the non-autonomous activation of the QE.

To test this, we first asked whether UAS-vg clones in ap⁰ discs ectopically express wg, as monitored by the expression of a wg-lacZ reporter gene. As shown in Fig. 3B, this was indeed found to be the case.

We next asked whether the long-range rescuing activity of such UAS-vg clones depends on their ability to ectopically express wg, using the MARCM technique(Lee and Luo, 2001) to make UAS-vg clones that are also wg⁰. In contrast to UAS-vg clones that were otherwise wild-type(Fig. 3A), wg⁰UAS-vg clones in ap⁰ discs failed to induce surrounding cells to express vg or the 1XQE-lacZ reporter gene (Fig. 3C).

Finally, we asked whether the long-range rescuing activity of UAS-vg clones depends on the long-range action of the ectopic Wg they secrete. To do this, we generated wg⁰ UAS-vg clones that co-overexpress either a UAS-wg or UAS-Neurotactin-wg(Nrt-wg) transgene, the latter encoding a membrane-tethered immobile form of Wg (Zecca et al.,1996). Co-overexpression of UAS-wg restored the ability to induce vg and QE reporter gene expression in surrounding cells up to several cell diameters away(Fig. 3D); co-overexpression of UAS-Nrt-wg did so as well, but only at short range - in cells near,or next to the clone (Fig. 3E).

We conclude that Vg-overexpressing cells have the capacity to induce neighboring cells to activate QE-dependent vg expression,but only if the responding cells also receive Wg.

The non-autonomous vg-inducing activity of UAS-vg clones might reflect either a normal role for vg in controlling its own expression in neighboring cells, or an artifactual outcome resulting from Vg overexpression within the clone. To distinguish between these possibilities,we made clones that express moderate levels of Vg in ap⁰discs, and asked whether they could induce neighboring cells to express vg in a Wg-dependent fashion.

Clones of cells that express vg under the direct control of the Tubα1 (αTub84B - Flybase) promoter(henceforth Tubα1>vg clones) generated moderate levels of Vg that were 2- to 3-fold below the normal peak of endogenous Vg expression (Fig. 4, Fig. 5A,C, Fig. 6A), but still sufficient to rescue wing development within the clone in vg⁰ wing discs (Zecca and Struhl,2007).

Tubα1>vg clones also autonomously rescued wing development in ap⁰ discs(Fig. 4A). However, they appeared to have little or no capacity to rescue expression of the endogenous vg gene (or that of BE or QE reporter genes),either inside or outside of the clones, except for a weak, cell-autonomous activation of QE reporter genes(Fig. 4A; data not shown). The failure of these clones to induce vg expression in surrounding cells correlates with the absence of wg-expressing border cells (normally specified by DSL-Notch signaling across the D-V compartment boundary, but absent from ap⁰ discs).

By contrast, when we repeated this experiment in ap⁰discs that express UAS-Nrt-wg uniformly throughout, we found that Tubα1>vg clones that were located within the prospective wing pouch were associated with robust activation of QEreporters as well as of the endogenous vg gene to normal levels(Fig. 4B). Significantly, this upregulation occurred not only within the clone, but also in surrounding cells up to many cell diameters away, providing a general rescue of wing development. By contrast, when we generated Tubα1>vg clones that co-express UAS-Nrt-wgin ap⁰ discs, the upregulation of QE-dependent vg expression was restricted to cells within the clone and their immediate wild-type neighbors, as expected given the restricted range of membrane-tethered Nrt-Wg (Fig. 5A,B).

Thus, cells that express moderate levels of exogenous Vg, well within the normal physiological range, can drive QE-dependent vgexpression in neighboring cells, provided that the responding cells also receive Wg signal.

The preceding experiments indicate that Wg is not sufficient to induce wing disc cells to express high levels of vg and develop as wing cells. Instead, Wg can only do so in the presence of cells that already express Vg. One explanation for this unexpected requirement is that Vg activity programs cells to make a second intercellular signal that is required in combination with Wg.

To test this, we analyzed ap⁰ wing discs that contained neighboring Tubα1>vg and UAS-Nrt-wgclones. As described above, ap⁰ discs that contain only one or the other type of clone do not express either vg or the 1XQE-lacZ reporter (except for weak cell-autonomous expression of the reporter detected within Tubα1>vg clones). However,in discs that contain both types of clones, Tubα1>vg clones were able to induce cells within neighboring UAS-Nrt-wg clones to express peak levels of both vg and the 1XQE reporter, provided that the two types of clones abutted and that the UAS-Nrt-wg cells were located in the prospective wing pouch (Fig. 5C,D). Importantly, neither output was limited to those UAS-Nrt-wg cells in direct contact with Tubα1>vg cells. Instead, vg and QE reporter gene expression typically propagated many cell diameters into the UAS-Nrt-wg clone and could extend to abutting wild-type neighbors on the opposite side of the clone(Fig. 5C,D). Similar results were obtained when we replaced clones expressing Nrt-Wg with clones expressing Arm^*, a truncated form of the cytosolic protein Armadillo that constitutively activates the Wg transduction pathway(Zecca et al., 1996). Tubα1>vg clones were able to induce cells in neighboring UAS-arm^* clones to express vg and QE reporter genes, much as they do when they abut UAS-Nrt-wgclones, except that expression did not propagate further to adjacent wild-type cells (as expected, because these UAS-arm^* cells should not provide Wg signal, in contrast to UAS-Nrt-wg cells).

Thus, Tubα1>vg clones can induce neighboring UAS-Nrt-wg and UAS-arm^* cells to activate QE-driven expression of vg, and the effect can then propagate from one such cell to the next, presumably via reiterative cycles of short-range induction of endogenous vg by cells already expressing Vg. In both cases, we only observed induction when the two types of clones abutted, and the effect could only propagate in cells in which the Wg transduction pathway was active, whether in response to Wg signal or to constitutive activation of Arm. We conclude (1) that Vg-expressing cells send an independent signal that acts combinatorially with Wg to induce QE-dependent vg expression in neighboring cells; (2) that this vg-dependent vg-inducing signal has a range of only one, or very few, cell diameters, possibly because it is contact-mediated; and(3) that this signal is integrated with Wg downstream of the cytosolic activation of Arm.

In the experiments described above, we have been able to substitute for the absence of border cells within the prospective wing pouch of ap⁰ discs by generating clones of ectopic Vg- and Wg-expressing cells: such clones restore QE-dependent vgexpression and rescue wing development. Here, we ask whether ectopic Vg- and Wg-expressing clones can activate the QE and induce the formation of ectopic wing primordia in more proximal portions of the wing disc that would normally give rise to the notum.

UAS-wg, UAS-Nrt-wg and UAS-arm^* clones in otherwise wild-type discs upregulate vg and QE reporter gene expression within the prospective wing pouch, but do not induce ectopic vg expression in the proximal hinge and notum primordia (data not shown) (Zecca et al., 1996; Klein and Martinez-Arias,1998; Baena-Lopez and Garcia-Bellido, 2003). By contrast, most UAS-vg clones activated QE reporter gene expression in the hinge and notum primordia, but unlike UAS-vg clones in the prospective wing pouch(Fig. 2), they did not express wg ectopically and did not induce QE-dependent vgexpression in surrounding cells (data not shown). Strikingly, UAS-wg UAS-vg clones that co-overexpressed Wg and Vg in these primordia were able to induce QE-dependent vg expression non-autonomously,seeding the formation of ectopic wing primordia(Fig. 7A,B) (see also Baena-Lopez and Garcia-Bellido,2003). To assess whether this reflects a spatially inappropriate,but otherwise normal, response of wing disc cells to the combination of Wg-and Vg-dependent signals, we performed the following two sets of experiments.

First, we used dpp-Gal4 to drive UAS-vg expression in a stripe of A compartment cells abutting the A-P boundary in the notum and asked whether these cells could induce QE-driven vg expression in neighboring P clones of Tubα1>Nrt-wg cells across the A-P boundary (Fig. 7C). Although such Tubα1>Nrt-wg clones expressed only moderate levels of Nrt-Wg and were normally unable to induce either vg or QE reporter gene expression in this portion of the disc, we found that they could do so if they were located next to the UAS-vg-expressing cells in the A compartment(Fig. 7D-F). Furthermore, vg and QE reporter gene expression could spread many cell diameters into the Tubα1>Nrt-wg clone, and extend outside of the clone to the adjacent wild-type cells(Fig. 7D-F). Propagation did not, however, extend more than ∼20-25 cell diameters from the A-P boundary(Fig. 7D,E), possibly reflecting a requirement for Dpp secreted by A cells along the compartment boundary.

In the second set of experiments, we expressed UAS-vg in clones,rather than under dpp-Gal4 control, and examined the response of UAS-arm^* as well as of UAS-Nrt-wg clones. We observed that UAS-vg-expressing clones can induce UAS-Nrt-wgas well as UAS-arm^* clones within the prospective notum to express vg, provided that they abut(Fig. 7G,H). Moreover, vg expression was able to spread through the responding clone, and even extended to neighboring cells outside in the case of UAS-Nrt-wgclones (but not UAS-arm^* clones).

Thus, Vg-overexpressing cells in the prospective notum can induce neighboring cells to express vg, provided that the latter also receive Wg input, and once initiated, vg expression can propagate to cells further away. Importantly, the levels of Vg expression and Wg signaling within such Tubα1>Nrt-wg clones fall within the physiological range, arguing that propagation is not an artifact of either Vg or Wg overexpression, but rather a normal signaling process operating in a spatially inappropriate context.

Following the D-V segregation, local DSL-Notch signaling across the compartment boundary induces the differentiation of specialized border cells that express vg, secrete Wg, and organize a dramatic ∼200-fold expansion of the wing primordium. In ap⁰ wing discs, D-V segregation fails to occur, border cells are not specified, and the early expression of vg that initially defined the wing primordium fades away. We have used this mutant condition to explore how vg and wg activity in border cells controls wing growth by asking what happens when we replace the missing border cells with cells that ectopically express Wg, Vg or both.

Our main finding is that Wg is not sufficient to sustain or induce vg expression in ap⁰ discs, even when the morphogen is overexpressed, continuously, in all cells. Instead, Wg can only drive vg expression in these discs when the responding cells are near or next to cells that express exogenous Vg. The clearest demonstration of this is the experiment in which we generated two types of clones in the same ap⁰ disc: clones that express Nrt-Wg, a membrane tethered immobile form of Wg, and clones that express moderate levels of exogenous Vg(Fig. 5C,D). Neither type of clone, alone, can restore normal expression of the endogenous vggene. However, ectopic Vg-expressing clones can induce neighboring Nrt-Wg-expressing clones to express vg, provided that they abut. Moreover, this vg expression can spread through the Nrt-Wg-expressing clone and extend to adjacent cells outside the clone.

These results indicate that vg-expressing cells send a short-range, possibly contact-dependent signal that is required to entrain neighboring cells to express vg in response to Wg. Furthermore, they indicate that once the responding cells express vg, they can in turn entrain their neighbors in the same way, propagating the recruitment of additional cells into the wing primordium. These findings establish the existence of a Wg-dependent feed-forward circuit of vg autoregulation and suggest that D-V border cells normally organize wing growth by providing Wg, as well as the initial Vg-dependent entraining signal that triggers reiteration of this autoregulatory circuit from one cell to the next(Fig. 8) [see also Fig. 7 in the accompanying paper (Zecca and Struhl,2007)]. Thus, feed-forward regulation in this context has a spatial component, mediating the expansion (in mass and cell number) of a developing primordium by a process of recruitment.

We note that our results are concordant with previous reports that Wg signaling cannot drive vg expression in the wing imaginal disc in the absence of border cells (Klein and Martinez-Arias, 1998; Klein and Martinez-Arias, 1999), and that co-overexpression of Wg and Vg can synergize to drive vg expression in surrounding cells(Klein and Martinez-Arias,1999; Baena-Lopez and Garcia-Bellido, 2003). However, our findings advance these results in three significant ways. First, we show that vg-expressing cells provide a discrete second signal, required together with Wg, to induce vg expression in surrounding cells. Second, we demonstrate that production of this signal can propagate from one cell to the next, establishing a feed-forward autoregulatory mechanism fueled by morphogen. Third, we show that physiologically normal levels of wg and vg activity are sufficient to initiate and propagate this feed-forward mechanism, establishing that it is a natural process and not an overexpression artifact.

Although we do not know the identity of the Vg-dependent feed-forward signal, our results argue that it should have a range of only one or very few cell diameters and that mutations that block its production or reception should prevent QE-dependent vg expression following D-V segregation. DSL ligands appear unlikely candidates, as high-level DSL-Notch signaling represses the QE (Go et al., 1998; Klein and Martinez-Arias, 1999; Nagaraj et al., 1999) (but see Baonza and Garcia-Bellido, 2000). Another possibility is an antagonist of the Drosophila Epidermal growth factor receptor (DER)/Ras pathway. DER/Ras and Wg pathways have generally opposing activities during wing disc development, with the DER/Ras pathway promoting notum and hinge development and the Wg pathway promoting wing blade development(Wang et al., 2000; Zecca and Struhl, 2002a; Zecca and Struhl, 2002b). Hence, vg-expressing cells might induce surrounding cells to upregulate vg by damping down activity of the DER/Ras pathway, which would otherwise block vg expression in response to Wg. A final,intriguing possibility would be a regulator of the Dachsous/Fat (Ds/Ft)pathway, which has been implicated in the control of proximal wing growth by prospective distal wing tissue (Cho and Irvine, 2004; Rodriguez,2004; Jaiswal et al.,2006). Because loss of either Ds or Ft causes extra wing growth,the feed-forward signal would presumably promote expansion of the wing primordium by acting as an antagonist of the pathway.

The capacity of Wg to drive recruitment of new cells into the wing primordium by fueling vg feed-forward autoregulation provides one mechanism for promoting wing growth. However, it appears to operate within the context of other mechanisms for promoting wing growth, as well as for limiting where and when such growth occurs.

We can distinguish at least three additional mechanisms for promoting wing growth, all dependent on Wg. First, in addition to recruiting new cells into the wing primordium, Wg acts continuously to retain cells that were previously recruited: wing cells in which Wg transduction is abrogated rapidly lose vg expression and either die, or sort out(Zecca et al., 1996; Neumann and Cohen, 1997). We suggest that retention, like recruitment, depends on the same Wg-dependent vg autoregulatory circuit. Specifically, we posit that the feed-forward signal is required both to induce vg expression in cells about to enter the primordiium, as well as to maintain vg expression in cells after they enter (Fig. 8).

Second, independent of its role in fueling vg autoregulation, Wg also appears necessary for the survival and proliferation of vg-expressing wing cells. As we describe in the accompanying paper,it is possible to bypass the requirement for Wg-dependent vgautoregulation by using a Tubα1>vg transgene to express exogenous Vg: nevertheless, such `rescued'Tubα1>vg wing cells still require Wg input to survive, grow and proliferate (Zecca and Struhl, 2007).

Third, cells are normally recruited into the vg-expressing population from a surrounding population defined by detectable expression of rn but not vg. Accordingly, the `rn-only'population must proliferate in conjunction with the growth of the wing primordium; otherwise, it would be depleted, limiting further recruitment and compromising the development of more proximal structures. In support, we find that the rescue of the wing primordium by Wg-dependent vgautoregulation is associated with the rescue and expansion of the surrounding population of rn-only cells (e.g. Fig. 2) (see also Liu et al., 2000; del Alamo Rodriguez et al.,2002; Kölzer et al.,2003; Cho and Irvine,2004). Hence, once cells are recruited into the wing primordium in response to Wg, they may send an additional signal to sustain the source population of rn-only cells from which additional wing cells will be recruited (Fig. 8) [see also Fig. 7 in the accompanying paper (Zecca and Struhl,2007)].

Conversely, we can distinguish at least three mechanisms that appear to constrain operation of the feed-forward circuit, limiting expansion of the wing primordium in space and time. First, is the early segregation of the wing imaginal disc into distinct distal (pre-blade) and proximal (pre-hinge/notum)compartments, only one of which, the pre-blade, is competent to engage the feed-forward autoregulatory circuit. This event, which occurs before D-V compartmental segregation, appears to be governed by an early burst of Wg signaling that selectively and heritably represses tsh expression in the founder cells of the putative pre-blade (tsh^OFF)compartment (Wu and Cohen,2002; Zirin and Mann,2004) [see Fig. 7in the accompanying paper (Zecca and Struhl, 2007)]. Although Wg-dependent vg autoregulation normally appears to operate only within the resulting pre-blade(tsh^OFF) compartment (which includes the rn-only domain, as well as the presumptive wing pouch), this limit can be exceeded if cells are exposed to ectopic Wg signal before they would otherwise segregate into the pre-hinge/notum (tsh^ON) compartment(Ng et al., 1996; Wang et al., 2000; Baena-Lopez and Garcia-Bellido,2003) (Fig. 6). We suggest that this ectopic Wg activity inappropriately blocks tshactivity in the prospective pre-hinge/notum, creating an ectopic pre-blade compartment in which feed-forward regulation can occur.

Second, is the availability of Dpp secreted by A compartment cells along the A-P compartment boundary. Dpp, like Wg, is essential for vgexpression and wing growth (Posakony et al., 1990; Zecca et al.,1995; Burke and Basler,1996; Kim et al.,1997; Guss et al.,2001). Hence, operation of the feed-forward mechanism might depend on the combined inputs of Wg and Dpp, centering the expanding domain of Wg-dependent vg expression on the intersection between the D-V and A-P compartment boundaries. In agreement, we only observe evidence for Wg-dependent feed-forward propagation in cells located within ∼25 cell diameters of the A-P boundary, the expected range of Dpp emanating from A cells along the boundary.

Third, operation of the vg feed-forward circuit might be temporally constrained. We find it striking that vg is initially expressed in ap-null discs up until the time the D-V compartmental segregation would normally occur; yet, flooding such discs with exogenous Wg signal (e.g. as in Fig. 1G,H)is not sufficient to sustain and propagate this early vg expression. By contrast, clones of Tubα1-vg cells generated in these same discs are effective in triggering the propagation of vgexpression in surrounding cells (Fig. 4B), suggesting that cells within the `pre-blade' only become competent to operate the feed-forward autoregulatory circuit after the time at which the D-V segregation normally occurs, concomitant with the differentiation of wg- and vg-expressing border cells.

Thus, we propose that following the D-V segregation, Wg drives wing growth by at least four distinct outputs: first, by recruiting new cells into the wing primordium; second, by maintaining the recruited cells and their descendents within the primordium; third, by sustaining the survival and proliferative growth of cells defined as `wing' by the selector activity of Vg; and finally, by acting through the agency of newly recruited wing cells to induce the expansion of the surrounding population of rn-only cells from which additional wing cells will be recruited. Counterbalancing these effects would be a requirement for heritable repression of tsh,availability of Dpp, and transition to a discrete phase of wing disc development during which the feed-forward circuit can operate. Within these constraints, the size of the wing primordium at any point following the D-V segregation would reflect the increasing range of Wg emanating from the D-V border cells via its capacity to propagate and sustain the vgautoregulatory circuit and, separately, its capacity to promote the proliferative growth of vg- and rn-only-expressing cells.

We thank X.-J. Qiu, A. Adachi and C. Bonin for technical assistance, S. Carroll, A. Guichard and E.Bier for fly stocks and reagents and U. Alon, L. Johnston, K. Irvine, P. A. Lawrence, R. Mann, M. Parter and A. Tomlinson for advice and comments on the manuscript. We also thank K. Basler for support during the initial stages of the study. M.Z. is a Research Associate and G.S. an Investigator of the Howard Hughes Medical Institute.