Functional subdivision of trunk visceral mesoderm parasegments inDrosophila is required for gut and trachea development

Interactions between germ layers are essential for the formation and development of tissues and organs in embryogenesis. Drosophilavisceral mesoderm would appear to be a good system for molecular and genetic study of these processes. Midgut visceral mesoderm consists of trunk and caudal visceral mesoderm, which, respectively, give rise to circular and longitudinal muscles (Campos-Ortega and Hartenstein, 1997). The formation and specification of trunk visceral mesoderm (VM), require overlaying ectodermal signals(Azpiazu et al., 1996;Riechmann et al., 1997), while it regulates the development of underlying endodermal tissue(Bienz, 1994). Gut digestion is carried out under the control of visceral muscles(Dettman et al., 1996).

During gastrulation and subsequent stages, future mesoderm invaginates,spreading laterally and dorsally to form a rather uniform cell monolayer(mesoderm), in close contact with overlaying ectoderm(Leptin and Grunewald, 1990). As with ectoderm, the mesoderm consists of parasegmental units (PSs), each divided into even-skipped (eve) and sloppy-paired(slp) domains along the anteroposterior (AP) axis(Borkowski et al., 1995;Azpiazu et al., 1996;Riechmann et al., 1997). The concerted action of Decapentaplegic (DPP), which emanates from the dorsal ectoderm, and Wingless (WG) and Hedgehog (HH), both of which are secreted from alternating ectodermal sources along the AP axis, must be available for further subdivision of mesodermal PS and subsequent mesodermal cell differentiation (reviewed by Baylies, 1998).

VM development begins with VM competent cell specification within the dorsoanterior quarter of each mesodermal parasegment. VM competent regions may be defined using expression of bagpipe (bap), a homeobox gene, which is positively regulated by hh, dpp and tinman(tin), and negatively by wg(Staehling-Hampton et al.,1994; Frasch,1995, Azpiazu et al.,1996; Lee and Frasch,2000). During stage 10, the anterior terminus of each VM competent region always appears to touch an anteroposterior compartment border (AP border) or ectodermal WG source; the putative posterior terminus, which is initially situated halfway between two flanking AP borders, approaches the neighboring ectodermal WG source at late stage 10(Borkowski et al., 1995;Azpiazu et al., 1996;Weiss et al., 2001). Each VM competent region splits into two regions that consist of peripheral columnar cells, which may constitute a VM progenitor region destined for founder myoblasts of circular muscles, and another comprising hexagonal cells, future fusion-competent visceral myoblasts (San Martin et al., 2001; Klapper et al., 2002). A string-like VM arch structure, positive for Fasciclin 3 (FAS3), is formed through head-to-tail connection of neighboring VM progenitor regions. VM progenitor cells eventually fuse with fusion-competent myoblasts to form circular visceral muscles(San Martin et al., 2001;Klapper et al., 2002).

As with mesoderm, VM possesses segmental modules (trunk visceral mesodermal parasegments; VM-PSs). Connection (CON), a cell adhesion molecule, is expressed in VM and its expression is well aligned with ectodermal parasegment borders (Bilder and Scott,1998). Bilder and Scott (Bilder and Scott, 1998) considered VM-PSs to consist of CON-positive and-negative regions, and suggested that con expression in VM-PSs is positively and negatively regulated by ectodermal hh and wg,respectively.

VM confers positional cues to the endoderm(Szuts et al., 1998).dpp and wg are expressed in VM-PS7 and VM-PS8, respectively. They are regulated by a positive feedback loop and, together, controllabial expression in underlying endoderm(Hoppler and Bienz, 1995;Bienz, 1997).

Attention in this study is directed to cell fate diversification within each VM-PS. We first show that VM-PSs are subdivided into five or six regions based on differences in the expression of five VM-PS genes, which includehh and branchless (bnl)(Sutherland et al., 1996). Shift-up/down experiments indicate that VM-PS gene expression is regulated by ectodermal HH and WG signals in different ways. Finally, we show thathh expression in VM-PS3 is required for normal gastric caecum development, while metameric expression of bnl in VM serves as a guidance of the initial budding of visceral branches (VB) of the trachea, an ectodermal organ.

Canton S was used as wild type. Mutant strains, enhancer-trap lines and GAL4-UAS lines used are:hh^13C,wg^CX4,hh^9K,wg^IL4,VM-hh-lacZ (K. T., unpublished), bap-lacZ (baplac4.5#23)(Azpiazu and Frasch, 1993),ftz-lacZ (Hiromi et al.,1985), UAS-hh (Suzuki and Saigo, 2000), UAS-wg, UAS-ciNZn(Hepker et al., 1997),UAS-dTCF-ΔN (Cavallo et al., 1998), twi-GAL4, 24B-GAL4, bap-GAL4(Zaffran et al., 2001) and 48Y-GAL4. See FlyBase for fly strains whose sources are not indicated. Genotypes of embryos were identified by the aid of lacZ or GFP balancers. Embryonic stages were determined according to Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1985). All the experiments other than temperature-shift-experiments were carried out at 25°C.

Immunohistochemistry was carried out as described previously(Suzuki and Saigo, 2000). Primary antibodies used are rabbit anti-HH (1:1000)(Tabata and Kornberg, 1994),mouse anti-WG (mab4D4, DSHB), rabbit anti-TIN (1:900; generously provided by Manfred Frasch), mouse anti-CON (1:5)(Meadows et al., 1994), mouse anti-PTC (1:200) (Capdevila et al.,1994), mouse anti-FAS3 (mab7G10, DSHB), mouse anti-FAS2 (1:10)(Grenningloh et al., 1991) and anti-lacZ protein [rabbit polyclonal (Cappel); mouse monoclonal(Promega)] antibodies. TSA indirect amplification kit (Renaissance) was used if necessary. Double fluorescence labeling with riboprobe and antibody was carried out as described by Suzuki et al.(Suzuki and Saigo, 2000). cDNA clones used as hybridization probes were bnl(Sutherland et al., 1996),bap (Azpiazu and Frasch,1993), Wnt4 (Graba et al., 1995), hh(Tashiro et al., 1993),dpp (Sato and Saigo,2000) and vein(Yarnitzky et al., 1997).

Temperature shift-up/down experiments were carried out essentially as described (Matsuzaki and Saigo,1996). hh^9K/hh^13C andwg^IL114 flies were used as hh^ts andwg^ts mutants, respectively. Developmental times for shift-up/down experiments are normalized to growth rate at 25°C and shown by hours after egg laying (AEL).

Larvae with appropriate genotypes were identified by the aid of the GFP-balancer and dissected guts were stained with phalloidin-FITC (Molecular Probes) as described previously (Hoppler and Bienz, 1994).

Enhancer analysis of hh demonstrated that one hh enhancer fragment (Sph-hh) is capable of inducing reporter gene expression segmentally in VM (K. T. and K. S., unpublished)(Fig. 1A2). lacZdriven by Sph-hh and also hh RNA and protein were expressed as ten VM patches, well aligned with overlaying ectodermal hh stripes(Fig. 1A1,B). Expression ofhh RNA in late stage 11 embryos almost entirely overlapped with that of lacZ driven by Sph-hh(Fig. 1A1-A3). Consequently,hh was concluded to be expressed in VM in a metameric fashion.hh expressed in VM is hereafter referred to as VM-hh. VM-hh RNA expression, which is initially detected at mid-stage 11,diminished at stage 12, but was weakly detectable until stage 15. Sph-hh-driven lacZ (VM-hh-lacZ) signals were detected up to stage 16. ptc, a general target gene of HH signaling(reviewed by Alcedo and Noll,1997), was expressed in or around hh-expressing VM cells at mid stage 12 (Fig. 1C),suggesting autocrine and paracrine functions of VM-HH. Even-numbered VM-PSs in early mesoderm is marked by ftz-lacZ expression(Fig. 1D2)(Tremml and Bienz, 1989).hh RNA staining of late stage 11 ftz-lacZ embryos disclosed VM-hh expression in the anterior terminal region of each VM-PS(Fig. 1D); VM-hhexpression was absent from VM-PS2.

Besides hh and ptc, four other genes appear to belong to the VM metameric gene family (Fig. 2A-D). Metameric CON expression occurs in VM between mid-stage 11 and stage 15 (Bilder and Scott,1998). Staining for CON and VM-hh-lacZ(Fig. 2E) demonstrated that CON is expressed in cells situated on the anterior side of each VM-PS. Thecon expression domain was broader than that of hh. conexpression was somewhat weak near the VM-PS border.

Metameric RNA expression of bnl, which encodes a ligand for Breathless FGF receptor (Sutherland et al., 1996), was first observed as 12 patches at mid stage 11(Fig. 2A). bnl RNA expression became homogeneous and then diminished during stage 12 (data not shown). tin is a homeobox gene that is required for dorsal mesodermal development (Azpiazu and Frasch,1993; Bodmer,1993). At early stage 10, tin is expressed throughout the dorsal mesoderm from which VM is derived. Metameric TIN expression became evident by early stage 11 (Azpiazu and Frasch, 1993) (Fig. 2B, Fig. 3L). TIN expression decreased during stage 12. Expression of bap, another homeobox gene required for VM development, can be monitored by bap 4.5#23 (bap-lacZ) (Fig. 2C). Staining for TIN and bap-lacZ or bnl RNA(Fig. 2F,G) indicated thattin, bap and bnl were co-expressed in VM-PS3-12 during stage 11; in VM-PS2, only bnl was expressed. Stage 11-12 VM was also stained for TIN and VM-hh-lacZ(Fig. 2H). TIN and VM-hh-lacZ expression partially overlapped. VM-hh expression in the anterior terminal region of VM-PSs (seeFig. 1D) indicated that eachtin/bnl/bap trio expression domain straddles the VM-PS boundary (seeFig. 2I).

In summary, VM-PSs in thorax and abdomen, respectively, was found to be subdivided into five or six regions with respect to differential expression of VM-metameric genes at stages 11-12 (Fig. 2I).

VM-PS cell fates may be governed by ectodermal HH and WG signals(Bilder and Scott, 1998). Thus,we examined the effects of changing hh or wg activity on VM-PS gene expression at late stage 11 to early stage 12. For misexpression ofhh or wg, 24B-GAL4 or twi-GAL4 were used as mesodermal-GAL4 drivers (Brand and Perrimon, 1993; Greig and Akam, 1993).

As shown in Fig. 3A-C,VM-hh expression was abolished in both hh^13C andwg^CX4 embryos, while VM-hh expression was expanded throughout VM in response to misexpression of wg(Fig. 3D) and some expansion was noted due to hh misexpression(Fig. 3N,O). As hh andwg expression is mutually regulated in early embryogenesis(Ingham, 1993), the above finding may indicate that either hh or wg, or both, is required for VM-hh expression. Should WG be a primary effector but HH not, VM-hh-lacZ misexpression may be expected to occur in hhmutants that misexpress wg. To provide clarification of this point,hh and wg were misexpressed on wg and hhmutant backgrounds, respectively. No positive VM-hh signals could be found in either case (Fig. 3E,F). Previous studies suggested that hh mutants with ectopic wg fail to form VM due to earlier patterning functions of these genes (Azpiazu et al.,1996). However, we found that VM, incompletely expressing TIN, is formed under our experimental conditions(Fig. 3G). It may thus follow that the concerted action of WG and HH, both of which serve as primary positive regulators, is required for VM-hh expression.

The effects of change in hh and wg activity on tin,bnl and bap expression in VM were very similar, if not identical, to each other. The expression of tin, bnl and bapat late stage 11 was significantly reduced or hardly present in the absence ofwg activity (Fig. 3H,I). The expression of these VM metameric genes expanded into nearly all VM cells on misexpression of wg(Fig. 3J), indicating that WG serves as a positive regulator of tin/bnl/bap expression in VM. HH may have little or no role in tin/bnl/bap expression in stage 11 VM. Neither appreciable expansion nor reduction in tin/bnl/bap domains could be observed with mesodermal hh misexpression (compareFig. 3M with 3L). Inhh^13C embryos, tin/bnl/bap expression appeared to have expanded in VM (Fig. 3K). But in hh^13C embryos, the number of FAS3-positive VM-PS cells was reduced by about a half (see below) and the number oftin/bnl/bap-positive cells per VM-PS was virtually the same as that in wild type (data not shown). The absence of tin/bnl/bap-negative cells in hh^13C embryos may thus be a reflection of partial failure of VM cell formation in hh mutants.

It is evident from the above that wg is required fortin/bnl/bap trio expression. Why was TIN was detected inhh^13C embryos, which possibly lack wg expression owing to mutual regulation between wg and hh? Inhh^13C embryos, only dorsalmost WG, which is required fortin/bnl/bap expression, was still found to be expressed inhh^13C embryos at stages 10-11, even with virtual elimination of other wg signals (data not shown).

The fact that hh misexpression does not change tinexpression in VM allowed us to examine whether VM-hh expression occurred anteriorly or posteriorly subsequent to hh misexpression. VM cells misexpressing hh were stained for TIN and VM-hh-lacZ. As shown in Fig. 3N,O,hh misexpression brought about the anterior expansion of VM-hh.

To determine when ectodermal HH and WG required for VM-metameric gene expression is produced, hh and wg activity was transiently altered using temperature sensitive hh (hh^9K/13C;hh^ts) and wg (wg^IL4;wg^ts) mutants. hh activity was eliminated by shifting-up hh^ts embryos from permissive (18°C) to non-permissive (29°C) temperatures at stage 10 to early stage 11. As shown previously for con expression(Bilder and Scott, 1998),metameric VM-hh (VM-hh-lacZ) expression, which normally becomes discernable at mid- to late stage 11, disappeared(Fig. 4A,B). By contrast,con and VM-hh expression was normal after temperature shift-up at mid stage 11 or later (Fig. 4A,C). With temperature shifted down from non-permissive to permissive temperatures at the end of stage 9 to early stage 10, expression patterns of VM-hh and all other VM genes examined here were very similar or even identical to those in wild type(Fig. 4A,D), although progenitor cell number per VM-PS unit (VM-PS cell number) was reduced to 60-70% that of wild type (data not shown). HH produced during stage 10 to early stage 11 should thus be considered essential for initiation of VM-hh as well as con expression. HH secreted before the end of stage 9 or after mid-stage 11 onwards should be dispensable at least for initiation of VM-hh expression. Similar results have been reported in the case of con expression by Bilder and Scott(Bilder and Scott, 1998). In contrast to VM-hh and con expression, the expression oftin and bnl and bap(Fig. 4A,E and data not shown)was normal with or without hh activity; this again demonstrated thathh is not involved in bnl, tin or bapexpression.

As in the case of HH, the shift-up study indicated that WG produced during stage 10 was required for hh and tin expression in VM(Fig. 4F,G,N) but not expression of con (Fig. 4F,H). Nevertheless, WG produced at early stage 11 is essential for tin expression but dispensable for that of VM-hh(Fig. 4F,I). WG produced at mid stage 11 or later was apparently unnecessary for either(Fig. 4F). Shift-down experiments indicated that WG produced at and before the end of stage 9 to early stage 10 is dispensable for VM-hh and tin expression(Fig. 4F,J) and it follows that WG produced during stage 10 to early stage 11 regulates the expression ofhh and tin expression.

hh and wg, expressed in ectoderm from stage 5 onwards(Tabata et al., 1992;Baylies et al., 1995), is not detectable in mesoderm by stages 9-10 (data not shown). Mesodermal hhwas again seen at mid-stage 11 (see Fig. 1A1,B), while that of wg is reported to start at mid-stage 11 only in VM-PS8 (Immergluck et al., 1990). Our results would thus mean that HH and WG emanating from the ectoderm during stage 10 to early stage 11 regulate the expression of VM-metameric genes.

Each tin/bnl/bap domain consists of a posterior terminal region of one VM-PS and anterior terminal region of its posterior neighbor (pand a regions, respectively; Fig. 2I, Fig. 4K). The latter is always close to the ectodermal AP border during early stage 10 to mid-stage 11, while the former is so only at late stage 10 to mid-stage 11(Borkowski et al., 1995) (see also Fig. 8). Consistent with this, examination of wg^ts embryos shifted-up at late stage 10 to early stage 11 indicated that an appreciable fraction of embryos possess VM with TIN and bnl signals only in the a region of VM-PSs(Fig. 4L,M). While notin expression occurred in embryos shifted-up from early stage 10 onwards (Fig. 4N), TIN signals were detected only in a few p-region cells near the VM-PS border in a fraction of embryos shifted-up at late stage 10 to early stage 11 and shifted-down at mid-stage 11 (see Fig. 7F1). In addition, in contrast to an earlier report(Azpiazu and Frasch, 1993), we found that, in wild-type VM, tin RNA signals change much more dynamically than do tin protein signals at late stage 10 to early stage 11 (compare Fig. 4O1 with 4O2,3). In slightly younger embryos, tin RNA signals were detected only in the a region of each VM-PS(Fig. 4O2), while, in older ones, tin RNA signals were detected in both a and pregions (Fig. 4O3). No corresponding protein signal change could be detected possibly because TIN is much more stable than tin RNA(Fig. 4O1 and data not shown). The tin/bnl/bap expression within each VM-PS is thus reasonably concluded under the control of WG signals emanating from two distinct sources at different times (see Fig. 8).

FAS3 is expressed throughout VM from mid stage 11 onwards(Fig. 3P). Thus, using anti-FAS3 antibody staining, VM-PS cell number in mid stage 11 embryos differing in HH and WG signaling activity was examined. Inhh^13C and wg^CX4 embryos, this cell number was reduced to about a half and two thirds that of wild type (17 cells), respectively (Fig. 3P-R), but was not affected by mesodermal overexpression ofhh or wg or transient loss of hh and wgactivity at stage 10 (Fig. 3Sand data not shown). Variation in hh or wg activity at stage 10 thus has no significant effect on VM-PS cell formation.

At mid-stage 11 and later, VM cells are situated far away from ectoderm and, consequently, HH signals possibly required for VM development from mid-stage 11 onwards may not come from hh ectodermal domains. The activity of VM-hh, whose expression starts at mid stage 11, was eliminated by shifting-up hh^ts embryos from permissive to non-permissive temperatures. A 90-minute elimination of hh activity during mid-stage 11 to mid-stage 12 resulted in loss of both VM-hhRNA and CON signals at stage 13 (Fig. 5A-D) but not those at stage 12. VM-hh RNA expression was also abolished by a 90-minute shifting-up from mid- to late stage 12 (data not shown). The VM expression of ptc, a general target gene of HH signaling, was also abolished upon elimination of hh activity from mid stage 11 onwards (data not shown). It may thus follow that VM-hh,the expression initiation of which is under the control of ectodermal HH, is essential for the maintenance of its own expression and that of ptc,and stage 13 con expression in VM.

Cubitus interruptus (Ci) is a downstream transcription factor of HH signaling and ectopic expression of the Ci repressor form in the posterior compartment of imaginal discs results in elimination of hh expression(Methot and Basler, 1999). Similarly, VM-hh was abolished by twi-GAL4-driven UAS-ciNZn encoding the Ci repressor(Hepker et al., 1997) (data not shown). 48Y-GAL4 is a GAL4 driver originally thought to be specific for endodermal expression (Martin-Bermudo et al., 1997) but capable of driving target genes in VM after mid stage 11 (data not shown). When UAS-ciNZn was driven by 48Y-GAL4,maintenance but not initiation of VM-hh expression was impaired(Fig. 5E,F). Thus, both initiation and maintenance of VM-hh appear to be negatively regulated by Ci repressor misexpression.

As with dpp and wg expressed in VM-PS7 and 8,respectively (Szuts et al.,1998), VM-hh may be required for endodermal cell differentiation. To clarify this point, VM-hh activity was abolished and possible morphological changes of VM derivatives were observed at 25 hours AEL. hh^ts embryos were shifted up from permissive to non-permissive temperatures for 3 hours during mid-stage 11 to late stage 12. The majority of embryos were associated with the short gastric caecum phenotype (Fig. 6A,D), while,in about 10% of the total (n=57), one or a few gastric caeca were lost and the size of those remaining was considerably reduced(Fig. 6F). A similar short gastric caecum phenotype was also observed in virtually all embryos with UAS-ciNZn driven by 48Y-GAL4 (Fig. 6G). As described above, in these embryos, VM-hh RNA expression is normally initiated but not maintained, while ectodermalhh expression appears normal, indicating that they possess normal ectodermal HH activity but lack VM-HH activity during most of late developmental stages. VM-hh thus appears essential for normal gastric caecum formation.

Gastric caeca are derivatives of precursors consisting of VM-PS3 cells and immediately adjacent endodermal cells(Reuter and Scott, 1990;Bate, 1993;Szuts et al., 1998) and all four gastric caecum invaginations become apparent by stage 17 (16 hours AEL)(Reuter and Scott, 1990). While a considerable fraction of gastric caecum precursors expressDWnt-4 (Graba et al.,1995), anterior VM-PS3 cells are positive for VM-hh(Fig. 6C). Because, in VM, PTC signal, which serves as a marker for the area with high levels of HH reception, was restricted to cells expressing VM-hh and their vicinity (see Fig. 1C), we believe that VM-PS3 must provide HH required for normal gastric caecum formation.

It should, however, be noted that the above conclusion does not necessarily exclude the possibility that ectodermal hh may have some essential roles in gastric caecum formation at earlier stages. Indeed, gastric caecum phenotypes more severe than those described above were found inhh^ts embryos shifted-up during stage 10-11, and in those with UAS-ciNZn driven by 24B-GAL4 or bap-GAL4(Zaffran et al., 2001)(Fig. 6I and data not shown). Gastric caecum was virtually completely abolished in hh^13Cand twi-GAL4-driven UAS-ciNZn embryos(Fig. 6J,L). We interpret these findings as suggesting that there are multiple steps requiring hhactivity for normal gastric caecum formation.

Short gastric caecum phenotypes have also been reported in dpphypomorphic mutants and vein (vn) mutants(Bate, 1993;Szuts et al., 1998).dpp is expressed in VM-PS3 from stage 11 onwards and its expression area virtually overlaps that of VM-hh-lacZ expression(Bate, 1993)(Fig. 6B), while VM-vn, whose expression is maintained by VM-PS3 dpp(Szuts et al., 1998), is restricted to VM-PS2-4 from stage 13 onwards(Szuts et al., 1998). The complete absence of dpp activity from VM-PS3 has been reported to induce the gastric caecum-less phenotype(Hursh et al., 1993;Masucci and Hoffmann, 1993). Thus, examination was made of whether dpp expression is reduced or abolished in hh^ts embryos shifted up during midstage 11 to late stage 12 and 48Y-Gal4-driven UAS-ciNZn embryos, along withhh^13C and twi-GAL4-driven UAS-ciNZnembryos. dpp expression was observed at stage 13. As shown inFig. 6B,E,H,K, the dppRNA expression levels in these HH signaling mutants were roughly proportional to the gastric caecum length. That is, in the hh^ts embryos shifted up during mid-stage 11 to late stage 12 and 48Y-Gal4-driven UAS-ciNZn embryos, both exhibiting the short gastric caecum phenotype(see Fig. 6D,G), VM-PS3dpp expression was significantly reduced(Fig. 6E,H), whereas, inhh^13C embryos and twi-GAL4-driven UAS-ciNZn embryos, both of which completely lack the gastric caecum(see Fig. 6J,L), nodpp signals were detected in VM-PS3(Fig. 6K and data not shown). Moreover, vn expression in VM-PS2-4 was significantly reduced or virtually completely abolished in these HH signaling mutants (data not shown). It may thus follow that the concerted action of HH secreted at stages prior to stage 11 and HH from VM-PS3 at later stages establishes the normal expression level of dpp in VM-PS3, and most, if not all, gastric caecum defects of these HH signaling mutants are attributed to reduction or loss ofdpp expression in VM-PS3.

Tracheal cells are formed from 10 clusters (Tr1-10) of ectodermal cells on each side of the embryo. Future tracheal cell clusters invaginate, and at late stage 11, migrate to produce six primary branches, one being VB(Manning and Krasnow, 1993;Samakovlis et al., 1996). The ordered development of trachea is a consequence of communication between tracheal cells and surrounding tissues (reviewed byZelzer and Shilo, 2000).bnl acts at least as a motogen for all primary branches(Sutherland et al., 1996). In parallel to and/or prior to bnl signaling, DPP, EGF, WG and HH are required for proper invagination of tracheal cells and guidance of specific branches (Glazer and Shilo,2001). Integrin proteins have been recently shown involved in VB guidance after VB reaches VM (Boube et al., 2001). Mesodermal cells may also be required for assisting or constraining tracheal branching(Franch-Marro and Casanova,2000; Wolf and Schuh,2000). No branch formation occurs in bnl mutants and ectopic bnl expression redirects some branches(Sutherland et al., 1996), but whether restricted bnl expression in VM is essential for VB guidance remains unclear.

VB cells, though not those in Tr3 or Tr9(Manning and Krasnow, 1993;Sutherland et al., 1996),begin budding towards corresponding bnl expression domains in a 1:1 fashion at late stage 11 (Fig. 7A). The tip of VB of Tr-(i-3) (i; integer)first appeared to touch the vicinity of the posterior end of the aregion or bnl/TIN-expressing anterior region of VM-PS-i(Fig. 7B) and then, VB/VM contact expanded to the entire VM-PS-(i-1) p-region and VM-PS-(i) a-region (Fig. 7C). During stage 12, with bnl expression becoming rather homogeneous throughout VM (data not shown), VB extends further anteriorly. This extension depends on integrin genes(Boube et al., 2001).

The expression of bnl was changed to examine how this would affect initial VB budding at stage 11 to early stage 12, which is independent of integrin genes (Boube et al.,2001). First, bnl was misexpressed in the entire VM withbap-GAL4. Should VM-BNL serve as a chemoattractant for VB guidance,considerable VB misrouting would be expected to occur upon bnlmisexpression. We found 48 VB bifurcations (23%) and five simple VB misroutings (2.4%) in 210 VM-PSs normally capable of coming into contact with VB (Fig. 7D,E). These defects appear to be persisted at least until stage 15.Fig. 7D,E also shows that no appreciable change in tin expression was induced by bnlmisexpression in VM. Thus, VB misrouting and bifurcation appear solely due tobnl misexpressed in VM.

The bnl/tin expression domain in VM-PSs was reduced in area in two ways, both dependent on WG being a positive regulator of metameric expression of bnl and tin in VM. As TIN and bnl are co-expressed in both wild-type and wg hypomorphic mutant conditions,TIN was used as an area marker for bnl expression.dTCF-ΔN is a dominant negative form of d-TCF and serves as an antagonist of WG signaling(Cavallo et al., 1998).twi-GAL4-driven UAS-dTCF-ΔN, which blocks WG signaling in a mesoderm-specific manner, restricted TIN/bnl expression only to cells in the vicinity of the VM-PS border(Fig. 7H,I). In these mutant flies, the remnant TIN/bnl-expressing p/a-border-cells served as targets for the initial VB contact(Fig. 7H); unlike the situation in wild type, no initial VB/VM contact was detected at the vicinity of the putative posterior end of the VM-PS a region(Fig. 7I).

A similar redirection of VB migration was also observed inwg^ts mutants shifted-up for 60 minutes from late stage 10 to stage 11. In this case, the remnant TIN/bnl-expressingp-region cells near the VM-PS border served as targets for the initial VB contact (Fig. 7F). Subsequently, VM/VB contact expanded, as in wild type(Fig. 7G). Based on these results, we conclude that metameric BNL within VM serves as a chemoattractant for initial VB migration.

Previous analysis of con expression by Bilder and Scott(Bilder and Scott, 1998)indicated that VM-PSs are subdivided by the concerted action of hhand wg into at least two domains. We now extended this finding through use of a greater number of trunk visceral mesodermal genes, includinghh and bnl. Detailed analysis of VM-hh, bnl, tinand bap expression in addition to con indicated that trunk visceral mesodermal genes are classified into three distinct groups —tin/bnl/bap, VM-hh and con — and each VM-PS is subdivided into five or six regions, becoming apparent during mid stage 11 to stage 12 (see Fig. 2I).

VM is presently considered to develop in two steps under the control of ectodermal HH and WG signals. First, by stage 10 (when four mesodermal primordia have become specified), VM competent or bap expression regions are promoted by hh but repressed by wg, via a direct targetor, slp (Lee and Frasch,2000). The second surge of hh and wg activity at stages 10-11 is responsible for subdividing VM-PSs into two regions:con positive and negative (Bilder and Scott, 1998). Our results indicate that the expression of other four VM-metameric genes, hh, tin, bnl and bap, is also regulated by the second surge of hh and wg activity at stages 10-11.

To examine the regulation of VM-metameric genes with changing the activity of hh and/or wg, it may be necessary to evaluate the effects of possible change in cell number on VM-PS subdivision or VM-PS cell specification. In temperature-sensitive mutants of hh and wgshifted-up from stage 10, the number of VM cells positive to FAS3 at mid stage 11 on was essentially identical to that of wild type, indicating that VM-cell number change is negligible under the conditions used, while the expression of some VM metameric genes appeared compromised. In hh^tsmutants, VM-hh and con were not expressed, though tin,bnl and bap were (see Fig. 4A). In wg^ts mutants, VM-hh andtin were not expressed, but con was (seeFig. 4F). All these observations are totally in agreement with those found in simple loss-of-function and overexpression experiments, as described in the Results;under these conditions, the formation of a VM competent region should be hindered. Thus, our results may indicate that ectodermal HH and WG regulate directly, but in different ways, the expression of metameric genes in VM;VM-hh expression requires both HH and WG. tin, bnl andbap are positively regulated by WG alone, and con is activated by HH and repressed by WG.

In view of morphological changes in a VM competent region and consideration of our findings on VM gene regulation, the following model for VM-PS cell specification may be proposed (Fig. 8).

At stage 10 to early stage 11, anterior terminal cells of VM-PSs are presumed to be always situated near an ectodermal AP border, where they are capable of continuously receiving WG and HH signals, and WG confers competence on these cells to express tin/bnl/bap. WG and HH are responsible for inducing VM-hh, and HH, for con expression. In the anterior-most cells, con expression is reduced, which would be expected in view of repression by high WG signal. The different thresholds ofhh for con and VM-hh expression may explain why thecon area expands more posteriorly compared with that of VM-hh. Posterior terminal VM cells, when formed, are situated far from WG expressed on the ectodermal PS border (see `early to late stage 10' inFig. 8). But as they migrate posteriorly and close to the posteriorly neighboring AP border by early stage 11, they become capable of receiving WG and acquire competence to expresstin/bnl/bap (see `late stage 10 to midstage 11' inFig. 8). Thus, thetin/bnl/bap domain would appear regulated by spatially and temporally distinct WG signals. The two-step induction of tin/bnl/bap expression is supported by experiments using the wg^ts mutant, where,either posterior or anterior expression within one patch could be differentially turned off (Fig. 4L-N). Indeed, we observed a stepwise activation oftin/bnl expression in VM-PSs around stage 11(Fig. 4O). As schematically shown in Fig. 4A, tinand bnl metameric expression became apparent almost simultaneously at mid-stage 11, and our preliminary experiments showed that neither tinnor bnl misexpression could induce the ectopic expression of any other metameric genes examined here. Thus, tin and bnlexpression might be initiated in a mutually independent manner.

This VM-PS subdivision model should be modified when applied to thoracic segments, where hh may not be the sole determinant of conexpression (Zaffran et al.,2001) (see Figs 2,5; data not shown).

Our study strongly suggested that metameric VM-hh is required for the maintenance of its own as well as metameric con expression (seeFig. 5C,D), although the latter becomes independent of VM-hh at late stages (C. H., unpublished). That PTC, a direct target of hh, is upregulated in each VM-hh expression domain at stage 12, at that time VM is far away from the epidermis or ectodermal HH sources (seeFig. 1C), is another evidence supporting the notion that hh signaling caused by metameric VM-hh is operative in VM.

Our results (Fig. 6) also showed that HH is required for gastric caecum development. HH may operate in multiple steps in mesoderm and its source for the last step is VM-HH emanating from VM-PS3, a part of the future gastric caecum region.Fig. 6 also indicated that most, if not all, gastric caecum defects found in HH signaling mutants may be due to the reduction or loss of VM-PS3 dpp, whose production is under the control of VM-PS3 HH and HH at stages prior to stage 11 (this work)(Hursh et al., 1993;Masucci and Hoffmann, 1993).vn expression, which is positively controlled by VM-PS3 dpp(Szuts et al., 1998), was also significantly reduced in HH signaling mutants (data not shown), whileDwnt4 expression was not seriously affected even in hh null mutants (C. H., unpublished). Thus, the effect of hh activity loss on gastric caecum formation may be due to partial changes in fate/transcription programs of VM-PS3 cell precursors.

Reiterative bnl expression in VM is likely to be a determinant of the particular mode of VB migration. The tip of VB first came in touch with the vicinity of the posterior end of the tin/bnl/bap expressinga region, where all the five metameric genes examined are expressed(Fig. 7A,B). BNL misexpression with VM-specific-GAL4 drivers induced VB misrouting and bifurcation(Fig. 7D,E) but neitherhh misexpression nor transient loss of HH activity during stage 11 had any effect on VB budding (C. H., unpublished). BNL misexpression brought about no significant change in expression of tin(Fig. 7D,E), while restriction of the tin/bnl/bap expression domain using eitherdTCF-ΔN or wg^ts caused a shift in the first VB/VM contact point (Fig. 7I). Furthermore, under a wg mutant condition, no change could be detected in VM-hh or in con expression (seeFig. 4F). Thus, only thebnl expression appears to be closely correlated with VB budding,strongly suggesting that BNL serves as a chemoattractant for initial VB migration.

We thank Drs R. White, M. Frasch, C. S. Goodman, M. Krasnow, T. Tabata, I. Guerrero, J. Pradel, T. Orenic, T. Volk, the DSHB, and the Bloomington Stock Center for antibodies, cDNA and fly strains. We thank all colleagues in the laboratory, especially T. Kojima. This study was supported in part by grants from the Ministry of Education, Culture, Sport, Science and Technology of Japan to K. S.