Three Drosophila EXT genes shape morphogen gradients through synthesis of heparan sulfate proteoglycans

Morphogens emanate from discrete sources and generate concentration gradients that elicit concentration-dependent responses in target cells(Wolpert, 1969). Secreted signaling molecules such as Hedgehog (Hh), Decapentaplegic (Dpp) and Wingless(Wg), have been identified as morphogens and their mode of action has been well-documented in Drosophila wing development (reviewed by Gurdon et al., 1998; Lawrence and Struhl, 1996; Neumann and Cohen, 1997a; Tabata, 2001).

Although the influence of Dpp and Wg extends to the edges of the wing disc(Capdevila and Guerrero, 1994; Neumann and Cohen, 1997b; Zecca et al., 1995) and Hh can signal across long distances as well (Chen and Struhl, 1996), these proteins share an unexpected characteristic; they are not readily soluble. The active form of Hh has cholesterol covalently bound at its C-terminus(Porter et al., 1996) and an N terminus that is palmitoylated (Pepinsky et al., 1998). Both modifications will probably anchor these proteins in the membranes of the cells in which they are produced. Both Wg and Dpp homolog TGF-β bind to matrix proteins(Reichsman et al., 1996; Taipale and Keski-Oja, 1997). In addition, Wg can also undergo lipid modification(Willert et al., 2003). Simple diffusion thus appears inadequate to distribute these proteins over even short distances.

Recently, several reports suggest that heparan sulfate proteoglycans(HSPGs) play a key role in morphogen transport and/or signaling (reviewed by Perrimon and Bernfield, 2000). HSPGs are abundant cell surface molecules and are part of the extracellular matrix. HSPGs consist of a protein core (such as syndecans and glypicans) to which heparan sulfate glycosaminoglycan (HS GAG) chains are attached. GAG chains are long unbranched polymers consisting of many sulfated disaccharides.(A GAG synthesis diagram is shown in Fig. 1G).

Genetic screens for mutations affecting morphogen signaling pathways in Drosophila have heretofore identified three genes that have sequence homologies with genes encoding vertebrate GAG biosynthetic enzymes. These are sugarless (sgl) (Binari et al., 1997; Hacker et al.,1997; Haerry et al.,1997), sulfateless (sfl)(Lin et al., 1999), and tout-velu (ttv)(Bellaiche et al., 1998; The et al., 1999), which encode proteins with homology to UDP-glucose dehydrogenase,N-deacetylase/N-sulfotransferase, and HS copolymerase with GlcAT-II and GlcNAcT-II activities, respectively. sgl mutations compromise signaling mediated by Wg (Binari et al.,1997; Hacker et al.,1997; Haerry et al.,1997) and Dpp (Haerry et al.,1997). The sfl mutation affects Wg and Hh signaling(Lin et al., 1999; The et al., 1999), and in somatic sfl mutant clones, Wg protein levels are diminished(Baeg et al., 2001). The ttv mutant was reported to selectively affect Hh signaling(Bellaiche et al., 1998; The et al., 1999). In addition to these three genes, notum, a gene that encodes a member of theα/β-hydrolase superfamily, has been reported to influence Wg protein distribution by destabilizing the HSPGs(Gerlitz and Basler, 2002; Giraldez et al., 2002). Finally, dally, which is proposed to encode a protein core of the HSPGs, has been shown to be required for Wg and Dpp activity(Jackson et al., 1997; Lin and Perrimon, 1999; Nakato et al., 1995; Tsuda et al., 1999; Fujise et al., 2003). Dally and related Dally-like (Dlp) have been shown to bind and stabilize Wg at the cell surface (Baeg et al.,2001; Strigini and Cohen,2000) and may provide a pool of Wg protein that can become available for receptor binding upon release from the HSPG.

Among the three GAG-biosynthetic genes, the ttv mutant is unique in its selective effects on Hh signaling. Wg-directed events were unaffected in ttv embryos and larvae(Bellaiche et al., 1998; The et al., 1999). A possible explanation accounting for this apparent selectivity is that HS GAGs synthesized by other EXT genes may be sufficient to allow Wg pathways to function (`quantitative model'). Alternatively, Hh-specific HSPGs may exist and ttv might be required for their synthesis, whereas other EXT genes might synthesize HSPGs specific for Wg signaling (`qualitative model').

Here we describe novel mutants of sister of tout-velu(sotv) and brother of tout-velu (botv), as well as new alleles of ttv. sotv and botv, like ttv, encode EXT family proteins. In the clones of mutant cells, HSPGs biosynthesis was severely affected, and the morphogen signaling activities of Hh, Dpp and Wg were impaired to various degrees. Morphogen molecules were rarely seen in mutant clones, and had been accumulated at the wild-type cells that reside next to a mutant clone. These results suggest that HSPGs are required for proper transport of morphogens.

y w; smo stc FRT^42Dy⁺sha/CyO males were mutated with ethylmethanesulfonate (EMS), crossed first to y w; Pin/CyO and then individually to y w FLP122;FRT^42Dy⁺. Males with phenotypes in wing development were retained and maintained as balanced stocks. The mutants and transgenic animals used were as follows: ttv⁵²⁴,ttv²⁰⁵, sotv³²⁶, botv⁴²³,botv⁵¹⁰ and botv⁵¹⁴ (isolated in this study); ttv^l(2)00681(Bellaiche et al., 1998);UAS-Dpp-GFP, a biologically active form of Dpp tagged with GFP(Entchev et al., 2000); dpp-GAL4 (Teleman and Cohen,2000); and UAS-sotv (made by injection of pUAST construct containing NotI-Asp718 fragment of the sotv cDNA, this study).

One of three complementation groups that had similar phenotypes, failed to complement the lethality of ttv, which has been reported to be a Drosophila EXT family gene(Bellaiche et al., 1998) and two other groups failed to complement Df(2R)Jp8 and Df(2R)P34 which delete 52F5-9; 53A1 and 55E6-F3; 56C1, respectively. In and around these regions there are two other Drosophila EXT genes, DEXT2(Han et al., 2002; The et al., 1999) and DEXT3 (Han et al.,2002; Han et al.,2001; Kim et al.,2002). We sequenced all predicted coding regions of their genomic DNA and identified SNPs in DEXT2 and DEXT3.

Clones of mutant cells were induced by Flippase-mediated mitotic recombination (Xu and Rubin,1993). First instar larvae were heat-shocked at 37°C for 1 hour and dissected 2-4 days after the heat-shock. Mutant clones were identified by the loss of GFP expression. Genotypes of the flies used for making mitotic clones were as follows: y w FLP122/+;FRT^42Dy⁺sha EXT gene(s)/FRT^42Dy⁺ Ub-GFP or wglacZ FRT^42D Ub-GFP or dpp-lacZFRT^42D Ub-GFP. For monitoring the Dpp-GFP distribution in the EXT mutant clone, we dissected wing discs of y FLP122/+;FRT^42Dttv⁵²⁴botv⁵¹⁰/FRT^42D hs-CD2; UAS-dpp-GFP/dpp-GAL4 larvae.

Imaginal disc staining was performed as described previously(Tanimoto et al., 2000),except for 3G10 staining: treatment of imaginal discs to expose the epitope recognized by the 3G10 monoclonal antibody (Seikagaku) was done by incubating fixed imaginal discs with 20 mU heparitinase I (Seikagaku) per 1 ml 100 mM NaOAc, 3.3 mM CaCl₂ for 1 hour at 37°C. After washing three times with PBT, the samples were stained with 3G10 antibody, 1:100 in 2%BSA/PBT.

The primary antibodies used were: rabbit anti-β-gal antibody, 1:2000(Cappel); mouse anti-CD2 antibody, 1:1000 (Cederlane); rat anti-Ptc antibody,1:150 (gift from R. Johnson); rat anti-Sal antibody, 1:250 (gift from R. Barrio); rabbit anti-p-Mad antibody PS1, 1:20,000(Persson et al., 1998); rabbit anti-Dll antibody, 1:300 (Panganiban et al., 1995); mouse anti-Wg antibody 4D4, 1:5000 (Developmental Studies Hybridoma Bank); rabbit anti-Hh antibody NHhI, 1:1000 (raised against the N-terminal amino acids 89-306 of Hh); rabbit anti-Dpp antibody, 1:5000(gift from M. Hoffmann).

Secondary antibodies used were anti-mouse Cy3, anti-rabbit Cy3, anti-Rat Cy3, anti-rabbit Cy5 and anti-rat Cy5 (Jackson).

Immunofluorescent images were observed on a Zeiss confocal laser microscope 510.

We performed a screen for mutations that affect wing patterning. Clones of cells homozygous for mutagenized chromosomes were generated, and adult wings were examined for patterning defects. In the wing blade, the area between longitudinal veins 3 and 4 is patterned by Hh signaling(Mullor et al., 1997; Strigini and Cohen, 1997) and the rest of the area is patterned by Dpp signaling(Bier, 2000; Sturtevant et al., 1997). We identified three independent loci on the right arm of chromosome 2 that delete tissue within the clone (Fig. 1B-E). This tissue-deletion phenotype was seen in all areas of the wing blade. These mutants were classified into three complementation groups. Two mutants of the first complementation group, ttv⁵²⁴ and ttv²⁰⁵, failed to complement ttv(Bellaiche et al., 1998) and are presumed to represent ttv mutant alleles. We sequenced all predicted coding regions of the ttv gene and identified SNPs in exons of ttv (Fig. 1F). ttv⁵²⁴ mutant has a nonsense mutation at Trp 301 that truncates 460 C-terminal residues, and ttv²⁰⁵ has a missense mutation at Gly 447 Asp (Fig. 1F). ttv is the Drosophila EXT family gene with highest homology to human EXT1 (amino acid sequence identity 50.4%) which has GlcNAcT-II and GlcAT-II activities that are required for the elongation of HS chains (Fig. 1G) (reviewed by Sugahara, 2000). Drosophila ttv mutant embryos exhibit strongly impaired HS biosynthesis (The et al.,1999; Toyoda et al.,2000), but it has not been demonstrated directly whether Ttv protein has GlcNAcT-II and GlcAT-II activities.

Crosses to deficiency strains mapped the other two loci to the interval 52F5-9; 53A1 and 55E6-F3; 56C1, respectively. Near these regions, two EXT-related genes have been identified by homology search: DEXT2,sotv (CG8433, 52E11) (Han et al.,2002; The et al.,1999) and DEXT3, botv (CG15110, 56B5)(Han et al., 2002; Han et al., 2001; Kim et al., 2002). However,evidence for the function of these genes in an intact organism has been lacking because no mutant has been identified. We sequenced all predicted coding regions of these genes and identified deletions and SNPs in exons of both DEXT2 and DEXT3(Fig. 1F).

The predicted ORF of sotv encodes a protein with 717 amino acids. Our Drosophila sotv³²⁶ mutant has a nonsense mutation at glutamic acid 626 that cuts off 92 C-terminal residues(Fig. 1F). This candidate gene must represent sotv, because a transgene harboring a full-length sotv cDNA driven by the ubiquitous promoter of the tubulin alpha1 gene was found to completely rescue the lethality of sotv326/Df(2R)Jp8 and other phenotypes associated with clones of sotvmutation (data not shown). The Sotv sequence is similar to human EXT2 (44.8%amino acid similarity), which also has GlcNAcT-II and GlcAT-II activities(Senay et al., 2000).

The predicted botv encodes a protein with 972 residues and shares 45.8% sequence identity with human EXTL3. Three mutants of our third complementation group, botv⁴²³, botv⁵¹⁰ and botv⁵¹⁴ have nonsense mutation at Leu 259, missense mutation at Pro 669 Leu, and Gly 821 Asp,respectively (Fig. 1F). Human EXTL3 has GlcNAcT-I and -II activities and is required for initiation of HS biosynthesis. Previous studies have shown that a truncated soluble form of Drosophila Botv protein has GlcNAcT-I and GlcNAcT-II activities when expressed in COS-1 cells (Kim et al.,2002).

Like most EXT family members, including Ttv, Sotv and Botv have three apparent domains: an N-terminal transmembrane domain; a DXD motif which is characteristic of glycosyltransferases and is required for binding divalent cations; and a large globular domain in the C-terminal region that will probably have enzymatic activity (Fig. 1F). Amino acid sequences of these Drosophila EXTs are related, with 26-32% amino acid overall similarity, and especially high homology in their C-terminal regions (Fig. 1H).

Animals transheterozygous for each EXT mutant(ttv⁵²⁴/ttv²⁰⁵, botv⁴²³/botv⁵¹⁰, sotv³²⁶/Df(2R)Jp8)die as pupae with small eyes and legs (data not shown). In all cases, third instar larvae have smaller discs than the wild type, and in particular, their wing discs were narrower in anterior-posterior direction (data not shown). In situ hybridization revealed that third instar larvae express sotv and botv ubiquitously in wing discs as well as ttv (data not shown).

To test whether sotv and botv function in HSPG biosynthesis in flies, we stained wing imaginal discs with antibody 3G10 that recognizes an epitope produced by heparitinase I digestion of HSPGs(David et al., 1992). In the wild-type wing discs, uniform staining was detected(Fig. 2A). However in discs with mutant clones, 3G10 staining was severely reduced in cells mutant for any of the EXT genes (Fig. 2B-D). These results indicate that HSPG biosynthesis in Drosophila requires each of these genes.

This and the following phenotypes described in this report are unlikely to be caused by possible secondary lesions on the mutated chromosome because the same phenotypes are associated with all ttv and botv alleles independently isolated in this screen.

In the wing imaginal disc, Hh is only expressed in the posterior compartment cells but diffuses into the anterior compartment(Basler and Struhl, 1994; Chen and Struhl, 1996; Lee et al., 1992; Tabata and Kornberg, 1994)where it upregulates expression of dpp and patched(ptc). dpp responds to low Hh activity and is observed as far as 15 cells away from the A/P boundary; ptc responds to higher Hh-signaling activity and can be detected only up to five cells away(Fig. 3A). We used dppand Ptc expression as reporters of the Hh signaling in EXT mutants.

In botv mutant clones located in the anterior compartment near the A/P boundary, Hh signaling was partially impaired, as indicated by reduced levels of Ptc protein and dpp expression in most parts of the clone(Fig. 3B, arrow). However, Ptc remained at the posterior edge of the clone next to the A/P boundary, and dpp expression in these cells also remained and was sometimes higher than in the corresponding wild-type cells(Fig. 3B). The reduction in dpp expression extended to the wild-type cells anterior to the clone(Fig. 3B, arrowhead). Similar observations were described in ttv mutant clones(Bellaiche et al., 1998) (data not shown). In sotv clones, Ptc and dpp levels were also reduced, but the decrease was not as significant as in botv or ttv clones (arrow in Fig. 3C). We interpret this as reflecting the remaining activity of Sotv protein encoded by sotv³²⁶ which has a mutation at the very C-terminus of the protein. These observations are consistent with the reduced distance between veins L3 and L4 that is associated with small mutant clones in this region of the wing blade(Fig. 1C).

Almost no change in level of dpp expression was detected when mutant clones grew along the boundary in the posterior compartment (arrow in Fig. 3D,E). This result suggests that neither sotv nor botv is required for Hh secretion, but rather that their functions are necessary for efficient movement of Hh. However on the same occasion, Ptc protein level was sometimes slightly decreased in the anterior cells along the A/P boundary(Fig. 3D,E). It is probably because ptc expression requires higher Hh signal than dppexpression, and is more sensitive to reduction in Hh protein level in posterior compartment, as described below(Fig. 6A). The same results were observed in ttv⁵²⁴ mutant clones (data not shown).

Effects on Dpp signaling were monitored by cataloging the distribution of Dpp target gene Spalt (Sal) protein, and the phosphorylated level of the cytoplasmic signal transducer, Mothers against dpp (p-Mad)(Tanimoto et al., 2000). The Sal and p-Mad levels are normally high in the central region of the wing disc(Fig. 4A); they are regulated by Dpp which is expressed by anterior compartment cells at the A/P boundary.

As shown in Fig. 4, Sal and p-Mad levels were decreased when any of the three EXT genes were mutated(Fig. 4B-D). Sal and p-Mad levels remained at the edge of the clone nearest to the Dpp source, and moreover their levels were sometimes higher, as in the case of Hh target genes(Fig. 4B-D). It should be noted that the Dpp reporter levels were reduced even in ttv mutant clones(Fig. 4D). As mentioned above, ttv has previously been reported to be exclusively involved in Hh signaling, thus it is the first evidence that ttv is also required for other morphogen signaling. These effects were observed both in anterior and posterior compartments, indicating that the observed effect on Dpp signaling was direct and did not reflect the reduction in Hh signaling, which is only active in the anterior compartment.

A functional connection between Wg and HSPGs has been suggested by the observed strong association of Wg proteins and sulfated proteoglycans in tissue culture cells (reviewed by Lin and Perrimon, 2000; Selleck,2001). In addition, Wg signaling is affected by defective HSPG biosynthesis in sgl and sfl mutants, and in dallyand dlp mutants that alter HSPG core proteins. Wg is normally expressed at the wing margin and controls patterning along the dorsal/ventral axis by regulating target genes such as distalless (dll) and vestigial (vg) in a concentration-dependent manner(Neumann and Cohen, 1997b; Zecca et al., 1996). We investigated Wg signaling in EXT mutant cells by monitoring the distribution of these proteins.

Levels of Dll protein were slightly reduced in the clones homozygous for botv⁴²³ or ttv⁵²⁴(Fig. 5B,D). In the cells mutant for sotv³²⁶, the effect was too subtle to assess definitively, probably because the sotv³²⁶ allele is weak,as described above. The reduction in Dll protein level extended to the wild-type cells near the clone (Fig. 5C,D, arrows). Furthermore, the Dll protein levels were sometimes higher at the edge of the clone nearest to the Wg source, as in the case of Hh and Dpp target genes (Fig. 5B-D). The similar phenotype was also observed for another target,Vg in botv⁴²³ and ttv⁵²⁴ mutant clones, but too subtle to assess in sotv mutant clones (data not shown). The previous study did not detect any effects of Ttv on the Wg signaling (The et al., 1999). Because the difference could be ascribed to the alleles used, we re-examined this finding using the same allele ttv^l(2)00681 that was used in the previous study (The et al.,1999) or our weak allele ttv²⁰⁵. We found that in these mutant clones Wg signaling was also slightly decreased. Thus, these results suggest that Drosophila EXT genes contribute to Wg signaling in the wing imaginal disc.

In addition to monitoring signaling in EXT mutant cells, we used antibodies that recognize Hh, Dpp and Wg, and a GFP-tagged version of Dpp to analyze whether the levels or distribution of these morphogens had been affected. We found that levels of each of these proteins were significantly reduced in the mutant, both in the morphogen-expressing region and in the receiving region(Fig. 6A-F). For Hh, Dpp and Wg, similar results were observed in cells mutant singly for any of the EXT genes, although we only show the results with the ttv⁵²⁴botv⁵¹⁰ double mutant clones in Fig. 6A-D,F. Single mutation was not tested for the distribution of Dpp-GFP(Fig. 6E).

In the morphogen-expressing region, hh expression was not downregulated, however levels of Hh protein were significantly decreased(Fig. 6A). This may indicate that Hh protein is destabilized and/or not retained efficiently on the cell surface in the absence of HSPGs. In contrast to hh, expression of the wg and dpp and levels of Wg and Dpp were decreased in the EXT clones (Fig. 6B,C). The decrease in dpp expression is easily accountable because Hh signaling is impaired in the absence of HSPGs (Fig. 3). In contrast, the decrease in wg expression is not as readily explainable: cut and wg are both targets of Notch signaling, however the protein level of Cut was not altered in EXT clones(data not shown). This suggests that wg is also regulated by unknown mechanism dependent on HSPGs.

In the morphogen-receiving region, each of these proteins was significantly decreased in the clones of cells mutant for EXT genes(Fig. 6D-F), although a little leakage of morphogen molecules was seen even in the clones doubly mutant for ttv and botv. This suggests two possible mechanisms that do not exclude each other: in the absence of HSPGs these three morphogens are 1)destabilized and/or are not retained efficiently on the cell surface, like Hh in morphogen-expressing region, or 2) prevented from diffusing efficiently into the region consisting of EXT mutant cells. Intriguingly, close observation of the distribution of Hh strongly suggested a function for HSPGs in morphogen movement. In the wild-type discs, Hh protein synthesized in the posterior compartment appears to flow into the anterior compartment, with a moderate concentration gradient starting from the middle of the posterior compartment (Fig. 7A-C). However, Hh abnormally accumulated in the posterior compartment when the EXT mutant clone was in the anterior compartment along the A/P boundary(Fig. 7D-I). This effect was seen both in the ventral region (Fig. 7D-F) and in the dorsal region(Fig. 7G-I). This suggests that Hh failed to move into the mutant cells and as a consequence accumulated in posterior cells instead. Dpp-GFP and Wg accumulation in front of the mutant clones was also apparent, however less pronounced compared with the case of Hh(Fig. 7J-M). Therefore we conclude that the HSPG-dependent diffusion is the common mechanism for the movement of these three morphogens.

In this report we identified novel Drosophila mutants of sotv and botv, and new alleles of ttv. Based on sequence homology, we suggest that sotv is a homologue of human EXT2 and that botv is a homologue of human EXTL3. We showed that each of these three mutants affected biosynthesis of HSPGs and also impaired signaling of the Hh, Dpp and Wg morphogens. Furthermore, we found that HSPGs are required for normal movement of morphogens across the disc epithelium.

Extensive studies of sgl, sfl and ttv showed that morphogen signaling and morphogenesis were affected in these HSPG mutants(reviewed by Lin and Perrimon,2000; Selleck,2001). sgl and sfl mutants had the broadest effects on signaling, compromising Wg- and Hh-regulated processes. In contrast, ttv seemed to have affected only Hh signaling, leaving Wg-mediated patterning unaffected in both embryonic and imaginal development(Bellaiche et al., 1998; The et al., 1999). These phenotypes suggested that HSPGs can distinguish between these different morphogens.

Findings on the three EXT mutants identified in this work, however, argue for a more general role for EXT genes in cell-cell signaling. We found that three EXT family genes, including ttv, affected Hh and Dpp signaling in a similar way. The effect on Wg signaling was not so pronounced compared with that on Hh and Dpp, but the subtle effect on target gene expression and clear reduction in the protein level in the mutant clones also suggest a role for the EXT genes in Wg signaling. The explanation for this contradiction might be that, as seen in Fig. 5, Wg signaling appears to depend on the EXT activity to a lesser degree than Hh and Dpp signaling, so previous study might have missed the role of EXT genes in Wg signaling. We therefore suggest that differences in the behavior of Hh, Wg and Dpp may be more quantitative than qualitative.

Although Ttv and Sotv are expected to have similar enzymatic activities,mutants defective in either of the ttv or sotv genes had dramatically impaired HSPG biosynthesis and morphogen signaling. Therefore,despite their similarities, they cannot compensate for each other. They apparently share this property with their human homologues.

EXT1 and EXT2, the human homologues of ttv and sotv, do not compensate for each other in vitro or in vivo(McCormick et al., 2000; Senay et al., 2000). In cultured cells, GFP-tagged EXT1 and EXT2 localize predominantly to the endoplasmic reticulum when expressed independently. However, GFP-tagged EXT1 and EXT2 formed hetero-oligomeric complexes that accumulated in the Golgi when expressed together in the same cell. Moreover, the Golgi-localized EXT1-EXT2 complex has substantially higher glycosyltransferase activity than EXT1 or EXT2 alone, which suggests that this complex represents the biologically relevant form. These findings provide a rationale to explain how inherited mutations in either of the two EXT genes can cause loss of activity, resulting in hereditary multiple extosis. In Drosophila, the observed lack of complementation is also consistent with an active enzymatic complex that consists of both Ttv and Sotv.

Gradients are formed by spread of morphogen from a localized source, but whether this occurs by simple diffusion or by more elaborate mechanisms is not known. Arguments against morphogen movement by diffusion have been raised by many, including Kerszberg and Wolpert(Kerszberg and Wolpert, 1998)who proposed that morphogens use a `bucket brigade' mechanism in which receptor-bound morphogen on one cell is transferred to receptors on an adjacent cell. Alternatively, Entchev et al.(Entchev et al., 2000) proposed transport through `planar transcytosis', a process by which morphogens move by repeated cycles of endocytosis and exocytosis in the plane of an epithelium. The notion that Dpp and other morphogens, such as Wg and Hh, all pass through tissues by transcytosis or similar processes has been championed by many (e.g. Greco et al., 2001; Moline et al., 1999; Narayanan and Ramaswami, 2001; Pfeiffer and Vincent, 1999),albeit not all (McDowell et al.,2001; Strigini and Cohen,2000) investigators. The third mechanism that has been considered includes serial passage through neighboring cells through GPI-anchored proteoglycans (glypicans) (The et al.,1999). GPI-linked proteins are inserted in only the outer leaflet of the plasma membrane, and have been shown to transfer from the plasma membrane of one cell to another when these cells are in contact, by flip-flap between adjacent outer leaflets.

We still do not know by which mechanism morphogen diffusion occurs in an intact organism. Our observation that morphogen signaling and protein levels are reduced in the mutant clone, and morphogens accumulated in front of the mutant clones, led us to suppose that HSPGs create an environment that supports the efficient movement of much of the morphogen molecules as scaffolds. However, we cannot exclude the possibility that other systems such as `bucket brigade' or `planar transcytosis' would require the recognition by HSPGs on cell surfaces for cell-to-cell transfer. In addition, we also observed a little leakage of morphogen molecules and signaling in the EXT mutant clones. A possible explanation accounting for these observations might be that in our EXT mutant clones a little amount of HSPGs is still produced,and they barely contribute the morphogen diffusion. Alternatively, other transport mechanisms that do not require HSPGs may play supplementary roles. In either way, all movement of the morphogens must occur within the contiguous plane of this surface environment, because tissues such as wing imaginal discs are folded in numerous places, and no evidence for skipping of morphogens over folds of epithelia was found. More work is necessary to find out how the morphogen gradient is established and maintained in developing organisms.

We thank Tom Kornberg and Alex Kouzmenko for critical reading of the manuscript, Kazuhide Tsuneizumi and the members of T.T.'s laboratory for discussion and fly maintenance. We are also grateful to M. Gonzalez-Gaitan, S. M. Cohen and the Bloomington Stock Center for fly strains, and to R. Johnson,R. Barrio, P. ten Dijke, S. Caroll, M. Hoffmann and the Developmental Studies Hybridoma Bank for antibodies. This work was supported by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan and grants from the Mitsubishi Foundation.