Cysteine Repeat Domains and Adjacent Sequences Determine Distinct Bone Morphogenetic Protein Modulatory Activities of the Drosophila Sog Protein

Corresponding author: Ethan Bier, University of California, San Diego, California 92093-0349., bier{at}biomail.ucsd.edu (E-mail)

Communicating editor: D. WEIGEL

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The Drosophila short gastrulation gene (sog) encodes a large extracellular protein (Sog) that inhibits signaling by BMP-related ligands. Sog and its vertebrate counterpart Chordin contain four copies of a cysteine repeat (CR) motif defined by 10 cysteine residues spaced in a fixed pattern and a tryptophan residue situated between the first two cysteines. Here we present a structure-function analysis of the CR repeats in Sog, using a series of deletion and point mutation constructs, as well as constructs in which CR domains have been swapped. This analysis indicates that the CR domains are individually dispensable for Sog function but that they are not interchangeable. These studies reveal three different types of Sog activity: intact Sog, which inhibits signaling mediated by the ligand Glass bottom boat (Gbb), a more broadly active class of BMP antagonist referred to as Supersog, and a newly identified activity, which may promote rather than inhibit BMP signaling. Analysis of the activities of CR swap constructs indicates that the CR domains are required for full activity of the various forms of Sog but that the type of Sog activity is determined primarily by surrounding protein sequences. Cumulatively, our analysis suggests that CR domains interact physically with adjacent protein sequences to create forms of Sog with distinct BMP modulatory activities.

SIGNALING mediated by the bone morphogenetic protein (BMP) pathway is required for a variety of cell fate choices during development and is regulated at many steps in the signaling cascade (reviewed in MASSAGUE and CHEN 2000 ; BALEMANS and VAN HUL 2002 ; CADIGAN 2002 ). An important extracellular mechanism for regulating BMP signaling is inhibition by extracellular antagonists such as the Drosophila Short gastrulation (Sog) protein (FRANCOIS et al. 1994 ) and its vertebrate homolog Chordin (SASAI et al. 1994 ). Sog and Chordin (Chd) inhibit BMP signaling in early Drosophila and vertebrate embryos, respectively, to promote neural over non-neural development (WILSON and HEMMATI-BRIVANLOU 1995 ; BIEHS et al. 1996 ; HAMMERSCHMIDT et al. 1996 ) by a mechanism that has been highly conserved during the course of evolution (PADGETT et al. 1993 ; HOLLEY et al. 1995 ; SASAI et al. 1995 ; SCHMIDT et al. 1995 ; BIER 1997 ; HEMMATI-BRIVANLOU and MELTON 1997 ). Sog and Chd bind to BMPs and prevent these ligands from activating their receptors (PICCOLO et al. 1996 ; CHANG et al. 2001 ; ROSS et al. 2001 ; SCOTT et al. 2001 ). The most likely BMP targets of Sog inhibition are Screw (Scw; ARORA et al. 1994 ) in the early embryo (NEUL and FERGUSON 1998 ; NGUYEN et al. 1998 ) and Glass bottom boat (Gbb) in the developing wing (HAERRY et al. 1998 ; KHALSA et al. 1998 ; YU et al. 2000 ).

Sog and Chordin are large molecules that each contain four copies of a cysteine repeat (CR) domain consisting of 70 amino acids defined by a fixed spacing of 10 cysteine residues and an invariant tryptophan residue between the first two cysteines (FRANCOIS and BIER 1995 ). The most amino-terminal CR (CR1), located a short distance after an internal type II signal sequence, is separated from a cluster of the remaining three CRs (CR2–CR4) by an intervening stretch of 565 amino acids (the stem; see Fig 1 in FRANCOIS et al. 1994 ). Domains similar to the CR repeats present in Sog and Chordin are also found in other extracellular proteins such as Thrombospondin and Pro-collagen (FRANCOIS et al. 1994 ; SASAI et al. 1994 ) and Crossveinless-2 (CONLEY et al. 2000 ).

View larger version (25K): In this window In a new window Download PPT slide   Figure 1. Activities of sog mutant constructs. (Columns 1 and 2) Names and diagrams of wild-type and mutant sog constructs indicating the positions of the cysteine repeat (CR) domains (open boxes), the conserved tryptophan residues (W), and type-II signal sequence (solid boxes). (Column 3) Each of these UAS-constructs was misexpressed in the developing wing, using the MS1096-GAL4 driver, which drives ubiquitous expression of UAS-transgenes in the wing (albeit at significantly higher levels on the future dorsal surface), and resulting wing phenotypes were scored as follows: Sog, mild vein truncation, but normal patterning of the A/P axis (see Fig 3B); Sog, reduced sog-like activity; Supersog, compression of the A/P axis in which there is partial or complete fusion of L2/L3 and L4/L5, significant vein loss, and modest reduction in overall wing size (Fig 3H); –, no activity; , ectopic veins anterior to L3 and posterior to L4 (Fig 3F). (Column 4) Each UAS-construct was coexpressed with UAS-tsg in the wing, using the MS1096-GAL4 driver to determine whether it could interact with Tsg to generate a Supersog-like phenotype (Fig 4B). +, positive interaction to generate Supersog-like phenotype; –, no interaction (e.g., Tsg-like activity observed); , greatly reduced interaction; , ectopic veins; nd, not done.

Two extracellular regulators of Sog activity have been characterized. The first is a metalloprotease encoded by the tolloid (tld) gene. Null loss-of-function tld mutants exhibit defects similar to those of moderate decapentaplegic (dpp) loss-of-function mutants (WHARTON et al. 1993 ). Consistent with Tld playing a role in promoting BMP signaling, Tld can cleave Sog in vitro and inactivate it (MARQUES et al. 1997 ). The other known regulator of Sog function is encoded by the twisted gastrulation (tsg) gene (MASON et al. 1994 ; CHANG et al. 2001 ; ROSS et al. 2001 ; SCOTT et al. 2001 ). When Tsg protein is combined with Tld and Sog in vitro, a different pattern of Sog processing is observed from that with Tld alone, resulting in the production of a 42-kD N-terminal fragment of Sog (YU et al. 2000 ). Although the exact set of fragments generated in vitro appears to depend on various reaction conditions (YU et al. 2000 ; SHIMMI and O'CONNOR 2003 ), a Sog fragment nearly identical to that produced by Tld and Tsg can be generated from full-length Sog in vivo in early Drosophila embryos as judged by molecular weight and isoelectric point (YU et al. 2000 ). This 42-kD Sog fragment contains both CR1 and part of the stem between CR1 and CR2 (YU et al. 2000 ). In addition to the 42-kD Sog fragment, multiple Sog species of varying sizes are generated during pupal wing development, when Sog plays a role in the vein vs. intervein cell fate choice (YU et al. 2000 ). Sog processing is developmentally regulated and is observed only during stages when endogenous Sog is expressed in a significant fraction of cells (YU et al. 2000 ). Vertebrate counterparts of Tld = Xolloid (PICCOLO et al. 1997 ) and Tsg (OELGESCHLAGER et al. 2000 ; CHANG et al. 2001 ; ROSS et al. 2001 ; SCOTT et al. 2001 ) have also been shown to play a role in early dorsal/ventral (D/V) patterning of Xenopus embryos, and Xolloid has been shown to serve a similar function as Tld by cleaving and inactivating Chordin (PICCOLO et al. 1997 ).

Recombinant forms of Sog truncated within the stem region, referred to as Supersog, have a broadened BMP inhibitory activity in comparison to intact Sog (YU et al. 2000 ). For example, during wing development, two BMPs, Dpp and Gbb, play roles in anterior/posterior (A/P) patterning, growth, and vein formation. Dpp (PADGETT et al. 1987 ) is required for each of these developmental events (SEGAL and GELBART 1985 ), while Gbb plays a more limited role (HAERRY et al. 1998 ; KHALSA et al. 1998 ). When Sog is coexpressed with these two divergent BMPs, it can block the effect of Gbb, but not Dpp (YU et al. 2000 ). Consistent with this selectivity, loss-of-function gbb alleles result in modest vein-loss phenotypes (HAERRY et al. 1998 ; KHALSA et al. 1998 ; RAY and WHARTON 2001 ) similar to those observed upon overexpression of Sog (YU et al. 1996 , YU et al. 2000 ). Supersog, on the other hand, can block the effects of both Gbb and Dpp in coexpression experiments (YU et al. 2000 ), in line with its ability to phenocopy dpp loss-of-function mutants.

A variety of evidence suggests that Tsg is involved in generating a Supersog-like activity during development. First, coexpression of Sog and Tsg results in a Supersog-like phenotype in the wing, whereas misexpression of these genes individually results in mild opposing phenotypes (YU et al. 2000 ). Second, Sog is cleaved into Supersog-like peptides in vitro by a combination of Sog, Tsg, and Tld (YU et al. 2000 ; SHIMMI and O'CONNOR 2003 ). Third, Sog is processed in vivo at developmentally appropriate stages into Supersog-like fragments, which are virtually indistinguishable from those generated in vitro (YU et al. 2000 ). Finally and critically, expression of Supersog in early tsg mutant embryos can partially rescue the tsg phenotype (YU et al. 2000 ). Full-length Sog is unable to rescue tsg mutant embryos in such experiments, however, consistent with Tsg being required to process or modify Sog to generate a Supersog-like activity. These findings in combination with those described above suggest a model in which Sog can be differentially cleaved by Tld to be either destroyed (i.e., in the absence of Tsg) or processed into a more active Supersog-like form (i.e., in the presence of Tsg).

In this article, we present a systematic structure-function study of Sog from which we extract the following major conclusions. First, we find that the individual CRs serve partially redundant functions. For example, an internally deleted form of Sog lacking all but the CR4 domain (SogCR4) retains significant Sog-like activity. This finding is consistent with a functional analysis of Chordin in which overlapping activities of the CR1 and CR3 domains were reported (LARRAIN et al. 2000 ). Second, we find that less truncated forms of Sog, which lack CR4 (SogCR4) or both CR3 and CR4 (SogCR1,2), exert activities that are distinct from those of Sog or Supersog. These SogCR4 and SogCR1,2 activities are similar to those resulting from activation of Gbb Sax signaling. Finally, we examine whether the CR domains determine the distinct activities of various forms of Sog by performing a series of CR domain swaps in the contexts of SogCR4 only, Supersog1, or SogCR1,2 constructs. Analysis of the activities of these chimeric molecules suggests that CRs are important for determining the level of activity and that surrounding amino acids are the chief determinants of the type of activity exerted by different forms of Sog. These results reinforce the view that Sog is a multifunctional modulator of BMP activity and focus attention on non-CR sequences as determinants of specificity in BMP recognition.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks:
Several independent lines of each pUAS construct were obtained by P-element-mediated germline transformation. The MS1096-GAL4 driver is ubiquitously expressed in the wing, but is expressed more strongly on the dorsal surface of the wing primordium during larval stages and becomes restricted to the dorsal surface during pupal stages (CAPDEVILA and GUERRERO 1994 ; LUNDE et al. 1998 ). The bicoid-GAL4/GCN4 stock (JANODY et al. 2000 ) was kindly provided by Claude Desplan, New York University. Other strains carrying genetic markers or chromosome balancers were obtained from either the Bloomington, Indiana, or Bowling Green, Ohio, Drosophila stock centers. All pUAS and pGAL4 constructs are supplied in a single copy unless indicated otherwise (e.g., Fig 6 and Table 2).

View larger version (51K): In this window In a new window Download PPT slide   Figure 2. Activities of sog mutant constructs assayed in blastoderm stage embryos. Expression of race (A–I) or zen (J–L) in embryos misexpressing intact sog (B and K) or various mutant UAS-sog constructs in the head region under the control of the bcd-GCN4/GAL4 driver (JANODY et al. 2000 ) is shown. The genotypes of the embryos and the gene expression patterns are as labeled. The genotype is indicated first in boldface black type. For example, bcd-GCN4/GAL4-driven expression of a wild-type UAS-sog construct in the head is indicated as bcd>sog. Expression of endogenous genes is indicated following the genotype and is color coded to match that shown in the embryos: brown type indicates a transcript/protein labeled with a biotin probe/antibody and detected with a brown HRP-reaction product, whereas blue type indicates a transcript labeled with digoxigenin and detected as a blue alkaline phosphatase reaction product. Embryos are viewed from a dorsal perspective with anterior to the left.

View larger version (84K): In this window In a new window Download PPT slide   Figure 3. Activities of sog mutant constructs assayed in developing wings. Adult wing phenotypes resulting from misexpression of intact (B) or mutant (C–J) UAS-sog constructs are shown. Labels are according to the GAL4 driver used (first term in boldface type, e.g., ubi, dpp, or en), followed by the symbol >, and then followed by the name of the gene being misexpressed (in boldface italics; e.g., ubiquitous expression of a wild-type UAS-sog construct in the wing using the MS1096-GAL4 driver is indicated as ubi>sog). Designations of sog mutant constructs are as indicated in Fig 1. The margin (M) of the wing and longitudinal veins (L2–L5) of a wild-type (wt) wing are indicated in A. Arrowheads in B indicate regions of missing veins in ubi>sog wings. In D (ubi>sogCR4) and F (ubi>sogCR1,2) arrowheads indicate ectopic veins. In J (en>supersog1) arrowheads indicate the locations where the L4 and L5 veins are truncated in wings misexpressing supersog1 in the posterior compartment under the control of the engrailed GAL4 driver. Bars, 0.2 mm in this and subsequent figures.

View larger version (54K): In this window In a new window Download PPT slide   Figure 4. Interactions between Sog and Tsg. (A–F) Adult wing phenotypes resulting from ubiquitous (ubi) coexpression of intact or mutant UAS-sog constructs or UAS-chordin with UAS-tsg using the MS1096-GAL4 wing driver. Designations for sog mutant constructs are as indicated in Fig 1. (G–J) Xenopus embryos were injected at the 4-cell stage with RNAs as indicated at the following concentrations: 1.5 ng sog, 1 ng of activated tld (tld*), and 1 ng tsg, and harvested at stage 30. (G) Uninjected embryo. (H) Embryo injected with sog. (I) Embryo injected with sog + tld*. (J) Embryo injected with sog + tld* + tsg. The arrowhead in H and J indicates the second axis. The embryos shown represent the major type found in each experiment (Table 1).

View larger version (47K): In this window In a new window Download PPT slide   Figure 5. Analysis of the sogCR1,2 phenotype. Gene expression patterns were examined in wild-type (WT) wing discs (A and B) or in wing discs ubiquitously misexpressing UAS-sogCR1,2 (C and D), UAS-sax*(E), or UAS-gbb (F), using the MS1096-GAL4 driver. We also examined rho expression in pupal ubi>sogCR1,2 wings (G). Expression of vein (rho) and intervein (Bs) genes was detected by in situ hybridization (rho) or by immunohistochemical staining (Bs). The adult wing shown in H is from an individual of the genotype HS-GAL4; UAS-sogCR1,2, which received a 2-hr heat shock during early pupal stages (e.g., 20–24 hr after pupariation) and was then shifted back to 25° until eclosion.

View larger version (65K): In this window In a new window Download PPT slide   Figure 6. Activities of Sog CR swap constructs. (A) Diagram of Sog CR domain swap constructs. (B–K) Adult wing phenotypes resulting from ubiquitous (ubi) expression of one (1X), two (2X) or four (4X) copies of UAS-sogCR-X, UAS-supersog-X = UAS-SS-X, or UAS-sogCR-X,Y domain swap constructs using the MS1096-GAL4 wing driver. Constructs are diagrammed in A, and the activities of these constructs are indicated in Table 2.

Generation of mutant sog constructs:
cDNAs encoding the full coding regions of sog, chordin (YU et al. 2000 ), and mutant forms of sog (this study) were inserted into the pUAS expression vector (BRAND and PERRIMON 1993 ), using appropriate restriction sites. The sequences and details of assembly of the various deletion, point mutation, and CR swap sog constructs diagrammed in Fig 1 and Fig 6 will be provided upon request.

Mounting fly wings:
Wings from adult flies were dissected in isopropanol and mounted in Canadian Balsam mounting medium.

In situ hybridization to whole-mount embryos or discs:
In situ hybridization to whole-mount embryos and wing imaginal discs was performed with digoxigenin-labeled RNA probes (visualized as a blue alkaline phosphatase precipitate; O'NEILL and BIER 1994 ). Stained embryos were mounted in Permount and examined and photographed using differential interference contrast optics under a compound microscope.

Microinjection of Xenopus embryos:
RNA was prepared using the mMessage mMachine kit (Ambion, Austin, TX) according to the manufacturer's protocol. RNA was injected once subequatorially on the ventral side of the embryo at the two- to four-cell stage. All RNAs were produced from CS2-derived vectors (TURNER and WEINTRAUB 1994 ). The sog-CS2 construct (sog25) was described previously (SCHMIDT et al. 1995 ) and activated tolloid (tld*; MARQUES et al. 1997 ) was the same as that used previously (YU et al. 2000 ). The Drosophila tsg gene was inserted into the EcoRI and XbaI sites of CS2 to produce CS-tsg.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The Sog CR1, CR2, and CR3 domains are individually dispensable:
As a first step in analyzing the roles of the various CR domains of Sog and the intervening stem between CR1 and the carboxy-terminal cluster of CR2–CR4 domains, we constructed a series of deleted forms of Sog lacking specific domains or combinations of domains (Fig 1). We also constructed point mutants in which the conserved tryptophan residues between the first two cysteines of the CRs were changed individually or in combination to alanine (Fig 1). To determine whether the various CR domains were individually required for Sog activity, we assayed the function of Sog mutants lacking each of the four CR domains (Fig 1) in blastoderm stage embryos (Fig 2) and adult wings (Fig 1 and Fig 3). In the blastoderm embryo, we drove expression of the various UAS-sog mutant transgenes, using a hybrid GAL4/GCN4 transactivating protein, the mRNA for which is tethered to the anterior portion of the embryo by bicoid RNA localization signals (JANODY et al. 2000 ). This bcd-GAL4/GCN4 transactivator consists of the DNA-binding domain of GAL4 and the transactivation domain of GCN4 and activates high levels of UAS-transgene expression in the head region of the embryo (Fig 2A). When wild-type sog is misexpressed with bcd-GAL4/GCN4, expression of BMP target genes such as race (Fig 2B), zen (Fig 2K), and rho (data not shown) is suppressed in dorsal anterior cells of the embryo. Similarly, expression of mutant forms of Sog deleted for either the CR1 (data not shown) or CR2 (Fig 2C) domains leads to suppression of race expression in the head and anterior trunk regions. Deletion of either CR3 (Fig 2D) or CR4 (data not shown), however, resulted in greatly reduced Sog activity in the embryo, although, as described below, these constructs had reduced but detectable Sog-like activity in the wing. These results indicate that the CR3 and CR4 domains individually contribute more to Sog activity in the embryo than do either the CR1 or CR2 domains.

We also analyzed the activity of all mutant UAS-sog constructs (Fig 1) in the wing by expressing them with the ubiquitous wing GAL4 driver MS1096 during late larval and pupal stages (Fig 3). In the wing it is possible to distinguish three distinct requirements for BMP signaling: growth of the imaginal disc, patterning of the A/P axis, and vein development (SEGAL and GELBART 1985 ; YU et al. 1996 ). Misexpression of full-length Sog results in only a mild vein-loss phenotype but does not alter the shape or size of the wing (Fig 3B, compare to Fig 3A). This phenotype most likely results from intact Sog selectively inhibiting Gbb but not Dpp signaling (YU et al. 2000 ). Similar vein-loss phenotypes were observed in response to misexpression of SogCR1, SogCR2 (Fig 3C), and SogCR3 although the SogCR3 phenotype was consistently weaker than that generated by misexpression of intact Sog (Fig 1). In addition, compound mutants lacking CR1 and the stem region (e.g., SogCR1,stem), CR1 through CR2 (e.g., SogCR3,4), or CR1 through CR3 (e.g., SogCR4; Fig 3E) had Sog-like activity in this assay (albeit the latter two constructs had reduced activity relative to intact Sog; Fig 3B), indicating that a single CR domain (e.g., CR4) can provide much of the Sog function. Some individuals misexpressing SogCR4 also exhibited a mild vein-loss phenotype (data not shown); however, others had ectopic veins (Fig 3D). This latter novel extra vein phenotype is discussed further below. These data indicate that any single CR domain is dispensable for the BMP inhibitory activity of intact Sog, although in the case of SogCR4, a conflicting gain-of-function activity partly obscures the underlying inhibitory Sog function.

Sog mutants in which each or all conserved tryptophan residues (Fig 3G) were mutated to alanine also caused vein-loss phenotypes similar to, although somewhat weaker than, those observed with intact Sog when misexpressed in the wing. This result indicates that the conserved tryptophan residues are dispensable for the Sog activity responsible for vein suppression. Similarly, individual W A mutants had a Sog-like activity in the embryo when driven with the bcd-GAL4/GCN4 driver, although the quadruple W A mutant had greatly reduced activity (Fig 2E).

Another question we addressed was whether there is functional significance to the fact that Sog has type II secretory signal located 56 amino acids from the amino terminus (FRANCOIS et al. 1994 ) rather than a typical amino-terminal signal sequence since, in principle, such an internal hydrophobic sequence could anchor Sog to the cell membrane. We replaced the endogenous internal Sog signal sequence with a standard signal peptide from the secreted Easter protease and found that this Ea-Sog construct was fully active when expressed in the wing (Fig 1). We also determined whether Supersog with its endogenous type II signal sequence can act in cells that do not express dpp. Since previous experiments have involved misexpressing Supersog in patterns that include those expressing dpp, it might be imagined that coexpression of a potentially membrane associated form of Supersog with Dpp could simply block Dpp secretion from cells. To address this question, we drove expression of UAS-supersog in posterior compartment cells with the en-GAL4 driver (Fig 3J) and observed a severe vein-loss phenotype in which the remaining L4 and L5 segments were nearly fused. This phenotype is very similar to that observed in posterior regions of the wing in response to ubiquitous expression of Supersog (Fig 3H) or with expression limited to Dpp producing cells (Fig 3I). These results indicate that the type II signal sequence of Sog is interchangeable with a standard signal sequence and that Supersog can function in cells both producing and responding to Dpp.

Conserved CR tryptophan residues and the stem region are required for the Sog-Tsg interaction:
A variety of evidence indicates that Tsg, which plays a necessary role in dorsal-ventral patterning in the embryo (MASON et al. 1994 ; YU et al. 2000 ), is involved in creating a Supersog-like activity (YU et al. 2000 ; SHIMMI and O'CONNOR 2003 ). For example, coexpression of intact Sog and Tsg generates a Supersog-like phenotype in the wing (Fig 4B, compare to Fig 3H), whereas expression of these genes individually generates only mild vein loss (Fig 3B) or a weak ectopic vein phenotype (Fig 4A), respectively. Consistent with previous evidence that vertebrate Chordin functions like Sog in the wing (Fig 4C; YU et al. 2000 ), coexpression of Chordin and Tsg also generates a Supersog-like phenotype (Fig 4D). We further tested the evolutionary conservation of the Sog/Tsg interaction by asking whether Sog and Tsg interact in Xenopus to create a Supersog-like activity. Previous experiments revealed that while injection of Sog or Supersog RNAs into ventral blastomeres of Xenopus embryos both can induce the formation of secondary axes (Fig 4H), Sog differs from Supersog in that the effects of Sog (Fig 4I), but not Supersog, can be reversed by co-injection with activated Tolloid (Tld*; YU et al. 2000 ). We asked whether co-injection of Sog with Tsg could similarly protect Sog from Tld* by injecting a mix of all three RNAs (Fig 4J, compare to Fig 4H and Fig 4I; see also Table 1). These experiments indicate that Sog and Tsg can indeed interact in vertebrate embryos and that Sog + Tsg, like Supersog, is immune to Tld* activity.

To identify potentially critical residues in Sog required for the interaction with Tsg, we misexpressed the collection of various Sog mutants with Tsg in the wing (Fig 1). In these experiments two types of Sog mutants exhibited compromised interactions with Tsg. The first of these mutants (Sog-A1,2,3,4), which is mutated at all four conserved tryptophan residues, generates a much weaker wing phenotype in combination with Tsg (Fig 4E) than the Supersog-like phenotype caused by coexpression of intact Sog with Tsg (Fig 4B). The Sog-A1,2,3,4 + Tsg coexpression phenotype (Fig 4E) is, however, somewhat stronger than that caused by Sog-A1,2,3,4 (Fig 3G) or intact Sog alone (Fig 3B), indicating that there is a residual interaction between Sog-A1,2,3,4 and Tsg. Since each of the single Sog W A mutants fully retained their ability to synergize with Tsg (Fig 1), it is likely that several or all of the conserved tryptophan residues in the CRs are involved in mediating the effect of Tsg. The second class of mutants that fails to interact with Tsg lacks the stem sequences between CR1 and CR2. For example, SogCR1, stem, which lacks both CR1 and the stem portion of Sog (or any other mutant lacking the stem region), fails entirely to interact with Tsg as revealed by the indistinguishable wing phenotypes resulting from misexpression of Tsg + SogCR1,stem (Fig 4F) or Tsg alone (Fig 4A). Since the Sog-A1 and SogCR1 mutants synergize fully with Tsg, it is likely that failure of SogCR1,stem to interact reflects a role of the stem region in this process.

We also found that coexpression of Supersog and Tsg generates an enhanced phenotype relative to that produced by a single copy of Supersog alone (Fig 1). Curiously, the interaction between Supersog and Tsg is also observed when the single remaining conserved tryptophan in Supersog is mutated to alanine (Supersog1-A1, Fig 1). This moderate genetic interaction may be mediated indirectly, however, since we observe a similar degree of interaction between Tsg and the non-Drosophila vertebrate antagonist Noggin (data not shown).

Deletion of the CR3 and CR4 domains creates a novel Sog activity:
As mentioned above, ubiquitous expression of a Sog mutant lacking the CR4 domain (SogCR4) results in a novel ectopic vein phenotype (Fig 3D). This ectopic SogCR4 phenotype is relatively mild and sometimes coexists with the typical vein-loss pattern caused by misexpression of intact Sog (data not shown). The complex and variable SogCR4 phenotype may reflect the sum of a vein-loss phenotype (e.g., as caused by intact Sog) and a novel extra-vein phenotype. Consistent with this possibility, removal of both CR3 and CR4 (SogCR1,2) induces a more extensive and penetrant ectopic vein phenotype without any vein loss (Fig 3F). It is possible that the stronger ectopic vein phenotype associated with SogCR1,2 overcomes a weaker Sog-like vein-loss phenotype. Alternatively, removal of both the CR3 and CR4 domains may severely reduce or eliminate the vein-inhibiting activity associated with intact Sog.

The ectopic vein phenotypes caused by misexpressing SogCR4 or SogCR1,2 are similar to those caused by hyperactive EGF-R (STURTEVANT et al. 1993 ; NOLL et al. 1994 ) or Dpp signaling (YU et al. 1996 ), suggesting that these forms of Sog may exert a positive influence on the EGF-R or Dpp pathways. We examined the basis for ectopic wing vein phenotypes induced by SogCR4 or SogCR1,2 by misexpressing these mutant forms of Sog during larval and pupal wing development and assaying the expression of various markers, including Dpp target genes (e.g., spalt and omb), vein markers (e.g., kni, abrupt, rhomboid or rho, Delta, and caupolican), and intervein markers (e.g., blistered or bs equals DSRF and vn). Consistent with the severity of their final ectopic wing vein phenotypes, SogCR4 and SogCR1,2 have similar effects on marker gene expression, with SogCR1,2 being the stronger of the two (data not shown). For comparison, we also examined gene expression patterns in discs misexpressing Sog, Supersog, BMP ligands, and dominant-negative or activated forms of BMP receptors and epidermal growth factor (EGF) receptors. Two primary conclusions can be drawn from this analysis. First, the activity of SogCR4 and SogCR1,2 differs from that of either intact Sog or Supersog. For example, misexpression of SogCR1,2 in third instar wing discs has no effect on expression of A/P patterning genes, but leads to ectopic expression of the vein-promoting gene rho (Fig 5C, compare to 5A) and downregulation of the vein-suppressing genes bs (Fig 5D, compare to 5B) and Delta (data not shown). In contrast, misexpression of intact Sog has no effect on expression of any positional or vein/intervein marker during larval stages, while misexpression of Supersog1 both inhibits expression of Dpp target genes (YU et al. 2000 ) and leads to suppression of rho expression (data not shown). Second, among the various BMP ligands and dominant-negative and activated forms of BMP receptors and EGF receptors, the SogCR1,2 misexpression phenotype most resembles that of activated Sax* or Gbb, in that there are no obvious A/P patterning defects, but there is localized ectopic expression of rho anterior to L3 and posterior to L4 (Fig 5E, compare to 5C) as well as reduced expression of the intervein genes bs (Fig 5F, compare to 5D) and Delta (data not shown). These data suggest that SogCR1,2 exerts a positive effect on BMP signaling, potentially via Gbb and the Sax receptor.

We also examined the effect of expressing SogCR4 and SogCR1,2 in pupal wings when the vein vs. intervein cell fate choice is finalized. We observed significant ectopic expression of the rhomboid (rho) gene (Fig 5G), which we have previously shown is a good marker for Dpp activity. BMP signaling efficiently activates expression of rho, whereas activation of the EGF-R pathway induces significant ectopic dpp expression, but only very weak ectopic rho expression (YU et al. 1996 ). The relatively weak ability of Rho/EGF-R signaling to positively autoregulate presumably is the result of dampening that accompanies the indirect mechanism of rho autoactivation (YU et al. 1996 ). We also found that heat-induced expression of SogCR1,2 during early pupal stages (when Dpp signaling is known to promote vein fates; YU et al. 1996 ) leads to ectopic vein formation (Fig 5H), indicating that this construct is active during pupal as well as larval stages.

As a final test of SogCR1,2 activity, we misexpressed UAS-sogCR1,2 in the embryo using the bcd-GCN4/GAL4 expression system described above (JANODY et al. 2000 ). In contrast to intact Sog, SogCR1,2 is unable to suppress expression of Dpp target genes such as race (Fig 2F), zen (Fig 2L), or rho (data not shown) in the head with the bcd-GCN4/GAL4 driver. Rather, sogCR1,2 misexpression has the opposite effect on zen expression in that refined zen expression in late blastoderm embryos expands from the narrow stripe observed in wild-type embryos (Fig 2J) to a stripe of nearly double width (Fig 2L).This broadening of the zen domain, however, depends on reducing the dose of endogenous sog, as it is observed only in sog–/+ heterozygous embryos, but not in embryos with the equivalent of two copies of sog. The pattern of zen expression is not significantly altered in sog–/+ control embryos (data not shown). Misexpression of SogCR1,2 does not have a significant effect on the expression of the dorsal markers race or rho, however.

A positive Sog activity in embryos has also been reported by others (ASHE and LEVINE 1999 ; DECOTTO and FERGUSON 2001 ). These authors observed that, in addition to suppressing expression of a variety of Dpp target genes at short range, sog acted at long range to rescue expression of race in mutants lacking endogenous sog function. In sog– embryos, race expression is broadened anteriorly but fades to low levels in more posterior regions (Fig 2G). Consistent with previously reported findings, we observed that loss of posterior race expression could be rescued by supplying Sog in the head (Fig 2H). SogCR1,2, however, was unable to rescue appreciable race expression in sog– mutants (Fig 2I). In addition, the positive function of SogCR1,2 broadened zen expression over a long range in sog–/+ heterozygous embryos, whereas wild-type Sog only inhibited zen expression (Fig 2K). These results suggest that the SogCR1,2 activity may be different from the race-promoting activity of intact Sog.

CR domains and surrounding sequences determine the level and quality of Sog activities:
The results of the CR deletion experiments described above suggest that no single CR is absolutely required for the activity of intact Sog (e.g., inhibition of Scw/Gbb), although certain CRs (e.g., CR3 and CR4) seemed to be more important than others for this function. Another way we assessed the degree to which CRs vs. surrounding protein sequences contribute to the strength and quality of Sog activities is by swapping the CR domains in the context of the backbones of SogCR4, Supersog1 (SS-1), and SogCR1,2. In these experiments, the native CR domains were replaced with each of the other three possible CRs (Fig 6A and Table 2). Several independently isolated transformant lines were then misexpressed in the wing and tested for activity as single insertions or combined in doses of two and four copies to increase levels of expression (Table 2). Two major conclusions can be drawn from these experiments. First, in the great majority of cases, altering the identity of the CRs lead to loss (Fig 6D, compare to 6B; Fig 6F and Fig H, compare to Fig 3H) or vastly reduced activity relative to the parent construct. The greatly attenuated activity of these chimeric proteins is not due to synthesis of unstable or abnormally processed protein products as we have examined induced protein expression from these various constructs in vivo by immunoblotting with an anti-Sog antibody and find that all the mutant proteins tested are made at levels comparable to those of the initial Sog constructs and that these proteins are of the predicted sizes (data not shown). These observations reveal that the CR domains are not interchangeable and minimally that they play a role in determining the magnitude of activity of the various Sog forms.

The second important result of the CR swap experiments is that when a CR swap construct had activity, it was always the same as that of the original construct. For example, in the context of the Sog-CR4 swaps, replacement of CR4 with CR2 (SogCR-2) or CR3 (SogCR-3) results in SogCR4-like phenotypes when these chimeric constructs are expressed in multiple copies (Fig 6E and Table 2). Likewise, constructs in which the CR1 domain of Supersog was substituted with CR3 or CR4 cause modest (Fig 6G, SS-3) or severe (Fig 6I, SS-4) Supersog-like phenotypes when expressed in several copies, and SogCR-4,2 (Fig 6K) has a modest SogCR1,2-like phenotype (Fig 6J) when expressed in two copies. These observations strongly suggest that the category of activity manifested by the various Sog forms is determined chiefly by the surrounding protein sequences and not by the CR domains, which act primarily in a quantitative rather than qualitative fashion to determine activity levels. The finding that deletion of either the CR1 domain or the neighboring stem region from Supersog eliminates its activity (Fig 1) is also consistent with the view that CR domains must interact with adjacent sequences to create a BMP inhibitory activity.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this study we examined the contribution of different Sog domains to three distinct activities of Sog: intact Sog, which selectively blocks signaling mediated by the BMP ligands Scw/Gbb; Supersog, which acts a broader spectrum BMP inhibitor; and SogCR1,2, which has a distinct potentially positive BMP-promoting activity. Analysis of deletion and point mutation sog constructs as well as CR domain swaps indicates that both the CR domains and surrounding sequences participate in defining the various distinct activities of Sog. The CR domains function in a partially redundant fashion to determine the activity level of different Sog forms, while the surrounding sequences appear to be the chief determinants of the type of Sog activity. These studies reinforce and extend the previous view of Sog as a multifunction modulator of BMP signaling and suggest that future studies should explore the mechanism by which CR domains modify the conformation of adjacent protein sequences to interact with particular BMP targets.

CR tryptophan residues and the stem region are required for Tsg to modify Sog activity:
The Sog CR domains are defined by a set of 10 cysteine residues with a conserved spacing and a single tryptophan residue located between the first two cysteines. We explored the function of the tryptophan residues by mutating them individually or all to alanine. The finding that all four single W A mutants have wild-type Sog function as assayed by misexpression in the wing, alone or when coexpressed with Tsg, indicates that none of these residues is individually essential for either Sog or Tsg + Sog (Supersog) activities. This finding is also consistent with the results of deleting the individual CR domains (see below). When all four tryptophans were mutated to alanine, however, we found that the Sog-like activity remained relatively unaffected but that this mutant was greatly compromised in its ability to interact with Tsg to generate a Supersog-like activity. These results suggest that the tryptophan residues in two or more CRs can mediate functional interaction with Tsg and that Sog residues outside of the four conserved tryptophans are not sufficient on their own to mediate this interaction. On the other hand, deletion of the stem region also eliminates the functional Tsg interaction since a mutant lacking the stem and CR1 fails to interact with Tsg while a mutant lacking just CR1 interacts fully. The requirement for the stem region in interacting with Tsg is consistent with both CR and stem sequences being essential for Supersog activity (Fig 1).

SogCR1,2 has a different activity than intact Sog or Supersog:
Truncated forms of Sog consisting of CR1, the stem, and CR2 behave differently than either Sog or Supersog constructs when misexpressed in the wing or embryo. The strongest form of this novel Sog activity is observed when both CR3 and CR4 are deleted (e.g., SogCR1,2). Several lines of evidence suggest that Sog CR1,2 functions by promoting BMP signaling. First, the effect of misexpressing SogCR1,2 on gene expression in wing discs is most similar to that of misexpressing an activated Sax receptor or the putative Sax ligand Gbb. This profile of gene response is quite distinct from that resulting from misexpression of activated or dominant-negative forms of the BMP receptor or EGF-receptor pathway, which is the other major signaling system regulating early vein development. Second, misexpression of SogCR1,2 in pupal wings by heat shock results in significant ectopic expression of the rho gene, which is a good measure of BMP vs. EGF-R pathway activation during this stage. Finally, expression of SogCR1,2 in the early embryo broadens the dorsal expression domain of the BMP target gene zen.

One attractive model for a positive function of Sog such as that potentially mediated by SogCR1,2 is that in addition to binding to BMPs and preventing them from gaining access to the receptors, Sog might also act as a carrier of BMPs to either protect them from degradation or possibly transport them dorsally (DECOTTO and FERGUSON 2001 ; ELDAR et al. 2002 ; SHIMMI and O'CONNOR 2003 ). Since the CR3 and CR4 domains appear to be those most critical to inhibition of Gbb/Scw activity, it is possible that the apparent positive activity of SogCR1,2 is the result of removing the Scw/Gbb inhibitory activity, while leaving a carrier function intact. The difference between the activity of SogCR1,2 and Supersog molecules, which lack the CR2 domain, might be explained if CR1 and CR2 can interact in SogCR1,2 to prevent CR1 and/or adjacent sequences from interacting with Dpp, thus neutralizing this remaining potential BMP inhibitory activity. It is currently unclear what relationship, if any, the SogCR1,2 activity has to the positive function of Sog required to sustain dorsal expression of race in the early embryo (ASHE and LEVINE 1999 ; DECOTTO and FERGUSON 2001 ). SogCR1,2, unlike intact Sog, is unable to rescue race at a distance, but can broaden dorsal zen expression, which intact Sog does not do. The apparent differences between these activities may be the result of threshold-dependent effects in the early embryo or may reflect a fundamentally different mode of action.

An important question is whether truncated forms of Sog similar to SogCR1,2 or SogCR4 are generated and function in vivo. It is known that Tld can cleave Sog in vitro to generate products of approximately the same size as these constructs (MARQUES et al. 1997 ; SHIMMI and O'CONNOR 2003 ), and Sog-reactive bands of approximately the same size are observed in early embryos and pupal wings (YU et al. 2000 ). In addition, forms of Chordin similar in structure to SogCR1,2 and Supersog are produced in humans as the result of alternative RNA splicing (MILLET et al. 2001 ). Further analysis of the production and activity of SogCR1,2-like molecules will be required to determine the relevance of such forms in vivo and to determine the mechanism by which they may act on the BMP pathway.

CR domains determine the level and adjacent sequences determine the type of Sog activity:
Analysis of Sog mutants in which individual CRs are deleted or single conserved tryptophan residues are mutated to alanine suggested that the CR domains perform partially overlapping functions since none of them is absolutely essential for intact Sog activity (e.g., inhibition of Gbb/Scw) or for interaction with Tsg to create a Dpp inhibitory activity. Nonetheless, these experiments also suggested that the CR domains are not equivalent. For example, deletion of CR3 or CR4 had much greater effects in reducing the SOG-like activity than did deletion of CR1 or CR2. Moreover, the fact that Supersog, which contains only CR1, can inhibit Dpp while SogCR4 apparently interferes selectively with Gbb, suggested that CR1 might bind Dpp while CR4 bound Gbb. The results of the CR swap experiments performed in the contexts of SogCR4, Supersog1, and SogCR1,2 are quite informative in resolving this question and provide a surprising answer, namely that sequences adjacent to CR domains are the primary determinants of BMP specificity.

While replacing a CR domain in swap constructs typically resulted in greatly reduced activity or inactivity of UAS-transgenes tested as single- or multicopy insertions, those that had activity generated phenotypes similar to those of the parent constructs. These data suggest that the CRs are required in a context-specific fashion to boost the activity levels of the various forms of Sog. These findings, although limited in scope, suggest that the primary determinant of the quality of Sog activity lies within the sequences surrounding the CRs rather than within the CR domains themselves. These non-CR sequences may correspond to identified repetitive motifs such as the SR repeats (FRANCOIS et al. 1994 ) or the circularly permuted SOG/CHRD domains (HYVONEN 2003 ). The simplest interpretation of this result is that CRs contribute to defining the strength and surrounding sequences determine the specificity of Sog activities. This model of Sog function is consistent with the finding mentioned above that deletion of any single CR does not eliminate Sog-like activity (e.g., inhibition of Scw/Gbb) or the ability to interact with Tsg (e.g., inhibition of Dpp) and that both the CR1 domain and adjacent stem sequences are required for Supersog activity. The fact that the conserved tryptophans in the CR domains are required in aggregate, but not individually, for interaction with Tsg and that stem sequences are also required for this interaction lends additional support to the view that the CRs function in a partially redundant fashion in conjunction with non-CR sequences. One potential explanation for the results reported here and by Larrain and colleagues (LARRAIN et al. 2000 ) is that the CR domains alone bind to BMPs, but do so with little selectivity. The surrounding sequences may provide specificity by interacting with only a subset of BMPs, thereby increasing the affinity of adjacent CRs for particular BMPs. Alternatively, surrounding sequences may alter the conformation of CR domains or sterically limit their interactions with BMPs, rendering them more selective. Further biochemical analysis will be required to determine the degree to which CRs affect affinity of the various Sog forms for particular BMPs and the mechanism by which surrounding sequences may confer specificity for binding particular BMPs.

	ACKNOWLEDGMENTS

We thank members of the Bier lab for helpful comments on the manuscript. This work was supported by National Institutes of Health grant R01 NS29870.

Manuscript received November 3, 2003; Accepted for publication December 11, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ARORA, K., M. S. LEVINE, and M. B. O'CONNOR, 1994  The screw gene encodes a ubiquitously expressed member of the TGF-beta family required for specification of dorsal cell fates in the Drosophila embryo. Genes Dev. 8:2588-2601.[Abstract/Free Full Text]

ASHE, H. L. and M. LEVINE, 1999  Local inhibition and long-range enhancement of Dpp signal transduction by Sog. Nature 398:427-431.[CrossRef][Medline]

BALEMANS, W. and W. VAN HUL, 2002  Extracellular regulation of BMP signaling in vertebrates: a cocktail of modulators. Dev. Biol. 250:231-250.[CrossRef][Medline]

BIEHS, B., V. FRANCOIS, and E. BIER, 1996  The Drosophila short gastrulation gene prevents Dpp from autoactivating and suppressing neurogenesis in the neuroectoderm. Genes Dev. 10:2922-2934.[Abstract/Free Full Text]

BIER, E., 1997  Anti-neural-inhibition: a conserved mechanism for neural induction. Cell 89:681-684.[CrossRef][Medline]

BRAND, A. H. and N. PERRIMON, 1993  Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development 118:401-415.[Abstract]

CADIGAN, K. M., 2002  Regulating morphogen gradients in the Drosophila wing. Semin. Cell Dev. Biol. 13:83-90.[CrossRef][Medline]

CAPDEVILA, J. and I. GUERRERO, 1994  Targeted expression of the signaling molecule decapentaplegic induces pattern duplications and growth alterations in Drosophila wings. EMBO J. 13:4459-4468.[Medline]

CHANG, C., D. A. HOLTZMAN, S. CHAU, T. CHICKERING, and E. A. WOOLF et al., 2001  Twisted gastrulation can function as a BMP antagonist. Nature 410:483-487.[CrossRef][Medline]

CONLEY, C. A., R. SILBURN, M. A. SINGER, A. RALSTON, and D. ROHWER-NUTTER et al., 2000  crossveinless 2 contains cysteine-rich domains and is required for high levels of BMP-like activity during the formation of the cross veins in Drosophila.. Development 127:3947-3959.[Abstract]

DECOTTO, E. and E. L. FERGUSON, 2001  A positive role for Short gastrulation in modulating BMP signaling during dorsoventral patterning in the Drosophila embryo. Development 128:3831-3841.[Abstract/Free Full Text]

ELDAR, A., R. DORFMAN, D. WEISS, H. ASHE, and B. Z. SHILO et al., 2002  Robustness of the BMP morphogen gradient in Drosophila embryonic patterning. Nature 419:304-308.[CrossRef][Medline]

FRANCOIS, V. and E. BIER, 1995  Xenopus chordin and Drosophila short gastrulation genes encode homologous proteins functioning in dorsal-ventral axis formation. Cell 80:19-20.[CrossRef][Medline]

FRANCOIS, V., M. SOLLOWAY, J. W. O'NEILL, J. EMERY, and E. BIER, 1994  Dorsal-ventral patterning of the Drosophila embryo depends on a putative negative growth factor encoded by the short gastrulation gene. Genes Dev. 8:2602-2616.[Abstract/Free Full Text]

HAERRY, T. E., O. KHALSA, M. B. O'CONNOR, and K. A. WHARTON, 1998  Synergistic signaling by two BMP ligands through the SAX and TKV receptors controls wing growth and patterning in Drosophila.. Development 125:3977-3987.[Abstract]

HAMMERSCHMIDT, M., G. N. SERBEDZIJA, and A. P. MCMAHON, 1996  Genetic analysis of dorsoventral pattern formation in the zebrafish: requirement of a BMP-like ventralizing activity and its dorsal repressor. Genes Dev. 10:2452-2461.[Abstract/Free Full Text]

HEMMATI-BRIVANLOU, A. and D. MELTON, 1997  Vertebrate embryonic cells will become nerve cells unless told otherwise. Cell 88:13-17.[CrossRef][Medline]

HOLLEY, S. A., P. D. JACKSON, Y. SASAI, B. LU, and E. M. DE ROBERTIS et al., 1995  A conserved system for dorsal-ventral patterning in insects and vertebrates involving sog and chordin.. Nature 376:249-253.[CrossRef][Medline]

HYVONEN, M., 2003  CHRD, a novel domain in the BMP inhibitor chordin, is also found in microbial proteins. Trends Biochem. Sci. 28:470-473.[CrossRef][Medline]

JANODY, F., J. REISCHL, and N. DOSTATNI, 2000  Persistence of Hunchback in the terminal region of the Drosophila blastoderm embryo impairs anterior development. Development 127:1573-1582.[Abstract]

KHALSA, O., J. W. YOON, S. TORRES-SCHUMANN, and K. A. WHARTON, 1998  TGF-beta/BMP superfamily members, Gbb-60A and Dpp, cooperate to provide pattern information and establish cell identity in the Drosophila wing. Development 125:2723-2734.[Abstract]

LARRAIN, J., D. BACHILLER, B. LU, E. AGIUS, and S. PICCOLO et al., 2000  BMP-binding modules in Chordin: a model for signalling regulation in the extracellular space. Development 127:821-830.[Abstract]

LUNDE, K., B. BIEHS, U. NAUBER, and E. BIER, 1998  The knirps and knirps-related genes organize development of the second wing vein in Drosophila.. Development 125:4145-4154.[Abstract]

MARQUES, G., M. MUSACCHIO, M. J. SHIMELL, K. WUNNENBERG-STAPLETON, and K. W. CHO et al., 1997  Production of a DPP activity gradient in the early Drosophila embryo through the opposing actions of the SOG and TLD proteins. Cell 91:417-426.[CrossRef][Medline]

MASON, E. D., K. D. KONRAD, C. D. WEBB, and J. L. MARSH, 1994  Dorsal midline fate in Drosophila embryos requires twisted gastrulation, a gene encoding a secreted protein related to human connective tissue growth factor. Genes Dev. 8:1489-1501.[Abstract/Free Full Text]

MASSAGUE, J. and Y. G. CHEN, 2000  Controlling TGF-beta signaling. Genes Dev. 14:627-644.[Free Full Text]

MILLET, C., P. LEMAIRE, B. ORSETTI, P. GUGLIELMI, and V. FRANCOIS, 2001  The human chordin gene encodes several differentially expressed spliced variants with distinct BMP opposing activities. Mech. Dev. 106:85-96.[CrossRef][Medline]

NEUL, J. L. and E. L. FERGUSON, 1998  Spatially restricted activation of the SAX receptor by SCW modulates DPP/TKV signaling in Drosophila dorsal-ventral patterning. Cell 95:483-494.[CrossRef][Medline]

NGUYEN, M., S. PARK, G. MARQUES, and K. ARORA, 1998  Interpretation of a BMP activity gradient in Drosophila embryos depends on synergistic signaling by two type I receptors, SAX and TKV. Cell 95:495-506.[CrossRef][Medline]

NOLL, R., M. A. STURTEVANT, R. R. GOLLAPUDI, and E. BIER, 1994  New functions of the Drosophila rhomboid gene during embryonic and adult development are revealed by a novel genetic method, enhancer piracy. Development 120:2329-2338.[Abstract]

OELGESCHLAGER, M., J. LARRAIN, D. GEISSERT, and E. M. DE ROBERTIS, 2000  The evolutionarily conserved BMP-binding protein Twisted gastrulation promotes BMP signalling. Nature 405:757-763.[CrossRef][Medline]

O'NEILL, J. W. and E. BIER, 1994  Double-label in situ hybridization using biotin and digoxigenin-tagged RNA probes. Biotechniques 17:870-875.

PADGETT, R. W., R. D. ST. JOHNSTON, and W. M. GELBART, 1987  A transcript from a Drosophila pattern gene predicts a protein homologous to the transforming growth factor-beta family. Nature 325:81-84.[CrossRef][Medline]

PADGETT, R. W., J. M. WOZNEY, and W. M. GELBART, 1993  Human BMP sequences can confer normal dorsal-ventral patterning in the Drosophila embryo. Proc. Natl. Acad. Sci. USA 90:2905-2909.[Abstract/Free Full Text]

PICCOLO, S., Y. SASAI, B. LU, and E. M. DE ROBERTIS, 1996  Dorsoventral patterning in Xenopus: inhibition of ventral signals by direct binding of Chordin to BMP-4. Cell 86:589-598.[CrossRef][Medline]

PICCOLO, S., E. AGIUS, B. LU, S. GOODMAN, and L. DALE et al., 1997  Cleavage of Chordin by Xolloid metalloprotease suggests a role for proteolytic processing in the regulation of Spemann organizer activity. Cell 91:407-416.[CrossRef][Medline]

RAY, R. P. and K. A. WHARTON, 2001  Context-dependent relationships between the BMPs gbb and dpp during development of the Drosophila wing imaginal disc. Development 128:3913-3925.

ROSS, J. J., O. SHIMMI, P. VILMOS, A. PETRYK, and H. KIM et al., 2001  Twisted gastrulation is a conserved extracellular BMP antagonist. Nature 410:479-483.[CrossRef][Medline]

SASAI, Y., B. LU, H. STEINBEISSER, D. GEISSERT, and L. K. GONT et al., 1994  Xenopus chordin: a novel dorsalizing factor activated by organizer-specific homeobox genes. Cell 79:779-790.[CrossRef][Medline]

SASAI, Y., B. LU, H. STEINBEISSER, and E. M. DE ROBERTIS, 1995  Regulation of neural induction by the Chd and Bmp-4 antagonistic patterning signals in Xenopus.. Nature 376:333-336.[CrossRef][Medline]

SCHMIDT, J., V. FRANCOIS, E. BIER, and D. KIMELMAN, 1995  Drosophila short gastrulation induces an ectopic axis in Xenopus: evidence for conserved mechanisms of dorsal-ventral patterning. Development 121:4319-4328.[Abstract]

SCOTT, I. C., I. L. BLITZ, W. N. PAPPANO, S. A. MAAS, and K. W. CHO et al., 2001  Homologues of Twisted gastrulation are extracellular cofactors in antagonism of BMP signalling. Nature 410:475-478.[CrossRef][Medline]

SEGAL, D. and W. M. GELBART, 1985  shortvein, a new component of the decapentaplegic gene complex in Drosophila melanogaster.. Genetics 109:119-143.[Abstract/Free Full Text]

SHIMMI, O. and M. B. O'CONNOR, 2003  Physical properties of Tld, Sog, Tsg and Dpp protein interactions are predicted to help create a sharp boundary in Bmp signals during dorsoventral patte