ABSTRACT
decapentaplegic (dpp) is a Transforming Growth Factor beta (TGF-β) -related growth factor that controls multiple developmental processes in Drosophila. To identify components involved in dpp signaling, we carried out a genetic screen for dominant enhancer mutations of a hypomorphic allele of thick veins (tkv), a type I receptor for dpp. We recovered new alleles of tkv, punt, Mothers against dpp (Mad) and Medea (Med), all of which are known to mediate dpp signaling. We also recovered mutations in the 60A gene which encodes another TGF-β -related factor in Drosophila. DNA sequence analysis established that all three 60A alleles were nonsense mutations in the prodomain of the 60A polypeptide. These mutations in 60A caused defects in midgut morphogenesis and fat body differentiation. We present evidence that when dpp signaling is compromised, lowering the level of 60A impairs several dpp-dependent developmental processes examined, including the patterning of the visceral mesoderm, the embryonic ectoderm and the imaginal discs. These results provide the first in vivo evidence for the involvement of 60A in the dpp pathway. We propose that 60A activity is required to maintain optimal signaling capacity of the dpp pathway, possibly by forming biologically active heterodimers with Dpp proteins.
INTRODUCTION
The Transforming Growth Factor-beta (TGF-β ) superfamily is a family of conserved polypeptide growth factors that regulate diverse biological activities (reviewed by Kingsley, 1994a; Massagué, 1996). In particular, within the TGF-β superfamily, the Bone Morphogenetic Proteins (BMPs) are critical regulators of cell proliferation, cell death, cell fate specification and organogenesis (reviewed by Kingsley, 1994b; Hogan, 1996). The versatile signaling capacity of TGF-β -related factors partially stems from extensive posttranslational modifications that occur during the maturation process. These factors are synthesized as large precursor molecules, which undergo homomeric or heteromeric dimerization and subsequent proteolytic cleavage to yield the bioactive carboxy terminal portion (Roberts and Sporn, 1990). Upon secretion, their association with extracellular binding proteins is important for regulating their access to cell surface receptors (reviewed by Miyazono et al., 1993). The interaction with a heteromeric transmembrane receptor system composed of two distinct serine/threonine kinases (type I and II receptors) adds one more level of complexity to TGF-β -related signaling (Massagué and Weis-Garcia, 1996). In order to understand the complex processes involved in TGF-β signaling, we focus on decapentaplegic (dpp), a Drosophila gene that is functionally interchangeable with mammalian BMP-2 and BMP-4 (Padgett et al., 1993; Sampath et al., 1993).
dpp has a dynamic expression pattern and multiple functions throughout Drosophila development. At the blastoderm stage, dpp transcripts are restricted dorsally (St. Johnston and Gelbart, 1987) where it functions as a morphogen to specify distinct dorsal structures (Ferguson and Anderson, 1992a,b; Wharton et al., 1993). Later during embryogenesis, dpp is expressed in the ectoderm (Jackson and Hoffmann, 1994), where it induces dorsal mesoderm differentiation (Staehling-Hampton et al., 1994). In the visceral mesoderm, dpp is expressed in discrete domains (Panganiban et al., 1990b) to regulate the expression of several homeotic genes in different tissue layers (Bienz, 1994). dpp is also expressed in specific positions in the larval imaginal discs (Masucci et al., 1990) to control the proliferation and patterning of adult appendages (reviewed by Brook et al., 1996).
In addition to dpp, 60A and scw are two other BMP-related genes in Drosophila (Arora et al., 1994; Doctor et al., 1992; Wharton et al., 1991). Scw proteins have been proposed to enhance dpp activity during early embryogenesis by forming heterodimers with Dpp (Arora et al., 1994). The role of 60A in development was unclear due to the lack of knowledge of the phenotypic consequences of disrupting 60A function.
In Drosophila, two type I receptors encoded by saxophone (sax) and thick veins (tkv) and one type II receptor encoded by punt have been shown to be functional dpp receptors (Penton et al., 1994; Brummel et al., 1994; Nellen et al., 1994; Ruberte et al., 1995; Letsou et al., 1995; Xie et al., 1994). Mothers against dpp (Mad), Medea (Med) and schnurri (shn) were identified through genetic interactions with dpp (Raftery et al., 1995; Sekelsky et al., 1995; Staehling-Hampton et al., 1995; Grieder et al., 1995; Arora et al., 1995). shn encodes a protein related to human zinc finger transcription factor PRDII/MBPI/HIV-EP1 (Staehling-Hampton et al., 1995; Arora et al., 1995; Grieder et al., 1995). Mad-related proteins (Smads) have been isolated from a wide range of distantly related organisms. Genetic and biochemical evidence indicated that Smads are key signal transducers, linking events between receptor activation and changes in target gene expression (Derynck and Zhang, 1996; Massagué, 1996; Wrana and Pawson, 1997).
The dosage-sensitive nature of dpp signaling prompted us to use modifier genetics to identify additional components in the dpp pathway. This approach exploits synergistic interactions between components in the same biological pathway. We sensitized the dpp pathway using a hypomorphic dpp receptor, tkv6, which had a mild visible phenotype. We reasoned that, in this genetic background where dpp signaling is below optimal level, a two-fold reduction in the activities of other signaling components in the pathway would produce a modified phenotype. So we screened for mutations that dominantly modified tkv6 phenotype. Such an approach has successfully identified components in several Drosophila signal transduction pathways, including the sevenless receptor tyrosine kinase pathway (Simon et al., 1991), the Abelson cytoplasmic tyrosine kinase pathway (Gertler et al., 1990) and the dpp pathway (Raftery et al., 1995).
In our screen, we recovered new alleles of tkv, punt, Mad and Med; all are known to mediate dpp signaling. Two other complementation groups were identified that potentially represent new components in the pathway. Most significantly, mutations in 60A, the Drosophila homologue of BMP-7, were recovered as enhancers for the sensitized dpp pathway. We describe the loss-of-function phenotypes of 60A and present the first in vivo evidence that 60A acts synergistically with dpp in several developmental processes.
MATERIALS AND METHODS
Drosophila stocks
Drosophila stocks were cultured on standard cornmeal yeast extract sugar medium at 25°C. Canton S. was used as the wild-type stock. sax5 was described in Twombly et al. (1996). MadP is described in Sekelsky et al. (1995). Med4 was described in Raftery et al. (1995). All other mutants and chromosomes are referenced in Flybase (http://cbbridges.harvard.edu).
Isolation and analysis of enhancer mutations of tkv6
The initial attempts to recover modifiers of tkv6 in an F1 screen was not successful because of greatly reduced viability and fertility of the flies with enhanced phenotypes. Thus an F2 screen was carried out.
pr cn was recombined onto the tkv6 chromosome and the stock was made isogenic for the second and the third chromosomes. tkv6pr cn/CyO males were mutagenized with 3.4 mM ethylnitrosourea (ENU) and mated to females with a translocation between CyO and TM6,B to force the co-segregation of the second and third chromosome. Individual male progeny were mated to tkv6/CyO; TM2/TM6,B females. Enhancement of the phenotypes of the imaginaldisc-derived structures were screened in the progeny homozygous for tkv6. The enhancer mutations were recovered from the siblings of the enhanced progeny and crossed to tkv6/CyO; TM2/TM6,B females for retesting and to establish balanced stocks.
The number of complementation groups was determined from inter se crosses among enhancers. For enhancers on the second chromosome, a tkv transgene on the third chromosome (Y. C. and F. M. H., unpublished data) was used to compensate for the tkv6 mutation present on the second chromosome. Allelism with known mutations was established by genetic non-complementation and by meiotic and deletion mapping. The Sp Bl LrmBc Pu2PinB chromosome was used for meiotically mapping enhancers on the second chromosome.
Sequencing of mutant alleles
Total RNAs were isolated from heterozygous Mad and punt females and heterozygous 60A males using the Tri Reagent (Molecular Research Center, Inc.). Gene-specific cDNAs were reverse transcribed and amplified using the Superscript Preamplification System (GIBCO BRL). The PCR products were subcloned into TA cloning vectors (Invitrogen) and sequenced on an automated sequencer (ABI 373). Sequencing multiple alleles of the same gene allowed identification of the polymorphisms specific to the mutagenized chromosomes. For 60AD4 and 60AD8, the genomic region of 60A was sequenced in a similar fashion because the mutant cDNAs were under-represented.
Preparation of larval gut, cuticle and adult appendages
Larval gut was dissected and mounted according to Masucci and Hoffmann (1993). Cuticles were prepared as described previously (Struhl, 1989). Wings and legs were mounted in Gary’s magic mounting media (Ashburner, 1989).
Antibody staining and identification of mutant embryos
Anti-Scr, Anti-Ubx, Anti-Lab antibodies were gifts of Dr Matthew Scott, Stanford University. Anti-Wg antibody was a gift of Dr Roel Nusse, Stanford University. Anti-Dpp antibody was described in Panganiban et al. (1990a). Antibody stainings were done as previously described (Panganiban et al., 1990b; Staehling-Hampton and Hoffmann, 1994; Reuter et al., 1990). All stocks used for antibody staining were balanced over a CyO chromosome with an elav-lacZ enhancer trap to allow unambiguous identification of mutant embryos.
RESULTS
shn and punt enhance tkv6 phenotypes
tkv6 is a mutation in a splice acceptor site that results in aberrant in-frame splicing, deleting two extracellular amino acids of the receptor. When expressed in COS1 cells, the mutant receptor fails to bind BMP-2 homodimers (Penton et al., 1994). However, tkv6 behaves genetically as a hypomorph. In contrast to the embryonic lethal tkv null alleles, tkv6 is homozygous viable and the only visible phenotype is the thickened wing veins (Fig. 1B). All other imaginal-discderived structures of tkv6 homozygotes appear normal (Figs 1C, 3A,C). Interestingly, tkv6/Df(2L)tkv2 flies are phenotypically identical to tkv6 homozygotes (data not shown). To test if tkv6 is a suitable genetic background for a modifier screen, we examined the effects of lowering the activity of other known dpp pathway components. We found that heterozygous mutations in shn or punt enhanced the tkv6 homozygous phenotype. In the tkv6 background, shnIB was a dominant enhancer of the venation pattern in the wing (Fig. 1D) and the proximal/distal patterning of the leg (Fig. 1E). In the wing, longitudinal vein 2 failed to reach the wing margin (Fig. 1D). In the leg, distal elements such as claws and distal tarsal segments were deleted (Fig. 1E). Such phenotypes were reminiscent of hypomorphic dpp phenotypes (Spencer et al., 1982). punt135 also enhanced the tkv6 phenotypes (data not shown). Based on these observations, we reasoned that the dpp signaling output through the mutant receptor tkv6 was near the threshold for proper patterning of the imaginal discs. It was therefore an appropriate genetic background for identifying new components essential for mediating dpp signaling.
shn enhanced tkv6 homozygous phenotypes. (A) Wild-type wing. (B) tkv6 wing. Thickening of the cross veins and the terminals of the longitudinal veins were evident (arrowhead). (D) tkv6shnIB/tkv6 wing. Besides the thickened wing veins, the longitudinal vein 2 was truncated (arrowhead). (C) Phenotypically normal tkv6 mesothoracic leg. (E) tkv6shnIB/tkv6 mesothoracic leg. The distal tarsal segments and the claws were missing (arrowhead). fm, femur; tb, tibia; ts, tarsal segments; cl, claws.
shn enhanced tkv6 homozygous phenotypes. (A) Wild-type wing. (B) tkv6 wing. Thickening of the cross veins and the terminals of the longitudinal veins were evident (arrowhead). (D) tkv6shnIB/tkv6 wing. Besides the thickened wing veins, the longitudinal vein 2 was truncated (arrowhead). (C) Phenotypically normal tkv6 mesothoracic leg. (E) tkv6shnIB/tkv6 mesothoracic leg. The distal tarsal segments and the claws were missing (arrowhead). fm, femur; tb, tibia; ts, tarsal segments; cl, claws.
Enhancers of tkv are phenotypically similar to dpp mutants
The modifier screen was conducted as outlined in Fig. 2. The F2 progeny were screened for enhanced phenotypes in the imaginal-disc-derived structures. Over 10,000 mutagenized genomes were screened and fourteen dominant enhancers defining seven loci were recovered (Table 1). The enhancers were recessive lethal in a wild-type background. tkv6 homozygotes that are heterozygous for the enhancer mutations had defects in imaginal disc development (Fig. 3). During pupal development, the dorsal proximal region of the two wing imaginal discs fuse to form the adult notum. In tkv6 flies, the notum appeared normal with a smooth contour and orderly oriented sensory bristles (Fig. 3A). This pattern was disrupted by heterozygous D1 mutation, resulting in a medial cleft in the notum, with abnormally parted bristles on both sides of the cleft (Fig. 3B). tkv6 flies had normally patterned legs (Fig. 3C). However, heterozygous D4 mutation caused deletions of distal and dorsal structures (Fig. 3D,E) and occasional duplication of ventrolateral structures such as sex combs on male prothoracic legs (Fig. 3E). These phenotypes were indistinguishable from those of dppdisk alleles (Spencer et al., 1982), suggesting that these enhancers act in the dpp signal transduction pathway.
Scheme for the F2 enhancer screen of tkv6. See Materials and Methods for detailed description of the procedure. ENU, Ethylnitrosourea; T(2;3), a translocation between CyO and TM6,B. Asterisks indicate the mutagenized chromosomes.
Heterozygous enhancer mutations enhanced tkv6 phenotypes. (A) Phenotypically normal tkv6 notum. (B) tkv6D1/tkv6 notum. The sensory bristles were parted to both sides and there was a profound medial cleft (arrow). The scutellum was often reduced in size (compare with A). (C) Phenotypically normal tkv6 male prothoracic leg. (D) tkv660AD4/tkv6 male prothoracic leg. The most distal tarsal segments and claws were truncated (arrow) and the ventral lateral sex combs were duplicated (arrow head). (E) tkv660AD4/tkv6 metathoracic leg. Note the severe truncation of distal structures and the curved appearance of the leg caused by the loss of dorsal tissues (arrow). fm, femur; tb, tibia; sc, sex combs; ts, tarsal segments; cl, claws.
Heterozygous enhancer mutations enhanced tkv6 phenotypes. (A) Phenotypically normal tkv6 notum. (B) tkv6D1/tkv6 notum. The sensory bristles were parted to both sides and there was a profound medial cleft (arrow). The scutellum was often reduced in size (compare with A). (C) Phenotypically normal tkv6 male prothoracic leg. (D) tkv660AD4/tkv6 male prothoracic leg. The most distal tarsal segments and claws were truncated (arrow) and the ventral lateral sex combs were duplicated (arrow head). (E) tkv660AD4/tkv6 metathoracic leg. Note the severe truncation of distal structures and the curved appearance of the leg caused by the loss of dorsal tissues (arrow). fm, femur; tb, tibia; sc, sex combs; ts, tarsal segments; cl, claws.
Additional evidence that the enhancers mediate dpp signaling came from their dosage-sensitive interactions with known mutations in the dpp pathway, including dpps5, tkv6, Df(2L)tkv2, sax5, punt135, MadP, Med4 and shnIB (Table 2). In many cases, the enhancers failed to fully complement these mutations and showed imaginal disc development defects ranging from gaps in wing veins to the notum and leg phenotypes described in Fig. 3. tkv6 alone showed no detectable heterozygous interactions with the dpp pathway mutations examined except for the thickened venation phenotype when in trans to tkv6 or Df(2R)tkv2, indicating that phenotypes observed were due to the presence of the enhancer mutations.
Dominant enhancers of tkv6: new alleles of tkv, Mad, Med, punt and 60A
Meiotic mapping and complementation tests established seven complementation groups for the enhancers. As expected based on the initial evaluation of the tkv6 genetic background, new alleles of tkv, Mad, Med and punt were recovered (Table 1). We sequenced the coding regions of the new Mad and punt alleles to establish their molecular identity. Of the five Mad alleles, three had point mutations in the coding region (Fig. 4A). Mis-sense mutations were found in both new punt alleles (Fig. 4B). The tkvD17 and MedD5 allelism was based on genetic non-complementation. In addition, it was found that a tkv transgene rescued the lethality of tkvD17 homozygotes, supporting the view that D17 is a tkv allele (data not shown).
Molecular lesions in new alleles of Mad, punt and 60A. The position and nature of the mutations are indicated below the schematic diagram of the protein. Asterisk, nonsense mutation. (A) Mutations in new Mad alleles. The hatched box represents the mutation hot spot. (B) Mutations in new punt alleles. The structural features shown are the putative signal peptide (oval box), the extracellular cysteine residues (vertical bars), the transmembrane domain (hatched box) and the kinase domain (open box). (C) Mutations in 60A alleles. Letter Cs within the mature domain represent conserved cysteine residues.
Molecular lesions in new alleles of Mad, punt and 60A. The position and nature of the mutations are indicated below the schematic diagram of the protein. Asterisk, nonsense mutation. (A) Mutations in new Mad alleles. The hatched box represents the mutation hot spot. (B) Mutations in new punt alleles. The structural features shown are the putative signal peptide (oval box), the extracellular cysteine residues (vertical bars), the transmembrane domain (hatched box) and the kinase domain (open box). (C) Mutations in 60A alleles. Letter Cs within the mature domain represent conserved cysteine residues.
Genetic and molecular characterizations of the D4 complementation group revealed that it corresponded to the 60A gene. 60A encodes a BMP-7 homologue isolated based on sequence homology (Doctor et al., 1992; Wharton et al., 1991). Its function remained unknown due to the lack of mutations in 60A. Three alleles of 60A were confirmed by sequencing the mutant alleles. 60AD8 and 60AD20 are nonsense mutations in the prodomain due to single nucleotide substitutions. 60AD4 has one nucleotide deletion, causing a frame-shift premature stop also in the prodomain (Fig. 4C).
Loss-of-function phenotypes of 60A
Animals lacking 60A function died at late larval/early pupal stages. One of the striking phenotypes of 60A mutant larvae is a transparent appearance due to the lack of fat body (Fig. 5B). In roughly 50% of the 60A larvae, the gastric caecae were reduced in length (Fig. 5D), consistent with the expression of 60A in the gastric caecae (Doctor et al., 1992). These phenotypes are similar to those of the Mad mutant larvae (Sekelsky et al., 1995).
Larval phenotypes of 60A. (A) Wild-type third instar larva. (C) 60AD4/60AD8 third instar larva had greatly reduced fat body and appeared transparent. (B) Wild-type third instar larval gut. Note the long and extended gastric caecae (arrowhead). (D) 60AD4/60AD8 third instar larval gut. The gastric caecae were short in about half of the 60A larvae (n=47) (arrowhead). Compared with the wild-type larva of the same stage, 60A mutants grew more slowly and were somewhat reduced in size.
Larval phenotypes of 60A. (A) Wild-type third instar larva. (C) 60AD4/60AD8 third instar larva had greatly reduced fat body and appeared transparent. (B) Wild-type third instar larval gut. Note the long and extended gastric caecae (arrowhead). (D) 60AD4/60AD8 third instar larval gut. The gastric caecae were short in about half of the 60A larvae (n=47) (arrowhead). Compared with the wild-type larva of the same stage, 60A mutants grew more slowly and were somewhat reduced in size.
During embryogenesis, 60A is expressed throughout the visceral mesoderm of the developing midgut (Doctor et al., 1992) suggesting a function for 60A in gut development. Indeed, embryos lacking 60A failed to form the first constriction (Fig. 6E,F). The homeotic gene Antennapedia (Antp) is expressed in the visceral mesoderm around the first constriction and is required for its formation (Reuter and Scott, 1990; Tremml and Bienz, 1989). We examined the Antp expression in 60A mutant embryos. Consistent with the lack of the first constriction, Antp expression was greatly reduced (Fig. 6B). Thus, 60A is required for the formation of the first constriction of the midgut, likely through positively regulating the expression of Antp in the visceral mesoderm.
60A mutants lacked the first constriction of the embryonic midgut. Lateral views of embryos stained with anti-Antp antibody (A-C) and anti-Lab antibody (D-F); stage 14 (A, B), stage 15 (C), stage 16 (D-F). Anterior is to the left and dorsal is up for this and subsequent figures. (A) Phenotypically normal visceral mesoderm expression of Antp in ps6 in tkv6 homozygous embryos (bracket). (B) 60AD4/60AD8 embryos lacked ps6 Antp expression (bracket). (C) tkv660AD4/tkv660AD8 embryos lacked Antp expression in ps6 (bracket) but acquired an ectopic Antp domain in ps7 (arrowheads) presumably due to lacking Ubx expression in ps7 (see Fig. 7L). (D) tkv6 homozygotes had normal lab expression (arrowhead) in the endoderm and formed three constrictions indicated by numbers. (E) 60AD4/60AD8 embryos failed to form the first constriction (asterisk) but the second and third constrictions still formed. lab expression was not altered (arrowhead). (F) tkv660AD4/tkv660AD8 embryos lacked lab expression (arrowhead) in the endoderm and only formed one constriction.
60A mutants lacked the first constriction of the embryonic midgut. Lateral views of embryos stained with anti-Antp antibody (A-C) and anti-Lab antibody (D-F); stage 14 (A, B), stage 15 (C), stage 16 (D-F). Anterior is to the left and dorsal is up for this and subsequent figures. (A) Phenotypically normal visceral mesoderm expression of Antp in ps6 in tkv6 homozygous embryos (bracket). (B) 60AD4/60AD8 embryos lacked ps6 Antp expression (bracket). (C) tkv660AD4/tkv660AD8 embryos lacked Antp expression in ps6 (bracket) but acquired an ectopic Antp domain in ps7 (arrowheads) presumably due to lacking Ubx expression in ps7 (see Fig. 7L). (D) tkv6 homozygotes had normal lab expression (arrowhead) in the endoderm and formed three constrictions indicated by numbers. (E) 60AD4/60AD8 embryos failed to form the first constriction (asterisk) but the second and third constrictions still formed. lab expression was not altered (arrowhead). (F) tkv660AD4/tkv660AD8 embryos lacked lab expression (arrowhead) in the endoderm and only formed one constriction.
60A maintains an optimal level of dpp signaling in the visceral mesoderm
The identification of mutations in 60A as dominant enhancers of tkv6, thus dpp signaling, in the imaginal discs raised the possibility that 60A is required for optimal signaling by the dpp pathway. To determine if there was a general requirement for 60A in dpp signaling, we examined the effects of 60A mutations on dpp signaling in the visceral mesoderm where both dpp and 60A are expressed.
dpp is expressed in two discrete domains in the visceral mesoderm (Panganiban et al., 1990b). The anterior domain of dpp coincides with the gastric caecae primordia, which are immediately anterior to the expression domain of Sex combs reduced (Scr) in parasegment (ps) 4. The failure to initiate dpp expression in ps3 in dppshv mutants results in anterior expansion of Scr expression and arrested outgrowth of the gastric caecae (Panganiban et al., 1990b; Hursh et al., 1993), indicating a role for dpp in repressing Scr in ps3. tkv6 homozygotes are homozygous viable, so it is not surprising that all the midgut gene expression patterns examined were essentially normal (Fig. 7A-D). Scr expression in tkv6 and 60A mutants was normal (Fig. 7A,E). However, in tkv6 and 60A double mutants, the Scr expression extended anteriorly into ps3 (Fig. 7I) as it did in dppshv mutants, suggesting that 60A activity is required in ps3 for optimal dpp signaling.
60A enhanced tkv6 midgut phenotypes. Lateral views of embryos stained with anti-Scr antibody (A, E, I); anti-Dpp antibody (B, F, J); anti-Wg antibody (C, G, K) and anti-Ubx antibody (D, H, L). A-D, tkv6 homozygotes; E-H, 60AD4/60AD8 embryos; I-L, tkv660AD4/tkv660AD8 embryos. C, G and K were at stage 14; all other embryos were at stage 15. Scr expression in ps4 was normal in tkv6 (A, bracket) or 60A mutants (E, bracket). In tkv660A embryos, Scr extended anteriorly into ps3 (I, bracket). The visceral mesoderm expression of dpp in the single mutants were normal (B, F, arrows, gastric caecae; arrowheads, ps7). Double mutants had greatly reduced expression of dpp in gastric caecae (J, arrow) and in ps7 (J, arrowhead). dpp expression in the other domains were unaffected. wg was expressed normally in the single mutants (C, G, arrowhead), however, ps8 expression is undetectable in the double mutants (K, arrowhead). Ubx expression in ps7 was normal in tkv6 (D, bracket) or 60A mutants (H, bracket), but was greatly reduced in the double mutants (L, bracket).
60A enhanced tkv6 midgut phenotypes. Lateral views of embryos stained with anti-Scr antibody (A, E, I); anti-Dpp antibody (B, F, J); anti-Wg antibody (C, G, K) and anti-Ubx antibody (D, H, L). A-D, tkv6 homozygotes; E-H, 60AD4/60AD8 embryos; I-L, tkv660AD4/tkv660AD8 embryos. C, G and K were at stage 14; all other embryos were at stage 15. Scr expression in ps4 was normal in tkv6 (A, bracket) or 60A mutants (E, bracket). In tkv660A embryos, Scr extended anteriorly into ps3 (I, bracket). The visceral mesoderm expression of dpp in the single mutants were normal (B, F, arrows, gastric caecae; arrowheads, ps7). Double mutants had greatly reduced expression of dpp in gastric caecae (J, arrow) and in ps7 (J, arrowhead). dpp expression in the other domains were unaffected. wg was expressed normally in the single mutants (C, G, arrowhead), however, ps8 expression is undetectable in the double mutants (K, arrowhead). Ubx expression in ps7 was normal in tkv6 (D, bracket) or 60A mutants (H, bracket), but was greatly reduced in the double mutants (L, bracket).
To test whether 60A also acts synergistically with dpp elsewhere in the midgut, we examined the gene expression of dpp and Ultrabithorax (Ubx) in ps7 and wingless (wg) in the adjacent ps8. It has been established that ps7 expression of dpp is activated by the homeotic gene Ubx (Immerglück et al., 1990; Panganiban et al., 1990b; Reuter et al., 1990; Capovilla et al., 1994; Sun et al., 1995) and maintained by an autostimulatory circuit involving dpp, Ubx and wg (Hursh et al., 1993; Thüringer et al., 1993a,b; Staehling-Hampton and Hoffmann, 1994). The proper expression of all three genes is interdependent and critical for maintaining a stable cellular differentiation commitment (Bienz, 1994). The expression of dpp and wg in the visceral mesoderm is required for the induction of the homeotic gene labial (lab) in the underlying endoderm (Immerglück et al., 1990; Panganiban et al., 1990b; Reuter et al., 1990). The absence of dpp function in ps7 disrupts the autoregulatory loop and reduces the expression of Ubx, wg, dpp and lab in ps7, leading to the absence of the second constriction in dpp mutant embryos (Immerglück et al., 1990; Panganiban et al., 1990b; Reuter et al., 1990).
We found that animals mutant for either tkv6 or 60A had normal expression of dpp, wg and Ubx (Fig. 7B-D,F-H). However, in tkv660A double mutants, dpp expression in ps3 and ps7 was greatly reduced (Fig. 7J). The initiation of dpp expression at earlier stages was not affected in the double mutants (not shown), suggesting that the reduction of dpp expression resulted from failure to maintain its expression at later stages. Similarly, in Mad mutants, the initiation of dpp expression in ps3 and ps7 is not affected but the maintenance of dpp expression does not occur (Newfeld et al., 1997). This is because dpp expression is activated directly by Ubx and only its maintenance requires positive feedback involving dpp signaling. In the double mutants, Ubx expression in ps7 and wg expression in ps8 were greatly reduced (Fig. 7K,L), suggesting the disruption of the positive regulatory loop. The reduction of dpp in ps3 in the double mutants may explain the observed derepression of Scr.
Ubx is required for repressing Antp in ps6. In Ubx mutants, the Antp domain is extended posteriorly into ps7 (Tremml and Bienz, 1989), indicating a homeotic transformation of ps7 into ps6. A similar phenotype was observed for tkv null embryos (Affolter et al., 1994). Consistent with the argument that lacking 60A compromises dpp signaling, there was also a posterior expansion of Antp in tkv660A double mutants (Fig. 6C). Interestingly, due to the 60A mutation, the endogenous Antp expression was absent, such that there was only ectopic Antp in ps7, where Ubx would normally be.
We also examined the expression of lab in the endoderm. Consistent with the gene expression changes in the visceral mesoderm, lab expression was not affected by tkv6 or 60A mutations. However, it was greatly reduced in tkv660A double mutants (Fig. 6F). The gut of the double mutants only formed two chambers instead of the normal four chambers (Fig. 6G, compare to 6D). This phenotype likely resulted from the failure to form the first constriction due to lacking 60A function and the failure to form the second constriction due to lacking dpp signaling. It is unclear why the position of the only constriction observed in the double mutants is somewhat more anterior than a normal third constriction.
The gene expression changes in the midgut of the tkv660A double mutants are consistent with 60A playing a role in augmenting dpp signaling.
60A enhances the ectodermal phenotypes of tkv6 homozygotes
Previous studies have established dpp’s role as a morphogen in patterning the embryonic ectoderm (Ferguson and Anderson, 1992a,b; Wharton et al., 1993). dpp signaling is also required for dorsal closure of the embryonic ectoderm (Hou et al., 1997; Riesgo-Escovar and Hafen, 1997). We therefore examined if the level of 60A affected the phenotype of the embryonic ectoderm.
We compared the cuticle phenotypes of single and double mutants. Since tkv6 homozygotes were viable and 60A mutants had no obvious defects until late in development, the cuticular patterns of these mutants were essentially normal (Fig. 8A,B). However, tkv660A homozygote embryos died and exhibited head defects and an excessive ventral curvature (Fig. 8C). Although the double mutant cuticles bore some resemblance to hypomorphic dpp mutants, they did not exhibit an obvious expansion of the ventral denticle belts (Wharton et al., 1993). The double mutant phenotype suggested, however, that, when dpp signaling was compromised in the embryonic ectoderm, removing 60A activity further attenuated dpp signaling. We considered that the relatively mild phenotype of the double mutant embryo might reflect partial rescue by maternally contributed wild-type Tkv receptors. Indeed, a quarter of the embryos produced by mothers homozygous for tkv6 and heterozygous for 60A exhibited a dorsal open phenotype similar (Fig. 8D) to that of zygotic tkv null embryos (Penton et al., 1994). Therefore, in the absence of maternally provided wild-type Tkv, tkv660A double mutant embryos exhibit a phenotype indicative of defective dpp signaling during the process of dorsal closure.
60A enhanced the tkv6 cuticle phenotypes. Phenotypically normal tkv6 (A) and 60AD4/60AD8 (B) cuticle. Note the fully internalized head skeleton (arrow). (C) tkv660AD4/tkv660AD8 cuticle. The partially deleted head skeletons remained external (arrow). (D) An embryo produced by tkv660AD4/tkv6 females mated to tkv660AD8/tkv6 males. Roughly a quarter of the embryos lacked dorsal hypoderm (arrow).
60A enhanced the tkv6 cuticle phenotypes. Phenotypically normal tkv6 (A) and 60AD4/60AD8 (B) cuticle. Note the fully internalized head skeleton (arrow). (C) tkv660AD4/tkv660AD8 cuticle. The partially deleted head skeletons remained external (arrow). (D) An embryo produced by tkv660AD4/tkv6 females mated to tkv660AD8/tkv6 males. Roughly a quarter of the embryos lacked dorsal hypoderm (arrow).
DISCUSSION
The haploinsufficiency of the dpp locus reflects the sensitivity of developmental processes to a reduction in dpp signaling. We have carried out a genetic screen to search for modifiers of tkv6, a hypomorphic type I dpp receptor. tkv has been implicated in all aspects of dpp signaling both in vitro and in vivo (Penton et al., 1994; Nellen et al., 1994; Brummel et al., 1994; Affolter et al., 1994; Burke and Basler, 1996; Singer et al., 1997). Therefore, the modifiers of tkv6 are most likely to be integral components of the dpp signal transduction pathway.
Identification of new alleles of tkv, Mad, Med, punt, 60A and two new loci as dominant enhancers of tkv6
New alleles of several genes known to mediate dpp signaling were identified, including tkv, punt, Mad and Med. Recovery of these mutations as dominant enhancers of tkv6 validated the specificity of the screen.
We isolated five alleles of Mad, a key signal transducer in dpp signaling (Newfeld et al., 1996; Kim et al., 1997). Three of them have point mutations in the coding region (Fig. 4A). The molecular lesions correlate with their phenotypic properties. MadD14 has a nonsense mutation predicted to produce a truncated protein with only the conserved MH1 domain. It behaves genetically as a null. MadD16 has a tyrosine-to-asparagine change in the divergent linker region and behaves as a hypomorph with residual activity (Table 2). This suggests that the amino acid change only partially affects the protein function. The MH2 domains of Smads are highly conserved.
The three-dimensional structure of the MH2 domain of Smad4 indicates that Smads form homotrimers whose intact conformation is essential for the assembly of a hexamer with a different Smad homotrimer in response to receptor activation (Hata et al., 1997; Shi et al., 1997). Many of the Smad mutations associated with tumors or affecting development map to the MH2 domain. Based on the crystal structure of the carboxy domain of Smad4 (Shi et al., 1997), the invariant aspartic acid mutated to asparagine in MadD24 maps to the trimer interface region critical for trimerization. The corresponding residue in Smad2 is mutated in colon cancers (Eppert et al., 1996). Interestingly, the MadD24 mutation resulted in dominant female sterility (Y. C. and F. M. H., unpublished data) which is not observed with Mad null alleles, suggesting that it has a dominant negative effect, perhaps by forming inactive oligomers with the wild-type proteins in the heterozygotes.
Analysis of the two new punt alleles also provides evidence for the in vivo importance of conserved structural motifs in this type II dpp receptor (Fig. 4B). puntD13 has a cysteine-totyrosine change in the extracellular cysteine cluster characteristic of all receptors for TGF-β superfamily members (Massagué et al., 1994). puntD18 changes the highly conserved glutamic acid to a valine in the catalytic core of the kinase domain, where another punt mutation, punt135, is mapped (Ruberte et al., 1995). Like punt135, both new punt alleles display some temperature sensitivity (Y. C. and F. M. H., unpublished data), suggesting that they are not protein nulls. No null mutations in the punt locus have been reported, suggesting that like dpp, punt may be haploinsufficient.
We did not isolate any new alleles of shn, which enhanced tkv6 in the initial test. The enhancement by shnIB may be allele specific, such that a particular form of mutant Shn protein is needed to produce an enhancement. Consistent with this, shnp, which makes no detectable protein (Staehling-Hampton et al., 1995), failed to enhance the tkv6 phenotype (data not shown). No dpp alleles were recovered either, possibly due to the haploinsufficiency associated with the locus and the fact that most hypomorphic dpp mutations affect regulatory regions, which are less likely to be affected by chemical mutagens such as ENU.
One unexpected locus identified in our screen is 60A, which encodes the BMP-7 homologue (Doctor et al., 1992; Wharton et al., 1991). The fact that in a screen of the entire genome, 60A mutations were recovered multiple times as dominant enhancers of a mutant dpp receptor provides strong evidence for its involvement in dpp signaling. Nonsense mutations were found in all three alleles of 60A in the prodomain of the precursor protein. Since these mutations are predicted to eliminate translation of the biologically active mature C-terminal domain, they most likely represent functional nulls of the 60A gene.
The developmental functions of 60A and its role in dpp signaling
Phenotypic analysis of 60A single mutants and tkv660A double mutants revealed both dpp-independent and dpp-dependent functions for 60A. 60A is expressed broadly throughout development, with enrichment in the developing gut (Doctor et al., 1992), suggesting a role for 60A in gut morphogenesis. 60A mutants lack the first constriction of the embryonic midgut and Antp expression in ps6, indicating that 60A is required for the formation of the first constriction, possibly through regulating Antp expression. This function is independent of dpp signaling, since mutations in dpp or its receptors only disrupt the formation of the second but not the first constriction (Panganiban et al., 1990b; Nellen et al., 1994; Ruberte et al., 1995). This also suggests that there is either redundancy or that a different receptor system is responsible for mediating 60A signaling to pattern the first constriction. It would be interesting to see if AtrI (Childs et al., 1993), a type I receptor and STKD (Ruberte et al., 1995), a type II receptor in Drosophila, both of unknown function, are mediators of 60A signaling at the site of the first constriction.
The fact that 60A mutations are dominant enhancers of a sensitized dpp pathway implicates 60A in potentiating dpp signaling. This is most obvious in the visceral mesoderm of the midgut where dpp signaling is required to regulate homeotic gene expression and to maintain its own expression through a positive feedback mechanism. Although dpp signaling in the visceral mesoderm appears intact in 60A mutants, a requirement for 60A is revealed in tkv660A double mutants. When dpp signaling is attenuated through a mutant tkv receptor, eliminating 60A function reduces the signaling to below threshold level. The derepression of Scr in the anterior midgut and the loss of expression of dpp target genes, wg, Ubx and dpp, in the visceral mesoderm and lab in the endoderm are consistent with inadequate dpp signaling. A similar requirement for 60A is observed during dorsal closure of the embryonic ectoderm. The enhanced phenotypes of the adult appendages closely resemble those of the dpp hypomorphic mutants (Spencer et al., 1982), suggesting that 60A activity is also required for imaginal disc patterning. It is interesting that the imaginal discs are more sensitive to the reduction of 60A function, as a 50% reduction in 60A function is sufficient to produce a phenotype in a tkv6 genetic background. This may reflect a differential threshold requirement for dpp signaling in different tissues. Taken together, our data argue for an involvement of 60A in dpp signaling at different developmental stages and in various tissues.
In a signaling system with multiple interacting dimeric ligands, the interpretation of any single mutant phenotypes must consider the effect of losing both homomeric and possible heteromeric ligands. Therefore, the functions of the dpp pathway may be a composite input from Dpp homodimers, and Dpp/Scw and Dpp/60A heterodimers. Alternatively, 60A homodimers may function in an additive fashion with Dpp homodimers at sites of overlapping expression. However, the loss-of-function phenotypes of dpp are as severe as the loss-of-function phenotypes of its downstream components, such as tkv or Mad (Padgett et al., 1987; Penton et al., 1994; Nellen et al., 1994; Newfeld et al., 1996, 1997), suggesting that there is very little signaling, if any at all, from 60A homodimers in dppdependent events. Therefore, we believe it is unlikely that 60A homodimers play a significant role in dpp-dependent processes. Rather, we favor the interpretation that Dpp/60A heterodimers form at sites of overlapping expression and participate with Dpp homodimers in multiple signaling events.
The overlapping expression patterns of murine BMPs have led to the suggestion that they may act combinatorially during development (Lyons et al., 1995). Given the dimeric nature of TGF-β superfamily ligands, one mechanism to achieve such a combinatorial effect is to form functional heterodimers. Heterodimers of Xenopus BMP-4 and BMP-7 have been generated in vitro (Hazama et al., 1995) and shown to be more potent in bone- (Aono et al., 1995) and mesoderm-inducing assays (Suzuki et al., 1997) than either homodimer. In Drosophila, the Scw protein is proposed to upregulate dpp activity by forming Dpp/Scw heterodimers in the dorsal/ventral patterning of the embryonic ectoderm (Arora et al., 1994). The broad distribution of 60A proteins provides an opportunity for forming Dpp/60A heterodimers. Unlike scw null mutations, no obvious disruption of dpp signaling is observed in 60A null mutants, suggesting that Dpp/60A heterodimers are not as limiting as Dpp/Scw heterodimers, but partially redundant with Dpp homodimers. The first constriction phenotype of 60A mutants is unique, suggesting that it may be a function of 60A homodimers.
The tkv6 receptor failed to bind homomeric ligands when expressed in COS cells (Penton et al., 1994). Since dpp signaling is intact in tkv6 mutants, it would be interesting to determine if the mutant receptor still binds Dpp/60A and/or Dpp/scw heterodimers. If this is the case, the severe midgut phenotypes of tkv660A double mutants could be due to additional loss of signaling from Dpp/60A heterodimers, which would be consistent with the proposal that 60A potentiates dpp signaling by forming heterodimers with Dpp. The cuticular phenotypes of tkv660A double mutants without maternal wild-type Tkv receptor do not exhibit the altered dorsal/ventral polarity observed in dpp null embryos, possibly due to Dpp/Scw heterodimer signaling through the tkv6 receptor during early embryogenesis.
In summary, we have isolated mutant alleles of genes involved in dpp signaling, including 60A. Mutations in 60A disrupt dpp signaling in multiple developmental processes when dpp signaling is compromised. We propose that 60A participates in dpp signaling by forming heterodimers with Dpp protein. Our data support both dpp-dependent and dppindependent functions for 60A. It remains to be determined if these distinct functions reflect the qualitative differences between different forms of the ligands and if they are mediated by differentially activated receptors or distinct cytoplasmic signal transducers. The availability of 60A mutations provides a genetic tool for dissecting the differential requirement of each component in this combinatorial signaling system.
ACKNOWLEDGEMENTS
We thank S. Ahern-Djamali for critical reading of the manuscript. We also thank other members of the Hoffmann laboratory for helping with the screen. Y. C. thanks H. Zhang for his support. This work was supported by Department of the Army grant DAMD17-94-J-4339 to F. M. H. and by Cancer Center Core support CA07175 to the McArdle Laboratory. F. M. H. is the recipient of a Faculty Research Award from the American Cancer Society.