Localized enhancement and repression of the activity of the TGF-P family member, decapentaplegic, is necessary for dorsal-ventral pattern formation in the Drosophila embryo

The result of the cascade of maternal gene action that patterns the dorsal-ventral axis of the Drosophila embryo is a gradient of nuclear localization of the product of the dorsal gene (Rushlow et al., 1989; Roth et al., 1989; Steward, 1989). High levels of the dorsal protein are present in ventral nuclei, while the dorsal protein is excluded from nuclei on the dorsal side of the embryo. The protein product of the dorsal gene is homologous to the transcription factor NF-KB and the rel oncogene (Steward, 1987; Kieran et al., 1990; Ghosh et al., 1990) and appears to specify pattern by the differential transcriptional activation or repression of specific downstream zygotic genes (Ip et al., 1991; Thisse et al., 1991).

While the mechanisms of action of the maternal genes are known in some detail, much less is known about how the early acting zygotic genes fix and elaborate embryonic dorsal-ventral pattern. Seven zygotically active genes, decapentaplegic (dpp), tolloid, screw, shrew, zerkni1llt (zen), twisted gastrulation and short gastrulation (sog), are required early in embryo genesis for the generation of normal pattern in the dorsal 40% of the embryo (Niisslein-Volhard et al., 1984; Wieschaus et al., 1984; Jiirgens et al., 1984; Wakimoto et al., 1984; Zusman and Wieschaus, 1985; Irish and Gelbart, 1987). At the syncytial blastoderm stage, the dpp, tolloid and zen genes are transcribed with apparent uniform intensity in the dorsal 40-50% of the embryo (St. Johnston and Gelbart, 1987; Shimell et al., 1991; Doyle et al., 1986; 1989), in the region where no dorsal nuclear protein is detectable. For zen, it has been demonstrated that dorsal-specific expression is caused by repression of zen transcription by the high levels of dorsal protein present in ventral nuclei (Rushlow et al., 1987; Ip et al., 1991). These results suggest that the gradient of dorsal protein may define the dorsal-ventral extent of the expression of the zygotic genes required for dorsal development, but may not affect the level of transcription within the region where they are expressed. Thus, if the gradient of dorsal protein only specifies the spatial extent of the transcrip tion of the zygotic genes required for dorsal develop ment, the pattern within the dorsal region of the embryo must be specified by interactions among these zygotic genes, rather than directly by dorsal.

Embryos that lack the dpp, tolloid, screw, shrew, zen or twisted gastrulation genes undergo abnormal mor phogenetic movements at the time of gastrulation, including failure of normal germ band extension, and subsequently differentiate larval cuticles that are miss ing characteristic subsets of dorsally and dorsolaterally derived cuticular structures. The gastrulation and cuticular defects of these mutants are similar to those observed in the ventralized embryos caused by a number of maternal effect mutations that ventralize the embryonic fate map (for example, see Rushlow and Levine, 1990; Rushlow and Arora, 1991). Since the early phenotypes of these mutations also appear to shift the fate map of the blastoderm embryo, we refer to them as zygotic ventralizing mutations.

The phenotypic and genetic analysis of the sag gene suggests that its mechanism of action is different from the six genes identified by zygotic ventralizing mu tations. Although sag mutants also fail to undergo normal germ band extension, the dorsal-most pattern elements are not completely deleted in sag mutant embryos (Zusman et al., 1988), suggesting that sag mutations affect the blastoderm fate map differently than the zygotic ventralizing mutations. In addition, genetic mosaic studies showed that the sag gene is required in the ventral embryonic cells to allow normal dorsal development (Zusman et al., 1988), indicating that the sag product acts cell non-autonomously to influence the fates of dorsally derived cells.

In this paper, we describe genetic studies on mutations in the zygotic genes required for normal dorsal development that allow us to begin to order the function of these genes. By studying the phenotypes of tolloid mutant embryos and genetic interactions be tween tolloid and dpp, we conclude that the function of the tolloid gene product is to increase the activity of dpp. Studies on the relationship of the other zygotically active genes to dpp lead us to conclude that dpp plays the central role in organizing the pattern of the dorsal region of the Drosophila embryo and that the activity of the dpp gene product is differentially regulated over the dorsal-ventral axis of the embryo.

The dpp gene encodes a member of the TGF-,β family of proteins (Padgett et al., 1987). Other TGF-,β family members are believed to be involved in an array of developmental events in vertebrate embryos (Melton, 1991). Our results suggest that during vertebrate development the activity of TGF-,β family members could be regulated by post-translational interactions with other gene products to organize the fates of fields of cells.

Fifteen to/laid alleles were isolated in the third chromosomal screen for zygotic pattern mutants (Jilrgens et al., 1984; Tearle and Niisslein-Volhard, 1986)-Seven tolloid alleles, t/d⁸¹, t/d⁸², tld⁸³, tld⁸⁴, tldw, t/d⁸, t/d⁸⁸, were isolated in various screens as zygotic lethal ventralizing mutations (K. V. A. and E. L. F., unpublished, P. Hecht and D. Morisato, personal communication). tlcf’8-62 is a P element-induced allele of tolloid (Shimell et al., 1991) that is a small deficiency that deletes the DNA encoding the to/laid transcript. Df(3R)XTA1, obtained from G. Jtirgens, deletes bands 94Cl-5 to 96B6-7. The tolloid alleles can be arranged in an allelic series from weak to strong as judged both by complemen tation behavior and by the progressive deletion of dorsally derived cuticular structures of the head and tail (see Fig. 3A} ve weak, tld5H, tld⁷M; weak, incompletely fienetrant, tld6, tld ; weak, completely penetrant, t/d⁷⁰, tld⁸, tid9° at 29°C, tld9Q¹; moderate alleles, tld⁸³, tfd9Q⁷, tl<f’3L, tld⁸¹, tld9⁸, t/d⁸⁷, tl<f6PI, tl<f6N, tld⁸²; and apparent null alleles, tld¹°F, tld⁷H, tldw, tJJ58-ol, tld10E, t/d84. AP element that contained a 14 kb insert of genomic DNA surrounding the tolloid gene and that was integrated on the X chromosome was used to increase tolloid+ gene dose (Shimell et al., 1991).

The recessive partial loss-of-function alleles of dpp, dpp”¹n-rl⁷ and dpP”¹n-r4, the haploinsufficient null allele of dpp, dppHtn4B, the deficiency of dpp, Df(2L) DTD2, and the dpp+ duplication, Dp(2;2) DTD48, are described either in Spencer et al. (1982) or in Irish and Gelbart (1987). Dp(2;2) DTD48 is the smallest chromosomal duplication of the dpp region, and it contains the dppd-ho mutation, a distal 3’ regulatory mutation that affects dpp expression in the wing disk. The dpp+ duplication, In (2LR) CyO P20, P[dppHm+], is a P-element construct that has a 20 kb insertion including the dpp transcription unit and surrounding regulatory se quences and is inserted into the second chromosome balancer, CyO (R. Padgett, personal communication).

The two putative null alleles of shrew, srw⁸⁴ and srw⁸⁵, isolated as zygotic lethal ventralizing mutations (P. Hecht and D. Morisato, personal communication) allelic to srw¹°K (Jtirgens et al., 1984), cause identical phenotypes that are stronger than any of the other four shrew alleles (data not shown). The phenotypes of these alleles are not enhanced in trans to a deficiency of the shrew gene, Df (3L) e’3 (obtained from M. Simon), suggesting that these mutations completely eliminate shrew gene activity. However, since embryos homozygous for the two known shrew deficiencies do not differentiate identifiable cuticular structures, we cannot prove that the srw⁸⁴ and srw’3⁵ mutations cause the complete absence of shrew activity.

The zen”‘36 allele appears to be null by phenotypic and molecular criteria (Wakimoto et al., 1984; Rushlow et al., 1987). Other alleles, zenMASJ, zenl7 and zen’62(M. Seeger, personal communication; Wakimoto et al., 1984), are partial loss-of-function alleles, as assayed by cuticular phenotype (Table 3).

The embryonic lethal alleles of short gastrulation are described in Wieschaus et al., 1984 and Zusman et al., 1988. Since sag maps on the X chromosome, allelism between these seven mutations had previously been assumed based on similarities of phenotype and map position. The weak, homozygous viable allele of sag that was isolated as a suppressor of the tld5Hjtld^10E phenotype (see Results) failed to complement all seven lethal sag mutations, confirming their allelism.

The twisted gastrulation alleles, tsgX⁸⁸⁶ and tsgYN9l, are described in Wieschaus et al. (1984) and the screw allele, scw¹G76, is described in Ni.isslein-Volhard et al. (1984).

Cuticle preparations were done as described (Wieschaus and Ntisslein-Volhard, 1986). In addition to the criteria of general morphology and ventral denticle band width, the severity of dpp, tolloid, shrew, screw, twisted gastrulation and zen mutant phenotypes was determined by scoring mutant cuticles for the presence and/or morphology of a number of dorsal and dorsolateral cuticular structures of the head and tail derived from known locations in the blastoderm fate map, as shown in Fig. 3A (based on Ji.irgens et al., 1986 and Ji.irgens, 1987). To examine the ventral sides of the normally tail-up sog embryos, a 23 gauge needle was used to tear the dorsal cuticle in the mid abdominal region and each embryo was oriented in the Hoyer’s -lactic acid mix to give a ventral view of its cuticle. So that the weight of the coverslip would not flatten the tom cuticles excessively, Sephadex G-50 beads were added to the mix.

Germ-line clones for a tolloid null mutation were induced by mitotic recombination in the developing germ lines of females heterozygous for the tolloid mutation tld¹⁰ErI. Females of genotype Tl¹⁰b, mwh e/tld¹⁰ErI, ru h th st sr bar-3 were irradiated with HXXJR in an Astrophysics Research Corpor ation (Long Beach, CA) X-ray source at the end of the first larval stage. Upon eclosion, these females were individually mated to males of genotype tld⁸⁴, th st cp in ri pP/TM3, Sb in egg-laying blocks, and the phenotypes of their eggs were examined. Tl¹⁰b is a dominant maternal-effect mutation that causes an extreme ventralization of the embryonic pattern such that all cells in embryos derived from females hetero zygous for the Tt¹⁰b mutation adopt the fate of the ventral most cells in the wild type and form mesoderm (Hudson, 1989). Because mesodermal cells do not secrete cuticle, in general no cuticular structures are visible in embryos derived from females carrying the Tt¹⁰b mutation. If, however, a mitotic recombination event is induced in a cell within the developing germ line of a female heterozygous for the Tt¹⁰b mutation, some descendants of that cell will be homozygous for the chromosome in trans to the Tl¹⁰b mutation and can be recognized by their ability to produce cuticle.

Approximately 600 individual irradiated females were examined. Fourteen females produced some embryos (ranging from 2-7) with cuticular structures, which were the result of a recombination event in the female germ line. Of 54 embryos with cuticular structures, 25 (of presumptive geno type tld¹⁰Erf/TM3) hatched, and 29 had a phenotype indis tinguishable from tld¹⁰Erl/tld⁸⁴ embryos derived from hetero zygous parents, indicating the absence of a detectable maternal contribution to the tolloid phenotype.

The onset of gastrulation, as evidenced by the formation of the ventral furrow, was used to precisely stage tolloid mutant embryos. To shift embryos prior to gastrulation, homozygous tlcfD flies were allowed to lay eggs on apple juice agar plates for two hours at either the permissive (18°C) or restrictive (30°C) temperature. The eggs were then transferred to a second plate at the opposing temperature, and all embryos that gastrulated within set intervals of time (every 15 minutes at 30°C and 30 minutes at 18°C) were put on separate plates. The developmental stage of the embryos at the time of the temperature-shift was calculated based on the amount of time from the temperature-shift until the onset of gastrulation. To shift embryos after the onset of gastrulation, ttcf⁰ embryos were collected at 18°C for four hours. The embryos were observed at 18°C and those embryos that gastrulated during each 30 minute period were put on a separate plate. All embryos were then simultaneously transferred to the restric tive temperature. The developmental stage of the embryos at the time of the temperature-shift was calculated based on the amount of time from the onset of gastrulation until the temperature shift.

Mutagenized males of genotype tld¹⁰E opa9c, h th st cu sr/TMJ, kni Me pP were mated to virgin females of genotype tl<f5H kn/ID, h sr e/TMJ, opa Me pP. Balancers with zygotic lethal mutations kni and opa were isolated by K. V. A. All four progeny genotypes from this cross normally die as embryos. Rare adult survivors should carry a second site mutation in tolloid that eliminates the antimorphic activity of the tld¹⁰E mutation, a dominant extragenic suppressor of the tl<f5H/IOE mutant phenotype, or a suppressor of odd paired (opa) or knirps(kni). EMS mutagenesis was performed according to the protocol of Lewis and Bacher (1968), using 40 mM EMS. For X-ray mutagenesis, males were irradiated with 4(XX) R. For P element mutagenesis, a P-cytotype stock of genotype .n2; ru h th st cu sr e Pr ca/TM], kni was constructed. Males from this stock were mated to females from the M cytotype strain tld¹⁰Eopa9c, h th st cu srjTM 1. The dysgenic F₁ progeny males of genotype tld¹⁰E opa⁹, h th st cu sr/TMJ kni were mated to the females described above. Approximately 25,CXXJ EMS treated F₁ zygotes, 50,(XX) X-ray treated F₁ zygotes and 150,(XX) hybrid dysgenic zygotes were tested. Zygotes that survived to adulthood were scored for visible markers and only the tl<f5H/IOE flies (h sr) were analyzed further. Adult flies of other genotypes appeared to be rare homozygous opa escapers or flies in which mutations on the original chromosomes had recombined with the TMJ balancer chromosome.

From 554 candidates, 16 lines were recovered in which more than 10% of the zygotes of genotype tld⁵H/IOE survived to adulthood. By segregation analysis, six lines carried a third chromosomal suppressor, nine lines carried a second chromo somal suppressor, and one line carried an X chromosome suppressor. All third chromosomal suppressors mapped within 2% of the tolloid locus and therefore were most likely second-site mutations within tolloid that eliminated the antimorphic activity of the tld¹⁰E allele. Based on 42 recombinants between vermilion (v) and forked (f), the X chromosomal suppressor mapped at 53.3 map units.

The nine second chromosomal dominant extragenic sup pressors of the tl<f5H/IOE mutant phenotype (12D, 26A, 26B, 28A, 35A, 39A,39D, 44F, 89D) appear to be mutations in dpp that elevate the activity of the gene or duplications of a region of the chromosome including the dpp transcript. All nine second chromosomal suppressors suppressed the haploletha lity of a dpp null allele. Based on the suppression of dpp haplolethality, seven of the suppressors (26A, 26B, 28A, 35A, 39A, 39D, 44F, 89D) were mapped between al (0.0) and dp (13.0) (dpp=4.0). Two suppressor chromosomes (12D and 28A) had chromosomal breaks in 22F, near dpp (22F2-3). Three suppressor strains had imaginal phenotypes like those associated with known alleles of dpp (Spencer et al., 1982; St. Johnston et al., 1990). The wings of 12D/dppHln-lS flies (dppHin4S is a haploinsufficient allele of dpp;Irish and Gelbart, 1987) were heldout. 39D/dppHm4S flies often had a central gap in the cuticle covering their notum (cleft notum) a defect associated with certain mutations in the 3’ regulatory region of dpp. Flies homozygous for the suppressor mutation 39A often had a terminal gap at the end of the L4 longitudinal wing vein (shortvein), a defect characteristic of mutations in the 5’ regulatory region of dpp (Segal and Gelbart, 1985). By Southern analysis (Maniatis et al., 1982), four suppressor lines (12D, 35A, 39A, 39D) had rearrangements in the 48 kb of the dpp gene surrounding the dpp transcript (+67kb to +115 kb on the molecular map of dpp;St. Johnston et al., 1990; kindly provided by M. Hoffmann). In three of the four lines, a single extra band was observed when probed with DNA from the 5’ (12D, 35A, 39A) or 3’ (35A, 39A) regulatory regions, which would be consistent with a duplication of the dpp coding region and part of the surrounding regulatory region of the gene. We hypothesize that those suppressor strains that displayed imaginal disk phenotypes characteristic of dpp alleles contained chromosomal rearrangements that both duplicated the coding region of dpp and disrupted the cis regulatory regions of the gene.

Zygotic ventralizing mutations cause the loss of a subset of structures normally derived from the cells in the dorsal 40% of the blastoderm embryo (Fig. 1). In the wild-type embryo, the dorsal-most blastoderm cells give rise to the extraembryonic amnioserosa, which does not contribute to the final cuticular structures of the larva, but is necessary for proper morphogenetic movements immediately after gastrulation (Lohs-Schardin et al., 1979). Most embryonic cells from 90 to 60% egg circumference (where 100% egg circumference is the dorsal midline) form the dorsal epidermis, character ized in cuticle preparations by a lawn of fine dorsal hairs.

Null mutations in six genes, dpp, tolloid, screw, shrew, zen and twisted gastrulation, delete different subsets of dorsal structures. The absence of dpp causes the most severe phenotype: these embryos lack the amnioserosa and the entire dorsal epidermis, which are replaced in the larval cuticle by ventral denticle belts that encircle the embryo (Fig. 1D, H; Irish and Gelbart, 1987). In embryos lacking to/Laid or screw activity, the amnioserosa and some, but not all, of the dorsal epidermal structures are deleted (Fig. lC, G; Arora and Ntisslein-Volhard, 1991). The partial loss of dorsal structures in to/Laid and screw embryos is accompanied by an expansion of the neurogenic ectoderm, as evidenced by an increase in the dorsal extent of each denticle band. Mutant embryos that lack shrew, twisted gastrulation or zen activity are missing the amnioserosa (Fig. 1 B, F; Zusman and Wieschaus, 1985; Wakirnoto et al., 1984). Partial loss-of-function mutations in dpp fall into an allelic series in which weak alleles delete the amnioserosa, like null alleles of shrew, twisted gastru lation or zen, while stronger alleles delete the same set of structures as null alleles of to/Laid and screw (Fig. 1 legend; K. Wharton, R. Ray and W. Gelbart, personal communication).

In contrast to zen, shrew and twisted gastrulation mutations, which affect only the amnioserosa, to/Laid mutations delete both the amnioserosa and part of the dorsal epidermis, suggesting that the tolloid gene product does not control differentiation of a particular cell type, but rather must be involved in organizing the overall pattern within the dorsal region of the embryo. We therefore chose to analyze the to/Laid mutant phenotypes in detail. These studies led us to investigate interactions between alleles of to/Laid and ventralizing mutations in other genes.

The temperature-sensitive period (TSP) of a tempera ture-sensitive tolloid allele, ttrP⁰, corresponds to the one hour immediately preceding gastrulation (Fig. 2). This suggests that wild-type to/Laid gene activity is required for normal patterning only during the cellular ization of the blastoderm. The to/Laid mutant pheno type is visible within 15 minutes after the end of the TSP: at that time in to/Laid embryos the normally lateral head fold begins to appear on the dorsal side, suggesting that the dorsal cells of tolloid embryos are behaving like the lateral cells of the wild-type embryo. The same shift in head fold position has been seen in dpp and zen mutants (Irish and Gelbart, 1987; Rushlow and Levine, 1990). Subsequently, to/Laid embryos do not extend their germ bands fully, as the dorsal-most cells of the embryo fail to form the squamous epithelial sheet characteristic of the amnioserosa.

The larval cuticular patterns of differentiated tolloid embryos lack a portion of the dorsal epidermis, as seen by the increased width of the ventral denticle bands. We wished to determine the exact regions within the dorsal epidermis that were affected by loss of tolloid activity; however, the thoracic and abdominal regions of the dorsal epidermis do not contain scorable cuticular markers that define different dorsolateral positions. Instead, we determined the severity of the phenotypes of tolloid mutants by scoring the specialized cuticular structures of the head and tail derived from stereotyped dorsal-ventral positions on the blastoderm fate map (Jurgens et al., 1986; Jurgens, 1987) (Fig. 3A). By these criteria, six of the 23 alleles examined resulted in as strong a mutant phenotype as does a deficiency of the tolloid locus and thus are likely to encode mutant polypeptides with little or no activity (Fig. 3A, Table 1, data not shown).

The remaining 17 tolloid mutations could be ordered in an allelic series (Table 1 and data not shown) in which weaker alleles (Fig. 3B) deleted only the most dorsally derived structures, while stronger alleles deleted dor sally and dorsolaterally derived structures (compare the dorsal-ventral extents of the top and bottom lines in Fig. 3A, which represent the structures deleted in weak and strong tolloid mutants, respectively). Thus, like dpp, tolloid is a component of a process that is more active in the dorsal-most regions of the embryo than in the dorsolateral regions.

Because lack of tolloid activity in the zygote deletes only a portion of the dorsal epidermis, it was possible that the maternal component of tolloid+ gene activity partially rescues the zygotic mutant phenotype. To test this hypothesis, female germ cells that lacked tolloid activity were produced by inducing mitotic recombi nation in the developing germ line of females hetero zygous for a to/laid null mutation (Materials and Methods). The lack of tolloid activity in the maternal germ line did not enhance the severity of the tolloid mutant phenotype (compare Fig. 3C to Fig. lC), indicating that there is no detectable maternal contribution to tolloid gene function in the developing embryo.

The tolloid gene is unusual among the zygotic lethal pattern genes because of a complex pattern of comp lementation among alleles (Jurgens et al., 1984). We repeated the complementation matrix and found that part of the apparent complexity of this pattern was caused by the incomplete penetrance of the four weakest tolloid alleles: a percentage of embryos carrying each of these alleles in trans to a tolloid deficiency survived to adulthood (Table 2 and data not shown).

The remaining complexity of the complementation pattern could be accounted for by the observation that three alleles, tlcfQI, tld6P4 and tld^{10 E}, behaved more strongly than a deficiency of the tolloid locus. For instance, fewer than 2% of the zygotes carrying anJ of these three alleles in trans to the weak alleles tld⁵ or tld7M survived to adulthood, while 20-35% of the zygotes carrying tld5H or tld^{7 M} in trans to a tolloid deficiency survived (Table 2). These three tolloid alleles also failed to complement a recessive, partial loss-of function dpp allele: embryos carrying one copy of both mutations died with a partially ventralized phenotype (Table 2 and data not shown). In contrast, embryos carrying a deficiency of the tolloid locus and the same dpp allele were fully viable, again indicating that the three tolloid alleles behaved more severely than a tolloid deficiency. The antimorphic (Muller, 1932) behavior of the tlcfQl, tld6P⁴ and tld¹⁰E mutations could be explained if their products block the remaining activity of the other allele, either by binding to the product of the other allele or by competing with that allele for association with the product of another gene required for the same process, possibly dpp.

To test whether the tld^10E allele did in fact encode an antimorphic product that interfered with normal dorsal development, we carried out a genetic selection to recover revertants that eliminated antimorphic activity of the tld^10E allele. If a second mutation were introduced into the tld^10E allele and eliminated all activity of the t/d^10E gene product, the resulting double mutant allele should behave less severely in genetic crosses than the t/d^10E gene product. For example, although only 0.25% of the embryos of genotype tld⁵Hjtld^10E survive, 20-35% of t!d5HjDJ embryos sur vive to adulthood. Thus, if a second -mutation that eliminated all to/laid activity were introduced into the tld^10E mutant allele, it should allow survival in trans to tld⁵H. Using this as the basis of a selection scheme (Materials and Methods), we recovered 16 lines from 225,000 mutagenized F zygotes in which more than 10% of the tlJ5Hjtld^10E zygotes survived. One of the revertant lines contained a large deletion encompassing the to/laid locus, indicating that the elimination of the antimorphic tolloid product leads to a less severe phenotype. The newly induced mutations in five of the other lines were inseparable from the original t/d^10E mutation (see Materials and Methods) and most likely are second site mutations within tolloid that inactivate the gene. In all six cases, the second site mutation eliminated the antimorphic activity of the t/d^10E mu tation both in interactions with weak to/loid alleles (the basis on which these mutations were selected) and in interactions with the dpp allele (Table 2 and data not shown).

The selection for intragenic revertants of the tld^10E mutation relied upon the fact that reversion of the antimorphic activity of the t/d^10E mutation resulted in the survival of the tld5H/tld^10E(*) flies to adulthood. However, this selection scheme could also be used to identify dominant extragenic suppressors of the t!d⁵Hjt!d^10E mutant phenotype. In fact, in ten suppres sor lines the mutations responsible for the survival of the tld⁵Hjtld^10E embryos were not linked to the tolloid locus and therefore defined dominant, extragenic suppressors of tol/oid. Nine of these suppressors mapped to the second chromosome, and one suppres sor mapped to the X chromosome. Because extragenic suppressors of a tolloid mutant could define genes that also affected dorsal development, we analyzed these mutations further.

In the course of characterizing the second chromosomal suppressors, we found that they could all suppress the haplolethality of null alleles of dpp (Irish and Gelbart, 1987), suggesting that they act either by elevating dpp activity or by partially bypassing the requirement for dpp. Although we have not character ized these suppressors in great detail, a variety of criteria suggested that most, if not all, of these mutations elevated dpp activity, possibly by duplicating the region of the dpp gene that included the open reading frame (see Materials and Methods).

To determine whether an elevation in df p gene dosage was sufficient to suppress the tld:5Hjtld¹⁰ mutant phenotype, we constructed tolloid mutant strains carrying either of two known duplications of the dpp+ gene, Dp(2;2) DTD48, which duplicates dpp+ and 8 known complementation groups (W. Gelbart,Jr,ersonal communication) or In (2LR) CyO P20, P[dpp n+], a P element construct with a 20 kb insert that includes the dpp transcription unit and surrounding regulatory sequences (R. Padgett, personal communication). In creasing the number of copies of the dpp transcription unit from two to three by introduction of either dur,lication resulted in a degree of suppression of the tld Hft/d¹⁰E mutant phenotype that was similar to the degree of suppression that we observed with the chromosome two suppressors (data not shown). Thus, we consider it likely that some or all of the chromosome two suppressors contain a duplication of the coding region of the dpp gene.

The results of other experiments indicated that an elevation of dpp+ gene dosage could suppress a variety of weak tolloid alleles. We constructed tolloid mutant strains that were homozygous for Dp(2;2) DTD48 and observed that doubling the dosage of the dpp+ gene completely suppressed the phenotypes of three weak tolloid alleles, tlcJ5H, tld6⁸ and tld⁷⁰. In each case, fewer than 2% of the homozygous tolloid embryos with four copies of dpp+ had a ventralized phenotype, compared to 60-100% ventralized embryos with two copies of dpp+. In addition, the degree of suppression was proportional to the dosage of the dpp+ gene: for several weak, completely penetrant tolloid allelic combi nations, three copies of dpp+ allowed some tolloid zygotes to survive to adulthood, while four copies of dpp+ promoted greater survival (data not shown). Thus, the higher the gene dosage of dpp+, the less tolloid activity was required.

The suppression of weak tolloid mutations by duplications of dpp+ indicated that the two gene products are required for the same process, but did not order their activities relative to one another. To try to order the activities of dpp and tolloid, we tested whether an elevation in the dosage of dpp+ could suppress the phenotype resulting from complete ab sence of tolloid gene activity. We reasoned that if dpp were required for the activity of tolloid or if dpp and tolloid were required independently for dorsal develop ment, an elevation in dpp+ activity would not suppress the phenotype of a tolloid null embryo. Conversely, if tolloid functioned to increase dpp activity, an elevation in dpp+ gene dosage could compensate for the loss of all tolloid activity. For two different tolloid deficiencies, DJ (JR) XTAJ and tlcf>S-6², and one phenotypically null allele, tld⁸⁴, the fraction of tolloid embryos that differentiated cuticular structures of the head and tail derived from dorsolateral positions on the blastoderm fate map increased with a doubling of the dpp+ dosage. For example, the percentage of tolloid null embryos in which some filzkorper material was observed in the tail of the embryo went from 31-41% of the embryos with two copies of dpp+ to 90-94% with filzkorper with 4 copies of dpp+ (at least 50 embryos of each genotype scored). Similarly, the fraction of tolloid null embryos differentiating the antenna) sense organs of the head more than doubled with a doubling of the dosage of dpp+. We then asked whether an increase in tolloid+ gene dosage could suppress a dpp mutant phenotype by constructing strains containing a dpp mutation and a P element insertion with tolloid genomic DNA (Shimell et al., 1991). We found that while doubling of the dose of tolloid+ could suppress a very weak dpp mutant phenotype, a doubling of tolloid+ gene dosage did not modify the phenotype caused by dpphin-r⁴, a partial Ioss of-function allele of dpp that resulted in a phenotype less severe than null alleles of tolloid (data not shown). Because an increase in the dosage of the dpp+ gene can partially bypass the requirement for tolloid, but an increase in tolloid gene dosage does not suppress a dpp mutant phenotype, we conclude that a normal action of tolloid is to elevate the activity of dpp.

The remaining extragenic suppressor of tolloid mapped to the X chromosome. The X-linked suppres sor did not suppress the haplolethality of a dpp null allele, and therefore was not a transposition of a wild type copy of the dpp gene to the X chromosome. The suppressor mapped to within 0.1 cM of the gene short gastrulation (sog) (Wieschaus et al., 1984; Zusman and Wieschaus, 1985). The suppressor mutation, which itself was homozygous viable, failed to complement all seven lethal sog alleles: only 1.5 -10% of flies carrying the suppressor in trans to any of the lethal sog alleles survived to adulthood. Thus, by the criteria of mapping and failure of complementation, the suppressor is a partial loss-of-function allele of sog. In addition, we found that a lethal sog allele acted as a dominant suppressor of the tolloid heteroallelic combination used in the reversion selection: 12% of the zygotes of genotype ttcJ5Hjtld¹⁰E; sogYS⁰⁶j+ survived to adulthood. Other weaker, but still completely penetrant, tolloid heteroallelic combinations were more strongly sup pressed by one copy of sogYS06 (data not shown). Thus the sog gene, while it has a different mutant phenotype than the zygotic ventralizing mutations (Zusman and Wieschaus, 1985), appears to be a component of the same pathway promoting dorsal development as tolloid.

shrew, zen and screw act in the same pathway as dpp The suppression of the phenotype of tolloid mutants by an increase in the dosage of dpp+ suggested that we could use increased dpp gene dosage to order the action of other zygotic ventralizing genes relative to dpp. Apparent null alleles of shrew (see Materials and Methods) cause completely penetrant phenotypes simi lar to those caused by weak alleles of tolloid. In shrew mutant embryos, the dorsally derived structures of the head skeleton are absent, and the dorsolaterally derived filzkorper of the tail are present, but are abnormal in morphology (Table 3). In shrew mutant embryos with four copies of the dpp+ gene, there was a restoration of the dorsally derived components of the head skeleton and the filzkorper were of normal mo1hology (Table 3). Since a doubling of the dose of dpp suppressed the phenotypes of putative null alleles of shrew, we conclude that shrew, like tolloid, functions upstream of dpp to increase dpp activity.

The homeobox-containing zen gene is required for the production of the amnioserosa cells. After the initial broad dorsal expression of zen defined by dorsal, zen expression becomes confined at the cellular blasto derm stage to the presumptive amnioserosa cells (Doyle et al., 1986; 1989). One of dpp’s functions in patterning the dorsal region of the embryo appears to be to activate zen transcription at this later stage (Rushlow and Levine, 1990) We found that doubling the dosage of dpp+ had no effect on the phenotype caused by a strong, apparently null allele of zen (Table 3), which is consistent with zen acting downstream of dpp. In contrast, doubling the dosage of dpp+ partially or completely suppressed the phenotypes caused by three weak alleles of zen (Table 3). This suppression of weak zen alleles suggests that the components of the dpp signal transduction pathway can transmit a doubling of the dosage of the dpp+ gene to increase the activity of a downstream target gene.

The two remaining zygotic ventralizing genes, screw and twisted gastrulation, could not be ordered unambiguously relative to dpp with this test. The screw gene was originally defined by a single partial Ioss-of function allele, scwc⁷⁶ (Niisslein-Volhard et al., 1984), that produces a weak ventralization of the embryonic pattern. The scwG⁷⁶ phenotype was completely sup pressed by two extra copies of dpp+ (data not shown), indicating that screw is also likely to function in the same process as dpp. However, since we did not determine whether null alleles of screw were suppressed by duplications of dpp+, we cannot conclude whether screw acts upstream or downstream of dpp. Two putatively null mutants of twisted gastrulation, which have a phenotype similar to shrew mutants, were not suppressed by extra copies of the dpp+ gene (data not shown), suggesting that twisted gastrulation acts down stream of, or in parallel with, dpp.

While doubling the gene dosage of dpp+ suppressed the phenotype of zygotic ventralizing mutations in four different genes, extra copies of the dpp+ gene enhanced the phenotype of sog mutants, causing profound alterations in the cuticular pattern of the neurogenic ectoderm. We observed that in most sog mutant embryos with four copies of dpp+ the ventral denticle bands were completely absent. Since doubling the dose of the dpp+ gene in sog+ embryos does not cause alterations in the embryonic cuticle, this phenotype indicated that the sog gene is required to inhibit dpp activity in the ventral regions of the embryo, and that ectopic dpp activity interferes with normal patterning in the neurogenic ectoderm.

To characterize the loss of ventral cuticular struc tures, we quantitated the left-right extent of the first and second abdominal denticle bands in wild-type embryos and in sog mutant embryos with two or three copies of the dpp+ gene (Table 4). We found that, compared to wild-type embryos (Fig. 4A), there was a reduction in the width of the denticle bands in sog mutants (Fig. 4B), and that an additional copy of the dpp+ gene in sog mutant embryos led to a further reduction in this width (Fig. 4C). This narrowing of the neurogenic ectoderm was accompanied by an expansion of the dorsal epidermis, as measured by a decrease in the left- to-right distance of the ventral-most dorsal hairs (data not shown), indicating that cells that would normally contribute to the ventral ectoderm were transformed to dorsal ectodermal fates.

We also observed cuticular defects within the neuro genic ectoderm in sog mutant strains. The Keilin’s organs are paired sensory structures located at stereo typed positions within the neurogenic ectoderm of the three thoracic segments. We found that the distance between the Keilin’s organs was reduced in sog embryos compared to the wild type and that increasing dpp dosage in sog embryos caused a further reduction in this distance (Table 4 and Fig. 5). These results suggest that in a sag mutant background, elevating dpp+ gene dose causes a ventral shift in the dorsal-ventral position of cuticular pattern elements within the neurogenic ecto derm and that the extent of this shift is correlated with the number of copies of the dpp+ gene.

The sog embryos with extra copies of dpp+ also appeared to have another patterning defect within the neurogenic ectoderm. For example, many sog embryos with four copies of the dpp+ gene had Keilin’s organs but completely lacked the second and third thoracic denticle bands (Fig. SB), suggesting that only ventral ectodermal structures at a particular anterior-posterior position within each segment were completely deleted. These phenotypes suggest that, in addition to dpp’s effect on dorsal-ventral patterning in these embryos, ectopic dpp activity in the ventral epidermis causes a segment polarity defect such that the cells that normally produce the denticle bands within each segment are transformed to produce naked cuticle.

This analysis and the work of Irish and Gelbart (1987) indicate that dpp, a member of the TGF-f:l family of proteins, plays a central role in patterning the dorsal region of the Drosophila embryo. dpp is the only gene whose activity is required for the production of all dorsal structures. We have found that a two-fold increase in dpp+ gene dosage has profound effects on dorsal-ventral patterning in various mutant back grounds. From these observations, we conclude that at least three other genes required for dorsal development act by regulating dpp activity.

Two genes required for normal dorsal development, tolloid and shrew, act by increasing the activity of dpp in the dorsal regions of the embryo (Fig. 6). Doubling the gene dosage of dpp+ can partially suppress the phenotypes caused by null mutations in both tolloid and shrew. Since increased dpp gene dosage can compen sate for the absence of tolloid or shrew, we conclude that tolloid and shrew function upstream of dpp to increase dpp activity. Our experiments do not allow us to conclude that the only functions of tolloid and shrew are to elevate dpp activity, since doubling the gene dosage of dpp+ does not completely suppress null alleles of either gene, and it will be interesting to test whether further elevation in dpp activity completely bypasses the requirement for both genes.

The sequence of the tolloid gene (Shimell et al., 1991) indicates that tolloid is likely to elevate dpp activity at a post-translational level. The tolloid protein is 41% identical over 620 amino acid residues to BMP-1, one of a group of bone morphogenesis proteins (BMPs) isolated as components of a protein complex that can induce ectopic bone morphogenesis in rats (Wang et al., 1988; Wozney et al., 1988; Celeste et al., 1990). The remaining BMPs are TGF-/3 family members, and one BMP, BMP-2, is 75% identical in its active carboxy terminus to the dpp protein (Wozney et al., 1988). Since the mammalian homologues of the tolloid and dpp proteins appear to be present in a complex, we postulate that the tolloid and dpp proteins are physi cally associated. Furthermore, this result raises the possibility that to/loid-like proteins could be involved in the signal-generating pathways used by other TGF-/3 family members.

The sequence of the to/loid protein provides a possible explanation for the relatively frequent isolation of antimorphic alleles of the gene. The predicted product of the tolloid gene has an N-terminal domain similar to a zinc-binding metallo-protease and a C terminal domain that contains EGF repeats and consequently may be involved in protein-protein inter-actions (Shimell et al., 1991). If tolloid acts as a protease, then the antimorphic alleles of tolloid could encode products that are able to form a complex with other proteins (perhaps including dpp), but render that complex inactive because they fail to cleave tolloid’s substrate.

In contrast to tolloid and shrew, the product of the short gastrulation (sog) gene appears to regulate dpp activity negatively (Fig. 6). While a twofold increase in the dosage of the dpp+ gene has no effect on the dorsal ventral pattern of sog+ embryos, the addition of extra copies of the dpp+ gene to sog mutant embryos causes multiple pattern abnormalities in the cuticular struc tures derived from the neurogenic ectoderm. These phenotypes indicate that one of the functions of the sog gene is to inhibit dpp activity in the ventral regions of the embryo. The ventral activity of sog observed in these experiments is consistent with the genetic mosaic data of Zusman et al. (1988) that the sog gene is required ventrally for embryonic survival.

Increasing the concentration of dpp activity in a sog mutant background appears to cause a greater number of cells to adopt dorsal fates. When we re-examined the cuticular defects of sog mutant embryos with two copies of the dpp+ gene, we observed similar, although much less severe, defects in these mutant embryos, indicating that in the wild-type embryo sog+ must inhibit dpp activity in the neurogenic ectoderm. Since the effects of sog mutations are visible at the time of germ band extension, we infer that sog inhibits dpp activity in the neurogenic ectoderm prior to or at the beginning of gastrulation. Because the initial transcription pattern of the dpp gene in the dorsal 40% of the embryo is not altered in sog mutants (Ray et al., 1991), the action of dpp on the cells of the neurogenic ectoderm of sog mutants must arise post-transcriptionally and be cell nonautonomous. We propose that in sog embryos the secreted dpp protein diffuses to ventral regions of the blastoderm embryo where it drives some ventrally located cells to differentiate along a dorsal epidermal pathway.

There are several possible mechanisms by which the sog gene product could inhibit dpp activity ventrally. The sog gene product could bind to and inactivate the dpp protein or it could inactivate any component of the dpp signal transduction pathway. Alternatively, the sog gene product could itself provide positional information in the ventral regions of the embryo, and the final fates of cells could reflect the local ratio of the activities of the sog and dpp gene products.

Small increases or decreases in dpp+ gene dosage can cause opposite shifts in the choice of cell fate along the dorsal-ventral continuum of the embryo. A 50% decrease in dpp gene dose is sufficient to cause the most dorsally located cells to adopt a more ventral fate (Irish and Gelbart, 1987). Conversely, in a sog background, a 50% elevation in dpp gene dose is sufficient to cause laterally located cells to adopt a more dorsal fate. Thus, the level of dpp activity can act as a developmental switch to specify cell fate, with high levels of dpp activity necessary for dorsal development and low levels of dpp activity necessary for ventral development.

The phenotypic analysis of partial loss-of-function alleles of dpp indicate that dpp is a component of a graded patterning process that is more active in the dorsal-most regions of the embryo than in the dorsolat eral regions (Irish and Gelbart, 1987; Fig. 1 legend; K. Wharton, R. Ray and W. Gelbart, personal communi cation). Because dpp plays the central role in this patterning process and because of the sensitivity of cells to incremental changes in dpp+ gene dose, we propose that the graded requirement for dpp within the dorsal epidermis reflects an actual gradient of dpp activity. In this case, a high level of dpp activity in the dorsal region of the embryo specifies the amnioserosa, intermediate levels of dpp activity promote dorsal epidermal differ entiation, and a lower level of dpp activity specifies the border between the dorsal epidermis and neurogenic ectoderm.

It is unclear what genes provide dorsal-ventral positional information within the neurogenic ectoderm (Ferguson and Anderson, 1991) and it is possible that a gradient of dpp activity could also contribute to positional information in that part of the embryo. In sog embryos with extra copies of the dpp+ gene, the ventral shift in the border between the dorsal epidermis and neurogenic ectoderm is accompanied by the loss of the ventral-most pattern elements of the neurogenic ecto derm, as measured by a decrease in the left-right distance between the Keilin’s organs. We conclude that positional information emanating from the dorsal region of the embryo influences, directly or indirectly, patterning within the neurogenic ectoderm. Thus it is possible that a gradient of dpp activity could provide positional information for patterning the entire dorsal 75% of the embryo.

If a gradient of dpp activity exists, it is likely to arise post-transcriptionally, since the dpp gene appears to be uniformly transcribed in the dorsal 40% of the syncytial blastoderm (St. Johnston and Gelbart, 1987). The data presented here suggest a gradient of dpp protein or activity could be produced by post-translational interac tions between dpp and the products of other genes. We have shown that the tolloid, shrew, sog and possibly screw genes are all required for normal levels of dpp activity. Because we have found that sog inhibits dpp activity ventrally, we propose that the effect of sog mutations on dorsal and dorsolateral cell fates (Zusman et al., 1988; Rushlow and Levine, 1990) are also caused by sog’s effect on dpp activity. The global and nonautonomous effects of sog mutations make sog the most attractive candidate for a gene that functions in the formation of a gradient of dpp activity.

Recently, the activins, TGF-/3 family members, have been shown to have dramatic effects on patterning the dorsal-ventral axis of both Xenopus (Thomsen et al., 1990; Smith et al., 1990; van den Eijnden-Van Raaij et al., 1990) and chick (Mitrani et al., 1990) embryos. Two groups have observed two different kinds of effect of the activins on early Xenopus embryos. Thomsen et al. (1990) found that exposure of regions of explanted animal cap to homogeneous concentrations of mam malian activin caused the production of embryoids with axial patterning, suggesting that an inherent polarity present in the animal cap is revealed upon induction by activin. In contrast, Green and Smith (1990) found that exposure of dissociated animal cap cells to activin caused these cells to adopt different fates according to the activin concentration, showing that activin can act as a dose-dependent morphogen. These two distinct interpretations of the role of activin in the patterning process could be reconciled if whole explants of animal cap tissue contain localized factors, possibly similar to the gene products that we have described here, that enhance or inhibit the activity of the applied activin to create a gradient of activin activity. We find it attractive to suppose that TGF-/3 family members act as graded morphogens to pattern the dorsal-ventral axis in both insects and vertebrates and that post-translational regulation is crucial in the establishment of activity gradients of TGF-/3 molecules.

We thank Kristi Wharton, Rob Ray, Bill Gelbart and Kavita Arora for communicating information prior to publi cation. We thank Rick Pad_gett for flies containing the In (2LR) CyO P20, P[dppHm+] construct. We thank Ed Espinoza for assistance in preparation of figures. We thank Mike O’Connor for many helpful discussions and Bill Gelbart, John Gerhart, Ulrike Heberlein, Rob Ray, Gerry Rubin, Mike Simon, Kristi Wharton and members of the Anderson lab for comments on the manuscript. E. L. F. was supported by a Helen Hay Whitney fellowship and by National Institutes of Health postdoctoral training grant HD07299. This work was supported by grant GM35437 from the National Institutes of Health to K. V. A.