Regulation of the Twist target gene tinman by modular cis-regulatory elements during early mesoderm development

Transcription

1 Development 124, (1997) Printed in Great Britain The Company of Biologists Limited 1997 DEV Regulation of the Twist target gene tinman by modular cis-regulatory elements during early mesoderm development Zhizhang Yin, Xiao-Lei Xu and Manfred Frasch* Brookdale Center for Developmental and Molecular Biology, Mount Sinai School of Medicine, Box 1126, One Gustave L. Levy Place, New York, NY 10029, USA *Author for correspondence ( SUMMARY The Drosophila tinman homeobox gene has a major role in early mesoderm patterning and determines the formation of visceral mesoderm, heart progenitors, specific somatic muscle precursors and glia-like mesodermal cells. These functions of tinman are reflected in its dynamic pattern of expression, which is characterized by initial widespread expression in the trunk mesoderm, then refinement to a broad dorsal mesodermal domain, and finally restricted expression in heart progenitors. Here we show that each of these phases of expression is driven by a discrete enhancer element, the first being active in the early mesoderm, the second in the dorsal mesoderm and the third in cardioblasts. We provide evidence that the early-active enhancer element is a direct target of twist, a gene encoding a basic helix-loop-helix (bhlh) protein, which is necessary for tinman activation. This 180 bp enhancer includes three E-box sequences which bind Twist protein in vitro and are essential for enhancer activity in vivo. Ectodermal misexpression of twist causes ectopic activation of this enhancer in ectodermal cells, indicating that twist is the only mesoderm-specific activator of early tinman expression. We further show that the 180 bp enhancer also includes negatively acting sequences. Binding of Evenskipped to these sequences appears to reduce twistdependent activation in a periodic fashion, thus producing a striped tinman pattern in the early mesoderm. In addition, these sequences prevent activation of tinman by twist in a defined portion of the head mesoderm that gives rise to hemocytes. We find that this repression requires the function of buttonhead, a head-patterning gene, and that buttonhead is necessary for normal activation of the hematopoietic differentiation gene serpentin the same area. Together, our results show that tinman is controlled by an array of discrete enhancer elements that are activated successively by differential genetic inputs, as well as by closely linked activator and repressor binding sites within an early-acting enhancer, which restrict twist activity to specific areas within the twist expression domain. Key words: Drosophila, tinman, mesoderm, patterning INTRODUCTION The molecular mechanisms that determine patterning and differentiation of the mesoderm are a major focus of current research. In Drosophila, genetic analysis has shown that the twistgene occupies a position at the top of a hierarchy of zygotically active genes that function in mesoderm development. twist encodes a basic helix-loop-helix (bhlh) protein that is expressed in the presumptive mesodermal cells on the ventral side of blastoderm embryos and, in the absence of twist function, no mesoderm is formed (Simpson, 1983; Thisse et al., 1988; Kosman et al., 1991; Leptin, 1991). In addition to its role in mesoderm formation, twist has a second function in stages after gastrulation, where it appears to be required for myogenesis of somatic muscles (Baylies and Bate, 1996). Because twist encodes a putative transcription factor, it is assumed to control mesoderm development through the activation of a large battery of target genes, either in the whole mesoderm or in specific portions of it. Candidates include the genes encoding the homeodomain proteins Tinman and Zfh-1 (Bodmer et al., 1990; Lai et al., 1991), the MADS-domain protein MEF-2 (Lilly et al., 1994; Nguyen et al., 1994; Taylor et al., 1995), the FGF-receptor Heartless (Shishido et al., 1993), the integrin PS2 (Leptin, 1991), the KH-domain protein Struthio (also named Who or How; Baehrecke, 1997; Zaffran et al., 1997; Lo and Frasch, 1997) and genes with as yet undefined functions (Casal and Leptin, 1996). The tinman gene has a key role in early mesoderm patterning and is essential for the formation of all dorsal mesodermal derivatives, including the heart, visceral musculature and dorsal somatic muscles (Azpiazu and Frasch, 1993; Bodmer, 1993). In addition, tinman is required for the formation of certain body wall muscles and glia-like cells that are derived from ventral areas of the mesoderm (Azpiazu and Frasch, 1993; Gorczyka et al., 1994). These functions of tin are reflected in its dynamic mesodermal expression pattern, which can be subdivided into three major phases. The early, twistdependent phase includes tin expression from late blastoderm until after gastrulation in all cells of the trunk mesoderm (Bodmer et al., 1990; Azpiazu and Frasch, 1993). During the

2 4972 Z. Yin, X.-L. Xu and M. Frasch second phase, upon internal spreading of the mesoderm, tin expression is restricted to a broad band of cells in the dorsal mesoderm. This restricted tin expression domain is controlled by inductive signals mediated through Dpp, which is secreted by dorsal ectodermal cells (Frasch, 1995). Most likely, it is during this phase when tin fulfills its major role in the determination of dorsal mesodermal derivatives. Finally, tinman expression is further restricted to heart precursors and may be involved in differentiation and diversification of cardioblasts and pericardial cells during this phase. In this study, we address the question of how mesodermautonomous regulation, involving activation by twist, and inductive inputs, which include those mediated by Dpp, are integrated molecularly at the promoter level of the tinman gene. Our functional dissection of regulatory regions from the tin locus reveals that tin is controlled by distinct enhancer modules, one of which is responsible for its early and broad mesodermal expression, another for its Dpp-mediated dorsal mesodermal activation, and a third for cardioblast-specific expression. We further show that the early-acting module is composed of positively and negatively acting subsequences that functionally antagonize each other. We propose that direct binding of Twist to its target sites within this element can activate tin expression in the whole mesoderm, but Twist activity is abrogated in defined mesodermal areas by the binding of negative regulators to adjacent sites. As a result, tin is activated in the trunk mesoderm which gives rise to visceral and somatic muscles, heart and glial-like cells, but not in areas of the head mesoderm that give rise to hemocytes. The exclusion of tinman from the prospective blood mesoderm is, at least in part, due to the activity of the head-patterning gene buttonhead. Moreover, negative inputs triggered by the binding of the pair-rule gene product Eve result in periodically modulated levels of tin along the anteroposterior embryo axis. MATERIALS AND METHODS Isolation of genomic clones and construction of P- transformation plasmids The tinman gene from Drosophila virilis was isolated from a genomic library obtained from W. Hanna-Rose through J.D. Licht (Hanna-Rose et al., 1997). The screen was performed with a tin cdna probe using hybridization conditions as in McGinnis et al. (1984). Washes were done three times 30 minutes at 50 C with 1 SSC, 0.5% SDS. Upstream fragments of tin (D.mel.) were cloned into P-transformation vectors with native orientations relative to the basal promoters. +1 bp of our map refers to the first base of the longest tin cdna available (Bodmer et al., 1990). Following upstream constructs were made and tested in vivo: tin-3436/tin1: A fragment from 1716 bp to +287 bp generated by PCR with BamHI-site containing primers and cloned into the BamHI site of pcasper AUG βgal (Thummel et al., 1988). tin-3668/tin1 contains a fragment from 984 bp to +287 bp that was generated by PCR and cloned as above. tin-pvu/tin1 contains a fragment from the PvuII site at 384 bp to +287 bp in the EcoRI/BamHI sites of the same vector. tin-3436/3437 was generated from a PCR fragment ( 1716 bp to 39 bp), with primers containing BamHI (5 ) and XhoI (3 ) sites, that was cloned into XhoI/BamHI of pcasper hs43 βgal (Thummel et al., 1988). tin-3436/3843 was generated similarly from a PCR fragment containing sequences from 1716 bp to 269 bp. tin-1.7 kb+int was generated from tin- 3436/3437 by inserting intron 1 into the NotI site. Intron fragments from D. melanogaster (Dm) or D. virilis (Dv) were cloned upstream of the basal hsp70 promoter of pcasper hs43 βgal. The final constructs were: tin-a/lac (Dm), a 500 bp HindIII/XhoI fragment cloned into NotI/XhoI; tin-b-374 bp/lac (Dm), a 374 bp fragment cloned into NotI/XhoI, previously generated by PCR from a pks+ clone containing the 2.5 kb HindIII fragment of tin, using the M13 primer (5 ) and ATGCGGCCGCCTGCGGGAAA (3 ) and cut with NotI and internally with XhoI; tin-b-180 bp/lac (Dm), a PCR fragment generated with the primer pair AGGAATTCGTCAACAT- GTGT (5 ) and AGCTCGAGTCTCGTATATGG (3 ) and cloned into EcoRI/XhoI; tin-b (Dv), a 390 bp PCR fragment generated with the primer pair AGGAATTCGACAAATCATCG (5 ) and AGGGATC- CTCGATTAGTTGGC (3 ) and cloned into EcoRI/BamHI. E1,2/lac and E3/lac were made from double stranded oligonucleotides. E1,2 was flanked by EcoRI (5 ) and BamHI(3 ) sites, E3 by BamHI (5 ) and XhoI (3 ) sites and cloned into the corresponding sites of pcasper hs43 β-gal. For E1,2-E3/lac, E3 was cloned into the BamHI/XhoI sites of E1,2/lac. For E1,2-R-E3/lac and E1,2-2R-E3/lac, one or two PCRderived fragments (using the primer BamHI site from the oligonucleotide GCGGATCCAGGGGCGTCCTT at 5 and the endogeneous BamHI site at 3 ) were added to E1,2-E3/lac, respectively. The final versions of constructs containing downstream fragments from the tin genes from D. mel. (Dm) and D.vir. (Dv) were made as follows: tin-c (Dm), a 300 bp PCR fragment generated with the primer pair AGGCGGCCGCCATGAACAGCTT (5 ) and AGGGATCCGAGGCAGGGAAA (3 ) and cloned into NotI/BamHI; tin-c (Dv), a 354 bp PCR fragment generated with the primer pair AGGCGGCCGCTCAGGCACGGATC (5 ) and AGCTCGAG- GCAAAACATTTTACAG (3 ) and cloned into NotI/XhoI; tin-d (Dm), a 346 bp PCR fragment generated with the primer pair GAGAATTCATGTCAAGTGGCACTA (5 ) and ACCTCGAG- GTGGGAGGCTCGCAGCT (3 ), cloned into EcoRI/XhoI; tin-d (Dv), a 2.3 kb SalI fragment cloned into XhoI. In vitro mutagenesis was performed with the Transformer TM mutagenesis kit from Clontech. P-element constructs were injected into embryos from yw flies together with 2-3 as the transposase source. Between two and five lines from each constructs were analyzed for expression. Embryo stainings In situ hybridizations were performed as described in Lo and Frasch (1997), and antibody stainings or combined antibody/in situ hybridization stainings as in Azpiazu et al. (1996). Tinman antibodies were raised in a rabbit against bacterially expressed His(6)-tagged proteins that were purified on a Ni-column (Quiagen). The expression construct contained a 1.1 kb BamHI (generated at the ATG of the first Met) to HindIII fragment in pqe11 (Qiagen). Peroxidase antibodies were obtained from L. Fessler (UCLA) and β-gal antibodies were from Sigma (mouse polyclonal) and Cappel (rabbit). Drosophila stocks Following mutant alleles were used: btd XG, Dfd 21, ems 7099, eve 1.27, otd D87. lacz-balancer chromosomes were used to identify homozygous mutant embryos, except for eve, where homozygous embryos were identified using Eve-antibodies. For ectopic expression of twist, we used a UAS-twist line (Baylies and Bate, 1996) and an en-gal4 line obtained from M. Baylies. DNaseI footprinting assays DNaseI protection assays were performed essentially as described by Heberlein et al. (1985) with following modifications: The reaction mixtures (50 µl) contained 110 mm KCl, 47.5 mm Hepes (ph 7.9), mm MgCl 2, 1 mm DTT, 17% glycerol, 0.05% NP-40, 1 µg poly[d(i-c)], and 5 ng of 3 end-labeled probe. Upon addition of purified GST/Twist fusion proteins (Ip et al., 1992a) or Eve proteins (gift from M. Biggin) and incubation for 60 minutes on ice, 50 µl of 10 mm MgCl 2/5 mm CaCl 2 were added, followed by 1 µl DNaseI (Boehringer) to a final concentration of 0.4 µg/ml. After 2.5 minutes

3 Regulation of tinman during Drosophila mesoderm development 4973 digestion on ice, the reaction was stopped by adding 90 µl of 1% SDS, 20 mm EDTA, 200 mm NaCl, 250 µg/ml yeast trna, extracted twice with phenol/chloroform and ethanol precipitated. Electrophoretic separation was done in 8% polyacrylamide/7.5 M urea gels. RESULTS Distinct cis-elements drive each aspect of endogenous tinman expression As an initial step in the identification of enhancer sequences of the tinman gene, we isolated the tinman homolog and its flanking sequences from a distantly related species, Drosophila virilis, with the assumption that essential cis-regulatory sequences would be evolutionarily conserved. The tinman genes from D. melanogaster and D. virilis share a high degree of sequence similarity in their open reading frames, are organized in three exons with similar lengths and display almost identical patterns of expression in the two species (Fig. 1; Z. Y. and M. F., unpublished data). As illustrated in Fig. 2, the temporal and spatial expression of D. melanogaster tin protein expression can be subdivided into four major aspects. During invagination and early migration of the mesoderm, tin is expressed in all mesodermal cells, except for a small area in the head mesoderm that is negative (Fig. 2A,J). This early expression of tin depends on the function of the twist gene (Bodmer et al., 1990). tin expression then becomes restricted to the dorsal portion of the mesoderm, and both tin mrna and protein disappear from ventral cells during elongated germ band stages (Fig. 2D). It has been shown that restricted tin expression in the dorsal mesoderm is triggered by Dppmediated induction events (Frasch, 1995). In subsequent stages of embryogenesis, tin expression is limited to cells of the dorsal vessel. As shown in Fig. 2G, tin products are detected in generally four out of six pairs of cardioblasts per segment and in pericardial cells during dorsal closure (Jagla et al., 1997). The only non-mesodermal domain of tin expression is located at the anterior tip of the embryo, with highest expression levels in cells that are fated to become part of the pharynx and esophagus (Fig. 2A,D,J). The region containing essential tin enhancer elements was defined by a genomic construct, encompassing sequences from 6.2 kb to +4.6 kb with respect to the presumed transcription start site, that rescued the lethality of null mutations of tin (Azpiazu and Frasch, 1993). Since the closest gene upstream of tin, mod(mdg4), starts at 1.6 kb (Z. Y. and M. F., unpublished data), we tested fragments from the entire region between 1.7 kb and +4.8 kb for enhancer activity. Genomic restriction fragments, or PCR mod (mdg4) C fragments generated with primer pairs spanning stretches of sequences that were conserved between the two Drosophila species, were cloned upstream of a basal promoter and a β- galactosidase (β-gal) reporter gene and transformed into the germline (see Materials and Methods). Fragments showing enhancer activity in transgenic embryos were further dissected and tested similarly in additional rounds of analysis. From these studies, we identified four distinct enhancer elements from D. melanogaster, with lengths varying between 180 bp and ~500 bp, that activated reporter gene expression in patterns resembling specific aspects of the tinman pattern (Fig. 1). Surprisingly, all of these elements were located at positions downstream of the transcription start site. Element B (180 bp), located in the first intron, activated β-gal expression in the whole trunk mesoderm during gastrulation and germ-band elongation (Fig. 2B). An element from D. virilis located at a similar position had an identical temporal and spatial activity when tested in D. melanogaster embryos (Figs 1, 2C). The intensity of reporter gene expression with the B elements from both species was modulated in a pair-rule fashion along the anteroposterior axis (see below). A second enhancer element, D (~350 bp), that was located ~2 kb downstream of the 3 end of tin activated reporter gene expression specifically in cells in the dorsal portion of the mesoderm (Fig. 2E). Element D was active between stage 11 and early stage 12 of embryogenesis, corresponding to the period when maintenance of tinman expression in the dorsal mesoderm requires dpp. An analogous element was identified downstream of the D. virilis tinman gene and showed identical activities in transformed D. melanogaster embryos (Figs 1, 2F). A third element from D. melanogaster, C (300 bp; Fig. 1), was active in the dorsal vessel. This element activated β-gal expression from stage 12 on in four out of six cardioblasts per hemisegment, similar to the cardioblast expression of the endogenous tin gene (Fig. tinman 5' 3' A B C D R C H Xh B H H R R B Xb C 5' 3' B C D Pv H B R H Xh RV RV SB P S 1kb D. mel. D. vir. Fig. 1. Genomic organization and location of enhancer elements of the tinman genes from Drosophila melanogaster and Drosophila virilis. Exons are shown as black and introns as white boxes. The putative location of the transcription start site of the D. melanogaster tin gene was based on the position of the first nucleotide of the longest cdna (Bodmer et al., 1990). The approximate positions of the 5 end, exon boundaries and 3 end of the D. virilis genes were deduced from sequence comparisons with the D. melanogaster gene. Enhancer elements A to D, as described in the text, are shown as patterned boxes. Abbreviations: A, anterior head element; B; broad mesodermal element, C, cardioblast element; D, dorsal mesodermal (dpp-response) element; B, BamHI; C, ClaI; H, HindIII; P, PstI; Pv, PvuII; R, EcoRI; Xb, XbaI; Xh, XhoI; mod(mdg4), modifier of midget4 gene.

4 4974 Z. Yin, X.-L. Xu and M. Frasch 2H). A cardioblast element with similar activity was again found at a corresponding position downstream of the D. virilis tin gene (Figs 1, 2I). Elements C from both species did not exhibit any expression in the pericardial cells of the dorsal vessel. However, further dissection of the D element from D. mel. showed it to contain sequences that are indeed able to drive expression in pericardial cells, but this activity is obscured by the general dorsal expression with the intact D element (data not shown). Finally, we located an element (A, ~500 bp; Fig. 1) in the 5 portion of the first intron of the D. mel. tin gene that displayed activity in the anterior tip of the head (Fig. 2K). After invagination of the stomodeum, the bulk of the cells that are marked with β-gal from this element form the roof of the pharynx (Fig. 2L). In summary, the four enhancer elements from the D. melanogaster tin gene reflect the four major aspects of endogenous tin expression. At least three of these elements are conserved between D.mel and D. vir. with respect to sequences, relative positions, as well as spatial and temporal activities. Importantly, we were unable to find any enhancer activity in the 5 flanking and 5 untranslated regions of tin, in spite of testing a variety of constructs both with homologous tin and heterologous hsp70 promoters. Neither did we detect any enhancement of expression when 5 flanking sequences were combined with the first intron that contained elements A and B (see Materials and Methods). These findings are in agreement with the virtually normal expression of tin in embryos that are homozygous for deletions of 5 flanking regions (tin and tin , which delete sequences between 1546 bp and 162 bp or 23 bp, respectively; [Azpiazu and Frasch, 1993]; data not shown). Three conserved Twist-binding sites in the tin B enhancer are functional in vivo The early expression of tinman in the whole mesoderm has been shown to depend on twist (Bodmer et al., 1990). Since the tin B enhancer elements mimicked this pattern of expression, we focused on these elements and tested whether they are direct targets for the Twist protein. Sequence comparisons between the B enhancers from D. melanogaster and Fig. 2. Comparisons of tinman protein expression and the reporter gene expression driven by tinman enhancer elements from D. melanogaster and D. virilis. (A,D,G,J) tinman antibody stainings of D. melanogaster embryos; (B,E,H,K,L) β-gal antibody stainings of D. melanogaster embryos transformed with D. mel.-derived enhancer constructs and (C,F,I) similar stainings of D. melanogaster embryos transformed with D. vir.-derived enhancer constructs. Anterior is to the left and ventral is down. (A) In gastrulating embryos, tinman expression is observed in all cells of the trunk mesoderm and in a cap at the dorsal tip of the head. (B,C) In embryos of similar stages, the tin B enhancers from D. mel. and D. vir. drive β-gal expression in all cells of the trunk mesoderm. (D) At stage 11, tin expressing is restricted to an undulating band of dorsal mesodermal cells (d.ms.) and is absent in the ventral mesoderm (v.ms.). Head expression is enhanced. (E,F) The tin D enhancers from D. mel. and D. vir. are active only during stage 11 and early stage 12 in dorsal portions of the mesoderm. (G) Dorsal view of a stage 14 embryo, showing tin expression in the majority of cardioblasts (cbs) and in pericardial cells (pcs). (H,I) β-gal expression driven by the tin C enhancers from D. mel. and D. vir. in about four cardioblasts per hemisegment. (J) High magnification view of tin expression in the anterior portion of a stage 8 embryo, showing expression in the anterior head cap (arrowheads) and absence of expression in a region of the head mesoderm (bracket). (K) β-gal expression driven by the tin A enhancer from D. mel. in a head cap (arrowheads). (L) Embryo as in K at stage 12, showing perduring β-gal proteins in pharyngeal structures. The corresponding portion of the intron from D. vir. tin was not tested for enhancer activity.

5 Regulation of tinman during Drosophila mesoderm development 4975 D. virilis showed strong similarities within multiple stretches of ~20 bp to ~50 bp length (Fig. 3A), which were candidates for conserved regulatory sequences. Importantly, the sequences corresponding to the minimal tin B enhancer (180 bp) from D. mel. include three conserved E-box sequences that could potentially bind the bhlh protein Twist. These E-boxes contain the sequences CATGTG or CATATG in both species, and while E-boxes 1 and E-boxes 2 are arranged in tandem, E- boxes 3 are located ~120 bp to ~150 bp further downstream (Fig. 3A). DNaseI footprinting assays with bacterially purified Twist fusion proteins showed clear protection in the vicinity of these sequences, demonstrating that Twist is able to bind to all three E-boxes in vitro (Fig. 4A,C). Binding specificity was confirmed by using fragments in which the E-boxes have been mutagenized in vitro. Upon altering the sequences of E-box 1 and 2 from D. mel. to TATGTA (E1,2mut I), a strong reduction of Twist-binding was observed, although there was still some protection of sequences containing the modified E-box 1 at the highest concentration of Twist that was tested (Fig. 4B). The modified sequence TATATA at the position of E-box 3 (E3mut I) lost its ability to bind Twist under these conditions (Fig. 4D). To test whether the in vitro Twist-binding sites are required for the activity of the tin B enhancer in vivo, we generated a series of reporter constructs with modified versions of B enhancer sequences and analyzed their activity in transgenic embryos. A B enhancer construct (374 bp; Fig. 3A) in which E-box 3 had been mutated to TATATA, showed reduced reporter gene expression (Fig. 5A; see Fig. 2B for comparison). Similar results were obtained with B enhancer constructs containing an intact E-box 3 but mutant E-box 1 and E-box 2 sequences (TATGTA; Fig. 5B). Most strikingly, the periodic modulation of β-gal expression, which was evident with the intact B element, became much more pronounced with the constructs containing one or two mutant E-box sequences. When all three E-boxes were mutated, enhancer activity was drastically reduced and only a few residual mesodermal cells expressed β-gal (Fig. 5C). The weak, residual activity may reflect the residual binding affinity of Twist seen in vitro with these modified sequences (Fig. 4B). Therefore, we tested a reporter construct in which the sequences of all three E-boxes were further modified (E1,2mut II and E3mut II; Fig. 3B). As predicted, this mutant version of B exhibited a complete lack of mesodermal enhancer activity (Fig. 5D). We next tested whether the conserved sequence blocks spanning Twist-binding sites are sufficient to activate reporter gene expression in the mesoderm. When the sequences E1,2 and E3 (Fig. 3B) were linked with each other, strong Fig. 3. Sequence comparison and molecular dissection of tin B enhancer elements from D. melanogaster and D. virilis. A. BESTFIT alignment 374 bp tin B enhancer from D. mel. with the 425 bp tin B enhancer from D. vir. The sequences corresponding to the 180 bp tin B enhancer from D. mel. are boxed. E-box sequences and consensus binding sites for homeodomain proteins (and more specifically, Evenskipped) are shaded in black. It may be significant that, in five of the six E-boxes, the 3 G is folllowed by a T (E-box 3 would be in reverse orientation). B. Shown are the normal sequences of the 29 bp E1,2 and E3 elements, as well as the 52 bp R element, that were used individually or in combination to test for enhancer activities. The modified sequences that were obtained in two steps of in vitro mutagenesis and tested in the context of the 374 bp tin B element are shown below the wild-type sequences.

6 4976 Z. Yin, X.-L. Xu and M. Frasch expression was observed in the whole mesoderm (Fig. 5E). Thus, a 58 bp element containing three putative Twist-binding sites was able to activate β-gal in all cells of the embryo that contain Twist. Even a 29 bp fragment, E1,2, containing the first two E-boxes was sufficient to drive reporter gene expression in this pattern, albeit at reduced levels (Fig. 3B; Fig. 5F). By contrast, the E3 sequence was inactive in the trunk mesoderm and showed expression only in small regions of the rostral and caudal mesoderm (Fig. 5G and unpublished data). In order to provide stronger support for the notion that Twist binds and transactivates gene expression through the three E-boxes in the B enhancer, we mis-expressed Twist with the binary GAL4 system (Brand and Perrimon, 1993) under the control of engrailed enhancers in embryos carrying the reporter construct with the combined E1,2 and E3 elements. As shown in Fig. 5H, ectopic expression of Twist in ectodermal stripes caused striped ectopic expression of β-gal in the ectoderm. Taken together, these experiments provide strong evidence that the three E-box sequences are in vivo targets of Twist and, in the context of surrounding sequences of ~50 bp, are responsible for the early activation of tinman in the trunk mesoderm. Periodic modulation of tin B enhancer activity depends on eve and an Eve-binding element During its earliest phase of expression, just prior to gastrulation, ventral tin expression is modulated in a pair-rule fashion along the anteroposterior embryo axis, with highest levels of expression seen in six mesodermal stripes (Fig. 6A). This feature may be reflected in the periodic expression that was observed with tin B enhancer elements, which was particularly obvious upon introducing mutations in individual E-box sequences that diminished Twist binding (Figs 2B,C, 5A,B). To determine the segmental register of reporter gene expression in embryos transformed with these constructs, we performed double stainings for β-gal and pair-rule gene products. As shown in Fig. 6B, the pattern of β-gal directed by the E1,2 mut I derivative of the B enhancer is complementary to that of the pair-rule gene even-skipped (Macdonald et al., 1986; Frasch et al., 1987). Peak levels of β-gal were observed in areas lacking Eve protein. Since this β-gal pattern could indicate repression of enhancer activity by Eve, we tested its activity in eve mutant embryos. Indeed, in the absence of eve activity, the same enhancer is able to drive strong expression throughout the trunk mesoderm without any periodic interruptions (Fig. 6C). A B C D E Fig. 4. DNA-binding of Twist and Even-skipped to tin B enhancer sequences. (A) DNase I footprinting with Twist using the 5 portion of the 374 bp tin B element (B1). Lane 1 shows a G + A sequence ladder, lanes 2 to 5 protection experiments with 200, 400, 800 and 1600 ng of GST- Twi fusion proteins, respectively, and lane 6 a control experiment with the GST fusion moiety only. The protected sequences, including E-box 1 and 2 (shadowed black), are shown on the left. (B) Similar footprinting experiment as in A, but with tin B sequences containing mutated E- boxes 1 and 2 (left column, shadowed sequences). Note that two times higher GST-Twi protein concentrations were used in each lane as compared to A. (C) Twist DNA-binding experiment as in A, using the 3 portion of the 374 bp tin B enhancer (B2) which includes E-box 3 (shadowed black). (D) Twist DNA-binding experiment as in B, using a tin B element with in vitro mutated E-box 3 (shadowed grey). (E) DNase I footprinting with Even-skipped (Eve) assaying sequences of the R portion of the 374 bp tin B element (BR). 50 ng, 100 ng, 200 ng, 400 ng, 800 ng and 1600 ng of Eve protein were used for lanes 2 to 7, respectively. The protected sequences, including a sequence matching a consensus binding site for Eve (shadowed black), are shown in the left column.

7 Regulation of tinman during Drosophila mesoderm development 4977 Fig. 5. Functional analysis of the in vitro Twistbinding sites from the tin B enhancer in vivo. (A) A 374 bp tin B enhancer element with E-box 3 mutated to TATATA shows reduced and periodically modulated levels of reporter gene expression along the anterioposterior axis. Weak anterior expression indicating the presence of a second head element in the 374 bp B enhancer is also evident. (B) A 374 bp tin B enhancer element with E-boxes 1 and 2 mutated to TATGTA, shows similar periodic reductions of expression as with the mutant element from A. (C) Mutation of all three E-box sequences (as in A and B) within the 374 bp element leads to a dramatic loss of enhancer activity. Arrows point to mesodermal cells with residual, weak expression. (D) Further modifications of E-box sequences within the 374 bp element (see Fig. 3B) completely abolish its enhancer activity. (E) A combination of the elements E1,2 and E3 with native orientations (see Fig. 3B) exhibits strong mesodermal enhancer activity. Ectopic expression in the head mesoderm is marked with arrowheads. (F) The 29 bp element E1,2 is sufficient to drive mesodermal gene expression. Note that the expression in E and F includes the head mesoderm and anterior and posterior endoderm primordia. (G) The 29 bp element E3 is not sufficient to drive expression in the trunk mesoderm, but expresses β-gal ectopically in a portion of the head mesoderm. (H) Ectopic Twist (expressed under the control of engrailed enhancers in ectodermal stripes) is sufficient to activate ectodermal β-gal expression in embryos with a reporter construct containing in vitro Twist-binding sequences only (E1,2-E3, as in E). Among the pair-rule mutant tested, this outcome was unique to eve. In mutants for fushi tarazu (ftz), hairy (h) and runt (run), we observed interruptions in the β-gal pattern along the anteroposterior axis, however the pattern of the interruptions was consistent with the previously reported alterations in the eve pattern in these mutants (data not shown; Frasch and Levine, 1987). Additional data indicated that Eve may repress B enhancer activity (and tin) through direct binding to sequences in this element. The conserved sequence block between E-box 2 and E-box 3 contained a sequence matching homeodomain binding sites, which was particularly close to a consensus binding site previously determined for Eve (Fig. 3A; Hoey et al., 1988). In vitro DNA-binding experiments confirmed that this sequence is able to bind Eve, both in the case of D. mel. (Fig. 4E) and D. vir. (data not shown). Enhancer constructs that lacked the R element containing this Eve-binding site showed uniform expression throughout the mesoderm (Fig. 5E, F). This pattern is similar to that of B enhancers with weakened Twist sites in eve mutant backgrounds (Fig. 6C). When the 52 bp R element was added to the E1,2 and E3 elements, with a configuration similar to that of the native B enhancer, periodic repression was restored (Fig. 6D). Moreover, addition of two R elements in tandem resulted in much stronger periodic repression and yielded a pattern of clean stripes of β-gal in the mesoderm. These results indicate that Eve represses B enhancer activity by direct binding to recognition sequences in the B enhancer and thereby reduces its activation by Twist in a periodic fashion. Regulation of tinman in trunk versus head mesoderm tin is expressed in the whole trunk mesoderm of late blastoderm and gastrulating embryos, but expression is excluded from portions of the mesoderm in the head (Figs 2A,J, 6A). This is in contrast to the expression of its upstream regulator twist, which is seen throughout the length of the mesoderm (Thisse et al., 1988), and indicates the presence of negative regulators of tin in the head mesoderm. Lack of head mesoderm expression was also observed with 180 bp tin B enhancer constructs and their derivatives, which suggested that, in addition to positively acting Twi-binding sites, there are negative elements that prevent activation by Twist in the anterior mesoderm. This possibility was confirmed with the 58 bp and 29 bp constructs, containing the Twist-binding sites plus immediately adjacent sequences, which lacked repression in the head mesoderm and showed a pattern identical to twist (Fig. 5E,F). Since the major difference between these latter constructs and the tin B enhancer constructs was the absence of the R element, it was likely that sequences within the R elements mediated the head repression. Indeed, when the R element was added to the E1,2 and E3 elements, repression in the head mesoderm was restored (Fig. 6D,E). These results show that the R elements mediate repression in the head

8 4978 Z. Yin, X.-L. Xu and M. Frasch mesoderm, in addition to their function in Eve binding and periodic repression of Twist activation. Since head repression was maintained in eve mutant embryos (Fig. 6C), additional negative regulators must act on the R element. The absence of tin in the head mesoderm may be functionally important, since these cells have different developmental fates (Tepass et al., 1994) and express serpent (srp), a gene of the GATA family. serpent is required for the normal differentiation of these cells into hemocytes (Rehorn et al., 1996). srp is expressed in the cells of the head mesoderm that lack tin expression (Fig. 7A,C, arrowheads; note that srp has a second, more anterior domain that overlaps with endodermal tin expression). To identify upstream genes involved in the regulation of gene expression in the head mesoderm, we examined tin and srp expression in embryos mutant for early head-patterning genes, including buttonhead (btd; Wimmer et al., 1991), Deformed (Dfd; Merrill et al., 1987), empty spiracles (ems; Dalton et al., 1989), orthodenticle (otd; Cohen and Jürgens, 1990; Finkelstein and Perrimon, 1990) and ems,otd double mutants. Interestingly, in gastrulating embryos that are mutant for btd, tin expression was expanded into the head mesoderm, whereas the mesodermal domain of serpent was absent (Fig. 7B). During blastoderm stages, we observed severely reduced, transient expression of srp in the head mesoderm of btd mutants (Fig. 7D). None of the other mutations tested produced any obvious alterations of the tin and srp domains, indicating a major role of btd in the patterning of the head mesoderm. Similar to tin, expression of β-gal from B enhancer constructs expanded into the head mesoderm in the absence of btd function (Fig. 7E,F). Taken together, these results suggest that btd acts through the R element to prevent tin activation by twist in the head mesoderm. In addition, btd is required for normal activation of srp in the same area. Since both these activities of btd may be essential for normal blood cell development, we examined the distribution of hemocytes in btd mutant embryos. Normally, a large number of hemocytes, which stain for peroxidasin, are scattered throughout the body cavity of late stage embryos (Fig. 7G; Nelson et al., 1994; Tepass et al., 1994). By contrast, in btd mutant embryos, the number of hemocytes was reduced and virtually all of the residual cells remained near their place of origin in the embryonic head (Fig.7H). DISCUSSION Fig. 6. Periodic abrogation of B enhancer activity by eve and the R element. (A) tin mrna expression in a wild-type embryo at blastoderm, showing periodic reductions of tin levels along the anteroposterior axis. (B) Stage 8 embryo carrying the E1,2mutI/lac derivative of the tin B enhancer and stained for β-gal (brown) and Eve (black). β-gal staining is observed between the Eve stripes, which have split up into a strong and a weak component at this stage (indicated by solid parts of bracket). The absence of β-gal expression in areas between strong and weak eve stripes could be explained by the repressing activities of more broadly distributed eve products prior to this stage (indicated by entire bracket). (C) Stage 8 eve mutant embryo (eve 1.27 ) carrying the same transgene and stained as embryo in B. Absence of eve results in a uniform mesodermal activity of this enhancer. (D) Stage 9 embryo carrying an E1,2-R- E3/lac transgene. The presence of the R element leads to periodic reductions of mesodermal β-gal expression (compare Fig. 5E). (E) Stage 9 embryo carrying an E1,2-2R-E3/lac transgene. The presence of two R elements leads to wide periodic gaps of β-gal expression and, thus, a pattern of six mesodermal stripes. The mesoderm-patterning gene tinman has three major phases of mesodermal expression, the first occurring in the whole trunk mesoderm, the second in broad dorsal subdomains of the trunk mesoderm and the third in heart progenitors at the dorsal mesodermal crest. This refinement from an initially broad expression domain to progressively smaller areas could be explained by several different molecular mechanisms, including selective mrna stabilization or successive waves of transcriptional activation. Our analysis of regulatory regions from the tin gene demonstrates that each of these aspects of tin expression is indeed controlled by a separate enhancer module and, thus, reflects a distinct transcriptional activation event. The modular character of these enhancers was also evident in experiments with larger elements containing two different enhancers (e.g. A+B, C+D; data not shown), which produced additive patterns of reporter gene expression. Candidates for upstream components that trigger tin activation through some of these elements have been identified in previous genetic experiments (Bodmer et al., 1990; Frasch, 1995). Our present study shows that twist acts through the early (B) element (Fig. 8). dpp is likely to act

9 Regulation of tinman during Drosophila mesoderm development 4979 through the dorsally active (D) element, a notion reinforced by our observation that this element is inactive in dpp mutant embryos (data not shown). The upstream activators acting on the heart element (element C) are not yet known. Although tin expression in the heart progenitors, and the formation of heart progenitors proper, have been shown to require wingless (Wu et al., 1995), the critical period of wg requirement occurs earlier (at stage 10-11) than the activation of this enhancer, which does not begin until stage 12. Furthermore, element C does not contain any sequences closely matching those of binding sites for dtcf/pangolin, which is the Drosophila homolog of LEF-1 and appears to mediate responses to the wingless signal (Brunner et al., 1997; van de Wetering et al., 1997; Z. Y. and M. F., unpublished results). Therefore we favor the idea that wingless is required early, together with tinman during its expression in a broad, dorsal domain, to determine heart progenitor identities. In a subsequent step, these cells activate tin through as yet undefined mechanisms that involve element C in the case of the cardioblasts. In order to obtain further insights into the molecular mechanisms of mesoderm patterning, Dppmediated induction and heart differentiation, it will be of major importance to functionally dissect the C and D elements to a similar extent to that done in this study with the B element, and to identify the corresponding binding factors. Since tinmanrelated genes are expressed in the developing heart of vertebrate embryos and their expression depends on bone morphogenetic proteins (BMPs) that are homologous to Dpp (for a review, see Harvey, 1996; Schultheiss et al., 1997), it is conceivable that some of the molecular mechanisms involving the C and D elements are conserved. This possibility can now tested through sequence comparisons and reporter gene assays in heterologous systems. In addition to the regulatory elements that are active in the mesoderm, our analysis has also revealed the presence of distinct elements driving tin expression in cells at the anterior tip of the embryonic head. Tagging these cells with β-gal shows that they give mainly rise to pharynx and anterior endoderm, consistent with previous fate map studies for this area (Technau and Campos-Ortega, 1985). Although this aspect of tin expression has previously received less attention, it is interesting to note that the pharynx and anterior endoderm are also prominent sites of expression of tinmanrelated genes from vertebrates (Lints et al., 1993; Tonissen et al., 1994; Evans et al., 1995; Lee et al., 1996; Brand et al., 1997). Thus, some of the upstream regulators acting through element A are likely to be evolutionarily conserved as well. In Drosophila, candidates include Bicoid and D-gsc, the Zn-finger protein Tailless, and the Torsodependent Ras pathway (Berleth et al., 1988; Hahn and Jäckle, 1996; Goriely et al., 1996; Sprenger et al., 1989). The focus of the present study is the regulation of tin expression in the early trunk mesoderm. Early tin expression has several important biological functions. Notably, it appears to be required for the specification of cell fates in the ventral portion of the mesoderm, such as those of distinct muscle precursors and of mesodermally derived glia-like cells at the ventral midline (Azpiazu and Frasch, 1993; Gorczyka et al., Fig. 7. The role of btd in tin and srp regulation in the head mesoderm and in hemocyte development. (A) Stage 9 wild-type embryo stained for Tin protein (brown) and srp mrna (purple). Tin is excluded from a portion of the head mesoderm (between arrowheads) that expresses srp. A few cells at the border between the Tin and srp domains may express both genes. Cells in a second domain of srp, which constitute prospective endoderm of the anterior midgut (anterior to the left arrowhead), co-express Tin. (B) In a stage 9 btd mutant embryo, Tin expression is expanded into the head mesoderm (between arrowheads) and mesodermal srp expression is absent. The endodermal domains of expression are unaltered. (C) srp mrna expression in a wild-type embryo at blastoderm stage. The mesodermal srp domain giving rise to hemocytes is marked by arrowheads. (D) srp expression in a btd mutant embryo at blastoderm. Mesodermal srp expression (arrowheads) is severely reduced. (E) The 374 bp tin B enhancer is not active in the head mesoderm (arrowheads). (F) In a btd mutant embryo, the activity of the 374 bp tin B enhancer is derepressed in the head mesoderm (arrowheads). (G) Stage 16 wild-type embryo stained with antibodies against peroxidasin to detect hemocytes. Fat body and ring gland are also stained. (H) Stage 15 btd mutant embryo, stained as in G. There is a reduced number of hemocytes and all of them remain in the embryonic head region. Abbreviations: fb, fat body; hc, hemocytes; rg, ring gland.

10 4980 Z. Yin, X.-L. Xu and M. Frasch 5' 3' HEAD (pharynx, ant.endoderm)? twi wg btd eve A B C D TRUNK MESODERM CARDIOBLASTS dpp DORSAL MESODERM Fig. 8. Summary of tinman enhancer elements and their regulatory inputs. 1994). It is further required in an autoregulatory fashion to allow tin expression at high levels in the second, dppdependent phase in the dorsal mesoderm (X.-L. X., Z. Y. and M. F., unpublished data). Our experiments demonstrate that early mesodermal tin expression is driven by a distinct regulatory module, element B, that is composed of closely linked positively and negatively acting sequences. The basic helixloop-helix protein Twist appears to be the principal activator of this element, which has three E-box-containing binding sites. Two of these are arranged in tandem, with the E-boxes being only seven base pairs apart. Strikingly, a minimal element, E1,2, containing these two E-boxes and only 17 base pairs of additional sequences is able to activate transcription in the mesoderm, suggesting that activation by Twist does not require many additional DNA-binding factors. Moreover, ectodermal expression of the reporter gene upon mis-expression of Twist suggests that Twist is the only mesodermally restricted factor that is required to activate tin through this module, although generally expressed proteins could act as co-factors. For example, given that bhlh proteins tend to form heterodimers (Murre et al., 1989b), it is possible that, in vivo, Twist binds as a heterodimer with ubiquitously expressed bhlh proteins to each of these E-box sequences. A candidate for a Twist partner could be the the ubiquitously expressed bhlh protein Daughterless (Da; Cronmiller et al., 1988; Murre et al., 1989a). However, we found that the early expression of tin in embryos lacking both the maternal and zygotic functions of da is normal and, thus, Da does not appear to be an essential Twist partner (M. F., unpublished data). Although Twist could bind as a homodimer, it probably requires one or more co-factors that bind to immediately flanking sequences for transcriptional activation. This is suggested by the inability of other fragments from the tin locus to activate early mesodermal gene expression, even though they contain E-box clusters that bind Twist in vitro with similar affinities (Lee et al., 1997; Z. Y. and M. F., unpublished data). Since vertebrate homologs of tinman are not broadly expressed in the early mesoderm, the Twist-dependent activation of tin-related genes may not be conserved between insects and vertebrates. However, Twist homologs are expressed in the early mesoderm of vertebrate embryos, which suggests that at least some molecular aspects of Twist function have been conserved. Previous studies have focused on inhibitory functions of mouse Twist during myogenesis, which appear to involve competition for binding of myogenic factors, titration of E proteins, and the formation of inhibitory complexes with promoter sequences (Spicer et al., 1996; Hebrok et al., 1997). It will be important to determine whether mouse Twist has the ability to also activate target genes, and our identification of an activating Twist-response element in Drosophila may provide a means to address this question. The 52 bp R element, which is interspersed between the E-boxes, counteracts the activating functions of the E1,2 and E3 sequences by preventing tin activation in specific areas of the twist expression domain. This arrangement is reminiscent of an enhancer module from the rhomboid (rho) promoter that contains closely linked positively and negatively acting binding sites (Ip et al., 1992b). In this case, the Dorsal protein serves as an activating factor, with a minor contribution of Twist, while in the mesoderm, Snail repressor molecules interfere with this activation, leading to defined on/off states of rho expression along the dorsoventral axis. In the case of tin, Twist is the key activator, and negative regulators determine off domains at specific positions along the anteroposterior axis. One repressor molecule, Even-skipped, reduces activation by Twist in odd-numbered parasegments. Eve has been previously described as a transcriptional repressor (Han and Manley, 1993). In the tin B element, it could interfere with the formation of an active complex, which includes Twist bound to the three E-boxes, or it could reduce transcriptional activation through interactions with basal transcription factors (Austin and Biggin, 1995; Um et al., 1995). We do not know whether the pair-rule modulation of tin is functionally relevant, or whether Eve is merely pirating a homeodomain binding site without any major consequences for mesoderm patterning. The major role of the R element may be to prevent activation of tin in the portion of the head mesoderm that is fated to form hemocytes. We have identified the head-patterning gene btd as a negative regulator of tin and as a positive regulator of the hematopoietic differentiation gene srp in this region. This and additional genetic experiments with tin and srp mutants (Z. Y. and M. F., unpublished data) suggest that the complementary patterns of srp and tin are not achieved by mutual inhibition, but rather by an overlapping set of upstream regulators that affect tin and srp expression in an opposite manner. It should be noted that other regulators in addition to btd must be involved, because we observe some residual activation of srp in a small mesodermal domain in btd mutants that disappears prematurely during gastrulation. Similarly, tin mrna expression is initially absent in the head mesoderm of btd mutants, but expands into this area after blastoderm. The small amount of srp products in btd appears to be sufficient to allow the formation of some hemocytes. However, the premature disappearance of srp expression and perhaps the ectopic expression of tinman seem to interfere with normal hemocyte differentiation, as indicated by the failure of the remnant cells to migrate into the embryonic body cavity. There is genetic evidence for a distinct enhancer driving srp expression in the blood mesoderm (Rehorn et al., 1996). Once available, it will be interesting to compare this enhancer element to the tin B enhancer and to determine how reciprocal patterns of tin and srp are achieved on the molecular level. Since we did not detect any specific binding of Btd protein to the R element (although it bound well to control sequences containing a SP1-binding site; data not shown), it is possible that tin repression by btd is