Complex regulatory region mediating tailless expression in early embryonic patterning and brain development

Expression of the Drosophila tailless (tll) gene provides an instructive example of what has now become a generally accepted principle: many developmental regulatory genes, particularly those encoding transcription factors, although initially identified on the basis of their role in a specific process, are utilized at different times and in different places during development to control a wide variety of events. Analysis of the mech-anisms responsible for the requisite spatial and temporal regulation is critical for understanding the function of these multifaceted genes. In many studies, a multiplicity of regulatory modules has been identified for a single gene; each module mediates a particular aspect of the overall expression pattern (reviewed by Kirchhamer et al., 1996). The problem of the organization and function of multiple distinct regulatory modules is of particular importance for understanding the regulation of gene expression required for normal brain development, as so many regulatory genes have been recruited over the course of evolution (presumably by the addition of novel regulatory modules) to serve the formation of this most complex organ.

tll was identified in a mutant screen as a gene required for establishment of the eighth abdominal segment (A8) and telson (Jürgens et al., 1984), and subsequently shown also to be required for establishment of the posterior gut and, at the anterior, for the development of parts of the head and the most anterior portion of the brain, the protocerebrum (Strecker et al., 1986, 1988; Pignoni et al., 1990; Younossi-Hartenstein et al., 1997). Molecular analysis showed that tll encodes a transcription factor of the nuclear receptor family and that it is expressed in domains consistent with the mutant phenotype: transiently in a posterior cap from which A8, telson and posterior gut arise, and throughout embryogenesis in the acron primordium, then procephalic neuroblasts and finally in the developing brain (Pignoni et al., 1990; Younossi-Hartenstein et al., 1997).

The first phase of tll expression is relatively transient and is required for patterning of the portion of the embryo that gives rise to A8 and the posterior gut. At the beginning of the syncytial blastoderm stage, local activation of the maternally encoded torso (tor) receptor tyrosine kinase (tor RTK) (reviewed by Duffy and Perrimon, 1994) leads to transcription of tll by a relief-of-repression mechanism. Specifically, two synergistically interacting regions within the upstream regulatory region mediate repression of tll throughout the embryo; this repression is lifted at the poles when the tor RTK pathway is locally activated (Liaw et al., 1995). Within the tll regulatory regions are consensus tor response elements (tor-REs) that mediate the tor-sensitive repression (Liaw et al., 1995). This earliest phase posterior cap of tll mRNA serves its function and disappears by the end of gastrulation (Pignoni et al., 1990).

Subsequent phases of tll expression are concerned almost entirely with development of the brain. The second phase of tll expression is initated only a few nuclear cycles later (still during the syncytial blastoderm stage), as the first phase anterior cap disappears rapidly and is quickly replaced, under control of the terminal system and of bicoid (bcd) and dorsal (dl), by a horseshoe-shaped stripe at the cellular blastoderm stage. Expression in this anterior stripe (rather than the anterior cap) is essential for normal anterior embryonic development (Pignoni et al., 1992).

A third phase of expression can be defined as the time when the anterior stripe is modulated into two roughly triangular dorsolateral domains from which the brain neuroblasts arise; portions of this expression pattern are maintained in late embryogenesis in a complex pattern in the dorsal-medial portion of the brain (Pignoni et al. 1990; Younossi-Hartenstein et al., 1997). A fourth phase consists of expression in the optic lobe primordia. The mechanism by which the complex pattern of tll expression in the brain primordium is established and maintained, and the requirement for this expression in normal embryonic brain development are unknown.

Analysis of tll regulation in the developing brain of the Drosophila embryo is likely to have far-reaching implications, since the existence of orthologous genes, and their similarities in spatial expression, suggest that there is a common ground plan for both the insect and the vertebrate embryonic brain (Arendt and Nübler-Jung, 1996). In both classes of organisms, proneural genes of the achaete-scute complex (ASC) promote commitment to the neural fate, while a set of orthologous genes defines specific regions in the anterior portion of the brain. tll, orthodenticle (otd) and empty spiracles (ems) are expressed in the procephalic region of the Drosophila embryo and are required for the development of different portions of the syncerebrum (composed of proto-, deutero- and tritocerebrum) (Pignoni et al., 1990; Hirth et al., 1995; Arendt and Nübler-Jung, 1996; Younossi-Hartenstein et al., 1997). The respective vertebrate orthologs, Tlx, Emx and Otx, are expressed in the pros- and mesencephalon, and Otx2 is required for development of these domains (reviewed by Cohen and Jürgens, 1991; Finkelstein and Boncinelli, 1994; Yu et al., 1994; Monaghan et al., 1995; Matsuo et al., 1995). These similarities compel further investigation into the roles that these genes play in brain development and into the mechanisms by which their expression is restricted to specific regions of the brain.

Like tll, the head gap genes otd, ems and buttonhead (btd) all encode transcription factors, are expressed in broadly over-lapping anterior stripes at the blastoderm stage and are required for normal head development (Finkelstein and Perrimon, 1991; Cohen and Jürgens, 1991; Gonzalez-Gaitan et al., 1994; Schmidt-Ott et al., 1994; Hirth et al., 1995; Younossi-Hartenstein et al., 1997). Our analysis of their expression domains at the blastoderm and later stages relative to that of tll suggests that otd and/or btd might play a role in regulating tll expression in the head. Another possible tll regulator is lethal of scute (l’sc), a proneural gene of the ASC that encodes a bHLH transcription factor expressed throughout the brain anlage (Alonso and Cabrera, 1988; Younossi-Hartenstein et al., 1996). In spite of these suggestive expression patterns, analysis of tll expression in embryos mutant for these various genes, as described here, does not support a required role for otd, btd or l’sc in the initial expression of tll in the brain primordium.

To understand the regulation of and requirement for tll expression in the procephalic neuroectoderm, we characterize both endogenous tll expression and that driven by a reporter construct carrying approximately 6 kb of tll upstream regulatory DNA. We show that the reporter construct drives a pattern of expression essentially identical to the endogenous tll expression pattern in the developing central nervous system (CNS) of the embryo. We map both endogenous tll mRNA expression and promoter driven β-galactosidase (β-gal) expression onto a recently established (Younossi-Hartenstein et al., 1996) fate map of the Drosophila embryonic brain.

To begin the process of identifying regulatory modules and, ultimately, relevant interacting proteins involved in tll regulation, we sequenced the roughly 6 kb tll regulatory region and used portions of it to generate lacZ reporter constructs that were transformed into the germline. From the results of these experiments, we identify a region containing additional tor-REs. Most significantly, we also identify a number of regions that behave as brain-specific modules, in particular a 350 bp sequence capable of driving expression exclusively in the neuroblasts that become the dorsal medial portion of the brain. Identification of modules mediating spatially localized expression of tll in the brain is an important advance that should allow identification of relevant transcription factors and may ultimately have implications for understanding the regu-lation of the vertebrate homolog.

Standard molecular techniques used were those of Sambrook et al. (1989). The 4.5 kb SalI-HindIII fragment from the Drosophila melanogaster tll 5′ regulatory region (Fig. 3) was subcloned into pBluescript SK (Stratagene). Nested deletions were generated using exonuclease III (Erase-a-base, Promega). Both strands were sequenced by the chain termination method (Sambrook et al., 1989). Sequences were compiled and analyzed with the Genetics Computer Group (GCG) Wisconsin Sequence Analysis Package (1995) and analyzed with the same package, as well as with MacVector (IBI) and MatInspector (Quandt et al., 1995). Similar techniques were used to sequence the Drosophila virilis tll 5′ regulatory region (Liaw et al., 1993), except that sequencing was only on one strand and internal primers, rather than nested deletions, were used. The sequences of the roughly 6 kb of upstream tll regulatory regions of Drosophila melanogaster and Drosophila virilis have been deposited in GenBank (accession numbers AF019362 andAF019361, respectively).

Various fragments of the tll 5′ regulatory region (Fig. 1) were cloned into the multiple cloning site of the lacZ P-element vectors PwHZ16 (Liaw and Lengyel, 1992) and PwHZ128; PwHZ128 is a derivative of PwHZ16, with a modified multiple cloning site (i.e., the addition of an M13 primer, an XhoI restriction site, and a single KpnI restriction site) (Liaw et al., 1995). Constructs P1 through P4 have been described previously (Liaw and Lengyel, 1992). Construct TM was generated by inserting the BstEII-XmnI fragment of the tll 5′ regulatory region (Fig. 3) into the XbaI (blunted)-NotI sites of PwHZ16. For constructs K1 through K10, a single copy of each fragment was subcloned into the P-element vector; for construct K11, fragments of the tll promoter region were amplified using PCR, then oligomerized in tandem as described (Liaw, 1994). Oligomers containing four tandemly repeated copies of different portions of the A4 region were inserted into the P-element vector PwHZ128. All constructs were injected into w1118 embryos as described (Rubin and Spradling, 1982). For each construct, at least five independent transformant lines were established and assayed for expression pattern by in situ hybridization (see below).

Endogenous btd, ems, otd and tll mRNA in wild-type embryos, and lacZ mRNA in embryos transformed with the various tll promoter-lacZ constructs, were detected by in situ hybridization (Tautz and Pfeifle, 1989) using digoxigenin-labeled (Boehringer Mannheim Biochemicals) btd (Wimmer et al., 1993), ems (Walldorf and Gehring, 1992), otd (Finkelstein et al., 1990) and tll (Pignoni et al., 1990) cDNA, or lacZ DNA as a probe. β-galactosidase in embryos from these lines was detected by antibody staining, using standard techniques (Ashburner 1989) and antibodies and reagents from Jackson Laboratories. Rabbit polyclonal (primary) antibody to β-galactosidase was used at a dilution of 1:10,000, and biotin-conjugated goat antirabbit (secondary) antibody at 1:2,000. The Vectastain ABC kit was used with diaminobenzidine staining to detect the secondary antibody. Stained embryos were dehydrated through an ethanol series and mounted in Epon (Polysciences). Embryonic stages were determined according to Campos-Ortega and Hartenstein (1985).

P-element constructs were maintained in a w1118 homozygous background. To determine the effect of alterations in tor activity on the expression driven by particular constructs, males from a line carrying a construct insert in homozygous condition were crossed to females heterozygous for the dominant gain-of-function allele torD4021 (Klingler et al., 1988). Alleles used to study effect on tll expression were otdh1, btdXG and Df(1)scB57, which genetically behave as nulls (Younossi-Hartenstein et al., 1997).

To provide the background information needed for dissection of the tll regulatory region, particularly for identification of modules controlling expression in the brain neuroblasts, we characterized expression of endogenous tll mRNA throughout embryogenesis.

Transcription of tll is initiated early in the syncytial blastoderm stage (stage 4) in two symmetrical caps; these two caps depend entirely on activity of the maternally encoded terminal system (Pignoni et al., 1992). Expression in the posterior cap is required for posterior patterning of the blastoderm stage embryo (Mahoney and Lengyel, 1987; Weigel et al., 1990). Although expression in both caps is transient, that in the posterior cap reaches a higher level and persists longer, i.e., through most of the cellular blastoderm stage. Expression in the posterior cap largely disappears during gastrulation and is undetectable by the end of germband extension (Pignoni et al., 1990).

By the beginning of the cellular blastoderm stage, the anterior cap is replaced by a horseshoe-shaped stripe that straddles the dorsal midline between 76 and 89% egg length (EL) (Pignoni et al., 1990, 1992). Projected onto the blastoderm fate map (Campos-Ortega and Hartenstein, 1985), this stripe covers the entire brain anlage, i.e., the procephalic neurectoderm; expression in this stripe continues in domains that will become the brain (Pignoni et al., 1990; Younossi-Hartenstein et al., 1997).

tll expression in the procephalic neurectoderm continues, with complex modulations, from the blastoderm stage into late stages of embryogenesis. We followed this expression pattern (by in situ hybridization to whole-mount embryos) and correlated it with the recently defined (on the basis of expression patterns of proneural and neurogenic genes) procephalic proneural domains from which the neuroblasts of the brain are derived. The nomenclature used below to refer to these domains, and the neuroblasts that delaminate from them, is that given by Younossi-Hartenstein et al. (1996). Confirming and extending earlier work (Younossi-Hartenstein et al., 1997), our analysis shows that tll mRNA is present in all proneural domains of the protocerebrum; furthermore, the level of tll mRNA in a domain is highest shortly before and during the stage at which neuroblasts delaminate from that domain.

During gastrulation, the horseshoe-shaped stripe of tll expression becomes tilted backwards and undergoes a process of internal differentiation into regions with different levels of expression. A group of cells in the dorsal midline gradually ceases expressing tll, resulting in a split of the horseshoe into two dorsolateral domains (Fig. 1A,C). Within each of these domains, a roughly triangle-shaped, anterodorsal region shows the highest level of expression and is designated HL; posteroventral to HL is a domain with a lower level of expression designated LL (Figs 1A,C, 2).

The HL domain during stages 7 and 8 covers the dorsocentral part of the protocerebral neurectoderm. From within this region, the first groups of protocerebral neuroblasts (Pc2, 3) delaminate during late stage 8 (Younossi-Hartenstein et al., 1996). During late stage 9 (Figs 1C, 2 and data not shown), the HL domain becomes restricted to the dorsoanterior portion of the procephalic region; this position corresponds to the anterior protocerebral domain from which neuroblast groups Pa3-4 delaminate during stage 10 (Younossi-Hartenstein et al., 1996). From stage 11 (Fig. 1E) onward, high levels of tll are found primarily in a bilaterally symmetric arc close to the dorsal midline; this arc largely overlaps the dorsomedial l’sc expression domain Pdm of Younossi-Hartenstein et al. (1996), but extends posteriorly to it.

tll is expressed in a spotty pattern in the LL domain, which surrounds the HL domain anteriorly, ventrally and posteriorly. LL covers the entire protocerebral neurectoderm, including the domains from which the Pc1/2 neuroblasts delaminate during stage 9, the Pa1/2 and Pp1/2 neuroblasts during stage 10, and the Pp3, 4 and 5 neuroblasts during stage 11. The relationship of the HL to the LL domain at stage 9, and to the position of the delaminating neuroblasts, is shown schematically in Fig. 2.

During stage 12, tll is expressed at a high level in a new region: the primordium of the optic lobe. This structure arises by invagination of the posterior procephalic ectoderm, a region that corresponds to the proneural domains Pp3-5 (Younossi-Hartenstein et al., 1996). tll expression remains high in the optic lobe throughout late embryonic development and can also be seen in the optic lobe in the late third instar larva (Fig. 1G and data not shown).

To identify regulatory elements responsible for the complex expression pattern described above, we began with reporter construct P1, which contains all of the 5′ tll regulatory region (5.9 kb) necessary to rescue a null tll mutation (Fig. 3; Pignoni et al., 1990). This construct has already been shown to express lacZ in a pattern that mimics endogenous tll expression at the blastoderm stage (Liaw and Lengyel, 1992). As described below, lacZ expression driven by P1 also mimics brain-specific tll expression.

During gastrulation, P1 drives expression at high levels in the HL domain, and lower levels in the LL domain, essentially coincident with the endogenous tll pattern (Fig. 1B,D). During later embryonic development, the dynamics of P1-driven lacZ mRNA expression within the HL and LL domains do not significantly differ from the pattern described above for the endogenous tll transcript (Fig. 1F,H). Thus the 5.9 kb of tll 5′ regulatory region drives expression transiently in all proto-cerebral neuroblasts (with the possible exception of the posteriormost groups Pp3-5), and in the clusters of neural precursors that segregate from the dorsomedial procephalon during stages 12-13 by mass delamination and invagination (domains defined by Younossi-Hartenstein et al., 1996). In addition to mimicking the endogenous tll expression in the HL and LL regions, the 5.9 kb region also drives expression at stage 12 in the optic lobe primordium. Like endogenous tll expression, P1-driven expression in the dorsomedial brain hemispheres and optic lobe persists until late embryonic stages (Fig. 1H).

These results indicate that the 5.9 kb of DNA immediately 5′ to the tll transcription unit has within it essentially all of the stage- and region-specific enhancers necessary to drive the endogenous tll expression pattern. In what follows, therefore, we refer to this DNA, contained within the P1 construct, as the ‘complete regulatory region’ and to embryos carrying the construct as [P1] embryos.

Antibody staining of [P1] embryos detects β-gal in essentially the same cells as lacZ mRNA (above), but present for longer (cf. Fig. 1F,H with Fig. 4A,B). This strong P1-driven β-gal expression pattern (which we refer to as tll P1 expression) allows us to test specific genes for their effect on expression of tll in the head.

Likely candidate regulators of tll are the head gap genes otd, ems and btd. These genes, in addition to their required role in patterning the gnathal segments and procephalic region (reviewed by Cohen and Jürgens, 1991), have each recently been shown to be required for the establishment of different specific neuroblast populations of the brain (Hirth et al., 1995; Younossi-Hartenstein et al., 1997). To evaluate their possible roles in regulation of tll brain expression, we review previous results and provide additional descriptions of the expression patterns of these head gap genes. At the cellular blastoderm stage, otd, ems and btd are expressed in broad overlapping stripes along the anterior-posterior axis of the embryo at positions on the dorsal midline of approximately 70-90, 69-75 and 67-75% egg length (EL), respectively (Dalton et al., 1989; Finkelstein and Perrimon, 1990; Wimmer et al., 1995). Note that, as described above, the anterior horseshoe domain of tll lies within the otd blastoderm expression domain.

Beginning at gastrulation (stage 6), expression of each of the head gap genes undergoes a different set of modulations. otd expression retreats from the ventral midline, but persists in a major portion of the deuterocerebral and protocerebral neurectoderm where it forms a horseshoe domain that overlaps with tll expression (Fig. 5A). btd expression becomes divided into two domains, the most anterior of which is a bilateral, anterior dorsomedial region overlapping partially with the tll HL domain (Fig. 5C). Both btd and ems are expressed in stripes just anterior to the cephalic furrow; these stripes are not part of the protocerebral neurectoderm, however, and lie posterior to the domain of tll expression (Fig. 5C,E). By stage 9, there is little or no expression of btd or ems in the protocerebral domains of the head, while otd expression persists in a fairly broad domain that covers most of the protocerebral domain of the head ectoderm (Fig. 5B,D,F).

From the preceding description of head gap gene expression, it is evident that only the otd and btd domains overlap with the tll protocerebral expression domain and might therefore affect tll expression. To investigate the role of otd or btd in establishing or maintaining tll expression in the brain primordium, the P1 construct was crossed into null otd and btd backgrounds.

In btd mutant embryos, development of the protocerebrum proceeds essentially normally (described in more detail by Younossi-Hartenstein et al., 1997), and the domain of tll P1 expression is initiated normally and remains normal in size and level into later stages (Fig. 4G,H). The smaller expression domain of tll at stage 14 is due to the absence of the optic lobe from btd embryos; therefore, we cannot assess the effect of btd on tll expression in this domain. We conclude that btd expression, even though partly overlapping with tll expression at stage 6, is not required for establishing or maintaining most of the tll expression pattern.

In otd mutant embryos, tll P1 expression is normal at stages 9 and 10 (i.e., no abnormal patterns were seen among progeny from the cross of otd heterozygous parents), indicating that otd is not required for tll expression until at least stage 10. Defects are first detectable in the brain primordium of otd embryos beginning at stage 11; these ultimately result in a protocerebrum that is severely reduced in size (Younossi-Hartenstein et al., 1997). Correspondingly, after stage 11, the tll P1 domain is reduced in size; the expression level within this domain, however, is similar to that seen in wild-type embryos (Fig. 4E,F). Thus although we cannot assess the role of otd in parts of the protocerebrum at later stages, we can say that otd is not required for tll expression prior to stage 11, or for tll expression in the portion of the protocerebrum that remains in older otd embryos.

Additional candidate genes for regulation of tll are the proneural genes within the ASC; all four of these genes encode bHLH transcription factors (Alonso and Cabrera, 1988). Of these four genes, however, only l’sc shows significant expression in the procephalic neurectoderm; this expression is seen in a broad domain in the protocerebrum between stages 7 and 11 that overlaps the domain of tll expression (Younossi-Hartenstein et al., 1996). Elimination of the activity of l’sc (as well as that of the three other genes of the ASC) by use of the deficiency Df(1)B57 did not, however, result in a loss of tll expression. At early stages, tll P1 expression in homozygous ASC mutants appears indistinguishable from expression in wild-type embryos. Beginning during stage 13 in embryos lacking l’sc, there is an increased amount of apoptotic cell death in the tll P1-expressing cells of the protocerebral ectoderm of l’sc embryos, as indicated by the appearance of numerous macrophages containing β-gal-positive cellular debris at the dorsal aspect of the developing brain. This excess cell death leads to a decrease in the size of the tll-expressing cell population (Fig. 4C,D). We conclude that l’sc, although required to maintain viability of certain tll-expressing cells in the protocerebrum, is not required to establish or maintain tll expression per se.

Finally, since tll itself encodes a nuclear receptor transcrip-tion factor (Pignoni et al., 1990), we can ask if tll regulates its own expression. Prior to stage 11, no difference is seen among embryos from crosses between heterozygous parents carrying a null tll allele (Pignoni, 1991). By stage 12, homozygous tll embryos are seen to lack a significant portion of the brain, the protocerebrum (Younossi-Hartenstein et al., 1997); these embryos still express tll, however, in a round placode-like domain that appears to constitute the uninvaginated optic lobe (Fig. 7L). These results indicate that the tll gene does not autoregulate.

We conclude that, even though their domains of expression in the blastoderm stage embryo overlap partially with the anterior domain of tll, the head gap genes btd and otd are not required for establishing or maintaining tll expression in the anterior stripe. Further, neither these genes nor l’sc is required to maintain tll expression in the brain neuroblasts; in particular, these genes do not regulate tll via the 5.9 kb regulatory region.

Previous analysis identified an 11 bp tor-responsive element (tor-RE) in the proximal region D3 (Fig. 3). D3 functions as a minimal regulatory element, or module; a module with similar function and interacting synergistically with D3 was inferred in the distal (upstream of −2291) region (Liaw and Lengyel, 1992; Liaw et al., 1993, 1995; see Fig. 3).

To localize the synergistically interacting distal tor-RE, constructs K1-K10 (Fig. 3) were generated. Of these, only K2 and K10 (which both carry the most proximal 635 bp of the distal regulatory region) drive polar expression (Figs 3, 6A,C). When K2 and K10 are introduced into embryos in which Torso is ectopically active, the posterior cap of expression is expanded dramatically (Fig. 6B,D), indicating that both constructs contain an element(s) that responds to activated Torso. Sequence comparison with the D3 region (Liaw et al., 1993) revealed two tor-REs in region −2291 to −2770 (Fig. 6F); this region, four-fold oligomerized to generate construct K11, drives strong polar expression (Figs 3, 6E). K11 drives only posterior and not anterior polar expression, perhaps due to direct repression by Bicoid via consensus binding sites in the −2291 to −2770 region (Fig. 6F); consistent with this interpretation, the D₃ region lacks Bicoid-binding sites and drives expression at both poles (Liaw and Lengyel, 1992).

To determine how tll expression is controlled in the HL, LL and optic lobe (OL) procephalic neuroblast domains, we examined expression driven by both previously and newly generated promoter-lacZ constructs (Liaw and Lengyel, 1992; Fig. 3). No single portion of the regulatory region was capable of driving the entire endogenous tll brain pattern through stage 13; rather, various brain expression modules are dispersed throughout the tll regulatory region.

Three regions were identified that drove some expression in the HL domain, starting at stage 6 (beginning gastrulation). One of these, deduced from the pattern driven by the P2 but not the P3 construct, lies in region B (−1253 to −2291, Fig. 3). Since expression in the HL domain driven by the P2 construct is only maintained to stage 11 (Fig. 7A-C), and since construct TM, containing region BCD₁ (−1253 to −2291) drives expression in the HL domain only early and transiently, there must be elements proximal to −763 that function synergistically with element(s) in BCD₁ to establish and maintain HL expression throughout embryo-genesis.

Two brain-specific elements were identified in the distal regulatory region. Constructs K3 and K4, but not K5, drive expression in a portion of the HL domain, localizing one brain-specific enhancer to a 243 bp region (Fig. 3; −3621 to −3864). Expression driven by K3 and K4 begins during stage 6 as a dorsal medial stripe (Fig. 7I) that overlaps with the portion of the HL domain where tll expression is strongest (cf. Fig. 1A,B). This expression by stage 8 has become a chevron-shaped domain that constitutes the most dorsal-medial portion of the HL domain (thus lying within the most dorsal part of the Pa domain and the dorsal portion of the Pc domain); we refer to this expression domain as HL_d (Fig. 7J). Expression driven by K3 and K4 continues in the HL_d domain into stage 11 (Fig. 7K) but not beyond. An important brain-specific enhancer is present in the 349 bp fragment carried by construct K7 (Fig. 3, −4400 to −4749). Construct K7 drives expression throughout the HL region at stage 6, in the HL_d region by stage 8, but drives little expression thereafter (data not shown). The fact that construct K9, which contains the HL enhancer between −3621 and −3864, cannot drive HL expression on its own, while construct K7 does, indicates that region −4400 to −4749 is the stronger of the two HL enhancers; these enhancers probably interact synergistically to drive brain neuroblast-specific expression.

A region containing an enhancer for the LL region can be deduced from the fact that constructs P2, P3 and P4 all drive expression in the LL region; the smallest of these constructs, P4, which contains the D region (Fig. 3, −908 to +460), drives expression during stages 6 through 11 in a pattern that overlaps the LL domain (cf. Figs 1A,C,E, 7E-G).

An optic lobe-specific enhancer within region −908 to −1253 is deduced from the fact that construct P3, but not P4, drives expression in the optic lobe as well as in immediately adjacent cells in the protocerebrum (Figs 3, 7D). No other elements mediating optic lobe expression were identified; specifically, the HL-specific enhancers identified above were not found to drive expression in the optic lobe (OL) domain.

Two head-specific repression elements can be inferred from comparing expression driven by K3 to K2, and K7 to K6. In each case, removal of a proximal fragment from a construct that does not drive anterior expression (K2 and K6) results in a construct that drives expression in the HL domain (K3 and K7) (Figs 3, 7). The regions mediating repression appear to be neuroblast-specific, since the K2 construct, which contains both of these regions, drives polar expression at the blastoderm stage. In addition, since they are immediately adjacent to HLspecific enhancers (described above) and the presence of one of them (−4215 to −4400) does not affect the enhancing effect of a more distant region (−3621 to −3864) (Fig. 3, constructs K3 and K4), these repression elements may only affect activation driven by an immediately adjacent enhancer. Since its addition to the P3 construct reduces ability to drive expression in the optic lobe but not at the poles during the blastoderm stage (Fig. 3), region −1253 to −2291 probably also contains a neuroblast-specific repression element.

Sequences conserved in the regulatory regions of distantly related species can provide suggestions of possible regulatory sites. Thus similarities between Drosophila melanogaster and Drosophila virilis in the −1 to −450 region of the tll 5′ regulatory region contributed to the identification of elements involved in regulation of the blastoderm polar caps (Liaw et al., 1993, 1995). To allow this rationale to be extended to the brain-specific modules, we obtained sequence for approximately 6 kb upstream of the tll transcription initiation sites for both Drosophila melanogaster and Drosophila virilis (GenBank accession numbers AF019362 and AF019361, respectively) and compared these to each other.

In addition to 450 bp promoter-proximal region of similarity, three additional regions of melanogaster tll regulatory DNA with similarity to virilis tll DNA were identified: −1300 to −1900, −2100 to −2900, and −3200 to −3650. How are these four regions of similarity related to the regulatory elements defined by promoter dissection? The −1 to −450 region overlaps with the LL element and one of the HL elements; the specific similar sequence elements in this region have been described (Liaw et al., 1993). The −1300 to −1900 region overlaps with a module that can independently drive early and transient expression in the HL and LL domains (Fig. 3); it is noteworthy that this region contains five 40 to 80 bp sequences that are highly conserved between melanogaster and virilis. The −2100 to −2900 region overlaps with the proximal neuroblast specific repressor. The −3200 to −3650 region does not overlap with regulatory elements defined here.

Surprisingly, for the two distal HL modules (−4400 to −4749; −3621 to −3864), one of which is capable of independently driving brain-specific expression, no region of sequence similarity was identified in the virilis tll regulatory DNA. In another approach, searching for repeated motifs within the melanogaster HL modules, we identified three repeats of the sequence TCTGG between −4611 and −4465 and four between −3763 and −3866. A number of these sites were identified in the virilis upstream sequence. While this repeat and the sequences in which it is embedded does not fit consensus binding sites of described Drosophila transcription factors, it does fit the consensus for a heterodimer of the two bHLH proteins E47 (the vertebrate homolog of the neurogenic Daughterless gene product) and Thing1 (Hollenberg et al., 1995).

Regulation of tll in the posterior of the Drosophila embryo under control of the tor RTK has been investigated in detail and has led to the definition of a tor-RE in the proximal tll regulatory region (Liaw et al., 1993; 1995). Dissection of the distal regulatory region described here has has allowed identification of additional, synergistically interacting tor-REs. In contrast to our relatively more detailed understanding of the early, tor-dependent regulation of tll, little is known about the regulation of tll in the developing brain. The work presented here provides a detailed description of tll expression throughout embryonic brain development, examines (and discards) possible regulators of tll, and most significantly, identifies specific enhancer elements (modules) that drive expression in different portions of the developing Drosophila brain.

We have mapped tll expression to a recently generated map of procephalic neuroblasts (Younossi-Hartenstein et al., 1996), and expanded an earlier description of tll expression in the brain (Younossi-Hartenstein et al., 1997). The earliest brain-related tll expression is in a horseshoe domain that covers the anlage of the entire brain; this domain then becomes split into two lateral expression domains containing high level (HL) and low level (LL) expressing regions. Later, tll is expressed most strongly in the earliest delaminating Pc2,3 and the later delaminating Pa3,4 protocerebral neuroblasts. Still later, tll is strongly expressed in the invaginating optic lobe. Neuroblasts that continue to express tll throughout embryogenesis are those in the most dorsal-medial protocerebrum and the optic lobe. Both the early horseshoe expression, as well as expression in the HL and LL domains (which together cover the entire protocerebral neurectoderm) are consistent with the tll mutant phenotype, which is the absence of the entire protocerebrum (Younossi-Hartenstein et al., 1997). The requirement for the later strong expression in the dorsal-medial protocerebrum and in the optic lobe remains to be addressed.

A number of genes that are expressed in the head, encode transcription factors and give brain phenotypes in mutant embryos were tested for their effect on tll expression, as assessed by in situ hybridization or expression of the P1 reporter construct. Neither the head gap genes otd and btd nor the genes of the ASC (the most relevant of which is l’sc) appeared to affect tll expression in the procephalic region, at least through stage 10. Furthermore, as tll mRNA is still seen in regions of the brain (optic lobe) that remain in a null tll mutant, it appears that tll does not regulate its own expression.

Regulation of tll in the embryonic brain thus remains something of a conundrum. Although it might be proposed that tll expression, once activated in the anterior horseshoe domain, is maintained (like the homeotic genes) by a complex of proteins of the Polycomb and Trithorax group (Orlando and Paro, 1995), the dynamic modulation of the tll expression domain into the HL and LL domains, and the appearance of tll expression at stage 11 in the OL domain renders such a mechanism improbable. Two other possibilities, not mutually exclusive, seem more likely: that additional genes regulating tll later during embryogenesis remain to be discovered, and/or that a constellation of partially redundantly functioning tran-scription factors, removal of any one of which does not have a strong effect, is required to generate the complex tll brain expression pattern.

Work described here, taken together with previous work primarily on the proximal region, provides a detailed map of the 5.9 kb tll regulatory region. Since this DNA can drive the endogenous expression pattern and also, as part of a construct containing the tll coding region (plus 2 kb of 3′ DNA of untested significance), rescue the null tll phenotype (Pignoni et al., 1990), it most likely contains all of the essential tll regulatory sequences. As well as identifying additional tor-REs mediating early expression in polar caps under control of the terminal system, we have identified both positive and negative elements mediating the HL, LL and OL domains of tll expression in the developing brain. The modules identified by our promoter dissection and genetic crosses are summarized in Fig. 8.

Like the previously described proximal region D₃ (yellow/orange in Fig. 8; Liaw et al., 1993), the distal region −2291 to−2770 (orange in Fig. 8) mediates polar expression in response to activation of the terminal system. Furthermore, this distal region contains two sequences similar or identical to the tor-RE identified in the proximal tll promoter region (D₃) (Liaw et al., 1995; Fig. 6F) and drives an expanded domain of expression in embryos with an ectopically active Torso receptor.

The novel result from our regulatory region dissection is the identification of elements (modules) that mediate different portions of the tll expression pattern in the procephalic neurogenic ectoderm. As the portion of the tll expression pattern that is conserved between Drosophila and vertebrates is the anterior, brain-specific expression (see below), identification of modules specifically mediating portions of this pattern could have more general significance. This is, to our knowledge, the first identification in either Drosophila or vertebrates of a brain-specific enhancer based on assays in the whole organism.

The fact that different DNA sequence elements drive different portions of the anterior pattern demonstrate that there is a complex basis for the pattern of tll expression in the procephalic region. Integration of results from regulatory region dissection with mapping of expression patterns allows us to conclude that there are at least three independently regulated domains of tll expression: HL, LL and OL. The phenotype of tll mutant embryos reveals that complete absence of tll activity starting from the syncytial blastoderm stage results in absence of the entire protocerebrum (Strecker et al., 1988; Pignoni et al., 1990; Younossi-Hartenstein et al., 1997). It is possible that continued expression in the distinct LL, HL and OL subdo-mains during later stages of embryogenesis is also required, and that absence of any one of these might result in more subtle defects in brain development.

The LL region encompasses most of the procephalic neurogenic area and is characterized by a relatively low and somewhat uneven expression of tll. The proximal regulatory region (+475 to −908) drives this expression (light blue, Fig. 8) throughout the procephalic neurogenic region, indicating that the HL domain is contained within the LL domain.

Strong expression in the triangular, dorsomedial HL region persists throughout much of embryogenesis (stage 6 to stage 13). Does the high level of tll expression in the HL domain make the HL cells different in some way from the surrounding LL cells? Since it covers only a portion of the protocerebral anlage, HL expression cannot function to define the (putatively) segmental protocerebrum as a unit. Furthermore, HL expression does not appear to result in an immediately obvious difference in cell behavior, since the morphology of neuroblast delamination in stages 10 and 11 is not noticeably different in the HL as compared to surrounding LL domain (V. H., unpublished observations). A third possibility is that HL expression plays a role in establishing a portion of the larval brain, the anlage for which has not yet been mapped in the embryonic procephalic neurogenic ectoderm. A candidate for such a brain region is the mushroom bodies, since they arise from a dorsomedial position in the protocerebrum (reviewed by Younossi-Hartenstein et al., 1996).

The ability to drive expression in the HL domain was localized to three different regions (two by subtraction); one of these regions, −4400 to −4749, is capable of independently driving expression in the HL domain. The lack of sequence similarity between Drosophila melanogaster and Drosophila virilis genomic DNA for these distal modules makes identification of significant sequences difficult. Multiple repeats of the sequence TCTGG, however, were found in the melanogaster HL elements (and also in the distal virilis genomic DNA). While TCTGG is not a described consensus binding site for a known Drosophila transcription factor, it is a site for a bHLH protein, one component of which is E47, the vertebrate homolog of the Drosophila Daughterless protein. It may be relevant that daughterless is required during Drosophila embryogenesis for commitment to the proneural fate, and that daughterless mutant embryos display brain defects (Caudy et al., 1988). The independently functioning −4400 to −4749 module provides a reagent for further investigation of the molecular and genetic basis of tll expression in the HL domain in the embryonic brain.

The Tlx protein of chickens and mice is 81% identical to the Tailless protein in the DNA-binding domain, and 41% identical in the ‘ligand binding’ (although no ligand is known for either) domain (Yu et al., 1994). These high levels of sequence identity, as well as the fact that more closely related genes have not been identified, indicate that Drosophila tll and vertebrate Tlx are orthologous genes.

In addition to this apparent common ancestry, related expression patterns suggest that the Drosophila Tailless and vertebrate Tlx proteins function in homologous regions of the brain. tll is expressed in the protocerebrum, the most anterior part of the primordial insect brain; similarly, Tlx is expressed in the most anterior portion of the vertebrate embryonic brain, the telencephalon and diencephalon (Yu et al., 1994; Monaghan et al., 1995). An additional similarity is that both tll and Tlx are expressed in visual centers of the brain. tll is expressed in the optic lobe of the stage 11/12 embryo and 4 days later in the optic lobe of the third instar larva (data not shown). Tlx is expressed in the optic vesicle and eye, which originate from the diencephalon; in the newborn mouse, Tlx expression is seen in in the optic tract and the portion of the brain giving rise to it (Monaghan et al.,1995).

Dissection of the tll regulatory region has identified multiple modules controlling expression in specific portions of the Drosophila embryonic brain. This information should pave the way for identification of brain-specific regulatory factors that could be studied both molecularly and genetically. Since, corresponding to the genes tll, otd and ems expressed in the Drosophila brain, there are orthologs Tlx, Otx and Emx similarly expressed in anterior regions of the vertebrate brain (Arendt and Nübler Jung, 1996), it is conceivable that insights gained into regulation of tll may increase our understanding of pathways of gene activity controlling development of the vertebrate brain.

We thank F. Pignoni and H. Kr ämer for providing some tll in situ hybridization data, and Wenjuan Silvia Yu for expert technical assistance. This work was supported by grants NIH HD09948 to J. A. L., NIH NS29367 to V.H., ACS NP-785 to J. A. L. and A. J. C., and fel-lowships from the UCLA Johnsson Cancer Center and NIH CA09956 to K. M. R. and USPHS NRSA GM-07014 to P. G.