The Drosophila tango gene encodes a bHLH-PAS protein that is orthologous to mammalian Arnt and controls CNS midline and tracheal development

The PAS domain is a large, multifunctional interaction domain found within a family of proteins that control a wide variety of biological processes in vertebrates, invertebrates, plants, and prokaryotes. These functions include neurogenesis, tubule formation, carcinogen metabolism, response to hypoxia, and biological rhythms. Most of the PAS proteins studied in mammals and Drosophila also contain a basic-helix-loop-helix (bHLH) domain that implicates them as DNA-binding transcription factors. One important aspect of PAS protein function is that some members can act as receptors and mediate cell signaling pathways. The mammalian aromatic hydrocarbon receptor complex (AHRC; also dioxin receptor) is one of the paradigmatic members of the bHLH-PAS family (Hankinson, 1995). The aromatic hydrocarbon receptor (Ahr) bHLH-PAS protein is localized in the cytoplasm in the ligand-free state. When ligand (e.g. dioxin) enters the cell, it binds Ahr, which then enter the nucleus. The Ahr-ligand complex binds another bHLH-PAS protein, Aromatic hydrocarbon receptor nuclear translocator (Arnt). The Ahr::Arnt heterodimer binds DNA sequence elements with a core GCGTG sequence, referred to as the xenobiotic response element (XRE), that reside within target genes such as that encoding Cytochrome P₁450, and activates their transcription. One of the key questions regarding PAS protein function is the generality of this paradigm. In particular, is the function of Arnt evolutionarily conserved, and does it function as a dimerization partner for multiple bHLH-PAS proteins?

The Drosophila single-minded (sim) bHLH-PAS gene is a master regulatory gene that controls the development of the neurons and glia that lie along the midline of the central nervous system (CNS). Genetic and ectopic expression studies reveal that sim function is required in the midline cells for activation of CNS midline gene expression (Nambu et al., 1990, 1991) and repression of lateral neuroectodermal expression (Chang et al., 1993; Mellerick and Nirenberg, 1995; Xiao et al., 1996). Thus, it acts as a genetic switch. When turned on in the neuroectoderm as cell lineages are being specified, sim causes the cells to enter the CNS midline cell lineage. Recently, two mammalian orthologues of sim (Sim1 and Sim2) have been discovered (Michaud and Fan, 1997). Both are expressed in the developing CNS. Sim2 is a candidate gene for Down Syndrome, based on its expression pattern and localization to chromosome 21. Transgenic and mutational analysis of target genes of Drosophila sim have identified a cis-regulatory CNS midline element (CME) that controls transcription of CNS midline cell transcription by sim (Wharton and Crews, 1993; Wharton et al., 1994). The CME has a core sequence of ACGTG. When multimerized, it is sufficient for CNS midline transcription in vivo. It was hypothesized, based on analogy to the AHRC, that Sim forms a dimer with a Drosophila Arnt-related protein to bind the CME and activate CNS midline transcription (Wharton et al., 1994). The work described in this paper describes a Drosophila Arnt-like protein and demonstrates that it functions as predicted.

More recently, several novel Drosophila and mammalian bHLH-PAS proteins have been identified. These include Drosophila trachealess (trh) (Isaac and Andrew, 1996; Wilk et al., 1996), the mammalian hypoxia inducible factor (HIF) (Wang et al., 1995), Drosophila similar (sima) (Nambu et al., 1996), and murine Clock (King et al., 1997). The trh gene plays a prominent role in tracheal development, by controlling the expression of genes involved in tracheal tubule formation. It also is involved in formation of the posterior spiracle and salivary gland duct. trh is specifically expressed in tracheal cells and the salivary gland. HIF controls the cellular response to hypoxia by activating the transcription of genes that protect the organism against oxygen deprivation. HIF consists of a heterodimer in which one subunit (HIF-1α) is related to Sim and the other (HIF-1β) is Arnt. Drosophila sima is most closely related to HIF-1α and is broadly expressed in the embryo. This suggests that Sima may play a role in Drosophila hypoxia induction, but sima mutations have not been isolated and its biological function remains unknown. The Clock gene controls the periodicity of biological rhythms in mice, and is postulated to activate circadian rhythm gene transcription. Although the Clock, Sim, Sima, Trh proteins are closely related, their interaction partners and mechanisms of gene control are not well understood.

In this paper, we report the cloning of the Drosophila tango (tgo) gene. Sequence, biochemical, and expression data indicate that tgo is highly related to mammalian Arnt. Tgo can dimerize strongly with both Sim, Sima and Trh and is able to activate gene transcription via the CME. Analysis of transgenic fly strains indicate that the CME acts in vivo as both a CNS midline and tracheal enhancer element, consistent with it being a regulatory element for Sim::Tgo and Trh::Tgo heterodimer binding. Mutations in the tgo gene reveal both CNS midline and tracheal defects, and genetic interaction studies suggest that sim and trh interact with tgo in vivo. These results provide in vivo evidence that Tgo/Arnt functions as a broadly expressed dimerization partner for a number of specifically expressed bHLH-PAS proteins that control a variety of developmental and physiological processes.

The following mutant loci were used: sim^H9 (null), trh⁸ (severe), neu^j6B12 (P[lacW] lethal insertion), and Df(3R)p-XT9 (84F14; 85CD). Enhancer trap and P[lacZ] transgenes used were: AA142 and X55 enhancer trap lines (Klämbt et al., 1991), P[w⁺; 4xCME-lacZ] transgenic line that contains a 4-fold multimerized Toll site 4 CME cloned into the C4PLZ enhancer tester vector (Wharton et al., 1994). The third chromosome sim^H9, tgo¹, and trh⁸ mutations were recombined individually onto a strain containing a P[4xCME-lacZ] that genetically maps to the distal tip of 3R.

The probe used to isolate the Drosophila tgo gene was derived from the bHLH region of the human ArntM1 cDNA clone (Hoffman et al., 1991). PCR was used to generate a fragment of 206 bp (nt 208-415) that included all of the Arnt bHLH region. Genomic clones were isolated from an EMBL3 bacteriophage library of isogenic dp cl cn bw DNA (R. Blackman, University of Illinois) using moderately stringent hybridization conditions consisting of incubation in 6× SSPE/50% formamide/10% dextran sulfate/0.1% SDS/0.1% NaPPi/5× Denhardt’s at 42°C for 18 hours with a 10⁷ disints per minute/ml probe followed by 3 washes at 25°C in 2× SSC/0.1% SDS for 15 minutes each and 2 washes at 55°C in 1× SSC/0.1% SDS for 30 minutes/each. Twelve clones (λ TG-1-12) were analyzed by restriction enzyme cleavage and shown to correspond to a single gene. The human ArntM1 bHLH probe was also used to screen a cDNA library derived from 4- to 8-hour embryonic RNA (N. Brown). Twelve cDNA clones (pCT-1-12) corresponding to tgo were isolated from 500,000 cDNA clones screened.

DNA was sequenced at the UNC-CH Automated DNA Sequencing Facility on a Model 373A DNA Sequencer (Applied Biosystems) using the Taq DyeDeoxy^™ Terminator Cycle Sequencing Kit (Applied Biosystems). Four tgo cDNA clones, pCT-5, 7, 8, and 12, were completely sequenced. The 1.8 kb HindIII-EcoRI and 2.6 kb BglII-HindIII restriction fragments that contain the entire tgo gene were also sequenced (data not shown). DNA was sequenced from two EMS-generated tgo mutants (tgo¹ and tgo²) and the parental strain they were generated from. Each tgo mutant was balanced with TM3 P[UbxlacZ]. Embryos were collected and stained for β-galactosidase expression using the X-gal reaction without fixation. Embryos that failed to turn blue were considered homozygous mutant for tgo, DNA was extracted, the entire tgo gene amplified by PCR, and sequenced. Three independent PCR reactions were subcloned and sequenced for each mutant strain to correct for PCR-generated artifacts.

To generate antibodies against Tgo, the bHLH PAS domain was cloned by PCR from the pCT-8 cDNA clone and inserted in-frame into pGEX-2T′6, a bacterial, glutathione S-transferase (GST) fusion protein expression plasmid. The Tgo fragment encodes amino acids 1-393 that includes the complete bHLH and PAS domains. Mice were immunized by subcutaneous injection of 50 μg of fusion protein, diluted in Ribi Adjuvant, every 21 days. Monoclonal antibodies were generated and screened by ELISA and embryo staining. Three different anti-Tgo monoclonal antibodies (mAbs) were isolated and designated mAb-Tgo1-3.

Antibody staining of embryos using HRP/DAB histochemistry was carried out according to standard protocols (Patel et al., 1987). Antibodies used were: mAb anti-β-galactosidase (Promega), mAb BP102, mAb 2A12, and mAb-Tgo1.

The yeast interaction system used in this study was the LexA system of Brent and colleagues (Brent, 1994). All bait constructs were cloned into the pEG202 vector, which fuses the sequence to be tested onto a LexA DNA binding domain. The prey plasmids were cloned into pJG4-5, which fuses the protein to be tested onto an acid blob transcriptional activation domain. Both bait and prey plasmids were transformed into the yeast strain EGY48, which contains the pSH18-34 lacZ reporter gene preceded by eight LexA binding sites. At least three independent colonies were picked for each bait-prey pair and assayed in the presence or absence of galactose for β-galactosidase expression using either solution or colony enzyme assays. β-galactosidase enzyme activity of transformants was quantitated using the chromogenic substrate o-nitrophenyl-β-D-galactoside (ONPG) in a standard enzyme assay, and the colony assay was a modification of the standard freeze-thaw assay.

Sf9 cells were infected individually or in combination with recombinant viruses containing full-length cDNA sequences of sim, tgo and trh using the Bac-to-Bac system (Gibco-BRL). The coinfection experiments were carried out at a m.o.i. of 2 for cells co-infected with Bacsim and Bac-tgo. Cells co-infected with Bac-trh and Bac-tgo, used a m.o.i. of 0.2 for Bac-trh and 20 for Bac-tgo. Individually infected cells utilized the same m.o.i. as coinfected cells. Cytoplasmic extracts were prepared and subjected to immunoprecipitation assays (Gradin et al.,1996) using mAb-Tgo1, rabbit α-Sim, rat α-Trh and Protein-A Sepharose (Pharmacia) followed by western blot analysis using the same antibodies except α-Sim, which was a rat antibody.

cDNA clone fragments containing the complete coding sequences of sim, tgo and trh were PCR-cloned into the vector pAct5C, which contains the Drosophila actin 5C promoter and poly(A) site (Han et al., 1989). The reporter plasmid contained 6 Toll site 4 CMEs fused to lacZ in C4PLZ. Standard transfections of Drosophila SL2 cells contained 10 mg copia-LTR-CAT as a transfection control and 10 mg of each transfected DNA. Each transfection was done in triplicate. 48 hours after transfection, cells were lysed and b-galactosidase and CAT activity assayed. CAT activity was quantitated using FAST-CAT (Molecular Probes) and b-galactosidase activity was quantitated using a LacZ/Galactosidase Quantitation Kit (Molecular Probes). Fluorescence units derived from the b-galactosidase assay were standardized using percentage conversion values of CAT assays as a control for transfection efficiency.

The starter strain for the local P-element transposition was neu^j6B12 which contained a P[w⁺; lacW] P-element in the 5′ region of the neu gene (Berkeley Drosophila Genome Project). The P[w^+; lacW] element was mobilized in the male germline by mating w; neu^j6B12/TM3 females to males containing the w; Δ 2-3 Sb/TM6 transposase source. Mosaic male progeny (w; neu^j6B12/Δ 2-3 Sb) were mated 2 per vial with 3 w; TM3/TM6B Tb Hu females to stabilize the mobilization events. Male or female progeny were scored for eye color changes from the original orange-eyed starter strain. The eye color changes can signify alterations in the copy number and/or positions of the P[w⁺; lacW] element with respect to its original site within the genome. An inverse PCR screen was performed to identify P-element insertions into the tgo gene according to the method of Dalby et al. (1995) with modifications by Bob Nelson (personal communication). Identification of the genomic DNA sites flanking the novel P-element insertions was achieved by plasmid rescue and subsequent hybridization and DNA sequence analyses.

EMS mutations in tgo were identified using an F₂ lethal screen in which EMS mutations were screened over the R65 P-element insertion line for lethality. Male w¹¹¹⁸ flies were fed 25 mM EMS in a 1% sucrose solution (Lewis and Bacher, 1968), and mated en masse to w; TM3 Sb e/TM6 Tb e female flies. Male progeny (2092) of the genotype w; */TM6 Tb e were mated individually with w; R65/TM6 Tb e females. Complementation to R65 was assayed by scoring the vials for the presence of Tb⁺ pupae. Vials containing Tb⁺ pupae were discarded and stocks established from w⁻ flies in the vials with only Tb flies. Flies from the seven stocks selected were tested for complementation inter se and with neu mutations. Four of the mutations failed to complement neu and the other 3 made up a novel complementation group shown to be tgo.

Sequence comparison of bHLH proteins between divergent species indicates that the bHLH region is often highly conserved with respect to the rest of the protein. Cloning of a Drosophila Arnt-related gene was carried out by screening a Drosophila genomic library with a human Arnt bHLH probe (Hoffman et al., 1991). Positive clones were identified, restriction mapped, and shown to correspond to a single gene, subsequently named ‘tango’. Sequence analyses of the complete gene and corresponding embryonic cDNA clones indicate that tgo is highly related to mammalian Arnt (Fig. 1), both in sequence and predicted structure. The bHLH region is near the N terminus, a feature common to all bHLH-PAS proteins, followed closely by the PAS domain, and then glutamine-rich C-terminal domains. The bHLH regions are 92% identical and PAS domains 53% (Fig. 1B). The C-terminal regions are generally unrelated in primary sequence, but share the occurrence of glutamine-rich sequences (18% in Tgo), which are activation domains in mammalian Arnt (Jain et al., 1994; Li et al., 1994), Drosophila Sim (Franks and Crews, 1994), and many other transcription factors (Mitchell and Tjian, 1989), and the entire region is proline-rich (15% in Tgo). In particular, the poly[glutamine] regions are closely flanked by proline residues. One interesting feature of the Drosophila Tgo C terminus not found in other Arnt proteins is the presence of a histidine-proline-rich region of unknown function found in a small number of other Drosophila transcription factors including the Paired segment polarity and Bicoid homeobox proteins. This region is called the Paired (PRD)-repeat (Frigerio et al., 1986), and is distinct from the ‘Paired domain’.

Sequence analysis of tgo genomic and cDNA clones revealed a relatively small gene of 2.9 kb with a single intron of 142 bp within the 5′-untranslated region (Fig. 2; nucleotide sequence data not shown). The simplicity of the exon-intron structure of the tgo gene is in contrast to the mammalian Arnt gene that is much larger with a more complex exon-intron structure (Maltepe et al., 1997). The sequence upstream of tgo revealed that the 5′-end of the longest tgo cDNA clone is 328 bp 3′-to the neuralized (neu) gene in the same orientation (Fig. 2). Since neu has been mapped by polytene chromosome in situ hybridization to 85C (Boulianne et al., 1991), tgo resides in this cytological vicinity. Furthermore, the PRD repeat sequence of tgo is identical to the previously reported prd7 gene sequence homology that was mapped cytologically to 85C (Frigerio et al., 1986; Price et al., 1993).

Embryonic expression of the tgo gene was examined by staining embryos with a monoclonal antibody generated against a bacterial Tgo fusion protein. Specificity of the reagent was indicated by the similarity between in situ hybridization (data not shown) and antibody staining with monoclonal and polyclonal antibodies, and reduction of expression in embryos homozygous for a deficiency of the tgo gene (Fig. 3F). Tgo protein is found in cell nuclei or cytoplasm depending on cell type and time of development (detailed description of subcellular Tgo localization will be published elsewhere). In the pre-cellular blastoderm, Tgo protein is uniformly distributed (Fig. 3A), presumably due to maternal contribution. Staining is more intense in the cytoplasm than the nucleus. During the extended germband stage, Tgo protein is detected in all three germ layers (Fig. 3B). As tracheal pits form, the cells surrounding the pits show enhanced levels of Tgo protein (Fig. 3C) and RNA (data not shown) compared to surrounding cells. Amounts of Tgo protein above the ubiquitous levels continue to be observed in the tracheal cells including the posterior spiracles from stage 11 to the end of embryogenesis at stage 17 (Fig. 3D,E). As the CNS forms, uniformly high levels of Tgo protein are found in the brain (supraesophageal ganglia) and ventral nerve cord (Fig. 3E). In summary, Tgo protein is found broadly distributed throughout embryogenesis, although certain cell types including trachea and CNS have enhanced levels.

Direct evidence that Tgo can dimerize with Drosophila bHLHPAS proteins was obtained using the yeast two-hybrid system. The full-length Tgo bHLH-PAS protein was used as a prey with bHLH-PAS protein baits of Drosophila Sim, Sima, and Trh. The results, shown in Fig. 4A, indicate that Tgo interacts strongly with all three bHLH-PAS proteins, and the levels of activation are comparable. Additional evidence for biochemical associations between Sim and Tgo and Trh and Tgo were revealed in co-immunoprecipitation experiments employing Sf9 cells infected with recombinant baculoviruses that express Sim, Tgo and Trh. These experiments showed that Sim and Tgo and Trh and Tgo formed stable complexes when co-infected (Fig. 4B). Since the cytoplasmic fraction, which does not include DNA, was analyzed, the associations are due to direct protein-protein interactions and not to independent binding of each protein to the same DNA molecules.

Additional two-hybrid interaction assays were carried out to determine the evolutionary conservation of Tgo/Arnt in its ability to interact with bHLH-PAS proteins, and to examine the rules that govern Drosophila bHLH-PAS protein interactions. rules that govern Drosophila bHLH-PAS protein interactions. Trans-species interactions with Tgo can occur since Tgo interacts strongly with murine Ahr (Fig. 4A). Additional experiments were carried out using Drosophila bHLH-PAS proteins and human Arnt in all pairwise combinations, both as bait and prey. The results (Table 1) indicate that the Drosophila Sim, Sima and Trh proteins interact strongly with human Arnt, as did murine Ahr. Homodimerization was observed with human Arnt, a result observed in some published accounts (Sogawa et al., 1995; Swanson et al., 1995), but not others (Reisz-Porszasz et al., 1994). However, no other bHLH-PAS protein examined was able to form homodimers. Heterodimerization was only observed between bHLH-PAS proteins and Tgo/Arnt, no heterodimers were formed between pairwise combinations of Sim, Sima, Trh and Ahr. These results show that both Drosophila and mammalian Arnt are not only highly related in sequence and expression, but can form heterodimers with other bHLH-PAS proteins, regardless of species.

Direct evidence that Tgo and other bHLH-PAS proteins can dimerize, bind DNA, and activate transcription through the CME was obtained using transient transfection assays with Drosophila SL2 cell culture. Western blot analysis indicated that Tgo protein is present in SL2 cells, but Sim and Trh are absent (data not shown). Full length cDNAs encoding each of these proteins were cloned into vectors that express them constitutively under the control of the actin5C promoter (Han et al., 1989), yielding the plasmids pAct-sim, pAct-tgo, and pActtrh. The reporter gene (P[6xCME-lacZ]) contained six copies of the CME (the putative Sim::Tgo and Trh::Tgo binding sites) fused to a P-element promoter driving lacZ. This is the same DNA construct that functions as a midline (Wharton et al., 1994) and tracheal (see below) enhancer in vivo, except that the in vivo studies were carried out with a reporter containing only 4 copies of the CME.

When pAct-sim and pAct-tgo were cotransfected with the P[6xCME-lacZ] reporter, lacZ transcription was induced to high levels (Fig. 5). Control experiments indicated that high levels of lacZ activation required the presence of the CME and overexpression of both Sim and Tgo. The magnitude of reporter gene activation in cultures transfected with pAct-sim and pAct-tgo compared to those transfected with only pActsim correlates with the magnitude of increased Tgo protein detected by western analysis (data not shown). The enhanced levels of lacZ reporter expression that were obtained when both Sim and Tgo were co-expressed at high levels suggests that the two proteins interact to activate CME-regulated transcription. Similar experiments were performed with pAct-trh and pActtgo; Trh and Tgo also interact to activate CME transcription.

tgo maps to 85C on the third chromosome. Mutations in tgo were generated using a two-step strategy. The first step involved hopping P-elements from the adjacent neu gene into the tgo gene (Dalby et al., 1995; Tower et al., 1992; see Materials and Methods). The P-element mutant tgo strain was then used to screen for EMS-induced tgo mutations. The local hop involved crossing P-element transposase into neu^j6B12 and identifying progeny flies with an eye color darker or lighter than the orange color of the parental strain. It is expected that some of these strains will have novel P-element insertions. Flies from 120 distinct lines that had altered eye color were screened using an inverse PCR/Southern blot procedure (Dalby et al., 1995). Three lines (R65, R91, and R94), all with red eyes, were found to have novel P-element insertions in or near the tgo locus (Fig. 2).

Sequence analysis indicated that R65 was located in exon 2 in the 5′-untranslated region 43 bp 5′ to the proposed initiator methionine (Fig. 2). This insertion is likely to disrupt transcription of the tgo gene, resulting in a severe mutation. The R65 strain retained the original P-element near the neu gene as well as the newly generated P-element insertion into tgo. Cellular analysis of the neu tgo double insertion strain consistently showed more severe embryonic defects than neu^j6B12 in the structure of the trachea (data not shown). The neu tgo double insertion line was used to screen 2092 EMS-mutagenized third chromosomes. Seven EMS mutations were obtained that were lethal over both the neu tgo double insertion strain and Df(3R)p-XT9, a chromosomal deletion of 85C. Complementation experiments revealed that 4 were new mutations of neu and the other 3 corresponded to a novel complementation group shown to be tgo.

Confirmation that the new lethal complementation group corresponded to the tgo gene was obtained by DNA sequence analysis of two of the tgo mutants (Fig. 1). The tgo¹ mutant has a termination codon at aa 532 that deletes a proline-rich region, the PRD repeat, and a glutamine-rich region from the protein. The tgo² mutant has a termination codon at aa 518 that deletes another poly[glutamine] stretch in addition to the region deleted in tgo¹. These experiments confirm that the tgo mutations correspond to the Arnt-related transcription unit and that tgo is a vital locus.

Mutant tgo embryos were examined for defects in CNS midline and tracheal development, which would be expected if Tgo is a dimerization partner for Sim and Trh. The tgo¹ mutant was used in all of the experiments described below. The tgo² and tgo³ mutants qualitatively show similar phenotypes to tgo¹, but quantitatively are more severe. tgo¹ and tgo² are likely to be hypomorphic, since they cause quantitatively more severe phenotypes when analyzed in trans to the Df(3R)p-XT9 chromosome that is deficient for tgo. Defects in CNS midline neurons and glia were examined using the AA142 and X55 enhancer trap reporters, respectively (Klämbt et al., 1991); expression of lacZ in both of the genes is absent in sim mutant embryos (Sonnenfeld and Jacobs, 1994). In wild-type embryos, the AA142 enhancer trap gene is expressed in an average of 3.5 midline glia/segment by stage 14 of embryogenesis (Fig. 6A). In tgo¹ mutant embryos, there was a reduction in the number of stained midline glia to approximately 1 cell per/segment (Fig. 6B). The X55 enhancer trap gene stains the ventral unpaired median neurons (VUMs) and the median neuroblast (MNB) and its progeny in the ventral region of the CNS (Fig. 6C). In tgo¹ mutant embryos, the number of VUM neurons and MNB progeny were reduced in number (60% of wild-type) and did not migrate into the ventral regions of the VNC (Fig. 6D). Mutations in trh result in severe defects in tracheal tubule formation (Younossi-Hartenstein and Hartenstein, 1993; Isaac and Andrew, 1996; Wilk et al., 1996). The role of tgo in tracheal development was examined by staining tgo¹ mutant embryos with monoclonal antibody 2A12 that stains the lumen of the tracheal tubes (Fig. 6E). The results revealed that tgo¹ mutant embryos have a variety of tracheal defects, some weak (Fig. 6F) and others more severe (Table 2). The relatively weak phenotypes of tgo¹ mutations compared to sim and trh mutations are likely due to phenotypic rescue by maternal tgo (Fig. 3A) and the hypomorphic nature of tgo¹.

Previous work on putative target genes that are expressed in the CNS midline cells revealed the existence of the CME, an enhancer element both required and sufficient for CNS midline transcription (Wharton et al., 1994). Multimerization (4×) of the CME drives CNS midline cell expression from a heterologous promoter (Fig. 7A). Further characterization of the P[4xCME-lacZ] transgene shows that it is also expressed in the developing and mature trachea, posterior spiracles, and salivary ducts (Fig. 7A,B). This expression pattern resembles the combined sim and trh expression patterns, and suggests that the CME is an in vivo target element of Sim, Trh and Tgo.

Tests of the genetic control of CME expression by trh, sim, and tgo were performed by crossing P[4xCME-lacZ] into the three mutant backgrounds. All CNS midline lacZ expression from P[4xCME-lacZ] was abolished in sim mutant embryos, although tracheal expression was normal (Fig. 7C). Similarly, all tracheal lacZ expression from P[4xCME-lacZ] was abolished in trh mutant backgrounds (Fig. 7D), while CNS midline expression was unaffected. In tgo¹ mutant embryos, both CNS midline and tracheal expression were reduced (Fig. 7E,F), although not as severely as observed in sim and trh mutant embryos, perhaps due to residual maternal tgo and the hypomorphic nature of tgo¹.

The CNS midline and tracheal defects observed in tgo mutant embryos are consistent with Tgo being a dimerization partner for Sim and Trh and binding to the CME in vivo. The less severe phenotypes observed in tgo¹ mutants compared to sim and trh are interpreted as being due to the occurrence of maternal tgo and the hypomorphic nature of tgo¹. We have used the relatively weak phenotypes observed with the tgo¹ mutant strain to genetically test for in vivo interactions between sim-tgo and trh-tgo. Both sim and trh were individually recombined onto tgo¹, and crosses done to generate embryos with variable copies of mutant and wild genes. Embryos were then assayed for CNS and tracheal defects.

mAb BP102 was used to examine the CNS for indication of midline defects (Thomas et al., 1988). 100% of the embryos that are homozygous mutant for sim show a severe collapsed phenotype (Fig. 6J; Table 2; Thomas et al., 1988). Embryos that are sim/+ showed 100% wild-type CNS (Fig. 6H; Table 2). Mutant embryos that are homozygous for tgo¹ are either wild-type (58%) or have a mild defect (42%) in axon morphology (Fig. 6I). None of the tgo₁ mutant embryos show a severe collapsed ‘sim’ phenotype (Table 2). In contrast, 57% of the embryos that are tgo¹sim/tgo¹ + have a severe collapsed CNS (Fig. 6K; Table 2). Corresponding results were obtained using P[4xCME-lacZ] as a CNS midline marker: most tgo¹sim/tgo¹ + embryos had a complete absence of CNS midline β-gal staining (Fig. 7G), whereas tgo¹ +/tgo¹ + embryos generally retained staining although it was reduced from wild-type (Fig. 7E). Similar results were obtained for mutant embryos that were homozygous for tgo₁ and heterozygous for trh. When stained with mAb 2A12, which visualizes the tracheal lumen, 96% of the embryos homozygous mutant for trh show a complete absence of trachea, and 100% of trh heterozygotes have wild-type trachea (Table 2). Most tgo¹ homozygous mutant embryos (86%) have mutant trachea, but the phenotypes are generally weak (Fig. 6F); 97% have some mAb 2A12-staining material, and only 14% have severe defects. In contrast, 42% of embryos that are trh tgo¹/+ tgo¹ completely lack tracheal staining (Fig. 6G) and another 19% had a severe phenotype. Thus, the loss of a single copy of trh in the tgo¹ homozygous mutant background resulted in greater than a 4-fold increase in severe tracheal defects. In summary, the results of the genetic experiments provide strong evidence that sim and trh interact with tgo in vivo to control CNS midline and tracheal transcription and development.

These results show that Arnt is a highly conserved protein that functions as a dimerization partner for a number of bHLH-PAS proteins. The Drosophila tgo gene is orthologous to mammalian Arnt, as indicated by the following criteria. (1) The primary sequence and protein domain organization of the two proteins are strongly conserved. (2) The expression patterns of tgo and mammalian Arnt (Abbott et al., 1995; Abbott and Probst, 1995) are also similar. They are both transcribed in most if not all cells of the embryo. Another mammalian Arnt gene, Arnt2, is expressed in a subset of cells in the brain (Hirose et al., 1996), and, for this reason, tgo resembles Arnt more than Arnt2. (3) Interaction studies indicate that Tgo and mammalian Arnt act interchangeably as dimerization partners for other phylogenetically diverse bHLH-PAS proteins. These observations demonstrate that Tgo/Arnt acts as a broadly expressed dimerization platform for bHLH-PAS proteins in both mammals and Drosophila. Recently, Arnt has been identified in fish (Pollenz et al., 1996) and C. elegans (Genome Project). Thus, Arnt constitutes an evolutionarily conserved transcriptional regulator found in most, if not all, multicellular animals.

Together, Arnt/bHLH-PAS protein complexes control a wide variety of developmental and physiological processes. Given the strong evolutionary conservation between Arnt, sim, their genomic target DNA sequence elements, and their accessory proteins including Hsp90, it is apparent that Arnt represents a central element in an evolutionarily conserved gene regulatory pathway. In some cases, these transcriptional regulatory pathways may be responsive to ligand::bHLH-PAS protein interactions (e.g. AHRC). Identification of tgo in Drosophila opens up the possibility of sophisticated genetic analysis of these regulatory and signaling pathways. These possibilities are realized in this work in which genetic interactions between tgo and other bHLH-PAS genes are clearly revealed.

Biochemical studies in mammals have implicated Arnt as a partner for Ahr and HIF-1α. We have provided evidence that Tgo associates in vivo with Sim and Trh using several criteria. (1) Tgo protein is expressed in all cells in the embryo and thus overlaps in expression with Sim in the CNS midline cells and

Trh in the tracheal cells. (2) Studies using the yeast two-hybrid assay indicate that Tgo can form dimers with Sim, Trh, and Sima. (3) Co-immunoprecipitation experiments with baculoviral-expressed proteins indicates that Tgo forms dimers with Sim and Trh. (4) SL2 transient expression assays suggest that Sim and Tgo and Trh and Tgo interact to bind DNA and activate transcription. (5) Mutations in tgo result in CNS midline and tracheal defects. (6) Gene dosage experiments further suggest that sim and tgo and trh and tgo interact to control CNS midline and tracheal cell transcription and development, respectively. This work firmly establishes that tgo functions in CNS cell fate specification and formation of a functional respiratory system.

The interaction experiments described in this paper and others indicate that Tgo/Arnt is a common dimerization partner for bHLH-PAS proteins. These bHLH-PAS proteins form heterodimers exclusively with Tgo/Arnt and not with each other. In addition, they do not form homodimers while Arnt does. Even though Arnt homodimers can be detected, no in vivo function has been reported. As more PAS proteins are discovered, exceptions to these rules may emerge (e.g. Hogenesch et al., 1997; Tian et al., 1997). Never-the-less, it is convenient to think of Arnt as a common PAS dimerization partner playing a role similar to that observed for E2A and Max for other classes of bHLH proteins.

The CME acts as an enhancer element for both CNS midline and tracheal transcription. CME transcription within the CNS midline is dependent on sim function, CME transcription in the trachea is dependent on trh function, and tgo function is required for CME transcription in both cell types. The CME is sufficient for transcription in these cell types since multimerized Toll site 4 is sufficient to drive transcription specifically within the CNS midline and tracheal cells. These results suggest that Sim::Tgo heterodimers control CNS midline transcription through the CME, and that Trh::Tgo heterodimers control tracheal transcription through the CME. Consistent with this view are the transient transfection results indicating that co-introduction of either Sim and Tgo or Trh and Tgo induce high levels of transcription from a multimerized CME reporter gene. Nevertheless, complete molecular dissection and identification of all possible protein::DNA interactions within the 20 bp Toll site 4 have not been attempted, and it is possible that regulatory proteins in addition to Sim::Tgo and Trh::Tgo may bind to this DNA and help control its CNS midline and tracheal enhancer function.

Our results suggest that for a gene to be expressed in both CNS midline precursor and tracheal cells, the only requirements are a promoter and multiple copies of the CME. Although synthetic genes of this type are expressed in both cell types, expression of authentic target genes of sim, trh, and tgo are more complex. Minimally, there are six classes of putative target genes that may utilize a CME. Their expression in vivo (including an example) are: (1) CNS midline and trachea (MT) breathless (Klämbt et al., 1992), (2) CNS midline only (M) sim autoregulation (Nambu et al., 1990), (3) trachea only (T) trh autoregulation (Wilk et al., 1996), (4) CNS midline glia (MG) – slit (Wharton et al., 1994), (5) sim midline repression (MR) – ventral nervous system defective (Mellerick and Nirenberg, 1995; Xiao et al., 1996), and (6) hypoxia (H) – HIF1α (Wang et al., 1995; Nagao et al., 1996). Since the CME by itself drives expression in both the CNS midline and trachea, CME-controlled expression would acquire specificity via the addition of repressor elements. We would predict that the MT class of genes would require only multiple copies of the CME to be expressed in both cell types. In contrast, the M and T classes may require the presence of cell-specific repressors to restrict the expression of CME-driven processes to only a single cell type. For example, the regulatory regions of genes such as sim and Toll, in which the CME has been directly implicated in controlling CNS midline, but not tracheal, transcription (Wharton et al., 1994) may possess binding sites for tracheal repressors. The presence of midline precursor and midline neuron repressors may also restrict expression of the slit MG enhancer to midline glia (Wharton and Crews, 1993).

The studies described in this paper establish that Tgo plays a role in CNS midline cell development and tracheal tubule formation. In particular, Tgo interacts with Sim just after gastrulation to activate midline precursor gene transcription. All genes expressed in the CNS midline cells that have been tested require sim (and presumably tgo) function. In this sense, tgo and sim control specification of the CNS midline cell lineage. trh and tgo, on the other hand, are not required for transcription of all tracheal genes, but only a subset that are hypothesized to control tubule formation (Isaac and Andrew, 1996).

There are likely to be additional roles for tgo. It is shown here that it interacts biochemically with the Drosophila Sima protein. The function of Sima is currently unknown, but its ubiquitous expression pattern and primary sequence suggest it may be related functionally to mammalian HIF-1α and control the Drosophila response to hypoxia. Drosophila cell culture experiments have revealed the existence of a CME-binding factor that is induced under hypoxic conditions (Nagao et al., 1996). Genetic analysis of tgo mutants should confirm whether tgo is implicated in controlling the hypoxia response, regardless of the role of sima.

Consistent with the possibility that Tgo has additional partners, tgo mutants are abnormal in their general embryonic morphology distinct from sim and trh defects; we are currently analyzing the defects in more detail. Models have also been proposed that suggest that bHLH-PAS proteins, including Clock and Arnt, may interact with the Per PAS protein to control the periodicity of biological rhythms (Huang et al., 1993; Lindebro et al., 1995; King et al., 1997). Thus, it seems likely that additional roles for tgo will be uncovered upon further analysis and with the generation of germline clones and postembryonic mosaics. In summary, it is proposed that Arnt/tgo is the centerpiece of an ancient metazoan regula-tory/signaling pathway that has diverged to carry out numerous biological roles.

The assistance of Satish Gopal, Jackie Stilwell, and Brad Tewes is gratefully acknowledged. We would like to thank Chen-Ming Fan, Oliver Hankinson, Bob Nelson, Mark Peifer and Michael Rosbash for stimulating and useful discussions. We would also like to thank Debbie Andrew, Ron Blackman, Gabriele Boulianne, Corey Goodman, Oliver Hankinson, Ruth Lehmann, Jim Manley, Nipam Patel, Michael Rosbash, Susan Shepherd and Benny Shilo for providing essential antibodies, clones and fly stocks, and Russ Finley and Roger Brent for two-hybrid yeast strains and plasmids. We would also like to acknowledge the North Carolina State University monoclonal antibody facility for production of Tango antibodies, the Bloomington Drosophila Stock Center for providing Drosophila stocks, and the NICHD Developmental Studies Hybridoma Bank for monoclonal antibodies. This work was supported by a Canadian MRC fellowship to M. S. and an American Cancer Society Fellowship to M. W. Research is supported by NIH grant RD25251 to S. T. C.