|Home | About | Journals | Submit | Contact Us | Français|
Steroid hormones fulfil important functions in animal development. In Drosophila, ecdysone triggers moulting and metamorphosis through its effects on gene expression1. Ecdysone works by binding to a nuclear receptor, EcR, which heterodimerizes with the retinoid X receptor homologue Ultraspiracle2,3. Both partners are required for binding to ligand or DNA4–6. Like most DNA-binding transcription factors, nuclear receptors activate or repress gene expression by recruiting co-regulators, some of which function as chromatin-modifying complexes7,8. For example, p160 class coactivators associate with histone acetyltransferases and arginine histone methyltransferases9. The Trithorax-related gene of Drosophila encodes the SET domain protein TRR. Here we report that TRR is a histone methyltransferases capable of trimethylating lysine 4 of histone H3 (H3-K4). trr acts upstream of hedgehog (hh) in progression of the morphogenetic furrow, and is required for retinal differentiation. Mutations in trr interact in eye development with EcR, and EcR and TRR can be co-immunoprecipitated on ecdysone treatment. TRR, EcR and trimethylated H3-K4 are detected at the ecdysone-inducible promoters of hh and BR-C in cultured cells, and H3-K4 trimethylation at these promoters is decreased in embryos lacking a functional copy of trr. We propose that TRR functions as a coactivator of EcR by altering the chromatin structure at ecdysone-responsive promoters.
TRR10 was identified as a protein that is closely related to the well-known HOX gene activator Trithorax (TRX). An antibody to TRR recognized a protein band of the expected size (relative molecular mass (M r) 260) in nuclear extracts from embryos (Fig. 1b), suggesting that TRR is a nuclear protein. Nuclear localization of TRR was confirmed by immunohistochemical staining of salivary glands from third instar larvae (Fig. 1c). TRR contains several protein domains that are conserved among transcriptional regulators (Fig. 1a), including a SET domain. This ancient domain is known to bind to several proteins, including histones11, and to have histone methyltransferase (HMTase) activity (reviewed in ref. 12). Indeed, TRR from nuclear extracts bound to a column containing the four core histones; and the isolated SET domain of TRR, like the closely related TRX SET11, bound to histones H3 and H4 (Supplementary Fig. 1a, b).
Purified TRR SET showed histone-H3-specific HMTase activity in vitro (Fig. 1d, e). It also methylated a histone H3 amino-terminal peptide that was either unmodified or trimethylated at lysine 9 (K9), but it did not methylate a similar peptide that was already trimethylated at K4 (Fig. 1d). To test whether other lysine residues known to be methylated in histone H3 are methylated by TRR SET, we tested histones mutated at K4, K27, K36 and K79 (ref. 13). Only mutation of K4 affected the ability of histone H3 to be methylated (Fig. 1e). K4 specificity was confirmed by a failure of antibodies specific for H3-MeK27 and H3-MeK36 to recognize the product of TRR methylation (data not shown). K4 was directly confirmed as a target of TRR methylation in histone H3 by Edman degradation of H3 methylated in vitro by TRR SET (Fig. 1f). Further analysis showed that TRR SET can trimethylate at H3-K4 (Fig. 1g). Because trimethylated H3-K4 is associated with high transcriptional activity14, these results indicate that TRR may function as a transcriptional coactivator.
Ash1 and TRX are also H3-K4 HMTases in Drosophila15,16. TRR seemed to be more abundant in embryos than either Ash1 or TRX (data not shown), and it was associated with many more sites on salivary gland polytene chromosomes (Fig. 2c). To test the magnitude of TRR’s contribution to H3-K4 methylation, we compared the amounts of H3-MeK4 in extracts from wild-type embryos with those from embryos homozygous mutant for the trr1 null allele10. Mutation of trr markedly reduced the quantity of both di- and trimethylated H3-K4 (Fig. 1h). By contrast, we detected only a small decrease in methylated H3-K4 in embryos homozygous for a trx null mutation (trxB11, data not shown). We conclude that TRR is a major H3-K4 HMTase in Drosophila.
Unlike TRX, TRR contains four sequences that match the nuclear-receptor-interacting motifs17 LXXLL or LLXXL (‘NR’ in Fig. 1a). We tested each TRR NR region for its ability to interact with the ecdysone receptor in vitro. The NR2 domain of TRR, which contains an LXXLL motif, bound to full-length EcR and Ultra-spiracle (USP) proteins; although EcR and USP are considered to be obligate heterodimer partners, either protein bound to the TRR NR2 domain alone (Fig. 2a). The three other NR domains of TRR showed very little if any interaction with EcR and USP (data not shown). An association between TRR and the ecdysone receptor in vivo was shown by co-immunoprecipitation from embryo extracts (Fig. 2b), in which TRR was associated with both components of the ecdysone receptor.
We confirmed the association of TRR and EcR in vivo by two additional experiments. Like TRX18, TRR bound to several sites on salivary gland polytene chromosomes; immunostaining of polytene chromosomes showed an almost complete colocalization of TRR with EcR-B1 (Fig. 2c), the predominant EcR isoform of the salivary gland. Although the stained bands were numerous, they constituted only a few of the bands visible by DNA staining (blue). By contrast, there was little overlap between the staining for TRR and either TRX or Polycomb (data not shown). In the eye imaginal disc, TRR was expressed in a pattern essentially identical to those of USP and EcR-A, the predominant EcR isoform of eye discs (Fig. 2d). As reported for USP19, both TRR and EcR-A were expressed primarily in the region posterior to the morphogenetic furrow and in smaller amounts in a row of cells just anterior to the morphogenetic furrow, and all three proteins were absent from most of the region anterior to the morphogenetic furrow. Thus, in both the salivary gland and the eye disc, there is marked spatial colocalization of TRR and EcR–USP.
Having established a physical association between TRR and the ecdysone receptor, we tested for a functional association in vivo, concentrating on the developing compound eye. Ecdysone and its receptor have complex roles in both progression of the morphogenetic furrow19,20 and post-furrow differentiation21. We showed that TRR is required for eye development by examining the phenotype of somatic trr1 clones in the adult eye. Two abnormalities were visible in the clonal patches: a notching of the anterior boundary of the eye, interpretable as a retardation of morphogenetic furrow progression; and a ‘rough eye’ phenotype, attributable to defects in photoreceptor differentiation (Fig. 2e).
Three approaches showed that trr interacts genetically with genes encoding components of the ecdysone receptor and the ecdysone synthesis pathway. First, heterozygosity for the null allele trr1, similar to the temperature-sensitive ecdysoneless allele ecd1 (at its partially restrictive temperature, 25 °C), enhanced the furrow-stop phenotype of the B1 mutation (Supplementary Fig. 2a, b). Introducing an extra copy of trr+ as a transgene10 reversed the trr1 effect (Supplementary Fig. 2a). Flies with a combination of ecd1, trr1 and B1 showed an additional defect in morphogenetic furrow progression, particularly in the ventral part of the eye (Supplementary Fig. 2b, bracket). trr3, in which the SET domain is deleted, was similar to trr1 in its interaction with B1 (data not shown). These genetic interactions are consistent with a role for TRR as a coactivator for the ecdysone receptor.
Second, expression of the dominant-negative mutant EcR-F645A in the post-morphogenetic furrow retinal epithelium by the GMR–GAL4 driver led to severe defects in eye differentiation (Fig. 2f and Supplementary Fig. 2d) and in a marked lethality before adult eclosion (ref. 10 and Table 1). Under the conditions used, lethality was apparently due largely to desiccation, presumably caused by defective cuticle secretion. The lethality associated with EcR-F645A was strongly enhanced in trr1 and trr3 heterozygotes (Table 1). Inclusion of an extra copy of trr+ partially suppressed both the lethality and the eye phenotype (Table 1 and Fig. 2f). A mutation in Smrter (Smr), which encodes a known co-repressor for the ecdysone receptor22, showed an interaction with EcR-F645A similar to that observed with an extra copy of trr+ (Table 1 and Fig. 2f). Thus, trr and Smr function antagonistically, which is again consistent with a role for TRR as a coactivator for the ecdysone receptor.
Third, heterozygosity for the EcR null allele EcRM554fs noticeably suppressed the furrow-stop phenotype of trr1/B1 flies (Fig. 2g, arrows). Homozygotes for the viable allele trr4, like the trr1 null clones, showed both defects in morphogenetic furrow progression and disorganization of the ommatidial array (Fig. 2h; trr4 did not complement trr1 (data not shown) and is associated with insertion of a P-element in the trr promoter region). Both trr4 phenotypes were suppressed by heterozygosity for EcRk06210, an isoform-A-specific hypomorphic mutation, which by itself caused no visible eye phenotype (Fig. 2h, left).
In contrast to the results from the first two approaches, these interactions between trr and a reduction in EcR function indicate that TRR may function in an opposite direction to EcR. These apparently paradoxical results are readily explained by reference to the fact that ecdysone converts its receptor from a transcriptional repressor to an activator. Both ecd1 and EcR-F645A specifically abolish the activation function without altering repression23, and both enhance the phenotype of trr mutations, consistent with a coactivator function for TRR. Mutations in EcR affect both receptor functions and suppress the phenotype of trr mutations. This suggests that EcR-mediated repression has an essential, dose-sensitive role that can be partially overcome by a decrease in coactivator (TRR) titre. Similarly complex effects in eye development have been seen for mutations in usp24.
We sought to clarify the role of TRR in furrow progression. Our analyses of genetic interactions between trr and both hedgehog (hh) and its downstream target decapentaplegic (dpp) indicated that TRR has a role in Hedgehog (Hh) signalling (Fig. 3). Like B1, trr1 was found to be a dominant enhancer of the eye-specific mutation dppblk (ref. 25 and Fig. 3a). Hh is involved in morphogenetic furrow progression and in post-furrow photoreceptor differentiation26, and trr mutations altered hh phenotypes in both processes. Although a reduction in trr dose had no apparent effect on either aspect of the hh1 phenotype (data not shown), an extra copy of trr visibly suppressed the furrow-stop phenotype and restored normal ommatidial differentiation in hh1 homozygotes. Flies heterozygous for both trr1 and hh13C also showed a rough-eye phenotype that was not produced by either mutation alone (Fig. 3b).
Because heterozygosity for trr1 enhanced the morphogenetic furrow progression defect of B1 flies (Supplementary Fig. 2a), we tested whether it also enhanced the reduction in dpp and hh expression caused by B1. Heterozygosity for trr1 alone had no effect on eye morphology or on hh and dpp expression (data not shown). B1 caused a decrease in expression of dpp in the middle of the eye disc, whereas dpp expression was completely abolished in this region in B1/trr1 transheterozygotes (Fig. 3c). trr1 also clearly enhanced the deficit in hh expression caused by B1 (Fig. 3d). The requirement of trr for hh expression was further shown by the considerable decrease in hh expression in the trr null clones posterior to the morphogenetic furrow (Fig. 3e). The spatial pattern of trr expression is consistent with TRR having a direct role in hh expression (Figs 2d and and3f),3f), whereas there is little if any overlap between the areas of expression of trr and dpp (Fig. 3f). Because hh functions upstream of dpp in directing morphogenetic furrow progression, these results indicate that TRR may positively regulate hh transcription, and that a decrease in hh expression in trr mutants may lead to a decrease in the expression of dpp.
Alternatively, TRR might upregulate dpp directly, in addition to its indirect effect through hh. These possibilities can be distinguished by use of the dominant hypermorphic mutation roDom: Hh stimulates ro expression, and roDom interferes with hh induction of dpp27–29. If trr directly upregulates dpp, then an extra copy of trr would be expected to suppress the roDom phenotype. If the effect of TRR on dpp expression is mediated completely by hh, then an extra copy of trr would be expected to enhance the roDom phenotype. We found that the latter is the case (Supplementary Fig. 2b). Therefore, the effects of TRR on dpp expression are likely to be predominantly a secondary consequence of its effects on hh expression.
The above experiments lead to the hypothesis that TRR functions as a coactivator for the ecdysone receptor, mediating ecdysone activation of transcription by modifying chromatin structure at target promoters. The results also indicate that hh may be a direct TRR target during eye development. This hypothesis predicts an ecdysone-dependent association of TRR with the ecdysone receptor and with ecdysone-responsive promoters, and an ecdysone induction of H3-K4 trimethylation at ecdysone-induced promoters. We tested these predictions as follows.
Fractionation of an embryo extract showed that TRR was present in very large complexes, M r ≈ 2,000,000 (2,000K), whereas EcR migrated as a broad band, predominantly at a much lower M r. Treatment of the extract with ecdysone caused the TRR peak to become confined to a higher Mr range and caused a marked shift in migration of EcR to higher M r, so that roughly half of the EcR co-migrated with TRR (Fig. 4a). Similar results were obtained with S2 cell extracts (data not shown). Antibody specific for EcR could precipitate TRR more efficiently from the 2,000K fraction of the hormone-treated embryonic extract than from the untreated extract (Fig. 4b). Thus, TRR and EcR are components of a hormone-dependent complex in vivo. Although the presence of USP in this complex has not been shown, the hormone dependence of the EcR–TRR interaction indicates that TRR may have a direct role in mediating transcriptional induction by the ecdysone receptor.
We then used S2 cells to examine the chromatin structure of two putative target promoters. We examined hh, which was suggested by our results to be a target of TRR activation in the eye disc, and BR-C, a gene whose transcription is ecdysone-responsive in many tissues, including the eye20,24. Transcription of both genes was induced by ecdysone in S2 cells (Fig. 4c). Chromatin immunoprecipitation (ChIP) assays showed that ecdysone treatment caused a marked increase in the association of both promoters with EcR, TRR and H3 trimethylated at K4 (Fig. 4d). A causal connection between the presence of TRR and histone methylation at these promoters was indicated by similar assays comparing extracts from wild-type embryos and from embryos homozygous for strong mutations at trr: transcription and the amount of H3-K4 trimethylation at both promoters was significantly reduced in embryos homozygous for either trr1 or trr3 (Fig. 4e, f). In addition, because the SET domain of TRR is deleted in trr3 (Fig. 1a), this indicates that TRR SET may be responsible for the ecdysone-induced H3-K4 methylation at these promoters.
Our experiments show that TRR is a major Drosophila HMTase that binds to EcR in an ecdysone-dependent manner and causes H3-K4 trimethylation at ecdysone-inducible promoters. In the developing compound eye, TRR has a crucial role in progression of the morphogenetic furrow and in post-furrow differentiation of the retinal epithelium, and one of its direct targets is the hh promoter. Taken together, our results imply that TRR is a coactivator for the ecdysone receptor.
Details of the genetic manipulations are given in the Supplementary Information.
The anti-TRR polyclonal antiserum T1 was raised in rabbits (Cocalico Biological) against a bacterially expressed 6 × His fusion protein containing TRR residues 568–828. The T1 antiserum was affinity purified, and its specificity was confirmed by the absence of staining in homozygous trr1 null mutant embryos (data not shown). Monoclonal antibodies to EcR-A (15G1a) and EcR-B1 (AD4.4) were a gift from C. Thummel. Immunostaining of polytene chromosomes and eye discs is described in the Supplementary Information.
Western blotting, co-immunoprecipiation, in vitro binding assays and size fractionation of nuclear extract were done by standard procedures (see Supplementary Information) with the following reagents: antibodies against histone H3 either dimethylated at K4 (Upstate Biotechnology) or trimethylated at K4 (Abcam); T1 antibodies against TRR; antibodies against glutathione S-transferase (GST; diluted 1:1,000; Amersham Pharmacia Biotech); DDA2.7 monoclonal antibodies against the EcR common region (1:1,000; a gift from C. Thummel); and monoclonal antibodies against USP (1:50; a gift from F. Kafatos).
Schneider S2 cells were grown in the presence of 10−6 M 20-hydroxy-ecdysone (20HE) for 24 h. RNA was extracted from roughly 106 S2 cells or 50 embryos with a High Pure RNA Isolation kit (Roche), and reverse transcription was done with random hexamers and AMV reverse transcriptase. To normalize the relative amounts of hh and BR-C Z1 RNAs, we first analysed samples by PCR with rp49-specific primers. Semi-quantitative PCR was achieved by preparing serial dilutions of input DNA. All samples were assayed in the range where the PCR products showed a linear correlation with input DNA. We used the following primers. For rp49: sense, 5′-cgg atc gat atg cta agc tg-3′; antisense, 5′-gaa cgc agg cga ccg ttg ggg-3′. For hh: sense, 5′-tga tcc caa cga tcc tgt-3′; antisense, 5′-gta aca gcg tct gtg ttc t-3′. For BR-C Z1: sense, 5′-ccg acc aat tgc aaa gat ct-3′; antisense, 5′-gta cat gga tct cag ctg ca-3′.
Experiments were done by standard protocols and a ChIP Assay kit (Upstate Biotechnology), using the same starting material as in the RT–PCR experiments. For normalization, the unprecipitated material was analysed by PCR using the primer set specific to the hh promoter. We used the following primers. For the hh promoter (−564 to −173): sense, 5′-tgg atg aaa gtg tgt gcc t-3′; antisense, 5′-acc act acc tga ggg gca tat-3′. For the BR-C Z1 promoter (−554 to −120): sense, 5′-tcc att tgc gat gcg ttg gct-3′; antisense, 5′-agg ttc act gca gtt cag agt-3′.
The SET domain of TRR (residues 2,199–2,410) was expressed in Escherichia coli as a GST fusion, and assayed for HMTase activity as described12. Recombinant mutant histones H3 were a gift from Y. Zhang. Edman degradation was done at the KCC Protein Facility on a 477A sequencer (Applied Biosystems).
We thank C. Thummel, F. Kafatos, J. Treisman, M. Fujioka, T. Kornberg and Y. Zhang for antibodies, mutant stocks and mutant histones; and J. Kumar and F. R. Turner for the scanning electron microscopy images. This work was supported by a grant from the National Cancer Institute (to A.M.); an NIH award (to J.B.J.); a grant from the NSF (to P.C.); and an NIH training grant (to E.C.).