dLKR/SDH regulates hormone-mediated histone arginine methylation and transcription of cell death genes

The N-terminal tails of histones are subject to diverse posttranslational modifications such as acetylation, methylation, phosphorylation, ubiquitination, and sumoylation (Kouzarides, 2007; Li et al., 2007). Distinct modifications act sequentially or in combination to form the histone code (Strahl and Allis, 2000). These modifications provide a platform recognition surface for nonhistone proteins that translate the histone code to ultimately affect many biological processes, including gene expression, DNA replication, and repair (Kouzarides, 2007; Li et al., 2007).

Although histone acetylation/deacetylation has been studied extensively, dynamics of histone methylation was the subject of considerable debate, as it was long thought to be a stable epigenetic modification because of the absence of a known histone demethylase. Chromatin immunoprecipitation (ChIP) studies have revealed that histone methylation on both lysine and arginine residues can be extremely dynamic on various promoters and in response to specific signals (Bannister et al., 2002; Metivier et al., 2003). Methylation of arginine residues on histone 3 (H3) by CARM1 and histone 4 (H4) by PRMT1 is known to occur in response to nuclear hormones (Chen et al., 1999; Wang et al., 2001). The discovery of PADI4 lent support to the dynamic nature of histone methylation, as it was demonstrated to deiminate both the unmodified and monomethylated forms of histone H3R2, R17, and R26, forming citrulline and thereby antagonizing histone arginine methylation (Cuthbert et al., 2004). This was followed by identification of the first lysine-specific demethylase, LSD1, a nuclear amine oxidase that was found to demethylate H3K4me1 and me2 in a flavin adenine dinucleotide–dependent oxidative reaction (Shi et al., 2004). This is consistent with the finding that LSD1 is part of the Co-REST repressor complex (Ballas et al., 2001). The discovery of JHDM1A (FBXL11), which specifically demethylates histone H3K36 using Fe (II) and α-ketoglutarate as cofactors (Tsukada et al., 2006), led to the establishment of the third family of demethylases, the JmjC domain–containing histone demethylases. This family is evolutionarily conserved with 30 proteins in humans and mice and 13 in Drosophila melanogaster (Klose et al., 2006; Shi and Whetsine, 2007). It is now apparent that the dynamics of histone methylation is a tightly regulated process involving both histone methyltransferases and histone-specific demethylases. As more demethylases will inevitably be discovered, the question arises as to how the rapid kinetics of histone arginine methylation/demethylation is controlled at the molecular level and how this relates to the dynamics of other histone modifications involved in the histone code in response to a given signal.

Nuclear receptors are ligand-activated transcription factors that bind to target gene promoters. In the presence of ligand, they recruit coactivators to modify and remodel the chromatin structure, which ultimately leads to the assembly of the RNA polymerase II–containing transcription complex (Tsai and Fondell, 2004). A ligand-induced conformational change results in direct recruitment of the p160 coactivators, further recruiting p300/CREB-binding protein (CBP) and coactivator-associated arginine methyltransferase (CARM1), which methylates histones H3R2, R17, and R26 (Chen et al., 1999; Schurter et al., 2001). In Drosophila, the major metamorphic transitions during development are mediated by the steroid hormone ecdysone, which binds its heterodimeric receptor, ecdysone receptor/ultraspiracle (EcR/Usp), and regulates proliferation, programmed cell death, and differentiation in a tissue- and developmental-specific manner (Lee et al., 2002; for reviews see Truman and Riddiford., 2002; Kumar and Cakouros, 2004; Yin and Thummel, 2005). An ecdysone pulse toward the end of the larval stage signals puparium formation and histolysis of the larval midgut. A second ecdysone pulse ∼12 h later initiates head eversion and histolysis of the larval salivary glands (Lee et al., 2002; for reviews see Truman and Riddiford, 2002; Kumar and Cakouros, 2004; Yin and Thummel, 2005). Previously, we discovered that the caspase dronc is dramatically up-regulated in dying larval salivary glands in response to ecdysone (Dorstyn et al., 1999; Cakouros et al., 2002), and this is partly the result of direct binding of EcR/Usp to the promoter that controls its temporal expression (Daish et al., 2003; Cakouros et al., 2004a). DRONC and its adaptor Apaf1-related killer (ARK) are essential for ecdysone-mediated cell death of larval salivary glands (Daish et al., 2004; Akdemir et al., 2006; Mills et al., 2006). We also found that the arginine histone methyltransferase CARMER/DART4 is associated with EcR/Usp (Cakouros et al., 2004b).

In this study, we identify a novel cofactor, Drosophila lysine ketoglutarate reductase (LKR [dLKR])/saccharopine dehydrogenase (SDH), that binds EcR/Usp and regulates hormone-mediated transcription. dLKR/SDH specifically inhibits CARMER-mediated H3R17me2 by directly interacting with histone H3 but does not inhibit H3K4me3, which is also induced by ecdysone. dLKR/SDH genetically interacts with EcR/Usp and is recruited to hormone-responsive promoters to control the kinetics of H3R17me2.

In Drosophila, the only known coactivators for EcR/Usp are Taiman (p160 coactivator), CARMER, and the trithorax-related protein (TRR; a histone lysine methyltransferase). To identify other factors, we used the immobilized template bead assay in which the EcR/Usp-binding element (BE) of the caspase dronc promoter was biotin tagged, immobilized onto streptavidin beads, and incubated with nuclear extracts from Drosophila l(2)mbn cells with or without ecdysone treatment. l(2)mbn cells have been used in the past for characterizing ecdysone-mediated transcription and apoptosis (Cakouros et al., 2002, 2004b). To verify recruitment of the EcR-B1 isoform to the dronc promoter, the bound extracts were subjected to immunoblotting, which demonstrated the recruitment of EcR-B1 to the template in both the absence and presence of ecdysone treatment (Fig. 1 A, bottom). However, recruitment to the mutated EcR/Usp BE template was reduced as expected. This mutation, which alters the consensus binding site for EcR/Usp, has been previously described (Cakouros et al., 2004a). Several proteins were identified by mass spectrometry to be specifically recruited to the wild-type EcR/Usp BE in the presence of ecdysone. These included CG10473, which is involved in chromatin condensation; CG10223 (topoisomerase II), which is involved in transcription from polymerase II promoters; a zinc finger protein, CG10084; an unknown methyltransferase, CG8545; a helicase, CG6375; as well as an uncharacterized protein (CG7144) of ∼110 kD that we have called dLKR/SDH (Fig. 1 A, top).

dLKR/SDH is an evolutionarily conserved protein with homologues in Arabidopsis thaliana (SDH or LKR/SDH), yeast (Lys9 and Lys1), Drosophila, mouse, and human (amino-adipate semialdehyde synthase). It is a dual activity enzyme with an N-terminal LKR domain and a C-terminal SDH domain (Sacksteder et al., 2000). Except in Saccharomyces cerevisiae, in which Lys1 encodes for the LKR and Lys9 encodes for the SDH enzyme, bifunctional LKR/SDH is encoded by a single protein. dLKR/SDH is 51% identical and 71% similar to mouse and human LKR/SDH. In A. thaliana, yeast, mice, and humans, this enzyme is involved in catabolism of lysine to saccharopine by the LKR domain (Fig. 1 B) and from saccharopine to amino adipate semialdehyde by the SDH domain (Bhattacharjee, 1985; Goncalves-Butruille et al., 1996; Sacksteder et al., 2000; Zhu et al., 2000). The reaction uses NADH and α-ketoglutarate in the first step, which is oxidized to NAD⁺, and then NAD⁺ is used for the conversion to amino adipate semialdehyde. In yeast, these reactions are reversible and can produce lysine from saccharopine; however, mammalian LKR/SDH is unable to do this (Sacksteder et al., 2000).

Using GST fusion proteins of both individual enzyme domains (LKR and SDH) and in vitro–translated EcR-B1, the direct association of the carboxyl SDH domain with the EcR-B1 isoform was observed both in the absence and presence of ecdysone, whereas the LKR domain displayed a much weaker interaction in both the absence and presence of ecdysone (Fig. 1 C). To further narrow the binding region, we sequentially deleted the SDH domain. Although the region spanning amino acids 618–928 bound EcR/Usp efficiently, further deletion to residue 803 diminished binding (Fig. 1 C). Endogenous EcR-B1 immunoprecipitated with dLKR/SDH both in the absence and presence of ecdysone (Fig. 1 D). An antibody raised in rabbits identified a 110-kD dLKR/SDH protein in l(2)mbn cells and Drosophila larvae. This protein is nearly ablated by dLKR/SDH RNAi in l(2)mbn cells, absent in a homozygous dLKR/SDH genetic deficiency, and reduced in heterozygous animals (Fig. 1 E).

To determine the subcellular localization of dLKR/SDH, fractionation experiments were performed using lysates from l(2)mbn cells. dLKR/SDH was readily detectable in the cytosol of both untreated and ecdysone-treated cells. dLKR/SDH was also seen in nuclear extracts, and the protein levels appeared to be dynamic upon ecdysone treatment, decreasing at 10–30 min of treatment and increasing by 1 h (Fig. 1 F). The same membrane was blotted for cytosolic proteins DRONC and DRICE as well as the nuclear heterochromatin protein 1 (HP1-β), demonstrating the purity of the fractions. Our antibody was unsuitable for immunohistochemical detection of the low levels of endogenous dLKR/SDH protein (unpublished data). Therefore, we further analyzed dLKR/SDH localization by expressing a tagged version of the protein in l(2)mbn cells. As shown in Fig. S1 , ectopically expressed dLKR/SDH localized to both the cytosol and nucleus.

To test the effect of dLKR/SDH on ecdysone-mediated transcription, we expressed and purified full-length His-tagged dLKR/SDH from insect cells. By using lysine as a substrate, we measured oxidation of NADH to NAD⁺ and observed a time-dependent decrease in NADH, implying that the full-length protein is enzymatically active (Fig. 2 A). l(2)mbn cells were therefore transfected with a promoter-reporter (pEcR-Luc) that is activated by direct binding of the preexisting EcR/Usp complex. This reporter was activated by ecdysone and repressed by coexpressing increasing amounts of dLKR/SDH, indicating a direct effect on the EcR/Usp complex (Fig. 2 B). Given the high conservation of dLKR/SDH throughout evolution, the putative NADH-binding domain, which is responsible for enzymatic activity, was identified based on homology to the extensively characterized yeast homologue. This domain was mutated (see Oligonucleotides and primers section), and its effect on reporter expression was examined. The mutated dLKR/SDH was enzymatically inactive (unpublished data). As shown in Fig. 2 C, there was no significant difference in the repressive ability of the mutant and wild-type proteins, suggesting that the repressive effect is independent of dLKR/SDH's catalytic activity. In contrast, overexpression of dLKR/SDH had no significant repressive effect on transcription mediated by an unrelated transcription factor, broad complex (BR-C), when driving expression of a BR-C–regulated promoter-reporter (Fig. 2 D). In support of its role as a repressor of ecdysone-mediated transcription, knockdown of dLKR/SDH resulted in a significant increase in ecdysone-induced reporter activity (Fig. 2 E), whereas no effect was evident on the BR-C reporter (Fig. 2 F).

To further study the function of dLKR/SDH on hormone-induced transcription, ecdysone-induced transcription of the cell death genes ark and dronc was assessed after knockdown of dLKR/SDH in l(2)mbn cells. As assessed by semiquantitative real-time PCR, dLKR/SDH expression was reduced by 80% (Fig. S2). Both ark and dronc transcription increased after ecdysone treatment (Fig. 3 A). When dLKR/SDH was knocked down, transcription of both dronc and ark was significantly increased in response to ecdysone treatment, with the greatest effect evident at 6 h of ecdysone treatment. Overall, the results suggest that dLKR/SDH negatively regulates hormone-induced transcription, and knocking down its expression alters the transcription kinetics of ecdysone-induced cell death genes.

Given that the histone demethylase PADI4 demethylates histone H3 on arginine 17 (R17) and also plays a role in amino acid catabolism, we postulated that dLKR/SDH might be involved in demethylating histones. This was examined using calf thymus histones, recombinant LKR or SDH domain, and appropriate cofactors. The LKR domain alone or together with the SDH domain did not demethylate H3K4me1, me2, me3, H3K9me2, me3, or H3R17me2 (H3R17me1 could not be detected; Fig. 3 B). The same results were obtained when using purified histones from HeLa cells (unpublished data). We next examined whether dLKR/SDH might inhibit histone methylation mediated by histone methyltransferases involved in hormone-mediated transcription. To test this, histones were incubated with increasing amounts of the LKR domain and were methylated with either TRR, which mediates H3K4me3, or CARMER, which mediates H3R17me2. The LKR domain had no significant effect on H3K4me3 but strongly inhibited H3R17me2 in a dose-dependent manner (Fig. 3, C and D). However, the SDH domain had no significant effect on the methylation of histone H3 (Fig. 3 E).

To determine whether dLKR/SDH modifies histones and therefore inhibits methylation, we incubated full-length purified dLKR/SDH with histone H3, NADH, and α-ketoglutarate. We did not detect any specific histone H3 modifications mediated by dLKR/SDH as determined by mass spectrometry, suggesting that dLKR/SDH-mediated inhibition of H3R17me2 is not driven by its enzymatic activity (unpublished data).

We next examined the direct interaction of the LKR domain with purified histones. GST-LKR specifically associated with histone H3 as shown by H3-specific antibodies (Fig. 4 A). Furthermore, protein staining showed that the LKR domain interacted specifically with H3 and H4 but not H2B or H2A (Fig. 4 A). The region of histone H3 that interacts with the LKR domain was determined by competition experiments using histone H3 peptides. Peptides corresponding to the first 10 amino acids (1–10) of the N terminus of histone H3 inhibited interaction of LKR with histone H3, whereas a peptide corresponding to the C terminus (amino acids 74–83) had no significant effect (Fig. 4 B). A peptide consisting of amino acids 10–20 also showed some competition for histone H3 (Fig. 4 B). This suggests that dLKR/SDH interacts with the N-terminal region of H3.

RNAi and overexpression data suggest that dLKR/SDH acts as a negative regulator of ecdysone-induced transcription. This is supported by the fact that histone H3 methylation by CARMER, a positive regulator of transcription, is specifically inhibited by dLKR/SDH. Therefore, we assessed the effect of the coexpression of dLKR/SDH and CARMER in a reporter assay (Fig. 4 C). Ecdysone treatment induced activity of the ecdysone-responsive reporter, and this was further enhanced by coexpressing CARMER (Fig. 4 C). When CARMER and dLKR/SDH were coexpressed, dLKR/SDH inhibited CARMER-mediated transcription (Fig. 4 C).

ChIP assays were performed to determine the mechanism of action for dLKR/SDH. In the absence of ecdysone, there was very little EcR-B1 or CARMER recruited to the dronc promoter, correlating with low levels of H3R17me2 (Fig. 5 A). After ecdysone treatment, there was an increase in EcR-B1 recruitment on the promoter, coinciding with the recruitment of CARMER and correlating with the appearance of H3R17me2 (Fig. 5 A). Carmer knockdown by RNAi led to a significant decrease in ecdysone-induced H3R17me2, suggesting that CARMER is responsible for H3R17me2 (Fig. 5 A). Detailed analysis of the kinetics of histone methylation showed recruitment of EcR-B1 to the promoter as early as 10 min after ecdysone treatment, which remained stable until 1 h (Fig. 5 B). H3R17me2 methylation appeared after 10 min of ecdysone treatment and was present at 30 min; however, by 1 h it had diminished, suggesting the dynamic nature of this modification (Fig. 5 B). As expected, CARMER was recruited by 10 min, coinciding with the appearance of H3R17me2; however, it had disengaged by 30 min and surprisingly reappeared by 1 h, when H3R17me2 was less abundant. A very low level of H3K4me3 was detected in the absence of ecdysone, but it rapidly increased by 10 min after ecdysone treatment. The levels of H3K4me3 increased with time up to 1 h, which is consistent with its role as a positive regulator of transcription. A high level of H3K9me3 was detected in untreated cells; however, in response to ecdysone treatment, H3K9me3 rapidly decreased with time consistent with its role as a repressor of transcription (Fig. 5 B). Recruitment of endogenous dLKR/SDH to the promoter was also assessed using the specific antibody. dLKR/SDH was present on the promoter in untreated cells but disappeared by 10 min of ecdysone treatment and reappeared at 1 h (Fig. 5 B), with its appearance inversely correlating with the appearance of H3R17me2.

We also examined the recruitment of these coactivators to the endogenous ark promoter. By electrophoretic mobility shift assay (EMSA), we found an EcR/Usp BE that lies 807 bp upstream of the transcription start site (Fig. 5 E). This site specifically binds the EcR-B1 isoform but not the A isoform (Fig. 5 C). ChIP analysis showed that a low level of EcR-B1 was associated with the ark promoter, which increased by 10 min of ecdysone treatment followed by a decrease at 30 and 60 min (Fig. 5 D). This is in contrast to the kinetics of recruitment to the dronc promoter. H3R17me2 methylation was also evident by 10 and 30 min of ecdysone treatment. However, by 1 h it had diminished. Flag-tagged dLKR/SDH was present on the promoter in untreated cells and cells treated with ecdysone for 10 min; however, a decrease was evident at 30 min and 1 h (Fig. 5 D). Recruitment of CARMER was evident by 10 min, reaching a maximum at 30 min, followed by a decrease by 60 min. H3K4me3 increased from 30 min onwards, and H3K9me3 decreased in response to ecdysone treatment with time. Collectively, these results demonstrate that by 10 min after ecdysone treatment, increased recruitment of EcR-B1 coincides with the recruitment of CARMER and the presence of H3R17me2. After 30 min of ecdysone treatment, this is followed by a rapid decline. Although CARMER is recruited again at 1 h (for dronc promoter), this does not coincide with H3R17me2 at that time, as dLKR/SDH presumably inhibits H3R17me2.

Common to both dronc and ark promoters, the presence of endogenous dLKR/SDH seems inversely proportional to H3R17me2, which is present mostly at the time when H3R17me2 does not occur or at the beginning of its appearance. It is also present at 1 h, as is CARMER, although H3R17me2 does not occur at this time. Based on these data, we conclude that dLKR/SDH inhibits H3R17me2 methylation when present on the promoter. This was confirmed by knocking down dLKR/SDH by RNAi and assessing changes in arginine methylation along the dronc promoter. Cells treated with control dsRNA showed the appearance of H3R17me2 by 7 min after ecdysone treatment, reaching maximum levels at 30 min and a decrease at 1 h (Fig. 6 A). Knockdown of dLKR/SDH resulted in a dramatic shift in the kinetics of H3R17me2 methylation, with levels at a maximum between 4–7 min and decreased again from 10 min onwards (Fig. 6 A). This experiment was validated by semiquantitative real-time PCR of the ChIP-enriched DNA and was compared with total input genomic DNA (Fig. 6 B). The results clearly show that H3R17me2 is generally low along the dronc promoter, with levels increasing at 10–30 min of ecdysone treatment and decreasing by 1 h. When dLKR/SDH was knocked down, in the absence of ecdysone, there is no significant difference in the levels of H3R17me2. However, as early as 4 min of hormone treatment, H3R17me2 levels increase sharply and then steadily decrease. These results indicate that dLKR/SDH inhibits H3R17me2 when occupying the promoter and controls kinetics of methylation at the promoter.

We next wanted to assess whether global levels of H3R17me2 were affected after dLKR/SDH knockdown. Given that the histone code states that individual histone modifications may affect the methylation status at other residues (Groth et al., 2007; Kouzarides, 2007), we also examined other histone methylation marks. Knockdown of dLKR/SDH showed a clear increase of H3R17me2 (Fig. 7). In contrast, the levels of H3K4me3 or H3K27me2 remained largely unchanged (Fig. 7).

dLKR/SDH expression was found to be dynamic throughout development, with transcript levels low in early embryos (0–8 h) and first instar larvae (L1) and highest at the time of larval/pupal metamorphosis (white prepupae; Fig. 8 A). In the salivary glands, the low levels of expression at 2 h before head eversion (Fig. 8 B, −2 hRHE) increased dramatically at head eversion (0 h) and decreased 2 h later (Fig. 8 B).

To study the in vivo function of dLKR/SDH in ecdysone-mediated transcription, we generated transgenic flies to knockdown or overexpress dLKR/SDH specifically in salivary glands. The larval salivary glands have been used as an experimental model to assess ecdysone-mediated transcription. Overexpression and knockdown of dLKR/SDH were confirmed by immunoblotting and real-time PCR, and the levels of dronc transcript were analyzed by real-time PCR (Fig. 8, C–E). In wild-type salivary glands, dronc transcript increased after the ecdysone pulse at the time of head eversion (0–1 h after head eversion; Fig. 8 E). Consistent with the results in l(2)mbn cells (Fig. 3 A), knockdown of dLKR/SDH in salivary glands resulted in an increase in dronc transcript 1 h before and 1 h after head eversion (Fig. 8 E), whereas overexpression of dLKR/SDH had no significant effect on the expression levels. Therefore, we conclude that dLKR/SDH mediates the correct timing of the developmental induction of ecdysone-regulated genes.

To further study the in vivo function of dLKR/SDH, we examined a gene deficiency that disrupts the dLKR/SDH locus. The homozygous deficiency (Df/Df) produced viable adult flies that exhibit significantly smaller (P < 0.0001) wing size (Fig. 8, F and G) and reduced overall body weight (not depicted). To confirm the role of dLKR/SDH in Drosophila development, we examined the effect of knocking down dLKR/SDH during wing development using the en-GAL4 driver. As in the Df/Df flies, the adult RNAi lines showed significantly smaller (P < 0.001) wings when compared with control flies (Fig. 8, F and G), confirming that the phenotype is caused by a specific loss of dLKR/SDH.

To study the endogenous dLKR/SDH and EcR interaction, we tested whether reducing dLKR/SDH by RNAi modified the phenotype of overexpression of a dominant-negative EcR allele (EcR^F645A) in the Drosophila eye using the GMR-GAL4 driver. As previously reported (Sedkov et al., 2003), specific overexpression of EcR^F645A in the eye disrupts differentiation of the ommatidia and eye patterning. This results in a decrease in pigment cells, collapse of the eye, and characteristic scarring (Fig. 8 H). Knockdown of dLKR/SDH partially restored the pigmentation defect and displayed a striking decrease of scarring. Therefore, reducing dLKR/SDH by RNAi resulted in suppression of the EcR^F645A rough eye phenotype. We also analyzed the genetic interaction between dLKR/SDH and EcR in the wing. Inhibition of EcR function by the expression of EcR^F645A had no effect on wing size; however, EcR^F645A expression restored the small wing phenotype caused by a knockdown of dLKR/SDH (Fig. 8 G). These results support the interaction of dLKR/SDH with EcR and the role of dLKR/SDH as a negative regulator of ecdysone-mediated transcription.

We have identified a novel cofactor involved in ecdysone-mediated regulation of gene expression. Surprisingly, this cofactor is an enzyme involved in lysine catabolism (Sacksteder et al., 2000). The fact that this gene is transcriptionally up-regulated in larval salivary glands undergoing programmed cell death in response to rising levels of ecdysone as well as in response to γ irradiation (Lee et al., 2003) prompted us to further characterize it. We established that dLKR/SDH binds directly to EcR/Usp in vitro, and this interaction occurs both in the absence and presence of ecdysone. This differed from the template assay, in which dLKR/SDH was recruited only in the presence of hormone. We believe that the complexes formed on oligonucleotides in this system are unstable, and therefore we could only detect dLKR/SDH and other factors in the presence of hormone, which is believed to increase the affinity of cofactors for their receptors.

With the use of reporter assays and RNAi in cells and in Drosophila, we found that dLKR/SDH is a repressor of ecdysone-induced transcription of dronc and ark. Its repressive effect is consistent with the fact that dLKR/SDH inhibits ecdysone-inducible H3R17me2. The greatest effect of dLKR/SDH knockdown on transcription was evident at 6 h of ecdysone treatment. This is likely the result of the active transcription of these genes occurring in the first 12 h of ecdysone treatment in l(2)mbn cells. Any effects on active transcription will therefore be more evident at earlier time points (i.e., 6 h) rather than later time points (i.e., 24 h). ark transcription is also more affected at 6 h than dronc, which may reflect different modes of ark and dronc regulation by ecdysone. For example, although ark seems primarily regulated by direct EcR/Usp binding, dronc is known to be additionally regulated by other ecdysone-inducible transcription factors such as BR-C and E93 (for review see Kumar and Cakouros, 2004).

Our experiments show that dLKR/SDH does not demethylate histones but potently inhibits arginine methylation of histone H3 by CARMER. Although we could not detect any interaction with CARMER, dLKR/SDH specifically interacted with H3 and H4 via its LKR domain but not H2B and H2A. This is intriguing given that arginine methylation occurs preferentially on H3 and H4 (Bedford and Richard, 2005). The absence of any modifications by full-length dLKR/SDH on histone H3 allows us to speculate that ketoglutarate might be important for dLKR/SDH to adopt a conformation required to sterically inhibit CARMER for methylating histone H3. This scenario is somewhat similar to the mechanism used by the corepressor C-terminal binding protein, which represses transcription using NADH (Zhang et al., 2002), and Sir2, which uses NAD⁺ (Imai et al., 2000). Given that dLKR/SDH and its homologues recognize lysine, it was surprising that it did not affect methylation of lysine on histone H3. In fact, we have mutated H3K4 and H3K9, and these mutations do not affect the interaction between dLKR/SDH and histone H3 (unpublished data). Therefore, it is highly likely that a group of amino acids within the N terminus are important for the interaction of dLKR/SDH with histone H3 and that the conformation adopted by dLKR/SDH specifically inhibits H3R17me2. We are currently examining the mechanisms involved in more detail.

Endogenous dLKR/SDH is present in the cytosol, consistent with its possible role in lysine catabolism, and in the nucleus, where its levels seem to fluctuate in a time-dependent manner in response to hormone treatment. Apart from histones, protein arginine methyltransferases are involved in methylating several other substrates (Bedford and Richard, 2005). In fact, the predominant arginine methyltransferase in human cells, PRMT1, is responsible for 85% of all arginine methylations and for methylating H4R3 when recruited by nuclear receptors (Pawlak et al., 2000; Tang et al., 2000). In addition, it has also been shown to methylate heterogenous nuclear RNP, which is involved in pre-mRNA processing and transport (Liu and Dreyfuss, 1995). PRMT1 lacks a nuclear localization signal, is localized preferentially in the cytoplasm, and shuttles into the nucleus by associating with other proteins entering the nucleus. Only when methylation is inhibited can PRMT1 be seen to accumulate in the nucleus (Herrmann et al., 2005). It is predicted that PRMT1 is attached to unmethylated substrates such as newly synthesized heterogenous nuclear RNP molecules (Herrmann et al., 2005). It is thus possible that dLKR/SDH is regulated in a similar manner, as it has no nuclear localization signal but does have a putative nuclear export signal. It is also possible that dLKR/SDH might regulate methylation in other cellular compartments. Given its ability to inhibit H3R17me2, it is not surprising that mechanisms exist to prevent its accumulation in the nucleus, where it could prevent H3R17me2 for prolonged periods of time and hinder the hormonal response. Although dLKR/SDH can bind EcR/Usp, our ChIP data show that dLKR/SDH is on the promoter spanning the BE before EcR/Usp is recruited. It is therefore more feasible that another constitutive factor is responsible for recruiting dLKR/SDH and interacting with histone H3, and its cycling back onto the promoter at 1 h of ecdysone treatment might be mediated by its ability to interact with EcR/Usp.

The ChIP data after dLKR/SDH knockdown demonstrated the premature appearance of H3R17me2 as well as the premature loss of H3R17me2, confirming that the role of dLKR/SDH is to control the kinetics of H3R17me2. Based on our in vitro data, this occurs by inhibiting methylation by the positive regulator CARMER. It has been documented that H3R17me2 serves to recruit RNA polymerase II to the promoter, and removal of H3 arginine methylation is needed for elongation of RNA polymerase II (Metivier et al., 2003). Therefore, this may explain why the change in H3R17me2 kinetics by dLKR/SDH knockdown results in a dramatic shift in transcription kinetics, as the initiation and elongation of the RNA polymerase could possibly occur more frequently when dLKR/SDH is knocked down. The only other enzyme capable of inhibiting R17 methylation is PADI4 in mammals, which is also involved in amino acid catabolism (Cuthbert et al., 2004). However, a PADI4 homologue is not present in Drosophila, which prompts us to conclude that dLKR/SDH may be functionally equivalent in flies.

Knockdown of dLKR/SDH in the salivary glands resulted in a premature increase in the expression of dronc but, in spite of this, showed no discernible alterations in tissue histolysis or metamorphosis. Furthermore, dLKR/SDH-deficient flies or ones harboring tissue-specific RNAi do not show strong phenotypes. Given that dLKR/SDH is a negative regulator of ecdysone-induced gene expression, these phenotypes suggest that dLKR/SDH plays a fine-tuning role as opposed to an essential role for ecdysone-regulated transcription. The mild phenotypes observed for alterations in dLKR/SDH are not surprising given the phenotypes observed for some other histone-modifying enzymes such as the histone demethylases LSD1 and Lid. Drosophila LSD1 homozygous mutants are viable but sterile despite showing alterations in the pattern of histone methylation (Rudolph et al., 2007), and overexpression of lid during eye differentiation had no effect on the adult phenotype (Secombe et al., 2007). Nevertheless, dLKR/SDH can genetically interact with the ecdysone receptor in the eye and wing, and loss of dLKR/SDH results in specific developmental phenotypes in the wing. The smaller wing size (in both deficiency and RNAi lines) is suggestive of defects in the regulation of cell growth, proliferation, and/or differentiation. We measured wing hair density in dLKR/SDH knockdown compared with control and observed no change, indicating that cell size is not altered (unpublished data). We are currently investigating alternative pathways that are mediated by dLKR/SDH and whether these effects are involved in the phenotypes observed in vivo.

Our current work supports a model shown in Fig. 9 in which dLKR/SDH binds the promoter region spanning the EcR/Usp BE. In response to hormone, there is an increase in EcR/Usp binding to the dronc and ark promoters followed by recruitment of CARMER within 10 min. However, the LKR domain of dLKR/SDH already bound to histone H3 shields H3R17 from methylation. dLKR/SDH is then removed by an unknown mechanism from the promoter, allowing CARMER to methylate H3R17. Between 30 and 45 min of hormonal stimuli, H3R17me2 is removed, followed by the recruitment of dLKR/SDH and CARMER by 1 h. dLKR/SDH therefore serves to control the correct developmental timing of ecdysone-induced genes such as dronc and ark in Drosophila at least partly by controlling histone H3R17 methylation kinetics.

l(2)mbn cells (a gift from A. Dorn, Johannes Gutenburg University, Mainz, Germany) were cultured as previously described (Cakouros et al., 2002). Ecdysone (Sigma-Aldrich) was used at 10 μM. Transfected SF21 insect cells were grown in SFII 900 complete medium and blasticidin at 20 μg/ml for selection. Antibodies to DRONC and DRICE have been described previously (Dorstyn et al., 2002). Antibodies to EcR-B1 (AD4.4), EcR-A (15G1a), EcR common (DDA2.7), HP1-β (C1A9), and armadillo (N27A1) were purchased from Developmental Studies Hybridoma Bank (University of Iowa). Histone H3 antibody was obtained from Cell Signaling Technology, H3R17me2 antibody was purchased from Abcam, and all other methylation-specific antibodies were obtained from Millipore.

CG7144 (dLKR/SDH) cDNA was prepared from 5 μg Drosophila embryo RNA using the Superscript II kit (Invitrogen). Full-length flag-tagged dLKR/SDH was cloned into the Drosophila pIE1-4 expression vector using restriction site SacII or pIB/V5-His with SacII for insect cell expression and purification. The dLKR/SDH mutant construct carrying Gly to Ala substitutions in residues 221, 223, and 228 was cloned into pIB/V5-His vector. These amino acids are conserved within the NADH-binding domain of various homologues. pIE CARMER expression vectors have been previously described (Cakouros et al., 2004a). EcoRI-containing primers were used to clone full-length dLKR/SDH into pUAST for tissue-specific expression in Drosophila. For tissue-specific RNAi, a hairpin was cloned into pUAST. pEcR-Luc consists of the EcR/Usp BE of the heat shock 27 promoter cloned in front of the luciferase reporter described previously (Cakouros et al., 2004a). BR-C Luc consists of the proximal promoter of dronc, which contains multiple BR-C BEs cloned in front of the luciferase reporter described previously (Cakouros et al., 2002). For expressing dsRNA in l(2)mbn cells, the region of dLKR/SDH amino acids 657–834 was cloned into pGEMT-easy and used for in vitro transcription using T7/Sp6. Control dsRNAi construct containing mammalian Ndfip2 cDNA has been previously described (Cakouros et al., 2002). For expression of recombinant LKR (amino acids 1–457) and SDH domains (amino acids 420–928) as GST fusion proteins, they were cloned into pGEX2TK. To generate a polyclonal antibody for dLKR/SDH, the region encoding amino acids 779–926 was cloned into pGEX2TK, and purified protein was used to immunize rabbits. Antibodies were purified by first depletion of the GST antibodies through a GST-agarose column followed by affinity purification on a dLKR/SDH (amino acids 779–926)-agarose column.

RNA extraction and Northern blotting was performed as previously described (Cakouros et al., 2002). An ark probe was generated by PCR using forward (5′-GGTGGCGTCGGCACACAGTGCC-3′) and reverse primers (5′-GCGGTGTCAAGGCGTCAGGG-3′). The RT-PCR and primers used were described previously (Cakouros et al., 2004b).

Total RNA was isolated from Drosophila salivary glands using TRIZOL reagent (Invitrogen). cDNA was synthesized using Superscript II Reverse transcription (Invitrogen) and oligo dT (Geneworks) from 5 μg of total RNA. Real-Time PCR was performed on a thermal cycler (Rotor-Gene 3000; Corbett Research) using RT² Real-Time SYBR green/ROX PCR MasterMix (Superarray) according to the manufacturer's instructions. The following primer sets were used: dLKR/SDH forward (5′-CGAGAACCTCAAGCAGAAGC-3′) and reverse (5′-ATTGATTCCACGCTCCTCAC-3′), dronc forward (5′-CAGCATCGTAGGCTCCTATTG-3′) and reverse (5′-ATATGGCACTATCCCCACCA-3′), and rp49 forward (5′-GCCACCAGTCGGATCGATATGCTA-3′) and reverse (5′-GGTTCTATTGAGAACGCAGGCGACCGTT-3′). Reactions were performed in triplicate, and the mRNA expression levels were normalized against the internal control gene rp49 using the ΔΔC_T method.

1 μg pEcR Luc reporter or pBR-C Luc was transfected using Cellfectin (Invitrogen) together with 100 ng EcR-B1 expression construct and different levels of pIB-dLKR/SDH. 400 ng pIE-CARMER was cotransfected where indicated with pIE-dLKR/SDH, pEcR-Luc, or pBR-C Luc. Where indicated, 2 μg pIE–BR-C expression construct was also used together with pIE-dLKR/SDH. Each transfection contained equal amounts of DNA using the pIE1-4 or pIB expression vector and was performed as described previously (Cakouros et al., 2002). All transfections were performed in parallel with pxpG empty reporter vector and values subtracted from those of the promoter-reporter construct.

Nuclear extracts and EMSA were performed as previously described (Cakouros et al., 2002, 2004b). The sequence of the probes used was 5′-CACACAAACATAACAAGTTCATTCACAGGGCAAACTCAGTGACG-3′.

Primers used for cloning dLKR/SDH into pIE were forward (5′-GACCCGCGGATGTGGCGAGTGATTCAACTGCGCGC-3′) and reverse (5′-GACCCGCGGTTACTTGTCATCGTCGTCATTTAGCCATCTGGAGGTTTCC-3′) containing flag tag sequence. Primers used for cloning dLKR/SDH into pIB/V5-His were forward (5′-ACCGAACCGCGGTGGCGAGTGATTCAACTGCGCG-3′) containing a SacII site and reverse (5′-ACCGAACCGCGGATTTAGCCATCTGGAGGTTTC-3′) containing a SacII site. The primer used to mutate amino acids 221, 223, and 228 to Ala was 5′-CTTACATTTGTGTTCACTGCCTCTGCAAATGTTTCCCAAGCTGCCCAGGAGGTG-3′. For cloning dLKR/SDH into pUAST, forward primer (5′-CAGGAATTCGAGATGTGGCGAGTGATTCAACTGCGC-3′) containing an EcoRI site and reverse primer (5′-CAGGAATTCTTACTTGTCATCGTCGTCATTTAGCCATCTGGAGGTTTCCG-3′) containing an EcoRI site and flag tag sequence were used. For the construction of tissue-specific RNAi for dLKR/SDH, the following primers were used: RNAi 1A (5′-CAGGAATTCATGGAGGGAAGTTCGGAGTCTGAC-3′) containing an EcoRI site, RNAi 1B (5′-CAGGGTACCCTGGCGCCACGTGACATCCGG-3′) containing a Kpn1 site, RNAi 2A (5′-CAGCTCGAGATCGAGGGAAGTTCGGAGTCTGAC-3′) containing an Xho1 site, and RNAi 2B (5′-CAGGGTACCGGAATCCCTGTTGGGGAAACC-3′) containing a Kpn1 site. To clone a portion of dLKR/SDH into pGEM for use in in vitro transcription, forward (5′-GGTCACCCAGGGGAGTACTTCT-3′) and reverse (5′-CCCGCTCGAAGGCAAGTCTCTT-3′) primers were used. Primers used for GST-dLKR/SDH were LKR forward (5′-CGCGAATTCCGTGGCGAGTGATTCAACTGCGCGCA-3′) and reverse (5′-CGCGAATTCTTACTTGTCATCGTCGTCGCCCTCCGTCAGCTGACCGTTGCT-3′ and 5′-CGCGAATTCTTACTTGTCATCGTCGTCATTTAGCCATCTGGAGGTTTCCGT-3′). For immobilized template bead assay, the primers used were wild type (5′-CAAAGGCCAAAGACCCACAACTTTTTC-3′). Oligonucleotides containing specific EcR/Usp BE mutations were also biotinylated at the 5′ end. The mutant primer was 5′-CAAAAACTAAAGACCCACAACTTTTTC-3′. dronc ChIP primer sequences used were forward (5′-CTATGCAATGCTAATTGGCAGCCCC-3′) and reverse (5′-AGGAATTTAGCACACAGAGC-3′). ark primers used were forward (5′-GCGTGCAGGTGAAAGCTTATGTACC-3′) and reverse (5′-GAGGAGGCGAGCAGTTACAGCATCT-3′). For rp49, forward (5′-GCCACCAGTCGGATCGATATGCTA-3′) and reverse (5′-GGTTCTATTGAGAACGCAGGCGACCGTT-3′) primers were used.

1.5 × 10⁸ l(2)mbn cells were treated with 10 μM ecdysone for 2 h, and nuclear extracts were prepared as previously described (Cakouros et al., 2002). Extracts were precleared with 100 μl streptavidin Sepharose beads (GE Healthcare) and 50 μg of salmon sperm DNA. Double-stranded oligonucleotides spanning the dronc promoter from region −1,308 to −1,281 bp relative to the transcription start site were biotinylated at the 5′ end of the sense strand. 500 μl streptavidin Sepharose beads were washed three times in buffer T (10 mM Tris, pH 7, 1 mM EDTA, 1 M NaCl, and 0.1% NP-40) and blocked for 1 h in 3.6 ml of buffer T supplemented with 50 mg/ml BSA. To 1 ml of slurry, 50 μg of double-stranded annealed oligonucleotide was added and incubated on a rotating wheel for 2 h at 4°C. Streptavidin bead–oligonucleotide complexes were then washed three times with buffer Z (25 mM Hepes, pH 7.8, 12.5 mM MgCl₂, 20% glycerol, and 0.1% NP-40). Five pull-downs were set up per treatment (with or without ecdysone), each with 100 μl of bead/oligonucleotide, 100 μg of nuclear extract, 800 μl of buffer Z, 50 μg of salmon sperm DNA, 100 mM NaCl, 20 μg poly(dIdC), and complete protease inhibitors. After overnight incubation at 4°C, complexes were washed twice in buffer Z supplemented with 100 mM NaCl, and beads were boiled in 2× SDS loading buffer and separated on a 10% SDS-PAGE. The gel was stained with Coomassie Brilliant blue, and appropriate bands were excised and identified by mass spectrometry at the Adelaide Proteomics Centre.

The excised bands were washed three times in 50 μl of ammonium bicarbonate (100 mM) containing 50% acetonitrile (HPLC grade; Merck) and again in 100% acetonitrile. 25 μl trypsin (500 ng of modified porcine trypsin, 2 mM acetic acid, 80 mM ammonium bicarbonate, and 8% acetonitrile [vol/vol]) was then added to rehydrate the spots. Proteolytic digestion was allowed to occur at 37°C for 16 h. Tryptic peptides were reduced by the addition of 1 μl of 0.1 M tris-(carboxyethyl)phosphine, and the tryptic digestion was stopped by the addition of 25 μl formic acid (1% vol/vol). The supernatant was collected, and the protein was further extracted in 25 μl acetonitrile (100%). The combined extracts were reduced in volume to ∼10 μl in a Speed-Vac system (Savant Instruments, Inc.) before liquid chromatography–mass spectrometry/mass spectrometry analysis. Peptides were introduced into a mass spectrometer (Micromass Q-Tof2; Waters) through a NanoSpray source coupled to a CapLC system (Waters). Injected samples were desalted on a C18 precolumn (0.3 × 3 mm, 5-μm packing; LC-Packings) before separation on an analytical column (0.075 × 50 mm, 3-μm packing; Atlantis dC18; Waters) using an acetonitrile gradient (5–70% containing 0.1% formic acid) at a flow rate of 200 nl/min.

Eluted peptides were ionized through the NanoSpray source with 3 kV. A 7-s survey scan of the eluting material identified multiple charged species (2, 3, and 4 charge states), and the most abundant (<4) were selected for fragmentation and mass spectrometry/mass spectrometry analysis. After signal had been accumulated for each of the fragment ion series for up to 21 s, the system reverted to the survey scan mode to identify further multiply charged species of potential interest. After acquisition, mass spectrometry/mass spectrometry data were processed and analyzed using ProteinLynx 2.0 software (Waters). De novo sequence tags were determined from the fragmentation data and used to search the National Centre for Biotechnology Information nonredundant protein database for protein identification matches. Matched proteins were assessed for quality and validity of the identification. In all cases, at least two separate peptide sequences were used to assign a match.

Recombinant dLKR and SDH were expressed as GST-tagged proteins in BL21 star LysS Escherichia coli, and thrombin was cleaved to remove the GST tag. In a 25-μl reaction, 4 μg of calf thymus histones was incubated with 100–500 ng of recombinant LKR or SDH in buffer (100 mM KH₂PO4, 1 mM EDTA, pH 6.8, 200 μM NADH, and 2 mM α-ketoglutarate). Where indicated, both NADH and α-ketoglutarate were omitted. After incubation at 30°C for 4 h, 30 μl histone methylation buffer (Cakouros et al., 2004b) was then added with 5 μl S-adenosyl [³H]Met and 1 μg GST-CARMER or TRR (provided by A. Mazo, Kimmel Cancer Center, Philadelphia, PA). Histone methylation was performed by incubation at 30°C for 90 min followed by SDS-PAGE in duplicate gels. One gel was Coomassie stained, whereas the other was fixed (25% isopropanol and 10% glacial acetic acid) for 20 min, incubated with Amplify (GE Healthcare) for 30 min, dried down, and exposed to hyperfilm for 5 d.

His-tagged full-length dLKR/SDH was cloned into pIB/V5His. The expression construct was transfected into SF21 cells, and stable transfectants were generated by maintaining cells with blasticidin at 20 μg/ml. For protein purification, 10⁸ cells were washed in PBS and resuspended in 20 ml of lysis buffer (50 mM Tris, pH 7.8, 150 mM NaCl, 1% NP-40, and complete protease inhibitors). Cells were lysed by freeze thawing, and cleared supernatant was incubated with 400 μl Ni²⁺–nitrilotriacetic acid agarose beads overnight at 4°C. Beads were washed in lysis buffer three times and protein eluted with 1 M imidazole. Purified protein was dialysed overnight in 100 mM NaCl/20 mM Tris-HCl, pH 7.6 (plus complete protease inhibitors).

5 μg of purified protein was incubated with 1 mM NADH, 2 mM α-ketoglutarate, and 14 mM lysine in a total volume of 50 μl (made up with 100 mM phosphate buffer, pH 6.8) in 96-well plates. Reaction was performed at room temperature for up to 4 h. Background levels not attributed to NADH were measured using buffer without NADH and values subtracted from the measurements with enzyme and NADH. Measurements were made at excitation and emission wavelengths of 340 nm and 455 nm, respectively, with an emission slit width of 2 nm using a luminescence spectrometer (LS55; PerkinElmer).

1–2 × 10⁷ l(2)mbn cells were cultured in 6-well dishes with or without 10 μM ecdysone. ChIP assay was performed as described previously (Shi et al., 2004). DNA was purified on spin columns (Ultraclean; MO BIO) and resuspended in 50 μl Tris/EDTA. 3 μl of purified DNA was used for PCR with annealing at 55°C and extension for 1.5 min. To detect EcR-B1 and flag-tagged CARMER or flag-dLKR/SDH recruitment, 37 cycles were used, whereas 33 cycles were used to detect H3R17me2, H3K4me3, and H3K9me3. 39 cycles were used for the Ig control, dLKR/SDH, and genomic input control.

Df(2L)ED496 fly stock that removes dLKR/SDH and a least one other gene was obtained from the Szeged Drosophila Stock Center. UAS-EcR^F645A, GMR-GAL4, sgs3-GAL4, and en-GAL4 stocks were obtained from the Bloomington Drosophila Stock Center. Wing analysis was performed using flies grown at 25°C by dehydrating adult flies in xylene and mounting wings in Canada Balsam. Samples were photographed using a microscope with a 4× UPlan Fl objective (BX51; Olympus) and a camera (DP70; Olympus) at room temperature. Images were captured with Olysia Bioreport software (Olympus) and compiled using Photoshop version 6.0 (Adobe). The wing areas were measured using the magnetic lasso tool and histogram function in Photoshop. Tissue-specific expression in the eye was driven using GMR-GAL4 and genetic interaction tests with UAS-EcR^F645A performed at 18°C because of the severe lethality of the UAS-EcR^F645A allele at 25°C. Eye images were acquired using a camera (DP11; Olympus) attached to a stereozoom microscope (SZ40; Olympus) with a 2× auxiliary objective at room temperature. Images were processed using Photoshop.

Fig. S1 shows localization of dLKR/SDH in transfected cells. Fig. S2 shows quantification of the RNAi-mediated knockdown of dLKR/SDH.

We thank C. Bagley (Adelaide Proteomics Centre, Adelaide, Australia) for proteomics support, A. Mazo for GST-TRR construct, the Bloomington and Szeged Stock Centers, the Australian Drosophila Biomedical Research Support Facility (OzDros) for fly lines, and Developmental Studies Hybridoma Bank (University of Iowa) for antibodies.

This work was supported by the National Health and Medical Research Council of Australia.