The Drosophila jumonji gene encodes a JmjC-containing nuclear protein that is required for metamorphosis


  • Nobuhiro Sasai,

    1.  Venture Laboratory, Kyoto Institute of Technology, Japan
    2.  Department of Applied Biology, Kyoto Institute of Technology, Japan
    3.  Insect Biomedical Research Center, Kyoto Institute of Technology, Japan
    Search for more papers by this author
    • Present address
      CNRS/UMR218, Institute Curie, Paris, France

  • Yasuko Kato,

    1.  Department of Applied Biology, Kyoto Institute of Technology, Japan
    2.  Insect Biomedical Research Center, Kyoto Institute of Technology, Japan
    Search for more papers by this author
  • Gaku Kimura,

    1.  Department of Applied Biology, Kyoto Institute of Technology, Japan
    2.  Insect Biomedical Research Center, Kyoto Institute of Technology, Japan
    Search for more papers by this author
  • Takashi Takeuchi,

    1.  Mitsubishi Kagaku Institute of Life Sciences (MITILS), Machida, Japan
    Search for more papers by this author
  • Masamitsu Yamaguchi

    1.  Department of Applied Biology, Kyoto Institute of Technology, Japan
    2.  Insect Biomedical Research Center, Kyoto Institute of Technology, Japan
    Search for more papers by this author

M. Yamaguchi, Department of Applied Biology, Kyoto Institute of Technology, Matsugasaki, Sakyo-ku, Kyoto 606-8585 Japan
Fax: +81 75 724 7760
Tel: +81 75 724 7781


Jumonji (Jmj) is a transcriptional repressor that plays important roles in the suppression of cell proliferation and development of various tissues in the mouse. To further clarify the roles of Jmj during development and gain insight into mechanisms of Jmj-mediated transcriptional regulation, we have taken advantage of Drosophila as a model organism. Drosophila Jmj (dJmj) shares high homology with mammalian Jmj in the JmjN, JmjC and AT-rich interaction domains, as well as in the N-terminal repression domain. dJmj localizes to hundreds of euchromatic sites but not to chromocenter heterochromatin on salivary gland polytene chromosomes. In addition, dJmj is excluded from regions stained with an antibody against Ser5-phosphorylated RNA polymerase II, suggesting a function of dJmj in transcriptionally inactive chromatin. Loss of djmj results in larval and pupal lethality with phenotypes similar to those observed in mutants of ecdysone-regulated genes, implying the involvement of dJmj in the repression of gene expression in the ecdysone pathway. Transgenic mouse Jmj mostly colocalizes with dJmj and partially rescues the phenotypes of djmj mutants, indicating that dJmj is a functional homolog of mammalian Jmj. Furthermore, mutation in djmj suppresses position effect variegation of the T(2;3)SbV rearrangement. These findings suggest that dJmj controls expression of developmentally important genes through modification of chromatin into a transcriptionally silenced state.


AT-rich interaction domain




Drosophila Jmj


glutathione S-transferase




little imaginal disks


mouse jumonji


RNA polymerase II

The basic unit of chromatin in eukaryotes is the nucleosome, which consists of 146 bp of DNA wrapped around an octamer of histones H2A, H2B, H3 and H4 [1]. Covalent modifications of histone tails, such as acetylation, methylation, phosphorylation and ubiquitination, modulate interaction affinities for chromatin-associated proteins, leading to the formation of either transcriptionally active or silent chromatin structures [2]. For example, methylation at Lys9 of histone H3 (H3-K9) by the su(var)3-9, enhancer of zeste, trithorax (SET) domain-containing protein SUV39H1 creates binding sites for the chromodomain-containing protein HP1, resulting in the establishment of heterochromatin [3]. In addition, methylation of H3-K27 and H4-K20 and hypoacetylation of histones are associated with transcriptionally silenced chromatin, whereas methylation of H3-K4 and hyperacetylation of histones are connected with active transcription [4].

The JmjC domain was initially characterized as a conserved domain among jumonji (Jmj) family proteins, including Jmj, RBP2 and SMCX, and has subsequently been identified in more than 100 proteins in prokaryotic and eukaryotic organisms [5–7]. JmjC-containing proteins have been shown to play important roles in various biological processes, including cellular differentiation, DNA repair and regulation of heterochromatin [8–10]. These JmjC-containing proteins are considered to regulate chromatin or transcription, as they are generally associated with chromatin- or DNA-binding domains, such as the plant homeo-domain (PHD) finger, the TUDOR domain, the AT-rich interaction domain (ARID) and the zinc finger motif [11–13]. Recent studies revealed that the JmjC-containing proteins are histone demethylases and that the JmjC domain is responsible for their enzymatic activity [14–19]. However, as several JmjC-containing proteins are predicted to be enzymatically inactive [11,20], additional mechanisms might be involved in JmjC-mediated regulation of chromatin or transcription.

The jmj gene was originally identified by a gene trap strategy in the mouse and shown to be required for the appropriate development of various tissues, including brain, liver, thymus and heart [7,21,22]. jmj encodes a transcriptional repressor containing the JmjC domain, JmjN domain and ARID. The latter two mediate the interaction of Jmj with A/T-rich DNA sequences [23]. Although the N-terminal region of Jmj itself is known to be responsible for its repressor activity [23,24], the mechanisms remain unknown. The JmjC domain of Jmj is predicted to be enzymatically inactive as a histone demethylase [11,12] and its function remains to be clarified.

Jmj appears to have an important role in suppression of cellular proliferation. In the developing heart, Jmj binds to the promoter and represses the expression of cyclinD1, which is essential for G1/S phase transition, thereby suppressing cell proliferation and regulating morphogenesis of cardiac cells [24]. Jmj also represses E2F activity and reduces cell cycle progression by associating with the Rb protein [25]. Furthermore, it represses expression of ANF, which encodes a hormonal mediator that is required for heart development, by counteracting the function of ANF activators Nkx2.5 and GATA4 [26]. As jmj is widely expressed and is required for the correct development of various tissues, involvement in the regulation of a diverse range of developmental programs, not limited to cardiac cells, is likely.

To further clarify the roles of Jmj during development and gain insight into mechanisms of Jmj-mediated chromatin regulation, we have taken advantage of Drosophila melanogaster as a model organism. We show here that loss of the Drosophila jumonji (djmj) gene results in larval and pupal lethality with phenotypes similar to those with ecdysone-regulated genes. On salivary gland polytene chromosomes, Drosophila Jmj (dJmj) localizes to euchromatic sites excluded from highly transcribed regions that are stained with an antibody against RNA polymerase II (PolII), suggesting a function of dJmj in transcriptionally inactive chromatin. Moreover, a djmj mutant suppresses the position effect variegation (PEV) of the T(2;3)SbV rearrangement. These observations suggest that dJmj controls expression of developmentally important genes through modification of chromatin into a transcriptionally silenced state.


The CG3654 gene encodes a Drosophila ortholog of mammalian Jmj

JmjC-containing proteins are classified into subgroups on the basis of their protein structures [11,17]. Jmj belongs to the JARID family, which is characterized by possession of the conserved domains, JmjN, JmjC and ARID [13]. Drosophila contains two JARID family proteins, little imaginal disks (Lid) and a novel protein CG3654 (Fig. 1A). Lid has been identified as a gene that enhances the phenotype of ash1 mutants, and is classified as a trithorax group gene [27]. Lid is considered to be a sole ortholog of mammalian JARID1 proteins, including RBP2, PLU-1, SMCX and SMCY, as all of them contain additional PHD fingers [12,13].

Figure 1.

 Identification of the Drosophila Jmj protein. (A) Schematic structures of mouse Jmj, Drosophila Jmj and Lid. The locations of the JmjN domain, JmjC domain, ARID, PHD, AT-hook domain and C5HC2 zinc finger domain are shown. (B) Amino acid alignment of the N-terminal repression domain of mouse and Drosophila Jmj. Identical and similar residues are shaded in black and gray, respectively.

Mouse Jmj (mJmj) and Drosophila CG3654 share 40%, 45% and 37% identities in the JmjN domain, JmjC domain and ARID, respectively (Fig. 1A). In addition to these conserved domains, mJmj contains a zinc finger motif at its C-terminus, whereas CG3654 possesses two AT-hook motifs (Fig. 1A). The N-terminal repression domain of Jmj is also conserved in CG3654 (Fig. 1B), but not in Lid. Therefore, we concluded that CG3654 is a Drosophila counterpart of mammalian Jmj and designated it as Drosophila jumonji (dJmj). Jmj proteins are also found in various species, from insects to mammals, but not in worms and yeasts. Importantly, all the Jmj proteins share high homology in the N-terminal region (data not shown), suggesting that this is important for Jmj function, probably acting as a repression domain.

djmj e03131 is a loss of function allele of djmj

The djmj gene localizes in the 67B9-10 cytological region and is composed of four exons, including 7053 bp of an ORF (Fig. 2A). To confirm the expression of dJmj protein, we generated a polyclonal antibody to dJmj by immunizing rabbits with the C-terminal region of dJmj (amino acids 1635–2351) as an antigen. Western blot analysis with affinity purified antibody to dJmj recognized a protein corresponding to the calculated molecular mass of dJmj (252 kDa) from embryo to adult stages, indicating continuous expression of dJmj throughout development (Fig. 2B, lanes 1–7). The lower band (120 kDa) detected by antibody to dJmj is evident in extracts of embryos (Fig. 2B, lanes 1 and 2) and embryo-derived Kc cells (Fig. 2B, lane 8). dsRNA-mediated knockdown of dJmj in Kc cells reduced the amount of the 250 kDa dJmj protein to an undetectable level at 4 days after dsRNA treatment, whereas that of the 120 kDa band was unchanged throughout dsRNA treatment (Fig. 2B, lane 9). Therefore, we concluded that the 120 kDa band is a nonspecific protein that is cross-reactive with the antibody. It should be noted that this cross-reactive 120 kDa band is undetectable in extracts from flies at later developmental stages.

Figure 2.

 Characterization of transposon-inserted djmj mutants. (A) The structure of djmj and the location of the transposon insertion in e03131 (piggyBac) is shown. The noncoding and coding regions of the djmj transcript are depicted as open and filled boxes, respectively. (B) Developmental western blot analysis of dJmj. Protein extracts from various developmental stages were probed with polyclonal antibody to dJmj. Anti-α-tubulin antibody was used to compare the amount of protein loading. An asterisk shows nonspecific bands. Lane 1: 0–12 h embryo. Lane 2: 12–24 h embryo. Lane 3: third larva. Lane 4: early pupa. Lane 5: late pupa. Lane 6: adult male. Lane 7: adult female. Lane 8: Kc cells. Lane 9: Kc cells treated with dsRNA. (C) Protein extracts from third instar larvae were subjected to western blotting with antibody to dJmj (upper). The same blot was reprobed with antibody to α-tubulin to compare protein loading (lower). Lane 1: wild type. Lane 2: djmje03131. Lane 3: djmje03131/Df(3L)AC1. (D) RT-PCR analysis of expression of djmj in third instar larvae from wild-type and djmje03131 mutants. Rp49 was used as an internal control. (E) Immunostaining for dJmj in whole salivary gland cells in wild-type and djmje03131 mutant larvae. DNA was visualized with DAPI. (F) Semiquantitative RT-PCR analysis of cell cycle regulators in wild-type and djmje03131 third instar larvae. Expression of rp49 was used as an internal control.

To clarify the in vivo roles of djmj, we analyzed transposon-inserted djmj mutants. Two fly strains that contain the P or piggyBac transposons in the djmj gene locus were identified. The djmjEY02717 allele is an insertion of the EY element [28] in the 5′-UTR of djmj. However, this insertion does not affect djmj expression, and homozygous djmjEY02717 flies proved to be viable and fertile (data not shown). The djmje03131 allele carries the insertion of the piggyBac construct RB, which contains the splice acceptor and an FLP recombination target (FRT) site [29], in the first intron of the djmj gene (Fig. 2A), and djmje03131 homozygotes, in contrast, showed a lethal phenotype. The dJmj protein was found to be absent in larval extracts of djmje03131 homozygotes or heterozygotes with the deficiency chromosome, Df(3L)AC1, which lacks a genomic region including the entire djmj locus (Fig. 2C). RT-PCR analysis also indicated a decrease of djmj transcripts in djmje03131 homozygotes (Fig. 2D). Immunostaining of whole salivary gland cells from third instar larvae showed predominant localization of dJmj protein in the nuclei of wild-type but not of djmje03131 homozygous cells (Fig. 2E).

As it has been reported that mammalian Jmj represses cyclinD1 expression via binding to its promoter [24], we investigated whether dJmj also represses the expression of cyclinD, the sole ortholog of mammalian cyclinD genes in Drosophila[30]. Semiquantitative RT-PCR analysis showed that cyclinD is not misregulated in djmje03131 mutant third instar larvae (Fig. 2F). The expression of other cell cycle regulators, including cyclinE, cdk4, E2Fs, Rbfs and stg, was also unaltered by loss of djmj (Fig. 2F and data not shown). These results suggest that dJmj does not play a dominant role in the repression of cell cycle regulators in Drosophila.

dJmj localizes to euchromatic regions on polytene chromosomes

The JmjC-containing proteins are thought to regulate chromatin or transcription [11,12]. To gain insight into the roles of dJmj in chromatin regulation, we analyzed its chromosomal localization by immunostaining of polytene chromosomes of salivary glands from third instar larvae (Fig. 3). DNA was visualized with 4′,6-diamidino-2-phenylindole (DAPI), which stains brightly at condensed DNA regions on euchromatic arms that are divided into bands and interbands and at chromocenter heterochromatin (Fig. 3A,D). Immunostaining of chromosomes with antibody to dJmj showed dJmj at hundreds of euchromatic sites with 10–20 bright signals (Fig. 3B,C). In contrast, no dJmj signals were detected in chromosomes of djmje03131 mutants (Fig. 3E,F). Higher magnification of merged images of dJmj and DAPI staining showed that dJmj was localized mostly to bands, but it was also observed in interbands and at band–interband boundaries, and no correlation was observed between dJmj localization and DNA density (Fig. 3G–I). dJmj was not localized in chromocenter heterochromatin, as confirmed by coimmunostaining of chromosomes with antibodies for dJmj and HP1, a marker of heterochromatin (Fig. 3J–L). These findings suggest that dJmj is involved in the regulation of specific target genes at euchromatin.

Figure 3.

 dJmj localizes to euchromatic regions on polytene chromosomes. (A–I) Polytene chromosomes of third instar larvae from wild-type (A–C, G–I) and djmje03131 mutants (D–F) were immunostained with antibody to dJmj (B, E, H). DNA was counterstained with DAPI (A, D, G). (C, F, I) Merged images of dJmj and DAPI staining. (G–I) Higher-magnification images of dJmj localization on polytene chromosomes of another spread. (J–L) Higher magnification of dJmj staining at chromocenter heterochromatin. Polytene chromosomes were coimmunostained with antibodies for HP1 (J) and dJmj (K).  (L) Merged image of dJmj and HP1 staining.

dJmj is excluded from highly transcribed chromatin regions

Given that mammalian Jmj functions as a transcriptional repressor [23,24], dJmj is likely to be associated with transcriptionally inactive chromatin. To investigate the correlation between dJmj localization and transcriptional activity, we performed coimmunostaining of polytene chromosomes with antibodies for dJmj (Fig. 4A,D) and PolII (Fig. 4B,E). Immunostaining with an antibody against Ser5-phosphorylated PolII detected numerous euchromatic bands in actively transcribed regions of the genome. Merged images of dJmj and PolII staining revealed no overlap in the distributions of these two proteins (Fig. 4C,F), suggesting that dJmj is associated with transcriptionally inactive chromatin.

Figure 4.

 dJmj is excluded from highly transcribed chromatin regions. (A–F) Polytene chromosomes from wild-type third larvae were stained with antibodies for dJmj (A, D) and PolII (B, E). Higher-magnification images of dJmj (D) and PolII (E) staining of another spread are also shown. (C, F) Merged images of dJmj and PolII staining.

djmj is a suppressor of position effect variegation

To address whether dJmj regulates the organization of chromatin structure, we examined the effect of djmj on position effect variegation (Table 1). In chromosomes with T(2;3)SbV rearrangement, the dominant Stubble mutation (Sb1), which results in a short bristle phenotype, is relocated close to pericentromeric heterochromatin, resulting in heterochromatin-induced silencing of Sb1 and a wild-type bristle phenotype [31]. Female flies of wild-type, djmje03131/TM6B and SUV4-20BG00814, a known suppressor of SbV variegation [32], were each crossed with T(2;3)SbV/TM3 males, and the bristles of the progeny were scored for Sb expression. On the wild-type genetic background, 29.7% of bristles showed the Sb phenotype. As a positive control, we confirmed that on the background of SUV4-20BG00814, Sb bristles were increased to 54.7%. In the djmje03131 mutant background the Sb bristles were significantly increased to 52.6%, indicating that djmje03131 acts as a suppressor of PEV. Similar results were obtained for the Df(3L)AC1 chromosome, which lacks a djmj locus in the genome. These results suggest the involvement of dJmj in the establishment and/or maintenance of the closed chromatin structure.

Table 1.   The djmj gene is a suppressor of position effect variegation of the T(2;3)SbV rearrangement.
GenotypeNumber of fliesTotal bristlesNumber of SbSb (%)

djmj is required for metamorphosis

To investigate in more detail the lethal phenotypes and lethal phases associated with djmj mutants, the djmje03131 allele was balanced with the green fluorescent protein-expressing balancer chromosome, and viable larvae were counted in each developmental stage. Almost all nonfluorescent djmje03131 homozygous larvae developed to the end of the third instar larvae, similarly to control animals. Approximately 95% of djmje03131 homozygous animals initiated pupation, but this was delayed for 2–3 days as compared to control animals, whereas the remaining animals continued to wander and did not undergo pupation. Of pupated djmje03131 homozygotes, 23% died in the early pupal stage (Fig. 5A,C). Other animals developed to the late pupal stage or pharate adults, with a few escapers that died shortly after eclosion (Fig. 5A). Precise excision of the piggyBac transposon reversed the lethality, indicating that the transposon insertion was indeed responsible for the phenotype (data not shown). Hemizygous djmje03131/Df(3L)AC1 animals also exhibited larval and pupal lethality and displayed similar phenotypes as homozygous djmje03131 mutants (Fig. 5A and data not shown), confirming that djmje03131 is a loss of function allele of djmj.

Figure 5.

 The djmj gene is required for metamorphosis. (A) Lethal phases were determined in animals with the following genotypes: +/+, djmje03131 and djmje03131/Df(3L)AC1. (B–F) Lethal phenotypes of djmje03131 homozygotes. (B) Wild-type control animal 4 days after pupation. (C, D) djmje03131 mutant animals 5 days after pupation. (E, F) djmje03131 mutants show a crooked leg phenotype. Third legs dissected from wild-type (E) and djmje03131 pharate adults (F) are shown.

Phenotypic characterization of pharate adults revealed some mutants to have defects in leg elongation and to show a crooked leg phenotype (Fig. 5D,F). These phenotypes are similar to those with loss of function of the genes involved in the ecdysone pathway [33,34], suggesting the participation of dJmj in ecdysone signaling.

The jmj gene is functionally conserved from flies to mammals

To investigate whether djmj is a functional homolog of mammalian jmj, we tested the chromosomal distribution of mJmj and its ability to rescue the phenotypes of the djmj mutants. To this end, transgenic flies that express FLAG-tagged full-length mJmj (FLAG–mJmj) under the control of the GAL4–UAS system [35] were established. To minimize the expression of FLAG–mJmj, the hsp70–GAL4 driver line was used without heat shock treatment, which results in leaky expression of FLAG–mJmj that is barely detected by western blotting with antibody to FLAG (Fig. 6A). Immunostaining of polytene chromosomes from FLAG–mJmj-expressing salivary gland cells detected numerous euchromatic bands (Fig. 6C,F), whereas no FLAG signals were detected in chromosomes without hsp70–GAL4 (Fig. 6H–J). Coimmunostaining of chromosomes with antibodies for dJmj (Fig. 6B,E) and FLAG (Fig. 6C,F) showed that most, but not all, mJmj sites colocalize with endogenous dJmj (Fig. 6D,G), suggesting that mJmj has similar function as dJmj on chromatin. The number of mJmj-binding sites was much greater than that for dJmj. This could be due to higher expression of FLAG–mJmj on transgenic lines as compared to endogenous dJmj or to stronger affinity of the antibody for FLAG.

Figure 6.

 Transgenic mouse Jmj mostly colocalizes with endogenous dJmj. (A) Western blot analysis of FLAG–mJmj expression in larval extracts with the indicated genotypes using antibody to FLAG (upper). The same blot was reprobed with antibody to tubulin to compare protein loading (lower). Lane 1: FLAG–mjmj/+. Lane 2: FLAG–mjmj/hsp70–GAL4. (B–J) Polytene chromosomes from FLAG–mjmj/hsp70–GAL4 (B–G) or FLAG–mjmj/+ (H–J) larvae were coimmunostained with antibodies to dJmj (B, E, H) and FLAG (C, F, I). (D, G, J) Merged images of dJmj and FLAG–mJmj staining. (E–G) Higher-magnification images of each staining.

We then expressed mJmj under the background of djmje03131 and investigated the lethal phases of the rescued flies (Table 2). As most djmje03131 homozygotes develop to the pupal stage (Fig. 5), third larvae with the desired genotype were picked up and tested for their lethal phases and phenotypes during pupal stages. Of the control flies that contain the either FLAG–mjmj (line 35) transgene or the hsp70–GAL4 driver under the background of the djmj mutation, 10.7–14.7% of pupae showed the abnormal leg phenotype and 0.6–7.1% of animals eclosed, which is similar to what was seen with djmje03131 homozygous mutants. In contrast, when mJmj was ubiquitously and modestly expressed by the hsp70–GAL4 driver, the abnormal leg phenotype was restored and 21.2% of rescued animals eclosed, indicating that mJmj can partially compensate for loss of djmj. The FLAG–mjmj transgene inserted in the independent genomic locus (line 19) showed similar, but less pronounced, effects on the rescue experiment. It is not possible to draw definitive conclusions regarding the degree to which mJmj can rescue the djmj mutant phenotype, as we have not yet succeeded in cloning the full-length cDNA for djmj to make djmj-expressing flies, due to its large size. However, these findings strongly suggest the functional conservation of the jmj gene from flies to mammals.

Table 2.   Transgenic mJmj partially rescues the phenotypes of djmje03131 mutants.
GenotypeLethal phase
Early pupaAbnormal legaLate pupaAdultTotal
  • a

     The number of late pupae that show the crooked leg phenotype.

+/hsp70–GAL4; djmje03131/djmje031316 (10.7%)6 (10.7%)40 (71.4%)4 (7.1%)56
FLAG–mjmj(35)/+ djmje03131/djmje0313113 (8.0%)24 (14.7%)127 (77.9%)1 (0.6%)163
FLAG–mjmj(19)/+ djmje03131/djmje031315 (4.9%)15 (14.7%)81 (79.4%)1 (1.0%)102
FLAG–mjmj(35)/hsp70–GAL4; djmje03131/djmje031310 (0.0%)2 (3.8%)39 (75.0%)11 (21.2%)52
FLAG–mjmj(19)/hsp70–GAL4; djmje03131/djmje031310 (0.0%)2 (4.9%)37 (90.0%)2 (4.9%)41


Although the Drosophila genome contains at least 13 genes encoding JmjC domain-containing proteins [11], little is known about their biological roles and their contributions to chromatin regulation. In this study, we showed that a novel JmjC-containing protein, dJmj, a Drosophila homolog of mammalian Jmj, is associated with euchromatic sites excluded from highly transcribed regions on polytene chromosomes and is required for metamorphosis during development.

The mjmj gene appears to be involved in many developmental pathways, as clarified by analysis of mutant mice that show various developmental abnormalities [7,21,22]. In the present study, loss of djmj function caused lethality during larval and pupal stages (Fig. 5), indicating that djmj is also important in Drosophila development. Jmj plays critical roles in suppression of cellular proliferation via repression of cyclinD1[24]. However, dJmj is not likely to regulate Drosophila cyclinD, as the expression of cyclinD was unchanged in djmj mutant larvae (Fig. 2F) and in dJmj-depleted Kc cells (data not shown). It is important to note that, unlike mammalian D-type cyclin proteins, Drosophila cyclin D is not required for G1/S phase transition but instead plays a role in cellular growth, whereas cyclin E plays an essential role in G1/S phase progression [36]. However, cyclinE and several other cell cycle-related genes were not misregulated in djmj mutant larvae (Fig. 2F and data not shown). Furthermore, dJmj depletion did not affect cell growth in Kc cells (data not shown). Therefore, cyclinD repression and subsequent suppression of cellular proliferation might be a mammal-specific event. However, these data do not rule out the possibility that dJmj might repress cyclinD expression in restricted tissues, which would not be detected by expression analysis of extracts of whole animals. In addition, although relatively high expression of dJmj was observed during embryonic stages (Fig. 2B), it remains unclear whether dJmj is required for the repression of cell cycle regulators during early development, as maternally deposited dJmj protein might contribute to embryogenesis in djmj mutants. Further studies are required to investigate the involvement of dJmj in cell cycle regulation during early embryonic development.

The detailed mechanism by which Jmj represses transcription remains to be clarified. Although it has been shown to counteract the function of DNA-binding transcription factors [25,26], Jmj directly binds to the cyclinD1 promoter to repress its expression [24]. As our data do not show direct evidence that dJmj has a transcriptional repression activity, we cannot conclude that dJmj is indeed a transcriptional repressor like mammalian Jmj. However, the observation that dJmj localizes on specific chromatin domains excluded from PolII sites on polytene chromosomes suggests that dJmj mediates transcriptional repression through modification of chromatin. In addition, djmj is not likely to affect global modification of histone tails that are associated with transcriptional activity (supplementary Fig. S1). Therefore, our findings suggest that dJmj is involved in the regulation of specific target genes at specific chromosomal loci in response to developmental signals rather than acting as a global regulator of chromatin.

The finding that the phenotypes of djmj mutants resemble those of Drosophila lacking ecdysone-regulated genes [33,34] suggests the involvement of dJmj in the ecdysone pathway. Expression of early and late puff genes are regulated in a direct or indirect manner by a subset of chromatin-modifying proteins, including NURF, p66, dGcn5, dAda2a, Bonus, Rpd3 and dG9a [37–43]. In addition, one property of JmjC-containing proteins is to associate with chromatin modification enzymes, such as the NCoR corepressor and histone deacetylase (HDACs) [8,44,45]. Investigation of whether dJmj links with these proteins to control metamorphosis is clearly warranted. The possible interaction domain of dJmj for these factors is the N-terminal repression domain, which is evolutionarily conserved among Jmj proteins (Fig. 1). Detailed analysis of the role of N-terminal and the JmjC domains in dJmj function may provide clues with which to address these issues.

Several studies have clarified that JmjC-containing proteins act as histone demethylases [11]. Lid, the closest protein to dJmj, was recently shown to be a histone demethylase that removes dimethyl and trimethyl K4 of H3 [46–48]. Although our results showed that the mutation in the djmj gene does not affect global modification of histone tails, including dimethyl K4 of H3 (supplementary Fig. S1), we cannot rule out the possibility that dJmj might demethylate histones at specific chromosomal loci or target a nonhistone protein as a substrate. However, importantly, both mammalian and Drosophila Jmj proteins are predicted to be catalytically inactive as histone demethylases because of the amino acid changes in the catalytic domain [11,12]. Several other JmjC-containing proteins are considered to be enzymatically inactive as histone demethylases [11]. Epe1 has been shown to counteract heterochromatin formation by interacting with Swi6, a yeast homolog of HP1. This event requires an enzymatically inactive JmjC domain, suggesting a novel function of the JmjC domain of Epe1 in heterochromatin formation [49]. As the JmjC domain is also found in bacteria, it might have diverse functions, and its analysis in dJmj should provide novel insights.

Despite the finding of djmj as a suppressor of PEV, the detailed roles of dJmj in chromatin organization remain unclear. Several different genes are reported to similarly act as suppressors, including Su(var)2-5, Su(var)3-7 and Su(var)3-9, which encode structural components of heterochromatin localizing to chromocenter heterochromatin [50,51], and Z4, which encodes a zinc finger protein that localizes to interbands of euchromatin and regulates chromatin organization at band–interband boundaries [52]. In addition, JIL-1 histone kinase functions to maintain euchromatic regions via antagonizing heterochromatinization by Su(var)3-9 [53,54]. On polytene chromosomes, dJmj signals were excluded from chromocenter heterochromatin, and heterochromatin components, including dimethyl K9-H3 and HP1, were not altered by loss of dJmj (data not shown). In addition, dJmj does not affect PEV of the whitem4 rearrangement (data not shown). Taken together, these findings strongly suggest that dJmj is not a structural element in heterochromatin and acts at particular domains rather than functioning as a general modifier of chromatin.

In conclusion, our data suggest that dJmj plays important roles during metamorphosis by regulating gene expression in response to developmental signals. As mJmj shows similar distributions to dJmj on polytene chromosomes (Fig. 6) and partially rescues the phenotypes of djmj mutants (Table 2), the Drosophila system could be a powerful tool with which to analyze Jmj functions in chromatin regulation and development.

Experimental procedures

Fly stocks

Fly stocks were raised at 25 °C on standard medium. Canton-S was used as the wild-type strain. The piggyBac-inserted djmje03131/TM6B fly was obtained from the Harvard stock center [29], and djmjEY02717, Df(3L)AC1 rnroe-1 pp/TM3, SUV4-20BG00814 and T(2;3)SbV, In(3R)Mo, Sb1, sr1/TM3Ser flies were from the Bloomington stock center. The hsp70–GAL4/CyO and whitem4 flies were obtained from the Drosophila Genetic Resource Center at Kyoto Institute of Technology.

Lethal phase analysis and phenotypic characterization

The djmje03131 and Df(3L)AC1 alleles were rebalanced with TM6BGFP and TM3GFP balancer chromosomes, respectively. Lethal phase analysis and phenotypic characterization were performed as previously described [34].

Generation of transgenic flies and rescue experiment

For constructing the pUAST–FLAG–mjmj vector, a cDNA for FLAG–mjmj in pBluescript was digested with ClaI, blunt-ended and inserted into the pUAST vector [35], which was blunt-ended after EcoRI digestion. Transgenic fly lines were generated as described previously [55,56], and three independent fly lines carrying the transgene on the second chromosome were established. The GAL4–UAS system [35] was used for ubiquitous expression of FLAG–mJmj using the hsp70–GAL4 driver.

For the rescue experiment, FLAG–mjmj (line 35), djmje03131/TM6B or FLAG–mjmj (line 19)/CyOGFP, djmje03131/TM6B females were crossed with hsp70–GAL4/CyOGFP, djmje03131/TM6B males at 25 °C. As control crosses, djmje03131/TM6BGFP females and males were mated with hsp70–GAL4/CyOGFP, djmje03131/TM6B males and FLAG-mjmj/(CyOGFP), djmje03131/TM6B females, respectively. Nontubby and nonfluorescent third larvae were picked up, and their lethal phases and phenotypes during pupal development were analyzed.

PEV analysis

To examine the effect of djmj on the whitem4 variegation, wm4/wm4 females were crossed with w/Y, djmje03131/TM6B males, and the eyes of wm4/Y, djmje03131/+males were scored and compared with those of wm4/Y, TM6B/+males. The effect of djmj on the SbV variegation was studied by crossing SUV4-20BG00814, djmje03131/TM6B, Df(3L)AC1/TM3SerGFP or Canton S females with T(2;3)SbV/TM3Ser males [31], and 14 defined bristles were scored as being wild type or Sb. Male and female scores were combined because no differences between sexes were observed.

Production of polyclonal antibody to dJmj

To construct an expression vector for the glutathione S-transferase (GST)-fused C-terminal region of the dJmj protein (dJmjC, amino acids 1635–2351), the djmj cDNA fragment was inserted into the SalI and NotI sites of the pGEX4T-1 vector. GST–dJmjC was expressed in the bacterial strain BL-21(DE3), affinity purified with a glutathione Sepharose column (GE Healthcare, Little Chalfont, UK), and injected into rabbits. The antiserum generated was applied to GST-conjugated sepharose, and this was followed by purification with GST–dJmj-conjugated sepharose.

Cell culture and knockdown experiments

Kc cells were cultured at 25 °C in M3 medium (Sigma, St Louis, MO, USA) supplemented with 2% fetal bovine serum. For dsRNA production, a 621 bp fragment spanning from nucleotide 6485 to the 3′-UTR (40 bp downstream of the stop codon) of djmj were amplified using 5′-CACGGGCGTATACCTCAAGC-3′ and 5′-TGTGCCTGAATCTTTCGTGC-3′ primers and cloned into the pGEM-T vector. Sense and antisense RNAs were synthesized in vitro and annealed. For knockdown experiments, 1 × 106 cells were plated on 6 cm dishes and transfected with 10 µg of dsRNA using cellfectin transfection reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. The cells were collected, directly suspended in SDS sample buffer, and subjected to western blotting.

Western blotting

Protein extracts were prepared by homogenization of animals in ice-cold SDS sample buffer followed by boiling for 5 min. After centrifugation at 12 000 g for 10 min at 4 °C, protein samples were separated by SDS/PAGE and transferred to poly(vinylidene difluoride) membranes (Millipore, Billerica, MA, USA). Antibodies used were anti-dJmj (1 : 2000), anti-α-tubulin (1 : 5000, Sigma), anti-FLAG (M2, 1 : 2000; Sigma), anti-acetyl H3 (06–599, 1 : 5000), anti-dimethyl K4-H3 (07-030, 1 : 2000), anti-monomethyl K9-H3 (07–450, 1 : 1000), anti-dimethyl K9-H3 (07–212, 1 : 1000), and anti-trimethyl K27-H3 (07–449, 1 : 1000) from Upstate (Lake Placid, NY, USA), and anti-H3 (1 : 1,000; Cell Signaling, Danvers, MA, USA). Horseradish peroxidase-conjugated anti-rabbit and anti-mouse IgGs (GE Healthcare) were used as secondary antibodies, and proteins were detected with ECL-plus (GE Healthcare).

Immunostaining of polytene chromosomes and whole salivary glands

For immunostaining of polytene chromosomes, salivary glands from wandering third instar larvae were dissected in 0.7% NaCl, fixed for 5 min, and squashed in 45% acetic acid/3.7% formaldehyde. The slides were frozen in liquid nitrogen and were then blocked in blocking buffer (5% skimmed milk in NaCl/Pi/0.1% Triton X-100) for 1 h at 25 °C. Slides were incubated with primary antibodies for 16 h at 4 °C. The antibodies used were anti-dJmj (1 : 400), anti-FLAG (M2, 1 : 5,000; Sigma), anti-PolII (H-14, 1 : 100; Covance, Princeton, NJ, USA) and anti-HP1 (C1A9, 1 : 100; Developmental Studies Hybridoma Bank at the University of Iowa). After being washed with NaCl/Pi/0.1% Triton X-100 twice for 15 min each, the slides were incubated with Alexa-488-conjugated anti-rabbit IgG, Alexa-488-conjugated anti-mouse IgM, or Alexa-594-conjugated anti-mouse IgG or anti-rabbit IgG (1 : 400) from Invitrogen for 2 h at 25 °C. DNA was visualized with DAPI. Preparations were mounted in FluoroGuard Antifade Reagent (Bio-Rad, Hercules, CA, USA), and images were obtained using an Olympus (Tokyo, Japan) BX-50 microscope equipped with a cooled CCD camera. Each staining experiment was performed at least three times, and representative spreads are shown.

For immunostaining of whole salivary glands, dissected glands were fixed in 4% formaldehyde/0.15% Triton X-100 for 20 min on ice. After blocking in NaCl/Pi containing 2% goat serum and 0.15% Triton X-100 for 30 min at 25 °C, the glands were incubated with antibody to dJmj (1 : 400) for 16 h at 4 °C, and this was followed by incubation with Alexa-488-conjugated anti-rabbit IgG (1 : 400) for 2 h at 25 °C. DNA was stained with DAPI.

Semiquantitative RT-PCR

Total RNA was extracted with Sepasol RNA I (Nacalai, Kyoto, Japan). First-strand cDNA was synthesized using oligo(dT)20 and Superscript III reverse transcriptase (Invitrogen). PCR reactions were performed over a range of cDNA dilutions to ensure exponential amplification. Primer sequences used were as follows: cycD-F, 5′-GGGATCCCACATTGTATTCG-3′; cycD-R, 5′-ACGGAGCTTTGAAGCCAGTA-3′; cycE-F, 5′-AAGGTGCAGAAGACGCACTT-3′; cycE-R, 5′-AATCACCTGCCAATCCAGAC-3′; cdk4-F, 5′-TACAACAGCACCGTGGACAT-3′; cdk4-R, 5′-TGGGCATCGAGACTATAGGG-3′; rp49-F, 5′-CGGATCGATATGCTAAGCTG-3′; and rp49-R, 5′-GAACGCAGGCGACCGTTGGGG-3′.


We would like to thank Haruki Shirato for providing the FLAG–mjmj plasmid and members of the Yamaguchi laboratory for helpful comments and advice. We also acknowledge the contribution of Malcolm Moore in critical reading of the manuscript. This work was supported in part by grants-in-aid from the Ministry of Education, Sciences, Sports and Culture of Japan.