Roles of jumonji and jumonji family genes in chromatin regulation and development


  • Takashi Takeuchi,

    Corresponding author
    1. Mitsubishi Kagaku Institute of Life Sciences (MITILS), Machida, Tokyo, Japan
    2. Graduate School of Environment and Information Sciences, Yokohama National University, Hodogaya, Yokohama, Kanagawa, Japan
    • Mitsubishi Kagaku Institute of Life Sciences (MITILS), 11 Minamiooya, Machida, Tokyo, 194-8511 Japan
    Search for more papers by this author
  • Yutaka Watanabe,

    1. Department of Population Genetics, National Institute of Genetics, Mishima, Shizuoka, Japan
    Search for more papers by this author
  • Toshiyuki Takano-Shimizu,

    1. Department of Population Genetics, National Institute of Genetics, Mishima, Shizuoka, Japan
    2. Department of Biosystems Science, Graduate University for Advanced Studies (SOKENDAI), Hayama, Kanagawa, Japan
    Search for more papers by this author
  • Shunzo Kondo

    1. Mitsubishi Kagaku Institute of Life Sciences (MITILS), Machida, Tokyo, Japan
    Current affiliation:
    1. JEOL Ltd. 3-1-2 Musashino, Akishima, Tokyo 196-8558 Japan
    Search for more papers by this author


The jumonji (jmj) gene was identified by a mouse gene trap approach and has essential roles in the development of multiple tissues. The Jmj protein has a DNA binding domain, ARID, and two conserved jmj domains (jmjN and jmjC). In many diverse species including bacteria, fungi, plants, and animals, there are many jumonji family proteins that have only the jmjC domain or both jmj domains. Recently, Jmj protein was found to be a transcriptional repressor. Several proteins in the jumonji family are involved in transcriptional repression and/or chromatin regulation. Most recently, one of the human members has been shown to be a histone demethylase, and the jmjC domain is essential for the demethylase activity. Meanwhile, more and more evidence indicating that the jumonji family proteins play important roles during development is accumulating. Many proteins in the jumonji family may regulate chromatin and gene expression, and control development through various signaling pathways. Here, we highlight the roles of jmj and jumonji family proteins in chromatin regulation and development. Developmental Dynamics 235:2449–2459, 2006. © 2006 Wiley-Liss, Inc.


In the past, we have attempted to identify novel genes that have important roles in development using a mouse gene trap method and have successfully identified jumonji (jmj; Takeuchi et al.,1995; Takeuchi,1997). The gene was named for the morphology produced by the normal neural groove and abnormal grooves on the neural plates of jmj mutant mice. The morphology resembles a cross, and “jumonji” means cruciform in Japanese (Takeuchi et al.,1995).

We first reported that Jmj protein has two conserved domains (Takeuchi et al.,1995). Later, one was further divided into two domains, a jmjN domain and an AT-rich interaction domain (ARID), because there are many proteins that have only one of these two domains (Balciunas and Ronne,2000; Kortschak et al.,2000). ARID is a DNA binding domain, and proteins that have ARID belong to the ARID family. Another conserved domain in Jmj is designated as jmjC (Balciunas and Ronne,2000).

More than 100 proteins from many species, including bacteria, fungi, plants, and animals have been shown to have the jmjC domain or both the jmjN and jmjC domains (Balciunas and Ronne,2000; Clissold and Ponting,2001; In the present study, we define a jumonji family as a family whose members have the jmjC domain.

We have demonstrated previously that Jmj is a transcriptional repressor and represses cyclin D1 transcription in the embryonic heart and that this repression is required for normal cardiogenesis (Toyoda et al.,2003). Several reports also suggested that jumonji family members are involved in transcriptional or chromatin regulation. In fact, most recently, Tsukada et al. showed that a human family protein (JHDM1A, previous name; FBXL11) is a histone demethylase and that the jmjC is required for the enzymatic activity (Tsukada et al.,2006). This finding revealed one function of the jmjC domain, and provided the first evidence that jumonji family members can act as enzymes for histone modification.

Our studies have shown that jmj has essential roles in the development of multiple tissues. In addition, many reports suggest that proteins in the jumonji family play important roles during development. Many proteins in the jumonji family may regulate chromatin and gene expression and control development through various signaling pathways. In this paper, we review the roles of Jmj and other jumonji family proteins in chromatin regulation and development.


The jumonji family proteins comprise a large family. We constructed a dendrogram of human, Drosophila, and yeast jumonji family proteins (Fig. 1) based on their jmjC domain sequences. Almost all proteins containing jmjC domain in Pfam (, SMART ( JmjC&BLAST=DUMMY), and InterPro ( data bases were analyzed after redundant, variant, hypothetical, or low score (e>1.0E-04, SMART program) proteins were removed. Recent structure analysis of jmjC of HIF1AN/FIH, which has a 2-oxoglutarate (2-OG)-Fe(II)–dependent dioxygenase activity, revealed residues binding to the cofactors 2-OG and Fe(II), and showed that the domain contains a double-stranded β-helix motif (Elkins et al.,2003). A sequence alignment of the jmjC domains of various jumonji family proteins suggested that these features are conserved (Elkins et al.,2003; Trewick et al.,2005). Recently, one jumonji family protein, FBXL11/JHDM1A was shown to be a 2-OG-Fe(II)–dependent histone demethylase (Tsukada et al.,2006). The alignment of sequences for the dendrogram was made by taking previously predicted conservative residues such as cofactor binding sites (Clissold et al,2001; Trewick et al,2005; SMART and Pfam databases; Fig. 2).

Figure 1.

Unrooted neighbor-joining tree for 44 jumonji family proteins based on jmjC domain sequences. The tree was constructed by using Poisson correction distance for amino acid sequences and the MEGA package. Percentage bootstrap values (≥ 90%) are shown on internal branches. Protein names in humans, Drosophila, and budding and fission yeasts are represented as black, magenta, blue and orange, respectively. Accession numbers are presented in Drosophila and yeast proteins; those of human proteins are given in Figure 2. Boxes beside protein names show domains that each protein contains. Colored boxes show characteristic domains in each cluster. Asterisks mark proteins that have been reported to be involved in gene repression or histone modification.

Figure 2.

Multiple sequence alignment of jmjC domains. The regions of jmjC domains in SMART database were analyzed. The alignment was performed by using the MEGA program with a few modifications by eye. In this figure, only human proteins are shown as examples. Asterisks and # mark the Fe (II) and 2-OG binding residues, respectively (predicted; Trewick et al.,2005), of which identical residues with HIF1AN/FIH are highlighted with a red background. Residues that the frequency of the most common residue is ≥ 0.5 or the frequency of the most common group of similar residues (Dayhoff et al., 1979) is ≥ 0.7 are highlighted with a blue background.

The family contains distantly related members, as shown in Figure 1. Note, where we could not place a root. We can see several distinct clusters in the dendrogram based on the high bootstrap values (≥ 99%). Human JMJD2A-D and JARID1 and −2 constitute a single major cluster, which is referred to as cluster 1 hereafter. We also pay attention to three clusters: a cluster containing UTX, UTY, and JMJD3 (cluster 2); one containing JMJD1A-C and Hairless (cluster 3); and one containing PHF2 and -8, and FBXL10, -11 (JHDM1A, B; cluster 4).

Figure 1 also shows the domains that each member contains. Of interest, most proteins in cluster 1, 2, 3, and 4 have jmjN, tetratrico peptide repeat (TPR), C2HC4 zinc-finger and plant homeodomain (PHD) finger domain domains, respectively (Fig. 1). Three proteins in the cluster 4 also have F-box, leucine-rich repeat (LRR1) and CXXC zinc-finger domains. Function of the jmjN domain remains unknown. The TPR domain mediates protein–protein interactions and the assembly of multiprotein complexes (D'Andrea and Regan,2003). PHD finger is a C4HC3 zinc-finger–like motif found in nuclear proteins thought to be involved in chromatin-mediated transcriptional regulation (Bienz,2006). Remaining proteins such as PTDSR or JMJD4 do not have any apparent distinct domains other than jmjC (Fig. 1).

Proteins in cluster 1 are further divided into two groups, depending on the presence of the DNA binding domain ARID (Fig. 1). Only cluster 1 members contain ARID as well as jmjN domains in the jumonji family (Fig. 1). ARID proteins are involved in transcriptional regulation and a variety of biological processes, including development (Kortschak et al.,2000; Wilsker et al.,2005). Recently, the human and mouse proteins that have jmjN, jmjC, and ARID sequences were named JARID proteins, which are composed of five proteins: JARID1A/RBP2, JARID1B/PLU-1, JARID1C/SMCX, JARID1D/SMCY, and JARID2/JUMONJI (Wilsker et al.,2005). In this review, the original names are used to avoid confusion. Human JARID1 proteins (RBP2, PLU-1, SMCX, and SMCY) are closely related to each other and also have a PHD finger domain and a C5HC2 type Zn-finger domain. Q9VMJ7, little imaginal discs (lid), may be a Drosophila common homologue of JARID1 proteins. On the other hand, JUMONJI protein has a C5HC2 type Zn-finger domain but not PHD finger. The amino acid sequence homologies of jmjN, jmjC, and ARID are not high between the jumonji and JARID1 proteins. In addition, the positions of jmjN and ARID in JARID1 proteins are very close to the N-terminal; however, those of JUMONJI are at the center in the protein sequence. This feature is shared by the Q9VT00 Drosophila protein. The relatively high homologies of jmjN, jmjC, and ARID of Q9VT00 with those of human and mouse jumonji protein suggest that the Q9VT00 protein is a Drosophila homologue of the JUMONJI protein.

Although all JARID1 proteins and JUMONJI protein have ARID, the dendrogram suggests that the jmjC sequences of JARID1 proteins are more similar to those of JMJD2 proteins, which do not have ARID, than to that of JUMONJI (Fig. 1). We hypothesize that the ancient protein had only the jmjC domain, and then some of the descendant proteins acquired jmjN and ARID. Although we could not come to a definite conclusion with the low bootstrap probabilities, the dendrogram suggests that the acquisition occurred before diversification of the cluster 1 genes and then the JMJD2 group lost ARID during the course of evolution. Balciunas and Ronne have proposed that domain swapping may have occurred during evolution of the jumonji family proteins (Balciunas and Ronne,2000). In the case of the JMJD2 group, JMD2A, B, and C have tudor domains instead of ARID (Fig. 1). Of interest, the tudor domain is found in many proteins that colocalize with ribonucleoprotein or single-strand DNA-associated complexes in the nucleus.

JMJD proteins scatter in the family and do not show a monophyly (Fig. 1). Because there is no apparent relationship between classification (nomenclature of A-D) of the proteins and positions in the dendrogram, the nomenclature should be reconsidered based on the phylogenetic relationships or functions.

Because the family contains many distantly related proteins, investigation of functional or structural conservation is required to ascertain whether these proteins form one functional family. The 2-OG-Fe(II)–dependent dioxygenase activity of jumonji family proteins is especially worth analyzing. HIF1AN/FIH and FBXL11/JHDM1A catalyze protein hydroxylation and histone demethylation, respectively, through the dioxygenase activity (Hewitson et al.,2002; Tsukada et al.,2006). In addition, the structure of jmjC involved in the enzymatic activity appears to be conserved in many members (Trewick et al.,2005; Fig. 2).


Many jumonji family proteins have domains involved in DNA binding, chromatin binding, or transcription, such as ARID (Kortschak et al.,2000; Wilsker et al.,2005), PHD finger (Bienz,2006), or Zn-fingers, and jmjC domains showing metalloenzyme-like structures, suggesting that jumonji family proteins regulate transcription or chromatin function or both (Balciunas and Ronne,2000; Clissold and Ponting,2001; Trewick et al.,2005).

In fact, the potential activities of Plu-1 or JMJD2A for transcriptional repression have been reported (Tan et al.,2003; Zhang et al.,2005). Hairless functions as a nuclear receptor corepressor (Potter et al.,2001). In addition, mouse Jmj protein was shown to be a transcriptional repressor that directly represses cyclin D1 expression (Toyoda et al.,2003). Lee's group also reported that Jmj is a transcriptional repressor that can repress the promoter activities of atrial natriuretic factor and alpha-cardiac myosin heavy chain genes (Kim et al.,2003,2004,2005). The region that is necessary for the repression is located in the N-terminal of Jmj (Kim et al.,2003; Toyoda et al.,2003). Further roles of the region are unknown; however, it is possible that the region binds to cofactors involved in repression, such as histone modification.

Covalent modifications of histone tails, such as methylation, acetylation, and phosphorylation have important roles in regulating chromatin dynamics (Strahl and Allis,2000). Indeed, many residues in histone tails are modified. For instance, several lysine and arginine residues are methylated. With regard to lysine, methylation occurs on at least five residues of H3 (H3-K4, -K9, -K27, -K36, and -K79) and a single lysine residue of H4 (H4-K20). The effects of methylation are dependent on the particular lysine residue (Martin and Zhang,2005). For example, methylation of H3-K4 and H3-K9 is linked to transcriptional activation and repression, respectively. In addition, the biological effect of methylation can differ, depending on the state of methylation (mono-, di-, or tri-methylation), even within the same lysine residue (Martin and Zhang,2005).

Although the extent of histone acetylation is determined by both acetyltransferases and deacetylases, it is still unclear whether histone methylation is also regulated by enzymes with opposing activities. Shi et al. showed that LSD1, a nuclear amine oxidase, found in several histone deacetylase complexes, specifically demethylates mono- or di-methyl-H3-K4 (H3-K4me1 or H3-K4me2, respectively) in a flavin adenine dinucleotide-dependent oxidase reaction (Shi et al.,2004). However, this type of enzyme cannot demethylate tri-methylated residues.

Trewich et al. proposed that certain jumonji family proteins may be other types of histone demethylases (Trewick et al.,2005). Their concept appears to be based on two points. First, a jumonji family protein Epe1 in the fission yeast is required for heterochromatin integrity. Inactivation of Epe1 promoted continuous spreading of heterochromatin-associated histone modifications such as methylation of H3-K9 (Ayoub et al.,2003), suggesting that Epe1 can act as a histone modifier. Second, jmjC domains in Epe1 and other jumonji family proteins can be modeled onto the structure of the jmjC domain of HIF1AN/FIH, a 2-OG-Fe (II)–dependent dioxygenase, that is a member of the jumonji family. HIF1AN/FIH mediates the hydroxylation of an asparagine residue in hypoxia-inducible factor α (HIFα; Hewitson et al.,2002) and inhibits the transactivation by HIFα. The structure of jmjC of HIF1AN/FIH (Elkins et al.,2003) and alignment of the amino-acid sequence of Epe1 and jumonji family proteins suggested that many jumonji family proteins have structures and enzymatic activities similar to those of HIF1AN/FIH. Escherichia coli AlkB family proteins, which remove methyl groups in DNA through oxidative demethylation (Falnes et al.,2002; Trewick et al.,2002), are also 2-OG-Fe (II)–dependent dioxygenases. The above hypothesis was proposed based on analogy of the mechanisms of demethylation by AlkB proteins.

Most recently, Tsukada et al. also speculated that the mechanism of a new type of histone demethylation is similar to that of DNA demethylation by the AlkB family proteins, and purified a jumonji family protein, FBXL11 as a histone demethylase from a human cell line using a biochemical assay (Tsukada et al.,2006). They renamed the protein JHDM1A (jmjC domain-containing histone demethylase 1A). In the presence of 2-oxoglutarate and Fe (II), JHDM1 demethylates H3-K36me2. The jmjC domain is critical for demethylase activity. The authors also showed that the same activity is detected in a highly related protein of JHDM1A, FBXL10 (which they renamed to JHDM1B), and a homologue in the budding yeast, scJHDM1.

Studies in yeast have shown that methylation of H3-K36 by SET2 recruits histone deacetylase to transcribed regions and links to phosphorylation of the C-terminal domain of RNA polymerase II and the process of transcriptional elongation (Carrozza et al.,2005; Joshi and Struhl,2005; Keogh et al.,2005). Of interest, Carrozza et al. reported that the histone deacetylation occurs at coding regions and suppresses intragenic transcription. These results suggest that JHDM1 modulates this regulation and controls transcription quantitatively or qualitatively.

The LSD1 protein family contains approximately 10 related proteins, and these proteins cannot demethylate tri-methylated residues. In contrast, the jumonji family is a large family as described above, and the 2-OG-Fe (II)–dependent dioxygenase activity of jmjC has the potential to demethylate tri-methylated residues, suggesting that the jumonji family proteins can demethylate various residues in three methylation states of histones (Tsukada et al.,2006).

The jumonji family proteins or the protein complexes may also be able to modulate histones in other ways. For example, a fission yeast jumonji family protein, Msc1, forms a complex exhibiting histone deacetylase activity and the lack of Msc1 enhanced acetylation of histone H3 tails (Ahmed et al.,2004). JMJD2A, a member of the JMJD2 subfamily, binds to pRb and HDAC (histone deacetylase) -1 and -3, and has the potential to perform pRb-mediated repression of E2F-regulated promoters (Gray et al.,2005). Hairless, a nuclear receptor corepressor, also binds to HDAC-1, -3, and -5 (Potter et al.,2001). Taken together, it is most likely that jumonji family proteins and/or their protein complexes can modulate the diverse range of histone modifications and regulate the expression of genes epigenetically and can control various biological events including development.


As described above, the jumonji (jmj) gene was identified by a mouse gene trap strategy (Takeuchi et al.,1995).The roles during development have been investigated most intensively among jumonji family genes.

jmj Expression Pattern During Development

jmj is expressed in a wide range of adult tissues in mice (unpublished data) and humans (Berge-Lefranc et al.,1996), and jmj is strongly expressed in embryonic stem cells. During development, very weak expression is observed in the whole embryonic body at embryonic day 8 (E8), after which, a band-like expression pattern is detected at the future midbrain–hindbrain boundary at E8.5. The expression becomes stronger and shows a ring-like pattern after neural tube closure at E9 (Takeuchi et al.,1995). In addition, weak expression is seen in the forebrain and stronger expression is also detected in the bulbous cordis (future outflow tract and right ventricle) of the heart, the primitive pharynx, and around the posterior neuropore in the tail. Subsequent to this, the regions where jmj is expressed expand gradually and the expression is detected in almost all adult tissues, although the intensities are different among cell types. It is difficult to identify the rules explaining all spatial and temporal expression patterns of jmj in various tissues and cell types; however, the expression pattern tends to be involved in cell proliferation or differentiation at least in several tissues. For example, strong expression can be detected in many neurons after final mitosis in the cerebrum and cerebellum (Takeuchi et al.,1995; Takeuchi,1997). jmj expression starts when the levels of cell proliferation start to decrease in cardiac myocytes in the ventricles (Toyoda et al.,2003). A more typical expression pattern can be seen in the lens (Fig. 3). jmj expression is not detected substantially at E11.5 when cells in whole lens vesicles proliferate (Fig. 3). However, the expression can be detected first in the posterior cells that exit the cell cycle and form the primary lens fibers, whereas expression cannot be detected in anterior lens epithelial cells that are still proliferating at E12.5 (Fig. 3). The lens epithelial cells then move toward the equatorial region of the vesicle and exit the cell cycle. These cells become gradually incorporated into the lens proper, and subsequently develop into secondary lens fibers. jmj expression starts at the equatorial regions and the intensity increases in lens fibers; however, the expression cannot be detected in the lens epithelial cells at E13.5 and E14.5 (Fig. 3). These patterns suggest a correlation of jmj expression with cell proliferation or differentiation of lens cells.

Figure 3.

Expression pattern of jmj gene in the lens. jmj expression was monitored using lacZ, which was introduced into the jmj gene. jmj expression is not detected substantially at embryonic day 11.5 (E11.5). The expression can be detected first in the posterior elongated cells forming the primary lens fibers at E12.5. jmj expression is also detected at the equatorial regions and the intensity increases in the lens fibers, but the expression cannot be detected in the lens epithelial cells at E13.5 and E14.5. Arrowheads and LE show the equatorial region of the lens vesicle and the lens epithelial cells, respectively. Scale bar = 50 μm.

jmj Mutant Strains and Genetic Backgrounds

The function of jmj gene has been investigated using its mutant mice that were also obtained by the gene trap strategy. Two jmj mutant strains have been used for functional analyses (Takeuchi et al.,1995; Baker et al.,1997). In the case of our mutant mice, we showed by expression analyses, including in situ hybridization (unpublished data) and Western blotting (Toyoda et al.,2000), that the trap vector disrupted the jmj gene in a whole body. On the other hand, jmj was disrupted in a limited number of tissues, such as the heart, in another jmj mutant strain (Lee et al.,2000). Although the mutant mice used by Lee's group showed abnormalities in only the heart (Lee et al.,2000) probably because of this limitation, our jmj mutant mice show developmental abnormalities in various tissues and die in utero (Takeuchi et al.,1995; Motoyama et al.,1997; Kitajima et al.,1999,2001; Takeuchi et al.,1999; Anzai et al.,2003).

jmj phenotypes are influenced by mouse genetic backgrounds. Therefore, we established jmj mutant strains on several genetic backgrounds. On a C3H/He background and probably a 129/Ola background, the mutant embryos show abnormal groove formation on the neural plate, neural tube defects, abnormal morphology in the right ventricle and enhanced proliferation of trabecular cardiac myocytes in the left and right ventricles, and die around E11.5 (Takeuchi et al.,1995,1999). All of these phenotypes were rescued by exogenous expression of jmj by transgenesis (Takahashi et al.,2004; Takahashi et al., unpublished data), indicating that the phenotypes result from mutation of jmj gene. In addition, a rescue experiment in only the heart showed that abnormalities in the heart cause lethality around E11.5 (Takahashi et al.,2004).

On BALB/c, C57BL/6, and DBA/2 backgrounds, the mutant embryos can survive until E15.5. Although most phenotypes observed in the mice on a C3H/He background cannot be observed, this survival enables us to analyze jmj phenotypes in midgestation. The mutant embryos show edema, hemorrhage, and hypoplasia of the liver, spleen, and thymus (Motoyama et al.,1997), and we found that definitive hematopoiesis is impaired (Kitajima et al.,1999).

Thus, jmj phenotypes are clearly divided into two types: C3H and BALB. However, mutant embryos on a C3H background would also have BALB-type phenotypes. Mutant embryos on a C3H background, with exogenous expression of jmj in the heart, survive but finally die around E15.5 and these embryos show similar phenotypes to mutant embryos on a BALB background (Takahashi et al.,2004). Most likely, BALB type phenotypes are masked by lethality around E11.5 in mutant embryos on a C3H background. Differences in the activities of unknown gene(s) (modifier) would cause in differences of phenotypes (Ohno et al.,2004).

jmj Functions in Cardiac Development

Abnormalities in the heart of jmj mutant mice have been investigated intensively. The cardiac ventricles are composed of two layers of cardiac myocytes: trabecular and compact layers. Cardiac myocytes in trabecular layer showed hyperproliferation in our jmj mutant mice (Takeuchi et al.,1999; Toyoda et al.,2003). From analysis of the hyperproliferation, we found that jmj expression starts at the stages when cell proliferation of cardiac myocytes starts to decrease in normal embryos and that cell proliferation and expression of cyclin D1, the gene that encodes one of the G1 cyclins, are not repressed in jmj embryos (Toyoda et al.,2003). jmj overexpression by transgenesis represses cyclin D1 expression in the heart. cyclin D1 overexpression by transgenesis causes hyperproliferation in the cardiac myocytes, but analysis of cyclin D1 and jmj double mutant mice showed that the absence of cyclin D1 in jmj mutant embryos rescues the hyperproliferation. Therefore, cyclin D1 is a critical downstream gene of jmj for repression of cardiac myocyte proliferation. From these results, we hypothesized that jmj protein directly represses transcription of cyclin D1. In fact, Jmj protein binds to cyclin D1 promoter in vivo and represses cyclin D1 promoter activity (Toyoda et al.,2003). Lee's group also reported that jmj has repressor activity with respect to several promoters (Kim et al.,2003,2004,2005) and that their jmj mutant mice on a mixed background showed hyperproliferation of trabecular myocytes and enhanced expression of cyclin D1 (Jung et al.,2005).

As well as hyperproliferation, cardiac myocyte differentiation is affected in jmj mutant embryos (Takeuchi et al.,1999). The expression levels of cardiac myocytes markers such as myosin heavy chain, myosin light chain, and α-cardiac actin decrease drastically. It is important to determine whether and how hyperproliferation or enhanced expression of cyclin D1 results in abnormalities in differentiation. Proliferation and differentiation are closely related to each other in many developing cell types. These analyses of jmj mutant embryos would be helpful for understanding the molecular mechanisms regulating proliferation and differentiation.

Most likely, a jmj-cyclin D1 pathway represses proliferation and maintains differentiation in both trabecular and compact layers, because jmj expression starts when the levels of cell proliferation start to decrease in the compact layer as well as in the trabecular layer (Toyoda et al.,2003) and cyclin D1 overexpression affected proliferation and differentiation in both layers (unpublished data).

jmj mutant embryos exhibit other abnormalities in the cardiovascular system. Lee et al. reported that their jmj mutant embryos on a mixed background (C57BL/6 X 129/sv) exhibited double-outlet right ventricle (DORV) and ventricular septal defect (VSD; Lee et al.,2000). We also observed these abnormalities in our mutant mice on a BALB/c background. These abnormalities are common human congenital heart defects and have been observed in numerous knockout (KO) mice. Defects in various cell types and various signaling pathways could cause the abnormalities. In addition, secondary defects should be noted because one defect in the heart, which works as an essential pump, could easily cause other defects. To clarify the roles of jmj in the cardiovascular system at the late stages, it is important to examine what cell types are primarily affected and what signaling pathways are involved.

jmj Functions in Brain Development

As described in the Introduction section, jmj mutant embryos have abnormal grooves at the future midbrain–hindbrain boundary on the neural plate at E8–E8.5 (Takeuchi et al.,1995). Expression of the lacZ knocked into the jmj gene is detected at the posterior region of the groove, indicating that jmj is expressed at the corresponding region in normal embryos (Takeuchi et al.,1995). We observed that cyclin D1 expression was enhanced at the region (Fig. 4B,D). Moreover, neural epithelial cells showed a round shape and abnormal cell accumulation and formed a multilayer at the posterior region of the groove (Fig. 4E). The neural epithelial cells of normal embryos and cells at other regions of jmj mutant embryos showed an elongated shape and form a monolayer. These results suggest that jmj represses cell proliferation by repression of cyclin D1 expression at the future midbrain–hindbrain boundary as in cardiac myocytes and that repression of cell proliferation is required specifically at the boundary.

Figure 4.

Enhanced expression of cyclin D1 and abnormal structure in the neural epithelial cells of jmj mutant embryos. A–D: Frozen sagittal sections of wild-type (Wt, A, C) and jmj mutant embryos (Hm, B, D) at embryonic day 8.5 (E8.5) were analyzed by immunohistochemistry using an antibody against cyclin D1. C,D: High-magnification images of A and B, respectively. E: Ultrastructural analysis of neural epithelial cells of jmj mutant embryos at E8.5. “p” and “a” represent posterior and anterior positions, respectively. B: The arrow shows the position of the abnormal groove. D,E: Bars indicate the posterior region of the abnormal groove where jmj expression is lost and cyclin D1 expression is enhanced. The neural epithelial cells showed abnormal cell accumulation at the posterior region of the groove. Scale bar = 50 μm.

In fact, recent studies have shown that the cells in the boundary region proliferate less rapidly than the surrounding cells at E10.5 (Trokovic et al.,2005). Of interest, a cyclin-dependent kinase inhibitor, p21Cip1 is also expressed at the same region as jmj. Expression of p21Cip1 and jmj decreased largely or disappeared while expression of cyclin D1 as well as cyclin D2 is enhanced in Fgfr1 KO embryos at E9.5–E10.5 (Trokovic et al.,2005). These results suggest that Fgfr1 signals enhance or maintain the expression of jmj and p21Cip1 and repress expression of cyclin D1 and D2 at a distinct region in the midbrain–hindbrain boundary, resulting in slow proliferation at the region. Because the Fgfr1 mutant embryos lack isthmic constriction, slow proliferation would be necessary for development of the isthmic constriction. It is conceivable that jmj contributes to the mechanisms by repression of cyclin D1.

jmj Functions in Cell Proliferation

As described above, jmj negatively regulates proliferation of cardiac myocytes (Toyoda et al.,2003) and would repress the proliferation of neuroepithelial cells at the midbrain–hindbrain boundary (see above and Trokovic et al.,2005). Further evidence for a role for jmj in cell proliferation has been obtained in other cells as follows.

First of all, transfection of jmj cDNA was found to down-regulate cell proliferation in NIH3T3 and COS cells (Toyoda et al.,2000). The number of megakaryocyte lineage cells increased in the liver of jmj mutant embryos, and a delay of growth arrest in these cells was observed in colony formation assays (Motoyama et al.,1997; Kitajima et al.,2001).

Together with the expression patterns negatively correlated with cell proliferation, these results suggest a negative role for jmj in cell proliferation during development. However, all phenotypes of jmj mutant embryos cannot be explained by this role. For example, despite the impaired differentiation, the proliferation of hepatocytes was not affected (Anzai et al.,2003). In addition, a lack of cyclin D1 did not rescue all phenotypes of jmj mutant embryos. For example, neural tube defects were not rescued (Toyoda et al.,2003). Although it is possible that other cell cycle regulators are affected, these results suggest that jmj has other roles in addition to the negative control of cell proliferation. Jmj likely regulates the expression of several or many genes; however, cyclin D1 is the only downstream target gene verified in vivo so far (Toyoda et al.,2003). Therefore, we know a portion of the functions of jmj. Finding other target genes and related pathways as well as further analyses of mutant phenotypes are necessary.

The early lethality of jmj mutant embryos prevents us from analyzing jmj functions in many tissues after midgestation. For example, we do not know the significance of jmj expression in neurons or cardiac myocytes after birth. Therefore, analysis of conditional KO mice is required.

Finally, the study of molecular mechanisms in transcriptional repression by Jmj is important. Because Jmj conserves almost no residues binding to the cofactors 2-OG and Fe(II) (Fig. 2), Jmj itself might not have any 2-OG-Fe(II)–dependent dioxygenase activity, especially histone demethylase. However, our recent studies showed that Jmj binds to several proteins highly involved in chromatin regulation (unpublished data), suggesting that Jmj protein complex(es) regulate transcription through epigenetic modification even if Jmj protein does not have histone demethylase activity.



RBP2 is a member of the JARID1 subfamily (Fig. 1). It is ubiquitously expressed in adult tissues and has jmjN, jmjC, PHD, and ARID sequences. Among the over 100 pRB binding proteins, RBP1 and RBP2 were identified as the first cellular pRB binding proteins (Defeo-Jones et al.,1991). Although the function has remained unknown for a long time, roles in cell differentiation were recently proposed (Benevolenskaya et al.,2005). It is well known that pRB regulates cell differentiation in addition to cell proliferation. Benevolenskaya et al. suggested that RBP2 represses expression of the genes required for differentiation of various types of cells such as myeloid, bone, and muscle, and binding to pRB neutralizes the functions of free RBP2 and then promotes differentiation. They also suggested that the binding is involved in euchromatin maintenance. It is interesting to consider that pRB may act as a negative regulator of RBP2 in cell differentiation. Activity in transcriptional repression and involvement in euchromatin maintenance suggest that RBP2 or RBP2 protein complex probably regulates chromatin remodeling. It would be interesting to examine whether RBP2 itself has histone demethylation activity.


PTDSR has only a jmjC domain as an apparent domain (Fig. 1). The protein was initially identified as the receptor of phosphatidylserine, a specific marker present at the surface of apoptotic cells, and is involved in apoptotic cell phagocytosis (Fadok et al.,2000). However, PTDSR is localized in the nucleus (Cikala et al.,2004; Cui et al.,2004). Moreover, an antibody against phosphatidylserine receptor, used for original screening, still recognizes the receptor in mice lacking PTDSR (Bose et al.,2004). These results strongly suggest that PTDSR is not the phosphatidylserine receptor and, therefore, that the name should be changed.

Several studies using KO or knockdown animals (worm, zebrafish, and mouse) have shown that PTDSR is required for clearance of apoptotic cells (Li et al.,2003; Wang et al.,2003; Hong et al.,2004; Kunisaki et al.,2004). However, PTDSR was not essential for clearance of apoptotic cells in an analysis of another PTDSR KO mouse line (Bose et al.,2004). Apart from the function related to the clearance of apoptotic cells, PTDSR would be required for normal embryogenesis, at least in zebra fishes and mice (Li et al.,2003; Bose et al.,2004; Hong et al.,2004; Kunisaki et al.,2004; Schneider et al.,2004). PTDSR KO mice died at the perinatal stage and showed various abnormalities in many tissues such as the heart, brain, lung, kidney, eye, and hematopoietic system. However, these phenotypes are also different among KO mouse strains. For example, neuroepithelial cells in the retina showed enhanced proliferation in Flavell's KO mice (Li et al.,2003). On the other hand, 14.1% of Lengeling's KO mice lacked eyes unilaterally or bilaterally (anophthalmia), or the retinal development was temporally delayed and morphology of the inner granular layer was abnormal, despite normal external eye structure. The abnormal morphology observed in Flavell's KO mice does not appear to be seen in these mice (Bose et al.,2004).

Together with the phenotypes regarding clearance of apoptotic cells, it is unknown why the phenotypes are quite different among KO mouse strains. Although the possibility that it resulted from differences in the genetic backgrounds could not be excluded, whether these phenotypes of the KO mice resulted from the lack of the PTDSR should be substantiated by rescue experiments using transgenic mice or other methods. In addition, it is important to elucidate the molecular functions of PTDSR, not as a transmembrane receptor but as a nuclear protein.


Hairless is in cluster 3 and has a jmjC domain as the only apparent domain (Fig. 1). The gene was identified as the responsible gene for the mouse mutant hairless (Cachon-Gonzalez et al.,1994). The hairless mouse was first recognized in 1926 for its characteristic hair loss phenotype, in which initial hair growth is normal, but after shedding, the hair does not grow back (Brooke,1926), suggesting abnormalities in hair follicle regeneration. Mutations of the human hairless gene revealed congenital hair loss disorders. Some mutant alleles in both mice and humans show skin wrinkling and papular rash (Panteleyev et al.,1998). Hairless null KO mice show both hair and skin phenotypes (Zarach et al.,2004). Analysis of the KO mice revealed increased proliferation and changes of cell types in the epidermis, suggesting that Hairless is required for normal balance of cell proliferation and differentiation in the epidermis cells (Zarach et al.,2004). The rescue experiments using transgenic mice showed that expression of Hairless in progenitor keratinocytes is required for hair follicle regeneration (Beaudoin et al.,2005). Hairless represses expression of Wise, a modulator of Wnt signaling, coincident with the timing of follicle regeneration. From these studies, a model in which Hairless regulates the precise timing of Wnt signaling required for hair follicle regeneration was proposed (Beaudoin et al.,2005).

Hairless protein functions as a nuclear receptor corepressor and interacts with multiple nuclear receptors such as thyroid hormone receptor, retinoic acid receptor-related orphan receptor α, and vitamin D receptor (Potter et al.,2001; Moraitis et al.,2002; Hsieh et al.,2003). Hairless represses transcriptional activities in the context of these nuclear receptors (Potter et al.,2001; Moraitis et al.,2002; Hsieh et al.,2003). Of interest, Hairless as well as Jmj have transcriptional repression activity, although these two proteins might not have 2-OG-Fe(II)–dependent histone demethylase activity, because cofactor binding sites are not conserved (Fig. 2). Because Hairless is a transcriptional repressor, altered expression of genes would cause the phenotypes of Hairless mutants. The expression of several genes such as keratinocyte differentiation markers and wise, was up-regulated in Hairless KO mice (Zarach et al.,2004; Beaudoin et al.,2005). Hairless represses the promoter activity of wise (Beaudoin et al.,2005). These studies link the molecular functions of Hairless to the functions in skin and hair development.


The molecular functions of jmjC and jumonji family proteins will be analyzed extensively and intensively. However, it will also be important to link the molecular functions to biological events such as development. Several points or questions that require attention when conducting these studies are discussed below.

What Molecular Functions Do Each Jumonji Family Protein or the Protein Complex Have?

There is little doubt that the histone demethylase activities of many jumonji family proteins will be examined. However, it is possible that jumonji family proteins and the protein complexes have functions of not only histone demethylases but also other chromatin modifiers, or other enzymes that are not directly related to chromatin regulation.

One of the jumonji family proteins, HIF1AN/FIH, hydroxylates an asparagine residue in HIFα, and it is unknown whether HIF1AN/FIH is active in histone demethylation or other chromatin modification. Several jumonji family proteins do not have conserved residues at the predicted binding sites for the cofactors, 2-OG or Fe(II) (Trewick et al.,2005; Tsukada et al.,2006; Fig. 2). Therefore, these proteins would not have 2-OG-Fe(II)–dependent dioxygenase activity that would be required for histone demethylation by JHDM1 and hydroxylation by HIF1AN/FIH. Of interest, jumonji family proteins exist in bacteria that do not have histones. These facts suggest that jumonji family proteins can have various functions other than as histone demethylases. For example, it is exciting if some members have DNA modification activity, such as a DNA demethylase. As described above, bacterial AlkB protein, a 2-OG-Fe(II)–dependent dioxygenase, is involved in DNA repair by DNA demethylation (Falnes et al.,2002; Trewick et al.,2002), suggesting the possibility. Because DNA modification as well as histone modification are very important events for making epigenetic signatures, the possibility is very interesting.

In addition, forming complexes with other proteins would add other functions in histone or chromatin modification to jumonji family proteins (see the section entitled Functions of Jumonji Family Proteins in Transcription and Chromatin Regulationdures). Several large protein complexes would have multiple or sequential functions (Ogawa et al.,2002), similar to that of “a factory.” It is possible that jumonji family proteins are one of the members in the “factory.” In this case, it is important to examine what distinct functions the jumonji family proteins have in the complexes.

What Molecules or Proteins, and Moreover, Which Sites in the Molecules Are Targets of Jumonji Family Proteins?

If a certain jumonji family protein or the protein complex has histone modification activities, including histone demethylation, which residues and states of premodification (for example, mono-, or di-, or tri-methylation) in the histones are modified by the proteins should be examined, because we can infer specific functions from modified residues and states.

Expression of What Genes Jumonji Family Members Regulate and How Is Expression Regulated?

If a certain jumonji family protein is shown to have a function in chromatin regulation, the next step is to answer the questions. It is also important to ascertain whether the jumonji family proteins regulate limited genes in a part of the genome, or many genes in more extensive regions. It would be interesting to study whether jumonji family proteins are involved in the regulation, such as formation or maintenance of euchromatin or heterochromatin, or imprinting or X-chromosome inactivation. In fact, Epe1 is required for heterochromatin integrity (Ayoub et al.,2003).

What Biological Events In Vivo Including Development Are Linked to Molecular Functions of Jumonji Family Proteins?

Basic biological events can be analyzed in cell cultures, however, studies in organisms are necessary to analyze more higher or complicated functions in vivo. Especially, using mutant organisms such as KO mice is one of the most powerful methods, because we can analyze which cell behaviors such as cell differentiation, cell proliferation, cell migration or cell adhesion are involved, and in which tissues the proteins are required. That mutant mice of two jumonji family proteins, Jmj and PTDSR, show abnormalities in multiple tissues (see above) suggests that at least some of the jumonji family proteins are required for the regulation of many genes or that these proteins have common essential functions in the development of diverse tissues.

Of course, it is difficult to elucidate the links between molecular functions and biological events completely by analysis of mutant organisms only. Therefore, a good combination of studies at the molecular and organism levels would be necessary. We believe that such studies on the jumonji family will contribute significantly to our understanding of the mechanisms of chromatin regulation and development.


Most recently, five jumonji family members, JMJD1A/TSGA (renamed to JHDM2A in the following paper) and JMJD2A-D have been shown to be H3K9 demethylases (Whetstine et al.,2006; Yamane et al.,2006). JMJD2A and JMJD2C are also H3K36 demethylases (Whetstine et al.,2006). Importantly, JMJD1A/TSGA can demethylate mono- and di-methyl H3-K9, and JMJD2 family can demethylate di- and tri-methyl H3-K9/K36. These results showed that jumonji family members can demethylate residues in three methylation states of histones.


We thank Ms. Mizuyo Kojima and Ms. Kuniko Nakajima for their excellent technical assistance.