Lysyl 5-Hydroxylation, a Novel Histone Modification, by Jumonji Domain Containing 6 (JMJD6)*

Background: JMJD6 hydroxylates U2AF65, but its role in histone modification has been obscure. Results: Our analysis of histones purified from JMJD6 knock-out mouse embryos reveals that JMJD6 hydroxylates histone lysyl residues. Conclusion: JMJD6 mediates histone lysyl 5-hydroxylation, which is a novel histone modification. Significance: Our study identifies a new function for Jumonji family proteins in epigenetic modification of histones. JMJD6 is reported to hydroxylate lysyl residues of a splicing factor, U2AF65. In this study, we found that JMJD6 hydroxylates histone lysyl residues. In vitro experiments showed that JMJD6 has a binding affinity to histone proteins and hydroxylates multiple lysyl residues of histone H3 and H4 tails. Using JMJD6 knock-out mouse embryos, we revealed that JMJD6 hydroxylates lysyl residues of histones H2A/H2B and H3/H4 in vivo by amino acid composition analysis. 5-Hydroxylysine was detected at the highest level in histones purified from murine testis, which expressed JMJD6 at a significantly high level among various tissues examined, and JMJD6 overexpression increased the amount of 5-hydroxylysine in histones in human embryonic kidney 293 cells. These results indicate that histones are additional substrates of JMJD6 in vivo. Because 5-hydroxylation of lysyl residues inhibited N-acetylation and N-methylation by an acetyltransferase and a methyltransferase, respectively, in vitro, histone 5-hydroxylation may have important roles in epigenetic regulation of gene transcription or chromosomal rearrangement.

JMJD6 is reported to hydroxylate lysyl residues of a splicing factor, U2AF65. In this study, we found that JMJD6 hydroxylates histone lysyl residues. In vitro experiments showed that JMJD6 has a binding affinity to histone proteins and hydroxylates multiple lysyl residues of histone H3 and H4 tails. Using JMJD6 knock-out mouse embryos, we revealed that JMJD6 hydroxylates lysyl residues of histones H2A/H2B and H3/H4 in vivo by amino acid composition analysis. 5-Hydroxylysine was detected at the highest level in histones purified from murine testis, which expressed JMJD6 at a significantly high level among various tissues examined, and JMJD6 overexpression increased the amount of 5-hydroxylysine in histones in human embryonic kidney 293 cells. These results indicate that histones are additional substrates of JMJD6 in vivo. Because 5-hydroxylation of lysyl residues inhibited N-acetylation and N-methylation by an acetyltransferase and a methyltransferase, respectively, in vitro, histone 5-hydroxylation may have important roles in epigenetic regulation of gene transcription or chromosomal rearrangement.
Jumonji domain containing 6 (JMJD6), 2 which possesses high binding affinity to single-stranded RNA, is reported to hydroxylate lysyl residues of an RNA splicing factor, U2AF65 (1,2). JMJD6 contains a JmjC domain that catalyzes lysyl hydroxylation of proteins in the presence of 2-oxoglutarate, Fe(II), and ascorbate. Proteins belonging to the JmjC family are classified into 2-oxoglutarate oxygenases (3). Among the known 2-oxoglutarate oxygenases, PLOD3 (procollagen-lysine, 2-oxoglutarate 5-dioxygenase 3) mediates hydroxylation of unmodified lysyl residues, yielding 5-hydroxylysine (4). Most JmjC family members catalyze hydroxylation of N-methyl groups at the ⑀-amino group of lysyl residues and generate hydroxymethyl groups, which are immediately processed to formaldehyde molecules, resulting in demethylation of methylated lysyl residues (5). However, JMJD6 does not add a hydroxyl group to the N-methyl group but adds it to one of the backbone carbons in a lysyl side chain and generates a stable 5-hydroxylysine (1). JMJD6 knock-out mice exhibited severe anemia, growth retardation, and a delay in terminal differentiation of the kidney, intestine, liver, and lung during embryogenesis, resulting in perinatal lethality (6,7).
In this study, we first identified JMJD6 as a novel UHRF1 (ubiquitin-like with PHD and RING finger domains 1) interacting protein. UHRF1 has important roles in transferring DNA methylation status and recognizes histone modification status (8). Therefore, we thought that JMJD6 might hydroxylate histone molecules through interaction with UHRF1. Using JMJD6 knock-out mice, we revealed that JMJD6 hydroxylates histone lysyl residues and generates 5-hydroxylysine in vivo. 5-Hydroxylation is a novel histone lysyl modification. Because it interfered with N-acetylation and N-methylation by an acetyltransferase and a methyltransferase, respectively, the modification may regulate transcription through these interactions with other histone modifications.

EXPERIMENTAL PROCEDURES
JMJD6 Wild-type and Knock-out Mice-Details of the JMJD6 knock-out mice were described elsewhere (6). C57BL/6 mice were used as wild-type control mice. JMJD6 knock-out embryonic day 14.5 (E14.5) embryos were obtained by crossing heterozygous JMJD6 mutant mice.
In Vitro Hydroxylation Assay-To perform the enzyme assay, GST-JMJD6 was prepared as described above. Extracted GST-JMJD6 was concentrated using a 50 K column (Millipore), and its buffer was replaced with 50 mM Tris-HCl (pH 7.5) by dialysis using EasySep (TOMY, Tokyo, JAPAN). Purity of GST-JMJD6 was assessed by Coomassie Brilliant Blue staining. The enzyme assay was performed in 50 mM Tris-HCl (pH 7.5) buffer containing 500 M ␣-ketoglutarate, 100 M L -ascorbate, 100 M Fe(NH 4 ) 2 SO 4 , 10 M GST-JMJD6, and 20 M histone peptides.
Protein purification and the enzyme assay were performed on the same day to avoid reduction of enzymatic activity of JMJD6.
Purification of Histones and Detection of 5-Hydroxylysine by Amino Acid Composition Analysis-Histones H2A/H2B and H3/H4 were separately purified from tissues or culturing cells using a histone purification kit (Active Motif, Carlsbad, CA) according to the manufacturer's instructions. The extracted histones were separated by SDS-PAGE, transferred to a membrane (Immobilon-P SQ , Millipore), and stained by Coomassie Brilliant Blue. The transferred histones were used for amino acid composition analysis to detect 5-hydroxylysine.
the presence of acetyl-coenzyme A (CoA, Sigma-Aldrich) at 37°C. The reaction was stopped by adding 100 l of quenching buffer (3.2 M guanidinium HCl, 100 mM Na 2 HPO 4 /NaH 2 PO 4 (pH6.8)) at the indicated times. Following this, 50 l of 2 mM 5,5Ј-dithiobis-2-nitrobenzoic acid (Sigma-Aldrich) in 100 mM Na 2 HPO 4 /NaH 2 PO 4 (pH 6.8) was added, and absorbance at 405 nm was read by a spectrophotometer (ARVO MX/Light 1420 Multilabel/Luminescence Counter, PerkinElmer Life Sciences). The transfer of an acetyl group from an acetyl-CoA to the ⑀-amino group of lysine residues was quantified by measurement of the thiol group of CoA. A standard curve was generated using ␤2-mercaptoethanol.
In Vitro Histone Methyltransferase Assay-The in vitro histone methyltransferase assay was performed as described previously (10), except for slight modifications. In brief, a fixed amount of purified baculovirus-produced recombinant SMYD3 (1 M) was incubated with indicated histone peptides, which were also used for the in vitro HAT assay, and 1 Ci of S-adenosyl-L-methionine (AdoMet; GE Healthcare) as the methyl donor in a mixture of 60 l of methylase activity buffer (50 mM Tris-HCl (pH 8.5), 100 mM NaCl, 10 mM dithiothreitol) at 30°C. The incorporated 3 H-labeled methyl groups in the substrates were measured by a scintillation counter after filter binding (units, cpm). The concentration

JMJD6 Effectively Hydroxylates Histone Lysyl Residues in
Vitro-During screening of UHRF1-interacting proteins, we identified JMJD6 as a novel binding partner of UHRF1 (data not shown). Because UHRF1 recognizes hemimethylated DNA and histone modifications, we assumed that JMJD6 might be recruited by UHRF1 to nucleosomes and modify histone lysyl residues. In vitro experiments showed that recombinant GST-JMJD6 possessed the ability to bind to histone H3 1-20 tail and histone H4 (Fig. 1, A and B) and hydroxylate multiple lysyl residues in the N-terminal tails of histone H3 1-20 and H4 1-30 , which was detected as of 16, 32, or 48 Da shifts by MS analysis (Fig. 1, C and D); subsequent MS/MS analysis revealed that JMJD6 mediates monohydroxylation of lysyl residues. As indicated by Webby et al. (1), JMJD6 preferentially hydroxylated lysyl residues in the basic peptides, and no apparent sequence preference was observed in vitro (data not shown).
Next, we established a sensitive hydroxylysine detection method based on amino acid composition analysis as an alternative to the MS-based method. For amino acid composition analysis, we briefly hydrolyzed peptides or proteins with HCl and separated each amino acid residue by reversed phase HPLC to detect 5-hydroxylysine. To evaluate this method, we first performed reversed phase HPLC using simplicial synthetic 2S,5R-hydroxylysine and synthetic racemic mixture of 5-hydroxylysine composed of 2S,5S (SS)-, 2R,5R (RR)-, 2R,5S (RS)-, and 2S,5R (SR)-stereoisomers ( Fig. 2A). We detected two peaks corresponding to SS/RR-and RS/SR-hydroxylysine by analyzing these synthetic 5-hydroxylysines without HCl treatment ( Fig. 2A). After HCl treatment of these synthetic 5-hydroxylysines, another peak was appeared ( Fig. 2A, arrow). This peak possibly corresponds to a lactone derivative, 3-amino-6-(aminomethyl)oxan-2-one, generated by dehydration condensation between C5 hydroxyl group and carboxyl group, which is described in a previous report (11). Next, we evaluated the method using unmodified H4 1-23 peptides and 5-hydroxylysine containing H4 1-23 peptides in which all the lysines at positions 5, 8, 12, and 20 were substituted with 5-hydroxylysines. After hydrolysis of these peptides, we detected two peaks corresponding to SS/RR-and RS/SR-hydroxylysine only in the 5-hydroxylysine containing peptides but not in the unmodified peptides (Fig. 2, B and C). We also detected the peak of the possible lactone derivative in the 5-hydroxylysine containing peptides by reversed phase HPLC performed in the same day of hydrolysis, but the peak disappeared in the next day of hydrolysis, indicating that the derivative is unstable. Because quantification of the derivative is technically difficult, we only quantified SS/RR-and RS/SR-hydroxylysine.
Using this method, we analyzed H3 21-40 and H4 1-23 peptides treated with or without recombinant GST-JMJD6. First, we confirmed hydroxylation of the peptides by GST-JMJD6 by MS analysis (Fig. 3A). Then, the peptides were separated from the enzyme reaction mixture, by reversed phase HPLC. The separated peptides were treated with HCl, and each amino acid residue was separated by reversed phase HPLC (Fig. 3, B and C). Comparison of the chromatograph between amino acids derived from the JMJD6-treated and -untreated peptides identified two additional peaks in the peptides treated with JMJD6, which are matched with the standard synthetic 5-hydroxylysine (Fig. 3, B and C).

JMJD6 Hydroxylates Histone Lysyl Residues in Vivo-To
investigate histone lysyl hydroxylation in vivo, we performed the amino acid composition analysis for analyzing a mixture of histone H2A/H2B and a mixture of histone H3/H4 proteins isolated from two JMJD6 wild-type and two JMJD6 knock-out whole embryos at E14.5 (Fig. 4A). JMJD6 knock-out was confirmed by qRT-PCR and Western blotting (Fig. 4, B and C). The results showed that 0.097 and 0.080% of total lysyl residues in histone H2A/H2B and 0.094 and 0.046% of those in histone H3/H4 were 5-hydroxylated in each of the two JMJD6 wild-type mice (Fig. 4, D and E), whereas 0.004 and 0.011% of total lysyl residues in histone H2A/H2B and 0.000 and 0.000% of those in histone H3/H4 were 5-hydroxylated in each of the two JMJD6 knock-out mice (Fig. 4, D and E), indicating that JMJD6 hydroxylates histone lysyl residues in vivo.

5-Hydroxylation Prevents N-Acetylation and N-Methylation of Histone Lysyl Residues in
Vitro-Because lysyl residues in histone tails are often subjected to N-acetylation and N-methylation, we examined whether 5-hydroxylation of lysyl residues affects modifications at the ⑀-amino groups. First, we examined the effect of lysyl 5-hydroxylation on histone H4 N-acetylation by p300, which catalyzes N-acetylation of the ⑀-amino group of lysyl residues, including histone H4K5 and H4K8, through its HAT domain (12). Kinetic analysis using the unmodified and the 5-hydroxylysine containing H4 1-23 peptides in which all the lysines were substituted with 5-hydroxylysines as substrates revealed that 5-hydroxylation largely interfered with the HAT activity of p300 in vitro (Fig. 6, A-C). Lineweaver-Burk plot analysis was performed to calculate the maximum velocity (V max ) and Michaelis constant (K m ) values (Table 1; Fig. 6C, R 2 ϭ 0.9981 and 0.9902). V max of the reactions in which p300 acetylated the 5-hydroxylysine-containing peptides (H4 1-23 OH) was 5-fold less than that of the control peptides (6.95 Ϯ 0.45 and 35.34 Ϯ 1.65 M/min, respectively), whereas K m of the two reactions was quite similar (204.51 Ϯ 10.12 and 200.37 Ϯ 14.32 M, respectively), indicating that 5-hydroxylation does not inhibit binding of lysyl residues to p300 but reduces the catalytic efficiency.

5-Hydroxylation of lysyl residue impairs N-acetylation and N-methylation in vitro.
A-C, the in vitro colorimetric HAT assay was performed using a fixed amount of p300 (0.44 M) and control H4 1-23 peptides (•) or 5-hydroxylysine-containing peptides (H4 1-23 OH, f). BSA was used as a negative control (ࡗ, OE). After the reactions, absorbance (405 nm) of the coproduct (CoA) was measured. A, reactions were terminated at the indicated time points, and the concentration of CoA was calculated on the basis of a standard curve that was generated from ␤2-mercaptoethanol. B and E, substrate concentration-velocity (s-v) plot. C and F, Lineweaver-Burk plot. The vertical axis is 1/velocity [v], and the horizontal axis is 1/substrate concentration [S]. D-F, the in vitro histone methyltransferase assay was performed using a fixed amount of SMYD3 (1 M) and control H4 1-23 peptides (•) or 5-hydroxylysine-containing peptides (H4 1-23 OH, f). AdoMet was used as a methyl donor. After the reactions, radioactivity (cpm) of the 3 H-methylated substrates was measured. The concentration of incorporated 3 H-methyl groups (nM) was calculated based on the basis of radioactivity (1 cpm was 0.02563 nM in the reaction). D, reactions were performed with fixed amounts of the peptides (141 M) and terminated at the indicated time points.

TABLE 1 Effect of 5-hydroxylation on N-acetylation of ⑀-amino group of lysyl residues
Lineweaver-Burk plots were used for estimation of the kinetic constants, V max , and K m . R 2 is the determination coefficient (see Fig. 6C).

TABLE 2 Effect of 5-hydroxylation on N-methylation of ⑀-amino group of lysyl residues
We also examined the effect of lysyl 5-hydroxylation on the histone methyltransferase activity of SMYD3, which catalyzes lysyl N-methylation of histone H3 (13) and also H4 (data not shown) through its SET (su(var) 3-9 enhancer-of-zeste trithorax) domain by the histone methyltransferase assay. The results showed that 5-hydroxylation at lysyl residues almost completely inhibited N-methylation catalyzed by SMYD3 (Table 2 and Fig. 6, D-F); V max values of the reactions with the control peptides (H4 1-23 ) and the 5-hydroxylysine-containing peptides (H4 1-23 OH) as substrates were 10.90 Ϯ 0.92 and 0.48 Ϯ 0.26 nM/min, respectively. Similar to the HAT assay, K m values of the two reactions were ϳ80.63 Ϯ 16.26 and 75.26 Ϯ 9.60 M, respectively.
Subsequently, we performed reciprocal experiments using H4 1-23 or H4 9 -23 peptides, in which all the lysines are either unmodified, N-acetylated, or N-monomethylated. JMJD6 effectively hydroxylated the control peptides (Fig. 7, A and B, left   panels); however, N-acetylation and N-monomethylation at the ⑀-amino group of the lysines completely blocked 5-hydroxylation by JMJD6 (Fig. 7, A and B, right panels).

DISCUSSION
We found a novel histone modification, 5-hydroxylation, by JMJD6. JMJD6 reportedly hydroxylates a splicing factor, U2AF65 (1). That study and another report (1,14) stated that evidence of histone lysyl hydroxylation was not found by MSbased analysis in vivo. In the present study, we developed an alternative method, amino acid composition analysis, to detect 5-hydroxylation of histone lysyl residues. As reported previously, we have not detected clear evidence of 5-hydroxylation of histone lysyl residues by MS-based analysis. We think that there are several causes for this. 1) The amount of 5-hydroxylysine is too small to detect by MS-based analysis. 2) Artificial methionine oxidation during preparation of samples for MS analysis FIGURE 7. N-Acetylation and N-methylation of lysyl residues impairs 5-hydroxylation by JMJD6 in vitro. A, the in vitro hydroxylation assay was performed using GST-JMJD6 (10 M) and 85 M of control H4 1-23 peptides or N-acetyl-lysine-containing peptides. B, the in vitro hydroxylation assay was performed using GST-JMJD6 (10 M) and 85 M control H4 1-23 peptides or N-monomethyl-lysine-containing peptides. BSA was used as a negative control. 5-Hydroxylation by JMJD6 was detected by MS analysis. makes detection of 5-hydroxylysine difficult. 3) 5-Hydroxylysine could be an intermediate form as it is in collagen, and a further unknown modification(s) such as glycosylation could be added; the final product of collagen is glucosylgalactosyl hydroxylysine (4). The collagen hydroxylase, PLOD3, possesses galactosyltransferase and glucosyltransferase activities. Unlike PLOD3, JMJD6 does not appear to possess any other enzymatic activities by domain search; therefore, it is difficult to assume possible further modification(s) by its protein structure. We may have been able to detect 5-hydroxylation in histone lysyl residues by amino acid composition analysis but not by the MS-based analysis because many modifications such as glycosylation or galactosylation could be removed during the hydrolysis process of amino acid composition analysis. By this analysis, we detected both SS/RR-and SR/RS-hydroxylysine in JMJD6-treated histone peptides and also in JMJD6 wild-type E14.5 embryos, ES cells, and the Dox-inducible JMJD6 stable HEK293 cells. Because the relationship between RR and RS and also between SS and RS is a diastereomer, we were able to distinguish them. However, because a relationship between SS and RR, and also between RS and SR is an enantiomer, we were not able to separate them by this method. Despite this, these two peaks are most likely SS-and RS-hydroxylysine because JMJD6 is reported to generate SS-hydroxylysine (11). The RS-hydroxylysine could be generated from SS-hydroxylysine through the lactone derivative, 3-amino-6-(aminomethyl)oxan-2-one, which is unstable and difficult to be quantified. Because of this difficulty, we only quantified SS/RR-and RS/SR-hydroxylysine in this report. Therefore, actual quantity of 5-hydroxylysine in the samples examined here could be a little higher.
Because we detected 5-hydroxylysines in the UHRF1 KO ES cells (data not shown), UHRF1 is not required for 5-hydroxylation of histone lysyl residues by JMJD6. Therefore, biological significance of the interaction between UHRF1 and JMJD6 remains unclear. Further analysis is also required to determine the biological significance of 5-hydroxylation of histone lysyl residues. In vitro experiments suggest that 5-hydroxylation can inhibit N-acetylation and N-methylation by p300 and SMYD3. The active site structure of the p300 and general control of amino acid synthesis 5 (GCN5) HAT domains showed that the side chain of the 5-hydroxylysine can invade the catalytic pocket; however, the 5-hydroxyl group may disturb active form formation of the substrate (Fig. 8, A and B). The catalytic site of SET domains of Neurospora crassa Dim-5 and human SETD7, which are structurally similar to SMYD3 (15), suggested that the side chain of 5-hydroxylysine can invade the catalytic pocket; however, the 5-hydroxyl group may disturb an active form formation of the substrate (16,17) similar to that in HAT domains (Fig. 8, C and D). Therefore, 5-hydroxylation could be important in the context of the histone code. It is known that histones H2A and H2B move more dynamically between the nucleosome and nucleoplasm. 5-Hydroxylation of these histones may have some effects for the movement because the modification was detected more in histones H2A/H2B than in histones H3/H4. The expression pattern of JMJD6 is also interesting. JMJD6 may play important role(s) in the testis, such as a role in histone-protamine exchange. We believe that our present finding provides a novel insight into epigenetic regulations of gene transcription and/or chromosomal rearrangement.