Biochemical Journal

Research article

Crystal structure of the catalytic core of Saccharomyces cerevesiae histone demethylase Rph1: insights into the substrate specificity and catalytic mechanism

Yuanyuan Chang, Jian Wu, Xia-Jing Tong, Jin-Qiu Zhou, Jianping Ding

Abstract

Saccharomyces cerevesiae Rph1 is a histone demethylase orthologous to human JMJD2A (Jumonji-domain-containing protein 2A) that can specifically demethylate tri- and di-methylated Lys36 of histone H3. c-Rph1, the catalytic core of Rph1, is responsible for the demethylase activity, which is essential for the transcription elongation of some actively transcribed genes. In the present work, we report the crystal structures of c-Rph1 in apo form and in complex with Ni2+ and α-KG [2-oxoglutarate (α-ketoglutarate)]. The structure of c-Rph1 is composed of a JmjN (Jumonji N) domain, a long β-hairpin, a mixed structural motif and a JmjC domain. The α-KG cofactor forms hydrogen-bonding interactions with the side chains of conserved residues, and the Ni2+ ion at the active site is chelated by conserved residues and the cofactor. Structural comparison of Rph1 with JMJD2A indicates that the substrate-binding cleft of Rph1 is formed with several structural elements of the JmjC domain, the long β-hairpin and the mixed structural motif; and the methylated Lys36 of H3 is recognized by several conserved residues of the JmjC domain. In vitro biochemical results show that mutations of the key residues at the catalytic centre and in the substrate-binding cleft abolish the demethylase activity. In vivo growth phenotype analyses also demonstrate that these residues are essential for its functional roles in transcription elongation. Taken together, our structural and biological data provide insights into the molecular basis of the histone demethylase activity and the substrate specificity of Rph1.

  • crystal structure
  • demethylation
  • H3K36 methylation
  • histone demethylase
  • histone modification
  • substrate specificity

INTRODUCTION

Histone methylation is an important post-translational modification that plays crucial roles in many important biological processes, including heterochromatin formation, X-chromosome inactivation, genomic imprinting and transcription regulation [1,2]. Histone methylation usually occurs on lysine residues of histone tails, including H3K4, H3K9, H3K27, H3K36, H3K79 and H4K20, which can be methylated by histone methyltransferases [3,4]. Different methylation sites and states can lead to different consequences. H3K4, H3K36 and H3K79 methylations are generally associated with active transcription, whereas H3K9, H3K27 and H4K20 methylations are associated with silenced transcription [2].

Budding yeast encodes three histone lysine methyltranferases, namely Set1, Set2 and Dot1 that methylate specifically H3K4, H3K36 and H3K79 respectively [58]. Methylations of these residues in budding yeast are tightly associated with the transcription process, and deposition of these modifications occurs mainly during the initiation and elongation stages of RNA Pol II (RNA polymerase II)-based transcription [5,911]. In particular, the methylations of H3K4 and H3K36 are associated with different phosphorylation states of the CTD (C-terminal domain) of the Rpb1 subunit of RNA Pol II [5,911]. It has been shown that the H3K4 methyltransferase Set1 can be recruited to the 5′ ends of genes when residue Ser5 of Rpb1-CTD is phosphorylated by Kin28, and the H3K36 methyltransferase Set2 can be recruited to the body of genes during elongation when residue Ser2 is phosphorylated by Ctk1 [5,1114]. The methylated H3K36 mark can be recognized by Eaf3 of the histone deacetylase complex Rpd3C(S) through its chromo domain, which recruits Rpd3C(S) to inhibit transcription elongation and cryptic initiation within ORFs (open reading frames) by RNA Pol II through histone deacetylation [1518]. This inhibition can be reversed by Bur1, a positive elongation factor that phosphorylates Ser2 of Rpb1-CTD and stimulates Ser2 phosphorylation by Ctk1 during transcription elongation [17,19].

Histone methylation can be removed by two classes of histone demethylases. One is the flavin-dependent amine oxidase LSD1 (lysine-specific demethylase 1), which is a specific demethylase for H3K9me2/1 (di-/mono-methylated H3K9) and H3K4me2/1 [2022]. The recently identified second class of histone demethylases possesses Fe2+- and α-KG [2-oxoglutarate (α-ketoglutarate)]-dependent oxygenase activity and is characterized by the presence of a JmjC (Jumonji C) domain [23,24]. The phylogenetic analysis of the JmjC-domain-containing proteins shows that they are evolutionarily conserved in eukaryotes ranging from yeast to humans [24]. Further domain alignment classifies these proteins into seven groups and shows that, in six of the groups, the proteins contain at least one domain in addition to the JmjC domain, which might contribute to the substrate specificities of the enzymes [24]. The JmjC-containing demethylases have displayed substrate diversity with respect to the methylation site and state of the lysine residue in the histone substrate [2328]. For example, mammalian JHDM1A (Jumanji-domain-containing histone demethylase 1A) can specifically demethylate H3K36me2 [23]. JMJD2A, JMJD2B and JMJD2C have dual specificity towards both H3K9me3/2 (tri-/di-methylated H3K9) and H3K36me3/2, whereas JMJD2D can only demethylate H3K9me3/2 [26,27]. As histone methylation is critical in the regulation of eukaryotic gene expression, it is not surprising that these enzymes play important roles in a variety of cellular processes, and have been found to be associated with various diseases, such as X-linked mental retardation [24,29].

In budding yeast, there are five JmjC-domain-containing proteins, namely Jhd1, Jhd2, Rph1, Gis1 and Ecm5 [30]. Jhd1 is a H3K36me2/1 demethylase which plays a role in keeping the fidelity of histone methylation states in transcriptional units [30,31]. Jhd2 has demethylase activity towards H3K4me3/2, which antagonizes the biological functions mediated by Set1 methylation and maintains the dynamic balance between the methylation and demethylation of H3K4 [32,33]. The demethylase activities of Gis1 and Ecm5 have not been determined yet. Rph1 is an orthologue of JMJD2A and has demethylase activity for H3K36me3/2 [30,34,35]. Rph1 can regulate the distribution and state of H3K36 methylation and enhance RNA Pol II occupancy in the actively transcribed regions by counteracting the methyltransferase activity of Set2 [34]. Surprisingly, Rph1 can also demethylate H3K9me3, although no H3K9 methylation exists in budding yeast, leading to the suggestion that budding yeast may have encoded an H3K9 methylation system during evolution and that Rph1 is a functional vestige of this modification system [35]. In addition, Rph1 contains two zinc fingers at the C-terminus that are not required for demethylase activity, but play critical roles in DNA-damage responses [36].

In the present paper, we report the crystal structures of the c-Rph1 (the catalytic core of Rph1) (residues 1–373) in apo form and in complex with Ni2+ and α-KG. Structural analysis and comparison with JMJD2A identify the key residues at the catalytic centre and in the substrate-binding cleft. The functional roles of these residues in the demethylation are verified by both in vitro enzymatic assays and in vivo growth phenotype analyses. These results provide insights into the molecular basis of the histone demethylase activity and the substrate specificity of Rph1.

MATERIALS AND METHODS

Antibodies

The sources of the antibodies used in the present study are as follows: anti-H3K36me1 antibody (Upstate), anti-H3K36me2 antibody (Upstate), anti-H3K36me3 antibody (Abcam), anti-H3 antibody (Abcam), goat anti-rabbit IgG antibody (Promega) and goat anti-mouse IgG antibody (Promega).

Expression and purification of the Rph1 catalytic core

The yeast genomic DNA corresponding to the catalytic core of Rph1 (residues 1–373) was amplified by PCR and cloned into the pET-28b–His6 expression plasmid (Novagen) using the NcoI and XhoI restriction sites. The plasmid was transformed into Escherichia coli strain BL21(DE3), and the bacterial cells were cultured in Luria–Bertani medium supplemented with 50 μg/ml kanamycin at 37 °C until the D600 reached 0.8. Protein expression was induced by adding 0.1 mM IPTG (isopropyl-β-D-thiogalactopyranoside) at 37 °C for approx. 4 h. The bacterial cells were harvested by centrifugation and resuspended in a lysis buffer (20 mM Tris/HCl, pH 8.0, 100 mM NaCl, 2 mM 2-mercaptoethanol, 0.1% NP40 and 1 mM PMSF), and then lysed by sonication on ice with a JY92-2D sonicator (Scientz) at power 400W and 50% pulse cycle for 1h. The cell debris was precipitated by centrifugation at 18000 g for 30 min, and the supernatant was collected for purification.

The target protein was purified with a Ni-NTA (Ni2+-nitrilotriacetate) agarose column (GE Healthcare) pre-equilibrated with binding buffer (20 mM Tris/HCl, pH 8.0, and 100 mM NaCl). The target protein was eluted with elution buffer (binding buffer supplemented with 300 mM imidazole). The elution sample was further purified by gel-filtration using a Superdex 16/60 (preparative grade) column (GE Healthcare) pre-equilibrated with binding buffer supplemented with 10 mM EDTA and 2 mM dithiothreitol. The target protein was collected and concentrated to approx. 25 mg/ml by ultrafiltration for structural and functional studies. The purified protein was of sufficient purity (>95%), as determined by SDS/PAGE (12% gel). Se-Met (selenomethionine)-substituted protein was expressed in the methionine auxotroph E. coli B834(DE3) (Novagen) and purified using the same method as for the native protein.

Constructs of the Rph1 mutants containing point mutations were generated using the QuikChange® Site-Directed Mutagenesis Kit (Stratagene) as described in the manufacturer's instructions and verified by DNA sequencing. Expression and purification of these mutants were the same as for the wild-type protein described above.

The yeast unmethylated, di- and tri-methylated H3K36 (residues 28–44) and H3K9 (residues 1–21) peptides were synthesized by HD Biosciences.

Crystallization and diffraction-data collection

The crystallization experiments for c-Rph1 were performed using the hanging-drop vapour-diffusion method at 4 °C. The crystals of both native and Se-Met c-Rph1 in apo form were grown in drops containing equal volumes of the protein solution (25 mg/ml) and the reservoir solution {0.1 M Mes, pH 6.0, 0.2 M ammonium sulfate and 7% PEG [poly(ethylene glycol)] 4000} to approximate dimensions of 0.2×0.2×2.0 mm3 in 20 days (Supplementary Figure S1A at http://www.BiochemJ.org/bj/433/bj4330295add.htm). The crystals of c-Rph1 in complex with Ni2+ and α-KG were prepared by soaking the apo-form crystals in the crystallization solution supplemented with 1 mM NiCl2 and 1.5 mM α-KG. The crystals belong to space group P64 with one c-Rph1 molecule in the asymmetric unit. All diffraction data were collected from flash-cooled crystals at 100 K at beamline BL-17U of Shanghai Synchrotron Radiation Facility, Shanghai, China. The SAD (single-wavelength anomalous dispersion) data were collected to 3.0 Å resolution (1 Å=0.1 nm). The native diffraction data of the apo form, which were collected to 2.5 Å resolution, were shown to be anisotropic and thus were anisotropically corrected using the Diffraction Anisotropy Server [37,38]. As indicated by the server, the diffraction data were truncated to 3.7 Å, 3.7 Å and 2.5 Å along the three reciprocal space directions a*, b*, and c* respectively. The diffraction data of the complex, which were also anisotropic, were collected to 2.2 Å resolution (Figure S1B) and were truncated to 3.3 Å, 3.3 Å and 2.2 Å along the three reciprocal space directions a*, b* and c* respectively (Figure S1C). Meanwhile, an anisotropic B-factor correction was applied to the diffraction data. All of the diffraction data were processed, integrated and scaled together with the HKL2000 suite [39]. The twin fraction of the diffraction data was tested using both the Scalepack2mtz program of the CCP4 package [40] and PHENIX software [41], but no twinning problem was detected. The statistics of the diffraction data are summarized in Table 1.

View this table:
Table 1 Crystallographic data and refinement statistics

N.a., not applicable.

Structure determination and refinement

The crystal structure of c-Rph1 was solved using the SAD method implemented in the program SOLVE [42]. c-Rph1 contains seven methionine residues (including the first methionine residue), and the SAD phases revealed five selenium sites in one asymmetric unit, suggesting that there is one c-Rph1 molecule in the asymmetric unit. The SAD phases were improved by statistical density modification, including solvent flattening and histogram matching using the program RESOLVE [43], increasing the overall figure of merit from 0.31 to 0.59 at 3.0 Å resolution. The resulting electron density map had very good quality and was used to build an initial model. RESOLVE automatically built 250 polyalanine residues out of 373 residues and successfully defined most of the secondary structural elements. Subsequent refinement and model building were performed with the PHENIX and Coot programs using the 2.5-Å-resolution native-diffraction data [41,44]. All of the other crystallographic manipulations were carried out with the CCP4 package [40]. The crystal structure of c-Rph1 in complex with Ni2+ and α-KG was solved by the molecular replacement method using the Phaser program, using the structure of the apo-form c-Rph1 as the search model followed by refinement and model building with PHENIX and Coot [41,44,45]. The statistics of the structure refinement and the quality of the structure models are summarized in Table 1.

SPR (surface plasmon resonance) binding analysis

The binding affinities of c-Rph1 with different histone H3 peptides were measured by SPR (surface plasmon resonance) experiments at 25 °C using a BIAcore 3000 instrument (GE Healthcare). The C-terminally biotinylated peptides of tri- and di-methylated H3K36 (residues 28–44) and H3K9 (residues 1–21), and unmethylated H3K36 (residues 28–44) and H3K9 (residues 1–21) were immobilized on a streptavidin-containing BIAcore SA chip. A channel without peptide was used as a negative control. c-Rph1 was used as the analyte for binding. The association was monitored over a period of 300 s, during which the protein in the running buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA and 0.002% surfactant P20) was passed over the immobilized peptide at a rate of 20 μl/min. The dissociation was monitored by the same buffer over the subsequent 600 s. The surface was regenerated with 5 mM NaOH. The Kd values were calculated by equilibrium analysis using the BIAevaluation 4.1 software. All experiments were performed at least three times.

In vitro demethylase activity assay

The demethylase activities of both wild-type and mutant c-Rph1 were assayed as follows: 20 μg of purified c-Rph1 protein was incubated with 2 μg of calf thymus type II-A histones (Sigma–Aldrich) in a reaction buffer consisting of 50 mM Tris/HCl, pH 7.5, 50 μM Fe(NH4)2(SO4)2, 1 mM α-KG and 2 mM ascorbate at 30 °C for 3 h. The demethylation reaction was stopped by adding an SDS loading buffer, and Western blot analysis was performed with the specific antibodies. All experiments were performed at least three times.

Growth phenotype analysis

The growth phenotype analyses of yeast strains with overexpression of wild-type and mutant Rph1 were performed as described in [34]. The yeast strain used in the present study was the BUR1 shuffle strain YSB787. To generate the pRS425 plasmid containing RPH1, the entire ORF, 500 bp of the upstream region and 500 bp of the downstream region were amplified by PCR using oligonucleotide primers that created terminal XhoI/NotI sites. The BUR1 shuffle strain was transformed with pRS425 plasmids (2 micron, TRP1) containing the full-length wild-type or mutant RPH1. The transformants were patched on synthetic complete medium lacking uracil (as a positive growth control) or synthetic complete medium containing 5-fluoro-orotic acid (to select against the BUR1/URA3 plasmid). The plates were incubated for 2–6 days. Spotting analyses were performed as previously described [17].

RESULTS AND DISCUSSION

Characterization of the histone demethylase activity of the Rph1 catalytic core

Previous biochemical results showed that Rph1 has demethylase activity towards H3K36me3/2, with slightly higher activity for H3K36me3, and that the demethylase activity of the Rph1 catalytic core is comparable with that of the full-length protein [30]. It was also shown that Rph1 also has demethylase activity towards H3K9me3, although there is no H3K9 methylation in budding yeast [35]. We first characterized the demethylase activity and the substrate specificity of c-Rph1. Our biochemical assay results confirm that c-Rph1 can demethylate H3K36me3/2, with slightly higher activity towards H3K36me3 (Figure 1A). We performed the SPR experiments to measure the binding affinities of c-Rph1 towards H3K36 and H3K9 peptides with different methylation states. The SPR results show that c-Rph1 binds to the tri- and di-methylated H3K36 peptides with a dissociation constant Kd of 0.52 μM and 0.57 μM respectively (Figures 1B and 1C, and Table 2), which are comparable to that of c-JMJD2A (catalytic core of JMJD2A) towards the H3K36me3 peptide (Kd=1.00 μM) [46]. In addition, c-Rph1 has slightly higher binding affinity to the H3K9me3 peptide (Kd=0.19 μM) than the H3K9me2 peptide (Kd=0.88 μM) (Supplementary Figures S2A and S2B at http://www.BiochemJ.org/bj/433/bj4330295add.htm, and Table 2), which are also comparable with that of c-JMJD2A towards the H3K9me3 peptide (Kd=1.00 μM) [46]. The binding affinities of c-Rph1 to the unmodified H3K36 and H3K9 peptides are relatively weaker (KD=0.94 μM and 1.12 μM respectively) (Figure 1D, Supplementary Figure S2C and Table 2).

Figure 1 Characterization of the histone demethylase activity of the Rph1 catalytic core

(A) Demethylase activity and substrate specificity of c-Rph1 based on in vitro demethylase assay. (BD) Characterization of binding of c-Rph1 with various H3K36 peptides using surface plasmon resonance: (B) H3K36me3 (28–44) peptide; (C) H3K36me2 (28–44) peptide; and (D) H3K36 (28–44) peptide.

View this table:
Table 2 Binding affinities of c-Rph1 with H3K36 and H3K9 peptides

Overall structure of c-Rph1

The crystal structure of c-Rph1 was solved using the SAD method at 3.0 Å resolution (Table 1). The diffraction data of both the apo form and cofactor-bound c-Rph1 were shown to be anisotropic, resulting in high Rmerge and low I/σ(I) values in the higher-resolution bins (Table 1) which are typical for the anisotropic diffraction data [38,47]. The original diffraction data were subjected to anisotropy correction using the Diffraction Anisotropy Server [37,38] and the resultant diffraction data were used for structure refinement (see the Materials and methods section). The structure of c-Rph1 in apo form was refined against the anisotropy-corrected 2.5-Å-resolution diffraction data, yielding an R factor of 20.6% and a free-R factor of 25.2% (Table 1). There is one c-Rph1 molecule in the asymmetric unit. The c-Rph1 structure is well defined except that three surface-exposed regions (residues 114–137, 206–219 and 343–373) are invisible in the electron density map. The structure of c-Rph1 in complex with Ni2+ and α-KG was refined against the anisotropy-corrected 2.2-Å-resolution diffraction data, yielding an R factor of 19.8% and a free-R factor of 23.6% (Table 1). Like other JmjC-domain-containing histone demethylases, Rph1 requires a Fe2+ ion for its enzymatic activity. Ni2+ mimics Fe2+ to bind at the active site, but cannot initiate the demethylation reaction. Structural comparison of c-Rph1 between the apo form and the complex [an RMSD (root mean square deviation) of 0.76 Å for 298 Cα atoms] indicates that the binding of the metal ion and α-KG does not cause evident conformational changes in the overall structure or at the active site.

Similar to human c-JMJD2A, the structure of c-Rph1 consists of a JmjN domain (residues 6–55), a long β-hairpin (residues 56–97), a mixed structural motif (residues 98–205) and a JmjC domain (residues 220–341) (Figures 2A and 2B). The JmjC domain is located at the centre of the structure, flanked by the JmjN domain on one side and the long β-hairpin and the mixed structural motif on the other. The JmjC domain of Rph1 is composed mainly of a jellyroll-like structure of two four-stranded β-sheets, a characteristic feature of the cupin fold that is widely found in many metalloenzymes [48,49], including Fe2+- and α-KG-dependent oxygenases and histone demethylases. In addition, it contains a long insertion between β9 and β11 which forms one β-strand (β10) and three α-helices (α9–α11) (Figure 2A). The JmjC domain of Rph1 comprises the catalytic active centre and part of the substrate-binding cleft (see the discussion later). The JmjN domain consisting of four short α-helices (α1–α4) and two parallel β-strands (β1 and β2) (Figure 2A) does not form part of the catalytic centre. Instead, it surrounds one of the four-stranded β-sheets of the JmjC domain and appears to stabilize the JmjC domain through extensive hydrogen bonds and hydrophobic interactions, in particular those between two anti-parallel β-strands (β2 and β12), which is in agreement with the biochemical results that deletion of the JmjN domain abolishes demethylase activity [35]. The other side of the JmjC domain is surrounded by the long β-hairpin and the mixed structural motif composed of several α-helices and loops (Figure 2A). The β-hairpin is located on the surface of the catalytic pocket and may be involved in the substrate binding. In addition, it is also involved in stabilizing the JmjC domain via connection of β4 with β10. Sequence alignment shows that compared with JMJD2A, Rph1 has one long and two short insertions in this region (Figure 2B). These insertions, some of which are disordered in this structure (residues 114–137 and 206–219), are located on the opposite side of the catalytic centre and appear not to be involved in the substrate binding and the catalytic reaction (Figure 2A).

Figure 2 Crystal structure of c-Rph1 and comparison with c-JMJD2A

(A) Overall structure of c-Rph1 in complex with Ni2+ and α-KG. c-Rph1 consists of a JmjN domain (residues 6–55, green), a long β-hairpin (residues 56–97, blue), a mixed structural motif (residues 98–205, purple) and a JmjC domain (residues 220–342, magenta). The bound Ni2+ is shown as a red sphere and α-KG is shown in a cyan ball-and-stick model. The disordered regions are indicated with dotted lines. Inset: interactions of Ni2+ and α-KG with the surrounding residues at the active site. (B) Sequence comparison of yeast c-Rph1 and human c-JMJD2A. Strictly conserved residues are highlighted in shaded red boxes and conserved residues in open red boxes. The secondary structure of c-Rph1 is placed above the alignment and that of c-JMJD2A is below the alignment. The critical residues involved in Fe chelation, α-KG interaction, H3K36me3 interaction and histone peptide binding are marked by magenta, cyan, green and black boxes, respectively. (C) Overall structure comparison between c-Rph1 (magenta) and c-JMJD2A (green; PDB code 2OS2). The bound Ni2+ and α-KG in c-Rph1 are shown as a red sphere and a cyan model respectively; those in c-JMJD2A are shown as a green sphere and a blue model respectively. The bound H3K36me3 peptide in the c-JMJD2A complex is shown as a yellow coiled model. (D) Electrostatic surface of c-Rph1 showing the substrate-binding cleft. The bound α-KG is shown as a cyan model. An H3K36me3 peptide with a U-shaped conformation as recognized by c-JMJD2A (yellow model) can be accommodated into the substrate-binding cleft of c-Rph1 very well. (EG) Comparison of the residues that are involved in binding to Ni2+ and α-KG (E), the trimethylated Lys36 (F) and the H3K36me3 peptide (residues H3K37–H3Y41) (G). The colour scheme is the same as in (C).

Structures of the catalytic active centre and the substrate-binding cleft

Like the other Fe2+- and α-KG-dependent histone demethylases, the active site of Rph1 is located in the centre of the jellyroll fold of the JmjC domain and is lined with three strictly conserved residues, His235, Glu237 and His323. The Ni2+ ion is chelated by these residues and also the C-1 carboxyl and C-2 oxo groups of α-KG (Figure 2A). In addition to co-ordinating the metal ion, α-KG forms three hydrogen bonds with the side chains of Tyr183, Asn245 and Lys253 (Figure 2A). The importance of the Ni2+-chelating residue His235 has been revealed by other studies [30,34,35]; however, the functional roles of the other residues have not yet been examined. Our biochemical assays show that mutations of any of these residues, either involved in Ni2+ chelation (H235A, E237A and H323A) or α-KG binding (Y183A, N245A and K253A), abolish the demethylase activity of c-Rph1 towards H3K36me3, confirming their functional roles in the demethylation reaction (Figure 3A).

Figure 3 In vitro and in vivo functional assays of wild-type and mutant Rph1

(A) In vitro demethylase activity assay of wild-type (wt) and mutant c-Rph1. Mutations of the residues that are involved in the binding of the metal ion, α-KG and the substrate peptide abolish the histone demethylase activity of c-Rph1. (B) Growth phenotype analysis of yeast strains with overexpression of wild-type and mutant full-length Rph1. Overexpression of wild-type Rph1 can rescue the poor growth phenotype of the bur1Δ strain. However, overexpression of mutant Rph1, containing point mutations of the residues that abolish the histone demethylase activity of c-Rph1 in the in vitro biochemical assay, cannot rescue the phenotype of the bur1Δ strain. SC–Ura, synthetic complete medium lacking uracil, SC+FOA, synthetic medium containing 5-fluro-orotic acid.

Although Rph1 and JMJD2A share low sequence identity (37%) (Figure 2B) and a different overall structure, with an RMSD of 5.5 Å for 264 Cα atoms, their JmjC domains are very similar, with an RMSD of 1.0 Å for 122 Cα atoms (Figure 2C). Both Rph1 and JMJD2A have dual substrate specificity towards H3K36me3/2 and H3K9me3. The substrate specificity and catalytic mechanism of c-JMJD2A (the catalytic core of JMJD2A) have been studied extensively [46,5053]. Thus structural comparison of c-Rph1 with c-JMJD2A can help us to predict the substrate-binding cleft of Rph1 and provide some insights into the molecular basis of the substrate specificity. The substrate-binding cleft of Rph1 is composed mainly of several structural elements (β6–β10, β13 and β14) of the JmjC domain, the long β-hairpin and the mixed structural motif. The trimethylated H3K36 peptide with a U-shaped conformation recognized by JMJD2A can be accommodated into the substrate-binding cleft of Rph1 without obvious steric clash and the trimethylated Lys36 can fit into the active site very well (Figure 2D). The residues involved in the binding of the metal ion and α-KG are strictly conserved and assume identical side-chain conformations between Rph1 and JMJD2A (Figure 2E). The methyl groups of the trimethylated Lys36 can be recognized by the hydroxy group of Ser335, the amine group of Asn337 and the phenol group of Tyr222, and also have interactions with Ni2+ and α-KG (Figure 2F). Mutations of these residues (Y222A, S335A and N337A) also abolish the demethylase activity of c-Rph1 (Figure 3A). In the structure of c-JMJD2A in complex with the H3K36 peptide, the N-terminal region of the substrate (residues Ala31–Val35) is recognized by parts of the mixed structural motif (residue Glu169) and the C-terminal domain (residues 311–316) of c-JMJD2A. The equivalent regions of c-Rph1, including residue Pro216 of the mixed structural motif and residues 358–363 of the C-terminal domain, are disordered, indicating high flexibilities of these regions that might be stabilized in the presence of the peptide substrate. Similar to JMJD2A, the C-terminal region of the substrate can be recognized by several conserved residues: Lys37 of the H3K36 peptide can form a hydrogen bond with Ser72 and have hydrophobic interactions with Ala185 and Asp186, and Arg40 of the H3K36 peptide can form a hydrogen bond with Asn89 (Figure 2G).

Although the majority of the residues forming the catalytic centre and the substrate-binding cleft assume similar conformations in c-Rph1 and c-JMJD2A, there are a few notable differences. In the structure of c-JMJD2A, the phenol of Tyr177 forms a hydrogen bond with the main chain of the trimethylated Lys36, whereas in the structure of c-Rph1, the equivalent residue Tyr224 is positioned further away from the trimethylated Lys36 (Figure 2F). In addition, the loop containing Lys288 and Met289 in c-Rph1 assumes a conformation different to that in JMJD2A (Figure 2G). In the structure of c-JMJD2A, the side chain of Lys241 (equivalent to Lys288 in c-Rph1) is orientated towards the active site and has interactions with residues Lys36, Pro38 and His39 of the H3K36 peptide [51]. In the structure of c-Rph1, Lys288 points its side chain away from the catalytic centre; instead, the side chain of Met289 is orientated towards the active site (Figure 2G). Biochemical assays show that mutations of Tyr224 and Lys288 to alanine residues impair the demethylase activity of c-Rph1 (Figure 3A), indicating that these residues are also involved in the demethylation reaction. It is possible that these conformational differences are due to the absence of the peptide substrate, and Tyr224 and Lys288 of Rph1 might assume similar conformations as these of JMJD2A in the presence of the substrate. It is also possible that the recognition and binding of the substrate might be slightly different between Rph1 and JMJD2A, accounting for the slightly different demethylase activity and substrate specificity of the two enzymes. The crystal structure of c-Rph1 in complex with the peptide substrate would provide the precise molecular basis of the substrate recognition by Rph1.

Previous biological data showed that depletion of the positive elongation factor Bur1 causes growth deficiency or death of the yeast cells, but that overexpression of Rph1 can circumvent a requirement for Bur1 [34]. To further verify the functional roles of the key residues at the catalytic centre and in the substrate-binding cleft in the demethylation of Rph1 in vivo, we performed the growth phenotype analyses of wild-type and mutant Rph1 on the growth of the yeast cells. As expected, overexpression of wild-type Rph1 can rescue the poor growth phenotype of the bur1Δ strain. However, overexpression of mutant Rph1 containing point mutations of the residues at the catalytic centre or in the substrate-binding cleft that can abolish the histone demethylase activity of Rph1 in vitro are unable to rescue the phenotype of the bur1Δ strain. These results further confirm that the histone demethylase activity of Rph1 is essential for bypassing the requirement for Bur1 in transcription elongation and that these residues play important roles in the demethylation reaction.

AUTHOR CONTRIBUTION

Yuanyuan Chang carried out the structural and biochemical studies and drafted the manuscript. Jian Wu participated in the structural studies. Xia-Jing Tong and Jin-Qiu Zhou participated in the in vivo functional studies. Jianping Ding conceived the study, participated in the experimental design and co-ordination, and wrote the manuscript.

FUNDING

This work was supported by grants from the Ministry of Science and Technology of China [grant numbers 2007CB914302 and 2011CB966301], the National Natural Science Foundation of China [grant numbers 30730028 and 30623002], the Chinese Academy of Sciences [grant numbers KSCX2-YW-R-107 and SIBS2008002], and the Science and Technology Commission of Shanghai Municipality [grant number 10JC1416500].

Acknowledgments

We thank the staff members at the Shanghai Synchrotron Radiation Facility (SSRF), Shanghai, China for technical support in diffraction data collection, and other members of our group for helpful discussions.

Footnotes

  • The structural co-ordinates of the Rph1 catalytic core alone and in complex with Ni2+ and α-KG will appear in the PDB under accession codes 3OPW and 3OPT respectively.

Abbreviations: c-JMJD2A, catalytic core of JMJD2A; c-Rph1, the catalytic core of Rph1; CTD, C-terminal domain; α-KG, 2-oxoglutarate (α-ketoglutarate); me1, monomethylated; Jmj, Jumonji; me2, dimethylated; me3, trimethylated; ORF, open reading frame; RNA Pol II, RNA polymerase II; RMSD, root mean square deviation; SAD, single-wavelength anomalous dispersion; Se-Met, selenomethionine; SPR, surface plasmon resonance

References

View Abstract