Structural and Functional Analysis of JMJD2D Reveals Molecular Basis for Site-Specific Demethylation among JMJD2 Demethylases

Structure

Article Structural and Functional Analysis of JMJD2D Reveals Molecular Basis for Site-Specific Demethylation among JMJD2 Demethylases Swathi Krishnan1 and Raymond C. Trievel1,* 1Department of Biological Chemistry, 1150 West Medical Center Drive, 5301 Medical Science Research Building III, University of Michigan, Ann Arbor, MI 48109, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.str.2012.10.018

SUMMARY

JMJD2 lysine demethylases (KDMs) participate in diverse genomic processes. Most JMJD2 homologs display dual selectivity toward H3K9me3 and H3K36me3, with the exception of JMJD2D, which is specific for H3K9me3. Here, we report the crystal structures of the JMJD2D,2-oxoglutarate,H3K9me3 ternary complex and JMJD2D apoenzyme. Utilizing structural alignments with JMJD2A, molecular docking, and kinetic analysis with an array of histone peptide substrates, we elucidate the specific signatures that permit efficient recognition of H3K9me3 by JMJD2A and JMJD2D, and the residues in JMJD2D that occlude H3K36me3 demethylation. Surprisingly, these results reveal that JMJD2A and JMJD2D exhibit subtle yet important differences in H3K9me3 recognition, despite the overall similarity in the substrate-binding conformation. Further, we show that H3T11 phosphorylation abrogates demethylation by JMJD2 KDMs. Together, these studies reveal the molecular basis for JMJD2 site specificity and provide a framework for structure-based design of selective inhibitors of JMJD2 KDMs implicated in disease. INTRODUCTION Protein lysine methylation is a dynamic posttranslational modification that is regulated through the concerted activities of lysine methyltransferases and lysine demethylases (KDMs). Presently, there are two known families of KDMs: the flavin-dependent LSD1 family, and the larger and more diverse Jumonji C (JmjC) family (Cloos et al., 2008; Mosammaparast and Shi, 2010). The JmjC KDMs are Fe(II)-dependent hydroxylases that catalyze lysine demethylation using the cosubstrates 2-oxoglutarate (2-OG) and molecular oxygen. To date, several subfamilies of JmjC KDMs have been identified and characterized, including JHDM1 (KDM2), JHDM2 (KDM3), JMJD2 (KDM4), JARID1, (KDM5), UTX/JMJD3 (KDM6), and PHF8/KIAA1718 (KDM7). These KDMs display distinct methylation site and state specificities for lysine residues in histones and in certain nonhistone

proteins (Allis et al., 2007; Cloos et al., 2008; Krishnan et al., 2011; Mosammaparast and Shi, 2010; Nottke et al., 2009; Shi, 2007; Shi and Whetstine, 2007; Upadhyay et al., 2011). Members of these families have been implicated in regulating diverse genomic processes, such as transcriptional activation, gene silencing, cell-cycle progression, heterochromatin maintenance, X chromosome inactivation, and development (Cloos et al., 2008; Mosammaparast and Shi, 2010; Nottke et al., 2009; Shi, 2007). The functional roles of the JmjC KDMs vary according to subfamily and are believed to be a consequence of their different histone methyllysine substrate specificities. The JMJD2 family of KDMs is conserved from budding yeast through mammals (Cloos et al., 2006; Fodor et al., 2006; Katoh and Katoh, 2004; Kim and Buratowski, 2007; Klose et al., 2006, 2007; Tu et al., 2007; Whetstine et al., 2006). Humans possess four JMJD2 homologs denoted JMJD2A, JMJD2B, JMJD2C, and JMJD2D. These enzymes have been implicated in transcriptional regulation, cell-cycle progression, nuclear hormone signaling, embryonic stem cell self-renewal, and development (Black et al., 2010; Iwamori et al., 2011; Lin et al., 2008; Loh et al., 2007; Lorbeck et al., 2010; Mosammaparast and Shi, 2010; Nottke et al., 2009; Shi, 2007; Strobl-Mazzulla et al., 2010). In addition, overexpression of several JMJD2 homologs has been linked to cancer. For example, JMJD2B has been implicated in breast, gastric, and colon cancers (Fodor et al., 2006; Fu et al., 2012; Kawazu et al., 2011; Li et al., 2011b; Shi et al., 2011; Toyokawa et al., 2011; Yang et al., 2010), whereas JMJD2A, JMJD2C, and JMJD2D have been reported to be overexpressed in prostate cancer, in which they function as androgen receptor (AR) coactivators to upregulate AR target genes (Shin and Janknecht, 2007; Wissmann et al., 2007). More recently, JMJD2D has been shown to function as a pro-proliferative protein and promote HCT116 colon cancer cell proliferation (Kim et al., 2012). In addition to cancer, upregulation of JMJD2A results in epigenetic changes in cardiac gene expression, promoting hypertrophic cardiomyopathy (Zhang et al., 2011). Despite the high sequence identity within their catalytic domains (>75%), there is a surprising degree of variability in the substrate specificities among JMJD2 homologs. Most JMJD2 enzymes display dual selectivity in demethylating H3K9me3, a mark associated with heterochromatin and gene silencing, and H3K36me3, a modification demarcating the coding regions of genes that are actively transcribed by RNA polymerase II (Mosammaparast and Shi, 2010; Whetstine et al., 2006). In contrast, JMJD2D efficiently demethylates di- and trimethyl H3K9

98 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Structural and Functional Analysis of JMJD2D

(H3K9me2/3), but is unable to recognize H3K36me3 (Whetstine et al., 2006). However, the molecular basis by which JMJD2D discriminates between H3K9me3 and H3K36me3 remains poorly understood. To gain insight into molecular determinants that distinguish the substrate specificities of different JMJD2 KDMs, we determined the crystal structure of a JMJD2D,2-OG,H3K9me3 ternary complex and the JMJD2D apoenzyme. Comparison of the structure of the JMJD2D ternary complex to JMJD2A,H3K9me3 and JMJD2A,H3K36me3 peptide complexes reveals structural variations between their substrate binding clefts that explain their different methylation site specificities. Molecular docking studies and kinetic analysis of JMJD2A and JMJD2D using a library of trimethylated histone H3 peptides corroborates these observations. Further, our studies demonstrate that recognition of H3T11 is important for H3K9me3 demethylation by JMJD2 KDMs, as mutations or phosphorylation of this residue disrupt activity. Collectively, our results define the molecular basis by which methylation site specificity is achieved in the JMJD2 family and provide a platform for designing substrate-mimetic inhibitors that exploit these structural variations to selectively target specific JMJD2 homologs. RESULTS Structure of the JMJD2D,2-OG,H3K9me3 Ternary Complex To elucidate the molecular basis for its discrimination between the H3K9me3 and H3K36me3 sites, we determined the crystal structure of the catalytic domain of JMJD2D (residues 12–342) bound to 2-OG and an H3K9me3 peptide (residues 1–15) at 1.8 A˚ resolution (Figure 1A; Table 1). Simulated annealing omit maps illustrate clear electron density for residues 6–15 in the H3K9me3 peptide substrate, the cosubstrate 2-OG, and Ni(II) that is bound in the Fe(II) coordination site within the active site (Figure 1B). We also determined a 2.5 A˚ crystal structure of the JMJD2D apoenzyme (Figure S1A available online; Table 1) containing Ni(II) in the active site. Alignment of the two structures reveals a high degree of similarity with no major structural variations induced by substrate binding (RMSDCa = 0.26 A˚ for all aligned Ca atoms). The JmjC domain (residues 146–312) of JMJD2D adopts a b-barrel fold that is highly homologous to the structures of other JMJD2 KDMs (Chen et al., 2006; Hillringhaus et al., 2011) (Figure 1A). The JmjC domain and C-terminal region of the enzyme (312–342) harbor a Zn-binding motif (Figure 1A) composed of residues His-244, Cys-238, Cys-310, and Cys312, a structural feature conserved throughout the JMJD2 family (Chen et al., 2006; Hillringhaus et al., 2011). The overall structure and domains of JMJD2D are very similar to those of JMJD2A, with an RMSDCa (261 atoms) of 0.4 A˚ (Figure 1C). Prior to the solution of these structures, two unpublished structures of the JMJD2D catalytic domain were deposited in the Protein Data Bank (PDB entries 3DXT and 3DXU) and have been cited in previous studies (Hillringhaus et al., 2011). In these structures, the C-terminal region adopts an alternative conformation that is neither observed in JMJD2A and JMJD2C, nor the JMJD2D structures reported here (Figures S1A and S1B). This alternative conformation is presumably stabilized by crystal-packing forces, as a structural alignment with the JMJD2D ternary complex illustrates

that this alternate conformation results in steric clashes between the N-terminal residues of the H3K9me3 peptide and the enzyme (Figures S1C and S1D). In addition, Gly-174, which forms part of the trimethyllysine binding pocket (Figures S1B and S2A), is shifted
12 A˚ out of the active site in the previously deposited JMJD2D structures, thus rendering this alternate conformation incompatible for trimethyllysine recognition and demethylation. In the active site of JMJD2D, Ni(II) occupies the Fe(II) binding site and is coordinated by His-192, Gly-194, and His-280 (Figure S2A). The methyl groups of K9me3 in the H3 peptide are coordinated by a network of CH,,,O hydrogen bonds to residues Tyr-181, Glu-194, and Gly-174 in JMJD2D. This binding mode is conserved in JMJD2A, with the exception that Ala-292 in JMJD2D is substituted by Ser-288 in JMJD2A (Figure S2B; Couture et al., 2007). The H3K9me3 peptide adopts a W-shaped conformation with two sharp bends when bound in the histonebinding cleft of JMJD2D, analogous to the H3K9me3 binding mode observed in JMJD2A (Couture et al., 2007; Ng et al., 2007). The first bend occurs at K9me3 and deposits the trimethyllysine substrate into a narrow channel leading to the active site, whereas the second bend corresponds to T11 and positions the threonyl side chain into a shallow pocket adjacent to the active site (Figure 1A). A combination of main-chain and sidechain hydrogen bonds and van der Waals contacts between the H3K9me3 peptide and JMJD2D facilitate optimal recognition of the substrate (Figure 1D). Many of these interactions are conserved in the JMJD2A,H3K9me3 peptide complex, but upon close inspection, several subtle, albeit significant, differences were observed in the H3K9me3 binding modes. H3K9me3 Binding Mode The H3K9me3 binding mode appears to be conserved between JMJD2A and JMJD2D, consistent with the similar catalytic efficiencies (kcat/Km values) they display toward an H3K9me3 peptide substrate (Krishnan et al., 2012; Table 2). Recognition of the H3K9me3 site by JMJD2A and JMJD2D involves hydrogen bonding or van der Waals contacts with residues R8, S10, T11, and G12 at the 1, +1, +2, and +3 positions in the H3K9 sequence (Figures 1D and 2). We first examined the interactions involving R8 recognition that had not been previously characterized in JMJD2 enzymes (Couture et al., 2007; Hillringhaus et al., 2011; Ng et al., 2007). R8 corresponds to the 1 position in the peptide sequence relative to K9 and is recognized by an intricate network of hydrogen bonds to Asp-139 and Tyr-179 in JMJD2D and to Glu-169 in JMJD2A (Figures 2A and 2B). In the case of both enzymes, an R8A substitution in the H3K9me3 peptide resulted in 6- to 10-fold decreases in the catalytic efficiencies (kcat/Km values) compared to demethylation of the wild-type (WT) peptide (Table 2; Figures 2A and 2B). These decreases were predominantly due to an increase in the Km values for the H3K9me3_R8A peptide that is suggestive of impaired binding to JMJD2A and JMJD2D, highlighting the importance of the interactions at the 1 position in the overall recognition of the H3K9me3 site. The residues that interact with R8 are conserved among all JMJD2 homologs (Figure S3), implying that recognition of R8 in the 1 position of the H3K9 sequence is conserved among these enzymes. With respect to residues in the C-terminal half of the H3K9me3 peptide, S10, T11, and G12, corresponding to the +1, +2, and +3

Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 99

Structure Structural and Functional Analysis of JMJD2D

Figure 1. Crystal Structure of JMJD2D (A) Structure of the JMJD2D,2-OG,H3K9me3 ternary complex. The JmjN domain (orange), JmjC domain (pink), mixed motif (blue), and C-terminal motif (red) are depicted in cartoon representation with the secondary structural elements labeled. The cyan sphere represents Ni(II) in the active site; the gray sphere represents Zn(II) in the Zn-binding motif. Stick representation of the cofactor 2-OG (black carbon atoms) and the substrate H3K9me3 peptide (green carbons) are shown. (B) Simulated annealing Fo-Fc omit maps are contoured at 3.0 s for 2-OG and Ni(II) (inset) and 2.0 s for the H3K9me3 peptide. *Two conformations were observed for S10 and each was modeled with an occupancy of 0.5. zThe side chain of K14 was not modeled due to a lack of clear electron density. (C) Alignment of JMJD2A (PDB entry 2OS2) (blue) and JMJD2D (pink) (RMSD Ca = 0.4 A˚). (D) Interactions between the H3K9me3 peptide substrate (green carbons) and JMJD2D (pink). Intrapeptide and enzyme-peptide hydrogen bonds are depicted as orange- and black-dashed lines, respectively. See also Figures S1 and S2.

positions, adopt a bent conformation that enables these amino acids to bind efficiently inside a pocket adjacent to the active site, as observed in the structures of JMJD2A,H3K9me3 peptide complexes (Figures 2C and 2D) (Couture et al., 2007; Ng et al., 2007). In JMJD2D, the bent conformation is maintained through two hydrogen-bond networks: (1) S10 adopts two alternate sidechain conformations that enable intrapeptide hydrogen bonding to either the G12 carbonyl oxygen and amide nitrogen or to the R8 carbonyl oxygen; and (2) a hydrogen bond between the side chains of T11 and Asp-139 in JMJD2D (Figure 2C). Mutation of S10 to alanine (S10A) or T11 to alanine (T11A) resulted in only a 2-fold reduction in the catalytic efficiency compared to the WT peptide with JMJD2D (Table 2; Figure 2C). A double S10A_T11A mutation, however, yielded a 6-fold decrease in catalytic efficiency, indicating that a combination of interactions to S10 and

T11 is important for optimal recognition of the H3K9me3 site by JMJD2D. The mode of recognition of S10 and T11 by JMJD2D starkly contrasts the interactions of these residues with JMJD2A. In a JMJD2A,H3K9me3 complex, the bent conformation of the C-terminal region of the peptide is stabilized only by intrapeptide hydrogen bonds between the side-chain hydroxyl group of S10 and the backbone oxygen and nitrogen atoms of G12 (Figure 2D). In addition, T11 is rotated away from Asp-135 and does not form hydrogen bonds with JMJD2A, unlike the orientation in JMJD2D (Figures 2C and 2D). An S10A mutation abrogated H3K9me3 recognition by JMJD2A with an
12-fold decrease in catalytic efficiency and an 8-fold increase in the Km value compared to the wild-type H3K9me3 peptide (Table 2; Figure 2D). Conversely, a T11A mutation did not affect recognition of the H3K9me3 site, consistent with the lack of

100 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Structural and Functional Analysis of JMJD2D

Table 1. Crystallographic Data and Refinement Statistics

RCSB PDB ID

JMJD2D,2OG,H3K9me3

JMJD2D Apoenzyme

4HON

4HOO

Data Collection Beamline

APS 21-ID-G

APS 21-ID-G

Wavelength (A˚)

0.9786

0.9786

Space Group

P212121

P32

Cell Dimensions a,b,c (A˚)

72.2, 79.7, 176.0

73.1, 73.1, 136.0

Resolution Range (A˚)a

25.0–1.80 (1.84–1.80)

25.0–2.50 (2.57–2.50)

Rmerge (%)a

6.00 (41.80)

5.90 (43.20)

I/sI

31.27 (5.25)

18.60 (2.40)

94.50 (100.0)

91.50 (87.0)

89,438

25,850

13.70 (10.60)

5.30 (3.10)

a

Completeness (%)a Unique Reflections Redundancya

Refinement and Validationb No. of Reflections

84,616

24,508

No. of Atoms

5891

5347

Protein Atoms

5347

5182

Ligand Atoms

133

4

Solvent Atoms

411

161

Rwork/Rfreec

18.9/21.4

20.7/24.6

Overall

20.5

39.0

Protein

19.7

39.5

Ligands

33.5

46.9

Waters

26.3

21.0

B-Factors (A˚2)

Root Mean Square Deviation Bond Length (A˚) 0.016

0.013

Bond Angles ( )

1.43

1.49

Clashscore (all atoms)d

6.85 (90th percentile n = 837)

7.94 (98th percentile n = 271)

Molprobity Score

1.41 (96th percentile n = 11,444)

2.23 (89th percentile n = 6960)

Resolution Range (A˚)

1.80 ± 0.25

2.50 ± 0.25

Favored (%)

98.5

95.7

Allowed (%)

1.5

4.3

Outliers (%)

0

0

MolProbity Scores

Ramachandran

a

Values in parentheses correspond to the highest-resolution shell. b Structures were refined in Refmac (Murshudov et al., 1997) using isotropic temperature-factor refinement. c Rwork = S jjFojjFcjj/S jFoj ; Rfree = 5% of the total reflections. d Clashscore is the number of serious steric overlaps (>0.4 A˚) per 1,000 atoms.

hydrogen bonding between T11 and JMJD2A (Figure 2D; Table 2). Interestingly, a T11S mutation resulted in a 6- to 10fold increase in the kcat/Km values for JMJD2A and JMJD2D compared to the WT H3K9me3 peptide (Table 2; Figure S4), primarily due to a decrease in the Km values for both enzymes

(Table 2). The presence of a serine residue at the +2 position corresponding to H3T11 may be more amenable to hydrogen bonding to the carboxylate group of Asp-139 in JMJD2D (Asp135 in JMJD2A) compared to the bulkier threonyl side chain, which has more limited conformational freedom due to steric constraints within the enzyme’s +2 binding pocket. In summary, the structural and kinetic data illustrate that R8 is recognized in a highly conserved manner in both JMD2A and JMJD2D, and they highlight major variations in the recognition of S10 and T11 by these enzymes, despite the homology of the H3K9me3 peptide conformation and the structural conservation of the +1 and +2 binding pockets in these homologs. Previous studies have demonstrated that other posttranslational modifications in the residues flanking H3K9me3 can influence K9me3 demethylation by JMJD2 KDMs. For example, phosphorylation of S10 (S10ph) has been shown to abolish H3K9me3 demethylation by JMJD2A (Ng et al., 2007), consistent with the importance of this residue in H3K9me3 recognition. In contrast, the presence of H3T11ph has been reported to enhance demethylation of H3K9me3 by JMJD2C and promote the upregulation of AR target genes (Metzger et al., 2008). Contrary to this finding, we observed that JMJD2A, JMJD2B, JMJD2C, and JMJD2D are essentially inactive toward an H3K9me3T11ph peptide, indicating that T11ph abrogates H3K9me3 demethylation by JMJD2 KDMs (Figure S5A). These data are corroborated by the structures of JMJD2A and JMJD2D in complex with H3K9me3 peptides, illustrating that the dimensions of the +2 binding pocket that recognizes H3T11 sterically preclude the binding of T11ph in the H3K9me3 site (Figure S5B). H3K36me3 Recognition and Discrimination Despite their high sequence identity (>75%) and structural homology, members of the JMJD2 family exhibit striking differences in their substrate specificities with respect to H3K36me3. JMJD2A, JMJD2B, and JMJD2C exhibit dual selectivity toward H3K9me3 and H3K36me3, whereas JMJD2D specifically demethylates H3K9me3 and is essentially inactive toward H3K36me3 (Table S1; Whetstine et al., 2006). To understand the molecular determinants underlying this varied specificity, we superimposed the structure of a JMJD2A,NOG,H3K36me3 complex (Chen et al., 2007) on the structure of our JMJD2D ternary complex. Based on this alignment, we identified several regions flanking the substrate binding cleft that differ in either sequence or structural conservation between the enzymes (Figure 3A). To ascertain whether these differences account for the preferential recognition by JMJD2D of H3K9me3 versus H3K36me3, we docked the H3K36me3 peptide into the substrate binding cleft of JMJD2D based on the superimposition of the JMJD2A,H3K36me3 and JMJD2D,H3K9me3 complexes (Figure 3A). The docking revealed minor steric clashes between JMJD2D and histone H3 residues N-terminal to K36 in the H3K36me3 peptide, whereas more severe steric clashes and electrostatic repulsion were observed with the histone H3 residues C-terminal to K36me3. Notably, there were pronounced clashes between H39 and R40 in the H3K36me3 peptide and His-90 and Leu-75 in JMJD2D (Figure 3B). The corresponding residues in JMJD2A, Asn-86 and Ile-71, adopt conformations that accommodate H39 and R40 through van der Waals interactions (Figure 3C). Further, Asp-135 in JMJD2A forms a salt bridge

Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 101

Structure Structural and Functional Analysis of JMJD2D

Table 2. JMJD2D and JMJD2A Kinetic Data JMJD2D

JMJD2A Km (mM)

kcat (min1)

53 ± 9

2.3 ± 0.2

45 ± 4

56 ± 9

1.7 ± 0.1

31 ± 4

H3K9me3_R8A

240 ± 27

1.6 ± 0.1

6.9 ± 0.9

340 ± 110

0.94 ± 0.19

2.9 ± 0.5

H3K9me3_S10A

96 ± 15

1.6 ± 0.2

17 ± 3

420 ± 19

1.0 ± 0.1

2.5 ± 0.1

H3K9me3_T11A

100 ± 13

2.5 ± 0.2

25 ± 2

40 ± 8

1.6 ± 0.1

43 ± 10

H3K9me3_S10A_T11A

110 ± 31

0.73 ± 0.13

6.8 ± 0.8

430 ± 32

1.1 ± 0.1

2.6 ± 0.1 320 ± 37

Km (mM) H3K9me3

H3K9me3_T11S

3.6 ± 0.1

kcat/Km 3 103 (min1 mM1)

kcat/Km 3 103 (min1 mM1)

230 ± 22

9.4 ± 1.4

2.9 ± 0.2

H3K36me3

NA

NA

NA

130 ± 6

1.2 ± 0.2

9.3 ± 1.4

H3K36me3_R40A

NA

NA

NA

830 ± 81

2.2 ± 0.2

2.7 ± 0.2

H3K9K36me3

NA

NA

NA

920 ± 150

0.57 ± 0.03

0.63 ± 0.07

H3K36K9me3

1400 ± 200

1.0 ± 0.1

0.75 ± 0.08

320 ± 28

0.82 ± 0.01

2.6 ± 0.3

280 ± 12

2.2 ± 0.2

7.9 ± 0.3

1.0 ± 0.1

62 ± 19

H3K36K9me3_V35R

16 ± 1

kcat (min1)

19 ± 8.8

Kinetic characterization of JMJD2D and JMJD2A. The amino acid sequences of the peptides used in the assay are listed in Table S2. NA, no activity detected. See also Tables S1 and S2.

with R40, an interaction that is pivotal for recognition of H3K36me3, as illustrated by the impaired demethylation of an H3K36me3_R40A peptide (Table 2; Figure 3C). Corroborating these observations, mutation of Ile-71 to a leucine (I71L) resulted in a 16-fold decrease in catalytic efficiency of JMJD2A toward H3K36me3 but no discernible effect on H3K9me3 demethylation (Figure S6). Additionally, the C terminus of the H3K36me3 peptide harbors a positively charged cluster of residues comprising H39, R40, and R42 that is juxtaposed to a positively charged patch on the surface of JMJD2D that includes Lys-91 and Lys-92, which may further preclude H3K36me3 binding via electrostatic repulsion (Figure 3D). In JMJD2A, Lys-91 and Lys-92 correspond to neutral amino acids Ile-87 and Gln-88 (Figure 3E). In addition to accommodating the H3K36me3 substrate without steric or electrostatic clashes, JMJD2A engages in several productive interactions along the length of the H3K36me3 peptide that promote substrate recognition (Chen et al., 2007). At the 1 position of the H3K36me3 site, the amide nitrogen of V35 engages in a hydrogen bond with the backbone carbonyl oxygen of Asp-311 in the loop linking helices a9 and a10 in the C-terminal region of JMJD2A (Figure 3F). This interaction facilitates the proper positioning of K36me3 substrate in the active site (Chen et al., 2007; Couture et al., 2007; Ng et al., 2007). Both the sequence and conformation of the loop that interacts with V35 vary substantially between JMJD2A and JMJD2D. In JMJD2A, residues Arg-309, Lys-310, Asp-311, and Met-312 form a broad U-shaped loop that bends sharply near the peptide-binding cleft, positioning the carbonyl oxygen of Asp311 for hydrogen bonding to the V35 amide (Figure 3F). In JMJD2D, this loop, consisting of residues Gly-313, Glu-314, Ala-315, and Arg-316 (denoted as the GEAR motif), is oriented away from the peptide-binding cleft and does not adopt a conformation that is conducive to hydrogen bonding to V35 in histone H3. Collectively, the structural comparisons between JMJD2A and JMJD2D combined with the JMJD2D,H3K36me3 docking analysis suggests that the loss of productive enzyme-substrate interactions coupled with multiple steric clashes and electrostatic repulsion precludes the recognition of the H3K36me3 site by JMJD2D.

Biochemical analysis of JMJD2D mutants based on sequence comparison with JMJD2A revealed that individual point mutations in the substrate-binding cleft were insufficient to enhance the activity of JMJD2D toward H3K36me3, and clustered mutations in JMJD2D significantly diminished enzyme stability (data not shown). Thus, to substantiate our structural observations, we assayed the catalytic activity of JMJD2D with hybrid peptide substrates bearing the N-terminal and C-terminal halves of the H3K9 and H3K36 methylation sites (Figure 4A). When assayed with a peptide composed of the N-terminal half of the H3K9 site and the C-terminal half of the H3K36 site (H3K9K36me3), JMJD2D exhibited no appreciable activity, consistent with its inability to demethylate the H3K36me3 peptide (Figure 4B). In contrast, JMJD2D displayed a substantial increase in its catalytic efficiency toward an H3K36K9me3 peptide composed of the N- and C-termini of H3K36 and H3K9 sites, respectively, although this activity was relatively weak compared to that observed for H3K9me3 (Table 2). These results corroborated observations from our docking studies, indicating that the residues C-terminal to K36 in histone H3 are primarily responsible for the discrimination of JMJD2D against this site. As noted earlier, the loop containing the GEAR motif in JMJD2D does not adopt a conformation that is conducive to hydrogen bonding with V35 in histone H3 (Figure 3F). To introduce a productive interaction at the 1 position of the demethylation site, V35 was mutated to an arginine (V35R) to mimic the interactions of R8 in the H3K9me3 peptide with JMJD2A and JMJD2D (Figures 2A and 2B). When assayed with the H3K36K9me3_V35R peptide, JMJD2D displayed a 10-fold increase in the kcat/Km compared to the H3K36K9me3 hybrid substrate (Figure 4B; Table 2). To complement these results, we next analyzed the kinetic parameters of JMJD2A using the H3 hybrid peptides. Consistent with JMJD2D, JMJD2A displayed weak activity toward the H3K9K36me3 peptide, suggesting that the residues preceding K36 in histone H3 contribute to substrate recognition (Table 2). Unexpectedly, when JMJD2A was assayed with H3K36K9me3 hybrid peptide, we observed a 4-fold reduction in the kcat/Km compared to WT H3K36me3 (Table 2). The decrease in catalytic

102 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Structural and Functional Analysis of JMJD2D

Figure 2. Recognition of H3K9me3 by JMJD2 KDMs Interactions at the 1, +1, and +2 positions observed in the JMJD2D (pink carbons),H3K9me3 (green carbons) complex (A and C) and in the JMJD2A (blue carbons),H3K9me3 (yellow carbons) complex (B and D). Enzyme-peptide hydrogen bonds and intrapeptide hydrogen bonds are shown as black- and orangedashed lines, respectively. (A and B) Recognition of R8 in the H3K9me3 substrate by JMJD2D (A) and JMJD2A (PDB entry 2OQ6) (B). The catalytic efficiencies (kcat/Km values) of the WT H3K9me3 and H3K9me3_R8A peptide substrates are illustrated as bar graphs in the insets. (C and D) Recognition of S10 and T11 in the H3K9me3 substrate by JMJD2D (C) and JMJD2A (D). The kcat/Km values for WT H3K9me3, H3K9me3_S10A, H3K9me3_T11A, and H3K9me3_S10A_T11A peptide substrates are represented in the insets. Error bars represent standard error calculated from three replicate assays. See also Figures S3–S5 and S7.

efficiency for the H3K36K9me3 peptide may have been a consequence of the loss of the salt bridge interaction between R40 in histone H3 and Asp-135 in JMJD2A (Figure 3C), as an R40A mutation in the H3K36me3 peptide impairs the catalytic efficiency of JMJD2A by 4-fold (Table 2). Conversely, JMJD2A exhibited a 7-fold increase in its catalytic efficiency when assayed with the H3K36K9me3_V35R peptide compared to WT H3K36me3 (Table 2). This increase in activity presumably reflects a compensatory effect in which the V35R mutation enhances interactions with JMJD2A through hydrogen bonding in the 1 position, as observed for R8 in the H3K9me3 site (Figure 2B). This interaction would offset the mutation of R40 to glycine in the H3K36K9me3 hybrid substrate (Figure 4A) that abolishes a salt bridge interaction with Asp-135 in JMJD2A (Figure 3C). In summary, the kinetic data corroborate our structural alignment and docking analysis of JMJD2A and JMJD2D, underscoring that steric clashes and electrostatic repulsion combined with the loss of productive enzyme-substrate interactions contribute to the discrimination against H3K36me3 demethylation by JMJD2D. DISCUSSION The functional diversity within the JMJD2 family has been attributed to their differential substrate specificities. (Cloos et al., 2008;

Mosammaparast and Shi, 2010; Nottke et al., 2009; Shi, 2007; Shi and Whetstine, 2007). JMJD2A, JMJD2B, and JMJD2C can efficiently demethylate H3K9me3 and H3K36me3, whereas JMJD2D is specific for the H3K9me3 site (Table S1). The structural studies reported here illustrate that JMJD2A and JMJD2D employ similar modes of recognition of the H3K9me3 site at the 1 position through hydrogen bonding to R8 (Figures 2A and 2B). However, these enzymes display unexpected differences in the recognition of residues S10 and T11 at the +1 and +2 positions of the H3K9me3 site (Figures 2C and 2D). S10 is pivotal for the recognition of the H3K9me3 site by JMJD2A, whereas in JMJD2D, this requirement is more flexible owing to a compensatory interaction formed between T11 and Asp-139 in JMJD2D. In addition, we also demonstrate that phosphorylation of T11 severely impairs demethylation of H3K9me3 by JMJD2 KDMs (Figure S5A), contrary to previous reports (Metzger et al., 2008). These findings correlate with prior studies illustrating that S10ph abrogates demethylation of H3K9me3 by JMJD2A (Ng et al., 2007) and are consistent with the overall acidic character of the S10-T11 binding cleft (Figures 3D and 3E), which is electrostatically incompatible with the binding of S10ph and T11ph. Our structural and biochemical results highlight key interactions at the 1, +1, +2, and +3 positions in the H3K9me3

Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 103

Structure Structural and Functional Analysis of JMJD2D

Figure 3. Differential Recognition H3K36me3 by JMJD2 Homologs

of

(A) Surface representation of JMJD2D (pink) containing the docked H3K36me3 peptide in stick representation (gray carbons). Purple patches highlight regions that lack sequence conservation with JMJD2A. (B) Steric clashes between H39 and R40 in H3K36me3 (gray carbons) and Asp-139, Leu-75, and His-90 in JMJD2D (pink carbons) are shown as red-dashed lines. (C) R40 in H3K36me3 (gray carbons) is recognized by a salt bridge interaction (black-dashed lines) with Asp-135 in JMJD2A (blue) (PDB entry 2P5B). Green dashes represent interatomic distances favorable for substrate binding. (D and E) Electrostatic surface representations of JMJD2D (D) and JMJD2A (E). Electrostatic potential is contoured from +3 kbT 3 e1 (blue) to 3 kbT 3 e1 (red). The backbone of the H3K36me3 peptide (gray) is depicted in cartoon form with the C-terminal basic residues H39, R40, and R42 represented in stick form (black). Yellow arrows denote the positively charged regions in JMJD2D that would electrostatically repel the basic C-terminal residues in the H3K36me3 peptide. (F) Hydrogen bond between Asp-311 in JMJD2A (blue) and V35 in H3K36me3 (gray carbons). The corresponding residue, Arg-316 (pink carbons), in JMJD2D is unable to form this hydrogen bond due to structural variability in the GEAR loop motif (inset). See also Figures S6 and S8.

sequence that are important for substrate recognition by JMJD2A and JMJD2D. These ‘‘signature’’ residues are also conserved in other known substrates of JMJD2 KDMs, most notably K26me3 of the linker histone H1.4 (H1.4K26me3) (Trojer et al., 2009) (Figure S7). In contrast to H3K9me3 and H1.4K26me3, JMJD2 homologs display distinct preferences for H3K36me3, which is efficiently demethylated by JMJD2A, JMJD2B, and JMJD2C but is not recognized by JMJD2D. Structural comparisons and docking studies of JMJD2A and JMJD2D (Figure 3) reveal that steric and electrostatic clashes combined with a lack of productive interactions contribute to the inability of JMJD2D to demethylate the H3K36me3 site. Consistent with these findings, the residues that can favorably accommodate H3K36me3 recognition are conserved among JMJD2A, JMJD2B, and JMJD2C but are divergent in JMJD2D (Figure S8). Despite their ability to demethylate H3K9me3, JMJD2 KDMs are inactive toward H3K27me3, a methylation site that mediates Polycomb group silencing (Cao et al., 2002; Kouzarides, 2007; Min et al., 2003). Although the H3K9 and H3K27 sequences harbor a conserved ARKS motif (where K is the methylated

lysine), sequence variations outside this ARKS motif, such as P30 in the +3 position, preclude recognition of the H3K27me3 site by the JMJD2 enzymes (Chen et al., 2007; Couture et al., 2007; Ng et al., 2007). A recently reported crystal structure of the H3K27me3-specific KDM UTX bound to an H3K27me3 peptide illustrates key differences in substrate recognition by UTX and the JMJD2 KDMs (Sengoku and Yokoyama, 2011). In JMJD2A and JMJD2D, the H3K9me3 peptide adopts a W-shaped conformation that is stabilized by interactions from residues R8–G12 (the –1 to +3 positions) in the H3K9 site. Conversely, UTX binds to an H3K27me3 peptide in an extended conformation through an extensive network of interactions spanning residues R17–G33, corresponding to the 10 to +6 positions in the H3K27 site. Despite these differences, UTX and JMJD2 KDMs share certain similarities in their substrate-binding modes. For example, UTX recognizes R26 in the 1 position of the H3K27me3 site through multiple hydrogen bonds. Moreover, an R26A mutation abolishes H3K27me3 demethylation by UTX, in agreement with the recognition of R8 in the 1 position of the H3K9me3 site by JMJD2A and JMJD2D (Figures 2A and 2B; Table 2). Similarly, phosphorylation of S28 at the +1 position abrogates demethylation of H3K27me3 by UTX, consistent with the mutual exclusivity of S10ph and K9me3 for demethylation by the JMJD2 enzymes (Ng et al.,

104 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Structural and Functional Analysis of JMJD2D

Figure 4. Substrate Specificity of JMJD2D toward Hybrid H3 Peptides (A) Amino acid sequences of the hybrid peptide substrates derived from the sequences of the H3K9 and H3K36 methylation sites. (B) Bar graph representing the kcat/Km values of JMJD2D toward H3K36me3 and the hybrid peptide substrates H3K9K36me3, H3K36K9me3, and H3K36K9me3_V35R. Error bars represent standard error calculated from three replicate assays.

2007). In summary, the different conformations adopted by the H3K9me3 and H3K27me3 sequences facilitate distinct recognition modes by their cognate JmjC KDMs. Aberrant expression of specific JMJD2 homologs contributes to the onset or progression of breast, prostate, colon, gastric, and squamous cell cancers, rendering these enzymes attractive targets for chemotherapeutic drug design (Cloos et al., 2006; Fodor et al., 2006; Fu et al., 2012; Ishimura et al., 2009; Kauffman et al., 2011; Kawazu et al., 2011; Kim et al., 2012; Li et al., 2011a, 2011b; Shi et al., 2011; Shin and Janknecht, 2007; Tan et al., 2011; Toyokawa et al., 2011; Wissmann et al., 2007; Yang et al., 2010). In recent years, several studies have reported synthetic inhibitors of JMJD2 KDMs, many of which possess moieties that mimic the cosubstrate 2-OG and coordinate the active-site Fe(II) to achieve high binding affinity (King et al., 2010; Luo et al., 2011; Rose et al., 2008; Thalhammer et al., 2011). More recently, a bisubstrate analog fusing 2-OG and an H3K9 peptide has been developed and crystallized with JMJD2A, illustrating the potential of bisubstrate peptidomimetics to inhibit JmjC KDMs (Woon et al., 2012). However, these inhibitors generally lack specificity for different JMJD2 homologs, an essential requirement in targeting cancers linked to specific JMJD2 enzymes. In JMJD2D, the loop in the C-terminal region of the catalytic domain possessing the GEAR motif is juxtaposed to the trimethyllysine binding cleft but is not conserved in sequence or structure with other JMJD2 homologs (Figures 3F and S8). It is conceivable that inhibitors can be designed with scaffolds that interact with this loop in different JMJD2 homologs, thus enhancing inhibitor specificity. In addition, the kinetic data illustrating that the T11S mutation in the H3K9me3 peptide augments apparent binding to JMJD2A and JMJD2D (Table 2; Figure S4) can be exploited in designing peptidomimetic inhibitors that incorporate a hydroxymethyl moiety that interacts with the T11 binding pocket in the substratebinding cleft. In conclusion, the structural and biochemical data presented here will aid in informing the structure-based

design of new demethylase inhibitors that display enhanced potency and selectivity toward specific JMJD2 KDMs. EXPERIMENTAL PROCEDURES Histone H3 Peptides The 15-mer H3K9me3 peptide that was cocrystallized with JMJD2D was purchased from New England Peptide and purified with an acetate counter ion. Methylated histone H3 peptide substrates used in the kinetic analyses of JMJD2A and JMJD2D were purchased from Anaspec and were purified with a chloride counter ion. Peptide concentrations were quantified by amino acid analysis, with the exception of peptides containing a tyrosine, whose concentrations were measured by their absorbance at 274 nm. The sequences of the peptides used in crystallization and kinetic experiments are listed in Table S2. Protein Cloning, Expression, and Purification For crystallization, the catalytic domain of human JMJD2D (residues 12–342) was cloned into a variant of pET15b (Millipore EMD Biosciences) with a tobacco etch virus protease cleavage site for removal of the N-terminal hexahistidine tag (Couture et al., 2007). The enzyme was overexpressed in E. coli BL21 DE3 Rosetta 2 cells (Millipore EMD Biosciences) and purified by Ni sepharose chromatography (GE Healthcare) followed by removal of the hexahistidine tag by tobacco etch virus protease and Superdex 200 gel filtration chromatography (GE Healthcare). The enzyme was concentrated to 10– 20 mg/ml, flash frozen, and stored at –80 C. Inductively coupled plasma high-resolution mass spectrometry analysis of these samples revealed >40% molar ratio of Ni present in the protein after extensive gel filtration, indicating the presence of Ni(II) in the active site (Krishnan et al., 2012). For demethylase assays, catalytic domains of JMJD2A (residues 1–350), JMJD2B (residues 9–357), JMJD2C (residues 1–350), and JMJD2D (residues 12–342) were expressed and purified as Strep(II)-tagged constructs to minimize contamination with inhibitory transition metal ions, as previously reported (Krishnan et al., 2012), with the exception that JMJD2B was purified in a single step by StrepTactin affinity chromatography. Crystallization and Structure Determination Crystals of JMJD2D,2-OG,H3K9me3 ternary complex were obtained by vapor diffusion at 4 C using 10 mg/ml JMJD2D, 1.5 mM 2-oxoglutarate and 1.5 mM H3K9me3 peptide. Crystals were grown in 7% polyethylene glycol 3350, 0.1 M sodium thiocyanate, and 0.35 M potassium nitrate. Crystals

Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 105

Structure Structural and Functional Analysis of JMJD2D

were harvested by serial transfer into the crystallization solution supplemented with 5%, 15%, and 25% 1,3-propanediol and subsequently flash frozen in liquid nitrogen. To obtain diffraction-quality crystals of the JMJD2D apoenzyme, a K93A/K94A mutant was generated based on sequence analysis by the surface entropy reduction server (Cieslik and Derewenda, 2009; Cooper et al., 2007; Derewenda and Vekilov, 2006; Goldschmidt et al., 2007). Crystals of the JMJD2D K93A/K94A mutant were grown at 4 C by vapor diffusion using 10 mg/ml protein in 0.2 M calcium acetate, 0.1 M HEPES 7.5, and 10% polyethylene glycol 8000. These crystals were harvested by serial transfer into the crystallization solution supplemented with 5%, 15%, and 25% glycerol and then flash frozen in liquid nitrogen. Diffraction data were collected at the LSCAT beamline 21-ID-G at the Advanced Photon Source Synchrotron (Argonne, IL). Data were processed and scaled using HKL2000 (Otwinowski and Minor, 1997). Molecular replacement was performed using MOLREP (Vagin and Teplyakov, 1997) with a JMJD2A structure (PDB entry 2Q8C) used as the search model for the JMJD2D ternary complex. The structure of the complex was used as a search model for the JMJD2D apoenzyme. Model building and refinement were conducted using Coot and Refmac, respectively (Emsley and Cowtan, 2004; Murshudov et al., 1997). TLS refinement was used to improve the electron density maps of the JMJD2D apoenzyme (Winn et al., 2001, 2003). Simulated-annealing omit maps were calculated using CNS (Brunger, 2007; Bru¨nger et al., 1998). After refinement, structures were validated using MOLPROBITY (Chen et al., 2010). Structural figures were rendered using PyMOL (Schro¨dinger) and the electrostatic surface was calculated using the Adaptive Poisson-Boltzmann Solver (APBS) plugin for PyMOL (Baker et al., 2001). Demethylase Assays KDM activity was measured using a formaldehyde dehydrogenase coupled assay as previously described (Krishnan et al., 2012). The assay was initiated by the addition of 1.0 mM 2-OG and a variable concentration of peptide substrate into the assay cocktail containing 50 mM HEPES (pH 7.5), 50 mM NaCl, 50 mM (NH4)2Fe(SO4)2, 1.0 mM L-ascorbic acid, 1.0 mM NAD+, 0.1 mM formaldehyde dehydrogenase (Roy and Bhagwat, 2007), and 1.0 mM JMJD2A or JMJD2D. NADH fluorescence was monitored (lex 340/lem 490) at 30 s intervals using a Sapphire 2 microplate reader (Tecan Group). Kinetic data were processed using GraphPad Prism (GraphPad Software), as previously described (Krishnan et al., 2012). ACCESSION NUMBERS The coordinates and structure factor files for the JMJD2D ternary complex and apoenzyme structures have been deposited in the Protein Data Bank under accession codes 4HON and 4HOO, respectively. SUPPLEMENTAL INFORMATION Supplemental Information includes eight figures and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.str.2012.10.018. ACKNOWLEDGMENTS We wish to thank John Tesmer and Paul Del Rizzo for reading the manuscript and providing insightful comments. We also acknowledge David Smith and Elena Kondrashkina (LS-CAT, Advanced Photon Source Synchrotron) for assistance in X-ray data collection. This work utilized the Protein Structure Core of the Michigan Diabetes Research and Training Center funded by grant DK020572 from the National Institute of Diabetes and Digestive and Kidney Disease. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Basic Energy Sciences, Office of Science, under contract No. DE-AC02-06CH11357. Use of the LS-CAT Sector 21 was supported by the Michigan Economic Development Corporation and the Michigan Technology Tri-Corridor for the support of this research program (Grant 085P1000817). S.K. received support through research grants from the University of Michigan’s Rackham Graduate School and Department of Biological Chemistry and from a Rackham Predoctoral Fellowship. This work was supported by funding from the University of Michigan’s

Biomedical Research Council and the Office of the Vice President of Research to R.C.T. Received: September 7, 2012 Revised: October 19, 2012 Accepted: October 31, 2012 Published online: December 6, 2012 REFERENCES Allis, C.D., Berger, S.L., Cote, J., Dent, S., Jenuwien, T., Kouzarides, T., Pillus, L., Reinberg, D., Shi, Y., Shiekhattar, R., et al. (2007). New nomenclature for chromatin-modifying enzymes. Cell 131, 633–636. Baker, N.A., Sept, D., Joseph, S., Holst, M.J., and McCammon, J.A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl. Acad. Sci. USA 98, 10037–10041. Black, J.C., Allen, A., Van Rechem, C., Forbes, E., Longworth, M., Tscho¨p, K., Rinehart, C., Quiton, J., Walsh, R., Smallwood, A., et al. (2010). Conserved antagonism between JMJD2A/KDM4A and HP1g during cell cycle progression. Mol. Cell 40, 736–748. Brunger, A.T. (2007). Version 1.2 of the Crystallography and NMR system. Nat. Protoc. 2, 2728–2733. Bru¨nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., GrosseKunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. Cao, R., Wang, L., Wang, H., Xia, L., Erdjument-Bromage, H., Tempst, P., Jones, R.S., and Zhang, Y. (2002). Role of histone H3 lysine 27 methylation in Polycomb-group silencing. Science 298, 1039–1043. Chen, V.B., Arendall, W.B., 3rd, Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. Chen, Z., Zang, J., Whetstine, J., Hong, X., Davrazou, F., Kutateladze, T.G., Simpson, M., Mao, Q., Pan, C.H., Dai, S., et al. (2006). Structural insights into histone demethylation by JMJD2 family members. Cell 125, 691–702. Chen, Z., Zang, J., Kappler, J., Hong, X., Crawford, F., Wang, Q., Lan, F., Jiang, C., Whetstine, J., Dai, S., et al. (2007). Structural basis of the recognition of a methylated histone tail by JMJD2A. Proc. Natl. Acad. Sci. USA 104, 10818–10823. Cieslik, M., and Derewenda, Z.S. (2009). The role of entropy and polarity in intermolecular contacts in protein crystals. Acta Crystallogr. D Biol. Crystallogr. 65, 500–509. Cloos, P.A., Christensen, J., Agger, K., Maiolica, A., Rappsilber, J., Antal, T., Hansen, K.H., and Helin, K. (2006). The putative oncogene GASC1 demethylates tri- and dimethylated lysine 9 on histone H3. Nature 442, 307–311. Cloos, P.A., Christensen, J., Agger, K., and Helin, K. (2008). Erasing the methyl mark: histone demethylases at the center of cellular differentiation and disease. Genes Dev. 22, 1115–1140. Cooper, D.R., Boczek, T., Grelewska, K., Pinkowska, M., Sikorska, M., Zawadzki, M., and Derewenda, Z. (2007). Protein crystallization by surface entropy reduction: optimization of the SER strategy. Acta Crystallogr. D Biol. Crystallogr. 63, 636–645. Couture, J.F., Collazo, E., Ortiz-Tello, P.A., Brunzelle, J.S., and Trievel, R.C. (2007). Specificity and mechanism of JMJD2A, a trimethyllysine-specific histone demethylase. Nat. Struct. Mol. Biol. 14, 689–695. Derewenda, Z.S., and Vekilov, P.G. (2006). Entropy and surface engineering in protein crystallization. Acta Crystallogr. D Biol. Crystallogr. 62, 116–124. Emsley, P., and Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132. Fodor, B.D., Kubicek, S., Yonezawa, M., O’Sullivan, R.J., Sengupta, R., PerezBurgos, L., Opravil, S., Mechtler, K., Schotta, G., and Jenuwein, T. (2006). Jmjd2b antagonizes H3K9 trimethylation at pericentric heterochromatin in mammalian cells. Genes Dev. 20, 1557–1562.

106 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved

Structure Structural and Functional Analysis of JMJD2D

Fu, L., Chen, L., Yang, J., Ye, T., Chen, Y., and Fang, J. (2012). HIF-1a-induced histone demethylase JMJD2B contributes to the malignant phenotype of colorectal cancer cells via an epigenetic mechanism. Carcinogenesis 33, 1664– 1673. Goldschmidt, L., Cooper, D.R., Derewenda, Z.S., and Eisenberg, D. (2007). Toward rational protein crystallization: a Web server for the design of crystallizable protein variants. Protein Sci. 16, 1569–1576. Hillringhaus, L., Yue, W.W., Rose, N.R., Ng, S.S., Gileadi, C., Loenarz, C., Bello, S.H., Bray, J.E., Schofield, C.J., and Oppermann, U. (2011). Structural and evolutionary basis for the dual substrate selectivity of human KDM4 histone demethylase family. J. Biol. Chem. 286, 41616–41625. Ishimura, A., Terashima, M., Kimura, H., Akagi, K., Suzuki, Y., Sugano, S., and Suzuki, T. (2009). Jmjd2c histone demethylase enhances the expression of Mdm2 oncogene. Biochem. Biophys. Res. Commun. 389, 366–371. Iwamori, N., Zhao, M., Meistrich, M.L., and Matzuk, M.M. (2011). The testis-enriched histone demethylase, KDM4D, regulates methylation of histone H3 lysine 9 during spermatogenesis in the mouse but is dispensable for fertility. Biol. Reprod. 84, 1225–1234. Katoh, M., and Katoh, M. (2004). Identification and characterization of JMJD2 family genes in silico. Int. J. Oncol. 24, 1623–1628. Kauffman, E.C., Robinson, B.D., Downes, M.J., Powell, L.G., Lee, M.M., Scherr, D.S., Gudas, L.J., and Mongan, N.P. (2011). Role of androgen receptor and associated lysine-demethylase coregulators, LSD1 and JMJD2A, in localized and advanced human bladder cancer. Mol. Carcinog. 50, 931–944. Kawazu, M., Saso, K., Tong, K.I., McQuire, T., Goto, K., Son, D.O., Wakeham, A., Miyagishi, M., Mak, T.W., and Okada, H. (2011). Histone demethylase JMJD2B functions as a co-factor of estrogen receptor in breast cancer proliferation and mammary gland development. PLoS ONE 6, e17830. Kim, T., and Buratowski, S. (2007). Two Saccharomyces cerevisiae JmjC domain proteins demethylate histone H3 Lys36 in transcribed regions to promote elongation. J. Biol. Chem. 282, 20827–20835. Kim, T.D., Oh, S., Shin, S., and Janknecht, R. (2012). Regulation of tumor suppressor p53 and HCT116 cell physiology by histone demethylase JMJD2D/KDM4D. PLoS ONE 7, e34618. King, O.N., Li, X.S., Sakurai, M., Kawamura, A., Rose, N.R., Ng, S.S., Quinn, A.M., Rai, G., Mott, B.T., Beswick, P., et al. (2010). Quantitative highthroughput screening identifies 8-hydroxyquinolines as cell-active histone demethylase inhibitors. PLoS ONE 5, e15535. Klose, R.J., Yamane, K., Bae, Y., Zhang, D., Erdjument-Bromage, H., Tempst, P., Wong, J., and Zhang, Y. (2006). The transcriptional repressor JHDM3A demethylates trimethyl histone H3 lysine 9 and lysine 36. Nature 442, 312–316. Klose, R.J., Gardner, K.E., Liang, G., Erdjument-Bromage, H., Tempst, P., and Zhang, Y. (2007). Demethylation of histone H3K36 and H3K9 by Rph1: a vestige of an H3K9 methylation system in Saccharomyces cerevisiae? Mol. Cell. Biol. 27, 3951–3961. Kouzarides, T. (2007). Chromatin modifications and their function. Cell 128, 693–705. Krishnan, S., Horowitz, S., and Trievel, R.C. (2011). Structure and function of histone H3 lysine 9 methyltransferases and demethylases. Chembiochem 12, 254–263. Krishnan, S., Collazo, E., Ortiz-Tello, P.A., and Trievel, R.C. (2012). Purification and assay protocols for obtaining highly active Jumonji C demethylases. Anal. Biochem. 420, 48–53. Li, B.X., Zhang, M.C., Luo, C.L., Yang, P., Li, H., Xu, H.M., Xu, H.F., Shen, Y.W., Xue, A.M., and Zhao, Z.Q. (2011a). Effects of RNA interference-mediated gene silencing of JMJD2A on human breast cancer cell line MDA-MB-231 in vitro. J. Exp. Clin. Cancer Res. 30, 90. Li, W., Zhao, L., Zang, W., Liu, Z., Chen, L., Liu, T., Xu, D., and Jia, J. (2011b). Histone demethylase JMJD2B is required for tumor cell proliferation and survival and is overexpressed in gastric cancer. Biochem. Biophys. Res. Commun. 416, 372–378. Lin, C.H., Li, B., Swanson, S., Zhang, Y., Florens, L., Washburn, M.P., Abmayr, S.M., and Workman, J.L. (2008). Heterochromatin protein 1a stimulates

histone H3 lysine 36 demethylation by the Drosophila KDM4A demethylase. Mol. Cell 32, 696–706. Loh, Y.H., Zhang, W., Chen, X., George, J., and Ng, H.H. (2007). Jmjd1a and Jmjd2c histone H3 Lys 9 demethylases regulate self-renewal in embryonic stem cells. Genes Dev. 21, 2545–2557. Lorbeck, M.T., Singh, N., Zervos, A., Dhatta, M., Lapchenko, M., Yang, C., and Elefant, F. (2010). The histone demethylase DmelA controls genes required for life span and male-specific sex determination in Drosophila. Gene 450, 8–17. Luo, X., Liu, Y., Kubicek, S., Myllyharju, J., Tumber, A., Ng, S., Che, K.H., Podoll, J., Heightman, T.D., Oppermann, U., et al. (2011). A selective inhibitor and probe of the cellular functions of Jumonji C domain-containing histone demethylases. J. Am. Chem. Soc. 133, 9451–9456. Metzger, E., Yin, N., Wissmann, M., Kunowska, N., Fischer, K., Friedrichs, N., Patnaik, D., Higgins, J.M., Potier, N., Scheidtmann, K.H., et al. (2008). Phosphorylation of histone H3 at threonine 11 establishes a novel chromatin mark for transcriptional regulation. Nat. Cell Biol. 10, 53–60. Min, J., Zhang, Y., and Xu, R.M. (2003). Structural basis for specific binding of Polycomb chromodomain to histone H3 methylated at Lys 27. Genes Dev. 17, 1823–1828. Mosammaparast, N., and Shi, Y. (2010). Reversal of histone methylation: biochemical and molecular mechanisms of histone demethylases. Annu. Rev. Biochem. 79, 155–179. Murshudov, G.N., Vagin, A.A., and Dodson, E.J. (1997). Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255. Ng, S.S., Kavanagh, K.L., McDonough, M.A., Butler, D., Pilka, E.S., Lienard, B.M., Bray, J.E., Savitsky, P., Gileadi, O., von Delft, F., et al. (2007). Crystal structures of histone demethylase JMJD2A reveal basis for substrate specificity. Nature 448, 87–91. Nottke, A., Colaia´covo, M.P., and Shi, Y. (2009). Developmental roles of the histone lysine demethylases. Development 136, 879–889. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. , J., Lie´nard, B.M., Bello, S.H., Sun, Z., Rose, N.R., Ng, S.S., Mecinovic McDonough, M.A., Oppermann, U., and Schofield, C.J. (2008). Inhibitor scaffolds for 2-oxoglutarate-dependent histone lysine demethylases. J. Med. Chem. 51, 7053–7056. Roy, T.W., and Bhagwat, A.S. (2007). Kinetic studies of Escherichia coli AlkB using a new fluorescence-based assay for DNA demethylation. Nucleic Acids Res. 35, e147. Sengoku, T., and Yokoyama, S. (2011). Structural basis for histone H3 Lys 27 demethylation by UTX/KDM6A. Genes Dev. 25, 2266–2277. Shi, L., Sun, L., Li, Q., Liang, J., Yu, W., Yi, X., Yang, X., Li, Y., Han, X., Zhang, Y., et al. (2011). Histone demethylase JMJD2B coordinates H3K4/H3K9 methylation and promotes hormonally responsive breast carcinogenesis. Proc. Natl. Acad. Sci. USA 108, 7541–7546. Shi, Y. (2007). Histone lysine demethylases: emerging roles in development, physiology and disease. Nat. Rev. Genet. 8, 829–833. Shi, Y., and Whetstine, J.R. (2007). Dynamic regulation of histone lysine methylation by demethylases. Mol. Cell 25, 1–14. Shin, S., and Janknecht, R. (2007). Activation of androgen receptor by histone demethylases JMJD2A and JMJD2D. Biochem. Biophys. Res. Commun. 359, 742–746. Strobl-Mazzulla, P.H., Sauka-Spengler, T., and Bronner-Fraser, M. (2010). Histone demethylase JmjD2A regulates neural crest specification. Dev. Cell 19, 460–468. Tan, M.K., Lim, H.J., and Harper, J.W. (2011). SCF(FBXO22) regulates histone H3 lysine 9 and 36 methylation levels by targeting histone demethylase KDM4A for ubiquitin-mediated proteasomal degradation. Mol. Cell. Biol. 31, 3687–3699. , J., Loenarz, C., Tumber, A., Rose, N.R., Thalhammer, A., Mecinovic Heightman, T.D., and Schofield, C.J. (2011). Inhibition of the histone demethylase JMJD2E by 3-substituted pyridine 2,4-dicarboxylates. Org. Biomol. Chem. 9, 127–135.

Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved 107

Structure Structural and Functional Analysis of JMJD2D

Toyokawa, G., Cho, H.S., Iwai, Y., Yoshimatsu, M., Takawa, M., Hayami, S., Maejima, K., Shimizu, N., Tanaka, H., Tsunoda, T., et al. (2011). The histone demethylase JMJD2B plays an essential role in human carcinogenesis through positive regulation of cyclin-dependent kinase 6. Cancer Prev. Res. (Phila.) 4, 2051–2061. Trojer, P., Zhang, J., Yonezawa, M., Schmidt, A., Zheng, H., Jenuwein, T., and Reinberg, D. (2009). Dynamic histone H1 isotype 4 methylation and demethylation by histone lysine methyltransferase G9a/KMT1C and the Jumonji domain-containing JMJD2/KDM4 proteins. J. Biol. Chem. 284, 8395–8405. Tu, S., Bulloch, E.M., Yang, L., Ren, C., Huang, W.C., Hsu, P.H., Chen, C.H., Liao, C.L., Yu, H.M., Lo, W.S., et al. (2007). Identification of histone demethylases in Saccharomyces cerevisiae. J. Biol. Chem. 282, 14262–14271. Upadhyay, A.K., Horton, J.R., Zhang, X., and Cheng, X. (2011). Coordinated methyl-lysine erasure: structural and functional linkage of a Jumonji demethylase domain and a reader domain. Curr. Opin. Struct. Biol. 21, 750–760. Vagin, A., and Teplyakov, A. (1997). MOLREP: an automated program for molecular replacement. J. Appl. Cryst. 30, 1022–1025. Whetstine, J.R., Nottke, A., Lan, F., Huarte, M., Smolikov, S., Chen, Z., Spooner, E., Li, E., Zhang, G., Colaiacovo, M., and Shi, Y. (2006). Reversal of histone lysine trimethylation by the JMJD2 family of histone demethylases. Cell 125, 467–481.

Winn, M.D., Isupov, M.N., and Murshudov, G.N. (2001). Use of TLS parameters to model anisotropic displacements in macromolecular refinement. Acta Crystallogr. D Biol. Crystallogr. 57, 122–133. Winn, M.D., Murshudov, G.N., and Papiz, M.Z. (2003). Macromolecular TLS refinement in REFMAC at moderate resolutions. Methods Enzymol. 374, 300–321. Wissmann, M., Yin, N., Mu¨ller, J.M., Greschik, H., Fodor, B.D., Jenuwein, T., Vogler, C., Schneider, R., Gu¨nther, T., Buettner, R., et al. (2007). Cooperative demethylation by JMJD2C and LSD1 promotes androgen receptor-dependent gene expression. Nat. Cell Biol. 9, 347–353. Woon, E.C., Tumber, A., Kawamura, A., Hillringhaus, L., Ge, W., Rose, N.R., Ma, J.H., Chan, M.C., Walport, L.J., Che, K.H., et al. (2012). Linking of 2-oxoglutarate and substrate binding sites enables potent and highly selective inhibition of JmjC histone demethylases. Angew. Chem. Int. Ed. Engl. 51, 1631–1634. Yang, J., Jubb, A.M., Pike, L., Buffa, F.M., Turley, H., Baban, D., Leek, R., Gatter, K.C., Ragoussis, J., and Harris, A.L. (2010). The histone demethylase JMJD2B is regulated by estrogen receptor alpha and hypoxia, and is a key mediator of estrogen induced growth. Cancer Res. 70, 6456–6466. Zhang, Q.J., Chen, H.Z., Wang, L., Liu, D.P., Hill, J.A., and Liu, Z.P. (2011). The histone trimethyllysine demethylase JMJD2A promotes cardiac hypertrophy in response to hypertrophic stimuli in mice. J. Clin. Invest. 121, 2447–2456.

108 Structure 21, 98–108, January 8, 2013 ª2013 Elsevier Ltd All rights reserved