Linking of 2-oxoglutarate and substrate binding sites enables potent and highly selective inhibition of JmjC histone demethylases

Article abstract of DOI:10.1002/anie.201107833

Select an isoform: Linking of cosubstrate and substrate binding sites enables highly selective inhibiton of isoforms of human histone lysine demethylases. The results should provide a basis for the development of potent and selective JmjC inhibitors, poss

Full text of DOI:10.1002/anie.201107833

Angewandte
Chemie
DOI: 10.1002/anie.201107833
Histone Demethylase Inhibition
Linking of 2-Oxoglutarate and Substrate Binding Sites Enables Potent
and Highly Selective Inhibition of JmjC Histone Demethylases**
Esther C. Y. Woon, Anthony Tumber, Akane Kawamura, Lars Hillringhaus, Wei Ge,
Nathan R. Rose, Jerome H. Y. Ma, Mun Chiang Chan, Louise J. Walport, Ka Hing Che,
Stanley S. Ng, Brian D. Marsden, Udo Oppermann, Michael A. McDonough, and
Christopher J. Schofield*
Ne-Methylation of histone lysine residues is an “epigenetic
modification” that can be either transcriptionally activating
or deactivating, depending on the position of the lysine, its
methylation state and the presence of other modifications.[1]
The largest family of demethylases, the JmjC enzymes,
employ 2-oxoglutarate (2OG) as a cosubstrate (Figure 1a).[2,3]
Some JmjC demethylases are targeted for cancer treat-
ment[4–6] and inflammatory diseases.[7] There are 5 JmjC
demethylase subfamilies, targeting histone lysines (H3K =
histone 3 lysine-residue) including at H3K4, H3K9, H3K27,
and H3K36 (Figure 1b). The factors determining JmjC
selectivities are emerging, and involve both catalytic[8–10] and
non-catalytic domains.[11–13] Although there are reports of
JmjC inhibitors,[14–19] to date there are no reported compounds
that are selective for subfamilies/isoforms. Here we report
that a strategy involving binding to both the 2OG and
substrate binding sites leads to selective and potent inhibitors
of the JMJD2 subfamily.
There are predicted to be four human JMJD2 enzymes (A
to D) and a “pseudogene” product JMJD2E. JMJD2A–C
accept both H3K9me3/me2 and H3K36me3/me2, whereas
JMJD2D–E only accept H3K9me3/me2.[20] Most, if not all,
reported JmjC inhibitors are 2OG analogues with limited or
undetermined selectivity, and with the exception of some
peptide-based inhibitors,[14] have not, at least rationally,
exploited the histone binding pocket.[14,21] We reasoned that
“two-component inhibitors” that bind to 2OG and histone
[*] Dr. E. C. Y. Woon,[+] Dr. A. Kawamura, Dr. L. Hillringhaus, Dr. W. Ge,
Dr. N. R. Rose, J. H. Y. Ma, M. C. Chan, L. J. Walport,
Dr. M. A. McDonough, Prof. C. J. Schofield
Chemistry Research Laboratory, University of Oxford
12 Mansfield Road, Oxford, OX1 3TA (UK)
E-mail: christopher.schofield@chem.ox.ac.uk
Dr. A. Tumber, K. H. Che, Dr. S. S. Ng, Dr. B. D. Marsden,
Prof. U. Oppermann
Structural Genomics Consortium, University of Oxford
Old Road Campus Research Building, Oxford, OX3 7DQ (UK)
[+] Current address: Department of Pharmacy, National University of
Singapore, 18 Science Drive 4, Singapore 117543 (Singapore)
[**] This work was supported by the Biotechnology and Biological
Sciences Research Council, the Wellcome Trust, the European
Research Council, the German Academic Exchange Service (DAAD)
(L.H.), the European Union, and the Structural Genomics Con-
sortium (registered charity no. 1097737).
Figure 1. The JmjC histone demethylases. a) Reactions catalyzed by
JmjC enzymes. b) Members of the five assigned subfamilies of human
JmjC enzymes. Substrate assignments are in progress for some
members. c) Active-site view from a crystal structure of JMJD2A bound
to a H3K9me3 fragment (orange) and 2OG (blue) (PDB ID 2Q8C).[9]
The cross-linking point identified by mass spectrometric studies
(Figure 2) is shown by the red arrow.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201107833.
Angew. Chem. Int. Ed. 2012, 51, 1631 –1634
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1631

.
Angewandte
Communications
binding sites may enable selectivity (Figure 1c). In the design
of JMJD2 selective inhibitors, we used DNOC 1 (N-oxalyl-d-
cysteine), which chelates to the iron as the 2OG-binding
component and substrate fragment H3K9me3 (7–14) as the
substrate-binding component (Figure 2a). To identify a
suitable cross-linking point, we employed the technique of
dynamic chemistry linked to MS analysis using the reversible
thiol–disulfide reaction (with multiple thiols, this is termed
dynamic combinatorial mass spectrometry, DCMS). Non-
denaturing electrospray ionization mass spectrometry (ESI-
MS) is used to identify members of equilibrating disulfides
that bind to the target protein.[17,22,23]
DNOC 1, LNOC 2, or NOG 3 (N-oxalylglycine). No Tm shift
was observed for the K9me2/me1/me0 T11C variants, includ-
ing those with a longer H3K9me3(1–15) sequence (Fig-
ure 3a), implying binding of the trimethylated form is
preferred for binding of the cross-linked molecules. In
contrast, varying the methylation state at K4 of H3K9me3-
(1–15) did not affect the Tm shift (Figure 3a). To study the
effects of substrate length, residues surrounding the
“Kme3SC motif” of the T11C peptides were systematically
removed (Supporting Information, Figure S1). Tm shift
analyses of these peptides with 1 in JMJD2A and JMJD2E
imply both Ala7 and Arg8 of T11C/4 are important for
binding (Figure S1).
A structure for JMJD2A in complex with disulfide 4 (from
DNOC 1 and T11C) (PDB ID 3U4S, Figure 3c, S2 and S3)
reveals how 2OG and histone binding sites can be occupied
simultaneously. When complexed with 1, the active site metal
(NiII for FeII) is coordinated by His188, Glu190 and His276
side chains. The N-oxalyl amide carbonyl oxygen and one of
the carboxylate oxygens of 1 coordinate the metal bidentately,
as does 2OG. The other carboxylate of 1 forms a salt bridge
with Lys206 side-chain amine (2.8 ꢀ) and hydrogen-bond
with Tyr132 side-chain hydroxy (2.5 ꢀ). Consistent with the
binding studies (Figure S1), both Ala7 and Arg8 of T11C
form apparently important interactions (Figure S2).
Analysis of JMJD2 structures[10] reveals a large pocket in
which the 2OG and H3K binding sites are contiguous. Our
approach to identifying a site for cross-linking involved
substituting each of the residues proceeding the trimethylated
lysine of a H3K9me3 fragment sequence (residues 7–14) with
a Cys residue 10–13, Figure 2a, then using MS to analyze for
cross-linking to DNOC 1 or LNOC 2 concomitant with
binding to JMJD2E (which is amenable to non-denaturing
MS analysis). The results reveal that only the combination of
peptide T11C (i.e. Thr11 replaced with Cys) and DNOC 1
form an apparent disulfide 4, capable of substantially binding
to JMJD2E.FeII (hereafter JMJD2E, Figure 2b). No binding
was observed for the other H3K9me3 cysteinyl variants:
DNOC 1/LNOC 2 combinations, implying a selection of 4 by
JMJD2E.
The T11C component of 4 adopts a “bent conformation”
as observed for an H3K9me3 peptide fragment in complex
with JMJD2A, which directs the N-trimethyl group. There is
no significant difference in the conformations of the T11C (4)
and H3K9me3 backbones (rmsd 0.3 ꢀ), except for the side-
chain position of Cys11 in T11C, which rotates by ca. 708 with
respect to that of H3K9me3 Thr11 to enable disulfide
formation with the thiol of 1. Modeling suggests that only
Differential scanning fluorimetry (“thermal shift” Tm)
assays[24] support the MS analyses, with large Tm shifts being
observed for the T11C:1 combination for JMJD2A (Tm
shift = 9.58C) and JMJD2E (Tm shift = 13.28C); T11C or 1
alone gave insignificant Tm shifts (Figure 2c). No Tm shifts
were observed for other combinations of cysteinyl peptides:
Figure 2. Mass spectrometric approach to identifying points for inhibitor–substrate cross-linking a) Sequences and structures of inhibitors. b) MS
analyses on JMJD2E.FeII (labeled E, z=13) in the presence of i) no compound, and compounds ii) N-oxalyl-d-cysteine 1, iii) disulfides 4, iv) cross-
linked compound 6. Masses (Da) in parentheses. c) Thermal shift with 1–3.
1632
www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 1631 –1634

Angewandte
Chemie
presence of 2OG; T11C alone showed < 5%
demethylation by JMJD2E under these
conditions (Figure S5b).
Profiling studies using luminescence
assays[26] and Tm analyses revealed selectivity
of 6 for the JMJD2 subfamily; 6 potently
inhibits both JMJD2A (IC50 = 0.27 mm, Tm =
+ 7.18C) and JMJD2E (IC50 = 0.09 mm, Tm =
+ 12.18C), with little or no inhibition or Tm
shift against other JmjC subfamilies (Fig-
ure 3b and S6a). The JMJD1A results are
notable because it also acts on H3K9. Tm
assays were carried out for the K9me3/me2/
me1/me0 Cys variants of H3K9 (1–15) in the
presence of 1 for JMJD1A; in all cases, no
Tm shifts were observed for the cross-linked
compounds (Figure S7a). Similarly, no shifts
were observed for the disulfides formed
between 1 and K9me3/me2/me1/me0 cys-
teine variants of K4me3K9 (1–15) for PHF8,
which acts on H3K9 (Figure S7b). 6 was also
tested against PHD2 (hypoxia-inducible
factor prolylhydroxylase 2, which is central
to the hypoxic response)[3] and FIH (factor
inhibiting hypoxia inducible factor)[3] with
IC50 values > 500 mm and > 1 mm, respec-
tively (Figure 3a). 6 is thus, in our assays,
> 300-fold selective for JMJD2A over the
other 2OG oxygenases tested (Figure 1b).
The Tm results revealed that the K9me3
form of the cross-linked inhibitor (as in 6) is
the preferred methylation state form for
inhibition (Figure 3a). This may reflect the
proximity of the cross-linking and the
methylated lysine arm as revealed in the
structure (Figure 3c). In the selectivity stud-
ies, JmjC demethylases with varying meth-
Figure 3. Selective inhibition of JMJD2A and JMJD2E. a) Tm shift studies of disulfide cross-
linking of DNOC 1 to T11C variants. b) Selectivity studies of 6 and 9 against representa-
tives of JmjC demethylases, PHD2 and FIH; NI100 =no inhibition at 100 mm. c) View from a
JMJD2A structure (white) bound to T11C (green) and DNOC 1 (salmon) (PDB ID 3U4S).
d) Superimposition of a view from a JMJD2A:T11C:1 structure with that of a structure of
JMJD2A complexed with H3K9me3 (orange)(PDB ID 2OQ6, NiII substituting for FeII).[10]
Thr11 in H3K9me3 is correctly positioned for reaction with
DNOC 1; cross-linking is not possible with cysteine replace-
ment at H3K9me3 Ser10, Gly12 or Gly13 (Figure 2). Notably,
the Lys14 side-chain position of T11C (4) is rotated by ca. 958
with respect to its position in H3K9me3; this directs the amide
of Gly13 in T11C (4) away from the backbone oxygen of
His240JMJD2A, weakening the hydrogen bond between them
compared to that with Gly13 of H3K9me3 (Figure S4).
His240JMJD2A coordinates an ion zinc, the binding of which is
proposed as important for substrate recognition by the
JMJD2 subfamily.[10,21]
ylation state and sequence selectivities (H3K9, H3K27,
H3K36) were tested. Note that no inhibition of JMJD1A
and PHF8 were observed; both these enzymes, like the
JMJD2 subfamily, act on H3K9, but prefer Kme2 (Figure 1b).
The results thus demonstrate that the selectivities of 4 and 6
We synthesized a stable analogue of the T11C:1 disulfide
À
4, wherein the disulfide is replaced with a CH2 S bond, using
a thiol–ene coupling as a key step,[25] i.e. addition of a thinyl
radical to an allylglycine-containing peptide 5 (Scheme 1a;
Supporting Information). The analogue 6 binds strongly to
JMJD2E by ESI-MS (Figure 2b); Kinetic analyses[18] reveal a
mixed mode of inhibition of JMJD2E by 6 with respect to
2OG (Kiapp = 114 Æ 46 nm, a = 2.3 Æ 1.4), suggesting it binds
both JMJD2E and JMJD2E.2OG (Figure S5a). No demethy-
lation of 6 by JMJD2E was observed by MALDI-MS[10] in the
Scheme 1. Synthesis of stable carbon analogue a) 6 and b) 9 by
utilizing a thiol–ene coupling reaction. Reagents and conditions:
1) Dimethoxyphenyl acetophenone, anhydrous MeOH, hn (365 nm,
2 h); 2) 30% (v/v) CF3CO2H/CH2Cl2, RT (3 h).
Angew. Chem. Int. Ed. 2012, 51, 1631 –1634
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
1633

.
Angewandte
Communications
[2] C. Martin, Y. Zhang, Nat. Rev. Mol. Cell Biol. 2005, 6, 838 – 849.
[3] N. R. Rose, M. A. McDonough, O. N. King, A. Kawamura, C. J.
Schofield, Chem. Soc. Rev. 2011, 40, 4364 – 4397.
[4] J. He, A. T. Nguyen, Y. Zhang, Blood 2011, 117, 3869 – 3880.
[5] J. Yang, A. M. Jubb, L. Pike, F. M. Buffa, H. Turley, D. Baban, R.
Leek, K. C. Gatter, J. Ragoussis, A. L. Harris, Cancer Res. 2010,
70, 6456 – 6466.
[6] P. A. Cloos, J. Christensen, K. Agger, A. Maiolica, J. Rappsilber,
T. Antal, K. H. Hansen, K. Helin, Nature 2006, 442, 307 – 311.
[7] F. De Santa, M. G. Totaro, E. Prosperini, S. Notarbartolo, G.
Testa, G. Natoli, Cell 2007, 130, 1083 – 1094.
[8] J. F. Couture, E. Collazo, P. A. Ortiz-Tello, J. S. Brunzelle, R. C.
Trievel, Nat. Struct. Mol. Biol. 2007, 14, 689 – 695.
[9] Z. Chen, J. Zang, J. Kappler, X. Hong, F. Crawford, Q. Wang, F.
Lan, C. Jiang, J. Whetstine, S. Dai, K. Hansen, Y. Shi, G. Zhang,
Proc. Natl. Acad. Sci. USA 2007, 104, 10818 – 10823.
[10] S. S. Ng, K. L. Kavanagh, M. A. McDonough, D. Butler, E. S.
Pilka, B. M. Lienard, J. E. Bray, P. Savitsky, O. Gileadi, F.
von Delft, N. R. Rose, J. Offer, J. C. Scheinost, T. Borowski, M.
Sundstrom, C. J. Schofield, U. Oppermann, Nature 2007, 448,
87 – 91.
[11] J. R. Horton, A. K. Upadhyay, H. H. Qi, X. Zhang, Y. Shi, X.
Cheng, Nat. Struct. Mol. Biol. 2010, 17, 38 – 43.
[12] C. Loenarz, W. Ge, M. L. Coleman, N. R. Rose, C. D. O. Cooper,
R. J. Klose, P. J. Ratcliffe, C. J. Schofield, Hum. Mol. Genet.
2010, 19, 217 – 222.
[13] L. Hillringhaus, W. W. Yue, N. R. Rose, S. S. Ng, C. Gileadi, C.
Loenarz, S. H. Bello, J. E. Bray, C. J. Schofield, U. Oppermann, J.
Biol. Chem. 2011, DOI: 10.1074/jbc.M11.283689.
[14] B. Lohse, A. L. Nielsen, J. B. L. Kristensen, C. Helgstrand,
P. A. C. Cloos, L. Olsen, M. Gajhede, R. P. Clausen, J. L.
Kristensen, Angew. Chem. 2011, 123, 9266 – 9269; Angew.
Chem. Int. Ed. 2011, 50, 9100 – 9103.
[15] X. Luo, Y. Liu, S. Kubicek, J. Myllyharju, A. Tumber, S. Ng,
K. H. Che, J. Podoll, T. D. Heightman, U. Oppermann, S. L.
Schreiber, X. Wang, J. Am. Chem. Soc. 2011, 133, 9451 – 9456.
[16] K. H. Chang, O. N. King, A. Tumber, E. C. Y. Woon, T. D.
Heightman, M. A. McDonough, C. J. Schofield, N. R. Rose,
ChemMedChem 2011, 6, 759 – 764.
arise from active-site structure coupled to sequence and
methylation state selectivity.
We investigated whether we could design JMJD2 isoform-
group selective inhibitors, using JMJD2A and JMJD2E as
representatives. JMJD2A–C accept both H3K9me3/me2 and
H3K36me3/me2 whereas JMJD2D–E only accept H3K9me3/
me2. We proposed that a H3K36-based inhibitor might select
for
JMJD2A
over
JMJD2E.
Modeling
of
a
JMJD2A:H3K36me3 structure suggests cross-linking of 1
with Pro38 of H3K36me3 (31–41) could be achieved (Fig-
ure S7). Synthesis of crossed-linked compound 9 was carried
out by thiol–ene coupling of an O-allylhydroxyproline-
containing peptide 8 with 1 (Scheme 1b). 9 demonstrated
60-fold selectivity for JMJD2A (IC50 = 1.5 mm, Tm shift =
5.88C) over JMJD2E (IC50 = 91 mm, Tm shift < 18C) and 9
did not bind to JMJD2E by non-denaturing ESI-MS. There is
little or no inhibitory activity or Tm shift of 9 against the other
tested JmjC subfamilies (Figure 3b and S6b). Notably, 9
showed > 35-fold selectivity against FBXL11 (IC50 = 55 mm,
Tm shift = 0.98C) and > 650-fold selectivity against PHF8
(IC50 > 1 mm, Tm shift = À0.38C).
Overall the results provide proof of principle that
selective inhibition of not only JmjC subfamilies, but also
isoform subgroups, is achievable by utilizing both 2OG and
substrate binding sites. With use of an appropriately func-
tional inhibitor, the cross-linking approach should be appli-
cable to other 2OG oxygenase subfamilies. By exploiting the
inherent substrate selectivity of the JmjC enzymes, compound
6, which is based on H3K9me3, exhibits a high degree of
selectivity and potency for the JMJD2 over other JmjC
subfamilies. Discrimination against human 2OG oxygenases
should be possible as shown by a lack of inhibition against
PHD2 and FIH. Importantly, isoform-subgroup selectivity
could also be achieved using this strategy, as demonstrated by
9 which is selective for JMJD2A over JMJD2E. It is likely
that the approach outlined here will be applicable to related
histone demethylases, and, more widely, other 2OG oxygen-
ases. The combined results, including crystallographic analy-
ses, should provide a basis for the development of potent and
selective JmjC inhibitors, suitable for use as functional probes
in cells and, possibly, for clinical use.
[17] N. R. Rose, E. C. Y. Woon, G. L. Kingham, O. N. F. King, J.
´
Mecinovic, I. J. Clifton, S. S. Ng, J. Talib-Hardy, U. Oppermann,
M. A. McDonough, C. J. Schofield, J. Med. Chem. 2010, 53,
1810 – 1818.
´
[18] N. R. Rose, S. S. Ng, J. Mecinovic, B. M. Liꢁnard, S. H. Bello, Z.
Sun, M. A. McDonough, U. Oppermann, C. J. Schofield, J. Med.
Chem. 2008, 51, 7053 – 7056.
[19] S. Hamada, T. Suzuki, K. Mino, K. Koseki, F. Oehme, I. Flamme,
H. Ozasa, Y. Itoh, D. Ogasawara, H. Komaarashi, A. Kato, H.
Tsumoto, H. Nakagawa, M. Hasegawa, R. Sasaki, T. Mizukami,
N. Miyata, J. Med. Chem. 2010, 53, 5629 – 5638.
[20] J. R. Whetstine, A. Nottke, F. Lan, M. Huarte, S. Smolikov, Z.
Chen, E. Spooner, E. Li, G. Zhang, M. Colaiacovo, Y. Shi, Cell
2006, 125, 467 – 481.
Experimental Section
Experimental details, including syntheses and characterizations, ESI-
MS methods, thermal shift and inhibition assays, protein purifications,
crystallization and structure solutions, are given in the Supporting
Information. The coordinates for JMJD2A in complex with T11C and
DNOC 1 is deposited as PDB ID 3U4S.
[21] R. Sekirnik, N. R. Rose, A. Thalhammer, P. T. Seden, J.
´
Mecinovic, C. J. Schofield, Chem. Commun. 2009, 6376 – 6378.
[22] B. M. R. Liꢁnard, R. Hꢂting, P. Lassaux, M. Galleni, J.-M. Frꢃre,
C. J. Schofield, J. Med. Chem. 2008, 51, 684 – 688.
[23] S.-A. Poulsen, J. Am. Soc. Mass Spectrom. 2006, 17, 1074 – 1080.
[24] F. H. Niesen, H. Berglund, M. Vedadi, Nat. Protoc. 2007, 2,
2212 – 2221.
Received: November 7, 2011
Published online: January 12, 2012
Keywords: 2-oxoglutarate · epigenetics ·
histone lysine demethylases · oxygenases · thiol–ene reaction
[25] A. Dondoni, A. Massi, P. Nanni, A. Roda, Chem. Eur. J. 2009, 15,
11444 – 11449.
[26] A. Kawamura, A. Tumber, N. R. Rose, O. N. King, M. Daniel, U.
Oppermann, T. D. Heightman, C. Schofield, Anal. Biochem.
2010, 404, 86 – 93.
.
[1] R. J. Klose, E. M. Kallin, Y. Zhang, Nat. Rev. Genet. 2006, 7,
715 – 727.
1634
www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 1631 –1634