Structure-activity relationship study of beta-carboline derivatives as haspin kinase inhibitors

Article abstract of DOI:10.1016/j.bmcl.2012.01.028

Haspin is a serine/threonine kinase that phosphorylates Thr-3 of histone H3 in mitosis that has emerged as a possible cancer therapeutic target. High throughput screening of approximately 140,000 compounds identified the beta-carbolines harmine and harmol as moderately potent haspin kinase inhibitors. Based on information obtained from a structure-activity relationship study previously conducted for an acridine series of haspin inhibitors in conjunction with in silico docking using a recently disclosed crystal structure of the kinase, harmine analogs were designed that resulted in significantly increased haspin kinase inhibitory potency. The harmine derivatives also demonstrated less activity towards DYRK2 compared to the acridine series. In vitro mouse liver microsome stability and kinase profiling of a representative member of the harmine series (42, LDN-211898) are also presented.

Full text of DOI:10.1016/j.bmcl.2012.01.028

Bioorganic & Medicinal Chemistry Letters 22 (2012) 2015–2019
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Structure–activity relationship study of beta-carboline derivatives as haspin
kinase inhibitors
Gregory D. Cuny a,b, , Natalia P. Ulyanova c, Debasis Patnaik c, Ji-Feng Liu d, Xiangjie Lin d, Ken Auerbach a,b
,
⇑
Soumya S. Ray e, Jun Xian b, Marcie A. Glicksman a,b, Ross L. Stein a,b, Jonathan M.G. Higgins c
a Laboratory for Drug Discovery in Neurodegeneration, Brigham & Women’s Hospital and Harvard Medical School, 65 Landsdowne Street, Cambridge, MA 02139, USA
b Partners Center for Drug Discovery, Brigham & Women’s Hospital and Harvard Medical School, 65 Landsdowne Street, Cambridge, MA 02139, USA
c Division of Rheumatology, Immunology and Allergy, Department of Medicine, Brigham & Women’s Hospital and Harvard Medical School, Smith Building, 5th floor,
1 Jimmy Fund Way, Boston, MA 02115, USA
d Aberjona Laboratories, Inc., 100 Cummings Center, Suite 242-F, Beverly, MA 01915, USA
e Center for Neurologic Diseases, Department of Neurology, Brigham & Women’s Hospital and Harvard Medical School, 65 Landsdowne Street, Cambridge, MA 02139, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Haspin is a serine/threonine kinase that phosphorylates Thr-3 of histone H3 in mitosis that has emerged
as a possible cancer therapeutic target. High throughput screening of approximately 140,000 compounds
identified the beta-carbolines harmine and harmol as moderately potent haspin kinase inhibitors. Based
on information obtained from a structure–activity relationship study previously conducted for an
acridine series of haspin inhibitors in conjunction with in silico docking using a recently disclosed crystal
structure of the kinase, harmine analogs were designed that resulted in significantly increased haspin
kinase inhibitory potency. The harmine derivatives also demonstrated less activity towards DYRK2 com-
pared to the acridine series. In vitro mouse liver microsome stability and kinase profiling of a represen-
tative member of the harmine series (42, LDN-211898) are also presented.
Received 19 December 2011
Accepted 10 January 2012
Available online 28 January 2012
Keywords:
Structure–activity relationship
Beta-carboline
Haspin
Kinase
Inhibitor
Ó 2012 Elsevier Ltd. All rights reserved.
The serine/threonine kinase haspin (Haploid Germ Cell-Specific
Nuclear Protein Kinase, also known as Germ Cell-Specific Gene-2;
Gsg2)1 functions in mitosis, where it phosphorylates histone H3
at Thr-3 (H3T3ph).2 During mitosis, this phosphorylation generates
a binding site on H3 for Survivin and thereby positions the
Chromosome Passenger Complex at centromeres to regulate chro-
mosome segregation,3,4 and it also displaces proteins such as TFIID
that normally bind to H3 through methylated Lys-4.5 Depletion of
haspin by RNA interference, or microinjection of H3T3ph antibod-
ies, causes chromosome alignment defects and failure of normal
mitosis.2,3,6
Human haspin has ATP-binding and catalytic sites structurally
similar to other members of the eukaryotic protein kinase (ePK)
superfamily with several notable exceptions. For example, the
highly conserved DFG motif involved in ATP-binding and the APE
motif involved in stabilizing the C-terminal lobe among ePKs are
altered or absent and the activation loop region is substantially
rearranged in haspin compared to other ePKs.7,8
CHR-6494 (1), that inhibits haspin displayed anti-tumor activity
in a mouse xenograft model.9 Also, 5-iodotubercidin (2) has been re-
ported as an effective haspin kinase inhibitor.7,10
We previously utilized a time-resolved fluorescence resonance
energy transfer (TR-FRET) high throughput screening (HTS) assay
to identify the acridine derivative 3 (LDN-192960) as another po-
tent haspin inhibitor (Fig. 1; IC50 = 0.010 l
M).11,12 This assay has
now also been used to discover the beta-carbolines harmine, 4,
and harmol, 5, as moderately potent haspin inhibitors with IC50
values of 0.59 and 0.77
been identified as an inhibitor of DYRK family kinases, with IC50
values between 0.03 and 0.35 M reported for DYRK1A, and
lM, respectively. Harmine has previously
l
approximately 50-fold lower potency toward DYRK2.13 Herein,
we describe the design, synthesis and improved potency of the
beta-carboline series for haspin inhibition utilizing the structure–
activity relationships previously determined for the acridine ser-
ies12 combined with in silico docking using a recently disclosed
crystal structure of the kinase.7 In addition, in vitro mouse liver
microsome stability and kinase profiling of a representative beta-
carboline analog are presented.
Haspin kinase inhibitors are expected to be useful probes for
elucidating the cellular roles of this protein and may have therapeu-
tic utility in treating cancer. A recently described small molecule,
A crystal structure of haspin bound to AMP was used for dock-
ing calculation.7,14 Analysis of this structure revealed key hydrogen
bonds between nitrogen atoms of the adenine ring of AMP and pro-
tein backbone atoms of residues E606 and G608 (Fig. S1).15 A 12 Å
docking grid was generated using the AMP center of mass as the
⇑
Corresponding author. Tel.: +1 617 768 8640; fax: +1 617 768 8606.
E-mail address: gcuny@rics.bwh.harvard.edu (G.D. Cuny).
0960-894X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2012.01.028

2016
G. D. Cuny et al. / Bioorg. Med. Chem. Lett. 22 (2012) 2015–2019
NH2
I
N
HN
N
N
N
OH
OH
H
N
O
N
n-Pr
N
N
OH
1
2
S
NH2
OMe
N
MeO
RO
N
H
Me
N
4 R = Me
5 R = H
3
Figure 1. Haspin inhibitors identified by radiometric, thermal stability shift and
TR-FRET HTS assays.
Figure 2. Molecular docking of 3 at the ATP-binding site of haspin. X1 and X2 are
the torsion angles that were sampled during the metadynamic calculation.
point of origin with a single hydrogen bond constraint on the back-
bone amide of G608. Docking calculations were performed on 3,
which demonstrated that this inhibitor was well accommodated
within the binding site and satisfied the hydrogen bonding
constraint on G608 (Fig. 2). In addition, the inhibitor also made a
hydrogen bond with K511, which likely disrupts a key salt bridge
between this residue and E535 that is necessary for closure of
the ATP-binding cleft enabling kinase activity. A metadynamic
simulation sampling the two torsion angles (X1 and X2) of the
alkylamine as collective variables was also conducted. One low
energy conformation was found (Fig. S2) that allowed a hydrogen
bond between the amine and D611 (Fig. 2).
Next, docking calculations were performed on four harmine ana-
logs (7a–b and 9a–b) that incorporate at two different positions
alkylamines similar to that present in 3. The two derivatives with
the alkylamine on the N9-position (9a and 9b) were well accommo-
dated within the binding site making a hydrogen bond with D687 as
well as the hydrogen bonding constraint on G608 (Fig. 3B and C). In
contrast, compounds 7a and 7b resulted in steric hindrance with the
region around F607 and G608 and were not well situated within the
ATP-binding site, suggesting that they would unlikely be haspin
Figure 3. (A) Free energy surface map from metadynamic calculations sampling
two torsion angles (X1 and X2) of 9a. Blue and red represent low and high energy
conformations, respectively. (B and C) Docking of 9a at the ATP-binding site of
haspin for the two conformations of the alkylamine side-chain based on metady-
namic calculations.
inhibitors. A similar metadynamic simulation for 9a sampling two
torsion angles (X1 and X2) of the alkylamine as collective variables
found two low energy conformations (Fig. 3A). One of these allowed
a hydrogen bond between the amine and D611 (Fig. 3B), similar to 3.
However, the second conformation permitted a hydrogen bond

G. D. Cuny et al. / Bioorg. Med. Chem. Lett. 22 (2012) 2015–2019
2017
R
NO2
NH2
R2
R2
R2
N
H
a
[ ]n
N
N
HO
a or
b
O
c
N
H
N
H
N
H
11
N
H
N
H
12
Me
Me
b
7a n = 1, R = H
7b n = 2, R = H
6a n = 1, R = Boc
6b n = 2, R = Boc
5
10
R2
R2
e
f
d
NH
R1
N
c
N
N
H
N
H
N
MeO
R1
MeO
N
N
[ ]n
Me
H
Me
13
14
R2
R2
4
R
g
N
N
d
9a n = 2, R = NH2
9b n = 3, R = NH2
8a n = 2, R = Phth
8b n = 3, R = Phth
N
[ ]n
N
[ ]n
R1
R1
O
Scheme 1. Reagents and conditions: (a) Br(CH2)nNHBoc, Cs2CO3, DMF, rt; (b) HCl,
MeOH, rt; (c) Br(CH2)nNPhth, NaH, DMF, rt; (d) NH2NH2ÁH2O, EtOH, 65 °C.
Phth = phthalimide.
N
H2N
O
15
16
Scheme 2. Reagents and conditions: (a) Me2NCH@CHNO2, TFA, rt, 30 min; (b)
POCl3, DMF, 0 °C to rt then MeNO2, NH4OAc, 100 °C, 1 h; (c) LiAlH4, THF, rt; (d)
R1CHO or CF3CHOH(OEt), MeOH, cat. HCl, rt; (e) MnO2, 5% Pd/C, DMF, microwave
(MW), 150 °C; (f) Br(CH2)nNPhth, NaH, DMF, rt; (g) NH2NH2ÁH2O, MeOH, DCM.
between the amine and the backbone carbonyl of I490 in the
glycine-rich loop (Fig. 3C).
The initial beta-carboline derivatives used in the in silico dock-
ing experiments were prepared using the procedure outlined in
Scheme 1. Harmol, 5, was O-alkylated with N-Boc-protected alkyl-
amines in the presence of cesium carbonate to give 6a and 6b. The
carbamate protecting group was removed under acidic conditions
to give 7a and 7b. Harmine, 4, could be N-alkylated. However,
phthalimide-protected alkylamines were required and a stronger
base (i.e. NaH) was also needed to give 8a and 8b. Removal of
the protecting group was accomplished using hydrazine hydrate
to yield 9a and 9b.
anticipated 7a and 7b were inactive haspin inhibitors at 10
However, 9a and 9b were both active demonstrating IC50 values
of 0.46 and 0.17 M, respectively (Table 1).
lM.
l
Based on these results additional beta-carboline analogs were
prepared using the general procedure outlined in Scheme 2. In-
doles 10 were either directly converted to derivative 11 using
Me2NCH@CHNO2 (generated from MeNO2 and Me2NCH(OMe)2 at
85 °C for 0.5 h) or in a two step procedure (i.e. Vilsmeier–Haak fol-
lowed by Henry reactions). The alkene and nitro groups were both
reduced in the presence of lithium aluminum hydride (LAH) to give
12. A Pictet–Spengler reaction between 12 and various aldehydes
generated 11. In the case where R1 = CF3, the hemiacetal
CF3CHOH(OEt) was used. Oxidation of the tetrahydro-beta-carbo-
line with manganese dioxide gave 14. Finally as previous outlined,
alkylation generated 15, which was de-protected to give 16.
Several other beta-carboline derivatives were also prepared
using the procedures outlined in Scheme 3. Intermediate 17 was
de-methylated using HBr in acetic acid to give phenol 18. This
material was converted to triflate 17 and then a methylsulfona-
mide was introduced utilizing a Pd-catalyzed reaction to yield
20. Removal of the protecting group gave amine 21. Likewise,
intermediate 16 was converted to the t-butyl ether 22 in the
presence of Me2NCH(O-t-Bu)2, which was again de-protected to
liberate amine 23. Intermediate 17 was also de-protected to give
24, which was converted to the tertiary amine 25 through
reductive amination. Finally, intermediate 24 was converted to
the secondary amine 27 via the formamide 26.
The additional compounds prepared were used to further ex-
plore the structure–activity relationship of 9b for haspin inhibition
(Table 1). Removal of the methyl at the 1-position (28) or replace-
ment with an isopropyl (29) were detrimental, where as replace-
ment with an ethyl (30) was equivalent. Replacing the methoxy
in the 7-position with a hydroxyl (31) was tolerated, but a fluorine
(32), methyl sulfonamide (33) or t-butoxy (34) were detrimental.
Transposition of the methoxy to the 5- or 6-positions (35 and 36)
resulted in loss of activity, while introduction of the methoxy to
the 8-position (37) was more tolerated. The 6,7- and 7,8-disubsti-
tuted analogs 38 and 39 demonstrated less activity. The primary
amine also appeared to be optimal, where the secondary (40)
Haspin kinase inhibitory activity for the various compounds
was assessed using the same assay utilized for the HTS, except in
the presence of varying concentrations of test compounds.11,15 As
Table 1
IC50 determinations for haspin and DYRK2 inhibition
5
R2
N
N
8
R1
[ ]n
R3
Compd
R1
R2
R3
n
IC50
Haspin
(l
M)
DYRK2
9a
9b
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Me
Me
H
i-Pr
Et
7-OMe
7-OMe
7-OMe
7-OMe
7-OMe
7-OH
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NH2
NHMe
NMe2
NH2
NH2
2
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
0.46
0.17
0.30
0.46
0.19
0.16
0.58
0.69
4.6
>10
>10
>10
>10
>10
>10
>10
>10
>10
7.0
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
Me
CF3
CF3
7-F
7-NHSO2Me
7-O-t-Bu
5-OMe
6-OMe
8-OMe
6,7-OCH2O
7-OMe, 8-Cl
7-OMe
7-OMe
7-OMe
7-OH
2.7
2.4
4.9
>10
4.8
0.22
0.50
0.83
0.34
0.47
0.10
0.22
8.0
>10
>10
15
8.5

2018
G. D. Cuny et al. / Bioorg. Med. Chem. Lett. 22 (2012) 2015–2019
N
N
MeO
HO
N
[ ]2
N
Me
Me
[ ]2
a
O
O
N
O
N
O
O
17
18
O
O
N
S
TfO
Me
N
N
H
N
[ ]2
Me
c
b
N
[ ]2
Me
N
R
Figure 4. A Venn diagram highlights kinases selectively inhibited by 3 (red) or 42
(blue) by P90% at 10 M. The overlapping region contains only six kinases
inhibited by both compounds.
O
19
20 R = NPhth
21 R = NH2
l
d
inhibited thirteen kinases, in addition to haspin, P90%.15,17 These
kinases were CaMK2b, CaMK2d, CDK7-CycH-Mat1, cGK2, CK1d,
CLK1, CLK2, DYRK1A, DYRK1B, DYRK3, PASK, PIM1 and PKD3.
Interestingly, many of these enzymes belong to the CMGC group
of kinases, unlike haspin which is a divergent member of the ePK
family.18 In addition, a comparison of the profiles of 3 and 42 sug-
gested that only six kinases, including haspin, were inhibited by
N
N
t-BuO
t-BuO
N
N
[ ]2
Me
Me
d
e
[ ]2
18
O
H2N
N
O
23
both compounds P90% at 10 l
M (Fig. 4).12 Profiling of additional
22
haspin kinase inhibitors, such as 1 and 2, may further reduce the
number of kinases, besides haspin, which are known to be potently
inhibited by all the compounds. In addition, the collective use of 3,
42 and potentially other haspin inhibitors in cell based assays may
allow for more concrete conclusions to be reached with regard to
haspin’s biological functions.
In conclusion, a structure–activity relationship study of the
beta-carbolines 4 and 5, identified utilizing a recently developed
HTS assay for haspin kinase inhibitory activity, was performed
guided by insights obtained from a previously optimized compound
series12 combined with in silico docking and metadynamic calcula-
tions. Increased potency was accomplished by introduction of a
tethered primary amine onto the N9-position of the beta-carboline.
Potency was further increased by replacing the methyl at the
1-position with a trifluoromethyl giving 42. In addition, this analog
demonstrated excellent in vitro metabolic stability in pooled
mouse liver microsomes. Kinase profiling of 42 suggested that it
was fairly selective and inhibited only six kinases (P90% at
N
N
MeO
MeO
N
[ ]2
N
[ ]2
Me
f or
g
Me
d
17
R
H2N
25 R = NMe2
26 R = NHCHO
27 R = NHMe
24
h
Scheme 3. Reagents and conditions: (a) HBr, AcOH, MW, 130 °C; (b) Tf2O, i-Pr2EtN,
DCM, 0 °C to rt; (c) MeSO2NH2, Pd2(dba)3, xPhOS, K3PO4, toluene, 110 °C, 1 h; (d)
NH2NH2ÁH2O, MeOH, DCM; (e) Me2NCH(O-t-Bu)2, DMF, 120 °C, 1 h; (f) CH2O,
HCO2H, MW, 150 °C; (g) EtOCHO, MW, 160 °C; (h) LiAlH4, THF, rt.
and tertiary (41) amines were less active. Finally, in order to
potentially improve metabolic stability, analogs (42 and 43) that
replace the methyl group in the 1-position with a trifluoromethyl
were evaluated. Gratifyingly, 42 displayed an IC50 value of
100 nM for inhibiting haspin. This compound was also assessed
for in vitro metabolic stability in pooled mouse liver microsomes
and demonstrated excellent stability with a t1/2 of 341 min and a
10 lM), including haspin, in common with the previously identi-
fied acridine inhibitor 3. The beta-carboline haspin inhibitor 42
(LDN-211898) described herein, along with other structurally
distinct inhibitors such as 1–3 provide valuable molecular probes
to study the cellular functions of haspin kinase and may have
potential therapeutic utility in treating cancer.
CLint of 3.8 lL/min/mg protein suggesting that this compound
may be useful as an in vivo probe.15,16
Acknowledgments
The beta-carboline series was also evaluated for inhibition of
human DYRK2 using a previously described assay.12 Harmine and
The authors thank Partners Healthcare for financial support.
This work was also supported in part by NIH Grant
R01CA122608 and a Leukemia and Lymphoma Society Scholar
Award to J.M.G.H.
harmol gave IC50 values for DYRK2 inhibition of 0.69 and 1.5 lM,
consistent with previous studies.13 Introduction of the tethered
amine onto the N9-position substantially reduced the potency for
DYRK2 inhibition in all cases examined (Table 1). For example,
42 had an IC50 value for DYRK2 of 15
fold selectivity for haspin over DYRK2.
Lastly, haspin inhibitor 42 was assessed against a panel of 292
kinases at 10 M. At this high concentration, the compound
l
M and demonstrated 150-
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2012.01.028.
l

G. D. Cuny et al. / Bioorg. Med. Chem. Lett. 22 (2012) 2015–2019
2019
10. Balzano, D.; Santaguida, S.; Musacchio, A.; Villa, F. Chem. Biol. 2011, 18, 966.
11. Patnaik, D.; Xian, J.; Glicksman, M. A.; Cuny, G. D.; Stein, R. L.; Higgins, J. M. G. J.
Biomol. Screen. 2008, 13, 1025.
References and notes
1. (a) Tanaka, H.; Yoshimura, Y.; Nishina, Y.; Nozaki, M.; Nojima, H.; Nishimune, Y.
FEBS Lett. 1994, 355, 4; (b) Tanaka, H.; Yoshimura, Y.; Nozaki, M.; Yomogida, K.;
Tsuchida, J.; Tosaka, Y.; Habu, T.; Nakanishi, T.; Okada, M.; Nojima, H.;
Nishimune, Y. J. Biol. Chem. 1999, 274, 17049.
2. (a) Dai, J.; Sultan, S.; Taylor, S. S.; Higgins, J. M. G. Gene Dev. 2005, 19, 472; (b)
Higgins, J. M. G. Chromosoma 2010, 119, 137.
3. Wang, F.; Dai, J.; Daum, J. R.; Niedzialkowska, E.; Banerjee, B.; Stukenberg, P. T.;
Gorbsky, G. J.; Higgins, J. M. G. Science 2010, 330, 231.
4. (a) Kelly, A. E.; Ghenoiu, C.; Xue, J. Z.; Zierhut, C.; Kimura, H.; Funabiki, H.
Science 2010, 330, 235; (b) Yamagishi, Y.; Honda, T.; Tanno, Y.; Watanabe, Y.
Science 2010, 330, 239.
5. Varier, R. A.; Outchkourov, N. S.; de Graaf, P.; van Schaik, F. M. A.; Ensing, H. J.
L.; Wang, F.; Higgins, J. M. G.; Kops, G. J. P. L.; Timmers, H. Th. M. EMBO J. 2010,
29, 3967.
6. Dai, J.; Kateneva, A. V.; Higgins, J. M. G. J. Cell Sci. 2009, 122, 4168.
7. Eswaran, J.; Patnaik, D.; Filippakopoulos, P.; Wang, F.; Stein, R. L.; Murray, J. W.;
Higgins, J. M. G.; Knapp, S. Proc. Natl. Acad. Sci. U.S.A. 2009, 48, 20198.
8. Villa, F.; Capasso, P.; Tortorici, M.; Forneris, F.; de Marco, A.; Mattevi, A.;
Musacchio, A. Proc. Natl. Acad. Sci. U.S.A. 2009, 48, 20204.
12. Cuny, G. D.; Robin, M.; Ulyanova, N. P.; Patnaik, D.; Pique, V.; Casano, G.; Liu, J.-
F.; Lin, X.; Xian, J.; Glicksman, M. A.; Stein, R. L.; Higgins, J. M. G. Bioorg. Med.
Chem. Lett. 2010, 20, 3491.
13. (a) Bain, J.; Plater, L.; Elliott, M.; Shpiro, N.; Hastie, C. J.; McLauchlan, H.;
Klevernic, I.; Arthur, J. S. C.; Alessi, D. R.; Cohen, P. Biochem. J. 2007, 408, 297;
(b) Seifert, A.; Allan, L. A.; Clarke, P. R. FEBS J. 2008, 275, 6268; (c) Göckler, N.;
Jofre, G.; Papadopoulos, C.; Soppa, U.; Tejedor, F. J.; Becker, W. FEBS J. 2009, 276,
6324; (d) Ogawa, Y.; Nonaka, Y.; Goto, T.; Ohnishi, E.; Hiramatsu, T.; Kii, I.;
Yoshida, M.; Ikura, T.; Onogi, H.; Shibuya, H.; Hosoya, T.; Ito, N.; Hagiwara, M.
Nat. Commun. 2010, 1, 86.
14. Protein Data Base (PDB) ID: 3DLZ. Docking experiments were conducted using
Glide XP v2.5 from Schrödinger Inc.
15. See Supplementary data for details.
16. Baranczewski, P.; Stan´ czak, A.; Sundberg, K.; Svensson, R.; Wallin, A.; Jannson,
J.; Garberg, P.; Postlind, H. Pharmacol. Rep. 2006, 58, 453.
17. Although the percent inhibition was conducted as a single point, this should
reflect relative potency for each kinase. However, dose–response IC50 value
determinations will ultimately be needed.
18. (a) Higgins, J. M. G. Protein Sci. 2001, 10, 1677; (b) Kannan, N.; Taylor, S. S.; Zhai,
Y.; Venter, J. C.; Manning, G. PLoS Biol. 2007, 5, e17.
9. Huertas, D.; Soler, M.; Moreto, J.; Villanueva, A.; Martinez, A.; Vidal, A.;
Charlton, M.; Moffat, D.; Patel, S.; McDermott, J.; Owen, J.; Brotherton, D.; Krige,
D.; Cuthill, S.; Esteller, M. Oncogene 2011. doi:10.1038/onc.2011.335.