Adenine derived inhibitors of the molecular chaperone HSP90 - SAR explained through multiple X-ray structures

Article abstract of DOI:10.1016/j.bmcl.2003.11.011

Multiple co-crystal structures of an adenine-based series of inhibitors bound to the molecular chaperone Hsp90 have been determined. These structures explain the observed SAR for previously described compounds and new compounds, which possess up to 8-fold

Full text of DOI:10.1016/j.bmcl.2003.11.011

Bioorganic & Medicinal Chemistry Letters 14 (2004) 325–328
Adenine derived inhibitors of the molecular chaperone
HSP90—SAR explained through multiple X-ray structures
Brian Dymock,* Xavier Barril, Mandy Beswick, Adam Collier, Nicholas Davies,
Martin Drysdale, Alexandra Fink, Christophe Fromont, Roderick E.Hubbard,
Andrew Massey, Allan Surgenor and Lisa Wright
RiboTargets Ltd, Granta Park, Cambridge, CB1 6GB, UK
Received 2 September 2003; revised 28 October 2003; accepted 7 November 2003
Abstract—Multiple co-crystal structures of an adenine-based series of inhibitors bound to the molecular chaperone Hsp90 have
been determined.These structures explain the observed SAR for previously described compounds and new compounds, which
possess up to 8-fold improved potency against the isolated enzyme.Anti-tumour cell potency and mechanism of action data is also
described for the most potent compounds.These data should enable the design of more potent Hsp90 inhibitors.
# 2003 Elsevier Ltd.All rights reserved.
1. Introduction
Hsp90 is an ATP-dependent chaperone protein essential
for the maturation and activity of a varied group of
proteins involved in signal transduction, cell cycle reg-
ulation and apoptosis.1,2 The ATPase activity can be
inhibited with some selectivity by natural product anti-
biotics such as geldanamycin or radicicol (Fig.1 ).3
Recent studies with geldanamycin derivatives suggest
that cancer cells are particularly sensitive to HSP90
inhibition.4 Further in vivo and clinical data supports
the hypothesis that the HSP90 family may be an
appropriate target for anti-cancer drug devel-
opment.5ꢀ10
Figure 1. Structures of published Hsp90 inhibitors.
Structures of the N-terminal domain (Nt-) of human
HSP90a have been published both as unliganded protein
and in complex with a variety of inhibitors.11 A seven-
stranded beta sheet forms the backbone of the protein
and four alpha helices are arranged such that they form a
compact cavity in which resides the ATP binding site.
and structure determination of a series of analogues
with enhanced enzyme inhibition.Detailed explanations
of the observed changes in activity of these and other
reported analogues of 113ꢀ15 are now possible.
Recently, Chiosis et al.described the design of a series
of purine analogues and measured binding affinity and
cellular activity.12 Here, we report the structure of Nt-
HSP90a complexed to inhibitor 1 (PU3) and the design
2. Synthesis
Scheme 1 depicts the preparation of compounds 1–12
starting from either 4,5,6-triaminopyrimidine, for 2-
position hydrogen, or 2,4,5,6-tetraaminopyrimidine, for
2-position fluoro.This route is similar to the published
synthesis of analogues of 113,16 but with modifications to
some reaction conditions.The first acylation proceeded
Keywords: HSP90; Chaperone; Adenine inhibitors; HSP70; Raf-1.
* Corresponding author.Fax: +44-1223-895556; e-mail: bd. ymock@
ribotargets.com
0960-894X/$ - see front matter # 2003 Elsevier Ltd.All rights reserved.
doi:10.1016/j.bmcl.2003.11.011

326
B. Dymock et al. / Bioorg. Med. Chem. Lett. 14 (2004) 325–328
mM).12 Despite its weak inhibition we were able to
obtain a crystal structure of 1 bound to Hsp90.
Interaction with the key conserved Aspartic acid
(Asp93) and a network of waters at the bottom of the
ATP-binding pocket is seen in the crystal structures of
the natural product inhibitors bound to Hsp90: the pri-
mary carbamate of geldanamycin and one of the
hydroxyls of radicicol donate hydrogen bonds to Asp93
and other parts of the inhibitor structures accept hydro-
gen bonds from at least three of the four waters.3 Like
ADP, 1 donates an H-bond to Asp93 through its primary
amine and interacts with the four waters as described
above (Fig.2 ).These interactions can be viewed as the
critical ‘warhead’ component of the binding mode since
any changes to these inhibitor groups abolish activity.
Scheme 1. Synthesis of purines from tri- or tetra-aminopyrimidine.
i
Reagents and conditions: (a) ArCH2CO2H, Pr2NEt, HBTU, DMF,
rt, 18 h; (b) thermal: NaOMe (15 equiv), EtOꢁH, reflux, O/N; micro-
.
wave: NaOMe (15 equiv), MeOH, mwave, 130 C, 30 min; (c) HF pyr
(10 equiv), pyr, 0 ꢁC, 5 min then tBuONO (1.5 equiv), ꢀ20 ꢁC to rt, 18
h; (d) R0X, NaOMe, DMF, 0 ꢁC, 5 min then to rt, 2 h.
most efficiently under HBTU conditions in DMF.Ring
closure could either be performed thermally with excess
methoxide or with a microwave reactor.HF in pyridine
installs the 2-fluoro substituent with no protection of
the purine required.The final alkylation step proved
sluggish and low yielding with the published Mitsunobu
conditions16 so we developed alternative alkylation
conditions, which gave good yields.17
When 1 binds Hsp90 a conformational change of the
backbone residues 104–111 occurs with the consequent
creation of a new pocket in which the phenyl ring of 1 is
bound.This lipophilic pocket is composed of Phe138,
Met98, Val150, Leu103, Leu107, Trp162 and Tyr139,
which remain in exactly the same conformation with
respect to the ADP-Hsp90 structure.The aromatic ring
of 1 is stacked between the side chains of Phe138 and
Leu107, and forms other favourable hydrophobic inter-
actions with Met98.Moreover, the methyls of the
methoxy groups make favourable hydrophobic contacts
with the aromatic rings of Trp162 and Tyr139, as well
as with the aliphatic carbons of Ala111, Leu103 and
Val150 (Fig.3 ).Such close hydrophobic interactions
lead to dramatic structure–activity relationships.
Scheme 1 requires the alkyl group to be installed in a
rather unsatisfactory final step and since the benzyl
group was of greater importance to us, we developed a
route, which not only installs the alkyl chain early in the
synthesis but also improves overall reaction yields.
Scheme 2 depicts this alternative synthesis starting from
4-6-dichloro-5-nitropyrimidine.Butylamine displace-
ment installs the butyl chain efficiently followed by
borohydride/nickel chloride reduction of the nitro
group and acylation of the resulting amine.Treatment
with ammonia in the microwave gave the target com-
pound.
The length and nature of the methylene linker between
the two ring systems is very important to provide 1 with
the appropriate shape and size to occupy both the ade-
nine binding site (Bergerat fold) and the new pocket.
Larger linkers, such as (CH2)2, result in a misplacement
of the phenyl ring, while conjugating groups, such as
nitrogen, drive the rings towards a more co-planar
orientation.
3. Medicinal chemistry
In a malachite green ATPase assay18 the IC50 of 1 was
measured at >200 mM (Table 1).This is consistent with
1 being 20-fold less active than 17AAG (IC50=13.4
A single meta-methoxy benzyl group confers sig-
nificantly better inhibitory activity than the trimethoxy
benzyl in 1 (3, IC50=75 mM) due to this methoxy group
being buried in a very hydrophobic cavity, aligned by
Met98 and Val150.Intriguingly, the single para-meth-
oxy benzyl group (2) does not inhibit enzyme activity at
Scheme 2. Synthesis of purines from 2,6-dichloro-5-nitropyrimidine.
t
i
Reagents and conditions: (a) BuNH2 (0.9 equiv) Pr2NEt (0.9 equiv),
.
THF, rt, 72 h; (b) NaBH4 (1.5 equiv), NiCl2 6H2O (2 equiv), MeOH,
rt, 1 h; (c) ArCH2CO2H (4 equiv), iPr2NEt (11 equiv), MeCN, rt 18 h;
(d) excess 0.88 NH3, mwave, 130 ꢁC, 30 min.
Figure 2. Key hydrogen bonds between 1, Asp93 and water network.

B. Dymock et al. / Bioorg. Med. Chem. Lett. 14 (2004) 325–328
Table 1. Purine analogues and enzyme inhibition data
327
Compd X R2
R3
R4
R5
R0
IC50 ATPase
(mM)a
AAG
Rd
1
2
3
4
5
6
7
8
9
10
11
12
13.4
<0.2
>200
>200
75
15.3
41
>200
>200
30
H
H
H
H
H
H
H
H
OMe
H
OMe
OMe OMe
n-Butyl
n-Butyl
n-Butyl
n-Butyl
n-Butyl
n-Butyl
OMe
H
H
H
Figure 4. Overlay of the phenyl rings of 2 (purple) and 3 (yellow) with
compound 4.The carbon atoms are displayed in green (protein) and
orange (ligand).
OCH2O bridge
H
OMe
H OMe
H
OMe
OMe
OMe
H
H
OMe OMe
H
H
F
F
F
F
F
Cl
Cl
Cl
OMe
OMe
OMe
H
OMe OMe Pent-4-ynyl
OMe OMe Pent-4-ynyl
assay and not in the more stringent functional ATPase
assay (6 and 7 are >200 mM in the ATPase assay).The
crystal structure of 7-Hsp90 shows that the chlorine
atom fits snugly making several hydrophobic contacts,
particularly with Met98 and Phe138.
H
H
H
OMe
OMe
H
n-Butyl
53.5
14.3
4.1
H
H
OMe Pent-4-ynyl
n-Butyl
OCH2O bridge
H
17.1
a All IC50 values are the average of at least two determinations.
The scope for substitution at C2 is very limited due to
the steric hindrance imposed by the backbone of Gly97.
Given the almost perfect fit of the adenine ring into its
cavity, it is unlikely that a bulky group would be toler-
ated in this position.
In our assay and as previously reported13 fluoro sub-
stitution at C2 results in improved potency.However,
we have observed a non-additive behaviour, ranging
from more than 7ꢂ for the less active compounds
(compare 7 and 8) to less than 2ꢂ in favour of the butyl
for the more active examples 12 and 4.While 4 is more
active than 5, the equivalent 2-fluoro-adenine com-
pounds (12 and 10) show very similar activities.In any
case, the analysis of the crystallographic structures
suggests that the purine ring and benzyl moieties add
binding free energy by different mechanisms, which
may or may not be complementary with specific
combinations.
Figure 3. Interactions of 1 in lipophilic pocket.
200 mM.The presence of a void in the lipophilic pocket
of the structure of 2, results in poor packing between the
ligand and its receptor.
Preference for linear and flexible substituents on N9 of
the purine14,15 ring may be explained by the presence of
Leu107 impeding the path out towards solvent (Fig.3 ).
The first few methylenes of the alkyl chain interact with
residues Leu107 and Met98, while the rest of the side
chain is completely solvent exposed explaining the rela-
tively flat SAR observed in this area.
Overlaying 3 with 2 led us to surmise that combination
of both these features into a dioxolane five-membered
ring would give an improved compound.Pleasingly, 4
had improved potency of around 5-fold over 3 (4,
IC50=15.3 mM).This compound exhibits the good fea-
tures of 3 (good packing) and 2 (water mediated hydro-
gen bonds with Tyr139, Trp162 and Leu103) (Fig.4 ).
Within the more active 2,5-dimethoxyl fluoro benzyl
series (compounds 9–11) the contribution of the 9-alkyl
group can be more clearly demonstrated.With no sub-
stitution the potency is quite weak but still measurable
at 53.5 mM (9).This compares favourably with the non-
fluoro series where 9-H compounds were not active even
at concentrations of 200 mM.When the butyl chain is
installed (10), the potency improves by around 4-fold
and by more than 14-fold with the pent-4-ynyl sub-
stituent (11).This may be due to an increased interac-
tion between the acetylene group and the hydrophobic
lip of the pocket.
Alternatively, installing an additional methoxy group in
the ortho-position of 3 led to 5 with a small improvement
in potency.The crystallographic structure of the complex
reveals a water-mediated hydrogen bond of the oxygen
of the ortho-methoxy with the nitrogen of Phe138,
which could explain the increased activity. ortho-Chloro
and bromo substitutions of the benzyl group in 1 have
been studied by Chiosis et al.14 Chloro has been repor-
ted to improve activity (compare 1 with 7) whereas the
larger bromo is approximately as active as hydrogen.14
These activity differences are only seen in a binding

328
B. Dymock et al. / Bioorg. Med. Chem. Lett. 14 (2004) 325–328
References and notes
1.Pearl, L.H;.Prodromou, C.
10, 46.
Curr. Opin. Struc. Biol. 2000,
2. Young, J. C.; Moarefi, I.; Hartl, F. U. J. Cell. Biol. 2001,
154, 267.
3. Roe, S. M.; Prodromou, C.; O’Brien, R.; Ladbury, J. E.;
Piper, P.W;. Pearl, L.H. J. Med. Chem. 1999, 42, 260.
4. Chiosis, C.; Huezo, H.; Rosen, N.; Mimnaugh, E.;
Whitesell, L.; Neckers, L. Mol. Cancer Ther. 2003, 2,
123.
5. Egorin, M. J.; Zuhowski, E. G.; Rosen, D. M.; Sentz,
D. L.; Covey, J. M.; Eiseman, J. L. Cancer Chemother.
Pharmacol. 2001, 47, 291.
6. Egorin, M.; Lagattuta, T.; Hamburger, D.; Covey, J.;
White, K.; Musser, S.; Eiseman, J. Cancer Chemother.
Pharmacol. 2002, 49, 7.
7. Soga, S.; Sharma, S.; Shiotsu, Y. S.; Makiko, T.;
Harunobu, Y.; Kazuo, I.; Yoji, M.; Chikara, T.;
Tatsuya, K.; Junichi, S.; Theodor, W.; Neckers, L. M.;
Akinaga, S. Cancer Chemother. Pharmacol. 2001, 48,
435.
Figure 5. Western blots analysis showing an increase in Hsp70 and a
decrease in c-Raf following exposure of HCT116 cells to 1- and 2-fold
GI50 concentrations of compounds 8, 10, 11 and 17AAG (control).
8. Banerji, U. O. D. A.; Scurr, M.; Benson, C.; Hanwell, J.;
Clark, S.; Raynaud, F.; Turner, A.; Walton, M.; Work-
man, P.; Judson, I. Proc. Am. Soc. Clin. Oncol. 2001, 20,
82a.
9. Munster, P. T. L.; Schwartz, L.; Larson, S.; Kenneson,
K.; De La Cruz, A.; Rosen, N.; Scher, H. Proc. Am. Soc.
Clin. Oncol. 2001, 20, 83a.
10. Wilson, R. H.; Takimoto, C. H.; Adnew, E. B.; Morrison,
G.; Grollman, F.; Thomas, R. R.; Saif, M. W.; Hopkins,
J.; Allegra, C.; Grochow, L.; Szabo, E.; Hamilton, J. M.;
4. Biology
The cellular activity of three of the most active ATPase
inhibitors was measured using a Sulforhodamine B (SRB)
growth inhibition assay on the human colon tumour cell
line, HCT116.All three compounds ( 8, 10, 11) exhibited
similar GI50 values (3.9, 3.4 and 3.3 mM, respectively).
This compares well with the GI50 described by Chiosis et
al.for 8 in MCF-7 cells.14 In order to distinguish between
generalised and HSP90 inhibitor-specific cell growth
inhibition, the upregulation of the co-chaperone HSP70
and down-regulation of the HSP90 client protein Raf-1
was measured at the cell growth inhibition GI50 value.It
has been shown previously that this change in protein
expression profile is due to specific HSP90 inhibi-
tion.18,19 When exposed to these compounds the
HCT116 cells demonstrated an increase in the levels of
HSP70 and decrease in the levels of Raf-1 (Fig.5 ).11
This indicates that cell growth inhibition is occurring by
a mechanism dependent on HSP90 inhibition.
Monahan, B.P;. Neckers, L.G;. Rem, J.L.
Soc. Clin. Oncol. 2001, 20, 82a.
Proc. Am.
11. Jez, J.; Chen, J.-H.; Rastelli, G.; Stroud, R.; Santi, D.
Chem. Biol. 2003, 10, 361.
12. Chiosis, G.; Timaul, M. N.; Lucas, B.; Munster, P. N.;
Zheng, F. F.; Sepp-Lorenzino, L.; Rosen, N. Chem. Biol.
2001, 8, 289.
13.Chiosis, G.and Rosen, N.PCT WO 02/36075, 2002;
Chem. Abstr. 2002, 136, 374828.
14. Chiosis, G.; Lucas, B.; Shtil, A.; Huezo, H.; Rosen, N.
Bioorg. Med. Chem. 2002, 10, 3555.
15. Kasibhatla, S. R.; Hong, K.; Zhang, L.; Biamonte, M. A.;
Boehm, M. F.; Shi, J.; Fan, J. PCT WO 03/037860, 2003,
Chem. Abstr. 2003, 356414.
16. Lucas, B.; Rosen, N.; Chiosis, G. J. Comb. Chem. 2001, 3,
518.
17.Data for 12: Purified by flash chromatography eluting
with CH2Cl2–2%MeOH/CH2Cl2. 1H NMR (CDCl3, 7.26)
d 0.98 (t, 3H, J=7.5 Hz), 1.39 (m, 2H), 1.68 (m, 2H), 4.05
(t, 2H, J=7.5 Hz), 4.23 (s, 2H), 5.99 (s, 2H), 6.04 (s, 2H),
6.76 (dd, 1H, J=1.5, 8.0 Hz), 6.80 (d, 1H, J=1.5 Hz),
6.86 (d, 1H, J=8.0 Hz). LC–MS retention time=2.31
min, m/z (EI): 344 (M++1).Anal.calcd for
C17H18FN5O2: C, 59.41; H, 5.24; N, 20.39. Found: C,
59.57; H, 5.34; N, 20.01.
5. Conclusion
Structure–activity relationships for a series of purine
inhibitors of the Hsp90 chaperone have been explained
by multiple X-ray structures of the ligands bound to the
target.Elucidation of the key hydrogen bonds and
hydrophobic interactions should enable more potent
inhibitors to be developed.
18. Aherne, G.; Maloney, A.; Prodromou, C.; Rowlands, M.;
Hardcastle, A.; Boxall, K.; Clarke, P.; Walton, M.; Pearl,
L.; Workman, P. Assays for HSP90 and Inhibitors. In
Methods in Molecular Medicine: Novel Anticancer Drug
Protocols; Adjei, A., Buolamwini, J., Eds.; Humana: NJ,
2003; p 149.
Acknowledgements
The authors are indebted to Joanne Wayne and Kate
Grant for enzyme assays and Adam Hold and Heather
Simmonite for analytical support.
19.Workman, P. Mol. Cancer Ther. 2003, 2, 131.