Bioorganic & Medicinal Chemistry Letters 24 (2014) 204–208

Contents lists available at ScienceDirect

Bioorganic & Medicinal Chemistry Letters

journal homepage: www.elsevier .com/ locate/bmcl

Correlation between chemotype-dependent binding conformationsof HSP90a/b and isoform selectivity—Implications for thestructure-based design of HSP90a/b selective inhibitors for treatingneurodegenerative diseases

0960-894X/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.bmcl.2013.11.036

⇑ Corresponding author. Tel.: +1 858 404 8409.E-mail address: Justin_Ernst@vrtx.com (J.T. Ernst).

Justin T. Ernst a,⇑, Michael Liu b, Harmon Zuccola c, Timothy Neubert a, Kevin Beaumont b, Amy Turnbull b,Adam Kallel a, Bryan Vought c, Dean Stamos a

a Vertex Pharmaceuticals, Department of Chemistry and Drug Innovation, 11010 Torreyana Road, San Diego, CA 92121, United Statesb Vertex Pharmaceuticals, Department of Biology, 11010 Torreyana Road, San Diego, CA 92121, United Statesc Vertex Pharmaceuticals, Department of Protein Sciences, 130 Waverly Street, Cambridge, MA 02139, United States

a r t i c l e i n f o a b s t r a c t

Article history:Received 11 October 2013Revised 12 November 2013Accepted 15 November 2013Available online 23 November 2013

Keywords:HSP90 inhibitorsIsoform selectivityStructure-based drug designGRP94TRAP1Huntington’s diseaseHuntingtin

HSP90 continues to be a target of interest for neurodegeneration indications. Selective knockdown of theHSP90 cytosolic isoforms a and b is sufficient to reduce mutant huntingtin protein levels in vitro. Chem-otype-dependent binding conformations of HSP90a/b appear to strongly influence isoform selectivity.The rational design of HSP90a/b inhibitors selective versus the mitochondrial (TRAP1) and endoplasmicreticulum (GRP94) isoforms offers a potential mitigating strategy for mechanism-based toxicities. Bettertolerated HSP90 inhibitors would be attractive for targeting chronic neurodegenerative diseases such asHuntington’s disease.

� 2013 Elsevier Ltd. All rights reserved.

1

HSP90 (90 kDa heat shock protein) is a key regulator of proteinfolding, playing an important role in the folding of newly synthe-sized proteins and the refolding or stabilizing of proteins that havebecome denatured due to cellular stress.1 The structure of HSP90contains 3 domains:2 a C-terminal domain which contains anATP-binding site thought to regulate HSP90 homodimerization, acharged middle linker region, and an N-terminal domain whichcontains an ATP-binding pocket (with ATP-ase activity) whichregulates the turnover of folded client proteins. Both the C- andN-terminal ATP binding sites have been explored as the targetfor inhibitors, however, the majority of HSP90 programs haveconcentrated on designing inhibitors aimed at the N-terminalATP-binding site.3 All of the HSP90 inhibitors currently in humanclinical studies are N-terminal inhibitors and to date all of thesestudies have been limited to treating oncology indications.4,5

Despite a strong rationale for HSP90 as a target of interest for neu-rodegeneration indications,6–8 poor solubility, low brain exposure,and modest tolerability profiles have limited the use of HSP90inhibitors in this area. HSP90 is reported to have greater than

200 client proteins; therefore the lack of selectivity in controllingclient protein degradation upon inhibiting HSP90 ATP-ase activityleads to potential mechanism-based toxicities. The apparent lack ofselectivity of this mechanism of action makes the use of currentHSP90 inhibitors less attractive for treating chronic indications.

It was recently reported that HSP90 inhibitors are effective inpromoting the clearance of mutant huntingtin protein (mHtt).8

Reducing levels of mHtt in the CNS of patients carrying the mutantallele is expected to slow or stop the progression of Huntington’sdisease (HD). Based on this finding we initiated an effort to identifyHSP90 inhibitors that would be suitable for a neurodegenerationindication such as HD. A critical element of the program was to ex-plore potential strategies that would address reducing the toxicityof HSP90 inhibitors and enable chronic administration.

This Letter describes one toxicity-mitigating strategy that in-volves selectively targeting specific isoforms of HSP90. HSP90 has4 known isoforms: inducible (a), constitutive (b), mitochondria(HSP75/TRAP1), and endoplasmic reticulum (GRP94).9,10 SiRNAknockdown studies indicated that targeting the cytosolic formsof HSP90 (a and b) is sufficientto reduce mHtt levels in vitro(Fig. 1).11 By selectively targeting the cytosolic forms of HSP90one may achieve sufficient clearance of mHtt and reduced toxicity

0

25

50

75

100

125

150

Non-TargetsiRNA

HSP90αsiRNA

HSP90βsiRNA

HSP90α/βsiRNA

p = 0.02p = 0.03

p = 0.004

mHttlevels

Figure 1. Effects of RNA silencing of the HSP90a and b isoforms on mutanthuntingtin protein levels in vitro (data generated using a full-length mHttexpressing HEK cell line11).

J. T. Ernst et al. / Bioorg. Med. Chem. Lett. 24 (2014) 204–208 205

by not affecting TRAP1 and GRP94 activities, which are critical formitochondria and ER viability, respectively. Based on this hypoth-esis we initiated an effort to explore the feasibility of identifyingHSP90a/b selective inhibitors. A review of the literature indicatedthat reported data for the isoform selectivity profiles of HSP90inhibitors are limited.12–14 In fact, no precedent was identifiedfor the rational design of HSP90a/b isoform specific inhibitors.Very recently, however, several publications have emerged thatdescribe the structure-based design of GRP94 selective inhibi-tors.15–17

An analysis of the primary sequences of the four HSP90 iso-forms indicates a high degree of homology, especially betweenHSP90a and b (93%, 59%, and 47% similarity for HSP90b, GRP94,and TRAP1 versus HSP90a, respectively).9 The key residues in theN-terminal active site of HSP90a reported5 for binding ATP-com-petitive inhibitors are heavily conserved across all four isoforms.9

In fact, only two residues, W162 and K112, are different and areY and R in TRAP1, respectively. However, in each case these aminoacid changes are subtle in regards to the binding interactions in-volved (hydrophobic interaction with W162 vs a Y in the case ofBIIB021 and SNX analogs; hydrogen bonding interaction withK112 vs an R in the case of 17-AAG analogs). The conservation ofkey binding residues was further confirmed by analyzing reportedor internally generated X-ray crystal structures of Apo and ligand-bound N-terminal HSP90a,b, and GRP94 proteins (no X-ray struc-tures exist for TRAP1).18–25 Given the similarities of the bindingsites, exploiting amino acid differences appears to be a challengingapproach to designing isoform selective compounds.

It has been previously shown that there is conformational flex-ibility of the sequence 104–111 in the N-terminal ATP bindingpocket of HSP90a.5 As shown in Figure 2 this region of the proteinis observed to be in a closed loop conformation in reported X-raystructures for Apo HSP90a.20,24 This sequence may adopt differentconformations depending on the chemotype of the ligand bound.Figure 2 shows the conformational changes to this area of theprotein resulting from the binding of chemically diverse HSP90inhibitors. Two classes of compounds identified as binding an ex-tended a-helix conformation in this region of the protein are thepurine-based inhibitors (i.e., BIIB021)25 and the benzamides (i.e.,SNX2112). These two classes of compounds take advantage of aunique arrangement of hydrophobic residues, including L107,I110, A111, F138, and W139, in the extended a-helix conformationto bind to HSP90a. In the case of the purine inhibitors, this mode ofbinding has also been shown with HSP90b.20 The benzamide seriesalso takes advantage of a key hydrogen bond in this region formedbetween the ketone of the indolone or indazolone motif and theside chain of Y139. Figure 3 illustrates these key interactions usingSNX0723 as an example. Other inhibitors bind to an open loop con-formation of HSP90a, such as the ansamycin derivative 17-DMAG(X-ray of 17-DMAG-bound HSP90a19 shown in panel C in Fig. 2).Inhibitors that do not strongly interact with the 104–111 regionof the protein, such as the bisaryl–isoxazole analog AUY922, bind

to a conformation that is very similar to Apo HSP90a (AUY922-bound structure18 shown in panel B in Fig. 2).

NH2O

OH

HN

O

NN

FFF

O

N

NH2

NCl

N N

N

O N O

HNHO

NO

NH2OHN

O

N

OF

HO

O

O

OOH O

O

NH

HN

N

OO

NH2

N

N S O

HN

H2N

Cl

Cl

ON

BIIB021 AUY922

SNX2112 SNX0723

17-DMAG

BEP800

Given the conformational dependence of the sequence 104–111in binding certain inhibitors to the N-terminal ATP binding pocketof the HSP90a and b isoforms, we hypothesized that the GRP94 andTRAP1 isoforms would need a similar a-helix conformation in thisregion to accommodate the preferred binding mode of these com-pounds. An analysis of the primary sequences for the four HSP90isoforms in the residue 104–111 region indicates a high degreeof homology.9 This sequence is identical for the HSP90a and b iso-forms (INNLGTIA). Only the first two amino acids of this sequenceare different for GRP94 (VK) and TRAP1 (VS) versus HSP90a and b.The side chains of these residues face away from the ATP-bindingpocket in the extended a-helix conformation and do not appearto directly interact with ligands that bind to this conformation.However, the Apo structure of canine GRP9421 (98% identical tohuman) reveals that there is a 5 amino acid insertion (QEDGQ) atposition 126 (HSP90a numbering) versus HSP90a/b that resultsin a slight variation of the tertiary structure of helices 1, 4, and 5which are adjacent to residues 104–111 (Fig. 4). Primary sequencedata indicate that an analogous two amino acid insertion (AE) ispresent at position 129 (HSP90a numbering) in TRAP1. It washypothesized that these structural differences may potentially pre-clude GRP94 and TRAP1 from accommodating the extended a-he-lix conformation in residues 104–111.

To test this hypothesis a set of structurally diverse, N-terminalHSP90 inhibitor chemotypes that bind to different conformationsof HSP90a was selected and analyzed in a set of isoform-specificfluorescence polarization-based (FP) binding assays.26 These com-pounds represent key HSP90 inhibitors currently in clinical studies(BIIB021, AUY922, SNX2112, 17-DMAG) or have reported brainexposure in preclinical models (SNX0723). The results of this anal-ysis are shown in Table 1. Several of the compounds show minimalto moderate selectivity versus GRP94 (3 to 100-fold) and TRAP1 (8to 300-fold). Interestingly, the compounds that utilize the a-helicalbinding conformation of HSP90a (BIIB021 and the SNX analogs)possess the greatest degree of selectivity versus both GRP94 andTRAP1. As stated earlier, there was no indication from the litera-ture that these chemotypes were rationally designed or optimizedfor HSP90ab isoform selectivity.

AUY922, which binds to the closed loop (Apo-like) conforma-tion, shows very little discrimination between the four isoforms.It should also be noted that recently the thienopyrimidine HSP90inhibitor BEP800 was reported to have improved HSP90a/b iso-form selectivity versus AUY922.14 The X-ray structure of BEP800bound to HSP90a has been previously disclosed27 and the residue

QEDGQ insertion

H1H4

H5

Figure 4. Overlay of X-ray crystal structures of Apo GRP94 (green; PDB 1YT2) andHSP90a (grey; PDB 1YER). The insertion sequence QEDGQ in GRP94 that influencesthe tertiary structure of helices 1, 4, and 5 is highlighted in red.

CBA

ED F

Figure 2. X-ray crystal structures of the N-terminal domain of Apo HSP90a (A; PDB 1YER) and HSP90a bound to AUY922 (B; 2VCI), 17-DMAG (C; 1OSF), BIIB021 (D; 3QDD),SNX2112 (E; 4NH7), and SNX0723 (F; 4NH8). Residues 104–111 of HSP90a are highlighted in red.

D93

T184M98 W162

L107 I110

Y139F138

V136

A111

Figure 3. X-ray crystal structure of SNX0723-bound HSP90a (PDB 4NH8). Keybinding residues are highlighted in blue.

Table 1Isoform selectivity profilesa of literature HSP90 inhibitors and the conformation ofresidues 104–111 in the ligand-bound structure of HSP90a

Compound Ki HSP90a(nM)

Ki HSP90b(nM)

Ki GRP94(nM)

Ki TRAP1(nM)

104–111ConformationHSP90a

BIIB021 2 4 176 62 a-HelixSNX2112 4 6 484 791 a-HelixSNX0723 3 4 375 1195 a-HelixAUY922 2 2 4 16 Closed loop17-DMAG 2 2 5 326 Open loop

a Ki values are means of duplicate experiments generated from isoform specificFP-based binding assays.

206 J. T. Ernst et al. / Bioorg. Med. Chem. Lett. 24 (2014) 204–208

104–111 sequence adopts the extended a-helix conformation,which is in line with our proposed hypothesis. 17-DMAG, whichbinds to an open loop conformation, demonstrates moderate selec-tivity versus TRAP1 (160-fold) but not GRP94 (3-fold). These data

indicate that despite the high level of homology of key residuesin the binding pockets of the four HSP90 isoforms, selectivity ver-sus GRP94 and TRAP1 can be achieved by taking advantage of a dis-tinct binding conformation of HSP90a and b that may not beenergetically favorable in the other isoforms.

An X-ray crystal structure was obtained of SNX0723 bound tohuman GRP94 (Fig. 5). This structure supports the hypothesis thatprotein conformation is a primary driver of selectivity. All of thekey binding residues outside of the sequence 104–111 are highlyconserved in space relative to HSP90a. However, the 104–111 se-quence remained unstructured and did not adopt the a-helix con-formation as observed in the case of SNX0723-bound HSP90a.Given the key role of a-helix residues L107, I110, and A111 inthe binding mode of SNX0723 (Fig. 3) it is quite plausible thatthe loss or partial loss of these key interactions can account forthe >100-fold selectivity versus GRP94.

The X-ray structure of SNX0723-bound HSP90a reveals that the3-Me group of the tetrahydroindolone moiety is directed towards ahydrophobic pocket that is formed from a-helix residues I110 andA111 as well as V136 from an adjacent loop to helix 5 (Figs. 3 and5). Removal of the 3-Me group results in an analog that is lessselective versus GRP94 (20-fold vs 100-fold in the case ofSNX0723) and less potent for HSP90a/b (40-fold), further illustrat-ing the importance of the a-helix conformation of sequence104–111 for HSP90a/b selectivity and potency (Fig. 6). Furtheroptimization of inhibitors, taking advantage of interactions with

D93

T184 M98 W162L107

I110

Y139F138V136

A111

Figure 5. Overlay of X-ray crystal structures of SNX-0723 bound to GRP94 (green with green ligand; PDB 4NH9) and to HSP90a (grey with pink ligand; PDB 4NH8). Keybinding residues are highlighted (blue in HSP90a and red in GRP94; HSP90a numbering used).

HN

O

N

R

OF

OH2N

125

4

Ki HSP90β(nM)

125

3

Ki HSP90α(nM)

2650H

375Me

Ki Grp94 (nM)

R

125

4

Ki HSP90β(nM)

125

3

Ki HSP90α(nM)

2650H

375Me

Ki Grp94 (nM)

R

Figure 6. Effect of 3-methyl group on HSP90a/b and GRP94 binding affinities.

J. T. Ernst et al. / Bioorg. Med. Chem. Lett. 24 (2014) 204–208 207

the residue 104–111 extended a-helix conformation, should leadto increased selectivity versus GRP94 and potentially TRAP1. Struc-ture–activity relationships further illustrating this point will bedisclosed in the near future.

In summary, an HSP90a/b isoform selectivity strategy to miti-gate potential mechanism-based toxicities of HSP90 inhibitorswas discussed. Chemotype-dependent binding conformations ofHSP90a/b in the region 104–111 appear to strongly influenceisoform selectivity. Optimization of HSP90a/b selective inhibitorstargeting HD and other neurodegeneration indications will bereported on in due course.

Acknowledgments

The Berkeley Center for Structural Biology is supported in partby the National Institutes of Health, National Institute of GeneralMedical Sciences, and the Howard Hughes Medical Institute. TheAdvanced Light Source is supported by the Director, Office ofScience, Office of Basic Energy Sciences, of the U.S. Department ofEnergy under Contract No. DE-AC02-05CH11231.

Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.bmcl.2013.11.036.

References and notes

1. Pratt, W. B.; Toft, D. O. Exp. Biol. Med. 2003, 228, 111.2. Pearl, L. H.; Prodromou, C. Adv. Protein Chem. 2001, 59, 157.

3. Drysdale, M. J.; Brough, P. A. Curr. Top. Med. Chem. 2008, 8, 859.4. Kim, Y. S.; Alarcon, S. V.; Lee, S.; Lee, M. J.; Giaccone, G.; Neckers, L.; Trepel, J. B.

Curr. Top. Med. Chem. 2009, 9, 1479.5. Biamonte, M. A.; Water, R. W.; Arndt, J. W.; Scannevin, R. H.; Perret, D.; Lee, W.

J. Med. Chem. 2010, 53, 3.6. Waza, M.; Adachi, H.; Katsuno, M.; Minamiyama, M.; Tanaka, F.; Doyu, M.;

Sobue, G. J. Mol. Med. 2006, 84, 635.7. Sloan, L. A.; Fillmore, M. C.; Churcher, I. Curr. Opin. Drug Discov. Devel. 2009, 12,

666.8. Baldo, B.; Weiss, A.; Parker, C. N.; Bibel, M.; Paganetti, P.; Kaupmann, K. J. Biol.

Chem. 2012, 287, 1406.9. Chen, B.; Piel, W. H.; Gui, L.; Bruford, E.; Monteiro, A. Genomics 2005, 86, 627.

10. Sreedhar, A. S.; Kalmar, E.; Scermely, P.; Shen, Y. FEBS Lett. 2004, 562, 11.11. HEK293 cells expressing inducible full-length Q73-htt protein were treated

with siRNA selective for either HSP90a, HSP90b, or both. Followingconfirmation of selective HSP90 knockdown by western blot analysis, theprotein levels of Q73 htt was determined by western blotting (mAB2166Millipore). Knockdown with either HSP90a or HSP90b resulted in a significantreduction of Q73 htt protein. Knockdown of both HSP90a and HSP90b resultedin an additive effect on reducing Q73 htt protein levels.

12. Eccles, S. A.; Massey, A.; Raynaud, F. I.; Sharp, S. Y.; Box, G.; Valenti, M.;Patterson, L.; de Haven Brandon, A.; Gowan, S.; Boxall, F.; Aherne, W.;Rowlands, M.; Hayes, A.; Martins, V.; Urban, F.; Boxall, K.; Prodromou, C.; Pearl,L.; James, K.; Matthews, T. P.; Cheung, K.-M.; Kalusa, A.; Jones, K.; McDonald, E.;Barril, X.; Brough, P. A.; Cansfield, J. E.; Dymock, B.; Drysdale, M. J.; Finch, H.;Howes, R.; Hubbard, R. E.; Surgenor, A.; Webb, P.; Wood, M.; Wright, L.;Workman, P. Cancer Res. 2008, 68, 2850.

13. Chandarlapaty, S.; Sawai, A.; Ye, Q.; Scott, A.; Silinski, M.; Huang, K.; Fadden, P.;Partdridge, J.; Hall, S.; Steed, P.; Norton, L.; Rosen, N.; Solit, D. B. Clin. Cancer Res.2008, 14, 240.

14. Massey, A. J.; Schoepfer, J.; Brough, P. A.; Brueggen, J.; Chene, P.; Drysdale, M. J.;Pfaar, U.; Radimerski, T.; Ruetz, S.; Schweitzer, A.; Wood, M.; Garcia-Echeverria, C.; Jensen, M. R. Mol. Cancer Ther. 2010, 9, 906.

15. Patel, P. D.; Yan, P.; Seidler, P. M.; Patel, H. J.; Sun, W.; Yang, C.; Que, N. S.;Taldone, T.; Finotti, P.; Stephani, R. A.; Gewirth, D. T.; Chiosis, G. Nat. Chem. Biol.2013, 9, 677.

16. Taldone, T.; Patel, P. D.; Patel, M.; Patel, H. J.; Evans, C. E.; Rodina, A.; Ochiana,S.; Shah, S. K.; Uddin, M.; Gewirth, D.; Chiosis, G. J. Med. Chem. 2013, 56, 6803.

17. Duerfeldt, A. S.; Peterson, L. B.; Maynard, J. C.; Ng, C. L.; Eletto, D.; Ostrovsky, O.;Shinogle, H. E.; Moore, D. S.; Argon, Y.; Nicchitta, C. V.; Blagg, B. S. J. J. Am. Chem.Soc. 2012, 134, 9796.

18. Brough, P. A.; Aherne, W.; Barril, X.; Borgognoni, J.; Boxall, K.; Cansfield, J. E.;Cheung, K.-M. J.; Collins, I.; Davies, N. G. M.; Drysdale, M. J.; Dymock, B.; Eccles,S. A.; Finch, H.; Fink, A.; Hayes, A.; Howes, R.; Hubbard, R. E.; James, K.; Jordan,A. M.; Lockie, A.; Martins, V.; Massey, A.; Matthews, T. P.; McDonald, E.;Northfield, C. J.; Pearl, L. H.; Prodromou, C.; Ray, S.; Raynaud, F. I.; Roughley, S.D.; Sharp, S. Y.; Surgenor, A.; Walmsley, D. L.; Webb, P.; Wood, M.; Workman,P.; Wright, L. J. Med. Chem. 2008, 51, 196.

19. Jez, J. M.; Chen, J.; Rastelli, G.; Stroud, R. M.; Santi, D. V. Chem. Biol. 2003, 10,361.

20. Wright, L.; Barril, X.; Dymock, B.; Sheridan, L.; Surgenor, A.; Beswick, M.;Drysdale, M.; Collier, A.; Massey, A.; Davies, N.; Fink, A.; Fromont, C.; Aherne,W.; Boxall, K.; Sharp, S.; Workman, P.; Hubbard, R. E. Chem. Biol. 2004, 11, 775.

21. Dollins, E. D.; Immormino, R. M.; Gewirth, D. T. J. Biol. Chem. 2005, 280, 30438.22. Soldano, K. L.; Jivan, A.; Nicchitta, C. V.; Gewirth, D. T. J. Biol. Chem. 2003, 278,

48330.23. Dollins, D. E.; Warren, J. J.; Immormino, R. M.; Gewirth, D. T. Mol. Cell 2007, 28,

41.24. Stebbins, C. E.; Russo, A. A.; Schneider, C.; Rosen, N.; Hartl, F. U.; Pavletich, N. P.

Cell 1997, 89, 239.25. Shi, J.; Van de Water, R.; Hong, K.; Lamer, R. B.; Weichert, K. W.; Sandoval, C.

M.; Kasibhatla, S. R.; Boehm, M. F.; Chao, J.; Lundgren, K.; Timple, N.; Lough, R.;

208 J. T. Ernst et al. / Bioorg. Med. Chem. Lett. 24 (2014) 204–208

Ibanez, G.; Boykin, C.; Burrows, F. J.; Kehry, M. R.; Yun, T. J.; Harning, E. K.;Ambrose, C.; Thompson, J.; Bixler, S. A.; Dunah, A.; Snodgrass-Belt, P.; Arndt, J.;Enyedy, I. J.; Li, P.; Hong, V. S.; McKenzie, A.; Biamonte, M. A. J. Med. Chem.2012, 55, 7786.

26. FP binding assay methods were adapted from Llauger-Bufi, L.; Felts, S.J.; Huezo,H.; Rosen, N.; Chiosis, G. Bioorg. Med. Chem. Lett. 2003, 3975. Human HSP90a(Enzo), HSP90b (Enzo), GRP94 (Prospec-Tany Technogene) or TRAP1(recombinantly synthesized at Vertex) were incubated at 10–100 nM with5 nM FITC-geldanamycin (Enzo) and varying concentrations of HSP90inhibitors in 50 mM KCl, 5 mM MgCl2, 20 mM HEPES, pH 7.4 containing2 mM DTT and 0.1% bovine gamma-globulin. Following 3 h incubation at 23 �C,

FP was measured in an Envision Reader (Perkin-Elmer). Ki values werecalculated from IC50s by the method of Nikolovska-Coleska, Z.; Wang, R.; Fang,X.; Pan, H.; Tomita, Y.; Li, P.; Roller, P.P.; Krajewski, K.; Saito, N. G.; Stuckey, J.A.; Wang, S. Anal. Biochem., 2004, 332, 261.

27. Brough, P. A.; Barril, X.; Borgognoni, J.; Chene, P.; Davies, N. G. M.; Davis, B.;Drysdale, M. J.; Dymock, B.; Eccles, S. A.; Garcia-Echeverria, C.; Fromont, C.;Hayes, A.; Hubbard, R. E.; Jordan, A. M.; Jensen, M. R.; Massey, A.; Merrett, A.;Padfield, A.; Parsons, R.; Radimerski, T.; Raynaud, F. I.; Robertson, A.; Roughley,S. D.; Schoepfer, J.; Simmonite, H.; Sharp, S. Y.; Surgenor, A.; Valenti, M.; Walls,S.; Webb, P.; Wood, M.; Workman, P.; Wright, L. J. Med. Chem. 2009, 52, 4794.