Structural basis for specificity of TGFβ family receptor small molecule inhibitors

Cellular Signalling 24 (2012) 476–483

Contents lists available at SciVerse ScienceDirect

Cellular Signalling

j ourna l homepage: www.e lsev ie r .com/ locate /ce l l s ig

Structural basis for specificity of TGFβ family receptor small molecule inhibitors

Abiodun A. Ogunjimi a, Elton Zeqiraj a, Derek F. Ceccarelli a, Frank Sicheri a,b,⁎,Jeffrey L. Wrana a,b,⁎⁎, Laurent David a,⁎⁎⁎

a Center for Systems Biology, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, 600 University Avenue, Toronto, Ontario, Canada M5G 1X5b Dept. of Molecular Genetics, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8

⁎ Correspondence to: F. Sicheri, Center for Systems BiolInstitute, Mount Sinai Hospital, 600 University Avenue,1X5. Tel.: +1 416 586 4800x2336; fax: +1 416 586 8869⁎⁎ Correspondence to: J. L.Wrana, Center for Systems Bio

Institute, Mount Sinai Hospital, 600 University Avenue, TorTel.: +1 416 586 4800x2791; fax: +1 416 586 8869.⁎⁎⁎ Correspondence to: L. David, Center for Systems BioloInstitute, Mount Sinai Hospital, 600 University Avenue,1X5. Tel.: +1 416 586 4800x2363; fax: +1 416 586 8869

E-mail addresses: [email protected] (F. Sicheri), [email protected] (L. David).

0898-6568/$ – see front matter © 2011 Elsevier Inc. Alldoi:10.1016/j.cellsig.2011.09.027

a b s t r a c t

Article history:Received 14 September 2011Accepted 24 September 2011Available online 1 October 2011

Keywords:ALK5SB431542ALK2DorsomorphinA8301

Transforming growth factor-β (TGFβ) receptor kinase inhibitors have a great therapeutic potential. SB431542 isone of the mainly used kinase inhibitors of the TGFβ/Activin pathway receptors, but needs improvement of itsEC50 (EC50=1 μM) to be translated to clinical use. A key feature of SB431542 is that it specifically targets receptorsfrom the TGFβ/Activin pathway but not the closely related receptors from the bonemorphogenic proteins (BMP)pathway. To understand the mechanisms of this selectivity, we solved the crystal structure of the TGFβ type Ireceptor (TβRI) kinase domain in complex with SB431542. We mutated TβRI residues coordinating SB431542to their counterparts in activin-receptor like kinase 2 (ALK2), a BMP receptor kinase, and tested the kinase activityof mutated TβRI.We discovered that a Ser280Thrmutation yielded a TβRI variant thatwas resistant to SB431542inhibition. Furthermore, the corresponding Thr283Ser mutation in ALK2 yielded a BMP receptor sensitive toSB431542. This demonstrated that Ser280 is the key determinant of selectivity for SB431542. This work providesa framework for optimising the SB431542 scaffold to more potent and selective inhibitors of the TGFβ/Activinpathway.

ogy, Samuel Lunenfeld ResearchToronto, Ontario, Canada M5G.logy, Samuel Lunenfeld Researchonto, Ontario, CanadaM5G 1X5.

gy, Samuel Lunenfeld ResearchToronto, Ontario, Canada [email protected] (J.L. Wrana),

rights reserved.

© 2011 Elsevier Inc. All rights reserved.

1. Introduction

TGFβ family proteins (TGFβ, BMPs, Activin/Nodal, GDFs andAMH) aresecreted morphogens that signal through heteromeric complexes oftransmembrane type I and type II serine/threonine kinase receptors toorchestrate a wide array of cellular signals in development and disease.

TGFβ family ligand–receptor interaction triggers the formation ofspecific type I and type II heterotetrameric receptor complexes. Thetype II receptors then phosphorylate the type I receptors, which inturn phosphorylate down-stream effectors, the receptor-activatedSmads (R-Smads). Phosphorylated R-Smads assemble with Smad4,accumulate in the nucleus and regulate transcription via interactionwith transcription factor partners [1]. The seven TGFβ family type Ireceptors cluster in two groups depending on the R-Smad they target.ALK4, TβRI (also referred to as ALK5) and ALK7 induce the phosphor-ylation of Smad2 and 3; this is called the “TGFβ/Activin pathway”.

ALK1, 2, 3 and 6 induce the phosphorylation of Smad1, 5 and 8, andthis is referred to as the “BMP pathway” (Fig. S1A). In these pathwaysit is the identity of the activated type I receptor that is critical in de-fining transcriptional outputs, as the two groups of R-Smads triggerdifferent, and often opposite, cellular responses.

Enhanced signalling output mediated via the intracellular kinasedomains of the TGFβ family receptors is important for a variety of path-ological processes such as cancer progression, fibrosis, aswell as parasiticinfection by Trypanosoma cruzi. Therefore, small molecule inhibitors ofTGFβ signalling that specifically target the kinase domains have greattherapeutic potentials.

Multiple inhibitors are commonly used to dissect themyriad biolog-ical and pathological functions of TGFβ family cytokines. During the firstattempt to find inhibitors for TβRI by using a TβRI kinase assay coupledto compound screening, Callahan et al. found that 2,4,5-substitutedimidazoles could be potent inhibitors [2]. This resulted in a series of2,4,5-substituted imidazoles [2–4]. Later on, virtual small moleculesscreening aimed atfindingnewmolecules targeting the TβRI ATP-bindingpocket yielded a series of 2,4,5-substitutedpyrazole inhibitors [5, 6]. 2,4,5-Substituted imidazoles and pyrazoles are the two main inhibitor familiesused to inhibit the TGFβ/Activin pathway (Fig. S1).

Recently, multiple inhibitors belonging to the substituted-imidazolegroup (SB431542 and its derivatives SB505124 or SB525334) orsubstituted-pyrazole group (LY364947 and its derivative A8301) havebeen benchmarked in a panel of 123 kinases [7]. This confirmed thatthese kinase inhibitors are specifically targeting the TGFβ- but not theBMP-pathway receptors, which makes them invaluable tools in the

Table 1Data collection, structure determination and refinement statistics for TβRI–SB431542complex.

Space group P21Unit cell a=41.8 Å, b=77.7 Å, c=90.1 Å α=β=γ=90°Number of molecules/asu 1Resolution range (Å) 30–1.7 (1.76–1.70)Observed reflections 183820Unique reflections 31955 (3236)Redundancy 5.8 (5.6)I/σI 39.8 (5.4)Completeness (%) 96.8 (99.8)Rsym 0.085 (0.383)Rwork/Rfree 0.172/0.205Rms deviations

Bonds (Å) 0.009Angles (°) 1.23

B-factor rmsd (Å2)Backbone bonds 0.96

Average B-factor (Å2)Protein 19.85Ligand (SB431542) 15.05Water 29.28

Ramachandran plotstatistics (%)Most favoured region 97.9%Additional allowed region 2.1%Disallowed region 0.0%

Values for the highest resolution shell are given in parentheses.asu = asymmetric unit.

477A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

laboratory. However, the molecular determinants conferring inhibitorspecificity of those inhibitors toward the TGFβ/Activin but not theBMP pathway are unknown. Understanding how this specificity is reg-ulated is needed to enhance the usefulness of these pathway specific in-hibitors through chemical engineering.

We have solved the structure of SB431542 bound to the TβRI proteinkinase domain, with the aim of understanding the molecular basis forSB431542 inhibitor specificity. The structure enabled a systematic mu-tational analysis of TGFβ- and BMP-receptors (TβRI and ALK2), coupledwith specific TGFβ- or BMP-receptor activity readouts to discern thebasis for inhibitor specificity. We have discovered that the gatekeeperresidue S280 in TβRI is the key determinant for the selectivity of bothSB431542 and A8301.

2. Materials and methods

2.1. Protein expression and purification

The TβRI kinase domain (residues 200–503) was cloned into apFastBac-HTa vector, expressed in Sf9 cells and purified using nickelaffinity chromatography. The 6×His purification tag was cleaved over-night by incubation with TEV protease at 4 °C. TβRI was then purifiedusing an S75 Superdex gel filtration column pre-equilibrated in25 mM Tris–HCl pH 7.5, 150 mM NaCl and 1 mM DTT.

2.2. Crystallisation, data collection and structure determination

TβRI was mixed with SB431542 at final concentrations of 0.25 mMand 1 mM respectively and incubated on ice for 60 min prior to crystal-lisation. Crystals were grown using the hanging drop vapour diffusionmethod by mixing 1.5 μl of TβRI–SB431542 complex with 1.5 μl ofmother liquor containing 100 mM imidazole pH 8.0 and 10% PEG8000. Full-sized crystals of the TβRI–SB431542 complex grew after5 days. Crystals were cryoprotected with mother liquor supplementedwith 25% glycerol (v/v) before flash freezing in liquid nitrogen. Diffrac-tion data were collected at NE CAT beamline 24-ID-C. The structure wassolved bymolecular replacement usingMOLREP [8] and the TβRI struc-ture (PDBID 1PY5, [9]) as a search model. The structure was refined byiterative rounds of refinement with REFMAC [10] and manual modelbuilding with the programme COOT [11].

2.3. Alignment and mutagenesis

Sequence alignmentswere performed usingMUSCLE [12] and editedusing ALINE [13]. Mutants of ALK2ca and TβRIca were generated usingconventional site directed mutagenesis technique [30]. Each mutantwas sequenced-verified.

2.4. Reporter gene constructs, expression plasmids and inhibitors

Reporter plasmids pGL3 (BRE)-luc was kindly provided by Dr P.ten Dijke (Leiden University Medical Center, Netherlands). 3TP-lux,TβRI and ALK2 were previously generated in the lab. SB431542 anddorsomorphin were purchased from Sigma-Aldrich; A8301 from Tocris.

2.5. Luciferase activity assay

NIH-3T3 cells were seeded in 24 well plates for 24 h, then trans-fected in Opti-MEM using lipofectamine and PLUS reagent (Invitrogen)with 0.1 μg pGL3 (BRE2)-luc or 0.2 μg p3TP-luc, 0.05 μg of pCMV5 β-galand 0.05 μg of either pCMV-GFP or a vector expressing ALK2 or TβRI.Four hours after transfection, cells were treated with inhibitors(SB431542, A8301 or dorsomorphin) at the indicated concentrationsfor 15 h. Cells were lysed, then Firefly luciferase and β-galactosidaseactivities were measured, using a luminometer (Berthold) or a platereader (Perkin-Elmer), respectively [14]. Values of fold activation

were calculated as relative luciferase activity compared with untreatedcells. Data are representative of at least three independent experiments.Error bars represent standard deviation calculated from data derivedfrom 3 individually transfected wells of NIH-3T3 cells.

2.6. Protein expression and Western blotting

293T cells were transfected with plasmids expressing ALK5ca andALK5ca S280T, by calcium phosphate. 48 h after transfection, cellswere lysed in 0.1% (v/v) TNTE and separated on SDS-PAGE. After transferon nitrocellulose membrane, the membranes were probed with anti-Phospho-Smad2 (Cell Signalling, 1/2500), anti-Smad2 (Cell Signalling,1/2500) anti-HA (Roche, 1/5000) and anti-Actin (Sigma, 1/10,000)antibodies.

3. Results

3.1. Structure of the TβRI–SB431542 complex

To investigate the basis for selectivity of SB431542 towards TGFβpathway receptors (EC50=1 μM) versus BMP pathway receptors (noinhibition at 20 μM), we co-crystallised SB431542 bound to TβRI kinasedomain (residues 200–503). Diffraction data collection to 1.7 Å resolu-tion, structure solution by molecular replacement and refinement,resulted in a final model with good statistics (Table 1). Residues 370–371 (within the activation segment) and 497–503 (corresponding tothe C-terminus) were disordered and are absent from the final model.TβRI displays the classical kinase domain organisation comprising asmall N-terminal lobe and a larger C-terminal lobe (Fig. 1A). The TβRIactivation loop (region between DFG and APE motifs) attains theextended conformation found in active protein kinases [15]. The Leuresidue of the DLG motif (equivalent to the DFG motif in most kinases)assumes the so-called “DFG in” conformation [16]. TβRI in complexwith SB431542 resembles the TβRI kinase conformation reported inthe FKBP12–TβRI complex (RMSD 0.3 Å over 207 Cα atoms) [17].Clear electron density was observed for the SB431542 compound (seeunbiased |Fo|− |Fc| electron density maps in Fig. 1B). SB431542 has a

ALK2

N-205

N-Lobe

C-Lobe

Hingeregion

G Loop

L45

αC

αD

αE

αF

αGαH

αI

αEF

β4

β1

β7

β8

β2β3

β5

C-497

B

A

TββRI G loop SB431542TβRI kinase domain

L278A230

Y282 V219K232 G214

D351N338

H283 L260

L340

L260

L340

S280

T283

L278A230

Y282 V219K232 G214

D351N338

H283S280

T283

Fig. 1. Crystal structure of TβRI–SB43154 complex. (A) Structure of TβRI Kinase domain (blue) in complex with SB431542. The structure of SB431542 (yellow) is shown as sticksoccupying the ATP binding cleft between the kinase N- and C-lobes. (B) Magnified stereo view of SB431542 compound coordination in the binding cleft showing residues contactingSB431542 and the conservation of the ATP binding cleft between TGFβ and ALK2. Hydrogen bonds are represented by dashed lines and a water molecule is depicted as a red sphere.

478 A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

heterocyclic ring structure (Fig. S1) and acts as a type I kinase inhibitor[18] by occupying the purine (ATP) binding region and adjacent hydro-phobic sites near the hinge region; the latter defined as the linkerbetween the N- and C-terminal kinase lobes (Fig. 1A). The first residueof the hinge region is referred to as the gatekeeper residue, whichoften fills a hydrophobic pocket adjacent to the ATP purine-binding site[19]. TβRI however contains a small residue at the gatekeeping position(S280), thus allowing for the hydrophobic pyridinyl ring of SB431542to be comfortably accommodated within this pocket (Fig. 1B). The posi-tion of the pyridinyl ring adjacent to the gatekeeper residue is furtherstabilised through hydrophobic interactions via residues A230 (β3sheet), L260 (αC-β5 loop) and L278 (β4 sheet). L260 and A230 contactboth the pyridinyl and benzodioxol rings, which adopt a non-co-planarconformation (Fig. 1B).

The benzodioxol ring of SB431542 is sheltered by hydrophobic resi-dues A230, L260, Y282 (hinge region) and L340 (β7 sheet) (Fig. 1B).A hydrogen bond links the benzodioxol oxygen of SB431542 andthe amide nitrogen of H283 from the hinge region of the TβRI kinasedomain (Fig. 1B), a feature common to many ATP competitive kinaseinhibitors. The classical ion pair between K232 (VAIK motif) andE245 (helix αC), reflective of an active-like kinase conformation, ismaintained and places K232 in hydrogen bonding distance to the

imidazole ring of SB431542, which is also stabilised by V219 (β7strand) through hydrophobic interactions as well as by coordinationby an ordered water molecule (Fig. 1B). The SB431542 benzamidering, sandwiched between the glycine-rich (Gly-rich, GXGXXG)loop and the activation segment, is coordinated in part throughhydrogen bonding to D351 (Fig. 1B). Compared to the other cyclicmoieties of SB431542, the benzamide ring appears loosely coordinated.The main SB431542 contacts contributing to binding involve thepyridinyl, benzodioxol and imidazole moieties interacting with residuesfrom the hinge region and N- and C-lobes of TβRI.

3.2. The gatekeeper residue is responsible for TGFβ selective inhibition bySB431542

To understand why SB431542 selectively inhibits TβRI but not itsclosely related counterpart ALK2, we superimposed the structure ofALK2 kinase domain (PDBID 3H9R) onto the structure of TβRI boundto SB431542, and looked for differences in residues contributing toSB431542 binding. Inspection of the SB431542-kinase contacts revealedthat of the 12 residues directly contacting the bound inhibitor, only thegatekeeper residue S280 differs between ALK2 and TβRI (Fig. 1B). S280

0

TββRIca TβRIca S280T

TβRI

00 201052.510.20.05201052.5

00 2.50 10 0 2.50 10 0 2.50 10 0 2.50 10 0 2.50 10 0 2.50 10

10.20.05SB431542 [µM]

Rel

. Lu

cife

rase

Act

ivit

y

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0

Rel

. Lu

cife

rase

Act

ivit

y

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

A

B

3TP-lux

3TP-lux

F216Y

SB431542 [µM]

mutations E228N--

wt

wt

ca

A368GH371P

S280T V279I

TβRI G loop

C

SB431542

TβRI kinase domainV279

A368

F216S280

Q208

E228

0

Fig. 2. TβRIca S280Tmutant is resistant to SB431542. (A) NIH-3T3 cells were transfectedwith 3TP-lux. β-gal and constructs expressing TβRIwt, or TβRIca with the indicated ATP bindingcleft mutations. After treatmentwith various doses of SB431542, Luciferase activity wasmeasured and results were normalised to β-gal activity. (B)Worm representation of the N-lobe ofTβRI showing residues selected formutagenesis based on lack of conservation among TGFβ type I receptors (see also sequence alignment in Fig. S2). (C) The experimentwas conducted asin (A), with the indicated constructs.

479A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

is conserved in all the TGFβ/Activin pathway receptors, while it isreplaced by T283 in all BMP pathway receptors (Fig. S2).

To assess if S280 indeed contributes to selective inhibitor bindingwemutated this position to threonine (its counterpart in ALK2), and testedfor the ability of SB431542 to inhibit TβRI. We used a constitutively

active form of TβRI (TβRIca, T204D), as our goal was to analyse theeffect of the mutation on the kinase activity, and not on ligand bindingor on type I/II transactivation.We co-transfected TβRIcawith a reportergene (3TP-Lux) that is specifically activated by Smad2 and 3, the down-stream targets of TβRI [20]. TβRIca strongly activated 3TP-lux when

480 A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

compared to the wild type TβRI (N15 fold) (Fig. 2A). This activationwas potently suppressed by SB431542 in a dose dependent mannerand TβRIca activity was diminished to basal levels at 20 μM SB431542(Fig. 2A). We then tested a TβRIca S280T mutant in the presence ofSB431542. We found that TβRIca (S280T) strongly activated 3TP-lux,but unlike the WT receptor, was completely resistant to SB431542inhibition when tested up to 20 μM (Fig. 2A). These results indicatethat S280 confers selectivity in SB431542-mediated inhibition ofTβRI-like kinases. We note that both TβRIca and TβRIca (S280T)have similar expression levels as shown by Western blotting (Fig.S3).

We next explored the possibility that additional residues could alsobe important for SB431542 selective inhibition. Sequence alignment of

TββRIca

00 10050251020.50A8301 [nM]

Rel

. Lu

cife

rase

Act

ivit

y

1

0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

A

B

TβRI -SB4

TBRI -A83

S280

H283

T283

Hingeregion

Fig. 3. Similarities between A8301 and SB431542. (A) Magnified ribbon representation of suplines: Hydrogen bonds; blue sphere: amide nitrogen of His283. (B) NIH-3T3 cells were transtreating cells with varying concentrations of A8301 compound.

TGFβ and BMP pathway receptor kinases revealed non-conserved resi-dues as possible contributors to TβRI–SB431542 specificity (Fig. S2).Namely Q208, F216, E228, V279, A368 and H371 corresponding toL211, Y219, N231, I282, G371 and P374 in ALK2 (Fig. 2B and Fig. S2).Although not in direct contact with SB431542 (Fig. 2B), we questionedif these differing residues contributed indirectly to the selective inhibi-tion of TβRI by SB431542. Additionally, the orientation of F216 is flipped≈180° toward the outside of the binding pocket in the presence ofSB431542 (compared to ALK5 apo structure PDBID 11AS — Fig. S4A).Assuming the same side chain flip will occur in ALK2, the correspondingresidue Y219 would be predicted to directly contact SB431542 (Fig.S4A). However, substitution of TβRIca residues to their ALK2 counter-parts, F216Y, E228N, V279I and A368G-H371P, all yielded mutant

TβRIca S280T

0200 20010050251020.5

3TP-lux

31542

01 precursor

D351

erimposed co-crystal structures of TβRI-A8301 and TβRI-SB431542 complexes. Dashedfected with 3TP-lux, β-gal and indicated constructs. Dose response was determined by

ALK2ca T283SALK2ca

ALK2ca T283SALK2ca

0 10510.50.10.010 0 0510.50.10.01Dorso [µM]

Rel

. Lu

cife

rase

Act

ivit

y

1

0

0.2

0.4

0.6

0.8

1.2

1.4

000 201052.510.1201052.510.1SB431542 [µM]

Rel

. Lu

cife

rase

Act

ivit

y

0.2

0.4

0.6

0.8

1

1.2

1.4A

B

C

BRE2-lux

BRE2-lux

ALK2 -Dorsomorphin

TββRI -SB431542

S280Hingeregion

T283

0

Fig. 4. ALK2 gatekeeper residue mutant T283S is sensitive to SB431542. (A) NIH-3T3 cells were transfected with BRE-Luc, β-Gal and plasmid expressing ALK2ca or ALK2ca T283S.Dose response was determined by treating cells with varying concentrations of SB431542 compound (A). (B) Superimposition of the ATP binding cleft of TβRI–SB431542 and ALK2–dorsomorphin complexes. (C) The experiment was performed as in (A), but with dorsomorphin.

481A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

482 A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

receptors that were equally sensitive to inhibition by 5 μM of SB431542(Fig. 2C). This suggests that these differing residues do not contribute toTβRI selective inhibition by SB431542. Analysis of combinational muta-tions further revealed no additional amino acids or group of amino acidsthat regulated SB431542 specificity (Fig. S4B).

3.3. TβRIca (S280T) is resistant to A8301 inhibition

Since the gatekeeper residue defines sensitivity to SB431542,we nextinvestigated if the same residue (S280) also regulates specificity tosubstituted-pyrazole compounds. Superimposition of TβRI–SB431542complex with the structure of TβRI bound to LY364947 (precursor ofA8301—PDBID 1PY5) [9], revealed similarities of TβRI-inhibitor binding(Fig. 3A). The two TβRI structures show little overall difference (RMSDof 0.3 Å over 293 Cα atoms) with the exception of a shift in the orienta-tion of the side chain of D351 (D351 interactswith the benzamide ring ofSB431542 and the imidazole ring of LY364947—Fig. 3). Comparison ofthe two structures led us to hypothesise that the S280T mutationwould also render TβRIca resistant to A8301 inhibition. Indeed, A8301inhibited TβRIca in a dose dependent manner with complete inhibitionat 50 nM A8301 (Fig. 3B), while TβRIca (S280T) was resistant to A8301inhibition under the same conditions (Fig. 3B). These findings demon-strate the importance of S280 in conferring selectivity of substituted-imidazole and -pyrazole small molecule inhibitors towards TβRI-likekinases.

3.4. T283S mutation renders ALK2 sensitive to SB431542

ALK2 has a threonine at the gatekeeper position and is resistant toSB431542, therefore, we mutated the ALK2 gatekeeper residue T283to the Ser residue found in TβRI and assessed sensitivity to SB431542.For this we used a constitutively active ALK2 mutant (ALK2ca, Q207D)and a BMP reporter gene (BRE2-luc) (Fig. S1A) [21]. We found thatthe single T283S substitution of the ALK2 gatekeeper residue resultedin ALK2ca inhibition by SB431542 (Fig. 4A). Notably, the EC50 ofSB431542 for ALK2ca (T283S) was similar to the EC50 of SB431542 onTβRI activity (EC50 ALK2ca T283S=2.5 μM, EC50 TβRIca=1 μM).These results confirm that the gatekeeper residue in TGFβ family recep-tors is critical for inhibitor specificity among the compounds we tested.

To further delineate themolecular basis for specific TGFβ versus BMPreceptor inhibition,we compared the bindingmode of SB431542 to TβRIwith that of the BMP inhibitor dorsomorphin to ALK2 (PDBID 3H9R)(Fig. 4B). We examined the co-crystal structure of ALK2 in complexwith dorsomorphin and observed that in contrast to TβRI-SB431542structure, dorsomorphin does not contact the gatekeeper position butinstead is mainly coordinated by hydrogen bonds with H286 in thehinge region of theALK2kinase domain (Fig. 4B). Thus,we hypothesisedthat inhibition by dorsomorphin would not be affected by the T283Smutation. Consistentwith this notion, dorsomorphin comparably inhib-ited both ALK2ca T283S and ALK2ca in the BMP pathway reporter assay(complete inhibition at 5 μM) (Fig. 4C). Moreover, none of the TβRIcamutants generated (see Fig. 2C and Fig S4B) was sensitive to dorsomor-phin (data not shown).

4. Discussion

In this study we show that the selective inhibitory function of thesubstituted-imidazole SB431542 on TGFβ- versus BMP-signalling de-pends on serine 280 at the gatekeeping position within the ATP bindingcleft of the kinase domain. Mutation of this serine to threonine (S280T)conferred complete resistance of TβRI to inhibitionby SB431542. Further-more, a reciprocalmutation of threonine to serine at this position in ALK2(T283S) rendered ALK2 sensitive to SB431542. Moreover, we showedthat the gatekeeper residue was also important for the substituted-pyrazole A8301 inhibition of TβRI. A number of co-crystal structures ofTβRI kinase domain-inhibitors with several types of core moieties are

available. To our knowledge, all co-crystallised inhibitors appear to occu-py the available space in the ATP-binding back-pocket of TβRI (Fig. S5).This suggests that S280 could be a key determinant of most TβRIinhibitors.

Altogether, our structural and functional studies point out the im-portance of having a small gatekeeper residue in TβRI that allows forSB431542 binding. Interestingly, 11 out of the 518 protein kinaseshave residues with small or no side chains (e.g. Ser, Ala and Gly) atthe gatekeeper position [Kinase Sequence Database (http://sequoia.ucsf.edu/ksd/)]. Six protein kinases have serine as a gatekeeper, threeof which are TGFβ family kinases sensitive to SB431542 inhibition(ALK4, TβRI and ALK7) and the other three are c-Abl, RETGC-1 andSTRADβ. Of these, c-Abl is not inhibited by SB431542 suggesting thatother binding pocket determinants, in addition to the small gatekeeperresidue, contribute to the selectivity of SB431542 inhibition [7, 22].

Substituted-imidazole TGFβ pathway inhibitors (SB4321542 and itsclose analogue SM16 — Fig. S1) have been successfully used in animaldisease models for fibrosis, Chagas disease and cancer. Usage of SM16resulted in decreased fibrosis and decreased myofibroblast inductionin the lung [23], kidney [24], liver [25] or blood vessels [4]. SB431542was successfully used in a mouse model of Chagas disease, whichresults from the infection of cardiomyocytes by T. cruzi parasites andled to decreased parasitaemia and prevented heart damage [26]. Finally,SM16 inhibited primary tumour progression and metastasis in mousebreast cancer and mesothelioma tumour models [27, 28]. Moreover, asubstituted-pyrazole, LY2157299 [29], is being tested in phase II onpatients with hepatocellular carcinomas (http://clinicaltrials.gov).

These studies bear great promises for therapeutical use of TGFβ/Activin pathway inhibitors based on substituted-imidazole or -pyrazolescaffolds. Our results pave the way forward for structural-based drugdesign aimed at improving the selectivity and potency of TGFβ/Activinpathway inhibitors, which will help their translation to the clinic.

5. Conclusions

- Mutation of the gatekeeper residue of TβRI (Ser280) to the equivalentresidue in ALK2 (Thr283) renders TβRI insensitive to SB431542, asubstituted imidazole selective TβRI inhibitor.

- The Ser280Thr mutant of TβRI is also insensitive to A8301, asubstituted pyrazole selective TβRI inhibitor.

- Analysis of the deposited structures of TβRI in complex with smallmolecules inhibitors suggests that Ser280 is a key determinant ofmost of the selective TβRI inhibitors.

Supplementary materials related to this article can be found onlineat doi:10.1016/j.cellsig.2011.09.027.

Accession codes

Coordinates and structure factors have been deposited in the ProteinData Bank (PDB) under accession code PDBID: 3TZM.

Financial disclosure

Supported by funds from CIHR to JLW (MOP14339) and FS(MOP36399). LD is recipient of a CIHR Fellowship and EZ is a SirHenry Wellcome Postdoctoral Fellow. JLW is an HHMI InternationalScholar and CRC Chair in Systems Biology and FS is a CRC Chair in Struc-tural Biology of Signal Transduction. The funders had no role in studydesign, data collection and analysis, decision to publish, or preparationof the manuscript.

Acknowledgments

We thank Andrew Hinck and Gerald Gish for critically reviewingthis manuscript.

483A.A. Ogunjimi et al. / Cellular Signalling 24 (2012) 476–483

References

[1] L. Attisano, J.L. Wrana, Science 296 (2002) 1646–1647.[2] J.F. Callahan, J.L. Burgess, J.A. Fornwald, L.M. Gaster, J.D. Harling, F.P. Harrington, J.

Heer, C. Kwon, R. Lehr, A. Mathur, B.A. Olson, J. Weinstock, N.J. Laping, Journal ofmedicinal chemistry. 45 (2002) 999–1001.

[3] G.J. Inman, F.J. Nicolas, J.F. Callahan, J.D. Harling, L.M. Gaster, A.D. Reith, N.J. Laping,C.S. Hill, Mol Pharmacol. 62 (2002) 65–74.

[4] K. Fu, M.J. Corbley, L. Sun, J.E. Friedman, F. Shan, J.L. Papadatos, D. Costa, F. Lutterodt, H.Sweigard, S. Bowes,M. Choi, P.A. Boriack-Sjodin, R.M. Arduini, D. Sun,M.N. Newman, X.Zhang, J.N. Mead, C.E. Chuaqui, H.K. Cheung, M. Cornebise, M.B. Carter, S. Josiah, J.Singh, W.C. Lee, A. Gill, L.E. Ling, Arteriosclerosis, thrombosis, and vascular biology.28 (2008) 665–671.

[5] J. Singh, C.E. Chuaqui, P.A. Boriack-Sjodin, W.C. Lee, T. Pontz, M.J. Corbley, H.K.Cheung, R.M. Arduini, J.N. Mead, M.N. Newman, J.L. Papadatos, S. Bowes, S. Josiah,L.E. Ling, Bioorganic & medicinal chemistry letters. 13 (2003) 4355–4359.

[6] M. Tojo, Y. Hamashima, A. Hanyu, T. Kajimoto, M. Saitoh, K. Miyazono, M. Node, T.Imamura, Cancer Sci. 96 (2005) 791–800.

[7] J. Vogt, R. Traynor, G.P. Sapkota, Cellular signalling. 23 (2011) 1831–1842.[8] A. Vagin, A. Teplyakov, Acta crystallographica. Section D, Biological crystallogra-

phy. 66 (2010) 22–25.[9] J.S. Sawyer, B.D. Anderson, D.W. Beight, R.M. Campbell, M.L. Jones, D.K. Herron,

J.W. Lampe, J.R. McCowan, W.T. McMillen, N. Mort, S. Parsons, E.C. Smith, M.Vieth, L.C. Weir, L. Yan, F. Zhang, J.M. Yingling, Journal of medicinal chemistry.46 (2003) 3953–3956.

[10] G.N. Murshudov, A.A. Vagin, E.J. Dodson, Acta Crystallogr D Biol Crystallogr. 53(1997) 240–255.

[11] P. Emsley, K. Cowtan, Acta Crystallogr D Biol Crystallogr. 60 (2004) 2126–2132.[12] R.C. Edgar, Nucleic Acids Res. 32 (2004) 1792–1797.[13] C.S. Bond, A.W. Schuttelkopf, Acta Crystallogr D Biol Crystallogr. 65 (2009)

510–512.[14] E. Labbe, C. Silvestri, P.A. Hoodless, J.L. Wrana, L. Attisano, Mol Cell. 2 (1998)

109–120.[15] B. Nolen, S. Taylor, G. Ghosh, Molecular cell. 15 (2004) 661–675.

[16] F. Zuccotto, E. Ardini, E. Casale, M. Angiolini, Journal of medicinal chemistry. 53(2010) 2681–2694.

[17] M. Huse, Y.G. Chen, J. Massague, J. Kuriyan, Cell. 96 (1999) 425–436.[18] J. Zhang, P.L. Yang, N.S. Gray, Nat Rev Cancer. 9 (2009) 28–39.[19] Y. Liu, K. Shah, F. Yang, L. Witucki, K.M. Shokat, Bioorganic & medicinal chemistry.

6 (1998) 1219–1226.[20] J.L.Wrana, J. Carcamo, L. Attisano, S. Cheifetz, A. Zentella, F. Lopez-Casillas, J. Massague,

Cold Spring Harb Symp Quant Biol. 57 (1992) 81–86.[21] O. Korchynskyi, P. ten Dijke, The Journal of biological chemistry. 277 (2002)

4883–4891.[22] M.W. Karaman, S. Herrgard, D.K. Treiber, P. Gallant, C.E. Atteridge, B.T. Campbell,

K.W. Chan, P. Ciceri,M.I. Davis, P.T. Edeen, R. Faraoni,M. Floyd, J.P. Hunt, D.J. Lockhart,Z.V. Milanov, M.J. Morrison, G. Pallares, H.K. Patel, S. Pritchard, L.M. Wodicka, P.P.Zarrinkar, Nature biotechnology. 26 (2008) 127–132.

[23] P. Bonniaud, P.J.Margetts, M. Kolb, J.A. Schroeder, A.M. Kapoun, D. Damm, A.Murphy,S. Chakravarty, S. Dugar, L. Higgins, A.A. Protter, J. Gauldie, American journal of respi-ratory and critical care medicine. 171 (2005) 889–898.

[24] E.T. Grygielko, W.M. Martin, C. Tweed, P. Thornton, J. Harling, D.P. Brooks, N.J.Laping, The Journal of pharmacology and experimental therapeutics. 313(2005) 943–951.

[25] A.C. de Gouville, V. Boullay, G. Krysa, J. Pilot, J.M. Brusq, F. Loriolle, J.M. Gauthier,S.A. Papworth, A. Laroze, F. Gellibert, S. Huet, British journal of pharmacology. 145(2005) 166–177.

[26] M.C. Waghabi, E.M. de Souza, G.M. de Oliveira, M. Keramidas, J.J. Feige, T.C. Araujo-Jorge, S. Bailly, Antimicrobial agents and chemotherapy. 53 (2009) 4694–4701.

[27] M.P. Rausch, T. Hahn, L. Ramanathapuram, D. Bradley-Dunlop, D. Mahadevan,M.E. Mercado-Pimentel, R.B. Runyan, D.G. Besselsen, X. Zhang, H.K. Cheung,W.C. Lee, L.E. Ling, E.T. Akporiaye, Anticancer Res. 29 (2009) 2099–2109.

[28] E. Suzuki, S. Kim, H.K. Cheung, M.J. Corbley, X. Zhang, L. Sun, F. Shan, J. Singh, W.C.Lee, S.M. Albelda, L.E. Ling, Cancer research. 67 (2007) 2351–2359.

[29] L. Bueno, D.P. de Alwis, C. Pitou, J. Yingling, M. Lahn, S. Glatt, I.F. Troconiz, Europeanjournal of cancer. 44 (2008) 142–150.

[30] T.A. Kunkel, J.D. Roberts, R.A. Zakour, Methods in Enzymology 154 (1987)367–382.