Galunisertib

Preclinical Evaluation of AZ12601011 and AZ12799734, Inhibitors of Transforming Growth Factor b Superfamily Type 1 Receptors

ABSTRACT
The transforming growth factor b (TGFb) superfamily includes TGFb, activins, inhibins, and bone morphogenetic proteins (BMPs). These extracellular ligands have essential roles in normal tissue homeostasis by coordinately regulating cell proliferation, differen- tiation, and migration. Aberrant signaling of superfamily members, however, is associated with fibrosis as well as tumorigenesis, cancer progression, metastasis, and drug-resistance mechanisms in a variety of cancer subtypes. Given their involvement in human disease, the identification of novel selective inhibitors of TGFb superfamily receptors is an attractive therapeutic approach. Seven mammalian type 1 receptors have been identified that have context-specific roles depending on the ligand and the complex formation with the type 2 receptor. Here, we characterize the biologic effects of two transforming growth factor b receptor 1 (TGFBR1) kinase inhibitors designed to target TGFb signaling. AZ12601011 [2-(2-pyridinyl)-4-(1H-pyrrolo[3,2-c]pyridin-1-yl)- 6,7-dihydro-5H-cyclopenta[d]pyrimidine]; structure previously undisclosed] and AZ12799734 [4-({4-[(2,6-dimethyl-3-pyridinyl) oxy]-2-pyridinyl}amino)benzenesulfonamide] (IC50 5 18 and 47 nM, respectively) were more effective inhibitors of TGFb-induced

Introduction
The transforming growth factor b (TGFb) cytokine super- family, including TGFb, activin, inhibins, nodal and bone morphogenetic proteins (BMPs), regulates developmental and cellular biology to control and maintain tissue homeostasis. In different contexts, the interaction of the dimeric ligands with their cognate receptors activates intracellular signaling path- ways to control cell death, survival, adhesion, differentiation,reporter activity than SB-431542 [4-[4-(1,3-benzodioxol-5-yl)-5- (2-pyridinyl)-1H-imidazol-2-yl]benzamide] (IC50 5 84 nM) and LY2157299 [4-[2-(6-methylpyridin-2-yl)-5,6-dihydro-4H- pyrrolo[1,2-b]pyrazol-3-yl]quinoline-6-carboxamide monohy- drate]] (galunisertib) (IC50 5 380 nM). AZ12601011 inhibited phosphorylation of SMAD2 via the type 1 receptors activin A receptor type 1B (ALK4), TGFBR1, and activin A receptor type 1C (ALK7). AZ12799734, however, is a pan TGF/BMP inhibitor, inhibiting receptor-mediated phosphorylation of SMAD1 by activin A receptor type 1L, bone morphogenetic protein receptor type 1A, and bone morphogenetic protein receptor type 1B and phosphorylation of SMAD2 by ALK4, TGFBR1, and ALK7. AZ12601011 was highly effective at inhibiting basal and TGFb-induced migration of HaCaT keratinocytes and, furthermore, inhibited tumor growth and metastasis to the lungs in a 4T1 syngeneic orthotopic mammary tumor model. These inhibitors provide new reagents for investigating in vitro and in vivo pathogenic processes and the contribution of TGFb- and BMP-regulated signaling pathways to disease states.movement, and deposition of components of the extracellular matrix (Shi and Massague, 2003). In human disease, the expression of, or cellular responses to, these factors may become deregulated. The resulting aberrant signaling can contribute to many disease pathologies, including cancer, fibrosis, atherosclerosis, and scarring (Blobe et al., 2000). TGFb, for example, can switch from tumor suppressor to tumor promoter functions depending on epigenetic/genetic changes occurring in the tumor cell (Inman, 2011).

Given their extensive role in human disease, the ligands, their receptors, and their downstream effectors are considered attractive therapeutic targets (Akhurst and Hata, 2012).Cytokine signaling is initiated by the formation of hetero- tetrameric complexes of six polypeptides comprising di- meric ligands, two constitutively active type 2 receptors, and two type 1 receptors. Upon complex formation, the serine/ threonine kinase type 1 receptors are phosphorylated and activated by the type 2 receptors, which initiates a canonical signaling cascade involving the recruitment, phosphorylation, and activation of receptor-regulated SMADs. The phosphory- lated SMADs bind to SMAD4, and the complexes accumu- late in the cell nucleus where they both positively and negatively regulate gene expression. In addition, a number of SMAD-independent noncanonical signaling pathways (e.g., mitogen-activated protein kinases, Rho GTPases, and phosphoinostide-3-kinases) are also activated which impact on gene regulation and cellular responses (Zhang, 2009). Five type 2 and seven type 1 receptors have been identified in mammalian cells. Their specificity for ligand and receptor complex formation determines biologic outcome (Schmierer and Hill, 2007). For instance, TGFb interacts primarily with transforming growth factor b receptor 2 and the ubiquitous type 1 receptor, trans- forming growth factor b receptor 1 (TGFBR1), to regulate cytostasis, apoptosis, and epithelial to mesenchymal transition, but both TGFb and growth differentiation factor 2 (BMP9) may bind to the endothelial cell restricted type 1 receptor ACVRL1 to regulate angiogenesis. In some circumstances, TGFb may also signal via TGFBR1 and activin A receptor type 1 (ALK2) to induce SMAD1/5 phosphorylation and a subset of TGFb- inducible genes involved in epithelial to mesenchymal transition (Ramachandran et al., 2018).

BMPs, meanwhile, bind selec- tively to BMPR2 and the type 1 receptors activin A receptor type 1L (ALK1), ALK2, bone morphogenetic protein receptor type 1A (BMPR1A), and bone morphogenetic protein recep- tor type 1B (BMPR1B) to regulate embryogenesis and bone formation (Wang et al., 2014). TGFb, activin, and nodal signal via TGFBR1, activin A receptor type 1B (ALK4), and activin A receptor type 1C (ALK7), respectively, predominantly to phos- phorylate and activate SMAD2 and SMAD3, whereas BMPs induce phosphorylation of SMAD1, SMAD5, and SMAD8.Given the role of TGFb in pathologic states, there have been numerous therapeutic approaches taken to blockade signal- ing, with some progressing through preclinical evaluation to clinical trial (Lahn et al., 2005; Akhurst and Hata, 2012). Some selective small-molecule inhibitors (SMIs) of the TGFBR1 kinase are well characterized (e.g., SB-431542, 4-[4-(1,3-benzodioxol- 5-yl)-5-(2-pyridinyl)-1H-imidazol-2-yl]benzamide) (Inman et al., 2002; Vogt et al., 2011) and have been used extensively in preclinical models to interrogate TGFb-regulated biological pathways. SMIs of TGFBR1 have shown therapeutic prom- ise in models of fibrosis (Gellibert et al., 2006; Park et al., 2015a,b), and two TGFBR1 inhibitors are currently in clinical trials, vactosertib (N-(2-fluorophenyl)-5-(6-methyl-2-pyridinyl)- 4-[1,2,4]triazolo[1,5-a]pyridin-6-yl-1H-imidazole-2-methanamine) (TEW-7197) and galunisertib (LY2157299, 4-[2-(6-methylpyridin- 2-yl)-5,6-dihydro-4H-pyrrolo[1,2-b]pyrazol-3-yl]quinoline-6- carboxamide monohydrate). The results of the first-in-human trials with vactosertib (ClinicalTrials.gov identifier: NCT02160106)are not yet reported.

Galunisertib, however, has progressed despite some initial toxicity concerns and is now considered safe enough when administered with careful dosing schedules (Fujiwara et al., 2015; Herbertz et al., 2015; Kovacs et al., 2015) to allow further clinical development in cancer patients with unmet need. TGFBR1 inhibitors, therefore, remain attractive leads for drug development.AZ12601011 2-(2-pyridinyl)-4-(1H-pyrrolo[3,2-c]pyridin-1- yl)-6,7-dihydro-5H-cyclopenta[d]pyrimidine; structure previ- ously undisclosed] and AZ12799734 [4-({4-[(2,6-dimethyl-3-pyr- idinyl)oxy]-2-pyridinyl}amino)benzenesulfonamide] (Goldberg et al., 2009; Anderton et al., 2011) are two selective TGFBR1 inhibitors. AZ12601011 and AZ12799734 inhibit TGFBR1 ki- nase activity (competition binding) with Kd values of 2.9 and 740 nM, respectively. Both compounds have some inhibitory activity against the related receptors ALK4 and BMPR1B but are only weakly active against ALK1, ALK2, and BMPR1A in in vitro kinase assays (Anderton et al., 2011). Here, we charac- terize their specificity and in vitro and in vivo biological activity in a variety of cell-based biochemical and functional assays in comparison with other well studied and clinically relevant TGFBR1 SMIs. We show that AZ12601011 is a selective in- hibitor of ALK4, ALK7, and TGFBR1. Importantly, AZ12601011 inhibited TGFb-induced epithelial cell migration at 10-fold lower concentrations than galunisertib and was effective in an in vivo tumor model system. Additionally, we identify AZ12799734 as a pan BMP/TGFb inhibitor providing a novel spectrum of single- agent inhibitory activity useful for analytical evaluation of TGFb superfamily-regulated processes.Synthetic procedures for AZ12799734 have been described pre- viously [compound 19 (Goldberg et al., 2009)].2-Pyridin-2-yl-1,5,6,7-tetrahydrocyclopenta[e]pyrimidin-4-one2- amidinopyridine hydrochloride (7.88 g, 50 mmol) and K2CO3 (6.91 g, 50 mmol) were added to methyl 2-oxocyclopentanecarboxylate (com- mercially sourced) (6.21 mL, 50 mmol) in ethanol (100 mL). The mixture was heated to reflux for 5 hours. While the mixture was still hot, it was acidified by the addition of 2 M aqueous HCl (20 mL) and water (30 mL).

The mixture was allowed to cool to room temperature overnight and was then partially concentrated in vacuo. The resulting precipitate was collected by filtration, washed with water and diethyl ether, and dried in vacuo to give the desired product. Additional product was obtained by concentrating the filtrate in vacuo and purifying the resulting residue by recrystallization from ethanol/water. The two batches were combined to afford the title compound (3.63 g, 34%) (1H NMR d6-DMSO): 2.04 (2H, quintet), 2.71 (2H, t), 2.88 (2H, t), 7.63 (1H, dd), 8.04 (1H, t), 8.30 (1H, d),8.73 (1H, d), NH missing; m/z MH1 214):4-Chloro-2-pyridin-2-yl-6,7-dihydro-5H-cyclopenta[d]pyrimidine phosphoryl chloride (4.72 mL, 50.7 mmol) was added to 2-pyridin-2-yl- 1,5,6,7-tetrahydro-cyclopenta[e]pyrimidin-4-one (2.7 g, 12.7 mmol). The mixture was heated at reflux for 1 hour and was then cooled to room temperature. The mixture was concentrated in vacuo and extracted with CH2Cl2/CH3OH. The organic layer was washed sequentially with 5% aqueous NaOH and sat. brine, then dried (MgSO4) and concentrated in vacuo to afford the title compound (2.60 g, 89%) as a solid [1H NMR (DMSO-d6) 2.20 (2H, quintet), 3.06 (2H, t), 3.16 (2H, t), 7.79 (1H, t), 8.27 (1H, dt), 8.53 (1H, d), 8.82 (1H, d); m/z MH1 232]:AZ12601011, 5-azaindole (1.32 g, 11.1 mmol), Cs2CO3 (6.05 g, 18.6 mmol), palladium(II) acetate (0.125 g, 0.56 mmol), and (R)-(2)-1-[(S)- 2-(dicylohexylphosphino)ferrocenyl]ethyldi-tert-butylphosphine (0.644 g, 1.11 mmol) were added to 4-chloro-2-(pyridin-2-yl)-6,7-dihydro-5H- cyclopenta[d]pyrimidine (2.15 g, 9.28 mmol) in 1,4-dioxane (46.4 mL), and the mixture was heated at reflux for 3 hours. The mixture was allowed to cool to room temperature and then concentrated in vacuo to give a brown gum which was partitioned between CH2Cl2/water.

The organic layer was isolated and dried (MgSO4) and concentrated in vacuo. The resulting crude mixture was purified by flash column chromatography, eluting with 0%–5% CH3OH in CH2Cl2, to afford the title compound (1.44 g, 50%) as a solid [1H NMR (DMSO-d6) 2.21 (2H, quintet), 3.15 (2H, t), 3.28–3.34 (2H, m), 7.01 (1H, dd), 7.58 (1H, dq),8.01–8.06 (2H, m), 8.45 (1H, d), 8.46 (1H, dt), 8.65 (1H, dt), 8.84 (1H, dq), 9.00 (1H, d); m/z MH1 314]:Plasmids. Plasmids expressing SMAD1 and SMAD2 and consti- tutively active ALK1, ALK2, BMPR1A, ALK4, TGFBR1, BMPR1B,and ALK7 used in transient transfections have been described previ- ously (Inman et al., 2002).Cell Culture. HaCaT, C2C12, T47D, VU-SCC-040, NIH3T3,NIH3T3-CAGA12-luc, C2C12-BRE-luc, and murine 4T1 [4T1-Luc (RRID:CVCL_J239)] cell lines were maintained in Dulbecco’s mod- ified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and glutamine. Medium for NIH3T3-CAGA12-luc cells was also supplemented with 400 mg/mL G418 ((2R,3R,4R,5R)-2- [(1S,2S,3R,4S,6R)-4,6-diamino-3-[(2S,3R,4R,5S,6R)-3-amino-4,5-dihydroxy-6-[(1R)-1-hydroxyethyl]oxan-2-yl]oxy-2-hydroxycyclohexyl]oxy-5-methyl- 4-(methylamino)oxane-3,5-diol), and medium for C2C12-BRE-luc was supplemented with 700 mg/mL G418. All cell lines were confirmed negative for mycoplasma contamination by the Institute’s myco- plasma testing service. Cells were exogenously stimulated with TGFb1 (PeproTech EC, Ltd. London, U.K.) resuspended in 1 mg/mL bovine serum albumin (BSA)/4 mM HCl. Inhibitors were diluted in DMSO at 1000–2000× final concentration required. Appropriate BSA/HCl and DMSO vehicle controls were used throughout. Osmotic shockwas induced by treatment of cells for 20 minutes with 0.7 M NaCl prepared in DMEM and was used to induce p38 mitogen-activated protein kinase (MAPK) phosphorylation (Davis, 2000).

Epidermal growth factor (EGF) treatment was carried out by exogenous addition of 30 ng/mL EGF for 5 minutes. The MEK1/2 (Mitogen-activated protein kinase kinase) inhibitor U0126 [1,4-Diamino-2,3-dicyano-1,4- bis[2-aminophenylthio]butadiene] (Tocris, Bristol, UK) was used at a concentration of 25 mM.C2C12-BRE-Luciferase and NIH3T3-CAGA12-Luciferase Re- porter Bioassays. The C2C12-BRE-luciferase (Herrera and Inman, 2009) and NIH3T3-CAGA12-luciferase stable cell lines (Spender et al., 2016) used in the bioassays have been described previously. Cells were seeded overnight at 1 × 104/well in 96-well plates. Cells were then pretreated for 20 minutes with a titration of SB-431542 (Inman et al.,2002) (Tocris), SB-525334 [6-[2-(1,1-Dimethylethyl)-5-(6-methyl-2- pyridinyl)-1H-imidazol-4-yl]quinoxaline] (Sigma-Aldrich, St. Louis, MO), AZ12601011 (Anderton et al., 2011), AZ12799734 (AstraZeneca, Cambridge, UK), or LY2157299 (galunisertib; Selleck, Houston, TX) followed by stimulation with exogenous TGFb (1 ng/mL) or BMPs. Cell lysates were analyzed by the Luciferase Assay system (Promega, Madison, WI) as recommended by the manufacturer. The mean percentage activity (6S.D.) of the reporter in the presence of the inhibitors relative to maximal activation induced by TGFb or BMPs (100%) is shown.Dose-Response IC50 (Dm) Calculation and Proliferation Curve Analysis. IC50 values were determined by curve fitting using either nonlinear regression (GraphPad Prism 7; GraphPad Software, La Jolla, CA) or CalcuSyn Version 2.11 software (Biosoft, Cambridge, UK), which reports median-effect plots, the median dose (Dm, IC50), and the correlation coefficient (r). The 95% confidence intervals of IC50 were determined using GraphPad Prism 7 software.

Statistical anal- ysis of cell proliferation curves was carried out by pairwise comparison between two or more groups of growth curves using the compare- GrowthCurves function from statmod (R Project, The R Foundation, Vienna, Austria). The number of permutations (nsim) was 10,000, and P values given were adjusted for multiple testing.Western Blotting and Antibodies. Cell lysates were analyzed by SDS-PAGE using the following antibodies: phospho-SMAD2 (Ser465/467) [rabbit polyclonal, #3101; Cell Signaling Technology (CST), Danvers, MA], phospho-SMAD1 (rabbit polyclonal, #9511; CST), phospho–extracellular signal-regulated kinase [ERK; rabbit polyclonal phospho p44/42 (Thr202/Tyr204) MAPK, #9101; CST], phospho-p38 (Thr180/Tyr182) (rabbit polyclonal, #9211; CST), SMAD1 (rabbit polyclonal #38-5400; Zymed, Paisley, UK), SMAD2 (mouse monoclonal, C16D3; CST), and SMAD2/3 (mouse monoclonal, Clone 18; BD Transduction Laboratories, Wokingham, UK). Second- ary horseradish peroxidase–conjugated antibodies (Dako, Stockport, UK) and enhanced chemiluminescence (GE Healthcare, Chicago, IL) were used to detect bound antibody. LI-COR (Cambridge, UK) infrared imaging was carried out using a LI-COR Odyssey system followingwestern blotting using rabbit monoclonal anti-phospho-SMAD2 (Ser465/ 467, 138D4, #3108; CST) and mouse monoclonal pan-SMAD2 antibody (L16D3, #3103; CST). Secondary antibodies were IRDye 680RD donkey (polyclonal) anti-rabbit IgG (H1L, #925-68073; LI-COR) and IRDye 800CW goat anti-mouse IgG (H1L, #926-32210; LI-COR). Analysis of signal intensities was carried out using LI-COR Image Studio version 4.0.21 software. Signal was normalized to pan-SMAD2 levels measured on channel 800 according to the manufacturer’s signal protein normalization strategy.Wound-Healing Assay, IncuCyte Zoom, and Data Analysis.

An IncuCyte Zoom live cell imaging microscope (Essen Bioscience, Welwyn Garden City, UK) with 10× objective and data manage- ment software was used to monitor cell migration in wound-healing assays. HaCaT cells were seeded overnight in 100 mL of growthmedium at 15,000 cells/well in 96-well ImageLock plates (Essen Bioscience). Cells were washed twice in phosphate-buffered saline (PBS) and serum starved for 24 hours in DMEM supplemented with 0.2% FBS to minimize proliferation, which would interfere with migration assays. Uniform wounds were then generated using a WoundMaker (Essen Bioscience), and the cells were washed twice in PBS and then pretreated for 20 minutes with titrations of LY2157299 (galunisertib), SB-431542, AZ12601011, or AZ12799734 in DMEM/ 0.2% FBS. Cells were then treated with BSA/HCl as a carrier control or TGFb (5 ng/mL). Closure of the wound was monitored by IncuCyte Zoom imaging over 4 days. Data presented are the mean 6 S.E.M. of a minimum of three replicate wells taken from a representative experi- ment. The mean 6 S.E.M. relative wound density was determined according to software processing definitions as recommended by the manufacturer.In Vivo 4T1 Syngeneic Orthotopic Mammary Fat-Pad Model for Efficacy and Pharmacokinetics/Pharmacodynamics. The 4T1 cell line was maintained in RPMI 1640 (Invitrogen, Paisley, UK) supplemented with 10% FCS (Sigma-Aldrich) and M1 (Egg Technol- ogies, Wiltshire, UK) with glutamine (Invitrogen) in 7.5% CO2 at37°C and was detached using 0.05% trypsin (Invitrogen). Cells were implanted into mammary fat pad 4 in 30 mL of PBS at 1 × 104 permouse. Female BALB/c mice (Charles River, Harlow, UK) weighing more than 18 g had tumor sizes monitored three times weekly by bilateral caliper measurements. Tumor volume was calculated at each measure with dosing from the day following implant (Shaw et al., 2017).

Controls received vehicle only (0.5% Methocel E4 Premium, [Dow Chemical Company, USA]/0.1% Polysorbate 80, [Sigma Aldrich, UK]) by oral gavage twice daily for the duration of the study. The treatment group received 50 mg/kg AZ12601011 twice daily in 0.5% Methocel/0.1% Polysorbate. Samples for pharmacokinetic and phar- macodynamic analysis were sampled at 1 and 6 hours post final morning dose on day 25. Additionally, lung samples were taken to assess the number of lung metastases. Each lobe was sectioned following formalin fixation, creating five serial sections per lobe, 200 mm apart. The sections were then stained with hematoxylin and eosin stain, and the number of metastases were counted using light microscopy. This study was run in the UK in accordance with UK Home Office legislation, the Animal Scientific Procedures Act 1986, and AstraZeneca Global Bioethics policy. Experimental details are outlined in Home Office Project License 40/2934. Additional dosing was undertaken for PK profile analysis for AZ12601011 and AZ12799734 at 50 mg/kg per oral dose in male nude mice in vehicle as detailed earlier. Statistical analysis was carried out using SPSS software (IBM, Armonk, NY).In Vivo Pharmacokinetic Analysis. Each plasma sample (25 mL) was prepared using an appropriate dilution factor and compared against an 11-point standard calibration curve (1–10,000 nM) prepared in DMSO and spiked into blank plasma. Acetonitrile (100 mL) was added with the internal standard, followed by centrifu- gation at 3000 rpm for 10 minutes. Supernatant (50 mL) was then diluted in 300 mL of water and analyzed via ultrahigh-performance liquid chromatography–tandem mass spectrometry.

The mass spec- trometer and ultrahigh-performance liquid chromatography parame- ters are detailed in Table 1, and the optimization parameters are presented in Table 2.In Vivo Pharmacodynamic Analysis of Phospho-SMAD2Using Enzyme-Linked Immunosorbent Assay. Mouse total SMAD 2/3 monoclonal antibody (catalog number 610843l; BD Biosci- ences, San Jose, CA) at 1.25 mg/mL in PBS was used to coat Greiner black high-bind 96-well plates with incubation for 1 hour at room temperature then inverted and blotted. Blocking buffer with 3% BSA (A-3912; Sigma-Aldrich) was added, and the incubation/blotting pro- cess was repeated. 4T1 tumor lysate was prepared at 100 mg/well in AP lysis buffer containing fresh protease/phosphatase inhibitors and then added to the wells for an overnight incubation at 4°C. Following this incubation, all wells were washed with Tris-buffered saline/ Tween 20. A plate was incubated with primary antibody (Rabbit Upstate phospho-SMAD2, 05-953), and a second plate was incubated with primary antibody (Rabbit CST Total SMAD 2/3, 3102; CST). Blocking buffer was used for the antibodies, and the plates were incubated for 2 hours at room temperature followed by washing with Tris-buffered saline/Tween 20. Anti-rabbit horseradish peroxidase conjugate (7074; CST) was added to the plates and incubated for 1 hour, followed by a further wash. The plates were developed with QuantaBlu (Thermo Fisher Scientific, Loughborough, UK) using the manufacturer’s recommended protocol and read on a Tecan Spectro- Fluor (Tecan, Switzerland) using excitation at 340/465-nm emission.

Results
AZ12601011 and AZ12799734 Inhibit Ligand-Activated SMAD3/4 Transcription. NIH3T3 cells stably expressing a TGFb-inducible CAGA12-luciferase reporter construct were used to determine the effect of AZ12601011 and AZ12799734 (compound structures are shown in Fig. 1A) on TGFb- induced transcription in comparison with the previously characterized TGFBR1 inhibitors SB-431542, SB-525334, and LY2157299 (galunisertib). The 12 tandem copies of the SMAD binding element CAGA (Dennler et al., 1998) regu- late luciferase reporter gene activity in response to TGFb receptor-regulated SMAD3 and SMAD4 binding. Luciferase activity in cell lysates increased in response to TGFb1 addition and was inhibited by all compounds tested (Fig. 1B). The IC50 values, determined by curve fitting from dose- response curves (Fig. 1C) of AZ12601011 (18 6 6 nM) and AZ12799734 (47 6 7 nM), were lower than those of the other inhibitors and were approximately 10-fold lower than the IC50 of galunisertib (384 6 170 nM), approved for use in clinical trials (Fig. 1, C and D). Phosphorylation of SMAD2 induced by exogenous TGFb addition, detected using a polyclonal rabbit antiphosphorylated SMAD2 (ser465/467) antibody, was com- pletely inhibited by the most active inhibitor, AZ12601011, in four cell lines at concentrations between 300 nM and 1 mM (Fig. 2A). The inhibition of SMAD2 phosphorylation in TGFb-treated HaCaT relative to pan-SMAD2 levels was quantified by LI-COR infrared imaging using a second monoclonal rabbit antiphos- phorylated SMAD2 (ser465/467) (138D4) antibody (Fig. 2, B–D). Representative blots are shown in Fig. 2B, and combined raw intensity traces from channel 700 and 800 scans of lane 2 (control 1 TGFb) measuring phospho-SMAD2 and pan-SMAD2, respec- tively, are shown in Fig. 2C. Normalized phospho-SMAD2 levels relative to uninduced levels (lane 1) are shown in Fig. 2D; these returned to baseline in the presence of 0.3 mM AZ12601011 and indicate levels of inhibition similar to those seen in Fig. 2A. Stability and Specificity of AZ12601011 and AZ12799734. We next tested the suitability, stability, and specificity of AZ12601011 and AZ12799734 as reagents for studying TGFb family signaling in cell-based assays.

Neither AZ12601011 nor AZ12799734 showed any reduction in their ability to block CAGA12-luciferase reporter activity follow- ing up to six rounds of freeze-thawing when used at concentrations selected to inhibit CAGA12 activity by approximately 50% (20 and 60 nM, respectively) (Fig. 3A). All four inhibitors tested maintained activity when in- cubated in tissue culture medium for up to 10 days (Fig. 3B) and effectively maintained efficacy in in vitro cell culture assays over this time period as assessed by the ability to block TGFb-mediated induction of SMAD2 phos- phorylation (Fig. 3C). AZ12601011 and AZ12799734 cantherefore be used to effectively block TGFb signaling in phenotypic assays over a 10-day time course.Previous drugs designed to target the TGFb type 1 receptor, such as SB-431542, effectively inhibit the activity of activin and nodal receptors ALK4 and ALK7 in addition to TGFBR1. Therefore, using transient transfection assays, we also assessed the effect of AZ12601011 and AZ12799734 on constitutively ac- tive ALK4, ALK7, and TGFBR1 receptors expressed in NIH3T3 CAGA12 reporter cells. Induction of CAGA12-luciferase re- porter activity induced by transient transfection of the consti- tutively active receptors is shown in Fig. 4A. Following transient transfection, cells were treated with a titration of AZ12601011 and AZ12799734, and the effect on CAGA12 reporter activity was monitored (Fig. 4B). AZ12601011 inhibited the activity of ALK4, ALK7, and TGFBR1 equally at concentrationsbetween 0.01 and 10 mM. AZ12799734 also equally inhibited the three constitutively active receptors at concentrations ranging from 1 to 10 mM. We next tested the effect of AZ12601011 and AZ12799734 on receptor-regulated SMAD phosphorylation by transient transfection of constitutively active receptors into NIH3T3 cells (Fig. 4C). Phosphoryla- tion of cotransfected SMAD1 and SMAD2 by the active recep- tors was monitored by western blot. All seven receptors induced phosphorylation of either SMAD1 or SMAD2; as expected, phosphorylation of SMAD1 occurred following transfection of ALK1, BMPR1A, and BMPR1B, and phosphor- ylation of SMAD2 occurred following transfection of ALK4, TGFBR1, and ALK7 (Fig. 4C).

AZ12601011 inhibited phos- phorylation of SMAD2 but not SMAD1, suggesting that (similar to SB-431542) AZ12601011 is a selective inhibitor of ALK4, TGFBR1, and ALK7.AZ12799734, however, inhibited phosphorylation of both SMAD1 and SMAD2 and is thus likely to be a pan BMP/TGF receptor inhibitor. The effect of AZ12799734 on BMP signaling was confirmed using a bioassay measuring the transcriptional activation of a BMP responsive element luciferase reporter construct (C2C12-BRE-LUC) (Herrera and Inman, 2009) after exogenous addition of BMP4, BMP6, or growth differentia- tion factor 2 (BMP9). As expected, given that there was no effect of AZ12601011 on SMAD1 phosphorylation, we saw no inhibition of BMP-induced activity of C2C12-BRE-LUC in AZ12601011-treated cells (Fig. 5A). Following treatment by AZ12601011, we observed no inhibition of receptor/MEK/ERK signaling cascade (Bogdan and Klämbt, 2001), which was efficiently blocked by the MEK1/2 inhibitor U0126, which was used as a positive control for MEK1/2 inhibition (Fig. 5B).AZ12601011 and AZ12799734 Inhibit TGFb-Induced Migration. AZ12601011 and AZ12799734 were assessed infunctional wound-healing assays using HaCaT epithelial cells that measure TGFb-inducible SMAD4-dependent migration responses (Levy and Hill, 2005). Confluent monolayers of serum-starved (0.2% FBS) HaCaT cells were uniformly scratched using a WoundMaker (Essen Biosciences), and the resulting wound closure was analyzed by real-time imaging (Fig. 6). Serum-starved DMSO control-treated cells migrated to close the wound by just over 80% after 80 hours of incuba- tion. Following treatment with TGFb, however, HaCaT cells migrated faster to close the wound entirely within 36 hours. When cells were pretreated with different concentrations of SB-431542, galunisertib, AZ12601011, or AZ12799734 prior to TGFb addition, we observed dose-dependent de- creases in TGFb-induced migration. SB-431542, galunisertib, AZ12601011, and AZ12799734 blocked the increase in migra- tion at concentrations of 500, 5000, 100, and 500 nM, respec- tively. At 500 nM, AZ12601011 reduced cell migration below baseline, which was only achieved at doses of $5 mM SB- 431542 and AZ12799734 (data not shown). AZ12601011, there- fore, was the most effective inhibitor in the wound-healingassay, inhibiting TGFb-induced migration at concentrations at least 10-fold lower than galunisertib.

AZ12601011 Inhibits Tumor Growth and Metastasis In Vivo. The selectivity profile of AZ12601011 for ALK4, TGFBR1, and ALK7 (similar in scope to galunisertib in clinical trials) made it a good representative probe molecule to understand the profile of a TGFBR1 inhibitor in vivo. We tested the pharmacokinetics of both drugs (total and free) in BALB/c mice following administration of a 50-mg/kg dose (Fig. 7A) and tested the effect of AZ12601011 on 4T1 cell proliferation in vitro prior to its use in an in vivo assay of preventative treatment of 4T1 syngeneic orthotopic tumor growth. AZ12601011 inhibited 4T1 proliferation with an IC50 of 400 nM (Fig. 7B). In the orthotopic in vivo assay, AZ12601011 inhibited signaling measured by a reduction in phosphorylated SMAD2 detected in tumor cell lysates 1 hour after administration (50 mg/kg) (Fig. 7C). AZ12601011 inhibited 4T1 cell growth in vitro (IC50 5 0.4 mM) (Fig. 7C). To assess efficacy in vivo, mice were treated twice daily with vehicle or AZ12601011 (50 mg/kg) starting the day after implantation of 4T1 cells into mammary fat pads. Tumor growth was monitored over 25 days (Fig. 7D). The average tumor size in the treated group (n 5 10) up to day 18 (when 13/13 mice in the control group remained in the study) was significantly inhibited by AZ12601011 (P 5 0.0001) compared with controls (n 5 13) (Fig. 7E). Two of 13 mice in the controlgroup had to be sacrificed before the end of the study at days 18 and 23 due to large tumor sizes. No tumors in the treated group exceeded the stipulated size limit, and one of 10 mice did not develop a measurable tumor. We noted a significant increase in event-free survival (P 5 0.002) and in the time taken to develop tumors exceeding 0.5 cm3 (P , 0.005) in the treated mice (Fig. 7F). Importantly, we also found a statisti- cally significant reduction (P 5 0.025) in the median lung metastasis score in mice treated with AZ12601011 (median 5 1) compared with the control group (median 5 4; U 5 26, z 5 22.267, using an exact sampling distribution for U).Therefore, AZ12601011 shows efficacy as a single-agent therapy in tumor model systems providing a TGFBR1 in- hibitor lead compound for future development.

Discussion
Our data describe the specificity and efficacy of two compounds designed to inhibit TGFBR1 kinase activity (AZ12601011 and AZ12799734) in a number of cell-based assays. We provide evidence that AZ12601011 targets ALK4 and ALK7 as well as TGFBR1, whereas AZ12799734 is a pan inhibitor of ALK1, ALK2, BMPR1A, ALK4, TGFBR1, BMPR1B, and ALK7. Thus, although not directly tested here, we predict that similar to SB-431542, AZ12601011 will also inhibit activin and nodal signaling through ALK4 and ALK7. AZ12799734, we anticipate, may be useful as a single-agent combined inhibitor of TGFb and BMP activity. The efficacy of AZ12601011 and AZ12799734 was also assessed in biological assays of TGFb-induced transcription and migration. TGFb-induced migration of HaCaT epithelial cells is, in part, reported to involve TGFb-mediated activation of a genetic program regulated by the ERK signaling pathway demonstrated by the observation that the MEK/ERK inhibitor U0126 can inhibit this response (Zavadil et al., 2001). Since 100 nM concentrations of AZ12601011 inhibited migration but had no effect on EGF-induced phosphorylation of ERK or p38 phosphorylation up to concentrations of 10 mM, we conclude that the effects of the inhibitors on migration are not a result of off-target inhibition of MEK/ERK but are likely a result of on-target TGFBR1 inhibition. In these assays, AZ12601011 and AZ12799734 were at least 10-fold more effective in blocking TGFb-induced transcription and migration than galunisertib, currently in clinical trials. Importantly, AZ12601011 reduced signaling via SMAD2 in vivo and inhibited tumor growth and metastasis in an orthotopic murine model of breast cancer, suggesting therapeutic potential. The development of novel inhibitors of TGFBR1 could provide new therapeutic options in TGFb-associated pathol- ogies, such as cancer and fibrosis.

The clinical use of TGFBR1 inhibitors, however, is not without concerns over patient safety. TGFb signaling via TGFBR1 has a role in cardiac development (Sridurongrit et al., 2008), and several heart defects are evident in TGFb2-null mice. Because increased TGFb expression is involved in cardiac remodeling in response to numerous stresses (Lim and Zhu, 2006), TGFBR1 inhibitors are predicted to be of benefit in combating the effects of cardiac remodeling in heart failure. However, administration of AZ12601011 to disease-free wild-type mice is associated with cardiac toxicity involving the development of aortic lesions (Anderton et al., 2011). Questions over these homeostatic functions of TGFb have thus delayed development of lead compounds, such as LY2157299 (galunisertib). Similar effects of galunisertib on small mammals were noted during detailed preclinical toxicity testing but were ameliorated by lower doses and careful dosing schedules (Herbertz et al., 2015). These concerns therefore appear to be largely overcome, and several cancer-related trials of galunisertib as a monotherapy, or in combination with other chemotherapeutics, are ongoing. The efficacy of galunisertib in models of cancer, however, appears limited by tumor environmental factors, and antitu- mor growth inhibitory effects are lacking in some xenograft models (Herbertz et al., 2015; Maier et al., 2015). In hepato- cellular cancer cells, galunisertib had minimal effects on proliferation but inhibited invasion (Serova et al., 2015). Most of the studies to date have focused on antiproliferative and tumor-killing potential in standard assays, and there is growing evidence that these inhibitors may be more effective in cancer models measuring metastasis, immune function, and cancer stem cell–like properties (Peñuelas et al., 2009; Anido et al., 2010; Maurantonio et al., 2011; Spender et al., 2016). For example, galunisertib has been shown to inhibit TGFb-induced migration and immune suppression in vitro (Yingling et al., 2017). In addition, SB-431542 inhibits mutant BRAF melanoma cell migration and stem cell–like properties of anchorage-independent growth and clonogenicity at low- cell density. These effects are overcome by increasing cell density, leading to the conclusion that the dependence of tumor cells on TGFb/TGFBR1 signaling is revealed under conditions of cellular stress. Therefore, TGFBR1 inhibitors may be most effective in inhibiting micrometastasis seeding or outgrowth, following removal of any clinically apparent tumor mass (Spender et al., 2016). This conclusion is supported by the inhibitory effect of AZ12601011 on HaCaT cell migration and the formation of 4T1 lung metastasis in our experimental cell-based assays. Hence, TGFb receptor inhibitors remain candidate therapeutic agents with therapeutic potential, de- spite the lack of significant tumor killing by galunisertib. Additionally, galunisertib has shown promise in ex vivo models of fibrotic disease (Luangmonkong et al., 2017). Therefore, AZ12601011 and AZ12799734 provide new tools for studying TGFb- and BMP-regulated biologic processes and lead com- pounds for clinical development.