Discovery and Characterization of a Highly Potent and Selective Aminopyrazoline-Based in Vivo Probe (BAY-598) for the Protein Lysine Methyltransferase SMYD2

Protein lysine methyltransferases have recently emerged as a new target class for the development of inhibitors that modulate gene transcription or signaling pathways. SET and MYND domain containing protein 2 (SMYD2) is a catalytic SET domain containing methyltransferase reported to monomethylate lysine residues on histone and nonhistone proteins. Although several studies have uncovered an important role of SMYD2 in promoting cancer by protein methylation, the biology of SMYD2 is far from being fully understood. Utilization of highly potent and selective chemical probes for target validation has emerged as a concept which circumvents possible limitations of knockdown experiments and, in particular, could result in an improved exploration of drug targets with a complex underlying biology. Here, we report the development of a potent, selective, and cell-active, substrate-competitive inhibitor of SMYD2, which is the first reported inhibitor suitable for in vivo target validation studies in rodents.


■ INTRODUCTION
SMYD2 is a catalytic SET domain containing protein methyltransferase reported to monomethylate lysine residues on histone and nonhistone proteins. 1 SMYD2 has been proposed as a potential therapeutic target in cancer. Its overexpression has been reported in cancer cell lines as well as in esophageal squamous carcinoma, bladder carcinoma, gastric cancer, and pediatric acute lymphoblastic leukemia patients. 2−6 In these studies, SMYD2 overexpression often correlated with lower survival rate and was suggested to be a clinically relevant prognostic marker. Knockdown of SMYD2 in overexpressing ESCC, bladder, and gastric cancer cell line models significantly reduced cell proliferation. 2 Initially, SMYD2 was characterized as methylating H3 lysine 36 7 and lysine 4 when interacting with HSP90a. 8 Methylation of histones by SMYD2 has been connected to increased transcription of genes involved in cellcycle regulation, chromatin remodeling, and transcriptional regulation. 8 In addition, several studies have uncovered an important role of SMYD2 methylation activity toward nonhistone proteins closely connected to cancer. This is in line with the emerging concept that posttranslational methylation of nonhistone proteins (e.g., of transcription factors) by protein methyltransferases can also substantially alter protein function. Thereby, a regulatory role of lysine methylation can probably be extended to multiple cellular pathways besides transcriptional regulation and histones. 9,10 So far, the best-characterized example of SMYD2 methylation of a nonhistone protein is the tumor suppressor transcription factor p53. 11−16 Transcriptional activity of p53 is inhibited by SMYD2-mediated posttranslational methylation at lysine 370 (K370). 13,17 The structural basis of p53 methylation by SMYD2 has been characterized by solving the crystal structure of a ternary complex with the cofactor product S-adenosylhomocysteine and a p53-derived substrate peptide. 16 It has been proposed that methylation at K370 reduces the DNA-binding efficiency of p53 and subsequently prevents the transcriptional activation of p53 target genes. 13 In the same study, a knockdown of SMYD2 and treatment with doxorubicin led to an increase in p53-mediated cell-cycle arrest and apoptosis. In line with these observations, low SMYD2 gene expression was suggested as a predictive marker for an improved response to neoadjuvant chemotherapy in breast cancer patients. 18 Besides p53, several other proteins have been identified as SMYD2 substrates, including the estrogen receptor (ER), 19,20 PARP1, 21 retinoblastoma protein (Rb), 4 and HSP90. 22,23 Mechanistically, methylation of HSP90 has been connected to the normal physiological role of SMYD2 in muscle biology, 24,25 as well as in cancer. 23 These studies indicate that SMYD2 has many substrates and various potentially tissue-specific physiological and pathogenic functions. SMYD2 therefore represents a very attractive target for further exploration in different disease-relevant models. Nevertheless, the biology of SMYD2 is still poorly understood and the molecular contribution of individual substrates to specific knockdown phenotypes remains largely unknown. For a more unbiased interpretation of biological experiments, fully profiled chemical probes can substantially contribute to preclinical target validation. 26,27 Although first cellular-active probe inhibitors of SMYD2 have been described (Figure 1), there is a need for structurally orthogonal chemical probes to enable cross-validation studies and thereby rule out unspecific effects. 26 The publication of 1 (AZ505) 12 was the first disclosure of a co-crystal structure of an inhibitor bound to SMYD2 and paved the way for further studies, leading to the discovery of 2 (LLY-507), 28 and 3 (A-893), 29 which have reported significantly improved potency. Analysis of the respective cocrystal structures reveals that all three inhibitors bind in a similar fashion, occupying the same binding pockets.
Furthermore, the cellular activity of the known inhibitors is limited, and no data about in vivo applicability have been published. Here we report the discovery of a potent and selective aminopyrazoline-based small-molecule inhibitor (S)-4 (BAY-598). 30, 31 We show that (S)-4 has a distinctly different binding mode compared to previous inhibitors, utilizing a dichlorophenyl moiety as so far unprecedented chemical motif for addressing the methyl-lysine binding pocket of SMYD2. For the first time, we are presenting in vivo xenograft and DMPK data for a SMYD2 inhibitor. In addition to previously described inhibitors, (S)-4 shows very potent cellular activity combined with reasonable DMPK properties ( Figure 1). Furthermore, we are indicating the potential that this inhibitor might offer to the field of protein methyltransferases in the quest to fully explore the underlying complex biology and therapeutic potential of SMYD2 by validating AHNAK protein 32 as a new cellular substrate.

■ RESULTS AND DISCUSSION
The potential link between SMYD2 and cancer motivated us to screen the Bayer compound collection, with the aim of identifying small-molecule inhibitors of the enzyme. To this end, a scintillation proximity assay (SPA) was set up using recombinant His-tagged SMYD2, a biotinylated p53-derived peptide substrate, and tritiated S-adenosyl-L-methionine ( 3 H-SAM) (Figure 2A) . Of the three million compounds tested in a primary HTS, we identified more than 2300 confirmed hits which inhibited SMYD2 with IC 50 values below 15 μM. Among the multiple structural clusters and singletons in the hit list, several offered starting points with low micromolar potency and tractable chemical matter, and our attention was drawn to pyrazolines such as compound 5. Initial hits and later derivatives of the

Journal of Medicinal Chemistry
Article pyrazoline series showed stabilizing effects in an SMYD2 thermal shift assay (TSA) ( Figure 2B). In addition, binding of compound 6 to SMYD2 was validated by isothermal titration calorimetry (ITC), which indicated a submicromolar binding constant (K d = 540 nM,

Journal of Medicinal Chemistry
Article Figure 2C) and a high enthalpic contribution to the binding energy. As the latter reduces the likelihood of nonspecific offtarget activities, the ITC data underscored the attractiveness of pyrazoline hit 5. 33,34 It transpired that compounds of this structural series had been prepared during the course of an in-house program as antagonists for protease-activated receptor 1 (PAR1), a G-protein-coupled receptor which is highly expressed in platelets and plays an important role in thrombin signaling and platelet aggregation. 35,36 Chemistry. The pyrazoline compounds of interest were synthesized according to previously described procedures. 37 Synthesis of the required intermediates 10a−f started from the commercially available 2-amino-1-phenylethanones 7 (Scheme 1).
Protection of the amines as the allyl carbamates 8 was followed by Mannich reaction with formaldehyde and piperidine, and the resulting product mixture was treated with hydrazine monohydrate to install the pyrazoline moiety (compounds 9). Subsequent reaction with diphenyl N-cyanocarbonimidate resulted in intermediates 10a−f.
Compounds 4 and 25−29 were prepared from intermediates 10b−f by the addition of 3-(difluoromethoxy)aniline, followed by installation of the N-ethylated hydroxyacetamide as described above (Scheme 3).
The synthesis of compounds 30−34 started from intermediate 10d by addition of the respective aniline derivatives, followed by introduction of the N-ethylated hydroxyacetamide (Scheme 4).
Compounds 4, 28, and 30−34 were separated into their enantiomers by preparative chiral HPLC or chiral supercritical fluid chromatography. For compounds 6 and 4, the integrity of the pyrazoline stereocenter was tested. The compounds were stable to racemization in aqueous solution at pH 7, as well as in mouse and human plasma at 37°C, for at least 48 h. However, clean racemization was obtained under basic conditions and microwave irradiation (DBU, THF, 90°C).
With respect to the amide moiety, it became clear that the presence and orientation of a hydrogen-bond donor has a large impact on potency toward SMYD2 ( Table 2). Variation of the amide N-alkyl substituents (R 1 ) revealed that the N-ethyl derivative is preferred: secondary amide 19 (IC 50 = 10.9 μM) is about 10-fold less active and the N-propyl derivative 20 is also less potent than the corresponding N-ethyl derivative 6 (IC 50 = 0.8 μM), whereas the larger cyclopropylmethyl substituent in 21 resulted in a loss of potency (IC 50 > 20 μM). Replacement of the hydroxyacetyl in compound 12 (IC 50 = 3.3 μM) by an aminoacetyl group gave the equipotent derivative 22 (IC 50 = 2.8 μM). In contrast, the methoxyacetyl derivative 23, lacking the hydrogen-bond donor, is inactive (IC 50 > 20 μM). In compound 24, where the hydrogen-bond donor is fixed in an oxoimidazolidine ring, there is significantly reduced potency (IC 50 = 6 μM) relative to the N-ethyl derivative 6. On the basis of the available amide derivatives, with BEI values in the same range (10−11), an improvement in the binding efficiency was not envisaged. Furthermore, alterations of the amide moiety did not offer a path forward to selectivity against PAR1. Compounds 22, 23, and 24 are in a similar potency range for PAR1 antagonism and greater than 10-fold more potent against PAR1 (IC 50 = 130, 30, and 100 nM, respectively) than SMYD2.
At this stage, we selected compound 6 for co-crystal structure determination with SMYD2 based on its potency and promising biophysical properties. Compound 6 was soaked into crystals of SMYD2 grown in the presence of SAM. The crystal structure revealed that compound 6 binds into the substrate peptide binding pocket of SMYD2; the observed binding mode is consistent with the previously established SAR. There is a very good steric and electrostatic fit of 6 to the substrate binding site of SMYD2 ( Figure 3A,B). The pyrazoline and the NH of the carboximidamide form hydrogen bonds to Gly183. The 4-chlorophenyl substituent inserts into the lysine binding channel and is engaged in π-stacking interactions with Phe184 and Tyr240. There is a good fit of the 3-(difluoromethoxy)phenyl

Journal of Medicinal Chemistry
Article substituent into the adjacent hydrophobic pocket-1, with the difluoromethoxy group pointing into a hydrophobic subpocket. Although racemic 6 was used for soaking, the density maps reveal that only the S-enantiomer is bound. The S configuration at the pyrazoline provides an optimal exit vector for the hydroxyacetamide substituent, which occupies pocket-2 and forms two hydrogen bonds with Thr185 ( Figure 3A). The N′-nitrile of the carboximidamide contributes to a waterbridged hydrogen bond with Ser196. On the basis of the crystal structure of compound 6, we envisioned that further exploration of the aniline and amide moieties would not lead to a significant improvement of potency and binding efficiency. Therefore, we elected to focus on derivatives of the 4-chlorophenyl substituent, which were underrepresented in our compound library.
With the co-crystal structure in hand, we employed molecular modeling for the prioritization of derivatives. In particular, WaterMap calculations, 39,40 which estimate the position and the thermodynamic properties of water molecules in the ligandfree structure, suggested the introduction of a second substituent at the 3-position of the 4-chlorophenyl group. To test this hypothesis, a few derivatives were synthesized (see Table 3). The unsubstituted phenyl derivative 25 is inactive, while the 4-bromo derivative 26 (IC 50 = 1.1 μM) is as potent as the 4-chloro derivative 6, indicating the importance of a hydrophobic substituent at the 4-position. As predicted by WaterMap calculations, introduction of a second substituent at the meta position, as exemplified by 3,4-dichloro derivative 4 (IC 50 = 0.08 μM, BEI = 13.5) and 4-chloro-3-methyl derivative 28 (IC 50 = 0.08 μM, BEI = 14), resulted in significantly improved potency (ca. 10-fold greater than 6) and binding efficiency. As highlighted in Figure 3C, the two chloro substituents are colocated with two calculated water sites that have high free energy, suggesting an optimal water displacement by 3,4-dichlorophenyl and thus a lower binding free energy than for the 4-chlorophenyl or the unsubstituted phenyl derivative, consistent with the observed potency difference for these derivatives. However, introduction of a third substituent leads to a decrease in potency compared to the 3,4-disubstituted derivatives, as exemplified by 4-chloro-3,5-dimethyl derivative 27 (IC 50 = 0.57 μM). Compound 29, with the bulkier 1,3-benzodioxole moiety, is also less potent (IC 50 = 4.2 μM), suggesting that the SAR at this position is rather steep.
Compound 4 was selected for further biological and crystallographic studies. The co-crystal was obtained by soaking (S)-4 into crystals of SAM-bound SMYD2. Structure determination revealed that (S)-4 features an almost identical binding mode as compound 6, as evidenced in the overlay of both structures ( Figure 3D). In line with the observed S configuration of compound 6 and (S)-4, we observed a greater  Table 4]. Introduction of the additional 3-chloro substituent also resulted in a dramatic decrease in the antagonistic effect on PAR1 [IC 50 = 1.7 μM for (S)-4 and >30 μM for (R)-4]. Although (S)-4 proved to be the active isomer for both SMYD2 inhibition and PAR1 antagonism, there is a greater than 50-fold selectivity for SMYD2 relative to PAR1.
The co-crystal structures of compound 6 and (S)-4 revealed that the pyrazoline-based SMYD2 inhibitors feature a different binding mode to other recently reported inhibitors. Thus, 1 12 and 2 28 ( Figure 4A) have published IC 50 values of 0.12 μM and <15 nM, respectively, and like (S)-4 they bind to the substrate binding site of SMYD2. The structures of the three inhibitors are superimposed in Figure 4B. All ligands occupy the lysine binding channel and the adjacent hydrophobic pocket-1; however, (S)-4 employs pocket-2 via hydrogen-bond interactions with its hydroxyacetyl moiety (cf. Figure 2A), which are not present in 1 or 2. 1 and 2, on the other hand, occupy a remote hydrophobic pocket-3 which is not addressed by (S)-4. Figure 4C provides a view into the lysine binding channel (the respective motifs are highlighted in color on the structures in Figure 4A). 2 binds with its N-alkylpyrrolidine moiety at this position, which is structurally closely related to the lysine side chain of a methylated substrate peptide. 1 and (S)-4 address the lysine channel with aromatic substituents. It is noteworthy that the 3,4-dichlorophenyl group of 1 and of (S)-4 bind to different positions.
On the basis of 3,4-dichlorophenyl as a novel lysine channel binding motif, we identified several pyrazoline derivatives as potent SMYD2 inhibitors (see Table 4). Aiming to identify a candidate for in vivo experiments, we focused on the potent S-enantiomers of these derivatives, profiling them in pharmacokinetic assays in vitro. (S)-4 showed moderate stability upon incubation with rat hepatocytes (CL blood = 2.5 L/h/kg) as well as moderate apparent permeability (34 nm/s) and a hint of active transport in the Caco2 assay (efflux ratio = 5). In comparison, the 4-chloro-3-methylphenyl derivative (S)-28 had similar permeability and efflux (33 nm/s, efflux ratio = 7) and slightly lower metabolic stability in rat hepatocytes (CL blood = 2.8 L/h/kg). Compounds (S)-30 and (S)-32−34 exhibited high metabolic stability (CL blood = 1.8−0.3 L/h/kg); however, these derivatives Calculated water sites are shown as spheres and colored based on their free energy. Only sites within the lysine channel are shown. WaterMap results are based on the crystal structure of the monochloro derivative 6 (PDB code 5ARF). The protein surface is depicted in gray. The modeled ligand structure is shown as colored sticks (chlorine, green; carbon, yellow; hydrogen, white; nitrogen, blue; oxygen, red; fluorine, light green). For clarity, protein residues and cofactors are not shown. (D) Overlay of the crystal structures of compound 6 (magenta) and (S)-4 (yellow, PDB code 5ARG), showing nearly identical binding modes.

Journal of Medicinal Chemistry
Article have very low aqueous solubility (<5 mg/L), thereby limiting their suitability for in vivo experiments. On the other hand, the methoxyphenyl derivative (S)-31 has high aqueous solubility (163 mg/L) and moderate permeability with a hint of active transport (62 nm/s, efflux ratio = 3) in the Caco2 assay; nevertheless, (S)-31 displayed low stability upon incubation with rat hepatocytes (CL blood = 3.2 L/h/kg), and thus bioavailability is expected to be low. On the basis of these data, (S)-4 was selected for further in vitro and in vivo studies. The in vivo pharmacokinetic properties of (S)-4 were first evaluated by a single-dose administration of (S)-4 at 0.4 mg/kg by iv bolus or 0.8 mg/kg po, respectively, to rats; there was moderate blood clearance (1.6 L/h/kg) and a low bioavailability of 24% (see Supporting Information Figure S1A). These data prompted us to assess the exposure of (S)-4 following oral administration. Hence, we treated mice with 10−100 mg/kg po qd which are well tolerated doses of (S)-4. As a result, unbound IC 50 [cellular methylation In-Cell Western (ICW) assay; see Figure 7A−C] is covered for ∼9 to ∼12 h at steady state when 10 and 100 mg/kg, respectively, are administered (see Supporting Information Figure S1B).
Potency, Selectivity, and Inhibition Mode. (S)-4 showed potent in vitro inhibition of SMYD2 with an IC 50 of 27 ± 7 nM in the biochemical SPA assay ( Figure 5A). To analyze the mode of inhibition and affinity of (S)-4, our SMYD2 protein preparation was characterized with respect to its apparent Michaelis−Menten constants for SAM and the p53 peptide substrate (data not shown). The K m(app) values of 60 nM and 1 μM, respectively, were in excellent agreement with the constants previously reported using a similar assay. 12 Then, we performed IC 50 determinations at increasing concentrations of one substrate and a fixed amount of the other ([S] = K m(app) ) and applied the Cheng−Prusoff relationship 41 as described elsewhere. 42 Increasing the concentration of peptide substrate resulted in a linear increase in IC 50 ( Figure 4B, upper panel), as would be expected for a competitive mode of inhibition. Fitting these data to the corresponding Cheng−Prusoff competitive inhibition model revealed a K i(app) of 8 ± 1 nM (SD). On the other hand, when SAM concentrations were titrated to saturation, we observed a decrease in IC 50 which converged to a constant value ( Figure 4B, lower panel). This type of behavior toward SAM corresponds to an uncompetitive mode of inhibition. Consequently, the IC 50 vs [S]/K m(app) plot fits best to the Cheng−Prusoff model for uncompetitive inhibition, yielding an inhibitor constant αK i(app) of 28 ± 3 nM (SD). Our data suggest that (S)-4 is a peptide-competitive, SAMuncompetitive inhibitor of SMYD2 methyltransferase activity, which preferably binds to the SMYD2−SAM substrate complex. Interestingly, SMYD2 has been reported to follow a random sequential Bi Bi mechanism of substrate binding, 12,17 but the uncompetitive mode of inhibition of (S)-4 regarding the SAM cofactor suggests an ordered sequential Bi Bi mode of substrate binding, where SAM would be required to bind before the peptide substrate. Similar results have recently been described for inhibitors of the SAM-dependent arginine methyltransferase PRMT5. 43 For further evaluation of selectivity, (S)-4 was tested on a panel of 32 additional methyltransferases, including closely related family members SMYD3, SUV420H1, and SUV420H2. As a result, (S)-4 displayed >100-fold selectivity for SMYD2, with very weak residual activity toward the closest related methyltransferase SMYD3 (IC 50 ∼ 3 μM) ( Figure 5C). In addition to the methyltransferases, (S)-4 was also profiled in the commercially available KINOMEscan (DiscoveRx) 44 and LeadProfilingScreen (Eurofins Panlabs) assay panels to fully determine relative selectivity and specificity for kinases and other primary molecular targets, including several CNS targets. Overall, we were able to confirm the high selectivity and specificity of (S)-4 for SMYD2 inhibition (see Supporting  Information Tables S1 and S2).
Cellular Methylation Activity on p53. The ability of (S)-4 to inhibit SMYD2 was tested by monitoring its effects on cellular p53 methylation using different cellular mechanistic assays. First, we generated a polyclonal antibody (SY46) for the specific detection of p53 monomethylation at lysine 370, as described elsewhere. 13 This antibody was then tested on recombinant p53 protein which had been in vitro methylated by SMYD2 in a Western blot. Specificity for methylated p53 was confirmed by the exclusive detection of the in vitro methylated p53 protein isoforms, whereas the nonmethylated p53 was not detected ( Figure 6A). Endogenous methylation of p53 protein was characterized by treatment of KYSE-150 esophageal cancer cells with increasing concentrations of (S)-4 for 5 days. The KYSE-150 cell line model was selected based on a described SMYD2 gene amplification 2 and a heterozygous R248Q mutation in p53 (COSMIC), leading to p53 protein accumulation without a stress stimulus. After treatment with (S)-4, a significant reduction of methylation was detected confirming that p53 is a cellular target of SMYD2-dependent methylation ( Figure 6B). Nevertheless overall endogenous detection of p53 protein methylation led to weak signals, hence this method was not useful for the determination of a cellular IC 50 . Therefore, we additionally employed an established system with a transient FLAG-tagged SMYD2 and FLAG-tagged p53 overexpression in HEK293T cells as benchmark assay. This assay has been used previously to characterize the structurally unrelated SMYD2 inhibitor 2. 28 As shown in Figure 6C, (S)-4 showed a concentration-dependent decrease in p53 methylation without affecting p53 total protein levels. A cellular IC 50 of 58 nM was determined ( Figure 6D), which confirms that Table 3. SAR of 3-Phenylpyrazoline Derivatives: Exploration of Phenyl Substituents (S)-4 is the most potent cellular-active SMYD2 inhibitor known to date (Figure 1).
Characterization and Inhibition of SMYD2-Mediated AHNAK Methylation. To further characterize the effects of our aminopyrazoline-based inhibitors on the cellular methylation activity of SMYD2, we generated a polyclonal cell line derived from KYSE-150 with stable N-terminal 2xc-myc-tagged SMYD2 overexpression to maximize cellular methylation activity.
In an immunofluorescence analysis, SMYD2 was localized mainly in the cytosol (Supporting Information Figure S2A), as reported by others. 22 Surprisingly, the antibody SY46 directed against methylated p53 showed a very strong signal specifically in the clones with SMYD2 overexpression in the immunochemistry staining not derived from p53 protein (Supporting Information Figure S2B,C). We validated the novel SMYD2 substrate giving rise to the strong immunochemistry staining signal by knockdown and overexpression experiments in additional cell lines to be derived from AHNAK protein 32 ( Figure 7B and Supporting Information Figure 2D−F). AHNAK methylation has been very recently reported in a proteomics study by other as an additional methylation substrate of SMYD2, further confirming our results. 45 Importantly, we used the strong AHNAK methylation signal to set up an In Cell Western (ICW) assay for cellular optimization of our lead

Journal of Medicinal Chemistry
Article series ( Figure 7A). The methylation signal was reduced by treatment with increasing concentrations of (S)-4, derivatives of the aminopyrazoline series or 1 ( Figure 7A). In addition, (S)-4 specifically reduced methylation of AHNAK without altering AHNAK protein expression in a Western blot ( Figure 7B). Importantly, IC 50 values in the cellular ICW assay for aminopyrazolines correlated with potency in the scintillation proximity assay ( Figure 7C) and were also comparable to the p53 methylation assay (see Figure 6D). Our results of the methylation of AHNAK and the recent identification of many additional methylation targets 45 clearly point toward additional roles beyond p53 regulation of SMYD2, and further studies supported by potent and selective inhibitors as (S)-4 are needed to fully elucidate the underlying biology.
Characterization of (S)-4 in Proliferation and Apoptosis Assays. To explore the potential effects of (S)-4 on proliferation, we tested a panel of 240 different cancer cell lines (OncoPanel 240/Eurofins Panlabs). Cell lines were long-term cultured with (S)-4 for 10 days to allow for a sufficient translation of demethylation of SMYD2 substrates to potential antiproliferative effects, which takes 48−72 h (Supporting Information Figure S2C). (S)-4 exposure resulted in only limited responses in a subset of cell lines; an antiproliferative response with IC 50 < 10 μM was seen in 21 cell lines (∼9%), however, most cell lines (83%) did not reach 50% proliferation inhibition with 20 μM of (S)-4 ( Figure 8A). In addition, there was no clear preference for a specific tissue origin of the responding cancer cell line. Thus, SMYD2 inhibition by (S)-4 has only limited proliferation effects in a small subset of cancer cell lines under the employed conditions. On the basis of the observation that p53 protein is methylated by SMYD2, which should lead to suppression of apoptosis, we were additionally interested in the effects of (S)-4 in combination with an apoptotic stimulus. KYSE-150, U2OS, and A2780 cell lines were pretreated with (S)-4 or inactive derivative 25 (see Table 3) for 2 days (demethylation phase), followed by treatment with doxorubicin (apoptotic trigger). (S)-4, but not 25, significantly improved caspase 3/7 activation in all three tested cell lines without inducing apoptosis alone ( Figure 7B). Thus, SMYD2 inhibition can enhance apoptotic responses.
The First Chemical Probe Suitable for In Vivo Characterization of SMYD2 Inhibition. Functional validation of novel potential cancer targets such as the protein methyltransferase SMYD2 relies on appropriate model systems in vitro as well as in vivo. Additionally, chemical probe inhibitors also suitable for in vivo applications are highly desirable. Most reported activities of SMYD2 are not directly involved in survival signaling of cancer cells. Hence, in vitro proliferation assays may not adequately cover the full phenotype of SMYD2 inhibition, and more complex (in vivo) assays are thus required. The research work characterizing SMYD2 in heart and skeletal muscle cells by knockdown experiments in vivo 15,24,25 clearly

Journal of Medicinal Chemistry
Article illustrates the complex underlying biology of SMYD2 and the necessity for chemical probes suitable for in vivo applications. Therefore, we were interested in establishing if SMYD2 methylation activity in tumor cells can be inhibited by (S)-4 in vivo. To this end, mice bearing subcutaneous tumor xenografts (tumor tissue derived from the SMYD2-overexpressing KYSE-150 cell line) were treated orally with 10, 30, 70, or 100 mg/kg (S)-4, or vehicle (PEG 400/water 8:2), once daily for 3 days. After the treatment period, tumors were harvested and analyzed ex vivo for methylation of AHNAK by dotblotting. For detection of the methylation signals, SY46 methylation antibody was used (see Figure 7A and Supporting Information Figure S2A−F). (S)-4 significantly reduced the methylation with doses starting from 30 mg/kg, with most significant effects in the 100 mg/kg treated group (P < 0.001, Student's t test) ( Figure 9A). Treatment with 10 mg/kg (S)-4 resulted in no significant effect on the methylation level. Exposure at 10 mg/kg is close to the level of the cellular IC 50 for ∼9 h, which may indicate a need for an even higher exposure as the IC 50 to achieve in vivo effects on demethylation.
Then the KYSE-150 esophageal xenograft model was used to evaluate if the observed improved apoptosis induction in the in vitro setting after treatment with doxorubicin ( Figure 8B) could translate to antitumor efficacy in vivo. Four groups of tumorbearing mice were treated as follows: Group 1 (control group) was only treated with vehicle (Solutol/ethanol/water 1:1:8) iv

Journal of Medicinal Chemistry
Article qd and once at day 4 with the vehicle used for doxorubicin (saline) iv; group 2 was treated with (S)-4 at 500 mg/kg po qd; group 3 was treated with 10 mg/kg doxorubicin iv once at day 4; group 4 was treated with a combination of (S)-4 and doxorubicin. There was a slight reduction in area ( Figure 9B) and weight ( Figure 9C) of tumors from mice treated with the combination of (S)-4 and doxorubicin relative to tumors from the control group. Combination treatment only resulted in a minor increase in treatment-related body weight loss ( Figure 9D). The combination treatment reached a T/C level (based on tumor weight) of 0.46, which is significant ( Figure 9C). Therefore, combination of a SMYD2 inhibitor with a chemotherapeutic agent resulted in reduced cancer cell growth in vivo. In comparison, the monotherapy groups treated with only doxorubicin or (S)-4 showed no significant antitumor efficacy relative to the vehicle control group. These data are consistent with the observed limited cellular proliferation effects of (S)-4 in the cell line panel ( Figure 8A) and indicate that, in contrast to SMYD2 knockdown, 2 a catalytic inhibition may be insufficient to induce cell death in the KYSE-150 esophageal model. Furthermore, the monotherapeutic approach with doxorubicin did not result in any antitumor efficacy in the KYSE-150 xenograft model.

■ CONCLUSION
In summary, we have identified (S)-4 as a potent, selective, and cell-active, substrate-competitive inhibitor of SMYD2. Our data show that SMYD2 inhibition can enhance efficacy of doxorubicin in vivo, which confirms our in vitro observation of higher caspase 3/7 activation ( Figure 8B). Our results (S)-4 are also in agreement with an earlier study, 13 where an increased apoptosis induction in cells with an SMYD2 knockdown was observed. Nonetheless, in our initial explorative in vivo study with (S)-4, effects on xenografted tumors were only moderate. In addition, high doses of (S)-4 were needed in vitro as well as in vivo relative to the concentration needed to achieve effects on methylation. Therefore, we cannot exclude the possibility that additional, so-far unexplored activities of SMYD2 might be responsible for the observed effects. This again underlines the necessity to identify suitable chemical probes for more extensive target validation campaigns to fully explore the complex biology of SMYD2 and other targets. In this regard, (S)-4 will be a highly valuable tool for the further exploration of SMYD2 biology, not only for in vitro but also for in vivo studies.
■ EXPERIMENTAL SECTION Chemistry. General Procedures. All reagents and solvents were used as purchased, unless otherwise specified. All final products were at least 95% pure, as determined by analytical HPLC.
Materials. Intermediate 10a and compounds 5, 6, and 12−24 were synthesized according to the methods described previously. 37 1 H NMR were determined to be >95%. 1 H NMR spectra were recorded on Bruker Avance III HD spectrometers operating at 300, 400, or 500 mHz. The chemical shifts (δ) reported are given in parts per million (ppm), and the coupling constants (J) are in hertz (Hz). The spin multiplicities are reported as s = singlet, br s = broad singlet, d = doublet, t = triplet, q = quartet, m = multiplet, and br = broad. The LC/MS analysis was performed on Waters Acquity UPLCMS SingleQuad with a Acquity UPLC BEH C18 column (1.7 μm, 50 mm × 2.1 mm) at 60°C, using water +0.1 vol % formic acid (99%) and acetonitrile as mobile phase at a flow rate of 0.8 mL/min and a DAD detector (210−400 nm). LC/MS/MS was performed on a CTC PAL autosampler, an Agilent 1200 HPLC, and a ABSciex 4000 mass spectrometer.
Assignment of Stereochemistry. For all separated enentiomers, it was assumed based on the co-crystal structure of 4, that the active enantiomer (SMYD2 inhibition) has S-configuration.

Allyl [1-{N′-Cyano-N-[6-(difluoromethoxy)pyridin-2-yl]-carbamimidoyl}-3-(3,4-dichlorophenyl)-4,5-dihydro-1H-pyrazol-4yl]carbamate (11q).
Compound 11q was prepared from 10d (5.0 g, 10.9 mmol) and 6-(difluoromethoxy)pyridin-2-amine. 11q was obtained as off-white solid (2.9 g, 43%). 1  To the amine product (2.30 g, 5.1 mmol) in MeOH (70 mL) at rt was added acetaldehyde (0.27 g, 6.1 mmol). The reaction mixture was stirred at rt for 1 h, and at 40°C for 1 h, then it was cooled to 0°C and NaBH 4 (0.22 g, 5.7 mmol) was added in small portions. The mixture was stirred at rt for 1 h, then poured over saturated NaHCO 3 solution (10 mL). The volatiles were removed by evaporation, and the resulting aqueous slurry was extracted with EtOAc (2 × 10 mL). The combined organic layers were washed with brine (100 mL), then dried over Na 2 SO 4 , filtered, and concentrated to yield a crude black oil. The To the debrominated byproduct (410 mg, 1 mmol) in DCM (15 mL) were added Et 3 N (0.15 mL, 1.1 mmol), acetoxyacetyl chloride (0.12 mL, 1.1 mmol), and DMAP (126 mg, 1 mmol), and the mixture was stirred at rt for 1 h. EtOAc (15 mL) was added, and the mixture was washed with saturated aqueous NH 4 Cl solution (20 mL) and brine (20 mL). The organic phase was dried over Na 2 SO 4 and concentrated. The crude product was purified by dry flash column chromatography (eluent: DCM, MeOH) to yield 504 mg of an impure intermediate, which was dissolved in MeOH (12 mL), K 2 CO 3 (140 mg, 1.01 mmol) was added, and the mixture was heated to reflux for 30 min. After cooling, to the reaction mixture was added saturated aqueous NH 4 Cl, and the mixture was extracted with EtOAc. The organic phase was dried over Na 2 SO 4 , concentrated, and the residue purified by flash column chromatography (eluent: DCM, MeOH) and subsequent preparative HPLC to yield N-  (27). Compound 27 was prepared as described for 4, starting from 11k (4.9 g, 9.6 mmol). 27 was obtained as gray solid (1.6 g, 32% over 3 steps). 1  Enzyme Activity and Inhibition Assays. SMYD2 enzyme kinetics and inhibitory activities of compounds were analyzed using a scintillation proximity assay (SPA) which measured methylation by the enzyme of the synthetic, biotinylated peptide Btn-Ahx-GSRAHS-SHLKSKKGQSTSRH-amide (Biosyntan) derived from the C-terminal domain of p53. The SMYD2 full-length enzyme with an N-terminal 6xHis tag was expressed in Escherichia coli and purified by affinity chromatography on a Ni-NTA Sepharose column, followed by sizeexclusion chromatography on a Superdex 200 16/60 column (GE Healthcare). Assays were conducted in 384-well microtiter plates in a buffer containing 50 mM Tris/HCl pH 9.0, 1 mM DTT, 0.01% (w/v) BSA, and 0.0022% (v/v) Pluronic, and a final volume of 5 μL. The SMYD2 concentration in the assay was 3 nM, while tritiated S-adenosyl-L-methionine ( 3 H-SAM) and the peptide substrate were present at 60 nM and 1 μM, respectively, to ensure "balanced" conditions. 46 Apparent Michaelis−Menten constants for SAM and the p53 peptide were determined by titrating one substrate to saturation at cosubstrate concentrations of 1 × K m(app) . Enzyme kinetics were followed over 2 h by quenching the reactions as described above at time points 0, 5, 10, 15, 30, 60, 90, and 120 min. Compounds were tested in 11-point, 3.5-fold dilution series ranging from 0.1 nM to 20 μM. Reactions were run for 30 min and quenched by adding Streptavidin PS SPA imaging beads (PerkinElmer) to a concentration of 3.12 μg/μL and 25 μM "cold" SAM. The amount of product was evaluated using a Viewlux (PerkinElmer) CCD plate imaging device [emission filter 613/55]. The data were normalized using two sets of control wells for high (= enzyme reaction with DMSO instead of test compound = 0% = minimum inhibition) and low (= all assay components without enzyme = 100% = maximum inhibition) SMYD2 activity. IC 50 values were calculated by fitting the normalized inhibition data to a four-parameter logistic equation using either a Bayer proprietary tool or Genedata Screener analysis software.
For mechanism-of-inhibition studies, IC 50 determinations were basically performed as described above, but incubation times were adjusted to guarantee kinetic linearity at each concentration of substrate analyzed. For peptide competition studies, IC 50 6. Crystallographic Data Collection, Structure Determination, and Refinement. The soaked crystal was briefly immersed in cryo buffer (0.1 M HEPES pH 7.0, 22% PEG 3350, 20% glycerol, 2 mM inhibitor) and shock-frozen in liquid nitrogen. Diffraction data sets were collected using beamline 14.1 at the Helmholtz-Zentrum Berlin at 100 K using a wavelength of 0.91841 Å and a PILATUS detector. The diffraction images were processed using the program XDS. 48 The crystals belonged to space group P2 1 2 1 2 1 with one molecule per asymmetric unit. The crystal form described here was first solved for an SMYD2:SAM crystal in the absence of an inhibitor, using the Molecular Replacement method (program Phaser 49 from the CCP4 program suite 50 and PDB entry 3TG5 as search model). The data sets described here were then solved by rigid body refinement using the SMYD2:SAM structure as starting model and the program Refmac 51 from the CCP4 program suite. 3D models for compound 6 and (S)-4 were generated using the program Discovery Studio, and parameter files for crystallographic refinement and model building were generated using the software PRODRG. 52 (S)-4 was manually built into the electron-density maps using Coot, 53 followed by several cycles of refinement with Refmac and rebuilding in Coot. For the data collection and refinement statistics, see Supporting Information Table S3.
WaterMap Calculations. WaterMap calculations were based on the crystal structure of the 4-chlorophenyl derivative 6 (PDB code 5ARF). The SMYD2−compound 6 complex structure was prepared using the Protein Preparation Wizard functionality in Maestro. 54−57 Preparation involved assignment of bond orders, addition of hydrogens, creation of zero-order bonds to metals, deletion of water molecules beyond 5 Å from heteroatoms, assignment of protonation states according to pH 7.0, and optimization of the hydrogen-bonding network and restrained minimization. Here, and for all further minimizations and simulations, the OPLS 2005 force field 58,59 was used.
Water sites and corresponding free energies were calculated using WaterMap. 39,40 The calculation involved the following (default) settings: water molecules within 10 Å of the ligand were investigated, the ligand as well as any crystallographic water molecules were removed prior to simulation, the protein was truncated beyond 20 Å from the ligand, and the simulation time was 2 ns. In brief (see ref 40 for details), the different stages of the calculation were solvation of the system in an orthorhombic box of TIP4P water 60 with a minimum distance between box edge and solute of 10 Å and a series of minimizations and short simulations to equilibrate the system, followed by a 2 ns production simulation. All nonwater heavy atoms were harmonically restrained during all minimizations and simulations using a force constant of 5 kcal mol −1 Å −2 . Coordinates were saved every 1.5 ps, yielding a total of 1334 snapshots for further analysis. Water molecules were subsequently clustered such that nonoverlapping spheres (i.e., hydration sites) with radius 1 Å were obtained. Thermodynamic properties of these sites (i.e., enthalpies, entropies, and thus free energies) were approximated as follows: Enthalpies were estimated as the difference between the average interaction energy of the water molecule with the rest of the system and the average interaction energy in bulk water. Entropies were estimated using inhomogeneous solvation theory. 61 Antibody Generation. To detect SMYD2-mediated methylation, we used a lysine monomethylation specific rabbit polyclonal antibody (SY46). The antibody was generated (Eurogentec) against a p53 peptide containing the monomethylated K370 epitope, as described elsewhere. 13 Antibody has been purified against unmethylated p53 peptide. Using this antibody in cellular systems revealed that it also recognized methylated AHNAK, which is also methylated by SMYD2 (see Supporting Information Figure S2).
Cell-Based Assay for the Detection of SMYD2Methylation Activity. For the detection of SMYD2 cellular methylation activity, an In-Cell Western (ICW) assay was established. For the ICW, KYSE-150 cells stably transfected with a construct expressing wild-type N-terminal 2xc-myc-tagged SMYD2 (NCBI reference sequence: NP_064582.2) were used. For further detection and validation experiments of AHNAK methylation, we additionally generated stable HeLa and MDA-MB-231 cell lines using the same construct. For conducting the ICW assay, 5000 SMYD2-engineered KYSE-150 cells/ well were seeded in 96-well plates (Sigma) and cultivated for 24 h at 37°C in 5% CO 2 . Nontransfected KYSE-150 cells were used as a control for maximal inhibition of methylation activity. Cells were grown in 49% RPMI 1640 and 49% Ham's F12 media supplemented with 2% heat-inactivated FCS. For the determination of SMYD2 inhibitory activity, cells were treated for 72 h in the presence of test compound (at a final concentration range of 3.9 × 10 −8 to 5 × 10 −6 M) or with DMSO. Media were removed, and 3.7% (w/v) formaldehyde in PBS was added for 20 min. After two washes with PBS, 0.25% (v/v) Triton X-100 in PBS was added for 15 min of permeabilization. After one wash with PBS, cells were blocked in 5% (w/v) nonfat dry milk in PBS for 1 h. Fixed cells were exposed to primary methylation antibody (SY46, 1:200) in 5% nonfat dry milk in PBS for 24 h. One row of cells on each plate was not exposed to methylation antibody (SY46) and was reserved for background control measurements. The wells were washed three times with PBS, then secondary IR800-conjugated antibody (LI-COR, no. 926−32211, 1:1000) and DNA-intercalating dye DRAQ5 (Thermo Fisher Scientific, no. 62251, 1:1000) were added for 3 h in blocking buffer. After five washes with PBS, the fluorescence in each well was measured on an Odyssey scanner (LI-COR) at 800 nm (SY46 methylation signal, 764 nm excitation) and 700 nm (DRAQ5 signal, 683 nm excitation). Fluorescence intensity was quantified and normalized to background and DRAQ5 signals. The normalized data were further analyzed by four-parameter logistic regression analysis using a Bayer proprietary tool to determine the IC 50 value for each tested compound. For IC 50 determinations, C 0 (= no inhibition) was defined as the signal measured for the DMSO-treated controls. C i (maximal inhibition) was defined as the signal measured for nontransfected KYSE-150 cells.
Proliferation Panel. For the characterization of proliferation effects, the OncoPanel 240 (Eurofins Panlabs) was used.
Mouse Tumor Xenograft Model. Animal experiments were conducted in accordance with the German animal welfare laws,

Journal of Medicinal Chemistry
Article approved by local authorities and in accordance with the ethical guidelines of Bayer AG. Eight-week-old female BALB/c nude mice obtained from Charles River Laboratories (Germany) were acclimated for at least for 24 h before tumor cell injection. KYSE-150 cells (4 × 10 6 ) were resuspended in 100% Matrigel (100 μL) and injected subcutaneously into the right flank region of the mice; 4 days after tumor cell inoculation, mice were randomized into four treatment groups. Treatment was started at day 4 after tumor inoculation. Group 1 (n = 12) was treated with vehicle only (Solutol/ethanol/water 1:1:8) iv qd and once at day 4 with saline iv; group 2 (n = 6) was treated with 500 mg/kg (S)-4 po qd; group 3 (n = 12) was treated with 10 mg/kg doxorubicin iv once at day 4; group 4 (n = 6) was treated with a combination of (S)-4 and doxorubicin. Tumors were measured three times per week for 14 days. Tumor weight is assumed to be log normally distributed, thus all tumor weights were logarithmically transformed (base = 2) prior to statistical inference. One-sided Dunnett's comparison of log 2 tumor weights of all treatment groups with the vehicle group were carried out on an overall significance level of 0.05. The results were transformed from the log 2 scale to the original scale so that the difference from the pairwise comparisons and the respective 95% confidence intervals calculated on the log 2 scale correspond to T/C ratios of geometric mean values plus respective 95% confidence intervals on the original tumor weight scale (mg). Data was analyzed with R 3.0. 1. Ex Vivo Methylation Detection by Dot Blot. Eight-week-old female BALB/c nude mice obtained from Charles River Laboratories (Germany) were acclimated for at least for 24 h before tumor cell injection. SMYD2-engineered KYSE-150 cells (4 × 10 6 ) were resuspended in 100% Matrigel (100 μL) and injected subcutaneously into the right flank region of the mice. Treatment was started when tumors reached a tumor area of 60−70 mm 2 (day 7 after inoculation). Mice (n = 12 per group) were treated with 10, 30, 70, or 100 mg/kg (S)-4 po qd for 3 days, or vehicle (PEG 400/water 8:2) po qd for 3 days. After the treatment period, tumor samples were immediately frozen in liquid nitrogen and stored at −80°C. Frozen tumors were mechanically homogenized using TissueLyser and stainless steel beads (Qiagen), and proteins were extracted as described for the Western blot method. Whole protein lysate (50 μg per sample) was transferred with the Dot-Blot system MiniFold-1 (Whatman) onto a nitrocellulose membrane (Invitrogen). Membrane were blocked in 5% milk PBS-T and immunoprobed with the SY46 methylation antibody. The secondary antibody used was goat antirabbit IRDye 800 CW (LI-COR, no. 926− 32211, 1:1000). Bands were detected and quantified with Odyssey Fc software (LI-COR Biosciences).
In Vivo Rat PK. Pharmacokinetic properties of (S)-4 were determined by administering the test compound as indicated. Blood samples were withdrawn at different time points, and plasma was separated by centrifugation. The samples were analyzed by LC-MS/MS.
Metabolic Stability in Rat Hepatocytes. Liver cells were distributed in Williams' medium E containing 5% fetal calf serum (FCS) to glass vials at a density of 1.0 × 10 6 vital cells/mL. The test compound was added to a final concentration of 1 μM. During incubation, the hepatocyte suspensions were continuously shaken and aliquots were taken at 2, 8, 16, 30, 45, and 90 min, to which an equal volume of cold MeOH was immediately added. Samples were frozen at −20°C overnight, then centrifuged (15 min, 3000 rpm), and the supernatant was analyzed with an Agilent 1200 HPLC system with LC-MS/MS detection. The half-life of a test compound was determined from the concentration−time plot. The intrinsic clearances were calculated from the half-life, together with additional parameters (liver blood flow, quantity of liver cells in vivo and in vitro). The hepatic in vivo blood clearance (CL) and the maximal oral bioavailability (F max ) were calculated using the following parameters: liver blood flow, 4.2 L/h/kg rat; specific liver weight, 32 g/kg rat body weight; liver cells in vivo, 1.1 × 10 8 cells/g liver; liver cells in vitro, 0.5 × 10 6 /mL.
Caco2 Drug Permeability Assay. Caco-2 cells (purchased from DSMZ Braunschweig, Germany) were seeded at a density of 4.5 × 10 4 cell per well on 24-well insert plates, 0.4 μm pore size, and grown for 15 days in DMEM medium supplemented with 10% fetal bovine serum, 1% GlutaMAX (100×, GIBCO), 100 μg/mL penicillin, 100 μg/mL streptomycin (GIBCO), and 1% nonessential amino acids (100×). Cells were maintained at 37°C in a humified 5% CO 2 atmosphere. Medium was changed every second to third day. Before running the permeation assay, the culture medium was replaced by an FCS-free Hepes-carbonate transport buffer (pH 7.2). For assessment of monolayer integrity, the transepithelial electrical resistance (TEER) was measured. Test compounds were dissolved in DMSO and added either to the apical or basolateral compartment in a final concentration of 2 μM. Before and after incubation at 37°C, samples were taken from both compartments. Analysis of compound content was performed following precipitation with methanol and LC/MS/MS analysis. Permeability (P app ) was calculated in the apical to basolateral (A → B) and basolateral to apical (B → A) directions. The apparent permeability was calculated using following equation: where V r = volume of medium in the receiver chamber, P 0 = measured peak area of the test drug in the donor chamber at t = 0, S = surface area of the monolayer, P 2 = measured peak area of the test drug in the acceptor chamber after incubation for 2 h, and t = incubation time.
The efflux ratio basolateral (B) to apical (A) was calculated by dividing P app (B−A) by P app (A−B). In addition, the compound recovery was calculated. Reference compounds were analyzed in parallel as assay controls. All samples were analyzed by LC/MS/MS.

* S Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jmedchem.5b01890. Additional figure and tables of PK and exposure data, selectivity data, and crystallographic data collection and refinement statistics (PDF) Molecular formula strings (CSV)

Accession Codes
The coordinates and structure factors for the described crystal structures have been deposited with the Protein Data Bank (PDB). The PDB accession codes are 5ARF (compound 6) and 5ARG (S)-4.