Design, Synthesis, and Biological Evaluation of PKD Inhibitors

Protein kinase D (PKD) belongs to a family of serine/threonine kinases that play an important role in basic cellular processes and are implicated in the pathogenesis of several diseases. Progress in our understanding of the biological functions of PKD has been limited due to the lack of a PKD-specific inhibitor. The benzoxoloazepinolone CID755673 was recently reported as the first potent and kinase-selective inhibitor for this enzyme. For structure-activity analysis purposes, a series of analogs was prepared and their in vitro inhibitory potency evaluated.


Introduction
Protein Kinase D (PKD) is a member of a novel family of serine/threonine kinases and diacylglycerol (DAG) receptors that have emerged as key regulators of many important cellular processes. The PKD family is now recognized as a subfamily of the Ca 2+ /calmodulin-dependent kinase superfamily. To date, three isoforms of PKD have been identified: PKD1 (formerly PKCµ) [1,2],

OPEN ACCESS
PKD2 [3], and PKD3 (formerly PKCν) [4]. All three isoforms share a highly homologous sequence and a distinct structure that includes a catalytic domain, a pleckstrin homology (PH) domain, which mediates protein-protein interactions and PKD autoinhibition, and an N-terminal cysteine-rich DAG/phorbol ester binding domain (C1 domain). Both the activation of PKD and the regulatory mechanisms that control PKD activity have been well documented. PKD is activated by DAGresponsive protein kinase C (PKC) isoforms via transphosphorylation of the conserved Ser 744 and Ser 748 within the activation loop of PKD [5][6][7]. Subsequent autophosphorylation at multiple sites, including Ser 916 , confers full, sustained activity [8,9]. PKD is also subject to spatial regulation by DAG or phorbol esters and, as a result, is capable of shuttling between different subcellular compartments [10][11][12]. Accordingly, this canonical DAG/PKC/PKD pathway is crucial to fundamental PKD function in cells.
Although extensive analysis of the role of PKD in biological processes has been hampered by the lack of both a three-dimensional structure and PKD specificity of early inhibitors (i.e. the isoquinoline sulfonamide H89 [39], staurosporine analogs [40,41], and resveratrol [42][43][44]), more potent and selective inhibitors have been reported recently [32,38,[45][46][47][48][49][50][51]. The first breakthrough in this area came in 2008 with the identification and characterization of CID755673 (Figure 1), a potent and PKDselective kinase inhibitor [52]. CID755673 was reported to inhibit all PKD isoforms with an IC 50 of 200-300 nM, but showed specificity toward PKD over several related kinases [52]. Interestingly, CID755673 was not competitive with ATP for enzyme inhibition, suggesting an alternate binding site on the enzyme, which may account for the selectivity for PKD compared to other protein kinases. This compound was also shown to effectively block PKD-mediated cell functions as well as the tumorpromoting functions of PKD in prostate cancer cells [52,53]. Despite the high specificity of CID755673 and its potent in vitro inhibition of PKD, its cellular activity was relatively weak (EC 50 = 11.8 µM) [52,53].
In addition to CID755673, promising ATP-competitive inhibitors continue to emerge in the literature [32,45,46,[48][49][50][51]. The novel 2,6-naphthyridine 1a (Figure 1) was identified by a high throughput screen (HTS) as a dual PKC/PKD inhibitor. Modification of this chemotype led to the orally available naphthyridine inhibitors 1b and 1c (Figure 1) [50]. Both 1b and 1c were able to block PKD phosphorylation and nuclear export of HDAC in the target tissue in vivo; however, their moderate kinase selectivity complicated data interpretation. In an effort to reduce the off-target effects and improve PKD selectivity, a series of amidobipyridyl-based analogs was generated [51]. From this series, BPKDi (Figure 1) was identified as a potent and selective inhibitor of all three isoforms of PKD with single digit nanomolar IC 50 values and improved selectivity relative to 1b [32,51]. Furthermore, BPKDi caused substantial inhibition of PKD1 signal-dependent phosphorylation and increased nuclear retention of class HDAC4 and HDAC5 in cardiomyocytes [32].  (Figure 1), was also reported to inhibit all isoforms of PKD with single-digit nanomolar IC 50 s. CRT0066101 was shown to block cell proliferation, induce apoptosis, and reduce the viability of pancreatic cancer cells both in vitro and in vivo. As with the aforementioned ATP-competitive inhibitors, the most attractive features of CRT0066101 are its reported orally availability and efficacy in vivo [46,48]. Lastly, preliminary studies of the structureactivity relationships (SAR) of a novel 3,5-diarylazole 2a (Figure 1), which was identified in a HTS as a moderately potent kinase inhibitor, led to a series of promising benzamide analogs [49]. One analog in particular, 2b (Figure 1), was found to inhibit all PKD isoforms with low nanomolar IC 50 s, while showing a 9-and 3-fold preference for PKD1 versus PKD2 and PKD3, respectively. Notably, 2b shows high selectivity for PKD against a panel of other kinases, and pharmacokinetic studies in rats indicate that compound 2b is orally available [49].
While all of these ATP-competitive, orally available PKD inhibitors represent valuable tools for further study of PKD signaling, their substantial off-target activity may likely be due to the high sequence homology near the hinge-binding regions of PKCs and PKDs. Recent evidence suggests additional targets of CID755673 [54], but as this lead structure is not competitive with ATP for PKD inhibition, it can provide an orthogonal approach to gain further understanding of the structure and function of PKD. In an effort to enhance the selectivity and potency for potential in vivo applications, small molecule analogs of CID755673 were generated by modification of the core structure as well as the side chains. We describe herein the complete SAR conducted thus far, which led to the discovery of a novel benzothienothiazepinone series. The improved PKD1 inhibitory activity of some of these analogs has already been highlighted in previous communications [53,55].

First generation SAR
Our investigations began with the chromenopyridine-based CID797718, a by-product of the synthesis of the parental compound, CID755673 (Table 1). This compound was 10x less potent at PKD1 inhibition than CID755673. Efforts to improve the activity of CID797718 by substitution of the phenolic hydroxyl group ( Polarization (FP) assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments.
Modifications to the benzofuroazepinone scaffold present in CID755673 proved to be more rewarding. The initial SAR analysis included modifications of the azepinone ring, substitution at the phenol and amide groups, and functionalization across the aryl moiety (Table 2). Changing the azepinone to either a 6-or 8-membered heterocycle resulted in a reduced inhibitory activity towards PKD1 (Table 2, entries 2-5). Substitutions of the phenolic hydroxyl group were not well tolerated either. Although both the methoxy and allyloxy substituents at that site showed moderate inhibitory activity towards PKD1 (Table 2, entries 6 and 7), the acetoxy and siloxy groups ( Table 2, entries 8 and 9) showed poor inhibition and complete loss in activity, respectively. These results indicate that the aryl binding pocket of the protein may be restrictive in terms of both the size and the polarity of the aryl substituents. Similarly, alkylation/acylation of the amide or bioisosteric replacement with a dihydroimidazole resulted in weak or no PKD1 inhibition (Table 2, entries [10][11][12][13][14]. These observations suggest that the amide functionality may provide crucial H-bonding interactions in the azepinone binding pocket that are required for optimal inhibitor-enzyme interactions. Polarization (FP) assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. c Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least two independent experiments; n = number of independent experiments.
Based on these results, subsequent synthetic analogs were biased to contain key features of the pharmacophore, i.e., the 7-membered ring azepine, the monosubstituted lactam, and a small R 2 substituent (Table 3). Derivatives bearing either a Cl or F atom at the R 1 position retained activity against PKD1, but did not show improved potency relative to CID755673 (Table 3, entries 1 and 2). In contrast, introduction of a Cl atom at either the R 3 position, or both the R 1 and R 3 positions, provided inactive compounds (Table 3, entries 4 and 5). Interestingly, allyl and propenyl groups were well tolerated in this region, despite the aforementioned putative size restrictions of the aryl binding region (Table 3, entries 6 and 7). The inhibitory activity of these analogs also provides further support for a relatively hydrophobic aryl binding pocket. Finally, all modifications of the azepinone moiety at the 5 position (Z) yielded analogs with reduced inhibitory activity ( Table 3, entries 8-12). Polarization (FP) assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. We next sought to investigate the replacement of the benzofuran core by a β-carboline skeleton (Table 4). Compound kb-NB123-57 (Table 4, entry 2), which retains all functionality of CID755673 except for the benzofuran oxygen being replaced with a nitrogen atom, exhibited comparable activity toward PKD1; however, there was no significant cellular activity. Variation of either the phenolic hydroxyl group or the lactam ring size in the β-carboline series did not provide any enhancement in activity (Table 4), and in cases where R 2 was replaced with O-benzyl (Table 4, entries 3 and 4) or Nacyl (Table 4, entries 7 and 8), all inhibitory activity was lost. These modifications, accordingly, confirm that the 7-membered azepinone represents an optimal size and that the binding pocket of the protein is sterically demanding at the aryl binding region. Polarization (FP) assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least two independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. c Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments; n = number of independent experiments.
In the next SAR iteration, the benzofuran was replaced with a benzothiophene while the azepine was replaced by a thiazepine ring. Gratifyingly, both the benzothienothiazepinone kb-NB142-70 and the methoxy analog kb-NB165-09 showed a significance increase in potency toward PKD1, with IC 50 values of 28.3 nM and 82.5 nM, respectively ( Table 5, entries 2 and 3). This increased potency was further confirmed in cell-based assays, for which the cellular EC 50 was lowered ca. 5-fold from 11.8 µM by CID755673 to 2.2 µM and 3.1 µM by kb-NB142-70 and kb-NB165-09, respectively. The remarkable activity of the methyl ether compared to the phenol indicates that a hydrogen bond donor at this position is not critical for activity. The lack of activity of the O-benzyl derivative for this series of analogs provides additional support for a sterically demanding aryl binding pocket. Polarization (FP) assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments with triplicate determinations at each concentration in each experiment; n = number of independent experiments. c Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments; n = number of independent experiments.

Second generation SAR
The previous screening cycles led to the discovery of the benzothienothiazepinone kb-NB142-70 as the most potent analog with an IC 50 of 28.3 nM, which was an almost 7-fold improvement in potency over the parent compound, CID755673. Furthermore, we had gained considerable understanding of the SAR of the tricyclic inhibitor scaffold. Accordingly, we updated our pharmacophore model based on the structure of kb-NB142-70, and devised 4 major structural zones ( Figure 2): zone I (aryl moiety), zone II (thiophene), zone III (thiazepinone), and zone IV (amide function). Each of these zones was subjected to selected structural modifications in order to elucidate essential activity relationships.
In zone I, the phenolic substituent and the substitution across the aryl moiety were modified ( Table 6). Most of the zone I analogs were less active than the lead, kb-NB142-70. Specifically, substitutions at R 4 were detrimental to in vitro activity ( Table 6, entries 7-9), at least when R 2 = H. In contrast, halogenation at R 1 ( Table 6, entries 5 and 6) and replacement of the phenolic hydroxy group with amine variants ( Table 6, entries 1-4) were surprisingly well tolerated. Notably, the azide analog, mcf292-08, maintained a high inhibitory activity both in vitro and in cells, with IC 50 values of 74.9 nM and 2.2 µM, respectively, thereby providing further support for the limited significance of a hydrogen bond donor at this position. The ortho-iodinated analog, kb-NB165-31, was also a potent inhibitor with an IC 50 value of 114 nM; however, its cellular activity was 4-fold lower than that of kb-NB142-70. Masking the hydroxyl group of the lead kb-NB142-70 was only slightly detrimental to the in vitro biological activity, as revealed by the methoxy analog kb-NB165-09. Moreover, preliminary results from an animal model suggest the glucuronidation of kb-NB142-70 at the phenolic position to be a major metabolic pathway.  kb-NB184-44 H H OMe 77.5 ± 3.6 (n = 2) n.d. n.d. a PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52].
Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments; n = number of independent experiments. Therefore, the methoxy analogs were systematically developed in addition to the corresponding phenols in order to potentially circumvent a rapid excretion scenario in vivo. As part of our zone II and III modifications, the oxidation state of the benzothiophene sulfur atom, the size of the thiazepinone ring, and the oxidation and substitution of the thiazepinone ring sulfur atom were explored (Table 7). Sulfur oxidations in zones II and III provided analogs with reduced activity ( Table 7, entries 1-3), while increasing the thiazepinone ring size by the addition of a methylene group had only minor effects on PKD1 inhibition for both the hydroxy and methoxy analogs ( Table 7, entries 5 and 6). Additionally, the replacement of the thiazepinone ring sulfur atom with an oxygen resulted in a loss of inhibitory activity (Table 7, entries 7 and 8), suggesting high hydrophobicity and polarizability to be preferred in zone III. Attempts to exchange the thiazepinone ring sulfur atom with a nitrogen atom were met with preparative difficulties. The desired diazepinone ring could not be formed, and, as a result, an acyclic precursor was submitted for testing (Table 7, entry 9). The lack of activity of this analog suggests that the zone III binding pocket may require the rigidity of a ring system for optimal binding interactions.
a PKD1 IC 50 was determined using a radiometric kinase activity assay as previously described [52].
Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least two independent experiments; n = number of independent experiments.
Furthermore, modifications to zone IV, which included functional group interconversions and replacement of the amide moiety, did not enhance the inhibitory activity (Table 8). These results suggest that the unique hydrogen bond donor-acceptor capability of the amide is critical for protein interactions within the zone IV binding pocket.
The results from both the initial and the second generation SAR firmly established that the benzothienothiazepinones displayed superior PKD1 inhibition compared to the benzofuroazepinones. The most potent analog, kb-NB142-70, inhibited PKD1 with an IC 50 of 28 nM, which is nearly a 7fold improvement in potency compared to the parental compound CID755673. This improved analog also demonstrated increased inhibition of PMA-induced autophosphorylation of endogenous PKD1 in LNCaP prostate cancer cells with an EC 50 of 2.2 µM, which represents a ca. 5-fold improvement in activity over CID755673, which has an EC 50 of 11.8 µM.  50 was determined using a radiometric kinase activity assay as previously described [52].
Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. However, despite the improvement of in vitro and cellular activities, in vivo studies revealed a short plasma half-life for this compound. In order to enhance the resistance of kb-NB142-70 towards phase I and II metabolism, modifications to zone I were explored with the goal to install a more electrondeficient heteroarene ring (Table 9). Table 9. Chemical structures and PKD1 inhibitory activities of analogs with zone I modifications to the pyrimidine scaffold. 50 was determined using a radiometric kinase activity assay as previously described [52].
Each IC 50 was calculated as the mean ± SEM with triplicate determinations at each concentration in each experiment; n = number of independent experiments. b Cellular IC 50 was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at S 916 in LNCaP cells as previously described [53]. Each IC 50 was calculated as the mean ± SEM of at least three independent experiments; n = number of independent experiments. Gratifyingly, the methoxypyrimidine kmg-NB4-23 exhibited an IC 50 of 124 nM, which represents only a slight decrease in activity compared to that of the parent compound, kb-NB165-09. This result not only validates our design, but also suggests that the zone I binding pocket is tolerant to a decrease in electron density in the aryl region. Surprisingly, the hydroxy analog kmg-NB4-69A showed a significant loss in activity relative to the parental compound, kb-NB142-70. This effect can be attributed to the susceptibility of the C-4 position of the pyrimidine towards nucleophilic attack [56]; the heterocycle spontaneously degrades in protic solvents, resulting in reduced inhibitory activity towards PKD1. More hydrolytically stable pyrimidine analogs containing a substituent at the R 4 position also exhibited a dramatic loss in activity, which is consistent with the general SAR results and provides additional support for a sterically limited zone I binding pocket. Unfortunately, the low aqueous solubility (<0.4 mg/mL even in the presence of lipophilic solubilizing agents) of kmg-NB4-23 prevented the in vivo evaluation of this compound.

Synthetic chemistry
CID755673 was initially formed in conjunction with CID797718 in a synthetic route starting with commercially available ε-caprolactam, which was dibrominated at the α-position [57], and then treated with piperidine to afford the known α-oxolactam enamine 4 (Scheme 1) [58]. HCl, reflux.
A Nenitzescu reaction between 4 and para-benzoquinone resulted in the formation of cyclic adduct 5, which was in turn subjected to heating in aqueous acid [59] to provide the desired CID755673. Although CID797718 was a byproduct of the synthesis of the target compound CID755673, it showed moderate activity and was therefore a desirable starting point for analog synthesis (Scheme 2). Introduction of a chlorine atom to the chromenopyridine scaffold was achieved by treating enamine 4 with 2-chloro-1,4-benzoquinone, followed by the same acid-promoted cis-trans isomerizationelimination as aforementioned in Scheme 1 (Scheme 3). The chlorinated analogs were isolated as a separable mixture of isomers. Unfortunately, all of the chromenopyridine-based analogs were inactive in the PKD1 inhibition assays, and therefore this scaffold was abandoned. Our studies on the modification of CID755673 began with the azepinone ring (Scheme 4). The syntheses of 6-and 8-membered ring analogs were carried out in an analogous manner to the sequence shown in Scheme 1. The acid-mediated piperidine elimination reaction of the 8-membered ring derivative 7b proceeded with complete chemoselectivity toward the formation of the desired product kb-NB96-53. In contrast, treatment of 7a with acetic acid provided the N-acetylated analog of the desired compound and the corresponding chromenopyrrole as the major products. Attempts to cleave the N-acetyl functionality to obtain the desired analog kb-NB123-23A were unsuccessful. Gratifyingly, subjecting 7a to mCPBA at elevated temperatures afforded the desired 6-membered azepinone ring analog, kb-123-23A, in good yield. With the 6-and 8-membered ring analogs in hand, we next substituted the phenolic hydroxyl group and alkylated the azepinone amide nitrogen atom (Scheme 4). Treatment of kb-NB123-23A with MeI and K 2 CO 3 provided the O-methylated compound, kb-NB123-32, while use of a stronger base such as KOt-Bu gave the desired dimethylated analog, kb-NB123-37, in good yield. The 8-membered dimethylated product kb-NB96-59 was obtained in a similar manner to kb-NB123-37, although only in modest yield.
Due to the lack of potency improvements for the 6-and 8-membered azepinone analogs, further modifications to the aryl region of the parental compound CID755673 were explored. Functionalization of the phenol was achieved by treatment of CID755675 with base in the presence of an electrophile (Table 10, entries 1-4). The dimethylated compound kb-NB96-04 and the diacylated compound kb-NB123-45-1 were obtained in a similar manner by the use of either a stronger base (KOt-Bu) or additional equivalents of the desired electrophile, respectively (Table 10, entries 5 and 6). Modification of the aryl region also included additions of halogens (Table 10, entries 7-9). These derivatives could be obtained by the treatment of CID755673 with either N-chlorosuccinimide in the case of the chlorinated analogs kb-NB77-88 and kb-NB96-43, or Selectfluor ® in the case of the fluorinated analog kb-NB96-21.
Further modifications of these analogs were carried out in order to obtain a more complete SAR for the benzofuroazepinone series (Scheme 5). O-Allylation and N-methylation were performed on kb-NB77-88 and kb-NB77-77, respectively. Additionally, the microwave-mediated Claisen rearrangement of the O-allylated-benzoxoloazepinolone kb-NB77-84 provided the target compound kb-NB96-02 with the allyl functionality at the R 1 position. Protection of the phenol of kb-NB96-02 with a silyl group provided a derivative which could be subjected to Ru-catalyzed olefin isomerization conditions. Subsequent TBS deprotection provided the isomerized product kb-NB96-30 in modest yield over the three steps. In addition to alkylations and acetylations of the amide moiety, we also synthesized kb-NB165-15 to evaluate the effect of the isoelectronic replacement of the amide with an imidazoline ring (Scheme 6). A TIPS protection of the parental compound CID755673 followed by N-alkylation provided 8. Debenzylation by transfer hydrogenation, mesylation of the primary alcohol, and displacement of the resulting mesylate with sodium azide afforded the desired alkyl azide 9 in good yield over the three steps. After protection of the phenol, the alkyl azide was reduced to the amine and protected, prior to treatment with Lawesson's reagent to furnish thiolactam 10. Cyclodethionation of thiolactam 10 [60] and subsequent TIPS deprotection delivered the desired imidazole containing analog, kb-NB165-15.
At this time, all modifications to the azepinone ring including altering the ring size, Nalkylations/acetylations, and isoelectronic replacements of the amide moiety had resulted in inhibitors with either reduced potency toward PKD1 or no inhibitory effect at all. Thus, we explored the introduction of functionality on the azepinone methylene groups to determine if this region of the molecule could be optimized further (Scheme 7). The synthesis of these analogs began with the oxidation of the O-acetylated derivative kb-NB123-36 with PDC-TBHP [61] in the presence of neutral alumina in order to install the desired ketone (11). The use of sonication proved to significantly improve both the yields and reproducibility of this oxidation. Subsequent acetyl deprotection in methanolic potassium carbonate solution provided kb-NB123-63. Access to further functionalized derivatives was achieved by acid-catalyzed condensation of kb-NB123-63 with O-benzylhydroxylamine or N-substituted hydrazines (Scheme 7). The newly installed keto functionality was reduced using sodium borohydride to furnish the hydroxy derivative kb-NB123-89.
We also synthesized a thio variant of the benzofuroazepinone scaffold to study the effects of sulfur incorporation (Scheme 9). This analog was constructed by a thionyl chloride-mediated Higa cyclization [67] of benzyl protected hydroxycinnamate 16 to a benzo[b]thiophene derived acid chloride, which was subsequently converted to the methyl ester 17. Treatment of 17 with cysteamine hydrochloride in the presence of DBU furnished the benzyloxy-benzothienothiazepinone kb-NB123-66 [68]. Deprotection of the aryl benzyl ether kb-NB123-66 with boron tribromide provided kb-NB142-70 in good yields, and subsequent O-methylation gave kb-NB165-09 in excellent yield (Scheme 9). Gratifyingly, kb-NB142-70 was identified as the most potent analog with an IC 50 of 28.3 nM for PKD1, which represents nearly 7-fold higher potency over the parental benzofuroazepinone CID755673. We therefore dissected this lead compound into four major structural zones in order to further probe the SAR, as shown in Figure 2.
In zone I, we modified the substituent on the phenolic hydroxyl group and the position of this group (Scheme 9). The positional isomer 18 could be isolated as a byproduct from the thionyl chloride- In contrast, the brominated analog kb-184-52 was synthesized from the corresponding sulfoxide kb-NB184-45 upon treatment with BBr 3 [Scheme 10 (b)]. Additional zone I modifications, which included replacement of the phenol group by nitrogen, required the development of an alternative route (Scheme 11 and Table 11). Nucleophilic aromatic substitution of methyl 2-chloro-5-nitrobenzoate by methyl thioglycolate anion followed by immediate Dieckman cyclization [63,69] afforded the benzothiophene precursor 19. Cyclization of the corresponding triflate 20 with cysteamine hydrochloride provided the desired tricyclic core 21 in 50% yield (67% based on recovered starting material 19, Scheme 11). Subsequent reduction of the nitro group furnished aniline mcf292-03, which was further functionalized by treatment with t-butyl nitrite and TMS-azide using Moses' method [70] to yield the aryl azide, mcf292-08 (Table 11, entry 1). The synthesis of the isothiocyanate mcf292-05 was realized by subjecting mcf292-03 to the modified Schotten-Baumann conditions reported by Nowick and coworkers [71] (Table 11, entry 2). Lastly, treatment of mcf292-03 with chlororacetyl chloride in the presence of 2,6-lutidine provided the desired chloroacetyl analog mcf292-09 in modest yield over two steps (   In our pursuit of zone II SAR, the sulfur ring atom of 17 was oxidized with trifluoroperacetic acid to the 3-chlorobenzo[b]thiophene-1-oxide 22 (Scheme 12). The vinylogous chloride 22 was then converted to the corresponding benzyloxybenzothienothiazepinone-6-oxide 23, which upon benzyl deprotection and methylation provided the desired sulfoxide analogs, kb-NB184-22 and kb-NB184-25, respectively (Scheme 12).  In an effort to further assess the SAR effects of modifying zone III, we investigated the synthesis of benzothiophene analogs linked to a three-carbon chain by an ether or amine function, instead of the thioether present in the lead compound. Key to accomplishing the synthesis of the ether analog kb-NB184- 36  In an analogous manner, the synthesis of diazepinone analog 31 was attempted; however, precursor 30 failed to undergo cyclization under both base-and Cu-mediated conditions [Scheme 15 (a)] [73]. Therefore, we opted for a nucleophilic displacement at the activated chlorine atom in 29 with aminopropanol, followed by a TMSI-mediated deoxygenation of 32 to provide kb-NB184-80 The N-methylated analog kb-NB165-17 was prepared from the benzyloxybenzothiazepinone kb-NB123-66 (Scheme 9) through an alkylation-deprotection sequence, while the dialkylated kb-NB165-16 was obtained in one step by treatment with NaH and MeI. The synthesis of kb-NB165-75 was accomplished via the N-alkylation of silyl protected kb-NB142-70 with the corresponding alkyl iodide, followed by functional group interconversions and a final deprotection to furnish the desired analog kb-NB165-75. The amide reduction of thiolactam 36 proceeded poorly upon treatment with Raney-Nickel in THF due to a competitive cleavage of the other C-S bonds present in this system (Scheme 17). Nonetheless, we were able to isolate kb-NB165-81 in low yields and debenzylate it to afford the desired phenol kb-NB165-83 (Scheme 17). As mentioned in Section 2.2, preliminary in vivo studies revealed that the lead compound kb-NB142-70 and the methoxy analog kb-NB165-09 possessed a short plasma half-life (data not shown). Therefore, we further explored a zone I modification of kb-NB142-70 to install a more electrondeficient pyrimidine moiety in place of the phenol ether, a known site of active phase I and II metabolism. The synthetic route to arrive at this new thiazepinothiophenopyrimidinone scaffold is summarized in Scheme 18. Starting with commercially available methyl 3-aminothiophene-2-carboxylate, formation of the pyrimidine moiety using potassium cyanate and chlorination with POCl 3 provided dichloride 38 [74]. Regioselective palladium-catalyzed hydrogenolysis of 38 in the presence of Na 2 CO 3 occurred exclusively at the C-4 position [75], and substitution of the remaining C-2 chloride with methoxide provided 40 in 79% yield over the two steps. Electrophilic bromination of 40 using bromine in acetic acid gave the desired C-7 bromo compound 41. Functionalization at C-6 was accomplished via selective metalation and trapping with Mander's reagent to provide the required cyclization precursor 42 [76]. Formation of the thiazepinone moiety was achieved by a one-pot nucleophilic displacementcondensation of 42 with cysteamine hydrochloride to provide the desired methoxypyrimidine kmg-NB4-23 in good yield. The structure of kmg-NB4-23 was confirmed by x-ray analysis (Figure 3; this structure has been deposited at the Cambridge Crystallographic Data Centre and allocated the deposition number CCDC 822403). When kmg-NB4-23 was subjected to 4 M HCl in 1,4-dioxane, the desired hydroxypyrimidine kmg-NB4-69A was formed as the hydrochloride salt (Scheme 18). These analogs were successfully synthesized in a similar manner to kmg-NB4-23 and kmg-NB4-69A (Scheme 18). Unfortunately, preliminary biological data indicate an absence of inhibitory activity for both kmg-NB5-13 and kmg-NB5-15.

Determination of cellular IC 50 for PKD1 inhibition
Inhibition of PKD1 in cells was determined by densitometry analysis of Western blotting data for PKD1 autophosphorylation at Ser 916 in LNCaP cells as previously described [52]. LNCaP cells were pre-treated with PKD inhibitors for 45 min at various concentrations, and then stimulated with 10 nM phorbol 12-myristate 13-acetate (PMA) for 20 min. Cells were then collected and lysed in lysis buffer containing 200 mM Tris-HCl, pH 7.4, 100 µM 4-(2-aminoethyl) benzenesulfonyl fluoride, 1 mM EGTA, and 1% Triton X-100. Cell lysates were probed by Western blot analysis using primary antibodies targeting p-S916-PKD1 (Millipore), PKD1 (Cell Signaling Technology), or GAPDH. Densitometry analysis of visualized bands was used to determine the cellular IC 50 values for PKD1 inhibition.

Statistical analysis
GraphPad Prism V software was used to determine statistical significance. Each assay was repeated two or three times with triplicate determination at each point. A p value of < 0.05 was considered significant.

Conclusions
An extensive SAR study starting with the HTS-based confirmed hit CID755673 led to the discovery of the benzothienothiazepinone kb-NB142-70 and its methoxy analog kb-NB165-09, which were ca. 7-fold more potent at PKD1 inhibition. A 4-zone pharmacophore model was developed based on these new lead compounds, and systematic modifications of each zone led to a novel series of benzothienothiazepinones with equal or greater potencies compared to CID755673. Five of these analogs were further investigated, and each of them inhibited PMA-induced autophosphorylation of PKD1 [53]. These analogs also caused dramatic arrest in cell proliferation and inhibited cell migration and invasion [53]. Furthermore, counterscreens on a panel of other kinases clearly demonstrated a high selectivity profile in this series for PKD inhibition [53]. In order to decrease the susceptibility of these compounds to metabolic pathways, a pyrimidine moiety was installed in zone I, leading to a promising new lead structure, kmg-NB4-23. Because of the inadequate solubility properties of the pyrimidinebased inhibitors, additional modifications to zone I are currently being explored in our laboratories. Overall, the results of this extensive SAR study suggest that the zone I binding region is restrictive in both size and polarity of the aryl substituents. Additionally, it appears that the hydrogen bond donoracceptor abilities of the amide moiety are critical for optimal interactions with the zone IV binding site. The synthesis of the azide analog mcf292-08 and the electrophilic reactive probe molecules mcf292-05 and mcf292-09 as tools for photoaffinity/affinity labeling studies was undertaken to gain key structural insights into the possible protein binding site. The latter information is currently pending and will be utilized for future SAR studies.