Design, Synthesis and Biological Evaluation of Isoxazole-Based CK1 Inhibitors Modified with Chiral Pyrrolidine Scaffolds

In this study, we report on the modification of a 3,4-diaryl-isoxazole-based CK1 inhibitor with chiral pyrrolidine scaffolds to develop potent and selective CK1 inhibitors. The pharmacophore of the lead structure was extended towards the ribose pocket of the adenosine triphosphate (ATP) binding site driven by structure-based drug design. For an upscale compatible multigram synthesis of the functionalized pyrrolidine scaffolds, we used a chiral pool synthetic route starting from methionine. Biological evaluation of key compounds in kinase and cellular assays revealed significant effects of the scaffolds towards activity and selectivity, however, the absolute configuration of the chiral moieties only exhibited a limited effect on inhibitory activity. X-ray crystallographic analysis of ligand-CK1δ complexes confirmed the expected binding mode of the 3,4-diaryl-isoxazole inhibitors. Surprisingly, the original compounds underwent spontaneous Pictet-Spengler cyclization with traces of formaldehyde during the co-crystallization process to form highly potent new ligands. Our data suggests chiral “ribose-like” pyrrolidine scaffolds have interesting potential for modifications of pharmacologically active compounds.


Introduction
To date, many drug discovery programs have been based on commercially or readily available building blocks and reagents to enable rapid synthesis of large compound libraries destined for various screening campaigns [1]. This approach often resulted in collections of large, "flat", achiral and low-diverse molecules [2,3]. In this context, the "quality" of compounds and building blocks in terms of complexity, carbon bond saturation, chirality or natural product-likeness has been intensively discussed in the medicinal chemistry community [4]. Usually the central heterocycle binds to the "adenine region" of the ATP binding site and undergoes hydrogen bonding to the so-called hinge region of the kinase, whereas the appended side chains of the heterocycle facilitate binding interactions within the "hydrophobic pocket I" and the solvent-exposed "hydrophobic region II" [5,6]. This approach inevitably gives rise to lipophilic, "flat" compounds that are often found to come with poor pharmakokinetic ADME parameters at the later stages of drug development [8].
In contrast, the more polar regions of the ATP binding site (ribose pocket and phosphate-binding region) have been exploited less frequently. This may be because binding to these hydrophilic waterexposed areas does not necessarily translate into an increased binding enthalpy of an inhibitor. Therefore, these regions are often only used to implement water-solubilizing groups (e.g., functionalized tetrahydropyran (1), pyrrolidine (2) or piperidine moieties) that enhance pharmacokinetic parameters, such as oral bioavailability ( Figure 2) [9][10][11][12][13]. However, residues that comprise the ribose pocket have also been successfully targeted through stereochemically defined interactions, to create highly potent and selective inhibitors [14][15][16][17][18]. Examples include Pfizer's interleukin-1 receptor-associated kinase 4 (IRAK4) inhibitor PF06650833 (3) [18], which is equipped with a chiral pyrrolidone scaffold, Merck's carboribose-functionalized diaminopyrimidine derivative 4 [17] and a series of p38 mitogen-activated protein kinase (MAPK) inhibitors (e.g., 5) developed by Laufer and co-workers [16]. A special case is the well-known microbial alkaloid staurosporine (6) that targets the ribose pocket with a chiral "glycosyl" subunit (cf. Figure 2). This moiety is reported to contribute to the high potency, but also to the very high kinase promiscuity of staurosporine (6) [19][20][21][22][23]. Interestingly, modification of the sugar-like moiety of several closely related indolocarbazole analogs resulted in enhanced protein kinase selectivity profiles (e.g., 7) [24][25][26]. Taken together, these Usually the central heterocycle binds to the "adenine region" of the ATP binding site and undergoes hydrogen bonding to the so-called hinge region of the kinase, whereas the appended side chains of the heterocycle facilitate binding interactions within the "hydrophobic pocket I" and the solvent-exposed "hydrophobic region II" [5,6]. This approach inevitably gives rise to lipophilic, "flat" compounds that are often found to come with poor pharmakokinetic ADME parameters at the later stages of drug development [8].
In contrast, the more polar regions of the ATP binding site (ribose pocket and phosphate-binding region) have been exploited less frequently. This may be because binding to these hydrophilic water-exposed areas does not necessarily translate into an increased binding enthalpy of an inhibitor. Therefore, these regions are often only used to implement water-solubilizing groups (e.g., functionalized tetrahydropyran (1), pyrrolidine (2) or piperidine moieties) that enhance pharmacokinetic parameters, such as oral bioavailability ( Figure 2) [9][10][11][12][13]. However, residues that comprise the ribose pocket have also been successfully targeted through stereochemically defined interactions, to create highly potent and selective inhibitors [14][15][16][17][18]. Examples include Pfizer's interleukin-1 receptor-associated kinase 4 (IRAK4) inhibitor PF06650833 (3) [18], which is equipped with a chiral pyrrolidone scaffold, Merck's carboribose-functionalized diaminopyrimidine derivative 4 [17] and a series of p38 mitogen-activated protein kinase (MAPK) inhibitors (e.g., 5) developed by Laufer and co-workers [16]. A special case is the well-known microbial alkaloid staurosporine (6) that targets the ribose pocket with a chiral "glycosyl" subunit (cf. Figure 2). This moiety is reported to contribute to the high potency, but also to the very high kinase promiscuity of staurosporine (6) [19][20][21][22][23]. Interestingly, modification of the sugar-like moiety of several closely related indolocarbazole analogs resulted in enhanced protein kinase selectivity profiles (e.g., 7) [24][25][26]. Taken together, these examples illustrate the potential of a stereochemical approach to developing potent and selective kinase inhibitors [19,27,28]. examples illustrate the potential of a stereochemical approach to developing potent and selective kinase inhibitors [19,27,28]. In 2009, our group reported isoxazole 8 (Scheme 1) as a nanomolar inhibitor of the protein kinase CK1δ [IC50 (CK1δ) = 0.033 μM] [29]. CK1δ is a Ser/Thr-specific protein kinase and is one of seven mammalian CK1 isoforms (α, β, γ1, γ2, γ3, δ and ε) within the CK1 family. All CK1 isoforms share a highly conserved kinase domain [30] and phosphorylate a high number of substrates, including the tumor suppressor protein p53 [30,31]. Based on the wide spectrum of substrate proteins CK1 family members play an important role in the diverse signaling pathways that are involved in cell division, apoptosis, membrane transport, immune response and inflammation, spindle and centrosomeassociated processes, DNA damage-related signal transduction and circadian rhythm. Deregulation and dysfunction of CK1 isoforms are associated with proliferative disorders, such as cancer [32]. Moreover, mutations, as well as disorders of CK1 expression and activity, have been observed in neurodegenerative diseases, such as Alzheimer's or Parkinson's disease, as well as sleeping disorders [32,33].
In the present study, we have modified diaryl-isoxazole 8 with functionalized enantiopure pyrrolidine scaffolds (Scheme 1) to promote selective binding interactions in the more hydrophilic areas of the ATP binding pocket. These kinds of pyrrolidines scaffolds have been frequently used in various medicinal chemistry projects, including the development of glycosidase inhibitors, azanucleosides and antiviral agents [34][35][36][37][38]. Especially hydroxy-functionalized pyrrolidines are remarkable scaffolds, as they can be regarded as sugar analogues with the furanose ring oxygen substituted by a nitrogen atom. Due to their structural and chemical similarity these "iminosugars" are effective carbohydrate mimics and are ideal scaffolds to address carbohydrate-related targets, including nucleoside or nucleotide-binding enzymes. In line with this notion, "iminosugars" provide access to polar active sites and thus to a chemical space that is usually inaccessible to Lipinskicompliant molecules [34]. In 2009, our group reported isoxazole 8 (Scheme 1) as a nanomolar inhibitor of the protein kinase CK1δ [IC 50 (CK1δ) = 0.033 µM] [29]. CK1δ is a Ser/Thr-specific protein kinase and is one of seven mammalian CK1 isoforms (α, β, γ1, γ2, γ3, δ and ε) within the CK1 family. All CK1 isoforms share a highly conserved kinase domain [30] and phosphorylate a high number of substrates, including the tumor suppressor protein p53 [30,31]. Based on the wide spectrum of substrate proteins CK1 family members play an important role in the diverse signaling pathways that are involved in cell division, apoptosis, membrane transport, immune response and inflammation, spindle and centrosome-associated processes, DNA damage-related signal transduction and circadian rhythm. Deregulation and dysfunction of CK1 isoforms are associated with proliferative disorders, such as cancer [32]. Moreover, mutations, as well as disorders of CK1 expression and activity, have been observed in neurodegenerative diseases, such as Alzheimer's or Parkinson's disease, as well as sleeping disorders [32,33].

Molecular Modelling
To understand how the lead compound 8 could be elaborated to exploit residues in the ribose Scheme 1. Inhibitor design concept. Based on protein kinase CK1δ inhibitor 8 the cinnamic acid side chain was modified to attach chiral pyrrolidine scaffolds.
In the present study, we have modified diaryl-isoxazole 8 with functionalized enantiopure pyrrolidine scaffolds (Scheme 1) to promote selective binding interactions in the more hydrophilic areas of the ATP binding pocket. These kinds of pyrrolidines scaffolds have been frequently used in various medicinal chemistry projects, including the development of glycosidase inhibitors, azanucleosides and antiviral agents [34][35][36][37][38]. Especially hydroxy-functionalized pyrrolidines are remarkable scaffolds, as they can be regarded as sugar analogues with the furanose ring oxygen substituted by a nitrogen atom. Due to their structural and chemical similarity these "iminosugars" are effective carbohydrate mimics and are ideal scaffolds to address carbohydrate-related targets, including nucleoside or nucleotide-binding enzymes. In line with this notion, "iminosugars" provide access to polar active sites and thus to a chemical space that is usually inaccessible to Lipinski-compliant molecules [34].

Molecular Modelling
To understand how the lead compound 8 could be elaborated to exploit residues in the ribose pocket, compound 8 was docked into a CK1δ-PF670462 co-crystal structure (pdb: 3UZP). According to the predicted binding mode the amidopyridinyl moiety occupies the adenine region of the ATP binding site and forms a bidentate interaction with the backbone NH and carbonyl oxygen of Leu-85 ( Figure 3). The isoxazole ring is predicted to pack between the sidechains of Ile-23 and Ile-148, and to form a hydrogen bond between the ring nitrogen atom and a structural water. The p-fluorophenyl substituent packs tightly in the hydrophobic pocket I, formed between the sidechains of Lys-38, Met-80 and the gate keeper residue Met-82. This characteristic "teardrop"-shaped [39] binding mode agrees with other related 4,5-diarylimidazoles that have been crystallized with CK1δ, as well as p38α MAPK [40]. Regarding isoxazole 8 this pharmacophore is extended towards the solvent-exposed hydrophobic region II by a cinnamic acid moiety. This configuration is projecting the aryl system to the solvent exposed area but leaving space for further interactions within the actual binding pocket.

Molecular Modelling
To understand how the lead compound 8 could be elaborated to exploit residues in the ribose pocket, compound 8 was docked into a CK1δ-PF670462 co-crystal structure (pdb: 3UZP). According to the predicted binding mode the amidopyridinyl moiety occupies the adenine region of the ATP binding site and forms a bidentate interaction with the backbone NH and carbonyl oxygen of Leu-85 ( Figure 3). The isoxazole ring is predicted to pack between the sidechains of Ile-23 and Ile-148, and to form a hydrogen bond between the ring nitrogen atom and a structural water. The p-fluorophenyl substituent packs tightly in the hydrophobic pocket I, formed between the sidechains of Lys-38, Met-80 and the gate keeper residue Met-82. This characteristic "teardrop"-shaped [39] binding mode agrees with other related 4,5-diarylimidazoles that have been crystallized with CK1δ, as well as p38α MAPK [40]. Regarding isoxazole 8 this pharmacophore is extended towards the solvent-exposed hydrophobic region II by a cinnamic acid moiety. This configuration is projecting the aryl system to the solvent exposed area but leaving space for further interactions within the actual binding pocket. To extend the pharmacophore of lead compound 8 towards the ribose pocket we replaced the susceptible double bond with a five-membered pyrrole ring to mimic and to rigidify the E-configured cinnamic acid moiety. Furthermore, the pyrrole nitrogen was used as a handle to attach the chiral To extend the pharmacophore of lead compound 8 towards the ribose pocket we replaced the susceptible double bond with a five-membered pyrrole ring to mimic and to rigidify the E-configured cinnamic acid moiety. Furthermore, the pyrrole nitrogen was used as a handle to attach the chiral scaffolds (Scheme 1). Docking of the designed ligands into CK1δ revealed that the key pharmacophore interactions described above are maintained. Additionally, the chiral scaffolds are predicted to pack between Ile-15 and Asp-91 and to crosslink the β1b strand of the N-terminal lobe and the αB helix of the C-terminal lobe with H-bonds ( Figure 4). Asp-91 is one of three amino acid residues of CK1δ (beside Asp-132 and Ser-88) that are deemed to be involved in ribose binding of ATP [42,43]. While compound 29d is predicted to form H-bonds between its hydroxyl groups and the carboxyl group of Asp-91 ( Figure 4A), enantiomer 29e is predicted to bind to Asp-91 with the NH-group of the pyrrolidine ring ( Figure 4B) scaffolds (Scheme 1). Docking of the designed ligands into CK1δ revealed that the key pharmacophore interactions described above are maintained. Additionally, the chiral scaffolds are predicted to pack between Ile-15 and Asp-91 and to crosslink the β1b strand of the N-terminal lobe and the αB helix of the C-terminal lobe with H-bonds ( Figure 4). Asp-91 is one of three amino acid residues of CK1δ (beside Asp-132 and Ser-88) that are deemed to be involved in ribose binding of ATP [42,43]. While compound 29d is predicted to form H-bonds between its hydroxyl groups and the carboxyl group of Asp-91 ( Figure 4A), enantiomer 29e is predicted to bind to Asp-91 with the NH-group of the pyrrolidine ring ( Figure 4B)
Molecules 2018, 23, x FOR PEER REVIEW 6 of 34 prepared from D-methionine its enantiomer (−)-20 was either synthesized from L-methionine or, particularly at larger scale, from the available 2,3-O-isopropylidene-L-lyxono-1,4-lactone, following published procedures [44][45][46][47][48][49]. Starting from the amino acids, D-and L-methionine were firstly converted into the corresponding tert-Butyloxycarbonyl (Boc) protected derivatives [50]. Upon reduction of the carboxylic acid functions [51] and subsequent tert-butyldiphenylsilyl (TBDPS)protection of the primary alcohols (S)-10 and (R)-10 [52] sulfides (S)-11 and (R)-11 were oxidized to the sulfoxides (S)-12 and (R)-12 with meta-chloroperoxybenzoic acid (m-CPBA) [53,54]. Thermal synelimination [52] of sulfoxides (S)-12 and (R)-12 yielded the vinylglycinol intermediates (S)-14 and (R)-14 which underwent N-allylation to (S)-15 and (R)-15 when reacted with sodium hydride and allyl bromide. The allylation reaction, however, turned out to be capricious as higher yields were only achieved by repeated addition of the same amounts of sodium hydride and allyl bromide. Attempts to deploy potassium tert-butoxide as a base did not lead to any satisfactory product formation in our hands and other reaction conditions were reported to be unsuccessful [55,56] The required pyrrole building blocks 24a,b (Scheme 3) were prepared from commercially available ethyl 4-bromopyrrole-2-carboxylate (21) in two steps following procedures published by Handy et al. [40,[59][60][61]. Previous studies showed that the use of unprotected pyrrole 21 in the envisaged Suzuki-Miyaura coupling results in extensive dehalogenation [61]. Accordingly, pyrrole 21 was initially N-Boc protected and subjected to palladium-catalyzed coupling with boronic acids 23a and 23b to give the desired pyrrole building blocks 24a and 24b. Now, mesylates (−)-20 and (+)-20, and pyrrole building blocks 24a,b were joint together in the presence of cesium carbonate at 80 • C to yield compounds 25d-f (Scheme 3) [62]. The enantiomeric purity of intermediates 25d and 25e was analyzed by asymmetric HPLC and determined to be >99%. In addition, the three-dimensional structure of 25d was confirmed by X-ray crystallographic analysis ( Figure S1). To synthesize inhibitor 29c without aryl substituent attached to the pyrrole ring 1H-pyrrole-2-carboxylate 24c [63] was coupled with mesylate (−)-20 employing the same methodology as before to afford compound 25c. Furthermore, we aimed to prepare inhibitors 28a,b in which the iminosugar unit is replaced by a methyl group. For this purpose, pyrrole building blocks 24a,b were methylated under standard reaction conditions to give compounds 25a,b.
Following saponification of the esters 25a-f the isoxazole building block 30 [34] was coupled stepwise to the corresponding carboxylic acids 26a-f via their N-hydroxybenzotriazole (HOBt)-activated esters 27a-f (Scheme 4) [64]. Thus, coupling products 28a-f were obtained in yields of up to 67%. Finally, cleavage of the protecting groups under acidic reaction conditions furnished 29c-f.

Biological Evaluation
Compounds 28a-c and 29d-f were initially screened in in-vitro kinase assays at a concentration of 10 μM for their ability to inhibit CK1δ and CK1ε. Compounds that showed only low residual activity at this concentration were taken forward to measure IC50 values (Table 1). In accordance with our molecular modelling studies 29d and 29e showed high activity in the CK1δ and CK1ε in-vitro kinase assays with IC50 values in the nanomolar range. Interestingly, no preference was observed between chiral moieties of 29d and 29e, while replacement of the pyrrolidine scaffolds by a methyl (28a and 28b) ablated activity against CK1δ and CK1ε. Omitting the aryl moiety (29c) also resulted in a considerable decrease in activity relative to 29d and 29e in accordance with previously reported CK1δ inhibitors [40]. Taken together, both the aryl moiety and the pyrrolidine scaffold seem to contribute to the binding affinity. On the one hand, interactions between the pyrrolidine scaffold and the active site might could play an important role in determining inhibitor engagement with the active site. On the other hand, the molecular configuration of the scaffolds (29d vs 29e) does not seem to contribute to affinity and CK1-isoform selectivity.

Biological Evaluation
Compounds 28a-c and 29d-f were initially screened in in-vitro kinase assays at a concentration of 10 µM for their ability to inhibit CK1δ and CK1ε. Compounds that showed only low residual activity at this concentration were taken forward to measure IC 50 values (Table 1). In accordance with our molecular modelling studies 29d and 29e showed high activity in the CK1δ and CK1ε in-vitro kinase assays with IC 50 values in the nanomolar range. Interestingly, no preference was observed between chiral moieties of 29d and 29e, while replacement of the pyrrolidine scaffolds by a methyl (28a and 28b) ablated activity against CK1δ and CK1ε. Omitting the aryl moiety (29c) also resulted in a considerable decrease in activity relative to 29d and 29e in accordance with previously reported CK1δ inhibitors [40]. Taken together, both the aryl moiety and the pyrrolidine scaffold seem to contribute to the binding affinity. On the one hand, interactions between the pyrrolidine scaffold and the active site might could play an important role in determining inhibitor engagement with the active site. On the other hand, the molecular configuration of the scaffolds (29d vs 29e) does not seem to contribute to affinity and CK1-isoform selectivity.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.
Molecules 2018, 23, x FOR PEER REVIEW 9 of 34 In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.
Molecules 2018, 23, x FOR PEER REVIEW 9 of 34 In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.
Molecules 2018, 23, x FOR PEER REVIEW 9 of 34 In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively ( Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells.  In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 μM, 10 μM and 20 μM, respectively (Table 2). EC50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells. In addition to the in-vitro kinase assays, all compounds were tested in cell viability assays against tumor cell lines. Since inhibition of CK1δ prolongs the survival of SV40 T-Ag/mutant CK1δ bitransgenic mice [65] and overexpression of CK1δ correlates with reduced survival rates of colorectal and breast cancer patients [66], we chose HT-29 and MCF-7 cells for biological testing [67,68]. These two cell lines are characterized on a molecular level in detail, exhibiting alterations in WNT and p53 signaling pathways, both being regulated by CK1 isoforms [30,69]. All compounds were initially screened at 5 µM, 10 µM and 20 µM, respectively (Table 2). EC 50 values were determined for compounds that showed a clear reduction in cell viability. There is a trend between the EC 50 data for both cell lines and enzyme inhibitory data against CK1, with compounds 29d-f exhibiting a modest effect on both, HT-29 and MCF-7 cell viability. These results are consistent with previous findings that inhibitory effects of CK1 specific inhibitors are dependent on the cellular background [29,[70][71][72]. Especially alterations in WNT and p53 signaling influence the effects of CK1 specific inhibitors [30,73]. The modest inhibitory effect on cancer cell viability may have several reasons: (i) The hydrophilic character of the pyrrolidine scaffolds might reduce their cellular uptake; (ii) efflux systems could contribute to a decreased availability of the inhibitors, (iii) compounds could be partly metabolized (e.g., the hydroxy groups of the pyrrolidine scaffolds) and, (iv) since these inhibitors are ATP competitive inhibitors and cellular ATP concentrations are much higher than the ATP concentration used in the in-vitro assays, higher compound concentrations are necessary to inhibit CK1δ within cells. Next, we screened effective inhibitor 29d in a panel of 320 kinases at a concentration of 1 µM to examine its selectivity ( Figure 5). Apart from CK1δ (residual activity = 1%) and CK1ε (residual activity = 4%) 29d hits only four other kinases, namely CK1α (residual activity = 16%), JNK2 (40%), JNK3 (15%), p38α (17%) thus resulting in an excellent selectivity score of 0.02 (number of kinases with residual activity < 50%/total number of tested kinases).  Next, we screened effective inhibitor 29d in a panel of 320 kinases at a concentration of 1 μM to examine its selectivity ( Figure 5). Apart from CK1δ (residual activity = 1%) and CK1ε (residual activity = 4%) 29d hits only four other kinases, namely CK1α (residual activity = 16%), JNK2 (40%), JNK3 (15%), p38α (17%) thus resulting in an excellent selectivity score of 0.02 (number of kinases with residual activity < 50%/total number of tested kinases).  Table S2.

X-ray Analysis of ligand-CK1δ Complexes
In the next step we sought to elucidate structural details of the binding of our chiral inhibitors via X-ray crystallography. For this, a C-terminally truncated version of CK1δ (1-294, tCK1δ) together with 29d and 29e, respectively, were used in co-crystallization trials. Crystals suitable for diffraction analysis grew in presence of both compounds in crystallization conditions containing 0.1 M MES (pH = 5.5), 10% polyethylengylcol (PEG) 4000 and 0.2 M Li2SO4. These crystals diffracted to a resolution limit of 1.8 Å and belong to the monoclinic crystal system. In each asymmetric unit, two tCK1  Table S2.

X-ray Analysis of ligand-CK1δ Complexes
In the next step we sought to elucidate structural details of the binding of our chiral inhibitors via X-ray crystallography. For this, a C-terminally truncated version of CK1δ (1-294, tCK1δ) together with 29d and 29e, respectively, were used in co-crystallization trials. Crystals suitable for diffraction analysis grew in presence of both compounds in crystallization conditions containing 0.1 M MES (pH = 5.5), 10% polyethylengylcol (PEG) 4000 and 0.2 M Li 2 SO 4 . These crystals diffracted to a resolution limit of 1.8 Å and belong to the monoclinic crystal system. In each asymmetric unit, two tCK1 molecules were present with an isoxazole-based ligand bound to each ATP binding pocket ( Figure 6). Surprisingly, the ligands in both protein-ligand complexes proved to be different from the actual target compounds submitted to co-crystallization. Since all analytical data for the originally employed inhibitors was consistent with the given structures, we postulate that the incorporation of an additional methylene group occurred via spontaneous Pictet-Spengler cyclization [74][75][76] during crystallization. In line with this notion, PEG, which has been used in our soaking solutions, is known to generate formaldehyde traces that have been shown in the past to affect the derivatization of both, proteins and ligands [77][78][79]. To further corroborate this hypothesis, we examined whether simple exposure of 29d/e to formaldehyde would give rise to the same Pictet-Spengler products 31a/b (Scheme 5). Indeed, when compounds 29d/e were treated with formaldehyde at room temperature 31a/b were formed readily and the obtained spectroscopic data was consistent with the predicted Surprisingly, the ligands in both protein-ligand complexes proved to be different from the actual target compounds submitted to co-crystallization. Since all analytical data for the originally employed inhibitors was consistent with the given structures, we postulate that the incorporation of an additional methylene group occurred via spontaneous Pictet-Spengler cyclization [74][75][76] during crystallization. In line with this notion, PEG, which has been used in our soaking solutions, is known to generate formaldehyde traces that have been shown in the past to affect the derivatization of both, proteins and ligands [77][78][79]. To further corroborate this hypothesis, we examined whether simple exposure of 29d/e to formaldehyde would give rise to the same Pictet-Spengler products 31a/b (Scheme 5). Indeed, when compounds 29d/e were treated with formaldehyde at room temperature 31a/b were formed readily and the obtained spectroscopic data was consistent with the predicted Pictet-Spengler products. In addition, aqueous solutions of 29d and 29e with and without PEG were analyzed by liquid chromatography-mass spectrometry (LC-MS) over a period of 7 days. Only in those solutions that had PEG present significant product formation of 31a and 31b was observed indicating that the degradation of PEG may have been the likely source of formaldehyde. Pictet-Spengler products. In addition, aqueous solutions of 29d and 29e with and without PEG were analyzed by liquid chromatography-mass spectrometry (LC-MS) over a period of 7 days. Only in those solutions that had PEG present significant product formation of 31a and 31b was observed indicating that the degradation of PEG may have been the likely source of formaldehyde.
Scheme 5. Proposed Pictet-Spengler reaction of originally employed compounds 29d and 29e with formaldehyde impurities to yield new compounds 31a and 31b. Traces of formaldehyde may originate from PEG reagent employed in ligand soaking approaches.
Following the discovery of new ligands as artefacts in the X-ray complexes, both these Pictet-Spengler products 31a and 31b were re-synthesized and subsequently assessed for their biological activities (Table 3). In the in-vitro kinase assays both compounds exhibited nanomolar activity against CK1δ and CK1ε with a five-fold preference for CK1δ over CK1ε. The cellular assays revealed 31a and 31b to be more potent than the originally designed compounds 29d and 29e with EC50 values below 2 μM. The higher cellular potency of both compounds is in a way surprising, because in contrast to compounds 29d and 29e the chiral moieties of the Pictet-Spengler products 31a and 31b are too far away to interact with residues in the ATP binding site ( Figure 6). Further studies will need to elucidate compound's kinase selectivity, and also their specificity in terms of addressing other targets. However, our results suggest that the chiral pyrrolidine scaffold attached to kinase inhibitors (and beyond) can be suitable for medicinal chemistry applications. Following the discovery of new ligands as artefacts in the X-ray complexes, both these Pictet-Spengler products 31a and 31b were re-synthesized and subsequently assessed for their biological activities (Table 3). In the in-vitro kinase assays both compounds exhibited nanomolar activity against CK1δ and CK1ε with a five-fold preference for CK1δ over CK1ε. The cellular assays revealed 31a and 31b to be more potent than the originally designed compounds 29d and 29e with EC 50 values below 2 µM. The higher cellular potency of both compounds is in a way surprising, because in contrast to compounds 29d and 29e the chiral moieties of the Pictet-Spengler products 31a and 31b are too far away to interact with residues in the ATP binding site ( Figure 6). Further studies will need to elucidate compound's kinase selectivity, and also their specificity in terms of addressing other targets. However, our results suggest that the chiral pyrrolidine scaffold attached to kinase inhibitors (and beyond) can be suitable for medicinal chemistry applications. Table 3. Biological activity of compound 31a and 31b in in-vitro kinase assays (CK1δ and CK1ε) and cellular assays (MCF-7-and HT-29). Results are presented as mean ± SD from experiments performed in triplicate (n = 3).

General Experimental Procedures
Melting points were determined either on a Stuart Melting Point (SMP3) apparatus and are uncorrected, or by differential scanning calorimetry (DSC) on a Mettler Toledo DSC1 instrument at a heating rate of 10 K·min −1 . Proton ( 1 H) and carbon ( 13 C) NMR-spectra were recorded on Bruker Avance (III)-500 or Avance (I)-300 spectrometers. 19  Electrospray ionization (ESI) mass spectrometry (MS) experiments were performed on a QTOF Premier mass spectrometer (Micromass, UK) under normal conditions. Sodium formate solution was used as a calibrant for high resolution mass spectra (HRMS) measurements. Elemental microanalyses were performed at The Campbell Microanalytical Laboratory, Department of Chemistry at the University of Otago on a Carlo-Erba EA 1108 elemental analyzer. Specific optical rotations were acquired on a Rudolph Autopol ® IV Automatic polarimeter at ambient temperature (20 • C), unless otherwise stated, λ = 589 nm and concentration (g/100 mL) in the solvent indicated, using a cell of 100 mm path length.

Cellular Assays and EC 50 Determination
In order to analyze the effects of selected inhibitors on the cell viability of HT-29 and MCF7 cells, MTT assays were performed. 1 × 10 4 cells/well were seeded in 96-well cell culture plates. After 24 h cells were treated with increasing concentrations of the indicated inhibitor compounds, with untreated and DMSO-treated cells as control. After an incubation period of 48 h, 10 µL of MTT solution (5 mg/mL in PBS) were added and cells were incubated for 4 h. MTT-containing medium was carefully removed and 100 µL acidic isopropanol (0.04 N HCl in isopropanol) per well were added. To dissolve formazan crystals, plates were shaken for 30 min on an orbital shaker in the dark. Finally, dissolved crystals were measured spectrophotometrically at 570 nm. All experiments were performed in triplicate with four technical replicates per assay. Results were normalized considering the mean optical density value of control wells as 100%. GraphPad Prism 6 software (San Diego, CA, USA) was used to calculate EC 50 values.
For co-crystallization of tCK1δ with 29d and 29e, protein stock solution (10 mg·mL −1 ) was mixed 30:1 with 10 mM of the respective compounds (solubilized in DMSO) and incubated for 30 min at room temperature. Sitting drop crystallization trials were set up at room temperature with drop ratios of 3 µL protein/inhibitor solution to 2 µL precipitant solution.
Crystals appeared after three to seven days in drops containing 0.1 M MES (pH = 5.5), 10% (w/v) PEG 4000 and 0.2 M lithium sulfate. For data collection, these crystals were cryo-protected by swiping them though reservoir solution supplemented with 25% (v/v) glycerol and subsequently flash frozen. Diffraction data were collected at beamline P13 at the PETRA/EMBL Hamburg, German Synchrotron Research Centre (DESY) campus, Hamburg, Germany (tCK1δ with 29d) and at beamline X06DA at the Swiss Light Source, Paul-Scherrer-Institute, Villigen, Switzerland. XDS [87] was used for data processing and the structures were solved by molecular replacement using the program PHASER [88,89] with a truncated crystal structure of CK1δ (pdb 4TWC [72]) as a search model. Between iterative cycles of refinement using phenix.refine [90] missing loops, as well as 31a/31b were manually built with Coot [91]. Restraints of 31a and 31b were calculated using phenix.elbow [92].

Conclusions
By using isoxazole-based CK1δ inhibitor 8 as model compound we implemented chiral pyrrolidine scaffolds to investigate specific interactions within the ATP binding pocket of this kinase. A synthetic approach leading to stereochemically defined compounds (29d-f) was established. Biological evaluation proved the compounds to be quite potent in in vitro CK1 kinase assays, but revealed only a minor impact of the different stereoisomers towards affinity differences and CK1 isoform-selectivity. However, selectivity profiling in a panel of 320 kinases showed compound 29d to be relatively specific for CK1. Compounds 29d-f showed also modest effects on HT-29 and MCF-7 cell lines. Surprisingly, X-ray crystallographic data revealed new cyclized compounds within the analyzed protein-ligand complexes, which were supposedly produced by spontaneous Pictet-Spengler cyclization implementing one equivalent formaldehyde during sample soaking procedures. In fact, this hypothesis could be confirmed by reacting the original compounds 29d/e with formaldehyde to yield the new ligands 31a/b. Taken together, functionalized pyrrolidines are expedient chiral scaffolds in medicinal chemistry and may have a potential for the development of potent and selective kinase inhibitors targeting the ribose pocket of the ATP binding site.

Supplementary Materials:
The following are available online, Figure S1: X-Ray crystal structure of compound 25d Table S1: Data collection, structure refinement and Ramachrandran plot results of protein crystallization, Table S2: Selectivity profile of compound 29d, Appendix: NMR spectra, HPLC chromatograms and DSC traces.
Funding: Funding for the synchrotron visit has been provided by the iNEXT initiative (proposal 1859, EU program Horizon 2020).