How to Separate Kinase Inhibition from Undesired Monoamine Oxidase A Inhibition—The Development of the DYRK1A Inhibitor AnnH75 from the Alkaloid Harmine

The β-carboline alkaloid harmine is a potent DYRK1A inhibitor, but suffers from undesired potent inhibition of MAO-A, which strongly limits its application. We synthesized more than 60 analogues of harmine, either by direct modification of the alkaloid or by de novo synthesis of β-carboline and related scaffolds aimed at learning about structure–activity relationships for inhibition of both DYRK1A and MAO-A, with the ultimate goal of separating desired DYRK1A inhibition from undesired MAO-A inhibition. Based on evidence from published crystal structures of harmine bound to each of these enzymes, we performed systematic structure modifications of harmine yielding DYRK1A-selective inhibitors characterized by small polar substituents at N-9 (which preserve DYRK1A inhibition and eliminate MAO-A inhibition) and beneficial residues at C-1 (methyl or chlorine). The top compound AnnH75 remains a potent DYRK1A inhibitor, and it is devoid of MAO-A inhibition. Its binding mode to DYRK1A was elucidated by crystal structure analysis, and docking experiments provided additional insights for this attractive series of DYRK1A and MAO-A inhibitors.


Introduction
Protein phosphorylation is a major regulatory mechanism of nearly every cellular process, and many human diseases involve aberrant protein kinase activities. Specific kinase inhibitors serve as valuable research tools to elucidate the individual roles of the more than 500 human protein kinases in physiological and pathological processes. DYRK1A (dual-specificity tyrosine phosphorylation-regulated kinase 1A) is a protein kinase that has important functions in neuronal development and cell cycle control, and attracts increasing attention as a possible drug target [1,2]. The DYRK1A gene is localized on human chromosome 21, and substantial evidence supports the hypothesis that DYRK1A overexpression

Results
Harmine, as well as the related alkaloids harmol and harmane were, inspired by the investigations of Frost et al. [24], included into our characterization. In the course of a systematic modification of harmine, easy-to-perform direct functionalizations of the native alkaloid were performed in a first series of synthesized analogues, typically by ring halogenation reactions, modification of the 1-methyl and 7-methoxy group, and alkylations or acylation at N-9. In a second series, analogues of harmine were prepared de novo, in order to achieve introduction of hitherto not accessible substituents at rings A and C, and preparation of 2-and 9-desaza analogues ( Figure 2A). Of special interest here was the generation of 7,8-dichlorinated β-carbolines, since our previous investigations on analogues of the dichlorinated 1-oxo-β-carboline alkaloid bauerine C provided potent and selective kinase (CLK1, DYRK1A, PIM1) inhibitors with a very unusual binding mode; namely, formation of a halogen bond with the hinge region in the active site [25][26][27] (Figure 2B).
Molecules 2020, 25, x FOR PEER REVIEW 3 of 34 Reniers et al. [18] already demonstrated that the MOA-A inhibitory activity of harmine can even be enhanced by replacing the methoxy group at ring C by lipophilic ethers (cyclohexyl, benzyl), whereas a polar (N,N-dimethylamino)ethoxy residue (see compound AnnH62 of this investigation) reduced MAO-A inhibition by a factor of 60. On the other hand, the co-crystal structure suggested that substituents at the central pyrrole nitrogen, which prevent formation of the hydrogen bond network, might lead to reduced MAO-A inhibition. This evidence was confirmed recently by Haider et al. with harmine-based 7-triazolylmethyl ethers [19]. In contrast, the methoxy group appeared essential for DYRK1A inhibition, while larger ether residues at C-7 should be detrimental due to steric hindrance in the pocket. Substitution at N-9 should be tolerated here from a steric point of view, and Drung et al. demonstrated a positive effect of an n-heptyl residue at N-9 for DYRK1A inhibition, combined with a significant, but not complete, reduction of MAO-A inhibition [20].
Based on the evidence available at the beginning of our project, structure modifications at multiple sites of harmine appeared worth investigating. Therefore, we set up a comprehensive synthesis program, which ultimately led to the development of harmine analogue AnnH75 as a potent DYRK1A inhibitor devoid of any MAO-A inhibitory potency. The central findings of this investigation were confirmed later by Balint et al. [21], who investigated a closely related, but significantly smaller, set of harmine analogues. The comprehensive pharmacological characterization of AnnH75 has already been published by our laboratory [22]. In 2018, Kumar et al. performed a number of variations at C-1 and C-3 of harmine, but this work was exclusively aimed at improving selectivity over other protein kinases, and MAO-A inhibition was not investigated [23]. In this paper, we describe the complete medicinal chemistry aspects of this project, including the enzyme kinetic characterization that provided new insight into structure-activity relationships, and led to the selection of AnnH75 as the most promising DYRK1A inhibitor for further pharmacological characterization.

Results
Harmine, as well as the related alkaloids harmol and harmane were, inspired by the investigations of Frost et al. [24], included into our characterization. In the course of a systematic modification of harmine, easy-to-perform direct functionalizations of the native alkaloid were performed in a first series of synthesized analogues, typically by ring halogenation reactions, modification of the 1methyl and 7-methoxy group, and alkylations or acylation at N-9. In a second series, analogues of harmine were prepared de novo, in order to achieve introduction of hitherto not accessible substituents at rings A and C, and preparation of 2-and 9-desaza analogues ( Figure 2A). Of special interest here was the generation of 7,8-dichlorinated β-carbolines, since our previous investigations on analogues of the dichlorinated 1-oxo-β-carboline alkaloid bauerine C provided potent and selective kinase (CLK1, DYRK1A, PIM1) inhibitors with a very unusual binding mode; namely, formation of a halogen bond with the hinge region in the active site [25][26][27] (Figure 2B).  Direct Modifications at Ring C Due to the interesting binding mode of the above mentioned dichlorinated kinase inhibitors from our previous work ( Figure 2B), we investigated the influence of additional halogen substituents on ring C of harmine. Following published methods [20,29,30], halogenation of harmine with Nchlorosuccinimide (NCS) and N-bromosuccinimide (NBS) in dichloromethane in the presence of silica gave 6-and 8-monosubstituted products (AnnH6_1, AnnH6_3_2, AnnH5_2, AnnH5_3) along with 6,8-dihalogenated harmines (AnnH6_3_1, AnnH5_1). Using an excess of NBS gave 3,6,8tribromoharmine AnnH5_4 in poor yield (4%).
As alkylation of N-9 promised reduction of MAO-A-inhibitory activity, we converted the three readily available monohalogenated compounds (AnnH5_3, AnnH6_1, AnnH6_3_2) into their Nbutyl derivatives AnnH9, AnnH11 and AnnH14 by N-deprotonation and subsequent reaction with n-butyl iodide (Scheme 1).
Further, we prepared aminoalkyl ether AnnH62, following the protocol of Reniers et al. [18] by O-alkylation of harmol. The authors found reduced MAO-A-inhibitory activity for this compound compared to harmine, but they did not investigate the influence of this modification on DYRK1A inhibitory activity (Scheme 1).

Scheme 1. Compounds obtained by modification of substitution patterns in rings A and C of harmine.
Direct Modifications at Ring C Due to the interesting binding mode of the above mentioned dichlorinated kinase inhibitors from our previous work ( Figure 2B), we investigated the influence of additional halogen substituents on ring C of harmine. Following published methods [20,29,30], halogenation of harmine with N-chlorosuccinimide (NCS) and N-bromosuccinimide (NBS) in dichloromethane in the presence of silica gave 6-and 8-monosubstituted products (AnnH6_1, AnnH6_3_2, AnnH5_2, AnnH5_3) along with 6,8-dihalogenated harmines (AnnH6_3_1, AnnH5_1). Using an excess of NBS gave 3,6,8-tribromoharmine AnnH5_4 in poor yield (4%).
As alkylation of N-9 promised reduction of MAO-A-inhibitory activity, we converted the three readily available monohalogenated compounds (AnnH5_3, AnnH6_1, AnnH6_3_2) into their N-butyl derivatives AnnH9, AnnH11 and AnnH14 by N-deprotonation and subsequent reaction with n-butyl iodide (Scheme 1).
Further, we prepared aminoalkyl ether AnnH62, following the protocol of Reniers et al. [18] by O-alkylation of harmol. The authors found reduced MAO-A-inhibitory activity for this compound compared to harmine, but they did not investigate the influence of this modification on DYRK1A inhibitory activity (Scheme 1).

Modifications at Ring B-Acetylation and Alkylation at N-9
The central pyrrole ring, especially the NH function, was of special interest, since the latter contributes significantly to binding of harmine to MAO-A via a hydrogen bond to a surrounding water molecule, but is obviously not involved in binding of the alkaloid to DYRK1A. Consequently, substitution at N-9 was expected to have a major impact on the selectivity profile of harmine derivatives. This prompted us to synthesize a large number of derivatives bearing residues at N-9 covering a broad spectrum of sizes and polarities. N 9 -Acetylharmine (AnnH59) was obtained accidentally by treating harmine with piperonal and acetic anhydride, following a method of Cao et al. [31], aimed at the synthesis of a (methylenedioxy)benzylidene derivative related to AnnH18.
Further, we prepared a considerable number of N-alkylated products. This work was inspired by the above-mentioned work of Drung et al. [20] and reports describing 9-ethylharmine as promising DYRK1A inhibitor [24,32]; further, Cuny et al. [33] described a series on N-aminoalkyl derivatives, but tested them only on Haspin kinase and DYRK2. In order to obtain a clear insight into the structure-activity relationships in this subclass, we introduced small and large alkyl residues, arylalkyl residues of different length, as well as polar groups, either neutral, acidic or basic in nature. These N-alkylations were generally easy to perform, as shown by Cao et al. [34] and in a patent [35] for a couple of compounds before, but yields differed strongly depending on the alkylation agent. Finally, we established three different protocols for these conversions: (A) using sodium hydride as base in DMF at 40 • C, B) using potassium tert-butoxide as base in DMSO at 80 • C, and C) using sodium hydride as base in THF under reflux), which provided the target compounds (Scheme 2) in acceptable yields.
Molecules 2020, 25, x FOR PEER REVIEW 5 of 34 Modifications at Ring B-Acetylation and Alkylation at N-9 The central pyrrole ring, especially the NH function, was of special interest, since the latter contributes significantly to binding of harmine to MAO-A via a hydrogen bond to a surrounding water molecule, but is obviously not involved in binding of the alkaloid to DYRK1A. Consequently, substitution at N-9 was expected to have a major impact on the selectivity profile of harmine derivatives. This prompted us to synthesize a large number of derivatives bearing residues at N-9 covering a broad spectrum of sizes and polarities. N 9 -Acetylharmine (AnnH59) was obtained accidentally by treating harmine with piperonal and acetic anhydride, following a method of Cao et al. [31], aimed at the synthesis of a (methylenedioxy)benzylidene derivative related to AnnH18.
Further, we prepared a considerable number of N-alkylated products. This work was inspired by the above-mentioned work of Drung et al. [20] and reports describing 9-ethylharmine as promising DYRK1A inhibitor [24,32]; further, Cuny et al. [33] described a series on N-aminoalkyl derivatives, but tested them only on Haspin kinase and DYRK2. In order to obtain a clear insight into the structure-activity relationships in this subclass, we introduced small and large alkyl residues, arylalkyl residues of different length, as well as polar groups, either neutral, acidic or basic in nature. These N-alkylations were generally easy to perform, as shown by Cao et al. [34] and in a patent [35] for a couple of compounds before, but yields differed strongly depending on the alkylation agent. Finally, we established three different protocols for these conversions: (A) using sodium hydride as base in DMF at 40 °C, B) using potassium tert-butoxide as base in DMSO at 80 °C, and C) using sodium hydride as base in THF under reflux), which provided the target compounds (Scheme 2) in acceptable yields.

De novo-Synthesis of Harmine Analogues
Harmine-Bauerine C hybrids-7,8-Dichloro Derivatives As pointed out above, the unique binding mode (halogen bridge) of ortho-dichlorinated indoles and β-carbolines ( Figure 1B) in the active site of kinases closely related to DYRK1A prompted us to Scheme 2. Harmine analogues prepared by N-acetylation (to give AnnH59) or N-alkylation. The yield as well as the respective N-alkylation method (A/B/C) are indicated in parentheses.

De novo-Synthesis of Harmine Analogues
Harmine-Bauerine C hybrids-7,8-Dichloro Derivatives As pointed out above, the unique binding mode (halogen bridge) of ortho-dichlorinated indoles and β-carbolines ( Figure 1B) in the active site of kinases closely related to DYRK1A prompted us to explore this motif here as well. We hypothesized that, with this modification, DYRK1A affinity might be conserved, while MAO-A inhibition would hopefully be eliminated. Synthesis of the target compounds started with 1-oxo-β-carboline AnnH65, a compound obtained by dehydration of 1, a precursor of our first total synthesis of bauerine C [36]. Compound AnnH65 itself was of interest, since some 3-aza analogues thereof (pyridazino [4,5-b]indol-4-ones) were recently identified as potent DYRK1A inhibitors by Bruel et al. [37]. The ring A pyridone moiety in AnnH65 was conveniently transformed into the 1-bromo derivative AnnH52 by treatment with POBr 3 , using anisole as solvent and bromine scavenger [38], by which undesired ring bromination at ring C was successfully suppressed. The 1-methyl group was introduced by a chemoselective Pd-catalyzed cross-coupling with trimethylaluminum [39]. The 1-bromo substituent reacted exclusively in presence of both chlorine residues, and the target compound AnnH63 was obtained in 92% yield.
Since our initial work on N 9 -alkylated derivatives of harmine (see Sections 2.1.1 and 2.2) clearly demonstrated a beneficial effect of small electron-withdrawing residues like cyanomethyl and related esters, we further performed N-alkylations with the respective alkyl halides on both the 1-bromoand 1-methyl-7,8-dichloro-β-carboline (AnnH52 and AnnH63) yielding products AnnH66, AnnH67, AnnH69 as well as AnnH79 and AnnH80; further aminopropyl derivative AnnH71 was prepared for getting deeper insight into structure-activity relationships in this subtype of harmine analogues (Scheme 3).
Molecules 2020, 25, x FOR PEER REVIEW 6 of 34 explore this motif here as well. We hypothesized that, with this modification, DYRK1A affinity might be conserved, while MAO-A inhibition would hopefully be eliminated. Synthesis of the target compounds started with 1-oxo-β-carboline AnnH65, a compound obtained by dehydration of 1, a precursor of our first total synthesis of bauerine C [36]. Compound AnnH65 itself was of interest, since some 3-aza analogues thereof (pyridazino [4,5-b]indol-4-ones) were recently identified as potent DYRK1A inhibitors by Bruel et al. [37]. The ring A pyridone moiety in AnnH65 was conveniently transformed into the 1-bromo derivative AnnH52 by treatment with POBr3, using anisole as solvent and bromine scavenger [38], by which undesired ring bromination at ring C was successfully suppressed. The 1-methyl group was introduced by a chemoselective Pd-catalyzed cross-coupling with trimethylaluminum [39]. The 1-bromo substituent reacted exclusively in presence of both chlorine residues, and the target compound AnnH63 was obtained in 92% yield. Since our initial work on N 9 -alkylated derivatives of harmine (see Sections 2.1.1 and 2.2) clearly demonstrated a beneficial effect of small electron-withdrawing residues like cyanomethyl and related esters, we further performed N-alkylations with the respective alkyl halides on both the 1-bromoand 1-methyl-7,8-dichloro-β-carboline (AnnH52 and AnnH63) yielding products AnnH66, AnnH67, AnnH69 as well as AnnH79 and AnnH80; further aminopropyl derivative AnnH71 was prepared for getting deeper insight into structure-activity relationships in this subtype of harmine analogues (Scheme 3).

Further Variations in Ring A
Elimination/Reduction of the Basicity of N-2: In order to investigate the role of the nitrogen atom at position 2, we prepared two analogues of harmine lacking a 2-N with significant basicity. In a first modification, lactam-type alkaloid harmalacidine (AnnH19), readily available from 6methoxytryptamine [40], was dehydrogenated with p-chloranil to give the ring A pyridone AnnH20. Our attempts to convert this 1-oxo-β-carboline into the 1-bromo compound using the POBr3/anisole couple (see Section 2.1.2) failed, since bromination at electron-rich ring C could not be suppressed here. Therefore, we performed a chlorination with POCl3, following published work [21,41], to receive the 1-chloro analogue AnnH24 of harmine (Scheme 4). The electronegativity of the lactam oxygen in the one, and of the electronegative chlorine substituent in the other compound were expected to change the electronic properties of N-2 significantly.

Further Variations in Ring A
Elimination/Reduction of the Basicity of N-2: In order to investigate the role of the nitrogen atom at position 2, we prepared two analogues of harmine lacking a 2-N with significant basicity. In a first modification, lactam-type alkaloid harmalacidine (AnnH19), readily available from 6-methoxytryptamine [40], was dehydrogenated with p-chloranil to give the ring A pyridone AnnH20. Our attempts to convert this 1-oxo-β-carboline into the 1-bromo compound using the POBr 3 /anisole couple (see Section 2.1.2) failed, since bromination at electron-rich ring C could not be suppressed here. Therefore, we performed a chlorination with POCl 3 , following published work [21,41], to receive the 1-chloro analogue AnnH24 of harmine (Scheme 4). The electronegativity of the lactam oxygen in the one, and of the electronegative chlorine substituent in the other compound were expected to change the electronic properties of N-2 significantly.
Further, we prepared a 2-desaza analogue of harmine, represented by the disubstituted carbazole AnnHOG3. This compound was obtained from known tetrahydrocarbazol-1-one 2 [42] through addition of methyllithium to the keto group. During work-up, dehydratisation and dehydrogenation took place to give the fully aromatic target compound in 25% yield (Scheme 5). Synthesis of 1-Ethyl Analogues of Harmine: Since even small changes in the size of a ligand of an enzyme may have significant influence on its affinity-for example, due to steric effects-we further prepared an analogue of harmine where the 1-methyl group was replaced by an ethyl group. Target compound AnnH89 was obtained with a 66% yield from 1-chloro derivative AnnH24 by Pdcatalyzed cross-coupling with triethylborane following the protocol we established for the synthesis of alkaloid 1-ethyl-β-carboline previously [39]. As a by-product, we obtained dechlorination product norharmine (AnnH90) at low yield. In an analogous manner, N-substituted 1-chloro compound AnnH76 was coupled with triethylborane to give the ester AnnH88 at 69% yield (Scheme 6).  Synthesis of 1-Ethyl Analogues of Harmine: Since even small changes in the size of a ligand of an enzyme may have significant influence on its affinity-for example, due to steric effects-we further prepared an analogue of harmine where the 1-methyl group was replaced by an ethyl group. Target compound AnnH89 was obtained with a 66% yield from 1-chloro derivative AnnH24 by Pd-catalyzed cross-coupling with triethylborane following the protocol we established for the synthesis of alkaloid 1-ethyl-β-carboline previously [39]. As a by-product, we obtained dechlorination product norharmine (AnnH90) at low yield. In an analogous manner, N-substituted 1-chloro compound AnnH76 was coupled with triethylborane to give the ester AnnH88 at 69% yield (Scheme 6). In the same manner as done before with harmine (Section 2.1.1.) and the 1-bromo-7,8-dichloro analogue AnnH52, we introduced polar residues at N-9 using the established N-alkylation protocols, ending up with compounds AnnH74, AnnH75, AnnH76, and AnnH77.
Further, we prepared a 2-desaza analogue of harmine, represented by the disubstituted carbazole AnnHOG3. This compound was obtained from known tetrahydrocarbazol-1-one 2 [42] through addition of methyllithium to the keto group. During work-up, dehydratisation and dehydrogenation took place to give the fully aromatic target compound in 25% yield (Scheme 5). Synthesis of 1-Ethyl Analogues of Harmine: Since even small changes in the size of a ligand of an enzyme may have significant influence on its affinity-for example, due to steric effects-we further prepared an analogue of harmine where the 1-methyl group was replaced by an ethyl group. Target compound AnnH89 was obtained with a 66% yield from 1-chloro derivative AnnH24 by Pdcatalyzed cross-coupling with triethylborane following the protocol we established for the synthesis of alkaloid 1-ethyl-β-carboline previously [39]. As a by-product, we obtained dechlorination product norharmine (AnnH90) at low yield. In an analogous manner, N-substituted 1-chloro compound AnnH76 was coupled with triethylborane to give the ester AnnH88 at 69% yield (Scheme 6).  As pointed out in the introduction of this manuscript, the NH functionality is important for the interaction of harmine with MAO-A, acting as a hydrogen bond donor for a water molecule, whereas a comparable interaction has not been reported for DYRK1A binding. Consequently, it was of interest to study the effect of replacing this NH group by a secondary alcohol (which is able to act as both a hydrogen bond donor and acceptor) and a keto group (which is exclusively a hydrogen bond acceptor) on the inhibitory activity on DYRK1A and MAO-A. This required a novel approach allowing the synthesis of appropriately substituted 2-azafluorenes.
The target 2-azafluorenone AnnHT2 was obtained starting from 4-arylnicotinate 3, which in turn was obtained by Hantzsch-type condensation [43] of 4-methoxycinnamaldehyde and ethyl 3-aminocrotonate, followed by dehydrogenation of the crude dihydropyridine intermediate with iodine. Cyclization to give the 2-azafluorenone AnnHT2 was accomplished as demonstrated before for the 4-azafluorenone alkaloid onychine [44], by intramolecular acylation upon heating with polyphosphoric acid. Since the yield of this reaction was only 19%, we also examined a significantly stronger acid, trifluoromethanesulfonic acid. Here, a 2-azafluorenone was obtained in 46% yield, but the product AnnHT6 was identified as the free phenol at C-7 (Scheme 7). Ketone AnnHT2 was readily reduced to the racemic secondary alcohol AnnHT3 with sodium borohydride. Unfortunately, diverse attempts to reduce the keto or carbinol function to a methylene group failed. As pointed out in the introduction of this manuscript, the NH functionality is important for the interaction of harmine with MAO-A, acting as a hydrogen bond donor for a water molecule, whereas a comparable interaction has not been reported for DYRK1A binding. Consequently, it was of interest to study the effect of replacing this NH group by a secondary alcohol (which is able to act as both a hydrogen bond donor and acceptor) and a keto group (which is exclusively a hydrogen bond acceptor) on the inhibitory activity on DYRK1A and MAO-A. This required a novel approach allowing the synthesis of appropriately substituted 2-azafluorenes.
The target 2-azafluorenone AnnHT2 was obtained starting from 4-arylnicotinate 3, which in turn was obtained by Hantzsch-type condensation [43] of 4-methoxycinnamaldehyde and ethyl 3aminocrotonate, followed by dehydrogenation of the crude dihydropyridine intermediate with iodine. Cyclization to give the 2-azafluorenone AnnHT2 was accomplished as demonstrated before for the 4-azafluorenone alkaloid onychine [44], by intramolecular acylation upon heating with polyphosphoric acid. Since the yield of this reaction was only 19%, we also examined a significantly stronger acid, trifluoromethanesulfonic acid. Here, a 2-azafluorenone was obtained in 46% yield, but the product AnnHT6 was identified as the free phenol at C-7 (Scheme 7). Ketone AnnHT2 was readily reduced to the racemic secondary alcohol AnnHT3 with sodium borohydride. Unfortunately, diverse attempts to reduce the keto or carbinol function to a methylene group failed. Scheme 7. Synthesis of 9-desaza analogues of harmine.
Since screening data of harmine derivatives bearing a cyanomethyl residue at N-9 were promising, we introduced a related cyanomethylene moiety into the obtained 9-desazaharmine scaffold. Horner-Wadsworth-Emmons olefination of ketone AnnHT2 with diethyl cyanomethylphosphonate/NaH gave E-configured cyanomethylene product AnnHT5 in 22% yield. The E configuration was confirmed by NOE experiments clearly indicating steric proximity of the olefinic methine hydrogen and the 1-methyl group (Scheme 7).

In Vitro Enzyme Assays
All analogues of harmine described above were subjected to in vitro tests for inhibition of MAO-A and DYRK1A. Further, most compounds were tested for inhibition of the protein kinase CLK1, another member of the CMGC group of protein kinases, since DYRK1A inhibitors frequently also show strong inhibition of this kinase [26,27]. Optimization of the compounds, however, aimed predominantly at elimination of MAO-A-inhibitory activity while keeping DYRK1A inhibitory Scheme 7. Synthesis of 9-desaza analogues of harmine.
Since screening data of harmine derivatives bearing a cyanomethyl residue at N-9 were promising, we introduced a related cyanomethylene moiety into the obtained 9-desazaharmine scaffold. Horner-Wadsworth-Emmons olefination of ketone AnnHT2 with diethyl cyanomethylphosphonate/NaH gave E-configured cyanomethylene product AnnHT5 in 22% yield. The E configuration was confirmed by NOE experiments clearly indicating steric proximity of the olefinic methine hydrogen and the 1-methyl group (Scheme 7).

In Vitro Enzyme Assays
All analogues of harmine described above were subjected to in vitro tests for inhibition of MAO-A and DYRK1A. Further, most compounds were tested for inhibition of the protein kinase CLK1, another member of the CMGC group of protein kinases, since DYRK1A inhibitors frequently Molecules 2020, 25, 5962 9 of 34 also show strong inhibition of this kinase [26,27]. Optimization of the compounds, however, aimed predominantly at elimination of MAO-A-inhibitory activity while keeping DYRK1A inhibitory potency, and not at achieving high selectivity for DYRK1A over CLK1. The tests were performed as described by us before in detail [22]. Due to the very large number of test compounds (more than 60), a first screening was performed at a single inhibitor concentration of 1 µM. The results are presented in Table 1. Only outstanding compounds from this screening were subjected to further pharmacological characterization, and these results have been published by us before [22]. Table 1. Results of in vitro screening of the harmine analogues on MAO-A, DYRK1A and CLK1 1 .

MAO-A DYRK1A CLK1
Residual activities in % at 1 µM inhibitor concentration Values >100% can be due to experimental variability or inhibition of luciferase activity in the Kinase-Glo ® assay.
"-": not tested. * Value not reliable due to very poor solubility, ** value not reliable due to fluctuating individual values. *** Experiments performed in duplicate.
The values of inhibition of MAO-A and DYRK1A for the most relevant inhibitors (including standard deviations, SD) are presented in Figure 3.
As described in our recent publication [22], the most attractive compounds from this program were subjected to a comprehensive pharmacological characterization, and the main results are summarized in Table 2.   Table 1 to illustrate the different degrees of selectivity for DYRK1A. Shown are inhibitors that reduce DYRK1A activity by at least 50%. For comparison, AnnH11 is included as an example for MAO-A selective inhibitor and AnnH35 as a non-selective inhibitor. Columns reflect the residual enzyme activities of DYRK1A (blue) and MAO-A (orange) after incubation with the indicated compounds at a concentration of 1 µM (means and SD).

Discussion of the Screening Results
Results obtained for the alkaloids harmane and norharmine (AnnH90) clearly indicated the importance of the 7-methoxy group for DYRK1A inhibition, but also 7-hydroxy compound harmol was a potent DYRK1A inhibitor (and even more potent on CLK1), notably devoid of MAO-A inhibition, as shown by Balint et al. previously [21]. 1-Styryl analogue AnnH18 was devoid of MAO-A inhibition, but at the cost of significantly reduced potency at DYRK1A.
Modifications at ring C: Bromination or chlorination at C-6 and/or C-8 eliminated DYRK1A inhibition completely. The monohalogenated products (including their N-butylated derivatives (AnnH9, AnnH11) retained significant MAO-A inhibition. This change in selectivity was exactly the opposite of what we intended to achieve. For the 7-aminoethyl ether AnnH62, we confirmed the published moderate MAO-A inhibitory potency [18] and found, not surprisingly, loss of kinase inhibition due to the known steric constraints at the small binding pocket of DYRK1A [19]. For this reason, we did not perform further modifications at C-7 with large substituents.
Modifications at N-9: Introduction of small (C1 to C5) alkyl/alkenyl/alkynyl residues preserved or even slightly improved DYRK1A inhibition, but led only to moderate reduction of MAO-A inhibition. Larger lipophilic residues (benzyl, arylalkyl) at N-9 typically reduced or eliminated both MAO-A and kinase inhibition, and the same holds for aminoalkyl residues. In contrast, small estercontaining residues derived from acetic or propionic acid and short cyanoalkyl groups led to significant reduction of MAO-A (and mostly also CLK1) inhibition, whereby DYRK1A inhibition was not affected. The free carboxylic acid AnnH25, however, showed an opposite shift in selectivity. This made aliphatic esters and even more nitriles at N-9 the most attractive modifications for the elimination of MAO-A inhibitory properties of harmine derivatives. The evidence gained from this first screening was fully confirmed for the cyanoalkyl derivatives AnnH31 and AnnH43, as demonstrated by the IC50 values determined for MAO-A and DYRK1A inhibition [22] (Table 2).
Harmine-bauerine C hybrids: The ring A pyridone analogue (AnnH65) of harmine could not be analyzed properly due to its very poor solubility. The 1-bromo analogue AnnH52 as well as its Nsubstituted derivatives AnnH79/AnnH80 no longer inhibited MAO-A, but also lost DYRK1A inhibitory potency significantly (AnnH52) or almost completely. In contrast, the analogous 1-methyl-7,8-dichloro-β-carboline AnnH63 was identified as a very potent DYRK1A inhibitor, but still showed  Table 1 to illustrate the different degrees of selectivity for DYRK1A. Shown are inhibitors that reduce DYRK1A activity by at least 50%. For comparison, AnnH11 is included as an example for MAO-A selective inhibitor and AnnH35 as a non-selective inhibitor. Columns reflect the residual enzyme activities of DYRK1A (blue) and MAO-A (orange) after incubation with the indicated compounds at a concentration of 1 µM (means and SD).

Discussion of the Screening Results
Results obtained for the alkaloids harmane and norharmine (AnnH90) clearly indicated the importance of the 7-methoxy group for DYRK1A inhibition, but also 7-hydroxy compound harmol was a potent DYRK1A inhibitor (and even more potent on CLK1), notably devoid of MAO-A inhibition, as shown by Balint et al. previously [21]. 1-Styryl analogue AnnH18 was devoid of MAO-A inhibition, but at the cost of significantly reduced potency at DYRK1A.
Modifications at ring C: Bromination or chlorination at C-6 and/or C-8 eliminated DYRK1A inhibition completely. The monohalogenated products (including their N-butylated derivatives (AnnH9, AnnH11) retained significant MAO-A inhibition. This change in selectivity was exactly the opposite of what we intended to achieve. For the 7-aminoethyl ether AnnH62, we confirmed the published moderate MAO-A inhibitory potency [18] and found, not surprisingly, loss of kinase inhibition due to the known steric constraints at the small binding pocket of DYRK1A [19]. For this reason, we did not perform further modifications at C-7 with large substituents.
Modifications at N-9: Introduction of small (C 1 to C 5 ) alkyl/alkenyl/alkynyl residues preserved or even slightly improved DYRK1A inhibition, but led only to moderate reduction of MAO-A inhibition. Larger lipophilic residues (benzyl, arylalkyl) at N-9 typically reduced or eliminated both MAO-A and kinase inhibition, and the same holds for aminoalkyl residues. In contrast, small ester-containing residues derived from acetic or propionic acid and short cyanoalkyl groups led to significant reduction of MAO-A (and mostly also CLK1) inhibition, whereby DYRK1A inhibition was not affected. The free carboxylic acid AnnH25, however, showed an opposite shift in selectivity. This made aliphatic esters and even more nitriles at N-9 the most attractive modifications for the elimination of MAO-A inhibitory properties of harmine derivatives. The evidence gained from this first screening was fully confirmed for the cyanoalkyl derivatives AnnH31 and AnnH43, as demonstrated by the IC 50 values determined for MAO-A and DYRK1A inhibition [22] (Table 2).
Harmine-bauerine C hybrids: The ring A pyridone analogue (AnnH65) of harmine could not be analyzed properly due to its very poor solubility. The 1-bromo analogue AnnH52 as well as its N-substituted derivatives AnnH79/AnnH80 no longer inhibited MAO-A, but also lost DYRK1A inhibitory potency significantly (AnnH52) or almost completely. In contrast, the analogous 1-methyl-7,8-dichloro-β-carboline AnnH63 was identified as a very potent DYRK1A inhibitor, but still showed significant MAO-A inhibition. Introduction of an aminoalkyl residue at N-9 in AnnH71 expectedly reduces DYRK1A inhibition, whereas a cyanomethyl residue in AnnH69 (and surprisingly also a propargyl group; AnnH67) resulted in very interesting compounds with significant DYRK1A inhibition and lack of activity on MAO-A. Docking studies aimed at the identification of the binding mode of AnnH69 are presented in Section 2.3.2.
Investigations on the effect of a significant basicity of N-2 in harmine showed that replacing the pyridine ring A with a (dihydro)pyridone (AnnH19, AnnH20) leads to complete loss of inhibitory activity on all enzymes of interest. The same holds for the 2-desaza analogue AnnHOG3, clearly demonstrating the relevance of N-2 with sufficient basicity for the interaction with DYRK1A. In contrast, the 1-chloro analogue AnnH24 of harmine was found to be a potent DYRK1A inhibitor, free of MAO-A-inhibitory activity. Similar effects were found for its N-cyanomethyl analogue AnnH75 and the N-propargyl compound AnnH74; ester analogue AnnH77 lost potency on DYRK1A, and 9-aminoalkyl derivative AnnH77 expectedly was virtually inactive, once again demonstrating the detrimental effect of this group. Detailed pharmacological investigation (Table 2 and [22]) showed that AnnH75 is the most attractive harmine modification from this project.
Exchanging the 1-methyl group in harmine by an ethyl group (also in combination with an ethoxycarbonylmethyl group at N-9 in AnnH88) did not lead to any beneficial changes in selectivity, but was accompanied by some loss of potency on DYRK1A.
The relevance of the NH group in the central pyrrole ring of harmine was investigated by introduction of other polar groups at this position. Introduction of carbonyl, carbinol and cyanomethylene residues led to products with inferior selectivity and poor DYRK1A inhibition. Only ketone AnnHT2 retained significant MAO-A inhibitory potency.
In conclusion, the following structure-activity relationships can be deduced from the results presented in Tables 1 and 2: 1.
The relevance of the main interactions of harmine with DYRK1A seen in the crystal structure [16] involving direct interactions with N-2 and the 7-methoxy group was confirmed here. Strong inhibition of DYRK1A requires the N-2 with significant basicity. The methoxy group at C-7 is a very well fitting substituent, larger ether groups at this position are not well tolerated. The methoxy group can be replaced by a chloro substituent, which most likely leads to a very similar binding geometry, involving a halogen bridge with the backbone in the hinge region (see Section 2.3.2), similar to previous reports on chlorinated β-carbolines bound to related kinases [25][26][27].

2.
For DYRK1A inhibition, the methyl group at C-1 can be replaced by chlorine, but not by bromine.

3.
Small alkyl groups as well as small polar groups (nitriles, esters, but not carboxylate, carboxamides and aminoalkyl groups) are tolerated at N-9. Replacement of N-9 by carbonyl or carbinol groups leads to a significant reduction of DYRK1A inhibition. 4.
MAO-A-inhibitory activity is most effectively reduced by introduction of small, polar residues (which do not significantly reduce DYRK1A inhibition) at N-9. Further, the 1-chloro motif at C-1 appears beneficial for achieving selectivity for DYRK1A over MAO-A.

5.
As expected, a clear dissection of DYRK1A inhibition from CLK1 inhibition could not be achieved despite the comprehensive structure variations presented here. A few β-carbolines (AnnH21, AnnH26, AnnH27) show some selectivity for CLK1, but on the other hand show insufficient selectivity over MAO-A.

X.ray Crystallography and Docking Studies
We next determined the complexed crystal structure of DYRK1A with AnnH75 (PDB 4YU2) and analyzed the binding mode [45][46][47][48][49][50][51][52]. By comparison, AnnH75 binds to the active site of DYRK1A in a very similar manner as described for harmine (PDB 3ANR) and L41 (PDB 4AZE) (Figure 4) with the pyridine N-2 moiety of AnnH75 engaging in a direct H-bond to the salt bridge (Lys188-Glu203). Nevertheless, the tricyclic core of AnnH75 was slightly shifted away from the catalytic salt bridge compared to harmine, hence resulting in a slightly longer distance between AnnH75 N-2 to Lys188 nitrogen (3.3 Å) than the equivalent moiety in harmine (~3 Å) ( Figure 4B). This may be due to the weaker basicity of the N-2 in AnnH75 (due to the electronegative chlorine substituent at C-1) or to steric constraints of the cyanomethyl residue at position N-9. These effects might contribute to the slight loss in inhibitory activity of AnnH75 compared to harmine (see Table 2). Notably, the cyanomethyl residue at N-9 did not make any direct or indirect interaction with DYRK1A in the structure, but it occupied a binding pocket located under the P-loop.
Molecules 2020, 25, x FOR PEER REVIEW 13 of 34 compared to harmine, hence resulting in a slightly longer distance between AnnH75 N-2 to Lys188 nitrogen (3.3 Å) than the equivalent moiety in harmine (~3 Å) ( Figure 4B). This may be due to the weaker basicity of the N-2 in AnnH75 (due to the electronegative chlorine substituent at C-1) or to steric constraints of the cyanomethyl residue at position N-9. These effects might contribute to the slight loss in inhibitory activity of AnnH75 compared to harmine (see Table 2). Notably, the cyanomethyl residue at N-9 did not make any direct or indirect interaction with DYRK1A in the structure, but it occupied a binding pocket located under the P-loop. Docking AnnH75 into the MAO-A structure (see Figure 1B) revealed that the cyanomethyl group is unfavorable, as this group would generate steric constraints within such a tight pocket where the harmine pyrrole N-9-H is located and forms a water-bridged hydrogen bond to Asn181 ( Figure  5). Upon alkylation of N-9, the molecule is further missing the essential NH group for the formation of the hydrogen bond network. In addition, chloro decoration at C-1 may not be optimal for MAO-A binding either, since it might create some charge clashes with the hydroxyl group of Tyr444. Therefore, this model can explain the loss in affinity of AnnH75 towards MAO-A.

Additional Docking Studies
We further performed docking studies on some other compounds that showed remarkable activity profiles. Harmine-bauerine C hybrid AnnH69 was found to have a selectivity profile very Docking AnnH75 into the MAO-A structure (see Figure 1B) revealed that the cyanomethyl group is unfavorable, as this group would generate steric constraints within such a tight pocket where the harmine pyrrole N-9-H is located and forms a water-bridged hydrogen bond to Asn181 ( Figure 5). Upon alkylation of N-9, the molecule is further missing the essential NH group for the formation of the hydrogen bond network. In addition, chloro decoration at C-1 may not be optimal for MAO-A binding either, since it might create some charge clashes with the hydroxyl group of Tyr444. Therefore, this model can explain the loss in affinity of AnnH75 towards MAO-A.
Molecules 2020, 25, x FOR PEER REVIEW 13 of 34 compared to harmine, hence resulting in a slightly longer distance between AnnH75 N-2 to Lys188 nitrogen (3.3 Å) than the equivalent moiety in harmine (~3 Å) ( Figure 4B). This may be due to the weaker basicity of the N-2 in AnnH75 (due to the electronegative chlorine substituent at C-1) or to steric constraints of the cyanomethyl residue at position N-9. These effects might contribute to the slight loss in inhibitory activity of AnnH75 compared to harmine (see Table 2). Notably, the cyanomethyl residue at N-9 did not make any direct or indirect interaction with DYRK1A in the structure, but it occupied a binding pocket located under the P-loop. Docking AnnH75 into the MAO-A structure (see Figure 1B) revealed that the cyanomethyl group is unfavorable, as this group would generate steric constraints within such a tight pocket where the harmine pyrrole N-9-H is located and forms a water-bridged hydrogen bond to Asn181 ( Figure  5). Upon alkylation of N-9, the molecule is further missing the essential NH group for the formation of the hydrogen bond network. In addition, chloro decoration at C-1 may not be optimal for MAO-A binding either, since it might create some charge clashes with the hydroxyl group of Tyr444. Therefore, this model can explain the loss in affinity of AnnH75 towards MAO-A.

Additional Docking Studies
We further performed docking studies on some other compounds that showed remarkable activity profiles. Harmine-bauerine C hybrid AnnH69 was found to have a selectivity profile very

Additional Docking Studies
We further performed docking studies on some other compounds that showed remarkable activity profiles. Harmine-bauerine C hybrid AnnH69 was found to have a selectivity profile very similar to AnnH75 (no inhibition of MAO-A, strong inhibition of DYRK1A and CLK1; see Table 1), but with slightly reduced inhibition of both kinases. Docking experiments suggested a binding pose in DYRK1A closely related to the one of AnnH75, with pyridine N-2 forming a hydrogen bridge with Lys144 and, as expected from the binding of our previously developed kinase inhibitors KH-CB19 and KH-CARB13 [25,26] (Figure 2B), an interaction of the chlorine substituent at C-7 with the backbone of Leu241 in the hinge region ( Figure 6). Docking AnnH69 with MAO-A did not give a favorable docking pose, most likely due to steric constraints of the N-cyanomethyl group.
Molecules 2020, 25, x FOR PEER REVIEW 14 of 34 similar to AnnH75 (no inhibition of MAO-A, strong inhibition of DYRK1A and CLK1; see Table 1), but with slightly reduced inhibition of both kinases. Docking experiments suggested a binding pose in DYRK1A closely related to the one of AnnH75, with pyridine N-2 forming a hydrogen bridge with Lys144 and, as expected from the binding of our previously developed kinase inhibitors KH-CB19 and KH-CARB13 [25,26] (Figure 2B), an interaction of the chlorine substituent at C-7 with the backbone of Leu241 in the hinge region ( Figure 6). Docking AnnH69 with MAO-A did not give a favorable docking pose, most likely due to steric constraints of the N-cyanomethyl group. The 9-desazaharmine analogues AnnHT2 (carbonyl group instead of NH) and AnnHT3 (corresponding secondary alcohol) showed strongly reduced inhibition of DYRK1A (and CLK1), but still strong (AnnHT2) or medium (AnnHT3) inhibition of MAO-A (Table 1). Docking both compounds into DYRK1A (PDB 4YU2) suggested binding poses very similar to those of harmine ( Figure 1A) and AnnH75 ( Figure 4) under involvement of N-2 and the 7-methoxy group ( Figure 7A,B). Calculation of the molecular electrostatic potentials, however, showed unfavorable electronegative potentials for the desaza analogues, whereas harmine is electropositive around N-9 ( Figure 7C). Most likely, this change in electrostatic potentials significantly reduces binding to the active site of DYRK1A.  The 9-desazaharmine analogues AnnHT2 (carbonyl group instead of NH) and AnnHT3 (corresponding secondary alcohol) showed strongly reduced inhibition of DYRK1A (and CLK1), but still strong (AnnHT2) or medium (AnnHT3) inhibition of MAO-A (Table 1). Docking both compounds into DYRK1A (PDB 4YU2) suggested binding poses very similar to those of harmine ( Figure 1A) and AnnH75 (Figure 4) under involvement of N-2 and the 7-methoxy group ( Figure 7A,B). Calculation of the molecular electrostatic potentials, however, showed unfavorable electronegative potentials for the desaza analogues, whereas harmine is electropositive around N-9 ( Figure 7C). Most likely, this change in electrostatic potentials significantly reduces binding to the active site of DYRK1A.
Docking AnnHT2 ( Figure 8A) and AnnHT3 ( Figure 8B) into MAO-A (PDB 2Z5X) revealed binding poses similar to those of harmine ( Figure 1B), with N-2 forming the same water-mediated hydrogen bond networks as harmine. The carbonyl (in AnnHT2) and carbinol (in AnnHT3) groups, however, are not oriented in the same direction as the 9-NH of harmine; rather, the molecules are oriented in an inverse manner, and do not form a second water-mediated hydrogen bridge network (with Ile207 and Asn181). This difference in orientation and different polar interactions might explain the moderate (AnnHT2) or stronger (AnnHT3) reduction of MAO-A inhibitory potency of these analogues. strong (AnnHT2) or medium (AnnHT3) inhibition of MAO-A (Table 1). Docking both compounds into DYRK1A (PDB 4YU2) suggested binding poses very similar to those of harmine ( Figure 1A) and AnnH75 (Figure 4) under involvement of N-2 and the 7-methoxy group ( Figure 7A,B). Calculation of the molecular electrostatic potentials, however, showed unfavorable electronegative potentials for the desaza analogues, whereas harmine is electropositive around N-9 ( Figure 7C). Most likely, this change in electrostatic potentials significantly reduces binding to the active site of DYRK1A.  Docking AnnHT2 ( Figure 8A) and AnnHT3 ( Figure 8B) into MAO-A (PDB 2Z5X) revealed binding poses similar to those of harmine ( Figure 1B), with N-2 forming the same water-mediated hydrogen bond networks as harmine. The carbonyl (in AnnHT2) and carbinol (in AnnHT3) groups, however, are not oriented in the same direction as the 9-NH of harmine; rather, the molecules are oriented in an inverse manner, and do not form a second water-mediated hydrogen bridge network (with Ile207 and Asn181). This difference in orientation and different polar interactions might explain the moderate (AnnHT2) or stronger (AnnHT3) reduction of MAO-A inhibitory potency of these analogues.

Discussion
The protein kinase DYRK1A has important functions in neuronal development and cell cycle control, and has attracted large attention as a possible drug target for treatment of neurodegenerative processes, certain cancers, diabetes, and other diseases [2]. The β-carboline alkaloid harmine is a potent DYRK1A inhibitor, but it suffers from undesired strong inhibition of MAO-A, which would prevent application as a drug or as an investigative DYRK1A inhibitor in vivo. We synthesized more than 60 analogues of harmine, either by direct modification of the native alkaloid or by de novo synthesis of the β-carboline and related scaffolds in order to investigate structure-activity relationships for DYRK1A inhibition and to separate this desired activity from undesired MAO-A inhibition. Based on published crystal structures of harmine bound to DYRK1A and MAO-A, we modified all relevant functionalities in the molecule systematically, and identified a number of promising harmine analogues, which are characterized by optimized substituents at position N-9 (which preserve DYRK1A inhibition and eliminate MAO-A inhibition) and for DYRK1A binding beneficial moieties at C-1 (methyl or chlorine). From this series, AnnH75 was identified as the most

Discussion
The protein kinase DYRK1A has important functions in neuronal development and cell cycle control, and has attracted large attention as a possible drug target for treatment of neurodegenerative processes, certain cancers, diabetes, and other diseases [2]. The β-carboline alkaloid harmine is a potent DYRK1A inhibitor, but it suffers from undesired strong inhibition of MAO-A, which would prevent application as a drug or as an investigative DYRK1A inhibitor in vivo. We synthesized more than 60 analogues of harmine, either by direct modification of the native alkaloid or by de novo synthesis of the β-carboline and related scaffolds in order to investigate structure-activity relationships for DYRK1A inhibition and to separate this desired activity from undesired MAO-A inhibition. Based on published crystal structures of harmine bound to DYRK1A and MAO-A, we modified all relevant functionalities in the molecule systematically, and identified a number of promising harmine analogues, which are characterized by optimized substituents at position N-9 (which preserve DYRK1A inhibition and eliminate MAO-A inhibition) and for DYRK1A binding beneficial moieties at C-1 (methyl or chlorine). From this series, AnnH75 was identified as the most attractive DYRK1A inhibitor, which completely lacks MAO-A inhibition.
The binding mode of this inhibitor was elucidated by crystal structure analysis (PDB 4YU2), and docking experiments gave additional insight into binding of this and related compounds to DYRK1A and MAO-A. The chemical space in the field of β-carbolines was widely explored here, but most recent achievements from our research group on the introduction of additional substituents into ring A [53,54] might even open additional opportunities for optimization.
AnnH75 is a valuable chemical tool for further investigation of the physiological role of DYRK1A, and the comprehensive evidence on structure-activity relationships from this project should further inspire future attempts in the development of DYRK1A inhibitors.

Enzyme Assays
Inhibitors were dissolved in DMSO at a concentration of 10 mM and stored at 20 • C. Working solutions were prepared to achieve a final concentration of 1 µM in the enzyme assays.
Recombinant kinases were purified from E. coli as GST (glutathione S-transferase) fusion constructs by affinity adsorption to glutathione-Sepharose. The expression constructs for GST-DYRK1A-∆C (containing amino acids 1-499 of rat DYRK1A) and GST-CLK1cat (containing amino acids 141-484 of human CLK1) have been described before [22]. Catalytic activity of the kinases was measured with the help of the Kinase-GLO TM Luminescent Assay from Promega. Assays were performed in a total volume of 10 µL in kinase buffer (25 mM Hepes pH 7.4, 0.5 mM DTT, 5 mM MgCl 2 , 5 µM ATP) with appropriate peptide substrates (20 µM DYRKtide for the DYRKs and 100 µM DYRKtide for HIPK2, RRRFRPASPLRGPPK; or 100 µM RS peptide for CLK1, GRSRSRSRSR). Kinases amounts in the assays were adjusted so that approximately 90% of the ATP in the assay was consumed in control samples without inhibitor. Kinase reactions were incubated at room temperature for 30 min before 10 µL Kinase-GLO reagent was added for luminometric measurement of the ATP that was not consumed by the kinases. After incubation at room temperature for an additional 10 min, luminescence was recorded for 1 s (Berthold Orion Microplate Luminometer). ATP consumption in the control samples without inhibitor was set as 100% catalytic activity, and the inhibitory effects of the test compounds were calculated relative to this value.
Catalytic activity of monoamine oxidase A (MAO-A) was measured by using the luminescence-based MAO-GLO TM Assay from Promega. Recombinant MAO-A converts a luminogenic MAO substrate into luciferin, which is, in turn, luminometrically detected in a luciferase reaction. The amount of light produced is directly proportional to MAO activity. Assays were run at room temperature for 1 h with 12 µU human recombinant MAO-A and 25 µM MAO-A substrate in a total volume of 20 µL. Inhibitory effects of the tested compounds were calculated relative to the activity of control reactions that were run in the absence of inhibitors.
Kinase and MAO-A assays were run with duplicate measurements, and the assay results in Table 1 and Figure 3 represent means of n = 3 experiments except for the following values: DYRK1A results represent n = 5 for AnnH43, AnnH44, n = 7 for AnnH31, MAO-A results represent n = 4 for AnnH38, AnnH43 and CLK1 results represent n = 4 for AnnH18, AnnH24, AnnH38, AnnH43, n = 5 for AnnH12 and n = 7 for AnnH31 and n = 2 for a number of compounds as indicated in the footnotes to Table 1.

Crystallization and Structure Determination Protein Production, Crystallization, Data Collection and Structure Determination
The recombinant DYRK1A kinase domain protein was expressed and purified as previously described [45]. The N-termial His 6 tag was removed, and the resulting cleaved protein was buffered in 50 mM HEPES pH 7.5, 500 mM NaCl and 5 mM DTT. The protein was concentrated to 15.6 mg/mL and was mixed with 1 mM AnnH75. Crystallization was performed using the sitting drop vapor diffusion method at 4 • C and the reservoir solution containing 31% PEG 400, 0.2 M Li 2 SO 4 and 0.1 M Tris pH 9.0. Diffraction data were collected at Diamond Light Source, beamline I04-1, and were processed and scaled with XDS [46] and Scala from CCP4 suite [47], respectively. The structure determination was achieved by molecular replacement using Phaser [48] and the coordinates of DYRK1A structure [45] as a search model. Manual model building was performed in COOT [49], and the structure was refined using Refmac [50] with a TLS model calculated from TLSMD server [51]. The final model was verified for its geometric correctness with MOLPROBITY [52]. The data collection and refinement statistics are summarized in Table 3.

Computational Methods
The GLIDE software version 9.3 was used for docking, whereas the calculation of all molecular descriptors and the analysis of the docking results were carried out in MOE20012.10 (Chemical Computing Group). All ligands were constructed using MOE2012.10 and energy minimized using the MMFF94 force field with a convergence criteria of 0.1 kcal/mol. All compounds were first docked into the binding pocket of the X-ray structure of DYRK1A (PDB code 3ANR complexed with harmine and which represents the active form of DYRK1A), CLK1 (PDB code 2VAG complexed with KH-CB19, which represents the active form of CLK1), and MAO-A (PDB code 2Z5X complexed with harmine).
The cocrystallized inhibitor was defined as centre of the enclosing box used for docking with a radius of 15 Å).
For docking into DYRK1A, two protein hydrogen bonds were defined, which are observed in the DYRK1A-harmine crystal structure: one to Leu241, which is part of the hinge region, and one to Lys188. No water molecules were considered for ligand docking. To test whether the used docking protocol is suitable for DYRK1A docking, the cocrystallized inhibitor was redocked into DYRK1A. Using GlideScore as scoring function, an RMSD value of 0.39 Å was derived for the top-ranked docking pose of harmin. For all docked inhibitors, the GlideScore was calculated and analyzed.
During the course of the project, we were able to cocrystallized AnnH75 with DYRK1A (PDB code 4YU2) and confirmed the predicted docking pose of AnnH75 (RMSD 0.22-0.26 Å, in the four chains of 4YU2). All studied compounds were then also docked to 4YU2 and similar docking poses as for 3ANR were obtained.
Docking to CLK1 was carried out using the crystal structure in complex with the small molecule inhibitor KH-CB19 which has a similar size and shape as harmine. Two hydrogen bonds were defined similar as in case of DYRK1A: one to the hinge region residue Leu244 and one to Lys191. To test whether the used docking protocol is suitable for CLK1 docking, the cocrystallized inhibitor was redocked into the ATP binding pocket. Using GlideScore as scoring function, an RMSD value of 0.61 Å was derived for the top-ranked docking pose of KH-CB19. For all docked inhibitors, the GlideScore was calculated and analyzed.
For docking into MAO-A, the cofactor and most of the water molecules were removed. A total of seven water molecules, found inside the binding pocket of MAO-A, were considered for docking. To test the applicability of the docking tool, a control docking was carried out using the cocrystallized inhibitor harmin. Using GlideScore as scoring function, an RMSD value of 0.94 Å was derived for the top-ranked pose of harmin. For all compounds under study, the GlideScore was calculated and analyzed.
To calculate the molecular electrostatic potential (MEP) of the studied inhibitors, ESP-AM1 partial charges were calculated using program MOE2012.10. The electrostatic potential was calculated using the MM-GBSA option and displayed on the van-der-Waals surface of the inhibitors.

General
All chemicals and solvents were used as purchased from commercial services (Sigma-Aldrich, Acros, Alfa Aesar) without further purification. The progress of all reactions was monitored by TLC on Machery nagel polygram sil plates G/UV 254 (0.2 mm, 40 × 80 mm). Flash column chromatography was performed on Merck silica gel Si 60 (40-63 µm). NMR spectra were recorded on a Jeol GSX 400 or Jeol JNMR-GX 500 (Jeol, Peabody, MA, USA), using TMS as internal standard. Chemical shifts are reported in ppm and given in δ units. Coupling constants are given in Hertz. The spectra were recorded at room temperature. Mass spectra (electron ionization, EI, 70 eV) were recorded using a Hewlett Packard 5989A mass spectrometer with a 59,980 B particle beam LC/MS interface (Agilent Technologies, PAO Alto, CA, USA) or Thermo Finnigan MAT95 mass spectrometer (Thermo Fischer Scientific, Waltham, MA, USA). Mass spectra (APCI) were recorded using a ABSciex API 2000 LC/MS/MS (AB SCIEX, Foster City, CA, USA). Mass spectra (ESI) was recorded using a Thermo Finnigan LTQ FT Ultra (Thermo Fischer Scientific, Waltham, MA, USA). HRMS (EI) were performed using a Jeol JMS GCmate II (Jeol, Peabody, MA, USA) or HRMS (ESI) using a Thermo Finnigan LTQ FT (Thermo Electron Corporation, Waltham, MA, USA). IR spectra were recorded as KBr disks on a Perkin Elmer FT-IR Parago 1000 (Perkin Elmer, Waltham, MA, USA) or JASCO FT/IR-410 (Jasco, Easton, PA, USA). Melting points were determined with a Büchi B-540 apparatus (Büchi, Flawil, Switzerland) and are uncorrected. Purity of all compounds was determined with a HP Agilent 1100 HPLC (diode array detector, Agilent Technologies, Waldbronn, Germany) equipped with an Agilent Poroshell column (120 EC-C18; 3.0 × 100 mm, 2.7 microns) using following conditions: eluent acetonitrile/water/THF 8:2:0.1 or 6:4:0.1, flow rate: 0.9 mL/min, column temperature 50 • C, detection at 210 nm and 254 nm, injection of 5 µL of 100 µg/mL solutions of the compound. Microwave-assisted reactions were conducted using a single-mode microwave reactor Discover SP (300 W, CEM, Matthews, NC, USA).

Synthesis of Compounds
General procedure A for the N-alkylation of varios β-carbolines with using of sodium hydride as base in DMF. To the appropiate β-carboline in dry DMF sodium hydride (60% in mineral oil) was added. The mixture was stirred at 40 • C for 20 min and the appropriate alkyl halide was added. The mixture was stirred at 40 • C for the specified time period and then over night at room temperature. The following detailed procedure for isolation of AnnH44 is representative for the preparation of AnnH44, AnnH31, AnnH70, AnnH73, AnnH35, AnnH36, AnnH39, AnnH40, AnnH38, AnnH12,  AnnH26, AnnH32, AnnH30, AnnH28, AnnH34, AnnH43, AnnH74, AnnH75, AnnH76, AnnH80, AnnH79, AnnH67, AnnH69 and AnnH66.
General procedure B for the N-alkylation of β-carbolines using potassium tert-butoxide as base in DMSO. The appropriate β-carboline and potassium tert-butoxide were dissolved in anhydrous DMSO and stirred for 30 min at 80 • C. The reaction mixture was cooled to room temperature and then the appropriate alkyl halide was added and the mixture stirred at 80 • C for the specified time period and then cooled to room temperature. After quenching with cold aqueous ammonia solution (10%, 10 mL), the mixture was treated with water until a solid was precipitated. The aqueos layer was extracted with ethyl acetate (3 × 50 mL). The combined organics were dried over sodium sulfate and evaporated to dryness. The residue was purified by column chromatography (silica gel, iso-hexane/ethyl acetate/ethanol 2:2:1) to afford the corresponding alkylated β-carbolines AnnH9, AnnH11, AnnH14, AnnH3, AnnH4, AnnH22 and AnnH27.
General procedure C for the N-alkylation of β-carbolines using sodium hydride as base in THF. A suspension of the appropriate β-carboline and sodium hydride (60% in mineral oil) in THF was stirred for 20 min, then the appropriate alkyl halide was added and the mixture was stirred under reflux for 65 h. The solution was allowed to cool to room temperature and water (10 mL) was added. The organic layer was separated and the aqueos layer was extracted with DCM (3 × 25 mL). The combined organic layers were dried over sodium sulfate and evaporated to dryness. The residue was purified by column chromatography (silica gel, iso-hexane/ethyl acetate/ethanol/triethylamine 2:2:1 +1%) to afford the corresponding alkylated β-carbolines AnnH16, AnnH21, AnnH53, AnnH55, AnnH57, AnnH71 and AnnH77.