Discovery and Evaluation of Enantiopure 9H-pyrimido[4,5-b]indoles as Nanomolar GSK-3β Inhibitors with Improved Metabolic Stability

Glycogen synthase kinase-3β (GSK-3β) is a potential target in the field of Alzheimer’s disease drug discovery. We recently reported a new class of 9H-pyrimido[4,5-b]indole-based GSK-3β inhibitors, of which 3-(3-((7-chloro-9H-pyrimido[4,5-b]indol-4-yl)(methyl)amino)piperidin-1-yl)propanenitrile (1) demonstrated promising inhibitory potency. However, this compound underwent rapid degradation by human liver microsomes. Starting from 1, we prepared a series of amide-based derivatives and studied their structure–activity relationships against GSK-3β supported by 1 µs molecular dynamics simulations. The biological potency of this series was substantially enhanced by identifying the eutomer configuration at the stereocenter. Moreover, the introduction of an amide bond proved to be an effective strategy to eliminate the metabolic hotspot. The most potent compounds, (R)-3-(3-((7-chloro-9H-pyrimido[4,5-b]indol-4-yl)(methyl)amino)piperidin-1-yl)-3-oxopropanenitrile ((R)-2) and (R)-1-(3-((7-bromo-9Hpyrimido[4,5-b]indol-4-yl)(methyl)amino)piperidin-1-yl)propan-1-one ((R)-28), exhibited IC50 values of 480 nM and 360 nM, respectively, and displayed improved metabolic stability. Their favorable biological profile is complemented by minimal cytotoxicity and neuroprotective properties.


Introduction
Glycogen synthase kinase-3 (GSK-3) is a ubiquitously expressed protein existing in two highly related isoforms, GSK-3α and GSK-3β [1]. This serine/threonine kinase is involved in several signal cascades and is assumed to phosphorylate more than 30 different substrates, highlighting its multifaceted role in intracellular processes [1,2]. Under physiological conditions, complex mechanisms ensure a strict regulation and proper functionality of this highly multitasking enzyme. In turn, aberrant activity of GSK-3 is assumed to be a critical factor for the development of diverse pathologies, including diabetes, cancer, bipolar disorder, and especially Alzheimer's disease (AD) [3]. In particular, GSK-3β is implicated in crucial mechanisms associated with AD pathology. These include tau hyperphosphorylation and Aβ generation, the two major histopathological hallmarks of the disease [4]. Therefore, GSK-3β is a potential target for novel disease-modifying AD therapeutics motivating drug discovery efforts in the field of small molecule kinase inhibitors.
Recently, we reported on the optimization of a novel class of 7-chloro-9H-pyrimido [4,5-b] indole-based glycogen synthase kinase-3β (GSK-3β) inhibitors, including compound 1 (Figure 1a) [5]. This tertiary alicyclic amine with a promising biological activity on the target enzyme, however, suffered from poor metabolic stability when exposed to human liver microsomes (HLMs). In the HLM stability assay, inhibitor 1 underwent rapid biotransformation, resulting in a limited half-life of approximately 30 min (Figure 1b). Tertiary alicyclic amines represent common yet metabolically vulnerable motifs in pharmaceutical agents. Their oxidative metabolization typically includes chemical modifications such as α-carbonyl introduction and ring opening as well as oxygenation or dealkylation of the nitrogen atom [6]. Correspondingly, the observed metabolic lability of 1 can be ascribed to the (2-cyanoethyl)piperidine substructure of the molecule. The mass spectrometry-based profiling of the metabolites formed from 1 in the HLM experiment indicated an extensive elimination of the cyanoethyl substituent through C-N bond cleavage. However, we found this moiety to be of utmost importance for the biochemical activity of 1 in our preliminary optimization study. This suggests that its removal would compromise the activity of this compound.
Herein, we present a strategy to improve the metabolic stability of this class of 9H-pyrimido [4,5-b] indole-based GSK-3β inhibitors while maintaining the biological activity. We hypothesized that the introduction of an acyl substituent on the piperidine nitrogen atom was a suitable approach to eliminate the potential metabolic hotspot. To this end, we refocused our attention on the reported amide derivative 2, which was 2.5-fold less active than 1 [5]. We used this compound as a template in order to design and optimize a series of novel amide-based GSK-3β inhibitors by applying a variety of structural modifications (Scheme 1). The obtained biological data established structure-activity relationships (SARs), which were substantiated by in silico approaches. The most promising candidates were assessed for their metabolic stability in the HLM experiment and further were characterized in cellular assays. Scheme 1. Development of novel amide-based Glycogen synthase kinase-3β (GSK-3β) inhibitors; IC 50 : halfmaximal inhibitory concentration.

Biological Evaluation
In our initial attempts to optimize 2 for GSK-3β inhibition, we focused on the piperidine nitrogen substituent and evaluated its effect on the compound potency ( Table 1). The application of bulky moieties including aromatic rings (8 and 10), an ethyl ester function (9), or a tert-butyloxy group (6c) led to inactive compounds. Analogs carrying substituents with longer aliphatic moieties such as butanoyl- (11), isovaleryl- (12), or cyclopropylacetyl- (15) were inactive or exhibited significantly higher IC 50 values than 2. Derivatives with shorter hydrocarbon chains, i.e., cyclopropanoyl-(39), acryloyl-(40), 3-(dimethylamino)propanoyl- (13), and tert-butanoyl-(38), were found to retain moderate biological activity. In agreement with this trend, a slight potency improvement was seen with the acetyl-and propanoyl-substituted compounds 14 and 16 compared to their cyanoacetyl counterpart. Perhaps surprisingly, the 3,3,3-trifluoropropanoyl derivative 17 was clearly less active than its propanoyl congener 16. This dramatical difference in potency can be rationalized by the larger van der Waals volume of the CF 3 group as compared to a CH 3 group [7]. These results indicate that only short-chained substituents such as acetyl-, cyanoacetyl-, or propanoyl-are tolerated in this position.
To assess the effect of the carbonyl group functionality, we replaced it with an oxetane ring (41). This four-membered heterocyclic motif has been successfully applied in medicinal chemistry programs in order to improve the metabolic stability of drug candidates [8]. The resulting compound 41, however, displayed a 4-fold higher IC 50 value than 2. An oxetane is considered to form weaker hydrogen bonds than an amide carbonyl group [9]. Thus, we ascertain that carbonyl group-mediated interactions are important for the binding affinity of 2.  [10,11] (for details, see the Supplementary Materials) and are the means of at least two independent experiments; b data taken from [5]; c n = 1.
The results obtained from the piperidine substituent series highlighted the suitability of both the cyanoacetyl and propanoyl substituent in this position. Based on these results, we elected to maintain both moieties for further optimization and shifted our focus to other positions of the scaffold that are amenable to modification ( Table 2). a IC 50 values were determined in an ADP Glo™ Kinase assay and are the means of at least two independent experiments; b data taken from [5].
Replacement of the chlorine atom in the 7-position of the tricyclic core (R 2 in Table 2) with different halides, i.e., fluorine (19), bromine (20), and iodine (21), as well as removal of this substituent (18) was well tolerated. In fact, the inhibitory potency within these compounds marginally increased alongside the van der Waals radii of the halogens. In contrast, the introduction of a methoxy (22) or CF 3 (23) group in this position gave inactive derivatives, presumably due to a steric clash at the target binding site. Similarly, the addition of a methyl group in the 2-position of the 7-chloro-9H-pyrimido [4,5-b]indole (24) resulted in a substantial loss in potency. Within this series, the removal of the N-methyl group (R 4 in Table 2) typically proved unfavorable to the activity (45-50), which prompted us to maintain this substituent.
Accordingly, the activity trend observed for the halogen series in Table 2 was maintained with the propionamides ( Table 3). The only exception was the iodine derivative (29), which exhibited a 5-fold reduced potency compared to its bromine counterpart 28. Unfortunately, no inhibition data could be generated for the methylated propionamide 27 due to insufficient aqueous solubility.
The relocation of the R 2 substituent to the adjacent 6-position retained potency in the case of chlorine (30 and 34) and bromine (31 and 35) ( Table 4). In contrast to compound 22, the methoxy substituent was also tolerated in this position (32 and 36). However, compounds 33 and 37 carrying chlorine in the 5-position of the tricyclic scaffold suffered from a decrease in activity relative to their regioisomers 2 and 15.  a IC50 values were determined in an ADP Glo™ Kinase assay and are the means of at least two independent experiments; b data taken from [5]; c not determined.
As observed in several instances [12,13], the bioactivity of enantiomers may be strikingly different. The compounds presented in this work have a stereocenter in the 3-position of the piperidine ring. We expected the conformations of the different enantiomers to be unidentical, which motivated us to examine the influence of stereochemistry on the compound activity. To this end, we prepared the enantiopure analogs (R)-2/(S)-2 and (R)-20/(S)-20 of inhibitors 2 and 20, respectively, and determined their IC 50 values. Within these two matched pairs, the respective (S)-configured enantiomers showed a dramatic loss in activity, while their (R)-configured stereoisomers displayed IC 50 values in the nanomolar range. Having identified the eutomer configuration, we also prepared (R)-28, which followed the same trend and exhibited an improved IC 50 value of 360 nM. These findings demonstrated that only the (R)-enantiomer displays GSK3β inhibition among these compounds.

Molecular Modeling
To gain insight into the binding interactions of these amide-based compounds within the ATP site of GSK-3β, we performed 1 µs molecular dynamics (MD) simulations for the most potent inhibitors (R)-2 and (R)-28. The observed binding modes for (R)-2 and (R)-28 are highly similar and provide plausible explanations for structure-activity relationships of the compound series (Figure 2a,b). Both compounds display extremely stable hydrogen bonding interactions to the hinge region residues Asp133 and Val135 (>95% frequency, Figure S1). The halogen substituent in the 7-position of the tricyclic scaffold (chlorine in (R)-2 and bromine in(R)-28) is pointing towards the hydrophobic region I of the kinase. Furthermore, the observed low values of ligand root-mean-square deviation (RMSD) highlight the stable binding conformation of both ligands. Putative binding mode of the described GSK3β inhibitors: (a) representative snapshot of the (R)-2 binding conformation during the 1 µs molecular dynamics (MD) simulation. All residues located within 4 Å of the ligand during the simulation are shown as sticks. Hydrogen bonds are depicted with cyan dashed lines; (b) representative snapshot of the (R)-28 binding conformation during the 1 µs MD simulation; (c) the ligands appear extremely stable in the shown conformation throughout the simulations as demonstrated by their root-mean-square deviation (RMSD) values; (d) the carbonyl oxygen points towards the polar residues Asp200, Lys85, Ser66, and Glu97 but does not form any specific direct interactions with GSK-3β; (e) the proton in the 2-position of the pyrimidine ring is in close proximity to the side chain of Tyr134. Here, the conformation of (a) is illustrated, where the distance is 2.87 Å. The (R)-2 and Tyr134 atoms are shown as spheres representing their van der Waals radius; and (f) comparison of the preferred quantum mechanics (QM)-derived conformations of the different enantiomers of 2. (S)-2 is unable to obtain the binding conformation as observed for (R)-2.
In the preferred (S)-2 conformation, the piperidine ring would clash with the β-sheet and Val70 and would not occupy the small lipophilic pocket formed by Leu188 and Thr138.
The carbonyl oxygen, which points towards the polar residues Asp200, Lys85, Ser66, and Glu87, does not form any direct interactions to any residues of GSK3β throughout the simulations (Figure 2d). Nevertheless, water-mediated interactions are evident for the carbonyl group ( Figure S1), which might explain the enhanced activity of 2 compared to its oxetane bioisoster 41.
The piperidine ring of both compounds occupies a small lipophilic pocket formed by Leu188 (and Thr138) in the bottom of the binding site (Figure 2a,b). The amide substituent on the piperidine nitrogen atom is oriented towards the solvent interface with polar residues, e.g., Lys183. We postulate that this space, which does not present any suitable binding sites for accommodation, especially of bulkier lipophilic substituents, provides an explanation for the observed activity trends in Table 1.
The proton in the 2-position of the pyrimidine ring is in close proximity and in contact with the side chain of hinge residue Tyr134 (Figure 2e). This observation offers an explanation for the inactivity of 2-methyl derivative 24, as this substitution would result in a steric clash with the side chain of Tyr134.
The remarkable activity difference among the stereoisomers can be realized via the comparison of the preferred quantum mechanics (QM)-derived conformations of (R)-2 and (S)-2 ( Figure 2f). The superimposition of the enantiomers clearly demonstrates that a favorable configuration of the piperidine is not possible for (S)-2. This enantiomer not only would be unable to occupy the small lipophilic pocket but also would clash with the β-sheet next to the G-loop (Figure 2f).

Microsomal Stability
As our initial aim was to improve the metabolic stability of the compounds, we evaluated the most potent eutomers (R)-2 and (R)-28 in our in-house microsomal stability assay. To this end, the compounds were incubated with pooled male and female HLM for 120 min, and the compound degradation as well as the formation of metabolites was monitored by liquid chromatography-mass spectrometry (LC-MS) analysis.
In contrast to the labile tertiary amine 1, the enantiopure cyanoacetamide compound (R)-2 displayed a favorable metabolic profile in the microsomal stability assay (Figure 3a). The metabolization of this inhibitor was characterized by a slower degradation rate and consequently a larger fraction of intact compound after the time span of 120 min. Some metabolites with an m/z ratio of 399 were detected in the LC-MS analysis and likely result from monooxygenation of the parent compound. However, no elimination of the piperidine nitrogen substituent was seen, which supports our initial hypothesis concerning stability of the amide bond.
The propanoyl analog (R)-28 even showed a slightly enhanced stability compared to its cyanoacetamide counterpart. Nearly 70% of unchanged inhibitors was detected after the incubation time of 120 min ( Figure 3b). We assume that this noticeable decrease in decomposition can be attributed to the lack of the metabolically vulnerable nitrile group. Similar to (R)-2, the LC-MS-based metabolite profiling indicates formation of oxygenated derivatives of (R)-28 yet no cleavage of the piperidine amide bond.

Cell Data
To extend the biological profile of inhibitor (R)-28, we evaluated the in vitro cytotoxic potential of this compound on a variety of cell lines. These included two wild-type cell lines (human lung fibroblast cell line MRC-5 and Chinese hamster ovary cell line CHO-K1) as well as three cancer cell lines (hepatocellular carcinoma cell line HepG2, human breast adenocarcinoma cell line MCF-7, and human neuroblastoma cell line SH-SY5Y). The cells were treated with different concentrations of (R)-28, and their viability was assessed by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay (for details, see Supplementary Materials). Even at the highest tested concentration of 10 µM, minimal to no cytotoxic effects were seen for (R)-28 on all cell lines, demonstrating a highly favorable cytotoxicity profile ( Figures S2 and S3). These results motivated us to further characterize compound (R)-28 by its ability to inhibit GSK-3β and to exert neuroprotective effects in neuronal SH-SY5Y cells (for details, see Supplementary Materials). Initially, we determined the concentrations of (R)-28 not associated to neurotoxicity by the MTT assay [14]. Therefore, concentrations of 1 and 5 µM were selected for the assays in SH-SY5Y cells, which were performed according to previously described protocols [15][16][17]. At the tested concentration of 1 µM, (R)-28 inhibited GSK3β activity in terms of inactive phospho-GSK3α/β (Ser21/9) increase and active phospho-GSK3α/β (Tyr279/Tyr216) decrease after 1 h of treatment in neuronal SH-SY5Y cells ( Figure S4).

Chemistry
We previously reported a synthetic route for the preparation of compound 2 [5]. This strategy demonstrated broad applicability within the herein presented study, as it provided access to the majority of final compounds with only minor alterations in the experimental protocols (Scheme 2). 4-Chloro-9H-pyrimido [4,5-b]indoles 3a-l were prepared in four steps from commercially available o-halonitrobenzenes according to modified literature procedures (for details, see Supplementary Materials) [5,[18][19][20]. These tricyclic intermediates were protected by a tosyl group on the indole nitrogen, resulting in 4a-l. The aliphatic side chain was then introduced by treatment of 4a-l with appropriate amines under basic conditions. In the case of 6a-l, we used racemic 1-Boc-3-(methylamino)piperidine, which was prepared as described previously [5]. For the synthesis of analogs (R)-5c,d and (S)-5c,d, enantiopure 1-Boc-3-aminopiperidine building blocks were utilized. This demanded an additional methylation step, which was carried out with methyl iodide under strictly anhydrous, basic conditions [21]. Subsequent cleavage of the orthogonal protecting groups furnished precursors 7a-l, (R)-7c,d, and (S)-7c,d, which were acylated on the piperidine nitrogen to access the final compounds. Searching for a generally applicable and regioselective amide coupling procedure, various reagents including 3-(ethyliminomethylidenamino)-N,N-dimethyl-propan-1-amine (EDCI), benzotriazol-1-yloxytripyrrolidinophosphonium hexafluorophosphate (PyBOP), and O-(benzotriazol-1-yl)-N,N,N ,Ntetramethyluronium tetrafluoroborate (TBTU) were tested. The application of EDCI suffered from a slow and inefficient conversion, while utilizing PyBOP required a laborious separation of the liberated tri(pyrrolidin-1-yl)phophine oxide. In contrast, the use of TBTU conveniently provided the final compounds in short reaction times and high purity and thus became the method of choice in the course of the project. In addition, compounds 38, 39, and 40 were prepared using commercially available acid chlorides applying common procedures. Compound 41 was accessible via Michael reaction of 7c with 2-(oxetan-3-yliden)acetonitrile (synthesized from (triphenylphosphoranyliden)acetonitrile [22] and oxetan-3-one according to the literature [9]) [23].

Conclusion
We used the moderately potent 9H-pyrimido [4,5-b]indole 2 as a lead structure to design a series of novel amide-based GSK-3β inhibitors. The newly synthesized compounds were evaluated for their biological activity on the targeted kinase in an ADP Glo ™ assay. Compounds 2 and 28 demonstrated an optimized scaffold decoration pattern for GSK-3β inhibition. Remarkably, we observed a strong influence of the stereoconfiguration on the activity of this compound series. The (R)-enantiomers were found to be nanomolar inhibitors of GSK-3β, while a substantial loss in activity was seen with the (S)-configured counterparts. Furthermore, the most potent nanomolar inhibitors (R)-2 and (R)-28 and their binding modes were examined by 1 µs molecular dynamics simulations, rationalizing the SARs of the series. Most importantly, these inhibitors exhibited enhanced stability in the HLM assay as well as a minimal toxicity along with neuroprotective effects in a cellular context. The favorable properties of these compounds motivate additional studies to further assess the biological effects of this class of GSK-3β inhibitors and to finally elucidate the binding mode to the target enzyme.

Molecular Modelling
All in silico work was conducted with Maestro (Schrödinger Release 2019-3/4: Maestro, Schrödinger, LLC, New York, NY, USA, 2019) using OPLS3e force field [24]. Illustrations were made with PyMOL (The PyMOL Molecular Graphics System, Version 2.2.3 Schrödinger, LLC, New York, NY, USA, 2020). First, we prepared the ligands (R)-2 and (R)-28 with LigPrep (Schrödinger, LLC) and then optimized their conformations with the QM Conformer & Tautomer Predictor tool (Schrödinger, LLC), which utilizes Jaguar [25]. In short, this tool optimizes conformations of the compounds with increasing levels of theory, starting with semiempirical method and using density functional theory (DFT) with M06-2X/cc-pVTZ(-f) in the final step. For more detailed description of the QM Conformer & Tautomer Predictor protocol, see the supplementary information in [26]. Next, the ligand of the Protein preparation wizard [27] prepared and energy-minimized GSK-3β structure (PDB ID: 4PTC) [28] was manually replaced with the lowest energy QM-derived structure of (R)-2 or (R)-28. After this, the new complex was prepared for simulations with the Protein preparation wizard [27]. For MD simulations, we used Desmond [29]. The systems were solvated in the cubic box with the minimum distance to the edges of 13 Å from the protein and neutralized with Clions, adding a total of 0.15 M KCl salt. The water molecules were described with the TIP3P water model [30]. The production simulations of 1000 ns were run with NpT ensemble (T = 310 K, Nosé-Hoover method; p = 1.01325 bar, Martyna-Tobias-Klein method) with the default Desmond settings as described previously [5]. Before the actual production run, the default Desmond relaxation protocol was applied for both systems. For the conformation comparison of different enantiomers of 2 (in Figure 2f), (S)-2 was also prepared with the QM Conformer & Tautomer Predictor tool with the same default settings as (R)-2, and they were superimposed by their hinge binding moieties.

General Information
All solvents and reagents were purchased from commercial sources and used without further purification, if not stated otherwise. Organic solvents used for analytical chromatography were generally of HPLC grade.
High performance liquid chromatography (HPLC) was performed on a Hewlett Packard HP1090 series II HPLC system (Hewlett-Packard, Palo Alto, CA, USA) or an Agilent 1100 series HPLC system (Agilent, Santa Clara, CA, USA) equipped with a diode array detector detecting at 230 nm and 254 nm. Method A consisted of elution using mobile phase A (MeOH) and mobile phase B (aqueous 0.01 M KH 2 PO 4 buffer, pH 2.3) in a flow of 1.5 mL/min on a Phenomenex Luna 5 µm C8(2) 100 Å RP column (150 × 4.6 mm) (Phenomenex, Torrance, CA, USA) and the gradient as follows: mobile phase A 40% to 85% during 8 min, mobile phase A 85% constant for 5 min, mobile phase A 85% to 40% during 1 min, mobile phase A 40% constant for 2 min; complete run time 16 min; injection volume 5 µL. Method B consisted of elution using the same mobile phases in a flow of 1.5 mL/min on an XBridge C18 5 µm RP column (150 × 4.6 mm) (Waters, Milford, MA, USA) and the gradient as follows: mobile phase A 45% to 85% during 10 min, mobile phase A constant for 6 min; complete run time 16 min; injection volume 10 µL. The purity of the final compounds was determined at 254 nm and was >95%.

Final compounds (R)-2/(S)-2, (R)-20/(S)-20, and (R)-28
were evaluated for their enantiomeric purity by chiral chromatography, which was performed on an Agilent 1290 Infinity series LC system (Agilent, Santa Clara, CA, USA) consisting of a binary pump, a thermostatted column compartment, an autosampler, and a diode array detector. The system was provided with an ultralow dispersion kit (including a Max-Light ultralow dispersion cartridge flow cell with an inner volume of 0.6 µL, an ultralow dispersion needle seat, and capillaries with 0.075-mm inner diameter from autosampler to column compartment (350 mm length) and from column compartment to DAD (220 mm length)) to minimize extra column volume. The method consisted of elution using the mobile phase (64% n-heptane/36% isopropanol) in a flow of 0.4 mL/min on a Chiralpak IA-U 1.6 µm column (100 × 3.0 mm) (Daicel, Osaka/Tokyo, Japan); injection volume was 5 µL. Samples of the stereoisomers and racemates were prepared by diluting 10 mM aliquotes in dimethyl sulfoxide (DMSO) by the factor 10 with the mobile phase to a total concentration of 1 mM. The purity was determined at 254 nm. The enantiomeric excess (ee) was >98% in all cases.
Thin layer chromatography-electrospray ionization-mass spectrometry coupled analysis (TLC-ESI-MS) was performed on an Advion expression s CMS coupling system (Advion, Ithaca, NY, USA). The parameters of the ESI+ mode were as follows: capillary temperature 250 • C, capillary voltage 180 V, source gas temperature 250 • C, and ESI voltage 3500 V. The parameters of the ESI− mode were as follows: capillary temperature 250 • C, capillary voltage 180 V, source gas temperature 250 • C, and ESI voltage 2500 V. The compounds were eluted from the TLC plate with MeOH. . Spectra were calibrated to residual peaks of the utilized deuterated solvents. Chemical shifts were reported in parts per million (ppm) relative to tetramethylsilane (δ = 0). NMR spectra of compounds with acyl substituents on the piperidine nitrogen frequently showed mixtures of amide bond rotamers resulting in complex reports. The ratio of rotamers was estimated from the respective integrals in the 1 H-NMR spectra.

General Procedures
The appropriate intermediate (3a,b and 3d-l) was suspended in dry tetrahydrofurane (THF). NaH was added, and the mixture was stirred at room temperature (rt) and under N 2 atmosphere for 15-30 min. p-Toluenesulfonyl chloride was added, and the mixture was stirred at rt and under N 2 atmosphere until reaction control by TLC indicated complete consumption of the starting material. The reaction mixture was poured into ice-cold, water and saturated NH 4 Cl solution was added. The precipitate was filtered off, rinsed with fresh demineralized water, and dried over P 2 O 5 in vacuo. The crude product was used in the next step without further purification.
(2) General Procedure B The appropriate intermediate (4a-l) was suspended in dry dimethylformamide (DMF). N,N-diisopropylethylamine (DIPEA) and the appropriate Boc protected secondary amine were added, and the mixture was stirred at 70-80 • C until reaction control by HPLC indicated sufficient consumption of the starting material. After cooling down to rt, the mixture was poured into ice-cold water and saturated NH 4 Cl solution was added. The resulting precipitate was filtered off, rinsed with fresh demineralized water, and dried over P 2 O 5 in vacuo. The crude product was used in the next step without further purification, if not stated otherwise.
(3) General Procedure C A solution of the appropriate intermediate ((R)-5c,d and (S)-5c,d) in dry DMF was stirred in a flame-dried Schlenk tube under Ar atmosphere and ice-cooling. NaH was added, and the mixture was left to stir for 30 min for deprotonation. Cooling was then switched to a MeOH ice bath prior to adding methyl iodide. The mixture was left to warm to 0 • C and then to rt and stirred under Ar atmosphere until HPLC indicated sufficient consumption of the starting material. The mixture was then poured into ice-cold saturated NH 4 Cl solution. The resulting precipitate was filtered off, washed with demineralized, and dried over P 2 O 5 in vacuo. The crude product was used in the next step without further purification, if not stated otherwise.
(4) General Procedure D The appropriate intermediate was dissolved in THF (dry or HPLC grade). KtBuO was added, and the mixture was stirred at rt and under N 2 atmosphere until reaction control by HPLC indicated complete consumption of the starting material. Saturated NH 4 Cl solution and ethyl acetate (EtOAc) were added to the reaction mixture, and phases were separated. The aqueous layer was extracted with EtOAc (2-3×). Combined organic layers were dried over Na 2 SO 4 and concentrated under reduced pressure. The residue was purified by flash column chromatography.
The appropriate intermediate (6a,b, 6d-l, (R)-6c,d, and (S)-6c,d) was stirred in a 17% (v/v) mixture of dry dichloromethane (DCM) and trifluoroacetic acid (TFA) at rt and under N 2 atmosphere until reaction control by HPLC indicated complete consumption of the starting material. The mixture was concentrated under reduced pressure, and saturated NaHCO 3 solution was added to neutralize residual TFA. The mixture was then extracted repeatedly with EtOAc. MeOH was added to improve the solubility of the product in the organic layer. Combined organic layers were washed with saturated NaHCO 3 solution (3×), dried over Na 2 SO 4 , and evaporated to dryness. The crude product was used in the next step without further purification, if not stated otherwise. and DIPEA in dry DCM was added to the activated carboxylic acids and the reaction mixture stirred at rt and under N 2 atmosphere until reaction control by HPLC indicated complete consumption of the starting material. The mixture was diluted with DCM, washed with saturated NaHCO 3 solution (2-3×) and saturated NH 4 Cl solution (2-3×), dried over Na 2 SO 4 , and concentrated under reduced pressure. The residue was purified by flash column chromatography.
The appropriate intermediate (4a-d and 4f,g), 44·HCl, and DIPEA were stirred in dry DMF at 70 • C overnight. After cooling down to rt, NatBuO was added and the mixture was stirred at rt overnight. Saturated NH 4 Cl solution was added, and the mixture was extracted thrice with EtOAc. Combined organic layers were dried over Na 2 SO 4 and concentrated under reduced pressure. The residue was purified by flash column chromatography.

Detailed Procedures
(1) Detailed Procedures for the Preparation of Intermediates 4a-l 4-Chloro-9-tosyl-9H-pyrimido [4,5- The title compound was prepared from 3a (500.0 mg, 2.45 mmol), p-toluenesulfonyl chloride (585.0 mg, 3.07 mmol), and NaH (147.3 mg of a 60% in mineral oil, 3.68 mmol) in dry THF (16 mL) according to general procedure A (reaction time 2 h); 753 mg of a yellow solid was yielded (82% crude yield) and used in the next step without further purification. ESI-MS: The title compound was prepared from 3b (250.0 mg, 2.05 mmol), p-toluenesulfonyl chloride (268.8 mg, 1.41 mmol), and NaH (67.7 mg of a 60% dispersion in mineral oil, 1.69 mmol) in dry THF (7.5 mL) according to general procedure A (reaction time 1.5 h); 412 mg of a yellow solid was yielded (97% crude yield) and used in the next step without further purification. ESI-MS: The title compound was prepared from 3c as described previously [5].
The title compound was prepared from 3d (560.0 mg, 1.98 mmol), p-toluenesulfonyl chloride (472.4 mg, 2.48 mmol), and sodium hydride (118.9 mg of a 60% dispersion in mineral oil, 2.97 mmol) in dry THF (15 mL) according to general procedure A (reaction time 20 min); 834 mg of a yellow solid was yielded (96% crude yield) and used in the next step without further purification. ESI-MS: (m/z) The title compound was prepared from 3e (421.0 mg, 1.35 mmol), p-toluenesulfonyl chloride (322.5 mg, 1.69 mmol), and sodium hydride (81.2 mg of a 60% dispersion in mineral oil, 2.03 mmol) in dry THF (10 mL) according to general procedure A (reaction time 1 h); 606 mg of a dark yellow solid was yielded (93% crude yield) and used in the next step without further purification. ESI-MS: (m/z) The title compound was prepared from 3f (710.0 mg, 3.03 mmol) and NaH (182.3 mg of a 60% dispersion in mineral oil, 4.56 mmol) in dry THF (10 mL) according to general procedure A (reaction time 30 min). p-Toluenesulfonyl chloride (695.1 mg, 3.65 mmol) was added as solution in THF (2 mL); 1.15 g was yielded (98% yield) and used in the next step without further purification.
NaH (176.7 mg of a 60% dispersion in mineral oil, 4.42 mmol) was added portion-wise to an ice-cooled stirring suspension of 3g (800.0 mg, 2.95 mmol) in dry THF (10 mL). The mixture was stirred under ice-cooling for 30 min. A solution of p-toluenesulfonyl chloride (673.8 mg, 3.53 mmol) in dry THF (2 mL) was drop-added, while the mixture was left to warm to rt and stirring continued at rt for 30 min. The mixture was poured into saturated NH 4 Cl solution (100 mL). The resulting precipitate was filtered off, washed with demineralized water, and dried over P 2 O 5 in vacuo; 1. The title compound was prepared from 3h (550.0 mg, 2.31 mmol), p-toluenesulfonyl chloride (550.2 mg, 2.89 mmol). and NaH (138.6 mg of a 60% dispersion in mineral oil, 3.47 mmol) in dry THF (18 mL) according to general procedure A (reaction time 40 min); 890 mg of a yellow solid was yielded (98% crude yield) and used in the next step without further purification. ESI-MS: The title compound was prepared from 3i (740.0 mg, 2.62 mmol), p-toluenesulfonyl chloride (624.2 mg, 3.27 mmol), and NaH (157.2 mg of a 60% dispersion in mineral oil, 3.93 mmol) in dry THF (33 mL) according to general procedure A (reaction time 1.5 h); 1.1 g was yielded (96% crude yield) and used in the next step without further purification. ESI-MS: The title compound was prepared from 3j (480.0 mg, 2.05 mmol), p-toluenesulfonyl chloride (489.6 mg, 2.57 mmol), and sodium hydride (123.3 mg of a 60% dispersion in mineral oil, 3.08 mmol) in dry THF (16 mL) according to general procedure A in a (reaction time 30 min); 750 mg of a brown solid was yielded (94% crude yield) and used in the next step without further purification. ESI-MS: 3k (600.0 mg, 2.52 mmol) was suspended in dry THF (20 mL), and NaH (151.2 mg of a 60% dispersion in mineral oil, 3.78 mmol) was added. The mixture was stirred at rt and under N 2 atmosphere for 15 min. p-Toluenesulfonyl chloride (600.6 mg, 3.15 mmol) was added and the mixture stirred at rt and under N 2 atmosphere for 40 min. Saturated NH 4 Cl solution (50 mL), EtOAc (30 mL), and some MeOH were added, and phases were separated. The aqueous layer was extracted with DCM (4 × 30 mL). Combined organic layers were dried over Na 2 SO 4 . Volatiles were removed under reduced pressure to yield 1.1 g of a brown solid (>100% crude yield), which contained excessive p-toluenesulfonyl chloride. The crude product was used in the next step without further purification.
The title compound was prepared from 3l (470.0 mg, 1.86 mmol), p-toluenesulfonyl chloride (444.3 mg, 2.33 mmol), and NaH (111.9 mg of a 60% dispersion in mineral oil, 2.80 mmol) in dry THF (16 mL) according to general procedure A (reaction time 20 min); 759 mg of a beige solid was yielded (100% crude yield) and used in the next step without further purification.
The crude material obtained from the first step was reacted with KtBuO (639.

7-Bromo-N-methyl-N-(piperidin-3-yl)-9H-pyrimido[4,5-b]indol-4-amine (7d)
6d (425.0 mg, 0.92 mmol) was stirred in dry DCM (9 mL) and TFA (1.5 mL) at rt for 1 h. The mixture was concentrated under reduced pressure. Residual TFA was neutralized by addition of saturated NaHCO 3 solution, resulting in a precipitate which was filtered off and washed with saturated NaHCO 3 solution and demineralized water and was then dried over P 2 O 5 in vacuo; 285 mg of an off-white solid was yielded (86% crude yield) and used in the next step without further purification. 1  6e (240.0 mg, 0.473 mmol) was suspended in dry DCM (7 mL), and TFA was added (1.5 mL). The mixture was stirred at rt for 30 min and then concentrated under reduced pressure. Residual TFA was neutralized by addition of saturated NaHCO 3 solution (30 mL), which resulted in a precipitate. The precipitate was filtered off, washed with saturated NaHCO 3 solution and demineralized water, and dried over P 2 O 5 in vacuo; 199 mg of a beige solid was yielded (>100% crude yield) and used in the next step without further purification. 1