Optimization of 3-Cyano-7-cyclopropylamino-pyrazolo[1,5-a]pyrimidines Toward the Development of an In Vivo Chemical Probe for CSNK2A

3-cyano-7-cyclopropylamino-pyrazolo[1,5-a]pyrimidines, including the chemical probe SGC-CK2–1, are potent and selective inhibitors of CSNK2A in cells but have limited utility in animal models due to their poor pharmacokinetic properties. While developing analogs with reduced intrinsic clearance and the potential for sustained exposure in mice, we discovered that Phase II conjugation by GST enzymes was a major metabolic transformation in hepatocytes. A protocol for co-dosing with ethacrynic acid, a covalent reversible GST inhibitor, was developed to improve the exposure of analog 2h in mice. A double co-dosing protocol, using a combination of ethacrynic acid and irreversible P450 inhibitor 1-aminobenzotriazole increased the blood level of 2h by 40-fold at a 5 h time point.


Introduction
The appearance of a novel coronavirus SARS-CoV-2 and the resulting pandemic have highlighted the need for effective treatments against COVID-19. 1 Host-directed therapies that target cellular pathways required for virus replication have emerged as a promising approach to combat viral infections. 2 Protein kinases have been proposed as potential targets for development of antiviral drugs. 3 Among these kinases, Casein Kinase 2α (CSNK2A) has been shown to play a role in the replication of β-coronaviruses (β-CoV), including SARS-CoV-2. 4-6 . Recent studies have demonstrated the efficacy of CSNK2A inhibitors in reducing the replication of β-CoV in vitro. 6 However, there has yet to be a demonstration of anti-COVID-19 efficacy in vivo.
3-cyano-7-cyclopropylamino-pyrazolo[1,5-a]pyrimidine (PZP) is a promising chemotype of CSNK2A inhibitor that has demonstrated potent activity in vitro. 7 PZPs possess good selectivity for CSNK2A over other kinases and display unique chemical features that contribute to their potent inhibitory activity ( Figure 1). Specifically, X-ray crystallography studies ( Figure 1A) revealed that the 4'-methyl substituent of the chemical probe SGC-CK2-1 (1a) enforces an s-cis (E)-conformation of the propionamide, allowing it to coordinate with a bound water molecule, which also interacts with the cyano group of the PZP core ( Figure 1B). Notably, the alkyl group A. X-ray cocrystal structure of SGC-CK2-1 (1a) bound to CSNK2A1 (PDB: 6Z83). Image created in PyMOL 2.4.0 with the CSNK2A1 protein shown as a green ribbon and 1a shown as sticks colored by atom type. Bound water molecules are shown as red spheres. Intermolecular hydrogen bonds are shown as yellow dashed lines. B. Illustration of the key interactions observed in the X-ray structure. The 7-cyclopropylamine and N-1 of the PZP form H-bond interactions with V116 in the hinge region. The 4'-substiuent (Me for 1a) causes the amide to adopt an s-cis conformation that allows the carbonyl group to interact with K68 and a bound water molecule. The ethyl group of the amide is placed in a pocket that increases selectivity over other kinases. The cyano group of the PZP also interacts with the bound water molecule, which is further coordinated through interaction with the backbone NH of D175. of the propionamide contributes to the exquisite selectivity of the PZP-based CSNK2A inhibitors over other kinases. 8 However, despite their promising in vitro activity, PZPs have shown limited bioavailability in rodents by oral dosing due to their low to moderate aqueous solubility, high first-pass metabolism, and rapid clearance. 9,10 In this paper, we report our initial studies to identify a PZP-based CSNK2A inhibitor with potent antiviral activity and high sustained blood levels for use in mouse-adapted model of COVID-19 developed for preclinical evaluation of potential drug therapies. 11,12 We discovered that the rapid clearance of PZPs in mice was due to pathways of both Phase I and Phase II metabolism. Co-dosing with a glutathione S-transferase (GST) enzyme inhibitor and an irreversible cytochrome P450 inhibitor greatly reduced intrinsic clearance of a PZP analog in mice and may be a viable approach to increase in vivo exposure for preclinical pharmacology studies.

Results
Antiviral PZPs with Reduced Phase I Metabolism. All compounds were screened for CSNK2A1 and CSNK2A2 activity using a NanoBRET assay in HEK293 cells and for inhibition of mouse β-CoV replication using an MHV-nLuc assay in DBT cells. 6 Kinetic solubility was determined using a miniaturized shake flask method. The stability of compounds to Phase I metabolism was determined by incubation with mouse liver microsome for 30 min. SGC-CK2-1 (1a) demonstrated low aqueous solubility and rapid metabolism in primary mouse liver microsomes (MLM), with only 40% of the parent compound remaining after 30 min (Table 1).
Since hepatic metabolism of 1a would likely lead to rapid in vivo clearance, we decided to focus In-cell target engagement by NanoBRET assay, n = 1. b Inhibition of MHV-nLuc replication, n = 3 with SE ± 0.1. c Values from ChemDraw Professional v16.0. d Miniaturized shake flask using CLND detection, n = 1; e Metabolism of compound in primary mouse liver microsomes (MLM) calculated as 100 -(% remaining), n = 1. f Compound previously reported in reference 6.
initially on increasing its stability in MLM. SMARTCyp, 13,14 a web-based predictor of metabolic hot spots, analyzes sites of metabolism using a ligand-based approach from a library of molecular fragments that can be applied across multiple species and isoforms of CYPs. 15 SMARTCyp suggested that that the cyclopropylamine and the 4'-methyl group might be sites of P450 oxidation ( Figure S1). As modification of the cyclopropylamine, which forms part of the hinge-binding motif ( Figure 1B), resulted in a large decrease in CSNK2A potency 6,8 we opted to study analogs with modification of the 4'-methyl group in order to lower logD, improve aqueous solubility, and possibly reduce P450 metabolism. 16  To build on these results we designed new analogs with 3-aminomethyl piperidine substituents at the 4'-position to combine the features of both the cyclic and linear diamines.
The synthesis employed the methods previously developed for analogs 1b-g (Scheme 1).
Addition of the (R) or (S)-enantiomer of the N-Boc-protected 3-aminomethyl piperidine to the acyl-2-fluoro-5-nitroaniline (x, R 1 = Et) followed by reduction of the nitro group yielded the aniline intermediate (y, R 1 = Et). Palladium-catalyzed Buchwald-Hartwig cross coupling with the 5-chloro-pyrazolo[1,5-a]pyrimidine core followed by N-Boc deprotection yielded the 3R-isomer (1h) and the 3S-isomer (1i). Both isomers, 1h and 1i, demonstrated good aqueous solubility and low metabolism in mouse liver microsomes compared to 1a ( Table 1). The 3S-isomer (1i) To further improve the potency, we synthesized 4'-substituted analogs in the acetamide series of PZPs ( Table 2). The compounds were synthesized by the established route from the acyl-2-fluoro-5-nitroaniline (x, R 1 = Me) and the appropriate secondary amine for the R 2substituent (Scheme 1). For those R 2 -substituents with chiral centers, the enantiomerically-pure amine was utilized as a building block. The 4'-methyl analog (2a) had shown sub-nanomolar activity on CSNK2A and good activity in the antiviral assay. 6 2a had a lower cLogP and slightly improved solubility and metabolic stability compared to the corresponding propionamide (1a) ( Table 2). Synthesis of the 3R-isomer (2b) and 3S-isomer (2c) of the 3-aminomethyl piperidine gave analogs with 10-fold improved potency in the CSNK2A and the antiviral assays while maintaining good metabolic stability. The 3S-isomer (2c) was again the more potent of the two enantiomers. Additional 4'-substituted analogs were synthesized in the acetamide series. The 3R-isomer (2d) and 3S-isomer (2e) of the 3-aminomethyl pyrrolidine also combined potency on CSNK2A and antiviral activity with good metabolic stability. Switching the orientation of the 3aminomethyl pyrrolidine gave analog 2f with similar potency but slightly poorer metabolic stability. The piperazine (2g) was less potent on CSNK2A and in the antiviral assay, but had good metabolic stability, as had been seen in the corresponding propionamide (1g). We had previously prepared and tested morpholine (2h), 6 which had improved antiviral activity compared to 2g. Although the morpholine (2h) showed low aqueous solubility its metabolic stability was much better than the parent analog 2a ( Table 2). Introduction of a hydroxy substituent in 2i maintained activity and improved aqueous solubility but decreased metabolic stability. The oxo-analogs 2j-n had lower activity in the antiviral assay and poor metabolic stability, despite some analogs (e.g. 2l) having low cLogP and good solubility.   Metabolite ID Study of 1i. To identify the metabolic pathway responsible for the rapid clearance in mouse hepatocytes, we performed a metabolite ID study using the 3S-aminomethyl piperidine (1i). Incubation of 1i for 4 h in primary cultures of mouse hepatocytes resulted in the appearance of 12 metabolites by ultrahigh performance liquid chromatography ( Figure 2A).
Analysis of the metabolites by mass spectroscopy and quantification by UV absorbance enabled the assignment of the primary pathways of metabolic transformation (Table 4, Figures 2B and   S2). Six of the metabolites (M1-4, M10-11) resulted from GSH conjugation of the PZPs ( Figure   2B), a Phase II metabolic transformation that could only occur in whole hepatocytes that contain the active GST enzymes. GSH conjugation was observed in the major fraction of the total metabolites as quantified by UV absorbance (Table 5) and is likely to be an important pathway of clearance of the PZP 1i ( Figure 2B). In addition to GSH conjugation, demethylation of the secondary amine on the piperidine and dealkylation of the 7-amino group on the PZP were found among the metabolites (Table 4 and Figure 2B). Notably, dealkylation was often found in combination with GSH conjugation. P450 oxidation of electron-rich aromatic rings to an epoxide intermediate is well known activation step in GST-catalyzed GSH conjugation, 18 making the electron-rich amino-substituted phenyl ring the most likely site of Phase II metabolism ( Figure   2C). Although we initially considered the nitrile as a potential site of conjugation, since it can undergo a Pinner reaction with free GSH, 19 there is no evidence that the reaction is catalyzed by GST enzymes. 20,21 The metabolite ID study implicated both cytochrome P450 and GST enzymes as likely to be responsible for the rapid clearance of the PZPs in mouse hepatocytes through a combination of demethylation, dealkylation, and GSH conjugation ( Figure 2C).

Hepatocyte Metabolism of Antiviral PZPs. Eight acetamide PZPs (2a-f,h-i) with IC 50
≤100 nM in the antiviral assay were selected for time course metabolic stability profiling in mouse and human and primary hepatocytes ( These results demonstrated that the metabolism of the PZPs in primary hepatocytes was  Figure S2). Molecular ions shown in parentheses. Doubly charged ions are indicated as z=2. The relative percent contribution of each metabolite taken from Table 5. C. Summary of the major sites of metabolism of 1i in mouse primary hepatocytes.
a predominantly a species-specific issue, with intrinsic clearance always much higher in mouse than human hepatocytes. Although this was a promising result for the future development of PZPs as therapeutic drugs for use in humans, it still limited their utility as pharmacological tools for studies in mice.

Inhibition of GSH Conjugation in Hepatocytes.
Given our long-term objective to identify a potent CSNK2A inhibitor for use in an in vivo mouse COVID-19 efficacy model in vivo to validate the host kinase as a potential antiviral drug target, we opted to explore the effect of inhibition of the Phase II GST enzyme on intrinsic clearance. Ethacrynic acid (EA) is an FDA-approved loop diuretic that is also a potent covalent, but reversible, inhibitor of GST enzymes 22 . EA has demonstrated potent inhibition of GST enzymes in perfused rat liver and human cancer cell lines, 22 but although it has been used in human clinical studies to block GSH conjugation 23 there are only a few reports of its use in rodents as a GST inhibitor. [24][25][26] To determine if EA could improve the metabolic stability of a PZP in primary mouse hepatocytes we tested the effect of co-dosing on the stability of 2h, which was one of the analogs with the highest intrinsic clearance ( Table 6). The intrinsic clearance of 2h in primary mouse hepatocytes was determined by measuring the decrease in the level of the parent compound over 2 h ( Figure 3). Parallel sets of incubations were performed in the presence of increasing doses of EA from 0-400 µM. In the absence of EA intrinsic clearance was >250 mL/min/kg. However, in the presence of doses of EA from 10-50 µM the intrinsic clearance was decreased by >50% ( Figure 3). At doses of EA above 100 µM, the intrinsic clearance of 2h was further decreased to <50 mL/min/kg. The co-dosing mouse experiments in primary mouse hepatocytes demonstrated that EA was effective at increasing the metabolic stability of PZP 2h, presumably by blocking its GSH conjugation.  (Table S1). Since GSH conjugation of electron-rich aromatic groups requires a prior cytochrome P450 oxidation, we also evaluated the in vivo pharmacokinetics of 2h in mice pretreated with 1-ABT for 2 h. 1-ABT was able to block the metabolism of 2h in primary mouse hepatocytes ( Figure S2). A decrease in the clearance of 2h was observed in mice pretreated with 1-ABT at 100 mg/kg p.o. with an increase in t 1/2 and AUC of 1.6-fold and 3.5-fold, respectively, compared to dosing of 2h alone. The larger increase in AUC with 1-ABT compared to EA may be due to both inhibition of Phase I P450 metabolism and indirect inhibition of Phase II GSH conjugation.
Importantly, the results demonstrate that it is possible to increase the circulating levels of the PZP CSNK2A inhibitor 2h by co-dosing with either a GST inhibitor or a P450 inhibitor. Finally, to

Discussion and Conclusions
The 3-cyano-7-cyclopropylamino-pyrazolo[1,5-a]pyrimidine (PZP) is the most potent and selective chemotype of ATP-competitive CSNK2A inhibitors, 7 with fewer kinase off-targets than silmitasertib and related compounds. 28 SGC-CK2-1 (1a) 8 has been established as high-quality chemical probe for studying CSNK2 signaling in phenotypic assay and human-derived primary cells. [29][30][31] However, the use of PZP-based CSNK2A inhibitors as chemical probes in rodent pharmacology models has been limited by their generally poor pharmacokinetic properties.
Although Dowling and coworkers reported that 2a had modest oral bioavailability in rats (F = 25%, C max = ~300 nM, t 1/2 = 2.6 h), 2a showed in vivo activity only at short time points after dosing. 9 After extensive work on the series, Dowling concluded that an analog with an ethylenediamine substituent at the 4'-position could be used as a pharmacological tool by i.v. or i.p. dosing, despite its rapid clearance in rats. 10 For the studies reported herein, we decided to focus on development of an in vivo pharmacological probe that would maintain high sustained exposure in mice following i.p. dosing, since this route of administration was preferred in many academic laboratories for preclinical target validation studies. 32 Initially, we examined the potential of the 4'-substitued analogs of SGC-CK2-1 (1a) to improve solubility and reduce its rapid metabolism in mouse liver microsomes. Testing several of the previously prepared analogs of 1a 6,8 confirmed prior observations 10 of reduced metabolism of linear and cyclic diamine analogs with cLogP <0.5 (Table 1). After synthesizing many new 4'-substituted PZPs in the acetamide (1) and propionamide (2) series, the chiral 3methylaminopipiperides 1h-I, 2b-c and 3-methylaminopyrollidines 2d-e were found to have the best balance of reduced MLM metabolism and good cellular potency in the antiviral assay.
However, despite their improved stability to Phase I metabolism 1i, 2c, and 2e showed rapid clearance in mice following i.v. or i.p. administration (Table 3).
After discovering that 1i, although stable in MLM, was rapidly metabolized in mouse primary hepatocytes, we performed a metabolic ID study that identified Phase II GSH conjugation as a major metabolic transformation (Table 4 and Figure 2). This result suggested that Phase II metabolism by GST enzymes was a previously underappreciated metabolic liability of PZPs containing electron-rich di-and tri-aniline rings. Unfortunately, all potent antiviral PZPs showed rapid metabolism in mouse primary hepatocytes (Table 6). Although the metabolic instability may be species-specific, since it was not observed in human primary hepatocytes, the rapid clearance in mouse primary hepatocytes predicted that none of the potent antiviral PZP analogs would be suitable for use in the mouse-adapted COVID-19 model.
To explore ways to reduce the Phase II metabolism of a PZP analog, we investigated codosing with EA, a well characterized broad-spectrum covalent, but reversible, inhibitor of rodent and human GSTs. 22 The studies were performed with the potent antiviral morpholine analog 2h, which exhibited very high clearance in primary mouse hepatocytes. Co-dosing of EA with 2h in primary mouse hepatocytes demonstrated that metabolism could be reduced in a dose dependent manner, with >90% reduction in intrinsic clearance at EA concentrations >100 µM ( Figure 3).
Translating the use of EA as a GST inhibitor to an in vivo setting in mice was not straightforward. EA is an inhibitor of NKCC2, a Na-K-Cl transporter, and is approved for use in humans as a loop diuretic for treatment of high blood pressure and edema. 33 (Table S1). Although the reason for the toxicity of EA by i.p. administration is unclear, the rapid onset suggests that it may be unrelated to inhibition liver metabolism enzymes. Nevertheless, we observed a reduction in the plasma clearance of 2h at both the 10 and 30 mg/kg i.p. co-doses of EA.
Furthermore, the clearance of 2h was also reduced by a 100 mg/kg p.o. dose of the irreversible cytochrome P450 inhibitor 1-ABT, which would be expected to block direct Phase I and indirect Phase II metabolism. The most effective strategy to decrease the clearance of 2h was the codosing protocol, using a combination of 30 mg/kg i.p. EA and 100 mg/kg p.o 1-ABT. With this approach the blood level of 2h was boosted ~40-fold at the 5 h time point.
In conclusion, our findings demonstrate that Phase II metabolism limits the in vivo exposure of PZP-based CSNK2A inhibitors in mice, which severely limits their utility as pharmacological probes despite their remarkable potency and selectivity in cells. Although the use of EA as a GST inhibitor provides an effective method to reduce the intrinsic clearance of the current molecules in mice, we are continuing to optimize the series to identify analogs with reduced potential for Phase II metabolism. Our goal is to find in vivo chemical probes for CSNK2A that can be dosed in mice without the additional complication of co-dosing with multiple inhibitors of drug metabolizing enzymes.

Experimental Section.
NanoBRET Assay. Assays were run with a modified version of the previously published protocols. 6   . To the product in CH 2 Cl 2 (5 mL) was added TFA (7 mL), and the mixture was stirred at 25°C for 2 h.
The mixture was degassed and purged with N 2 and heated in a microwave reactor at 130°C for 0.5 h. The reaction mixture was concentrated in vacuo and the residue was purified by flash silica gel chromatography (eluent of 0-3%, MeOH/CH 2 Cl 2 ). To the product in CH 2 Cl 2 (5 mL) was added TFA (7 mL), and the mixture was stirred at 25°C for 2 h. The reaction mixture was concentrated in vacuo and the residue was purified by prep-HPLC  [Heptane-EtOH (0.1% NH 3 /H 2 O)]; B%: 10-50%, 5min) to give the product as a yellow solid. To the product in CH 2 Cl 2 (5 mL) was added TFA (7 mL), and the mixture was stirred at 25°C for 2 h. The reaction mixture was concentrated in vacuo and the residue was purified by prep-HPLC    chromatography (eluent of 0-2%, MeOH/CH 2 Cl 2 ) to give the product as a brown solid. To the product in CH 2 Cl 2 (5 mL) was added TFA (7 mL), and the mixture was stirred at 25°C for 2 h.
The reaction mixture was concentrated in vacuo and the residue was purified by prep-HPLC  Compound 2l (18.2 mg, 9.71% yield) was obtained as a white solid. 1