Pyrazole-Based Acid Ceramidase Inhibitors: Design, Synthesis, and Structure–Activity Relationships

Abstract Acid ceramidase (AC) is a lysosomal cysteine amidase responsible for the cleavage of ceramide into sphingosine, which is then phosphorylated to sphingosine 1-phosphate. AC regulates the intracellular levels of ceramide and sphingosine, and AC inhibition may be useful in the treatment of disorders, such as cancer, in which ceramide-mediated signaling may be dysfunctional. Despite their potential experimental and therapeutic value, the number of available small-molecule inhibitors of AC activity remains limited. In the present study is described the discovery of a class of potent pyrazole carboxamide-based AC inhibitors, which were identified using the atomic property field (APF) approach and developed through systematic SAR investigations and in vitro pharmacological characterization. The best compound of this series inhibits AC with nanomolar potency and causes ceramide accumulation and sphingosine depletion in intact G361 proliferative melanoma cells. By expanding the current armamentarium of AC inhibitors, these results should facilitate future efforts to unravel the biology of AC and the therapeutic potential of its inhibition.

The sphingolipids are ubiquitous lipid constituents of eukaryotic cell membranes. 1 In addition to this essential structural function, sphingolipids are also thought to play multiple signaling roles and to be involved in the pathogenesis of cancer, 2 inflammation, 3 and neuropathic pain. 4 In particular, two key sphingolipids -ceramides and sphingosine 1-phosphate (S1P) -have been shown to regulate cell fate in opposite manner: ceramides by mediating cellular senescence 5 and apoptosis, 6 and S1P by promoting cell survival and proliferation. 7 The ceramides are a highly heterogeneous family of Nacylated sphingosines with long-chain fatty acids. 8 They hold a central role in sphingolipid metabolism and are the metabolic precursors of S1P; conversion of one into the other has been proposed to serve as a 'rheostat' in the regulation of cell fate. 9 The lysosomal lipid amidase, acid ceramidase (AC, encoded by the ASAH-1 gene) is a key element of this rheostat. It catalyzes the hydrolysis of ceramides into fatty acid and sphingosine, which is then converted into S1P by sphingosine kinase (SK) (Scheme 1). 10 Daniele Piomelli is the holder of Louise Turner Arnold Chair in Neurosciences and Professor of Anatomy and Neurobiology, Pharmacology and Biological Chemistry at the University of California, Irvine. In addition, he is the founding director of the unit of drug discovery and development (D3) at the Italian Institute of Technology in Genoa, Italy. Dr. Chiara Pagliuca earned her Ph.D. in organic chemistry from Florence University, where she worked on the synthesis and development of new molecules for targeted tumor diagnosis and therapy. She has most recently completed a series of post-doc assignments leading drug development projects for international research institutions, including the Italian Institute of Technology. Her scientific investigation has ranged from hit identification, hit-to-lead, and lead optimization programs to the preclinical development of new molecular entities. She has co-authored 16 publications and holds 4 patent applications. Dr. Pagliuca is currently working as a Scientist at a global player of the chemical industry. Dr. Luca Goldoni received his Ph.D. in drug discovery from the University of Genova (Italy) in 2016, focused on metabolomics by NMR spectroscopy, under the guidance of Professor Piomelli. He gained 12 years of experience as an NMR technician in several chemical and pharmaceutical companies. In 2011 he joined the Italian Institute of Technology (IIT) as NMR senior specialist where he is currently responsible for the structure elucidation of new chemical entities, and their purity evaluation. From 2013 his research interest also included metabolomics by NMR spectroscopy.
Dr. Natalia Realini got her Master's degree in biology applied to medical science and her Ph.D. in medical pharmacology, toxicology and chemotherapy. In 2010, she joined Professor Piomelli's group at the University of California, Irvine and she's currently a senior postdoctoral associate at the Italian Institute of Technology, Genova, Italy. Her research interests include sphingolipid biology, mainly focused on the biological role of acid ceramidase in health and diseases.
Since inhibition of AC increases the cellular levels of ceramides and concomitantly decreases those of sphingosine and S1P, it represents a potential strategy to address disorders associated with abnormal levels of these bioactive lipids -including certain types of cancer and inflammation.
Structural modifications of these molecules have led to the discovery of significantly more potent compounds, such as DP24a (4) 14 (IC 50 = 1.3 μM on human recombinant AC) and SABRAC (5), 15 which inhibit AC activity in intact cells and cell lysates with nanomolar potency.
Among the AC inhibitors that are not structurally related to ceramide, Ceranib-1 and 2 (6,7) 16 show micromolar potency towards intracellular ceramidases, inhibit cell proliferation and induce cell death in certain types of cancer, alone or in combination with paclitaxel. In addition, cystatin SA (encoded in humans by the CST2 gene) is a physiological inhibitor of AC (K i ≈ 5 μM), which binds the enzymes and creates an allosteric interference with the active site domains. 17 Previous work in our laboratory has identified two new classes of potent small-molecule inhibitors of AC ( Figure 1). The first includes substituted 2,4-dioxopyrimidine-N1-carboxamides such as carmofur (8), which inhibits AC activity with nanomolar potency. 18 Members of this class act synergistically with standard anti-neoplastic drugs to stop proliferation of various types of cancer cells, suggesting a potential clinical use as chemosensitizers. 18 Recently, we have shown that N3-carbamoyl uracil derivatives may be useful in normalizing ceramide levels and concomitantly sensitize proliferative melanoma cells to the cytotoxic actions of various anti-tumoral agents. These results are suggestive of the potential utility of AC blockade as chemosensitizing strategy in proliferative melanoma. 19 Another class of AC inhibitors identified by our laboratory includes benzoxazolone-carboxamide derivatives, which display improved chemical and metabolic stability, compared to carmofur and its congeners. 20 Focused structure-activity relationship (SAR) studies around the benzoxazolone-carboxamide scaffold have led to the identification of 6-(4-fluorophenyl)-2-oxo-N-(4-phenylbutyl)-1,3-benzoxazole-3-carboxamide (ARN14974, 9, IC 50 = 0.079 μM) as a highly potent systemically active inhibitor of intracellular AC activity. 20 Expanded SAR investigations confirmed the potential for benzoxazolone-carboxamide derivatives to provide stable and systemically active inhibitors of AC. 21

Feature Syn thesis
Despite these advances, the number of molecular scaffolds that have demonstrated utility as AC inhibitors remains very limited. Yet, broadening chemical diversity in this class of molecules is essential not only to expand our ability to study the roles of ceramide and S1P in living animals, but also to support possible preclinical and clinical development efforts.
To identify novel molecular scaffolds for AC inhibition, in the present study, we utilized the atomic property field (APF) method 22 using carmofur (8) as a template. Five distinct heterocyclic scaffolds 10-13 and 14a were selected, which were functionalized as N-carboxamides to yield a small set of 12 compounds: 15-22, 23a-24a, 25a-26a (Figure 2).
The ability of these molecules to inhibit AC activity was examined using lysates of HEK293 cells that overexpressed human AC (h-AC). We obtained promising results with the pyrazole derivatives 23a-26a (Table 2, vide infra), and selected this chemical class for further expansion. Our initial structure-activity relationship (SAR) studies were focused on compounds 23a and 24a (Scheme 2), investigating the effects of various substituents at positions 3 and 5 of the pyrazole ring as well as the role of the acyl function and side chain length. When two isomeric pyrazole N-carboxamides were obtained, the regiochemistry-based SARs were also evaluated. Finally, the potent inhibitor 24n was selected for in vitro pharmacology studies and its ability was tested to interfere with AC activity in intact G361 cells, a human melanoma cell line characterized by a proliferative phenotype.
Chemistry In analogy to hit compounds 8 and 9, we coupled the selected heterocyclic scaffolds 10-13 and 14a with hexyl and phenylbutyl isocyanate to obtain the corresponding N-car-

Feature Syn thesis
boxamide derivatives 15-22 and 23a-26a, in good yields (Scheme S1, Supporting Information and Table 2). The acylation of asymmetrically substituted pyrazolol 14a takes place competitively at both nitrogen atoms of the pyrazole ring giving a mixture of two regioisomers 23a-24a and 25a-26a, in overall good yield and in a ratio of 8:92 and 7:93, respectively. The regioisomers were separated by column chromatography, and the structures assigned by spectroscopic analysis. No traces of O-acylated products were found.
Next, the syntheses of a series of unsubstituted and differently substituted pyrazole N-carboxamide derivatives are outlined in Scheme 3 and Table 1. A small set of compounds 30-35, bearing N-carboxamide alkyl chains of different length, was prepared as shown in Scheme 3. The unsubstituted pyrazole 14b was coupled to various alkyl isocyanates in pyridine at room temperature, in the presence of a catalytic amount of DMAP, to give derivatives 30-33. For compounds 34 and 35, 14b was first activated with triphosgene followed by in situ quenching with n-decylamine and N-methylhexylamine, respectively (Scheme 3). Then, we focused on the synthesis of mono-and disubstituted pyrazole N-carboxamides. Table 1 shows the reaction conditions adopted for the coupling of a series of substituted pyrazoles 14c-t with hexyl isocyanate to afford compounds 23c-o, the corresponding regioisomers 24c-f, 24h, 24k-o, and derivatives 36-40 as single isomers, in moderate to good yields.
For the asymmetrically substituted pyrazoles 14c-o, a mixture of two regioisomers 23 and 24 was obtained in all cases, with the exception of 3-phenyl-, 3-trifluoromethyland 3-nitro-1H-pyrazoles 14g,i,j, which reacted forming only one isomer 23g,i,j (Table 1). The ratio between the two N-acyl compounds 23 and 24 is found to be affected by a combination of electronic and steric factors of the substituents in position 3 and 5 of the pyrazole ring. 23 Indeed, in the presence of an electron-donating group, such as methyl, both pyrazole nitrogen atoms undergo acylation giving a mixture of the two possible regioisomers, in which the predominant one is the less sterically hindered compound, namely 23d and 24d (Table 1). In contrast, the effect of an electron-withdrawing group, such as trifluoromethyl, decreases the reactivity of the neighboring nitrogen atom leading to the formation of only one regioisomer, which also corresponds to the less bulky derivative, namely 23i (Table 1). In the cases of pyrazoles having substituents comparable in size but with different electronic properties, such as methyl and trifluoromethyl groups, both nitrogen atoms participate in the acylation reaction, which also occurs at the nitrogen atom neighboring the electronwithdrawing moiety, as for 23n and 24n (Table 1).
We observed that acylation of one nitrogen atom of the pyrazole system implies (i) a marked 1 H-downfield shift of the adjacent substituent due to anisotropic effect exerted by the carbonyl group of the urea moiety, 24 (ii) a 15 N-high-

Feature Syn thesis
field shift of the acylated nitrogen atom (≈ 220 ppm), while the non-acylated nitrogen resonates at lower field (≈ 290 ppm), 25 and (iii) a 13 C-downfield shift of the carbon atom distal to the urea moiety ( Figure 3). Cross-correlation of these evidences, consistently found throughout regioisomer series 23 and 24, allowed building a robust spectroscopic method for structural elucidation of these compounds. In those cases when one regioisomer is formed, structural assignment was based on the comparison between spectroscopic data collected for the single isomer and the corresponding non-acylated pyrazole (see the Supporting Information for complete structural assignment of representative substituted pyrazole-N-carboxamides).
Pyrazoles that were not commercially available were prepared as shown in Schemes 4 and 5. 3-Methoxy-1H-pyrazole (14f) and 3-methoxy-5-methyl-1H-pyrazole (14l) were synthesized through a three-step sequence starting from the corresponding hydroxyl pyrazoles 14c and 14a, respectively (Scheme 4).   These compounds were first reacted with 1 equivalent of acetic anhydride to afford N1-acetyl pyrazoles 41 and 42 as single isomers. Methylation with iodomethane under basic conditions proceeded smoothly to give compounds 43 and 44, which upon cleavage of the acetyl group delivered the desired methoxypyrazoles 14f and 14l in overall good yield. Finally, a Knorr-type reaction was adopted to synthesize pyrazoles 14o, 14q, and 14r. 26 Cyclocondensation of the proper 1,3-dicarbonyl compounds 46-48 with hydrazine led to the desired pyrazoles in good to excellent yields (Scheme 5). Scheme 5 Synthesis of 3,5-disubstituted pyrazoles 14o, 14q and 14r. 1,3-Diketone 48 was obtained in situ from the corresponding commercially available ketone and acid chloride.

Results and Discussion
The primary objective of the present study was to discover novel chemical classes of small-molecule inhibitors of lysosomal AC activity. We utilized the computer-assisted APF method -starting from carmofur (8) as a template -to identify novel scaffolds that, while structurally different, match the pharmacophore fingerprint of 8. From ZINC, 27 we retrieved a library of over 10 million rule-of-five compliant compounds ( Figure 2). 28 To guarantee synthetic accessibility, the selected dataset encompassed only commercially available derivatives. First, the library was pre-filtered by means of a simple two-dimensional query, retrieving only those molecules bearing a ureidic group matching the one displayed by 8. A total of 26,651 compounds were selected, and assigned a pseudoenergy score that expressed their fit into the APF fields generated using 8 as template (see Computational Methods). This score expressed the best possible overlap of the following six properties: hydrogen bond donor propensity, hydrogen bond acceptor propensity, sp 2 carbon atom hybridization, lipophilicity, size, and electronegativity. Since 8 is a neutral molecule, the charge field usually included in APF calculations was not taken into consideration. To select the compounds to prioritize, we arbitrarily set a threshold for the pseudoenergy value at -1200 score units. This value, which is two-thirds of the 1800 score units assigned to compound 8 into its own APF field, represented a good compromise between prioritizing a limited number of compounds and extending the selection beyond those derivatives just trivially resembling 8 without introducing any element of true novelty. Eventually 2317 compounds proceeded to medicinal-chemistry-guided visual inspection. Five distinct molecular scaffolds were selected based on (i) structural diversity, (ii) ease of preparation, and (iii) efficient access to structural analogues for SAR purposes (Figure 2). Scaffolds 10-13 and 14a, displaying monocyclic and bicyclic as well as heteroaliphatic and heteroaromatic systems, were functionalized as N-carboxamide derivatives 15-22, 23a-26a and screened against h-AC in a fluorescence-based assay 29 (Table 2).
Compounds 15-20 having a heterocyclic scaffold such as hydantoin, oxazolidinone, and piperazinone did not significantly inhibit AC at 10 μM. Similarly, no inhibition was seen for compounds 21 and 22, obtained from the heteroaromatic bicycle dihydrobenzoimidazolone. By contrast, Ncarboxamides of 3-hydroxy-5-methylpyrazole 23a and 24a and 25a and 26a showed a marked degree of inhibitory activity. These compounds showed median inhibitory concentration (IC 50 ) values in the low micromolar range, and an interesting difference in activity between regioisomers (e.g., 23a, IC 50 = 3.137 μM vs 24a, IC 50 = 0.455 μM, Table 5, vide infra). 30 Based on these results, we selected 23a and 24a as starting points for our SAR analysis.
We first investigated the role of acyl function and side chain. This study was conducted on the unsubstituted pyrazole ring to better appreciate possible changes in potency while avoiding influences due to substituents on the heterocycle and regioisomer formation. The unsubstituted pyrazole-N-carboxamide 29, bearing a six-carbon alkyl chain, was found to be moderately potent (IC 50 = 0.481 μM, Table 3).
The corresponding amide 27 and carbamate 28 derivatives showed no inhibitory activity. Similarly inactive was the N-Me derivative 35, where the 1-carboxamide NH moiety of 29 was substituted with a methyl group (Table 3).
Having assessed that the 1-carboxamide NH is mandatory for activity, we investigated the influence of side-chain length by preparing a small set of derivatives 30-34 with different number of carbon atoms in the side chain. A decrease in chain length, as in N-pentylpyrazole-1-carboxamide (30) (IC 50 = 1.39 μM), was associated with a decrease in inhibitory potency. Conversely, increasing the length of the chain from six to nine methylene units led to progressively

Feature Syn thesis
higher potency, with the N-nonyl derivative 33 showing the best IC 50 value (0.197 μM) within this set of compounds. Further extension of the chain length to ten methylene units had an opposite effect (34, IC 50 = 0.993 μM).
Next, we focused on substitutions of the pyrazole ring. Considering that the initial hit compounds, 23a and 24a, bear a hydroxyl and a methyl group at positions 3 and 5, respectively, we started our analysis by removing one group at the time to obtain the monosubstituted 3-and 5-hydroxy 23c and 24c and 3-and 5-methyl derivatives 23d and 24d (Table 4).
The 3-substituted compounds 23c and 23d showed an inhibitory potency in the micromolar range; the corresponding regioisomers 24c and 24d turned out to be 6-to 12-fold more potent than 23c and 23d, displaying IC 50 values of 0.196 μM and 0.169 μM, respectively, and 2-fold more potent than hit compound 24a (IC 50 = 0.455 μM). To further explore the effect of an electron-donating group on the pyrazole ring, the amino and methoxy-substituted de-rivatives 23e and 24e, and 23f and 24f were synthesized. Interestingly, we did not observe significant difference in potency between regioisomers 23e and 24e, which inhibited AC with an IC 50 of 0.216 μM and 0.522 μM, respectively. Although regioisomer 23e was more potent than 24e, the difference in IC 50 was only two-fold. Conversely, methylation of the hydroxyl group caused a substantial drop in potency, as the 3-methoxy derivative 23f only partially inhibited AC at 10 μM; its isomer 24f showed a limited chemical stability and was not tested. Replacement of the hydroxyl group with a bulkier, but planar moiety, such as a phenyl, led to the single isomer 23g, which inhibited AC with moderate potency (IC 50 = 0.766 μM).
Continuing our study on the effect of monosubstitution of the pyrazole, we turned to electron-withdrawing groups, as shown with compounds 23h-23i, obtained as single isomers (Table 4). The presence of bromine at position 3 led to a compound which was equipotent to 23c (23h, IC 50 = 2.49

Feature Syn thesis
μM), while a trifluoromethyl group turned out to be well tolerated, as derivative 23i showed an IC 50 value of 0.110 μM, being the most potent compound of this set.
As a continuation of the SAR study, we investigated the effect of disubstitution at position 3 and 5 of the pyrazole ring ( Table 5). As methyl-substituted derivatives 23d and 24d were found to be slightly more potent than the pyrazolol carboxamides 23c and 24c, and with the objective of removing an additional chemical handle suitable for acylation, we maintained the methyl substituent and replaced the hydroxyl group with those electron-donating and electron-withdrawing moieties studied in the monosubstituted series of compounds. We observed that disubstitution of the pyrazole ring generally enhances inhibitory potency against AC and is associated to significant differences in activity between the two isomeric series of compounds. The combination of methyl group with both electron-donating and electron-withdrawing moieties was generally well tolerated, delivering compounds with submicromolar potencies, with the exception of derivative 23l, which showed a 4-fold drop in potency with respect to hit 23a, confirming the detrimental effect of a methoxy group on this scaffold for AC inhibition. Of particular interest were the results obtained with the 3-methyl-5-bromo-and 3-methyl-5-trifluoromethyl-substituted derivatives 24m and 24n, which showed IC 50 values of 0.011 μM and 0.014 μM, respectively, representing the most potent compounds within this series (Table 5).
All the compounds showed double-digit nanomolar potency toward AC indicating that steric hindrance is generally well tolerated, with the exception of the less potent 3,5dipropyl derivative 38 (IC 50 = 0.294 μM). The highest activity within this small series was observed with compound 40 bearing a tert-butyl group at positions 3 and 5 (IC 50 = 0.011 μM).
At the outset, we set as goal of our project the diversification of the available arsenal of small-molecule inhibitors of AC that can be used as tools to investigate the biology of this enzyme and, possibly, as starting points for therapeutic candidates. Having elucidated the main SAR features of this new chemical class, our efforts were focused on its pharmacological characterization. The highly potent inhibitor 24n was selected for this purpose and its stability was assessed in buffer and mouse plasma (Table 7).
Compound 24n displayed adequate chemical stability both at acidic and neutral pH, as well as in cell culture media. Conversely, the compound had very limited stability in mouse plasma (half-life, t 1/2 = 9 min), making it a promising candidate for 'soft drug' applications (e.g., topical skin or lung delivery).
Accordingly, the compound was tested in the proliferative melanoma cell line G361. We selected this line based on previous work, which had shown that AC regulation of ceramide signaling is important in melanoma cell viability. 19 The cells were incubated in the presence of 24n (0.1-20 μM) and AC inhibition and sphingolipid levels were measured after various incubation times (Figure 4).

Feature Syn thesis
The compound caused a rapid concentration-dependent inhibition of AC activity in these cells. Notably, 24n inhibited cellular AC activity at concentration as low as 0.1 μM, and the inhibition was already appreciable after 15 minutes of incubation and persisted for 6 hours, with a partial recovery of enzyme activity after 24 hours (Figure 4, A and B).
This level of efficacy compares well with those reported for other classes of AC inhibitors. 19 As expected, AC inhibition was accompanied by accumulation of various species of ceramides: specifically, we observed a significant increase of those (dihydro)ceramides, which are preferred substrates for AC (namely, d18:0/16:0 and d18:1/16:0) and a concen-

Feature Syn thesis
tration-dependent decrease in sphingosine (Figure 4, C). At the highest concentration of 24n tested (20 μM), we observed a time-dependent accumulation of ceramide, which reached a maximum after 6 hours, and a rapid decrease in sphingosine levels, which was already detectable after 15 minutes (Figure 4, D). Increases in sphingomyelins and glucosylceramide were also observed after 15 min ( Figure S4 in the Supporting Information). The results demonstrate that 24n is able to effectively inhibit AC in a complex cellular environment.

Conclusions
In the present study, we utilized the APF-based computer-assisted approach to identify novel scaffolds that may serve as starting points for the design of small-molecule inhibitors of AC activity. Using carmofur (8) as a template, we probed the ZINC library and retrieved compounds that, while structurally different, match the pharmacophore fingerprint of 8. Medicinal chemistry-guided visual inspection of a set of computationally selected compounds identified five distinct scaffolds that, when appropriately functionalized, revealed substituted pyrazole carboxamides as promising novel AC inhibitors. A series of differently substituted

Feature Syn thesis
derivatives were synthesized and an NMR-based model was built to unambiguously assign the regiochemistry of N-acylation on the pyrazole ring. A systematic SAR study around the new scaffold highlighted structural features that are important for AC inhibition. We observed that an unsubstituted nitrogen atom in the carboxamide moiety is mandatory for activity, and assessed the best length for the alkylic side chain as nine methylene units. The electronic nature of substituents on the pyrazole ring does not significantly impact the inhibitory potency toward AC, as both electron-donating and electron-withdrawing moieties are tolerated, with the exception of the methoxy group that leads to inactive or unstable compounds 23f-24f, 23l-24l. Conversely, the regiochemistry of acylation turned out to be critical, as a 2-to 10-fold difference in potency was generally observed between the two series of regioisomers 23 and 24. Finally, one of the most potent compounds identified, 24n, was selected for further characterization. Stability analysis of this derivative suggested its potential use for 'soft drug' applications and, thus, it was tested on proliferative melanoma cells G361. The compound 24n effectively engages AC in these cells leading to the expected variations in the sphingolipid profile. This novel chemical class further expands and complements the available armamentarium of chemical tools to study the biology of AC and may offer promising candidates for the development of topical treatments for cancers and other hyperproliferative disorders of the skin.

-5, Scheme 3)
The properly substituted pyrazole (1.0 equiv) was dissolved in anhyd pyridine. DMAP (0.1 equiv) was added, and the reaction mixture was stirred under N 2 for 30 min. The appropriate isocyanate (1.1-1.5 equiv) was then added, and the resulting mixture was stirred for 12 h. The solvent was evaporated under reduced pressure and the crude was purified by silica gel column chromatography, eluting with a mixture cyclohexane/EtOAc.

APF Field Generation
First, compound 8 was generated in three-dimensional structure within ICM (ICM3.7, Molsoft LLC, San Diego, CA, USA). 31 Cartesian coordinates were translated into internal coordinates and MMFF 32 force field atom types and charges were assigned. The starting conformation of compound 8 was obtained assigning an elongated orientation to the acyl side chain. Then, continuous three-dimensional pharmacophoric grid potentials were calculated accounting for relevant physico-chemical features of 8. According to the original implementation reported by Totrov, seven APF potentials were selceted: hydrogen bond donor and acceptor propensity, lipophilicity, size, charge, hybridization, and electronegativity. These grids encompassed the template molecule plus a 5Å margin.

APF Screening
We retrieved a library of over ten million compounds from ZINC ('In-Stock', 'Drug-like' subset as of March 2011). 27 In ICM, this molecular dataset was prescreened by means of a 2D substructure query to retrieve all those compounds bearing a ureidic group having one nitrogen atom included in a ring system and the other as unsubstituted NH group. Selected compounds were automatically assigned three-dimensional internal coordinates, bond orders, tautomeric forms, stereochemistry, hydrogen atoms, protonation states, MMFF force field atom types, and charges by the ICM converting procedure. Then, the conformation of each compound was globally optimized using the biased probability Monte Carlo method within the precalculated APF field. 33 Torsional variables describing flexible rings were sampled explicitly. The intramolecular energy contribution was also considered to avoid unrealistically strained solutions. To provide a simulation of suitable length, the thoroughness parameter was set equal to the previously validated value of 1. Based on the fit of its optimized conformation into the APF field, each compound was assigned a pseudo-energy value and the library was ranked accordingly.

Pharmacology
Preparation of Enzyme-Enriched Lysate: Hek293 cells stably expressing human AC (Hek293-hAC) 19 were suspended in 20 mM Tris·HCl (pH 7.5) with 0.32 M sucrose, sonicated and centrifuged at 800 g for 30 min at 4 °C. Supernatants were then centrifuged at 12 000 g for 30 min at 4 °C. Pellets were re-suspended in PBS buffer (pH 7.4) and subjected to three freeze-thaw cycles at -80 °C. The suspension was finally centrifuged at 105 000 g for 1 h at 4 °C and protein concentration was measured in the supernatant with bicinchoninic acid based protein assay.