Nonclassical Phenyl Bioisosteres as Effective Replacements in a Series of Novel Open-Source Antimalarials

: The replacement of one chemical motif with another that is broadly similar is a common method in medicinal chemistry to modulate the physical and biological properties of a molecule (i.e., bioisosterism). In recent years, bioisosteres such as cubane and bicyclo[1.1.1]pentane (BCP) have been used as highly e ﬀ ective phenyl mimics. Herein, we show the successful incorporation of a range of phenyl bioisosteres during the open-source optimization of an antimalarial series. Cubane ( 19 ) and closo -carborane ( 23 ) analogues exhibited improved in vitro potency against Plasmodium falciparum compared to the parent phenyl compound; however, these changes resulted in a reduction in metabolic stability; unusually, enzyme-mediated oxidation was found to take place on the cubane core. A BCP analogue ( 22 ) was found to be equipotent to its parent phenyl compound and showed signi ﬁ cantly improved metabolic properties. While these results demonstrate the utility of these atypical bioisosteres when used in a medicinal chemistry program, the search to ﬁ nd a suitable bioisostere may well require the preparation of many candidates, in our case, 32 compounds.


■ INTRODUCTION
Phenyl rings are ubiquitous in medicinal chemistry, appearing in all manners of biologically relevant molecules. Yet phenyl groups may not always be the optimal motif: they are relatively nonpolar; they may be involved in π−π stacking interactions that can contribute to low aqueous solubility or have limited bioavailability; and they can be a metabolic liability. 1,2 Fortunately, there are several well-established strategies for addressing solubility issues. 3,4 Perhaps the most common strategy is to append charged or polar groups such as alcohol or amine moieties to the parent compound. Such modifications will typically reduce the hydrophobicity of the compound, but can often dramatically worsen potency. Similar effects may be seen through the use of heterocyclic ring replacements, but again these may alter the potency of the desired compound. 5 Alternatively, modifications can be made to alter the crystal packing of a compound by either removing aromaticity via removing the phenyl rings or changing the molecular geometry or topology. For example, replacement of a phenyl ring by an alkyl group (linear, cyclic, or caged) may help improve solubility by eliminating π−π stacking interactions and have the added advantage of introducing further options for chemical derivatization that may not be accessible with a phenyl ring. Again, however, there is the caveat that these changes can heavily influence the binding interactions between a drug and its ultimate biological target.
Our recent experiences in dealing with such issues are related to an antimalarial medicinal chemistry program. Malaria is one of the most prevalent infectious diseases affecting low-income countries, with 219 million cases reported worldwide in 2017, 2 million more cases than the previous year. This translates to around 1200 deaths per day, mostly involving children. 6 With an increasing number of reports of resistance to current treatment and prevention strategies, new medicines must be discovered to combat the disease. 7,8 To help address this, the Open Source Malaria (OSM) consortium was created, with the aim of discovering new antimalarial medicines using an inclusive community operating on open-science principles, where all data and experiments are shared in real time (for example, every experiment involved in the current study is freely available to view online). 9 The most recent, fourth series examined by the consortium, the so-called Series 4 triazolopyrazine antimalarials, originated from a high-throughput screen performed at Pfizer in collaboration with the Medicines for Malaria Venture (MMV). The series was then passed to TCG Lifesciences for further optimization before being donated to the OSM consortium in 2013. 10 The series has produced many compounds with potent (<100 nM) activity against Plasmodium falciparum, promising physicochemical properties, and low toxicity ( Figure 1). Two compounds from the series were shown to have high potency in an in vivo mouse model (full results to be reported elsewhere). 11 While the potency of compounds in the series has been frequently high, issues still remain, particularly with regard to poor solubility and moderate clearance rates. As such, modification of the phenyl groups present in the structure was seen as an attractive strategy for mitigating these issues. Careful consideration of such modifications was required as the series is in the lead optimization stage of development and displays high sensitivity to functional group changes.
The approach adopted was based on isosteres, a concept first described by Harris Friedman in 1950 as "compounds or groups that possess near-equal molecular shapes and volumes, approximately the same distribution of electrons, and which exhibit similar physical properties". 12 The use of a bioisostere (isosteres that have the same type of biological activity) can be an extremely useful strategy in medicinal chemistry programs for altering the physical and biological properties of a compound. 13 Classical bioisosteres focus on the use of structurally simple atoms, groups, and ring equivalents (e.g., replacement of a phenyl ring with a pyridine ring), while nonclassical bioisosteres may differ quite dramatically from the original group. 14−17 One of the more interesting nonclassical phenyl bioisosteres is cubane, a motif that has been used in medicinal chemistry projects spanning a range of applications. 18−20 The size and shape of cubane mimic the rotational volume and shape of a phenyl ring. 21 Several biologically active molecules have been evaluated for their ability to tolerate a cubane as a replacement for an aromatic ring, examining changes in biological activity, solubility, metabolism, stability, and synthetic tractability (Figure 2A). 21 The five compounds evaluated included those with chemotherapeutic (Vorinostat), anesthetic (Benzocaine), and neotropic (Leteprinim) applications. It was found that four of the five showed equivalent or improved potency when the phenyl ring was replaced with a cubane. In these four cases, the slight increase in log P values with the cubane replacements (largest difference of log P ∼ 0.5) did not have a significant impact on compound solubility. However, in the one case that led to decreased potency, a much larger increase in log P (∼1.4) was seen for the cubane analogues that translated to a large reduction in solubility. A more recent evaluation of a further five pharmaceuticals revealed a mixed effect of the cubane replacement on both potency and solubility. 22 In a similar manner to cubane, the bicyclo[1.1.1]pentane (BCP) motif has recently found use as a nonclassical phenyl ring bioisostere. 23 While the size of the BCP motif does not match that of benzene as closely as cubane (vide infra), its use as an effective phenyl bioisostere has been demonstrated in a number of cases. For example, the phenyl ring in a γ-secretase inhibitor currently in development was replaced with the BCP motif resulting in a compound with not only equipotent enzyme inhibition but also improved passive permeability, aqueous solubility, and oral absorption characteristics. 24 In another example, a BCP replacement in a LpPLA 2 inhibitor resulted in maintenance of potency while improving the compound's physicochemical properties (B, Figure 2). 25 Increasing work has  been done in recent years with the synthesis of BCP building blocks allowing for more possibilities for its use as a phenyl bioisostere. 26,27 Perhaps less immediately obvious to medicinal chemists is the use of the boron-rich icosahedral clusters known as carboranes and their potential application as phenyl and adamantyl bioisosteres. 28 One prominent example is the closo-carboranyl derivative (closed form with all carbon and boron vertices) of tamoxifen. 29 The resulting carboranyl compounds showed high activity as antiestrogen agents and were even found to be more stable to degradation than tamoxifen (C, Figure 2). Another example is that of asborin, a closo-1,2-carborane derivative of aspirin. 30 Interestingly in this case, the pharmacological profile of the resulting carboranyl derivative was changed drastically. Instead of acting as a selective cyclooxygenase enzyme inhibitor (like aspirin), asborin was instead found to be a potent inhibitor to the unrelated aldo/keto 1A1 reductase family. 31 Other examples of carboranyl bioisosteres of polycycles such as adamantanes can be found in medicinal chemistry, including their in vivo application as central nervous system (CNS)modifying agents. 32 While it is not unusual for a medicinal chemistry campaign to explore select phenyl bioisosteres (typically cubane and BCP), 33,34 it is much less common to see the wider range of possible bioisosteres used in a single lead optimization program. In the case of the OSM Series 4 triazolopyrazines, the changes required to achieve our goal (improve the aqueous solubility and metabolic clearance parameters, but not significantly change the potency) were not immediately clear, in part because the project was a phenotypic drug-discovery program with no confirmed biological target. While the calculation of cLog P values may give an indication of the potential aqueous solubility of a compound, these values can only be used as a general guideline and do not necessarily provide a good indication of bioavailability (note that in this study, cLog P was prioritized over other possibly relevant parameters, such as molecular weight or total polar surface area; these values may be found in the Supporting Information (SI), though they appear to show no correlation with experimental potency). This is particularly notable for some of the phenyl bioisosteres that are mentioned above, where the calculated log P values can be inconsistent across different software platforms, 35 particularly where the carborane motif is involved; in such cases, it may be necessary to measure solubility experimentally. 22 Accordingly, we investigated a wide range of possible phenyl bioisosteres for Series 4, without relying heavily on the calculated cLog P values as a guide.

■ RESULTS AND DISCUSSION
Chemistry. The synthesis of the triazolopyrazine core was achieved in three steps from the commercially available 2,6dichloropyrazine (Scheme 1). Initial displacement of a chlorine atom with hydrazine hydrate gave 2-chloro-6-hydrazinylpyrazine (1 cyclized to give the chlorinated triazolopyrazine core (3a−g). Final nucleophilic displacement of the chlorine atom with the appropriate alcohol produced the desired target compounds 4− 31. A number of additional transformations were performed on select target compounds. Deprotection of N-Boc compounds 14, 26, and 28 was carried out using standard trifluoroacetic acid (TFA) conditions to give the corresponding free amine compounds 32, 33, and 34, respectively. Compounds 16 and 19 were used to generate late-stage biofunctionalized derivatives using dog or rabbit liver microsomes (see the Experimental Section). In the case of the norbornene compound 16, oxidation of the norbornene double bond led to a mixture of epoxides and E and/or Z diol compounds (35 and 36). In the case of the cubane compound 19, two metabolites were isolated and were found to be hydroxylated on the cubane framework (37 and 38). The reaction of compound 23 with CsF in EtOH gave the hydrophilic nido-7,8-carboranyl derivative as its cesium salt 39.
In Vitro Activity. All compounds were evaluated for in vitro antiplasmodial activity against the 3D7 strain of P. falciparum. The activities, indicated by their IC 50 values, are summarized in the SI and shown graphically in Figure 3. While the inherited dataset for OSM Series 4 suggested an ether chain length of two methylene groups between the heterocyclic core and the northwest pendant phenyl ring was optimal, the specific compounds to prove this relationship were absent from the dataset. In order to validate this hypothesis, a set of compounds with an ether methylene chain length from zero to three was evaluated, clearly showing that this separation of the phenyl from the core was indeed important to maintain potency. Specifically, any chain length other than two carbon atoms (6) resulted in reduced potency with complete inactivity seen with a phenol side chain (4).
Surprisingly, when the saturated heterocyclic derivatives (9− 14, 26−29, and 32−34) were evaluated, it was found that they were all inactive with complete loss in potency compared to the parent phenyl compound. This was seen in both the northwest and northeast cases. The increased flexibility of these saturated heterocycles is presumably contributing to this loss of activity.
Encouragingly, the comparable two-and three-methylene linker cubane analogues 19 and 20, respectively, were only slightly less potent than the phenyl compounds but also followed the trend of decreased potency with longer chain length. The recovery of activity compared to the saturated heterocyclic derivatives provides strong evidence of the ability of cubane to mimic the rotational volume of a phenyl ring. It was also verified that the mechanism of action had not changed following these modifications: the potent bioisosteric compounds showed activity in an ion regulation assay, which is linked to the malarial target PfATP4. 36,37 However, on the contrary, the northeast cubane derivative 31 was found to be completely inactive, whereas the analogous phenyl compound 30 remained potent. This surprising result suggests that the ability of cubane to mimic a phenyl ring is reduced if the phenyl ring possesses restricted rotation (as is assumed to be the case here), in which case the volume occupied by the phenyl ring is not well mimicked by the bulk of the cubane. Alternatively, the change may have removed a key πstacking interaction. Considering that all northeast phenyl replacements resulted in completely inactive compounds, the remaining investigation was focused on the further exploration of the northwest pendant phenyl ring.
The BCP derivative 22 was relatively well tolerated compared to the analogous benzylic ether compound 3 with only a slight decrease in activity. This loss in potency may be due to the smaller size of the BCP motif; however, it is noted that many examples of the use of BCP as a phenyl bioisostere tend to result in more significant changes to the metabolic and pharmacokinetic properties instead, as will be discussed below.
The larger adamantyl derivatives were less well tolerated with the one-and two-carbon chain compounds (17 and 18, respectively) showing potencies approximately 2−9 times lower than the corresponding phenyl compounds (5 and 6, respectively). This result may be attributed to the larger volume of adamantane (vide infra) compared to either a cubane or a rotating phenyl ring. 38 Other hydrocarbon-caged derivatives including nopol, norbornene, and trishomocubane (15,16, and 21, respectively) were better tolerated; however, these were still found to be less potent than the parent phenyl compounds.
As a final investigation point, the closo-1,2-, 1,7-, and 1,12carborane isomers were installed in place of the northwest phenyl position. Unexpectedly, the closo-1,2-and 1,7-carborane isomers (23 and 24, respectively) showed potencies greater than not only all of the other previous derivatives but also, more importantly, the parent phenyl compound. A significant drop in activity was seen between the closo-1,7-and 1,12-carborane isomers (24 and 25). The potency was observed to decrease from closo-1,2-< 1,7-< 1,12-carborane. This trend is likely related to the differences in polarity (and hydrophobicity) of the three carborane isomers. 39 The more hydrophobic and less polar the closo-carborane isomer is (as pertaining to the relative positions of the two carbon atoms in each carborane cage and the resulting dipole moments), the less potent the final compound was found to be.
In a similar manner to closo-1,2-carborane, the hydrophilic nido-7,8-carborane cage (open form with one boron vertex removed) has also shown potential as a phenyl bioisostere both in a carborane variant of tamoxifen 40 and trimethoprim, 41 for example. In the latter case, the nido form was found to be less toxic than the closo form; however, it also performed less usefully as a boron neutron capture therapy (BNCT) agent, with poorer tumor retention and lower selectivity ratios for boron distribution in tumor versus normal tissues. In our case, the nido-7,8-carborane 39 was also found to be less active, with its potency decreasing significantly compared to the closo-1,2carborane 23. Selective removal of a single BH group from the 3 (or 6) position of the closo-carborane cage affords a hydrophilic, anionic species, which would likely show diminished binding to a hydrophobic receptor pocket.
With these encouraging results, select compounds were further evaluated for a number of other properties including toxicity, solubility, and metabolic stability.
Toxicity. Having synthesized a number of highly potent compounds that possessed unusual motifs, it was important to check for any associated toxicity. While a number of examples in the literature show no significant increases in toxicity with cubane bioisosteres, 42,43 carborane-associated toxicity has been far less studied. Encouragingly, when closo-and nido-carborane compounds 23 and 39 were evaluated for potential cytotoxicity in HepG2 cells, both compounds were found to be inactive, with IC 50 values of >10 μM. Compounds 6, 19, and 23 were additionally evaluated for their human ether-a-go-go-related gene (hERG) activities with the phenyl compound 6 showing moderate activity (IC 50 = 7.41 μM) and the cubane and carborane analogues showing slightly greater potencies (IC 50 = 4.26 and 3.62 μM, respectively). While these values are not entirely desirable, further hERG optimization to lower inhibition may be achieved through further derivatization.
Metabolic and Physicochemical Properties. The reported effects on the metabolic and physicochemical properties of phenyl bioisosteric replacement by cubane have been mixed. While a number of cases have reported improvements in solubility following cubane substitutions, 34 the trend for metabolic stability is less consistent. There have been reports of both increased 21 and decreased 33 metabolic stability as a result of cubane substitution. In the present case, when compounds 6, 19, and 23 were evaluated for their metabolic and physicochemical properties, they were found to perform poorly compared to the parent phenyl compound ( Table 1). The solubility at pH 6.5 was low for 19 and 23 (<1.6 μg/mL) compared to 6 (6.3−12.5 μg/mL). Both 19 and 23 also exhibited high clearance and short half-lives in human liver microsomes (HLM) and mouse liver microsomes (MLM). In rat cryopreserved hepatocytes (RCH), 19 and 23 showed intermediate and low degradation rates with low clearance rates and short half-lives. It is likely that clearance of the parent compound occurs via benzylic oxidation 44 and that the removal of this phenyl ring should slow down the rate of clearance due to the absence of the benzylic position. Our results suggest that the Journal of Medicinal Chemistry pubs.acs.org/jmc Article incorporation of cubane in place of phenyl has led to the cubane becoming a metabolic liability. This appears to be in contrast to the notion that cubane derivatives are more metabolically stable to hydroxylation than the corresponding phenyl analogues (increased s-character from the strong and hindered tertiary C− H bonds). 45 In our case (vide infra), it appears that the metabolic "hotspot" has shifted onto the cubane itself. Encouragingly, the BCP derivative 22 was found to possess significantly improved metabolic and physicochemical properties compared to the parent phenyl compound. While the solubilities were within the same range, 22 exhibited lower clearance and shorter half-life values across human, mouse, and rat liver microsomes. Late-Stage Biofunctionalization. Further light is shed on these results by the isolation of the four metabolites (35−38), which were found to be inactive. A similar metabolism of phenylsubstituted compounds leads to oxidation at the benzylic position, as expected (these will be reported separately for this series), whereas for cubane, the hydroxylation occurred on the hydrocarbon cage instead, i.e., we observe preferential hydroxylation of the benzylic position in the phenyl case but core hydroxylation for cubane. This is unexpected. Experiments on enzymatic oxidation of methylcubane have indicated that hydroxylation on the methyl group is favored over hydroxylation on the cubane framework with only small amounts of the latter products being identified. 46,47 If the cubane does become hydroxylated, it is thought that such products are unstable and lead to rearrangement to a ring-opened ketene product. 48 This possibly explains why there are a small number of reports of hydroxycubanes in the literature. 49,50 The present isolation of hydroxycubane derivatives is therefore unusual.
Comparison of Phenyl Bioisosteres. In literature reports where phenyl bioisosteres have been used, a size comparison between the bioisostere and the phenyl ring is often given. There is some variation between these values. For example, in one report, the diagonal C−C atom distances for benzene and cubane have been quoted as the same (2.70 Å), 51 while another report has quoted them as different (2.82 and 2.72 Å for benzene and cubane, respectively). 34 Similar disagreements have been seen for BCP comparisons as well (vide infra). For meaningful comparisons, a unified approach was preferable. We have therefore examined deposited crystal structure files from The Cambridge Crystallographic Data Centre (CCDC) and calculated the dimensions of key phenyl bioisosteres with UCSF Chimera, allowing for a direct comparison ( Table 2).
While the literature atom distances are, for the most part, consistent with those calculated with UCSF Chimera, there is a notable increase of about 10−15% between the calculated and literature values for surface area and volume. It remains largely unclear as to which methods provide calculated values closest to those measured from X-ray structures; accurate calculations would be useful for calculating the dimensions of novel bioisostere motifs. Most literature sources only provide a small selection of calculations, and specifically for volume calculations, none take into account the free rotation of the bioisostere. However, as these comparisons are often made to show the relative dimensions, calculations performed using the same method or program would be the most reliable, such as in this case with UCSF Chimera. Nevertheless, it is interesting to note how different the dimensions can be for phenyl and an effective bioisostere. For example, the closo-carborane core was found experimentally to be an excellent mimic, yet its volume is significantly larger than that of a phenyl ring, meaning the functional significance of bioisosteric size comparisons must be treated with care.

■ CONCLUSIONS
While phenyl rings can be found in all manners of biologically important molecules, they may not always be the ideal motif for druglike molecules. Using classical and nonclassical bioisostere replacements, the biological properties of a compound may be altered in dramatic ways. These changes may not always result in the desired improvements, and accurate predictions can be challenging. In phenotypic projects like this one, which are now so common in early-stage drug discovery, rational modifications, such as the substitution of a rotatable phenyl group with a cubane, may or may not lead to improvement in the overall properties of the molecule given how many other potential interactions may be involved before the molecule reaches its intended target. As seen in this work, equivalent substitutions for different phenyl rings in the same molecule can lead to widely different potencies. In any given case, it may well be necessary to Surface areas and volumes were calculated using a model of the solvent-excluded molecular surface with hydrogen atoms included. b R = terminal alkyne. c Alkyne C−C distance. d Surface area and volume calculated without alkyne substituents. e Carborane C−B distance. f Where they exist, literature values are shown in parentheses. C−C and H−H distances are indicated in purple and orange, respectively. explore a considerable range of possible bioisosteres before arriving at an effective solution. In our case, we have identified a series of novel closo-carborane containing compounds that possessed potencies greater than that of the parent phenyl compound. To get to this point required the synthesis of 30 variations of the initial hit. More rational substitutions may be possible in cases where the structures of the biological targets are known, particularly through consideration of the volumes occupied by the various isosteres.
■ EXPERIMENTAL SECTION General Information. Reagents were purchased from Sigma-Aldrich, Alfa Aesar, Acros Organics, Merck, Fischer Scientific, Matrix Scientific, Ajax, or Fluorochem. Unless otherwise specified, the reagents were used without further purification. Anhydrous conditions: glassware was dried at >130°C for >12 h, assembled hot, and allowed to cool under a high vacuum where appropriate or purged with inert gas. Anhydrous solvents were obtained from the PureSolv system or by drying over activated 3 Å molecular sieves. Nitrogen gas was dried over silica and calcium chloride. Argon gas was used as acquired. The phrase in vacuo corresponds to ∼1 mbar on a Schlenk line. Reduced pressure means under rotary evaporation at 40°C from 900 to 50 mbar. Flash chromatography was performed on Davisil Grace Davison 40−63 μm (230−400 mesh) silica gel or on a Biotage Isolera One. Analytical thinlayer chromatography was performed on Merck Silica Gel 60 F 254precoated aluminum plates (0.2 mm) and visualized with UV irradiation (254 nm) and potassium permanganate, anisaldehyde or ninhydrin staining. High-temperature reactions were carried out in silicone oil baths, controlled by temperature probe in the oil bath.
Melting points (mp) were recorded on a Stanford Research Systems OptiMelt at 1°C min −1 (capillaries ø = 1.5−1.6 mm, 90 mm) or a Stuart SMP10 at 2°C min −1 (capillaries ø = 1.8−1.9 mm, 100 mm). Infrared spectroscopy was carried out on a Bruker α-E (attenuated total reflectance) without atmospheric compensation and processed using OPUS 7.0 software. Samples were analyzed neat. Nuclear magnetic resonance spectroscopy was carried out at 300 K on  (Hz). Integrals are relative. app = apparent when the multiplicity was unexpected, e.g., coincidental or unresolved. Lowresolution mass spectrometry (m/z) was carried out on a Finnigan quadrupole ion-trap mass spectrometer using electrospray ionization (ESI) or atmospheric-pressure chemical ionization (APCI). Highresolution mass spectrometry (HRMS) was performed on a Bruker 7T FT-ICR using ESI or APCI. Positive and negative detection is indicated by the charge of the ion, e.g., [M + H] + indicates positive-ion detection.
Purity of all compounds was >95% as determined by NMR spectroscopy (provided for all compounds evaluated biologically).
2-Chloro-6-hydrazinylpyrazine (1). To a solution of 2,6-dichloropyrazine (22.3 g, 150 mmol, 1 equiv) in EtOH (428 mL, 0.35 M) was added hydrazine monohydrate (14.7 mL, 299 mmol, 2 equiv). The reaction was heated to 80°C until completion, as indicated by thinlayer chromatography (TLC). The solution was allowed to cool to rt, and the solvent was removed under reduced pressure. H 2 O and EtOAc were added, and the organic layer was separated. The aqueous layer was extracted with EtOAc (3×), and the combined organic layers were washed with brine, dried (Na 2 SO 4 ), filtered and concentrated under reduced pressure to give 1 as fine yellow needles (19.7 g, 93%); R f 0.12 (30% EtOAc in hexanes); mp 133−135°C (lit. 56 56 General Procedure for the Synthesis of 2a−g. Compound 1 (1 equiv) was stirred into EtOH (112 mM). Aldehyde (1 equiv) was added, and the reaction was stirred at rt until completion, as indicated by TLC. The solvent was removed under reduced pressure to give compounds 2a−g, which were carried forward without further purification unless otherwise stated.
Cytotoxicity Studies (Drug Discovery Unit, University of Dundee). 70 In vitro cytotoxicity studies were carried out using HepG2 (Human Caucasian hepatocyte carcinoma, ECACC; cat. no. 85011430) as indicators for general mammalian cell toxicity. HepG2 in vitro cytotoxicity can be assessed using the assay procedure, as described. 71 Kinetic Solubility Estimation using Nephelometry (Centre for Drug Candidate Optimization, Monash Institute of Pharmaceutical Sciences). Compound in DMSO was spiked into either pH 6.5 phosphate buffer or 0.01 M HCl (pH ∼ 2.0) at seven concentrations with the final DMSO concentration of each being 1%.
After 30 min, the samples were then analyzed for precipitation via nephelometry to determine the solubility range. 72 Distribution coefficient was estimated using chromatography (Centre for Drug Candidate Optimization, Monash Institute of Pharmaceutical Sciences). Partition coefficient values (Log D) of the test compounds were estimated at pH 7.4 by correlation of their chromatographic retention properties (mean of two injections) against the characteristics of a series of standard compounds with known partition coefficient values. The method employed is a gradient HPLCbased derivation of the method developed by Lombardo. 73 In Vitro Metabolic Stability (Centre for Drug Candidate Optimization, Monash Institute of Pharmaceutical Sciences). The metabolic stability assay was performed by incubating each test compound with liver microsomes at 37°C and a protein concentration of 0.4 mg/mL. The metabolic reaction was initiated by the addition of an NADPH-regenerating system and quenched at five time points over a 60 min incubation period by the addition of acetonitrile containing diazepam as the internal standard. Control samples (containing no NADPH) were included (and quenched at 2, 30, and 60 min) to monitor potential degradation in the absence of cofactor. All liver microsomes used in these experiments were purchased from XenoTech. Microsomal incubations were performed at a substrate concentration of 1 μM. The half-life and intrinsic clearance were determined by exponential regression of the concentration vs time data.
Late-Stage Biofunctionalization of Compounds 16 and 19 (Obach Lab, Pfizer). Compound 19 (40 μM) was incubated with dog liver microsomes (1 mg/mL) in a total volume of 40 mL of potassium phosphate buffer (0.1 M; pH 7.4) containing MgCl 2 (3.3 mM) and NADPH (1.3 mM). The incubation was done in a 500 mL Erlenmeyer flask in a shaking water bath maintained at 37°C open to the air. After 1 h, the incubation was terminated by addition of MeCN (40 mL) and the precipitated protein was removed by spinning at 1700g for 5 min. The supernatant was partially evaporated in a vacuum centrifuge for 1.5 h, and to the remaining liquid was added formic acid (0.5 mL), MeCN (0.5 mL) and H 2 O to a final volume of 50 mL. This mixture was spun in a centrifuge at 40 000g for 30 min. The clarified supernatant was applied to a Varian Polaris C18 column (4.6 mm × 250 mm; 5 μm) at 0.8 mL/ min through a Jasco HPLC pump. After the entire solution was applied and another 5 mL of 0.1% formic acid in H 2 O was washed through the system, the column was moved to an HPLC-UV-MS (Thermo Velos LTQ mass spectrometer, equipped with a Waters Acquity HPLC-UV system) in line with a fraction collector (Leap Analytics). A mobile phase gradient was applied at 0.8 mL/min beginning with 2% MeCN in 0.1% aqueous formic acid, raised to 20% MeCN at 1 min, held until 5 min, then increased linearly to 55% MeCN at 80 min, followed by a 10 min wash at 95% MeCN and 10 min reequilibration to initial conditions. Fractions were collected every 20 s. Fractions predicted to contain products of interest eluted around 39.5 and 42.0 min and were individually analyzed on a Thermo Orbitrap Elite UHPLC-UV-HRMS system to ascertain identity and purity for pooling and qNMR analysis. Pooled fractions were evaporated in a vacuum centrifuge, and the residue was taken up in DMSO-d 6 (0.05 mL). Compound 16 was subjected to a similar procedure except that rabbit liver microsomes were used as the source of enzyme, the substrate concentration was 25 μM, and the incubation time was 45 min. The fractionation was carried out similarly, except that the MeCN composition was increased to 70% instead of 55%.

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c00746. 1 H and 13 C NMR spectra of the synthesized compounds; graphical table of in vitro potencies for the target compounds; and NMR spectra of the synthesized compounds (PDF)