Towards Selective Mycobacterial ClpP1P2 Inhibitors with Reduced Activity against the Human Proteasome

ABSTRACT Mycobacterium tuberculosis is responsible for the greatest number of deaths worldwide due to a bacterial agent. We recently identified bortezomib (Velcade; compound 1) as a promising antituberculosis (anti-TB) compound. We showed that compound 1 inhibits the mycobacterial caseinolytic proteases P1 and P2 (ClpP1P2) and exhibits bactericidal activity, and we established compound 1 and ClpP1P2 as an attractive lead/target couple. However, compound 1 is a human-proteasome inhibitor currently approved for cancer therapy and, as such, exhibits significant toxicity. Selective inhibition of the bacterial protease over the human proteasome is desirable in order to maintain antibacterial activity while reducing toxicity. We made use of structural data in order to design a series of dipeptidyl-boronate derivatives of compound 1. We tested these derivatives for whole-cell ClpP1P2 and human-proteasome inhibition as well as bacterial-growth inhibition and identified compounds that were up to 100-fold-less active against the human proteasome but that retained ClpP1P2 and mycobacterial-growth inhibition as well as bactericidal potency. The lead compound, compound 58, had low micromolar ClpP1P2 and anti-M. tuberculosis activity, good aqueous solubility, no cytochrome P450 liabilities, moderate plasma protein binding, and low toxicity in two human liver cell lines, and despite high clearance in microsomes, this compound was only moderately cleared when administered intravenously or orally to mice. Higher-dose oral pharmacokinetics indicated good dose linearity. Furthermore, compound 58 was inhibitory to only 11% of a panel of 62 proteases. Our work suggests that selectivity over the human proteasome can be achieved with a drug-like template while retaining potency against ClpP1P2 and, crucially, anti-M. tuberculosis activity.

ClpP1P2, proteasome, and bacterial-growth inhibition assays. We employed two target-based whole-cell assays, the mycobacterial-ClpP1P2 and human-proteasome inhibition assays, in order to evaluate the selectivity of derivatives of compound 1 for the bacterial target. The ClpP1P2 inhibition assay measures the intracellular accumulation of RFP-SsrA as a result of ClpP1P2 inhibition (11). The principle of the assay is as follows. Under undisturbed conditions, ClpP1P2 recognizes and degrades the RFP-SsrA protein to a background level of fluorescence. An inhibitor of ClpP1P2, like compound 1, binds to the catalytic sites of the protease and prevents the degradation of the RFP-SsrA protein, resulting in its accumulation and a gain-of-fluorescence signal (Fig. 3A). Similarly, the human-proteasome inhibition assay makes use of a proteasome-specific cleavage tag (Z-LLVY) fused to an aminoluciferin molecule. Under undisturbed conditions, the chymotrypsin-like catalytic site of the proteasome recognizes the tag and cleaves it, releasing the aminoluciferin molecule. The free aminoluciferin is a substrate for the luciferase enzyme in a reaction that produces luminescence. In the presence of a proteasome inhibitor, like compound 1, the cleavage is blocked, preventing the release of the aminoluciferin and the subsequent emission of luminescence (Fig. 3B). In order to determine both the ClpP1P2 and proteasome 50% inhibitory concentrations (IC 50 s) for all derivatives of compound 1 (i.e., the concentration required to inhibit 50% of the ClpP1P2 or proteasome activity), we tested new compounds in a dose-response analysis in both assays, from which we determined the IC 50 s. We further evaluated the growth inhibition potency of each derivative using a standard turbidity-based growth inhibition assay. Exponentially growing cultures of M. smegmatis ⌬prcAB were exposed to increasing concentrations of a given derivative. After 24 h of exposure, culture turbidity (i.e., growth) was assessed by measuring optical density at 600 nm (OD 600 ) and plotted as a function of compound concentration. MIC 50 s were then determined (i.e., the concentration required to inhibit 50% of growth compared to the growth of an untreated control). In all assays, compound 1 was used as a positive control and a reference compound.
From these three whole-cell biological assays, we could evaluate the activities of our derivatives both against the bacterial target, ClpP1P2 (ClpP1P2 IC 50 ), and against bacterial growth itself (MIC 50 ) as well as against the human proteasome (proteasome IC 50 ). Measuring both ClpP1P2 and bacterial-growth dose responses ensured that new compounds remained on target and were correlated with desired phenotypic disease Without any interference, the proteasome recognizes the Z-LLVY tag and cleaves it. The aminoluciferin is used as a substrate by the luciferase enzyme to generate luminescence. In the presence of a proteasome inhibitor (compound 1), the cleavage of Z-LRR is prevented. The lack of the luciferase substrate results in a reduced luminescence emission. RFU, relative fluorescence units; RLU, relative luminescence units.
activity. Using cell-based assays for the assessment of new compounds ensures that bacterial penetration is built in and, hence, increases the drug-like character of new compounds.
Design of compounds. At the outset of this project, our primary goal was to identify compounds which inhibit bacterial ClpP1P2 in a bacterial cell but which have reduced potency against the human proteasome compared to that of compound 1. Modeling studies suggested that a larger P1 substituent could be tolerated by the ClpP1P2 P1 pocket but should be less well accommodated in the P1 pocket of the human proteasome (Fig. 4). We therefore prioritized exploration of both aromatic and saturated rings directly attached to the P1 backbone carbon of the inhibitor. Given the uncertainty in the precise nature of the P1 group, we also planned to prepare a range of larger substituents connected at P1 via progressively longer linkers. For initial studies, we planned only minimal variations of the P2 phenylalanine side chain but wider variations of the CAP group. Our rationale was that CAP group changes would be well tolerated in this area of the molecule, and given the significant role that bacterial cell penetration is likely to play, the CAP group represents a good opportunity for tuning of physicochemical properties and hence of ClpP1P2 and antibacterial activity.
Chemistry. Preparation of final compounds with the P2 substituent fixed as phenylalanine (benzyl side chain) and the CAP group fixed as pyrazine was accomplished via acid intermediate 6, prepared from L-phenylalanine (compound 2) and 2-pyrazinecarboxylic acid (compound 4) (Fig. 5). Silylation of compound 2 was (Left) Molecular surface of M. tuberculosis ClpP1P2 and the human proteasome in gray, blue, and red, indicating neutral, positive, and negative electrostatics, respectively, with substrate sites 1 to 3 (S1 to S3) labeled in yellow. There is more available space in the S1 and S3 sites of ClpP1P2 than in the human proteasome. (Right) Selected residues of M. tuberculosis ClpP1P2 and the human proteasome are shown as a thin tube, with gray carbon and hydrogen bonds as dashed magenta lines. Residues are from the same protein subunit unless marked by a suffix indicating the PDB chain. The hydrogen bond network between compound 1 and the human proteasome is retained when compound 1 binds to M. tuberculosis ClpP1P2, and consequently, we did not try to modify the backbone of compound 1. The orientation of the P2 side chain differs, but there is little interaction between this and either protein, and modeling indicates that it is free to move around in the binding site.
Structure-activity relationships. In order to understand the roles of the P1, P2, and CAP groups of the boronic acids with regard to ClpP1P2 targeting, antimycobacterialgrowth inhibition, and selectivity against the human proteasome, various substituents on the P1, P2, and CAP positions were studied. In these assays, compound 1 has an IC 50 of 6 M in the ClpP1P2 cell reporter assay, which translates to the same concentration as its MIC 50 for the growth inhibition of M. smegmatis ⌬prcAB, whereas potency for the human proteasome was in the single-digit nanomolar range (IC 50 ϭ 5 nM) (Table 1). This profile translates to a selectivity preference for the human proteasome of 1,200fold, not surprising given the fact that compound 1 is a highly optimized proteasome inhibitor. Hence, the objective of this work was to prepare new compounds with a reduced preference for the human proteasome while maintaining or improving the potency against ClpP1P2. Guided by modeling, the influence of substituents at the P1 position was initially studied.
Replacing the iso-butyl in compound 1 with a less hindered straight-chain n-pentyl (compound 27) resulted in a 2-fold improvement in antibacterial activity, while ClpP1P2 potency was maintained at 4 M. Encouragingly, this compound was 6-fold-less potent than compound 1 in the proteasome assay (see the proteasome potency ratio in Table  1), with an IC 50 of 30 nM. Increasing the steric bulk of P1 with a cyclohexyl group directly attached (compound 28) further reduced proteasome activity to 155 nM (30-fold less active than compound 1). It appeared that our strategy of increasing the size of the P1 group did influence selectivity. However, disappointingly, the potency against ClpP1P2 of compound 28 was reduced to only 33 M, with corresponding reduced antibacterial activity.
Inserting a methylene linker to position the cyclohexyl further from the inhibitor backbone (compound 29) produced a similar result, but adding one additional methylene (cyclohexylethyl 30) further reduced proteasome activity to 0.5 M, an overall 100-fold reduction compared to the activity of compound 1. Unfortunately a significant drop in the potency against ClpP1P2 was also observed. Moreover, this compound still exhibited bacterial-growth inhibition (MIC 50 ϭ 3.5 M), suggesting that this activity was not due to ClpP1P2 inhibition. We were also concerned about increasing hydrophobicity and poor aqueous solubility with lipophilic cyclohexyl derivatives. Therefore, we next tested aromatic P1 groups. However, only 10-fold-less proteasome activity over that of compound 1 was achieved with benzyl P1-substituted compound 31, and growth-inhibitory potency was also reduced to 16.5 M. Potency against ClpP1P2 and the growth MIC 50 were maintained with a phenethyl P1 group compound (compound 32) which had a 14-fold-lower potency for the proteasome (IC 50 ϭ 70 nM) than compound 1. Various groups at the CAP position were next studied to explore their influence on the ClpP1P2 and proteasome potency ( Table 2). Removal of the entire CAP group and the P2 amine (compound 25) abolished both the bacterial-growth inhibition and the ClpP1P2 activity. Replacing only the P2 amine, without CAP, gave compound 26, with similar results. These non-CAP compounds still retained proteasome activity, demonstrating the considerable challenge of reducing activity against the proteasome in this series while retaining ClpP1P2 potency.
Upon reducing the size of the CAP group to methyl (compound 33), ClpP1P2 activity partially returned (42 M), indicating that a CAP group is strictly required for minimal ClpP1P2, as well as antibacterial activity. Potency was further regained with benzyl (compound 34), suggesting that a bulky CAP is required for ClpP1P2 potency. However, no reduction in proteasome activity at all was achieved with this compound. An increase of 2-fold ClpP1P2 activity was obtained with phenyl analogue 35, compared to the ClpP1P2 activity of pyrazine compound 1. Unfortunately proteasome activity was unchanged.
With a 3-pyridyl CAP group (compound 36) the proteasome activity was reduced by 6-fold, with retention of ClpP1P2 activity, while with a 2-pyridyl (compound 37),  both ClpP1P2 potency and proteasome potency were reduced. These data indicate that the specific position of the nitrogen in the aromatic CAP is important for ClpP1P2 activity and may have a role to play in proteasome activity as well.
More-bulky heterocyclic groups were also screened, with significant SAR findings. 5-Substituted 2-phenylpyrimidine (compound 38) maintained activities against ClpP1P2 and bacterial growth; however, in contrast, the isomer 4-substituted 6-phenylpyrimidine (compound 39) completely lost potency in the target assays but fully retained proteasome potency. We speculated that this might be due to either poor bacterial-cell penetration or poor target binding due to repulsive interactions in the ClpP1P2 active site. However, modeling studies revealed that this compound binds the ClpP1P2 active site with no apparent repulsive interactions. We therefore concluded that poor bacterial-cell penetration (with maintenance of mammaliancell penetration) might be the reason for the poor ClpP1P2 and antibacterial potency. 3-Indolizine (compound 40) retained target potency but was only 5-foldless active for the proteasome than compound 1. However, 2-indolizine (compound 41) was slightly less active against ClpP1P2 and more potent against the proteasome and hence did not offer clues as to a way forward. No change in selectivity or potency was observed with 2-substituted 5-oxazole (compound 42). This compound has a substitution pattern similar to that of compound 39, which was not active. Benzothiazole (compound 43) was quite potent, with a 2-fold increase in ClpP1P2 and bacterial-growth inhibition compared with that of compound 1. Inserting a P2 phenylalanine and maintaining the pyrazine CAP ("extended pyrazine," compound 44) led to complete loss of ClpP1P2 potency and no improvement in proteasome activity. However, 5-fold-reduced proteasome activity was achieved with piperidine urea 45, but with reduced ClpP1P2 and bacterial-growth inhibition. This series of CAP changes does not provided a clear way forward but does offer options of CAP groups for subsequent P1-P2-CAP combinations for the next phase of SAR exploration.
To obtain increased ClpP1P2 potency and reduced activity against the proteasome, combinations of the best-performing P1 groups, such as benzyl, cyclohexylethyl, and phenethyl, and selected CAP groups were next screened (Table 3).
Encouragingly, a 43-fold reduction in proteasome potency was obtained when benzyl P1 and methyl CAP were combined (compound 46). However, poor target potency was achieved with this compound. This confirmed our previous observation that a bulky CAP is strictly required for ClpP1P2 potency. Disappointingly, no activity was detectable with the combination of cyclohexylethyl with ClpP1 and CAP groups, such as benzothiazole (compound 47), phenyl (compound 48), 2-fluorophenyl (com- Moreira et al. Antimicrobial Agents and Chemotherapy pound 49), and a hindered pyrrole analogue (compound 50). We suspected that the LogP (the partition coefficient of a molecule between an aqueous and a lipophilic phase) of these compounds was too high, reducing aqueous solubility. Interestingly, with phenethyl as ClpP1 and phenyl as CAP (compound 51), proteasome activity was reduced by 33-fold compared to that of compound 1 (0.005 M), with an IC 50 of 0.167 M. Furthermore, this compound is also slightly more active against ClpP1P2 and it has greater bacterial-growth inhibition. However, reduced antibacterial activity was observed with 4-(trifluoromethyl) phenyl (compound 52) and 2-fluorophenyl (compound 53), albeit with reduced proteasome activity. Pyrrole analogue (compound 54) exhibited 100-fold-reduced potency against the proteasome but unfortunately was only weakly potent against ClpP1P2 and bacterial growth. A larger CAP group, 5-substituted 2-phenylpyrimidine (compound 55), was 10-fold-less active than compound 1 against the proteasome but maintained ClpP1P2 and bacterial-growth inhibition. With this study of ClpP1-CAP combinations, we obtained compounds whose potencies against the target were retained but whose activities against the proteasome were reduced. We then considered changes at the P2 side chain in efforts to further optimize the series. P1/P2 dual modifications were screened with pyrazine as the CAP group (Table 4). When the P2 side chain was reduced to a methyl (compound 56), the proteasome activity decreased 45-fold to 0.222 M; however, ClpP1P2 inhibition and antibacterial activity significantly dropped as well. As with the CAP SAR from Table 2, this result indicates that a bulky P2 group is required for ClpP1P2 activity and bacterial-growth inhibition. Increasing the hydrophilicity at the P2 position of compound 32 with tyrosine in place of phenylalanine, giving compound 57, resulted in reduced proteasome activity (107-fold-less active than compound 1) and reduced retention of target activity compared to those of compound 1. Introducing a nitrogen into the phenyl ring to give a 2-pyridyl phenylalanine derivative (compound 58) improved potency against ClpP1P2 4-fold, and bacterial-growth inhibition improved just over 2-fold compared to that of compound 1. With 74-fold-reduced proteasome activity (IC 50 ϭ 0.367 M) compared to that of compound 1, compound 58 demonstrates that it is possible to retain anti-ClpP1P2 and antibacterial activity while reducing potency against the human proteasome.
Activity against M. tuberculosis. To confirm activity against M. tuberculosis, we tested a subset of the most promising compounds against M. tuberculosis H37Rv (Table  5). MIC 50 s and MIC 90 s followed the same trend against M. tuberculosis as they did against the M. smegmatis ΔprcAB strain used for SAR studies, with the exception of the MIC 50 of compound 57, 0.8 M, which represents an encouraging 5-fold improvement over results produced with compound 1. The MIC 90 of compound 58 required a Ͻ3-fold-higher concentration (6 M), while the other tested compounds required 3-to 6-fold-higher concentrations. Finally, we tested compound 58 for its bactericidal activity against M. tuberculosis. The MBC 99.9 (minimum bactericidal concentration required to kill 99.9% of the bacterial population, i.e., to induce a 1,000-fold kill) was 50 M against M. tuberculosis, similar to that of compound 1.
Molecular modeling. Docking of compound 58 to the binding site of ClpP1P2 indicates that the hydrophobic S1 residues Ile71, Met75, Met99, Phe102, Pro125, Leu126, and Met150 make close contacts with the P1 phenethyl side chain (Fig. 9). Important hydrogen bonds are formed between the P2 amine and the backbone carbonyl of Leu126 and between the CAP carbonyl and the backbone amine of Ile71. The pyridyl P2 side chain is close to Ser70, and the CAP pyrazine group appears to be orientated in space, suggesting further potential for tuning of molecular properties. Protease selectivity. ClpP1P2 is a serine protease, and as such, any inhibitor of ClpP1P2 has the potential to inhibit other serine proteases or other proteases of other classes, which in turn may lead to off-target effects or even toxicity. To assess the broader protease activity of preferred compound 58, we tested it against a panel of 62 diverse proteases representative of the proteome (Table 6).
Only 2/62 (3%) of the panel were inhibited with a submicromolar IC 50 and 5/62 (8%) with an IC 50 between 1 to 10 M. Both chymotrypsin and proteinase K are serine proteases with a preference for large aromatic or aliphatic P1 groups and were, not surprisingly, the most potently inhibited at IC 50 s of 120 and 350 nM, respectively. Of the others, which were inhibited in the micromolar range, chymase, kallikreins, and proteinase A/K are all serine proteases.
In vitro ADME properties. Compound 58 was selected for further profiling in in vitro absorption, distribution, metabolism, and excretion (ADME) assays (Table 7). Calculated parameters, such as molecular weight and the numbers of hydrogen bond donors and acceptors, are all in the accepted range for an oral small-molecule drug. The calculated LogP of the compounds (cLogP) is on the low side, in agreement with the topological polar surface area (TPSA), which is slightly high but not uncommonly so for antibacterial agents. These data explain the good aqueous solubility (Ͼ0.4 mg/ml at pH 7.4); however, permeability is still in the acceptable range, which is supported by the observed cellular potency. There is clearly space to increase the cLogP with further optimization, which may further improve permeability and maintain solubility in an acceptable range. However, Log D (octanol/PBS partition coefficient measured at pH 7.4) is in a preferred range of 2.96, so any increases in lipophilicity should be minimal in order to maintain the favorable solubility profile. Plasma protein binding (PPB) was determined to be moderate, with little difference between species. Bound levels of 90.46% (mouse) and 89.07% (human) indicate a significant free fraction. To confirm that antibacterial inhibitory potency was preserved in the presence of plasma, the MIC 50 for Mycobacterium bovis BCG, with the addition of 10% fetal bovine serum (FBS), was determined to be 1.8 M for compound 1 and 1.5 M for compound 58. Human liver microsome stability was moderate, with a half-life of just over 24 min. However, high clearance in mouse microsomes was observed, with a half-life of about 8 min. Inhibition of cytochrome P450 enzymes was not detected at the highest concentration tested (10 M), reducing concerns regarding drug-drug interactions, an important concern for an anti-TB drug, which is likely to be used in combination. Cytotoxicity data. The 50% growth inhibition concentration (GI 50 ; a determination of the minimal concentration of a compound that inhibits the growth of the cells by 50%) of compounds 1 and 58 were determined against Vero or HepG2 cells and found to be 250 or 400 M (compound 1) and 500 or 500 M (compound 58), respectively, by an MTS [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-

2H-tetrazolium] assay. This provides for a large therapeutic window in cells and reduces toxicity concerns.
In vivo pharmacokinetics. The pharmacokinetic profile of compound 58 was determined in mice following a single intravenous (i.v.) dose of 10 mg/kg of body weight and single oral doses of 10 and 100 mg/kg (Fig. 10A) Following oral administration at 10 mg/kg in mice, concentrations of compound 58 could be quantified for up to 24 h. The mean maximal plasma concentration (C max ) was 798 ng/ml, observed at 0.5 h after dosing, indicating rapid absorption. The mean exposure as calculated by an AUC from time zero to infinity (AUC 0 -inf ) was 2,873 ng · h/ml, resulting in an absolute oral bioavailability of 14%. Although bioavailability was low, significant exposures were achieved orally, suggesting that higher doses may result in higher oral exposures, assuming dose linearity. The highest tolerated dose that is likely to be employed to maximize efficacy and dose linearity is an important concern for an antibacterial agent; hence, the higher dose of 100 mg/kg was studied with a determination of tissue concentrations in lung and brain (Fig. 10B). At 100 mg/kg, high concentrations of compound 58 were quantifiable for up to 24 h in plasma and lung and for up to 8 h in brain. The C max was 6,906 ng/ml in plasma, 3,374 ng/g in lung, and 93 ng/g in brain. Peak concentrations were observed in a short time (time to maximum concentration of a drug in serum [T max ] ϭ 10 min), indicating rapid absorption. The mean exposure as calculated from the AUC 0 -t(last) was 22,798 ng · h/ml in plasma, 16,253 ng · h/g in lung, and 286 ng · h/g in brain. Plasma concentrations were above the murine M. tuberculosis MIC 50 for about 8 h. Lung concentrations were above the murine M. tuberculosis MIC 50 for about 2 to 3 h but persisted longer than in plasma (half-life [t 1/2 ] ϭ 0.3 h), indicating good penetration of the target organ. Concentrations in brain, although measurable, were much lower than in plasma or lung and not quantifiable after 8 h. Dose linearity was very good, as assessed by comparing the plasma concentrations at 10 and 100 mg/kg for both the C max (Fig. 10C) and the AUC to infinity (AUC 0 -inf ) (Fig. 10D). The maximum tolerated dose was not determined. This preliminary work supports compound 58 as a drug-like template suitable for further optimization.
Concluding remarks. In this work, a series of novel analogues of the potent human-proteasome inhibitor, compound 1, were prepared as mycobacterial caseino-

Boronic Acid Mycobacterial ClpP1P2 Inhibitors
Antimicrobial Agents and Chemotherapy lytic protease ClpP1P2 inhibitors. Compounds were characterized in a ClpP1P2 targetbased cell reporter assay, confirming protease inhibition and, importantly, bacterial-cell penetration. All compounds were also tested in an antibacterial assay using a proteasome knockout strain of M. smegmatis, allowing growth activity SAR to be established without interference from bacterial-proteasome inhibition. In most cases, growth inhibition tracked ClpP1P2 activity, reassuring us that there were no other off-target effects.
A key focus of this study was to gain an understanding of human-proteasome SAR in an effort to identify the first lead compounds with reduced proteasome activity, and hence less toxicity, while retaining ClpP1P2 activity. A preferred compound, compound 58, had low, single-digit, micromolar ClpP1P2 and growth-inhibitory activity while having 74-fold-reduced potency against the human proteasome. The growth-inhibitory potency of compound 58 against M. tuberculosis was gratifyingly in the same lowmicromolar-concentration range, while GI 50 s were 200-fold higher in both the Vero and HepG2 cell lines. Compound 58 was active (Ͻ10 M) against only 11% of a panel of 62 proteases and active (Ͻ1 M) against only 2 proteases (3% of the panel). Compound 58 has favorable in vitro ADME properties for an antibacterial agent, including moderate protein binding, indicated by its MIC not being affected by additional serum. Oral/i.v. pharmacokinetics indicated moderate clearance and low bioavailability. However, oral exposures were linear between 10 and 100 mg/kg with plasma concentrations above the murine M. tuberculosis MIC 50 for up to 8 h at the highest dose tested. This work demonstrates that potent inhibitors of ClpP1P2 are possible with reduced proteasome activity and paves the way for further studies, ultimately leading to new treatment options for drug-resistant M. tuberculosis.

MATERIALS AND METHODS
Chemistry and general experimental details. All reactions requiring anhydrous conditions were carried out under a nitrogen atmosphere using oven-dried glassware (105°C), which was cooled under vacuum. All reaction solvents used, such as dichloromethane, toluene, and diethyl ether, were freshly collected from a solvent purification system (PureSolv MD-4; Innovative Technology, MA). All final compounds (boronic acid) were stored in a Ϫ20°C freezer under nitrogen. Proton nuclear magnetic resonance ( 1 H NMR) and carbon NMR ( 13 C NMR) spectra were recorded in CDCl 3 , CD 3 OD, and dimethyl sulfoxide-d 6 (DMSO-d 6 ), unless otherwise stated. 1 H NMR (400 MHz) and 13 C NMR (101 MHz) with complete proton decoupling were performed on a 400-MHz Bruker Ultra Shield NMR spectrometer. Chemical shifts were reported as ␦ in units of parts per million downfield from tetramethylsilane (␦, 0.00), with the residual solvent signal used as an internal standard, as follows: DMSO (2.50 ppm for 1 H and 39.5 ppm for 13 C NMR), CDCl 3 (7.26 ppm for 1 H and 77.2 ppm for 13 C NMR), and CD 3 OD (3.31 ppm for 1 H and 49.0 ppm for 13 C NMR). Multiplicities were abbreviated as s (singlet), d (doublet), t (triplet), q (quartet), quintet, m (multiplet), dd (doublet of doublets), ddd (doublet of doublet of doublets), dt (doublet of triplets), dq (doublet of quartets), dm (doublet of multiplets), and br (broad). Coupling constants (J) were recorded in hertz. The 13 C signal for the boron-attached carbon was very weak and broad, which was observed by heteronuclear multiple-quantum-coherence (HMQC) NMR. High-resolution electrospray ionization (ESI) mass spectra were obtained on a Bruker micrOTOF-Q II mass spectrometer. All tested compounds were Ͼ95% pure, as determined by reverse-phase high-pressure liquid chromatography (HPLC) on a Shimadzu SPD-20A HPLC system. The test compound was dissolved in methanol (MeOH) and injected through a 100-l loop at a flow rate of 1.0 ml/min, with UV detection at 254 and 220 nm. Separation was carried out on a Zorbax SB-C 18 column (250 mm by 4.6 mm, with a 5-m inside diameter; Agilent). The purity of each compound was assessed from the area of the major peak in comparison with the total area of peaks obtained on the chromatogram. Boronic acids undergo facile dehydration and/or decomposition upon being heated (28). Therefore, melting points of boronic acids were not determined. All commercial reagents were purchased from Sigma-Aldrich, Fluka, Alfa Aesar, Merck, TCI, or Acros and were of the highest purity grade available. They were used without further purification unless specified. Compound 1 was purchased from Selleckchem.
Compound 25. Compound 25 was (R)-(3-methyl-1-(3-phenylpropanamido)butyl)boronic acid. To a round-bottom flask, 2-methylpropylboronic acid (222 mg, 2.17 mmol, 5.0 eq) and 1 N HCl (1.0 ml) were added to a biphasic mixture of ester 19 (150 mg, 0.437 mmol, 1.0 eq) in methanol (3 ml) and pentane (3 ml). The round-bottom flask was closed tightly, and the biphasic mixture was stirred vigorously at room temperature (rt) for 18 h. The reaction mixture was diluted with methanol (50 ml). The methanolic phase was washed with hexane (50 ml three times), and the combined hexane layer was extracted with methanol (50 ml once). The combined methanolic layers were evaporated in vacuo. By using a procedure similar to that used for compound 25, the following boronic acids were synthesized. Note that final isolated yields were generally low due to the required preparative HPLC purification on a small scale.
Pharmacokinetics. Female ICR mice (aged ϳ6 to ϳ8 weeks; 3 animals per time point) were used for all studies. Studies were performed as per approved internal protocols for animal care and use. Doses were administered as clear solutions in 50% polyethylene glycol 400 (PEG400) plus 50% dextrose and 5% water (D5W). Animals were sacrificed by overdose of CO 2 , and blood was collected through cardiac puncture at 5 min (10 min per os [p.o.]), 30 min, and 1, 2, 4, 8, 16, and 24 h after administration in tubes containing K3 EDTA as an anticoagulant. The samples were centrifuged, and the plasma was separated and stored at Ϫ70°C until analysis. Plasma samples were processed and analyzed by LC-tandem MS. Pharmacokinetic parameters were estimated by noncompartmental methods using WinNonlin (version 5.2; Pharsight, CA).