Binding of α,α-Disubstituted Amino Acids to Arginase Suggests New Avenues for Inhibitor Design

Abstract

Arginase is a binuclear manganese metalloenzyme that hydrolyzes L-arginine to form L-ornithine and urea, and aberrant arginase activity is implicated in various diseases such as erectile dysfunction, asthma, atherosclerosis, and cerebral malaria. Accordingly, arginase inhibitors may be therapeutically useful. Continuing our efforts to expand the chemical space of arginase inhibitor design, and inspired by the binding of 2-(difluoromethyl)-L-ornithine to human arginase I, we now report the first study of the binding of α,α-disubstituted amino acids to arginase. Specifically, we report the design, synthesis, and assay of racemic 2-amino-6-borono-2- methylhexanoic acid and racemic 2-amino-6-borono-2-(difluoromethyl)hexanoic acid. X-ray crystal structures of human arginase I and Plasmodium falciparum arginase complexed with these inhibitors reveal the exclusive binding of the L-stereoisomer; the additional α-substituent of each inhibitor is readily accommodated and makes new intermolecular interactions in the outer active site of each enzyme. Therefore, this work highlights a new region of the protein surface that can be targeted for additional affinity interactions, as well as the first comparative structural insights on inhibitor discrimination between a human and a parasitic arginase.

Introduction

Arginase is a ubiquitous manganese metalloenzyme that catalyzes the hydrolysis of L-arginine to form L-ornithine and urea. In mammals, two isozymes have been identified with distinct tissue distributions and subcellular localizations.^1–5 Arginase I is a cytosolic enzyme found predominantly in the liver, and arginase II is a mitochondrial enzyme found at highest concentrations in the kidney.^6–9 However, both isozymes are constitutively expressed or induced in other tissues to regulate cationic amino acid homeostasis in three critical metabolic pathways: (1) the regulation of L-arginine levels for nitric oxide (NO)^a biosynthesis;^10–15 (2) the regulation of L-ornithine levels for L-proline biosynthesis to promote collagen production;^16,17 and (3) the regulation of L-ornithine levels for polyamine biosynthesis to facilitate cellular proliferation.^2,18–20

Significantly, expression of arginase I and/or arginase II is upregulated in certain diseased tissues and cell types. For example, given that arginase competes with NO synthase for their common substrate, L-arginine, aberrant arginase II activity in the penile corpus cavernosum of the diabetic male attenuates NO biosynthesis and therefore compromises the NO-dependent relaxation of cavernosal smooth muscle required for penile erection.²¹ Moreover, arginase I levels in the corpus cavernosum increase with age and similarly compromise penile erection.²² Thus, erectile dysfunction can result from the upregulation of either arginase I or arginase II in different etiologies of the same disease. In another example, arginase I is upregulated in the asthmatic lung by Th2 cytokines such as interleukin-13,^23–27 and arginase I and arginase II single-nucleotide polymorphisms are identified in atopic asthma.^28,29 Arginase activity contributes to asthma pathology through three metabolic functions:^30–32 (1) arginase compromises NO-dependent relaxation of airway smooth muscle, leading to bronchoconstriction; (2) arginase enhances L-proline and collagen biosynthesis, leading to the accumulation of fibrotic tissue in the chronic asthmatic airway; and (3) arginase enhances polyamine biosynthesis, which stimulates cellular proliferation and hyperplasia, e.g., of airway smooth muscle cells, in the asthmatic airway. These disease manifestations can be blocked by the use of arginase inhibitors in ex vivo and in vivo experiments.^{15,21,22,33–35} Thus, the arginase isozymes are increasingly considered as vital pharmaceutical targets for the treatment of erectile dysfunction, asthma, and cardiovascular diseases linked to aberrant arginase activity, such as atherosclerosis.³⁶

The first high-affinity inhibitor of arginase to be reported was the boronic acid analogue of L-arginine, 2-(S)-amino-6-boronohexanoic acid (ABH; Figure 1),³⁷ which binds to human arginase I (HAI) with K_d = 5 nM and human arginase II with K_i = 8.5 nM.^38,39 ABH is also the most potent inhibitor known for Plasmodium falciparum arginase (PFA), to which it binds more weakly with K_d = 11 μM.⁴⁰ The X-ray crystal structures of HAI and rat arginase I complexed with ABH,^15,38 as well as the structure of the PFA-ABH complex,⁴⁰ reveal that the boronic acid side chain undergoes nucleophilic attack by the metal-bridging hydroxide ion observed in the unliganded enzyme^41,42 to yield a tetrahedral boronate anion that mimics the tetrahedral intermediate and its flanking transition states in catalysis. The binding of the ABH analogue S-(2-boronoethyl)-L-cysteine (BEC; Figure 1) to HAI and human arginase II occurs through an identical mechanism.^33,38

Figure 1.

The amino acid substrate and product of the reaction catalyzed by arginase, L-arginine and L-ornithine, respectively, are compared with DFMO and the boronic acid inhibitors BEC, ABH, MABH, and FABH.

Structural comparisons of arginase-ABH and -BEC complexes reveal a conserved hydrogen bond network responsible for the molecular recognition of the inhibitor α-amino and α-carboxylate groups; these interactions contribute significantly to enzyme-inhibitor affinity, as demonstrated in recent mutagenesis studies with rat arginase I.⁴³ Given that ABH analogues in which the side chain is modified exhibit radically diminished affinity,^37,44,45 there is only one remaining possibility for the derivatization of the ABH scaffold to yield new inhibitors: the amino acid Cα-H atom, which in the enzyme-inhibitor complex is solvent exposed and oriented toward a region of the protein surface that is currently “uncharted” with regard to inhibitor binding. The design and synthesis of α,α-disubstituted amino acids based on the high-affinity ABH scaffold (Figure 1) may thus enable the generation and development of a new class of arginase inhibitors to expedite the search for new arginase-directed therapies. Notably, such therapies may target indications beyond those outlined above. For instance, the inhibition of PFA has been recently proposed as a potential adjuvant therapy for accelerating the recovery of malaria patients.^40,46

We now report the first study of the binding of α,α-disubstituted amino acids to arginase. We demonstrate initial proof-of-principle with the X-ray crystal structures of HAI complexed with 2-(difluoromethyl)-L-ornithine (DFMO) and L-ornithine to show that an additional α-substituent does not perturb the intermolecular interactions of the amino acid product of the arginase reaction. We then describe the design, synthesis, and assay of two new α,α-disubstituted derivatives of ABH: 2-amino-6-borono-2-methylhexanoic acid (MABH) and 2-amino-6-borono-2-(difluoromethyl)hexanoic acid (FABH) (Figure 1). X-ray crystal structures of their complexes with HAI and PFA, together with in vitro evaluations of inhibitory potency, illuminate new avenues toward the development of α,α-disubstituted amino acids as arginase inhibitors.

Results

HAI DFMO Complex

While DFMO is reported⁴⁷ to be a weak inhibitor of arginase activity in human colon carcinoma cells with K_i = 3.9 mM, and while racemic DFMO was used in crystal soaking experiments with HAI, the omit map in Figure 2a clearly shows that the L-stereoisoimer of DFMO is bound exclusively will full occupancy. This reflects the stereoselectivity of the HAI active site for the binding of L-amino acids, e.g., L-arginine is a substrate for arginase whereas D-arginine is not.⁴⁸ Inhibitor binding does not cause any significant conformational changes in the active site or elsewhere in the protein structure, and the r.m.s. deviation is 0.26 Å for 314 Cα atoms in comparison with the unliganded enzyme. As for DFMO, the catalytic product L-ornithine binds in the HAI active site without causing any significant conformational changes. The r.m.s. deviation is 0.22 Å for 314 Cα atoms in comparison with the unliganded enzyme. An omit map showing the binding of L-ornithine is found in Figure 2b. Comparison of the two complexes reveals a generally similar binding mode for DFMO and L-ornithine (Figure 2c). Direct and water-mediated hydrogen bonds with the α-amino and α-carboxylate groups are identical in each complex. These interactions are crucial affinity determinants⁴³ and α,α-disubstitution clearly does not perturb them.

Figure 2.

(a) Simulated annealing omit electron density map of the HAI0DFMO complex (cyan, contoured at 3.0σ), in which DFMO was omitted from the structure factor calculation. Atoms are color-coded as follows: C = yellow, O = red, N = blue, F = black; Mn²⁺ ions and water molecules are purple and red spheres, respectively. Metal coordination and hydrogen bond interactions are indicated by red and green dashed lines, respectively. (b) Simulated annealing omit electron density map of the HAI-L-ornithine complex (cyan, contoured at 3.3σ). Atoms are color-coded as in (a). (c) Superposition of the HAI-DFMO complex (red) and the HAI-L-ornithine complex (blue).

Chemistry

Since the active site of HAI readily accommodated the additional α-substituent of DFMO without perturbing the binding interactions or conformation observed for L-ornithine, we synthesized the corresponding α,α-disubstituted derivatives of the archetype HAI inhibitor, ABH (MABH and FABH, see Figure 1). As shown in Scheme 1, the key intermediate in the syntheses of MABH and FABH was ethyl 2-N-(diphenylmethyleneamino)hex-5-enoate (2). While this compound was previously reported,⁴⁹ we developed an alternative synthetic pathway with a comparable overall yield. We synthesized 2 from the commercially available N-(diphenylmethylene)glycine ethyl ester (1), using the classic mild deprotonation with potassium tert-butoxide (tBuOK),^50,51 followed by alkylation with 4-bromo-1-butene. The use of benzophenone imines of glycine alkyl esters as precursors of α-amino acids is a well-known⁵² synthetic strategy to ensure selective monoalkylation of the amino acid Cα atom. We employed the same mild base (tBuOK) to further deprotonate the Cα in the resulting derivative 2. Subsequent one-pot reaction with methyl iodide and chlorodifluoromethane,⁵³ respectively, yielded the corresponding methyl- (3) and difluoromethyl- (4) derivatives in good yields, as new compounds (Scheme 1). Hydroboration with pinacolborane catalyzed by complexes of cyclooctadiene iridium chloride dimer, [Ir(cod)Cl]₂, with 1,1-bis(diphenylphosphino)methane (dppm)⁵⁴ introduced the synthetic equivalent of the boronic synthon in the terminal olefins 3 and 4, generating correspondingly the intermediates 5 and 6. Complete deprotection of derivatives 5 and 6 with 6N HCl in tetrahydrofuran (THF) led to the targeted compounds MABH (7) and FABH (8), respectively.

Scheme 1.

Synthesis of 2-amino-6-borono-2-methylhexanoic acid (MABH) and 2-amino-6-borono-2-difluoromethylhexanoic acid (FABH).

Binding Affinities of MABH and FABH

Dissociation constants for the S-isomer of ABH, and racemic MABH and FABH, against HAI determined by surface plasmon resonance (Figure 3), and inhibition constants against PFA determined by kinetic assay, are recorded in Table 1. Curiously, both α,α-disubstituted derivatives of ABH exhibit weaker affinity compared with ABH. For HAI, MABH and FABH bind 49-fold and 1889-fold less tightly, respectively; for PFA, MABH and FABH exhibit inhibitory potencies based on K_i that are 26-fold and 200-fold weaker than that observed for ABH. The structural basis for affinity loss is not immediately evident for the binding of α,α-disubstituted amino acids to HAI based on crystal structures of enzyme-inhibitor complexes; however, that only the L-stereoisomer of each inhibitor binds indicates that the affinity losses are only one-half those measured for the racemic mixture (vide infra). Even so, it is quite possible that affinity can be recaptured and enhanced with longer α-substituents capable of making favorable interactions in the active site. In contrast with HAI-inhibitor complexes, crystal structures of PFA-inhibitor complexes reveal intersubunit interactions with the α-substituents of MABH and FABH.

Figure 3.

Surface plasmon resonance sensorgrams for HAI showing the binding of: (a) ABH, K_d = 18 ± 1 nM; (b) MABH, K_d = 0.88 ± 0.02 μM; and (c) FABH, K_d = 34 ± 2 μM.

Table 1.

α,α-Disubstituted Amino Acids Synthesized as Inhibitors of Human Arginase I (HAI) and Plasmodium falciparum Arginase (PFA)

Inhibitor	HAI Kd (μM)a	PFA Kt (μM)b
ABH (S-isomer)	0.018 ± 0.001	10 ± 1c
MABH (racemic)d	0.88 ± 0.02	260 ± 20
FABH (racemic)d	34 ± 2	2000 ± 200

^aSurface plasmon resonance.

^bKinetic colorimetric assay. Errors are standard deviations of experiments run in triplicate.

^cFrom reference 40.

^dSince only the L-stereoisomer is observed to bind in each crystal structure, the actual K_d or K_i for the inhibitory stereoisomer is likely to be one-half that measured for the racemic mixture.

Crystal Structures of HAI-MABH and HAI-FABH Complexes

The 1.60 Å resolution crystal structure of the HAI-MABH complex reveals that inhibitor binding does not cause any significant conformational changes in the active site or elsewhere in the protein structure. The r.m.s. deviation is 0.24 Å for 313 Cα atoms in comparison with the HAI-ABH complex. The electron density of MABH is well defined and clearly indicates that the L-stereoisomer of racemic MABH used in cocrystallization experiments is exclusively selected for binding in the HAI active site (Figure 4a). Superposition with the 1.29 Å resolution crystal structure of human arginase I-ABH complex shows excellent overlap (Figure 4c), and enzyme-inhibitor hydrogen bonds are identical in both complexes.

Figure 4.

(a) Simulated annealing omit electron density map of the HAI-MABH complex (cyan, contoured at 3.8σ), in which MABH was omitted from the structure factor calculation. Atoms are color-coded as follows: C = yellow, O = red, N = blue, B = green; Mn²⁺ ions and water molecules are purple and red spheres, respectively. Metal coordination and hydrogen bond interactions are indicated by red and green dashed lines, respectively. (b) Simulated annealing omit electron density map of the HAI-FABH complex (cyan, contoured at 3.0σ), in which FABH was omitted from the structure factor calculation. Atoms are color-coded as in (a), with F = black. (c) Superposition of the HAI-ABH complex (blue), HAI-MABH complex (red), and HAI-FABH complex (green).

The 1.7 Å resolution electron density map of the HAI-FABH complex similarly reveals the exclusive binding of the L-stereoisomer of racemic FABH used in cocrystallization experiments (Figure 4b). Inhibitor binding does not cause any significant overall conformational changes in the HAI structure, and the r.m.s. deviation for 313 Cα atoms is 0.29 Å when compared with the structure of the HAI-ABH complex. However, each fluorine atom of the α-difluoromethyl group accepts hydrogen bonds from solvent molecules, which in turn donate hydrogen bonds to D183; additionally, it appears that T136 undergoes a 128° conformational change about side chain torsion angle χ₁ to accommodate the binding of one of these solvent molecules (Figure 4c). Enzyme-inhibitor hydrogen bond interactions are otherwise identical to those observed in the HAI-ABH and HAI-MABH complexes.

Crystal Structures of PFA-MABH and PFA-FABH Complexes

The 1.9 Å resolution crystal structure of the PFA-MABH complex and the 2.0 Å resolution crystal structure of the PFA-FABH complex (Figure 5) each reveal the exclusive binding of the L-stereoisomer from the racemic inhibitor mixtures used in cocrystallization experiments. Surprisingly, however, each inhibitor binds to PFA with an intact boronic acid side chain, i.e., neither inhibitor binds in the tetrahedral boronate anion form as observed in the PFA-ABH complex,⁴⁰ or the HAI-MABH or HAI-FABH complexes (Figure 4). Attempted refinement with the tetrahedral boronic acid form of each inhibitor resulted in negative electron density peaks on both the boron atom and the adjacent carbon atom. Such spurious peaks are not observed when each inhibitor is refined with an intact boronic acid moiety. Clear electron density for each inhibitor shows that a planar boronic acid moiety binds so as to preclude the binding of the metal-bridging solvent molecule typically observed in unliganded arginase structures.^41,42 Electron density for a water molecule coordinated to Mn_A²⁺ is ~2 Å from the boron atom of MABH and FABH (e.g., see Figure 5b). Thus, MABH and FABH do not bind as transition state analogues. Moreover, neither inhibitor binds as a true substrate analogue, since the trigonal planar boronic acid moiety of each inhibitor adopts a binding conformation with a Cδ-Cε-Bζ-Oη₁ dihedral angle of ~90°, which is very different from the corresponding Cδ-Nε-Cζ-Nη₁ dihedral angle of 180° required by the planar guanidinium moiety of L-arginine. This structural aberration may contribute to the weaker affinity of MABH and FABH compared to ABH.

Figure 5.

(a) Simulated annealing omit electron density map of the PFA-MABH complex (cyan, contoured at 3.5σ), in which MABH was omitted from the structure factor calculation. Atoms are color-coded as follows: C = yellow (C = gray for H381 from the adjacent monomer), O = red, N = blue, B = green; Mn²⁺ ions and water molecules are purple and red spheres, respectively. Red and green dashed lines indicate manganese coordination and hydrogen bond interactions, respectively. (b) Simulated annealing omit electron density map of the PFA-MABH complex (cyan, contoured at 3.5σ), in which MABH and the Mn²⁺_A-bound water molecule were omitted from the structure factor calculation. The view is zoomed in on the binuclear manganese cluster and oriented differently from that in (a) to clearly show that the electron density is most consistent with the binding of a trigonal planar boronic acid moiety and a separate Mn²⁺_A-bound water molecule. Similar electron density characterizes the PFA-FABH complex (data not shown). Atoms are color-coded as in (a). (c) Simulated annealing omit electron density map of the PFA-FABH complex (cyan, contoured at 3.5σ), in which FABH was omitted from the structure factor calculation. Atoms are color-coded as in (a), with F = black. (d) Superposition of the PFA-ABH complex (blue), PFA-MABH complex (red), and PFA-FABH complex (green).

Inhibitor binding to PFA generally does not cause any significant conformational changes in the protein structure. However, active site solvent structure is slightly different, in that the binding of α,α-disubstituted amino acids displaces a water molecule that is observed in the PFA-ABH complex (Figure 5d). In comparison with the PFA-ABH complex, the r.m.s. deviations are 0.10 Å for 308 Cα atoms and 0.12 Å for 308 Cα atoms for the PFA-MABH and PFA-FABH complexes, respectively. However, some local structural changes accommodate the binding of α,α-disubstituted amino acids. In particular, an alternate conformation with 50% occupancy is observed for D272 in the PFA-MABH complex; this residue hydrogen bonds with H381 in the PFA-ABH complex. The α-methyl group of MABH is 3.0 Å from the Nε2 atom of H381, and H381 rotates ~5° about side chain torsion angle χ₁, so this van der Waals contact may destabilize the H381-D272 interaction. Higher thermal B factors are observed for this portion of the inhibitor and the side chain of H381, suggesting that there is some residual disorder in this region of the enzyme-inhibitor complex.

Inhibitor binding conformations are very similar in PFA complexes with MABH and FABH (Figure 5d). Interestingly, the α-difluoromethyl group of FABH does not trigger the conformational change of D272 as observed in the PFA-MABH complex. H381 undergoes a slightly greater conformational change (~12°) than that observed in the PFA-MABH complex, perhaps due to a hydrogen bond interaction with one of the fluorine atoms of FABH (F---N separation = 2.6 Å). However, as observed in the PFA-MABH complex, higher thermal B factors characterize this portion of the inhibitor and the side chain of H381, suggestive of some residual disorder in this region of the enzyme-inhibitor complex. Such residual disorder may compromise enzyme-inhibitor affinity.

Discussion

In the current study, we sought to determine the structural characteristics and inhibitory properties of the most potent arginase inhibitor known to date, ABH, when the amino acid Cα-H atom is substituted with alkyl groups. Based on the structure of the HAI-DFMO complex, we selected the simplest alkyl substituents, CH₃ and CHF₂, to initiate this study. As discussed in detail in a recent review,⁵⁵ the replacement of hydrogen by fluorine is a commonly used strategy in medicinal chemistry to increase the lipophilicity and metabolic resistance of organic molecules. The introduction of fluorine can confer greater pharmacokinetic and pharmacodynamic stability, and can also influence binding affinity in that the C–F group of a fluorinated inhibitor in an enzyme-inhibitor complex can be a weak hydrogen bond acceptor.

The structures of HAI and PFA complexed with the corresponding inhibitors, MABH and FABH, reveal that the new α-substituent of each inhibitor is directed out toward a new region of the active site not previously identified to interact with bound inhibitors. In HAI, this region of the active site is in the vicinity of T136, so we now refer to this previously “uncharted” region of the protein surface as the “T136 region”. Importantly, the current work is the first to demonstrate that the arginase active site can accommodate the binding of an α,α-disubstituted amino acid without compromising the molecular recognition of the α-amino and α-carboxylate groups, critical determinants of enzyme-inhibitor affinity.⁴³ Both C–F groups of FABH accept hydrogen bonds from active site solvent molecules in the HAI-FABH complex, but these interactions presumably do not contribute to enzyme-inhibitor binding affinity.

The residue corresponding to T136 of HAI in PFA is P228, so the T136 region of HAI and the corresponding “P228 region” of PFA differ in their three-dimensional contours and polarities. These regions additionally differ due to the proximity of H381 and its associated polypeptide chain in the adjacent monomer, which constricts the P228 region of PFA relative to the T136 region of HAI. If an α,α-disubstituted amino acid inhibitor can ultimately be designed to interact with the P228 region of PFA, it may be possible to develop a species-specific inhibitor of PFA that does not cross-react with HAI. Although a C–F group of FABH accepts a hydrogen bond from H381 in the PFA-FABH complex, this interaction does not enhance inhibitor binding affinity.

It is surprising that while the α-amino and α-carboxylate groups of MABH and FABH bind similarly within the active sites of HAI and PFA, both exhibit decreased inhibitory potency in comparison with ABH (Table 1). Moreover, it is particularly surprising that while each inhibitor binds with a tetrahedral boronate anion side chain to HAI, each inhibitor binds to PFA with a trigonal planar boronic acid side chain. However, neither MABH nor FABH bind to PFA as substrate analogues due to a side chain conformation that would be unattainable for substrate L-arginine. This structural feature, as well as interactions between the newly introduced α-substituent and H381 in an adjacent subunit, presumably contribute to the aberrant inhibitor binding modes and lower affinities with PFA compared with HAI. While previous studies⁵⁶ suggest that PFA may have a significant role in immune evasion and infection by the parasite P. falciparum, and PFA inhibition is suggested⁴⁶ as an adjuvant therapy in the treatment of malaria patients, the current work demonstrates that boronic acid analogues of ABH are not particularly effective against PFA.

The decreased affinity of MABH and FABH in comparison with the parent inhibitor ABH may be rationalized by analysis of the crystal structures of their complexes with HAI. Here, all three inhibitors clearly bind as analogues of the tetrahedral intermediate and its flanking transition state, and this feature significantly contributes to high affinity. Given the presence of potential hydrogen bonding groups in the T136 region of the HAI active site, it appears that a more polar substituent would be more favorably accommodated in this region of the active site. Additionally, a longer α-substituent might be better suited for capturing additional binding interactions in the protein landscape of T136 and beyond. Indeed, a recent report demonstrates the success of this particular strategy, with a variety of α-substituents such as –(CH₂)₄NH₃⁺, –(CH₂)₃OPh, or –(CH₂)₃OPh-p-F yielding racemic ABH derivatives with K_i values ranging 10⁻⁸ - 10⁻¹⁰ M against both HAI and human arginase II.⁵⁷ Future studies will continue the exploration of structure-affinity relationships for α,α-disubstituted amino acid inhibitors bearing substituents capable of making favorable interactions in the T136 region, and such studies clearly promise to advance the development of arginase inhibitors as possible therapeutic agents.

Conclusions

This work reports the design, multistep synthesis, assay, and structural evaluation of two α,α-disubstituted amino acids, MABH and FABH, in complex with HAI and PFA. This is the first study describing the binding of α,α-disubstituted amino acids to human and parasitic arginases, and X-ray crystal structures show that the additional α-substituent is readily accommodated in each enzyme active site. These structures highlight new regions of the protein surfaces of HAI and PFA that can be targeted for additional affinity interactions. Finally, this work provides the first comparative structural insights on inhibitor discrimination between a human arginase and a parasitic arginase, which may facilitate the development of species-specific inhibitors in the search for new arginase-directed drugs.

Experimental

Synthesis of α,α-Disubstituted Amino Acids. General Procedures

All reagents were of at least 95% purity, purchased from Sigma Aldrich Co. and Fisher Scientific, and used as received. All solvents were of HPLC grade (Fisher Scientific or Sigma Aldrich Co). For anhydrous conditions, solvents were freshly distilled under N₂ (CH₂Cl₂ from P₂O₅, and THF from Na/benzophenone). Reactions were monitored by TLC with Sigma-Aldrich aluminum plates (silica gel F₂₅₄, 60 Å-0.250mm), visualized by quenching under UV light, equilibration in a glass chamber containing iodine, and/or stained with ninhydrin solution. Flash column chromatography was performed using Fisher Scientific silica gel 60 (230–400 mesh). High resolution mass spectrometry (HRMS) was carried out using an instrument from LCT Premier XE Micromass/Waters MS Technologies. Purities of all synthesized and tested compounds were greater than 95% based on HPLC analysis. ¹H and ¹³C NMR spectra were recorded on Bruker DMX 360 and DRX 500 spectrometers at 360 and 500 MHz for ¹H; 90.6 and 125.6 MHz, respectively for ¹³C; 282 MHz for ¹⁹F; and 128 MHz for ¹¹B NMR. Assignments were made based on chemical shifts, signal intensity, COSY, and HMQC sequences. ¹H, ¹³C, ¹⁹F, and ¹¹B NMR chemical shifts (δ) are reported in ppm relative to the residual solvent peaks. ¹H NMR coupling constants (J) are reported in Hz, and multiplicities are denoted as follows: s, singlet; d, doublet; t, triplet; m, multiplet; bs, broad singlet. ¹¹B NMR spectra are decoupled.

Ethyl 2-N-(diphenylmethyleneamino)hex-5-enoate (2)

A 10% solution of tBuOK (0.9 g, 8.25 mmol) in dry THF was added dropwise to a 10% THF solution of N-(diphenylmethylene)glycine ethyl ester (1) (2 g, 7.5 mmol) at −78 °C, with stirring under N₂. After 15 min, 4-bromobutene (2.4 mL, 22.5 mmol) in dry THF (5 mL) was added at −78 °C. After warming to room temperature, the reaction mixture was stirred for 20 hrs and then quenched with NH₄Cl (20 mL) and water (5 mL). The resulting immiscible layers were separated and the aqueous layer extracted with CH₂Cl₂ (4 × 20 mL). The combined organic extracts were washed (brine), dried (Na₂SO₄), and rotoevaporated. Purification by flash column chromatography (hexane/ethyl acetate gradients) afforded 2 as a light-yellow oil (1.8 g, 75 %). ¹H and ¹³C NMR (CDCl₃) spectra confirmed literature data.⁴⁹

Ethyl 2-N-(diphenylmethyleneamino)-2-methylhex-5-enoate (3)

A solution of 2 (1.8 g, 5.6 mmol) in dry THF (20 mL) was treated successively with 10% THF solutions of tBuOK and methyl iodide, respectively (molar ratio 2:tBuOK:CH₃I 1:1.1:3). Subsequent work-up and flash column chromatography (hexane/ethyl acetate gradients) yielded pure 3 as a colorless oil (1.5 g, 80%). ¹H NMR (CDCl₃): δ 7.63-7.54 (m, 2H, Ph), 7.39-7.22 (m, 6H, Ph), 7.20-7.05 (m, 2H, Ph), 5.94-5.83 (m, 1H, -HC=CH₂), 5.04 (dq, J = 1.6, 17.1, 1H, H_A, -HC=CH_AH_B), 4.98 (dd, J = 1.9, 10.2, 1H, H_B, -HC=CH_AH_B), 3.85-3.60 (m, 2H, -O-CH₂-), 2.40-2.25 (m, 1H, H_A, (CH₃)C-CH_AH_B-CH₂-), 2.22-1.94 (m, 3H, H_B, H_C, H_D, (CH₃)C-CH_AH_B-CH_CH_D-), 1.41 (s, 3H, (CH₃)C-), 1.12 (t, J = 7.1, 3H, -CH₂-CH₃). ¹³C NMR (CDCl₃): δ 175 (CO), 166.8 (C=N), 141.5 (C_q, C_1-Ph), 139 (-CH=CH₂), 137.7 (C_q, C_1'-Ph), 130.3, 128.9, 128.8, 128.7, 128.2, 128.0 (all from Ph), 114.7 (-CH=CH₂), 66.3 (O-C), 60.6 (N-C(CH₃)), 42.6 ((CH₃)C-CH₂-), 28.8 ((CH₃)C-), 24.6 ((CH₃)C-CH₂-CH₂-), 14.1 (-CH₂-CH₃). HRMS m/z 336.1960 (calcd for M+H, 336.1963).

Ethyl 2-N-(diphenylmethyleneamino)-2-methyl-6-(4,4,5,5-tetramethyl-[1,3,2]dioxaborolan-2-yl)-hex-5-enoate (5)

The catalyst, a mixture of [Ir(cod)Cl]₂ (150 mg, 0.2 mmol) and dppm (172 mg, 0.4 mmol), was weighed in the glovebox and dissolved with stirring in dry CH₂Cl₂ (10 mL) under N₂ at room temperature. Pinacolborane (2.6 mL, 18 mmol), and then a solution of 3 (1.5 g, 4.5 mmol) in dry CH₂Cl₂ (5 mL), were added. ¹H NMR monitoring showed the disappearance of signals for the olefinic protons after 18 hrs, and the reaction mixture was then quenched with water (5 mL). After separation, the aqueous layer was extracted with Et₂O (4 × 10 mL). The combined organic layers were washed with water (5 mL), dried (Na₂SO₄), and rotoevaporated. The resulting crude orange oil was further purified by flash column chromatography (hexane/ethyl acetate gradients) and yielded 5 as a colorless oil (1.2 g, 60%). ¹H NMR (CDCl₃): δ 7.60-7.51 (m, 2H, Ph), 7.40-7.22 (m, 6H, Ph), 7.18-7.08 (m, 2H, Ph), 3.78-3.55 (m, 2H, -O-CH₂-), 2.05-1.80 (m, 2H, (CH₃)C-CH₂0), 1.52-1.45 (m, 3H, H_B, H_C, H_D, (CH₃)C-CH₂-CH_AH_B-CH_CH_D-), 1.42-1.35 (bs, 3H, (CH₃)C-), 1.25-1.15 (m, 1H, H_A, (CH₃)C-CH₂-CH_AH_B-CH_CH_D-), 1.23 (s, 12H, 4CH₃ from pinacolboranyl), 1.1 (t, J = 7.1, 3H, CH₂-CH₃), 0.81 (t, J = 7.5, 2H, -CH₂B). ¹³C NMR (CDCl₃): δ 174.4 (CO), 166.2 (C=N), 141.6 (C_q, C_1-Ph), 137.5 (C_q, C_1'-Ph), 130.2, 129.3, 129.0, 128.9, 128.3, 128.0 (all from Ph), 83.2 (2C, C_q from pinacolboranyl), 66.8 (O-C), 60.5 (N-C(CH₃)), 43.2 ((CH₃)C-CH₂), 27 ((CH₃)C-), 25.2 (4C, 4CH₃ from pinacolboranyl), 24.9, 24.6 (C(CH₃)-CH₂-CH₂-CH₂-), 14.2 (-CH₂-CH₃), 11.5 (bs, CH₂-B). ¹¹B NMR (CDCl₃): δ 38.2. HRMS m/z 464.2979 (calcd for M+H, 464.2972).

2-Amino-6-borono-2-methylhexanoic acid (MABH, 7)

The protected derivative 5 (1.2 g, 2.6 mmol) was stirred with 6N HCl:THF 2:1 (4 hrs, 70 °C), cooled down to room temperature, and extracted with Et₂O (4 × 25 mL) to remove the benzophenone. The aqueous layer was rotoevaporated to dryness, retaken in 6N HCl (50 mL), and stirred at 80–90 °C until TLC monitoring (MeOH:CHCl₃:NH₄OH 8:2:0.5) revealed complete deprotection (after ~24 hrs). Rotoevaporation at 40 °C, followed by flash column chromatography (CHCl₃/MeOH/NH₄OH gradients) afforded 7 as white powder (0.46 g, 93%). ¹H NMR (D₂O): δ 1.94-1.87 (m, 1H, H_A, (CH₃)C-CH_AH_B-), 1.80-1.65 (m, 1H, H_B, (CH₃)C-CH_AH_B0), 1.47 (s, 3H, (CH₃)C-), 1.48-1.32 (m, 3H, H_B, H_C, H_D, (CH₃)C-CH₂-CH_AH_B-CH_CH_D-), 1.29-1.13 (m, 1H, (CH₃)C-CH₂-CH_AH_B-CH_CH_D-), 0.79 (t, J = 7.5, 2H, -CH₂B). ¹³C NMR (D₂O): δ 177 (CO), 61.6 (N-C(CH₃)), 36.9 ((CH₃)C-CH₂), 25.9 ((CH₃)C-0), 23.5, 22.5 ((CH₃)C-CH₂-CH₂-CH₂0), 13.9 (bs, CH₂-B). ¹¹B NMR (D₂O): δ 36.25. HRMS m/z 190.1258 (calcd for M+H, 190.1250).

Ethyl 2-N-(diphenylmethyleneamino)-2-(difluoromethyl)hex-5-enoate (4)

A solution of 2 (1.8 g, 5.6 mmol) in dry THF (10 mL) was cooled to −78 °C and treated with 20% tBuOK/THF (molar ratio 2:tBuOK 1:1.1), under stirring and N₂ purging. After 20 min, chlorodifluoromethane (4.8 g, 55.5 mmol) was added to the reaction flask via needle. The resulting mixture was stirred at −78 °C (10 min), then at room temperature (45 min), quenched with brine (5 mL), and extracted with Et₂O (4 × 15 mL). The combined organic extracts were washed (brine), dried (Na₂SO₄), and rotoevaporated. The resulting yellow oil was purified (flash column chromatography; hexane/ethyl acetate gradients) to generate 4 as colorless oil (1.4 g, 70%). ¹H NMR (CDCl₃): δ 7.69-7.05 (m, 10H), 6.17 (t, J = 55, 1H, -CHF₂), 5.91-5.70 (m, 1H, -HC=CH₂), 5.04 (dq, J = 1.6, 17.1, 1H, H_A, -HC=CH_AH_B), 4.95 (dd, J = 1.7, 10.2, 1H, H_B, -HC=CH_AH_B), 3.85-3.63 (m, 2H, -O-CH₂-), 2.45-2.20 (m, 2H, H_A, H_B, (CHF₂)C-CH_AH_B-), 2.18-2.09 (m, 1H, H_C, (CHF₂)C-CH₂-CH_CH_D), 2.06-1.95 (m, 1H, H_D, (CHF₂)C-CH₂-CH_CH_D), 1.12 (t, J = 7.1, 3H, -CH₂-CH₂). ¹³C NMR (CDCl₃): δ 176.6 (CO), 169.8 (C=N), 140.3 (C_q, C_1-Ph), 138.5 (-CH=CH₂), 136.7 (C_q, C_1'-Ph), 130.9, 130.3, 129.2, 129.0, 128.5, 128.31, 128.27, 128.1 (all from Ph), 117.6 (t, ¹J_CF = 250, CHF2), 114.9 (-CH=CH₂), 71.1 (bs, C(CHF₂)-N), 61.4 (O-C), 32.0 ((CHF₂)C-CH₂), 28.3 ((CHF₂)C-CH₂-CH₂-), 13.8 (-CH₂-CH₃). ¹⁹F NMR (CDCl₃): δ 0129.0 (ABX system 8 lines, ²J_FF = 271, ²J_HF = 56). HRMS m/z 372.1764 (calcd for M+H, 372.1775).

Ethyl 2-N-(diphenylmethyleneamino)-2-difluoromethyl-6-(4,4,5,5 tetramethyl-[1,3,2]dioxaborolan-2-yl)-hex-5-enoate (6)

A solution of 4 (1.4 g, 4.2 mmol) in dry CH₂Cl₂ (5 mL) was reacted with a mixture of [Ir(cod)Cl]₂/dppm and pinacolborane following the same experimental procedure as described for the synthesis of 5. Flash column chromatography (pentane/Et₂O gradients) generated 6 as colorless oil (0.95 g, 45%). ¹H NMR (CDCl₃): δ 7.65-7.11 (m, 10H), 6.16 (t, J = 56, 1H, -CHF₂), 3.81-3.65 (m, 2H, -O-CH₂-), 2.1-1.80 (m, 2H, H_A, H_B, (CHF₂)C-CH_AH_B-), 1.60-1.38 (m, 4H, H_C, H_D, H_E, H_F, (CHF₂)C-CH₂-CH_CH_D-CH_EH_F-), 1.18 (s, 6H, 2CH₃ from pinacolboranyl), 1.16 (s, 6H, 2CH₃ from pinacolboranyl), 1.1 (t, J = 7.1, 3H, CH₂-CH), 0.85–0.70 (m, 2H, -CH₂B). ¹³C NMR (CDCl₃): δ 170.0 (CO), 169.5 (C=N), 140.3 (C_q, C_1-Ph), 136.8 (C_q, C_1'-Ph), 130.8, 129.1, 129.0, 128.4, 128.2, 128.1 (all from Ph), 83.1 (2C, C_q from pinacolboranyl), 71.1 (t, ³J_CF = 21, (CHF₂)C-), 61.3 (O-C), 32.9 ((CHF₂)C-CH₂), 26.4 ((CHF₂)C-CH₂-CH₂-), 25.1 ((CHF₂)C-CH₂-CH₂-CH₂-), 25.0 (2CH₃ from pinacolboranyl), 24.9 (2CH₃ from pinacolboranyl), 13.9 (-CH₂-CH₃), 11.3 (CH₂-B). ¹¹B NMR (CDCl₃): δ 32.5. ¹⁹F NMR (CDCl₃): δ 0129.3 (ABX system 8 lines, ²J_FF = 274, ²J_FF = 56.5). HRMS m/z 500.2772 (calcd for M+H, 500.2783).

2-Amino-6-borono-2-(difluoromethyl)hexanoic acid (FABH, 8)

Intermediate 6 (0.9 g, 1.8 mmol) was deprotected by the same experimental protocol utilized for the deprotection of 5. Flash column chromatography (CHCl₃/MeOH/NH₄OH/iPrOH gradients) afforded 8 as a white precipitate (0.37 g, 92%). ¹H NMR (D₂O): δ 6.19 (t, J = 54, 1H, -CHF2), 1.96-1.88 (m, 1H, H_A, (CHF₂)C-CH_AH_B), 1.78-1.68 (m, 1H, H_B, (CHF₂)C-CH_AH_B), 1.53-1.30 (m, 3H, H_B, H_C, H_D, (CHF₂)C-CH₂-CH_AH_B-CH_CH_D), 1.30-1.08 (m, 1H, (CHF₂)C-CH₂-CH_AH_B-CH_CH_D-), 0.76 (t, J = 7.5, 2H, -CH₂B). ¹³C NMR (D₂O): δ 171.5 (d, ⁴J_CF = 6.1, CO), 116.0 (t, ¹J_CF = 246, CHF₂), 65.5 (dd, ³J = 16.5, 20, (CHF₂)C-), 31.2 (d, ⁴J_CF = 3.2, (CHF₂)C-CH₂), 25.1 ((CHF₂)C-CH₂-CH₂-), 23.4 ((CHF₂)C-CH₂-CH₂-CH₂-), 13.7 (bs, CH₂-B). ¹¹B NMR (D₂O): δ 21.6. ¹⁹F NMR (D₂O): δ −130.0 (ABX system 8 lines, ²J_FF = 278, ²J_HF = 54). HRMS m/z 226.1057 (calcd forM+H, 226.1062).

Surface Plasmon Resonance (SPR)

The binding affinities of MABH and FABH to HAI were determined by SPR on a Biacore 3000 instrument according to a previously reported procedure,⁵⁸ except that all measurements were made at pH 8.5 and inhibitor concentrations ranged 0–200 nm for ABH, 0–25 μM for MABH, and 0–500 μM for FABH. In this measurement, HAI is immobilized on carboxymethylated dextran over the gold surface of Sensor Chip CM5 and the inhibitor is added to a buffer solution that continuously flows over the sensor surface. Inhibitor binding to the immobilized enzyme causes a change in the angle of reflection of polarized light used to interrogate the glass sensor support. Angular changes are recorded in real time as “response units” in the sensorgrams shown in Figure 3; these changes are proportional to the concentration of enzyme-bound inhibitor and thus allow for real time monitoring of the enzyme-inhibitor interaction as a function of inhibitor concentration. Dissociation constants are calculated using a 1:1 Langmuir interaction model.⁵⁹

Enzyme Inhibition Assay

The inhibition of PFA by MABH and FABH was analyzed using a colorimetric method,⁶⁰ as previously described.⁴⁰ Briefly, the reaction of urea with α-isonitroso propiophenone was measured at a wavelength of 550 nm using the Envision plate reader (courtesy of Dr. Scott Diamond, Institute of Medicine and Engineering, University of Pennsylvania). The arginase reaction was performed in 50 mM Tris (pH 8.0), 0.5 mM TCEP, 1 mM MnCl₂, 0.2 μM protein, 0–100 mM inhibitor, and 20 mM L-arginine for 5–16 min at 37 °C. Assay mixture and protein were preincubated at 37 °C for 1 min before initiating the reaction. The assay mixture (20 μL) was stopped with a sulfuric-phosphoric acid/α-isonitroso propiophenone mixture (140 μL). Reaction points were developed in a thermocycler (90 °C, 1 hr), followed by incubation at 21 °C (15 min). The K_i values for the racemic mixtures of MABH and FABH were calculated using the Cheng-Prusoff equation⁶¹ with the software Graphpad Prism (2008).

Crystallography, HAI Complexes

Crystals of the HAI-DFMO and HAI-L-ornithine complexes were prepared by soaking crystals of unliganded HAI prepared as described⁴² in 0.1 M bis-Tris (pH 6.5), 20% PEG monomethyl ether, and 20 mM DFMO or 20 mM L-ornithine for 2 days. Crystals of the HAI-MABH and HAI-FABH complexes were prepared by cocrystallization in hanging drops at 21 °C. Drops containing 3 μL of protein solution [3.5 mg/mL HAI, 50 mM bicine (pH 8.5), 2 mM MABH, 100 μM MnCl₂] and 3 μL of precipitant solution [0.1 M HEPES (pH 7.0), 22–28% Jeffamine] were equilibrated against a 1 mL reservoir of precipitant solution. Crystals appeared overnight and grew with typical dimensions of 0.5 mm × 0.2 mm × 0.2 mm. All crystals were cryoprotected in a precipitant solution containing 32% Jeffamine prior to flash cooling in liquid nitrogen.

X-ray diffraction data from all crystals were collected at GM/CA-CAT beamline 23-ID-D or NE-CAT beamline 24-ID-C at the Advanced Photon Source (APS, Argonne, IL). Diffraction intensities measured from crystals of HAI complexes exhibited symmetry consistent with apparent space group P6 (unit cell parameters a = b = 90.3 Å, c = 69.3 Å). Intensity data integration and reduction were performed using the HKL2000 suite of programs.⁶² Data reduction statistics are recorded in Table 2. As with crystals of other HAI complexes,^38,42 deviations from ideal Wilson statistics were observed with ‹I²〉/
I²〉 = 1.5, indicating perfect hemihedral twinning.⁶³ The structure of each enzyme-ligand complex was solved by molecular replacement using the program Phaser⁶⁴ with chain A of unliganded HAI (PDB accession code 2ZAV, less water molecules)⁴² used as a search probe against twinned data. In order to calculate electron density maps, structure factors amplitudes (|F_obs|) derived from twinned data (|I_obs|) were deconvoluted into structure factor amplitudes corresponding to twin domains A and B (|F_obs/A| and |F_obs/B|, respectively) using the structure-based algorithm of Redinbo and Yeates⁶³ implemented in CNS.⁶⁵ Electron density maps were visualized with the graphics software COOT.⁶⁶ After initial rigid body refinement with CNS,⁶⁵ refinement of each complex was performed against twinned data using PHENIX.⁶⁷ After water molecules were located and fit into the map, gradient omit maps revealed the bound ligand in the active site of each monomer in the asymmetric unit. Ligand atoms were refined with full occupancy and exhibited atomic B factors consistent with the average B-factor calculated for the entire protein. The quality of each final model was assessed using PROCHECK.⁶⁸ Final refinement statistics are recorded in Table 2.

Table 2.

X-ray Crystallographic Data Collection and Refinement Statistics

Complex	HAI-DFMO	HAI-L-Orn	HAI-MABH	HAI-FABH	PFA-MABH	PFA-FABH
Resolution limits, Å	50 – 1.70	50 – 1.43	50 – 1.60	50 – 1.70	50 – 1.90	50 – 2.0
Total/unique reflections	129926/67814	233126/116771	82598/8046	128641/68384	404894/44684	412761/38494
Completeness (%) (overall/outer shell)	99.9/100.0	99.8/100.0	98.4/95.6	99.0/98.2	99.9/99.2	99.9/98.7
Rmerge (overall/outer shell)a	0.101/0.412	0.067/0.640	0.050/0.298	0.148/0.577	0.081/0.444	0.090/0.587
I/σ(I) (overall/outer shell)	11.0/2.7	35.6/2.0	25.1/2.7	11.6/2.3	23.0/3.4	23.6/2.7
Refinement
R/Rfree	0.161/0.202b	0.147/0.185b	0.129/0.163b	0.136/0.178b	0.163/0.182c	0.167/0.183c
Protein atomsd	4782	4782	4764	4778	2416	2416
Manganese ionsd	4	4	4	4	2	2
Ligand atomsd	24	18	28	32	13	15
Water moleculesd	345	393	765	760	212	203
R.m.s. deviations
Bond lengths, Å	0.005	0.006	0.006	0.006	0.007	0.006
Bond angles,°	1.3	1.4	1.0	1.0	1.1	0.9
Dihedral angles,°	23	23	17	18	13	13
PDB accession code	3GN0	3GMZ	3SJT	3SKK	3SL1	3SL0

^aR_merge = Σ| I-〈I〉 | /ΣI, where I is the observed intensity and 〈I〉 is the average intensity calculated for replicate data.

^bR_twin = Σ| [| F_calc/A|²+| F_calc/B|²]^1/2-|F_Obs| |/Σ| F_Obs| for reflections contained in the working set. | F_obs| is the observed structural factor amplitude, and | F_calc/A| and | F_calc/B| are the structure factor amplitudes calculated for twin domains A and B, respectively. R_twin underestimates the residual error in the model over the two twin-related reflections by a factor of approximately 0.7. The same expression describes R_twin/free, calculated for test set reflections excluded from refinement.

^cR = Σ| | F_obs| -| F_calc| |Σ| F_obs|, where |F _Obs| and | F_calc| are the observed and calculated structure factor amplitudes, respectively. The same expression describes R_free, calculated for test set reflections excluded from refinement.

^dPer asymmetric unit.

Crystallography, Plasmodium falciparum Arginase Complexes

Recombinant PFA was prepared and purified as described, and crystals of complexes with MABH and FABH were prepared using conditions similar to those employed in the structure determination of the PFA-ABH complex.⁴⁰ Briefly, a 2 μL sitting drop of enzyme-inhibitor complex [5 mg/mL PFA, 50 mM Tris (pH 8.0), 200 mM NaCl, 1 mM MnCl₂, 1 mM tris(2-carboxyethyl)phosphine (TCEP), 5 mM inhibitor] was mixed with a 2 μL drop of precipitant solution [1.2 M sodium/potassium phosphate (pH 8.0)] and equilibrated against 500 μL of precipitant solution in the reservoir at 21 °C. Crystals appeared within 1–2 days and were harvested, cryoprotected in 30% Jeffamine ED-2001, 0.1 M HEPES (pH 7.0), 50 μM inhibitor, and flash cooled in liquid nitrogen. Diffraction data were measured on NE-CAT beamline 24-ID-E at APS. Data collection statistics are recorded in Table 2.

Data were indexed and merged using HKL2000.⁶² Molecular replacement calculations were performed with PHASER⁶⁴ using the atomic coordinates of PFA less inhibitor and solvent atoms (PDB accession code 3MMR)⁴⁰ as a search probe for rotation and translation function calculations. Iterative cycles of refinement and model building were performed using PHENIX⁶⁷ and COOT,⁶⁶ respectively, in order to improve each structure as guided by R_free. Refinement included 6 defined TLS groups identified by the TLS Motion Determination Server.⁶⁹ As also observed in the PFA-ABH complex,⁴⁰ the N-terminus was disordered (continuous main chain electron density starts at K22) and residues G72-N153 in the L2 loop were disordered. The quality of each final model was assessed using PROCHECK.⁶⁸ Refinement statistics are recorded in Table 2.