Crystal Structures of Complexes with Cobalt-Reconstituted Human Arginase I

Abstract

The binuclear manganese metalloenzyme human arginase I (HAI) is a potential protein drug for cancer chemotherapy, in that it is capable of depleting extracellular l-Arg levels in the microenvironment of tumor cells that require this nutrient to thrive. Substitution of the native Mn²⁺₂ cluster with a Co²⁺₂ cluster in the active site yields an enzyme with enhanced catalytic activity at physiological pH (~7.4) that could serve as an improved protein drug for l-Arg depletion therapy. A different catalytic mechanism is proposed for Co²⁺₂-HAI compared with that of Mn²⁺₂-HAI, including an unusual Nε---Co²⁺ coordination mode, to rationalize the lower K_M value of l-Arg and the lower K_i value of l-Orn. However, we now report that no unusual metal coordination modes are observed in the cobalt-reconstituted enzyme: the X-ray crystal structures of unliganded Co²⁺₂-HAI determined at 2.10 Å resolution (pH 7.0) and 1.97 Å resolution (pH 8.5), as well as the structures of Co²⁺₂-HAI complexed with the reactive substrate analogue 2(S)-amino-6-boronohexanoic acid (ABH, pH 7.0) and the catalytic product l-Orn (pH 7.0) determined at 1.85 Å resolution and 1.50 Å resolution, respectively, are essentially identical to the corresponding structures of Mn²⁺₂-HAI. Therefore, in the absence of significant structural differences between Co²⁺₂-HAI and Mn²⁺₂-HAI, we suggest that a higher concentration of metal-bridging hydroxide ion at physiological pH for Co²⁺₂-HAI – a consequence of the lower pK_a of a Co²⁺-bound water molecule compared with a Mn²⁺-bound water molecule – strengthens electrostatic interactions with cationic amino acids and accounts for enhanced affinity as reflected in the lower K_M value of l-Arg and the lower K_i value of l-Orn.

Arginase is a binuclear manganese metalloenzyme that catalyzes the hydrolysis of l-arginine (l-Arg)¹ to form l-ornithine (l-Orn) and urea (1-5). This metalloenzyme is becoming increasingly prominent in the exploration and development of new approaches to cancer chemotherapy. For example, arginase activity is upregulated in certain human colon cancer and breast cancer cell lines, resulting in decreased l-Arg levels and increased l-Orn levels; since l-Orn is a biosynthetic precursor of polyamines that facilitate tumor cell growth and proliferation, arginase inhibition decreases l-Orn levels and disrupts the l-Orn supply for polyamine biosynthesis, thereby inhibiting cell proliferation (6-9). Conversely, some cancer cells are auxotrophic for l-Arg and depend on extracellular l-Arg to thrive; such cancer cells can be targeted with l-Arg depletion therapy (10, 11). For example, hepatocellular carcinoma cells (12) and prostate carcinoma cells (10, 13) require extracellular l-Arg to thrive, in the absence of which they undergo apoptosis (14). Accordingly, arginase is a potential protein drug for cancer chemotherapy insofar that it can efficiently deplete extracellular l-Arg levels in the tumor microenvironment. To this end, the substitution of the native Mn²⁺₂ cluster with a Co²⁺₂ cluster in the active site of human arginase I (HAI) yields an enzyme with enhanced catalytic activity at physiological pH (~7.4) that can potentially serve as an even more effective agent for l-Arg depletion therapy (14).

The catalytic mechanism of Mn²⁺₂-HAI is believed to be initiated by the nucleophilic attack of a metal-bridging hydroxide ion at the guanidinium carbon of l-Arg to form a neutral tetrahedral intermediate, which subsequently collapses to form products l-Orn and urea (Figure 1a) (5, 15, 16). A key feature of this mechanistic proposal is a non-metal binding site for the guanidinium group of l-Arg in the precatalytic enzyme-substrate complex. This hypothesis is consistent with structure-activity relationships established for rat arginase I mutants in which the metal binding sites are perturbed by substitution of the metal ligands (17). These mutants exhibit nearly invariant K_M values regardless of whether Mn²⁺_A or Mn²⁺_B binding is perturbed; a more significant effect on K_M would be expected if an inner-sphere substrate-metal coordination interaction occurred in the precatalytic enzyme-substrate complex. This hypothesis is also consistent with the X-ray crystal structures of HAI as well as rat arginase I complexed with boronic acid substrate analogue inhibitors (16, 18). Specifically, the trigonal planar boronic acid moieties of substrate analogues 2(S)-amino-6-boronohexanoic acid (ABH) (19) and S-(2-boronoethyl)-l-cysteine (20) undergo nucleophilic attack to form negatively charged tetrahedral boronate anions. Each tetrahedral boronate anion coordinates to the Mn²⁺₂ cluster in the HAI active site in much the same manner expected for the tetrahedral intermediate in catalysis (Figure 1b); a similar binding mode is also observed for 2(S)-amino-7-oxoheptanoic acid, an aldehyde amino acid analogue of l-Arg, which binds as the tetrahedral gem-diol (Figure 1c) (21). The geometry required for the binding of such boronic acid or aldehyde hydrates as analogues of the tetrahedral intermediate is not readily compatible with a substrate-metal coordination interaction in the precatalytic enzyme-substrate complex. However, coordination of the Nη2 atom of l-Arg to Mn²⁺_A can occur during nucleophilic attack: formation of the C–O bond breaks the “Y”-shaped guanidinium π system and localizes a lone electron pair on the Nη2 atom, which can then coordinate to Mn²⁺_A (Figure 1a) (5).

Figure 1.

(a) Proposed mechanism of Mn²⁺₂-HAI (5). (b) The binding of a boronic acid inhibitor as the tetrahedral boronate anion mimics the tetrahedral intermediate and its flanking transition states in catalysis (16, 18, 20). (c) The binding of an aldehyde inhibitor as the tetrahedral gem-diol mimics the tetrahedral intermediate and its flanking transition states in catalysis (21). (d) Proposed mechanism of Co²⁺₂-HAI (14).

Curiously, Co²⁺₂-HAI is proposed to catalyze l-Arg hydrolysis through a significantly different mechanism than Mn²⁺₂-HAI (14). Specifically, the Nε atom of l-Arg is proposed to coordinate directly to an unspecified Co²⁺ ion in the precatalytic enzyme-substrate complex, after which the Co²⁺-bound guanidinium group undergoes nucleophilic attack by a hydroxide ion bound to the adjacent Co²⁺ ion to form the tetrahedral intermediate; this intermediate is stabilized by coordination of the Nε and Oη atoms to the Co²⁺ ions and ultimately collapses to yield l-Orn and urea (Figure 1d). This proposal emanates from two observations. First, the K_M of l-Arg at pH 8.5 with Co²⁺₂-HAI is 11-fold lower than that measured with Mn²⁺₂-HAI, which implicates the Co²⁺₂ cluster in a stronger substrate binding interaction. Second, the inhibition constant (K_i) of l-Orn is 26-fold lower than that of l-leucine against Co²⁺₂-HAI at pH 8.5, whereas the corresponding K_i values are relatively unchanged against Mn²⁺₂-HAI at pH 8.5. These data are interpreted to suggest that the side chain Nε atom of l-Orn coordinates directly to a Co²⁺ ion (14).

To better understand structure-mechanism relationships in the cobalt-reconstituted enzyme, we now report the X-ray crystal structures of Co²⁺₂-HAI and its complexes with a boronic acid substrate analogue as well as the catalytic product. Specifically, we describe the structures of metal-free HAI at 1.64 Å resolution, unliganded Co²⁺₂-HAI at pH 7.0 (2.10 Å resolution) and pH 8.5 (1.97 Å resolution), Co²⁺₂-HAI complexed with ABH at 1.85 Å resolution, and Co²⁺₂-HAI complexed with l-Orn at 1.50 Å resolution. These structures show that Co²⁺ substitution does not trigger any significant structural changes in the active site of HAI. Moreover, these structures do not provide any evidence for different enzyme-substrate, enzyme-intermediate, and enzyme-product binding modes compared with corresponding structures of Mn²⁺₂-HAI.

Experimental Procedures

Preparation of Crystalline Co²⁺₂-HAI and Complexes

Recombinant HAI was expressed in Escherichia coli and purified as described (16), and crystallized in the unliganded state according to published procedures (22). Metal-free HAI crystals were prepared by soaking Mn²⁺₂-HAI crystals for 7 days in 15 mM EDTA, 15 mM dipicolinic acid, 100 mM HEPES (pH 7.0), 30% (v/v) Jeffamine ED-2001. Essentially complete metal removal was confirmed by X-ray crystal structure determination, which revealed the absence of both metal ions in the active site. Crystals of metal-free HAI were reconstituted with Co²⁺ at pH 7.0 by soaking in a buffer solution containing 20 mM CoCl₂, 100 mM HEPES (pH 7.0), 30% (v/v) Jeffamine ED-2001 for 25 hours. Metal-free HAI crystals were reconstituted with Co²⁺ at pH 8.5 by soaking in 20 mM CoCl₂, 100 mM Bicine (pH 8.5), 32% (v/v) Jeffamine ED-2001 for 21 hours. The Co²⁺₂-HAI-ABH complex was prepared by soaking a Co²⁺₂-HAI crystal in 20 mM ABH, 5 mM CoCl₂, 100 mM HEPES (pH 7.0), 30% (v/v) Jeffamine ED-2001 for 39 hours. The Co²⁺₂-HAI-l-Orn complex was prepared by soaking a Co²⁺₂-HAI crystal in 20 mM l-Orn, 5 mM CoCl₂, 0.1 M HEPES (pH 7.0), 30% (v/v) Jeffamine ED-2001 for 24 hours. All crystals were flash-cooled in liquid nitrogen with their corresponding mother liquor solutions serving as cryoprotectants.

X-ray Crystal Structure Determinations

X-ray diffraction data from single crystals of metal-free HAI, Co²⁺₂-HAI, the Co²⁺₂-HAI-ABH complex, and the Co²⁺₂-HAI-l-Orn complex were collected on beamline X29 (λ = 0.9795 Å) of the National Synchrotron Light Source at Brookhaven National Laboratory (Upton, NY). Diffraction data were indexed, integrated, and scaled using the HKL-2000 suite (23). These twinned arginase crystals belonged to the apparent space group P3, as reported for the Mn²⁺₂-HAI-ABH complex (16), and their unit cell dimensions were very similar (Table 1).

Table 1. Data collection and refinement statistics.

		Co2+2-HAI
	metal-free HAI	pH 7.0	pH 8.5	Co2+2-HAI-ABH	Co2+2-HAI-l-Orn
Data collection
resolution limits (Å)	50.0 - 1.64	50.0 - 2.10	50.0 - 1.97	50.0 - 1.85	50.0 - 1.50
total/unique reflections   measured	837181/78451	174332/37395	179056/40321	247148/53857	558194/102235
unit cell dimensions
a, b, c (Å)	90.86, 90.86, 69.88	90.41, 90.41, 69.74	87.42, 87.42, 67.25	90.43, 90.43, 69.43	90.49, 90.49, 69.69
α, β, γ (deg)	90, 90, 120	90, 90, 120	90, 90, 120	90, 90, 120	90, 90, 120
R merge a,b	0.080 (0.430)	0.126 (0.417)	0.097 (0.608)	0.078 (0.586)	0.109 (0.550)
I/σ(I)a	35.18 (5.46)	12.25 (4.18)	14.98 (2.20)	19.55 (2.32)	14.63 (2.89)
completeness (%)a	99.3 (95.4)	100 (100)	98.9 (96.3)	99.3 (98.4)	99.9 (100)
Refinement
reflections used in   refinement/test set	70907/6432	34308/1833	36032/1913	46410/5045	94996/4852
twinning fraction	0.50	0.45	0.30	0.50	0.45
R twinc c	0.155	0.147	0.144	0.195	0.148
R twin/free c	0.197	0.209	0.193	0.239	0.178
solvent moleculesd	207	127	223	233	277
ligand moleculesd	0	0	2	2	3
Co2+ ionsd	0	4	4	4	4
Root mean square deviations e
bonds (Å)	0.008	0.008	0.007	0.007	0.007
angles (deg)	1.5	1.5	1.5	1.5	1.6
Average B-factors (Å2) f
main chain	24	24	24	27	20
side chain	26	25	26	28	21
solvent	26	24	28	28	23
ligand	--	--	36	22	23
Co2+ ions	--	20	25	20	15
Ramachandran plot (%) e
allowed	89.2	85.7	88.4	86.9	90.4
additionally allowed	10.2	13.9	11.2	12.7	9.2
generously allowed	0.6	0.0	0.4	0.2	0.2
disallowed	0.0	0.4	0.0	0.2	0.2
PDB accession code	3TF3	3TH7	3THE	3THH	3THJ

^aValues in parenthesis are for the highest resolution shell.

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

^cR_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 structure 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.

^dPer asymmetric unit cell

^eCalculated using PROCHECK (35)

^fCalculated using MOLEMAN (36)

Structures were solved by molecular replacement using the program Phaser (24) as implemented in CCP4 (25), with the chain A structure of the Mn²⁺₂-HAI-ABH complex (PDB 2AEB) (16) less inhibitor, Mn²⁺ ions, and solvent atoms used as the search probe for rotation and translation function calculations. Each refinement was performed with CNS (version 1.2) (26) and model building was performed with Coot (version 0.6.1) (27). Hemihedral twinning operation parameters used in the refinement were -h, -k, and l while the twinning fraction depended on the dataset (Table 1).

Crystallographic refinement of each structure against twinned intensity data was performed as previously described (16). Water molecules were included in the later stages of each refinement. For the Co²⁺₂-HAI-ABH and Co²⁺₂-HAI-l-Orn complexes, gradient omit maps clearly showed ligands bound to the active site of each monomer in the asymmetric unit, and ligand atoms were added and refined with full occupancy. Thermal B factors for ligands were consistent with the average B factor calculated for the entire protein (Table 1). Disordered segments M1–S5 and P320–K322 at the N- and C-termini, respectively, were excluded from all final models. Ramachandran plots revealed Q65 with a disallowed conformation in certain structures. For Co²⁺₂-HAI (pH 7.0), this included monomers A and B; in the Co²⁺₂-HAI-ABH and Co²⁺₂-HAI-l-Orn complexes, this included only monomer A. Generally speaking, Q65 was characterized by well-defined electron density in these structures, so its conformation was not ambiguous. Moreover, this residue adopted a similar conformation in Mn²⁺₂-HAI (PDB 2PHA) (22), so its unfavorable conformation was not likely to be an artifact. Data collection and refinement statistics for all structure determinations are recorded in Table 1.

Results

Metal-free HAI

The structure of metal-free HAI is the first structure of any arginase in which both active site metal ions are absent (Figure 2a). The overall structure is comparable to that of unliganded wild-type HAI (Mn²⁺₂-HAI, PDB 2PHA (22)) with a root-mean-square (r.m.s.) deviation of 0.24 Å for 313 Cα atoms. Metal binding residues D128, D124, H126, D232, D234 and H101 remain fully ordered despite the loss of Mn²⁺_A and Mn²⁺_B; however, H101 and D232 undergo slight conformational changes to form a hydrogen bond upon metal dissociation (Figure 2b). Thus, the metal binding site is nearly pre-formed with optimal geometry for metal binding, i.e., minimal conformational changes are necessary for binding the two metal ions required for catalysis.

Figure 2.

(a) Final |F_o| - |F_c| map of metal-free HAI contoured at 3.0σ (green) and −3.0σ (red). Only spurious noise peaks are observed, and none correspond to residual metal ions. A water molecule (red sphere), confirmed in a simulated annealing omit map contoured at 3.0σ (blue), is hydrogen bonded (green dashed lines) to D124, D128, and D232. Atoms are color-coded as follows: C = yellow, N = blue, O = red. (b) Superposition of chains A for metal-free HAI (pH 7.0, color-coded as in (a)) and Mn²⁺₂-HAI (pH 7.5) (PDB 2PHA, all atoms and interactions colored light blue). Apart from small conformational changes of D232 and H101, protein residues in the apoenzyme are nearly pre-formed for ideal metal coordination interactions.

Co²⁺₂-HAI

The overall fold of unliganded Co²⁺₂-HAI at pH 7.0 and pH 8.5 is essentially identical to that of unliganded Mn²⁺₂-HAI at pH 7.5 (PDB 2PHA (22)), with an r.m.s. deviation of 0.26 Å for 313 Cα atoms or 0.47 Å for 313 Cα atoms, respectively. Interestingly, no metal-bound solvent molecules are observed at pH 7.0, except for a single solvent molecule bound to the Co²⁺_A ion of chain A (Figure 3a), which could imply weaker coordination or disorder. However, metal-bound solvent molecules are fully visible in the structure of Co²⁺₂-HAI determined at pH 8.5 (Figure 3b) and are identical to those observed in the structure of unliganded Mn²⁺₂-HAI (22) (Figure 3c). Parenthetically, we note that a bicine buffer molecule is observed to bind at the mouth of the Co²⁺₂-HAI active site at pH 8.5 (data not shown).

Figure 3.

(a) Simulated annealing omit map (green) of the Co²⁺_A-bound solvent molecule in chain A of Co²⁺₂-HAI at pH 7.0, contoured at 3.5σ. Atoms are color-coded as follows: C = yellow, N = blue, O = red, Co²⁺ = magenta spheres, solvent = red sphere. Metal coordination and hydrogen bond interactions appear as blue and green dashed lines, respectively. (b) Simulated annealing omit map (blue) of the Co²⁺-bound solvent molecules in Co²⁺₂-HAI at pH 8.5, contoured at 4.2σ. Atoms and intermolecular interactions are color-coded as in (a). (c) Superposition of chains A for Co²⁺₂-HAI (pH 8.5) (color-coded as in (b)) and Mn²⁺₂-HAI (pH 7.5) (PDB 2PHA, all atoms light blue).

Co²⁺₂-HAI-ABH complex

The structure of the Co²⁺₂-HAI-ABH complex at pH 7.0 is essentially identical to the structure of the Mn²⁺₂-HAI-ABH complex (PDB 2AEB) (16) in both the overall fold (r.m.s. deviation = 0.25 Å for 313 Cα atoms) and the intermolecular interactions of ABH in the active site. A simulated annealing omit map is shown in Figure 4a along with a superposition of the Mn²⁺₂-HAI-ABH complex. The boronic acid side chain of ABH undergoes nucleophilic attack, probably by the metal-bridging hydroxide ion, to yield a tetrahedral boronate anion that coordinates to the Co²⁺_A and Co²⁺_B ions. The α-carboxylate of ABH accepts hydrogen bonds from N130, S137, and two water molecules, and the α-amino group donates hydrogen bonds to D183 and two water molecules. Water molecules hydrogen bonded with the α-carboxylate and the α-amino groups also hydrogen bond with active site protein residues as summarized in Figure 4b.

Figure 4.

(a) Simulated annealing omit map (grey) of the inhibitor ABH bound in the active site of Co²⁺₂-HAI, contoured at 4.0σ. The boronic acid moiety of ABH binds as a tetrahedral boronate anion that mimics the tetrahedral intermediate and its flanking transition states in the reaction catalyzed by Co²⁺₂-HAI. Atoms are color-coded as follows: C = yellow, N = blue, O = red, B = pink, Co²⁺ = magenta spheres, solvent = red spheres; metal coordination and hydrogen bond interactions are indicated by blue and green dashed lines, respectively. (b) Scheme illustrating average distances (Å) of intermolecular interactions in the Co²⁺₂-HAI-ABH complex (black numbers) and Mn²⁺₂-HAI-ABH complex (PDB 2AEB) (orange numbers).

Co²⁺-HAI-l-Orn complex

The structure of the Co²⁺₂-HAI-l-Orn complex determined at pH 7.0 has an overall fold similar to that of the Mn²⁺₂-HAI-l-Orn complex determined at pH 6.5 (28), and the r.m.s. deviation for 314 Cα atoms is 0.17 Å. A simulated annealing omit map is shown in Figure 5. The terminal amino group of l-Orn donates hydrogen bonds to D128, the backbone carbonyl of H141, and two solvent molecules, one of which is the metal-bridging hydroxide ion. The α-amino and α-carboxylate groups of l-Orn make interactions similar to those observed for ABH in the Co²⁺₂-HAI-ABH complex (Figure 4) and the Mn²⁺₂-HAI-ABH complex (16). Parenthetically, we note that in the Co²⁺₂-HAI-l-Orn complex, an additional l-Orn molecule binds to the surface of monomer A in the asymmetric unit, where the backbone carbonyl groups of T134 and T135 accept hydrogen bonds from the α-amino group and the ε-amino group of l-Orn, respectively (data not shown). This second l-Orn molecule is characterized by weak electron density consistent with reduced occupancy; since Stone and colleagues do not report any nonlinearity in product inhibition (14), l-Orn binding to this second site does not appear to significantly affect catalysis.

Figure 5.

Simulated annealing omit map (grey) of the catalytic product l-Orn bound in the active site of Co²⁺₂-HAI at pH 7.0, contoured at 3.0σ. Atoms are color-coded as follows: C = yellow, N = blue, O = red, Co²⁺ = magenta spheres, solvent = red spheres; metal coordination and hydrogen bond interactions are indicated by blue and green dashed lines, respectively.

Discussion

The first successful preparation of a crystalline metal-free arginase described in this work allows for the substitution of metal ions other than the native Mn²⁺ ions in the active site. Previous studies with rat arginase I show that only one Mn²⁺ ion is readily extracted from the Mn²⁺₂ cluster (29), so the human enzyme offers a distinct advantage for the preparation of the metal-free apoenzyme. Significantly, Co²⁺₂-HAI is said to have ideal functional properties for use in the treatment of l-Arg auxotrophic tumors, e.g., as assayed against human melanoma and hepatocellular carcinoma cell lines: a decreased pK_a for metal-bound water, a decreased K_M for substrate l-Arg, and a decreased K_i for product l-Orn (14).

Functional studies of Co²⁺₂-HAI (14) are interpreted to reflect the proposed mechanism in Figure 1d, which includes the following chemical steps: (1) deprotonation of l-Arg, (2) tautomerization of the neutral guanidinium group and coordination of the Nε atom to a Co²⁺ ion, (3) nucleophilic attack at the guanidinium carbon by a Co²⁺-bound hydroxide ion, (4) formation of the tetrahedral intermediate in which the Oη and Nε atoms are coordinated to separate Co²⁺ ions, (5) collapse of the tetrahedral intermediate to yield product l-Orn with its Nε atom coordinated to one Co²⁺ ion, and a water molecule that displaces urea to coordinate to the other Co²⁺ ion, and (6) ionization of Co²⁺-bound water to regenerate the nucleophilic Co²⁺-bound hydroxide ion.

X-ray crystal structures of Co²⁺₂-HAI allow us to evaluate certain aspects of this mechanistic proposal. First, it is intriguing to consider the possibility that the Nε atom of l-Arg or l-Orn coordinates to an active site metal ion. While it is unusual to consider the inner-sphere coordination of the guanidinium Nε or Nη atom of l-Arg to a metal ion, such interactions are occasionally observed (30). However, since the Nε atom of l-Orn does not coordinate to metal ions in the active site of Co²⁺₂-HAI (Figure 5) or Mn²⁺₂-HAI (28), the observed binding mode is at odds with the Nε---Co²⁺ coordination mode proposed for l-Orn, and by inference the Nε---Co²⁺ coordination mode proposed for l-Arg, in Figure 1d (14). Since l-Orn adopts an identical binding mode in the active sites of Mn²⁺₂-HAI and Co²⁺₂-HAI (Figure 6), the lower K_i value of l-Orn for binding to Co²⁺₂-HAI must therefore arise from an indirect effect of the metal ions, e.g., the hydrogen bond between the l-Orn side chain and the metal-bridging solvent molecule (Figure 5). The concentration of metal-bridging hydroxide ion is greater for Co²⁺₂ than for Mn²⁺₂ at physiological pH (14), which in turn would strengthen its hydrogen bond and electrostatic interactions with the positively charged side chain of l-Orn.

Figure 6.

(a) Superposition of the Co²⁺₂-HAI-ABH complex (color-coded as in Figure 4a) and the Mn²⁺₂-HAI-ABH complex at pH 7.5 (PDB 2AEB, all atoms and metal coordination interactions light blue). Apart from a 0.5 Å shift of the side chain Cγ atom of ABH, the structures of these complexes are essentially identical. (b) Superposition of the Co²⁺₂-HAI-l-Orn complex (color-coded as in Figure 5) and the Mn²⁺₂-HAI-l-Orn complex (PDB 3GMZ, all atoms and interactions light blue).

The binding of ABH as the tetrahedral boronate anion likely mimics the binding of the tetrahedral intermediate and its flanking transition states in catalysis by Mn²⁺₂-rat arginase I (18), Mn²⁺₂-HAI (16), Co²⁺₂-HAI (this work), and Mn²⁺₂-arginase from Plasmodium falciparum (31). The binding conformation of ABH to these enzymes is essentially identical (as illustrated in Figure 6 for Co²⁺₂-HAI and Mn²⁺₂-HAI), such that the amino acid side chain is extended into the active site with all C-C bonds adopting trans or nearly trans conformations. Identical binding modes for ABH to these enzymes suggest identical binding modes for the corresponding tetrahedral intermediate and its flanking transition states, i.e., identical catalytic mechanisms.

Further analysis of the binding mode of ABH to Co²⁺₂-HAI suggests that the boronate anion hydroxyl groups O2 and O3 correspond to the two hydroxyl groups of the native boronic acid, whereas boronate anion hydroxyl group O1 corresponds to the metal-bridging hydroxide ion of the native enzyme. This binding mode is most consistent with a mechanism in which the trigonal planar boronic acid moiety of ABH (or the trigonal planar guanidinium group of the substrate) enters the active site of Co²⁺₂-HAI and subsequently undergoes nucleophilic attack by a metal-bridging hydroxide ion to yield the tetrahedral boronate anion (or tetrahedral intermediate), with a metal-bridging hydroxyl group (i.e., the former metal-bridging hydroxide ion) and the terminal O2 hydroxyl group (or Nη2 amino group) coordinated to Co²⁺_A (Figure 7a). This mechanistic model is identical to that outlined for Mn²⁺₂-HAI (5) and derives from that initially proposed by Kanyo and colleagues for Mn²⁺₂-rat arginase I (15), which satisfies the principle of least nuclear motion (Figure 7b) (32). Alternative mechanistic models, such as that initially proposed for Co²⁺₂-HAI (14), involve additional chemical or conformational steps that would add seemingly unnecessary nuclear motions to the mechanistic sequence.

Figure 7.

(a) Attack of the nucleophilic metal-bridging hydroxide ion at the boronic acid moiety of the inhibitor ABH results in the formation of a tetrahedral boronate anion that mimics the formation of the tetrahedral intermediate in catalysis. While this chemistry is illustrated for Co²⁺₂-HAI, it is identical to that of Mn²⁺₂-HAI. Selected atoms discussed in the text are indicated with red labels. (b) Model of l-Arg superimposed on the experimentally determined structure of ABH bound in the active site of Co²⁺₂-HAI, which mimics the binding of the tetrahedral intermediate and its flanking transition states in catalysis. Atoms are color-coded as follows: C = yellow (protein), black (ABH), or grey (l-Arg), N = blue, O = red, Co²⁺ = magenta spheres. Solvent molecules are omitted for clarity; metal coordination and hydrogen bond interactions for ABH are indicated by blue and green dashed lines, respectively. The side chain Nη1 and Nη2 atoms of l-Arg correspond to the boronate O3 and O2 atoms (atom labels are shown in (a)). Nucleophilic attack of a metal-bridging hydroxide ion (which would correspond to the position of the boronate O1 atom) at the planar guanidinium group of l-Arg satisfies the principle of least nuclear motion, and simply requires the pyramidalization of the guanidinium carbon (or the boron atom of the boronic acid) as it transitions from sp² to sp³ hybridization. This conclusion is valid for both Co²⁺₂-HAI and Mn²⁺₂-HAI.

The observed binding mode of the tetrahedral boronate anion form of ABH to Co²⁺₂-HAI is not consistent with the proposed binding mode of the tetrahedral intermediate that involves metal coordination by the Nε atom of l-Arg (Figure 1d) (14). A key structural feature that prevents a closer approach of the Nε atom of l-Arg or l-Orn to the binuclear metal cluster is the extensive array of hydrogen bond interactions that anchor the α-amino and α-carboxylate groups of amino acids bound in the arginase active site (e.g., see the scheme summarizing the binding mode of ABH in Figure 4b). This hydrogen bond array ensures the precise molecular recognition of l-amino acids in the arginase active site. Modification or deletion of these enzyme-substrate hydrogen bonds significantly compromises catalysis, either by mutagenesis in the enzyme active site (33) or by modification of substrate structure or stereochemistry (34).

In closing, while the proposed mechanism of Co²⁺₂-HAI summarized in Figure 1d (14) is inconsistent with the X-ray crystal structures of Co²⁺₂-HAI reported herein, the fact remains that functional studies clearly indicate a decreased K_M value for substrate l-Arg and a decreased K_i value for product l-Orn (14). If functional differences between Co²⁺₂-HAI and Mn²⁺₂-HAI do not arise from structural differences in the binding of substrate, tetrahedral intermediate, and product, from what do they arise? Given that the metal-bridging solvent molecule exhibits a lower pK_a in Co²⁺₂-HAI, there would be a higher concentration of the negatively charged metal-bridging hydroxide ion at physiological pH in comparison with Mn²⁺₂-HAI (14). As previously mentioned, this would enhance the hydrogen bond and electrostatic interactions with the positively charged side chain of l-Orn and thereby enhance affinity, which would account for the lower K_i value measured against Co²⁺₂-HAI. Similarly, we suggest that the lower K_M value measured for l-Arg with Co²⁺₂-HAI results from an enhanced electrostatic interaction (but not a hydrogen bond interaction) between the positively charged guanidinium side chain of l-Arg and the negatively charged metal-bound hydroxide ion in the precatalytic Michaelis complex. Thus, in the absence of significant structural differences between Mn²⁺₂-HAI and Co²⁺₂-HAI complexes, we suggest that a simple electrostatic interaction dependent upon the predominant ionization state of a metal-bound water molecule – not a direct metal coordination interaction – could explain the functional differences between Co²⁺₂-HAI and Mn²⁺₂-HAI.