Structural and Kinetic Analyses Reveal the Dual Inhibition Modes of Ornithine Aminotransferase by (1S,3S)-3-Amino-4-(Hexafluoropropan-2-Ylidenyl)-Cyclopentane-1-Carboxylic Acid (BCF₃)

Abstract

Hepatocellular carcinoma (HCC) is the most common form of liver cancer and the leading cause of death among people with cirrhosis. HCC is typically diagnosed in advanced stages when tumors are resistant to both radio- and chemotherapy. Human ornithine aminotransferase (hOAT) is a pyridoxal-5′-phosphate (PLP)-dependent enzyme involved in glutamine and proline metabolism. Because hOAT is overexpressed in HCC cells and a contributing factor for the uncontrolled cellular division that propagates malignant tumors (Ueno, A. et al., J. Hepatol. 2014, 61, 1080–1087), it is a potential drug target for the treatment of HCC. (1S,3S)-3-Amino-4-(hexafluoropropan-2-ylidenyl)-cyclopentane-1-carboxylic acid (BCF₃) has been shown in animal models to slow the progression of HCC by acting as a selective and potent mechanism-based inactivator of OAT (Zigmond et al., ACS Med. Chem. Lett. 2015, 6, 8, 840–844). Previous studies have shown that the BCF₃-hOAT reaction has a bifurcation in which only 8% of the inhibitor inactivates the enzyme while the remaining 92% ultimately acts as a substrate and undergoes hydrolysis to regenerate the active PLP form of the enzyme. In this manuscript, the rate-limiting step of the inactivation mechanism was determined by stopped-flow spectrophotometry and time-dependent ¹⁹F-NMR experiments to be the decay of a long-lived external aldimine species. A crystal structure of this transient complex revealed both the structural basis for fractional irreversible inhibition and the principal mode of inhibition of hOAT by BCF₃, which is to trap the enzyme in this transient but quasi-stable external aldimine form.

Graphical Abstract

Introduction

Aminotransferases represent a group of PLP-dependent enzymes that use a “Bi-Bi, Ping-Pong” mechanism to catalyze the transfer of an amino group, typically donated by an amino acid to an acceptor α-keto acid^1–3. In the first half-reaction, an amino acid is oxidized with the concomitant reduction of the PLP coenzyme to pyridoxamine-5’-phosphate (PMP). In the second half-reaction, the amino group from PMP is transferred to a suitable α-keto acid acceptor with the concomitant oxidation of PMP to regenerate the PLP coenzyme. The transamination mechanism of hOAT exhibits these two canonical half-reactions with the conversion of L-ornithine and PLP to glutamate-γ-semialdehyde and PMP followed by the conversion of α-ketoglutarate and PMP to L-glutamate and PLP (Scheme 1).

Scheme 1.

The Chemistry of OAT

hOAT in mammals is transported into the mitochondrial matrix and is in high abundance in the brain, intestines, kidneys, and liver. It functions at a nexus for multiple metabolic pathways including glutamine metabolism, proline synthesis, and the urea cycle⁴. Knockout experiments of the oat gene revealed that OAT may not have an important role in mitosis within normally developing cells⁴. However, studies by Wang et al. have implicated hOAT in the assembly of the mitotic spindle in human cancer cells⁵. Complex coordination of the components of the mitotic spindle is essential for successful separation of the sister chromatids. Moreover, inhibition of OAT is expected to diminish cellular glutamate and proline, both of which are important metabolites in neoplastic cells. Therefore, inhibition of hOAT is one path to slow cell division in cancer cells.

Mehta et al.² performed a sequence alignment of 14 aminotransferases, including GABA aminotransferase (GABA-AT), L-aspartate aminotransferase (Asp-AT), and hOAT, and found that while they share low sequence identity, the available structures indicate highly similar tertiary folds^{1, 6}. The highly conserved residues in the active sites of different aminotransferases make it difficult to selectively inhibit one enzyme without generating off-target reactivity and associated toxicity. Known irreversible inhibitors for both OAT and GABA-AT, such as gabaculine and 4-amino-5-hexynoic acid lack desired selectivity⁶. Although active site residues may be conserved, aspects of the chemical mechanism of each enzyme are unique. Therefore, the key to the rational design of a selective inactivator of an aminotransferase is based on an understanding of the specific chemistry involved.

(1S,3S)-3-Amino-4-(hexafluoropropan-2-ylidenyl)-cyclopentane-1-carboxylic acid (named here as BCF₃ after the (Bis)trifluoromethyl feature) is an irreversible inhibitor of hOAT that showed promising results for suppression of HCC tumor growth in athymic mice implanted with human-derived HCC⁷. BCF₃ covalently and irreversibly modifies both the catalytic lysine 292 and PLP, forming a tethered tertiary adduct that inactivates the enzyme. The currently proposed inactivation mechanism of BCF₃ lacks experimental proof for the proposed intermediate complexes, and the rate-limiting step was hitherto yet to be identified⁸. To better understand the nature of hOAT as a drug target and facilitate the rational design of agents active against HCC, it is essential to quantitatively establish and verify the mechanism of BCF₃ inhibition. This work presents an investigation of hOAT inactivation by BCF₃ using a combined approach of transient-state kinetics and X-ray crystallography. The conclusions we have drawn incorporating the data from these additional methodologies are summarized in (Scheme 2).

Scheme 2.

Proposed Inactivation Mechanism of hOAT by BCF₃

RESULTS AND DISCUSSION

Mechanism-based inactivators commandeer the chemical mechanism of an enzyme to covalently and irreparably alter functional groups in the active site⁹. The rational design of covalent inactivators based on a detailed understanding of the chemical mechanism has great potential for drug discovery, as such an approach builds in both potency and specificity simultaneously. Methods to verify the chemistry of such inactivators most often include inhibition assays, mass spectrometry and X-ray crystallography^10–14. Although these approaches have been successful in describing the inactivation mechanisms of PLP-dependent enzymes, the amenability of such enzymes to transient-state approaches allows for more comprehensive evaluation, particularly those that exhibit multiple intermediates and partitions.

PLP-dependent transaminase reactions are often studied by spectrophotometric transient-state measurements as they exhibit predictable sequences of spectrophotometrically identifiable intermediates with absorption transitions from 300–650 nm^15–17. However, the identity of new intermediates or reaction endpoints cannot be based solely on absorption spectra. For such species, it is necessary to employ complementary approaches in parallel to define the chemistry that is occurring. The kinetics of the BCF₃ inhibition of hOAT has rates that span seven orders of magnitude (Scheme 2) and can be studied by spectroscopic methods that have fast and slow integration times. Here we employ rapid mixing spectrophotometric, time-dependent NMR, and steady-state X-ray crystallography/crystal-soaking methods to reveal the detailed inhibition mechanism of hOAT reacting with BCF₃.

Transient State Measurements of hOAT Inactivation by BCF₃:

A combination of spectrophotometric and NMR measurements was used to capture the kinetics of the inhibition of hOAT by BCF₃. This inhibition process had previously been characterized by mass spectroscopy and X-ray crystallography⁸. The approaches used take advantage of the highly conjugated species that occur sequentially, the unique NMR shifts of ¹⁹F correlated with additional X-ray structural data (see below). These measurements provide a more complete assignment of the inactivation chemistry of hOAT by BCF₃. Titration of hOAT with BCF₃ modulated the rate of accumulation of a quinonoid intermediate observed maximally at 567 nm (Figure 1).

Figure 1.

Transient state absorption changes observed at 567 nm for hOAT reacting with BCF₃. hOAT (9.84 μM) was allowed to react with BCF₃ (63, 127, 254, 508, 1016, 2035, 4065 μM) and CCD spectra were collected on two timeframes (0.009–12.3 seconds for concentrations 63, 127, 254 μM, and 0.02 – 24 sec for concentrations 508, 1016, 2035, 4065 μM). The extracted 576 nm data were fit to a linear model shown in orange in Scheme 3 and returned the rate constants shown in Scheme 2.

The formation and decay of this intermediate could be fit to a reversible three-step model shown in orange in Scheme 3 and the rate constants returned from the fit are shown in Scheme 2. The observed rate of formation of the quinonoid is a function of the extent of population and rate of decay of the preceding external aldimine species (see below). As such these data report both on the individual rate constants for formation and decay of three species and the prior equilibrium for formation of the initial external aldimine. The global fit gave rate constants of k₁ = 1.5×10⁵ ± 3×10⁴ M⁻¹s⁻¹, k₋₁ = 421 ± 64 s⁻¹ for formation of the external aldimine I (Scheme 2). The dissociation constant for the reversible formation of this species was therefore 3.0 ± 0.5 mM. The subsequent formation of the quinonoid species was described by rate constants k₂ = 6.18 ± 0.24 s⁻¹ and k₋₂ = 0.81 ± 0.02 s⁻¹. The irreversible decay of the quinonoid was fit to yield a rate constant of k₃ = 0.74 ± 0.01 s⁻¹.

Scheme 3.

Four step model for the intermediates observed via CCD 340–700 nm and ¹⁹F-NMR spectrometry. The first three reaction steps (orange box) were observed for a range of BCF₃ concentrations and analyzed globally by numerical integration at a single wavelength (567 nm). The rate constants associated with the second and third steps (green box) were derived from CCD data using singular value decomposition (SVD). The data for the last step (blue box) were acquired with ¹⁹F-NMR spectrometry and analyzed using SVD.

Three-dimensional CCD data recorded for a single concentration of BCF₃ (Figure 2A) were fit globally using SVD to deconvolute and define pure spectra of the intermediates observed (Figure 2B). These data were fit to a simplified three-step irreversible model and thus did not return evidence of reversibility. The rate constants obtained from the fit agree well with those determined from the single wavelength analysis at 567 nm (k₂ = 5.13 ± 0.03 s⁻¹ and k₃ = 0.51 ± 0.01 s⁻¹).

Figure 2.

Transient-state CCD data for hOAT reacting with BCF₃. A. hOAT (9.84 μM final) was mixed with BCF₃ (4065 μM) and CCD spectra were collected on two timeframes (0.0025 – 12.5, 15–2500 sec) and spliced together at ~12 sec to form a single data set for analysis. The data were fit using the SVD routine available in KinTek Explorer software to a linear model shown in green in Scheme 3, where the last rate constant was fixed to that determined from the time-dependent NMR data shown in Figure 3 and returned rate constants shown in Scheme 2. B. Spectra shown are the deconvoluted spectra assigned as follows: red; external aldimine I, blue; quinonoid I, green; external aldimine II, and orange; pyridoxal 5’-phosphate.

As described above, the final step k₄ was fixed to the value measured by transient state ¹⁹F-NMR spectrometry. Time-dependent ¹⁹F-NMR signals were fit using the SVD module of KinTeK Explorer software (Figure 3) and returned a rate constant of 0.00015 ± 0.00003 s⁻¹. This value is a net rate that describes a bifurcation and includes events for the decay of the external aldimine II species that liberates PLP (Scheme 2) and the rate of formation of the covalently modified hOAT. Prior data indicated that in any one turnover, 92% of the reaction yields PLP⁸; as such, the two rate constants diverging from the bifurcating step are summed from this rate and are k_cat = 0.00014 s⁻¹ and k_inact = 0.00001 s⁻¹. These values describe the rates for substrate and inhibitor roles of BCF₃, respectively. The dominant path for decay of external aldimine II reinstates the PLP form of the cofactor permitting subsequent turnovers with BCF₃. This net rate constant is four orders of magnitude slower than any other step which dictates that the BCF₃ turnover rate is defined by this value (k_cat). That no other spectrophotometrically conspicuous species, such as a subsequent quinonoid state, are observed suggests that the measured k_inact can be ascribed specifically to the first step in the inhibitory path for decay of external aldimine II.

Figure 3.

Transient-state nuclear magnetic resonance changes observed for hOAT reacting with BCF₃. hOAT (333 μM final) was mixed with BCF₃ (247 μM) and ¹⁹F spectra (16 scans per spectrum) were collected every 1 min for 8 h. The spectra were culled to approximate a logarithmic time base and were fit simultaneously to a linear A->B model that returned a rate constant of 0.00015 ± 0.00003 s⁻¹.

The transient-state single-turnover NMR spectral data report only on the slow steps of the hOAT BCF₃ reaction. The initial spectrum obtained at ~300 sec is assigned as external aldimine II. This state of the reaction has eliminated one fluorine (by integration) as a fluoride that is observed at −120 ppm (Figure S1). The elimination of the single fluorine conceivably could occur to yield the E or Z configuration about the exocyclic vinylic bond of external aldimine II. That a single set of NMR signals is observed at 300 sec indicates that the external aldimine II species is permitted free rotation within the active site (Figure S2). These signals resolve to products with two configurations by 480 min, both of which have a predicted mass equivalent to the mass determined for this product(s) in prior studies with BCF₃⁸. While the proportion of E and Z configurations of the turnover products cannot be known by these methods, the structure of the covalently inhibited state indicates that only the E configuration formed by nucleophilic attack by Lys292 is able to bring about irreversible hOAT inhibition (Scheme 2)⁸.

Crystal structure of hOAT inhibited by BCF₃:

Holoenzyme hOAT crystals were obtained in 3 days and had hexagonal prism morphology with the largest dimension of ~ 0.4 mm. The crystals were stable during soaking and diffracted to 1.96 Å resolution. After molecular replacement and refinement, the best refined molecular model had three monomers in an asymmetric unit in the P3₂21 space group (Table S1). The biological assembly of hOAT is a typical homodimer which can be reconstructed using the symmetrically related subunit in the crystal lattice (Figure S3).

The decay of external aldimine II is exceedingly slow compared to all prior steps observed. For this reason, we attempted to trap this intermediate by soaking. Crystallized hOAT was soaked with 1 mM BCF₃ (1 μL added to a 4 μL hanging drop) for 40 min and then frozen in liquid nitrogen. Given that the decay of the external aldimine II intermediate is rate limiting for both BCF₃ inactivation and turnover, the majority fraction of hOAT will accumulate as external aldimine II in the presence of excess BCF₃. The active site of the structure of hOAT soaked with BCF₃ is shown in (Figure 4) and compared with the structure of the covalently inhibited state of hOAT determined previously⁷ (pathway b, Scheme 2). The density observed in the active site was interpreted as the external aldimine II intermediate. The electron density supports that PLP has been covalently modified while the conserved Lys292 is detached from the PLP-inactivator adduct. The resolution of this structural data (1.96 Å) permitted assignment of the five remaining fluorine atoms in this adduct (Figure S4). At this stage of the reaction, Lys292 is available for a nucleophilic attack either at the PLP imine carbon or the carbon of a difluoromethylenyl group (pathway a and b respectively in Scheme 2).

Figure 4.

Comparison of the hOAT structures of the external aldimine II state and the covalently inhibited complex, external aldimine V. Left: crystal structure of external aldimine II (intermediate, PDB code: 7JX9). The proposed attack distances are shown as dashed lines in cyan and potential hydrogen bonding distances are shown as dashed lines in black. A simulated annealing composite map (2Fo-Fc) at 1.1 σ level is shown in gray color for Lys292 and the external aldimine II intermediate. Right: crystal structure of external aldimine V (final product, PDB code: 6OIA). A water molecule is labeled as WAT.

The external aldimine II intermediate interacts with key residues in the hOAT active site (Figure 4). The carboxylate group forms apparent hydrogen bonds with Tyr55 whose native function is to bind the α amino group of ornithine¹⁸. The distance between the carboxylate and the amino group of BCF₃ is shorter than the distance between the ε amino group and the α carboxylate of ornithine. To ensure the formation of external aldimine I, it is required that the carboxylate of BCF₃ be more proximal to PLP than ornithine. As a result, the conserved Arg180, which is proposed to interact with the α carboxylate group of ornithine, interacts instead with BCF₃ indirectly via an ordered water molecule (Figures S5 & 4). The difluoromethylenyl group is observed to interact with Arg413 and Gln266. The native function of Gln266 is to interact with PLP via the pyridyl 5’-hydroxyl group. In the native reaction Arg413 is prevented from interacting with the ornithine carboxylate by a salt bridge with Glu235. In the native OAT reaction this salt bridge can be broken by the dicarboxylic substrates, α-ketoglutarate or glutamate (Scheme 1). The trifluoromethyl group of the external aldimine II state is crowded against the aliphatic portion of the Glu235 side chain constraining its conformation away from complementary interaction with the guanidinium of Arg413. In addition to the aliphatic portion of Glu235, two other residues contribute to the binding of external aldimine II via hydrophobic interactions: Phe177 and Tyr85. Collectively, these interactions define an active site cavity that closely contours about the BCF₃ derived moiety, providing specific molecular recognition (Figure S6).

The primary difference between the crystal structure of the external aldimine II intermediate, and the final adduct is covalency of the syn-allylic carbon of what was initially the hexafluoropropan-2-ylidine group with Lys292. The structure of this intermediate indicates that the covalent linkage to Lys292 is yet to form. This unliganded lysine and the five remaining fluorine atoms are the defining features of the density that identifies the external aldimine II state. The loss of one fluorine atom from this structure correlates well with the fluoride ion observed to accumulate within the first scans in single turnover monitored by ¹⁹F-NMR spectrometry (Figure S2). Moreover, the free fluoride ion signal was unchanging throughout subsequent data acquisition, indicating that only one fluorine atom is displaced prior to the reformation of the PLP. The difluoromethylene group in external aldimine II was converted to form the amide bond in aldimine V linked to Lys292. The transition to the covalently inhibited state leaves the remaining trifluoromethyl group to interact with Arg413 and Gln236. Glu235 remained oriented away from interaction with Arg413 and no longer crowded against the trifluoromethyl group of the inactivator.

Integrating Discussion:

Formulating the mechanism for inhibition of hOAT by BCF₃ required the use of complementary time-dependent spectroscopic methods. These approaches provide redundancy to the cumulative observations facilitating definitive assignment of the reaction coordinate intermediates. hOAT reacts with BCF₃ to form an initial external aldimine species that accumulates in the deadtime of the stopped-flow instrument. The dependence of the decay of this species to the quinonoid provided a means of characterizing the equilibrium that forms the external aldimine species. These data indicate that the dissociation constant of this species is relatively weak (~3 mM), consistent with BCF₃ being a nonnative substrate. Decay of external aldimine I resulted in characteristic spectral transitions associated with a quinonoid species (Figure 2, Scheme 2). The fit of the formation and decay of this species returned rate constants that indicate ~75% accumulation of a quinonoid at the highest concentrations of BCF₃ with an extinction coefficient for this species at 567 nm of 36,000 M⁻¹cm⁻¹. These observations yielded a relatively good kinetic description for the steps that led to the formation of the quinonoid and its ensuing decay. It is the decay of quinonoid I that forms the quasi-stable and key intermediate for hOAT inhibition, external aldimine II. Prior characterization of the interaction of BCF₃ with hOAT identified that only 8% of any turnover reaction resulted in irreversible inactivation of the enzyme (pathway b in Scheme 2)¹³. Consequently, BCF₃ is for the majority fraction a substrate for hOAT, and only repeated turnovers with fractional partitioning will inactivate the entire enzyme population. Previous studies had also suggested that PLP is hydrolyzed from external aldimine V, leaving Lys292 covalently modified by BCF₃. The basis for this conclusion was the detection of PLP in inactivated hOAT solutions by mass spectrometry⁸. The origin of this PLP is now known to be from the substrate mode of BCF₃, as the final inhibited state of the enzyme was since shown by X-ray crystallographic characterization to be external aldimine V⁸.

Different interactions with the active site for the external aldimine II intermediate and the external aldimine V product indicate that the initial binding mode of BCF₃ is significantly different from the orientation of the final product (Figures 4 & S5). The structure reported herein for external aldimine II provides evidence of the initial binding orientation of BCF₃. It is likely that the carboxylate group of BCF₃ interacts with the hydroxyl group of Tyr55 throughout the inactivation reaction. This is the result of a shorter distance between the carboxylate and the amino group in BCF₃ compared to the α-carboxylate and the ε-amino group of ornithine. For the Schiff base formation between BCF₃ and PLP, the carboxylate pivots toward the PLP cofactor, interacting with Tyr55 instead of Arg180, which has been proposed to interact with the α-carboxylate of ornithine (Figure S5).¹⁸

In previous studies, two mono-trifluoromethyl derivatives similar to BCF₃ were tested as hOAT inhibitors (Figure 5).^{7, 8} The data indicated that the bistrifluoromethyl compound BCF₃ is a better inactivator than either the syn- or anti-monotrifluoromethyl compounds. Based on the structures reported in this work and previously⁸ (Figure 4), the role of the bistrifluoromethyl group in improved potency can be attributed to three factors. First, BCF₃ possesses a larger volume that better fills the active site cavity of hOAT, confining the chemical possibilities. Second, both trifluoromethyl groups contribute to binding in the initial complex. The anti-positioned trifluoromethyl group disrupts a salt bridge between Arg413 and Glu235. Arg413 in turn interacts initially with the syn-positioned trifluoromethyl group and the difluoromethylenyl group in the external aldimine II intermediate. Third, elimination of one fluoro group as fluoride, can yield either the E or Z isomer. The E and Z states of external aldimine II both advance on dominant pathway a and yield two products that differ only in their configuration about the exocyclic vinylic bond (Scheme 2). The structure of the covalently inhibited enzyme indicates that only the E-isomer of external aldimine II can advance to form the covalently inhibited complex. The remaining CF₃ group can reassume the role of interacting with Arg413, thereby contributing to the formation of the final adduct external aldimine V. As a direct consequence of these interactions, BCF₃ displays improved specificity against hOAT compared to prior drug candidates with similar structures such as gabaculine, canaline, and FCP.^{12, 19, 20}

Figure 5.

BCF₃ and previously developed mono-trifluoromethyl inhibitors. Top: Structures of the inhibitors. Bottom: Structures of the first intermediates (external aldimines) after initial binding.

The electronegative nature of fluorine atoms has substantial impact on the orientation of BCF₃ during initial binding and during the ensuing chemistry. The fluorine has a propensity to form hydrogen bonds that are weaker than those formed with oxygen or nitrogen atoms. Consequently, in many cases hydrophobic interactions with fluoro groups can have a significant impact on the stabilization of enzyme-substrate complexes compared to fluorine hydrogen-bonding interactions²¹. Vulpetti et al. hypothesized that relatively shielded fluorine atoms tend to form hydrogen bonds when bound to proteins, while relatively deshielded fluorines prefer hydrophobic interactions^{22, 23}. Initially, during inactivation of hOAT by BCF₃, the fluorine atoms of both trifluoromethyl groups are deshielded by proximity to the other fluorines. The anti-CF₃ group thus interacts with the aliphatic carbons of the Glu235 sidechain via hydrophobic interactions, breaking the salt bridge between this residue and Arg413. As the reaction proceeds, either the anti- or syn-CF₃ group loses one fluorine atom partially shielding the remaining two fluorines at this site. Therefore, the difluoromethyl group of external aldimine II preferentially forms hydrogen bonds with Arg413 and Gln266. (Figures 4 & S6). This interaction may in turn project weak shielding effects over the distant CF₃ group that would preferentially participate in hydrophobic interactions with Glu235. Conversely, in the external aldimine V state, the local chemical environment of the vicinal CF₃ group has changed. The amide linkage to Lys292 is expected to shield this group so that it prefers to form hydrogen bonds with Arg413 and Gln266.

According to the proposed chemical mechanism the irreversible inactivation of hOAT proceeds via a series of chemical oscillations between aldimine and quinonoid states (Scheme 2). The nucleophilic attack of Lys292 to the difluoromethylene of external aldimine II tethers only the E-configuration of the inhibitor and each subsequent decay of ensuing quinonoid state yields one fluoride ion. This repeating process ultimately denudes all fluorines from one of the CF₃ groups, solidifying the covalent link to Lys292 and eliminating a path to reinstate an active form of the cofactor. This oscillation terminates with a chemically isolated aldimine, external aldimine V (Figure 3). In contrast, in the dominant substrate pathway (pathway a) for BCF₃, Lys292 attacks the carbon of the PLP imine bond of external aldimine II and subsequent hydrolysis results in the regeneration of the active pyridoxal form of the active enzyme. The latter pathway does not regenerate BCF₃ so that multiple turnovers are required to bring about complete inactivation. The decay of aldimine II is slow, resulting in nearly complete fractional accumulation of this species in the steady state, which facilitated capture of this state in crystallo (Figures 4 & S4). This structure not only provides direct structural evidence for the proposed inactivation mechanism, but also a foundation for design of molecules that bias the partition of this species toward covalent inhibition. Collapse of the iminium ion of external aldimine II to form quinonoid II would require the Lys292 ε-amino group to be positioned at an angle of ~107° from the plane of the π-system of the difluoromethylenyl group²⁴. Although the crystal structure does indicate that the position of the ε-amino group of Lys292 is proximal (3.3 Å, Figure 4), it is in the plane of the difluoromethylenyl group and one of the fluorine atoms partially occludes the electrophilic carbon. While neither pathway is particularly favored by the observed intermediate’s position, it is these conformational restrictions that dictate the order of magnitude lower probability for the pathway that inactivates hOAT relative to PLP regeneration.

The above analysis serves as a revision of the previously proposed inactivation mechanism of hOAT by BCF₃. Within ~20 seconds the interaction of BCF₃ with hOAT would result in the accumulation of external aldimine II and inhibit the target enzyme effectively by stabilizing this state. Inactivation via aldimine V would require an excess amount of BCF₃ compared to the enzyme so that multiple turnovers bring about the irreversible inactivation⁷. The rate constants determined for the decay of external aldimine II state can be modelled using numerical integration to show that covalent inactivation requires ~20-fold excess of BCF₃ and then only asymptotically approaches complete inactivation over ~10 days. By itself, this slow rate of inactivation to aldimine V is unlikely to be effective in suppressing hOAT activity in vivo, as the high rate of expression in cancer cells is very likely to exceed the rate of covalent inhibition. Therefore, the primary and medically relevant mode of BCF₃ inhibition is trapping the vast majority of hOAT as aldimine II, and formation of aldimine V only modestly contributes to the long-term potency of the compound. In regard to iterative rational design, it may be desirable to create the next generation of drug candidates that bias the conversion from external aldimine II to quinonoid II, thereby altering the partition ratio and requiring less inhibitor for complete covalent inactivation. Transient inhibition of hOAT using stoichiometric doses of BCF₃ will trap the enzyme as aldimine II and return the majority fraction to an active state within a few days. Therefore, BCF₃ also has the potential to be used as a chemotherapy sensitizer²⁵ that transiently halts the activity of hOAT to increase the susceptibility of cancer cells to other chemotherapeutic agents.

Conclusions:

Conventional methods to characterize mechanism-based inhibition involving combinations of mass spectrometry and ligand co-crystallization are effective to both identify the final adduct and to propose overall mechanisms. However, for amenable systems transient kinetic studies are an effective additional method to reveal further mechanistic details. The characteristic intermediate spectra observed for hOAT reacting with BCF₃ and the slow kinetics under single turnover conditions provide considerably greater detail of the inactivation mechanism, prompting a revision of BCF₃’s potency that is now ascribed to the formation of the external aldimine II species and not the irreversible covalent external aldimine V species as previously described. The accumulation of this quasi-stable intermediate in a short time frame (~20 sec) provides a high steady-state concentration of this species such that crystal soaking could be employed to trap a near homogenous intermediate in the crystal structure during turnover. The crystal structure obtained provides the structural basis for rational design of future inhibitor/inactivator candidates. We confirm that BCF₃ is a prominent drug candidate that relies on its overall structure for specificity but requires an excess to ensure irreversible inactivation. Such a requirement is a result of a non-ideal attack angle during the conversion from external aldimine II to quinonoid II. However, the slow decay of aldimine II results in essentially all of the enzyme being trapped in this state and thus effectively inhibited soon after administration.

MATERIALS AND METHODS

Materials:

All chemicals were purchased from Millipore-Sigma, unless otherwise noted. All enzymes used for cloning were purchased from New England Biolabs.

Expression, and Purification of Human OAT:

hOAT was expressed and purified according to previously published protocols¹². See supporting information.

Transient State Methods:

The reaction of BCF₃ with hOAT was observed in transient state using three approaches. These methods captured signals for both relatively fast and slow chemical transformations. The first three steps in the reaction were captured using a Hitech Scientific (TgK) stopped-flow spectrophotometer in combination with charged coupled device detection (260–800 nm). In these experiments 9.84 μM hOAT was allowed to react with 2-fold serially diluted BCF₃ concentrations (63, 127, 254, 508, 1,016, 2,035, 4,065 μM) in 50 mM HEPES, 200 mM NaCl, pH 7.5 at 10 °C (all reported concentrations are after mixing in a 1:1 ratio).

Individual reaction traces at 567 nm for each BCF₃ concentration were extracted from the multiwavelength datasets and combined into a single file. These data were fit globally to the biege portion of the model depicted in (Scheme 3) using KinTek Explorer software (KinTek Corp). The CCD spectral datasets for singular value decomposition (SVD) deconvolution were collected at a concentration of BCF₃ equal to 2.08, 4.16 and 8.32 mM. Duplicate shots were acquired using a log time base for two timeframes (0.0025 – 12.4 seconds and 0.0025 – 2,480 seconds). Duplicate datasets for any one timeframe were averaged and spliced together at 12 seconds to form one dataset with sufficient time resolution to accurately describe fast and slow processes. The hybrid dataset was fit globally to the green portion of the model depicted in (Scheme 3) using the Spectrafit SVD module of KinTek Explorer software. In this model the rate constant for k₄ was fixed to the value measured using transient state NMR.

¹⁹F-NMR spectral observation of the reaction was used both to measure the slowest step observed and to verify the chemistry through spectral assignment of the species that decay and accumulate. A single turnover condition was established by mixing hOAT (333 μM) with BCF₃ (247 μM), and ¹⁹F-NMR spectra (16 scans per spectrum) were acquired at 1 min intervals for 8 hours (for acquisition parameters, see Supporting Information). The time of the first spectrum equated to ~300 seconds after the reaction commenced and so captures data for the external aldimine II species onward. Spectra were culled to approximate a logarithmic time base and to eliminate intervening portions of the ppm axis that contained no data. The time dependence of these NMR spectra were fit globally to a one-step model that is depicted as the blue portion of (Scheme 3).

Crystallization and Crystal soaking of hOAT with BCF₃:

Purified hOAT was buffer exchanged into 50 mM HEPES, pH 7.5 and concentrated to a protein concentration of 6 mg mL⁻¹. The previously reported crystallization conditions¹² were further optimized using the hanging drop vapor diffusion method by varying PEG 6000 (8–12%), NaCl (100–250 mM), and glycerol (0–10%) with a fixed buffer concentration of 100 mM Tricine, pH 7.8. For each hanging drop, either 2 or 3 μL of protein solution was mixed with equal volume of well solution. Several rounds of seeding were applied to improve crystal size and quality. The crystals with the best morphology and size grew in a final condition containing 10% PEG 6000, 100 mM NaCl, and 10% glycerol. BCF₃ (0.5 μL, 1 mM final) was added to the hOAT crystal drop and allowed to soak for 40 min. After soaking, crystals were transferred to a cryo-protectant solution (well solution supplemented with 30% glycerol and 1 mM BCF₃) and frozen in liquid nitrogen.

X-ray Diffraction and Data Processing:

Monochromatic X-ray diffraction data were collected at the LS-CAT beamline 21-ID-D at the Advanced Photon Source at Argonne National Laboratory. Data were collected at a wavelength of 1.127 Å and a temperature of 100 K using a Dectris Eiger 9M detector. Data sets were processed and analyzed with autoPROC software²⁶.

Model Building and Refinement:

The hOAT structure was solved by molecular replacement using PHASER²⁷ in Phenix. The starting search model was the previously published structure of hOAT (PDB code: 1OAT²⁸). The model building and refinement were accomplished in Coot²⁹ and Phenix³⁰, respectively, as an iterative process until the lowest possible R_free/R factor values were attained. Structural depiction figures were prepared using UCSF Chimera³¹.

Accession Code:

7JX9 (PDB code)