Rational Design, Synthesis, and Mechanism of (3S,4R)-3-Amino-4-(difluoromethyl)cyclopent-1-ene-1-carboxylic Acid Employing a Second-Deprotonation Strategy for Selectivity of Human Ornithine Aminotransferase over γ-Aminobutyric Acid Aminotransferase

Abstract

Human ornithine aminotransferase (hOAT) is a pyridoxal 5′-phosphate (PLP) dependent enzyme that contains a similar active site to that of γ-aminobutyric acid aminotransferase (GABA-AT). Recently, pharmacological inhibition of hOAT was recognized as a potential therapeutic target for hepatocellular carcinoma. In this work, we first studied the inactivation mechanisms of hOAT by two well-known GABA-AT inactivators (CPP-115 and OV329). Inspired by the inactivation mechanistic difference between these two aminotransferases, a series of analogues was designed and synthesized, leading to the discovery of analogue 10b as a highly selective and potent hOAT inhibitor. Intact protein mass spectrometry, protein crystallography, and dialysis experiments indicated 10b was converted to an irreversible tight-binding adduct (34) in the active site of hOAT, as was the unsaturated analogue (11). The comparison of kinetic studies between 10b and 11 suggested that the active intermediate (17b) was only generated in hOAT and not in GABA-AT. Molecular docking studies and pK_a computational calculations highlighted the importance of chirality and the endocyclic double bond for inhibitory activity. The turnover mechanism of 10b was supported by mass spectrometric analysis of dissociable products, and fluoride ion release experiments were carried out. Notably, the stopped-flow experiments were highly consistent with the proposed mechanism, suggesting a relatively slow hydrolysis rate for hOAT. The novel second-deprotonation mechanism of 10b contributes to its high potency and significantly enhanced selectivity over other aminotransferases.

Graphical Abstract

INTRODUCTION

Ornithine aminotransferase (OAT, EC2.6.1.13) is a pyridoxal 5′-phosphate (PLP)-dependent enzyme¹ that catalyzes two coupled transamination reactions (Figure 1).² In the first half-reaction, OAT catalyzes the conversion of PLP and ornithine to pyridoxamine phosphate (PMP) and glutamyl-5-semialdehyde (L-GSA), which spontaneously cyclizes to Δ1-pyrroline-5-carboxylate (P5C)² and can be further converted to L-proline by pyrroline-5-carboxylate reductase (PYCR).³ In the second half-reaction, PMP and α-ketoglutarate (α-KG) are converted to PLP and L-glutamate (L-Glu).² Recent evidence indicates that proline metabolism plays an important role in metabolic reprogramming to sustain cancer cell proliferation by the upregulated synthesis of P5C as a central intermediate.^4–6 Furthermore, the glutamate generated from hOAT can be converted to glutamine by glutamine synthetase (GS) to support de novo nucleotide biosynthesis and anabolic cell programs (Figure 1).⁷ Hepatocellular carcinoma (HCC) is the second leading cause of cancer death worldwide.^8–11 This disease is highly prone to malignancy and typically refractory to systemic treatment with the standard-of-care receptor tyrosine kinase inhibitor, sorafenib, and radiotherapy.^12–15 Human OAT (hOAT) and glutaminogenic enzymes were found to be strongly activated and commonly overexpressed in HCC due to oncogenic Wnt/β-catenin signaling.^{16, 17} hOAT has been recognized as a potential metabolic regulator of HCC progression via modulation of the flux through proline metabolic pathways.¹⁸ Metabolic reprogramming in HCC is characterized by hydroxyproline accumulation and accelerated proline consumption, which induces a hypoxia-inducible factor-1α (HIF1α) transcriptional program and sorafenib resistance.¹⁹ Encouragingly, the pharmacological inhibition of hOAT exhibited potent in vivo anti-tumor activity in the HCC mouse model, along with dramatically reduced α-fetoprotein (AFP, a biomarker for HCC) levels.²⁰ More recently, hOAT was also found to be upregulated in non-small cell lung cancer (NSCLC), and the specific knockdown of hOAT in NSCLC suppressed in vitro cell proliferation and in vivo tumor growth.²¹ Overall, hOAT is a promising therapeutic target for HCC and other related cancers.

Figure 1.

Metabolic role of hOAT

γ-Aminobutyric acid aminotransferase (GABA-AT) belongs to the same enzyme subgroup as OAT, demonstrating a similar active site and catalytic mechanism. Distinct from OAT, GABA is converted to succinic semialdehyde (SSA) in the first half-reaction of GABA-AT; however, these two aminotransferases share the same type of second half-reaction (Figure 1). Over the last few decades, our research group has been focusing on the development of mechanism-based inactivators (MBIs) of GABA-AT for the potential treatment of epilepsies and addictions.²² MBIs initially act as substrates and induce chemical transformations that form off-pathway intermediates that result in the inactivation of the target enzymes, mostly by formation of covalent or tight-binding complexes.²³ For example, (1S,3S)-3-amino-4-(difluoromethylene)cyclopentane-1-carboxylic acid (1, CPP-115, Scheme 1)^{24, 25} was designed and synthesized as a GABA-AT inactivator based on the structure and inactivation mechanism of vigabatrin, an FDA-approved drug. Later, a mechanistic study²⁶ demonstrated that Schiff base intermediate 2 was formed from 1, followed by its tautomerization to intermediate 4 (much as the native substrate does). The highly electrophilic intermediate (4) reacted with water molecules in the active site of GABA-AT, assisted by the catalytic lysine,²⁷ leading to the formation of tight-binding adduct 5 (Scheme 1). The mono-fluorine analogue (6a) inactivated GABA-AT via a similar pathway but resulted in formation of a tight-binding aldehyde adduct,²⁸ while the non-fluorine analogue (6b) failed to serve as a MBI of GABA-AT.²⁵ The introduction of a double bond (7, OV329) maintained the inactivation mechanism but greatly enhanced the potency, potentially as a result of the reduced acidity of the γ-proton.^{27, 29} Because of the structural similarity between these two aminotransferases, the aforementioned analogues also inactivated hOAT.^{20, 27} To improve the potency and selectivity toward hOAT over GABA-AT, six-membered ring analogue 8³⁰ was designed and synthesized, taking into account the relatively more flexible and larger active site of hOAT. Interestingly, analogue 8 was found to form a covalent adduct with attachments to both nearby Lys292 and *Thr322 (from the other subunit of a biological homodimer) in the catalytic pocket of hOAT, as indicated by cocrystal X-ray structures and protein mass spectra (MS).³⁰ Although analogue 8 displayed satisfactory inhibitory activities against hOAT with improved selectivity, it still inactivated GABA-AT at high concentrations.³⁰ Therefore, we set a goal to discover novel hOAT inactivators with more favorable selectivity over GABA-AT.

Scheme 1.

Mechanism of inactivation of GABA-AT by 1

Herein, we present the inactivation mechanisms of hOAT by analogues 1 and 7 via cocrystal X-ray structures, which reveal strikingly different inactivation mechanisms from those observed with GABA-AT. Based on the above observations, we rationally designed and synthesized analogues 9a-9b and 10a-10c, along with analogue 11 as a comparison. Among them, compound 10b was revealed as the most potent hOAT inactivator (k_inact/K_I = 4.73 mM⁻¹ min⁻¹) with excellent selectivity over other aminotransferases. We elucidated the inactivation and turnover mechanisms for 10b using various biochemical methods such as mass spectrometry (MS), protein crystallography, dialysis experiments, turnover experiments, and fluoride ion release experiments. Subsequently, a stopped-flow experiment was conducted for hOAT and 10b, for which the results were consistent with the mechanistic hypothesis and the proposed mechanism. These findings, based on the structural differences between the two aminotransferases, provide a novel strategy for the further rational design of selective hOAT inactivators.

RESULTS AND DISCUSSION

Dialysis and X-ray Crystallography of hOAT inactivated by 1 and 7

Analogues 1 and 7 were initially discovered as partially irreversible inhibitors of GABA-AT but later found to inactivate hOAT with relatively low efficiency (k_inact/K_I ratio).²⁷ It was found that they were converted to tight-binding adducts in the active site of GABA-AT via water molecule additions to the warhead of the Michael acceptor intermediates (Scheme 1).^{26, 27} However, the inactivation mechanisms for hOAT by 1 and 7 have remained unreported. The enlarged-ring strategy has proven to successfully enhance the potency and selectivity for hOAT, but six-membered ring analogues failed to prevent time-dependent inhibition of GABA-AT.^{30, 31} Notably, analogue 8 generated a covalent adduct in the active site of hOAT, even though the ring size is the only structural difference between 7 and 8. To elucidate whether the mechanistic difference results from the ring size (5 vs. 6) or the enzymatic machinery (hOAT vs. GABA-AT), we conducted dialysis experiments and X-ray crystallography with hOAT inactivated by 1 and 7.

After hOAT activity was partially or fully abolished by 2−17 equiv of 1 or 3−8 equiv of 7, it was dialyzed, and aliquots at different time intervals were collected and assayed for return of enzyme activity. Interestingly, no enzyme activity was recovered after 48 h of dialysis (Figure S1). The observation of complete irreversible inhibition against hOAT by 1 and 7 differs from the slow, partially irreversible inhibition of these molecules against GABA-AT.^{26, 27}

To investigate the inhibited states of hOAT with 1 and 7, we obtained cocrystal structures for both compounds incubated with hOAT. The hOAT cocrystals with 1 and 7 were grown via the hanging drop vapor diffusion method. The complex formed with 1 diffracted to 2.7Å (Figure 3B), while the hOAT cocrystal with 7 diffracted to 2.0Å (Figure 3A). Both structures were solved by molecular replacement (search model PDB code: 1OAT) and refined using Phenix.³² As shown in Figure 3, both compounds 1 and 7 formed covalent bonds with the catalytic Lys292 residue in the active site of hOAT, which differs from the noncovalent complexes observed for both in the case of GABA-AT.^{26, 27} Similar to its six-membered ring analogue (8)³⁰, compound 7 was converted to a covalent adduct (12, Scheme S1) with the attachments to nearby residues Lys292 and *Thr322 from the adjacent subunit (Figure 3A). Based on the previous report, a plausible inactivation mechanism for hOAT by 7 is proposed in Scheme S1. Michael acceptor intermediate S6 was principally formed because of the potential steric hindrance between the fluorine atom of the warhead and the internal H-bond in S3, followed by the sequential nucleophilic attacks from Lys292 and *Thr322 to form the final adduct (12). Notably, the endocyclic double bond is assumed to play an important role in transforming S2 to S5, as it was found in the mechanism for 8 and hOAT³⁰. In contrast, compound 1 generated a diamine adduct (13, Scheme S2) in the catalytic pocket of hOAT (Figure 3B), in which Lys292 was linked to the C4` position of PLP and one of the fluorine atoms was cleaved from the original warhead. Accordingly, a potential inactivation mechanism was proposed as shown in Scheme S2. As with compound 7, potential steric hindrance disfavored the conversion of 3 to Michael acceptor 4 and instead led to the formation of tautomer S11 in the absence of the endocyclic double bond. The final adduct (13) was generated by attack of Lys292 with release of a fluoride ion, which is reminiscent of the first step of the enamine inactivation mechanism for vigabatrin.²⁶ Interestingly, no water molecule was involved in the inactivation pathways for hOAT with either compound (Schemes S1 and S2). In the case of GABA-AT, two water molecules react with “unfavored” intermediate S9 or 4 (Schemes S1 and S2), leading to the generation of a dicarboxylate tight-binding adduct. Overall, both dialysis and crystallography results demonstrated that the mechanism differences between analogues 7 and 8 are derived from the difference in enzymatic machinery of the two aminotransferases rather than the ring size. Furthermore, we hypothesize that hOAT exhibits relatively low rates for the hydrolysis step than GABA-AT, which could be utilized in the rational design of selective hOAT inactivators.

Figure 3.

Co-crystal structures of hOAT inactivated by 7 (A, PDB ID: 7LNM) and 1 (B, PDB ID: 7TFP). For both crystal structures polder (F_o-F_c) maps are shown at 3.0 σ.

Design of Novel hOAT Inactivators

Aminotransferase inactivators usually form Schiff bases (aldimines) with PLP, followed by the conversion to active intermediates (ketimines) that lead to inactivation.²³ On the other hand, ketimines could be alternatively turned over to re-generate active enzyme after releasing the formed products. Based on the above observation and hypothesis, we speculated that a ketimine intermediate would be more stable in the active site of hOAT compared to GABA-AT, resulting in selective inhibition and greater potency against hOAT, possibly because of the slower rate of hydrolysis. The more stable ketimine has a better chance to be further elaborated by hOAT to generate an active intermediate and then lead to the specific inactivation observed. Thus, analogues 10a-10c were designed; their potential inactivation mechanisms are shown in Scheme 2. Schiff bases 14a-14c are initially generated from 10a-10c with the internal PLP aldimine complex, followed by deprotonation at the γ-position by the catalytic lysine; protonation at the C4` position of intermediates 15a-15c then affords tautomers 16a-16c. Based on the hypothesis that the water addition step is relatively slow in hOAT, intermediates 16a-16c should be relatively more stable and not readily hydrolyzed to give the corresponding ketones and PMP. Considering the electron-withdrawing effects of the nearby fluorine atoms and the imine moiety, the δ-proton could then be abstracted. The subsequent release of a fluoride ion would generate 17a-17b, which could react with nearby water molecules or residues, leading to the formation of either tight-binding or covalent adducts.

Scheme 2.

Possible inactivation mechanisms for 10a-10c

Unlike previous aminotransferase inactivators, two deprotonation steps are proposed to be involved during the inactivation process, in which the chirality of the γ/δ positions should play an important role. To evaluate the influence of the chiral centers, molecular docking studies for 14b/16b and their enantiomers (14b`/16b`) were conducted to examine binding poses at the catalytic pocket of hOAT. As shown in Figure S2A, these modelings indicate that the carboxylate moiety of 14b establishes stable hydrogen bonds with Tyr55 and Arg180, while its γ-proton is positioned close (3.4 Å) to catalytic Lys292. Intermediate 16b demonstrates a similar binding pose (Figure S2B), and the catalytic Lys292 is the closest basic residue (3.3 Å) to the δ-proton, indicating its potential involvement in the second deprotonation. Interestingly, the enantiomers of 14b and 16b maintain interactions with nearby residues, but the inversion of chirality forces their protons to face away from the catalytic lysine (Figures S2C and S2D, respectively). The docking results were consistent with the proposed inactivation mechanism, which prompted the synthesis of enantiomerically-pure analogues.

Synthesis

The synthetic route to analogues 9a-9b and 10a-10b is shown in Scheme 3. Ketones 18a/18b were obtained from the chirally-pure Vince lactam³³ via a known procedure.³⁴ The ketones were treated with 2-PySO₂CF₂H and ^tBuOK to give the difluoromethylene analogues 19a/19b.³⁵ Deprotection of PMB by CAN and protection with Boc₂O afforded intermediates 20a/20b. The desired product (9b) was obtained by selective hydrogenation to give 21, followed by ring-opening under acidic conditions. The selective hydrogenation of 20b and ring-opening under basic conditions yielded intermediate 22, which was subsequently converted to the desired product (10b). Iodine intermediates 23a/23b were obtained by hydrazone iodination and elimination under basic conditions. Treatment of the intermediates with CuI and MSFDA³⁴ yielded trifluoromethyl intermediates 24a/24b, respectively. Following a similar approach as 9b and 10b, desired products 9a and 10a were obtained from 24a and 24b, respectively. The synthetic route to analogues 10c and 11 is shown in Scheme 4. Ketone 18b was treated with PhSO₂CFHPO(OEt)₂ and LHMDS to give sulfonyl intermediate 29 as the major product. Deprotection of the sulfonyl group and PMB afforded intermediate 30, which was subsequently protected with a Boc group (31). Selective hydrogenation of 31 and ring-opening under basic conditions yielded intermediate 32, which was converted to analogue 10c under acidic conditions. Compound 11 was obtained by ring-opening of 31 to give 33, which was deprotected under acidic conditions.

Scheme 3.

Synthesis of analogues 9a-9b and 10a-10b

Conditions and reagents: (a) 2-PySO₂CF₂H, ^tBuOK, DMF, −60 °C; (b) CAN, CH₃CN, H₂O, r.t., 4 h; (c) Boc₂O, DMAP, DIPEA, DCM, r.t., overnight; (d) Pd(OH)₂, MeOH, r.t., overnight; (e) HCl (4M), AcOH, 80 °C, overnight; (f) K₂CO₃, MeOH, r.t., 2 h; (g) N₂H₄ (aq. 55%), EtOH, reflux, 3 h; (h) I₂, Et₃N, benzene, r.t., 6 h; (i) ^tBuOK, THF, 0 °C - r.t., overnight; (j) MFSDA, CuI, NMP, 160 °C, 1 h; (k) MFSDA, CuI, DMF, 2,5-lutidine, 100 °C, 4 h.

Scheme 4.

Synthesis of analogues 10c and 11

Conditions and reagents: (a) PhSO₂CFHPO(OEt)₂, LiHMDS, THF, −78 °C; (b) HgCl₂, Mg, MeOH, 0 °C-r.t.; (c) CAN, CH₃CN, H₂O, r.t., 4 h; (d) Boc₂O, DMAP, DIPEA, DCM, r.t., overnight; (e) K₂CO₃, MeOH, r.t., 2 h; (f) HCl (4M), AcOH, 80 °C, overnight; (g) Pd(OH)₂, MeOH, r.t., overnight.

Kinetic Studies

As shown in Table 1, analogues 10a and 10b demonstrated time-dependent inhibitory activities against hOAT, whereas analogues 9a, 9b, and 10c showed either no or only weak inhibition at a concentration of 10 mM. Considering the structural similarity among the analogues, the difference in potency may result from the electron-withdrawing effects of the fluorine atoms and the conjugated carboxylate during the deprotonation steps. Among them, the most potent compound (10b, k_inact/K_I = 4.73 min⁻¹mM⁻¹) is 5.4 times more efficient as an inactivator of hOAT than 6c (Figure 2, k_inact/K_I = 0.88 min⁻¹mM⁻¹), a compound that exhibited potent in vivo antitumor efficacy.²⁰ Satisfactorily, analogue 10b demonstrated weak inhibitory activity against other human aminotransferases (Asp-AT, Ala-AT, and GABA-AT) even at high concentrations (Figure S3). For a better understanding of the capabilities of this aminotransferase and for future rational design of new inactivators, the inactivation mechanism of 10b was studied by dialysis experiments, intact protein MS, and X-ray structures of co-crystallized complexes.

Table 1.

Kinetic constants for the inactivation of hOAT by 6c, 9a-9b and 10a-10c^a

Compound	hOAT
	KI (mM)	kinact (min−1)	kinact/KI (mM−1min−1)
9a	-b
9b	-b
10a	0.048 ± 0.011	0.072 ± 0.005	1.50
10b	0.022 ± 0.004	0.104 ± 0.005	4.73
10c	47% inhibition @ 10 mM
6c 31	0.065 ± 0.010	0.057 ± 0.003	0.88

^ak_inact and K_I values were determined by the equation: k_obs = k_inact*[I]/(K_I +[I]) and are presented as means and standard errors.

^bNo inhibition at 10 mM concentration

Figure 2.

Structures of GABA analogues 1, 6–11

Intact Protein MS and Dialysis for hOAT Inactivated by 10b

Intact protein MS is an efficient approach to distinguish inactivation mechanisms for aminotransferases with molecular specificity.^{30, 31, 36, 37} Indeed, if inactivation of an enzyme proceeds through a covalent modification pathway, a mass shift corresponding to the molecular weight of the adduct would be observed relative to the native, untreated enzyme. However, noncovalent inactivation adducts are lost under the denaturing liquid chromatography conditions used by this technique. After complete inactivation of hOAT by 10b, the inactivated enzyme exhibited the same intact mass as the untreated enzyme when analyzed by denaturing MS (Figure S4), which indicates the formation of a tight-binding adduct rather than a covalent adduct.

Difluoromethylene analogues 1 and 7 were shown to form tight-binding adducts in the active site of GABA-AT that resulted in only partially irreversible inhibition because of hydrolysis of the ketimine intermediates.^{26, 27} However, these molecules appear to be irreversible inhibitors of hOAT, forming stable covalent adducts (Figure S2). Likewise, for 10b, no enzyme activity was recovered after 91 h of dialysis when treated with varying equivalents of 10b (Figure S1).

X-ray Crystallography of hOAT inactivated by 10b or 11

The intact protein MS and the dialysis experiment suggested analogue 10b inactivates hOAT via the generation of a tight-binding adduct. As shown in Figure S5, analogue 11 was found to inactivate hOAT with a similar potency (k_inact/K_I = 2.97 min⁻¹mM⁻¹) to that of analogue 10b (k_inact/K_I = 4.73 min⁻¹mM⁻¹). To better elucidate the inactivation mechanism, protein crystallography of hOAT inactivated by 10b and 11 was conducted according to the same procedure as that for hOAT-7. hOAT-10b crystals diffracted to 1.9 Å resolution and hOAT-11 crystals diffracted to 2.6Å resolution. Both structures were solved by molecular replacement (search model PDB code: 1OAT) and were refined using Phenix³².

The refined models for hOAT-10b and hOAT-11 are shown in Figure 4 (polder maps). Both inactivators are covalently linked to the PLP but free from Lys292 and other active site residues. In both structures, the aldehyde group of the ligand (34, Figure 4) forms a hydrogen bond with Gln266. One of the oxygen atoms on the carboxylate group of both inactivators forms a strong hydrogen bond with Tyr55 (≤ 2.5Å), along with interactions with Arg180, resulting in high stability for the protein-ligand system. Overall, the two cocrystal structures are highly similar, except for the one water molecule observed close to the carboxylate group in the case of hOAT-10b (Figure S6), although fewer water molecules are resolved due to lower resolution. Recent work by Carugo et al. has shown that at least a resolution of 1.6Å is required to observe a continuous hydration layer at the protein surface.^{38, 39} Thus, a water molecule resolved at 1.9Å in the hOAT-10b cocrystal structure could be present, but not observed at 2.6Å in the cocrystal of hOAT-11. Several possible tautomers of hOAT-10b and hOAT-11 products were built into the model and refined using the same strategy. Among them, endocyclic adduct 34 (Figure 4C) was chosen since it had the lowest B factors for the ligand in all three subunits within the asymmetric unit, although other possible tautomers cannot be dismissed.

Figure 4.

A) Co-crystal structure of hOAT inactivated by 10b (PDB ID: 7TEV); B) Co-crystal structure of hOAT inactivated by 11 (PDB ID: 7TED); for both crystal structures polder (F_o-F_c) maps are shown at 3.0 σ. C) Structure of final adduct 34

On the basis of the proposed inactivation mechanism for 1 and 7, active intermediate 17b was assumed to be formed from analogue 11 in the active site of hOAT (Scheme S3), followed by water attack to afford tight-binding adduct 34 (Figure 4C). This result is consistent with the hypothesis that there is a potential for steric hindrance between the fluorine of the warhead and the internal H-bond in the cases of difluoromethylene analogues 1 and 7. Considering the similarity between the above cocrystal structures, the same intermediate (17b) is expected to be generated during the inactivation of hOAT by 10b via the proposed mechanism (Scheme 2). The subsequent attack by water on the fluorinated methylene leads to the formation of final adduct 34 (Figure 4C).

Effects of Endocyclic Double Bonds on Inhibitory Activities

The above experiments suggested final adduct 34 was generated during the inactivation process of hOAT by 10b, which involved a second deprotonation step to form active intermediate 17b. With the exception of an endocyclic double bond, 9b is identical to 10b, yet 9b demonstrated no inhibitory activity against hOAT up to a concentration of 10 mM. To evaluate the influence of the endocyclic double bond, we conducted molecular docking studies for the intermediates of 9b and calculated the theoretical pK_a values for the protons at the γ/δ positions using the hybrid DFT/B3LYP method⁴⁰. As shown in Figures S2 and S7, intermediates S15 and S16 maintain the same binding poses as their corresponding olefin intermediates 14b and 16b, in which the catalytic Lys292 residue shows a similar accessibility to the γ/δ protons for the deprotonation steps (Figure S2). However, the endocyclic double bonds have dramatic effects on the pK_a of the γ-protons for intermediates 14b and S15 (7.71 vs 8.42) and the pK_a of the δ-protons for intermediates 16b and S16 (5.32 vs 7.67) (Figure S7). The deprotonation step usually plays an important role and is generally the rate-determining step in the inactivation mechanism for an aminotransferase inactivator²², and this series of analogues requires two deprotonation steps to generate the active intermediates. Considering that 6a was recognized as a MBI of hOAT,²⁰ a similar Michael acceptor might not be generated from 9b as is derived from 10b, which could result from the significantly reduced acidity of the γ/δ positions in the absence of an endocyclic double bond.

Turnover Mechanism

MBIs typically act as substrate analogues for target enzymes and often bifurcate, such that they are fractionally converted to dissociable products during the inactivation process.²³ Analogue 10b was shown to generate stable tight-binding adduct 34 via tautomerization, HF elimination, and water attack (Scheme 2, Scheme 5). Accordingly, three possible turnover pathways (a-c) were proposed based on hydrolysis occurring at different stages as shown in Scheme 5, along with the release of PMP and products 36, 37, and 38, respectively. To identify which turnover pathway is dominant, we carried out partition ratio and fluoride ion release experiments, along with MS analysis of products.

Scheme 5.

Proposed turnover mechanism of 10b by hOAT or GABA-AT

The partition ratio is the ratio of turnover to inactivation, which is calculated by titrating the enzyme with varying equivalents of the inactivator. Since this number includes the one molecule of inactivator required to inactivate one enzyme monomer, the partition ratio is equal to the number of turnovers minus one. hOAT was incubated with varying equivalents of 10b, and from the remaining activities the partition ratio was determined to be 2.38 (Figure S8).

Different equivalents of fluoride ions would be released in turnover pathways a-c. According to the partition ratio and the inactivation mechanism, the theoretical equivalents of fluoride ions released per active site via different turnover pathways in the presence of α-KG can be calculated. As shown in Table S1, pathway a would release only 2.0 equivalents of fluoride ions per active site from the inactivation of hOAT, while pathways b and c would release 4.38 and 6.76 equivalents, respectively. A fluoride ion-selective electrode was used to determine that 4.42 equivalents of fluoride ions were released after inactivation (Table S1), which is consistent with the theoretical number for pathway b leading to 37.

Different products also are released by turnover pathways a-c and could be distinguished by untargeted LC-HRMS and confirmed by tandem MS. However, none of the above products (36-38) was detected by LC-HRMS in the 10b-inactivated hOAT sample, possibly because of poor ionization or chemical instability. As indicated from the number of fluoride ions released, product 37 is most likely to be generated, which contains a highly electrophilic Michael acceptor. Therefore, to improve the sensitivity to detect this potential product, β-mercaptoethanol (β-ME) was added during the incubation of hOAT and 10b. This additive yielded the mass of product 39 which was further confirmed by its unique isotopic distribution and fragmentation spectrum (Figure 5), confirming the release of product 37 in pathway b.

Figure 5.

Confirmation of product 39 by high-resolution MS. A) Structure, mass, and abundance of product 39 detected by HRMS within a 2 ppm window with and without β-ME treatment. B) Theoretical and experimental mass and isotope distributions of product 39. C) HCD fragmentation spectrum for m/z 215.037 as confirmation of product 39.

Analogue 10b could also be degraded by GABA-AT via the above turnover mechanisms (Scheme 5). Notably, analogue 11 displayed similar time-dependent inhibitory activities (Figure S5) against GABA-AT (k_inact/K_I = 2.51 min⁻¹mM⁻¹) as hOAT (k_inact/K_I = 2.97 min⁻¹mM⁻¹). Considering the similar structure of 11 to analogues 6a and 7, it probably inactivates GABA-AT via Michael acceptor intermediate 17b (Scheme S3). However, 10b is structurally similar to 11 but was identified as a weak reversible inhibitor of GABA-AT (Figure S3), indicating that intermediate 17b may not be formed by 10b in the active site of GABA-AT. Thus, for the turnover mechanism of 10b by GABA-AT, the simplest explanation is that 16b is rapidly hydrolyzed to release 36 and PMP rather than being converted to 17b. This result is consistent with our hypothesis that a relatively slow hydrolysis step occurs for GABA analogues with hOAT than for GABA-AT.

Plausible Mechanism for 10b

Based on the above inactivation and turnover mechanism studies, a modified pathway for 10b with hOAT and GABA-AT is proposed in Scheme 6. Initially, analogue 10b reacts with the Lys-PLP complex, as native substrates do, to generate Schiff base 14b. The ensuing abstraction of the γ-proton gives 15b, and the re-protonation at the PLP-C4` position yields ketimine 16b. In the case of hOAT (pathway a), deprotonation occurs at the δ position by catalytic residue Lys292, along with the release of fluoride ion via either an E1cB or E2 elimination mechanism, to form Michael acceptor intermediate 17b. This second deprotonation could result because of the slower hydrolysis of ketimine 16b by hOAT compared with GABA-AT, indicated by the mechanistic difference between 1 and 7 with these two aminotransferases. Molecular docking studies and computational calculations of pK_a values indicate that the chirality of the γ/δ position and the presence of the endocyclic double bond play critical roles in the deprotonation steps. The water attack on the fluoromethylene group of 17b (pathway c) leads to the formation of tight-binding adduct 34, which accounts for ~30% of the reaction according to the partition ratio (1/3.38). The structure of the final adduct was well supported by the intact protein MS and the X-ray crystal structure of hOAT inactivated by 10b. The remaining ~70% of 17b undergoes ketimine hydrolysis to release product 37 and PMP (pathway d), suggested by the fluoride ion release experiment (4.42 equiv) and untargeted LC-HRMS. Intermediate 16b is assumed to be formed in the active site of GABA-AT, but it is quickly hydrolyzed (pathway b) rather than being converted to active intermediate 17b, which is suggested by a comparison with analogue 11 in the kinetic studies. Therefore, the mechanistic differences observed for 10b may result from changes in the hydrolysis rates for these two aminotransferases.

Scheme 6.

Plausible mechanisms for 10b with hOAT and GABA-AT

Transient State Measurements of hOAT Inhibited by 10b

As shown in Scheme 6, various transient states are proposed to be involved in the mechanism of hOAT inhibition by 10b. For a better interpretation of this process, we performed rapid mixing absorption measurements to detect spectrophotometric evidence for the intermediate sequence. Initially, singular value decomposition analysis was performed on a spliced composite data set collected from two-time frames using a charge-coupled device (CCD) for a single concentration of 10b (500 μM). The model-free analysis indicated the presence of five components, but one of them was deemed to be noise and was culled. The data were thus fit to a three step, four species linear irreversible model (Figure 6). The wavelength of components (Int. Ald, Ext. Ald, M1, M2 and P1) observed in the spectra matched well with the corresponding intermediates (Lys-PLP, 14b, 16b, 17b, and 34 + PMP) proposed in Scheme 6 (see Scheme 7). To further investigate the wavelength of final adduct 34 and the composition of P1, it was mixed with excess amounts of α-KG (250 μM final) in the presence and absence of excess 10b, correspondingly. These data indicated that peaks at ~330 nm and ~380 nm in P1 were both increased when treated with excess 10b and α-KG (Figure 7, P2). This was interpreted as the reverse and forward half-reactions, consuming the residual PMP by conversion to PLP and successive fractional conversions to 34 that ultimately lead to complete inactivation. This suggested that there are two absorption maxima for 34, possibly as a result of its high conjugation (Scheme 7, P2). On the other hand, the peak at ~330 nm was greatly decreased, and the peak at ~380 nm was shifted toward the internal aldimine (~420 nm) when only treated with excess α-KG (Figure 7, P3), which could be explained by the dominant conversion of PMP to PLP with limiting 10b (Scheme 7, P3). Overall, the deconvoluted spectra are highly consistent with the proposed mechanism.

Figure 6.

Spectral deconvolution of the rection of hOAT with 10b. A) hOAT (16.1 μM final) was mixed with 10b (500 μM final), and spectra were recorded with a logarithmic spacing for two time frames: 0.0025 – 12.4 sec and 0.0025 – 1280 sec. These datasets were spliced together at 12.4 sec, and the combined dataset was deconvoluted by fitting to a linear three-step model using singular value decomposition. B) Deconvoluted, noise-filtered spectra. The progression of species is indicated in the inset, and the spectrum of the resting internal aldimine of hOAT is shown in black and represents the zero-time spectrum.

Scheme 7.

Transient kinetics of the reaction of 10b with hOAT

Figure 7.

Spectra observed during inactivation. The black spectrum is the resting PLP state of hOAT, included here for reference. The orange spectrum is hOAT after reaction with excess 10b. The purple spectrum is obtained when the form shown in orange is allowed to react with both α-KG (250 μM) and 10b (250 μM) for 250 sec and presumably is the product of multiple turnovers in the forward and reverse directions that ultimately leads to complete covalent inhibition of hOAT. The blue spectrum is a successive composite state observed when the orange species is allowed to react with α-KG (250 μM) alone.

After confirmation of proposed components by spectroscopy for the reaction of hOAT with 10b, we measured the rate constant for each step. Single wavelength traces extracted from CCD detector spectral datasets were fit to linear combinations of two exponentials based on pseudo-first order enzyme:inhibitor ratios. The data at 320 nm and 410 nm report principally on the formation of an intermediate state and the decay of the PLP forms of the enzyme, respectively. In each case, the subsequent phase incorporated the contribution of additional small amplitude changes that were poorly resolved at these wavelengths. For the case of 410 nm (Figure S9), the dependence of the observed rate constants indicated that the rate of the first phase titrated hyperbolically with the concentration of analogue 10b with a limit of 1.69 ± 0.15 s⁻¹, which is the net rate constant for the formation of M1 (k₂’, Scheme 7), as well as a dissociation constant of 2.48 ± 0.54 mM (K_d) for 10b combining to form external aldimine 14b (Scheme 7, Figure S9B). These data were interpreted as a reversible and weak association of the inhibitor with the internal aldimine form of hOAT, followed by the latter step (k₄, Scheme 7) that converts the enzyme to the PMP state (P1). The dependence of the observed rate constant for the second phase showed no clear trend with data scattered about an average of 0.02 s⁻¹ (Figure S9C), which was assigned to the formation of P1 from intermediate M2 (Scheme 7) for this reaction that predicts only absorption changes for k₂ and k₄ at this wavelength (see below). The data obtained at 320 nm report on the latter steps in the forward reaction of hOAT with 10b (Figure S10). The dependence of the observed rate constant at this wavelength is also described by a hyperbolic curve according to equation 2 (k_1obs = k₁[10b]/(K_10b + [10b]), indicating the influence of reversibility in the preceding step (k₂, k_−2,) that consumes the external aldimine. The limit of the dependence indicates a net rate constant (k₃’, Scheme 7) of 0.26 ± 0.10 s⁻¹ for the formation of intermediate M2. The data at this wavelength also showed small increases in optical density beyond ~500 sec, which were not assigned in this analysis. Notably, the fit of the CCD data sets (500 μM of 10b) indicated successive rate constants of 0.22, 0.21, 0.03 s⁻¹ qualitatively in agreement with the observed rate constants for k₂’, k₃’, k₄ (Figures S9 and S10). The spectra obtained are shown in Figure 6B and are overlaid with the internal aldimine spectrum acquired from the resting enzyme that serves as a representation of the time zero state of the reaction. As shown in Scheme 7, this sequence of spectra combined with the concentration dependencies indicate that the external aldimine (14b) forms an equilibrium accumulation rapidly and reversibly within the deadtime of the stopped-flow instrument with a weak binding constant of ~2.5 mM (K_d). The first phase observed is the decay of the external aldimine (M1) with a rate constant of ~1.7 s⁻¹ (k₂’) to yield a weakly absorbing intermediate state that then decays at ~0.26 s⁻¹ (k₃’) to form a second intermediate species (M2) with a prominent shoulder at 320 nm. This state then decays at ~0.02 s⁻¹ (k₄) to form the PMP state (P1) of the enzyme (Figure 6B, Scheme 7). Although abstraction of the γ-proton was previously proven to be the rate-determining step for the reaction of GABA-AT and inactivators, kinetics measurements for the reaction of hOAT and 10b showed that the hydrolysis step (k₄) from M2 to P1 is much slower than the other two deprotonation steps (k₂’ and k₃’), which supports our hypothesis that a relatively slow hydrolysis step in the catalytic process of hOAT might be responsible for the difference between the inactivation mechanisms with hOAT and GABA-AT.

CONCLUSIONS

Human ornithine aminotransferase (hOAT) is a pyridoxal 5′-phosphate (PLP) dependent enzyme that has a similar active site to that of γ-aminobutyric acid aminotransferase (GABA-AT). Over the last few years, selective inhibition of hOAT has been recognized as a potential treatment for cancers, especially hepatocellular carcinoma (HCC). In this work, we first elucidated the inactivation mechanisms of hOAT by two well-known GABA-AT inactivators, CPP-115 (1) and OV329 (7). Interestingly, irreversible covalent adducts (12 and 13) were generated from them in the active site of hOAT, while 1 and 7 were identified as partially irreversible inhibitors of GABA-AT with the formation of noncovalent, tight-binding adducts. This observation might result from a potential enzymatic machinery difference between these two aminotransferases leading to a relatively slow hydrolysis rate with hOAT. Inspired by the above findings, a series of analogues (9a, 9b, and 10a-10c) were designed and synthesized. Among them, the best compound (10b, k_inact/K_I = 4.73 min⁻¹mM⁻¹) is 5.4 times more efficient as an inactivator of hOAT than 6c (k_inact/K_I = 0.88 min⁻¹mM⁻¹), which exhibited potent in vivo antitumor efficacy. Furthermore, analogue 10b demonstrated weak reversible inhibitory activity against other human aminotransferases (Asp-AT, Ala-AT, and GABA-AT), even at high concentrations. Intact protein mass spectrometry, protein crystallography, and dialysis experiments showed that 10b was converted to active intermediate 17b via a second-deprotonation process, leading to the formation of a tight-binding adduct (34) and the irreversible inhibition of hOAT. Notably, the chiral centers and the presence of the endocyclic double bond played important roles in the inactivation process as indicated by molecular docking studies and pK_a theoretical calculations. The turnover mechanism of 10b was supported by mass spectrometric analysis of the products and fluoride ion release experiments, suggesting that the inactivation and turnover processes were determined by water molecule attack at different electrophilic centers of active intermediate 17b. Interestingly, the same active intermediate could not be generated in the active site of GABA-AT, indicated by a comparison with analogue 11. To further elucidate the mechanistic details of hOAT and 10b, we carried out stopped-flow experiments, which revealed the identity of intermediates and reaction rates for each step. Not only was this result highly consistent with the proposed mechanism (Scheme 6) but it also identified the slow step in the mechanism as the hydrolysis of 17b for hOAT, which supports our hypothesis for why the mechanisms of inactivation of hOAT are different from those with GABA-AT, as we demonstrate for inactivators 1 and 7 (Schemes S2 and S1, respectively). The novel second-deprotonation mechanism for 10b also contributes to its high potency and significantly enhanced selectivity over other aminotransferases, especially GABA-AT.

ABBREVIATIONS

^tBuOK

potassium tert-butoxide

CAN

cerium (IV) ammonium nitrate

Boc₂O

di-tert-butyldicarbonate

DMAP

4-dimethylaminopyridine

DIPEA

N, N-diisopropylethylamine

DCM

dichloromethane

MFSDA

methyl fluorosulfonyldifluoroacetate

NMP

N-methylpyrrolidone

THF

tetrahydrofuran

β-ME

β-mercaptoethanol