Apple 1-Aminocyclopropane-1-carboxylate Synthase in Complex with the Inhibitor l-Aminoethoxyvinylglycine

The 1.6-Å crystal structure of the covalent ketimine complex of apple 1-aminocyclopropane-1-carboxylate (ACC) synthase with the potent inhibitorl-aminoethoxyvinylglycine (AVG) is described. ACC synthase catalyzes the committed step in the biosynthesis of ethylene, a plant hormone that is responsible for the initiation of fruit ripening and for regulating many other developmental processes. AVG is widely used in plant physiology studies to inhibit the activity of ACC synthase. The structural assignment is supported by the fact that the complex absorbs maximally at 341 nm. These results are not in accord with the recently reported crystal structure of the tomato ACC synthase AVG complex, which claims that the inhibitor only associates noncovalently. The rate constant for the association of AVG with apple ACC synthase was determined by stopped-flow spectrophotometry (2.1 × 105 m −1 s−1) and by the rate of loss of enzyme activity (1.1 × 105 m −1 s−1). The dissociation rate constant determined by activity recovery is 2.4 × 10−6 s−1. Thus, the calculatedK d value is 10–20 pm.

The plant hormone ethylene, which regulates many aspects of plant growth, development, and senescence is synthesized in the Yang cycle (1). The biosynthetic precursor of ethylene, 1-aminocyclopropane-1-carboxylate (ACC), 1 is produced by a pyridoxal phosphate (PLP)-dependent enzyme, ACC synthase (S-adenosyl-L-methionine methylthioadenosine lyase, EC 4.4.1.14). This enzyme catalyzes the ␣,␥-elimination of methylthioadenosine (MTA) from S-adenosyl-L-methionine (SAM) to produce ACC (Scheme 1). The production of ACC from SAM is the committed and rate-determining step in ethylene biosyn-thesis; therefore, the reaction and inhibition mechanisms of ACC synthase attract considerable interest for agricultural applications. Among the most potent inhibitors of ACC synthase is L-aminoethoxyvinylglycine (AVG) (2), which is widely used for plant physiology studies and agricultural applications. The structure of unliganded ACC synthase from Malus domestica to 2.4-Å resolution is available (3). Here we describe the structure of the complex of apple ACC synthase with AVG at 1.6-Å resolution together with the rate constants characterizing its formation and decomposition. The structure results disagree with those reported by Huai et al. (4) at lower resolution (2.7 Å) on the tomato ACC synthase AVG complex.

EXPERIMENTAL PROCEDURES
Protein Purification and Crystallization-WT and E47Q ACC synthase were purified for biochemical studies as described previously (5). Recombinant V435STOP apple ACC synthase was prepared as described previously (6) and crystallized at 293 K by the vapor diffusion method (sitting drop). 1 l of protein solution containing 15 mg/ml enzyme, 50 mM HEPES, pH 7.9, 10 M PLP, 1 mM DTT, and 4 mM AVG was mixed with an equal amount of reservoir containing 30% 2-methyl-2,4-pentanediol (MPD) (v/v) and 50 mM MES, pH 6.5.
The space group of the crystals is C2 with cell parameters a ϭ 103.3 Å, b ϭ 61.1 Å, c ϭ 77.1 Å, and ␤ ϭ 123.4°. The asymmetric unit contains a monomer of ACC synthase, corresponding to a solvent content of 38%.
Data Collection and Processing-The data to 1.6 Å were collected at 100 K using a MAR Research (Hamburg, Germany) imaging plate at the Swiss-Norwegian Beamline at the European Synchrotron Radiation Facility (Grenoble, Switzerland). Data were reduced with DENZO and other programs of the CCP4 suite (7).
Structure Solution and Refinement-The structure was solved by molecular replacement with AMoRe (8) using one subunit from the native apple ACC synthase structure (Protein Data Bank code 1B8G) (3). The structure was refined with X-PLOR and CNS (9), alternating cycles of torsion angle-simulated annealing and individual B-factor refinement with sessions of manual rebuilding with O (10). The data between 20 and 1.6 Å were used. No -cutoff was applied. A mF o Ϫ DF c map calculated after the first round of simulated annealing and thermal refinement exhibited clear difference density corresponding to a covalent adduct between the PLP cofactor and AVG (Fig. 1). A model for the adduct (residue code PPG) was built into the density and included in the refinement. Subsequently, 332 water molecules and a molecule of MPD were added to the model. Refinement was performed with X-PLOR 3.85 with topology and parameter files TOPHCSDX.PRO and PARHCSDX.PRO (11).
The final model was validated by WHAT_CHECK (12) and Procheck (13). The final refinement statistics are given in Table I. The atomic coordinates have been deposited with the Protein Data Bank with entry code 1M7Y.
Spectroscopic Characterization of the AVG⅐PLP Complex-Steadystate absorbance spectra were recorded with an Agilent Technologies 8453 UV-visible diode array spectrophotometer. Absorbance values at 421 and 341 nm (the PLP-aldimine and ketimine species, respectively) measured during the course of reactivation of AVG-inactivated ACC synthase were corrected for changes in protein concentration over time by normalization of the absorbance of each sample at 280 nm.
Rapid reaction studies monitoring the rate of formation of the PLP-ketimine species upon incubation of WT and E47Q ACC synthase with AVG were carried out with a temperature-controlled Applied Photo-Physics model SF.17MV stopped-flow spectrophotometer. Ketimine formation was monitored at 341 nm, and data were fit to first-order kinetics.
Kinetic Characterization of ACC Synthase Inactivation by AVG-ACC synthase activity in the presence of AVG was monitored by a continuous coupled assay (14). Reactions were initiated by the addition of ACC synthase to mixtures of AVG, SAM, and coupling enzyme preequilibrated in buffer at 25°C. Resulting reaction progress curves were fit to Equation 1 (15), where P is the product concentration, v s and v o are the steady-state and initial reaction velocities, respectively, and k obs is the observed rate constant for inactivation. The rate constant for association of the enzyme with AVG was determined from a plot of k obs versus inhibitor concentration.

RESULTS
Structure Determination-The structure of V435 STOP apple ACC synthase, cocrystallized in the space group C2 with AVG, was determined by molecular replacement with the unliganded enzyme structure as search model. The crystals contain one monomer of the ACC synthase dimer per asymmetric unit with the two subunits related by a crystallographic 2-fold axis. The structure was refined to 1.6-Å resolution and an R-factor of 20.5% (R-free 21.9%) with excellent stereochemical parameters (Table I). Difference Fourier map analysis from the first step of refinement revealed that AVG is bound covalently to PLP (Fig. 1). Lys-273 is clearly displaced from the cofactor. The distance between the C4A atom of PLP and the side-chain amino group of the lysine is 3.7 Å. The final refined model for the apple ACC synthase-AVG complex encompasses 424 residues, the covalent adduct AVG⅐PLP, a molecule of MPD that was used in the crystallization solution, and 332 water molecules. The geometry of the refined AVG⅐PLP is that of a ketimine adduct with the AVG C ␣ atom and the C4A atom of PLP exhibiting sp 2 and sp 3 hybridization, respectively.  refined model (PDB code 1IAY) of that complex contains no covalent adduct between the cofactor and the inhibitor. They find that AVG is bound near the entrance of the active site with its ␣-amino group ϳ4 Å from the PLP C4A. In 1IAY, AVG binds to the protein through its ␣-carboxylate group, which accepts two hydrogen bonds, one from the backbone nitrogen of Ala-54 and the other from a guanidinium nitrogen of Arg-412. A third hydrogen bond is donated to the ␣-carboxylate of AVG by a water molecule. This mode of AVG binding, involving a few weak interactions, cannot be reconciled with our crystallographic results. Our refinement strategy was aimed at avoiding model bias; thus, neither cofactor nor AVG was included in the first steps of refinement against the 1.6-Å diffraction data. Simulated annealing at 5000 K ensured additionally that residual model bias from the internal aldimine of 1B8G was removed. A difference Fourier map after one cycle of individual B-factor refinement contoured at 3 exhibited a very clear and strong signal corresponding to the AVG⅐PLP covalent adduct shown in Fig. 1. Clausen et al. (16) and Krupka et al. (17) also find covalent AVG⅐PLP adducts in Escherichia coli cystathionine-␤-lyase (CBL) and in Treponema denticola cystalysin (PDB codes 1CL2 and 1C7O), respectively. Biochemical Evidence for a Ketimine AVG⅐PLP Adduct-Incubation of ACC synthase with AVG results in the rapid disappearance of the PLP internal aldimine absorption band at 421 nm with concomitant appearance of an enzyme species that absorbs at 341 nm (Fig. 2). The extinction coefficient for the latter is ϳ34,600 M Ϫ1 cm Ϫ1 . We propose that the absorption band at 341 nm is consistent with the formation of an AVG⅐PLP ketimine adduct. For comparison, the max and ⑀ values for the aspartate-PLP ketimine species formed by aspartate aminotransferase are 323 nm and 6,770 M Ϫ1 cm Ϫ1 (18). An AVG⅐PLP ketimine species (Scheme 1), however, would contain an addi- tional conjugated bond and heteroatom, thus producing the observed bathochromic shift and hyperchromic effect. Moreover, the structurally characterized AVG-ketimine in the CBL complex also absorbs maximally at 341 nm, and its ⑀ value is 27,500 M Ϫ1 cm Ϫ1 (16).
Rate constants for the appearance of the ketimine absorption band at 341 nm were measured as a function of AVG concentration. A linear plot of the data yields k a ϭ 2.1 (Ϯ 0.2) ϫ 10 5 M Ϫ1 s Ϫ1 (Fig. 3). k a was also determined by measuring the rate of enzyme inactivation in the presence of AVG. Initial rates of enzyme-catalyzed product formation are independent of the concentration of AVG, but the enzyme is inactivated over time in an AVG-dependent manner (Fig. 4). Observed first-order inactivation rate constants (k obs ) were fit to Equation 2 (15), where K m is the experimentally determined Michaelis-Menten constant for SAM under the assay conditions described in the figure legend, and the value of k a obtained by this method is 1.1 ϫ 10 5 M Ϫ1 s Ϫ1 , in reasonable agreement with that determined spectrophotometrically. ACC synthase activity is recoverable from the AVG⅐PLP complex (Fig. 5A). Dialysis of AVGinactivated enzyme against buffer containing 10 M PLP results in the loss of the 341-nm absorption band and complete recovery of the internal aldimine absorption with a rate constant of 2.4 (Ϯ 0.5) ϫ 10 Ϫ6 s Ϫ1 . These spectral changes are accompanied by recovery of enzyme activity with the same rate constant (Fig. 5B). Dialysis of the inactivated enzyme against buffer lacking PLP yields only ϳ35% of the absorbance change at 341 nm observed in the presence of the cofactor, and this decay in absorbance at 341 nm is not accompanied by an increase in absorbance at 421 nm. Furthermore, only 15% of the original enzyme activity is recovered. Therefore, dissociation of AVG is accompanied by PLP loss. The rate constants for disappearance of absorption at 341 nm and recovery of activity are 1.6 (Ϯ 0.3) ϫ 10 Ϫ6 s Ϫ1 and 3.0 (Ϯ 1.4) ϫ 10 Ϫ6 s Ϫ1 , respectively in the absence of PLP in the dialysis medium. Although the disappearance of the ketimine absorption band was not accompanied by the appearance of the aldimine absorption band, recovery of activity was observed because of the presence of PLP (1 M) in the activity assay mixture.

Comparison of Unliganded and AVG-inhibited ACC Synthase-Superposition of the unliganded (PDB code 1B8G) and
AVG-inhibited ACC synthase structures shows that association of AVG with the enzyme does not induce large conformational changes (Fig. 6C). The root mean square deviation between the two structures is 0.7 Å. The only large-scale differences are found in the small domain. 1) A part of the small domain encompassing mainly helix ␣1, strand ␤2, helix ␣13, and strand ␤13 (nomenclature is according to Ref. 3) is slightly displaced toward the large domain of the other subunit. The program DYNDOM identifies this conformational change as a rotation of the aforementioned "subdomain" by ϳ7°around an axis that crosses the small domain and is approximately parallel to the dimer axis. However, whether one considers the small domain as a whole, no overall domain movement can be identified by visual inspection or by DYNDOM (19). 2) The main chain conformation at the beginning of helix ␣1 is rearranged so that Tyr-19 points inside the active site in the AVGbound structure (see following paragraph).
Interactions of the Ketimine Adduct with the Protein-The AVG⅐PLP adduct is tightly bound in the active site of ACC synthase (K d ϭ 10 -20 pM based on the ratio of the k a and k d values described above). Fig. 7 is a LIGPLOT scheme (20) that shows all hydrogen-bonding interactions that contribute to AVG association, and Fig. 6A presents a stereoview of the active site. The PLP moiety of the adduct retains the strong interactions between the phosphate group and the enzyme side chains that are present in the unliganded structure (PDB code 1B8G) (3). The position of the phosphate is only slightly modified with respect to 1B8G, whereas the pyridine ring of the cofactor appears to be tilted by ϳ16°. Thus, the phosphate moiety acts as an anchor. It is noteworthy that the formation of the external aldimine or ketimine not only tilts the cofactor but pushes the ring 1.1 Å down toward the carboxylate of Asp-230 to form a relatively strong hydrogen bond (2.7 Å) in the complex. These crystallographic results confirm the theoretical model of a reaction intermediate of ACC synthase (external aldimine with ACC) described previously (3), which predicted that such a hydrogen bond would be established upon external aldimine formation.
The carboxylate group of the AVG⅐PLP adduct interacts  and open symbols, respectively). A, the spectra of the enzyme were recorded over time, and the absorbancies of the PLP-aldimine (squares, 421 nm; active enzyme) and AVG-ketimine (circles, 341 nm; inactive enzyme) species are plotted. The data were fit to first-order kinetics yielding rate constants of 2.4 (Ϯ 0.5) ϫ 10 Ϫ6 s Ϫ1 in the presence of PLP (solid lines; average of both ketimine and aldimine absorption curves), and 1.6 (Ϯ 0.3) ϫ 10 Ϫ6 s Ϫ1 in the absence of PLP (dashed line; ketimine absorption only). B, the activity of the enzyme with SAM was measured (14) over time, and is plotted as a fraction of the activity of the enzyme before inactivation with AVG. The data were fit to firstorder kinetics yielding rate constants of 2.5 (Ϯ 0.6) ϫ 10 Ϫ6 s Ϫ1 and 3.0 (Ϯ 1.4) ϫ 10 Ϫ6 s Ϫ1 in the presence and absence of PLP, respectively. strongly with the protein. It forms a doubly hydrogen-bonded salt bridge with Arg-407 and accepts a hydrogen bond from the carboxamide nitrogen of Asn-202. Notably, the AVG⅐PLP ␣-carboxylate receives a fourth hydrogen bond from the OH of Tyr-19. The side chain of Tyr-19 points toward the solvent in the unliganded ACC synthase structure, whereas in the ACC synthase-AVG complex there is a rearrangement involving residues 18 -20 that results in the phenolic ring of Tyr-19, reaching the inner active site to interact with the AVG⅐PLP ␣-carboxylate group.
The ⑀-amino group of the AVG side chain in the complex makes a triple electrostatic interaction with Glu-47 and Lys-273. The side chain of Glu-47 is found in a different orientation with respect to the unliganded enzyme, and its ␥-carboxylate is the pivot for the formation of this interaction. The side chain amino groups of both AVG and Lys-273 donate hydrogen bonds to the Glu-47 ␥-carboxylate, which also receives another hydrogen bond from the carboxamide nitrogen of Gln-83 of the other subunit (Gln-83*). A comparison of the AVG-bound ACC synthase and CBL active sites (Fig. 6B) (16) provides a structural   FIG. 6. A, superposition of the active sites of unliganded (sea green) and AVG-bound (yellow) ACC synthase. Only selected residues are shown. B, the active sites of AVG-bound ACC synthase (gray) and cystathionine ␤-lyase (sea green). C, the C␣ traces of unliganded (yellow) and AVG-bound (blue) ACC synthase. This illustration was prepared with DINO. basis for the different AVG affinities of the two enzymes (K i ϭ 1.1 M for CBL; K d ϭ 10 -20 pM for ACC synthase). The two AVG⅐PLP adducts superimpose very well; however, the side chain of AVG⅐PLP interacts with CBL much more weakly than with ACC synthase because the AVG side chain in the CBL⅐AVG complex points toward the active site entrance and its amino group is held in place mainly by weaker interactions with the hydroxyl group of Tyr-111 and with two water molecules.
A molecule of the precipitating agent MPD is found at the entrance of the active site of AVG-bound ACC synthase. Interestingly, the MPD association is stereoselective. Racemic MPD was used for crystallization, but only the R-enantiomer is found in the present structure. The two OH groups of R-MPD are the points of contact. One (O4 in the crystallographic atom naming) accepts a hydrogen bond from the guanidinium group of Arg-150, whereas the second (O2) is involved as donor in a bifurcated hydrogen bond with the carboxamide oxygen atom of Gln-83* and the ether atom of AVG.
Kinetics Describing the Association of AVG with E47Q ACC Synthase-The addition of AVG to E47Q ACC synthase results   a Reported k a values were calculated from stopped-flow spectrophotometric measurements of ketimine formation rates. k d values were obtained by monitoring recovery of the internal aldimine absorbance band following removal of AVG from reaction mixtures. CBL data are from Clausen et al. (16). in a rapid loss of the PLP internal aldimine absorption band at 421 nm, and concomitant appearance of an absorption band at 341 nm with an extinction coefficient of ϳ35,200 M Ϫ1 cm Ϫ1 (data not shown). The rate constants for appearance of the absorption band at 341 nm were measured as a function of AVG concentration, yielding a second-order rate constant of 7.2 (Ϯ 0.7) ϫ 10 4 M Ϫ1 s Ϫ1 for that reaction (see Fig. 3).
The association of AVG with E47Q ACC synthase is reversible (Fig. 8). Dialysis of AVG-bound E47Q ACC synthase against buffer containing 1 M PLP results in a loss of the ketimine absorption band at 341 nm and recovery of the internal aldimine absorption band at 421 nm with a rate constant of 8.2 (Ϯ 1.7) ϫ 10 Ϫ5 s Ϫ1 . The rate at which catalytic activity is recovered could not be determined because of the extremely low activity of the E47Q mutant for ␣,␥-elimination on SAM (5).

DISCUSSION
The AVG⅐ACC Synthase Complex Is a Ketimine-We present crystallographic and biochemical evidence that AVG inhibits ACC synthase from M. domestica by forming a covalent ketimine complex with the PLP cofactor. The high resolution crystallographic evidence supported by spectroscopic and kinetic analyses cannot be reconciled with the findings of Huai et al. (4) for the complex of Lycopersicum esculentum ACC synthase with the same inhibitor (PDB code 1IAY). That report claims that a noncovalent complex is formed between AVG and the holoenzyme and that it is stabilized only by two hydrogen bonds. An examination of electron density for 1IAY as calculated by the Uppsala University Electron Density Server (portray.bmc.uu.se/eds/index.html) shows that the AVG ligand is associated with the density only in its carboxylate moiety (assuming that the structure factor file deposited for 1IAY and used by the server corresponds to the ACC synthase-AVG complex). Our estimated K d for apple ACC synthase⅐AVG is 10 -20 pM. A weak non-covalent enzyme-inhibitor interaction as proposed by Huai et al. (4) would suggest a much higher K d value. Finally, the 341-nm absorption maximum is incompatible with a conjugated aldimine. The identity level between the M. domestica and the L. esculentum (LE-ACS 2) sequences is very high (56% with 67% similarity), and all of the residues that could be involved in binding are conserved; thus, the very different structural results with AVG cannot be ascribed to differences between the tomato and the apple enzyme. In our structure, the very strong affinity of AVG toward ACC synthase nicely correlates with the formation of a ketimine adduct held in place by a dense network of interactions with the protein. This network involves both the side chain and ␣-carboxylate of AVG. Notably, there is a charge interaction between the side chain nitrogen of AVG and Glu-47 of ACC synthase. That residue has been shown to be functionally critical for the reaction mechanism (5).
The Active Form of ACC Synthase Is a Dimer-Furthermore, the contention of Huai et al. (4) that monomeric ACC synthase may be active is incompatible with hybrid reconstitution results on apple ACC synthase (21), which demonstrated unambiguously that the dimer or theoretically a higher oligomer is the functional form of the enzyme. The mutated residues (Lys-273 and Tyr-85) employed in the hybridization experiments are conserved throughout plant ACC synthase sequences. Finally, it should be emphasized that none of the characterized PLPdependent enzymes of this class functions as a monomer.
Comparison of ACC Synthase and CBL AVG-Ketimine Adducts-A covalent AVG-ketimine adduct is also present in two other PLP-dependent enzyme structures, CBL (16) and cystalysin (17). The kinetics and equilibria of the reaction of AVG with CBL and ACC synthase are compared in Table II. Each enzyme reacts to form an AVG⅐PLP ketimine species that has a high affinity for the enzyme active site. Although the mechanisms for inhibition of both CBL and ACC synthase appear to be similar, the binding affinities of the two enzymes for AVG differ significantly. The K d for the CBL complex is in the range of 0.3-1 M, 2 whereas that for ACC synthase is 10 -20 pM. As shown in Table II, the lower affinity of CBL for AVG is expressed in both the association and dissociation rate constants. The value of k a for ACC synthase is 1,400 times greater than that for CBL, whereas k d for CBL is ϳ15 times greater than that for ACC synthase. The much lower affinity of CBL for AVG can be explained by the weaker binding modes of the AVG side chain and ␣-carboxylate group to CBL. The structure of the CBL⅐AVG complex shows hydrogen bonds among the AVG ⑀amino group, two water molecules, and Tyr-111, which probably make much weaker contacts than those present in the triple electrostatic interaction between the AVG ⑀-amino group and Glu-47 and Lys-273 in ACC synthase. Furthermore, the interaction between the ␣-carboxylate group of AVG and ACC synthase is strengthened by the presence of a hydrogen bond to Tyr-19, which is not present in CBL.
The Contribution of Glu-47 to AVG-Ketimine Complex Stability-k a and k d values for E47Q ACC synthase were determined to elucidate the value of the contribution of the carboxylate of Glu-47 to the stability of the complex (see Table II). The replacement of the Glu-47 carboxylate ion with an amide group results in a 100-fold decrease in affinity. k a is three times lower for the E47Q mutant than for WT, and k d is 34 times greater for the E47Q⅐ACC synthase complex than it is for the corresponding WT ACC synthase complex. Although this decrease in affinity is significant, the K d for the E47Q⅐ACC synthase complex remains 300-fold less than that of the WT CBL complex which also lacks the Glu-47-AVG interaction, thus suggesting that other protein-inhibitor interactions such as the hydrogen bond between Tyr-19 and the ␣-carboxylate group are also important for the tight binding of AVG to ACC synthase. The fact that the E47Q mutation elicits such a large reduction (ϳ3 kcal/mol) in affinity of the enzyme for AVG supports the earlier conclusion that the carboxylate group of Glu-47 plays an important role in the recognition of the sulfonium group of the natural substrate SAM as suggested by mutational, kinetic, and modeling studies (5).