Analysis of the Decarboxylation Step in Mammalian Histidine Decarboxylase

We report a hybrid quantum mechanics/molecular mechanics theoretical study on the reaction mechanism of mammalian histidine decarboxylase that allows us to obtain valuable insights on the structure of the cofactor-substrate adduct (external aldimine) in the active site of rat histidine decarboxylase. By means of molecular dynamics simulations, we traced the potential of mean force corresponding to the decarboxylation reaction of the adduct both in the active site of the enzyme and in aqueous solution. By comparing this process in both media, we have identified the key electrostatic interactions that explain the lowering of the free energy barrier in the enzyme. Our analysis also offers a validation of Dunathan's hypothesis (Dunathan, H. (1966) Proc. Natl. Acad. Sci. U. S. A. 55, 712–716) regarding the role of stereoelectronic effects in the enzymatic decarboxylation process.

We report a hybrid quantum mechanics/molecular mechanics theoretical study on the reaction mechanism of mammalian histidine decarboxylase that allows us to obtain valuable insights on the structure of the cofactor-substrate adduct (external aldimine) in the active site of rat histidine decarboxylase. By means of molecular dynamics simulations, we traced the potential of mean force corresponding to the decarboxylation reaction of the adduct both in the active site of the enzyme and in aqueous solution. By comparing this process in both media, we have identified the key electrostatic interactions that explain the lowering of the free energy barrier in the enzyme. Our analysis also offers a validation of Dunathan's hypothesis ( Computational approaches to investigation of enzymatic catalysis rely on the ability to obtain a proper initial structure for the system under study. Three-dimensional structures are best determined by experimental methods such as x-ray diffraction, nuclear magnetic resonance (NMR) spectroscopy, or even electron microscopy (1). However, experimental methods cannot always be applied. Many proteins are too large for NMR analysis and cannot be expressed or purified in an amount substantial enough to be crystallized for x-ray diffraction; consequently, there are no corresponding three-dimensional structures for the majority of protein sequences. In such cases prediction of the protein structure by computational methods can frequently result in a useful model. However, the application of a comparative model depends on its accuracy. A strong predictor is the sequence identity between the target and the known protein structure used as a template; there is a broad application spectrum based on this identity. The high end of the accuracy spectrum corresponds to models based on 50% iden-tity or more; the average accuracy of these models is about 1 Å root mean square for the main-chain atoms, which is comparable with that of low resolution x-ray structures (2).
Mammalian histidine decarboxylase (HDC 4 ; EC 4.1.1.22) is one of the enzymes for which there is no experimentally solved structure due to its instability in vivo, in crude extracts, and in purified solutions (3,4). This is a pyridoxal 5Ј-phosphate (PLP)dependent enzyme that synthesizes histamine by decarboxylation of histidine. The importance of this enzyme is obvious considering that the biogenic amine histamine (2-(4-imidazolyl)ethylamine) is a multifunctional intercellular mediator involved in many different physiological responses in animals (hematopoiesis, gastric secretion, neurotransmission) (5)(6)(7)(8). It is also related to many pathological processes (gastric ulcer, allergy and many other inflammatory diseases, bone lost, tumor progression, learning deficiency, and epilepsy, etc.) (9 -13). Thus, the control of its synthesis should be interesting for therapy against these pathologies that affect a high population rate at any stage of their lives. Nevertheless, nowadays, antihistamine pharmacology is based on agonists and antagonists of the histamine reception in target cells but not on the development of modulators of its synthesis. This is mainly due to the lack of information about the proper targets for intervention at the level of its synthesis in the different histamine-producing cell types (3).
Gram-positive bacteria express a non-homologous pyruvoyldependent HDC, but Enterobacteria express a PLP-dependent HDC that is evolutionary related to the mammalian enzyme (14). Besides, several mammalian cell types (mainly neurons) also express a highly homologous PLP-dependent enzyme that is responsible for the synthesis of several neurotransmitters, named DOPA decarboxylase or L-aromatic amino acid decarboxylase (DDC/AADC) (15)(16)(17). According to these facts, it has been difficult to develop specific modulators of mammalian histidine decarboxylase activity. In actuality, synthetic substrate analogs such as fluoromethylhistidine and histidine methyl ester can inhibit both the Gram-negative and the mammalian HDC (18); other inhibitors, such as epigalocathequine-3-gallate, have been described as inhibitors of both mammalian HDC and DDC/AADC (19). As a consequence, a deeper structural characterization of the mammalian enzyme is needed to locate the best targets for selective intervention.
The first three-dimensional model of a mammalian enzyme (rat HDC (rHDC)) was obtained in 2003 based on a pig DOPA decarboxylase (pDDC) structure as the template (sequence identity of 52%) (18). This structure was then validated with experimental results that were obtained from more than 25 direct mutants (20,21) so that the functional data on the enzyme fit well with the proposed model of the enzyme (3). The agreement between our proposed rHDC structure and the experimental data available has led us to perform a detailed computational study aimed to unravel special topics of the catalytic mechanism, namely the location of the cofactor-substrate adduct in the active site as well as its correct conformation for decarboxylation. Such an effort offers additional and definitive validation of our model as a reliable representation of the structure of mammalian histidine decarboxylase and contributes to extending the scope of computational focus on enzyme catalysis.
Catalysis by PLP-dependent enzymes has been extensively reviewed (23,24). The PLP-dependent decarboxylase catalytic cycle begins with the cofactor covalently bound to the ⑀-amino group of a specific and conserved Lys residue (Lys-308 in rHDC) in a Schiff base known as internal aldimine. After substrate binding, the external aldimine intermediate (PLP-His) is formed via a reversible transaldimination process; then, decarboxylation occurs, histamine is released from the catalytic site, and the enzyme recovers the internal aldimine state (Fig. 1). Any of these processes may be the rate-limiting step of the entire reaction rate. In fact, k cat of the mammalian enzyme is extremely low (lower than 0.1 s Ϫ1 ), in contrast to those higher values (2-3 orders of magnitude) reported for both homolo-gous enzymes, mammalian DDC/ AADCs, and HDC from Enterobacteria (25,26).
As revealed by structure/function relationship studies carried out in our and other laboratories, both mammalian HDCs and DDC/ AADCs are described as homodimeric enzymes (18 -21). Their active sites must be similar and involve residues of both monomers. PLP binding residues are mainly located in one of them, but both monomers contribute to the substrate binding site (16,18,21). By means of different biophysical techniques (circular dichroism, electrophoresis under semidenaturing conditions, and gel filtration chromatography), it was detected that local conformational changes occur after histidine binding and then are transmitted to the global conformation of the mammalian histidine decarboxylase (18). These facts have not been observed in other homologous enzymes, at least in such extent (21, 26 -28). Transaldimination leads to formation of the external aldimine, whereas the conserved lysine residue is released. This external aldimine can exist in different prototropic forms, although the predominant form seems to be that which presents the pyridine and the amide nitrogen atoms as protonated, whereas the 3Ј-hydroxyl group is deprotonated (29) (see Fig. 2).
There are two possible ways of histidine binding to PLP that lead to two different conformations for the external aldimine. We have named them as conformation 1, for the conformation in which the side chain of the substrate points toward the re face of the imino sp2 center (i.e. opposite to the PLP-binding lysine), and conformation 2, for that in which the substrate points toward the si face of the cofactor. Kern et al. (30) asserted that the PLP-His position in the group II amino acid decarboxylase folding family corresponds to the conformation 1. When the structure of pDDC was solved, a conformation similar to conformation 2 was observed for the PLP-carbiDOPA adduct in the catalytic site (16).
In this work we carry out a theoretical analysis of external aldimine decarboxylation in the active site of rHDC. The structure of the enzyme, obtained from homology models, is used to explore the catalytic properties of the enzyme by means of quantum mechanics/molecular mechanics (QM/MM) simulations. The first part of the study was devoted to investigation of an appropriate model of the external aldimine in the active site. We investigated different possible structures for this external aldimine, concluding that conformation 1 seems to be more appropriate for the external aldimine considered in this case. In the second part, we simulated the decarboxylation reaction in the active site and compared it to the uncatalyzed process in water.

EXPERIMENTAL PROCEDURES
A homology model for the structure of rat histidine decarboxylase reported by the authors (18) and validated by directed mutagenesis (3,21) was used to manually fit the Schiff base formed by the histidine substrate and the cofactor PLP into the active cleft in two different ways; this was done to obtain the two starting conformations of the external aldimine.
Calculations were performed by means of a QM/MM computational scheme (31)(32)(33) using the CHARMM 27 package (34). The external aldimine was chosen as the QM subsystem (described using the AM1 hamiltonian) (35), whereas the MM subsystem was composed of both protein residues and water molecules. To achieve a reasonable agreement between performance and calculation time for this large protein (15,054 atoms), a simulation protocol was followed (36,37). First, the coordinates of the hydrogen atoms of the protein were determined using the HBUILD program of the CHARMM package, where all the ionizable groups were set to their normal ionization state at pH 7. Then, the protein was relaxed using a potential restraining scheme with a low force constant (25 kcal⅐mol Ϫ1 Å Ϫ2 ) applied only on the peptidic backbone. Finally, the optimized protein was solvated with a 24-Å radius sphere of TIP3P water molecules using atom C4 of the substrate (see Fig.  2 for atom numbering) as the geometrical center of the system. Water molecules that were within a distance of 2.5 Å of any non-hydrogen atom were removed. The resulting system was resolvated four more times using different relative orientations between the protein and the water sphere to ensure good solvation of the system. Then, water positions were optimized using the Adopted Basis Newton-Raphson (ABNR) algorithm followed by optimization of that part of the system included in the sphere of 24 Å around C4. Finally, a molecular dynamics simulation of the water molecules (5 ps) was carried out to relax energetically unfavorable contacts, and the 3-fold cycle of superposition, deletion, and rotation was then repeated to fill in additional cavities generated during the dynamic simulation. Thus, the final model has 16,352 atoms; 15,054 of them are atoms from the protein, 41 from the cofactor-substrate adduct, and 1,257 atoms are from the water molecules. The residues lying outside of this 24 Å sphere were kept fixed for subsequent calculations, whereas the residues inside the 20-Å sphere from C4 were totally free. A buffer zone was described for the residues and water molecules in the zone between 20 and 24 Å. The movement of the atoms in this buffer region was restrained following a protocol detailed in Poulsen et al. (36). In addition, a switched cut-off radius from 12 to 13.5 Å was used for all interactions. The final structure was first heated to 310 K for 30 ps and then equilibrated using hybrid QM/MM Langevin-Verlet molecular dynamics (NVT) over 500 ps. We also studied the decarboxylation step in solution using the same QM/MM methodology. In this case the aldimine (QM subsystem) was placed in a cavity deleted from a 55.8-Å cubic box of water molecules, described using the classical TIP3P potential (a total of 5804 water molecules). For all calculations performed in water, periodic boundary conditions were applied, making use of the same switched cut-off radius. The conformational changes and decarboxylation reactions were studied through computation of the corresponding potentials of mean force (PMFs) using the canonical ensemble (NVT). A PMF was traced using the dihedral angle (see Fig. 2) as the distinguished coordinate to investigate the relative orientation of the carboxylate group and the pyridine ring. A total of 12 simulation windows were run to cover coordinate values from Ϫ80 to Ϫ180 degrees for this transition. Each window was started from the final conformation of the previous window and consisted of 10 ps of equilibration and 80 ps of production using a time step of 1 fs. Changes in the selected reaction coordinate were restrained using an umbrella potential with a force constant of 1.2 kcal⅐mol Ϫ1 degree Ϫ2 and an additional biasing potential (39). A potential of mean force was also computed using the distance between CA and C (see Fig. 2) as the reaction coordinate. In this case we ran 17 windows to cover a reaction coordinate range from 1.4 to 3.0 Å. Each consecutive window consisted of 10 ps of production at 310 K, with a time step of 1 fs and a force constant of 500 kcal⅐mol Ϫ1 Å Ϫ2 . In both cases, the reference temperature was 310 K. The different values of the variable sampled during the simulations were then pieced together by means of the weighted histogram analysis method (38) to construct the distribution functions from which the PMFs were derived.
Finally, the same chemical step was studied in water solution. The reaction coordinate, defined also as the distance between the CA and C atoms, was sampled along a total of 65 windows. Langevin-Verlet NVT molecular dynamics were applied to each window, running for a total of 10 ps for relaxation followed by 15 ps for production. The umbrella constant applied on each particular value of the reaction coordinate was 600 kcal⅐mol Ϫ1 ⅐Å Ϫ2 , and the temperature of the bath was 310 K. As before, weighted histogram analysis method was used to render the PMF.

RESULTS AND DISCUSSION
External Aldimine Conformation-The external aldimine is formed in the pDDC active site from PLP and the inhibitor carbiDOPA, the substrate analog pointed toward the si face of the cofactor in a conformation similar to conformation 2 in Fig.   FIGURE 2. Representation of the external aldimine with atom labeling. is the C-CA-N-C4A dihedral angle that describes the relative orientation between substrate and cofactor. 3 (16). An interesting discussion has been reported on the suitability of binding the substrate to PLP toward the re or si face (40). Matsuda et al. (40) suggest that because the inhibitor car-biDOPA has an ␣-hydrazino group with an extra nitrogen atom (compared with the ␣-amino group of DOPA and His), the PLP-carbiDOPA complex is expected to adopt a significantly different conformation from that of PLP-DOPA and, consequently, the PLP-His in the HDC active site. This issue has important consequences for the relative orientation of the substrate with respect to the enzyme. In the conformation 2, the side chain of the substrate points toward the interior of the protein, whereas in the conformation 1 it points outwards the enzyme (see Fig. 3, panels C and D).

Analysis of the Decarboxylation Step in Mammalian HDC
To solve this question, we built both conformations for the external aldimine complex. The two conformations of the external aldimine were fully minimized in the active site using the QM/MM hybrid potential described in the previous section. After geometry optimization, we obtained the conformations presented in Fig. 3.
It is important to note that conformation 2 presents a tooshort distance between two negatively charged groups of the molecule; that is, the phosphate and the carboxylate groups. This high energy conformation of the external aldimine is then very unlikely to be appropriate for describing the product of the transaldimination. To confirm this point, the full system (cofactor-substrate complex, protein, and water molecules) was heated to 310 K and then 450 ps of NVT/MD simulation were run for each of the two conformations. Fig. 4 displays the time evolution of some key distances for both cases; that is, the distance between the carbon atom of the carboxylate group and the phosphorous atom of the phosphate group (C-P), the distance between the imino nitrogen atom and the negatively charged O3 atom (N-O3), and the CA-C bond length. As can be observed, the conformation 2 evolves through spontaneous decarboxylation after 100 ps. The only way found in this conformation for diminishing the electrostatic repulsion between the phosphate and carboxylate groups is to lengthen the CA-C distance, leading to decarboxylation of the transaldimination product. On the other hand, the conformation 1 remains stable during the simulation. Another noticeable difference between the two conformations is that the first form retains a strong and biologically relevant intramolecular hydrogen bond between the imino nitrogen and the O3 atom (23) (the O3-N distance fluctuates around 2.2 Å), whereas in the other form, the N atom forms a hydrogen bond with a water molecule. We can, thus, consider that the conformation 2 does not properly represent the external aldimine, whereas the first conformation may be considered as a realistic model in agreement with the proposal of Matsuda et al. (40); therefore, subsequent studies were performed on this conformation.
Active Site Description-Once the conformation of the external aldimine in the active site seemed to be well established, we could concentrate on the most important interactions observed with the active site residues. We monitored distances between heteroatoms of PLP-His and those of the closest-neighbor amino acids to localize important hydrogen bonds. Fig. 5 shows the position of some important residues of the active site interacting with the external aldimine in the conformation 1.
Residues located less than 5 Å from the PLP moiety are (A and B are used to distinguish both monomers) V100A, T152A, V153A, S154A, H197A, S199A, K202A, T247A, T250A, T251A, D276A, K308A, F314A, P355B, and L356B; those closer to the substrate moiety are Y83A, Y84A, I439A, F331B, M350B, H351B, and I354B. There is experimental evidence for an essen- For clarity, panels A and B only show the external aldimine conformations with respect to two essential residues of the catalytic center (Lys-308 and His-197), and the rest of the protein has been omitted (in these upper views, the entrance of the catalytic site would be located on the right side of the pictures). In panels C and D, the surface of both monomers have also been represented, and the views have been clock-wise rotated by 90°approximately on the longitudinal axes with respect to the upper panels, so that the catalytic hole entrance is now in front of the viewer. tial role in mammalian HDC activity for those residues highlighted in italics (21).
The active site is located in the monomer-monomer interface but is mainly composed of residues from one monomer; as in pDDC (16), residues of the opposite monomer participating in the active site are mainly hydrophobic and noteworthy, and there are a number of water molecules participating in direct interactions with both the cofactor and substrate. There are several nonpolar residues that are meant to increase the hydrophobicity of the active site to facilitate the introduction of an aromatic substrate such as histidine. This is also a common feature of aromatic amino acid decarboxylases (23). It is also noteworthy that there are a number of water molecules (only one is shown in the figure) participating in direct interactions with both the cofactor and the substrate.
Lys-308 is the residue that binds PLP covalently to form the internal aldimine. This residue is shifted during transaldimination and is responsible for product release (23). Interestingly, in the external aldimine state, this lysine together with Lys-202 interacts with the phosphate group of PLP, providing extra binding energy to the adduct.
Lys-202 can establish this interaction after a conformational change of its side chain. The charge-charge interactions established between these two lysine residues and the phosphate group of PLP play a very important role in providing a favorable environment for the external aldimine. The averaged distances from the N ⑀ atoms of lysine residues to the closest oxygen atoms of the phosphate groups are 2.73 and 3.27 Å for Lys-308 and Lys-202, respectively.
His-197, another highly conserved residue, is stacked parallel to the cofactor pyridine ring. Its analogue position in pDDC (His-192) is predicted to form a hydrogen bond with the carboxylate group of carbiDOPA (16). The authors assume that the natural substrate, L-DOPA, should be located even closer to the imidazole ring of His-192, so they suggested a direct catalytic role for this residue. In our model, taking into account the different location of the histidine substrate in the conformation 1, His-197 cannot interact with the carboxylate group of the substrate, so an active catalytic role is not expected for this residue. This finding is supported by the previous observation that a H197G direct mutant still retains 10% of the wild-type HDC activity (21).
Asp-276 and its counterpart in pDDC Asp-271 were predicted to contact the pyridine ring of PLP (16,18). In the case of  the pDDC crystal structure, Asp-271 can form a hydrogen bond with the protonated nitrogen of the pyridine ring (N1). This direct interaction is not observed in rHDC, although Asp-276 can still stabilize the protonated state of the pyridine ring by a water-mediated hydrogen bonding interaction (see Fig. 5). The external aldimine is free to move in the active site cavity, and before the molecular dynamics simulations, it is allowed to relax and adopt a suitable position in the active site. The permanence of this water molecule, acting as a bridge between Asp-276 and the external aldimine, over the equilibration and production molecular dynamics trajectory, indicates that this water molecule has a structural (and functional) role. Moreover, a translation of the substrate cofactor adduct deeply in the active site cavity toward Asp-276 would mean the loss of important stabilizing interactions such as those established by Lys-308 and Lys-202 with the phosphate group of PLP and clashes between the side chain of the substrate and other active site residues.
According to a previously reported model of the structure of rat (rAADC) and the subsequent directed mutagenesis analysis (40), Asp-271 corresponding in rAADC to Asp-276 in rHDC is also essential for rAADC activity. Interestingly, a D271E mutant conserves 2% of the wild-type activity, whereas noting that lengthening the distance between the PLP and the negatively charged group leads to an important reduction of the enzymatic activity. The assumed role for this interaction is to enhance the electron delocalization along the conjugated system, which is commonly accepted to stabilize PLP-dependent enzyme intermediates (32,41). The distance between the carboxylate group of Asp-271 and the N1 atom in the pDDC crystal structure is 2.50 Å, much lower than the corresponding distance between Asp-276 and the N1 in our model (4.38 Å), where a water molecule is placed in between the carboxylate group of aspartate and PLP. As said, this larger separation could partly explain the lower catalytic efficiency of rHDC when compared with pDDC and rAADC.
The equivalent residue to Asn-305, Asn-300 in pDDC, was predicted to play a relevant role in catalysis due to its close interaction to the phosphate group (16). This residue has an identical counterpart in other mammalian group II decarboxylases, such as glutamic acid decarboxylase, in addition to pDDC (21). In our system Asn-305 is not predicted to have such importance in the external aldimine state. This residue is found 5 Å away from the oxygens of the phosphate group of PLP. As far as we know there is no experimental support for a role of this residue either in pDDC or in rHDC.
Dunathan's Hypothesis-Another important question related to the structure of the external aldimine in the rHDC active site is the orientation of the bond to be broken during decarboxylation (CA-C bond) with respect to the conjugated system. Dunathan (42) suggested that PLP-dependent enzymes can favor the reaction by means of stereoelectronic effects, orienting the nascent p orbital parallel to the p orbitals of the electronic system. Crystal structures of several PLP-dependent enzymes (43,44) support this observation. In the gas phase there are three main contributions which determine the relative positioning of the CA-C bond relative to the plane of the conjugated system; they are the intramolecular hydrogen bond between the carboxylate and the protonated imino groups (which should favor a dihedral angle of 180°), the electrostatic repulsion between the carboxylate group and the O3 atom (favoring a dihedral angle of 0°), and the stereoelectronic effect (which stabilizes a dihedral angle of Ϯ90°). Using a reduced model for the external aldimine (where the histidine moiety is substituted by a hydrogen atom), we determined that the gas phase AM1 energy minima appeared at the (C-CA-N-C4A) dihedral angle of approximately Ϯ50°. It seems that the value of this dihedral angle in the gas phase minimum is dominated by electrostatic repulsion and stereoelectronic effects, although the minimum energy orientation is far from the preferred value for decarboxylation (Ϯ90°).
To investigate Dunathan's hypothesis, we traced the free energy profile associated with the dihedral angle in the enzyme. This PMF, presented in Fig. 6, shows two free energy minima placed at approximately Ϫ105 and Ϫ140°, with the latter being the more stable by about 2.2 kcal/mol. This result means that the enzymatic environment induces considerable changes in the conformation of the external aldimine. Intermolecular interactions in the protein residues and water molecules very efficiently shield the electrostatic repulsion between the negatively charged carboxylate group and the O3 atom. In this way, both the formation of an intramolecular hydrogen bond between the carboxylate and the protonated imino groups and the stereoelectronic effect dominate the energy profile associated with the dihedral angle in the enzyme. In the free energy minimum appearing at Ϫ140°, the position of the carboxylate group is favored by means of intramolecular hydrogen bonds with the protonated imino nitrogen  nearly perpendicular to the conjugated system, and thus, according to Dunathan's hypothesis, the decarboxylation process should be favored in this case. Although this is not the most stable conformation, the free energy difference with respect to the absolute minimum is not too high, and a significant proportion of external aldimine complexes should present a value of the dihedral angle optimum for the decarboxylation. Nonetheless, we decided to investigate the reaction paths that began with both conformers (see below).
Decarboxylation Reaction and Origin of the Catalytic Effect-Experimental data on the reaction mechanism and its kinetics have been obtained on a fully active recombinant version of the rat enzyme by using different spectroscopic approaches (UVvisible, fluorescence, and circular dichroism). A k cat of 0.077 s Ϫ1 was determined at 310 K (25) from which we can estimate an activation free energy of 19.7 kcal/mol by direct application of the transition state theory (assuming a transmission coefficient equal to unity). These studies pointed out that the rate-limiting step in the catalytic cycle is one of the steps occurring after transaldimination (26). Decarboxylation is, therefore, an excellent candidate for the rate-limiting step of the global catalytic process. Simulations of this step also offer an excellent chance to validate Dunathan's hypothesis, an experimentally unaffordable task. To this end we have traced the free energy profile associated with this process in terms of a PMF obtained as a function of the CA-C distance. We considered the reaction from the two different conformations obtained for the external aldimine. This is (with two different orientations of the leaving carboxylate group with respect to the plane of the conjugated system) the absolute energy minimum complex (with a dihedral angle of Ϫ140°) and the Dunathan's complex (with a dihedral angle value of Ϫ105°). The results are shown in Fig. 7. The calculated barriers for the decarboxylation processes are 20.4 kcal/mol for the Dunathan's complex and 24.7 kcal/mol for the other. Thus, the relative reactivities of these two conformers are in agreement with Dunathan's hypothesis that the con-former with a perpendicular orientation of the CA-C bond relative to the conjugated system is the most reactive.
However, it must be taken into account that to estimate the total activation free energy associated with the decarboxylation process through the Dunathan's conformer, one must add the free energy difference with respect to the absolute minimum energy complex (2.2 kcal/mol, see Fig. 6), thus resulting in a total free energy barrier of 22.6 kcal/mol. This result compares satisfactorily with the experimental estimation (19.7 kcal/mol). Although we are using a low-level semiempirical Hamiltonian to describe the process, it has been previously shown that the AM1 Hamiltonian gives energy barriers in very good agreement with those obtained using much higher theoretical treatments (such as post-Hartree-Fock or density functional based methods) for this particular type of reaction (45). Thus, the agreement of the reported barrier with the experimental activation free energy gives us confidence in our model for the rHDC active site.
The presence of two external aldimine conformations could justify a mechanism selected by evolution to restrict the efficiency of the mammalian enzyme to maintain homeostasis of the tissues. Intracellular concentrations of the substrate can be Ͼ10-fold that of the HDC K m value (1-2 mM compared with 0.2-0.4 mM) (26, 46), which means that the enzyme could work at its maximum activity. The product, histamine, can elicit its physiological roles at very low concentrations (in the nano-and micromolar order) in target cells (47). To ensure a minimum of histamine efflux from the major histamine-producing cells, the amine needs to be stored in secretion granules until the proper external stimulus (pro-inflammatory, neuroendocrine, etc.) is received (47,48). Because there is no need of a high enzymatic activity, mammalian HDC is under a low evolutionary pressure to achieve optimal catalysis. The cell can coordinate both histamine synthesis and storage, minimizing histamine efflux to biological fluids, which could have lethal effects on the whole organism.
To quantify the catalytic power of the rHDC enzyme in the decarboxylation process, we also traced the PMF corresponding to the decarboxylation of the external aldimine in aqueous solution. In this case we verified that in the initial conformation of the substrate, the CA-C bond was perpendicular to the conjugated system, similar to the Dunathan's conformer in the enzyme active site, and that this conformation remained stable during simulation. As shown in Fig. 7, the free energy barrier obtained was 31.2 kcal/mol. This result is in good agreement with experimental and theoretical determinations of the activation free energy barriers for the decarboxylation of similar compounds (44,45). It must be stressed that the uncatalyzed reaction presents a free energy barrier 10.8 kcal/mol above the best of the enzymatic pathways, which translates to a rate acceleration of 4.38 ϫ 10 7 at 310 K caused by the enzyme.
There is growing consensus that enzymes catalyze chemical reactions by selective transition state stabilization by means of specific electrostatic interactions (46,47). In the reaction we are considering the carboxylate group, which formally supports a charge of Ϫ1, leaves the substrate as a neutral entity, with an electron being transferred to the rest of the molecule. Thus, there is a great charge reorganization when passing from the  MAY 2, 2008 • VOLUME 283 • NUMBER 18 reactant state to the transition state that should be favored in the enzymatic environment with respect to the aqueous solution. To analyze this point we have computed the averaged atomic Mulliken charges for the reactant and transition states in aqueous solution and in the enzyme (in this last case for the most reactive pathway taking place through the Dunathan's conformer). Although these charges cannot be taken as a quantitative measure of the electronic distribution, they are very informative when employed for comparative purposes. For the sake of simplicity, the charges have been gathered to represent different groups of the external aldimine and are represented in Fig. 8.

Analysis of the Decarboxylation Step in Mammalian HDC
According to the results presented in Fig. 8, we can observe that the PLP acts as an electron sink, which is the role usually assumed for this cofactor (23,34). It must be noted that this effect is substantially larger in aqueous solution than in the enzyme. Effectively, the electron charge on the pyridine ring becomes more negative when going from the reactant to the transition state, by Ϫ0.68 atomic units in solution and Ϫ0.21 atomic units in the enzyme. Although the largest portion of the charge transferred from the carboxylate group goes to the pyridine ring in solution, in the enzyme active site the electron density increases noticeably on the CA atom (B group in Fig. 8), and the charge of this atom becomes more negative by Ϫ0.35 atomic units.
After decarboxylation occurs, a quinoid structure seems to be favored in solution, resulting from the addition of an electron to the pyridine ring, whereas the enzyme seems to preferentially stabilize a form that localizes the negative charge on the B group. According to Fig. 8, the electron flow that occurs during the decarboxylation in aqueous solution diminishes the partial positive charge on the pyridine ring. Consequently, the hydrogen bonds established between the protonated nitrogen atom of the pyridine ring and the water solvent molecules are weakened in the transition state relative to the reactant state. This effect obviously contributes to increase the activation energy in aqueous solution because a stabilizing interaction is partially lost when passing from reactants to the transition state.
In the enzyme active site the electrostatic environment is quite different from that found in solution. The quite rigid water-mediated interaction between Asp-276 and the protonated nitrogen atom of the pyridine ring (see Fig. 5) should favor the presence of a formal positive charge on this atom; then the quinoid structure is not especially favored. The averaged distance between the pyridine nitrogen atom and the oxygen atom of the water molecule is nearly the same in the reactant and the transition states (2.86 and 2.85 Å respectively). The net effect, comparing the reaction in solution and in the enzyme is a relative transition state stabilization in this latter environment through the interactions established with the protonated nitrogen atom of the pyridine ring. It is also interesting to notice that the different polarization induced by the enzymatic environment, favoring the localization of negative charge on the B group, is expected to also favor proton transfer to this atom, which must take place in the next step of the catalytic cycle (see Fig. 1).
The enzyme effect on the carboxylate group can be interpreted as the consequence of an active site electrostatically designed to favor the transition state charge distribution, where the charge is transferred toward the B group and to the pyridine ring. Positioning of Lys-308, Lys-202, and Asp-276 is the origin of this electrostatic complementarity with the transition state.
The changes observed in the reactant state are not an exclusive feature of this reaction. In fact, it has been often observed that the enzyme can induce changes on the substrate to drive it (geometrically or electronically) to the transition state (48). This effect has been recently recognized as the consequence of an active site electrostatically designed to accommodate the transition state (49 -52).
Summarizing, in the present work QM/MM-based molecular simulations on a previously validated mammalian HDC model made possible solving some questions that are hard or impossible to tackle by classical experimentation. We have analyzed in detail the decarboxylation reaction step in HDC, and we have been able to validate Dunathan's hypothesis about stereoelectronic effects contributing to the reactivity of the external aldimine. The presence of two positively charged lysine residues (Lys-308 and Lys-202) and a water-mediated hydrogen bond interaction with Asp-276 seem to be the key electrostatic factors contributing to the reduction of the activation free energy in the active site.