Crystal Structure of Histamine Dehydrogenase from Nocardioides simplex*

Histamine dehydrogenase (HADH) isolated from Nocardioides simplex catalyzes the oxidative deamination of histamine to imidazole acetaldehyde. HADH is highly specific for histamine, and we are interested in understanding the recognition mode of histamine in its active site. We describe the first crystal structure of a recombinant form of HADH (HADH) to 2.7-Å resolution. HADH is a homodimer, where each 76-kDa subunit contains an iron-sulfur cluster ([4Fe-4S]2+) and a 6-S-cysteinyl flavin mononucleotide (6-S-Cys-FMN) as redox cofactors. The overall structure of HADH is very similar to that of trimethylamine dehydrogenase (TMADH) from Methylotrophus methylophilus (bacterium W3A1). However, some distinct differences between the structure of HADH and TMADH have been found. Tyr60, Trp264, and Trp355 provide the framework for the “aromatic bowl” that serves as a trimethylamine-binding site in TMADH is comprised of Gln65, Trp267, and Asp358, respectively, in HADH. The surface Tyr442 that is essential in transferring electrons to electron-transfer flavoprotein (ETF) in TMADH is not conserved in HADH. We use this structure to propose the binding mode for histamine in the active site of HADH through molecular modeling and to compare the interactions to those observed for other histamine-binding proteins whose structures are known.

Histamine is an essential biogenic amine present in prokaryotes and tissues of animals and plants. In humans, histamine acts as a neurotransmitter, mediates allergic reactions, plays a role in cell proliferation, and is important in signaling the release of gastric acid into the stomach (1). Histamine receptors are the targets of drugs that treat allergies and stom-ach acidity, but there is very little structural information on the histamine-binding sites of these proteins. There is interest in isolating histamine-metabolizing enzymes that could be used as biosensors, and such enzymes may also prove useful for understanding histamine recognition (2). Although histamine oxidase from Arthrobacter globiformis (a topaquinone-containing copper amine oxidase) has been used as a histamine sensor, this enzyme was found to be more sensitive for dopamine and tyramine than for histamine (2). Trimethylamine dehydrogenase (TMADH) 5 from Methylophilus methylotrophus (3) and the quinoprotein methylamine dehydrogenase from Paracoccus denitrificans (4) have been used in general amine-sensing electrodes without the complication of O 2 chemistry, but neither are as effective as histamine sensors.
Histamine dehydrogenase (HADH) can be isolated from cultures of Nocardiodes simplex grown on histamine as the sole nitrogen source (5). HADH is a homodimer of ϳ76-kDa subunits and catalyzes the oxidative deamination of histamine to give imidazole acetaldehyde (Scheme 1) where the enzyme is remarkably selective for histamine (5,6) thereby showing great potential for use in a biosensor. HADH was originally classified as a quinone-containing amine dehydrogenase (5). Subsequent work by our group (6) and others (7) showed HADH to be a homologue of TMADH and the closely related dimethylamine dehydrogenase (DMADH) from M. methylotrophus, sharing ϳ40% sequence identity and 56% similarity with both proteins. The cofactors in this small family of proteins are the uncommon 6-S-cysteinyl-FMN, or 6-S-Cys-FMN ( Fig. 1) and [4Fe-4S]. The oxidized form of HADH displays a UV-visual spectrum with a max at 440 nm and a shoulder at 340 nm characteristic of the oxidized flavin, 6-S-Cys-FMN ox , as does TMADH (6). TMADH binds one molecule of ADP per subunit whose function is not known (8). The conserved ADP binding domain is also found in HADH (Lys 392 -Leu 423 ) (6) and DMADH and aligns with the FAD binding domain of other flavoproteins, such as Escherichia coli 2,4-dienoyl-CoA reductase (9).
The oxidation of histamine by HADH leaves the flavin in its reduced form, 6-S-Cys-FMN red , which is reoxidized by a stepwise electron transfer reaction through the [4Fe-4S] to either a mediator, such as phenazinemethosulfate (6) or, in TMADH, an electron transferring flavoprotein (ETF) (10). If either HADH or TMADH are reduced under single turnover conditions, a disproportionation reaction occurs between 6-S-Cys-FMN red and [4Fe-4S] 2ϩ to produce a semiquinone, 6-S-Cys-FMN sq , and [4Fe-4S] ϩ (Scheme 2) (6). In TMADH, the two spins form a triplet that is detected by EPR as a strong half-field signal (11)(12)(13). Under steady-state conditions at high substrate concentrations, the 6-S-Cys-FMN sq is stabilized by the binding of a substrate cation and can no longer be oxidized by [4Fe-4S], leading to substrate inhibition (6,14,15).
Although HADH is a homolog of TMADH, there are many differences that show that HADH is a functionally distinct enzyme. The substrate specificity of HADH differs significantly from both TMADH and DMADH, as secondary and tertiary amines are not substrates for HADH. HADH can only oxidize histamine (K m ϭ 31 M, k cat /K m ϭ 2.1 ϫ 10 5 M Ϫ1 s Ϫ1 ), agmatine (K m ϭ 37 M, k cat /K m ϭ 6.0 ϫ 10 4 M Ϫ1 s Ϫ1 ), and putrescine (K m ϭ 1280 M, k cat /K m ϭ 1500 M Ϫ1 s Ϫ1 ) (6). The trimethylamine-binding site of TMADH is well characterized, showing that the (CH 3 ) 3 NH ϩ is held in place by -cation interactions with Tyr 60 , Trp 265 , and Trp 365 (16). Two additional residues, Ser 74 and Trp 105 , along with Tyr 60 , have been shown to control the selectivity for trimethylamine over dimethylamine. Analogous to TMADH, monoamine oxidase and polyamine oxidase use aromatic amino acids to form -cation interactions with their amine substrates (17,18). The equivalent positions in HADH from an amino acid sequence alignment are Gln 59 , Glu 79 , Ala 110 , Trp 264 , and Asp 364 (6). These differences suggest that the histamine-binding site of HADH is significantly different from other flavin-dependent amine metabolizing enzymes. DMADH seems to provide a bridge between TMADH and HADH, containing residues Gln 59 , Thr 74 , Phe 105 , Trp 271 , and Trp 357 . Although sequence alignment suggests that the active sites of TMADH and HADH are very different, EPR shows that the magnetic interactions between the 6-S-Cys-FMN and [4Fe-4S] also differ between TMADH and HADH. In contrast to TMADH, the cofactors are not strongly spin-coupled in HADH suggesting that there are structural or conformational differences between the two enzymes (6).
The pH dependence of the steady-state kinetic parameters of histamine oxidation by HADH yielded two pK a values (pK a 1 ϭ 5.6 Ϯ 0.3 and pK a 2 ϭ 5.4 Ϯ 0.4) (6). The imidazole group of histamine must be neutral when it binds to the active site but the protonation state of the primary amino group of histamine was not conclusive as we were not able to determine the kinetics above pH 9 (6). The second pK a may represent the ionization of an active site residue, where an anionic residue in the active site could be responsible for stabilizing the monoprotonated histamine (6). HADH exhibits a significant primary kinetic deuterium isotope effect using ␣,␣,␤,␤-d 4 -histamine on k cat /K m (Ͼ7.0) that is absent in k cat (6). This suggests that C-H bond cleavage, and therefore 6-S-Cys-FMN reduction, is rate-limiting at subsaturating concentrations of histamine but that the subsequent steps, possibly electron transfer, define k cat . This contrasts with TMADH, where the effect of using (CD 3 ) 3 NH ϩ can only be examined by stopped-flow methods, and there is no isotope effect in the steady state (19).
Taken together, these studies strongly suggest that HADH is unique compared with TMADH and that they do not share a common mechanism. Detailed structure/function studies on HADH are needed to understand the substrate specificity and the fundamental chemistry underlying catalysis in this uncommon amine oxidizing flavoprotein. Toward this goal, we report the first three-dimensional structure of a histamine dehydrogenase.

MATERIALS AND METHODS
Expression and Purification of HADH-A recombinant form of HADH was expressed and purified as described previously (20). Briefly, HADH was expressed in E. coli Rosetta 2 (DE3) cells (Novagen) using a pET-21b vector (Novagen). HADH was purified by Toyopearl DEAE, Toyopearl Butyl-650 (Tosoh Bioscience), and HiLoad Superdex 200 16/60 sizing column (GE Healthcare) chromatography. Full flavinylation of HADH was confirmed by the ratio of A 444 /A 382 ϭ 1.38 (7). The activity of HADH against histamine was determined using the standard assay (5). The selenomethionine-substituted HADH (SeMet-HADH) was prepared and purified as described previously (20) and the substitution of 10 of the 13 Met with SeMet was confirmed by matrix-assisted laser desorption ionization  Crystallization and Data Collection-The crystallization of HADH and SeMet-HADH, subsequent data collection, and data processing were carried out as described previously (20). In short, crystals of HADH were obtained by equilibration versus 0.1 M HEPES, pH 7.4, 2.0 M ammonium sulfate, 2% polyethylene glycol 400, and 4% glycerol using hanging-drop vapor diffusion. Glycerol (25%) was used as a cryoprotectant and data were collected from cryocooled crystals. A data set was collected to 2.7-Å resolution from crystals grown from SeMet-HADH at beamline BL9-2 at the Stanford Synchrotron Radiation Laboratory (SSRL) (21). The estimated maximal error is 0.165 Å and the average coordinate error estimated from the Luzzati plot is 0.294 Å. Data collection was performed at 100 K. Data integration and scaling were performed with MOSFLM (22) and SCALA (23).
Structure Solution, Refinement, Rebuilding, and Model Analysis-As described previously (20), structure determination by SOLVE/RESOLVE (24,25) using the complete threewavelength multiwavelength anomalous dispersion data set was not immediately successful. However, using only the single-wavelength 0.98-Å data, we were able to determine the structure by molecular replacement using the program Phaser (26). The search model was the polypeptide monomer of TMADH (PDB accession number 1DJN, 2.2-Å molecule A only) (8), which had nonconserved residues pruned back to the last common atom. A log likelihood gain value of 247 and a translation function Z-score of 18.1 were obtained in Phaser (26) for two solutions without packing clashes. The x-ray diffraction data used to solve the HADH structure was collected to a resolution of 2.7 Å with 99.7% completeness and an R merge of 16%. The Ramachandran plot showed 88.4% of residues in the most favored region, 11.1% in the additional allowed regions, 0.5% in the generously allowed regions, and no residue was found in the disallowed region. The crystal belongs to an orthorhombic space group P2 1 2 1 2 1 , with unit cell parameters a ϭ 101.14, b ϭ 107.03, c ϭ 153, 35 Å. Each asymmetric unit contained two molecules of HADH. The Matthew's coefficient was 2.79 Å 3 Da Ϫ1 , giving a solvent content of 55.96% (27). NCS restraints were imposed during the early phases of refinement and relaxed during later phases of refinement. In the final rounds of refinement, NCS restraints were removed. The refinement statistics are summarized in Table 1. The initial model of HADH was iteratively refined using the data set of 2.7-Å resolution with REFMAC (28) and built manually using COOT (29). The model was validated using PROCHECK (30) and WHAT IF (31). Superimposition of structures of HADH and TMADH and all molecular figures were generated by PyMOL (32). The PDB deposition code is 3K30.
Modeling Study-Structural models for Arg 444 (HADH) and Tyr 442 (TMADH) were obtained by extracting the relevant amino acids directly from their corresponding crystal structure using SYBYL 8.0 (Tripos International, St. Louis, MO). The models were then protonated in SYBYL according to the following specifications: Arg 444 was assumed to be cationic, whereas Tyr 442 was assumed to be neutral; both amino acids were assigned neutral backbones with amine and aldehyde termination at the N and carbonyl ends, respectively. The HOMO-LUMO gaps for each were computed in Gaussian 03 (Gaussian, Inc., Wallingford, CT) at a B3-LYP level of representation (33, 34) using the 6-31G* basis set (35).
Docking Study-The ligands (monoprotonated and neutral forms of histamine and trimethylamine) were sketched in SYBYL 8.0 (Tripos International) and refined via molecular mechanics optimization using the Tripos Force Field (36) and Gasteiger-Marsili electrostatics (37). The two protonated states of histamine ligands were docked into the HADH active site (receptor) by FlexX (38), retaining for explicit consideration all residues and crystallographically resolved heteroatomic moieties within 8.0 Å of the receptor center. All other docking controls were left at default settings. Three ligands (trimethylamine and two protonated states of histamine) were docked into the TMADH active site (receptor) by Surflex (39), as guided by a receptor-based protomol file defined by the position of the 6-S-Cys-FMN cofactor molecule (i.e. ligand binding modes that interact with this molecule are preferentially sought). Ligands via space-filling spheres were scaled according to van der Waals radii. All other docking controls were left at default settings.

RESULTS AND DISCUSSION
Primary Characterization of HADH-The recombinant HADH is fully flavinylated judging from the ratio of A 444 /A 382 ϭ 1.38, and exhibits identical UV-visual and EPR spectroscopic properties as the native HADH (6) (data not shown). The kinetic parameters of histamine oxidation by recombinant HADH and the native HADH are in good agreement, and furthermore, like the native form, histamine shows substrate inhibition with recombinant HADH (supplemental Table S1). These results show that recombinant HADH can be used in place of native HADH in the structure determination.
The Overall Structure of HADH-Recombinant HADH was crystallized as a homodimer with two molecules per asymmetric unit (Fig. 2a). The overall structure is very similar to that of  Fig. S1), a large domain (residues 7-385), a medium domain (residues 386 -491 and residues 622-690), and a small domain (residues 492-621). The large domain contains an N-terminal triosephosphate isomerase barrel, the most common tertiary fold observed in protein crystal structures (41). The 6-S-Cys-FMN is located at the opening of the barrel and surrounded by ␣ helices and large excursions at the end of ␤ strands of the barrel that serve to cover and bury the 6-S-Cys-FMN. The 6-S-Cys-FMN is covalently linked to Cys 35 positioned after the first parallel ␤-strand of the barrel. The [4Fe-4S] cluster is positioned outside the barrel and connected to the end of an ␣-helix close to the medium domain. The [4Fe-4S] cluster is coordinated to four Cys residues (Cys 348 , Cys 351 , Cys 354 , and Cys 366 ). As predicted from the sequence alignment (6), one molecule of ADP is bound per protein molecule, located in the medium domain. ADP is exposed to the surface, next to the interface of the medium and small domains, but is not covalently linked, as is observed for TMADH (8,42).
The Active Site of HADH-The electron density for the 6-S-Cys-FMN is very clear (Fig. 2b), consistent with a fully flavinylated enzyme. By comparison, crystals of recombinant TMADH lacked much of the electron density at the flavin site. The estimated average occupancy for the FMN in the recombinant TMADH (PDB code 1DJN) was reported as 0.55 (43). The electron density suggested that the isoalloxazine ring of the 6-S-Cys-FMN in recombinant HADH is not planar but in a "butterfly bend" conformation centered at the N5 and N10 positions as seen in the recombinant TMADH (43). An overlap is shown of FMN from TMADH (cyan carbons), FMN from HADH without planar restraints (green carbons) and with planar restraints (magenta carbons)  on the isoalloxazine ring system in Fig. 2c. When the FMN in HADH is refined without planarity restraints on the isoalloxazine ring, the deviation from planarity is ϳ15°, which is slightly less than that in TMADH (20°) (43).
Although the factors contributing to the butterfly bend in TMADH have not been defined yet, the cysteinyl cross-link formation has been shown not to cause the bend as a similar bend was seen in the x-ray structure of the C35A mutant of TMADH (39). We have looked into the active site of HADH surrounding the benzene ring of FMN where the bend has been seen (Fig. 2c) in comparison to the corresponding area in TMADH, but we have not been able to find particular interactions that contribute to the extent of butterfly bend. Further study using higher resolution crystals and site-directed mutagenesis is necessary.
The redox potentials of HADH have been determined to be ϩ34 mV (fully oxidized form to one electron-reduced form) and ϩ30 mV (one electron-reduced form to fully reduced form), respectively (44), whereas the reported values of TMADH are ϩ44 and ϩ36 mV, respectively (45). Theoretical studies on lumiflavin and C6-methylsulfanyllumiflavin suggested that the butterfly bend raises the twoelectron reduction potential, making the flavin more reactive to substrate oxidation (46,47). The changes in the redox potential of the 6-S-Cys-FMN cofactors in HADH and TMADH are relatively small, yet support the theoretical study where the redox potential of the former is slightly more negative than the latter.
The electron density for the [4Fe-4S] cluster is also very clear (supplemental Fig. S2). The spatial arrangement of 6-S-Cys-FMN and the [4Fe-4S] cluster in HADH is very similar to that in TMADH (8) (Fig. 3). The distance between the 8␣-methyl carbon of 6-S-Cys-FMN and the closest iron in [4Fe-4S] is 5.7 Å, and to the closest Cys ligand (Cys 354 ) of the cluster is 4.1 Å, respectively. It has been proposed that the electrons from substratereduced flavin pass through the 8␣-methyl to the closest Cys ligand upon reoxidation in TMADH (8). The oxidation of 6-S-Cys-FMN red in HADH may go through a similar electron transfer mechanism.  In the native HADH, the spin coupling of the semiquinone of 6-S-Cys-FMN and [4Fe-4S] ϩ are not as strong as that observed in TMADH (6). We previously proposed that the spatial arrangement of the two cofactors or the surrounding residues could affect the spin coupling of the two cofactors (6). However, the active site structure of HADH revealed that the distances between the two cofactors are almost identical in HADH and TMADH (Fig. 3). We saw small differences in the location of two of the triad residues in HADH. The hydroxyl of Tyr 176 resides 3.6 Å away from the O 2 carbonyl of 6-S-Cys FMN, which is ϳ1 Å further away than the hydroxyl of Tyr 169 in TMADH. His 179 is 0.5 Å closer to the pyrimidine moiety of 6-S-Cys-FMN than His 172 in TMADH. Arg 222 in TMADH is 2.6 Å from the O 2 carbonyl of 6-S-Cys-FMN and is proposed to stabilize the negative charge building up on the N1 and C2 of the isoalloxazine ring during the 6-S-Cys-FMN biogenesis (50). Arg 230 in HADH is spatially conserved with Arg 222 in TMADH and is 3.6 Å away from the O 2 carbonyl of the 6-S-Cys-FMN. Further studies are necessary to define whether the small differences of Tyr 176 and His 179 are the origin of the weak spinspin coupling observed in HADH. We expect that F176Y exhibits spin-spin coupling to the same extent as the wild-type but W176Y may exhibit stronger coupling only if the mutation does not disrupt the rest of the active site environment.
In the proposed catalytic mechanism of HADH and TMADH, a water molecule plays a catalytic role, i.e. to produce corresponding aldehydes from the imine products. However, no water molecule has been found in the active site of HADH, probably due to 2.7-Å resolution, but this could also be due to high mobility of water molecules as no water molecule has been detected in the active site of TMADH (2.2 Å) either.
Electron Transfer Pathway-X-ray crystal structures of TMADH-ETF complex (2.0 and 3.7 Å) revealed that ETF interacts with a shallow surface groove of TMADH, comprised of nine amino acids, namely Ser 391 , Leu 393 , Thr 414 , His 416 , Gln 462 , Ala 464 , Tyr 478 , Gly 479 , and Ala 480 (Fig. 4a and  supplemental Fig. S3), through its proposed recognition peptide sequence (51). Leu 194 in the ␤ subunit of ETF is proposed to act as an anchor and positions ETF into the groove of TMADH via van der Waals interactions with Leu 393 and Tyr 478 .
The presence of an ETF-like protein has not yet been determined in N. simplex. By inspection of the crystal structure of HADH, there seems to be a similar shallow groove on the surface (Fig. 4b).
The surface residue Tyr 442 in TMADH has been shown to play a key role in the electron-transfer reaction between the [4Fe-4S] center and ETF (52)(53)(54). Tyr 442 resides in the surface groove ( Fig. 4a and supplemental Fig. S4a) different from that discussed earlier. Mutations of Tyr 442 to Phe, Leu, Cys, and Gly lead to perturbation of the interaction between TMADH and ETF, resulting in retardation of the electron-transfer reaction (52). It has been proposed that Tyr 442 transiently pairs with Arg 237 in the ␣ subunit of ETF to stabilize productive electrontransfer configurations. The residue corresponding to Tyr 442 is Arg 444 in HADH. Molecular orbital calculations suggest that the Arg 444 residue in HADH may be better suited to stabilize the charge transfer complex than the Tyr 442 in TMADH. Specifically, the HOMO-LUMO gap of Arg 444 is 4.49 eV, whereas the HOMO-LUMO gap of Tyr 442 is 4.91 eV, thus Arg 444 requires less energy to accept a negative charge than Tyr 442 would.
In TMADH, Val 344 resides at the bottom of the surface groove where Tyr 442 resides and engages in van der Waals interactions with Cys 345 and Tyr 442 (Figs. 3 and 4a). Val 344 has been shown to participate in the electron transfer to ferricenium, an artificial electron acceptor (52). From the molecular modeling study, it has been proposed that ferricenium binds in this hydrophobic surface groove and can be in close contact with Val 344 . Mutations of Val 344 to amino acids with smaller side chains (Ala, Gly, or Cys) resulted in increasing the rate of electron transfer and it was proposed that shortening the electron-transfer pathway and/or changes in packing density accelerates the reaction. Mutation of Val 344 to bulkier side chains (Ile or Tyr) to reduce the size of the groove resulted in an impaired electron transfer reaction. In HADH, the corresponding residue to Val 344 is Glu 347 making this groove hydrophilic rather than hydrophobic (supplemental Fig. S4). Taken together, we found that the electronic environment of these surface grooves involved in ETF binding and electron transfer reaction in TMADH is not strictly conserved in HADH, where both grooves in the latter contain charged residues.
Two possible electron-transfer pathways from the [4Fe-4S] to the surface Tyr 442 are proposed for TMADH, namely Cys 345 3 Glu 439 3 Tyr 442 or Cys 345 3 Val 344 (48) (supplemental Fig. 5a). The first path is used to transfer electrons to ETF and the second path is used to transfer electrons to ferricenium. Interestingly, HADH has a very analogous pair of choices, namely Cys 348 3 Glu 347 3 Arg 444 or Cys 348 3 Ala 441 . These pathways are geometrically comparable with the two pathways in TMADH, except the positions of the possible intermediates (Glu 437 in HADH and Gly 439 in TMADH) are reversed (supplemental Fig. 5b). Although we have not identified an ETF-like protein in N. simplex, it is possible that HADH utilizes the Cys 348 3 Glu 347 3 Arg 444 pathway to its electron acceptor protein.
Substrate (Histamine)-binding Site-Although the overall three-dimensional fold of TMADH and HADH are very similar, there are some significant differences between the two active sites. In TMADH, substrate binds to the "aromatic bowl" created by Tyr 60 , Trp 264 , and Trp 355 (55). The binding of trimethylamine in TMADH involves cation-interaction of the methylammonium ion and the aromatic bowl. The corresponding residues in HADH are Gln 65 , Trp 267 , and Asp 358 as predicted from the sequence alignment (6), as shown in Fig. 3.
Modeling of Histamine into the Active Site of HADH-To gain insight into the histamine selectivity of HADH, histamine was modeled into the active site in both neutral and monoprotonated forms. The modeling was chosen over crystal structure determination of the histamine-bound form of HADH as we have not found the optimal condition to prevent oxidation of histamine to the corresponding aldehyde in the active site of HADH yet. Both forms bound in the same fashion, with the latter predicted to be more energetically favored than the former (Ϫ21.5 kcal/mol versus Ϫ14.4 kcal/mol). Histamine is positioned over the si-face of the isoalloxazine ring of 6-S-Cys-FMN (Fig. 5). The positively charged amino group of histamine forms salt bridges or hydrogen bonding interactions with the carboxylate groups of Asp 358 and Glu 79 . The carbonyl oxygen of the Asn 115 side chain is in hydrogen bond interaction with the NE2 atom of the imidazole ring. The amido group of Gln 65 is in hydrogen bonding distance from the CD2 proton of the imidazole ring. In addition to the hydrogen bonding residues, a number of hydrophobic side chains (Ile 74 , Phe 77 , Tyr 181 , and Trp 267 , respectively) line the binding pocket and have van der Waals contact with histamine. Among them, the aromatic ring of Tyr 181 is in -stacking interaction with the neutral imidazole ring of histamine and Phe 77 is packed edge-to-face against the imidazole ring. The closest approach of the plane of the ring is 3.6 and 3.7 Å for Tyr 181 and Phe 77 , respectively. This position places the protonated amino group of histamine close to the isoalloxazine ring, as would be required for catalysis to the imidazole acetaldehyde. The histamine-binding site is located at the center of the dimer interface and a channel connects both binding sites to the surface (Fig. 6).
Modeling of Histamine into the Active Site of TMADH-To gain further insight into histamine selectivity of HADH, we docked trimethylamine and monoprotonated and neutral forms of histamine into the active site of TMADH. Trimethylamine successfully docked into the proposed aromatic bowl in the active site, comprising Tyr 60 , Trp 264 , and Trp 355 (supplemental Fig. 6a) (56). However, both forms of histamine ligands failed to dock into the active site, as docking efforts consistently placed histamine outside of the active site. When manually superimposing histamine onto the aromatic bowl, where trimethylamine binds, we found that histamine is too bulky to fit into the aromatic bowl (supplemental Fig. 6b) leading to insurmountable steric clashes. These results indicate that the substrate-binding site of HADH and TMADH are evolved to best accommodate each substrate by specific interactions with amino acid side chains as well as the size of the cavity.
Comparison of Histamine Binding Motifs in Histamine-binding Proteins-The histamine binding motif in the active site of HADH (Fig. 7a) from the modeling study correlates closely with the x-ray crystallographically determined binding sites of histamine N-methyltransferase from human (HNMT) (Fig. 7b) FIGURE 6. The proposed substrate-entry channels in HADH. There are two channels connecting the surface to each active site (only one channel is shown). The bound histamine predicted from the docking study is shown in magenta. a, a top view of the channel. b, a side view of the channel.
(57) and histamine-binding proteins (HBPs) from the saliva of blood sucking insects, such as the ticks Rhipicephalaus appendiculatus (Fig. 7c) (58) and Argas monolakensis (supplemental Fig. 7a) (59), and the malaria mosquito, Anopheles gambiae (supplemental Fig. 7b) (60). HBPs bind histamine that is released by the host during blood feeding to suppress inflammatory response of the host. HNMT inactivates histamine by transferring a methyl group from S-adenosyl-L-methi-onine to the imidazole ring. In the HBPs, the negatively charged residues such as Asp and Glu stabilize the positively charged amino group of histamine, whereas -stacking interactions with Tyr, Phe, and Trp help to stabilize the imidazole ring of histamine (58 -60). The selective binding of histamine in HNMT is achieved by ionic or hydrogen bonding interactions of Glu, Gln, and Asp with the positively charged amino group and -stacking interaction of Tyr and Trp with the imidazole ring (57). Even though the function and mechanism of HBPs, HNMT, and HADH differ, the histamine-binding motif is clearly conserved and this contributes to their high histamine selectivity. HNMT and HBPs are not sequentially homologous to HADH, suggesting that the similarities in histaminebinding motifs have arisen through convergent evolution.
The sequences of the human histamine receptors (HHRs) are known (61)(62)(63)(64)(65), but the crystal structures have yet to be determined. Molecular modeling studies of HHRs have provided insights into the binding mode of histamine into these receptors (66 -69). In all four HHRs, a salt bridge between a conserved Asp and the positively charged amino group of histamine is proposed to be essential for histamine binding. HHR1 and HHR4 both contain an Asn residue that is predicted to be in hydrogen bond interaction to the imidazole nitrogen of histamine. In HHR1, a Tyr residue is predicted to stabilize histamine binding by forming a -stacking interaction with the imidazole ring (70). The binding motif of HADH demonstrates significant similarities with HHRs, and especially the binding motif of HHR1. Conclusion-We have solved the crystal structure of a recombinant form of HADH to 2.7-Å resolution. The overall structure, domain organization, and the spatial arrangement of 6-S-Cys-FMN and [4Fe-4S] are very similar to those of TMADH but we found significant differences at the substratebinding site as well as on the surface proposed to be from contact with electron acceptors like ETF in TMADH. From the docking study, we predict that monoprotonated histamine preferentially binds to the active site in the following manner. The anionic residues (Glu 79 and Asp 358 ) stabilize the protonated amino group of histamine by salt bridge, whereas the aromatic residues (Phe 77 and Tyr 181 ) stabilize the imidazole ring via a -stacking interaction. This binding mode of histamine is very similar to those found in the structures of other histamine-binding proteins and consists of a model for understanding the selectivity of HHRs, in particular HHR1. The similar binding mode of histamine into HADH and HHR1 could make the HADH substrate recognition site a potential model for HHR1.