Crystallographic and Spectroscopic Studies of Native, Aminoquinol, and Monovalent Cation-bound Forms of Methylamine Dehydrogenase from Methylobacterium extorquens AM1*

Various monovalent cations influence the enzymatic activity and the spectroscopic properties of methylamine dehydrogenase (MADH). Here, we report the structure determination of this tryptophan tryptophylquinone-containing enzyme from Methylobacterium extorquens AM1 by high resolution x-ray crystallography (1.75 Å). This first MADH crystal structure at low ionic strength is compared with the high resolution structure of the related MADH from Paracoccus denitrificansrecently reported. We also describe the first structures (at 1.95 to 2.15 Å resolution) of an MADH in the substrate-reduced form and in the presence of trimethylamine and of cesium, two competitive inhibitors. Polarized absorption microspectrophotometry was performed on single crystals under various redox, pH, and salt conditions. The results show that the enzyme is catalytically active in the crystal and that the cations cause the same spectral perturbations as are observed in solution. These studies lead us to propose a model for the entrance and binding of the substrate in the active site.

Subsequently, electrons are transferred to the membranebound terminal oxidase, cytochrome aa 3 via a series of soluble electron carrier proteins (3). The oxidative deamination of the methylamine to the aldehyde occurs in two steps. The first step leads to substrate oxidation with the subsequent formation of a TTQ-aminoquinol. The second step involves the reoxidation of the protein cofactor and the release of the aldehyde product.
Two classes of MADH can be differentiated based on biochemical characterization and sequence comparison. The first class comprises the MADH from Methylobacterium extorquens AM1 (MADH-AM) (4), Thiobacillus versutus (MADH-TV) (5), and Paracoccus denitrificans (MADH-PD) (6) that interact with the cupredoxin amicyanin to form a transient electron transfer complex (7). The second class includes the MADH from Methylophilus methylotrophus W3A1 (MADH-WA) (8) and Methylobacterium flagellatum KT (MADH-KT) (9), which possess distinct biochemical and functional properties such as a different electron transfer partner. Light and heavy subunit sequences are 87 and 68% identical, respectively, for any two MADH belonging to the same subfamily, whereas these sequence identity levels drop to 72 and 49% for proteins from two distinct subfamilies.
The crystal structures of two MADH from the first class, MADH-TV (resolution 2.2 Å) (10) and MADH-PD (resolution 1.75 Å) (11), have been solved and show a characteristic protein fold and the unique cofactor. These quinoproteins consist of a homodimer of heterodimers. The mature heterodimer is composed of one heavy subunit (H; ϳ400 amino acids) and one light subunit (L; ϳ130 amino acids). The structure of the heavy subunit is composed of seven ␤-sheets containing four antiparallel ␤-strands each. The heavy subunit fold, known as a ␤-propeller, possesses pseudo 7-fold symmetry and is conserved among the MADHs, the central domain of galactose oxidase (12), and the ␤1-subunit of the signal-transducing G protein (13). Pairwise comparisons of the common core (160 -180 amino acids) of the distinct proteins from this structural superfamily show a very low level of sequence identity (12-15%) and a large root mean square (r.m.s.) deviation (2.8 -3.2 Å). The mature light subunit comprises the TTQ cofactor and six disulfide bridges. It is also mainly a ␤-structure but with no related fold yet identified.
The specific cofactor of the MADH is formed by oxidation of a tryptophan to an ortho-quinone, the indole-6,7-dione, and by formation of a covalent link to another tryptophan (2,14). The N ⑀1 and O-7 of the first tryptophan are hydrogen bonded to the backbone, whereas the O-6 carbonyl group is involved in the enzymatic reaction (15). Steady-state kinetic studies of MADH using phenazine ethosulfate as electron acceptor support a ping-pong mechanism (16). The reductive half reaction proceeds through a substrate imine intermediate at C-6 of TTQ followed by abstraction of a methyl proton by an active site nucleophile to give a carbanion intermediate concomitant with reduction of the cofactor (17). The resulting product imine intermediate is hydrolyzed to yield the aldehyde product and the reduced aminoquinol form of TTQ. In the oxidative half reaction, electrons are passed one at a time to an electron carrier while the TTQ forms an aminosemiquinone intermediate (18). Ammonia release then follows transfer of the second electron and reoxidation of TTQ to the quinone.
MADH is known to interact with a number of monovalent cations that cause spectral perturbations of the TTQ prosthetic group (19,20). These cations are of two types. Type I causes a red shift of the oxidized spectrum of MADH, whereas type II causes a bleaching of the visible spectrum and an increased absorbance in the near ultraviolet, similar to that observed during reduction of the enzyme by methylamine. Analogous spectral perturbations have been observed with another TTQcontaining enzyme, aromatic amine dehydrogenase (21). The type I binding site is specific for larger cations such as Cs ϩ , Rb ϩ , and NH 4 ϩ as well as di-, tri-, and tetramethylammonium ions. Their K d values decrease with pH, typically ranging from 10 -50 mM at pH 7 to 2-10 M at pH 10. The type II site binds smaller cations such as K ϩ and Na ϩ , with K d values typically much higher, ranging from 0.2-1 M at pH 7 to 0.5-1 mM at pH 10. Some ions bind to both sites with different affinities, causing a red shift at lower concentrations and bleaching at higher concentrations (19).
The type I cations act as competitive inhibitors of MADH reduction by substrate (22,23). A transient, red-shifted intermediate is also observed in the absorption spectrum of MADH prior to its reduction by methylamine (23). From these results, the binding site for these monovalent cations was predicted to be located at the active site in the vicinity of the cofactor atom O-6 and to mimic the first step of the reaction. The same ions were also shown to enhance the rate of oxidation of reduced MADH by amicyanin (24). To gain new insights into the role of these cations, we have investigated the crystal structures of MADH-AM and of some of its crystalline adducts at low ionic strength.
We present here the crystallographic studies of the native MADH-AM, its substrate-reduced form after reaction with methylamine, and its complexes with two monovalent cation inhibitors. We also present spectra of MADH recorded in the presence and absence of these monovalent cation inhibitors, which demonstrate that the spectral properties of the protein behave the same in solution and in the crystalline states. In the light of our results, we discuss the first steps of the MADH enzymatic reaction.
Mass Spectrometry-A matrix-assisted laser desorption ionization (MALDI) experiment was performed at the Mass Spectrometry Facility of Washington University (St. Louis, MO).
Measurement of Absorption Spectra in Solution-Titration of MADH-AM with cesium chloride and trimethylamine was carried out essentially as described earlier (19). Spectra were recorded with a Hewlett-Packard 8451A diode array spectrophotometer using a 1-ml cuvette at 25°C. Titrant solutions of CsCl and (CH 3 ) 3 NHCl were 1.0 -5.0 M in water. All spectra were corrected for dilution for each addition of titrant. Before use, the enzyme was treated with 50 mM K 3 Fe(CN) 6 overnight at 4°C to reoxidize a small amount of reduced enzyme. The K 3 Fe(CN) 6 was removed by seven cycles of washing with 25 mM bis-Tris propane buffer using a Centricon-30 centrifuge concentrator (Amicon).
Crystallization-Crystallization conditions were identified in a sparse matrix search by the hanging drop method using the Crystal Screen I and II (Hampton Research, Inc.). Crystals were obtained in different PEGs (2-8 K) and at various precipitant concentrations (15-25%) in acetate buffer at acidic pH. The best crystals were grown in sitting drops using a protein concentration of 8 -15 mg/ml in PEG4000 (19%) and sodium acetate buffer (70 mM, pH 4.1). Cations or substrate were introduced at 10 mM concentration into the mother liquor either prior to setting up the crystallization drops (CsCl) or later after the crystals were grown (CH 3 NH 3 ϩ and (CH 3 ) 3 NH ϩ ). Polarized Single Crystal Absorption Microspectroscopy-The methodology of polarized absorption spectroscopy of single crystals is well established (27,28). The spectra for this study were recorded using a Zeiss MPM800 microspectrophotometer. The crystals were placed in a flow cell with quartz windows, and the aqueous surroundings of the crystal were varied by flowing fresh solutions of various pH containing a variety of salts, reducing or oxidizing agents.
Data Collection-Two data sets (native and the cesium derivative) were collected on a Hamlin multiwire area detector, and two other data sets (methylamine-and trimethylamine-soaked crystals) were collected on a Rigaku Raxis II imaging plate, all at room temperature, using a Rigaku RU200 rotating anode x-ray source. The crystals were all monoclinic, space group P2 1 , and diffracted beyond 2-Å resolution. The unit cell dimensions for the native crystal were a ϭ 55.3 Å, b ϭ 94.0 Å, c ϭ 107.5 Å, and ␤ ϭ 99.9°and were roughly identical (within 1%) in all derivatives. The data recorded on the Hamlin detector were processed using software developed at the University of California at San Diego (29). The data recorded on the Raxis II were processed using the package DENZO/SCALEPACK (30).
Structure Analysis and Refinement-The structure of native MADH-AM was determined by molecular replacement using the program AMoRe (31). A search model consisting of backbone atoms (residues 59 -410 of the H subunit and 63-186 of the L subunit) and strictly conserved side chain atoms from the MADH-PD structure (11) was used. This partial model was constructed automatically with the program TITO (32), which incorporates the sequence alignment of all known MADHs and also estimates the validity of the model by pseudoenergy calculations. A complete model for MADH-AM was subsequently built by reorienting the output coordinates from TITO according to the transformation matrix obtained from the AMoRe solution using an in-house program. 2 Refinement of the native structure was performed by standard procedures for minimization and simulated annealing using the program X-PLOR (version 3.8.1) (33). Non-crystallographic symmetry restraints were not applied to the two independent HL dimers in the asymmetric units of any of the crystals. Restraints on the TTQ cofactor geometry, in addition to those normally applied to tryptophan side chains, were that the angle tryptophylquinone (Trq)-112C 3 -Trq-112C ⑀3 -Trp-163C ␦1 and the angle Trq-112C ⑀3 -Trp-163C ␦1 -Trp-163N ⑀1 (in the light subunit) be equal and that the O-6 and O-7 atoms be coplanar to the indole ring of Trq-112. The dihedral angle between the two indole rings of TTQ was not restrained. The distance between Trq-112C e3 and Trp-163C ␦1 (light subunit) was also not restrained except for the methylamine-soaked crystal, where unrealistic values for this distance were obtained in the absence of this restraint. Manual rebuilding was performed using the program O (33). Water molecules were added using the X-PLOR automatic procedure and verified individually using O.
The data quality for the substrate-reduced and cation-inhibited crystals was verified by molecular replacement with AMoRe. As these crystals were isomorphous, their structures were then determined by direct refinement of the MADH-AM structural model after stripping it of solvent. Modeling of solvent and ligands was performed using O; simulated annealing refinement and gradient refinement were carried out with X-PLOR.

RESULTS
Crystallization-Crystals grow in 1 week in 19% PEG4000 at low salt concentration (70 mM acetate buffer, pH 4.1). The MADH-inhibitor complexes were obtained either by soaking or growing the crystal with the desired salt. The aminoquinol form was obtained by soaking native oxidized crystals with methylamine rather than by growing crystals directly from substrate-reduced protein, which resulted in crystals that were too small. The complexation of the cations and the reaction with methylamine were qualitatively monitored through visual observation of changes in the crystal color. The stability of the crystals upon addition of the reagents suggested that little conformation change within the protein was taking place. Similarly, change of pH did not affect the appearance of the crystal. The monoclinic unit cell contains one heterotetramer, (HL) 2 , in the asymmetric unit, and the solvent occupies 55% of the cell volume.
Native Structure-Determination of the structure of MADH-AM was achieved using molecular replacement. The starting model was based on the refined structure of MADH-PD (11). Construction of the starting model was greatly simplified using TITO to automate the procedure of sequence comparison and generation of coordinates. Refinement of the native AM1 structure to an R factor of 16.5% (R free 20.9%) was done using several runs of simulated annealing, manual rebuilding, and addition of solvent molecules. The data collection and refinement statistics are shown in Table I.
The numbering scheme adopted for the amino acid sequence of the H and L subunits used in the crystal structure analysis is based on the sequences of the mauB and mauA genes in M. extorquens AM1, which encode polypeptides 411 and 186 residues in length, respectively (4). MALDI was performed to analyze the post-translational processing of the two subunits of this periplasmic enzyme. The observed peaks (not shown) correspond to each monomer, the dimer, and the complete tetramer. The molecular masses deduced from MALDI (L ϭ 13,962; H ϭ 41,170) are consistent with polypeptide cleavage after position 57 of the mauA gene product and after position 36 of the mauB gene product, resulting in an L subunit of 129 amino acids and an H subunit of 375 amino acids (computed molecular masses, L ϭ 13,967, H ϭ 41,157). In the crystallized protein, the first four residues of both mature H subunits and the first five residues and last two residues (185 and 186) of both mature L subunits are not seen in the electron density (contoured at 1). The refinement revealed, however, the structure of 16 amino acids at the N-terminal end of the heavy subunit (Gly-42(H) to Ala-58(H)) not seen in the crystal structure of MADH-PD. Continuous density is observed in the 2F o Ϫ F c map when contoured at 1.3 showing clearly the backbone and the side chains. Despite weakness in this part, the density was sufficient to build a 12-amino acid helical segment similar to that observed in MADH-TV (10) and in the MADH-PDamicyanin binary complex (7).
The heterotetrameric structure is composed of two heterodimers related by a non-crystallographic 2-fold axis roughly oriented perpendicular to the 7-fold axes of the pseudosymmetrical heavy subunits. The two cofactors are separated by a distance of 38 Å (O-6 -O-6Ј). The area of the solvent-accessible surface of the HL dimer, which is buried upon tetramer formation, is 3790 Å 2 , suggesting that the tetramer is stable in solution and represents the active form of the enzyme. The stabilization of the tetramer is mainly due to H1L2 and H2L1 contacts, whereas the H1H2 interface accounts for 715 Å 2 . The surface area buried at the interface of the H and L subunits in each dimer (H1L1 or H2L2) is also large (1600 Å 2 ). The Nterminal tail of a heavy subunit from one HL dimer is wrapped around the small subunit of the other HL dimer. This results in the L subunit being sandwiched by two distinct heavy subunits and about half buried. The remainder of the accessible surface of the small subunit comprises part of the cupredoxin binding site.
The r.m.s. deviation measured between C ␣ atoms of the two dimers, H1L1 and H2L2, of the of MADH-AM structure (which was refined without using any non-crystallographic symmetry restraints) is low (0.14 Å). The r.m.s. deviation between the C ␣ atoms of two light subunits is lower (0.10 Å) than that between the two heavy subunits (0.15 Å). This is probably due to the seven intrasubunit covalent bonds between side chains (six disulfide bridges plus TTQ in 130 amino acids for L, compared with one disulfide in 400 amino acids for H). A comparison of the common C ␣ atoms of the structures of MADH-AM and MADH-PD gives a low r.m.s. deviation (light subunit, 0.27 Å/121 amino acids; heavy subunit, 0.51 Å/345 amino acids; dimer, 0.46 Å/466 amino acids; tetramer, 0.52 Å/932 amino acids). In this case, their close structural similarity is probably due to their high sequence identity (89% for L subunits and 68% for H subunits). The conservation of the secondary structural elements in the heavy subunit, which comprise seven four-stranded antiparallel ␤-sheets common to other MADHs, was verified by secondary structure assignment based on the C␣ positions (35). The only major difference between the two heterodimers H1L1 and H2L2 resides in an external segment (residues 231-241) of the heavy subunits. The discrepancy is due to crystal packing involving this segment in one subunit (H2). In both H1 and H2, this segment comprises a ␤-turn (residues 233-236) and an extended structure (residues 237-241) on the surface of the protein. Continuous density is observed in the 2F o Ϫ F c map when contoured at 1.5 (but not at 2) in H2, whereas in H1 the density is much weaker, especially for the ␤-turn. In the absence of strong stabilizing interactions with the remainder of the structure, the residues 233-240 of subunit H1 and residues 233, 234, and 239 of subunit H2 seem partly disordered. The r.m.s. deviation of this 10-amino acid-long segment in the two different environments is very high (2.20 Å). Except for this segment, the two HL heterodimers are virtually identical, and no distinction between them will be made during the remainder of this article.
Only two residues, Ile-128(H) ( ϭ 65, ϭ Ϫ70) and His-209(H) ( ϭ 72, ϭ 161), lie outside the allowed zones of the Ramachandran plot. Otherwise, the overall quality of the structure is high according to PROCHECK (36). The electron density for these two residues is very well defined, and both are tightly packed in the protein interior. Their identity, conformations, and molecular environments are highly conserved in MADH-PD (11). Ile-128(H) belongs to a ␤-turn, which is close to the H1H2 interface. His-209(H) forms side chain hydrogen bonds to a carbonyl oxygen and a buried water; the two peptide planes adjacent to it are sandwiched by the side chains of Met-117(H) and Met-220(H). One cis-proline (Pro-184(H)) is also observed in MADH-AM and is present within a distorted polypeptide chain reversal as in MADH-PD (11). What role, if any, these strained peptide groups may play in the structure and function of MADH is not clear.
The side chains of residues Met-89, Val-111, Tyr-125, Leu-253, Val-364, Val-406, and Met-407 in the heavy subunits and Ser-115, Ser-128, Ile-127, and Met-172 in the light subunits were each modeled with two alternate side chain conformations. These include the sulfur atoms of three methionines that were each modeled to occupy two positions (separated by up to 2 Å) to fit weak and/or donut-shaped electron density. A concerted conformation change seems to take place for the side chains of Met-89(H) and Val-111(H), as their C ⑀ and C ␥ , respectively, are in close van der Waals contact distance.
Active Site Structure-The active site of the native enzyme is similar to that of MADH-TV (15) and MADH-PD (11). It resides at the interface of the H and L subunits. Although the heavy subunit contributes only one residue (Phe-80(H)) directly to the entrance of the substrate binding site, other residues from that subunit help stabilize the surroundings of the active site.
The cofactor is built up from the covalently linked Trq-112(L) and Trp-163(L). The refined TTQ geometry is presented in Table II   There are two solvent molecules in the active site pocket of subunit L1 of MADH-AM, whereas there is only one in subunit L2. These sites are weakly occupied, having average solvent temperature factors greater than 45 Å 2 , compared with an overall solvent B-factor of 30 Å 2 . One of the solvent sites of subunit L1 corresponds closely to one of the two solvent sites in MADH-PD (11).
Aminoquinol Enzyme Structure-The structure of MADH-AM reacted with methylamine was refined to 2.1-Å resolution with an R factor of 17.8 (R free 23.6%) (see Table I). The TTQ cofactor may bear a positive charge at position N-6 of the aminoquinol but may become neutral at physiological pH. The only major structural changes arising from substrate reduction occur in the active site pocket where some side chain reorientation occurs. The dihedral angle of TTQ increases by 10° (Table II) with all the rotation taken by Trq-112(L), whereas Trp-163(L) remains fixed (Fig. 2). The carbonyl oxygen of Asp-87(L) moves slightly as does the main and side chain of Asp-131(L). The result is that N-6 is now within hydrogen bond distance of the backbone carbonyl oxygen of Asp-87(L) but is farther from Thr-177(L) O ␥1 than in its native conformation, suggesting elimination or a marked weakening of the latter hydrogen bond. The Asp-131(L) oxygens O ␦1 and O ␦2 remain about the same distance from the N-6/O-6 atom. These changes seem to arise mainly from a different arrangement of hydrogen bonds and charges. A single, well defined water molecule (average B ϭ 25 Å 2 ) is present in the active sites of both HL dimers of the reduced protein, hydrogen bonded to the TTQ N-6 (distance 3.1 Å) and the main chain oxygen of Ile-161(H) (distance 2.8 Å). The remainder of the structure, after reduction by methylamine, is only slightly affected with respect to the native oxidized form (r.m.s. deviation of 0.15 Å).
Cation-bound Structures-The two complexed forms of the enzyme were also refined to 2.1-Å resolution, with an R factor of 15.5% (R free 20.4%) and 17.9% (R free 23.3%), respectively. For the CsCl crystal, two peaks appeared in the F o Ϫ F c map, at 11, and 10, respectively, in the active site pocket (Fig. 3A). The distance separating the peaks was 2.7 Å, indicating that the two sites are mutually exclusive and are partially occupied. The two cesium positions were refined with a partial occupancy of 0.3 to lower the B factors to 27 Å 2 (the average value for solvent) from 48 Å 2 , which was obtained with 0.5 occupancy each. This suggests that the concentration of cesium chloride in the crystals was not saturating. One site, the "proximal position," is about 3.3 Å from O-6 of TTQ, within 2.8 -3.6 Å of 3 carbonyl oxygen atoms (Asp-87(L), Asn-159(L), and Ile-161(L)), and about 3.9 Å from the carbonyl oxygen of Asp-160(L) (Fig.  3B). The other site, at the "distal position," is in van der Waals contact with the aromatic rings of Tyr-174(L) and Phe-80(H) (distance from the cesium to the ring carbon atoms of 3.3-4.2 Å) and is about 3.0 Å from Asp-87(L) O. This carbonyl oxygen is shared between the two sites, being the most distant carbonyl oxygen from the proximal Cs ϩ site. The carbonyl oxygen of Asp-87(L) moves about 0.5 Å closer to the two Cs ϩ ions from its native position. In the native structure (subunit L1), the two water molecules occupy similar positions to those of Cs ϩ but with high temperature factors.
Following refinement of the MADH-trimethylammonium complex, a 2F o Ϫ F c electron density map (Fig. 4A) revealed the binding of TMA in the active site at a position about halfway between the two Cs ϩ sites. The protonated nitrogen of TMA is hydrogen bonded to the carbonyl oxygen of Asp-87(L) (2.8 Å), which moves about 0.8 Å closer to the site to form the bond (Fig.  4B). The nitrogen atom is located 1.1 Å from the proximal Cs ϩ position at 3.8 Å from the O-6 of Trq. One inhibitor methyl group is directed toward O-6 of TTQ. The observed orientation of TMA in the active site explains the low reactivity of MADH with the ternary amines. Another methyl group makes van der Waals contact with the two aromatic rings of Tyr-174(L) and Phe-80(H). It occupies a position 0.5 Å from the distal Cs ϩ site. The inhibitor methyls are in van der Waals contact with Ile-162(L) C ␦1 and with the side chain of Asp-87(L). The side chains of Ile-162(L), Asp-87(L), and Phe-80(H) move by about 0.4, 1.2, and 1.3 Å, respectively, to make room for the bound TMA (Fig.  4B). Movement of the latter two side chains has the effect of opening up the active site "gate," which partitions the active site cavity from bulk solvent.
The remainder of the structure is only slightly affected upon the inhibitor binding (r.m.s. deviation 0.08 Å for both Cs and trimethylamine complexes) to the native oxidized form. The two inhibited forms differ from each other and from the reduced form by an r.m.s. deviation of 0.09 and 0.15 Å, respectively.
Single Crystal Microspectrophotometry Studies-Crystals of MADH-AM were found to be very stable from pH 4 to pH 10.5. They can be transferred to 40% PEG with no requirements for salt or buffer, and the pH can be adjusted within this range by titration with sodium hydroxide, requiring only small concentrations of the reagent. The polarized spectrum of the oxidized crystal is similar to that of the crystalline MADH-PD-amicyanin apobinary complex (34) and to MADH-AM in solution, with an absorption peak at ϳ 440 nm. Crystals titrated with sodium dithionite undergo transition through a semiquinone with max ϳ 425 nm and then to the reduced state with max ϳ 330 nm. Titration with methylamine yields the reduced state only, again like the MADH-PD apobinary complex, at a rate that is limited by diffusion through the crystals.
The pH has little effect on the polarized spectrum of the unliganded crystals of MADH-AM, but, as in solution, the spectral effects caused by both type I and type II cation binding are enhanced at high pH. The effects of the type I monovalent cations, cesium and trimethylammonium ion, on the absorption spectrum of MADH-AM as a function of concentration at pH 7.0 are shown in Fig. 5, A and B, respectively. The spectral shifts and absorbance increases are very similar to those observed for MADH-WA (34). In single crystals of MADH-AM, the binding of Cs ϩ and TMA ϩ (Fig. 5, C and D), also at ϳpH 7, are qualitatively similar to the results in solution, showing a 25-30-nm red shift and an absorbance increase with increased cation concentration. Binding to single crystals of MADH-AM at 300 mM Na ϩ , a type II cation, reduces the absorbance of the ϭ 440 peak 25% and gives increased absorbance at 330 nm (data not shown). These results also closely approximate the results of analogous experiments with MADH-WA in solution (20). Divalent cations have little effect, which is consistent with observations in solution. DISCUSSION MADH from M. extorquens AM1 is the first MADH structure for which the aminoquinol reaction intermediate of the enzyme has been analyzed. It is also the first for which monovalent cation binding has been structurally characterized. The resolution of the native MADH-AM structure, 1.75 Å, is comparable with that of MADH-PD (11). The structures of these two forms of the enzyme are remarkably similar, even to the extent of maintaining the same three residues of the heavy subunit in a strained conformation, Ile-128(H), His-209(H), and cis-Pro-184(H). This similarity exists despite the differences in sequence identity for the H and L subunits (32% and 13%) and the differences in ionic strength during crystallization (ϳ3 M ammonium sulfate versus ϳ20% PEG4000).
The structure of the aminoquinol form of MADH shows ex-tremely close correspondence to the oxidized form of the enzyme. All of the conformational differences are localized to the active site pocket and seem to result from changes in the redox state of TTQ and/or substitution of the quinone oxygen by an amino group. Movement is restricted to a 10°rotation of the Trq-112(L) side chain and replacement of a hydrogen bond from O-6 of oxidized TTQ to O ␥ of Thr-177(L) by a hydrogen bond from N-6 of reduced TTQ to the carbonyl oxygen of Asp-87(L). Similar changes were observed in two hydrazine derivatives of MADH-TV (15). In the latter structures, the positions of both the N ␣ and N ␤ azo nitrogen atoms of the bound inhibitors could be located, although additional atoms of the methyl or trifluoroethyl hydrazine moieties could not be seen. A similar rotation of the Trq group (ϳ6°) was observed, and the distance from the N/O-6 atom to O ␥ of Thr-177(L) increased from 3.1 Å to ϳ3.5 Å. However, both carboxylate oxygen atoms of Asp-131(L) (compared with only one in MADH-AM) seemed to be within hydrogen bonding distance of N ␣ , whereas the carbonyl oxygen of Asp-87(L) seemed to be hydrogen bonded to N ␤ of the azo group. MADH contains two classes of binding sites for monovalent cations that modulate the TTQ spectrum and the catalytic activity. In the case of MADH-PD, it has been shown (37) that catalysis of the reductive half-reaction is dependent on the presence of a metal ion. The high ionic strength required for crystallization of MADH-PD (ϳ3 M ammonium sulfate) (38) obviates the use of this enzyme for the study of cation binding since ammonium ion (a cation and an inhibitor of the enzyme) is present at such high concentration. Similarly, its binary and ternary complexes with amicyanin and with amicyanin plus cytochrome c 551I are grown in 2.2 M sodium-potassium phosphate buffer with high sodium and potassium ion concentrations (38,39). Likewise, MADH-TV crystals are grown from approximately 40% saturated ammonium sulfate solution (40) and are similarly unsuitable for studying these effects. We recently discovered that crystals of MADH-WA could be grown from PEG under low salt conditions and attempted to carry out cation binding studies in this system. 3 However, we found that these crystals were very difficult to grow, did not diffract well, and were not suitable for single crystal microspectrophotometry. Subsequently, we discovered conditions to grow crystals of MADH-AM that are reproducible, contain low concentrations of salt, are stable from pH 4 to pH 10, and are well suited for microspectrophotometry. These polarized single crystal microspectrophotometry studies were critical for verifying that the structural changes observed in the crystals reflect the changes that occur in solution upon binding cations.
It is interesting that specific cation binding occurs in MADH-AM crystals at pH 4.1. Although at pH 7.0 the K d for Cs ϩ is about 7 mM and for TMA ϩ about 0.3 mM for MADH-AM, it has been shown that monovalent cation affinity decreases as the pH is lowered. In the case of MADH-WA, for example, the K d values for Cs ϩ and TMA ϩ increase by an order of magnitude for each 2-pH unit decrease (19). The lowered affinity is due to protonation of a group on MADH so that if the pH is decreased enough, the affinity should level off as this group becomes fully protonated. At pH 4.5, the polarized spectrum of a single crystal of MADH-AM in the presence of 50 mM CsCl (not shown) exhibits approximately the same spectral shift as 3 mM CsCl at pH 7.1 (Fig. 5C). This indicates that significant cation binding can occur at 10 mM concentrations at pH 4.1 but may also account for the subsaturation of the Cs ϩ sites observed in the crystal.
The two cesium positions differ in their electrostatic and chemical environments. In the proximal position, nearer the O-6 atom, several carbonyl or carboxylate groups point toward the positively charged ion. In contrast, the other Cs ϩ position is involved in van der Waals contacts mainly with the planes of the aromatic side chains of Tyr-174(L) and Phe-80(H). This environment helps explain the specificity of MADH for large monovalent cations rather than polyvalent ones (41,42). The first crystallographic investigation of the binding of Cs ϩ to a protein was with rhodanese (43). Two internal binding sites were identified. One of these involves coordination of Cs ϩ by five oxygen atoms at an average distance of 2.7 Å in a square pyramidal configuration. The other involved van der Waals interaction of Cs ϩ with the -bonding system of the six-membered ring of a tryptophan side chain plus coordination by two oxygen atoms at an average distance of 3.0 Å. By comparison, the proximal Cs ϩ in MADH-AM is coordinated to five oxygens at an average distance of 3.3 Å, whereas the distal Cs ϩ is coordinated to one oxygen atom (3.0 Å) and to the ring systems of two aromatic side chains. The distal Cs ϩ position is occupied by a methyl group in the trimethylamine complex. The main difference between the two cation complexes is the interaction of the inhibitor with the surrounding residues, especially the side chains of Phe-80(H) and Asp-87(L). These two side chains are displaced to accommodate a methyl group of the trimethylamine. This demonstrates that the flexibility of these two side chains is sufficient to allow the entry of substrate into the active site pocket and to accommodate larger substrates such as longer primary amines (C1 to C5 n-alkylamines).
The study of these two complex structures suggests the binding of the substrate in a similar orientation upon entrance into the active site pocket. The methyl group of methylamine is likely to bind in the hydrophobic half of the active site, whereas the hydrogens will point toward the carbonyl groups lying in the hydrophilic half. Modeling such a complex suggests that the methylamine nitrogen atom may occupy a position intermediate to that of the analogous nitrogen atom of the trimethylamine and that of the proximal cesium position. This will favor hydrogen bonding to the carbonyl groups TTQ O-6, Ile-1610, and Asn-1590 (light subunits) not allowed in the TMA complex. The cavity entrance would be more tightly closed in the absence of the supplementary methyl groups.
In the two inhibited structures as well as in the native, neither the solvent nor inhibitor molecules bound into the active site are in contact with Asp-131(L) or with Thr-177(L). The catalytic functioning of these two side chains during the reductive half reaction seems to rely on the stabilization or activation of the TTQ cofactor. The active site cavity contains no side chain atom involved either in the recognition of the amine moiety or the activation of the substrate nitrogen except for the TTQ O-6.
At the low salt concentration (10 mM) and pH (4.1) used in the cation-binding experiments, we expected to observe only the type I cation binding. However, the observation of two mutually exclusive binding sites for Cs ϩ suggests that these sites may represent the type I and type II sites. In the cationbinding studies with MADH-WA (19), Cs ϩ was found to occupy both type I and type II sites with a negative cooperative factor of about 1000. The distal site would probably correspond to the type I site because it involves the interactions with the aromatic rings of Tyr-174(L) and Phe-80(H), which are common to the TMA type I interactions at the active site. The proximal Cs ϩ site would then correspond to the type II binding site. Further crystallographic studies at high pH with the type II cations will be needed to verify this hypothesis. In the reduced form of the MADH, a water molecule is hydrogen bonded to the N-6 atom and to the main chain carbonyl oxygen of Ile-161(L). The position of this water molecule is very similar to that occupied by the cesium, interacting with the O-6 in the oxidized MADH. It has been shown that cation binding to reduced MADH enhances its oxidation rate by amicyanin (24). The electron transfer to the cupredoxin is gated by a rate-limiting proton transfer as shown by stopped flow kinetic studies of the reductive half-reaction using deuterated substrate (37). The later study demonstrates a primary kinetic isotope effect of approximately 9 -17 (at 30°C), considerably higher than the semiclassical limit for proton abstraction (16). This binding site has the same affinity and the specificity as the type I site of the oxidized MADH (23,24). We postulate that the water molecule may be replaced by the cation that may partially occupy the proximal position. The presence of this positively charged element would repel the hydrogen atoms bonded to the N-6 toward the Asp-131(L) O ␦1 and O ␦2 . Such a mechanism will explain the enhanced electron transfer rate. Confirmation of this interaction will require the determination of the reduced MADH structure in the presence of salt. In that case, the active site would be partitioned by the TTQ in two halves: a binding site cavity and a transfer cavity comprising the catalytic triad Tyr-174(L), Thr-177(L), and Asp-131(L).