NikD, AN UNUSUAL AMINO ACID OXIDASE ESSENTIAL FOR NIKKOMYCIN BIOSYNTHESIS: STRUCTURES OF CLOSED AND OPEN FORMS AT 1.15 AND 1.90 Å RESOLUTION

Summary

NikD is an unusual amino acid oxidizing enzyme that contains covalently bound FAD, catalyzes a 4-electron oxidation of piperideine-2-carboxylic acid to picolinate and plays a critical role in the biosynthesis of nikkomycin antibiotics. Crystal structures of closed and open forms of nikD, a two-domain enzyme, have been determined to resolutions of 1.15 and 1.9 Å, respectively. The two forms differ by an 11° rotation of the catalytic domain with respect to the FAD-binding domain. The active site is inaccessible to solvent in the closed form; an endogenous ligand, believed to be picolinate, is bound close to and parallel with the flavin ring, an orientation compatible with redox catalysis. The active site is solvent accessible in the open form but the picolinate ligand is approximately perpendicular to the flavin ring and a tryptophan is stacked above the flavin ring. NikD also contains a mobile cation binding loop.

INTRODUCTION

NikD is a flavoenzyme that catalyzes a key step in the biosynthesis of nikkomycins. Nikkomycins are a closely related group of peptidyl nucleoside antibiotics that resemble the natural substrate of chitin synthase and act as potent and specific competitive inhibitors of this enzyme (Fig. 1) (Fiedler et al., 1982). Chitin is a homopolymer of N-acetyl-D-glucosamine that is an essential structural component of the cell wall in fungi and the exoskeleton of invertebrates. Chitin is not found in mammals. Nikkomycins act as potent antifungal agents against several important human pathogens and are also useful in agriculture as insecticides that are non-toxic for bees and mammals (Hector, 1993). The dramatic rise in the population of immunocompromised patients has spurred great interest in nikkomycins and related antifungal agents.

Figure 1.

Structure of nikkomycin and streptogramin antibiotics (top panel) and the nikD-catalyzed oxidation of P2C to picolinate (bottom panel). The peptidyl portion of each antibiotic is shown in red with a black box indicating atoms derived from L-lysine. The variable nikkomycin nucleoside moiety is shown in blue. The DHP intermediate shown in the bottom panel is one of 6 possible isomers.

The peptidyl and nucleosidyl portions of nikkomycins are synthesized in separate pathways and then linked by a peptide bond (Bormann et al., 1989). The nikkomycin peptide is synthesized via a nonribosomal pathway and contains a N-terminal pyridyl moiety, derived from L-lysine, that is essential for antibiotic activity. Biosynthesis of the pyridyl moiety is initiated by a L-lysine α-aminotransferase that converts the amino acid to piperideine-2-carboxylate (P2C), a compound that can exist in imine and enamine tautomeric forms. NikD catalyzes a remarkable aromatization reaction that converts P2C to picolinate, accompanied by the reduction of 2 mol of oxygen to hydrogen peroxide. The 4-electron oxidation of P2C proceeds without apparent release of the labile dihydropicolinate (DHP) intermediate (Bruckner, R. C. et al., 2004; Venci et al., 2002) (Fig. 1). A nikD-like reaction is implicated in the biosynthesis of the N-terminal pyridyl moiety found in streptogramin antibiotics that are used to treat infections caused by multi-drug resistant gram positive bacteria (Fig. 1) (Barriere et al., 1998).

NikD from Streptomyces tendae has 389 residues, covalently binds one molecule of FAD through a cysteinyl linkage to Cys321 (Venci et al., 2002) and exists as a monomer in solution. NikD acts as an obligate two-electron acceptor in a process that generates the fully reduced form of its flavin prosthetic group. The enzyme exhibits two absorption maxima in the visible region, as expected for a flavoenzyme, plus an unusual long wavelength absorption band due to charge transfer interaction of the flavin with Trp355 (Bruckner, R.C. et al., 2007). NikD is a member of a family of FAD-containing amino acid oxidases that includes monomeric sarcosine oxidase (MSOX) (Trickey et al., 1999). NikD shares about 25% sequence identity with MSOX. Other members of this family include N-methyltryptophan oxidase (MTOX), pipecolate oxidase (PIPOX) and heterotetrameric sarcosine oxidase (TSOX) (Wagner et al., 1999). As compared with other family members, nikD is unique with respect to its ability to catalyze an overall 4-electron oxidation reaction. In this paper we report the crystal structure of two forms of nikD and discuss their potential catalytic significance.

Results and Discussion

Structure determination

Three crystals of the product of the nikD gene, prepared under slightly different conditions (Table 1), were used in this study. The crystal structure was first determined by MAD phasing at a resolution of 1.75 Å by the replacement of the five methionine residues of the mature protein (lacking the amino-terminal methionine) with selenomethionine (SeMet). Four of these selenium atoms refined to full occupancy, while the fifth selenium atom refined to approximately half occupancy. The partially occupied selenium site corresponds to the residue Met110, which is the first of two consecutive methionine residues in the nikD sequence. The unbiased experimental electron density map calculated with MAD phases was of high quality, and was interpretable for residues 3-385 as well as the FAD prosthetic group. The map (Fig. S1) clearly showed the site of covalent attachment at the C(8) methyl group of the flavin ring to residue Cys321 of the protein, consistent with biochemical studies (Venci et al., 2002). In addition, electron density for an endogenous small molecule was found near the isoalloxazine ring of FAD. Although initially modeled as a benzate anion, the ligand was subsequently refined as the isostructural compound picolinate, an assignment consistent with the finding of picolinate in nikD extracts (Bruckner, R.C. et al., 2007). The ligand in all three nikD crystals has been refined at full occupancy. However, the isolated enzyme is found to contain less than 0.1 mol of picolinate, a difference that has been attributed to selective crystallization of the enzyme•picolinate complex (Bruckner, R.C. et al., 2007). The final model of SeMet nikD consists of residues 3-385, one FAD molecule, 264 water molecules, and one molecule of picolinate.

Table 1.

Data collection and refinement statistics

Crystal type	SeMet				Closed	Open
Crystallization
Precipitant	PEG 1500 (~30%)				PEG 1500 (~30%)	MPD (~25%)
Buffer	HEPES (50 mM) pH 6.6				ADA (100 mM) pH 6.5	MES (100 mM) pH 5.7
[Na+]	~5 mM				~100 mM	~100 mM
Cell parameters	a=78.52 Å b=96.12 Å c=78.05 Å β=118.49°				a=78.24 Å b=95.60 Å c=77.90 Å β =118.30°	a=87.71 Å b=90.57 Å c=85.40 Å β =118.54°
Data collection
Wavelength (Å)	0.9791	0.9794	0.95000	1.0000	0.9000	0.9000
Data set	λ1 (peak)	λ2 (inflection)	λ3 (remote high)	λ4 (remote low)
Resolution range (Å)a	50-1.75	50-1.75	50-1.75	50-1.75	1.15	1.9
# Reflections	384455	376632	374339	374960	851037	155334
# Unique reflections	51037	50987	50823	50708	160538	43421
Rmerge (%)b	6.8 (30.4)	7.3 (36.6)	7.0 (47.3)	7.0 (42.5)	4.7 (20.5)	3.9 (20.0)
Completeness (%)	99.7 (99.5)	99.6 (98.0)	99.3 (94.7)	99.3 (95.7)	90.0 (46.1)	92.8 (60.2)
Anomalous completeness (%)	98.7	98.4	97.5
<I/σ(I)>c	33.6 (5.9)	30.9 (4.1)	28.7 (2.8)	28.3 (3.4)	27.3 (4.5)	29.8 (5.0)
Redundancy	3.8	3.7	3.7	3.7	5.3	3.6
Phasing
Ranomalousd	0.37	0.34	0.43
(wavelengths)	λ1-λ2	λ1-λ3	λ2-λ3
Rcentrice	0.72	0.66	0.76
Phasing power	1.3	1.9	1.2
Refinement
Resolution range	50-1.75 Å				50-1.15 Å	50-1.9 Å
Rworkf	0.191				0.127	0.185
# reflections	47300				152457	40971
Rfreeg	0.228				0.153	0.217
# reflections	2501				8054	2142
# protein atoms	3006				3028	3081
# FAD atoms	53				53	53
#ligand atoms	9				9	9
#other atomsh	0				1	16
# water molecules	264				431	266
B-factors
Protein (Å2)	28.1				20.2	27.6
Ligands (Å2)	18.0				14.4	26.7
Water (Å2)	38.8				34.8	35.5
All (Å2)	28.8				21.9	28.2
RMSD
Bonds (Å)	0.010				0.019	0.010
B-factor bonds (Å2)	2.9				3.6	2.2
Δ B main-main (Å2)	2.1				2.1	1.7
Δ B side-side (Å2)	3.7				5.0	2.6
Δ B main-side (Å2)	2.9				3.5	2.0
Ramachandran plot
Most favored	90.7%				92.3%	92.2%
Additional allowed	9.0%				7.4%	7.5%
Generously allowed	0.3%				0.3%	0.3%

^aValues in parentheses refer to the highest resolution shell.

^bR_merge (I)= Σ|I_i - <I>| /ΣI_i, where I_i is the intensity of the i^th observation, <I> is the mean intensity of the reflection, and the summation extends over all data.

^cI/σ(I) is the average signal to noise ratio for merged reflection intensities.

^dMAD phasing statistics for f“

^eMAD phasing statistics for f’

^fR-factor= Σ||F_o|-|F_c||/Σ|F_o|, where |F_o| is observed structure factor amplitudes and |F_c| is calculated structure factor amplitudes and the summation extends over all data.

^gR_free is the R-factor obtained for a test set of reflections, consisting of a randomly selected 5% subset of the diffraction data, not used during refinement.

^hThe closed form of nikD contains one sodium ion and the open form contains two molecules of MPD

Wild-type nikD was obtained in two crystalline forms: closed and open. The closed form is isomorphous with the SeMet-substituted enzyme, and diffracted to a maximum resolution of 1.15 Å. Two of the four C-terminal residues absent from the SeMet nikD model were visible in the closed form model. Also, the ligand molecule tentatively identified as picolinate was observed (Fig. 2A); in addition, a sodium ion not present in the isomorphous SeMet form of the enzyme was identified. The structure the wild type closed form of nikD is virtually the same as that of SeMet nikD except for a short polypeptide segment from residues Phe260 to Phe270 (see below). The final model of wild-type closed form nikD consists of residues 3-387, one FAD, 508 water molecules, one molecule of picolinate and one sodium ion.

Figure 2.

Stereo views of the electron density of closed and open forms of nikD. The electron density surrounding the flavin ring of FAD, its covalent attachment to Cys321 and the endogenous ligand modeled as picolinate is shown. The atoms are colored with nitrogen blue, oxygen red, sulfur yellow and carbon green. (A) Refined |2F_o|-|F_c| electron density at 1.15 Å resolution of the closed form of nikD, contoured at 1 σ (cyan) and 3 σ (red). (B) Refined |2F_o|-|F_c| electron density at 1.90 Å resolution of the open form of nikD, contoured at 1 σ (cyan).

The wild-type open form of nikD was obtained from slightly different conditions (Table 1) and was non-isomorphous with the closed form, which necessitated solution by molecular replacement. The structure analysis revealed that the catalytic domain was rotated by about 11° relative to the flavin binding domain. In addition, electron density for a ligand molecule close to the flavin ring was observed, but in a different orientation from that found in the closed forms of nikD (Fig. 2B; see below). The final model of the wild-type open form of nikD consists of residues 2-389, the first five residues of the C-terminal 6-His tag, FAD, 266 water molecules, two molecules of MPD and the small molecule built as picolinate.

All three structures exhibited excellent geometry with over 90% of the amino acids in the most favored region of the Ramachandran plot, none in the disallowed region (Table 1).

Structure of the wild-type closed form of nikD

NikD is folded into two domains (Figs. 3A, S2, S3). One domain binds the FAD cofactor while the other binds the substrate. The FAD-binding domain consists of residues Glu3 near the N-terminus to Ile87, residues Gly148 to Glu225, and residues Cys321 to Ser387 near the C-terminus (Fig.S3). The catalytic domain consists of residues Gly88 to Gly147 and Met226 to Thr320. The flavin-binding domain has a classic FAD-binding motif (Dym and Eisenberg, 2001). The catalytic domain forms an eight-stranded β-sheet flanked on either side by a β-hairpin and short α-helix and a pair of α-helices (Fig. S3).

Figure 3.

Closed form of nikD. (A) Stereo ribbon drawing with the flavin binding domain colored magenta and the catalytic domain colored cyan. The FAD and picolinate ligand are also included as stick figures with the same coloring scheme as in Fig. 2. (B) Closeup stereo view of the active site in the closed form of nikD. The putative picolinate ligand sits above the flavin ring in a pocket defined by the aromatic side chains of Trp355, Phe242, Phe244, Tyr258, Phe260. Above picolinate and the aromatic cluster are the side chains of Asn100 and Asp276. Oxygen and nitrogen atoms are colored red and blue, respectively; for FAD and picolinate, carbon atoms are colored green and for the side chains they are colored yellow. Hydrogen bonds are represented by dashed lines. (C) Close-up stereo view of the active site of MSOX containing the inhibitor pyrrole-2-carboxilic acid (PCA) (pdb code 1eli), oriented in a manner similar to that of the closed form of nikD, as shown in Fig. 3B. Sulfur is yellow and all other atoms are colored using the same scheme as in Fig. 3B. Side chains that surround PCA and correspond approximately in position to the aromatic side chain cluster Trp355, Phe242, Phe244, Tyr258, Phe260 found in nikD are included. Hydrogen bonds are represented as black dashed lines.

In the closed form of wild-type nikD, the adenosine pyrophosphate and ribityl portions of FAD extend into the interior of the flavin binding domain and upward toward the interface between the flavin binding and catalytic domains. The flavin ring is located at the interface with its si face in van der Waals contact with the side chain of Arg50. The N(1) atom of the flavin ring forms a hydrogen bond with the main chain nitrogen of Phe357, O(2) with the main chain oxygen of Trp355, the main chain nitrogen of Phe357 and the main and side chain nitrogens of Lys358; atoms N(3) and O(4) form hydrogen bonds with the main chain oxygen and nitrogen, respectively, of His51.

The ligand, modeled as picolinic acid, is located just above the re face of the flavin ring in the closed form of nikD (Fig. 3B). The plane of the ligand is approximately parallel (~20°) to the plane of the flavin ring of FAD and is separated from it by about 3.4 Å. The carboxylate oxygens of the ligand form three hydrogen bonds with nearby side chain atoms, one oxygen with Arg53 NE (2.78 Å) and Glu101 OE2 (2.65 Å) and the other oxygen with Lys358 NZ (2.65 Å). The side chains of Arg53 and Glu101 also form a hydrogen bond between each other (Arg53 NH2 to Glu101 OE2, 3.12 Å) so that the two basic side chains and the carboxylate groups of Glu101 and the ligand form a stable charge-neutral cluster in the protein interior. In addition to these three side chains and the flavin ring, the ligand is surrounded by a cluster of five aromatic side chains, Phe242, Phe244, Tyr258, Phe260 and Trp355 (Fig. 3B). Of these, the closest to the ligand (~3.4 Å) is Trp355 which is nearly perpendicular (~85°) to the ligand plane; it also is tilted by about 65° to the flavin ring and approaches within about 3 Å of it. Above the picolinate and aromatic cluster are the side chains of Asn100 and Asp276 that are tightly hydrogen bonded together; Asn100 also forms a hydrogen bond (3.03 Å) to atom NE1 of Trp355. Other aromatic interactions near the active site include the imidazole group of His51 that stacks on the aromatic ring of Phe242 and the side chains of Tyr258 and Phe260 which also are in van der Waals contact with the flavin ring.

The orientation of the pyridine ring of the putative picolinate ligand is arbitrary since the N(1) and C(3) atoms are crystallographically indistinguishable from each other. However, in the orientation shown in Fig. 3B, the ring nitrogen atom is in approximate van der Waals contact with the plane of Trp355 and potentially would be able to interact in its zwitterionic form with the pi electron system of the indole ring.

Structure of the wild-type open form of nikD

The open form of nikD consists of a two-domain structure as does the closed form. However, when the structures of the two forms are compared (Fig.4A), the catalytic domain of the open form appears to have undergone a rigid-body rotation of about 11° with respect to the flavin binding domain. As a result the orientation of the picolinate ligand now differs from that found in the closed form, where it is parallel to the flavin ring, to one nearly perpendicular to the flavin ring (~80°) (Fig. 5A). At the same time, the orientation of the side chain of Trp355 has changed to one that is nearly parallel to the flavin ring (~15°) and in van der Waals contact with it (Fig. 5A). The parallel stacking of the flavin and indole rings promotes partial π-π charge transfer from Trp355 to FAD, accompanied by the appearance of a charge transfer absorption band in the long wavelength region (Bruckner, R.C. et al., 2007). One of the carboxylate oxygens of picolinate has maintained its hydrogen bonding interactions with the side chains of Glu101 (2.67 Å) and Arg53 (3.11 Å) and forms an additional hydrogen bond with a water molecule (Fig. 5A); however, the salt bridge from Lys358 to the other carboxylate oxygen is now mediated by a bridging water molecule and a new hydrogen bond is formed between the carboxylate oxygen and NE1 of Trp355 (3.12 Å). Although still crystallographically indistinguishable from atom C(3), the ring nitrogen of picolinate, N(1) as now modeled, is within hydrogen bonding distance of O(4) of the flavin ring (3.13 Å). The salt bridge between Arg53 and Glu101 found in the closed form is no longer maintained.

Figure 4.

Comparison of the open and closed forms of nikD. (A) Cα trace of the open form of nikD (blue) superimposed on that of the closed form (magenta). Included are stick figures of FAD and picolinate for the closed form only, colored as in Fig. 2. Carbon, nitrogen, oxygen and phosphorus atoms are colored green, blue, red and gold, respectively. (B) Surface representation of the closed form of nikD. There is a small channel visible that leads into the active site. (C) Open form surface representation of nikD. The indole ring of Trp355 (shown as a stick figure) is exposed to the solvent.

Figure 5.

Active sites of open and closed forms of nikD. (A) Close-up view of the active site in the open form of nikD. The chemical groups shown and the atom coloring scheme used are the same as in Fig. 3B, except for the addition of two water molecules (red spheres). Hydrogen bonds are represented by dashed lines. (B) Superimposition of the active sites in the open and closed forms of nikD. Carbon atoms are colored green and yellow in the open and closed forms, respectively. In each form, the same color is used for oxygen (red) or nitrogen (blue).

The flavin interaction with the open form of nikD is virtually the same as in the closed form except for the absence of the Lys358 hydrogen bond to flavin O(2) and the presence of the additional hydrogen bond from the picolinate ring nitrogen to flavin O(4). The flavin ring has now changed position by about 0.6 Å, approximately in the direction of the long axis of the ring (Fig. 5B); the picolinate ring has moved by about 2.5 Å. Four of the five aromatic groups that surround the picolinate ligand in the closed form, Phe242, Phe244, Tyr250 and Phe260 still do so in the open form, differing in position by an average of about 2 Å. Phe260 maintains it stacking interaction with His51, but it no longer makes van der Waals contact with the flavin O(4) atom.

Comparison of the closed and open forms of nikD indicate that the flavin-binding domain, except for the segment from Trp52 to Leu64, behaves as a rigid group, exhibiting an rms deviation of about 0.45 Å for the 237 matched Cα atoms between the open and closed forms. Likewise, the catalytic domain, except for the segment from Pro281 to Asp293 behaves as a rigid group, displaying an rmsd of about 0.37 Å for 128 matched residues. The 281-293 segment, a loop located between strand β9’ and helix α3’ of the catalytic domain extends from the main body of the catalytic domain to a nearby portion of the flavin binding domain, helix α4, β12 and β13 where a few hydrogen bonds and hydrophobic interactions cause it to adhere to the flavin binding domain during the closed to open transition.

Segment 52-64 in domain 1 does not remain rigid when comparing the closed and open forms, but undergoes local deformation. The first five residues (52-56) move with the catalytic domain, adhering to it mainly through hydrophobic interactions, while the remainder of the segment (57-64) gradually relaxes back to the closed conformation of the flavin-binding domain.

On going from the closed to open forms of nikD, a large cavity opens in the molecule that exposes the ligand binding pocket to the environment (Figs. 4B, 4C). Surface area calculations for the open and closed forms of nikD show that when the catalytic domain rotates away from the FAD-binding domain, the amount of new surface exposed to solvent is about 600 Ų. This rotation exposes portions of residues Thr57, Asn100, Met226, Glu278, Leu325, and Trp355. These residues line the opening through which the substrate can enter and exit the active site. Some additional electron density was discovered in this opening and was modeled as two molecules of 2-methyl-2,4-pentanediol, the precipitant used for crystallizing the open form of nikD.

NikD Structural Homologs

Comparison of nikD (closed form) with other protein structures deposited in the Protein Data Bank using DALI (Holm and Sander, 1993) reveals six with high structural similarity as indicated by a DALI Z-score ranging from 46.9 to 24.7 (Table 2). The next closest matches have Z-scores of 16.2 or below. All six of these closely matched homologues are members of a FAD-containing amino acid oxidase family, with the closest being MSOX. When comparing the closed form of nikD to MSOX using LSQMAN (Kleywegt, 1999), the overall rmsd of C-α positions is 1.38 Å over 319 equivalent residues (using the default maximum separation of 3.5 Å). However, the homology of their flavin binding domains is stronger than for their catalytic domains. Within the former, 173 equivalent residues match with 1.13 Å rmsd of C-α atoms of which 48 are identical (28%). Comparison of their catalytic domain show 156 equivalent residues matching with an rmsd for C-α atoms of 1.62 Å, of which 31 are identical (20%).

Table 2.

Structure comparisons of nikD with other proteins using Dali (Holm and Sander, 1993).

Pdb code	Dali Z scorea	Rmsd (Å)	No. equivalent residues	% Identity	Enzyme
1el5-A	47.0	1.8	365	22	monomeric sarcosine oxidase
1y56-B	37.7	2.8	355	18	L-proline dehydrogenase
2gag-B	37.2	2.6	349	20	heterotetrameric sarcosine oxidase
1ng3-A	35.5	2.9	344	17	glycine oxidase
1pj5-A	34.4	3.0	352	15	N,N-dimethylglycine oxidase
1an9-A	24.5	3.2	304	14	D-amino acid oxidase

^aThe DALI Z-score is a measure of the strength of structural similarity in standard deviations above expected with only values above 2 reported.

The interactions of FAD with MSOX within the flavin-binding domains of nikD are reasonably well conserved. The flavin ring in MSOX is covalently bound through its C(8) methyl atom to Cys315 (Cys321 in nikD). Significantly, an arginine side chain (Arg49 in MSOX, Arg50 in nikD) makes van der Waals contact with the si face of the flavin ring in both proteins. The interactions of flavin atoms N(1), O(2), N(3) and O(4) with main chain atom of Phe347, His345, Lys348 and Ile50 in MSOX are maintained in nikD (to Phe357, Trp355, Lys358 and His51, respectively); the side chain interaction of flavin O(2) with the lysine is maintained in both as well.

The structures of several complexes of MSOX with bound ligands are known (Wagner et al., 2000). The ligand most similar to picolinate is pyrrole-2-carboxylic acid (PCA) in which a 5-membered heterocyclic aromatic ring replaces the 6-membered heterocycle of picolinate (Fig. 3C). In the MSOX-PCA complex the carboxylate of PCA is hydrogen bonded to the side chains of Arg52 and Lys348 which are homologous to Arg53 and Lys358 in nikD. However, in MSOX there is no homologous residue for Glu101 of nikD that makes an additional hydrogen bond to the ligand carboxylate; Glu101 in nikD is part of a hairpin loop, Asp94-Gly102, that is absent in MSOX and blocks access to the substrate binding pocket in the closed form. Four of the five aromatic residues that surround the picolinate ring, Phe242, Phe244, Tyr258 and Phe260 are only partially conserved in MSOX, being replaced by Ile 50, Met245, Tyr254 and Phe256. The fifth aromatic residue in nikD in contact with picolinate is Trp355. Its position in MSOX is occupied by the main chain atoms of Gly344-His345. The carbonyl oxygen of Gly344 plays an important role in MSOX since it serves as a hydrogen bond acceptor for binding PCA (Fig. 3C). Above the PCA and the group of largely hydrophobic side chains is His269 which corresponds to Asp276 of nikD.

The open form of nikD differs in structure from MSOX to a considerably greater extent than does the closed form, showing an rmsd of 1.66 Å for 279 equivalent Cα atoms. The structural difference in this case arises almost entirely from the domain movement of open form nikD relative to the closed form of the enzyme. In the open form of nikD, the hydrogen bond from flavin O(2) to the conserved lysine (Lys358 in nikD, Lys348 in MSOX) is lost and a new hydrogen bond from flavin O(4) to the picolinate ring nitrogen is formed. These two differences arise from the marked difference in ligand binding between the open form of nikD on one hand and MSOX and the closed form of nikD on the other. When a carboxylate-containing ligand binds to MSOX, the plane containing the carboxylate group and most of the other atoms is parallel to the plane of the flavin ring (Trickey et al., 1999; Wagner et al., 2000). as it is in the closed form of nikD. However, in the open form of nikD, the picolinate ligand is approximately perpendicular to the flavin ring, as described above, leading to the observed alterations in the hydrogen bonding interactions of the flavin ring. The unusual conformation of the ligand in the open form of nikD suggests that the complex is not a redox-active form.

Sodium binding to wild type nikD

With the exception of nikD, all members of the amino acid oxidizing flavoenzyme family of known structure (Table 2) contain a hydrogen bond donor to the N(5) atom of the flavin ring. In DAAO (Miura et al., 1997), TSOX (Chen et al., 2006), glycine oxidase (Settembre et al., 2003) and L-proline dehydrogenase (Tsuge et al., 2005), the hydrogen bond donor to the flavin N(5) is a backbone nitrogen atom adjacent to the si face of the flavin ring. In DMGO (Leys et al., 2003), the hydrogen bond donor is the hydroxyl group of Tyr259 which is located adjacent to the re face of the flavin ring. In the case of MSOX (Trickey et al., 1999), the donor is a water molecule which forms part of a chain of waters that extends to the protein surface in the ligand free enzyme. These hydrogen bond donors increase the electron deficiency at flavin N(5) and are generally expected to increase the oxidative power of the coenzyme (Fraaije and Mattevi, 2000; Ghisla and Thorpe, 2004).

Examination of the environment of the N(5) atom in both the closed and open forms of wild-type nikD indicates that the N(5) atom is surrounded by one or two aromatic side chains, the aromatic portion of the bound ligand, and the aliphatic part of the side chain of Arg50. Thus there is no chemical group available that can form a hydrogen bond to the N(5) atom of the flavin ring in nikD. However, in SeMet nikD (closed form), the conformation of the region from Phe260-Phe270 is different from that of the wild-type, closed form nikD structure. In crystals of wild-type closed form nikD, grown at [Na⁺]~100 mM, a sodium ion is coordinated to the backbone carbonyl groups of Ala266 and Gly268 as well as to four water molecules (Fig. 6A). Also, Asn263 is part of an asparagine turn conformation (Dempsey et al., 2000) stabilized by a bifurcated hydrogen bond between the side chain oxygen and the backbone nitrogen atoms of Trp265 and Ala266. In addition, His262 forms a salt bridge with the side chain of Glu269. In crystals of SeMet nikD (Fig. 6B), grown at [Na⁺]~5 mM, there is no bound sodium atom, the asparagine turn is not present, and His262 no longer forms the salt bridge but now occupies part of the sodium coordination site and forms hydrogen bonds to the backbone carbonyl groups of Ala266, Gly268 and Phe270. This rearrangement of His262 pulls the strand containing Phe260 away from FAD, resulting in an opening of the space between FAD and Phe260 large enough to accommodate a water molecule. With wild-type nikD, the conformation of the 260-270 loop in the open form differs from that of the closed form and the crystal does not contain bound sodium despite sodium being present at the same level (~100 mM) in the crystallization medium. These differences are attributed to the movement of domain 2 which destroys the sodium binding site in the open form.

Figure 6.

Configuration of the 259-271 segment in the sodium-containing wild-type and sodium-free SeMet closed forms of nikD. (A) Wild-type closed nikD with a sodium ion (violet) coordinated by carbonyl oxygen atoms of Ala266 and Gly268 plus 4 water molecules, and His262 forming a salt bridge with Glu269. Nitrogen, oxygen and carbon atoms are colored blue, red and yellow, respectively. Sodium coordination and hydrogen bonds are shown as black dashed lines. The non-bonding distance from Phe260 to the flavin N(5) atom is shown as a red dashed line. (B) SeMet nikD without bound sodium in which the His262 side chain replaces the sodium ion and forms hydrogen bonds to the carbonyl oxygens of Ala266, Gly268 and Phe270. The atom coloring is the same as in part A.

Binding of potassium causes a change in the conformation of the 260-270 region of closed form nikD similar to that observed for the sodium complex (unpublished results). The possible catalytic significance of the monovalent cation-induced conformational changes in nikD is currently under investigation. It is worth noting that the mobile cation-binding loop found in nikD has no equivalent in the other members of the flavoprotein amino acid oxidase family.

Substrate Access

The amino acid oxidase family of flavoproteins has a variety of strategies for allowing access to the enzyme active site. In the case of L-proline dehydrogenase, the active site is open to the environment. In TSOX (Chen et al., 2006) and DMGO (Leys et al., 2003), there is a tunnel between the FAD active site and a second active site that binds folic acid. Structures of DAAO from both fungal (Umhau et al., 2000) and mammalian (Miura et al., 1997) origin have been reported. In the porcine DAAO, where product release is rate-limiting, the active site is covered by a polypeptide loop that forms a lid; this lid is not present in the enzyme from the fungus Rhodotorula gracilis. In MSOX, substrate access appears to be controlled by a loop from Tyr55 to Tyr61. This loop contains Glu57 that forms a salt bridge with Arg52, which in turn binds the carboxylate of the substrate and product of the enzymatic reaction. This loop has been observed to be in two different conformations in which the Arg52-Glu57 salt bridge moves and access to the active site is controlled. Glycine oxidase from B. subtilis has a similar strategy to MSOX, but is slightly more complicated.

In nikD, the equivalent residue to Arg52 in MSOX is Arg53, which also binds to the carboxylate group of the substrate and product. There is no loop in nikD that is equivalent to the Tyr55-Tyr61 loop in MSOX. The loop containing Glu101 in nikD (absent in MSOX) makes extensive contacts with the β strands β7’-β8’ of the catalytic domain. This β structure is significantly longer in nikD than it is in the other members of the amino acid oxidase family. These two strands contain the cluster of aromatic residues that surround the picolinate ligand. It appears that maintenance of the salt bridge between Arg53 and Glu101, as well as the van der Waals contacts between the ligand and the aromatic cluster, cause the entire catalytic domain to rotate upon the change in conformation of Trp355. This domain rotation in turn appears to control access to the enzyme active site.

Catalytic Implications

NikD exhibits two unique features as compared with other members of the amino acid oxidase family of flavoproteins: 1) NikD catalyzes an overall 4-electron oxidation process, unlike the 2-electron oxidation reactions observed for other family members; 2) NikD exhibits two distinct modes for substrate binding, as judged by the observed structures for the open and closed forms of the nikD complex with picolinate, the product of the physiological catalytic reaction. The flavoenzymes choline oxidase, thiamine oxidase and glycolate oxidase catalyze an overall 4-electron oxidation of alcohol substrates to the corresponding carboxylic acids (Ghanem et al., 2003; Gomez-Moreno and Edmondson, 1985; Schuman and Massey, 1972). These reactions all proceed via two successive 2-electron steps, as observed with nikD. However, the alcohol oxidase reactions occur at the same carbon atom whereas the nikD reaction appears to require oxidation of two spatially discrete bonds in P2C (see Fig. 1).

Substrate access to the active site of nikD is possible only in the open conformation but substrate oxidation requires the closed conformation where the substrate is positioned just above the re-face of the flavin ring. This suggests that substrate binding involves formation of an initial redox-inactive “open complex”, followed by a conformational change to generate the redox-active “closed complex”, as we have previously proposed (Bruckner, R.C. et al., 2007).

The substrate atom being dehydrogenated in flavoenzyme-catalyzed reactions is typically found within a well-defined region just in front of flavin, as judged by the distance to flavin N(5) (≤ 3.8 Å) and the angle defined by the substrate atom with flavin atoms N(10) and N(5) (96 to 117°) (Fraaije and Mattevi, 2000). There are six possible paths for the initial oxidation of P2C to DHP, depending in part on whether the enzyme oxidizes the imine or enamine tautomer of P2C. Inspection of the “closed complex” of nikD with picolinate in the orientation shown in Fig. 3B, shows that the atom C(6) in the ligand ring is located within the region described above (Fig. 7). Using the picolinate complex as an approximate model for the enzyme•substrate complex, the results suggest that the bond between N(1) and C(6) in the enamine tautomer of P2C is the site of the initial 2-electron oxidation reaction. Interestingly, this is the only possible path that involves oxidation of a carbon-nitrogen bond, consistent with the observed homology of nikD with other amine-oxidizing enzymes. This path is also consistent with the fact that nikD oxidizes a carbon-nitrogen bond in 3,4-dehydro-L-proline at an apparent rate that is only ~4-fold slower than observed for P2C oxidation (Venci et al., 2002).

Figure 7.

The reaction scheme shows the mechanism postulated for conversion of P2C to picolinate, as discussed in the text. The inset in the bottom right-hand corner shows the positions of FAD and picolinate, in the same orientation as shown in Fig. 3B. Distances (Å) between N(5) of FAD and the indicated atoms in picolinate are shown in magenta. Angles (°) defined by N(5) and N(10) in FAD and the indicated atoms in picolinate are shown in cyan.

The second 2-electron oxidation step is more problematic because the enzyme•DHP complex, generated as described above, would not contain any atom in a position favorable for substrate dehydrogenation. This apparent obstacle may be overcome by a process involving conversion of the initially formed protonated DHP imine to the corresponding neutral enamine, a facile ionization likely to exhibit a pK_a around 8, as judged by the value determined with P2C (Bruckner and Jorns, unpublished results). A 180 degree “flip” of the neutral enamine would place C(4) of the intermediate in a position favorable for dehydrogenation in a reaction that would be functionally equivalent to the direct oxidation the of the N(1)-C(6) bond in P2C (Fig. 7).

The postulated re-orientation of the DHP intermediate is unlikely to occur within the “closed complex” where movement is restricted by hydrogen bonding to the ligand carboxylate and the “cage” of aromatic residues that surround the ligand ring. Instead, we postulate that movement of DHP is achieved by transient formation of an “open complex”, followed by re-formation of a “closed complex” with the ligand in the desired “flipped” orientation. This scenario provides a functional rationale for the two unique features observed with nikD: the presence of two substrate binding modes and the enzyme’s ability to catalyze an aromatization reaction that requires an overall 4-electron oxidation of its physiological substrate.

Concluding remarks

The structures of a closed form and an open form of nikD have been determined at 1.15 and 1.90 Å resolution, respectively; the two forms differ by an 11° rotation of the catalytic domain with respect to the FAD-binding domain. In the closed form the active site is inaccessible to solvent; an endogenous ligand, believed to be the natural, 4-electron oxidation product, picolinate, is bound close to and parallel with the flavin ring. The active site is accessible to solvent in the open form. However, the picolinate ligand and a nearby tryptophan side chain have reoriented in the open form so that the picolinate is approximately perpendicular to the plane of the flavin ring and the tryptophan in now stacked on the flavin ring, shielding it from solvent. The ligand binding mode in the closed, but not the open, form is compatible with redox catalysis and resembles that observed with ligand-bound forms of MSOX. A cluster of aromatic side chains surround the ligand binding site in both forms of nikD. No chemical groups are located close by that could act as an active site base for proton abstraction during substrate oxidation. The lack of a catalytic base and the observed orientation of the picolinate ligand close to the N(5) atom of the flavin ring suggest that the initial 2-electron oxidation of the physiological substrate occurs by hydride transfer from the carbon adjacent to the ring nitrogen in the enamine form of P2C. The second 2-electron oxidation step, in which DHP is oxidized to picolinate, could occur by a similar mechanism after tautomerization from an imine to an enamine form of the DHP intermediate and flipping the DHP ring by 180° for optimal placement of an equivalent tetrahedral CH bond near the flavin N(5) atom. This flip could most easily be achieved during the transient formation of the open form of the enzyme, suggesting that domain rotation not only regulates active site accessibility but also plays a critical role within the catalytic cycle.

Experimental Procedures

Preparation of nikD and SeMet nikD

Recombinant nikD (unsubtituted) was isolated from cells (E. coli BL21(DE3)/pDV101) grown in LB or Terrific Broth and purified in a single step by using a Talon metal affinity resin (Clontech) (Bruckner, R.C. et al., 2007; Venci et al., 2002). SeMet-substituted enzyme was prepared by growing the same strain in the presence of SeMet under conditions that suppress the endogenous biosynthesis pathway for methionine (Doublie, 1997; Nunn et al., 2002). The growth medium was similar to that described by Nunn et al. (Nunn et al., 2002) except that the antibiotic was changed to carbenicillin (0.1 mg/mL), the medium was supplemented with riboflavin (100 mg/L) and contained a higher concentration of SeMet (48 mg/L). SeMet-substituted nikD was purified as indicated above except that the affinity column eluate was re-chromatographed on a Ni-NTA (Qiagen) column.

Crystallization and data collection

NikD crystals were obtained by the sitting drop vapor diffusion method (Mcpherson, 1999) after the protein was first exchanged with 10 mM TRIS buffer, pH 7.6 and concentrated to 10 or 20 mg/mL using a Centricon 30 membrane filter. Crystals of SeMet-substituted nikD were obtained by mixing 4 mL protein (at 20 mg/mL) with 4 mL 28-30% PEG-1500 in 50 mM HEPES buffer, pH 6.6 followed by seeding with microcrystals obtained from an initial screen. Two forms of wild type nikD crystals were obtained. One form, subsequently identified as “closed”, was obtained by mixing 4 mL wild type nikD (at 20 mg/mL) with 4 mL 28-30% PEG-1500 in 100 mM ADA buffer adjusted to pH 6.5 by addition of sodium hydroxide. The second form, subsequently identified as “open”, was obtained by mixing 4 mL nikD (at 10 mg/mL) with 4 mL 20-30% 2-methyl-2,4-pentanediol (MPD) in 100 mM MES buffer (pH 5.7). In each case the free acid form of the buffer was titrated to the final pH with sodium hydroxide which produced final sodium ion concentrations of ~5 mM for the SeMet crystal and ~100 mM for both wild type crystal forms. All crystals were bright yellow and formed rods about 400×60×40 microns in size for the SeMet substituted enzyme, prisms about 300×200×150 microns in size for the closed form and plates about 200×150×20 microns in size for the open form. The crystals are monoclinic, space group P2₁, with the SeMet and closed form crystals isomorphous. The crystallization conditions and cell dimensions are summarized in Table 1.

Multiwavelength anomalous diffraction (MAD) data were collected from a SeMet nikD, crystal that was transferred briefly to Paratone N oil (Hampton Research, Laguna Niguel, CA) using a nylon loop and then to a cryostream at 110 K after removal of excess mother liquor from the crystal surface. Data were recorded on a MAR-345 CCD at Beamline 14-ID-B at BIOCARS, Argonne, IL, at four different wavelengths, the peak and inflection points of the selenium edge and two remote wavelengths above and below the peak wavelength. Cryoprotection and flash-freezing of a wild-type closed form nikD crystal were accomplished in the same manner as for the SeMet crystal and the data were collected using an ADSC Quantum-4 CCD at Beamline 14-BM-C at BIOCARS, Argonne, IL. Two passes were made at the same wavelength; the first pass was 200° with steps of 1° processed to 1.5 Å resolution, and the second pass was 200° with steps of 0.5° processed to 1.15 Å resolution. Diffraction data were collected from a crystal of wild-type open form nikD that was transferred directly from a crystallization drop to the cryostream at 110 K. Data were collected on an ADSC Quantum-4 CCD at Beamline 14-BM-C at BIOCARS, Argonne, IL. A single pass of 180° with steps of 0.25° was sufficient to collect a complete dataset to 1.9 Å resolution. Scaling and merging of all data sets, including the two passes from closed form wild type nikD was carried out using HKL2000 (Otwinowski and Minor, 1997). The X-ray wavelengths used and the data collection and scaling statistics are shown in Table 1.

Phasing and refinement

Five selenium atoms were located by Patterson search methods from the anomalous differences at the peak wavelength using SHELX (Sheldrick and Schneider, 1997). Using those selenium positions, phases were calculated to 2.5 Å resolution using the program XHEAVY in the XtalView package (Mcree, 1999). DM (Cowtan and Main, 1998) of the CCP4 package (Ccp4, 1994) was then used for solvent flattening and phase extension to 1.75 Å. Phasing statistics are shown in Table 1. The model was then built in the XTALVIEW package using the maps generated from solvent flattening.

Refinement of the SeMet nikD model against the data collected at the low remote energy (λ=1.0000 Å, Table 1) was performed using CNS (Brünger et al., 1998). Simulated annealing was performed on the model, with a starting R_work of 43.1% and R_free of 44.3%. Subsequent positional and temperature factor refinement revealed all but the two N-terminal residues, the four C-terminal residues, and the C-terminal His-tag.

For the closed form of wild-type nikD, rigid-body refinement was performed on the polypeptide chain and FAD from the model of SeMet nikD, which is isomorphous (Table 1). The starting R-factor prior to rigid body refinement was 31.5%. Alternating cycles of refinement using REFMAC of the CCP4 package and model building in XTALVIEW was then done until R_work and R_free dropped to 17.7% and 18.9% respectively. Refinement of anisotropic temperature factors was performed using SHELX against all data to 1.15 Å resolution. After anisotropic refinement and addition of hydrogen atoms to the model, R_work dropped to 12.7% and R_free to 15.3%.

The structure of the open form of wild-type nikD was solved by molecular replacement using the program CNS. The refined coordinates of SeMet nikD were used as a search model. Prior to the translation search, Patterson correlation refinement (Grosse-Kunstleve and Adams, 2001) was performed to optimize the orientations of each of the two domains found in nikD. The starting R-factor prior to positional refinement was 45.2%. Wild-type open form nikD refinement was done by a round of simulated annealing in CNS, followed by alternating rounds of atomic refinement in CNS and model building in XTALVIEW. Refinement statistics can be found in Table 1.

Structural calculations and drawings

Structural comparisons were carried out using LSQMAN of the Uppsala Software Factory (Kleywegt, 1999)and DALI (Holm and Sander, 1993). Surface area and geometrical calculations were made using AREAIMOL and GEOMCALC of the CCP4 package (Ccp4, 1994). Structural diagrams were rendered using PYMOL (available on the World Wide Web at www.pymol.org).

Data Bank accession numbers

The atomic coordinates for the crystal structures of the closed and open forms of the native protein and of the SeMet substituted protein can be accessed through the RCSB Protein Data Bank under accession numbers 2H2V, 2H2L and 2Q6U, respectively.