To investigate the single crystal spectroscopic properties of choline oxidase as a function of x-ray exposure, we harvested yellow crystals of oxidized enzyme from aerobic mother liquor. After transferring crystals to a cryoprotectant (3.5 M Na-Malonate), they were individually mounted in nylon loops and flash-cooled by plunging them into liquid N2
. We performed several types of experiments with enzyme crystallized in space groups P
2 or P
from ammonium sulfate. The exposure to 1 Å (12.398 KeV) x-ray photons (~4 × 1010
photon/s through a 200 μm diameter collimated beam) at 100 K was performed on stationary crystals or with crystals rotated through 180° at beamline X26-C of the National Synchrotron Light Source. The results are very reproducible and typical spectra are illustrated in . The absorption spectrum of choline oxidase crystals at 100 K prior to x-ray exposure reveals two maxima centered at 460 nm and 485 nm. We and others have observed that single crystal spectra at low temperature are anisotropic and depend critically on crystal orientation, especially when planar chromophores such as flavins are present within the crystal (23
). For example, AMO has collected single crystal spectra from several oxidized flavoenzymes (data not shown) including nitroalkane oxidase, xenobiotic reductase A, cholesterol oxidase, and thioredoxin reductase, which all yield two resolved peaks near 450 nm. In most of these flavoenzymes and in choline oxidase in particular, the two maxima at 460 nm and 485 nm move up and down along with baseline excursions, as well as change in their relative intensities to each other as a function of the rotation angle. Thus, the single crystal spectra of choline oxidase are better resolved than for the enzyme in solution at 300 K (for a typical example, see Figure S2
), which yields a peak at 450 nm and a shoulder at about 480 nm. After x-ray diffraction data collection, the optical spectrum of choline oxidase crystals yields a single absorption peak at approximately 400 nm. The difference spectra (after - before) clearly shows an absorption band with λmax
at 400 nm and features that extend from about 510 nm to longer wavelengths. The 400 nm feature represents the vast majority of the flavin species present in the region of the crystal exposed to x-rays and is remarkably similar to spectra obtained from transient flavin C4a-OO(H) or C4a-OH intermediates (3
). For example, these types of C4a adducts typically have a maximum absorbance at ~380 nm; however, in low temperature single crystals of choline oxidase the absorbance spectrum for the C4a adduct is red-shifted approximately 20 nm. The longer wavelength features at 510 nm indicate that a small fraction of the enzyme contains a FAD semiquinone species as previously observed for choline oxidase in solution (25
Figure 1 Spectroscopic changes observed in single crystals of choline oxidase upon x-ray irradiation at 100 K. (A) Optical spectra measured from a single crystal of choline oxidase before (blue) or after (red) an x-ray diffraction data set was collected with 180° (more ...)
We next evaluated the time-dependent process of adduct formation in a single crystal at 100 K. Spectra were collected every ten seconds from a stationary crystal of choline oxidase during exposure to the monochromatic synchrotron x-ray beam at X26-C. As illustrated in , the difference feature at 400 nm increases in an exponential process with a t1/2
of approximately 40 seconds. Thus, the appearance of the spectral feature at 400 nm is nearly complete within the time required to collect only a small fraction of the unique x-ray diffraction data (equivalent to less than 10° of crystal rotation), even at the relatively modest x-ray intensity of beamline X26-C of the NSLS.1
This process is also approximately commensurate with the decrease of the 460 and 485 nm features attributed to oxidized FAD (t1/2
of approximately 100 seconds, ). Optical spectra from enzyme in solution at room temperature show that the flavin semiquinone species (i.e.
one electron reduced) has a spectrum that overlaps that of oxidized FAD (25
). In contrast, the hydroquinone FAD species (i.e.
reduced by 2 electrons), or the C4a-adduct observed here, each have very little absorbance in the 450 - 500 nm region. Consequently, the exponential decay seen at 485 nm appears to be slower than the observed rate of formation of the 400 nm feature assigned to the C4a adduct. All taken together, the spectroscopic analyses of single crystals at low temperature show that the decrease in the concentration of oxidized FAD in the region of the crystal exposed to x-rays is correlated to an increase of the concentration of the C4a-adduct.
The atomic structure and 1.86 Å resolution simulated-annealing omit electron density maps for the FAD are shown in . The flavin isoalloxazine ring is not planar, as anticipated for oxidized choline oxidase, or bent along the N5-N10 axis as is often observed in reduced flavoproteins (26
). The electron density for the dimethylbenzene and piperazine rings indicates that each ring is flat and coplanar with one another. However, the plane of the pyrimidine ring is at an approximately 120° angle to the plane of the other two rings (Table S1
). The electron density for the pyrimidine ring indicates that it adopts a “half-boat” configuration in which the C4a atom is approximately 0.5 Å above the pyrimidine ring plane. Moreover, throughout the refinement process a greater than 4σ positive difference peak associated with the C4a atom persisted. This feature clearly indicates that the C4a atom is sp3
hybridized and, therefore, that a covalent flavin adduct is present in the crystal structure. The electron density for the difference feature is large enough to accommodate only one or possibly two atoms in a covalently linked FAD C4a-adduct.
To our knowledge, the materials and aerobic crystallization conditions do not include any reagents that can readily form a C4a adduct with oxidized FAD. Moreover, none of the reagents used to crystallize the enzyme alter the optical spectrum of oxidized choline oxidase indicating that the reagents do not perturb significantly the FAD electronic environment (see Figure S2
). In contrast, reduced flavins in the semiquinone or hydroquinone states do react with O2
and can form C4a-oxygen adducts (2
). Therefore, we followed a conservative approach and refined a FAD C4a-OH atomic model, which converged well with REFMAC5
to yield a 1.45 Å C4a-O bond distance ( and Table S1
). Next we modeled an O2
molecule bound to the C4a atom of reduced FAD. After refinement with REFMAC5
bond lengths are 1.4 Å ( and Table S1
) with a C4a-Op
bond angle of 116°. Although the estimated coordinate error of the model is approximately 0.1 Å, these parameters agree well with those determined by quantum mechanical calculations for the model FAD-C4a-OOH intermediate in p
-hydroxybenzoate 3-hydroxylase and phenol hydroxylase (27
), as well as the DFT calculations discussed below. Therefore, the C4a-OH and C4a-OO(H) models for the FAD adduct in choline oxidase have reasonable geometry for sp3
hybridization and the appropriate bond lengths and angles for a C4a-adduct. However, the observed electron density for the distal oxygen atom (Od
) is weaker than for the proximal atom oxygen (Op
) and, consequently, the former atom also has a higher B-factor. This suggests that either the Od
atom may be partially disordered, possibly due to precession about the C4a-Op
bond, or that only one oxygen atom is present in the adduct.
The simulated-annealing omit maps provide the most unbiased view of the FAD adduct. During this procedure, the FAD, DMSO, and C4a-adduct were omitted from the model. The resulting mFo
electron density maps () calculated with reflections between 50 and 1.86 Å resolution were superimposed with refined models containing either the C4a-OH or the C4a-OO(H) adduct. This analysis prompts us to conclude that the FAD in choline oxidase is most likely a C4a-OH (hydroxy) flavin adduct. However, as discussed above, we can not unambiguously rule out the C4a-OO(H) (peroxy or hydroperoxy) adduct. The resolution of the structure, quality of the refined atomic models, and the fits to the observed electron density for choline oxidase are comparable to those of the recently reported dioxygen complexes of cytochrome P450cam
), naphthalene dioxygenase (30
), superoxide reductase (31
), homoprotocatechuate 2,3-dioxygenase (32
), and amine oxidase (33
The x-ray crystal structure shows that the active sites are completely sequestered within each subunit of the dimeric enzyme such that there is no direct access of bulk solvent to the FAD isoalloxazine ring. Nevertheless, a DMSO molecule, an additive in the crystallization solution, is observed in the solvent excluded cavity within each active site (Figure S3
). In addition to the covalent bond between the dimethylbenzene ring and His99, several deduced hydrogen bonds stabilize the isoalloxazine ring configuration and the C4a-OH (or C4a-OO(H)) adduct. For example, the FAD pyrimidine ring forms a network of hydrogen bonds with protein backbone atoms from Asn100, Cys102, Ile103, and Asn512. In addition there are side chain interactions between the pyrimidine ring and Asn100 and Ser101. Consequently, nearly every atom of the isoalloxazine ring in this unusual configuration with the potential to participate in hydrogen bonds does so with either a protein residue2
or the DMSO. Furthermore, the structure indicates that the atoms of either C4a adduct are stabilized by hydrogen bonds with the side chains of His351 and Asn510. Finally, the DMSO molecule is located adjacent to the C4a adduct and also hydrogen bonds with the side chain of Ser101 and the FAD N5 moiety.
To gain a better understanding of the electronic structure of the unique isoalloxazine ring configuration observed in the x-ray structure of choline oxidase, we performed a number of DFT calculations. Each structure was optimized using Gaussian 03 program (17
) at the B3LYP/6-31G(d) level in the gas phase. The final energies were calculated using a large 6-31+G(d,p) basis set, which included dielectric effects (for ε = 4.3) from the surrounding protein using the self-consistent reaction field (IEFPCM) method at the B3LYP/6-31G(d) level. The largest calculations (ca. 180 atoms) included the FAD isoalloxazine ring and most of the first shell residues comprising the active site (see Figure S3
; His99, Asn100, Ser101, Cys102, Ile103, His310, His351, His466, Asn510, Pro511, Asn512, and a DMSO molecule). Several flavin electronic states within the enzyme active site were computed including: a)
one-electron reduced flavin semiquinone (FADsq
a reduced flavin C4a hydroperoxy complex (FAD-C4a-OOH) c)
a flavin C4a-hydroxy complex (FAD-C4a-OH), and d)
flavin C4a-peroxy complex (FAD-C4a-OO-
). The metrics for the observed x-ray structure and the DFT optimized structures for each flavin species are summarized in Table S1
The DFT calculations unequivocally indicate that formation of the FAD C4a adduct, and the first shell hydrogen bond interactions with the flavin moiety are both necessary to reproduce the observed x-ray structure (, S3
). Unfortunately, the accuracy of the DFT calculations and the uncertainty of the crystal structure do not support a clear discrimination between the C4a-OH or C4a-OO(H) structures. In contrast, gas-phase optimizations of proteinfree flavin adducts yield less distorted FAD-C4a-OOH and FAD-C4a-OH structures (Figure S4-D
). By comparison of these structures, we estimate that the energies required to stabilize the x-ray structure of the enzyme are ~29 kcal/mol for the FAD-C4a-OH adduct and 22 kcal/mol for the FAD-C4a-OOH adduct. Either of these total energetic contributions that stabilize the distorted flavin C4a adduct is consistent with the crystal structure and the numerous active site hydrogen bonds deduced between the distorted FAD and active site residues. In this context, our computational analysis suggests that the hydrogen bonding interactions of the flavin with the side chains His99, Asn100, Ser101, Cys102, Ile103, Asn510, His466 and His351 (Figure S3
) are essential for the stabilization of the distorted flavin C4a adduct. This is underscored by our additional computations (not shown) suggesting that removal of any of these residues in silico
yields a more planar flavin C4a-adduct.
We also considered alternative scenarios including degradation of the isoalloxazine ring system similar to that reported by Iwata et al.
), but in this case being initiated by reduction of the flavin in the x-ray beam. Indeed, due to the formation of C4a=O double bond some of the DFT optimized models resulted in ring-open products similar to the 10a-spirohydantoin characterized by Iwata. However, we ruled out this hypothesis based upon structural and spectroscopic evidence. First, our high resolution electron density maps are of sufficient quality to establish sp3
hybridization at the flavin C4a and C10a atoms, respectively. In contrast, the best characterized 10a-spirohydantoin degradation product has sp3
hybridization at atoms analogous to flavin C10a and C4a, respectively. Our single crystal spectroscopic analysis also supports the C4a adduct. Indeed, while the optical spectrum for the choline oxidase adduct has a λmax
near 400 nm, the optical spectrum for the degradation products referred to by Iwata have λmax
near 310 nm.
Solvated electrons are generated in biological samples by synchrotron x-ray irradiation on the time scale of electronic transitions (see for example (35
) and references therein). Rapid electron transfer over significant distances can be facilitated by aromatic side chains, protein backbone atoms, and protein secondary structure (36
). Consequently, it is very likely that the FAD in choline oxidase is reduced in the x-ray beam. We propose that the high reduction potential for the enzyme bound FAD provides a strong thermodynamic driving force to capture the solvated electrons (, path a
). This conclusion is consistent with the negative features observed in the difference spectra at 450 and 480 nm, as well as the 510 nm features ascribed to the FAD semiquinone as a minor, but persistent species. Our DFT calculations indicate that the unpaired electron in an anionic FAD semiquinone radical, that is stabilized in choline oxidase, is delocalized between the C4a (0.19e), C9 (0.18e), C7 (0.08e), C8 (-0.13e) and C8M (0.58e) atoms of the isoalloxazine ring system. Thus, the spectroscopic results show that the exposure of the crystal to the x-ray beam at low temperature is commensurate with a decrease in concentration of oxidized FAD in the sample.
Figure 3 The proposed reaction scheme for the C4a adduct formation in single crystals of choline oxidase at 100 K. Radiolysis of solvent by x-rays yields solvated electrons and several types of reactive oxygen species (path a). An electron migrates to the high (more ...)
Hydroxyl and superoxide radicals, as well as other reaction products, are also generated in biological samples by x-ray exposure, but on a time scale of microseconds to milliseconds (35
). However, the migration of hydroxyl and superoxide radicals, dioxygen, and protons within proteins at cryogenic temperatures is less well understood than electron transfer processes. Nevertheless, this raises the possibility that the C4a-adduct may form by one of several scenarios, which differ principally in the origin of the oxygen atom in the C4a-adduct (). For example, our DFT calculations show that a hydroxyl radical does react with the C4a position of a FAD semiquinone yielding a flavin C4a-OH species, but only if the two radicals are near each other (ca. 2.0 - 2.2
), and only if the flavin N5 is unprotonated (, path b
). Similarly, our calculations indicate that a reaction between superoxide radical and the FAD semiquinone yielding a flavin C4a-OO(H) species is also possible (, path c
). Alternatively, the C4a-OO(H) adduct may derive from a reaction between a two-electron reduced flavin (FADH-
) and a ground-state O2
(, path d
). Previous experimental and theoretical studies have demonstrated that a flavin radical can react with dimethylsulfide to yield a flavin C4a-OH adduct and DMSO (37
). Our crystallization conditions include 10% v/v DMSO, which likely has dimethylsulfide as a minor contaminant component. Consequently, a trace amount of dimethylsulfide could promote conversion of the semiquinone to yield the C4a-OH adduct, but in this case the resulting oxygen atom in the observed adduct originates from O2
rather than solvent. Each of the scenarios presented here requires that for a C4a-adduct to form, the appropriate reactants (with the exception of solvated electrons) must be either present or generated in the vicinity of the flavin isoalloxazine ring. Indeed, diffusion of bulky species over long-distances or through the protein and/or crystal matrix to the interior FAD C4a position is very unlikely at 100 K. In contrast, the proposed short-range migration of radical species or ground-state O2
within the choline oxidase active site is analogous to the well documented results of carbon monoxide photodissociation of myoglobin-CO complexes (39
Irrespective of the mechanism of formation of the flavin C4a-OH or C4a-OO(H) adduct, it seems likely that once formed neither breaks down to release water or H2
because the cryogenic conditions do not establish the correct proton inventory on the surrounding residues and the flavin to facilitate product dissociation.3
In contrast, during catalytic turnover with choline (11
), the enzyme coordinates the extraction of two protons and two electrons from the substrate with their subsequent delivery to O2
to yield H2
(see Figure S5
). Indeed, during the reductive half reaction, the net transfer and storage of two proton equivalents from the substrate to the active site base and the flavin N5 atom is commensurate with the two-electron reduction of the flavin. In the subsequent oxidative half reaction, the two electrons are transferred from the reduced flavin to O2
along with the delivery of the two stored protons. Therefore, an incorrect proton inventory on the active site residues may stall the oxidative half-reaction. Thus, the x-ray exposure at low temperature creates an improper proton inventory scenario that consequently yields the structure of choline oxidase containing the C4a-adduct. To date, we have not obtained evidence for a similar C4a adduct species in choline oxidase in experiments conducted in solution. However, very recently Sucharitakul et al
reported that pyranose 2-oxidase from Trametes multicolor
does form a transient C4a-hydroperoxyflavin intermediate in its oxidative half-reaction. Moreover, that enzyme exhibits spectroscopic characteristics for the C4a-adduct in solution that are very similar to the adduct we observe in single crystals of choline oxidase. In addition, the pyranose 2-oxidase from Trametes multicolor
) is structurally related to choline oxidase and other members of the Glucose-Methanol-Choline (GMC) oxidoreductase enzyme superfamily that catalyze the oxidation of alcohols.
Flavoenzyme mechanistic schemes often invoke FAD C4a-OO(H) or C4a-O(H) intermediates, but they have heretofore eluded structural characterization and have only rarely been detected by transient kinetic and spectroscopic techniques (1
). However, the structural environment in the active site of choline oxidase appears to be ideally suited to stabilize a C4a adduct involving an oxygen species. This study is therefore the first direct observation by x-ray crystallography of a hydroxy-flavin or peroxo-flavin intermediate in any flavoenzyme, despite decades of effort by many researchers in the field, and numerous mechanistic proposals that invoke such a species. Our multidisciplinary experimental approach and correlation with complementary theoretical analysis thus provides direct evidence for an important oxygen intermediate in the reaction cycle of flavin-dependent enzymes.