Resonance Raman Spectroscopic Study of the Interaction Between Co(II)rrinoids and the ATP:Corrinoid Adenosyltransferase PduO from Lactobacillus reuteri

Abstract

The human-type ATP:corrinoid adenosyltransferase PduO from Lactobacillus reuteri (LrPduO) catalyzes the adenosylation of Co(II)rrinoids to generate adenosylcobalamin (AdoCbl) or adenosylcobinamide (AdoCbi⁺). This process requires the formation of “supernucleophilic” Co(I)rrinoid intermediates in the enzyme active site that are properly positioned to abstract the adeonsyl moiety from co-substrate ATP. Previous magnetic circular dichroism (MCD) spectroscopic and X-ray crystallographic analyses revealed that LrPduO achieves the thermodynamically challenging reduction of Co(II)rrinoids by displacing the axial ligand with a non-coordinating phenylalanine residue to produce a four-coordinate species. However, relatively little is currently known about the interaction between the tetradentate equatorial ligand of Co(II)rrinoids (the corrin ring) and the enzyme active site. To address this issue, we have collected resonance Raman (rR) data of Co(II)rrinoids free in solution and bound to the LrPduO active site. The relevant resonance-enhanced vibrational features of the free Co(II)rrinoids are assigned on the basis of rR intensity calculations using density functional theory to establish a suitable framework for interpreting rR spectral changes that occur upon Co(II)rrinoid binding to the LrPduO/ATP complex in terms of structural perturbations of the corrin ring. To complement our rR data, we have also obtained MCD spectra of Co(II)rrinoids bound to LrPduO complexed with the ATP analogue UTP. Collectively, our results provide compelling evidence that in the LrPduO active site, the corrin ring of Co(II)rrinoids is firmly locked in place by several amino acid side chains so as to facilitate the dissociation of the axial ligand.

Introduction

Adenosylcobalamin (AdoCbl), also known as coenzyme B₁₂, is one of the biologically active corrinoid species.¹ These species are characterized by the presence of a naturally occurring tetrapyrrole macrocycle termed the corrin ring that chelates a low-spin Co(III) ion in a tetradentate manner. On the “upper” and “lower” faces of the corrin ring of AdoCbl, the Co(III) ion is axially coordinated by, respectively, an adenosyl (Ado) moiety and a 5,6-dimethylbenzimidazole (DMB) group that is tethered to the macrocycle through a nucleotide loop.² AdoCbl serves as a cofactor in numerous enzymes, such as ribonucleotide tri-phosphate reductase,^3–5 glutamate mutase,^6,7 diol dehydratase,^8–11 ethanolamine ammonia lyase,^12–14 and methylmalonyl-CoA mutase,^15,16 all of which dramatically (by up to 12 orders of magnitude) accelerate the rate of homolytic Co–C bond cleavage of AdoCbl.^6,15,17,18 The Ado^• radical that is generated in this process (along with cob(II)alamin [Co(II)Cbl], Fig. 1) initiates radical-mediated enzymatic catalysis by abstracting a hydrogen atom from substrate.^19–23

Figure 1.

Chemical structure of cob(II)alamin [Co(II)Cbl]. In cob(II)inamide [Co(II)Cbi⁺], the nucleotide loop with its DMB tail is absent and the Co(II) center is axially ligated by a water molecule. In AdoCbl, an Ado moiety is bound axially to the central Co(III) ion trans to the DMB group.

Lactobacillus reuteri is one of few prokaryotic species that can synthesize AdoCbl de novo, employing a PduO-type ATP:corrinoid adenosyltransferase (LrPduO)²⁴ to catalyze the formation of the Co–C(Ado) bond. LrPduO utilizes Co(II)rrinoids and ATP as substrates, transferring the Ado moiety from ATP to a transiently generated Co(I)rrinoid species.²⁵ The generation of this Co(I)rrinoid intermediate is a thermodynamically challenging task, given that the reduction potentials of free Co(II)rrinoids [e.g., E° (SHE) = −610 mV for Co(II)Cbl]²⁶ are considerably lower than those of putative in vivo reducing agents [e.g., E° (SHE) = −440 mV for the semiquinone/reduced flavin couple of flavodoxin A (FldA)].²⁷ Nevertheless, several in vitro studies of LrPduO and other ATP:corrinoid adenosyltransferases (ACATs) complexed with co-substrate ATP have revealed that these enzymes can accomplish the Co(II) → Co(I)rrinoid reduction when supplied with FldA or even free dihydroflavins [e.g., E° (SHE) = − 228 mV for FMN at pH 7.5].^28–30

Previously, the mechanisms by which the CobA-type ACAT from Salmonella enterica,^31,32 the PduO-type human ACAT,³³ LrPduO,³⁴ and the EutT-type ACAT from Salmonella enterica^35,36 can overcome the thermodynamically challenging Co(II) → Co(I)rrinoid reduction were investigated using magnetic circular dichroism (MCD) and electron paramagnetic resonance (EPR) spectroscopies. These studies led to the proposal that the ACAT/ATP complexes convert Co(II)rrinoids to effectively four-coordinate, square-planar species, thereby stabilizing the redox-active Co 3d_z²-based molecular orbital (MO) and raising the reduction potential into the physiologically accessible range.³¹ In support of this proposal, the X-ray crystal structure of LrPduO complexed with Co(II)Cbl and MgATP²⁵ revealed that the side chain of a phenylalanine residue (F112) displaces the DMB moiety from the Co(II) ion upon Co(II)Cbl binding to the active site. By providing this hydrophobic environment below the corrin ring, the active site of the LrPduO/ATP complex is capable of converting both Co(II)Cbl and cob(II)inamide [Co(II)Cbi⁺, a natural cobalamin precursor that lacks the nucleotide loop with its DMB tail and instead features an axially bound water molecule] to four-coordinate species.³⁴

These enzyme-bound, four-coordinate Co(II)rrinoid species exhibit distinct spectral signatures, most notably an intense, positively signed band in the near-IR spectral region of the MCD spectrum and a series of widely spread resonances in the low-field region of the EPR spectrum, which reflect the presence of a significantly stabilized Co 3d_z²-based MO. While the MCD features associated with the corrin π→π* transitions are also notably perturbed relative to those displayed by five-coordinate free Co(II)rrinoids, relatively little is currently known about the effects of corrin ring conformational changes on the electronic structure of Co(II)rrinoids. Yet, it is obvious that non-bonding (electrostatic/hydrophobic) interactions between the corrin ring and the enzyme active site must play an important role in the mechanism of axial ligand removal, as amino acid substitutions in LrPduO that disrupt the interactions between the peripheral amide groups of the corrin ring and the enzyme active site result in a significantly decreased yield for the five- to four-coordinate Co(II)rrinoid conversion and thus slow down the formation of AdoCbl.^37,38

Because corrinoids exhibit intense electronic absorption (Abs) features associated with corrin π→π* transitions in the UV/visible region, resonance Raman (rR) spectroscopy provides a particularly sensitive probe of geometric perturbations of the corrin ring and has been used to investigate interactions between AdoCbl and AdoCbl-dependent enzymes.^39–41 The strong coupling between the corrin π→π* transitions and the stretching motion of the conjugated C–C and C–N bonds in the corrin ring leads to the predominant enhancement of corrin-based vibrational modes upon laser excitation in the UV/visible region. However, assigning these corrin-based vibrational modes is far from straightforward due to the structural complexity and lack of symmetry of the corrin ring. Originally, the two most strongly enhanced features in the rR spectra of Co(III)rrinoids were assigned on the basis of their rR enhancement patterns.^42–46 The lower-energy feature at ~1500 cm⁻¹ was found to be predominantly enhanced upon laser excitation into the so-called α-band, which arises from the lowest-energy corrin π→π* transition that is polarized along the C5⋯C15 axis (the long axis, LA in Fig. 1), while the higher-energy feature at ~1550 cm⁻¹ was preferentially enhanced upon laser excitation into the so-called γ-band, which arises from a corrin π→π* transition polarized along the Co⋯C10 axis (the short axis, SA in Fig. 1). On the basis of these findings, the ~1500 and ~1550 cm⁻¹ features were assigned as vibrational modes that primarily entail stretching motion of the conjugated C–C and C–N bonds oriented along the LA (the LA-stretching mode) and the SA (the SA-stretching mode), respectively.^45,46 In support of these assignments, the ~1550 cm⁻¹ feature was shown to undergo a significant down-shift upon H/D exchange at C10, while the ~1500 cm⁻¹ feature was largely unaffected by this isotopic substitution.⁴⁵ However, on the basis of a density functional theory (DFT)-assisted normal mode analysis for methylcobalamin (MeCbl),⁴⁷ Kozlowski and coworkers suggested that the frequency of the SA-stretching mode is in fact lower than that of the LA-stretching mode. This finding raised questions regarding the validity of the normal mode assignments that were based on the rR enhancement patterns.

An equally puzzling finding in previous studies of corrinoids was that the rR enhancement patterns of the corrin-based vibrational modes vary significantly as a function of the cobalt oxidation state.⁴⁸ In the rR spectra of Co(II)rrinoids, two features associated with corrin-based vibrational modes are similarly enhanced upon laser excitation in resonance with the LA-polarized “α-band-type” corrin π→π* transition.⁴⁹ Notably, one of these enhanced features is considerably down-shifted upon H/D exchange at C10, indicating that it arises from a vibrational mode predominantly involving SA-stretching motion. Consistent with these observations, a recent DFT-based analysis of the relative rR intensities of the relevant corrin-based vibrational modes revealed that the relative coupling strengths of these modes to a given corrin π→π* transition strongly depend on the exact compositions of the donor and acceptor orbitals.⁴⁸ As a result, vibrations mainly involving stretching of conjugated C–C and C–N bonds oriented along one axis of the corrin ring may, in fact, couple to a perpendicularly electronic transition.

In the present study, we have used rR spectroscopy to investigate how the conformation of the corrin ring is perturbed upon Co(II)rrinoid binding to the active site of the LrPduO/ATP complex. To aid in the interpretation of our experimental data, we computed the frequencies and relative rR intensities for models of Co(II)Cbl, Co(II)Cbi⁺, and four-coordinate Co(II)rrinoids. Additionally, we obtained MCD spectra of Co(II)Cbl bound to LrPduO in the presence of the ATP analogue UTP, which allowed us to discriminate between different possible interpretations of our computational results. Collectively, the spectroscopic and computational data obtained in this study provide significant new insight into the nature of the interaction between four-coordinate Co(II)rrinoids and the LrPduO/ATP complex.

Materials and Methods

Sample preparation and spectroscopy

All chemicals including aquacobalamin (H₂OCbl⁺), dicyanocobinamide [(CN)₂Cbi], potassium formate (HCOOK), sodium borohydride (NaBH₄), deuterium chloride (DCl), and deuterium oxide (D₂O) were purchased from Sigma and used as obtained. Dihydroxocobinamide [(OH)₂Cbi] was prepared as described previously.⁴⁹ Briefly, an aqueous solution of (CN)₂Cbi was reduced to Co(I)Cbi with NaBH₄, while purging the reaction mixture with N₂(g) to remove the nascent HCN(g). The resulting Co(I)Cbi solution was then re-oxidized in air and loaded onto a C18 SepPack cartridge to quench the remaining NaBH₄ with distilled water, and the (OH)₂Cbi product was subsequently eluted with methanol. Solutions of free Co(II)Cbl and Co(II)Cbi⁺ were obtained anaerobically by first degassing aqueous solutions of H₂OCbl⁺ and (OH)₂Cbi, respectively, under vacuum and subsequently reducing with HCOOK. H/D isotopic labeling at C10 was accomplished by incubating H₂OCbl⁺ or (OH)₂Cbi in a DCl/D₂O solvent mixture.⁵⁰ The excess deuterium ions in the solutions of H₂OCbl⁺(d-10) and (OH)₂Cbi(d-10) were quenched with NaOH dissolved in D₂O before the reduction to Co(II)Cbl(d-10) and Co(II)Cbi⁺(d-10), respectively, was initiated with HCOOK.

The protocol for producing and purifying LrPduO is described in references (9) and (18). In short, a pTEV3 plasmid encoding the LrPduO protein with an N-terminal His₆ tag was overexpressed in Escherichia coli, and the tag-free enzyme was obtained using rTEV protease and a HisTrap FF column (Amersham Biosciences). Samples of enzyme-bound, four-coordinate Co(II)rrinoids were prepared by mixing solutions of free (i.e., five-coordinate) Co(II)Cbl or Co(II)Cbi⁺ and LrPduO, which was contained in pH 8 Tris-HCl buffer with at least an 8-fold molar excess of MgATP, in a mole ratio of 0.9:1.0. The Co(II)Cbl + LrPduO/UTP sample was prepared in the same way, except that excess MgUTP instead of MgATP was added to the buffer containing LrPduO.

All rR spectra were collected on samples contained in NMR or EPR tubes that were submersed in either an ice bath or liquid N₂ and upon excitation with a Coherent I-305 Ar⁺ ion laser with 10–50 mW laser power at the sample. The ~135° backscattered light was dispersed by an Acton Research triple monochromator equipped with 1200 and 2400 groves/mm gratings and analyzed with a Princeton Instruments Spec X:100BR deep depletion, back-thinned CCD camera. The HCOO⁻ peak observed at 1348 cm⁻¹ (with the ice peak aligned at 228 cm⁻¹) was used as an internal standard.^51,52 Low-temperature Abs and MCD spectra were collected on a Jasco J-715 spectropolarimeter in conjunction with an Oxford Instruments SM-4000 8T magnetocryostat. All MCD spectra presented in this work were obtained by taking the difference between spectra collected with the magnetic field oriented parallel and antiparallel to the light propagation axis to remove contributions from the natural CD and glass strain.

Computational models and frequency calculations

The initial geometries of all Co(II)rrinoid models were derived from the X-ray crystal structure of Co(II)Cbl⁵³ by replacing all peripheral side chains with H atoms at a distance of 1.1 Å from adjacent C atoms, except for the methyl groups on the C5 and C15 methine bridges of the corrin ring (see Fig. 1 for the atom numbering scheme used in this study). These latter methyl groups were preserved because some corrin-based vibrational modes of interest involve significant methine-bridge movement.⁵⁴ Also, to prevent excessive flattening of the corrin ring during the geometry optimization of the Co(II)Cbl model,⁴⁹ the methyl group on the B5 position of the DMB group was maintained, while that on B6 was replaced by a H atom. In the Co(II)Cbi⁺ model, the oxygen atom of the coordinated water molecule was placed at the position originally occupied by the ligating nitrogen atom of the DMB group in the Co(II)Cbl model, while for the four-coordinate Co(II)rrinoid model the axial ligand was removed altogether. The geometries of these truncated models were optimized by performing DFT energy minimizations using the ADF 2006.01 program. The optimized structures were then subjected to analytical frequency calculations to ensure that the geometry optimizations converged to true minima of the potential energy hypersurfaces.^55–60 In each case, the Vosko-Wilk-Nusair (VWN)⁶¹ local density approximation and the Perdew-Burke-Ernzerhof (PBE)⁶² gradient-corrected exchange and correlation functionals were employed, and the integration constant was set to 5.0. TZP (core double-ξ and valence triple-ξ with polarization) Slater-type orbital basis sets were used for all atoms and core orbitals were frozen through 1s (O, N, C) and 2p (Co). The ADF calculated eigenvector representations of the normal modes of vibration were visualized using the Jmol program.⁶³

Time-dependent DFT (TDDFT) and DFT excited-state calculations

TDDFT calculations were performed using the ORCA 2.6 program developed by Dr. Frank Neese.⁶⁴ The PBE exchange correlation functional⁶² was employed along with the RI approximation^65–71 for accelerating the calculations. The valence double-ξ polarized Gaussian-type orbital basis set in conjunction with DeMon-J Coulomb fitting basis set were used for all atoms except Co, for which Ahlrichs' valence triple-ξ with polarization function basis set was used.^72,73 At least 60 excited states were calculated within an orbital energy window of ± 3 hartrees using the Tamm-Dancoff approximation.⁷⁴ On the basis of these TDDFT calculations, the corrin π₇- and corrin π₈*-based MOs were identified as the donor and acceptor MOs, respectively, for the “α-band-type” transition. MOs were plotted with the gOpenMol program developed by Laaksonen using an isodensity value of 0.03 au.^75–77

The excited electronic configuration associated with the “α-band-type” transition was specified using the block keyword “Occupations” in subsequent ADF calculations aimed at determining the excited-state potential-energy curves along the relevant corrin-based normal modes for the three Co(II)rrinoid models described above. In these calculations, the models were systematically distorted along the modes of interest, and for each nuclear configuration a single-point DFT excited-state calculation was performed using the same functionals and basis sets as those employed for the geometry optimizations and frequency calculations.

Results and Analysis

Abs spectra of Co(II)rrinoids

Fig. 2 shows the low-temperature Abs spectra of Co(II)Cbl and Co(II)Cbi⁺ free in solution and bound to the LrPduO/ATP complex. In each case the dominant Abs band in the visible region is centered at ~470 nm (21000 cm⁻¹). This band was previously assigned to a transition with predominant corrin π₇ → corrin π₈* character on the basis of spectroscopic and computational results.^48,49 Because this same transition is responsible for the so-called α-band in the Abs spectra of Co(III)rrinoids, it is generally referred to as the “α-band-type” transition. Previous TDDFT studies of Co(II)rrinoids revealed that the corrin π₇-based donor MO contains a sizable contribution from the Co 3d_z² orbital, making this orbital weakly σ-antibonding with respect to the axial ligand–Co(II) bond.⁴⁹ Because the DMB moiety is more basic than a water molecule, this σ-antibonding interaction is weaker in Co(II)Cbi⁺ than in Co(II)Cbl, and the “α-band” of Co(II)Cbi⁺ is thus slightly blue-shifted from that of Co(II)Cbl (cf. dashed gray traces in the top and bottom panels of Fig. 2). A more pronounced blue-shift of the “α-band” is observed upon binding of Co(II)Cbl or Co(II)Cbi⁺ to the LrPduO/ATP complex (cf. dashed gray and solid black traces in Fig. 2), reflecting the large stabilization of the Co 3d_z² orbital and, thus, greatly diminished contribution from this orbital to the corrin π₇-based MO in response to the axial ligand dissociation from the Co(II) ion.

Figure 2.

Electronic absorption spectra at 4.5 K of free Co(II)Cbl (top, dashed gray), Co(II)Cbl bound to the LrPduO/ATP complex (top, solid black), free Co(II)Cbi⁺ (bottom, dashed gray), and Co(II)Cbi⁺ bound to the LrPduO/ATP complex (bottom, solid black). The laser excitation wavelength (465.8 nm) used for collecting the rR spectra presented in Figure 3 is indicated by an arrow.

rR spectra of Co(II)rrinoids

A series of rR spectra were collected for each Co(II)rrinoid species described above using numerous Ar⁺ ion laser lines between 514.5 and 457.9 nm. Consistent with a previous report,⁴⁹ the rR excitation profiles of the corrin-based vibrational modes were found to trace the “α-band” envelope in the corresponding Abs spectra (Supporting Information, Fig. S1). The rR spectra obtained with laser excitation at 465.8 nm, which is close in energy to the “α-band” peak position of four-coordinate Co(II)rrinoids, are presented in Fig. 3. The rR spectrum of free Co(II)Cbl is dominated by two similarly intense features at 1482 and 1587 cm⁻¹ (Fig. 3a, dashed gray trace). Our recent DFT-based analysis of the corrin-based vibrational modes of this species revealed that the normal mode associated with the 1482 cm⁻¹ feature (referred to as the symmetric ν_s(#1) mode) primarily involves SA-stretching motion along the N22–C9 and N23–C11 bonds coupled in-phase with methine-stretching motion along the C5–C6 and C15–C14 bonds. Alternatively, the mode responsible for the 1587 cm⁻¹ feature (the symmetric ν_s(#3) mode) mainly involves LA-stretching motion along the N21–C4 and N24–C16 bonds coupled out-of-phase with the same methine-stretching motion (vide infra).⁴⁸

Figure 3.

(a) rR spectra of free Co(II)Cbl (dashed gray, top), Co(II)Cbl + LrPduO/ATP (solid black, top), free Co(II)Cbi⁺ (dashed gray, bottom), and Co(II)Cbi⁺ + LrPduO/ATP (solid black, bottom). (b) rR spectra of Co(II)Cbl (solid black, top), Co(II)Cbl(d-10) (dashed gray, top), Co(II)Cbi⁺ (solid black, top) and Co(II)Cbi⁺(d-10) (dashed gray, bottom). All spectra were obtained with 465.8 nm laser excitation.

While all of the Co(II)rrinoid rR spectra presented in Fig. 3a exhibit two resonance-enhanced features at ~1490 and ~1590 cm⁻¹, the relative intensities of the ~1590 cm⁻¹ band of the four-coordinate Co(II)Cbl and Co(II)Cbi⁺ species bound to the LrPduO/ATP complex are greatly diminished (cf. solid black and dashed gray traces). Additionally, the frequency of the ν_s(#1) mode varies noticeably as a function of axial ligation. From Co(II)Cbl to Co(II)Cbi⁺ and the four-coordinate Co(II)Cbl/Co(II)Cbi⁺ species, the ν_s(#1) mode up-shifts by as much as 11 cm⁻¹ (Fig. 3a). This trend reflects the decrease in charge donation from the cobalt center to the empty corrin π* frontier orbitals upon replacement of the axial DMB ligand by a more weakly donating water molecule and the removal of the axial ligand, respectively. No such simple correlation exists between the frequency of the ν_s(#3) mode and the electron density on the cobalt center. Specifically, removal of the axial ligand by the LrPduO/ATP complex results in an up-shift of the ν_s(#3) mode of Co(II)Cbl by 5 cm⁻¹, but in a down-shift of this mode by 3 cm⁻¹ in the case of Co(II)Cbi⁺. As a result, the rR spectra of four-coordinate Co(II)Cbl and Co(II)Cbi⁺ are nearly superimposable. This observation is consistent with the fact that in the X-ray crystal structure of Co(II)Cbl bound to the LrPduO/ATP complex,⁹ the DMB group is excluded from the enzyme active site and no longer interacting with the cobalt center.

H/D exchange at C10 causes the ν_s (#1) mode to down-shift by ~4–6 cm⁻¹, while the ν_s(#3) mode is barely affected (Fig. 3b). The fact that a vibrational mode showing strong enhancement upon laser excitation into the “α-band” is sensitive to isotopic labeling at C10 implies that significant coupling occurs between corrin SA-stretching motion and an LA-polarized electronic transition. Although it had generally been assumed that the corrin-centered vibrational modes mainly involving SA-stretching motion should only couple to SA-polarized electronic transitions,^45,46 our recent DFT-based rR intensity calculations revealed that these modes can in fact also couple to LA-polarized electronic transitions.⁴⁸

Flanked by the rR features associated with the ν_s(#1) and ν_s(#3) modes, two additional features can be discerned whose intensities vary marginally as a function of laser excitation wavelength. The feature at ~1530 cm⁻¹ was previously attributed to a symmetric mode [the ν_s(#2) mode] involving LA-stretching motion coupled in-phase with methine-stretching motion and out-of-phase with SA-stretching motion. Despite its totally symmetric character, the ν_s(#2) mode is only weakly enhanced upon laser excitation across the entire UV/vis region, but down-shifts upon H/D exchange at C10 due to the involvement of SA-stretching motion (Table 2). The feature at ~1560 cm⁻¹ was assigned to an anti-symmetric vibrational mode (the ν_as mode) primarily involving LA-stretching motion, with out-of-phase stretching of the bonds related by the pseudo-mirror plane perpendicular to the corrin ring. Due to the significant involvement of C10 motion, the ν_as mode shows the largest shift upon H/D isotopic exchange at C10 (Table 2).

Table 2.

Experimental and DFT-computed frequencies (in cm⁻¹) of the relevant corrin-based vibrational modes depicted in Fig. 5. The isotopic shifts for H/D exchange at C10 are given in parentheses; their estimated error is ±1 cm⁻¹.

		νs (#1)	νs (#2)	ν as	νs (#3)
Free Co(II)Cbl	Exp.	1482 (− 6)	1529 (− 3)	1555 (− 9)	1587 (− 2)
	Calc.	1487.4 (− 4.9)	1530.0 (− 2.3)	1556.8 (− 5.7)	1584.9 (− 0.6)
Free Co(II)Cbi+	Exp.	1486 (− 4)	1533 (− 6)	1564 (− 6)	1594 (0)
	Calc.	1485.2 (− 4.7)	1528.1 (− 2.0)	1557.1 (− 5.5)	1584.2 (− 0.9)
Four-coordinate Co(II)rrinoids	Co(II)Cbl	1493	1533	1561	1592
	Co(II)Cbi+	1491	1535	1564	1591
	Calc.	1490.6 (− 5.0)	1528.1 (− 2.2)	1558.2 (− 5.1)	1584.6 (− 1.1)

DFT geometry optimizations and frequency calculations

To aid in the interpretation of the experimental data presented above, computational models of Co(II)Cbl, Co(II)Cbi⁺, and the four-coordinate Co(II)rrinoid species were generated via DFT energy minimization (Fig. 4) and subsequently subjected to analytical frequency calculations. As expected, the corrin ring in the optimized Co(II)Cbl model is considerably folded (see LA fold angle, θ, in Table 1), whereas that of the four-coordinate Co(II)rrinoid model is nearly planar due to the lack of steric interactions with an axial ligand. This prediction is consistent with the X-ray crystal structural changes observed from free Co(II)Cbl to Co(II)Cbl bound to the LrPduO/ATP complex, which revealed that θ decreases by more than a factor of 2 upon removal of the axial DMB ligand from five-coordinate Co(II)Cbl.^25,53

Figure 4.

Geometry-optimized computational models of (a) Co(II)Cbl, (b) Co(II)Cbi⁺, and (c) the four-coordinate Co(II)rrinoid species.

Table 1.

Corrin fold angles and axial ligand bond distances for various Co(II)rrinoid species predicted computationally and observed by X-ray crystallography

	Method	θ (°)a	ϕ (°)b	Co–Axial ligand
Co(II)Cbl, Fig. 4a	DFT	10.4	8.0	2.21 Å
Co(II)Cbi+, Fig. 4b	DFT	4.5	4.6	2.33 Å
Four-coordinate Co(II)rrinoid, Fig. 4c	DFT	1.8	1.0
free Co(II)Cblc	X-ray	16.2	3.6	1.93 Å
Co(II)Cbl + LrPduO/ATPd	X-ray	7.3	10.8

^aθ corresponds to the LA fold angle and is defined as the angle between the plane that is composed of N21, C4, C5, C6, and N22 and the vector connecting the Co center and C15.

^bϕ corresponds to the SA fold angle and is defined as the angle between the plane that is composed of N22, C9, C10, C11, and N23 and the vector connecting the Co center and the midpoint between C1 and C19.

^cFrom ref 33.

^dFrom ref 9.

Despite the variation in axial coordination, the DFT-computed eigenvector descriptions of the corrin-based normal modes are virtually identical for all three Co(II)rrinoid models (Fig. 5). Four corrin-based vibrational modes, including three totally symmetric vibrational modes in the parent C_2v point group, corresponding to ν_s(#1~3), and an anti-symmetric vibrational mode, ν_as, are predicted in the 1480–1600 cm⁻¹ range (Fig. 3a and Table 2). Experimentally, the frequency of the ν_s(#1) mode increases with increasing positive charge on the cobalt center due to the consequent decrease in charge donation from the cobalt center to the empty corrin π* frontier orbitals (vide supra). Consistent with this experimental trend, the computed frequency of the ν_s(#1) mode is larger for the four-coordinate Co(II)rrinoid model than for the five-coordinate models, though DFT predicts a higher frequency for the Co(II)Cbl model than for the Co(II)Cbi⁺ model. The DFT-predicted isotopic shifts for the ν_s(#1) mode upon H/D exchange at C10 are in excellent agreement with the experimentally observed shifts (Table 2).

Figure 5.

DFT-computed vector representations of the relevant corrin-centered normal modes of Co(II)Cbl (top), Co(II)Cbi⁺ (middle), and four-coordinate Co(II)rrinoid (bottom).

The computed frequencies for the ν_s(#2) and ν_as modes also agree well with our experimental values (Table 2). The ν_as mode mainly involves LA-stretching motion whereby atoms belonging to one half of the corrin ring move out-of-phase relative to their counterparts belonging to the other half. Consistent with the anti-symmetric nature of this mode, it carries substantial “off-resonance” intensity in the experimental rR spectra (Supporting Information, Fig. S1). The large contribution from C10 motion to the ν_as mode leads to a sizable H/D isotope shift computed for this mode, which agrees well with the experimental observation that the ν_as mode displays the largest shift upon H/D exchange at C10 (Table 2). Lastly, the computations predict nearly identical frequencies and negligible H/D isotope shifts for the ν_s(#3) mode of all three Co(II)rrinoid models, consistent with our experimental data. Overall, the good agreement between the experimentally observed and computationally predicted frequencies and isotope shifts for the relevant corrin-centered normal modes lends support to our DFT-based normal-mode analysis for the different Co(II)rrinoids.

TDDFT and DFT excited-state calculations

While the computed eigenvector representations of the three ν_s modes vary only slightly as a function of axial ligation, the experimental rR spectra of four-coordinate Co(II)rrinoids bound to the LrPduO/ATP complex show a significantly diminished intensity of the ν_s(#3) mode relative to that of the ν_s(#1) mode. To explore the origin of this puzzling result, TDDFT and single-point DFT excited-state calculations were carried out for all three Co(II)rrinoid models.

Because the methyl groups at C5, C15, and B5 are expected to affect the vibrational frequencies of the corrin-centered normal modes, they were included in our Co(II)rrinoid models.⁴⁸ However, because these groups contribute minimally to the corrin-based frontier MOs, the TDDFT-computed Abs spectra for our Co(II)Cbl and Co(II)Cbi⁺ models (Supporting Information, Figure S2) are nearly identical to those reported previously for models lacking the peripheral methyl groups.⁴⁹ The most intense band in the visible region of the computed Abs spectra for all three Co(II)rrinoid models considered in this study arises from the “α-band-type” transition. The donor MO involved in this transition, the corrin π₇-based MO, is symmetric with respect to reflection about the pseudo-mirror plane oriented along the SA, while the acceptor MO, the corrin π₈*-based MO, is anti-symmetric (Fig. 6, left). Therefore, the “α-band-type” transition is LA-polarized.

Figure 6.

Results from TDDFT and single-point DFT excited-state calculations used to estimate the Δ values for four-coordinate Co(II)rrinoids. (a) Experimental (red) and computational (solid black) Abs spectra of four-coordinate Co(II)rrinoids. The calculated LA- and SA-polarized contributions to the Abs envelope are shown by dashed black and gray lines, respectively. (b) Molecular orbitals involved in the LA-polarized “α-bandtype” (corrin π₇→π₈*) transition. (c) Computed potential-energy curves of the ground state and excited state associated with the corrin π₇→π₈* transition for the four-coordinate Co(II)rrinoid model along the ν_s(#1) mode (yellow triangles), the ν_s(#2) mode (green circles), and the ν_s(#3) mode (blue squares).

The experimental rR excitation profiles for both the ν_s(#1) and ν_s(#3) modes of Co(II)rrinoids roughly follow the Abs envelope in the region of the LA-polarized “α-band-type” transition (Supporting Information, Figure S1). To compute the relative rR intensities of the three ν_s modes for excitation in resonance with this transition, the displacement (Δ) of the corresponding excited state from the ground state along these modes was estimated by systematically distorting the computational models along each ν_s mode and performing DFT calculations for the distorted models in the excited-state associated with the corrin π₇ → π₈* transition. Since in ADF the eigenvectors of all normal modes are expressed in terms of normalized displacements of the atoms in non-mass-weighted Cartesian coordinates, the distorted models were generated by adding a variable fraction of the ν_s normal mode displacements to the optimized ground-state geometry in Cartesian coordinates. The excited-state potential-energy curves obtained using this approach are shown in Fig. 6.⁷⁸ Note that the Δ values (Table 3) are not dimensionless; rather, they correspond to the fraction of the normal mode eigenvectors added to the ground-state geometry in Cartesian coordinates.

Table 3.

Calculated displacements Δ_n of the excited state associated with the LA-polanzed “α-band-type” transition along the ν_s(#n) modes (n=1, 2, 3) and relative rR intensities of the ν_s(#n) modes obtained using the relationship I_n ∝ Δ_n²ν_s(#n)⁴, along with the corresponding experimental intensity ratios

Species	Δ 1	Δ 2	Δ 3	I1 : I2 : I3
				Calc.	Exp.
Co2+Cbl	0.023	0.014	− 0.012	100 : 41 : 35	86 : 16 : 100
Co2+Cbi+	0.031	0.015	0.001	100 : 26 : 0	100 : 24 : 60
Four-coordinate	0.030	0.013	− 0.020	100 : 21 : 57	100: 11 : 43 (Cbl)
Co2+corrinoid					100 : 18 : 39 (Cbi+)

In agreement with our previous vibrational analysis of the normal modes of free Co(II)Cbl,⁴⁸ the excited state associated with the “α-band-type” (corrin π₇ → π₈*) transition is displaced more substantially along the ν_s(#1) mode than along the other ν_s modes, regardless of the identity of the axial ligand (Table 3). This computationally predicted coupling between the ν_s(#1) mode, which primarily involves SA-stretching motion, and an LA-polarized electronic transition explains the puzzling observation that the rR band associated with the ν_s(#1) mode is enhanced upon laser excitation into “the α-band” while at the same time it down-shifts upon H/D exchange at C10.

The relative rR intensities of the other ν_s modes were estimated using the semi-classical theory of Shorygin^79,80 for the A-term intensity, $\mathrm{I}\propto {\Delta }^2{\stackrel{‒}{\nu }}^4$ (thus, we assumed that the ratio of Δ values is unaffected on going from dimensionless units to fractions of normal mode displacements in Cartesian coordinates, which is reasonable because the same atoms are involved in the vibrations of interest). While the intensity of the ν_s(#2) mode relative to that of the ν_s(#1) mode, I₂/I₁, is correctly predicted for the Co(II)Cbi⁺ and four-coordinate Co(II)rrinoid models, it is overestimated for Co(II)Cbl, primarily because of the small |Δ₁| value predicted for this model (Table 3). The I₃/I₁ intensity ratio is underestimated for all three Co(II)rrinoid models, but most significantly for Co(II)Cbi⁺, as for this model the ground and excited states are predicted to be minimally displaced along ν_s(#3) (i.e., |Δ₃| ≈ 0).

In the experimental rR spectra (Fig. 3), the intensities of the corrin-based modes of the four-coordinate Co(II)rrinoid species bound to the LrPduO/ATP complex are considerably different from those observed for their unbound, five-coordinate counterparts. In particular, the I₃/I₁ ratio decreases by a factor of ~2 upon binding of Co(II)Cbl or Co(II)Cbi⁺ to the enzyme active site. Yet, the DFT-predicted I₃/I₁ intensity ratio is substantially larger for the four-coordinate Co(II)rrinoid model than for both of the five-coordinate models. This discrepancy may be rationalized in two ways. First, it is conceivable that the computational approach employed in this study to calculate the relative rR intensities of the corrin-based normal modes is not sufficiently accurate to replicate the experimentally observed variations among the different Co(II)rrinoid species. This seems unlikely, though, considering that we successfully used the same approach to reproduce the key experimental trends regarding the frequencies and relative rR intensities of the corrin-based modes from CNCbl to Co(II)Cbl and Co(I)Cbl⁻.⁴⁸ Second, it is plausible that our four-coordinate Co(II)rrinoid model does not accurately mimic the actual conformation of the enzyme-bound Co(II)rrinoid species in the LrPduO/ATP complex. In this scenario, the diminished I₃/I₁ intensity ratio and, thus, |Δ₃|/|Δ₁| ratio reflect a conformational change of the coring ring imposed by hydrogen-bonding, electrostatic, and/or hydrophobic interactions with amino acid residues in the enzyme active site. To discriminate between these two possibilities, we used MCD spectroscopy as a complementary probe of the corrin ring conformation of Co(II)rrinoids bound to the active site of LrPduO.

Corrin ring perturbations in the active site of LrPduO

The MCD spectrum of Co(II)Cbl in the presence of LrPduO incubated with the ATP analogue uridine-5'-triphosphate (UTP) lacks the features characteristic of four-coordinate Co(II)Cbl species (Fig. 7a). However, in the region of the corrin π→π* transitions (~20,000 – 30,000 cm⁻¹), this spectrum is much more similar to that of the four-coordinate Co(II)Cbl species bound to the LrPduO/ATP complex than the MCD spectra of Co(II)Cbl free in solution and in the presence of LrPduO devoid of nucleotide [which is unable to bind Co(II)Cbl].¹⁸ These similarities suggest that the MCD spectral changes in the region of the corrin π→π* transitions observed upon Co(II)Cbl binding to the LrPduO/ATP complex primarily reflect perturbations to the corrin ring conformation induced by non-bonded interactions with active site residues, rather than the dissociation of the axial DMB ligand.

Figure 7.

(a) MCD spectra collected at 7 T and 4.5 K and (b) rR spectra obtained at 77 K upon 465.8 nm laser excitation of free Co(II)Cbl (solid gray), five-coordinate Co(II)Cbl in the presence of LrPduO/UTP (solid black), and four-coordinate Co(II)Cbl in the presence of LrPduO/ATP (dashed gray). The major spectral changes in the region of the corrin π→π* transitions that occur upon binding of Co(II)Cbl to LrPduO complexed with ATP or UTP are indicated by arrows.

To test this hypothesis, we also collected rR spectra for the sample of Co(II)Cbl in the presence of the LrPduO/UTP complex. As shown in Fig. 7b, the intensity of the ν_s(#3) mode relative to that of the ν_s(#1) mode is substantially lower in the rR spectrum of Co(II)Cbl + LrPduO/UTP than in the free Co(II)Cbl spectrum, indicating that a significant decrease in the |Δ₃|/|Δ₁| ratio occurs upon Co(II)Cbl binding to the LrPduO/UTP complex (though not quite as large as in the case of Co(II)Cbl + LrPduO/ATP). Yet, the frequency of the ν_s(#3) mode is virtually unaffected when Co(II)Cbl binds to LrPduO/UTP while the ν_s(#1) mode is up-shifted by a mere 2 cm⁻¹, providing further evidence that Co(II)Cbl retains the axial DMB ligand in the active site of the LrPduO/UTP complex. Thus, we conclude that the unusually small |Δ₃|/|Δ₁| ratio observed for the four-coordinate Co(II)Cbl species generated in the active site of the LrPduO/ATP complex stems primarily from perturbations imposed on the corrin ring, rather than the dissociation of the axial ligand.

Discussion

Corrin-centered normal modes of Co(II)rrinoids

Two symmetric corrin-centered vibrations, the ν_s(#1) and ν_s(#3) modes, are simultaneously enhanced in the rR spectra of Co(II)rrinoids obtained with laser excitation into the “α-band”. The ν_s(#1) mode down-shifts upon H/D exchange at C10, indicating that this mode primarily entails C–C/N stretching motions along the corrin SA (Fig. 5). This coupling between a vibrational mode mainly involving SA-stretching motion and an LA-polarized electronic transition is reproduced quite well by our DFT-assisted rR intensity calculations (Table 3). The rR enhancement patterns of the corrin-based modes of Co(II)rrinoids in the region of the “α-band” are strikingly different from those observed for Co(III)rrinoids and Co(I)rrinoids, even though the eigenvector representations of the ν_s modes and the compositions of the corrin π- and π*-based MOs involved in the “α-band-type” transition differ marginally among these species.⁴⁸ This variation in the degree of coupling between corrin ring vibrations and corrin π→π* transitions as a function of the Co oxidation state reflects the changes in the extent of mixing between the Co 3d valence orbitals and corrin frontier orbitals that occur from Co(III)- to Co(II)- and Co(I)rrinoids. With decreasing cobalt oxidation state, the filled Co 3d orbitals shift closer in energy to the empty corrin π* frontier orbitals, which leads to an increased mixing between these orbitals. Because in the lowest-energy corrin π*-based MOs the Co ion and N atoms of the corrin ring engage in π-antibonding interactions, the way by which the corrin ring distorts upon population of these MOs in electronic excited states will depend on the extent of Co 3d → corrin π* “backbonding”, which increases from Co(III)- to Co(II)- and Co(I)rrinoids. It is due to this delicate interplay between bonding and antibonding interactions between the Co 3d and corrin π/π* frontier orbitals that the direction of the corrin ring distortion in the excited state corresponding to the “α-band-type” transition varies as a function of the Co oxidation state.⁴⁸

Spectral probes of corrin ring distortions and axial ligation changes in Co(II)rrinoids

The frequency of the ν_s(#1) mode of corrinoids decreases with decreasing formal oxidation state of the cobalt center; thus, this mode is observed at lower frequencies for Co(II)rrinoids than for Co(III)rrinoids. In the case of Co(II)rrinoids, even relatively subtle changes in the electron density on the Co(II) center caused by axial ligand substitutions have a noticeable effect on the frequency of the ν_s(#1) mode [i.e., Co(II)Cbl < Co(II)Cbi⁺ < four-coordinate Co(II)Cbl and Co(II)Cbi⁺, Table 2]. Therefore, rR spectroscopy can be used as a probe of axial ligand–Co bonding interactions in Co(II)rrinoids to complement the information obtained from EPR and MCD studies.

No obvious correlation exists between the charge on the Co(II) ion and the frequency of the ν_s(#3) mode. The DFT computed frequencies of the ν_s(#3) mode for the three different Co(II)rrinoid models are nearly identical (Table 2), suggesting that the small differences observed experimentally from Co(II)Cbl to Co(II)Cbi⁺ and the four-coordinate Co(II)rrinoid species are not a direct consequence of the changes in axial ligation. Indeed, our rR spectra obtained for Co(II)Cbl bound to the LrPduO/ATP and LrPduO/UTP complexes reveal that the frequency of the ν_s(#3) mode is sensitive to changes in the corrin ring conformation; i.e., in both spectra the ν_s(#3) mode is observed at ~1590 cm⁻¹ despite the differences in the Co(II) coordination number (4 and 5, respectively). These spectra also show that the intensity of the ν_s(#3) mode relative to that of the ν_s(#1) mode decreases substantially upon Co(II)Cbl binding to the LrPduO active site, even when the axial DMB ligand is retained. The diminished relative rR intensity of the ν_s(#3) mode is accompanied by significant changes of the MCD features associated with the corrin π→π* transitions, in support of our hypothesis that the frequency and relative intensity of the ν_s(#3) mode are modulated by changes in the corrin ring conformation.

As expected, the frequency of the ν_s(#1) mode remains essentially unchanged when Co(II)Cbl binds to the LrPduO/UTP complex, because the axial DMB ligand is retained in this process (Fig. 7a). Likewise, minimal differences in the region dominated by LF transitions are observed between the MCD spectra of Co(II)Cbl free in solution and bound to the LrPduO/UTP complex (Fig. 7b), and the EPR spectra obtained for these two species are almost identical (Supplementary material, Fig. S3 and Table S1). These findings indicate that the relative energies of the Co 3d-based MOs in five-coordinate Co(II)Cbl are insignificantly affected by changes in the corrin conformation and confirm that the ν_s(#1) and ν_s(#3) modes provide selective probes of perturbations to the axial ligand–Co(II) bonding interaction and the corrin ring conformation, respectively.

Corrin ring conformation of Co(II)rrinoids in the active site of LrPduO

Our rR and MCD data provide evidence for a distorted coring ring conformation in the four-coordinate and five-coordinate Co(II)Cbl species bound to the LrPduO/ATP and LrPduO/UTP complexes, respectively (Fig. 7). Because of the largely diminished |Δ₃|/|Δ₁| ratios observed for these species, we conclude that the corrin ring is primarily distorted along the ν_s(#3) mode, which will reduce the displacement of the ground-state potential relative to the excited-state potential associated with the “α-band-type” transition along this mode. In support of this conclusion, inspection of the X-ray crystal structure of Co(II)Cbl bound to the active site of the LrPduO/MgATP complex²⁵ (Fig. 8) reveals that three of the four edges of the corrin ring are in close contact with active site residues, while the remaining edge that connects the A and D pyrrole rings (Fig. 1) faces the substrate access channel. Thus, the active site can modulate the conformation of the corrin ring along the LA, but not along the SA. Because the ν_s(#1) and ν_s(#3) modes primarily involve C–C/N stretching motions along the SA and LA of the corrin ring, respectively, only the latter is significantly affected by Co(II)rrinoid binding to the LrPduO active site.

Figure 8.

X-ray crystal structure of Co(II)Cbl bound to the active site of the LrPduO/MgATP complex (PDB code: 3CI1⁹).

Although the unique conformation of the corrin ring imposed by the enzyme active site may aid in the conversion of five- to four-coordination Co(II)rrinoids, probably the most important finding from this study with regards to the mechanism of substrate activation by LrPduO is that certain residues in the enzyme active site serve to lock the corrin ring in place, as evidenced by the unusually small |Δ₃|/|Δ₁| ratios and perturbed corrin π→π* transitions observed for Co(II)Cbl bound to the LrPduO/ATP and LrPduO/UTP complexes (Fig. 7). This tight binding of the corrin ring is expected to be critical for preventing the Co(II)rrinoid substrate from repositioning itself while retaining the axial ligand when the side chain of F112 moves into the position originally occupied by the DMB and H₂O ligands of Co(II)Cbl and Co(II)Cbi⁺, respectively. In support of this proposal, we have recently shown that alanine substitutions of active site residues located in the plane of the corrin ring drastically lower the relative yield of four-coordinate Co(II)rrinoid formation achieved by the LrPduO/ATP complex.³⁸