Spin Densities in Flavin Analogs within a Flavoprotein

Abstract

Characterization by electron paramagnetic resonance techniques of several variants of Anabaena flavodoxin, where the naturally occurring FMN cofactor is substituted by different analogs, makes it possible to improve the details of the spin distribution map in the isoallosazine ring in its semiquinone state. The analyzed variants were selected to monitor the effects of intrinsic changes in the flavin ring electronic structure, as well as perturbations in the apoflavodoxin-flavin interaction, on the spin populations. When these effects were analyzed together with the functional properties of the different flavodoxin variants, a relationship between spin population and biochemical parameters, as the reduction potential, could be envisaged.

Introduction

The flavin ring (7,8 dimethyl-isoalloxazine (see Fig. 1)) is a heterocyclic molecule that acts as the cofactor in a large family of proteins called flavoproteins. These proteins are ubiquitous in living systems and participate in a large number of processes including aerobic and anaerobic metabolisms, cell respiration, photosynthesis, denitrification, DNA photorepair, light sensing, light emission, synthesis of natural products, scavenging of reactive oxygen species, and apoptosis. In flavoproteins the isoalloxazine ring is able to stabilize three oxidoreduction states: fully oxidized (ox), partially reduced by one electron (semiquinone (sq)), and fully reduced by two electrons (hydroquinone (hq)). Although the sq state is poorly stabilized in free flavins under physiological conditions, some protein environments considerably contribute to its stabilization, conferring on flavoproteins the capacity of transferring either just one electron in each redox step or two electrons at once (1, 2, 3, 4, 5, 6).

Figure 1.

Molecular structure of the flavin isoalloxazine ring in the neutral sq state. The relevant atomic positions are labeled. For FMN (the cofactor of Fld), the residue, R, bound to C1′ consists of a ribityl chain followed by a phosphate group. One of the two protons bound to C1′ is labeled in the text as H^sC1′ and corresponds to the one with the strongest hyperfine couplings. The methyl substituents at positions 7 and 8, and the proton at position 3 have been modified in the different analogs studied in this work (see Materials and Methods).

The flavodoxin (Fld) from Anabaena is a flavin-mononucleotide (FMN)-dependent flavoprotein involved in the photosynthetic electron transfer (ET) from the donor side of photosystem I (PSI) to the flavoenzyme ferredoxin NADP⁺ reductase (FNR). To accomplish this, its FMN cofactor cycles between the Fld_sq and the Fld_hq states, although a physiological role for the Fld_ox/Fld_sq pair cannot be ruled out. The involvement of Fld in this ET system has been widely studied from both the structural and functional points of view (6, 7, 8, 9, 10, 11). Additionally, the stability and reversible denaturation properties of this Fld also make it a very good model for studying protein folding, as well as the influence of the apoprotein portion in determining the physicochemical properties of the flavin ring within the protein environment (10, 12).

The functionally relevant Fld_sq state contains an unpaired electron, making it paramagnetic and allowing the use of electron-paramagnetic-resonance (EPR)-related techniques in the characterization of spin distribution within the flavin ring. Previous studies have revealed details about the electronic structure of Fld and its implications for ET processes in which Fld is involved (13, 14, 15, 16). To elucidate the influence of the flavin isoalloxazine ring environment on its chemical function and reactivity, Fld mutants that substitute residues in close vicinity to the isoalloxazine ring (7, 8, 10, 14, 17, 18), as well as Fld variants that include modified flavin molecules (6, 11, 14), have been used. Characterization of Fld variants in which FMN analogs (where atoms or groups of this molecule are modified) substitute for FMN allows determination of the influence of different portions and substituents of the flavin ring on its reactivity and identifies key interactions between the flavin cofactor and the apoprotein (19). The study of such Fld variants by EPR techniques is particularly useful for extending this knowledge. In this study, we use continuous-wave (CW) EPR, hyperfine sublevel correlation spectroscopy (HYSCORE), and pulse electron nuclear double resonance (ENDOR) to further determine the influence of methyl groups at positions 7 and 8 of FMN (for atom numbering, see Fig. 1) on the properties of the flavin cofactor within the Fld environment. The structural and functional properties of these Fld variants have been previously described (6, 11), allowing direct correlation with the observations presented here. Additionally, an Fld variant including an FMN analog in which the proton at position N(3) of the isoalloxazine ring is replaced by a methyl group is studied to analyze the influence of introducing this bulky substituent on the properties of the reconstituted protein.

Materials and Methods

Sample preparation

Wild-type (WT) Anabaena Fld was over-expressed in Escherichia coli and purified as described elsewhere (11). Its apoprotein form, ApoFld, was prepared by treatment with 3% trichloroacetic acid at 4°C in the presence of dithiothreitol (13, 14). The precipitated ApoFld was separated from FMN by centrifugation and dissolved in 500 mM MOPS, pH 7.0, before dialysis against 50 mM MOPS, pH 7.0 (13, 14). 3-methyl-FMN (3-CH₃-FMN), 8-nor-Cl-FMN (8-Cl-FMN), 7,8-nor-7,8-Cl-FMN, (7,8-diCl-FMN), 7-nor-7-Cl,8-nor-FMN (7-Cl,8-H-FMN), and 7-nor-8-nor-8-Cl-FMN (7-H,8-Cl-FMN) were produced from the corresponding riboflavin analogs and used to reconstitute Fld as previously described (6, 11).

EPR measurements

Samples with a protein concentration of 400–800 μM in 50 mM MOPS, pH 7.0, were placed in 3 mm EPR tubes and anaerobically reduced under an argon atmosphere to the neutral Fld_sq state by stepwise dithionite addition. Samples were then frozen and stored in liquid nitrogen (at 77 K) until used in EPR measurements. Low temperature CW-EPR and HYSCORE measurements were performed in an ELEXSYS E-580 spectrometer from Bruker Biospin (Karlsruhe, Germany) operating in X-band (9–10 GHz) provided with a continuous-gas-flow CF935 cryostat from Oxford Instruments (Eynsham, United Kingdom). CW-EPR spectra were taken at 60 K. Microwave (μw) power and modulation amplitude were adjusted to get spectra free of saturation or distortion effects. Typical conditions were μw frequency, 9.60 GHz; μw power, 10⁻² mW; and modulation amplitude, 0.2 mT.

HYSCORE spectra were taken around 15 K, with the μw frequency at ∼9.77 GHz and the magnetic field set at ∼348 mT, corresponding to the center of the CW-EPR signal. The μw pulse sequence for the HYSCORE experiments was π/2−τ−π/2−t₁−π−t₂−π/2. Suitable phase cycling was applied to remove unwanted echoes (20). In each experiment, the τ value was fixed, and the t₁ and t₂ values were varied independently. τ values used were 96 ns, 128 ns, 168 ns, or 208 ns. Typical time steps were Δt₁ = Δt₂ = 16 ns (ν_Nyquist = 31.25 MHz). Shot repetition times were selected to be as long as 12 ms to avoid saturation. Once the time domain experimental data were acquired, the frequency domain spectrum was obtained as follows: polynomial base lines in both time dimensions were subtracted to eliminate zero-frequency contributions; a Hamming window function was applied to improve the signal/noise ratio; data were zero-filled until 256 × 256 points; Fourier transforms in the t₁ and t₂ domains were applied, and the modulus of this spectrum was calculated.

X-band pulse ENDOR experiments involving μw and radio frequency (rf) pulses were performed in an ELEXSYS E-680 spectrometer from Bruker Biospin (Karlsruhe, Germany). The resonator, also from Bruker, was an ER MD4-EN dielectric ring of ∼9.77 GHz resonance frequency equipped with an ENDOR coil. The temperature in the sample space was regulated using a continuous gas-flow ESR900 cryostat from Oxford Instruments. The experiments were performed using the Davies ENDOR experiment (20), with pulse sequence (π)_μw – T₁- (π)_rf – (π/2)_μw − τ − (π)_μw − τ – echo and μw pulse lengths t_π/2 = 100 ns and t_π = 200 ns. The interpulse delays were T₁ = 17 μs and τ = 208 ns. An rf pulse, π, of length 15 μs was applied during T₁ and its frequency was varied between 4 and 40 MHz. The entire pulse sequence was repeated with a frequency of 200 Hz at 80 K, at lower temperatures, the repetition frequency had to be decreased due to slower relaxation of the electron spin, and at 15 K, it was only 33 Hz.

Analysis of the ENDOR and HYSCORE spectra

The flavin ring in the sq state shows many ¹H and ¹⁴N nuclei displaying hyperfine interactions. Assignment of the CW-EPR, ENDOR, and HYSCORE signals might in principle be a difficult task. Fortunately, several previous studies have resolved the assignment in other flavoproteins on the basis of the flavin structure, selective isotopic labeling, and quantum chemical calculations. As the differences between our spectra and others previously reported are small, we can make use of those assignments. Details of these assignments can be seen in the Supporting Material.

To obtain accurate data on the hyperfine parameters, all the ENDOR and HYSCORE spectra have been simulated. The spectra simulations have been obtained by using the EasySpin package, version 5.0.9 (21). A comparison of the measured and simulated spectra can be found in the Supporting Material.

Structural and functional characterization of 3-CH₃-Fld

3-CH₃-Fld was functionally and structurally characterized according to the protocols previously reported for 8-Cl-Fld, 7,8-diCl-Fld, 7-Cl,8-H-Fld, and 7-H,8-Cl-Fld (6, 11). Midpoint reduction potentials for the ox/sq and sq/hq couples (E_ox/sq and E_sq/hq), as well as the overall midpoint reduction potential, E_m, were determined at 25°C in 50 mM MOPS, pH 7.0 (11). The dissociation constant (K_d) of the ApoFld:3-CH₃-FMN_ox complex was determined fluorometrically after quenching of the CH₃-FMN fluorescence upon titration with ApoFld according to procedures previously reported (6, 17). This K_d, together with the determined E_ox/sq value, was also used to determine free energies for the formation of the ApoFld:3-CH₃-FMN_ox and ApoFld:3-CH₃-FMN_sq complexes (as explained in the Supporting Material). The complex association constant (K_a) and the first-order rate constant for ET (k_et) from PSI_rd to 3-CH₃-Fld were determined by laser-flash absorption spectroscopy at 25°C (11). K_d, extinction coefficient changes (Δε), and free-energy changes (ΔG) for complex formation with WT FNR_ox were obtained at 25°C, by difference absorption spectroscopy (11). The apparent observed rate constants (k_ap) for the ET from WT FNR_hq to 3-CH₃-Fld were determined in 50 mM MOPS, pH 7.0, at 25°C under anaerobic conditions using a stopped-flow spectrophotometer (11). In silico 3-CH₃-Fld structural models and structural calculations, as well as their analysis and representation, were done using procedures and software previously validated with the WT and the other reconstituted Fld variants (6, 11).

Results

Effects of the substitutions on the spin distribution in the isoalloxazine ring

Systematic CW-EPR, HYSCORE, and pulse ENDOR measurements were taken for 8-Cl-Fld, 7-Cl,8-H-Fld, 7-H,8-Cl-Fld, 7,8-diCl-Fld, and 3-CH₃-Fld.

CW-EPR

The substitution of the different analogs for FMN causes changes in the overall shape of their CW-EPR spectra related to the naturally occurring Fld (Fig. 2). Replacement of CH₃ groups at positions C7 and/or C8 results in a better resolution of the shoulders related to the largest hyperfine couplings, as seen in Fig. 2 (positions Q1, P1, Q2, and P2). The distance between the outer shoulders, Q1 and Q2, displays information about couplings of N5, H5, and N10 (see the Supporting Material):

$$\Delta B_{\text{out}}^{\text{wt}}=\left[B\left(\text{Q}2\right)-B\left(\text{Q}1\right)\right]\approx C\left\{2\left[A_{\parallel }\left({}^{14}\text{N}5\right)+A_{\parallel }\left({}^{14}\text{N}10\right)\right]+A_2\left({}^1\text{H}5\right)\right\},$$ (1)

where $\Delta B_{\text{out}}^{\text{wt}}$ is the splitting between the two outer shoulders for WT Fld (distance between Q1 and Q2 field positions in Fig. 2); A_||(¹⁴N5), A_||(¹⁴N10), and A₂(¹H5) are the hyperfine constants for the corresponding nuclei in the direction perpendicular to the flavin ring plane; and C has a value of 3.57 × 10⁻² mT/MHz.

Figure 2.

X-band CW-EPR spectra of the WT and reconstituted Fld variants at 60 K. Dashed lines labeled Q1, Q2, P1, and P2 indicate the positions of the outermost shoulders, from which we obtained information about some hyperfine constants, recorded in Table 1 (see text).

In addition, the distance between the first and second shoulder (Q1 and P1, or, equivalently, P2 and Q2) is due to the smallest of these three couplings, namely, A₂(¹H5) (22, 23). Table 1 shows the detected CW-EPR splittings. Although the values obtained for A₂(¹H5) are not very accurate, they are fully coherent with those obtained from HYSCORE and pulse ENDOR experiments. A small reduction of the splitting between the outer shoulders (Q1 and Q2) in all the variants with chlorine substitution is detected, pointing to a slight reduction (5–10%) of the couplings of nuclei within the pyrazine ring.

Table 1.

Splittings in the CW-EPR Signals of WT and Reconstituted Fld Variants and the Corresponding Hyperfine Constants

	ΔBout = Q2 − Q1 (mT)	{2 × [A||(14N5) + A||(14N10)] + A2(1H5)} (MHz)	A||(14N5)a (MHz)	ΔBsmall1 = P1 − Q1 (mT)	ΔBsmall2 = Q2 − P2 (mT)	A2(1H5)b (MHz)
WT Fld	7.0 ± 0.4	196 ± 10	54 ± 5	unresolved	unresolved	–
3-CH3-Fld	6.9 ± 0.3	193 ± 8	52 ± 4	unresolved	unresolved	–
8-Cl-Fld	6.7 ± 0.2	188 ± 5	50 ± 3	0.9 ± 0.2	1.0 ± 0.2	26 ± 4
7-Cl,8-H-Fld	6.6 ± 0.2	185 ± 5	48 ± 3	0.9 ± 0.2	0.9 ± 0.2	25 ± 4
7-H,8-Cl-Fld	6.8 ± 0.2	190 ± 5	51 ± 3	1.0 ± 0.2	1.0 ± 0.2	28 ± 4
7,8-diCl-Fld	6.4 ± 0.2	179 ± 5	45 ± 3	1.0 ± 0.2	1.0 ± 0.2	28 ± 4

Positions Q1, Q2, P1, and P2 refer to Fig. 2.

^aValues estimated by considering the hyperfine constants obtained for ¹⁴N10 and ¹H5 from HYSCORE measurements (see below).

^bThese values can be directly compared to the ones obtained from HYSCORE experiments, collected in Table 2.

HYSCORE

An example of the HYSCORE spectra obtained for the Fld variants is shown in Fig. 3. HYSCORE measurements display features related to hyperfine couplings of several nitrogen and hydrogen nuclei, as indicated in the figure. A comparison of HYSCORE and ENDOR results for H6 and H^sC1′ couplings indicate that some of the hyperfine constants obtained from the HYSCORE features could be overestimated (see the Supporting Material). On the other hand, HYSCORE spectra display the best data for determining the hyperfine coupling constants of N1, N3, N10, and H5.

Figure 3.

HYSCORE spectra in X-band (9.77 GHz) obtained for 3-CH₃-Fld in the neutral sq state. The figure was obtained from the superposition of spectra taken with four different τ values: 96 ns, 128 ns, 168 ns, and 208 ns. The field position is B = 348 mT. Some level curves have been eliminated to make the picture clearer. Spectral features are assigned to specific hyperfine interactions on the basis of previously reported data (15) and simulations (see the Supporting Material). Spectra for the rest of the Fld variants are similar.

For HYSCORE features of nitrogen couplings, it is found that changes in the hyperfine or quadrupolar parameters, if any, are too small for the HYSCORE resolution, and the positions and shapes of these features for the Fld variants and WT Fld are the same. Features of N3 coupling for 3-CH₃-Fld are the exception. Substitution of the hydrogen at position N3 for a methyl group causes a detectable change in the nitrogen coupling, allowing the assignment of the spectral features to each of the individual nuclei N1 and N3 in Fld. This assignment, described in detail in the Supporting Material (15), had not been possible on the basis of any existing experimental data or calculation to date.

For the H5 features, small changes are detected, as shown in Table 2.

Table 2.

Hyperfine Parameters for the Positions where Differences between WT Fld and Its Variants Are Detected

	H5	H6	CH38	HsC1′
A1, A2, A3 (MHz)	A1, A2, A3 (MHz)	A||A, A⊥A (MHz)	A1, A2, A3 (MHz)
A||E, A⊥E (MHz)d
WT Flda	−1.1 ± 0.3b, −27.6 ± 0.2b, −39.5 ± 0.5c	−4.2 ± 0.5c, −5.5 ± 0.2c, −6.1 ± 0.3c	9.1 ± 0.4c, 8.0 ± 0.1c	+9.3 ± 0.5c, +11.1 ± 0.2c, +14.8 ± 0.6c
17.5 ± 1.0c, 15.6 ± 0.5c
3-CH3-Fld	+0.0 ± 0.3b, −26.2 ± 0.4b, −38.6 ± 0.8c	−4.1 ± 0.5c, −5.5 ± 0.2c, −6.2 ± 0.3c	9.0 ± 0.4c, 7.6 ± 0.2c	+8.2 ± 0.5c, +10.0 ± 0.5c, +13.6 ± 0.6c
16.4 ± 1.0c, 14.7 ± 0.5c
8-Cl-Fld	−1.4 ± 0.4b, −26.7 ± 0.4b, −38.6 ± 0.5c	−3.6 ± 0.5c, −6.2 ± 0.2c, −6.8 ± 0.3c	–	+8.8 ± 0.4c, +11.4 ± 0.3c, +14.7 ± 0.6c
7-Cl,8-H-Fld	−1.7 ± 0.3b, −26.7 ± 0.3b, −38.1 ± 0.6c	−3.8 ± 0.5c, −5.8 ± 0.2c, −6.4 ± 0.3c	–	+8.7 ± 0.5c, +10.4 ± 0.3c, +14.7 ± 0.6c
7-H,8-Cl-Fld	−1.7 ± 0.3b, −27.3 ± 0.2b, −39.1 ± 0.6c	−3.5 ± 0.5c, −6.1 ± 0.2c, −6.8 ± 0.4c	–	+8.7 ± 0.5c, +10.5 ± 0.3c, +14.4 ± 0.6c
7,8-diCl-FMN	−1.3 ± 0.8b, −26.2 ± 1.0b, −38.2 ± 0.6c	−3.5 ± 0.5c, −6.5 ± 0.2c, −7.0 ± 0.4c	–	+8.6 ± 0.5c, +11.1 ± 0.3c, +14.5 ± 0.6c

Data were obtained from pulsed ENDOR and HYSCORE experiments.

^aValues obtained from Martínez and colleagues (15, 16).

^bValues obtained from HYSCORE measurements.

^cValues obtained from ENDOR measurements.

^dSuperscripts A and E of the hyperfine constants refer to the two ENDOR signals associated with the CH₃8 dynamics (see Martínez et al. (16)).

Pulse ENDOR

Davies ENDOR spectra of the Fld variants are displayed in Fig. 4. From their analysis (16), proton hyperfine coupling parameters are obtained as shown in Table 2. Suitable manipulation of the spectra of the different Fld variants allowed us to obtain the complete set of hyperfine principal values for H6 and H^sC1′ (for details, see the Supporting Material). For the 7-H,8-Cl-Fld and 7-Cl,8-H-Fld variants, hyperfine constants for H7 and H8 couplings, respectively, are also determined (see the Supporting Material).

Figure 4.

Pulsed-ENDOR spectra of reconstituted Fld variants measured in the magnetic field position corresponding to the center of the CW-EPR flavin feature at 80 K. Each spectral feature has been labeled on the basis of previously reported ENDOR measurements on flavoproteins. Matrix protons are protons from surrounding water or protein residues, and weakly interacting flavin protons like H3, CH₃7, or H9; signals of CH₃8 protons only appear for WT Fld and variants where this methyl group is not substituted. For 7-Cl,8-H-Fld and 7-H,8-Cl-Fld, additional weak signals are seen (marked with an asterisk in the figure). These signals are assigned to the hyperfine interaction of the protons bound to C8 and C7, respectively (for details, see the Supporting Material). Analysis of these signals (see Materials and Methods and the Supporting Material) provides the hyperfine parameters in Table 2.

ENDOR measurements reveal several differences between WT and the Fld variants. The largest relative changes correspond to the H6 parameters for FMN analogs incorporating chlorine in position 7 or 8. In general, these samples show an increase of the absolute values of both the isotropic,

$$A_{\text{iso}}=\left(A_1+A_2+A_3\right)/3,$$ (2)

and the anisotropic,

$$T=\left[A_1-\left(A_2+A_3\right)/2\right]/3,$$ (3)

components of the H6 hyperfine tensor. Additionally, they show a reduction of the hyperfine constants for H5 and H^sC1′, consistent with the displacement of a moderate amount of spin density to the ring containing chlorine, a powerful electron-withdrawing substituent.

On the other hand, 3-CH₃-Fld displays no change in the H6 hyperfine constants, but relevant reductions of the absolute values of the isotropic hyperfine constants of H5 and CH₃8 are seen. Changes in H^sC1′ hyperfine parameters are also detected for this variant. The isotropic hyperfine constant is reduced ∼10% with respect to WT Fld, whereas anisotropy remains the same. On the other hand, HYSCORE measurements show very similar N10 hyperfine parameters for 3-CH₃-Fld and WT Fld. Although a small change in the spin populations cannot be ruled out, all this evidence points to the change in the H^sC1′ isotropic hyperfine constant being related to a reorientation of the C1′-H bond around the N10-C1′ axis (16, 24). It is known that this constant depends on the angle between the direction perpendicular to the flavin plane and the projection of the C1′-H bond over the plane perpendicular to the N10-C1′ axis. In the case of WT Fld, this angle is close to 60°. A slight reorientation of the C1′-H bond, keeping the same spin population at the N10 position, will explain the change in the H^sC1′ isotropic constant preserving the hyperfine anisotropy for N10.

At low temperatures (below 35 K), a thermal evolution of the ENDOR spectrum of 3-CH₃-Fld occurs. This behavior is related to the dynamics of the CH₃8 methyl group (16). Details of the corresponding analysis can be seen in the Supporting Material.

Structural and functional properties of 3-CH₃-Fld

Our results regarding the spin populations from EPR experiments on the different Fld variants can be connected with their structural and functional properties. For 8-Cl-Fld, 7,8-diCl-Fld, 7-Cl,8-H-Fld, and 7-H,8-Cl-Fld, these properties have been described elsewhere (6, 11). 3-CH₃-Fld is here characterized in the same manner.

Reduction potentials for 3-CH₃-Fld

The weak affinity of Fld for 3-CH₃-FMN hq prevented the determination of E_sq/hq and E_m. However, a value of 270 ± 5 mV for E_ox/sq was determined (Table 3). As for FMN, upon 3-CH₃-FMN binding to ApoFld, E_ox/sq was shifted to a value less negative than that of the free analog. However, although the free 3-CH₃-FMN E_ox/sq was expected to be the same as that for FMN (25), the shift in E_ox/sq upon ApoFld binding was considerably smaller than that for FMN or for the rest of the analogs here studied (Table 3), with the E_ox/sq for 3-CH₃-Fld being ∼75 mV more negative than that for WT Fld.

Table 3.

Midpoint Reduction Potentials for the Different FMN Derivatives Both Free and Bound to ApoFld

	Em (mV)			Eox/sq (mV)			Esq/hq (mV)
Free	HoloFld	Embound − Emfree	Free	HoloFld	Eox/sqbound − Eox/sqfree	Free	HoloFld	Esq/hqbound − Esq/hqfree
FMNa	−207	−312.5	−107	−313	−194	119	−101	−431	−331
8-Cl-FMNa	−152	−236	−84	−258	−134	124	−46	−337	−288
7,8-diCl-FMNa	−126	−156	−30	−232	−89	143	−20	−222	−202
7-H,8-Cl-FMNa	−144	−245	−101	−250	−128	122	−38	−362	−324
7-Cl,8-H-FMNa	−128	−200	−72	−234	−101	133	−22	−298	−276
3-CH3-FMN	−207b	–	–	−313	−270	43	−101	<−320	>−220

E_m data for the free flavins were at pH 7.0. The error in the determination of midpoint potentials is ±5 mV.

^aData were obtained from Frago et al. (11).

^bConsidered to be the same as FMN, since the introduction of a methyl group at position 3 in several flavins, including the isoalloxazine, hardly alters E_m,pH7 (25).

Stability of the ApoFld:3-CH₃-FMN complex in different redox states

Quenching of the oxidized flavin fluorescence upon titration with ApoFld was used to evaluate the stability of the ApoFld:3-CH₃-FMN_ox complex. 3-CH₃-FMN_ox required larger ApoFld concentrations than FMN_ox to reach saturation, and when this point was reached, only ∼50% fluorescence quenching was achieved versus >90% for the native form (Fig. 5 A). These two observations suggest considerably weaker binding of 3-CH₃-FMN_ox, as well as formation of a complex in which the flavin ring is more exposed to the solvent than in WT Fld_ox. Thus, the K_d of the ApoFld:3-CH₃-FMN_ox complex was 74 ± 7 nM, ∼100-fold larger than those reported for WT Fld_ox (Fig. 5 A) and the other Fld_ox variants (6). Therefore, binding of 3-CH₃-FMN_ox to ApoFld is considerably weaker than binding of FMN_ox or the other FMN analogs. In WT Fld, the FMN_sq is bound more tightly than the FMN_ox, whereas reduction to the hq state considerably destabilizes the flavin-protein interaction, providing a V-shape energy profile (Fig. 5 B) (6, 7, 17). This characteristic V energy profile cannot be observed for ApoFld:3-CH₃-FMN, since free-energy values are remarkably less negative (Fig. 5 B) and stability of the ApoFld:3-CH₃-FMN_hq complex is considerably diminished (thus preventing determination of the interaction parameters or of E_sq/hq for 3-CH₃-FMN). Additionally, the ApoFld:3-CH₃-FMN_sq complex is only ∼1 kcal·mol⁻¹ more stable than the ApoFld:3-CH₃-FMN_ox, compared to the ∼2.75 kcal·mol⁻¹ difference for WT Fld. This differs also from the other Fld variants, which showed ΔG values for the complex formation in the three oxidation states very similar to those for WT Fld (6). Therefore, and in contrast to the other FMN analogs here studied, 3-CH₃-FMN binds weakly to ApoFld.

Figure 5.

Binding of FMN and the 3-CH₃-FMN analog to ApoFld. (A) Fluorometric titrations of ∼200 nM samples of FMN_ox and 3-CH₃-FMN_ox with ∼10 and ∼20 μM ApoFld, respectively. (B) Binding energy profiles of FMN (squares) and 3-CH₃-FMN (circles) versus the reduction states of the reconstituted Flds.

3-CH₃-Fld as an ET protein

The ability of 3-CH₃-Fld to interact and exchange electrons with its physiological partners, PSI and FNR, was also evaluated.

Similar to the case for WT Fld, the nonlinear dependencies of the k_obs for the one-electron reduction of 3-CH₃-Fld_ox by PSI on protein concentration indicate formation of a transient PSI:Fld_ox complex before ET and allow determination of the K_a for the PSI:Fld_ox complex and k_et. 3-CH₃-Fld_ox showed a one order of magnitude decrease of the K_a compared with WT Fld (5.0 × 10⁴ M⁻¹ vs. 2.2 × 10⁵ M⁻¹), but a slight increase in the ET rate (500 s⁻¹ vs. 313 s⁻¹). Also similar to the case for WT, a rearrangement of the initial PSI:Fld_ox interaction to achieve an optimal ET is reflected in the biphasic dependence of k_obs with MgCl₂ concentration (not shown) (26). On the other hand, important deleterious effects in k_et were reported for some of the other variants here analyzed (11).

Difference spectra obtained when using 3-CH₃-Fld_ox to titrate FNR_ox showed perturbations similar to those reported for the WT complex (18, 27), as well as a saturation profile that allowed determination of the K_d and ΔG parameters in the range of WT Fld (Table 4). Therefore, replacement of the hydrogen at the N3 of FMN by a CH₃ does not affect binding of 3-CH₃-Fld_ox to the FNR_ox flavin environment, as also reported for the rest of the Fld_ox variants (11). When analyzing fast ET from FNR_hq to 3-CH₃-Fld_ox, a two-step process was observed, again similar to that reported for the WT and the rest of the variants here studied. These processes are related to initial production of Fld_sq and FNR_sq, followed by the reduction of a second Fld_ox molecule by the FNR_sq produced in the first step (6, 18, 27). However, when compared with the WT protein, 3-CH₃-Fld_ox showed an improved ability to accept electrons from both FNR_hq and FNR_sq (Table 4). Nevertheless, such increments in k_et were lower than those described for the other Fld reconstituted variants, in agreement with their more positive midpoint potentials (Tables 3 and 4).

Table 4.

Interaction Parameters Obtained by Differential Spectroscopy and Kinetic Parameters from Stopped-Flow Analysis for the Processes of FNR with the Different Fld Variants

	Parameters for the FNRox:Fldox Interaction					Fast Kinetic Parameters for the Reduction of Fldox by FNRhq
Kd (μM)	Δεband I (mM−1 cm−1)	Band II (nm)	Band I (nm)	ΔGoxb (kcal mol−1)	kap1 (s−1)	kap2 (s−1)
WT Flda	0.65	2.67	392	466	−8.4	1.6	0.3
8-Cl-Flda	0.55	2.9	393	464	−8.5	100	20
7,8-diCl-Flda	0.55	2.84	396	465	−8.5	220	30
7-H,8-Cl-Flda	0.63	2.2	395	464	−8.5	18	2
7-Cl,8-H-Flda	0.88	2.5	397	465	−8.3	100	25
3-CH3-Fld	0.32	2.1	393	469	−8.9	33	10

Data were obtained in 50 mM MOPS, pH 7.0, at 25°C. Errors in the estimated ΔG, K_d, Δε, and k_ap values were <±15% of the value.

^aData were obtained from Frago et al. (11).

^bValues were calculated according to ΔG_ox = −RTln(1/K_d).

Structural model for 3-CH₃-Fld

The calculated molecular electrostatic potentials of 3-CH₃-FMN_ox, 3-CH₃-FMN_sq, and 3-CH₃-FMN_hq, indicated no changes related to the corresponding FMN states in the charge distribution upon replacement of the N3 proton of FMN with a methyl group (Fig. S5). The 3D quantum mechanics/molecular mechanics models derived for 3-CH₃-Fld_ox, 3-CH₃-Fld_sq, and 3-CH₃-Fld_hq indicate that to fit the bulky CH₃ group inside the protein, some of the in-plane H-bond network interactions and, therefore, the π-stacking interactions provided by the 50s and 90s loops must be displaced (Fig. 6). Thus, in-plane interactions of N1, O2, N3, and N5 of the pyrazine and pyrimidine portions of the isoalloxazine ring with protein residues such as Thr⁵⁶, Asn⁵⁸, Ile⁵⁹, and Asn⁹⁷ are expected to disappear or become severely impaired. In fact, the H-bond between the H at N3 and the Ans⁹⁷ carbonyl group in ApoFld cannot be formed anymore. This interaction is a universal feature in flavoproteins, and particularly in Flds (28). As a consequence, a configuration change of the flavin ring occurs within the flavin binding pocket, inducing increased accessibility of the benzene isoalloxazine surface to the solvent. Additionally, a slight bending of the ribityl chain is also proposed, in accordance with our results for the hyperfine interaction of H^sC1′ proton (see above). Finally, the orientation of the two aromatic residues π-stacking the flavin ring, Trp⁵⁷ and Tyr⁹⁴, is also predicted to be modified, as observed particularly in the hq state (Fig. 6). All these features contrast with the structural model calculations for the other Fld variants, which showed just a modest rearrangement of the FMN environment, basically preserving all the main in-plane and π-stacking interactions of WT Fld (6).

Figure 6.

Isoalloxazine environment showing selected residues at 3–4 ns of the quantum mechanics/molecular mechanics MD simulation of (A) Fld_ox, (B) Fld_sq, (C) Fld_hq, (D) 3-CH₃-Fld_ox, (E) 3-CH₃-Fld_sq, and (F) 3-CH₃-Fld_hq. In (A)–(C), the H-bonds that stabilize binding of FMN and are affected by introduction of 3-CH₃-FMN are indicated. Although much longer production runs will be required to visualize liberation of 3-CH₃-FMN_hq, the simulation time allows us to predict such a tendency.

Discussion

Only considering the EPR results reported here, one can say that the study of the FMN-substituted Fld variants has contributed to the characterization of the general sq state of the flavin, since some details about the spin-density map were not obtained before. Worth mentioning are (see the previous section and the Supporting Material for details): 1) the assignment of the HYSCORE spectral features, previously referred to as NA and NB nuclei, to N1 or N3, respectively (13, 15); 2) the assignment of the complete set of hyperfine principal values for H6 and H^sC1′ protons; and 3) new information about the hyperfine constants of H7 and H8 in the corresponding Fld variants. Moreover, the more accurate determination of as many hyperfine parameters as possible in the flavin sq state contributes to complete the picture of the flavin spin populations and, therefore, to improve our understanding of the relationship between the flavin sq electronic structure and its functionality within each particular protein environment (15). In this line, our study is an example of the application of CW-EPR, ENDOR, and HYSCORE techniques to study the details of the flavin-ring spin population in the sq state when it is differentially substituted and is embedded within the protein environment (see Tables 1 and 2).

To discuss our results, it should first be considered that the relationship between spin populations in the flavin sq state and flavoprotein functionality is not straightforward. Previous studies showed that some mutations in the closest protein environment of the flavin ring substantially affect the flavoprotein ET ability while having only minor effects on the spin distribution (14, 18). On the other hand, it is also known that deprotonation at position N5 of the sq state when going from neutral to anionic sq produces important redistribution of the electronic spin (29). Moreover, it should be also kept in mind that flavoproteins are a large and versatile family of proteins with many different functions, mechanisms, and sophisticated relationships between flavin electronic and molecular structures and protein function. Thus, spin distribution in the sq state can be considered as a major factor related to function, as it offers a map of the orbital occupied for the electron to be transferred. Nevertheless, other elements must be also considered relevant, such as the atomic structure of the flavin isoalloxazine ring itself, the noncovalent interactions contributing to fit the flavin ring into the flavin binding pocket of each particular flavoprotein, the flavin solvent accessibility, and the interaction with other molecules (including protein substrates, partners, ligands, etc…).

Mutational analysis in the Anabaena Fld indicated that in the physiological ET reaction, the initial relative orientation of the Fld molecule and its FNR partner is driven by the alignment of their dipole moments, which contribute to the formation of several alternate binding modes competent for ET, whereas further reorganization is required upon initial electrostatic interaction with PSI (9, 10). Therefore, as the short-range Fld-protein interaction does not appear to limit Fld biological ET efficiency, two physicochemical properties must be mainly considered as determinants for ET efficiency: midpoint reduction potentials for the two independent ET steps, E_ox/sq and E_sq/hq, and electronic distribution within the flavin rings. Previous studies showed a close relationship between those two properties for isolated flavin analogs (30). Additionally, internalization of the FMN isoalloxazine ring in the ApoFld binding pocket dramatically influences its reduction potentials (7, 8, 11, 17). Residues contributing to stabilization of the isoalloxazine ring concentrate in the 50s and 90s loops of ApoFld and participate in its binding through both parallel π-stacking and in-plane H-bond interactions (Fig. 6) (7, 8, 17, 18). The parallel π-stacking ensures an isoalloxazine electronegative environment and allows tight FMN_hq binding while making formation of its anion thermodynamically unfavorable, thus contributing to establish the typical very negative E_sq/hq of Fld (10). Tyr⁹⁴ stacking against the FMN si-face modulates midpoint reduction potentials by limiting solvent accessibility, whereas Trp⁵⁷ (at the FMN re-face) modulates E_ox/sq by influencing the strength of the H-bond between the N5 and the Asn⁵⁸-Ile⁵⁹ peptide bond (8). Rearrangement of this peptide bond allows a main-chain carbonyl to flip conformation (Fig. 6, A–C). The flip involves breaking an H-bond present in Fld_ox between the N5 of FMN_ox and the NH of Ile⁵⁹ and forming an H-bond between this carbonyl and the N5H of the isoalloxazine in the neutral Fld_sq (8). This backbone rearrangement provides a versatile mechanism for modulating the strength of FMN binding, as well as of E_ox∕sq and E_sq∕hq (6, 8). Finally, Fld has a large excess of acidic residues, which produce a strong dipole that orients its negative end toward the FMN isoalloxazine (7, 8, 10). Experimental data using EPR-related techniques as well as calculations have also shown that in-plane interactions influence the spin distribution in larger extension than π-stacking interactions (14, 31, 32).

In this context, the characterization by EPR-related techniques of ApoFld samples reconstituted with FMN analogs having different electronic properties further contributes to our understanding of the intricate relationships among spin-density distribution and functionality. The first group of samples selected in our study consists of Flds reconstituted with FMN analogs modified in CH₃7 and CH₃8, namely, 8-Cl-Fld, 7-H,8-Cl-Fld, 7-Cl,8-H-Fld, and 7,8-diCl-Fld. Their functional and structural characterization have been previously reported (6, 11). Most differential structural and functional properties of these variants compared with WT Fld can be attributed to the flavin structure modification. Thus, differences in the flavin analog molecular electrostatic potentials, when compared with the FMN, demonstrate the effect of electron withdrawal by chlorine ions from the flavin pyridine and pyrazine moieties toward the benzene ring (6). Accordingly, the hyperfine couplings reported here for all these Fld_sq variants (Tables 1 and 2) follow the effect determined by their molecular electrostatic potentials when compared with the WT. This includes a reduction of the spin distribution at position 5 (as indicated by the N5 and H5 hyperfine anisotropies in Tables 1 and 2), an increase at position 6 (isotropic hyperfine constant and hyperfine anisotropy of H6; see Table 2), and possibly also an increase at position 7 (hyperfine constant of H7 in the 7-H,8-Cl-Fld variant; see the Supporting Material). Although those changes can be considered modest (<10%), they are relevant given the robustness of the flavin spin distribution over changes of its structure and environment (31). These changes in spin distribution have no major effect on affinity of the cofactor:ApoFld complexes (6). Nevertheless, they are clearly related to the lower effect of ApoFld in influencing midpoint reduction potentials, particularly E_sq/hq, in the analogs when compared with FMN and to the protein environment enhancing the substituent withdrawal effect. In fact, a correlation between N5 hyperfine constants and E_sq/hq is evident (see Tables 1 and 3).

Notably, the larger changes in the spin-distribution population are concentrated at positions N5, C6, and C7 of the flavin ring, which, together with Asn⁵⁸, are proposed to be determinant in the formation of competent complexes with partners, with a key role envisaged for C6 and the Asn⁵⁸ backbone in the ET process (6). These changes have a limited functional effect. In all these variants, ET from PSI to Fld is just somehow impeded at the complex rearrangement step due to the substituent electron withdrawal effect on the magnitude and orientation of the molecular dipole moment of Fld, whereas ET from FNR_hq is more efficient than in WT Fld, as expected from their less negative midpoint reduction potential (Table 4).

We have additionally chosen to study 3-CH₃-Fld. For this variant, the bulky methyl group prevents formation of the H-bond between N3 of the flavin and Asn⁹⁷ and also causes other in-plane and π-stacking interactions to either disappear or become weaker (Fig. 6). Thus, replacement of FMN with 3-CH₃-FMN in Fld considerably impairs flavin binding, producing a protein that stabilizes a considerably smaller amount of the sq state and whose E_ox/sq is relatively close to that of the free flavin (Table 3; Fig. 5). Although 3-CH₃-FMN keeps the electronic structure of the naturally occurring FMN ring (Fig. 5), 3-CH₃-Fld_sq shows changes in spin distributions, with a slight decrease at position 5 (as indicated by the value of the isotropic hyperfine constant of H5; see Table 2) and a more prominent decrease at position 8 (hyperfine parameters of CH₃8; see Table 2). There is also a change in the H^sC1′ hyperfine parameters, which may be due to a slight ribityl reorientation around the N10-C1′ bond as it is also related to changes in the accommodation of 3-CH₃-FMN within the Fld flavin pocket (see Fig. 6).

Since none of the spin-population changes seems to be related to changes in the flavin electronic structure itself, the interaction with the protein must be responsible for them. In this line, previous calculations showed the effect of in-plane and π-stacking interactions on the flavin spin distribution (6, 31, 32). The changes in hyperfine interactions of H5 and CH₃8 here reported are in better qualitative agreement with the effect of a weakening of the in-plane interactions. These observations are also in agreement with the ApoFld:3-CH₃-FMN_ox complex being destabilized by 2.3 kcal/mol compared to WT FMN. Such destabilization considerably increases upon reduction and therefore produces dissociation of the hq from ApoFld (Fig. 5). All these facts contribute to a considerable increase in the benzene isoalloxazine surface accessibility to the solvent and to the weaker flavin:ApoFld interaction. Therefore, they point to the observed changes in 3-CH₃-Fld being attributable not only to the disruption of the N3-Asn⁹⁷ H-bond but also to the alteration of other H-bond interactions between the isoalloxazine ring and ApoFld. Therefore, this network plays a fundamental role not only in holding the isoalloxazine ring inside the protein, but also in setting the spin distribution within the protein environment.

Conclusions

Our study of Anabaena Fld reconstituted with several FMN analogs adds to the experimental quantitative information describing the flavin spin population in the sq state. This is a relevant property, as it maps the singly occupied molecular orbital involved in the ET processes. Thus, characterization of these FMN-substituted Fld variants by combining different EPR-related techniques provides a better understanding of the factors influencing the electronic structure of the flavin ring within the protein, separating the effect of intrinsic changes in the electronic distribution in the isoallosazine ring from those related to the ApoFld-FMN interaction. The electronic withdrawal effect of Cl substituents at positions 7 and/or 8 produces a decrease of the spin population at position 5 and an increase at positions 6 and 7, whereas perturbation of the ApoFld-FMN interaction reduces spin populations at positions 5 and 8. The behavior of the spin distributions can be related to changes in the reduction potentials of the variants relative to those of the WT Fld, and these potentials constitute the main driving force of the protein functionality.

Finally, the protocol here presented can be used to further deepen the relationship of the spin distributions of the flavin ring in the sq state to other physicochemical functionally relevant properties. These include physical properties, such as solvent exposure or both long-range and short-range electrostatic interactions with substrates, together with chemical interactions and changes in the strength of specific H-bonds or residue configurations.

Author Contributions

J.I.M. and M.M. designed the research; J.I.M. and I.G.-R. performed EPR experiments; J.I.M. and P.J.A. analyzed the EPR results; S.F. performed biochemical characterization experiments; I.L. performed calculations; M.M., S.F., and I.L. analyzed the results of the biochemical characterization; and J.I.M., P.J.A., and M.M. wrote the manuscript.

Editor: H. Jane Dyson.

Supporting Citations

References (33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44) appear in the Supporting Material