Probing the Backbone Dynamics of Oxidized and Reduced Rat

12320

Biochemistry 1998, 37, 12320-12330

Probing the Backbone Dynamics of Oxidized and Reduced Rat Microsomal Cytochrome b5 via 15N Rotating Frame NMR Relaxation Measurements: Biological Implications† Lucia Banci, Ivano Bertini,* Christine Cavazza, Isabella C. Felli, and Dionysios Koulougliotis Department of Chemistry, UniVersity of Florence, Via Gino Capponi 7, 50121 Florence, Italy ReceiVed April 20, 1998; ReVised Manuscript ReceiVed July 2, 1998

Rotating frame 15N relaxation NMR experiments have been performed to study the local mobility of the oxidized and reduced forms of rat microsomal cytochrome b5, in the microsecond to millisecond time range. Measurements of rotating frame relaxation rates (R1F) were performed as a function of the effective magnetic field amplitude by using off-resonance radio frequency irradiation. Detailed analysis of the two data sets resulted in the identification of slow motions along the backbone nitrogens for both oxidation states of the protein. The local mobility of reduced and oxidized cytochrome b5 turned out to be significantly different; 28 backbone nitrogens of the oxidized form were shown to participate in a conformational exchange process, while this number dropped to 12 in the reduced form. The correlation time, τex, for the exchange processes could be determined for 21 and 9 backbone nitrogens for oxidized and reduced cytochrome b5, respectively, with their values ranging between 70 and 280 µs. The direct experimental evidence provided in this study for the larger mobility of the oxidized form of the protein is consistent with the different backbone NH solvent exchangeability recently documented for the two oxidation states [Arnesano, F., et al. (1998) Biochemistry 37, 173-184]. Our experimental observations may have significant biological implications. The differential local mobility between the two oxidation states is proposed to be an important factor controlling the molecular recognition processes in which cytochrome b5 is involved. ABSTRACT:

Solution structures obtained through NMR are quite relevant as they constitute the starting point for other studies; in particular, they can be further explored in terms of local mobility. In electron transfer proteins, the structure and mobility can be meaningfully compared between the two redox partners (1-8). This has become possible after the progress in the solution structure determination of paramagnetic molecules (2, 9-11) since, in most cases, at least one redox partner in electron transfer proteins is paramagnetic (12). Mobility may be related to molecular recognition (13). The solution structures of both oxidized and reduced rat microsomal cytochrome b5 have recently become available (7, 8). Cytochrome b5 (cyt b5 hereafter) is a largely widespread electron transfer protein found in a variety of mammalian and avian species, and it has been extensively studied (12, 14, 15). It contains a b-type heme which, together with two histidine side chains, provides the coordination sphere for the iron ion. The latter cycles between the +3 and +2 oxidation states. Since the heme is not covalently bound to the protein matrix, it is present in two orientations, usually indicated as the A (major) and B (minor) † This work was supported by the European Community (Program BIOMED, CT BMH4-CT96-1492), the Italian CNR (Biotecnologie e biologia molecolare, CTB CNR 95.02860.CT14), and MURST-ex 40%. D.K. thanks the TMR Programme of the European Union for a postdoctoral fellowship (Contract ERBFMBICT960994). * Corresponding author: Department of Chemistry, University of Florence, Via Gino Capponi 7, 50121 Florence, Italy. Fax: +39 55 2757555. Telephone: +39 55 2757549. E-mail: [email protected].

forms. The relative populations of the two forms vary according to the specific isoenzyme and are 6:4 in the rat microsomal isoenzyme. Cyt b5 is mainly found as a membrane-bound protein (16). However, a hydrophilic domain can be identified. It can be either isolated through tryptic cleavage or expressed in Escherichia coli, and it still retains its activity (16-19). Due to its moderate dimensions as well as to the relatively easy procedure through which it can be obtained, it has become the subject of many studies with a variety of different methods. Therefore, it represents an ideal system for addressing fundamental problems related to how the electron transfer process occurs in living organisms (20-22) and, more generally, to how energy is transferred in living organisms. The careful comparison of the solution structures of cyt b5 in both oxidation states (7, 8) revealed small but significant differences indicating structural rearrangements that could be important in determining the stability of the protein in the two redox states and thus be related to the redox potential. Further inspection of NMR data revealed a differential behavior of the two redox states with respect to exchange of labile protons with the bulk solvent (7). Since the structure of the protein in the two oxidation states is not so different to justify a different solvent accessibility, this fact could be interpreted as being due to differential mobility in the two oxidation states (7), which would produce different instantaneous accessibility. Local mobility has often been thought, from indirect evidence, to be important in controlling the electron

S0006-2960(98)00885-X CCC: $15.00 © 1998 American Chemical Society Published on Web 08/08/1998

Backbone Dynamics of Oxidized and Reduced Cytochrome b5 transfer process (22-27). Local mobility can indeed influence the molecular recognition process (13, 28-31) as well as the reorganization energy λ (20-22), both features being necessary to achieve electron transfer in biological systems. Moreover, differential mobility of some regions of the protein can provide insight about the recognition process between the two redox partners and in the electron transfer process (7). However, direct experimental evidence of local mobility for electron transfer proteins in both oxidation states is still quite rare (32-34). Therefore, a further characterization of the local mobility of cyt b5 in both oxidation states could provide useful information. Analysis of heteronuclear NMR relaxation has revealed a unique method for obtaining information on local mobility (35-37). In particular, rotating frame relaxation experiments have recently attracted the attention of several researchers (38-40) due to their potential to reveal motions in an intermediate time range, that is from microseconds to milliseconds (36, 41-45). This method has recently been applied on the HiPIP I from Ectothiorhodospira halophila, and it allowed the identification of a loop exhibiting an exchange process (46). We report here a characterization of the local mobility of rat microsomal cyt b5 in both oxidation states, in the microsecond to millisecond time range, by exploiting 15N off-resonance rotating frame relaxation. EXPERIMENTAL PROCEDURES Sample Preparation. 15N-labeled cyt b5 was isolated and purified as previously described (16) from cultures of E. coli strain TOPP 1 harboring the recombinant pUC13 plasmid containing the gene encoding the 98-amino acid polypeptide corresponding to the soluble part of rat microsomal cyt b5 (kindly provided by S. G. Sligar). The cultures were grown in minimal M9 medium containing (15NH4)2SO4 (3 g/L) as the sole nitrogen source. After cell lysis with sonication, the purification was carried out in two steps, including a CL6B anion exchange column followed by a P100 (Bio-Rad) chromatography column. Approximately 15 mg of pure protein (obtained from a 5 L culture) was exchanged through ultrafiltration (YM10 membranes, Amicon) with 10 mM phosphate buffer at pH 7.0. The sample volume was reduced to ca. 500 µL with a final concentration of ca. 3 mM. Under these conditions, the majority of the protein (>95%) is in the oxidized form. To keep the protein in the oxidized state over the course of several days (needed for the performance of the NMR experiments), the NMR tube was periodically (every 40-45 h) flashed with oxygen for 10-15 min. The protein was reduced by the anaerobic addition of 20 µL of 1 M sodium dithionite solution buffered at pH 6.9 (10 mM phosphate). NMR Experiments. All NMR experiments were carried out at 298 K, on a Bruker Avance 600 NMR spectrometer operating at a proton Larmor frequency of 600.13 MHz. The 15N off-resonance rotating frame relaxation rates (R1FOFF) were measured as a function of the effective magnetic field amplitude (ωeff) by using the pulse sequence described recently (39). The 15N RF irradiation was applied with an amplitude ω1 and with an offset ∆ω with respect to the center of the amide nitrogen resonances to lock the magnetization along an axis making an angle θ ) arctan-

Biochemistry, Vol. 37, No. 35, 1998 12321 (∆ω/ω1) ) 35° with a resulting effective magnetic field amplitude ωeff. The amplitude of the 15N RF irradiation during the pulse sequence was increased and decreased gradually in a trapezoidal fashion to achieve adiabatic rotation of the magnetization to the effective magnetic field axis (39). During the application of the continuous wave spin-lock field at the 15N frequency, decoupling of the protons is achieved with the Waltz-16 pulse decoupling sequence (47) to avoid creation of antiphase 15N magnetization via cross-correlation processes (of the type NxIz, where N and I are the nuclear spin operators of the 15N and the directly bound 1H nuclei, respectively). The decoupling power was set to 2670 Hz in all experiments. The INEPT transfer delay was set to 2.5 ms (48). The applied effective magnetic field amplitude values ωeff for the measurement of R1FOFF were as follows: 1870, 2090, 2300, 2590, 2890, 3210, and 3550 Hz. For each spin-lock power, a series of two-dimensional (2D) experiments was performed in which the relaxation delay T was set to the following values: 10, 20, 36, 50, 60, 86, 100, 150, 200, 300, and 10 ms. A further 2D spectrum without the period of spin-lock field was also acquired to measure the initial magnetization. All experiments were recorded with a spectral width of 2130 Hz in the F1 (15N frequency) dimension and of 8390 Hz in the F2 (1H frequency) dimension. A total of 160 experiments in t1, each of 2048 real data points, were recorded. Every free induction decay was comprised of 16 scans. Quadrature detection in F1 was obtained by using the TPPI method of Marion and Wu¨thrich (49). The longitudinal relaxation rate, R1, was measured via an inversion recovery experiment of the 15N nuclear spin introduced to a 2D heteronuclear correlation spectrum as described by Peng and Wagner (50). The following relaxation delays were used: 5, 10, 20, 40, 80, 120, 240, 320, 500, 1000, and 1500 ms. The acquisition parameters were equivalent to those used for the R1FOFF measurement except that a total of 192 experiments were recorded for each 2D spectrum, each being comprised of eight scans. Sequence-specific assignments of cyt b5 in both oxidation states were available in the literature (7, 8, 51, 52). Data Processing. All 2D NMR data were processed on a Silicon Graphics workstation using the Xwinnmr Bruker software. Only the downfield part of the spectra (in the 1H dimension), containing the HN-N connectivities (5-12 ppm), was kept for the data analysis. All spectra were transformed with 4K × 1K points in the F2 and F1 dimensions, respectively. All spectra acquired with a common spin-lock power were processed by using the same processing parameters (phasing parameters, baseline correction, etc.). Subsequent integration of cross-peaks for all spectra was performed by using the standard routine available from Bruker. Determination of Relaxation Rates. Relaxation rates R1 and R1FOFF were determined by fitting the cross-peak intensities (I) measured as a function of the relaxation delay T to a single-exponential decay by using the Levenberg-Marquardt algorithm (53, 54) according to the following equation:

I(T) ) A + B exp(-RT)

(1)

A, B, and R are adjustable fitting parameters. A program

12322 Biochemistry, Vol. 37, No. 35, 1998

Banci et al.

which uses a Monte Carlo approach (39, 55, 56) to estimate via the Levenberg-Marquardt algorithm the rates and their errors was used for this purpose. For R1FOFF, the phase cycle was chosen so that the magnetization relaxes to zero for long relaxation delays. Thus, in this case, A was set equal to zero in the fitting procedure. The error bars correspond to the experimental error averaged over all powers. The error value for R1FOFF,cor (see later) includes the errors in the measurement of both R1FOFF and R1. Analysis of Relaxation Rates. The magnetization of a 15N spin i, when an RF irradiation of frequency ω and amplitude ω1 is applied in the vicinity of its resonance frequency ωi, is aligned along the effective magnetic field, which is characterized by an amplitude ωeff,i of (∆ωi2 + ω12)1/2 and a direction making an angle θi of arctan(ω1/∆ωi) with the static magnetic field axis. The relaxation rate R1FOFF of the ith spin along the effective magnetic field is given by the following expression (40, 44, 57, 58)

R1FOFF ) R1 cos2 θi + R1FON,∞ sin2 θi + K sin2 θi

τex 1 + τex2ωeff,i2

(2)

in the presence of chemical and/or conformational exchange between two conformations and a fast regime with respect to the chemical shift separation, i.e., δΩτex , 1, where δΩ is the chemical shift difference between the resonating nucleus in the two sites. The variables have the following meanings. R1 is the longitudinal relaxation rate of the ith spin. R1FON,∞ is the on-resonance rotating frame relaxation rate of the ith spin for an infinitely large effective field amplitude (where all exchange contributions are dispersed to zero). τex is the correlation time for the exchange process involving the ith spin. K is a constant equal to papb δΩ2, where pa and pb are the relative populations of the two states a and b, respectively, between which the exchange process occurs. Even if the rates and τex are of course relative to a specific spin i, this will not be indicated explicitly so the nomenclature can be kept as simple as possible. If the angle θi were kept constant, a dependence of R1FOFF on ωeff would be indicative of the presence of an exchange process involving the 15N nucleus under investigation. However, for an applied RF irradiation with an amplitude ω1, the angle θi will be strictly constant only for one spin i with frequency ωi. Therefore, for all the other spins, when R1FOFF, R1, and i are known, the following correction should be applied:

R1FOFF - R1 cos2 θi sin2 θi

) R1FON,∞ + τex ) R1FOFF,cor (3) K 2 2 1 + τex ωeff,i

Thus, information on fast exchange processes can be obtained from the dependence of R1FOFF,cor on the effective magnetic field amplitude. If the experimental R1FOFF,cor values are fit to a Lorentzian function (eq 3), the correlation time τex for the exchange process can be estimated. If, on the other hand, no exchange process is present, R1FOFF,cor is independent of ωeff. The lower limit value for R1FOFF,cor is R1FON,∞ when

ωeff,iτex . 1, i.e., τex . 10-4 s (for the ωeff used in this work), and the upper limit is R1FON,∞ + Kτex when ωeff,iτex , 1, i.e., τex , 10-4 s (in the limit of ωeff f 0, R1FON,∞ f R2). In addition, δΩτex , 1 must always hold. These limits are of help in determining τex from the experimental data even if the latter values span a limited part of the Lorenzian curve. In our particular case, since ωeff ranges between 1870 and 3550 Hz, we have access to τex values in the range of approximately 10-300 µs. A dependence of R1FOFF,cor on ωeff is an indication of the presence of an exchange process; however, the estimate of τex is not very accurate for τex values larger than approximately 150 µs, and the data on τex are discussed qualitatively (see later). RESULTS The off-resonance rotating frame relaxation rate, R1FOFF, was determined for each amide backbone 15N spin at seven different effective magnetic field amplitudes, ωeff. Except if differently indicated, the data are reported and analyzed for the A form of the protein. The experiments were performed by changing simultaneously the values of ω1 and ∆ω (∆ω is the offset between the frequency of the applied RF field and the center frequency of the amide region) to keep the angle θ ) arctan(∆ω/ω1) ) 35° for a nitrogen spin resonating at a frequency corresponding to the center of the amide region. For 15N spins resonating at a different frequency, the angle θi will undergo a small but systematic variation corresponding to the changes in the values of ω1 and ∆ωi. This was accounted for, as described above, to obtain R1FOFF,cor values according to eq 3. These values are hereafter analyzed. For rat microsomal cyt b5, the largest increases in θi between the lowest and highest values of ω1 used were 4° and 4.8° (for the amide spin of residue 52) for the reduced and oxidized forms, respectively. The largest decreases in θi were 14.6° (reduced cyt b5) and 14° (oxidized cyt b5), for the backbone amide of residue 94. It should be noted, however, that the angular dispersion in the rotating frame (off-resonance) is always smaller than the corresponding dispersion when on-resonance irradiation is applied. The results on R1FOFF,cor are discussed below separately for each oxidation state of cyt b5. Oxidized Cyt b5. The longitudinal relaxation rates R1 for 68 backbone amide nitrogen spins, which could be precisely integrated as they are not overlapping with other signals, are shown in Figure 1A. These rates display a uniform behavior along the protein sequence, and their average value is 1.9 ( 0.1 Hz, consistent with previously reported data on the bovine isoenzyme (59). These values were subsequently used in eq 3 to obtain R1FOFF,cor values. The dependence of R1FOFF,cor on ωeff was examined for each of the 68 backbone amides, and for 47 of these, the R1FOFF,cor values were found independent of the effective magnetic field amplitude. The average values of R1FOFF,cor (estimated using all the values obtained at the seven effective magnetic field amplitudes employed in the experiments) for the 47 backbone amides, which were found independent of ωeff, are shown in Figure 2A (unmarked residues). These relaxation rates show overall a quite uniform behavior with an average value for R1FOFF,cor ) R1FON,∞ ) 6.8 ( 0.9 Hz, except of the ones corresponding to the backbone nitrogens of residues 27, 31,

Backbone Dynamics of Oxidized and Reduced Cytochrome b5

FIGURE 1: Longitudinal relaxation rates of 15N backbone amides per residue for the oxidized (A) and reduced (B) forms of cyt b5.

34, 44, 54, 72, and 80. The latter seven residues have R1FOFF,cor rates which are 40-150% larger than those of the remaining 40. An increase in R1FON,∞ due to higher “rigidity” of the NH bond vector in the sub-nanosecond time scale should be smaller than 40% (on the order of 25%), and thus, the data obtained for the above seven residues can be interpreted in terms of exchange processes. These findings can be interpreted as follows. For the 40 residues with a uniform relaxation rate R1FOFF,cor (6.8 ( 0.9 Hz), one can conclude either that there is no exchange process (τex ) 0) or that it is not large enough to sizably contribute to R1F. On the other hand, the behavior observed for the other seven residues can be an indication of the presence of a conformational exchange process. However, its time scale is faster that the one accessible with the available powers of the spin-lock field (ωeff), and thus, R1FOFF,cor appears to be constant as a function of ωeff but with a higher value than in the absence of an exchange process. Since the highest effective magnetic field corresponds to a frequency of 3550 Hz, the exchange process operating for residues 27, 31, 34, 44, 54, 72, and 80 should have a correlation time τex of