---

1. Introduction

1.1 NMR versus EPR

Imagine opinion polls had been taken every ten years since 1980 and questionnaires had been distributed among scientists from physics, chemistry and biology asking for an answer to the question “what is magnetic resonance?” Up to 10 years ago, most of these scientists, no matter in which specific discipline they had been working, would have responded in detail, but probably with reference to NMR (nuclear magnetic resonance) only. The question “what is EPR?” (electron paramagnetic resonance), however, being specifically asked next in the questionnaire, would probably have been left unanswered by most biologists, and many chemists and physicists would have known only the basic principles of conventional continuous wave X-band EPR spectroscopy (operating at a microwave frequency of around 9 GHz). In fact, in contrast to EPR, NMR has been established over the last 20 years as a multi-frequency tool of highest significance in multi-disciplinary material and life sciences for spectacular applications in structure determination and imaging. Not surprisingly, therefore, as many as four Nobel prizes for NMR methodology and applications have been awarded within the last 13 years (to R. R. Ernst in Chemistry in 1991, K. Wüthrich in Chemistry in 2002, to P. C. Lauterbur and P. Mansfield in Physiology and Medicine in 2003). EPR, on the other hand, seemed to be hopelessly lagging behind its famous (and younger) sister NMR, and it is only during the last decade that the chemistry, biology and physics communities witness a dramatic catching up of EPR because of technological breakthroughs in pulsed microwave, sweepable cryomagnet and fast data acquisition instrumentation. In fact, modern EPR is booming now rather similar to what had happened with NMR 10 years earlier. Hence, imagine there will be another opinion poll these days and again the question will be asked “what is magnetic resonance?” It might turn out that many chemists, biologists and physicists will answer by mentioning both NMR and EPR, both exhibiting unique potentials in elucidating structure and dynamics of complex systems.

Both EPR and NMR were invented by continuous wave (cw) experiments, EPR in 1944 by E. K. Zavoisky at Kazan University in the USSR,1,2 NMR in 1946 by E. M. Purcell, H. G. Torrey and R. V. Pound at Harvard3 and, independently, by F. Bloch, W. Hansen and M. E. Packard at Stanford.4 These classical NMR experiments were honoured as early as 1952 by the Nobel Prize in Physics to Bloch and Purcell. Zavoisky’s discovery of EPR, on the other hand, was only inadequately recognized on the western side of the Iron Curtain (in contrast to the eastern side5), and it was as late as 1977 when Zavoisky was finally honoured at least by the prestigious ISMAR Award of the International Society of Magnetic Resonance at its 6th ISMAR symposium in Banff, Alberta, Canada—posthumously, as he had died in October 1976 just after having been informed in Moscow about the decision of the Prize Committee.

Up to the 60’s, both NMR and EPR remained to be cw methods, i.e., the samples sitting in a static magnetic field were irradiated by continuous radiofrequency (rf) and microwave (mw) electromagnetic fields to drive NMR and EPR transitions, respectively. But it was as early as 1949 when E. L. Hahn6,7 at Urbana applied rf pulses and invented nuclear spin-echo detection, thereby opening the arena for pulse NMR with all its future potential for beautiful multi-dimensional spectroscopy of complex biosystems in the liquid and solid state.8,9 The realization of such pulsed multi-dimensional NMR experiments, however, with their sophisticated pulse sequences and appealing acronyms like COSY, TROSY, NOESY or ROESY, EXSY and SECSY, INEPT and INADEQUATE, had to wait for the introduction of powerful computer and fast Fourier transformation techniques in the late 70’s, before the primary time-domain data could be recorded and translated into spectral information on structure. Nowadays, pulse NMR has completely replaced cw NMR, and high-field cryomagnets and resonance frequencies close to 1 GHz for protons have dramatically improved the detection sensitivity and chemical-shift separations, and even formerly exotic nuclei have become routinely observable.

Despite the breakthroughs in mm and sub-mm microwave technologies in the last decade, in EPR the cw versus pulse situation is very different from that in NMR. The prognosis is that a peaceful coexistence between cw and pulse EPR will continue to persist and will be determined entirely by the characteristics of the sample under study, i.e., the sample properties and relaxation times ultimately dictate the preference for either a cw or pulse experiment to be performed.

Why is there such a discrepancy between the requirements for pulsed NMR and pulsed EPR? The answer is related to the different time scales of the NMR and EPR experiments. They are determined by the nuclear and electron resonance frequencies (in the rf and mw domains, respectively), the characteristic frequency separations in the respective spectra (Hz versus MHz) and the relaxation times T1, T2 (ms versus μs). Because of the long nuclear T1 and T2 times in diamagnetic molecules, NMR pulses need not be shorter than 1 μs, which to coherently generate and detect does not pose technical problems. The electronic relaxation times are typically in the μs range or shorter (T2) and, consequently, in EPR the mw pulses have to be as short as a few ns. To generate them poses great technical problems even today in terms of mw sources and fast computers, as it does to detect and handle the transient signals in the ns time scale. Thus, the instrumental requirements for Fourier-transform (FT) EPR are much tougher to fulfil than for FT-NMR, and this explains why pulse EPR has become popular only since the late 80’s. Pulse EPR spectroscopy has many fathers, and a rich variety of pulse sequences and sophisticated experiments is found in recent reviews and text books (see, for example, ref. 10). It is probably fair to say that truly pioneering work in pulse EPR was done among others by W. B. Mims at Bell Telephone Laboratories,11,12 Yu. D. Tsvetkov in Novosibirsk,13 J. H. Freed at Cornell14 and A. Schweiger at ETH Zurich.10 This work represents a milestone in modern EPR spectroscopy.

1.2 From EPR basics to high-field EPR and ENDOR

Let us take a sample containing molecular radicals with unpaired electron spins with S = 1/2 and place it into an EPR spectrometer. The electron and nuclear spins will align with respect to the total magnetic field composed of the external “Zeeman” field, B0, the local field originating from “spin–orbit coupling” in the radical (leading to effective g-tensor components shifted from the free-electron value) and local “hyperfine” fields from nearby magnetic nuclei, for example protons with nuclear spin I = 1/2. For this sample, the static spin Hamiltonian, Ĥ0, that describes the time-independent spin-interaction energies, consists of three terms

	(1)

i.e., the field-dependent electron and nuclear Zeeman interactions and the field-independent electron-nuclear hyperfine interactions (h: Planck constant; μB, μK: Bohr and nuclear magnetons; gn: nuclear g-factors; Ŝ, Î: electron and nuclear spin vector operators; the summation is over all nuclei).

The interaction tensors  and Ãi are probing the electronic structure (wave function) of the molecule either globally ( tensor) or locally (hyperfine tensors). The tensors contain isotropic and anisotropic contributions. In isotropic fluid solution, only the scalar values, Tr()/3 and Tr(Ã)/3, are observed. In frozen solutions, powders or single crystals, on the other hand, also anisotropic tensor contributions become observable provided appropriate resolution conditions prevail. For this situation, the information content of the EPR spectra is considerably enhanced, for example in terms of molecular orientation with respect to B0.

In the strong-field approximation, the energy eigenvalues of eqn. (1) are classified by the magnetic spin quantum numbers, mS and mI, and are given, to first order, by

	(2)

where the scalar quantities g′ and A′ contain the desired information about magnitude and orientation of the interaction tensors.

As an example, let us consider the simple case of a doublet radical (S = 1/2) in isotropic fluid solution containing four symmetry-equivalent protons in a strong B0 field. Fig. 1 shows the energy level scheme according to eqn. (2), in which for isotropic solutions only scalar interaction parameters are observable due to fast molecular tumbling. Following the EPR selection rules ΔmS = ±1, ΔmIi = 0, five EPR lines are observed with binominal intensity distribution owing to the first-order transition-frequency degeneracies of equivalent nuclei.

For single-crystal samples, the complete tensor information can be extracted from the angular dependence of the resonance lines when the crystal is rotated in its three symmetry planes (“rotation patterns”). If the  and Ãi tensors are collinear, i.e., have the same principal axis system (α,β,γ), their rotation patterns will have the same angular dependence. When the crystal is mounted in such a way that the rotation axis is the x-axis of the molecular axes system (x,y,z) and is perpendicular to the B0 direction, the rotation occurs in the yz plane defining an angle θ between the z-axis and the field direction. Then, g′ takes the form:

g′ = (g2yysin2θ + g2zzcos2θ + g2yzsin 2θ)1/2	(3)

(notice that gij2 denotes the components of the squared  tensor). An analogue expression holds for A′i in the limit of small g-anisotropy, as is typical for bioorganic systems as long as they do not contain paramagnetic transition-metal ions. For the other two symmetry planes, zx and xy, the corresponding tensor components are found by cyclic replacement y → z → x → y.

For large low-symmetry radicals, such as those occurring in photosynthesis, with each set of inequivalent nuclei the number of EPR lines obviously increases in a multiplicative way, according to the selection rules ΔmS = ±1, ΔmIi = 0. This results in strongly inhomogeneously broadened EPR spectra because individual hyperfine lines can no longer be resolved in the available spectral range (which is fixed due to the normalization condition for the unpaired electron spin density). For such cases, one can resort to electron–nuclear-double resonance (ENDOR) techniques.10,15,16 It was as early as 1956 when G. Feher17 at Bell Labs, demonstrated that by ENDOR the spectral resolution can be greatly improved. This first ENDOR experiment was technically feasible only because the sample, phosphorus-doped silicon, was studied at low temperature, where all the relaxation times are sufficiently long to easily obtain the necessary saturation of the EPR and NMR transitions. For the radicals in liquid solution, however, these relaxation times are much shorter, on the order of 10−5–10−7 s, and consequently ENDOR-in-solution experiments are technically much more sophisticated, since much larger saturating mw and rf fields have to be applied. This is the reason why the first ENDOR-in-solution experiments required many more years before they were successful. The pioneering work was performed by A. L. Cederquist18 in 1963, who studied metal–ammonia solutions, and by J. S. Hyde and A. H. Maki19 in 1964, who investigated an organic radical in solution. The further development of ENDOR-in-solution spectroscopy was highly stimulated by J. H. Freed,20 whose general theory of saturation and double-resonance proved to be adequate for describing amplitude, width and shape of ENDOR lines in great detail. Certain drawbacks of ENDOR concerning sensitivity and relative line intensities can be overcome by extending it to electron-nuclear-nuclear triple resonance.21 This was proposed by G. Feher22 and J. H. Freed,23 but experimentally realized for a radical in solution first in the Möbius group at FU Berlin by K. P. Dinse et al.24 (“Special TRIPLE”). As was demonstrated in the same group by R. Biehl et al.,25 additional information about relative signs of hyperfine couplings of radicals in solution can be obtained by generalizing the triple resonance experiment to include NMR transitions of different nuclei in the radical (“General TRIPLE”). The analogue of this experiment for solid-state samples at low temperature was performed earlier by R. J. Cook and D. H. Whiffen.26 The advantages of TRIPLE over ENDOR, enhanced sensitivity and resolution, information about multiplicity and relative signs of hyperfine couplings from line intensity variations, often justify the extra experimental efforts inherent in the triple resonance spectroscopy.21

In ENDOR, the sample is irradiated simultaneously by two electromagnetic fields, a mw field (to drive EPR transitions ΔmS = ±1) and an rf field (to drive NMR transitions ΔmIi = ±1). Under appropriate experimental conditions, which are different for cw and pulse irradiation schemes.10,15,27 ENDOR signals are observed by monitoring the changes of EPR line amplitudes when sweeping the rf field through the nuclear resonance frequencies. Thus, every group of equivalent nuclei, no matter how many nuclei are involved and what their spin is, contributes only two ENDOR lines because, within an mS manifold, the hyperfine levels are equidistant to first order. The gain in resolution of ENDOR versus EPR, therefore, becomes very drastic for low-symmetry molecules because, with increasing number of groups of nuclei, the number of ENDOR lines increases only in an additive way. This gain in resolution is particularly pronounced when nuclei with different magnetic moments are involved. Their ENDOR lines appear in different frequency ranges, and from their Larmor frequencies these nuclei can be immediately identified. In the case of accidental overlap of ENDOR lines from different nuclei at X-band (9.5 GHz, 0.34 T) they can be separated when working at higher mw frequencies and Zeeman fields, for instance at 95 GHz, 3.4 T (W-band) or even at 360 GHz, 12.9 T (see Fig. 2).

For a doublet radical, the ENDOR lines of a particular group of equivalent nuclei appear, to first order, at

	(4)

with the free nuclear Larmor frequency νn = (gnμK/h)B0, and A′ being defined analogous to eqn. (3). In isotropic solution, the hyperfine couplings (hfc’s) are given by A′ = Tr(Ã)/3. Hence, in essence, ENDOR is a variant of NMR on paramagnetic systems, the unpaired electron serving as highly sensitive detector for the NMR transitions. Double resonance excitation thus offers the advantage of detecting low-intensity rf transitions by high-intensity mw transitions, i.e., by means of quantum transformation.

To illustrate the power of the ENDOR method, monomeric BChl a˙+ is chosen as example28–30 (Fig. 3). The low-symmetry BChl a˙+ is involved as an electron-transfer partner in bacterial photosynthetic reaction centres (RCs) from Rb. sphaeroides (see below). It contains many inequivalent sets of protons and nitrogens and, consequently, the EPR spectrum, even in liquid solution at room temperature, is strongly inhomogeneously broadened with only an indication of hyperfine-structure resolution (see Fig. 3A). This gets even worse for a dimer of bacteriochlorophylls like P865˙+, the primary-donor cation in RC protein complexes, for which only a Gaussian envelope EPR line can be observed, with no indication of hyperfine structure at all. ENDOR-in-solution and TRIPLE21 at elevated temperatures are superior to EPR in this respect because the linewidths are smaller, owing to the motional averaging of anisotropic hyperfine interactions in fluids, so that several isotropic hfc’s from the magnetic nuclei in the radicals can be obtained (see Fig. 3B). For BChl a˙+ up to 12 proton hfc’s could be determined and assigned to molecular positions.31,32 Furthermore, all four nitrogen hfc’s could be obtained by ENDOR in 15N-labeled BChl a˙+,32–34 whereas in the 14N-containing species only up to three 14N hfc’s were resolved.31 First liquid-state ENDOR and TRIPLE experiments on P865˙+ in RCs from Rb. sphaeroides under physiological conditions (water, 25 °C), carried out in this laboratory35,36 showed that the overall rotation of the RC protein is fast enough in this environment to effectively average out the hyperfine anisotropy of the various nuclei in the radical. Consequently, a highly resolved ENDOR spectrum was obtained, and at least seven proton hfc’s could be extracted (see Fig. 3C). By comparing the hfc’s of monomeric BChl a˙+ and light-induced P865˙+, the “special pair” hypothesis of Norris et al.37 and Feher et al.38 could convincingly be confirmed, which had been deduced from solid-state EPR and ENDOR experiments.

Well below room temperature, the overall rotation of the protein complex becomes so slow that powder-type spectra are obtained. Nevertheless, under certain circumstances which depend on the magnitude of the anisotropy of the interactions in the spin Hamiltonian in comparison to the inhomogeneous line width, even from disordered powder-type EPR spectra single-crystal like information can be extracted by applying orientation-selective ENDOR.39–42 In the case of transition-metal complexes the hyperfine anisotropy may provide this orientation selectivity from the entire orientational distribution of the molecules. In the case of organic radicals with small hyperfine interactions one has to resort to the anisotropy of the Zeeman interaction which can become large enough at high B0 fields to provide the desired orientation selectivity for ENDOR experiments42 (see below). The best approach for elucidating molecular structure and orientation in detail is, of course, to study single-crystal samples. Unfortunately, to prepare them is often difficult or even impossible for large biological complexes.

2. Why high-field EPR and ENDOR?

We now turn to high-field/high-frequency EPR and ENDOR experiments in more detail to show what can be additionally learned about biological systems when going beyond conventional X-band EPR/ENDOR. From the spin Hamiltonian (eqn. (1)) one sees that some interactions are magnetic field-dependent (the Zeeman interactions), while others are not (the hyperfine interactions). Obviously, in complex biological systems it will be necessary to measure at various field/frequency settings in order to separate these interactions from each other. Up to now, continuous-wave (cw) and time-resolved (tr) EPR studies on biological samples have been concentrated on standard X-band frequencies (9.5 GHz), extensions to lower (S-band, GHz) and higher microwave frequencies (K-band, 24 GHz; Q-band, 35 GHz; W-band, 95 GHz, or even higher frequencies) are exceptions.

For low-symmetry systems, particularly in frozen solution samples, standard EPR suffers from low spectral resolution due to strong inhomogeneous line broadening. Such problems arise, for instance, because several radical species or different magnetic sites of rather similar g-values are present or a small g-tensor anisotropy of the paramagnetic system does not allow canonical orientations of the powder EPR spectrum be observed. In such a case, even X-band ENDOR may not be sufficiently orientation-selective to provide single-crystal type information of the hyperfine structure. For improving the spectral resolution by high-field EPR, we have to define a lower limit of the mw frequency and corresponding magnetic field B0. For “true” high-field EPR experiments, properties of the spectrometer have to be related with properties of the sample: For all cases of delocalized spin systems, in which unresolved hyperfine interactions dominate the inhomogeneous EPR linewidth, a true high-field experiment must fulfil the condition

	(5)

i.e., the anisotropic electron Zeeman interaction must exceed the inhomogeneous line broadening, ΔBhf1/2. For example, for deuterated samples, Q-band EPR might already fulfil this condition in the case of semiquinone radicals with rather large g-anisotropy, whereas for protonated samples with inherently larger linewidths, it does not. On the other hand, in the case of chlorophyll ion radicals, due to their small g-anisotropy, even W-band EPR might not meet the high-field condition for protonated samples. Then deuteration of the sample will be necessary or, as an alternative, to further increase the mw frequency and B0 field, for instance by resorting to 360 GHz EPR (see below). Except for transition-metal complexes, most bioorganic systems have g = 2, and relative g variations and anisotropies, Δg/g, rarely exceed 10−4–10−3. At X-band frequencies the corresponding ΔB0 values are only 0.03–0.3 mT, which can easily be masked in disordered samples with typical EPR linewidths around 1 mT. Fortunately, for many non-metallic protein systems the increase of ΔB0 with increasing Zeeman field directly translates into an increase of spectral resolution, because often no noticeable line broadening occurs with increasing B0. For the primary donor cation radical in RCs from Rb. sphaeroides, for example, up to 24 T were applied,43 and g-strain broadening was found to be negligible. Double-resonance extension to high-field ENDOR then has the additional advantage of providing single-crystal-like hyperfine information in the reference frame of the g-tensor, even from disordered samples with very small g-anisotropy. Fig. 4 shows, as an example, the Zeeman magnetoselection of W-band EPR and ENDOR on the radical anion of ubiquinone-10 in frozen perdeuterated propan-2-ol at T = 115 K.42 Pronounced orientational selectivity of the g-tensor components is achieved by W-band EPR which is inaccessible by X-band EPR (compare Fig. 4A and 4B). This is further exploited by performing W-band pulsed Davies-type ENDOR at the well-separated B0 values corresponding to the canonical g-tensor peaks to obtain dominant proton hfc’s (for the Davies-ENDOR pulse sequence, see Fig. 6 later). At least for the gxx and gzz canonical field positions, the ENDOR spectra are single-crystal-like with narrow lines. Accordingly, the representations of the orientational selections of molecules show narrow distributions (see right side of Fig. 4C). These representations follow from the simulations of the spectra on the basis of the spin Hamiltonian of eqn. (1). The orientational distribution of molecules is considerably broader for the gyy value, which reflects its still insufficient resolution by W-band EPR. When varying the solvent (protic and aprotic, with and without perdeuteration), characteristic changes of hyperfine- (predominantly along the y-direction) and g-tensor components (predominantly along the x-direction) could be discerned. They are attributed to hydrogen-bond formation at the lone-pair orbitals of the oxygens.42

Also for high-spin systems (S > 1/2) such as enzymatic proteins with several transition metal cofactors, high-field EPR might be of advantage. An example is the nitrogenase MoFe protein that is spin-coupled to S = 3/2.44,45 For such systems a “fine-structure” term Ĥss has to be added to the spin Hamiltonian, i.e., eqn. (1) has to be extended by

Ĥss/h = Ŝ··Ŝ − JŜ1·Ŝ2	(6)

with the total spin S1 + S2 = S to consider the interactions between a pair of paramagnetic ions with spins S1 and S2 coupling to the total spin S. Here  is the traceless zero-field splitting tensor, and J is the isotropic exchange interaction parameter (in standard convention J < 0 corresponds to antiferromagnetic coupling, J > 0 to ferromagnetic coupling). In the principal axis system of , the anisotropic part in eqn. (6) is normally rewritten in terms of the zero-field parameters D, E:

	(7)

with D = (X + Y)/2 − Z, E = −(X − Y)/2, where −X, −Y, −Z are the zero-field energy values.

Although in the spin Hamiltonian the fine-structure term is not field dependent it leads, in combination with the electron Zeeman term, to field-dependent mixing of the electron spin eigenfunctions. When the external field values B0 lead to electron Zeeman splittings comparable in magnitude to the fine-structure splittings, the mixing gets very severe. This intermediate region requires elaborate calculations to analyse the EPR spectrum. Hence, one more reason to perform high-field EPR is to simplify the spectrum analysis for high-spin systems.

Moreover, it might happen that EPR transitions of high-spin systems with large zero-field splittings cannot be observed at all at X-band frequencies because the energy of the mw quantum is too small. For such cases the higher quantum energy of high-frequency microwaves can drive the transitions. An example for such a biological high-spin system is metmyoglobin with S = 5/2 ferric heme, for which EPR transitions at 130 GHz became observable that were undetectable at X-band due to the large zero-field splitting.46 Another benefit is that the EPR lines of high-spin systems usually get narrower at higher magnetic fields than in X-band EPR because of second-order effects: If we take Mn2+ centres (S = 5/2, I = 5/2) in disordered protein samples as an example, the EPR transitions are strongly broadened by contributions from the zero-field tensor. Their linewidth, ΔB1/2, is determined by second-order contributions from the zero-field coupling D, ΔB1/2 ∝ D2/B0, and “needle sharp” manganese hyperfine lines are observed in 95 GHz high-field EPR spectra of Mn2+ protein complexes even in disordered samples. A challenge for future high-field EPR experiments is the water-splitting manganese complex in photosystem II of green plants. Structure and enzymatic function of the Mn cluster site in PS II are still subject of controversial discussions, and additional information from high-field EPR is desirable. Up to now, however, no high-field EPR signals from the PS II Mn cluster have been reported, but promising multi-frequency high-field EPR experiments on relevant binuclear Mn(III)Mn(IV) mixed valence complexes have already been performed.47,48 The line-narrowing behaviour is often exploited in high-field EPR by using Mn2+ ions doped into MgO powder as reference sample for precise g-factor and field-calibration measurements.49

With these advantages in mind, over the last 20 years a small number of dedicated laboratories met the technological challenge to construct mm and sub-mm high-field EPR and ENDOR spectrometers, thereby opening a promising new research area. The physical principles and technical aspects have been published by the laboratories involved. The pioneering high-field EPR work was done by Ya. S. Lebedev in Moscow.50 Appropriate references to these laboratories are included in recent overview articles, for instance.14,51–61 Some details of the laboratory-built 95 GHz and 360 GHz EPR/ENDOR spectrometers at FU Berlin are presented below (see also refs. 49, 62–64).

3. High-field EPR and ENDOR spectrometers developed at FU Berlin

3.1 The 95 GHz spectrometer

The development of a W-band (95 GHz) EPR spectrometer at FU Berlin started at the beginning of the 80’s with the aim to construct a multi-purpose high-sensitivity machine. The continuous improvement of the first spectrometer version65 went in parallel with the improvements of commercial superconducting magnets, mm microwave components, control instrumentation as well as computer technology in the last 20 years.49,58,62

Increasing the mw frequency for mm high-field EPR causes two main technical problems: First, the required magnetic field strength increases proportionally to the frequency increase and, at 95 GHz, reaches, for molecules with g = 2, a value of 3.4 T. This field strength cannot be generated any more by resistive electromagnets, but requires superconducting cryomagnets. Second, at 95 GHz and higher frequencies, the performance of the mw components deteriorates, especially of mw sources and amplifiers. Additionally, higher precision is required for manufacturing the mechanical parts, such as waveguides and resonator probeheads due to the decreasing wavelength.

In our 95 GHz spectrometer a superconducting magnet (Cryomagnetics, USA) provides a B0 field of up to 6 T at a warm-bore diameter of 114 mm, see Fig. 5. The field homogeneity of 1 ppm (about 3 μT at 3.4 T) over a spherical volume of 10 mm diameter is sufficient for EPR applications because of the intrinsic EPR linewidth (typically 0.1 mT or larger for solid-state samples at such high magnetic fields) and the small sample size (typically smaller than the mw wavelength of 3 mm). The magnetic field can be linearly swept up to ±0.1 T by controlling the current in additional superconducting sweep coils inside the magnet Dewar.

In the early days of high-field EPR development the challenging task was to fight the low performance of mw components in the mm and sub-mm regions. This situation has recently improved by novel developments in mm-wave radar and radio-astronomy projects, and high-performance mw components are becoming commercially available.57,58,63,66 Nowadays, W-band EPR spectroscopy has been established as a powerful technique which can favourably compete with conventional X-band EPR in terms of detection sensitivity, reaches comparable pulse-scheme versatility, and is clearly superior in spectral and time resolution.

Fig. 5 shows the detailed block diagram of the last upgrade of the EPR microwave bridge developed at FU Berlin. The heterodyne detection bridge works at an intermediate frequency (IF) of 4 GHz. It is provided from the dielectric-resonator oscillator (DRO) which allows convenient manipulation of phase and amplification. The signal-oscillator output power at 95 GHz is used to irradiate the sample and, when the EPR resonance condition is met, the reflected power from the resonator is mixed with the 99 GHz local-oscillator signal. This is obtained by up-conversion of the 95 GHz signal-oscillator power with the 4 GHz DRO. Down-conversion to 4 GHz IF is achieved by mixing the 95- and 99 GHz microwaves. This 4 GHz power is amplified with a low-noise preamplifier and then down-mixed with the same 4 GHz DRO as local oscillator. The resulting quadrature signal is guided either to a lock-in amplifier or, after optional further amplification, to a fast digitizing scope. The newest developments in semiconductor microwave technology allowed us to replace the klystron, which was used many years as an optimum 95 GHz source,49,62 by a specially designed solid-state oscillator source. It utilizes four low-frequency oscillators, three fixed-frequency, temperature-stabilized DROs and one voltage-controlled oscillator (VCO), operating at 7.3 GHz. The frequency of this mw field is up-converted by a factor of 13 using an IMPATT-active frequency multiplier. An additional gain of the 95 GHz power could be obtained by a two-stage cw IMPATT amplifier in the transmitter channel.

Fig. 6 shows selected irradiation schemes of the various cw and time-resolved EPR and ENDOR experiments on laser-pulse initiated transient intermediates that can currently be performed with our W-band spectrometer. For the experiments several probehead configurations, including multimode Fabry–Perot resonators and single-mode TE011 cylindrical cavities (see Fig. 7), have been designed and optimized for EPR or ENDOR on a variety of different samples, e.g., lossy aqueous or non-lossy frozen-solution samples.49,62,67–69 In cw EPR experiments, a 10 kHz modulation of the magnetic field is achieved by additional coils placed around the EPR cavity. They provide up to 1 mT modulation amplitude at the sample. For both cw and pulsed ENDOR experiments, the TE011 cavity is slotted to reduce eddy currents induced by the rf field of the ENDOR coil placed around the cavity. For proton ENDOR the rf circuit is tuned to and matched for 140 MHz (with 40 MHz bandwidth) corresponding to the proton Larmor frequency at 3.4 T. An rf power of 300 W in cw and 2.5 kW in pulse mode is available. It yields 2 μs long rf π-pulses in proton ENDOR experiments. ENDOR on other nuclei can also be performed by using a short-cut rf circuit optimized for 40 MHz central frequency (with 30 MHz bandwidth).

For light-excitation experiments either the Nd-YAG laser beam (355 nm, 532 nm wavelengths) or the light of a 100 W halogen lamp is guided to the cavity by means of a quartz fibre of 0.8 mm diameter. Temperature control of the sample between 90 and 350 K is achieved by a stabilized nitrogen gas-flow system in the probehead. In addition, a continuous-flow system for liquid solutions to be pumped through the W-band cavity allows to perform photochemical experiments on systems with irreversible sample depletion.70 For single-crystal samples to measure the complete rotation patterns, a W-band Fabry–Perot resonator has been constructed containing a goniometer with three rotation axes.67

For small single-crystal samples in the TE011 cavity, the cw EPR sensitivity is about 108 spins/mT at 1 Hz detection bandwidth, i.e., about two orders of magnitude higher than at standard X-band frequencies.71 This number corresponds to a minimum detectible concentration of a nitroxide spin label in frozen solution of 0.1 μM (200 K, sample tube ID 0.6 mm, active sample volume 1 μl). For direct-detection time-resolved (tr) EPR measurements the time resolution of the heterodyne detection channel of about 2 ns and the short ringing time of the EPR cavity, QL(πν)−1, of about 6 ns (cavity loaded quality factor QL = 2000 at ν = 95 GHz) provide a time resolution one order of magnitude higher than with X-band tr EPR. In contrast to pulsed X-band spectrometers with their high-power TWT mw amplifiers, the mw power required for pulsed W-band experiments is drastically smaller due to the small cavity dimensions. The shortest measured π/2-pulse length of 22 ns is achieved with a 95 GHz power of only 100 mW at the TE011 cavity (QL = 2500). For comparison, to obtain the same pulse length in a standard X-band TE102 cavity with the same QL, about 20 W mw power incident on the cavity is required. Another advantage of a pulsed W-band EPR spectrometer is its short dead time. This time is most relevant for practical applications. It is defined as the time interval from the moment when the mw power is switched on or off to the moment when the EPR signal can be recorded without any distortions. At W-band the dead time is more than an order of magnitude shorter than at X-band. This is due to the much lower mw power required for sample excitation at 95 GHz and the higher ringing frequency (shorter ringing time) of the W-band cavity. For our W-band spectrometer configuration with quadrature detector, the dead time is measured to be 10QL/1000 ns when using the full excitation mw power. The technical specifications of the present version of our TE011-cavity based W-band spectrometer ought to be compared with the most recently published specifications of the new W-band spectrometer at Cornell.66 This high-power pulsed spectrometer is based on a Fabry–Perot resonator probehead and incorporates an “extended interaction klystron” (EIK) as microwave source which is used to amplify the mm-wave pulses to the 1 kW power level.66

3.2 The 360 GHz spectrometer

The 360 GHz EPR spectrometer built at FU Berlin has been previously described in detail.64,72 The most notable difference to an EPR spectrometer operating at lower fields and frequencies, for example at X- or W-band, is the use of quasi-optical mw components of the heterodyne mw bridge and a corrugated waveguide in the transmission line to the probehead. This is indispensable, since the losses in conventional fundamental mode waveguides become prohibitive at mw wavelengths in the sub-mm range, while in a quasi-optical setup the use of free-space Gaussian beam propagation leads to very low losses.

A further advantage of the Gaussian beam optics is that it greatly facilitates an induction-mode design of the mw bridge (see Fig. 8). In an induction-mode detection scheme, the EPR signal is discriminated from the excitation microwave by letting it excite orthogonal degenerate resonator- and transmission-line modes. In our spectrometer, the different modes are provided by the perpendicular polarization of the mw Gaussian beams. They are separated by a wire-grid polarizer yielding an overall isolation of ∼30 dB. Elliptical mirrors with off-axis beam alignment are chosen as focusing elements in favour of lenses because they do not generate standing waves in the system and have lower insertion losses. The excitation mw at ν = 360.03 GHz is generated by a phase-locked tripled 120.01 GHz Gunn source (Farran Technology, Ireland). In the heterodyne detection channel, a sub-harmonic mixer with a doubled 90.31 GHz local oscillator (Farran) down-converts the signal reflected from the probehead to an IF signal of 1.21 GHz with a subsequent power-detection or quadrature-detection stage.

Inside the magnet warm bore, the limited space prevents the use of reflective focusing optics. Instead, we employ a circular corrugated waveguide. The incoming linearly polarized Gaussian beam couples with high efficiency to the hybrid HE11 mode of the corrugated waveguide which also has linear polarization and a Gaussian intensity profile. Both the quasi-optical components and the corrugated waveguide exhibit low cross-polarization that is essential for the induction-mode scheme. To couple the microwave into the resonator, the corrugated waveguide has a tapered bottom end.

The original brass probehead contains a semi-confocal Fabry–Perot (FP) resonator (see Fig. 9), consisting of a free-standing copper mesh that constitutes the flat mirror, and a spherical gold-coated quartz mirror. The FP resonator operates in a cylindrically symmetric TEM006 mode. It is set up in reflection with the microwaves coupled to the resonator through the mesh, and the sample deposited onto the surface of the spherical mirror. The resonator can be tuned via three straight rods by a micrometer screw on top of the probehead. Light excitation of the sample can be achieved by introducing a quartz light fibre close to the coupling mirror to illuminate the sample through the mesh. A new probehead for low-temperature experiments was constructed recently. The body of the probehead containing the FP resonator is machined from TUFNOL (Tufnol Ltd, UK), a cotton reinforced epoxy laminate with very good machining and thermal properties. It allows quick cooling of the cryostat while avoiding electromagnetic shielding or heating problems typical for metallic probeheads. This is essential for ENDOR experiments. The holders for the coupling mesh and for the modulation and ENDOR coils are fabricated from MACOR (Corning Inc., USA), a machinable glass ceramic. Due to the compact design of the probehead, B0-field modulation amplitudes up to ∼3 mT can be applied without generating excessive microphonics.

The magnet is a superconducting Teslatron-H system (Oxford Instruments, UK) with a B0-field strength of up to 14 T (12.9 T for g = 2 systems), a homogeneity of 3 ppm in a 10 mm sphere and a warm-bore diameter of 88 mm. The integrated superconducting sweep coil provides a sweep range of ±0.1 T with the main coil in persistent mode. The static-flow cryostat (Oxford) of the spectrometer (inside diameter 62 mm) supports temperatures between 4 K and room temperature. Since at B0 = 12.9 T the temperature corresponding to the Zeeman energy is 17.3 K, high Boltzmann polarization is achieved already at higher temperatures compared to low-field EPR. Since the current-field relation of the superconducting sweep coil is non-linear, the Mn2+ ion in a MnO/MgO polycrystalline matrix has been used as field standard for calibration of B0.49,72 The standard is dispersed as a powder into a polystyrene matrix and placed directly adjacent to the sample on the Fabry–Perot mirror to minimize effects of residual field-inhomogeneities and to prevent standard and sample from mixing.

The overall detection sensitivity of the spectrometer has been measured to be 1.5 × 1010 spins/mT at 1 Hz detection bandwidth by evaluating the signal-to-noise ratio of a known number of spins in a Mn2+ sample. Because of this high detection sensitivity the 360 GHz spectrometer is particularly suited to resolve extremely small g-anisotropies of samples that are available only in minute quantities. Prime examples, site-specific mutants of the chlorophyll cofactor radicals in photosynthetic reaction centres, are presented below.

The basic cw version of the 360 GHz spectrometer has been extended recently for operation both in pulsed EPR and ENDOR modes. For ENDOR measurements, the probehead can be fitted with an additional rf circuit, consisting of a single loop coil around the FP resonator and an rf-oscillator circuit with tuneable capacitors to allow for impedance matching of the ENDOR coil. The new TUFNOL resonator construction has been optimized to reduce heating effects that were disturbing the experiments in our earlier version of the ENDOR setup, with first tests in progress.

For pulsed EPR operation, the multiplied cw Gunn-diode source can be exchanged for a novel pulsed vacuum-tube source, a 360 GHz Orotron (Gycom Ltd, Russia). At its core, the Orotron is a Smith–Purcell non-relativistic free-electron maser embedded in a high-quality Fabry–Perot resonant structure to achieve high output power and frequency stability. The design employed in our novel setup extends previous cw Orotron constructions by introducing a high-voltage gating grid close to the electron-beam generating cathode. This opens the possibility to generate pulsed microwave radiation by selectively blocking the electron beam via pulsed application of high voltage to the gating grid. A detailed description of the operating principle and technical realisation of our pulsed Orotron has been presented in a recent publication.73

4. Computational methods for data interpretation

The development of multifrequency EPR spectroscopy has provided a powerful tool to the exploration of molecular structure and dynamics complementary to other strategies. Despite this development the explorer needs the additional aid of modern quantum chemistry theory. This statement applies also to reliable computational methods to calculate the interaction parameters of the spin Hamiltonian, in particular electron-nuclear hyperfine-, quadrupole- and g-tensor components. It is this combination of new technologies for measuring different molecular parameters and the enormous increase in computing power that now allows biophysicists to describe quantitatively proteins in actions, and to visualize complex biosystems in their three-dimensional structure and functional dynamics. But even today the level of molecular theory chosen remains to be a compromise between reliability of the predictions and affordable computational time, and the compromise is dictated by the size of the molecular system.

In comparison to hyperfine tensors the theory for g-tensor calculations of complex molecular systems was rather neglected for a long time because of missing experimental g-tensor data of sufficient precision. Given the improved spectral resolution and information available from high-field EPR, it has become increasingly desirable now to calculate in particular g-tensors on a thorough quantum-mechanical basis with sufficient accuracy and reliability. This impetus, combined with improvements in quantum-chemical programs and computer hardware, has in the past years tremendously stimulated the development in quantitative calculations of g-tensors.

After earlier INDO or AM1 type semiempirical quantum-chemical approaches to calculate g-tensors of bioorganic systems,74–76 our laboratory has recently joined the few existing groups in the high-field EPR community that apply modern and quite pretentious density functional theory (DFT) methods to g-tensor calculations. One of the two first modern DFT approaches to the calculation of magnetic resonance parameters has been reported by van Lenthe and coworkers77 and is implemented in the widely used Amsterdam Density Function (ADF) code.78 A serious limitation of this “zero order regular approximation” (ZORA) is the use of a spin-restricted treatment which is inappropriate for the calculation of nuclear hfc’s, the evaluation of which along with the g-tensor is at least equally important.

An alternative modern DFT approach which allows the calculation of hfc’s and g-tensors at the same spin-unrestricted levels, has been developed by Schreckenbach and Ziegler.79 Two related implementations have been reported by Kaupp and coworkers80 and by Neese.81 These three approaches do all employ second-order perturbation theory based on the unrestricted Kohn–Sham equations.82 They incorporate relativistic contributions to the g-tensor at different levels of approximation. The SZ-method79 has the distinct advantage of using gauge including atomic orbitals (GIAOs) or “London orbitals”.83 These orbitals avoid the undesirable gauge dependence (choice of origin) in the calculation of g-tensors introduced by the orbital Zeeman operator.

In our recent study of g shifts in the primary donor P˙+ in Rb. sphaeroides mutants (see below), we used the SZ-method at the “relativistic scalar Pauli” level employing double-zeta atomic basis sets and the Becke exchange84 and Lee–Yang–Parr correlation85 options for the generalized gradient approximation (GGA) part of the exchange–correlation (XC) functional.79 Appropriate computer programs with extensive documentation are commercially available as part of the ADF program package.78

As a general statement one can say that the theory of g-tensor calculations is much more involved for organic systems with many carbon atoms and strongly delocalized electrons than for systems with dominant heavy-atom contributions, such as metallo-proteins, quinones or nitroxides.86–88 Therefore, in this review we will refrain from describing in length the details of g-tensor theory for strongly delocalized systems, like the dimeric chlorophyll donor in photosynthetic systems, and rather present only the results of our computation. On nitroxide-labelled proteins, on the other hand, with strong localization of the unpaired electron on the N and O atoms, we will sketch some more details of the theory behind the computations (see below).

5. Characterization of molecular switches in paradigmatic proteins

In this overview we focus on a selection of protein systems that were previously crystallized and for which high-resolution X-ray structures have been made available by now: bacterial photosynthetic reaction centres (RCs) for light-induced electron transfer across the membrane; bacteriorhodopsin (BR), the light-driven transmembrane proton pump; colicin A, the transmembrane ion-channel forming bacterial toxin. These proteins have been characterized in detail over the last years by powerful spectroscopic techniques including, for example, ultra-fast laser spectroscopy,89 FT-IR,90–93 solid-state NMR94–96 and multifrequency EPR.53,56,97–99 In addition, sophisticated theoretical studies had been performed to elucidate their light-induced electron and proton transfer characteristics. Hence, these proteins represent paradigm systems of general interest. They are well suited for new multifrequency high-field EPR experiments to study functionally important transient states during biological action. Furthermore, we prefer EPR as the method of choice because of the accessibility of functionally important time windows for the transient states. These states are either directly detectable by EPR by their transient paramagnetism as, for instance, the radical-ion and radical-pair states of cofactors involved in photosynthetic electron transfer in RCs or, more indirectly, by site-specific nitroxide spin-labelling of the protein as, for instance, in site-directed mutants of BR and Colicin A. It will be shown that high-field EPR experiments allow specific protein regions to be identified and characterized as molecular switches for vectorial transfer processes across membranes. Such switches are envisaged to function by controlling the electronic structure of cofactors, for example by invoking weak cofactor-protein interactions via H-bonds, by polarity gradients or even by substantial cofactor and/or helix displacements. RCs and BR are examples of involving molecular switches of this kind to control their vectorial transmembrane electron and proton transport. Colicin A toxin, on the other hand, requires massive protein refolding for transmembrane channel formation with subsequent ion depletion and cell death.

The high-field EPR work on the RC, BR and Colicin A paradigm proteins presented in this overwiew was performed in collaboration with the research groups of W. Lubitz (Mülheim/Ruhr) for the RC studies and of H.-J. Steinhoff (Osnabrück) for the BR and Colicin A studies.

5.1 Bacterial photosynthetic reaction centres (RCs) of Rb. sphaeroides

Photosynthesis is the most important process that enables life on Earth by converting the energy of sunlight into electrochemical energy needed by higher organisms for synthesis, growth and replication. The so-called primary processes of photosynthesis are those in which the incoming light quanta, after being harvested by “antenna” pigment/protein complexes and channelled to the RC complexes by ultra-fast energy transfer, initiate electron-transfer (ET) reactions between protein-bound donor and acceptor pigments across the cytoplasmic membrane. The successive charge-separating ET steps between the various redox partners in the transmembrane RC have very different reaction rates, kET. The lifetimes, t1/2 = (kET)−1, of the transient charge-separated states range from less than 1 ps for neighbouring donor–acceptor pigments to more than 1 ms for large donor–acceptor separations on opposite sides of the membrane (ca. 40 Å). The cascade of charge-separating ET steps of primary photosynthesis competes extremely favourably with wasteful charge-recombination ET steps thereby providing almost 100% quantum yield. The largest impact of photosynthesis on life on Earth is due to green plants and certain algae in whose RCs a reversible catalytic ET photocycle occurs for which water serves as electron donor. Carbon dioxide is fixed in the form of carbohydrates, and oxygen gas is released as a by-product thereby stabilizing the composition of the Earth’s atmosphere.

Three billion years before green plants evolved, the Earth and its atmosphere were very different from what they are today. Nevertheless, photosynthetic energy conversion could be achieved by certain bacteria, for instance the purple bacterium Rhodobacter (Rb.) sphaeroides. These early photosynthetic organisms are simple, one-cellular protein-bound donor–acceptor complexes that contain only one RC for light-induced charge-separation. They cannot split water, but rather use hydrogen sulfide or organic compounds as electron donors to reduce CO2 to carbohydrates with the help of sunlight and bacteriochlorophyll as biocatalyst.

In Fig. 10 the structural arrangement of the RC of Rb. sphaeroides is shown according to the 3 Å resolution X-ray structure.100–102 The cofactors are embedded in the L, M, H protein domains forming two ET branches, A and B. The RC of the carotinoid-less strain R26 of Rb. sphaeroides contains nine cofactors: the primary donor P865 “special pair” (a bacteriochlorophyll a (BChl) dimer), two accessory BChls (BA, BB), two bacteriopheophytins a (BPhe: HA, HB), two ubiquinones (QA, QB), one non-heme iron (Fe2+).

As a dominant motif in the evolution of photosynthetic bacteria, an approximate C2 symmetry prevails for the cofactor arrangement in the RC. It is intriguing that, despite the apparent two-fold local symmetry of the cofactor arrangement, the primary ET pathway is one-sided along the A branch, as indicated by the arrows in Fig. 10. The origin of this “unidirectionality” enigma of bacterial ET is not yet fully understood despite the numerous elaborate studies, both experimentally and theoretically, performed over the last 10 years. As a matter of fact, the high-resolution X-ray structure reveals that C2 symmetry does not hold for the protein environment of the cofactors, but is broken by different amino acids along the two ET branches. Thereby, the relative energetics and H-bond properties of the cofactors along the two branches will be different. They control the participation of cofactors as intermediate states in the ET cascade. From numerous genetic engineering experiments it turned out that such a “unidirectionality switch” in the RC system is very robust against point mutations of the amino acid environment of the cofactors.103 It was only recently demonstrated that specific double-site mutations in the vicinity of the primary donor and an accessory BChl can significantly change the partition of ET between the A and B branches.104,105 This is strong indication that in the wild-type system the breakage of symmetry in the ET pathways is largely due to the finely tuned energetics and electronic couplings of the primary donor and the intermediary acceptors.

For further reading about structure and spectroscopy of photosynthetic reaction centres, a large number of overviews is available, for example refs. 106–108.

5.1.2 Results of g-tensor computations of P˙+

All previous experimental results on the M202 mutants from X-ray structural analyses,100,129,130 ENDOR/TRIPLE110 and optical and FT-IR measurements131 strongly indicate that the primary donors of both the HL(M202) and the HE(M202) mutants are heterodimers of bacteriochlorophyll a and bacteriopheophytin a, BChl : BPhe, in contrast to the homodimeric species (BChl)2 of wild-type and R26 RCs. Both EPR studies and energetic considerations then lead to the conclusion that this structural modification of the primary donor in its oxidized paramagnetic state is characterized by an almost complete localisation of the spin density on the BChl(L) half bound to the L protein subunit.110,116 From this it seems straightforward to expect a shift of the g-tensor components of the mutants towards those of an isolated BChl˙+ monomer, as was also concluded earlier from a comparison of HL(M202) with R26.113 The g-component shifts observed in our 360 GHz EPR experiments, however, are very surprising in this respect: While for P˙+ of HL(M202), as noted above, the overall g-tensor anisotropy Δg = (gxx − gzz) becomes smaller (more like that of the monomer), for HE(M202) Δg increases to a value considerably larger than that for R26 (see Table 1).

This unexpected behaviour of Δg strongly suggests a local structural change of BChl(L) as a consequence of the M202 ligand mutations. This conclusion is also supported by the observation of considerable rearrangements in the spin density distributions on BChl(L) in the two mutants. For example, previous ENDOR/TRIPLE studies110 yielded quite different ratios RL = a(121L)/a(21L) of the isotropic methyl proton hfc’s at positions 121 and 21 on BChl(L) (see Fig. 3):

RL(HL(M202)) = 1.29; RL(R26) = 1.44; RL(HE(M202)) = 1.66.

Earlier semiempirical MO studies132 and experimental ENDOR/TRIPLE results on P˙+ in RCs115,118,133 suggest that such a variation of RL might originate in different torsional angles θac(L) of the acetyl group at position 3 with respect to the plane of the adjacent aromatic ring A (see Fig. 3). Different values of θac(L) for R26, HL(M202) and HE(M202) are, in fact, expected since the acetyl-methyl groups point towards the region of the molecular centres of the dimeric counter parts. This is BChl(M) in the case of R26, and BPhe in the case of HL(M202) and HE(M202), where the changes in the chemical structures due to the mutations occur. Obviously, the experimental ratios RL, as given above, correlate with the corresponding values of Δg (see Table 1). This observation urged us to perform theoretical calculations of the g-tensors of the three P˙+ species as a function of θac(L). Since earlier semiempirical MO calculations based on AM1 or ZINDO/S wavefunctions74 failed to reveal any changes in Δg as a function of θac (but rather revealed changes in g-tensor axes orientation), we have applied more advanced relativistic DFT methods employing the ADF package,78 and the SZ-approach79 for the calculation of magnetic resonance parameters. Details of the employed packages can be found in our previous publication.114

The results are summarized in Table 2. The calculated values of RL show a marked monotonic increase with increasing values of θac(L). The calculated principal components gxx, gyy and gzz show only moderate quantitative agreement with the experimental values, something that is frequently observed for DFT-based calculations of g-tensor components.79,88,114,134 However, a correct qualitative prediction of trends of gii as a function of θac(L) should be provided by the present DFT results.

---

The essential conclusion from the current computational results is that our model predicts a range of increasing angles θac(L) between 0° and 30°–35° in which both RL and Δg increase parallel to each other, in accordance with our experimental results. The increase in Δg in this series is entirely due to a decrease of gzz. This is reasonable, since the rotation of the acetyl group perturbs the planarity of the electronic π-system and, therefore, primarily leads to changes in gzz. In the same angular region, the value of gyy − gzz also increases parallel to RL and Δg, in accordance with the experimental results (Table 1).

To model the remaining experimentally observed strong shift of the gxx component, one needs to take into account H-bonding interactions with the environment. An H-bond of the acetyl-group oxygen atom to the neighbouring histidine shifts gxx to larger values, and we conclude that the strength of this H-bond, i.e., the distance r(H⋯Oac), varies in the three mutant species according to r(HE) < r(R26) < r(HL). This conclusion cannot be tested directly by means of the X-ray structure because of the insufficient precision of atomic positions from X-ray crystallography. Further DFT calculations are in progress with the aim to gain a deeper insight into the general influence of polar side groups, superimposed electric fields and H-bond effects on g-tensor components in photosynthetic systems.

Also further 360 GHz EPR experiments with RC mutants, in which the H-bond network around the L-half of the special pair has been altered, are in preparation. The high sensitivity of very-high-field EPR to subtle changes in the structure and microenvironment of the ET chromophores should provide valuable insight into the influence of structural features and energetics on the effectiveness and directionality of the electron transfer steps.

5.2 Bacteriorhodopsin (BR)

Apparently, Nature has invented photosynthesis, i.e., the strategy to use sunlight as energy source to synthesize ATP, twice: In the photosynthetic reaction centre (RC) protein complex of purple bacteria this is initiated by light-induced primary electron transfer between bacteriochlorophyll and quinone cofactors, mediated by the protein microenvironment. In the bacteriorhodopsin (BR) protein complex this is set going by light-initiated primary proton transfer between amino acid residues, mediated by conformational changes of the only cofactor, the retinal.

BR is a 26 kDa protein complex located in the cell membrane of halophilic archae-bacteria such as Halobacterium salinarium. High-resolution (1.6 Å) X-ray crystallography coordinates are available for the ground-state structure153 (see Figs. 16 and 20). Seven transmembrane helices (A-G) enclose the chromophore retinal which is covalently attached to the amino acid lysine, K216, on helix G via a protonated Schiff base. Absorption of 570 nm photons initiates the all-trans to 13-cis photoisomerization of the retinal. The Schiff base then releases a proton to the extracellular medium and is subsequently reprotonated from the cytoplasm. Transient intermediates of this catalytic photocycle can be distinguished by the different absorption properties of the retinal, and a sequence of intermediates (J, K, L, M, N and O) has been characterized by time-resolved absorption spectroscopy.154 Double-flash experiments revealed that the M intermediate is divided into two substates, M1 and M2155,156 During this photocycle conformational changes of the protein (and the retinal) occur, as has been detected by a variety of experimental techniques (for a review see, e.g., ref. 157). The physiological role of such changes is discussed to ensure that release and uptake of protons do not occur from the same side of the membrane, but rather enable BR to work as a vectorial transmembrane proton pump. To this end, in wild-type BR conformational changes associated with the M1 to M2 transition are suggested to function as a “reprotonation switch” required for the vectorial proton transport. Hence, it is believed that during the lifetime of the M state the accessibility of the Schiff base for protons is switched from the extracellular to the cytoplasmic side of the membrane. Detailed analyses of the nature of the conformational changes, for example by neutron diffraction,158,159 electron microscopy,160–163 X-ray diffraction,164,165 solid-state NMR166,167 or EPR spectroscopy,168–173 agree in the major changes to be localized at the cytoplasmic moieties of helices F and G. These helix movements in wild-type BR, which can be observed to be present already in the M intermediates with the Schiff base deprotonated, have been shown to result in an “opening” of the protein to protons on the cytoplasmic end of the transmembrane proton channel161 and, thus, should allow proton transfer to occur from the internal aspartic-acid proton donor, D96, to the Schiff base during the M to N transition. The reprotonation of D96 from the cytoplasm occurs during the recovery of the BR initial state. A detailed inspection of the structure of the unilluminated state of the protein reveals that certain amino acid side chains block the proton pathway from the cytoplasm to D96.161 The region between D96 and the Schiff base is largely nonpolar, packed with bulky amino acid residues. Hence, in this unilluminated state the Schiff base is effectively inaccessible to protons from the cytoplasm. In the light-driven M1 to M2 transition, this region is opened for access of protons to the Schiff-base nitrogen atom. However, the question remains, and is in fact still hotly debated,161,174 whether the large conformational changes observed in the photocycle of the wild-type and many BR mutants are a prerequisite for vectorial proton transport, i.e., really represent the proposed reprotonation switch.

This on-going discussion got strong momentum in favour of an alternative, or additional, conformational change mechanism, involving reduction of the retinal curvature that emerged from the results of recent FT-IR studies of the BR triple mutant D96G/F171C/F219L.174 This functionally intact BR triple mutant has been shown to provide already in the dark the full extent of the light-driven conformational changes of the protein helices at the cytoplasmic entrance of the putative proton channel.161,162 In other words: In the triple mutant the wild-type structure around the cytoplasmic proton channel is altered to such an extent that a pseudo M2 state is generated with the helix conformation of the late M intermediate, but which is stable in the dark. Surprisingly, FT-IR spectroscopy monitoring the helix movements during proton transfer in comparison with those of wild-type BR, revealed that efficient vectorial proton translocation in the photocycle of the BR triple mutant occurs in the absence of substantial conformation changes of the protein.174 These results indicate that the reprotonation switch operates even without large protein conformational changes, but rather involves retinal unbending. However, despite this evidence in disfavour of large protein rearrangement for the switch mechanism, it is likely that the last word has not been spoken on the subject. For example, on the biological function of the large conformational changes of protein helices that occur during the photocycle of wild-type BR. Obviously, additional spectroscopic experiments are needed on wild-type and strategically constructed BR mutants. They should distinguish themselves by sufficient sensitivity and resolution to follow also small conformational changes of protein and cofactor during the photocycle. In this respect, good news appears to be on the horizon with the combination of site-directed mutagenesis for spin labelling and high-field EPR for resolving structural changes of proteins at work.

For further reading about structure and spectroscopy of bacteriorhodopsin, a large number of overviews is available, for example.157,175

5.3 Colicin A bacterial toxin

Colicin A is a member of a family of bacterial toxins that form transmembrane ion channels. Seven members of the colicin family are currently identified and sequenced, denoted E1, A, B, K, N, Ia, Ib.188–190 They are water-soluble proteins, mostly about 70 kDa in size, and have varying homology among themselves. Colicin A kills unprotected cells of attacked organisms by inserting specific portions of protein sub-domains into the cytoplasmic membrane forming a voltage-gated ion channel. The open channel leads to electrical depolarization of the membrane and depletion of intracellular ion pools, which ultimately leads to cell death. The mechanisms of protein entry into membranes and transmembrane channel formation are fundamental aspects of membrane biology and toxicology. To understand these mechanisms on the molecular level is currently also of biomedical interest, since insertion of proteins into membranes and subsequent channel formation are common to many toxic proteins of bacterial pathogens found in organisms ranging from bacteria to humans, such as the diphtheria, tetanus and cholera toxins.189–193

The colicin toxins have to overcome the protecting barriers of the attacked cell and, hence, the mechanisms of cell killing involve three distinct functional steps:188,194 (i) binding of the toxic protein to a receptor at the surface of the outer membrane, (ii) translocation across the outer membrane and periplasm and (iii) insertion of specific helical segments of the toxin sub-domain into the cytoplasmic membrane to form the ion-conducting channel. Accordingly, Colicin A consists of three functional protein domains: (i) the central receptor domain, R; (ii) the N-terminal translocation domain, T, to penetrate the outer membrane (aided by receptor proteins of the target cell) and to traverse the periplasm (aided by several membrane translocation proteins of the target); and (iii) the C-terminal channel-forming domain, C, to penetrate the inner membrane which protects the cytoplasm of the cell.190,195 These distinct functions of the colicins are reflected by their three-dimensional shape. The X-ray crystallographic structure of a complete (T–R–C) Colicin Ia protein was recently determined to 3.0 Å resolution,196 now refined to 2.3 Å,190 and reveals a harpoon-shaped molecule, 210 Å long, with the three functional domains well separated from each other thereby overcoming the protection barriers of the attacked cell.

The isolated C-domain retains its channel-forming ability in aqueous solutions of artificial membranes,197,198 such as lipid vesicles. Hence, details of refolding processes of the C-domain can be studied by in vitro experiments (see below). The X-ray crystal structure of the 204-residue (21 kD) channel-forming C-domain of Colicin A in its water-soluble conformation is available to 2.4 Å resolution199 (see Fig. 22). Ten α-helices, eight amphiphilic, two hydrophobic, are arranged in such a way that the amphiphilic helices surround the hydrophobic hairpin (helices 8 and 9) deeply buried in the centre of the protein. Thereby the 8,9-hairpin is shielded from contact with the water solvent.

In view of the difficulties encountered with determining X-ray structures of transient protein states in action, numerous spectroscopic techniques are being used to study colicins on the molecular level. Despite many years of work on colicins by solid-state NMR,200,201 FT-IR,202 FRET203 and EPR194,204,205 the details of the refolding processes upon membrane association and channel formation are not yet known. Since Colicin A is now so well characterized, it may serve as a paradigm to answer an intriguing question of general interest in molecular biology and medicine: What is the mechanism to switch on the insertion of water-soluble pore-forming proteins into the nonpolar lipid environment of a membrane?

6. Conclusions and outlook

In this overview it is shown that modern multifrequency EPR spectroscopy at high magnetic fields provides detailed information about structure and dynamics of transient radicals and radical pairs occurring in biological electron and ion transfer processes. Thereby our understanding of the relation between structure, dynamics and function of molecular switches is considerably improved. This holds with respect to the fine-tuning of electronic properties of donor and acceptor cofactors by means of weak interactions with their protein and lipid environment, such as H-bonding to specific amino acid residues. This also holds with respect to the massive protein refolding in the course of transmembrane ion-channel formation.

We have restricted our overview to work performed in the high-field EPR laboratory at FU Berlin rather than providing an exhaustive review on advanced EPR in biophysical/biochemical research. Nevertheless, we claim that our conclusions are equally based on the work of our esteemed colleagues in the field. An excursion through biological magnetic resonance literature reveals a strong increase of high-field EPR applications with exciting news on nm and sub-nm structures of proteins and on the dynamics of their transient states while in biological action. These applications give new insights complementary to the results of other spectroscopic techniques, and some of them provide even unique information inaccessible by other “classical” spectroscopic techniques. This applies, for example, to large biosystems that lack long-range order, such as glasses and frozen solutions, for which distances and orientations of protein fragments and domains can be measured over a broad distance range, say 0.3 nm < r < 6 nm, even in time-resolved mode for short-lived intermediates in enzymatic reaction cycles. The limitations in high-field EPR instrumentation capabilities for short life times or relaxation times that existed until recently are currently being remedied.14,63,70,73 This refers to fast-flow probeheads for chemically instable transient70 or to sufficient output power of mm and sub-mm microwave sources for fast manipulation of the electron spins by short mw pulses.14,73 Consequently, wide avenues are being opened for fascinating new applications.

Below, some summarizing conclusions are given, relevant to biological systems in general:

1. Many organic cofactors in transfer proteins have only small g-anisotropies and, therefore, require much higher magnetic fields than available in X-band EPR to resolve the canonical g-tensor orientations in their powder spectra. Thereby, even on disordered samples, orientation-selective hydrogen bonding and polar interactions in the protein binding sites can be traced. This is important information complementary to what is available from high-resolution X-ray diffraction of protein single crystals.

2. In electron-transfer processes often several organic radical species are generated as transient intermediates. To distinguish them by the small differences in their g-factor and hyperfine interactions, high Zeeman fields are required and, hence, high-frequency EPR becomes the method of choice.

3. Often high-purity protein samples can be prepared only in minute quantities, for example of site-directed mutants or isotopically labelled cofactors. The problem of small concentration of paramagnetic states is even more dramatic for protein single crystals. Accordingly, to study them by EPR or ENDOR very high detection sensitivity is needed, which often can be accomplished only by dedicated high-field/high-frequency spectrometers. Optimum concentration sensitivity appears to be attainable in the W-band region, but at much higher mw frequencies a compromize has to be made between gain in resolution and loss of sensitivity which, for technical reasons, turns out to be unavoidable.

4. High-field/high-frequency cw EPR generally provides, by lineshape analysis, shorter time windows down into the ps range for studying correlation times and fluctuating local fields over a wide temperature range. They are associated with characteristic dynamic processes, such as protein, cofactor or lipid motion and protein folding.

5. Pulsed high-field/high-frequency EPR as, for example, two-dimensional field-swept electron spin-echo spectroscopy, provides real-time access to specific cofactor and/or protein motions in the ns time-scale. Motional anisotropy can be resolved which is governed by anisotropic interactions, such as hydrogen bonding along specific molecular axes within the binding site.

6. ENDOR at high Zeeman fields takes additional advantage of the orientation selection of molecular sub-ensembles in powder or frozen-solution samples. Thereby, even in the case of small g-anisotropies, ENDOR can provide single-crystal like information about hyperfine interactions, including anisotropic hydrogen bonding to the protein.

7. In metalloprotein high-spin systems, such as the Mn2+ proteins, the EPR spectrum analysis can be drastically simplified at high Zeeman fields due to the suppression of second-order effects. In case of large zero-field splittings, EPR transitions might not be observable at all at X-band, but become accessible at higher quantum energies of mm or sub-mm microwaves.

8. High-field EPR adds substantially to the capability of “classical” spectroscopic and diffraction techniques for determining structure-dynamics-function relations of biosystems, since transient intermediates can be observed in real time in their working states at biologically relevant time scales.

9. High-field EPR noticeably extends the applicability of the site-directed spin-labelling technique in bioengineering. Owing to the resolution of both the hyperfine- and g-tensor components of nitroxide spin probes, changes in the protein microenvironment, including conformational changes, can be identified as potential molecular switches for vectorial transmembrane transport processes in proteins at work. In this respect, the recent progress made in the quantum theoretical interpretation of the EPR observables, in particular of the g-tensor, has to be recognized.86–88

10. Most exciting developments in the theory and computation of magnetic resonance parameters of large biosystems are envisaged for the near future. High-field EPR instrumentation development will remain a challenging task for experimental physicists. Their exertion in measuring molecular parameters with high precision has cross-fertilized the efforts of theoreticians to interpret the experimental data in terms of molecular structure and dynamics, and vice versa. Cross-fertilization of ideas becomes also apparent in the efforts of molecular biochemists to construct point mutants with site-specifically altered properties. Such mutants are often strategically designed for elucidating the structure and dynamics of biosystems by EPR techniques.

On the horizon, non-evolutionary optimized protein chemistry appears feasible following reaction pathways that have been explored by advanced spectroscopic techniques. Such an approach will provide not only deeper insight into biological processes of utmost importance on Earth, such as photosynthesis, but will provide also new ideas how to take technical advantage of strategies developed by Nature. As a promising and appealing example the world-wide activities are mentioned to synthesize biomimetic model complexes for efficient photoinduced charge separation in analogy to evolutionary optimized primary photosynthesis. Interdisciplinary research on artificial photosynthesis will certainly continue to open new avenues of enquiry. Remarkable progress with biomimetic donor–acceptor complexes has been achieved during the last decade,207–212 and the high spirits of photovoltaic research based on organic compounds is obvious from glancing through specialist and general journals. Most striking in this respect is the recent report on successful integration of photosynthetic reaction-centre proteins in solid-state photovoltaic cells with high quantum yield and conversion efficiency.213

Imagine there is a breakthrough in realizing biomimetic photovoltaic cells with a capacity high enough to be attractive for practical applications. Multifrequency high-field EPR might have had a significant impact to make this dream come true. Advanced EPR techniques are principally able to provide the essential ingredients of structure-function relations for the electron-transfer intermediates of the photocycle. However, continuation of concerted efforts of biologists, chemists and physicists will be needed to achieve this goal. Fortunately, the high-field EPR community has an excellent record in interdisciplinary research. Expertise and instrumental capabilities of individual groups are often shared in cooperative projects to push molecular spectroscopy to new frontiers in biology, chemistry and physics. Besides photosynthetic reaction centres also the other two examples discussed in this review, the site-specifically spin-labelled bacteriorhodopsin proton pump and the colicin ion-channel forming toxin, should give a flavour of the potentiality of high-field EPR in characterizing the key factors that control vectorial transfer processes in proteins. Understanding those factors will enable the molecular engineering community to modulate such transfer processes by point mutations at will. Bright prospects for the future of advanced EPR spectroscopy, 60 years after Zavoisky’s discovery of the electron paramagnetic resonance phenomenon!

Acknowledgements

Over the years, numerous coworkers—students, postdocs, colleagues—from different parts of the world have contributed to our interdisciplinary high-field EPR work. We have greatly benefited from these collaborations and discussions, which were sometimes controversial, but were always stimulating. To all these coworkers we want to express our gratitude and appreciation. The expensive instrumental developments and spectroscopic research activities in our laboratory have enjoyed generous financial support by the Volkswagen-Foundation and the Deutsche Forschungsgemeinschaft (in recent years in particular by the DFG priority program SPP 1051 and the DFG collaborative research center SFB 498), which is gratefully acknowledged.

References