Accurate Extraction of Nanometer Distances in Multimers by Pulse EPR

Abstract Pulse electron paramagnetic resonance (EPR) is gaining increasing importance in structural biology. The PELDOR (pulsed electron–electron double resonance) method allows extracting distance information on the nanometer scale. Here, we demonstrate the efficient extraction of distances from multimeric systems such as membrane‐embedded ion channels where data analysis is commonly hindered by multi‐spin effects.

Nanometer distance restraints obtained through electronp aramagnetic resonance (EPR) have attracted increasing attention in structural biology. [1] Pulsed methods such as pulsede lectron-electron double resonance (PELDORo rD EER for double electron-electron resonance) [2] and double quantum coherence [3] (DQC)-based experiments have shown potential for measuring distances up to 10 nm and beyond. [4] Thesel ongrange restraints are very informativef or assigning ap rotein's conformational state. This is very valuable for large membrane protein complexes, where obtaining multiple X-ray crystal or NMR structures, necessarytodescribe changes during function, is challenging. However,c ombining as tructure with EPR measurements appears to be ap romising strategy for ion channels and transporters. [5] Many of these systems fold in an active state as dimers or higher-order multimers. Commonly,s ite-directed spin-labeling [6] of these systems is used for introducing one spin-label per protomer.F or homo-multimers this entails the number of spin-labels per protein complex n equals the number of monomers. This introduces an additional challenge for EPR distance measurements, as extracting all distances present in such multiply labeled nano-objects is complicated by multi-spin contributionstot he dipolar coupling. [7] The 4-pulseP ELDOR experiment is mostly used for distance measurements. [8] Briefly,o ne set of spins (A-spins) is probedb y ad etection pulse sequence while the dipolarc oupling is selectively introduced by inverting as econd set of spins (B-spins) with ap umpp ulse (usually placed on the most intense feature of the spectrum).
While ar egularization artifact and impropera mplitudes in the distance distribution of at etraradical were attributed to multi-spine ffects, [9] Jeschke et al. providedt he first systematic study using three-spin models ystemsa nd relieving the problem by reducing the probability of multiple excitation. [7] Similar experiments had been proposed for determining the numbers of interacting monomeric units [10] and had shown somei mprovement in applicationso na lbumin. [11] Explicit treatment of multi-spine ffects allowed quantitative simulations in tetrameric KcsA. [12] Further recent applications include the heptameric mechanosensitive channel of small conductance (MscS) [13] and the octamerico uter membrane protein Wza from E.coli. [14] In both cases only the modald istances( between one monomer and the next in rotational symmetry) were interpreted.T he other distance data were ignored as unreliable, af act attributed to truncation of time traces (i.e.,t oo short observation of the dipolar evolution). Broadening of the shortest distance and suppression of all other distances was later shown to be an intrinsicp roblemo fm ulti-spins ystems of C n symmetry even in cases when time traces weren ot truncated. [15] As shown on three-spin model systems, by reducing the flip angle of the pump pulse in aP ELDORe xperiment [7] the probability l of pumping B-spins reduces, thus diminishing multispin effects exponentially (with l n ). However, thisa pproach can substantially reduce sensitivity (i.e.,d ipolar modulation to noise ratio). [16] Recently,t he post-processing approach of 'power-scaling' was introduced;t his diminishes spurious peaks from multi-spin contributions to the dipolar coupling( termed' Ghost Peaks') without havingt or educe l (and sensitivity).T his has been experimentally demonstrated for systems withu pt ot hree spins. However,p erformance of this approachr educes with increasing l and n. [7,15] For more than four spinsacombination of reducing l and power-scalingh as been recommended. [16] In parallel, experiments on MscS and the hexameric proteorhodopsin show that power-scalingw ithout reducing l is insufficient for giving reliable distance distributionsf or the non-modal distances. [17] While sparse labeling [18] has led to improved distance distributions in proteorhodopsin, the reduced excitationp robability that comes with Gd III comparedt on itroxide spin-labeling has proveneven more beneficial for multi-spinsystems. [17a] In this work, we quantify the effects of combining powerscaling and choice of l for effectively diminishing multi-spin effects without overly compromising on sensitivity.F urthermore, alternative spectral positions for pulse excitation are explored.
To wards this we have employedt wo tetraradical abaci [19] to investigate this approach in more detail.T etraradical 1 [20] (Figure 1a)i sb ased on an adamantane core with six almost equal distances. The distance measurement data on 1 given in Figure 2c learly demonstrate that reducing l decreasest he modulation depth (D)a nd diminishes as econd distance that provesa'Ghost Peak'. The corresponding suppression function (power-scaling) in DeerAnalysis2013 [16,21] yields similars uppression of this extra distance peak ( Figure 2c). However,p erforming the same experiment on MscS S196R1 (Figure 1b and Figure 2) clearly demonstrates that superior results are obtained by using ar educed l.H ere, reliable distance intensity beyondt he modald istance is recovered, as confirmed by comparisonw ith crystal-structure-based models( see the Supporting Information for detailso fs tructuralm odeling and validation of distance distributions).
These results stillp ossess as ignificant uncertainty related to the larger distances owing to limited observation times. Nevertheless, the same trend could be reproduced using simulations based on geometric modelsm imicking 1 as at etrahedron and MscS S196R1 as ar egularc onvex heptagon [15] (see Figure 3 and the Supporting Information). This provest hat 'Ghost Peaks' and improper amplitudes in the distance distributions obtainedw ith increasing l arise from multi-spine ffects, as the observation time in the simulations was chosen long enough to avoid truncation artifacts. These findings could also be con-firmedo nan ewly synthesized rectangular tetraradical and an octameric Wza [14] (see the Supporting Information). Furthermore,b ys ystematically exploring the recovery of distance distributionsf rom simulated data for equilateral triangles to octagons shows that power-scaling can reliably recover the 'true' distance distribution as long as l is kept below the maximum of the two-spin contribution (1/(n-1)). [15,16] Noise-free simulations suggestt hat even larger l might be tolerated but this is not confirmed upon addition of 1t o3%n oise to the simulations (see the Supporting Information).
While decreasing l yields much-improved distance distributions with increasing n, it is important to note that the resulting reduction in D decreases the modulation effect with respect to noise (though not necessarily the signal-to-noise), which compromises sensitivity (see the Supporting Information). [22] However,i ti sn ecessary to reduce l in combination with power-scaling to diminish multi-spin effectsi np rotein hepta-and octamers (see Figure 2, Figure 3a nd the Supporting Information). Initially,r esults with reduced l were obtained using a5mm dielectric ring (MD5) resonator that hasl arge concentration sensitivity (when the sample amount is not limiting). However,t he maximumm icrowave field-strengthi ss mall-  www.chemeurj.org er than in the 3mms plit-ring resonator (MS3) that is commonly recommended due to its absolute sensitivity (when the sample amount is al imiting factor)a nd highera chievable modulation depth. [23] These resultsa re described in the Supporting Information.
The smaller D that comesw ith reducing multi-spin effects through lowering l prompted us to revisitt he experimental parameters used for distance measurements. Commonly,t he maximum of an itroxide spectrum is inverted by the pump pulse while the second most intense feature is chosen for the detection frequency.I natwo-spin system one trades the detected number of spins for larger D.H owever,r educing l on 1 and MscS S196R1 clearly shows that in am ulti-spin system sensitivity cannot always be traded for D as multi-spin effects scale with the latter (Figure 2). Thus, we decided to interchange the spectral positions in af irst instancea nd also the approximate excitation widths of detection and pump pulses in as econd stage. Implications of frequency positions with respect to the resonator mode have been discussed recently. [24] While the reduction in D is unequivocal, the performance of these experiments based on a( spectrally) minor broad inversion [25] showsc onsistent improvement in distance distributions, but appears to give mixed resultsi nt erms of sensitivity.T his is quantified in the Supporting Information. Briefly,t he relative performance of the standard and frequency-interchanged experiments can be reliably predicted, with the latter especially promising for MscS S196R1.
For the precise resolution of multimodald istance distributions, long observation times and excellent signal-to-noise ratios are required. Recently,t echnical advances have allowed distance measurements exciting significant parts of nitroxide EPR spectra to be expanded from 9GHz to 34 GHz [26] and 95 GHz. [27] We have tested the experiment interchanging detectiona nd pump excitationsa t3 4GHz (Q-band). Figure 4 showst he significant improvements that can be achieved utilizing the superior sensitivity at higher field. Though the 'Ghost Peak' in 1 is less pronouncedi nt he standarde xperiment at Qband owing to slightly lower achievable l and ad ifferent spectral shape, MscS S196R1 still lacks the intensity of the two longer distances. Thisc an be recovered by the frequency-interchanged experiment in line with experiments systematically reducing l (see Figure 4a nd the Supporting Information). Statistical analysisr eveals that all three expected distance peaks for MscS S196R1 can be extracted with high reliability and in good agreement with am odeled distance distribution (see the Supporting Information).
In conclusion we have demonstrated that fully labeled multi-spins ystems with up to eight spins can be reliably measured by PELDOR experiments in combination with power-scaling during post-processing as long as pumping multiple spins is reducedb yk eeping l < 1/(n-1). This is in excellent agreement with earlier predictions by vonH agens et al. [16] We extend their findings to the reliable extraction of experimental distance distributions in heptameric complexes and the estimate of the largest feasible l.F urthermore, interchanging the positions of pump and detection pulses andt heir approximate excitation widths in the nitroxide spectrum allows an alternative route to significant reduction of multi-spin effects.  The performance of these alternative experiments can be reliably predicted before the actual measurements, allowing efficient use of instrument time for the most promising experiment. To gether,o ur resultss hould be especially significant for the rising interest in distance measurements in multimeric membrane transporters.