DNP Enhanced Frequency-Selective TEDOR Experiments in Bacteriorhodopsin

Abstract

We describe a new approach to multiple ¹³C-¹⁵N distance measurements in uniformly labeled solids, frequency selective (FS) TEDOR. The method shares features with FS-REDOR and ZF- and BASE-TEDOR, which also provide quantitative ¹⁵N-¹³C spectral assignments and distance measurements in U-[¹³C,¹⁵N] samples. To demonstrate the validity of the FS-TEDOR sequence, we measured distances in [U-¹³C,¹⁵N]-asparagine which are in good agreement with other methods. In addition, we integrate high frequency dynamic nuclear polarization (DNP) into the experimental protocol and use FS-TEDOR to record a resolved correlation spectrum of the arg-¹³C_γ-¹⁵N_ε region in [U-¹³C,¹⁵N]-bacteriorhodopsin. We observe six of the seven cross peaks expected from this membrane protein.

1. Introduction

Heteronuclear dipole Hamiltonians are simple in comparison with their homonuclear counterparts in that they do not contain flip-flop terms; therefore, heteronuclear recoupling experiments are not subject to the phenomenon of dipolar truncation [2]. Accordingly, successful approaches to heteronuclear distance measurements exploit this simplicity and are usually based on REDOR [3] and its variants such as TEDOR [4; 5; 6]. However, even though dipolar truncation is absent, there remain complications when distance measurements in uniformly ¹³C-¹⁵N labeled systems are involved. In particular, evolution under the homonuclear ¹³C-¹³C J couplings imposes an overall modulation on the buildup or dephasing during the recoupling period. J-couplings can generate anti-phase coherences, which in turn give rise to phase-twisted line shapes. In order to circumvent these limitations, Jaroniec, et. al. introduced J-decoupled [7]and frequency-selective REDOR[8] experiments in which the detrimental effects of J-coupling are removed through the application of frequency-selective pulses. These effects were also consider by Mehta and Schaefer [9] in the development of the RDX version of REDOR. Another logical extension of these techniques has been the 3D ZF- and BASE TEDOR experiments [1; 10], in which multiple ¹³C-¹⁵N distances are measured simultaneously in a broadband or a band selective manner in the context of a ¹³C-¹⁵N correlation experiment. In these cases, the effects of J-coupling are removed through the application of a coherence filter or frequency selective ¹³C pulse, respectively. This family of sequences has been successfully employed in the de novo structural determination of a peptide [11], an amyloid fibril [12], and quantitative distance measurements in a membrane protein via MAS NMR [13], and is therefore potentially an important part of the repertoire of experiments for determining structures of proteins with MAS NMR.

However, even where the recoupled dipolar Hamiltonian for individual spin pair interactions commutes, strong dipolar couplings can compromise the sensitivity of the experiment. For example, in the TEDOR experiment, the intensity corresponding to a single ¹⁵N site is distributed over all ¹³C-¹⁵N cross peaks and is a function of the dipolar couplings, their mutual orientations, homonuclear J couplings, and the mixing time. Thus, in the practically important case of distance measurements between backbone ¹⁵N’s and ¹³C’s of an amino acid side chain, the presence of strong ¹³C_β-¹⁵N_α couplings (2.5Å) in the TEDOR dynamics reduces the maximum intensity of polarization transfer due to weaker couplings (~3.3Å) to 1–3% of the initial intensity [12; 14]. This sometimes compromises the application of broadband techniques to membrane proteins and other systems in which sensitivity is limited. This effect is also particularly acute in amino acids whose side chains contain nitrogen (e.g. asparagine, glutamine lysine, arginine), as a single ¹³C site may be subject to several strong couplings that dominate the spin dynamics [15]. However, when the backbone or side chain ¹⁵N^’’s are resolved, then a ¹⁵N-frequency-selective analogue of the 3D TEDOR experiment can simplify the spin dynamics and allow high sensitivity measurement of weak couplings.

2. Results and discussion

These deficiencies and opportunities have stimulated us to develop the frequency-selective (FS) TEDOR experiment illustrated in Figure 1 that shares many features with the FS-REDOR and BASE and ZF-TEDOR methods discussed above. Following ramped, ¹H-¹³C cross polarization, we apply a REDOR train to generate ¹³C-¹⁵N antiphase coherence which evolves under the heteronuclear dipolar couplings. During the excitation period, a frequency selective pulse is applied on the ¹⁵N channel, such that only those ¹⁵N nuclei within the bandwidth contribute appreciably to the spin dynamics during the REDOR mixing period. Following the excitation period, optional frequency labeling with the ¹⁵N chemical shift occurs during t₁. A second, frequency selective REDOR period converts the resulting coherence into observable magnetization on the ¹³C channel, which is detected during t₂. A delay, τ, is necessary to ensure that the time between REDOR periods is an integer number of rotor cycles during which the signal is transferred to the ¹³C’s. Finally, ¹³C-¹³C J-couplings give rise to undesired zero-quantum and double-quantum coherences that, after the reconversion period, result in severe anti-phase distortions to the line shape and spurious cross-peaks in 2D spectra. Two variable-length coherence filters (z-filters) are applied to dephase these undesired coherences, resulting in undistorted and purely absorptive spectra. TPPM and CW decoupling is used where appropriate [16]. In addition, for the experiments on the membrane protein bacteriorhodopsin (bR), we enhance the overall signal sensitivity of the experiments by integrating dynamic nuclear polarization (DNP) into the protocol, as illustrated in Figure 1. In particular, we suspend the bR sample in 60% glycerol/H₂O that forms a stable glass and serves to disperse the biradical polarizing agent TOTAPOL[17]. Irradiation with 250 GHz microwaves [18] leads to a polarization enhancement of ~43 in the signal intensities.

Figure 1.

Pulse sequence for ¹⁵N frequency selective TEDOR transfer. Only ¹⁵N spins within the bandwidth of the selective pulses contribute to the spin dynamics under REDOR. The sequence is described in the text. The REDOR pulses are phase cycled according to the XY-8 scheme and the overall phase cycle is documented elsewhere[1]. CW microwave irradiation was employed in the version of the experiment used to acquire the DNP enhanced spectrum.

As an initial test of FS-TEDOR, all heteronuclear distances in [U-¹³C, ¹⁵N]-asparagine were measured in two experiments, with a selective pulse applied at the frequency of the backbone (-¹⁵NH₃⁺) or side chain ¹⁵N (-¹⁵NH₂). The results are summarized in Figure 2 and Table 1. All data were recorded on a 500 MHz (¹H) home-built spectrometer (Cambridge Instruments) using a Varian triple-resonance MAS probe and in a sample of asparagine diluted to 10% in its natural abundance analogue to avoid the confounding effects of intermolecular couplings. The data were fit using a multispin simulation performed with the program SPINEVOLUTION [19]. Our FS-TEDOR distance is in good agreement with the distance measured by frequency-selective REDOR [8] which is a true two-spin experiment and with the neutron diffraction structure. We note that the FS-REDOR and FS-TEDOR will likely be more accurate that SPECIFIC-CP [20] experiments for distance measurements. A careful comparison with results from PAIN-CP [21] is in progress.

Figure 2.

FS-TEDOR experiment in 10% U-[¹³C,¹⁵N]-Asparagine. We employed 100 kHz TPPM decoupling during REDOR and other periods, a 1 ms Gaussian refocusing pulse (switched between the NH₃ and NH₂ sites), 50 kHz REDOR pulses. ω_r/2π = 10 kHz

Table 1.

Heteronuclear distances measured with FS-TEDOR experiment as applied to [U-¹³C, ¹⁵N]-Asparagine

	FS-TEDOR (Å)	FS-REDOR (Å)	Neutron diffraction (Å)
N-Cα	1.50	1.50	1.49
Nδ2-Cα	3.56	3.58	3.75
N-Cβ	2.46	2.49	2.48
Nδ2-Cβ	2.41	2.44	2.42

To demonstrate the utility of FS-TEDOR as an approach for spectral editing in a more challenging and interesting case, we used the FS-TEDOR sequence, together with DNP, to record 2D ¹³C-¹⁵N spectra of the seven arginine side chains in the membrane protein bacteriorhodopsin (bR), a light-driven ion pump that has been studied extensively by magnetic resonance[22; 23; 24; 25; 26; 27; 28];[29]. Owing to their hydrophilicity, many of the lysine and arginine side chains in proteins are located at or near the surface. In contrast, those buried in the hydrophobic interior are frequently functionally important. bR contains a number of such examples, including Lys-216, to which the retinal chromophore is attached via a Schiff base linkage, and Arg-82, which is part of the complex counterion and is perturbed following deprotonation of the Schiff Base, a key step in the L->M_o photostate transition of bR [30]. In this and other systems, approaches to spectral editing based on selective cross polarization [20; 31; 32], magnetization preparation and/ or selective dephasing [13] , or both in combination[33], have been used to achieve similar ends. The optimal sequence to use in each case depends on the spin topology (e.g. ¹³C coupled to multiple ¹⁵Ns or the reverse) and desired selectivity.

In this case, addition of DNP, which consists of CW irradiation of the EPR spectrum of TOTAPOL with 250 GHz microwaves to the FS-TEDOR experiment, enhances the signal sensitivity by a factor of ~43. Together with the lower temperatures, we obtain a factor of ~90 increase in signal intensity over our previous experiments on bR at 200 K. Figure 3 shows the 1D 15N spectrum of U-¹³C,¹⁵N-bR.

Figure 3.

1D DNP enhanced ¹⁵N spectra of bR illustrating the excellent S/N available with DNP that permits observation of the ¹⁵N signal of the single Schiff Base (K216). The Arg resonances irradiated with the selective pulse are labeled and lie at ~80 ppm.

We recorded correlation spectra, focusing on the ¹³C_γ-¹⁵N_ε region of arginine. Details of instrumentation[18; 34; 35; 36], MAS probes[37], polarizing agents[38; 39], and sample preparation for applications of DNP to bacteriorhodopsin[33; 40; 41] are published elsewhere and not reviewed here. We empirically and separately optimized the TPPM decoupling [16] during the inter-pulse periods of the REDOR sequence, the selective pulse, and detection, to minimize the transverse relaxation (T₂), avoid interference effects between the decoupling and multiple pulse sequence, and to maximize the sensitivity. We chose a Gaussian refocusing pulse whose length was optimized in the range 1.98 ms ± 0.265 ms [42](see Figure 4) and whose frequency was centered on the arginine side chain region (~90 ppm for ¹⁵N_ε) (Figure 4). A pulse of this length sufficiently suppresses the amide backbone resonances, which otherwise would compromise the dynamic range of the experiment and obscure the arginine cross peaks, which are an order of magnitude less intense. Further, we selected a TEDOR mixing time of 4 ms, in which polarization transfer over two-bond distances (e.g. ~2.4–2.5 Å between arg-¹³C_γ-¹⁵N_ε ) in U-¹³C,¹⁵N compounds is maximized.

Figure 4.

2D ¹⁵N-¹³C selective correlation experiment focused on the Arg-¹³C_γ-¹⁵N_ε of bR. The sequence described in Figure 1 was employed with a 1.98 ms Gaussian refocusing pulse, and TPPM decoupling optimized to reduce transverse dephasing. The REDOR mixing time of 4 ms was chosen to favor short distances (~2.4–2.5 Å). The spinning frequency was 7.576 kHz, and all DNP experiments were conducted in dark-adapted bacteriorhodopsin (90 K), as described elsewhere.

The spectrum in Figure 4 reveals ¹⁵N-¹³C correlations which we assign to six of the seven expected Arg-¹³C_γ-¹⁵N_ε cross peaks. Our assignment is based on our choice of parameters (a selective pulse centered at 90 ppm and a 4 ms TEDOR recoupling time) which favor Arg-¹³C_γ-¹⁵N_ε magnetization transfer over Arg-¹³C_δ-¹⁵N_ε,η1,η2, due to the mixing time and the presence of multiple ¹³C_δ-¹⁵N couplings. In addition, the chemical shifts observed agree with typical values for Arg-¹⁵N_ε and Arg-¹³C_γ, indicated by amino acid chemical shift statistics for proteins [43; 44]which show Arg-¹⁵N_ε chemical shifts as 90 ± 19 ppm vs. 75 ± 14 ppm and 75 ± 16 ppm for Arg-¹⁵N_η1 and Arg-¹⁵N_η2, respectively, and 27.2 ± 2.0 ppm for Arg-¹³C_γ vs. 43.2 ± 1.8 ppm for Arg-¹³C_δ. However, while both the ¹⁵N and ¹³C chemical shifts are only consistent with the Arg-¹³C_γ-¹⁵N_ε assignments, their eventual validation will require site-specific assignment of all the Arg side chains, which we did not attempt here.

3. Conclusions

In summary, we have demonstrated that the FS-TEDOR experiment, a close analogue of the frequency-selective REDOR experiment, can be used quantitatively and qualitatively for ¹⁵N-¹³C correlation spectroscopy in crystalline solids and membrane proteins. Using DNP, we have recorded resolved arg-¹³C_γ-¹⁵N_ε correlation spectra, tentatively identifying six of the seven expected resonances. Similar methods might be contemplated in lieu of specific isotopic labeling or suppression[47] to simplify the spin dynamics, or in contexts where quantitative distance information is required from a resolved ¹⁵N site, such as in distance measurements from the Schiff base of bR and other rhodopsins, or in peptides containing ¹⁵N-rich side chains[15; 48]. The decision to use FS-TEDOR over other methods depends on the topology of the spin system under consideration, the degree of selectivity desired, and the dephasing of transverse magnetization during the evolution times of the experiment.

Table 2.

Chemical shifts and intensities of crosspeaks in the Arg-¹³C_γ-¹⁵N_ε region of 2D fs-TEDOR spectra bR (Figure 4). The chemical shifts and intensities were fit using SPARKY [45] following processing in NMRPipe [46]

Peak	15N Chemical Shift (ppm)	13C Chemical Shift (ppm)	Peak Volume (arbitrary units)
A	82.9	30.9	2.9
B	87.3	31.7	2.7
C*	87.8	31.0	2.7
D	89.9	30.7	3.1
E	89.4	29.3	3.0
F	87.6	27.9	3.5
G	82.1	26.0	3.1

^*Cross peak C is not resolved from D.