Combined Multi-Band Decoupling in Biomolecular NMR spectroscopy

Summary:

Multi-resonance NMR experiments are powerful analytical and structural tools. Their conceptualization assumes that RF fields may be combined independently to manipulate spin interactions. However, practical implementation can compromise performance. One limitation is the generation of combination bands when two or more RF fields are applied simultaneously within the NMR probe. The combination bands can lead to significant interference with the detection circuitry. A facile approach to combined multi-band decoupling can resolve these problems and increase sensitivity two-fold (or more), by time sharing the application of the individual frequencies rather than time sharing decoupling and data acquisition.

Spin decoupling is an important and essential part of modern magnetic resonance spectroscopy. From the early discovery of scalar couplings, the ability to remove the presence of spin coupling in the observed NMR spectra has had powerful ramifications. Assignment of signals was made by selective decoupling one resonance and removing the couplings observed in the connected resonance ^1,2, and one of the most powerful applications was the extensive development of broadband decoupling of one spin-type from another^3–5. These tools were largely aimed at providing broad bandwidth coverage over the chemical shift range and simultaneously removing the relatively large one-bond scalar couplings, predominantly focusing on ¹H-¹³C couplings. With the development of indirect ¹H detection of heteronuclear correlations, related decoupling patterns were applied to the heteronucleus, typically ¹³C or ¹⁵N, during acquisition periods. The latter application has fueled the dramatic developments in multi-dimensional spectroscopy in the study of biomolecules^6,7. Recent examples of tailored spin decoupling have also been developed to address band-selective homonuclear decoupling ^8–14. Many of the examples referred to involve decoupling of single spin types within an experiment; however, the advent of perdeuteration in triple resonance biomolecular NMR^15,16 also led to combinations of ²H and ¹⁵N decoupling within sequences, although normally at different segments of the sequence and usually not simultaneously. Similarly, in other areas of application of NMR, such as inorganic chemistry, numerous atoms within the system under study may possess NMR-active nuclei and it is desirable to decouple the spin-couplings both individually and collectively. The ability to have flexible decoupling of any or all spins within a molecule and at any desired time within the pulse sequence for a given experiment is a powerful and significant component of modern NMR spectroscopy.

Modern NMR spectrometers are multi-channel, highly programmable electronic systems that can create arbitrarily complex RF outputs to manipulate spin-spin interactions. Indeed, modern NMR spectroscopy depends on these capabilities; however, unfortunately, the nature of NMR detection hardware (the NMR probe) has a limited number of resonators (coils) to apply these RF fields to the sample (Figure 1). The probe generally relies on inductors, or inductive resonator structures¹⁷, to create magnetic fields responding to the applied RF from the console. The physical geometry of the probe and magnets limit the system to two orthogonal coils, which are then also orthogonal to the magnetic field. Orthogonality is used to minimize inductive coupling between the two coils. To apply more than two RF frequencies, the coils are double- or multiply-tuned to provide irradiation at multiple frequencies¹⁷. The designs have become quite sophisticated and efficient; nevertheless, it cannot be avoided that the experiment is an induction process and the two coils (inductors), although designed to be orthogonal, have finite coupling between them. Hence, as is known from basic electronics, there will be some transformer-like effects that lead to combination frequencies created in the probe when two frequencies are applied simultaneously. The probe circuitry can act as a mixer wherein combination sum and difference frequency RF will result from simultaneous application of any two (or more) RF frequencies. The relative intensity of the applied RF voltage compared to the detected NMR signal voltage (either observe or lock channel) is at least on the order of 10⁶:1; hence, the very minor coupling between coils will lead to signals much larger than the desired signal. If these sum or difference frequencies happen to fall within the bandwidth of the associated electronics for the detected nucleus, then considerable interference can be observed during simultaneous decoupling and detection. The interference can occur for either the observe channel or the lock channel.

Figure 1.

Common probe coil arrangement. Two orthogonal coils are used to apply RF fields generated by RF patterns created in the console. A cartoon of the double tuned circuit (right) presents how the coils can act like transformer/mixers to create sums and differences of the frequencies applied to the coil inputs. This design applies for ¹H frequencies up to approximately 600 MHz. Above 600 MHz, the circuits often have ¹H alone on the inner coil (to optimize sensitivity) and ²H, ¹³C and ¹⁵N are multi-tuned on the outer coil.

Examination of the resonance frequencies of nuclei that are commonly investigated in biomolecular NMR, e.g. ¹H, ²H, ¹³C, ¹⁵N, and ³¹P, reveals a number of complicating interference bands that arise due to these sum and difference frequencies (Figure 2). The majority of triple resonance ¹H-detected biomolecular NMR experiments avoid these combination bands and use only ¹⁵N decoupling during acquisition. For the case of simultaneous ¹³C- and ¹⁵N-edited sequences¹⁸, the combination band is well out of the ¹H frequency range, and the ¹H detection channel of most spectrometers is bandwidth limited to avoid interference. Hence, other than power handling complications, little interference would be observed in the presence of simultaneous ¹⁵N- and ¹³C-decoupling. However, there is an opportunity for interference on the ²H lock channel, as the ¹⁵N and ¹³C difference frequency is near the deuterium frequency. For example, interference on the ²H lock channel can be observed in an experiment where simultaneous ¹³C- and ¹⁵N-decoupling is applied (Figure S1). The bandwidth of most ²H lock channels is approximately +/− 2 MHz; hence, it is not possible to filter out this interference. Additionally, in cases of perdeuterated proteins, many of the conventional triple-resonance experiments may utilize ²H decoupling, but it previously has been applied during evolution times within the pulse sequence that are separate from the detection period and the amide ¹H is detected in the presence of ¹⁵N decoupling. Consequently, pulse sequences that employ simultaneous ¹³C- and ¹⁵N-decoupling (or ²H decoupling) typically employ a blanking procedure that puts the lock circuitry in a ‘LOCK_HOLD’ mode during the duration of such simultaneous sequences. The lock is switched back to normal ‘LOCK_ON’ operation during the recovery time.

Figure 2.

Combination frequencies occurring between some typical NMR frequencies in biomolecular corresponding to a ¹H resonance frequency of 600 MHz.

A critical problem arises for the cases where the application of two decoupling frequencies create combination frequencies near or within the bandwidth of detection for the observed nucleus. Referring to figure 2, it is apparent that this may occur for ¹³C {¹⁵N, ²H}, ¹⁵N {¹³C,²H}, ²H {¹³C,¹⁵N}, ²H {³¹P,¹³C}, ¹³C {³¹P,²H}, or ³¹P {¹³C,²H}, where the nuclei within brackets are decoupled during the acquisition of the first nucleus. One of the most common experiments that is impacted by this interference is ¹³C {¹⁵N,²H}, corresponding to ¹³C-detected spectra in the context of ¹⁵N and ²H-labeled proteins¹⁹ These applications arise for large and/or paramagnetic proteins,^20,21 particularly when C^α-detection is employed. Using a sample of uniformly ¹³C-labeled, ¹⁵N-labeled, ²H-labeled, and methyl-protonated valine (Cambridge Isotopte Laboratories, Inc.), the interference created from simultaneous ¹⁵N- and ²H-decoupling is illustrated in Figure 3. In the absence of decoupling, the FID and the transformed spectrum are of high quality and the full multiplicities of the resonances can be observed (Figure 3A,C and Figure 5). However, in the presence of combined composite pulse decoupling, using WALTZ for ²H and GARP for ¹⁵N, there is significant interference present in the FID, and the transformed spectrum is full of spikes and exhibits low quality (Figure 3B,D).

Figure 3.

¹³C-detection of the one-pulse FID for (*CH₃)₂*CD*CD(*NH₂)*COOH valine in D₂O, where *C represents ¹³C labeling. A, C) Direct detection of the ¹³C-FID and FT-spectrum acquired with no decoupling. B, D) Direct detection of ¹³C-FID and FT-spectrum acquired with simultaneous ²H and ¹⁵N decoupling.

Figure 5.

¹³C-detected one-dimensional spectra of (*CH₃)₂*CD*CD(*NH₂)*COOH valine for different combinations of decoupling. Decoupling is applied using broadband WALTZ decoupling for ¹H and the simultaneous COMB decoupling according to Figure 6 for ¹⁵N and ²H. The decoupling applied is indicated from top to bottom as {¹H,²H,¹⁵N}, {²H,¹⁵N}, {¹H,²H}, {¹H,¹⁵N}, {²H}, {¹⁵N}, {¹H}, and none.

One approach to the interference in the ¹³C {¹⁵N,²H} experiment would consider gating the acquisition sequence, wherein the decoupling frequencies are gated off for the sampling of each data point, as has been utilized in band-selective homonuclear decoupling^8,9,13,14. While modern spectrometers have the flexibility to accomplish this programming during the acquisition of the FID, the additional programming complexity may place some restrictions on the spectral width, dwell time, and ease of operation. Time sharing each dwell point also inherently leads to sampling artifacts and a loss of sensitivity which is proportional to the off time of the receiver^8,9.

We have explored a general solution to the problem that involves an interleaved, or time-shared, synchronous decoupling scheme applied to two hetero-nuclei, which avoids the creation of combination bands. The concept is illustrated in Figure 4. For the case of ¹³C {¹⁵N,²H,¹H}, decoupling of ¹⁵N is initiated in a synchronous fashion at the start of the acquisition. Then, gaps are programmed into the decoupling pattern such that accurate, synchronously timed decoupling can be applied to ²H in the gap duration. The sequences then repeat for the full acquisition time, and there is no further programming required. This sequence is readily programmed as shown in Figure S2, and the synchronous timing is controlled by the command ‘cpds’ in the pulse program operating in TopSpin, which ensures that the sequence starts at the same point in the waveform at the beginning of each acquisition. We refer to this approach as Combined Multi-Band (COMB) decoupling. Since there is no simultaneous application of both ²H and ¹⁵N frequencies, there is no combination sum band at the frequency of ¹³C, and the acquisition of the FID can proceed normally without interference and without any explicit acquisition programming. The latter characteristic simplifies experimental setup and definition of data sizes in the acquisition domain. The small delay times, which are 6 us in Figure S2, may be adjusted depending on the generation and characteristics of the hardware; however, this value has worked on a range of Bruker AVIII and newer consoles operating TopSpin 3.x or 4.x software.

Figure 4.

Simultaneous decoupling of ¹H, ²H, and ¹⁵N during direct ¹³C-detection. A) Conventional broadband sequences applied asynchronously simultaneously to the respective probe circuit. B) Combined Multi-Band (COMB) decoupling applied simultaneously and synchronously to ²H and ¹⁵N channels along with asynchronous broadband ¹H decoupling.

Using the sample of uniformly ¹³C-labeled, ¹⁵N-labeled, ²H-labeled, and methyl-protonated valine, a series of one-dimensional ¹³C spectra were acquired for all possible single, dual, and triple combinations of ¹H, ²H, and ¹⁵N decoupling during acquisition (Figure 5). The dramatic impact of combined ²H and ¹⁵N decoupling on the C^α resonance is evident. These spectra are acquired free of interference artifacts and retaining the full sensitivity of the spectrometer. One can readily envision that these high-quality spectra may then be further simplified or manipulated using various forms of homonuclear ¹³C decoupling²². Examples may include the IPAP or DIPAP tools developed by Bermel and coworkers¹⁹ for studying intrinsically disordered proteins or different forms of band selective decoupling^8,9,13,14. These tools are fully compatible with multi-dimensional triple- and quadruple-resonance experiments

An important requirement in the application of COMB is the actual decoupling pattern used in the interleaved manner. Most optimized decoupling schemes, ranging from MLEV, WALTZ, and GARP to more sophisticated optimal-control-theory generated sequences^23,24 are window-less sequences. The COMB approach is reminiscent of the CRAMPS methodology in ssNMR^25,26, and it is important to consider the bandwidth complications of window-ed or gap-decoupling sequences. It is well established that decoupling sequences have been constructed using repetitive inversion of spins, and the common sequences, such as MLEV, WALTZ, and GARP^3,5 (including rectangular and adiabatic pulses (e.g. WURST²⁷)) were developed using phase cycling of spin inversion pulses. By inserting gaps to enable application of an inversion on the second spin, the behavior of the decoupling bandwidth can be impacted. We have explored several options and find that, for ¹⁵N decoupling, use of BIP720 pulses²³, phase cycled according to the P5M4 supercyle^28,29 provides an excellent compromise between RF power requirements and decoupling bandwidth. Using a 1.8 ms BIP720 pulse applied at an effective RF field of 800 Hz (equivalent to a rectangular π/2 pulse of 320 μs) provides very adequate bandwidth, while leaving a gap of up to 450 μs for a ²H π pulse. The effectiveness and bandwidth of decoupling for both ¹⁵N and ²H is demonstrated in Figure 6. Observation of the C^α resonance of the labeled Val sample, during on-resonance ²H decoupling and stepwise offsetting the ¹⁵N decoupling frequency, illustrates a bandwidth of approximately 90 ppm in offset at the ¹⁵N frequency. Similarly, observation of the offset dependence for ²H decoupling, while on-resonance ¹⁵N decoupling, illustrates a bandwidth of approximately 8.5 ppm. These bandwidths provide complete decoupling under conditions appropriate for protein NMR spectroscopy. Further comparisons of conventional continuous decoupling and the gapped decoupling in COMB are provided in the Supporting information Figures S3 to S25. These data illustrate bandwidth and residual linewidth comparisons for gap-less, gapped, and COMB sequences.

Figure 6.

Decoupling bandwidth for ¹⁵N and ²H under COMB conditions. A) Offset dependence of the Ca resonance of Valine under conditions of on-resonance ²H decoupling combined with stepped offset of ¹⁵N decoupling using a 1.8 ms BIP720 pulse incremented in steps of 5 ppm. B) Offset dependence of the Ca resonance of Valine under conditions of on-resonance ¹⁵N decoupling combined with stepped offset of ²H decoupling using a 428 ms square p pulse incremented in steps of 0.5 ppm.

As a practical example, we chose the ¹³C direct-detected HNCOCA experiment¹⁹ acquired on a sample of uniformly ¹³C-, ¹⁵N- and ²H-labeled human carbonic anhydrase II (HCA-II) at 16.45 T (Figure 7). In this experiment, the acquisition dimension is set on the alpha carbons, and the ¹J_C-C couplings of approximately 53 Hz and 35 Hz to C’ and C^β, respectively, were removed by a double IPAP scheme^19,22. Previously, this experiment was acquired with only ¹⁵N-decoupling, for the reasons presented above, which leaves the ¹J_C-D couplings (~20 Hz) active. Without simultaneous decoupling of ¹⁵N and ²H, the ¹³C^α signals are split into complex multiplets. Each spectrum was acquired as the C^α-N plane: direct ¹³C^α acquisition dimension 2048 points (1024 complex), spectral width 28.4 ppm (5000 Hz) and acquisition time 0.2048 sec; indirect ¹⁵N dimension 138 points ×4 for double IPAP (69 complex – limited by the constant time), spectral width 40 ppm (2838.6 Hz) and acquisition time 0.02043 sec. The spectra differ only in the decoupling scheme applied during acquisition. Figure 7 demonstrates the gain in signal intensity and resolution by successive incorporation of the respective modes of decoupling. The ¹J _C-N, ²J_C-N and the ¹J_C-D couplings were removed by COMB decoupling during the acquisition time (Figure 7A). The low average power of the COMB decoupling scheme also allows longer acquisition times and permits even higher resolution in these spectra. The sensitivity difference between ¹⁵N only and optimally COMB decoupled spectra corresponds to approximately a factor of ~2.5. Furthermore, the raw data may be processed without application of the linear combination of data (DIPAP), thus yielding the fully coupled spectrum and allowing very clear observation of the J(Cα,C’) and J(Cα,Cβ) couplings.

Figure 7.

Comparison of decoupling modes in ¹³C-detected HNCOCA. ¹⁵N, ¹³C^α planes acquired with (A) COMB decoupling of both ¹⁵N and ²H, (B) ²H decoupling only, and (C) ¹⁵N decoupling only. The data were acquired as 2D experiments, and the ¹³C-¹³C couplings between C’-C^α and C^α-C^β were removed with a double IPAP scheme. The insets provide a clearer visualization of peak shapes. The dashed lines correspond to traces shown to the right of each panel.

We have explicitly dealt with the combination bands relative to common labeling schemes in biomolecular NMR; however, there are numerous examples of combinations of frequencies that will lead to interference in the general application of NMR spectroscopy. These frequencies arise due to the ratios of the gyromagnetic ratio of nuclei, and these will occur at all magnetic field strengths. The approach of COMB can be extended and adapted to most of these cases. A bandwidth complication may be encountered for decoupling ¹³C in the cases of ¹⁵N {¹³C,²H}³⁰ and ³¹P {¹³C,²H}. Achieving a sufficient bandwidth with windowed decoupling to cover the full ¹³C chemical shift range might prove difficult; however, when the chemical shift range of coupled ¹³C sites is moderate, COMB will be quite effective. The ³¹P {¹³C,²H} case may be easily optimized by selection of different windowing, since J³¹_P-²_H tend to be rather small and may require either no or minimal decoupling. The application of simultaneous ¹³C and ²H decoupling during acquisition for proteins would be envisioned only for samples where amide nitrogens were deuterated and direct ¹⁵N-detection was utilized. This case would represent a very low sensitivity case and is not likely of practical concern. Furthermore, for cases within a pulse sequence where simultaneous ¹³C and ²H decoupling is desired, COMB decoupling would still be an effective approach to avoid power handling problems depending on the specific probe configuration.

In conclusion, the Combined Multi-Band decoupling method provides a simple, efficient scheme to achieve simultaneous decoupling of two hetero-nuclei, and COMB avoids significant RF interference that can result in large artifacts and compromise the sensitivity of the experiment. COMB allows the greatest flexibility, elimination of sampling artifacts, and up to a 2.5-fold gain in sensitivity for demanding ¹³C-direct detection {¹⁵N,²H} experiments.