Multi-Receiver Solid-State NMR using Polarization Optimized Experiments (POE) at Ultrafast Magic Angle Spinning

Abstract

Ultrafast magic angle spinning (MAS) technology and ¹H detection have dramatically enhanced the sensitivity of solid-state NMR (ssNMR) spectroscopy of biopolymers. We previously showed that, when combined with polarization optimized experiments (POE), these advancements enable the simultaneous acquisition of multi-dimensional ¹H- or ¹³C-detected experiments using a single receiver. Here, we propose a new sub-class within the POE family, namely HC-DUMAS, HC-MEIOSIS, and HC-MAeSTOSO that utilize dual receiver technology for the simultaneous detection of ¹H and ¹³C nuclei. We also expand this approach to record ¹H-, ¹³C-, and ¹⁵N-detected homonuclear 2D spectra simultaneously using three independent receivers. The combination of POE and multi-receiver technology will further shorten the total experimental times of ssNMR experiments for biological solids.

1. Introduction

Magic angle spinning solid-state NMR (MAS ssNMR) spectroscopy is maturing as a central technique for chemical, biochemical, and biophysical research, spurring its application to the characterization of the structure, dynamics, and ligand binding of protein fibrils, crystalline proteins, and membrane proteins at atomic resolution^1–9. Higher magnetic field strengths and multidimensional experiments have provided powerful new tools for investigating larger biomacromolecules^10,11. In the past few years, there has been a significant effort to accelerate MAS ssNMR experiments using dynamic nuclear polarization (DNP), paramagnetic relaxation enhancement (PRE), and ultra-fast spinning speeds^12–15. Several technological developments have increased the scope of ssNMR applications to large biomolecular complexes. These advancements include low-E or E-free probes that increase the sensitivity of experiments, while avoiding heating caused by high-power RF pulses and enabling the analysis of temperature-sensitive biological samples under physiological conditions^16,17. More recently, fast and ultrafast MAS probes as well as spectrometers equipped with field-gradient coils and multiple receivers enabled the development of cross-polarization (CP) and INEPT-based experiments for biosolids, with a dramatic increase in sensitivity and resolution^18–21. During the past decade, our group has developed a general approach, namely, polarization optimized experiments (POE), which enables the concatenation of various pulse sequences into single experiments^22,23. Using different strategies (i.e., DUMAS, MEIOSIS, and MAeSTOSO), we implemented the simultaneous acquisition of multiple 2D and 3D spectra at slow and fast spinning rates using either ¹H- or ¹³C- detected single-receiver experiments^22,24–27. Along with auxiliary developments made by other research groups, the POE family now includes ¹H-detected experiments under fast MAS conditions, afterglow, dipolar-edited versions of DUMAS pulse sequences, and, more recently, hybrid pulse sequences combining CP and INEPT transfer periods^27–34. These experiments reinforce the importance of developing multi-acquisition pulse sequences for ssNMR that utilize orphan spin operators (unused polarization) to enhance signal intensity and/or collect several datasets simultaneously²³.

In ssNMR, ¹³C- and ¹⁵N-detected experiments have been paramount to the high-resolution spectral analysis of biomolecular systems^1,2,35,36. While ¹H-detected experiments using fast MAS have provided a significant boost in signal-to-noise, the resulting spectra are generally less resolved due to narrower chemical shift dispersion and line broadening from strong ¹H-¹H homonuclear dipolar couplings. In this work, we extend the POE methods using ultra-fast MAS and multi-receiver technologies to combine the benefits of ¹H, ¹³C and ¹⁵N detection. Specifically, we present the multi-receiver implementation of three POE strategies, i.e., ¹H- and ¹³C-detected DUMAS, MEIOSIS and MAeSTOSO (namely HC-DUMAS, HC-MEIOSIS, and HC-MAeSTOSO), and a triple-receiver (¹H, ¹³C, and ¹⁵N) HCN-DUMAS. While specific combinations of POE are illustrated, the HC-DUMAS, HC-MEIOSIS, HC-MAeSTOSO, and HCN-DUMAS strategies can also be used to concatenate several other ¹H, ¹³C, and ¹⁵N detected pulse sequences to speed up data acquisition. We demonstrate the performance of these new multi-acquisition POE methods, using a fully protonated uniformly U-¹³C,¹⁵N labeled microcrystalline GB1 protein and fMLF tripeptide samples.

2. Materials and methods

All the pulse sequences were implemented on a Bruker 600 MHz spectrometer equipped with a 1.3 mm fast MAS probe and Avance NEO® console with multi-receiver technology. The spectrometer was operated using TOPSPIN 4.0.2 software. U-¹³C and ¹⁵N labeled microcrystalline GB1 (β1 immunoglobulin binding domain of protein G) protein sample was prepared from the protocol (crystal form A) described previously by Rienstra and co-workers^37,38. Approximately 2 to 3 mg of GB1 microcrystals, in residual precipitant solution, were packed into a 1.3 mm rotor. All the spectra were acquired with a MAS rate of 65 kHz. A sample temperature of 25 °C was maintained by setting the RF coil temperature to −20 °C to compensate for the heat induced by fast spinning as measured from the water resonance frequency.

All the experiments carried out on GB1 protein were acquired with a 2 s recycle delay, whereas a 3 s recycle delay was used for recording fMLF spectra. The 90° pulse length for ¹H was set to 1.25 μs, whereas 90° pulses of 3 μs were used for ¹³C and ¹⁵N. For the GB1 sample, ¹H solvent suppression was obtained by the MISSISSIPPI sequence (without gradients) with a ¹H RF amplitude of 30 kHz applied for 200 ms (represented by t_supp in Figures 2A, 3A, 4A, and 5A)³⁹. For HC-MAeSTOSO-4 and HC-MAeSTOSO-10 pulse sequences (Figures 4A and 5A), an additional water suppression (0.25*t_supp) period was used²⁸. For the SIM-CP preparation period, simultaneous ¹H-¹³C and ¹H-¹⁵N Hartmann-Hahn matching conditions were obtained by using constant amplitude RF pulses on ¹³C and ¹⁵N, while the RF pulse on ¹H was linearly ramped from 90 to 100%^22,28,40. The optimized SIM-CP RF amplitudes for ¹H, ¹³C, and ¹⁵N were 88.9, 23.7, and 23.6 kHz, respectively. For ¹³C-¹H and ¹⁵N-¹H back-CP periods, the optimal transfer was obtained using an 83.8 kHz RF amplitude on ¹H with a 90 to 100% linear ramp, whereas RF amplitudes were set to 23.7 kHz for ¹³C and ¹⁵N. During NCA and NCO SPECIFIC-CP transfer, the ¹⁵N RF amplitude was set to 43 kHz, while the RF of ¹³C was linearly ramped from 85 to 100% with a maximum amplitude of 17 kHz⁴¹. Homonuclear DREAM polarization transfer was obtained with a linear RF ramp of 80 to 100% for CACO, and 85 to 100% for CACB using a 31.9 kHz ¹³C RF amplitude⁴². The ¹³C offset was set to 57 ppm during NCA and CACO, 175 ppm for NCO, and 45 ppm for CACB transfer periods. Mixing times for HH-RFDR sequence (Figures 5 and 7), were set from 0.5 to 2 ms^43–45. During the t₁ and t₂ periods, a WALTZ16 sequence was used for heteronuclear decoupling with 10 kHz RF amplitude⁴⁶. The RF carrier frequency for ¹H, ¹³C and ¹⁵N were set to 4.7, 42, and 122 ppm, respectively. The contact times for CP and SIM-CP periods, were set to 1 ms, whereas ¹³C-¹H and ¹⁵N-¹H back-CP periods were set to 200 and 500 μs, respectively. The duration of NC SPECIFIC-CP and CACB DREAM transfer was set to 9.5 and 10 ms, respectively. The CXCX TOCSY mixing was obtained from a 12 ms WALTZ16 period with ¹³C RF amplitude set to 10 kHz. The t₂ acquisition period for ¹H (t₂^H) was set to 8.3 ms, whereas the ¹³C and ¹⁵N acquisition times were set to 25 ms (t₂^C and t₂^N). The total experimental time for the 2D HC-DUMAS, HC-MEIOSIS, and HC-MAeSTOSO-4 spectra of GB1 sample were 0.4, 3, and 3 h, respectively (Figures 2–4, and Table 1); whereas the HCN-DUMAS 2D spectra on fMLF sample was acquired in 1.4 h (Figure 7, and Table 2). For the HCN-DUMAS pulse sequence carried out on fMLF, suppression of the ¹H background signal in the HH 2D experiment was achieved using a spin-echo sequence, τ_r-180°-τ_r before ¹H acquisition, where τ_r corresponds to a rotor period of 15.38 μs.

Figure 2:

(A) 2D HC-DUMAS pulse sequence for simultaneous acquisition of (H)CACO and (H)NH experiments. The phase cycle is set to, ϕ1=y, −y; ϕ2=x, x, −x, −x; ϕrec=y, −y, −y, y. (B) One-dimensional (H)CACO and (H)NH spectra of GB1 protein acquired using the first increment of the pulse sequence setting t₁^C and t₁^N to zero, and n=2 (Table 1). (C) 2D (H)CACO and (H)NH spectra of GB1 acquired simultaneously using HC-DUMAS pulse sequence. (D) 1D cross sections for ¹H and ¹³C dimensions extracted from the 2D (H)CACO and (H)NH spectra. The corresponding single receiver (¹H or ¹³C) experiments were shown in references^22,28.

Figure 3:

(A) HC-MEIOSIS pulse sequence for simultaneous acquisition of two pairs of ¹³C and ¹H detected dual-receiver experiments. The phase cycle is set to, ϕ1=y, −y; ϕ2=x, x, −x, −x; ϕrec=y, −y, −y, y. (B) 1D HC-MEIOSIS spectra of GB1 protein obtained from the first increment of HC-MEIOSIS pulse sequence, and n=4 (Table 1). The corresponding integrated S/N values are indicated. (C) 2D HC-MEIOSIS spectra of GB1 protein, (H)CXCX, (H)N(CA)CX, (H)CA(N)H, and (H)NH acquired simultaneously using the pulse sequence in A. The 1D cross sections along the dotted lines are shown for ¹³C and ¹H detected 2D spectra. The corresponding single receiver experiments were shown in references^24,28.

Figure 4:

(A) Dual receiver HC-MAeSTOSO-4 pulse sequence for simultaneous acquisition of one ¹³C-detected and three ¹H-detected experiments. The phase cycle is set to, ϕ1=y, -y; ϕ2=x, x, −x, −x; ϕ3=x, x, x, x, −y, −y, −y, −y; ϕrec=y, −y, −y, y, −y, y, y, −y. (B) 1D HC-MAeSTOSO-4 spectra of GB1 protein obtained by setting t₁^C and t₁^N to zero. (C) Simultaneous acquisition of 2D (H)CXCX, (H)N(CACB)H, (H)N(CA)H, and (H)NH spectra of GB1 using HC-MAeSTOSO-4 pulse sequence. (D) 1D cross sections along ¹³C (55.6 ppm) and ¹⁵N (124.2 ppm) dimensions and corresponding S/N and FWHM values. The corresponding single receiver experiments were shown in references^25,28.

Figure 5:

(A) Dual receiver HC-MAeSTOSO-10 pulse sequence for the simultaneous acquisition of four ¹³C detected, and six ¹H detected experiments. The phase cycle is set to, ϕ1=y, −y; ϕ2=x, x, −x, −x; ϕrec=y, −y, −y, y. (B) HC-MAeSTOSO-10 1D spectra of GB1 protein obtained from the first increment (t₁^C, t₁^N=0). The integrated S/N values for each 1D spectra are shown in Figure 5B. For ¹H spectra with HH RFDR mixing, the S/N values are calculated for H^N (6 to 10 ppm) and H^C (0 to 4 ppm) spectral regions, and respectively shown on the left and right sides of the 1D spectra. The ¹H peak at 4.7 ppm, indicated by *, corresponds to water signal.

Figure 7:

(A) Triple receiver HCN-DUMAS pulse sequence for simultaneous acquisition of 2D HH, CC, and N(HH)N homonuclear correlation experiments. The phase cycle is set to, ϕ1=y, −y; ϕ2=x, x, −x, −x; ϕrec=y, −y, −y, y. (B) 2D HH, CC, and N(HH)N spectra of fMLF tripeptide sample using HCN-DUMAS pulse sequence. (C) 1D cross-sections taken along ¹H, ¹³C, and ¹⁵N dimensions, and corresponding S/N and FWHM values.

Table 1:

Experimental t₁ evolution parameters for the POE reported in Figures 2–4.

POE Method	13C detected experiments	1H detected experiments	13C t1 evolution (t1C)			15N t1 evolutio n (t1N)				ns	nseff =n* ns	Texp	~TCex
			dw (t1C)	ni (t1C)	T(t1C)	dw (t1N)	ni (t1N)	T(t1N)	n
HC-DUMAS	(H)CACO	(H)NH	200 μs	74	14.6 ms	400 μs	37	14.4 ms	2	4	8	23 min	42 min
HC-MEIOSIS = (ϕ = x, −x)	(H)CXCX (H)N(CA)CX	(H)CA(N)H (H)NH	100 μs	148	14.7 ms	400 μs	37	14.4 ms	4	16	64	3 h	7.9h
HC-MAeSTOSO-4	(H)CXCX	(H)N(CACB)H (H)N(CA)H (H)NH	100 μs	148	14.7 ms	400 μs	37	14.4 ms	4	16	64	3 h	6.3 h

dw = dwell time; ni = no. of t₁ increments; T (maximum t₁ evolution time) = dw*(ni-1); n = no. of ¹⁵N t₁ loops; ns = no. of scans; ns_eff = Effective number of scans for ¹⁵N edited spectrum; T_exp = Total experimental time; T^C_exp= Estimated experimental time using conventional single acquisition pulse sequences; Condition for parallel ¹³C and ¹⁵N t₁ evolutions: ni(t₁^C) = n*ni(t₁^N)^22,26.

Table 2:

Experimental t₁ evolution parameters for the HCN-DUMAS spectra reported in Figure 7.

CXCX				HH					N(HH)N					Texp	~TCexp
dw (t1C)	ni (t1C)	T(t1C)	ns		ni (t1H)	T(t1H)	n1	nseff =n1*ns	dw (t1N)	ni (t1N)	T(t1N)	n2	nseff =n2*ns
80 μs	100	7.9 ms	8	80 μs	25	1.9 ms	4	32	320 μs	25	7.7 ms	4	32	1.4 h	3.2 h

Condition for parallel ¹H, ¹³C, and ¹⁵N t₁ evolutions: ni(t₁^C) = n1*ni(t₁^H) = n2*ni(t₁^N).

All the 2D spectra were acquired with States mode during t1 evolution, by switching the phase ϕ* between y and −x⁴⁷. Both t₁ and t₂ dimensions were processed with a 90° shifted sine bell window functions using NMRpipe scripts as previously described for multiple acquisition experiments^26,48. The 2D data sets were analyzed using Sparky⁴⁹. All multi-acquisition pulse sequences utilize multiple WALTZ16 decoupling periods on ¹H, ¹³C, and ¹⁵N channels. As reported previously, we recommend to use different names for each WALTZ16 period in the programming code to avoid sharing of the pulse profile²⁸. For reasons unknown at this point, we had to add three ¹³C (t₂^C) and one ¹H (t₂^H) dummy acquisition periods at the end of the pulse program to successfully execute the HC-MAeSTOSO-4 experiment. Similarly, the HC-MAeSTOSO-10 pulse program required an additional two ¹³C and one ¹H dummy t₂ acquisition periods. Note that these dummy t₂ acquisition periods contain no signal and are discarded during processing. We hope to rectify these issues with future software/hardware upgrades. The POE pulse programs will be available to users upon request.

3. Results

3.1. Polarization transfer efficiencies of CC and NC mixing periods

To design efficient and sensitive POE, it is important to optimize the transfer efficiencies for homo- and hetero-nuclear recoupling sequences. Figure 1 shows the comparison of optimized ¹³C spectral intensities obtained from the homo- and hetero-nuclear polarization transfer schemes, using the conventional 1D pulse sequences with ¹³C detection and WALTZ16 ¹H decoupling. Each of the 1D spectra was acquired using 512 scans. The complete experimental parameters are reported in section 2. To quantify the efficiencies of CC and NC transfer periods, the integrated intensities of the spectra in Figures 1B–F were normalized with respect to the Cα spectral region of ¹H-¹³C CP spectrum (Figure 1A)⁵⁰. In comparison to ¹H-¹³C CP, the integrated intensities of SPECIFIC-CP NCA and NCO are 31 and 44%, respectively (Figures 1B and C). In general, the transfer efficiency of NCA is lower than NCO due to the high density of aliphatic protons and the CACB couplings. Note that typical NC transfer efficiencies range from 20–50%, depending on the RF parameters, MAS rate, spectral dispersion of CA and CO regions (i.e., B₀ field) as well as sample conditions such as protein heterogeneity and extent of deuteration. The CACO and CACB DREAM transfer was optimized using the ¹⁵N CP-NCA-CACO (or CACB) pulse sequence, and the respective signal intensities are 19 and 10% (Figures 1D and E). The DREAM transfer is based on the double quantum recoupling Hamiltonian and inverts the phase of transferred polarization (CO or Cβ spectra of Figures 1D and E) with respect to the source polarization, Cα⁴². Figure 1F shows the ¹³C spectrum obtained with ¹⁵N-CP-NCA-CACX, where the CACX transfer is achieved using 12 ms TOCSY mixing period. Unlike CACB, the CACX TOCSY mixing period transfers the polarization from Cα to all intra-residue side chain carbons. The integrated intensities for CACB and CACX spectra were measured from 10 to 36 ppm, and 65 to 73 ppm, which correspond to Cβ resonances of Ser and Thr residues.

Figure 1:

Comparison of ¹³C spectral intensities obtained from CC and NC polarization transfer periods, with respect to ¹³C CP. The spectra were recorded using conventional ¹³C detected 1D pulse sequences with WALTZ16 ¹H decoupling.

3.2. ¹H- and ¹³C- detected DUMAS (HC-DUMAS) with dual receiver

DUMAS (DUal acquisition Magic Angle Spinning) pulse sequences begin with a SIM-CP preparation period that transfers the polarization from the ¹H spin bath to ¹³C and ¹⁵N nuclei simultaneously^22,51. The polarization transfer efficiency of CP and SIM-CP for GB1 protein is shown in Figure S1 (Supporting Information). The high efficiency of transfer for SIM-CP has been demonstrated previously for both microcrystalline and membrane proteins at slow and fast MAS rates^22,28. The relative intensities of ¹³C spectra are almost identical for CP and SIM-CP schemes, while a minimal signal loss (5–8%) is observed for ¹⁵N. Using a single receiver, we previously showed that the ¹³C and ¹⁵N polarization generated by SIM-CP enables the simultaneous acquisition of pairs of ¹³C- or ¹H-detected spectra with DUMAS and H-DUMAS strategies^22,28. Analogously, the SIM-CP scheme can be used to develop HC-DUMAS sequences in which ¹³C- and ¹H-detected experiments are acquired on two separate channels. Figure 2A shows the HC-DUMAS pulse sequence for the simultaneous acquisition of 2D ¹³C-detected (H)CACO and ¹H-detected (H)NH spectra in 1^st and 2^nd acquisition periods, respectively. The SIM-CP preparation period is followed by ¹³C (t₁^C) and ¹⁵N (t₁^N) evolution periods. After t₁ evolution, the ¹⁵N polarization is stored along the z-direction and a CACO DREAM mixing is applied on the ¹³C channel followed by the 1^st acquisition period ($t_2^C$) that records a 2D (H)CACO spectrum. After the 1^st acquisition, the solvent suppression is achieved by applying the MISSISSIPPI sequence for a time t_supp. The ¹⁵N z-polarization is then transferred to amide protons by applying a 90° pulse followed by a ¹⁵N-¹H back-CP period, and subsequently used to obtain a 2D (H)NH spectrum in the 2^nd acquisition period ($t_2^H$), which is detected on the ¹H channel. A key feature of POE is that the t₁ evolution periods (t₁^C and t₁^N) used for ¹³C and ¹⁵N nuclei can be optimized separately as shown in Table 1²⁶. This enables the operator to optimize the t₁ dwell-times and total evolution periods for each experiment. For the GB1 sample at 600 MHz ¹H frequency, the indirect spectral widths for ¹³Cα and ¹⁵N were 5000 and 2500 Hz, respectively. Therefore, the ¹³Cα t₁ dwell-time (t₁^C) for the (H)CACO experiment was set to 200 μs and 74 t₁ increments were collected for a maximum ¹³C evolution time of 14.6 ms (Table 1). On the other hand, ¹⁵N dwell-time (t₁^N) was set to 400 μs, and two identical t₁ evolution periods (n=2 in Figure 2A), each with 37 increments were recorded for a maximum t₁ evolution period of 14.4 ms. As shown in Table 1, the t₁ evolution parameters satisfy the condition, ni(t₁^C) = n * ni(t₁^N), where ‘ni’ represent number of t₁ increments²². Therefore, rather than oversampling the ¹⁵N spectral width (t₁), these cycles are instead used to acquire two data sets for the (H)NH experiment, which are added together during processing to essentially double the effective number of scans with respect to the (H)CACO experiment^22,26.

Figure 2B shows the 1D CO and H^N spectra of GB1 protein obtained from the first t₁ increment (t₁^C and t₁^N = 0) of the HC-DUMAS pulse sequence using 4 scans. Note that the effective number of scans for H^N spectrum correspond to 8 (ns_eff of Table 1). The integrated signal to noise ratios (S/N) are 23 and 213 for CO and H^N, respectively. During the DREAM spin lock, the aliphatic ¹³C signals are dephased out and display weak intensities (Figure 2B). Figure 2C shows the 2D (H)CACO and (H)NH spectra acquired simultaneously using the HC-DUMAS pulse sequence with 4 scans per t₁ increment and a total experimental time of 23 min. Several resolved peaks in the 2D (H)CACO and (H)NH spectra were assigned using the published data by Pintacuda and co-workers⁴⁴. Figure 2D shows the cross-sections for E56 extracted from the 2D (H)CACO and (H)NH data sets with S/N values of 7 and 59, respectively. Overall, the 2D (H)CACO spectrum has lower S/N but has superior resolution in the direct dimension compared to the 2D (H)NH spectrum. The average CO and H^N full-width-at-half-maximum (FWHM) line widths in the 2D (H)CACO and (H)NH spectra are 55 and 242 Hz, respectively. Note that the longer T₂ relaxation times (or narrow line widths) of the CO nuclei resulted in the splitting of the resonances caused by CACO homonuclear J-couplings (Figures 2C and 2D). To further narrow the CO line widths, it is possible to use the BASHD acquisition method with ¹³C-¹³C homonuclear decoupling⁵². Note that for HC-DUMAS experiments, the high signal intensity of ¹H can be exploited to record other 2D spectra, such as a (H)NHH experiment, by incorporating a HH mixing period prior to the 2^nd acquisition. Similarly, the CACO DREAM mixing can be replaced by CACB or CXCX TOCSY sequences for ¹³C-detected experiments. Overall, the HC-DUMAS strategy is flexible enough to incorporate other mixing methods and optimize the sensitivity of ¹³C- and ¹H-detected spectra⁵³.

3.3. ¹H- and ¹³C-detected MEIOSIS (HC-MEIOSIS) with dual receiver

In the MEIOSIS (Multiple ExperIments via Orphan SpIn operatorS) and H-MEIOSIS (¹H-detected Multiple ExperIments via Orphan SpIn operatorS) methods two pairs of ¹³C- and ¹⁵N-edited spectra are recorded with either ¹³C- or ¹H-detection using a single receiver^24,28. A version of the HC-MEIOSIS (¹H- and ¹³C- detected MEIOSIS) pulse sequence using two receivers is shown in Figure 3A. After SIM-CP, the ¹³C and ¹⁵N polarization pathways are evolved during the t₁ period followed by bidirectional SPECIFIC-CP transfer from N to Cα and Cα to N (NCA_trans and CAN_trans). Note that some residual polarization remains on both ¹³C and ¹⁵N after the SPECIFIC-CP sequence (CC_res and NN_res)^24,29,31. The NCA_trans and CC_res pathways are encoded within the ¹³C polarization, whereas the CAN_trans and NN_res pathways are encoded in the ¹⁵N polarization. To recover these pathways, both ¹⁵N and ¹³C spin operators are flipped to the z-direction by applying 90° pulses after SPECIFIC-CP. The CC_res and NCA_trans pathways evolved under TOCSY mixing, followed by a 90° readout pulse that enables the acquisition of 2D (H)CXCX and (H)N(CA)CX spectra in the 1^st acquisition period using the ¹³C receiver. The polarization from NN_res and CAN_tran pathways is transferred to amide protons by applying a 90° pulse and NH back-CP period followed by detection on the ¹H channel. This 2^nd acquisition period gives 2D (H)NH and (H)CA(N)H spectra. As shown previously for the MEIOSIS and H-MEIOSIS experiments, the phase (ϕ) flip of the ¹⁵N SPECIFIC-CP pulse does not affect the residual polarization of the NN_res and CC_res pathways, but it does invert the phase of the CAN_trans and NCA_trans pathways²⁴. Therefore, the two polarization pathways in each of the 1^st and 2^nd acquisition periods are decoded by adding and subtracting the two data sets recorded with ϕ set to x and −x. Of course, the sensitivity of HC-MEIOSIS spectra depends on the intensities of the respective polarization pathways. Figure S2 shows the 1D MEIOSIS pulse sequence and corresponding intensities of the four polarization pathways (CC_res, CAN_trans, NN_res, and NCA_trans) for GB1 using ¹³C detection. At the optimal SPECIFIC-CP transfer period (τ = 9.5 ms), approximately 30% of the ¹⁵N residual polarization (NN_res) is observed; whereas the residual polarization for ¹³Cα and side chain atoms (Cβ, Cγ etc. or ¹³C_{side chain}) is 52 and 74%, respectively. Figure 3B shows the 1D spectra of ¹³C and ¹H for GB1 obtained from the first increment (t₁ = 0) of the HC-MEIOSIS experiment collected with 8 scans and a TOCSY mixing period of 12 ms. In the 1D (H)CXCX and (H)N(CA)CX spectra, the S/N of the ¹³C resonances (0–70 ppm) are 41 and 15, respectively. As expected, the 1D ¹³C spectrum from the (H)N(CA)CX experiment has the lowest S/N since it includes both heteronuclear (NCA) and homonuclear CACX TOCSY transfer periods. The integrated S/N of the ¹H resonances between 6 and 12 ppm for the second set of 1D HC-MEIOSIS spectra, i.e., (H)CA(N)H and (H)NH, were 80 and 394, respectively. Using single receiver MEIOSIS or H-MEIOSIS experiments, the CC_res pathway showed the highest S/N in the 2D experiments (¹³C-detected DARR or ¹H-detected (H)CH)^24,28. On the other hand, the CC_res pathway in the HC-MEIOSIS experiment with dual receiver (i.e., ¹³C-detected (H)CXCX) shows a lower S/N compared to the NN_res and CAN_trans pathways that are recorded via ¹H-detected experiments, i.e., (H)CA(N)H and (H)NH. Even though the (H)NH spectrum is recorded from the residual ¹⁵N polarization (~30%, Figure S2), it displays higher S/N with respect to the (H)CA(N)H spectrum, which requires an additional CAN transfer period. The 2D HC-MEIOSIS spectra of GB1 were recorded with 8 scans and consist of two data sets acquired in an interleaved manner with ϕ = x, −x (Figure 3C). For the ¹³C evolution (t₁^C), the dwell-time was set to 100 μs and 148 increments were collected, corresponding to a maximum evolution time of 14.7 ms (see Table 1). For the ¹⁵N indirect dimension, the t₁ dwell-time was set to 400 μs with four identical t₁ evolution periods (n=4 in Figure 3A) each with 37 increments and corresponds to a maximum evolution of 14.4 ms. The four ¹⁵N-edited 2D (H)N(CA)CX and (H)NH data sets were then added using post-acquisition processing scripts²⁶. Essentially, the effective number of scans (ns_eff in Table 1) for the ¹⁵N-edited 2D (H)N(CA)CX and (H)NH spectra resulted to be four times higher than the ¹³C-edited 2D (H)CXCX and (H)CA(N)H spectra. Figure 3C shows the 1D cross-sections from the four 2D HC-MEIOSIS spectra. ¹H-detected 2D (H)NH and (H)CA(N)H spectra have higher S/N, but their limited resolution can often prohibit residue-specific assignment. Conversely, the ¹³C-detected 2D (H)CXCX and (H)N(CA)CX spectra resolve many more peaks, but with lower S/N. The latter is evident from the NCA region of the N(CA)CX spectrum, which gives rise to 43 resolved peaks of GB1. In the (H)CXCX spectrum of Figure 3C, the intensities of the upper off-diagonal cross-peaks associated with residual polarization of ¹³C_sidechain (72%) are relatively higher compared to the bottom off-diagonal cross-peaks originating from ¹³Cα residual polarization (52%). A similar observation was made for the 2D DARR spectrum obtained from the single receiver MEIOSIS pulse sequence²⁴. The cross-peak intensities of the 2D (H)CXCX and (H)N(CA)CX spectra also depend on the efficiency of the homonuclear CC mixing period, which can be further improved using faster MAS rates (100–110 kHz)⁴⁴. Similarly, the HC-MEIOSIS experiment can also be implemented with other CC mixing sequences such as CACO, CACB, PAR, and CHHC that can provide intra- and inter-residue CC correlations^54–56. The four 2D HC-MEIOSIS spectra in Figure 3C were recorded in approximately 3 h (T_exp). While the 2D (H)N(CA)CX and (H)CA(N)H spectra were recorded using transferred polarization, the 2D (H)CXCX and (H)NH spectra originate from the residual polarization, 72% for ¹³C_{side chain} and 30% for ¹⁵N (i.e., a factor of k = 0.72 and 0.3). To obtain these spectra at the same S/N using conventional single-acquisition sequences, the estimated time (k²*T_Exp) for each experiment is 1.6 h for the (H)CXCX experiment (~0.72²*3 h), 0.3 h for the (H)NH experiment (~0.3²*3 h) and 3 h for each of the (H)N(CA)CX and (H)CA(N)H experiments for a total of 7.9 h (3+3+1.6+0.3). Therefore, the HC-MEIOSIS multi-acquisition sequence, which requires just 3 h, provides a net savings of 62% in experimental time.

3.4. ¹H- and ¹³C-detected MAeSTOSO (HC-MAeSTOSO) with dual receiver

Multiple acquisitions via sequential transfer of orphan spin polarization (or MAeSTOSO) strategy exploits ¹⁵N and ¹³C residual polarization originating from NC SPECIFIC-CP and CH or NH back-CP periods^25,28. Figure 4A shows an example of the dual receiver HC-MAeSTOSO-4 pulse sequence. In this experiment, after SIM-CP, the ¹³C and ¹⁵N polarization is evolved during t₁. Subsequently, the¹⁵N polarization is stored along the z-direction and the ¹³C polarization is used to acquire a ¹³C-detected (H)CXCX TOCSY experiment (1^st acquisition period). After the 1^st acquisition, a MISSISSIPPI sequence is used on the ¹H channel to suppress the solvent signals. Then, the ¹⁵N polarization is transferred to Cβ using an NCA SPECIFIC-CP followed by a CACB DREAM transfer. A ¹³C-¹H back-CP transfers the Cβ polarization to the covalently attached Hβ atoms, providing a ¹H-detected 2D (H)N(CACB)H spectrum in the 2^nd acquisition. At the end of the first SPECIFIC-CP period, ~30% of residual polarization remains on ¹⁵N and is stored along the z-direction by applying a 90° pulse (Figure S2). After the 2^nd acquisition, the ¹⁵N residual polarization is transferred to Cα and then to Hα using another SPECIFIC-CP followed by a ¹³C-¹H back-CP period. The Hα polarization is then detected in the 3^rd acquisition to yield a 2D (H)N(CA)H spectrum. Note that the residual polarization at the end of the second SPECIFIC-CP period is approximately 9% (30% of the 30% remaining after the first SPECIFIC-CP), which is stored along the z-direction by a 90° pulse. After the 3^rd acquisition, this 9% residual polarization is transferred to the amide protons using a ¹⁵N-¹H back-CP and then detected in the 4^th acquisition to give a 2D (H)NH spectrum. Both NCA and CACB transfer periods are applied prior to the 2^nd acquisition since the intensity of the ¹⁵N polarization stored after the 1^st acquisition is relatively high. On the other hand, the 3^rd acquisition uses 30% residual ¹⁵N polarization, therefore only the NCA transfer period was incorporated prior to ¹³C-¹H back-CP period. To recover the 9% ¹⁵N residual polarization a ¹⁵N-¹H back-CP is used prior to the 4^th acquisition period. A similar strategy was used for single receiver MAeSTOSO pulse sequences to optimize the sensitivity of multiple 2D spectra acquired simultaneously with NC and CC mixing periods²⁵.

Figure 4B shows ¹³C and ¹H 1D spectra of GB1 recorded from the first increment (t₁ = 0) of the HC-MAeSTOSO-4 experiment using 16 scans. The 1D integrated S/N for 2D (H)CXCX, (H)N(CACB)H, (H)N(CA)H, and (H)NH are 51, 22, 34, and 100, respectively. In spite of the weak ¹⁵N residual polarization, the S/N of H^N detected in the 4^th acquisition is higher than the Hα and Hβ spectra recorded in the 2^nd and 3^rd acquisitions, respectively. This is because the Hβ spectrum requires NCA and CACB transfer steps, while the Hα spectrum only requires a NCA transfer step prior to ¹³C-¹H back-CP period. In fact, these additional transfer periods significantly lower the signal intensities of Hα and Hβ resonances. The 2D (H)CXCX, (H)N(CACB)HB, (H)N(CA)HA, and (H)NH spectra of GB1 (Figure 4C) were acquired using 16 scans per t₁ increment for a total of 3 h, as reported in Table 1. The DREAM CACB transfer inverts the phase of Cβ with respect to Cα peaks, therefore, the corresponding Hα and Hβ peaks in the 2D (H)N(CACB)HB spectrum have opposite signs⁴². Note that all three ¹H-detected spectra of Figure 4C share the same ¹⁵N evolution period (t₁^N), and thus detect the same ¹⁵N chemical shift evolution for Hα, Hβ, and H^N resonances. As an example, Figure 4C shows selected ¹⁵N intra-residue correlations with the corresponding Hα, Hβ, and H^N atoms. The 1D cross-sections of the A26 resonances with respective S/N and FWHM values are shown in Figure 4D. The S/N for the Cβ, Hα, Hβ, and H^N peaks of A26 are 24, 10, 12, and 27, respectively. In general, the required number of scans for recording multi-dimensional ssNMR experiments are estimated from the 1D integrated S/N from the 1^st increment. However, the relative S/N of resulting 2D or 3D peaks may not follow the same proportionality. This is because the relative S/N of various residues depend on the efficiency of homo- and hetero-nuclear polarization transfer, which is variably affected by protein dynamics and spectral parameters such as the offset frequency and linewidths. As shown in Figures 3B and 3D, the 1D integrated S/N in the (H)NH spectrum is three times higher than the (H)N(CA)H, but the S/N of A26 peak in the 2D (H)NH spectrum is only 2.2 times higher compared to (H)N(CA)H. The linewidths of Hα, Hβ, and Hγ resonances are in the range of 200 to 270 Hz, while ¹³C linewidths in the 2D (H)CXCX spectrum were considerably narrower (~25–70 Hz). Also note that the (H)CXCX spectrum from the HC-MEIOSIS experiment (Figure 3A) uses ~50 to 70 % of the ¹³C residual polarization, whereas the (H)CXCX spectrum in the HC-MAESTOSO-4 (Figure 4A) is generated from the full ¹³C polarization. Therefore, the relative sensitivity of the (H)CXCX spectrum acquired from HC-MAeSTOSO-4 is higher than the one obtained from HC-MEIOSIS. The four 2D HC-MAeSTOSO-4 spectra of Figure 4C were recorded in only 3 h (T_exp). To further summarize, while the (H)CXCX and (H)N(CACB)H spectra utilize the polarization transferred from the initial SIM-CP, the (H)N(CA)H and (H)NH spectra are recorded from 30 and 9% (or a k factor of 0.3 and 0.09) residual polarization originating from the SPECIFIC-CP periods. Therefore, using the single acquisition experiments, each of the (H)CXCX and (H)N(CACB)H spectra would require approximately 3 h, whereas the (H)N(CA)H with 30% (k = 0.3) and (H)NH with 9% (k = 0.09) polarization require 0.27 h (~0.3²*T_exp) and 0.02 h (~0.09²*T_exp)²⁸. In other words, the total experimental time for all four spectra acquired separately would be 6.29 h (3.0+3.0+0.27+0.02), whereas using the HC-MAeSTOSO-4 strategy they were acquired in 3 h for a 53% reduction in acquisition time.

Recently, we demonstrated that it is possible to recover the residual polarization from both NC SPECIFIC-CP, and NH (or CH) back-CP periods²⁸. This led to the simultaneous acquisition of ten ¹H-detected experiments (H-MAeSTOSO-10). The dual receiver HC-MAeSTOSO-10 is shown in Figure 5A and acquires four ¹³C- and six ¹H-detected experiments using five acquisition periods. In each acquisition period, a pair of ¹³C- and ¹⁵N-edited 2D spectra are recorded using two polarization pathways (represented by arrows in Figure 5A). Since these spectra are acquired with alternate phases (ϕ set to x and −x), the two polarization pathways in each acquisition period are decoded by adding and subtracting the data sets (similar to HC-MEIOSIS of Figure 3A). Note that the ¹³C detection periods (1^st and 2^nd acquisition) of the HC-MAeSTOSO-10 experiment resembles the MEIOSIS pulse sequence at slow spinning speeds²⁴. In the HC-MAeS-TOSO-10 experiment, after first SPECIFIC-CP period, four MEIOSIS polarization pathways, CC_res, CAN_trans, NN_res, and NCA_trans (color-coded in Figure 5A), are encoded in the ¹³C and ¹⁵N polarization. The CC_res and NCA_trans pathways are subjected to a TOCSY mixing followed by the 1^st acquisition period recording the ¹³C- and ¹⁵N-edited (H)CXCX and (H)N(CA)CX spectra. After the 1^st acquisition period, the ¹⁵N polarization from NN_res and CAN_trans pathways is transferred to CO via SPECIFIC-CP followed by 2^nd acquisition period that records the (H)CA(N)CO and (H)NCO spectra. At the end of second SPECIFIC-CP, the CAN_trans and NN_res pathways are encoded in the ¹⁵N residual polarization. Note that the application of two SPECIFIC-CP periods reduces the residual polarization to less than 10%. Nonetheless, this residual polarization is enough to acquire ¹H-detected experiments. After the 2^nd acquisition period, a MISSISSIPPI sequence is used for solvent suppression. Next, the two pathways encoded in the ¹⁵N polarization are transferred to amide protons by the NH back-CP scheme followed by HH RFDR mixing. The latter makes it possible to record the (H)NHH and (H)CA(N)HH spectra under ¹³C and ¹⁵N WALTZ16 decoupling applied during the 3^rd acquisition. The ¹⁵N residual polarization after the NH back-CP period is stored along the z-direction by a 90° pulse on ¹⁵N. The residual ¹⁵N polarization from the NH back-CP is recovered after the first ¹H acquisition by adding an additional WALTZ16 period (τ₁), which keeps the ¹⁵N polarization in the z direction²⁸. After the 3^rd acquisition and the following τ₁ period, the ¹⁵N residual polarization is exploited for two additional pairs of ¹³C and ¹⁵N edited (H)NHH and (H)CA(N)HH spectra in the 4^th acquisition, and (H)NH and (H)CA(N)H spectra in the 5^th acquisition. Compared to single acquisition experiments, the time saving of HC-MAeSTOSO-10 is similar to HC-MEIOSIS, i.e a time saving of up to 62%. In other words, additional acquisition periods (3^rd, 4^th and 5^th) only contribute to marginal time savings because of weak residual polarization (less than 10%). Figure 5B shows ten 1D spectra of GB1 acquired using the dual-receiver HC-MAeSTOSO-10 pulse sequence with 32 scans and using the first increment (t₁ = 0). The integrated S/N values of ¹³C- and ¹H-detected spectra are shown in Figure 5B. In the ¹H spectra, the water peak (indicated by an asterisk in Figure 5B) is not completely suppressed, which is probably due to the additional ¹H pulses required for the RFDR mixing period. As shown in Figure 5B, in spite of the weak residual polarization, the S/N for the ¹H-detected spectra in the 3^rd and 4^th acquisitions are comparable to those obtained with the ¹³C-detected experiments recorded in the 1^st and 2^nd acquisitions. Note that the application of (H)NHH and (H)CANHH 2D or 3D experiments requires higher proton resolution (~100 Hz ¹H linewidth), which is achievable using combination high field NMR and fast MAS rates. In fact, 3D (H)NHH and (H)CHH experiments were utilized successfully to measure H-H distances for GB1 protein at 1 GHz using MAS rate of 100 kHz⁴⁴.

3.5. ¹H-, ¹³C-, and ¹⁵N-detected DUMAS (HCN-DUMAS) with triple receiver

Previous POE were designed using ¹³C and ¹⁵N polarization obtained from the ¹H spin bath via a SIM-CP preparation period²³. Using a triple receiver, we are able to assess both ¹³C and ¹⁵N transferred polarization as well as the residual ¹H polarization that remains at the end of a SIM-CP period. This experiment is similar to the first multi-acquisition CP experiment on small organic molecules carried out with residual ¹H polarization under CW decoupling⁵⁷. In general, following a CP or SIM-CP period a residual polarization is left on ¹H nuclei. This phenomenon can be understood by analyzing the CP dynamics in DQ (double quantum) and ZQ (zero quantum) sub-spaces^58–60, where only a portion of the ¹H spin polarization, namely the ZQ operator, is perpendicular to the ZQ dipolar Hamiltonian and evolves during CP. The oscillatory nature of the transferred ¹³C and ¹⁵N polarization is significantly damped by strong ¹H-¹H spin diffusion and T₁ρ relaxation mechanisms, which leads to incomplete polarization transfer from ¹H to ¹³C and ¹⁵N⁶¹. Even with fast MAS rates, strong HH dipolar couplings of fully protonated samples affect the CP transfer efficiency. In fact, these CP oscillations can be observed in samples where dipolar couplings are significantly scaled-down as for liquid crystals^59,61.

Figure 6A shows the SIM-CP pulse sequence for simultaneous acquisition of residual ¹H and transferred ¹³C and ¹⁵N polarization. After a SIM-CP period, the ¹H receiver detects the residual polarization; whereas the ¹³C and ¹⁵N transferred polarization is stored along the z-direction by 90° pulses. After the ¹H acquisition, another pair of 90° pulses on ¹³C and ¹⁵N creates transverse polarization that is simultaneously detected using ¹³C and ¹⁵N receivers, while decoupling the ¹H channel with WALTZ16 sequence. Note that for protein samples, ¹H detection requires solvent suppression using hard pulses (MISSISSIPPI), which dephase ¹H residual polarization. For this reason, we chose a powdered sample of fMLF that does not require solvent suppression⁶². The pulse sequence of Figure 6A was carried out on the fMLF sample by varying the SIM-CP contact time (t). The integrated intensities for ¹H, ¹³C, and ¹⁵N resonances were measured at various t and plotted in Figure 6B. The 1D spectra at t=0 and 1 ms are shown in Figure 6C. At t=0, maximum ¹H intensity was observed, whereas ¹³C and ¹⁵N show no signal. The SIM-CP intensities of ¹³C and ¹⁵N gradually increases at higher t. For t=1 ms, both ¹³C and ¹⁵N intensities reached a plateau. Interestingly, the integrated ¹H intensity from residual polarization drops down to 32% for side chain protons (−3 to 3 ppm), and to 18% for protons resonating between 3 to 10 ppm.

Figure 6:

(A) 1D triple receiver SIM-CP experiment for monitoring residual ¹H, and transferred ¹³C and ¹⁵N polarization. (B) Plot of the normalized intensities for SIM-CP polarization pathways obtained from the integrated intensities of ¹H, ¹³C, and ¹⁵N 1D spectra of fMLF tripeptide sample, at contact times ranging from 0 to 2 ms. (C) ¹H, ¹³C, and ¹⁵N 1D spectra at SIM-CP contact times 0 and 1 ms.

The SIM-CP sequence can be exploited to acquire three 2D homonuclear correlation experiments (Figure 7A) using two acquisition periods. In this pulse sequence, namely HCN-DUMAS, ¹³C and ¹⁵N polarization is stored along the z-direction after SIM-CP. The residual ¹H polarization is evolved during a t₁ period followed by a 90° pulse, a ¹H-¹H RFDR mixing, and another 90° pulse. Then, the ¹H receiver records a 2D ¹H-¹H correlation experiment in the 1^st acquisition period (t₂^H). After the ¹H acquisition, the ¹³C polarization is evolved during t₁ followed by a 90° pulse and TOCSY mixing. At the end of the TOCSY mixing period, the ¹³C polarization is kept along the z-direction by setting the duration of the TOCSY mixing to ‘n/4’ times of the WALTZ16 period, where ‘n’ is an integer²⁸. After the ¹³C TOCSY mixing, the ¹⁵N z-polarization is flipped into the transverse plane by a 90° pulse and then evolved for a t₁ period. Unlike ¹H and ¹³C, the homonuclear dipolar- and J- couplings for ¹⁵N nuclei are very weak. Therefore, homonuclear ¹⁵N-¹⁵N transfer is achieved through protons using the NHHN sequence, which consists of a CP transfer from ¹⁵N to ¹H followed by a HH RFDR mixing and another CP from ¹H to ¹⁵N^56,63,64. After the NHHN transfer steps, a 90° pulse on ¹³C flips the z-polarization to the transverse plane, and ¹⁵N and ¹³C polarization is then detected in the 2^nd acquisition period by separate receivers under ¹H decoupling, yielding 2D ¹³C-¹³C and ¹⁵N-¹⁵N homonuclear correlation spectra. Figure 7B shows the 2D HH, (H)CXCX, and N(HH)N homonuclear correlation spectra of the fMLF peptide, acquired simultaneously using the HCN-DUMAS pulse sequence. The successful implementation of double and triple receiver experiments relies on proper optimization of t₁ evolution parameters for all nuclei. The t₁ dwell-time is inversely proportional to the spectral width, whereas the number of t₁ increments depends on T₂ relaxation.

As for other POE, the HCN-DUMAS experiment can be implemented with separate t₁ evolution parameters for ¹H, ¹³C, and ¹⁵N nuclei (t₁^H, t₁^C, and t₁^N in Figure 7A). As shown in Table 2, the t₁ evolution parameters satisfy the condition, ni(t₁^C) = n1 * ni(t₁^H) = n2 * ni(t₁^N). The spectra of Figure 7B were recorded using 80 μs t₁ dwell time for ¹H and ¹³C (Table 2). The ¹³C polarization was evolved for 7.9 ms with a maximum t₁ evolution obtained with 100 t₁ increments. For ¹H evolution, four identical t₁ periods (n₁=4) were implemented using 25 t₁ increments that corresponds to a maximum ¹H t₁ evolution of 1.92 ms. Similarly, for ¹⁵N t₁ evolution, the dwell-time was set to 320 μs and four identical t₁ periods (n₂=4) were implemented using 24 increments with a maximum t₁ evolution of 7.68 ms (see Table 2). Note that for fully protonated samples, ¹H T₂ relaxation is relatively short compared to ¹³C and ¹⁵N. Therefore, the HH 2D spectrum of Figure 7B was recorded with a shorter t₁ evolution time (1.92 ms). The four identical 2D N(HH)N and HH data sets were added during the processing. The ¹H-¹H RFDR mixing times for HH and NHHN experiments were set to 0.96 and 1.92 ms, respectively. The ¹³C-¹³C TOCSY mixing was set to 12 ms, corresponding to five cycles of WALTZ16. Inter- and intra-residue correlations are shown in Figure 7C together with 1D cross-sections of the indirect dimensions for H, Cα, and N resonances of the Leu residues. The 2D N(HH)N spectrum of fMLF showed intense cross-peaks between the neighboring residues pairs, namely M-L and L-F, and relatively weaker cross-peaks between the 1^st and 3^rd residues (M and F). In the 2D CC spectrum, we observed total correlation of intra-residue ¹³C atoms for each of the three residues. In general, the peak intensities of homonuclear correlation spectra are similar for lower and upper off-diagonal cross-peaks. However, in the 2D HH spectrum obtained from HCN-DUMAS, the intensities of lower off-diagonal cross peaks (between 3 to 10 ppm along indirection ¹H dimension) are much lower than the upper off-diagonal cross-peaks (0–3 ppm) that originated from high intensity sidechain protons (Figure 6C). Nevertheless, the high sensitivity of ¹H detection leads to ¹H cross peaks in both upper and lower off-diagonal regions of 2D spectra, as shown in the 1D cross sections of Figure 7C.

Using the HCN-DUMAS strategy, other combinations of 2D experiments can be designed. For example, the ¹H residual polarization can be exploited to acquire a DQSQ experiment in the 1^st acquisition⁶⁵. Similarly, inter-residue CC correlations can be obtained in the ¹³C acquisition by replacing the TOCSY with a C(HH)C mixing period. The 2D HCN-DUMAS spectra of the fMLF sample (Figure 7B) were acquired using 8 scans for a total experimental time of 1.4 h. As shown in Figure 7C, in spite of using residual ¹H polarization, the high sensitivity of ¹H channel leads to higher S/N of HH cross peaks compared to the (H)CXCX and N(HH)N spectra. While the N(HH)N spectrum can be highly informative from having narrower linewidths, its feasibility for protein samples would be limited since longer experimental times are required to overcome the poor sensitivity associated with ¹⁵N-detection. The HCN-DUMAS can also be implemented with multiple ¹H- and ¹³C-detected experiments, while acquiring the same N(HH)N experiment with ¹⁵N detection. In this experimental setup, HCN-DUMAS will have to be coded as two separate pulse sequences in two separate scripts, each with different HH and CC mixing times, but sharing the same NHHN mixing time. The two 2D HH and CC experiments would be processed as separate experiments, while the two resulting N(HH)N data sets can be added for signal enhancement. A similar experimental setup was previously described for a single receiver ¹³C-detected DUMAS experiment, which records 2D NCA and NCO spectra while acquiring a single DARR spectrum⁵³.

4. Discussion

Multi-receiver technology represents a stepping-stone to expand the applications of POE methods to various multi-dimensional ssNMR experiments. To determine the efficiency of multi-receiver experiments, we used microcrystalline GB1 protein and fMLF tripeptide samples. The relative sensitivity of multi-receiver experiments depend on different parameters such as the gyromagnetic ratio, length of pulse sequences, as well as probe and RF console features such as quality factor (Q) and receiver digitization^66–68. In general, the main drawback of multi-receiver experiments is the intrinsic low sensitivity of low γ nuclei such as ¹³C and ¹⁵N. Therefore, dual and multi-receiver experiments need to be acquired with a high number of scans to accumulate sufficient polarization for ¹³C and ¹⁵N-detected spectra, while highly sensitive ¹H spectra would require far fewer scans^68,69. Nevertheless, POE methods provide new ways to optimize the sensitivity, and therefore, make full use of multi-receiver technology.

As shown in Figure 2, the 2D (H)NH is a highly sensitive experiment, therefore, a ¹H-detected (H)NH experiment is combined with another highly sensitive ¹³C-detected experiment such as CACO using HC-DUMAS strategy. On the other hand, the ¹³C-detected (H)CXCX and (H)N(CA)CX are low-sensitivity experiments due to NCA and CXCX transfer periods. Therefore, in the HC-MEIOSIS experiment (Figure 3), (H)CXCX and (H)N(CA)CX are combined with another pair of low-sensitivity ¹H detected experiments (H)CA(N)H and (H)NH that utilize transferred and residual polarization from the SPECIFIC-CP period. Similarly, in the HC-MAeSTOSO-4 experiment (Figure 4), the low-sensitivity ¹³C-detected (H)CXCX experiment is combined with three other low-sensitivity ¹H-detected experiments, (H)N(CACB)H, (H)N(CA)H, and (H)NH¹¹. In comparison to single acquisition experiments, the time saving of HC-DUMAS is ~50%; whereas the simultaneous acquisition of four to ten spectra with HC-MEIOSIS and HC-MAeSTOSO lead to approximately 53 to 62% reduction in acquisition time. The application of triple receiver HCN-DUMAS experiment requires highly concentrated samples to achieve sufficient S/N for ¹⁵N-detected spectra. An alternative way to optimize the experimental time for HCN-DUMAS is to record multiple 2D HH and CC experiments, while acquiring a single N(HH)N experiment⁵³. However, for fully protonated samples, broad ¹H linewidths (200–400 Hz) represent a major drawback and lead to poorly resolved HH cross-peaks (Figure 7C). On the other hand, the ¹H linewidths can be significantly narrowed for perdeuterated protein samples that give high resolution ¹H spectra with 30 to 50 Hz ¹H linewidths even at lower MAS rates of 30 to 40 kHz^70–72. In comparison to fully protonated samples, the efficiency of SIM-CP is lower for perdeuterated samples due to the diluted ¹H spin bath. As a future development of our approach, we will be testing the efficiency of SIM-CP and POE methods on fractionally perdeuterated protein samples.

High sensitivity and resolution are fundamental spectral requirements for the atomic resolution analysis of proteins by NMR spectroscopy. In spite of the low sensitivity, ¹³C-detected experiments are routinely used in solution- and ssNMR. In fact, ¹³C-detection offers unique advantages such as narrower linewidths and broad spectral dispersion. For example, in solution NMR, ¹³C-detected CON and CAN based experiments are used for analyzing protein samples (such as IDPs or intrinsically disordered proteins) that are poorly dispersed in the ¹H dimension^73,74. Similarly, ¹H-detected fast MAS ssNMR spectra, such as CP-HSQC, suffer from broader ¹H linewidths in fully protonated samples. For the GB1 spectra shown in Figures 2–4, the number of resolved peaks in a ¹H-detected (H)NH spectrum is 19 (Figure 2C). Whereas for Cα and CO spectral regions in the ¹³C-detected (H)N(CA)CX (Figure 3C) and (H)CACO (Figure 2C) spectra, number of resolved peaks are 43 and 26, respectively. Note that the ¹H line widths of microcrystalline GB1 are much narrower compared to more challenging heterogeneous systems such as fibrillar and membrane-bound proteins^75,76. In other words, the ¹H resolution can be even more compromised for such large systems. In practice, before proceeding to spectral analysis, it is necessary to test several ¹H- and ¹³C-detected 2D experiments to optimize both the sensitivity and resolution. In addition, we recommend testing a range of protein sample conditions. For example, in the case of membrane proteins, it is important to optimize temperature, spinning speed, lipid composition, protein-to-lipid ratio as well as hydration levels. To this extent, the multi-receiver 2D experiments can provide multiple protein fingerprints, giving a comprehensive view of the ¹H, ¹³C, and ¹⁵N spectral quality, which will be instrumental to optimize sample preparation/conditions as well as designing 3D experiments. Similarly, the 2D fingerprints are routinely used for screening protein samples with various mutations as well as by titrating with ligand or drug molecules. Such applications can be envisioned with POE to evaluate biological processes by recording multiple protein fingerprints simultaneously.

POE using fast MAS rates and multi-receiver technology can unleash the full potential of ssNMR applications to biomacromolecules. Several solution-NMR pulse sequences, such as PANSY and afterglow, have been developed for multi-receiver experiments^66,69,77,78, while relatively few have been developed for ssNMR due to the limitations of ¹H-detected experiments for solid samples. However, recent developments in fast MAS probes and perdeuteration have increased the sensitivity and resolution of ¹H-detected experiments. In this work, we implemented the POE strategies to various multi-receiver experiments and evaluated relative S/N and FWHM of ¹H, ¹³C and ¹⁵N detected spectra. To the best of our knowledge, this is the first experimental demonstration of DUMAS, MEIOSIS, and MAeSTOSO methods using multi-receiver technology and ultrafast magic angle spinning. Dual and multi-receiver experiments have also been demonstrated by other research groups. For example, the simultaneous acquisition of ¹³C- and ¹⁵N-detected experiments were shown for spectrometers equipped with two receivers at slow MAS rates⁶⁸. Also, a recent study shown the application of triple receiver experiments (¹⁹F, ³¹P, and ²⁷Al) to organometallic compounds⁷⁹. Similarly, window ¹H detection was used for dual receiver DUMAS and MEIOSIS experiments at slow MAS rates³². Dual receiver experiments were also implemented by concatenation of a ¹H-detected 3D with ¹³C-detected 2D experiments at fast MAS rates using CP as preparation period⁸⁰. More recently, we also demonstrated the application of POE methods using single receiver ¹H detected ultrafast MAS experiments²⁸. Multiple polarization pathways can also be detected in a single acquisition using Hadamard encoding and decoding of spin operators⁸¹. Such experiments have been reported for oriented (static) as well as for MAS pulse sequences^{24,30,60,82,83}. On the other hand, the pulse sequences reported in this manuscript, are based on the multi-acquisition and enables the incorporation of various CC, NC, and HH mixing periods. The experiments reported in this work will help select and design novel ultrafast MAS POE methods using dual and triple receiver experiments. Although in principle it is possible to concatenate several different pulse sequences, it is important to select the POE method based on their performance. Therefore, the assessment of the relative S/N for ¹H, ¹³C, and ¹⁵N detected experiments presented here will be instrumental to choose and tailor the experiments that fit the features of the protein sample under analysis.

Since HC-MEIOSIS and HC-MAeSTOSO utilize multiple transferred and weak residual polarization pathways, these methods are suited for combining two to five pairs of low and high sensitivity experiments. On the other hand, the HC-DUMAS method uses high intensity transferred polarization pathways, and therefore more suitable for combining a pair of high sensitivity experiments such as (H)CACO and (H)NH. The HC-MAeSTOSO methods incorporate the elements of HC-DUMAS and HC-MEIOSIS, but uses weak residual polarization for the 3^rd through 5^th acquisitions periods (Figures 4 and 5). For this reason, the 1^st and 2^nd acquisitions of HC-MAeSTOSO, are used for recording low sensitivity experiments that require a high number of scans, and enables the accumulation of enough residual polarization for the 3^rd, 4^th, and 5^th acquisitions. Finally, the dual- or multi-receiver experiments reported here can be implemented on spectrometers equipped with a single receiver setup similar to solution NMR UTOPIA pulse sequences⁷⁸ by switching the single receiver back and forth for ¹H, ¹³C, and ¹⁵N acquisitions. Considering the advantages of ¹³C detection, we anticipate that the benefits of multi-receiver POE methods will be further enhanced when fast MAS probes are developed with improved ¹³C sensitivity.

5. Conclusions

POE methods allow the concatenation of almost all types of biomolecular ssNMR experiments, boosting the capability of ssNMR spectrometers at least two-fold. The combination of POE and multiple receivers opens up new horizons for pulse sequence development and propels the spectroscopic characterization of complex biomolecular systems. In this work, we have developed a new sub-class of POE (HC-DUMAS, HC-MEIOSIS, HC-MAeSTOSO, and HCN-DUMAS) with multiple receivers and ultrafast magic angle spinning to combine the benefits of ¹H-, ¹³C-, and ¹⁵N-detected experiments. We anticipate that these experiments will be even more powerful when combined with other methods such as PRE, DNP, and non-uniform sampling.