Modest Offset Difference Internuclear Selective Transfer via Homonuclear Dipolar Coupling

Homonuclear dipolar recoupling is routinely used for magic-angle spinning NMR-based structure determination. In fully protonated samples, only short proton–proton distances are accessible to broadband recoupling approaches because of high proton density. Selective methods allow detection of longer distances by directing polarization to a subset of spins. Here we introduce the selective pulse sequence MODIST, which recouples spins that have a modest chemical shift offset difference, and demonstrate it to selectively record correlations between amide protons. The sequence was selected for good retention of total signal, leading to up to twice the intensity for proton–proton correlations compared with other selective methods. The sequence is effective across a range of spinning conditions and magnetic fields, here tested at 55.555 and 100 kHz magic-angle spinning and at proton Larmor frequencies from 600 to 1200 MHz. For influenza A M2 in lipid bilayers, cross-peaks characteristic of a helical conformation are observed.


SIMULATIONS
In this section we give a brief numerical comparison of SPR and MODIST pulses. However recoupling spins of the same type of functional groups (for example, 1 H N -1 H N ) with these pulses is more challenging, since it is difficult to achieve pure double quantum or zero quantum recoupling for all correlations. Therefore, starting with SPR5 we came up with a new sequence, The difference in the phase cycling changes the properties of SPR pulses, as detailed below. Figure S1B shows a simulated four spin systemtwo amide protons ( N H) and two aliphatic protons ( α H). This spin system was used to obtain Figures S1 and S3-S6. Figures S1C-E show how the offset values of amide protons determine the transferred signal between amide protons (C), the total amide signal (D, the signal of the first amide signal + transferred to the second amide), and the undesired transferred signal from amide proton to the closest aliphatic proton (D). All signals were simulated using 2.016 ms mixing. On the basis of the simulations in Figure S1C, the transferred signal reaches at least 50% of the maximal transfer with respect to the signal with zero offset difference if the offset difference (∆ ) does not exceed ~0.64 kHz. However as shown below in Figures S6-S8, this value can be increased by setting the carrier frequency position to about 3 ppm or even -1 ppm without substantially decreasing the efficiency of MODIST pulses. Figure S1D displays the high efficiency of MODIST in retaining the initial signal (at least ~98%), whereas Figure S1E shows the low intensity of the undesired transferred signal (at most ~1%). Figure S1F shows the dependence of the MODIST signal on the distances between amide protons, when the offset difference is lower (0 ppm / 0 Hz, solid lines) and larger (1 ppm / 0.8 kHz, dashed lines) than the optimal threshold of 0.7 kHz. In case of 0 offset difference, it is possible to detect long distances, however, for 0.8 kHz offset difference (dashed line) only detection of distances up to about 5 Å can be expected. For any quantification of MODIST signals, the offset differences must be considered.  Similar to Figure S1C-E, we simulated the behavior of SPR54 ( Figure S2) and π/4-SPR42 ( Figure S3-4). For SPR54 we observe only anti-diagonal transfers as was shown before. 1 Under the same conditions as in Figure S1, SPR54 transfers a higher fraction of the total signal than MODIST, however, the loss of total signal is very high in this case (up to 28%) ( Figure S2B), and the undesired signal (transfer to aliphatic protons) is higher than in the case of MODIST (up to ~16.5% in the absolute value). The carrier frequency was set to 8.2 ppm, and the simulated signals were measured at 0.864 ms mixing. The offset values (in kHz) were simulated using 800 MHz proton Larmor frequency.
Decreasing the flip angle of SPR42 pulses from π/2 to π/4 in the simulations, we observe diagonal and anti-diagonal transfers ( Figure S3A). It is also shown in Ref. [ 2 ] for SPR51 with a flip angle of π/4. For p-SPR42 pulses (p indentifies the flip angle 2 ) with p= π/2 and π/4 the transfer efficiency is worse than that of other SPR pulses. The antigonal transfers achieve ~17% (in absolute value) in the best case with π/2, while for SPR54 it achieves 60% in the absolute value S6 ( Figure S2A). The loss of the total signal is especially high around zero offset (~8.2 ppm, Figure   S3B), however not as so high as for SPR54. With respect to SPR54, π/4-SPR42 shows less undesired transfer to aliphatic protons ( Figures S2C and S3C). One of our solutions for increasing the efficiency of π/4-SPR42 is setting the carrier frequency (CF) away from the amide and aliphatic range, e.g., at 13.5 ppm. Although the supression of the undesired transfers as good as for MODIST ( Figure S4C), the weak antidiagonal ( Figure S4A) is still observed. Under these conditions π/4-SPR42 pulses have 33% smaler Δ than MODIST pulses. Additionally, the tranfer efficiency of π/4-SPR42 pulses depends on the offset values of the spins with respect to position of CF. For example, around 8.2 ppm the transferred signals are ~1.15 fold lower than around 9.5 ppm region ( Figure S4A). This makes π/4-SPR42 less efficient than MODIST for the amide-amide transfer.  Increasing the flip angle of MODIST pulses from π/4 to π/2, ∆ increases from 0.64 kHz to 1.7 kHz ( Figure S5A). However, the maximal transfer efficiency is decrease (~1.4 fold) and the total amide signal retains less ( Figure S5B). The comparison of 2D SH3 (H)N(H)H MODIST spectra with π/4 (blue) and π/2 (red) flip angles ( Figure S5C) confirms the loss of efficiency seen in simulations. With flip angles of π/2 (red) the amide region appears with significantly lower intensities than with π/4 (blue) (it is clearly seen in slice I in Figure S5C). Additionaly, undesired amide-aliphatic transfers are also observed in the first case.  The MODIST simulations in Figure S1 were obtained setting the CF position to 8.2 ppm. Figure     Simulated parameters were the same as in Figure Figure S6G) and the loss of the signal is higher ( Figure S6H) than for the region around 8 ppm ( Figure S6G). It is interesting that experimentally we observed higher retention of the total amide signal with -1 ppm ( Figure S7B, red) of CF setting than with 12.7 ppm ( Figure S7A, red). Also the undesired amide-aliphatic correlations are observed with 12.7 ppm CF setting ( Figure S7A, red).  MODIST transferred signal is shown in red. Simulated parameters were the same as in Figure Figure S10 shows the experimental performance of MODIST (6.48 ms mixing) at 100 kHz MAS for fully protonated SH3 microcrystals. The spectrum in Figure S10A shows qualitatively similar transfer characteristics, and confirms the simulations. At this mixing time, amid-amide transfer is observable, and signal starts to build up in the alpha region of the aliphatic protons.
The inset shows a comparison of the 1D signal at 0 and 6.48 ms mixing. High preservation of the total amide signal of 80% is observed.

Sample Preparation
Fully protonated microcrystalline chicken alpha-spectrin SH3 was prepared according to the protocols in the next references 11,12 and influenza A M2 protein, residues 18-60, was prepared S28 according to the protocols in the references [ 13,14 ]. Deuterated alpha-spectrin SH3 microcrystals were prepared according to the published protocols 5,15 . Each sample was packed into a Bruker

Solid state NMR spectroscopy
The rf-field power of MODIST pulses was optimized using a single pulse calibration, detected in the 1D (H)NH spectrum. The width of the first proton pulse was set to one rotor period of duration (for example, 18 μs for 55.555 kHz MAS). The optimal rf-field power was obtained for zero (H)NH signal (180 o -pulse).             The width of hard pulses are in μs, while the total duration of CP and decoupling are in ms. MODIST and BASS-SD were used for dipolar recoupling. For Figure S10A and B 152 and 96 scans were used, respectively.