Sequential acquisition of multi-dimensional heteronuclear chemical shift correlation spectra with 1H detection

RF pulse schemes for the simultaneous acquisition of heteronuclear multi-dimensional chemical shift correlation spectra, such as {HA(CA)NH & HA(CACO)NH}, {HA(CA)NH & H(N)CAHA} and {H(N)CAHA & H(CC)NH}, that are commonly employed in the study of moderately-sized protein molecules, have been implemented using dual sequential 1H acquisitions in the direct dimension. Such an approach is not only beneficial in terms of the reduction of experimental time as compared to data collection via two separate experiments but also facilitates the unambiguous sequential linking of the backbone amino acid residues. The potential of sequential 1H data acquisition procedure in the study of RNA is also demonstrated here.

)NH}, that are commonly employed in the study of moderately-sized protein molecules, have been implemented using dual sequential 1 H acquisitions in the direct dimension. Such an approach is not only beneficial in terms of the reduction of experimental time as compared to data collection via two separate experiments but also facilitates the unambiguous sequential linking of the backbone amino acid residues. The potential of sequential 1 H data acquisition procedure in the study of RNA is also demonstrated here. N MR spectroscopy is a powerful technique for the study of biomolecular structure and dynamics, both in solution as well as in the solid state. In protein NMR a variety of multi-dimensional heteronuclear chemical shift correlation experiments are typically used for resonance assignment and for extraction of 1 H-1 H distance restraints, e.g. HACANH 1,2 , HNCAHA 3-5 , HACACONH 2 , HNCOCAHA 4,5 , HCCNH 1,6 and 15 N-edited 1 H-1 H NOESY, respectively. These multi-dimensional spectra are based on different magnetisation transfer pathways and are customarily collected individually. As a result, the time for the acquisition of all required data sets can in many cases become exceedingly long. In this context, a variety of techniques are currently being explored for reducing data acquisition times [7][8][9][10] . One of the approaches that has received considerable attention in the study of proteins, both in solution [11][12][13][14] and in the solid state [15][16][17] , involves the simultaneous collection of different chemical shift correlation spectra. For example, using dual receivers with 1 H and 13 C acquisition in the direct dimension, the simultaneous collection of chemical shift correlation spectra, e.g. 12 and sequential 13 data acquisition procedures. However, as noted already in the literature 11,12 , all other aspects being equal, e.g. magnetisation transfer and relaxation characteristics, the signal intensities seen in the 1 H and 13 C detected data sets would differ because of the difference in the gyromagnetic ratios of the two nuclei. Furthermore, the utility of such experiments is limited as the solution state NMR probes are typically optimised either for 1 H or for 13 C detection only. In addition, many of the chemical shift correlation experiments of interest do not require dual receivers and instead may involve 1 H or 13 C acquisition only. In this context, RF pulse schemes enabling the collection of multi-dimensional data sets with a single receiver is of great interest [18][19][20][21] . Recently we have reported RF pulse schemes involving dual sequential 1 H acquisition with only amide proton detection and making use of dual 15 N-13 C mixing steps for achieving protein resonance assignment 22 . While such sequences can be easily adapted to the study of large 2 H-labeled protein samples, here we present RF pulse schemes that were developed in the context of moderately sized proteins. In such systems the relaxation losses during 15 N-13 C mixing periods are not expected to be significant even in a fully protonated sample. The pulse sequences presented here make use of the availability of both the HA and HN protons and employ only a single 15 N-13 C mixing step to achieve sequential resonance assignments in protonated protein samples. The efficacy of the approach is experimentally demonstrated by the 'one-shot' collection of representative protein NMR spectra. We also show that this approach is useful for the study of RNA.

{HACANH & HACACO} and {HACACO & HACACONH}, has been demonstrated employing both parallel
Results and discussion {HA(CA)NH & HA(CACO)NH}. The triple resonance HACANH experiment is frequently used for sequential resonance assignment of the backbone 13 C a , 15 N, 1 H N and 1 H a nuclei and involves throughbond magnetisation transfer between the directly coupled nuclei via the pathway 1 H a -. 13 C a -. 15 N-. 1 H N . Interresidue cross peaks arising from transfer of magnetisation from the 13 C a spin of residue (i) to the 15 N spin of residue of (i 1 1), resulting from 2 J CaN couplings, are observed in this experiment. Interresidue cross peaks can usually be distinguished from intraresidue cross peaks based on their respective signal intensity. However, to achieve unambiguous resonance assignment, the HACACONH experiment involving magnetisation transfers via the pathway 1 H a -. 13 C a -. 13 CO-. 15 N-. 1 H N and leading only to interresidue cross peaks is generally carried out in addition. In the HACANH experiment, however, the 13 C a -. 15 N magnetisation transfer is the critical step as it relies on weak intraresidue 1 J CaN couplings (,11 Hz). Although heteronuclear magnetisation transfers are typically carried out via INEPT type transfers, in-phase magnetisation transfers via heteronuclear cross polarization schemes have also been successfully used to enhance sensitivity in triple resonance NMR experiments such as in HACANH 1 . Here, we have implemented RF pulse schemes for the sequential collection of different correlation spectra using 15 N,-. 13 C cross polarization schemes. The RF pulse scheme given in Fig. 1a permits the 'one-shot' acquisition of 3D HA(CA)NH and 3D HA(CACO)NH data sets. The initial transverse 1 H magnetisation generated by the first 90u pulse is allowed to evolve under its chemical shift during the t 1 (HA)/t 1 9(HA) period and under the one bond heteronuclear 13 C-1 H coupling for a period of 2D 0 to generate antiphase 1 H magnetisation. The anti-phase 1 H magnetisation is then converted into antiphase carbon magnetisation by the 90u pulses applied to the two nuclei. The antiphase 13 C a polarisation is allowed to refocus during the interval 2D 1 to generate ( 13 C a ) x magnetisation and then subjected to a period of 13 C a -. 15 N magnetisation exchange via the application of a band-selective het-TOCSY mixing sequence. The residual 13 C transverse magnetisation remaining after the 13 C a -. 15 N transfer step is flipped to the z axis and the 15 N magnetisation generated after 13 C a -. 15 N mixing is allowed to evolve under its chemical shift during the t 2 (N) period and transferred to the attached proton via an INEPT step and is observed in the t 3 period under 15 N decoupling to generate the 3D HA(CA)NH spectrum. The WATERGATE sequence 29 is used for water suppression. After the completion the first 1 H acquisition, the residual 13 C a magnetisation is brought to the transverse plane and subjected to 13 C a -. 13 CO cross polarisation. This transverse 13 CO magnetisation is then subjected to a period of 13 CO-. 15 N magnetisation exchange via the application of a band-selective het-TOCSY mixing sequence. The resulting transverse 15 N magnetisation is allowed to evolve during the t 2 9 (N) period and then transferred to the attached proton via the INEPT procedure. The 1 H signals are acquired in t 3 9, under 15 N decoupling to generate the 3D-HA(CACO)NH data. The cross peak intensities observed in Gradients with a sine bell amplitude profile were used. Durations and strength with respect to the maximum strength of 50 G/cm are: G 1 5 1 ms (60%), G 2 5 1 ms (80%).   the HA(CACO)NH spectrum is dependent on the amount of residual 13 C a magnetisation present after the 13 C a -. 15 N mixing period and hence related to the duration of the mixing period and the performance characteristics of the mixing sequence employed.
The residual 13 C a magnetisation has to be kept along the z axis until the first data acquisition is completed and this may affect the signal intensities observed in the HA(CACO)NH spectrum due to relaxation losses. However, in the systems studied here, significant variation in signal intensities were not observed when the residence time of the 13 C a magnetisation along the z axis was varied over a range of 0-100 ms (Fig. S1). The optimal length of the 15 N-13 C het-TOCSY mixing period was found to be ,50 ms ( Here, unlike the case in the RF pulse schemes given in Fig. 1a, the initial transverse magnetisation generated from both 15 N and 13 C attached protons by the first 90u pulse is allowed to undergo chemical shift evolution during the t 1 (HN)/t 1 9(HA) period. These evolve under the one bond heteronuclear 15   C a -13 CO anisotropic cross polarisation, respectively. 13 C a -13 CO mixing was carried out keeping the 13 C RF carrier at 115 ppm, with a peak RF power level of ,11 kHz and for a total duration of 17.92 ms by repeating the basic sequence twice (8.96 ms * 2). 15 N-13 C mixing was carried out by keeping the 13 C RF carrier either at 55 ppm or at 175 ppm for achieving band-selective 15  with the 13 C a magnetisation along the z axis. In both data sets, intraand inter-residue peaks arising, respectively, due to 1 J CaN and 2 J CaN couplings are observed. With a single 15 N-13 C mixing step in the RF pulse sequence (Fig. 1b), the simultaneous collection of H(N)CAHA and HA(CA)NH spectra delivers the chemical shifts of the backbone 13 C a;i , 15 N i , 1 H N;i and 1 H a;i nuclei. Additionally, the ( 15 N, 1 H) backbone chemical shifts of the adjacent i11 residue and the ( 13 CA, 1 HA) chemical shifts of the preceding i21 residue are also obtained. This facilitates the unambiguous linking of three amino acid residues, i.e. i-1, i and i11. With this approach we have successfully acquired a combined data set comprising the HA(CA)NH and H(N)CAHA experiment (Fig. 3). Data collected with a cryoprobe are provided in the supplementary material (Fig. S5, S6) to illustrate the performance of the sequence at lower protein concentrations.

{H(N)CAHA & H(CC)NH}.
In addition to resonance assignment of backbone nuclei a modification of the RF pulse scheme given in Fig. 1b allows to simultaneously acquire the 3D H(CC)NH and H(N)CAHA correlation spectra and to obtain chemical shift information on the protein side chain as well as on the backbone nuclei. In a simple HACANH experiment the 13 C a magnetisation used for 13 C a -. 15 N mixing arises only from the magnetisation transfer from directly attached 1 H a protons. However, the 13 C a magnetisation in the HCCNH experiment is also generated starting from the side chain protons via the 1 H sc -. 13 C sc -. 13 C a magnetisation transfer pathway. This is achieved by introducing a 13 C-13 C longitudinal TOCSY mixing period just before the heteronuclear cross polarisation step (Fig. 1c), with the remainder of the pulse sequence essentially the same as in Fig. 1b. Obviously, one can design the RF pulse scheme to obtain either 1 H or 13 C side chain chemical shift information. The spectral widths in the indirect dimension in the two data sets can also be independently adjusted by appropriate scaling of the t 2 (CA)/t 2 9 (N) increments and spectral folding in one data set does not lead to resonance overlaps in the other, as the two data sets are effectively independent. As in the case of HACANH experiment, interresidue side chain cross peaks arising from transfer of magnetisation from 13 C a spin of residue (i) to the 15 N spin of residue of (i 1 1) are observed in the HCCNH spectrum. The HCCNH and HNCAHA spectra (Fig. 4) were acquired in one shot via the pulse scheme given in Fig. 1c.
The results presented here clearly demonstrate that it is possible to achieve simultaneous acquisition of multidimensional data sets in solution using the sequential data acquisition procedure, akin to recently reported solid state NMR studies of proteins [15][16][17] . Additionally, the unambiguous sequential linking of backbone nuclei (i 2 1, i, i 1 1) is achieved in one shot. The basic strategy with sequential data acquisition procedure is that two different experiments leading to correlation spectra arising from different magnetisation transfer pathways are simultaneously started and at a defined intermediate stage the relevant magnetisation belonging to one of the pathways is kept along the z axis. Depending on the type of data to be sequentially collected, it can be either 15 N or the 13 C nuclei that are to be kept as longitudinal polarisation. After this, magnetisation transfers followed by the first data acquisition are carried out to complete the experiment via the first pathway. Subsequently, appropriate mag- N- 13 CA mixing was carried out by keeping the 13 C RF carrier at 55 ppm, employing 15 N/ 13 C peak RF power level of ,3.6 kHz and for a duration of 50 ms by repeating the basic sequence twice (25 ms * 2). The 13 C-. 1 H het-TOCSY was carried out with one cycle of the AK2-JCH aniso1 sequence having a duration of 7.2 ms, employing 1 H/ 13 CA peak RF power level of ,12.5 kHz. The 1 H RF carrier was kept at 4.7 ppm. The 1 H RF carrier was kept at 3 ppm during t 1 and subsequently switched back to the water position (4.7 ppm). D 0,1,2 5 2.58, 1.79, 0.79 ms, 2t 1 5 2D 0 and 2t 2 5 3 ms.   netisation transfers allow for the second data acquisition to complete the correlation experiment via the second pathway. Central to this approach is, that the magnetisation which is stored along the z axis for usage in the second experiment should not be disturbed during the completion of the first experiment. The order in which both multi-dimensional data sets are sequentially acquired has to be chosen appropriately in this context. For medium sized molecules, which are the focus of the present study, and with short acquisition times on the order of ,50 ms in the direct dimension, sequential acquisition of correlation spectra do not suffer from significant relaxation losses irrespective of whether 15 N or 13 C nuclei are kept along the z axis. Although the acquisitions of only a few representative spectra are demonstrated here, the approach outlined can be extended to acquire simultaneously other types of correlation spectra. For example, utilizing the HNCAHA experiment for the sequential assignment of the backbone 13 C a , 15 N, 1 H N and 1 H a nuclei allow to exploit the residual 15 N magnetisation after the 15 N-. 13 C a transfer for generating simultaneously a 15 N edited 1 H-1 H NOESY spectrum (Fig. S7, S8, S9). As the application of het-TOCSY mixing schemes over long periods of time may lead to sample heating and hence might pose problems in the study of temperature sensitive samples, one may take recourse to the INEPT procedure to effect 15 N-13 C magnetisation transfers. Although the relative merits in the context of sequential data acquisitions are yet to be fully assessed, good quality spectra are obtained by implementing INEPT type transfers 31 for 15 N,-. 13 C mixing (Fig. S10).
The sequential data acquisition procedure can also be effectively used for simultaneously generating heteronuclear correlation spectra of RNA 32,33 . For example, triple resonance NMR experiments such as HCNCH/HCNH are often used for achieving intra-nucleotide correlation of the sugar protons with the base protons. The HCNCH experiment involves through-bond magnetisation transfers between the directly coupled nuclei via the pathway 1 H 1 9-. 13 C 1 9-. 15 N 1,9 -. 13  In both experiments, that are typically carried out in D 2 O, the residual 13 C 1 9 magnetisation after the 13 C 1 9-. 15 N 1,9 transfer step can efficiently be exploited to obtain simultaneously COSY/TOCSY data for sugar protons (Fig. S11, S12). For RNA samples dissolved in H 2 O, the through-bond HNCCH experiment is often used for correlating the H3 imino protons of uridines with the H5/H6 base protons via the pathway 1 H im -. 15 N im -. 13 C 4 -. 13 C 5,6 -. 1 H 5,6 . Employing the residual magnetisation after the 15 N im -. 13 C 4 transfer step, one can simultaneously obtain the NOE correlation spectrum of the imino protons in RNA (Fig. S11, S13). The sequential data acquisition strategy presented here may also be combined with other approaches for further reducing the data acquisition time, e.g. sparse sampling in the indirect dimension 34 . Furthermore, such dual sequential acquisition procedure may also be applied to collect 3D data with direct 13 C detection.

Methods
Uniformly ( 13 C, 15 N)-labelled samples of the 82 amino acid MCM C-terminal winged helix domain of Sulfolobus solfataricus and the 94 amino acid N-terminal region of human hnRNP C proteins were expressed and purified as reported earlier 23,24 . For development, a room temperature triple resonance probe was used and the final protein concentrations were 7 mM and 3 mM, respectively. In addition studies were also undertaken with a cryoprobe utilizing protein samples at lower concentrations (0.8 mM and 1.2 mM, supplementary material). The uniformly ( 13 C, 15 N)-labelled RNA (59-GGCGUUCGCUUAGAACGUC-39), referred to as BEVSLD5, was prepared as described 25 and a final concentration of 0.9 mM was used here. Multidimensional chemical shift correlation experiments for proteins were carried out with a Bruker 600 MHz narrow-bore Avance III NMR spectrometer equipped with pulse field gradient accessories, pulse shaping units and a triple resonance probe. Sample temperature was set to 303 K. For the BEVSLD5 RNA a triple resonance cryoprobe was used; sample temperature was set to 293 K for experiments on non-exchangeable and 288 K for experiments on exchangeable protons, respectively. Homo-and heteronuclear magnetisation transfers were achieved using amplitude and phasemodulated mixing sequences (AK2-JCH anisol , AK2-JCaC' aniso : 26 ; AK2-JCC 27 ). Where required, RF field strength and duration of the mixing period were scaled appropri-ately. The States procedure 28 was applied for phase-sensitive detection in the indirect dimensions. Standard phase cycling procedures were employed to select signals arising from desired magnetisation transfer pathways.