Assignment of aromatic side-chain spins and characterization of their distance restraints at fast MAS

Graphical abstract

Highlights

• •Strategy for the assignment of aromatic side-chain spins is reported.
• •Low power TOBSY $\left(\text{C}{9}_{48}^1\right)$ is used for ¹³C^aro → ¹³C^α polarization transfer to link aromatic spins to the backbone atoms.
• •¹H–¹H distance restraints for aromatic spins were generated using RFDR and selective BASS-SD recoupling sequences.
• •The obtained restraints provided important contacts between atoms in the α-helix and the β-sheet secondary structure elements.

Abstract

The assignment of aromatic side-chain spins has always been more challenging than assigning backbone and aliphatic spins. Selective labeling combined with mutagenesis has been the approach for assigning aromatic spins. This manuscript reports a method for assigning aromatic spins in a fully protonated protein by connecting them to the backbone atoms using a low-power TOBSY sequence. The pulse sequence employs residual polarization and sequential acquisitions techniques to record H^N- and H^C-detected spectra in a single experiment. The unambiguous assignment of aromatic spins also enables the characterization of ¹H–¹H distance restraints involving aromatic spins. Broadband (RFDR) and selective (BASS-SD) recoupling sequences were used to generate H^N-Η^C, H^C-H^N and H^C-H^C restraints involving the side-chain proton spins of aromatic residues. This approach has been demonstrated on a fully protonated U-[¹³C,¹⁵N] labeled GB1 sample at 95–100 kHz MAS.

Introduction

With recent improvements in probe hardware, solid-state nuclear magnetic resonance (ssNMR) at a magic angle spinning (MAS) frequency of ∼100 kHz is routinely employed in several labs. Under these conditions, ¹H-detection allows protein characterization using sub-milligram amounts of protein (Agarwal et al., 2014). Structures of microcrystalline proteins (Agarwal et al., 2014; Andreas et al., 2016), membrane proteins (Lakomek et al., 2017, Schubeis et al., 2020), amyloids (Daskalov et al., 2021) and large protein assemblies (Lecoq et al., 2019, Lecoq et al., 2020) have been solved using ¹H-detection. Protein dynamics, protein–ligand and protein interactions with other biomolecules have also been studied using these methods (Ahlawat et al., 2022a, Ghassemi et al., 2022, Le Marchand et al., 2022).

Nevertheless, spectral mapping of peaks to atoms in a molecule remains the main bottleneck in these studies. The protocols for backbone assignments at fast MAS have been reported and reviewed extensively in the literature (Andreas et al., 2015, Fraga et al., 2017, Higman, 2018, Penzel et al., 2015, Sharma et al., 2020, Stanek et al., 2020, Xiang et al., 2015). Recording multiple assignment spectra are time-intensive and the experimental data suffers from instrumental instabilities (sample temperature, drifts in the magnetic field, fluctuations in magic angle settings) and possibly sample degradation. The duration of data acquisition can be significantly reduced by employing multiple acquisition approaches to acquire multiple spectra in a single experiment. The multiple acquisition experiments use various elements such as simultaneous CP (sim-CP) (Herbst et al., 2008), multiplexing (Ivchenko et al., 2003), bidirectional CP (Gopinath and Veglia, 2012a), residual polarization (Banigan and Traaseth, 2012, Gopinath and Veglia, 2012a), multiple receivers (Herbst et al., 2008, Takeda et al., 2012) and sequential and simultaneous acquisition (Bjerring et al., 2012, Gopinath and Veglia, 2012b) to record signals from multiple coherence pathways in a single experiment. However, backbone assignments are insufficient for most biomolecular studies and side-chain information is often required. Side-chain assignments have been performed by linking side-chain ¹³C and ¹H^C atoms to the backbone atoms (Agarwal and Reif, 2008, Kulminskaya et al., 2015). Linking multiple ¹³C side-chain atoms to backbone ¹³C^α/¹⁵N atoms requires multistep polarization transfer. The polarization transfer during first-order dipolar recoupling sequences like DREAM (dipolar recoupling enhanced by adiabatic modulation) (Verel et al., 2001) and RFDR (radio frequency driven recoupling) (Bennett et al., 1992) suffers from dipolar truncation and are ill-suited for linking multiple carbon atoms in long side-chains of Leu, Ile, Lys etc. A second-order dipolar recoupling sequence, AL FRESCO (adiabatic linearly frequency swept recoupling), shows multistep polarization transfer in the fast MAS regime but employs mixing times on the order of a second (Wi and Frydman, 2020). The side-chain ¹³C atoms have also been assigned in HET-s fibrils by recording a ¹H-detected 3D ¹³C DQ-SQ (double quantum-single quantum) spectrum (Tolchard et al., 2018). The t₁ dimension encodes the DQ evolution of the two directly bonded ¹³C spins, which is used to assign ¹H^C spins. The larger dispersion in the t₁ dimension results in better separation of the peaks but comes at the expense of larger spectral width in the indirect dimension.

In solution-state NMR, multistep homonuclear polarization transfer along the amino acid side-chain is efficiently achieved via scalar (J) couplings (Cavanagh et al., 2010). Scalar coupling transfer is isotropic, impervious to motions and particularly in solids, independent of crystallite orientation and MAS frequency. Scalar couplings are approximately two-orders of magnitude weaker than dipolar couplings and require effective proton decoupling for efficient transfer. At fast MAS, J-couplings-based transfer can be achieved by solution-state sequences such as WALTZ-16 (Kulminskaya et al., 2016, Stanek et al., 2016) or equivalent solid-state TOBSY (total through bond correlation spectroscopy) sequences (Baldus and Meier, 1996, Hardy et al., 2003, Hardy et al., 2001). Symmetry sequences enable efficient scalar coupling transfer despite the two-order smaller magnitude of the interaction (Levitt, 2007). Recently, ${\text{C}9}_{39}^1$ and ${\text{C}9}_{48}^1$ symmetry sequences were shown to provide high polarization (∼70–80 %) transfer efficiencies at fast MAS while employing low rf fields (Tan et al., 2018). These symmetry numbers also efficiently decouple heteronuclear dipolar and scalar coupling interactions, thereby eliminating the need for ¹H decoupling during mixing, especially in the fast MAS regime. At fast MAS, these TOBSY sequences show better polarization transfer efficiency between aliphatic ¹³C side-chain spins than other solution-state sequences such as WALTZ-16 (Paluch et al., 2022). The pulse sequences utilizing J-based transfers have been used to acquire 3D, 4D or 5D spectra for assigning side-chain spins (Klein et al., 2022a, Klein et al., 2022b, Schubeis et al., 2021, Stanek et al., 2016). Alternatively, INADEQUATE (incredible natural abundance double quantum transfer experiment) (Lesage et al., 1997, Lesage et al., 1999), UC2QF-COSY (uniform-sign cross-peak double-quantum-filtered correlation spectroscopy) (Mueller et al., 2002, Olsen et al., 2003) and CTUC-COSY (constant-time uniform cross-peak correlation spectroscopy) (Chen et al., 2006) have also been used to link one/two bond separated side-chain atoms but require long transverse coherences.

Although these developments have resulted in the routine assignment of aliphatic side-chains, the assignment of aromatic side-chains remains an issue. Selective labeling/unlabeling and mutagenesis are still the most commonly used methods for aromatic side-chain assignments (Ghosh and Rienstra, 2017, Kainosho et al., 2006, Shcherbakov and Hong, 2018, Torizawa et al., 2005). Side-chain ¹³C^aro spins have higher deshielding resulting in distinct chemical shifts (∼110–140 ppm) compared to aliphatic spins (10–70 ppm). Large chemical shift anisotropy (CSA) and complex dynamics in aromatic side-chains can hinder efficient polarization transfer via dipolar coupling-based sequences and significantly limit the possibility of assignment. This fact is evident in the BMRB database, with an exceedingly low number of assignments deposited for the aromatic side-chains. There are few reports of aromatic side-chain assignments in small proteins by employing WALTZ-16, DQ-SQ and AL-FRESCO sequences (Andreas et al., 2016, Tolchard et al., 2018, Wi and Frydman, 2020). Generally, only the aromatic spins with distinct chemical shifts are assigned with these sequences. In a few cases, the complete side-chain assignment was possible because of the single occurrence of a particular aromatic residue in the protein sequence. An efficient ¹³C^aro → ¹³C^α polarization transfer sequence is required to assign the aromatic spins. The ${\text{CN}}_n^v$ symmetry-based TOBSY sequences are well suited for this purpose and are explored for the aromatic side-chain spins assignment in this manuscript. Obtaining spatial contacts for the aromatic spins has also remained challenging but can be extremely useful in structural applications. In addition to proposing a strategy to obtain assignments of aromatic residues, we also propose methods for observing a large density of spatial contacts involving aromatic proton spins.

In this study, we show the assignment of the aromatic side-chain atoms using ${\text{C}9}_{48}^1$-based TOBSY mixing on U-[¹³C,¹⁵N] labeled GB1 at ∼95.23 kHz MAS. The aromatic side-chain ¹H/¹³C spins are correlated to the backbone spin pairs, such as C^αH^α and NH and side-chain spins, such as C^βH^β. These 2D C^αH^α, NH and C^βH^β spin-pairs show the largest dispersion and are routinely used as high-resolution 2D-fingerprint spectra in multidimensional solid-state NMR spectroscopy. Multispin simulations of aromatic residues were used to assess the performance of WALTZ-16, ${\text{C}9}_{39}^1$ and ${\text{C}9}_{48}^1$ symmetry sequences for ¹³C^aro → ¹³C^α/¹³C^β transfer. The simultaneously acquired (H)C^aroNH^N-TOBSY and (H)C^aroCH^C-TOBSY spectra link the aromatic side-chain atoms to the backbone ¹⁵N, ¹H^N, ¹³C^α and ¹H^α atoms. The ¹H^aro spins can also be independently assigned from the (H)C^aroCH^C-TOBSY spectrum. The ¹H–¹H restraints for aromatic residues were obtained using the broadband recoupling sequence RFDR and selective recoupling sequence BASS-SD (Band-selective spectral spin diffusion) (Jain et al., 2017). These two mixing sequences were used in a pulse sequence that combines four experiments to generate H^N ↔ H^N, H^N ↔ H^aro and H^aro ↔ H^aro distance restraints. In the case of GB1, the aromatic restraints provide important contacts in the hydrophobic core and assist in arranging the α-helix and β-sheets in the 3D structure.

Result and discussion

The strategy of linking aromatic side-chain spins to backbone atoms is analogous to that employed in the aliphatic side-chain assignment (Agarwal and Reif, 2008, Kulminskaya et al., 2015). Typically, 3D/4D HCNH^N and HCCH^C spectra have been used to assign the aliphatic side-chains. The same 3D spectra can be used to assign aromatic spins with an efficient multistep homonuclear polarization transfer sequence such that aromatic spins are linked to ¹³C^α/¹³C^β spins. Acquiring a single experiment for both aliphatic and aromatic side-chain spins is possible but suffers from practical challenges. The large differences in the isotropic chemical shifts, CSAs and scalar couplings of aliphatic and aromatic side-chain carbon atoms result in marginally different optimum conditions for ¹H–¹³C CP and multi-bond ¹³C–¹³C scalar coupling transfers. In addition, the need for acquiring two indirect dimensions with large spectral widths can be avoided in separate experiments. Anyways, the significant difference in the signal-to-noise observed for aromatic and aliphatic spins requires different extents of signal averaging. A pulse sequence linking only aromatic side chains to backbone atoms can also be employed for spectral editing and identifying backbone peaks of aromatic residues in 2D HC and HN heteronuclear correlation spectra. An independent 3D experiment optimized only for aromatic residue correlations is discussed in the next section.

Pulse sequence for aromatic side-chain assignment

The pulse sequence for aromatic side-chain assignment is largely similar to the reported 3D (H)C(C^α)NH^N sequence used to assign side-chain spins in the perdeuterated and fully protonated samples (Kulminskaya et al., 2016). However, two key differences from the previous report are the selection of only aromatic signals and the employment of ${\text{C}9}_{48}^1$ symmetry sequence for scalar transfer. Linking side-chains to the peaks in the NH^N plane is preferred over the C^αH^α plane because of the larger dispersion of ¹H-¹⁵N peaks. As previously reported, the ¹³C^α-¹⁵N SPECIFIC-CP block (employed for heteronuclear polarization transfer in the 3D (H)CNH^N pulse sequence) transfers only ∼30 % of the ¹³C^α polarization to the ¹⁵N spins (Loening et al., 2012, Nielsen et al., 2013). This implies that after ¹³C^α-¹⁵N SPECIFIC-CP, ∼70 % of the initial polarization on ¹³C^α and ∼90–100 % of polarization on the other ¹³C atoms is intact and can be utilized to record a 3D (H)CCH^C-TOBSY spectrum.

Fig. 1A shows the pulse sequence to simultaneously record 3D (H)C^aroCH^C-TOBSY and (H)C^aroNH^N-TOBSY. A 180° selective pulse after the ¹H–¹³C CP (shown in green in Fig. 1A) combined with two-step phase cycling retains all the aromatic ¹³C magnetization, while the non-aromatic ¹³C magnetization is cancelled out after adding two scans. The ¹³C^aro resonances are encoded in the t₁ dimension. The TOBSY block enables homonuclear ¹³C polarization transfer amongst the directly bonded ¹³C atoms. In our case, after the ¹³C^α-¹⁵Ν SPECIFIC-CP block, ∼73 % of the initial ¹³C^α magnetization remains on ¹³C^α, implying that at best ∼27 % of the initial magnetization is transferred to ¹⁵N spins. The other aliphatic carbons retain ∼90–95 % of the initial magnetization. In the second evolution period (t₂), the polarization on ¹⁵N and the residual polarization on ¹³C atoms evolve simultaneously (t_2N and t_2C). The π/2-pulses applied on ¹³C and ¹⁵N channels store the magnetization along the z-direction. The τ_d2 and τ_d2′ are the delays for maintaining a constant duty cycle after t_2C and t_2N, respectively. The water suppression block minimizes the solvent signals and a subsequent π/2-pulse followed by ¹³C–¹H back CP enables the detection of ¹H^C magnetization (red pathway in Fig. 1A). For this entire duration, ¹⁵N magnetization is stored along the z-direction. H^N polarization is detected after CP from ¹⁵N following a π/2-pulse after applying the second water suppression block (blue pathway in Fig. 1A). H^C polarization is detected before H^N because of the shorter carbon T_1. The τ_d2 and τ_d2′ delays, in addition to maintaining a constant duty cycle, also allow arbitrary spectral widths and independent evolution blocks for the ¹⁵N and ¹³C spins. The pulse sequence stores the H^C and H^N signals in an interleaved fashion (See supplementary information for the pulse program code implemented on Bruker Spectrometers). The first FID encodes signals from the (H)C^aroCH^C-TOBSY pathway (depicted by a red line in Fig. 1A), while the second FID encodes signals from the (H)C^aroNH^N-TOBSY pathway (depicted by the blue line in Fig. 1A). The FIDs from the (H)C^aroCH^C-TOBSY pathway are stored as odd FIDs, while the FIDs from the (H)C^aroNH^N-TOBSY pathway are stored as even FIDs. Therefore, the 3D dataset has a factor of two larger t₁ points than the actual size of the individual spectrum.

Fig. 1.

(A) The pulse sequence for the aromatic side-chain assignment. The polarization transfer pathways for (H)C^aroCH^C-TOBSY and (H)C^aroNH^N-TOBSY spectra are shown in red and blue, respectively. The pulse sequence stores H^C- and H^N-detected spectra separately so multiplexing is not required. The individual spectrum is obtained by separating the even and odd FIDs along the t₁ dimension. The SLAP (Solvent suppression using adiabatic pulses) sequence is used for water suppression (Matsunaga et al., 2021). The rCW^ApA sequence was used for ¹H decoupling while WALTZ-16 heteronuclear decoupling was used during ¹H detection. The phase cycling for the pulse sequence is: Φ₁ = x, −x; Φ₂ = x, x, y, y; Φ₃ = y; Φ₄ = x, x, −x, −x; Φ₅ = −y, −y, −y, −y, y, y, y, y; Φ_rec1 = x, −x, −x, x, −x, x, x, −x and Φ_rec2 = −x, x, −x, x, x, −x, x, −x. (B, C) The aromatic residues filtered 2D NH^N (blue) and C^αH^α (red) heteronuclear correlation spectra overlaid with the unfiltered heteronuclear correlation spectra. The filtered spectra were recorded with the pulse sequence (A) without t_1C evolution as pseudo-3D. (B) The overlay of the 2D (H)(C^aro)NH^N-TOBSY spectrum (blue) with the 2D NH^N-HSQC spectrum (black). The 2D (H)(C^aro)NH^N-TOBSY spectrum has five peaks belonging to 4 residues (3Y, 43W, 45Y, 52F and 43W^ε1) are labeled in red. (C) The overlay of the 2D (H)(C^aro)CH^C-TOBSY spectrum (red) with the 2D C^αH^α-HSQC spectrum (black). The 2D (H)(C^aro)CH^C-TOBSY spectrum has four peaks belonging to (3Y, 43W, 45Y and 52F), which are labeled in blue. The peaks of 30F and 33Y are not visible in both spectra. The complete overlay of (H)(C^aro)CH^C-TOBSY spectrum (red) with CH^C-HSQC (black) is shown in Fig. S2. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

TOBSY block was independently optimized for ¹³C^aro → ¹³C^α transfer using a 1D (H)C^aroCH^C-TOBSY spectrum. The optimum TOBSY (${\text{C}9}_{48}^1$) mixing time was 20 ms with ¹³C carrier frequency at 90 ppm (See Table S3). The offset, rf amplitude and mixing time of the TOBSY block were optimized to maximize the intensity in the H^α region (Fig. S1). The pulse sequence (Fig. 1A) is also an aromatic residue filtering experiment. The 2D spectra recorded without t₁ evolution can also be used to identify the backbone peaks of aromatic residues in the 2D ¹⁵N-¹H^N and ¹³C^α-¹Η^α heteronuclear correlation spectra (Fig. 1B, C and Fig. S2).

Resonance assignment from the (H)C^aroNH^N-TOBSY and (H)C^aroCH^C-TOBSY spectra

The H^N-detected spectrum, (H)C^aroNH^N-TOBSY, generally has better resolution because of the NH plane, while the H^C-detected spectrum, (H)C^aroCH^C-TOBSY, has higher sensitivity. In the (H)C^aroNH^N-TOBSY spectrum (Fig. 2; blue spectrum), ¹³C^aro side-chain atoms in the F₁ dimension are linked to peaks in the 2D ¹H-¹⁵N spectra. In contrast, in the (H)C^aroCH^C-TOBSY spectrum (Fig. 2; red spectrum), the ¹³C^aro side-chain atoms in the F₁ dimension are linked to peaks in the 2D ¹H–¹³C CP-based one-bond heteronuclear correlation plane. The backbone ¹⁵N, ¹H^N, ¹³C^α and ¹H^α chemical shifts must be assigned before assigning side-chain aromatic atoms. These backbone chemical shifts were assigned using the multi-acquisition approach and the backbone walk is shown in Fig. S3 of the supplementary information (Sharma et al., 2020).

Fig. 2.

The F₁-F₃ strips of the spectra obtained from the pulse sequence in Fig. 1A for the aromatic side-chain assignment. The 3D (H)C^aroNH^N-TOBSY spectrum is colored blue and the 3D (H)C^aroCH^C-TOBSY spectrum is colored red. The F₁-F₃ strips for residues (A) 43W, (B) 3Y and 45Y are shown. The ¹³C^aro side-chain atoms are assigned from the F₁ dimension of the 3D (H)C^aroNH^N-TOBSY spectrum (blue) using the NH^N plane. The side-chain peaks are connected with the vertical dotted line in the 3D (H)C^aroNH^N-TOBSY spectrum. The assigned ¹³C^aro atoms are used to assign the ¹H^aro chemical shifts in the 3D (H)C^aroCH^C-TOBSY spectrum (red). The F₁-F₃ plane at the ¹³C^aro chemical shifts of side-chain atoms provides the assignment of ¹H^C atoms. The horizontal dotted line connects the ¹³C^aro side-chain atoms between the strips of (H)C^aroNH^N-TOBSY (blue) and (H)C^aroCH^C-TOBSY (red) spectra. The vertical dotted lines assign ¹H^C atoms in the red strips. The green cross (x) indicates the peaks with lower intensity. These peaks are visible at lower contour levels and are shown in Fig. S5B. The experiment was recorded on U-[¹³C,¹⁵N]-GB1 on a 700 MHz spectrometer at 95.238 kHz MAS. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The ¹³C^aro and ¹H^aro side-chain peaks were assigned from the 3D (H)C^aroNH^N-TOBSY and 3D (H)C^aroCH^C-TOBSY spectra, respectively. In the 3D (H)C^aroNH^N-TOBSY spectrum, the ¹³C^aro peaks were assigned in the F₁ dimension (dotted horizontal lines in Fig. 2A and B). The 3D (H)C^aroCH^C-TOBSY spectrum verifies the assigned peak using the C^αH^α plane. The ¹H^aro chemical shifts were assigned from the (H)C^aroCH^C-TOBSY spectrum. Using the chemical shifts of an assigned ¹³C^aro atom, the ¹H^aro (dotted vertical lines in Fig. 2A and B) spins were identified and assigned. Fig. 2A shows the strips for aromatic side-chain assignments of 43W. Using the ¹⁵N and ¹H^N chemical shifts, the C^δ1 and C^ε3/C^ζ3 peaks are assigned in the (H)C^aroNH^N-TOBSY spectrum (the blue strip in Fig. 2A). In the (H)C^aroCH^C-TOBSY spectrum, the C^αH^α peak in the HSQC plane (F₂-F₃) is linked to C^δ1 and C^ε3/C^ζ3 in F₁. Peaks for the entire aromatic side-chain of 43W are correlated to the N^ε1-H^ε1 in the (H)C^aroNH^N-TOBSY spectrum (the blue strip in Fig. 2A). From the (H)C^aroCH^C-TOBSY spectrum, the ¹³C^aro chemical shifts in the F₂ dimension were used to assign ¹H^aro spins. This approach assigned all the ¹³C^aro and ¹H^aro peaks for the 43W (Fig. 2A). The aromatic spins for 3Y and 45Y were also similarly assigned (Fig. 2B). The C^γ, C^δ1/δ2 and C^ε2/ε3 along with H^δ1/δ2 and H^ε1/ε2 spins, were assigned for 3Y and 45Y residues. The peaks corresponding to C^ζ and H^η spins were not observed in both spectra. The H^η atom is not directly bonded to C^ζ so its peak will not be visible. For 52F, two peaks around 131 and 140 ppm are visible (Fig. S4). The peak at 140 ppm belongs to C^γ while the peak at 131 ppm could belong to either C^δ1/δ2(∼131.5 ppm), C^ε1/ε2 (130.7 ppm) and C^ζ (∼129.2 ppm) or all of the atoms. The parentheses report the average chemical shift of these atoms from the BMRB database with a standard deviation of 1.2–1.4 ppm. The poor chemical shift dispersion of the C^δ1/δ2, C^ε1/ε2 and C^ζ side-chain atoms can be remedied by selective labeling (Gauto et al., 2022, Gauto et al., 2019, Teilum et al., 2006). Except for the –OH group, the F and Y residues have a similar side-chain architecture, so we expect polarization transfer profiles along the side-chain atoms to be similar.

In principle from the 3D (H)C^aroCH^C-TOBSY, the C^β-H^β plane can also be used to assign aromatic spins. The aromatic residue assignments are listed in Table S1 in the supporting information. The strips for the ¹³C^β chemical shifts are shown in Fig. S5A. Numerical simulations (shown below) indicate that in the best-case scenario, ∼10 % polarization is transferred from aromatic carbons to C^α and C^β spins. We do not observe the C^αH^α and NH^N peaks for 30F and 33Y in either the (H)C^aroCH^C-TOBSY or the (H)C^aroNH^N-TOBSY spectra, i.e., when the initial magnetization is on the aromatic spins. However, we observe the C^αH^α and NH^N peaks for the 30F and 33Y residues in the regular 2D heteronuclear correlation spectra (black spectrum in Fig. 1B and C). Both residues are present in the α-helix region and face the β3-strand. Previously, it has been shown that the residues in the α-helix facing the β-strands have large to moderate amplitude motions (Shi and Rienstra, 2016). Motional broadening or poor ¹H–¹³C dipolar transfer in the aromatic side-chain could explain the absence of these residues in the spectra.

Fig. 5.

The F₁-F₃ strips from the (A) (H)N(H^N)(H^C)CH^C (blue), (B) (H)C(H^C)(H^N)NH^N (red) and (C) (H)C(H^C)(H^C)CH^C (green) spectra, showing ¹H–¹H restraints for aromatic residues. The restraints spectra were acquired with the 4X-restraint pulse sequence using 6 ms of BASS-SD mixing. The experimentally observed restraints between α-helix and β-sheets were plotted on the structure of GB1 (PDB: 2QMT) for residues (D) 43W, (E) 3Y and 52F. The arrow indicates H^N ↔ H^aro (blue), H^aro → H^N (red) and H^aro → H^aro (green) contacts. The H^N atoms are colored cyan and the H^aro atoms are colored purple. The H^aro atoms are labeled without considering the ring flips. The arrowhead indicates the direction of the transfer. The double-headed arrow indicates that two symmetrical cross-peaks are observed between two spins, and the arrowhead's color shows the transfer's direction. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The pulse sequence in Fig. 1A can also be used to acquire spectra for the assignment of aliphatic side-chain spins. The spectra are acquired without a selective pulse and TOBSY is optimized for transfer within the aliphatic region. The strips for the 3D (H)C^aliNH^N-TOBSY (blue) and 3D (H)C^aliC^aliH^ali-TOBSY (red) spectra are shown in Fig. S6.

TOBSY transfer efficiency in numerical simulations

Numerical simulations were used to assess the transfer efficiency from ¹³C^aro → ¹³C^α spins with various scalar coupling transfer sequences such as ${\text{C}9}_{39}^1$, ${\text{C}9}_{48}^1$ and WALTZ-16. SIMPSON simulations (Bak et al., 2000, Tošner et al., 2014) were performed for histidine, phenylalanine and tyrosine spin systems and details of the spin system are reported in Figs. S7 and S8 in supplementary information. At fast MAS, the detrimental effects of protons during the TOBSY irradiation are minimized (Tan et al., 2018, Paluch et al., 2022). Hence, protons were omitted from the spin systems to reduce the computational time. For the histidine spin system, the initial polarization was selected to be on the C^γ and C^δ2 aromatic spins. In contrast, for the phenylalanine and tyrosine residues, the initial polarization was limited to C^γ, C^δ1 and C^ε1 spins. The initial polarization was limited to spins that correlated to ¹³C^α spins in the 3D experimental spectra. Simulations were performed at 95.238 kHz MAS frequency and a static field of 16.44 T. All other simulation parameters are reported in the supplementary information. Points in the simulated magnetization buildup curves were acquired after 39, 48 and 96 rotor periods for ${\text{C}9}_{39}^1$, ${\text{C}9}_{48}^1$ and WALTZ-16 sequences, respectively.

Fig. 3 shows symmetry sequences have better transfer efficiencies than WALTZ-16 with ${\text{C}9}_{48}^1$ TOBSY being the best. The optimum duration for ¹³C^aro → ¹³C^α scalar coupling transfer is ∼20 ms and is similar to experimentally optimized duration and was also used in recording 3D spectra (Table S3 in supplementary information). The transfer efficiencies from aromatic carbons to ¹³C^α and ¹³C^β atoms with 20 ms of ${\text{C}9}_{48}^1$ TOBSY mixing is ∼9–10 % for phenylalanine and tyrosine residue, while for histidine, the same is higher due to a shorter spin-system. The detailed transfer efficiencies are summarized in Table S2 in supplementary information.

Fig. 3.

Simulated polarization transfers build up from side-chain aromatic atoms of histidine (A and D), phenylalanine (B and E) and tyrosine (C and F) residues to ¹³C^α atom (A-C) and ¹³C^β atoms (D-F). Simulations are performed at 95.238 kHz MAS at a static field of 16.4 T (700 MHz) using ${\text{C}9}_{48}^1$ (black), ${\text{C}9}_{39}^1$ (red) and WALTZ-16 (teal) sequences, respectively. The y-axis is normalized with the cumulative initial magnetization on the aromatic atoms. The dotted vertical blue line indicates the transfer intensity at 20 ms of mixing time. The polarization transfer efficiencies from individual atoms to other atoms for the three aromatic residues are shown in supplementary information (Fig. S6). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

¹H–¹H restraints involving aromatic protons

High-resolution NMR structure calculation requires a large set of distance restraints (∼15 per residue) (Russell et al., 2019). The general approach is to record multiple second-order dipolar recoupled spectra. The observed peaks are evidence of spatial proximity within the limit of upper distance restraints determined by the details of the recoupling sequence and mixing time (Manolikas et al., 2008). Both broadband (RFDR) and selective recoupling sequences such as BASS-SD (band-selective spectral spin-diffusion), SERP (selective recoupling of protons) and MODIST (modest offset difference internuclear selective transfer) have been used to observe ¹H–¹H spatial correlations interpreted as structural restraints in the limit of upper distances (Andreas et al., 2016, Bennett et al., 1992, Jain et al., 2017 Duong et al., 2018; Potnuru et al., 2020; Nimerovsky et al., 2022). The lack of aromatic assignment has limited the engagement of aromatic protons as structural restraints (Klein et al., 2022b). After the complete ¹H and ¹³C assignment of aromatic side-chain spins in the previous section, the restraints involving aromatic spins can be observed and assigned.

Ideally, four experiments, (H)N(H^N)(H^N)NH^N, (H)N(H^N)(H^C)CH^C, (H)C(H^C)(H^N)NH^N and (H)C(H^C)(H^C)CH^C are required to generate a comprehensive of H^N-H^N, H^N-H^C, H^C-H^N and H^C-H^C distance restraints. We previously demonstrated that these four experiments could be combined into a single experiment that generates four spectra for the duration required to record one H^C-H^C restraint spectrum (Ahlawat et al., 2022b). We refer to this pulse sequence as a 4X-restraint experiment, as the four spectra encode four different restraints. Here, we use the 4X-restraint pulse sequence to map out ¹H–¹H restraints involving aromatic residues: H^N-H^N, H^N-H^aro, H^aro-H^N and H^aro-H^aro. We recorded two independent 4X-restraint spectra, one with a broadband (RFDR) recoupling and the second with a selective recoupling (BASS-SD) to generate ¹H–¹H aromatic restraints. This also allowed us to evaluate the performance of broadband and selective recoupling sequence in mapping out restraints involving aromatic spins.

H^N-H^aro, H^aro-H^N and H^aro-H^C restraints observed with RFDR mixing

RFDR is a broadband recoupling that recouples the entire proton bath, including the aromatic residues, so a selective pulse on aromatic spins is not required. Without the selective pulse, the ¹³C spectral width is around 140 ppm (including the aliphatic and aromatic side-chain). The spectra were recorded by folding the ¹³C^aro peaks. The distinct aromatic proton chemical shifts assist in unfolding the aromatic peaks. The restraints obtained for aliphatic protons are shown in Fig. S9 and are not explicitly discussed here.

In the (H)N(H^N)(H^C)CH^C (blue) and (H)C(H^C)(H^N)NH^N (red) spectra (Fig. 4A and B), the contacts are observed only for the 43W(H^ε1) and 43W(H^δ1). A long-range 3.2 Å contact between 54V(H^γ2)-43 W(H^ε1) is visible in the (H)C(H^C)(H^N)NH^N (red) spectrum (Fig. 4B). The (H)C(H^C)(H^C)CH^C (green) spectrum (Fig. 4C) has both intra- and inter-residue contacts. The inter-residue contacts are long-range in the upper distance limit of 4–4.5 Å. The observed long-range contacts are 43W(H^ζ2)-31K(H^α), 43W(H^ε3)-52F(H^α), 43W(H^ε3)-52F(H^β), 52F(H^aro)-26A(H^β), 52F(H^aro)-24A(H^α), 52F(H^aro)-23A(H^α), 52F(H^aro)-27E(H^α) and 52F(H^aro)-27E(H^β). These contacts are plotted on the structure of GB1 (Fig. 4D). Most of the observed contacts are between aromatic side-chain residues attached to the β-sheets and the aliphatic side-chain protons in the α-helix that together form the hydrophobic core in the case of GB1. Identifying several spatially proximal aromatic and aliphatic side-chain protons should help restrain the hydrophobic core of GB1.

Fig. 4.

The F₁-F₃ strips from the (A) (H)N(H^N)(H^C)CH^C (blue), (B) (H)C(H^C)(H^N)NH^N (red) and (C) (H)C(H^C)(H^C)CH^C (green) spectra highlight the H^N-H^aro, H^aro-H^N and H^aro-H^aro restraints for aromatic residues. The restraints spectra were acquired with the 4X-restraint pulse sequence using 920 μs of RFDR mixing. The peaks in the aromatic region were folded in the spectrum, subsequently unfolded during processing for assigning cross-peak, and marked with an asterisk (*). (D) The experimentally observed restraints between α-helix and β-sheets were plotted on the structure of GB1 (PDB: 2QMT). Only the inter-residue contacts are plotted in the figure. The H^N atoms are colored in cyan, H^α in brown, H^ali in orange and H^aro in purple. The H^ali atoms are shown for only those residues for which aromatic contacts are observed. The arrowhead indicates the direction of the transfer and the double-headed arrows are used to show contacts with two-cross peaks. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

H^N-H^aro, H^aro-H^N and H^aro-H^aro restraints observed with BASS-SD

In contrast to broadband ¹H–¹H recoupling sequences, selective recoupling sequences have been shown to provide cross-peaks with better intensities and long-range contacts in the case of proteins (Duong et al., 2020, Jain et al., 2017, Zhang et al., 2020). BASS-SD recoupling is a low-power sequence tailored to recouple a subset of proton spins in fully protonated molecules. The amide and aromatic protons have a similar chemical shift range; therefore, BASS-SD recoupling is ideally suited to generate H^N-H^aro, H^aro-H^N and H^aro-H^aro restraints.

The 4X-restraint experiment was recorded with BASS-SD mixing. In the ¹³C dimension, the ¹³C^aro spins were selected by employing the selective pulse (Q3; Gaussian Cascade) (Emsley and Bodenhausen, 1992). Experimentally, BASS-SD mixing was optimized to maximize H^aro → H^N transfer in the 2D (H)C^aro(H^aro)-H spectrum. The optimum BASS-SD condition for H^N ↔ H^aro transfer was a 6 ms tangential spin lock pulse with variable rf amplitude of 3.77–5.66 kHz at the carrier frequency of 8.15 ppm. A detailed list of experimental parameters is provided in the supplementary infomration (Table S3). The four spectra obtained by employing BASS-SD recoupling in the 4X-restraint pulse sequence are (H)N(H^N)(H^N)NH^N, (H)N(H^N)(H^aro)C^aroH^aro, (H)C^aro(H^aro)(H^N)NH^N and (H)C^aro(H^aro)(H^aro)C^aroH^aro. The (H)N(H^N)(H^N)NH^N spectrum gives only H^N-H^N restraints and has been discussed previously in literature for both BASS-SD and its variant (Jain et al., 2017, Ahlawat et al., 2022b). We do not further discuss H^N-H^N restraints in this manuscript because no aromatic side-chain spins are involved. The other three relevant spectra (H)N(H^N)(H^aro)C^aroH^aro (blue), (H)C^aro(H^aro)(H^N)NH^N (red) and (H)C^aro(H^aro)(H^aro)C^aroH^aro (green) are shown in Fig. 5.

Fig. 5A shows H^N → H^aro contacts (blue strips). The H^N’s of several residues (31K, 34A and 35N) in the α-helix show contact with aromatic protons (H^ζ2, H^ζ3 and H^η2) of 43W, located in the β3 strand. The 43W(H^δ1) also shows contacts with neighboring residues (42E and 41G). Most of these contacts observed between 43W and other residues are 4–5 Å away. Similarly, the H^N’s of residues in the α-helix show contact with residues 3Y and 52F present in the β1 and β4 strands, respectively. 26A and 27E H^N’s show contact with 52F (H^ε1/ε2 and Η^ζ). The high-resolution X-ray diffraction structure shows the distance between 52F aromatic protons and 27E and 26A H^N’s range from 2.9 Å to 5.2 Å. 52F also shows contact with 4K and 5L in the β1 strand. The residue 3Y shows contact with residues 26A and 20A present in the α-helix and turn region, respectively. In Fig. 5B, the red (H^aro → H^N) strips show most contacts observed in the blue (H^N → H^aro) spectrum but in the opposite direction. These contacts are plotted on the GB1 structure (PDB: 2QMT) and are color-coded as blue in Fig. 5D and E. The arrowhead color denotes the color of the spectrum in which the cross-peak is present. Double-headed arrows with red and blue colors depict the contacts between H^aro → H^N and H^N → H^aro. The H^aro-H^aro spectrum (green) in Fig. 5C mostly has intra-residue cross-peaks. Two inter-residue contacts between 52F and 3Y(H^ε1/ε2) and 43W(Η^ζ2)-45Y(H^δ1/δ2) are visible in the spectrum. The observed 43W(Η^ζ2)-45Y(H^δ1/δ2) contact (green dotted lines in Fig. 5) corresponds to a distance of about 10.6 Å in the crystal structure (PDB: 2QMT).

The two restraints spectra (RFDR and BASS-SD mixing) generate an exhaustive set of H^N-H^aro, H^aro–H^N and H^aro-H^C restraints. The two spectra demonstrate that RFDR is suited to map out H^C–H^aro restraints, where H^C represents aliphatic side-chain protons. In contrast, BASS-SD is suited to observe symmetrical long-range H^N-H^aro and H^aro-H^N restraints. The RFDR spectra should be acquired by folding ¹³C^aro peaks to avoid two dimensions with large ¹³C spectral widths. In contrast, the BASS-SD recoupled experiment is designed only to evolve ¹³C^aro resonances and requires lesser spectral width, thereby reducing the experimental time.

Conclusion

To conclude, the approach proposed in this study functions as an aromatic residue filtering sequence and can be used to identify backbone atoms of aromatic residues in the 2D NH^N and 2D C^αΗ^α heteronuclear correlation spectra. The optimized ¹³C^aro → ¹³C^α transfer via the TOBSY block links aromatic spins to the backbone atoms for assigning side-chain aromatic spins. The proposed pulse sequence allows the simultaneous acquisition of (H)C^aroNH^N-TOBSY and (H)C^aroCH^C-TOBSY spectra in a single experiment. In combination with backbone assignment experiments, the (H)C^aroNH^N-TOBSY spectrum assists in assigning ¹³C^aro spins, while the (H)C^aroCH^C-TOBSY spectrum assists in assigning ¹H^aro spins in fully protonated proteins. RFDR and BASS-SD recoupling, in combination with the 4X-restraint pulse sequence, generate an extensive set of distance restraints involving aromatic side-chain protons. RFDR is particularly suited for establishing H^aro ↔ H^C contacts, while BASS-SD is better for identifying H^N ↔ H^aro contacts. In the case of GB1, the RFDR and BASS-SD restraints provide important contacts between residues in the hydrophobic core of the protein. The assignment of aromatic side-chain spins in this study allows characterizing of aromatic amino acids in fully protonated samples by ssNMR. In larger proteins with more aromatic residues, additional 3D H(C^aro)NH^N-TOBSY and H(C^aro)CH^C-TOBSY spectra or 4D HC^aroNH^N-TOBSY and HC^aroCH^C-TOBSY spectra might be required to assign all aromatic residues unambiguously.

Method and materials

The experiments were recorded on a 700 MHz Avance III spectrometer using the 0.7 mm HCN probe. All the experiments were recorded on U-[¹³C,¹⁵N]-GB1 microcrystals at MAS frequency of 95–100 kHz at a probe temperature of 250 K. The chemical shift values and experimental details of rf fields, mixing times and offsets are mentioned in the Supplementary information (Tables S1–S6). rCW^ApA decoupling was used for ¹H decoupling (∼19 kHz) and WALTZ-16 decoupling was used to decouple ¹³C (∼10 kHz) and ¹⁵N (∼4 kHz) during the ¹H acquisition. The solvent suppression was performed using either the MISSISSIPPI or the SLAP sequences. The spectra were processed in Topspin and analyzed in CCPN software (Stevens et al., 2011).

The pulse program for side-chain assignment acquires two spectra (H)C^aroCH^C-TOBSY and (H)C^aroNH^N-TOBSY, in a single experiment. The pulse program separately stores H^C- and H^N-detected FIDs as odd and even points along the t₁ dimension. The individual (H)C^aroCH^C-TOBSY and (H)C^aroNH^N-TOBSY spectra are obtained by splitting the odd and even FIDs using the “split” command in the topspin 3.5.7. The 4X-restraint pulse sequence and script for splitting and combining FIDs resulting in spectra encoding four different types of restraints have been previously reported (Ahlawat et al., 2022b).

CRediT authorship contribution statement

Sahil Ahlawat: Conceptualization, Data curation, Investigation, Writing – original draft, Writing – review & editing. Subbarao Mohana Venkata Mopidevi: Software, Investigation, Data curation, Writing – review & editing. Pravin P. Taware: Investigation, Resources. Sreejith Raran-Kurussi: Resources. Kaustubh R. Mote: Methodology, Data curation, Writing – review & editing. Vipin Agarwal: Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Edited by Bauke W. Dijkstra

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.