Alternate ¹³C-¹²C labeling for complete mainchain resonance assignments using Cα direct-detection with applicability towards fast relaxing protein systems

Abstract

Experiments that use direct ¹³C detection and take advantage of the slower relaxation of ¹³C magnetization compared to 1H offer an attractive strategy for extending the limits of NMR to include larger, highly dynamic, or paramagnetic proteins. Because carbonyl carbons (¹³C′) suffer from serious relaxation enhancement as a consequence of their large chemical shift anisotropy, deuterated alpha carbons are the preferred nuclei for ¹³C detection in large and/or fast relaxing systems. However, direct detection of ¹³Cα is not straightforward owing to the presence of one-bond ¹³C-¹³C couplings with ¹³C′ and ¹³Cβ that split the signals into multiples and hence reduce the sensitivity. Here we present the use of ¹³C enrichment at alternating carbon sites and deuteration at the Cα position to overcome these difficulties. The desired labeling pattern is achieved by expressing the protein in E.coli in D2O with either [2-¹³C]- or [1,3-¹³C]-glycerol as the carbon source. With this labeling strategy, we show that complete assignment of the main chain (including prolyl residues) can be achieved with a single CaN HSQC experiment. This approach offers advantages for the detection of NMR signals from sites with fast nuclear relaxation and offers promise for investigations of larger proteins and/or protein complexes that are inaccessible by proton-detected experiments.

Sequence specific NMR protein assignments typically use uniform ¹³C/¹⁵N labeling with ¹H-detected triple resonance experiments and have been applied successfully to proteins up to ~80 kDa^1–5. However, fast relaxation of ¹H nuclei hinders the application of ¹H-detected experiments for higher molecular weight systems or signals close to paramagnetic centers. To overcome some of these problems ¹³C-detected experiments have been used, which take advantage of slower relaxation of ¹³C coherence due to its lower gyromagnetic ratio. After the pioneering work by Markley and co-workers^6–8 for studies of small diamagnetic proteins, ¹³C-detected experiments were largely abandoned when the more sensitive ¹H detected heteronuclear experiments were introduced. However, increased interest in paramagnetic proteins and the recent arrival of ¹³C detected cryogenic probes have led to a revival of direct detection experiments^9–11.

Carbonyl carbon (C′) direct detection has been the preferred choice as this nucleus is coupled only to alpha carbon (Cα)^9–12. In larger molecular weight systems, particularly at higher magnetic fields, however, carbonyl carbons suffer from fast relaxation due to their large chemical shift anisotropy (CSA). In contrast, Cα nuclei have a small CSA and relax slower if samples are deuterated. Thus, deuterated Cα would be the better nuclei for ¹³C detection experiment in large and/or fast relaxing systems.

However, direct Cα detection is complicated due to the scalar couplings with C′ and Cβ causing crowded spectra and reducing sensitivity due to splitting peaks into multiplets. Spectral complexity can be avoided by computational deconvolution of the spectra¹³ or by selecting a single component within the split peaks using IPAP or S³E schemes¹⁴. While this strategy clearly showed the potential of Cα detection for high molecular weight systems, it requires more complicated pulse programs and/or processing.

Here we present the use of ¹³C-¹²C alternate labeling in Cα-detected triple-resonance experiments to overcomes one bond ¹³C-¹³C coupling by isotopic enrichment at alternating carbon sites. This strategy uses an isotopic labeling scheme similar to the procedure established by LeMaster et al.^{15, 16}. It enables alternate ¹³C-¹²C labeling at most positions by expressing the protein in E.coli using either combinations of [2-¹³C] glycerol and NaH¹³CO₃ or [1,3-¹³C] glycerol and NaH¹²CO₃ as carbon source. Figure 1A (left) shows the Hα-Cα region in a ¹H-¹³C HSQC spectrum of uniformly ¹³C labeled SH3 domain and illustrates the spectral complexity in the indirect dimension due to ¹³Cα-¹³C′ and ¹³Cα-¹³Cβ couplings. The same protein labeled with [2-¹³C] glycerol has well resolved singlets indicating that neighboring carbons are not ¹³C labeled (Fig. 1A right). Only the valines (red arrows) exhibit doublets due to the ¹³Cα-¹³Cβ couplings, as is expected from the metabolic pathways¹⁵. Figure 1B and 1C, report the ¹³C/¹²C labeling ratios at the Cα positions in the [2-¹³C] or [1,3-¹³C] glycerol labeled SH3 domain. [2-¹³C] and [1,3-¹³C] glycerol labeling yielded inverse labeling patterns. As expected from amino acid synthetic pathways, ¹³Cα-labeling was more abundant with [2-¹³C] glycerol labeling compared to [1,3-¹³C] glycerol labeling. Group I residues (red bars in Figure1B) were more than 80 % ¹³Cα labeled with [2-¹³C] glycerol, while Group III residues were primary ¹³C labeled with [1,3-¹³C] glycerol (green bars in Figure1B). Group II residues were partially labeled in both labeling schemes. Complete deuteration at Cα sites is critical for taking advantage of reduced dipole relaxation. This was readily achieved by culturing E.coli in 100% D₂O media with protonated [2-¹³C] or [1,3-¹³C] glycerol (see supplemental information).

Figure 1.

Labelling profile at Cα sites with alternate ¹³C labeling. (A) The Hα-Cα region in a ¹H-¹³C HSQC for uniformly ¹³C labeled (left) and [2-¹³C] glycerol labeled (right) Nck SH3.1. (B) ¹³Cα labeling percentage with [2-¹³C] (upper) or [1,3-¹³C] (lower) glycerol (see supplemental information for detailed procedures). (C) Classification of amino acids based on the ¹³Cα labeling percentage (see text for classification criteria). Cys, His, Met, Phe, and Pro are classified based on earlier literature¹⁵.

A 2D CaN HSQC experiment optimized for the deuterated and alternately-¹³C labeled proteins correlates in a straightforward way the chemical shifts of Cα nuclei with the shifts of the two neighboring nitrogen nuclei (Cαⁱ-Nⁱ and Cαⁱ-Nⁱ⁺¹) (See supplemental information for pulse program). Thus, it allows sequential linking of the backbone nuclei. The pulse sequence was applied to uniformly ²H¹⁵N and alternately ¹³C labeled Nck 1^st SH3 domain (Nck SH3.1). All experiments were recorded in D₂O to minimize the relaxation of ¹⁵N coherence. The measuring time was reasonably short at 1 mM concentration (20–36 hrs). As shown in Figure 2A, the larger dispersion of ¹³C with respect to ¹H enabled unambiguous assignment for most of the sites in the protein. As shown in Figure 2B (red bar), [2-¹³C] glycerol labeling resulted in the observation of 91% of the intra-residue correlations and 86% of the sequential connections. Missing resides (4 Leu and 2 Arg, 1 Gln, and 1 Ser) belong to Group III, which has a lower ¹³C incorporation with [2-¹³C] glycerol, except for Ser53 as discussed below. On the other hand, 72% of intraresidue and 58% of sequential connections were observed with [1,3-¹³C] glycerol labeling (green bars), respectively. The smaller number of correlation observed in [1,3-¹³C] glycerol labeling is due to the fact that ¹³C incorporation from the 1, 3 position is less efficient compared to position 2. In total, however, the information from both glycerol-labeling schemes established all intra and sequential correlations from the structured region of the protein including proline, except for that between Ser53 Cα and Asn54 N (which became observable when using a 2.5mM sample, data not shown). For the protonated protein, only 98% and 87% of intra and sequential correlations are observed, respectively, in experiments for comparable conditions, indicating that full deuteration is a key factor for the Cα detection experiment.

Figure 2.

Sequential main chain assignments in a CaN HSQC experiment using Cα direct detection. (A) CaN HSQC spectrum recorded for 1 mM [ul-²H¹⁵N, [2-¹³C] glycerol] Nck SH3.1. Arrows indicate the sequential walk. Note the proline ¹⁵N signal (residue 52) is folded from its high frequency position. (B) Signal to noise ratio (S/N) for intra-residue (upper) and sequential (lower) cross peaks. Red and green bars are for samples labeled with [2-¹³C] and [1,3-¹³C ] glycerol, respectively.

The low intensity of the Ser53-Asn54 cross peak is likely attributed to chemical exchange. Asn54 is not observable in a regular ¹H-¹⁵N HSQC, and conventional HN detection experiments failed to assign Asn54 N¹⁷. In contrast, the procedure described here successfully assigned both N and Cα resonance of the residue based on the Asn54 Cα-Tyr55 N and Asn54 Cα-N correlations. This indicates the clear advantage of Cα detection over proton detection experiment at an exchange-broadened site. A further crucial advantage is that prolines can be assigned easily. Pro52 in the Nck SH3.1 is assigned based on the Val51 Cα-Pro52 N and Pro51 Cα-Asn52 N correlations.

To demonstrate the general applicability of this approach, we recorded the same experiments at low temperatures in a viscous buffer to model slower tumbling. In 30% glycerol at 288K, the Nck SH3.1 exhibited a diffusion coefficient corresponding to a 150K protein. Even at this condition of slow tumbling, all the correlations assigned in Figure 2 were observed (Supplemental Figure 3), which clearly indicates that this approach is applicable to fast-relaxing or large molecular weight systems.

In summary, ¹³C-¹²C alternate labeling with deuteration at Cα sites is a key strategy to optimize the resolution and sensitivity in the Cα direct-detection experiment without using any spin-state selective schemes. While we analyze the [2-¹³C] and [1,3-¹³C] glycerol labeled samples separately for clarity, one can also combine the two samples to obtain all the information at once. In this case, however, the Cα labeling rate would be half for all residues. Thus, the CaN experiment optimized with the described labeling strategy can establish main chain assignments in a single experiment and without using protons. Similar data could be obtained from a HMQC-type CaN experiment. In addition, the alternate labeling provides complementary ¹³C labeling in the C′ position when Cα is not ¹³C labeled. Thus, it would also enable recording of simple C′N correlated spectra without Cα-C′ coupling, which might be of interest for smaller systems. While the 2D spectrum shown here provided sufficient resolution, one can also apply non-uniform sampling to improve resolution without extending experimental time. In addition, a 3D version of this experiment, which would avoid signal overlap in larger proteins, is under development. In summary, the alternate ¹³C-labeling strategy has clear advantages for higher molecular weight proteins and/or protein complexes that are inaccessible by proton-detected experiments.

Supplementary Material

1_si_001. Supporting Information Available.

Detailed experimental procedures including sample preparation, experimental condition, and pulse program are described. This material is available free of charge via the Internet at http://pubs.acs.org.