High-Resolution Homonuclear 2D NMR of Carbon-13 Enriched Metabolites and their Mixtures

Abstract

High-resolution 2D ¹³C-¹³C NMR correlation spectra of uniformly ¹³C-labeled molecules in solution are obtained by homonuclear ¹³C-decoupling along both dimensions by the application of indirect covariance NMR to constant-time NMR spectra. The spectra are optimally suited for chemical structure elucidation and molecular identification of the components of complex mixtures, such as ones from uniformly ¹³C-labeled cell cultures.

The characterization of chemical mixtures and their components is an essential task in many areas of Chemistry. NMR spectroscopy is a powerful tool enabling quantitative studies of complex mixtures, e.g. in metabolomics, to monitor changes both in terms of chemistry and concentrations without requiring extensive chromatographic fractionation. The vast majority of applications is based on 1D and 2D ¹H NMR taking advantage of the high sensitivity afforded by 100% abundance of protons and their large gyromagnetic ratio. Although the benefits of 2D ¹³C-¹H HSQC correlation spectroscopy of complex mixtures at ¹³C natural abundance have been demonstrated [1, 2], the absence of correlation information between different C-H pairs impedes identification of entire spin systems and thereby limits applications to compounds catalogued in NMR databases, such as the BMRB [3] and HMDB [4]. To establish correlations between all spins in a molecule or a spin system in an efficient manner, the use of ¹³C-enriched metabolite samples hold significant promise. Such samples can be produced, for example, by uniformly ¹³C-labeled cell cultures and organisms. Unfortunately, the presence of large ¹³C-¹³C ¹J-couplings (>30 Hz) generates broad multiplet structures, which leads to substantial spectral crowding and cross-peak overlap when applied to complex mixtures. ¹³C constant-time (CT) spectroscopy [5-9] can help overcome the resolution issue along the indirect dimension, but the problem persists along the detection dimension. Here, we demonstrate how the combination of constant-time ¹³C-¹³C TOCSY with indirect covariance processing [10-14] produces homonuclear-decoupled high-resolution, high-sensitivity ¹³C-¹³C TOCSY spectra suitable for studying complex mixtures. This strategy has recently been demonstrated for the resolution enhancement of pure shift ¹H correlation spectra [15, 16].

NMR Samples

A uniformly ¹³C-labeled algal amino acid mixture, purchased from Sigma-Aldrich, was prepared by dissolving 0.5 mg mixture in 2 ml D₂O. The resulting suspension was centrifuged and the supernatant was used for measurements. Uniformly ¹³C-labeled glucose, purchased from Cambridge Isotope Laboratories, Inc., was prepared as a 1 M solution in D₂O.

NMR Experiments and Processing

All 2D ¹³C-¹³C CT-TOCSY [9] and 2D ¹³C-¹³C TOCSY [17] data sets were collected with 512 N₁ and 1024 N₂ complex data points, with 56 ms DIPSI-2 mixing [18] for TOCSY and 56 ms FLOPSY-16 mixing [19] for CT-TOCSY. All NMR spectra were collected using a cryogenically cooled TCI probe (from Bruker Biospin) at 700 MHz proton frequency with 110 pm ¹³C spectral widths at 298 K temperature. The NMR data were zero-filled to 1024 (N₁) and 2048 (N₂), apodized using shifted sine-bell functions, and Fourier transformed, phase and baseline corrected using NMRPipe [20], and converted to a Matlab-compatible format for subsequent processing and analysis. The experimental time of the CT-TOCSY spectra was 5 hours for glucose and 20 hours for the algal amino acid mixture.

Indirect covariance processing takes the 2D FT NMR spectrum F as input and generates the new spectrum C according to [10, 13]:

$$\mathbf{\text{C}}={(\mathbf{\text{F}}\cdot {\mathbf{\text{F}}}^{\mathbf{\text{T}}})}^{1/2}$$ (1)

In the present context, the method takes advantage of the high resolution of F along the indirect dimension ω₁ endowed by the constant-time scheme by mapping it onto the direct dimension. This results in a symmetric 2D spectrum C that is homonuclear decoupled along both dimensions.

Since CT-experiments are susceptible to the appearance of (minor) intermittent extra-peaks [5-9], we apply an exclusion mask to C [21, 22]. For this purpose, a 2D FT CT-TOCSY spectrum F′ is computed from the same 2D time-domain data with zero-filling along ω₁ to the same total number of datapoints N₂ as along ω₂ . Application of Gaussian line broadening (40 Hz) along the ω₂ dimension and symmetrization by selecting min{|F′_ij|,|F′_ji|} from spectral points F′_ij and F′_ji at positions that are symmetric with respect to the main diagonal yields a medium-resolution spectrum. Line broadening is required, otherwise cross-peaks whose multiplets have no signal at the center frequency will disappear by the symmetrization procedure. Regions in C are then set to zero for which F′ lies below a given threshold. The application of the mask suppresses peaks that are not present in the original 2D FT spectrum while retaining the narrow peak shapes of the indirect covariance spectrum, which is demonstrated below.

Figure 1 illustrates the method for uniformly ¹³C-labeled glucose. The two panels on the far left show a standard ¹³C-¹³C 2D TOCSY spectrum (bottom) with a zoomed region at the top. The presence of homonuclear ¹J_CC, ²J_CC, ³J_CC leads to prominent peak splittings that by far exceed the intrinsic line width. Nuclei that are bonded to two adjacent carbons show characteristic 1:2:1 multiplet patterns that cover a wide spectral range of 2·¹J_CC ≈ 70-90 Hz, which makes these cross-peaks naturally prone to overlaps. The middle two panels depict the corresponding ¹³C-¹³C 2D FT CT-TOCSY spectrum, which shows good decoupling along the ω₁ dimension, while the ω₂ dimension is fully impacted by ¹³C-¹³C J-coupling effects. Application of the covariance method of Eq. (1) to the ¹³C-¹³C 2D CT-TOCSY spectrum in the middle leads after masking to the spectrum on the right, which is effectively homonuclear decoupled along both dimensions. Although glucose does not suffer from cross-peak overlaps, Figure 1 illustrates the method.

Figure 1.

2D ¹³C-¹³C TOCSY (left), ¹³C-¹³C CT-TOCSY (middle), and indirect covariance ¹³C-¹³C CT-TOCSY spectra (right) of ¹³C labeled glucose. The 3 top panels depict expansions of the boxed spectral regions in the lower panels. The double-arrows indicate selected cross-sections that belong to the α and β forms of glucose.

The chance of overlaps increases with the number of cross-peaks present in the spectrum as is the case for the amino-acid mixture shown in Figure 2. This figure is laid out analogously to Figure 1 comparing the standard ¹³C-¹³C 2D TOCSY (left panels) with the corresponding ¹³C-¹³C 2D CT-TOCSY spectrum (middle panels) and the indirect covariance ¹³C-¹³C 2D CT-TOCSY spectrum (right panels). The resolution improvement from left to right is striking as can be seen in the zoomed regions in the three panels at the top. Cross-peak clusters that evade direct analysis in the standard 2D FT spectrum (left) become partially resolved in the CT spectrum (middle) and fully resolved after indirect covariance processing (right). The resolution enhancement over standard ¹³C-¹³C 2D FT amounts to a factor 4, improving the average resolution from 70 Hz to 17 Hz along both dimensions. This trait considerably facilitates analysis of cross-peak connectivities, spin system assignment, component identification, or even chemical structure elucidation. Although a skilled NMR spectroscopist can still extract useful information from the standard ¹³C-¹³C 2D TOCSY spectrum (left panels), the covariance ¹³C-¹³C spectrum (right panels) permits a straightforward interpretation and is directly amenable to automated analysis. Figure 3 illustrates complete ¹³C-spin system identification for isoleucine taking optimal advantage of the resolution gain afforded by the CT-covariance approach.

Figure 2.

2D ¹³C-¹³C TOCSY (far left), ¹³C-¹³C CT-TOCSY (middle), and indirect covariance ¹³C-¹³C CT-TOCSY spectra (far right) of a uniformly ¹³C- labeled amino-acid mixture. The 3 top panels depict expansions of the boxed spectral regions in the lower panels.

Figure 3.

Indirect covariance ¹³C-¹³C CT-TOCSY spectrum of a uniformly ¹³C-labeled amino acid mixture determined according to Eq. (1). The ¹³C-spin connectivity network for isoleucine is indicated by blue lines.

Homonuclear decoupling of 2D NMR experiments along both dimensions has been a long-standing challenge, which has elicited both experimental and computational approaches. Composite pulse decoupling [23, 24] and adiabatic decoupling [25, 26] can alter resonance positions and they work best for relatively small frequency ranges. The maximum entropy method [27] has been successfully applied to Cα resonances of amino acids that have a ¹J-coupling to Cβs, but its performance for ¹³C nuclei with more than one coupling partner has not been demonstrated. Spin-state selective methods, such as S³E [28], IPAP [29], and DIPAP [30, 31] are tailored to a fixed number of coupling partners and their performance is sensitive to the exact magnitudes of the involved couplings. A homonuclear decoupling scheme based on pulsed-field gradient slice selection (‘pure shift spectroscopy’) [32] has recently been combined with covariance NMR to improve the resolution of ¹H-¹H 2D TOCSY, COSY, and NOESY spectra [15, 16]. Although this approach can in principle achieve better decoupling than CT-based methods, the resolution gain is offset by a sensitivity loss caused by slice selection. The CT-covariance spectrum, on the other hand, retains the inherent sensitivity of the 2D FT parent spectrum.

Our results demonstrate that dramatic cross-peak sharpening can be achieved along both ¹³C dimensions by the CT-covariance method producing high-resolution homonuclear-decoupled ¹³C-correlation spectra with qualitatively improved spectral behavior. While this property significantly assists the analysis complex organic molecules and mixtures of moderate complexity, as is the case for the amino-acid mixture of Fig. 2, it will be even more beneficial for studying extracts and lysates from ¹³C-labeled organisms, such as bacteria, yeast, and plants, for metabolic profiling, flux analysis, and de novo structure determination of metabolites [33].

Highlights.

• 2D TOCSY spectra of ¹³C-labeled metabolites could be decoupled along both dimensions

• Indirect covariance processing of constant-time 2D NMR spectrum provides resolution enhancement

• High-resolution ¹³C-¹³C TOCSY spectra of complex mixtures display reduced cross-peak overlaps