Multidimensional High-Resolution Magic Angle Spinning and Solution-State NMR Characterization of 13C-labeled Plant Metabolites and Lignocellulose

Lignocellulose, which includes mainly cellulose, hemicellulose, and lignin, is a potential resource for the production of chemicals and for other applications. For effective production of materials derived from biomass, it is important to characterize the metabolites and polymeric components of the biomass. Nuclear magnetic resonance (NMR) spectroscopy has been used to identify biomass components; however, the NMR spectra of metabolites and lignocellulose components are ambiguously assigned in many cases due to overlapping chemical shift peaks. Using our 13C-labeling technique in higher plants such as poplar samples, we demonstrated that overlapping peaks could be resolved by three-dimensional NMR experiments to more accurately assign chemical shifts compared with two-dimensional NMR measurements. Metabolites of the 13C-poplar were measured by high-resolution magic angle spinning NMR spectroscopy, which allows sample analysis without solvent extraction, while lignocellulose components of the 13C-poplar dissolved in dimethylsulfoxide/pyridine solvent were analyzed by solution-state NMR techniques. Using these methods, we were able to unambiguously assign chemical shifts of small and macromolecular components in 13C-poplar samples. Furthermore, using samples of less than 5 mg, we could differentiate between two kinds of genes that were overexpressed in poplar samples, which produced clearly modified plant cell wall components.


Results
Analysis of 13 C-poplar metabolites without extraction. We prepared a poplar sapling with 13 C-plant biomass for multidimensional NMR measurements. First, an intact poplar was measured by HR-MAS. Figure 2 shows the aliphatic region in the HR-MAS 2D 1 H-13 C-HSQC spectrum of the 13 C-poplar sample. Although these 1 H-13 C signals were slightly broadened in the 1 H dimension by residual 1 H-1 H dipolar interactions, chemical shift dispersion could be adequately resolved in the intact tissues. In the HR-MAS 1 H-13 C-HSQC spectrum of the sample, various amino acids, ethanol, malate, choline, ethanolamine, and glucose were assigned by matches using the 1 H and 13 C chemical shift database SpinAssign 50,51 . The SpinAssign database, which is a database of standards dissolved in potassium phosphate buffer, is useful to analyze the HR-MAS HSQC spectrum of intact 13 C-poplar because of similarities in chemical shifts when comparing the potassium phosphate buffer with an intracellular environment. During the process of matching chemical shift data for each signal, we frequently observed multiple candidate metabolites in the database, indicating that the candidates may include false-positive matches. For example, three signals, (3.757 of 1 H, 57.160 ppm of 13  SpinAssign database (unpublished data). However, by integrating the matched signals, the individual signals were initially matched to 10, 3, and 2 candidate metabolites, respectively, in the results of the SpinAssign search. The candidate metabolites did include the chemical shifts of glutamine α , β , and γ , i.e., true-positive matches. The occurrence of false-positives matches was discussed in a previous study 33 . 13 C-lipid, which is unregistered in SpinAssign, was measured by HR-MAS, and then matched with chemical shifts for 13 C-poplar. To assign the 2D 1 H-13 C-HSQC signals on the basis of the information from 1 H-13 C-13 C-1 H correlations, the 3D HCCH-COSY spectrum (Fig. 3) of the sample was recorded. Figure 3 shows 4 1 H-1 H planes slicing the 3D 1 H-1 H-13 C spectrum of the intact tissue sample at different points along the 13 C axis. Various amino acids, ethanol, malate, choline, ethanolamine, and glucose were assigned on the basis of 1 H-13 C-13 C-1 H correlations by 3D HCCH-COSY to avoid false-positive matches.
We analyzed the samples to detect polysaccharides using HR-MAS and solid-state NMR techniques. In the HR-MAS analysis, the HSQC spectrum of a cell-wall-rich sample prepared from 13 C-poplar was compared with those three commercial pectin standards. The NMR signals of the sample matched to 18 out of 20 apple pectin signals, 27 out of 31 rhamnogalacturonan-I signals, and 27 out of 35 arabinogalactan signals ( Figure S1). In the solid-state NMR analysis, 2D 13 C-13 C refocused Incredible Natural Abundance DoublE QUAntum Transfer Experiment (INADEQUATE) NMR spectra 52 of 13 C-poplar were obtained using two different τ delay times to detect optimal signal intensities of polysaccharides [ Figure  S2(b)] and aliphatic compounds [ Figure S2(c)]. The NMR signals were characterized using chemical shift data for cellulose Iβ 40 , amorphous cellulose 53 , lipids 54,55 , and amino acids 50,56 . The MAS-J-HMQC 57 signals of 13 C-poplar were also used to characterize J-based 13 C-1 H cross peaks of the signals assigned using the characterization [ Figure S2  Application to comparison of mutants [overexpression of VASCULAR-RELATED NAC-DOMAIN 6 (VND6) and VND7]. Based on the assignment of the signals of both metabolites and polymeric components, we applied these NMR studies to mutant characterization. We prepared two transgenic poplars by overexpression of VND6 and VND7 genes 58 . VND6 and VND7 are reported to be master regulator genes involved in the formation of vessels during the development of the metaxylem and protoxylem, respectively, because overexpression of these genes can induce trans-differentiation of various cells into metaxylem-and protoxylem-like vessel elements, respectively, in both Arabidopsis and poplar 58 .  VND genes are anticipated to play an important role in the study of biomass production 59 . HR-MAS analyses of these two mutant samples revealed a dramatic difference in their metabolic profiles, based on their 2D NMR spectra ( Figure S3). Furthermore, polymeric components of the two transgenic poplars were compared. Figure 6 shows spectra of the anomeric and aliphatic regions in the HSQC experiments of VND6  Discussion 3D NMR measurements combined with 13 C isotope labeling techniques are an effective approach for identifying polymeric components and metabolites. Using 3D NMR measurements such as HCCH-COSY and HCCH-TOCSY improves peak resolution compared with 2D NMR measurements, and thus, complete assignment of polymeric components and metabolites is anticipated. However, the low natural abundance of 13 C produces low signal intensity that inhibits detection of 13 C-13 C correlations in 3D NMR measurements. By enhancing the signals by 13 C isotope labeling, we were able to completely assign polymeric components and metabolites (e.g., glucopyranose) within polysaccharides.
A number of peaks from the HR-MAS 2D 1 H-13 C-HSQC spectrum of 13 C-poplar were matched to particular metabolites using the SpinAssign database, including glutamine and arginine. However, we also encountered a problem with this approach, which may list matched metabolites with many false-positive candidates. Two solutions to the false-positive assignment have been proposed 33 : (1) to combine multiple 2D NMR experiments that are used to identify compounds in complex mixtures 60,61 and (2) to introduce heteronuclear 3D NMR spectra for reducing the number of ambiguous assignments 34 .
Thus, by introducing 3D NMR measurements for reducing false-positive assignments, we were able to assign signals of metabolites such as arginine and asparagine in the 2D 1 H-13 C-HSQC spectrum on the basis of 13 C-13 C correlations observed in the 3D HCCH-COSY spectrum. However, 1 H-13 C-13 C-1 H correlations for saccharides could not be determined in the 3D HCCH-COSY except for glucose, because a substantial number of peaks associated with monosaccharides and oligosaccharides were detected in the sugar region (3.0-4.7 in the 1 H dimension and 50-90 ppm in the 13 C dimension) of the 2D spectrum (Fig. 2). Assignment of these peaks will likely require the development of new, higher resolution NMR methods.
Although the HR-MAS technique is capable of measuring intact samples, many metabolites were not detected using this approach. However, we were able to identify these metabolites by comparing the results of HR-MAS with those obtained by solution-state NMR techniques. By comparing the HR-MAS spectrum from our intact sample (Fig. 2) with the solution-state NMR spectrum from the sample extracted by potassium phosphate buffer ( Figure S4), we observed the localization of metabolite signals that were undetected by HR-MAS, but detected by solution-state NMR. For example, asparagine, citrate, and unassigned metabolites were detected in the solution-state NMR spectrum ( Figure S4), but undetected in the HR-MAS spectrum (Fig. 2), indicating that their metabolites were localized in a particular organelle. In the HR-MAS technique using an intact sample, this phenomenon may be attributed to the limited motion reflected by localization in a particular organelle such as the mitochondrial membrane 17,62-65 . Conversely, since solution-state NMR requires the extraction of components, their composition is varied and depends on the solvents used for extraction. However, this limitation is not present with HR-MAS because it works with intact samples. This allows us to observe lipid chains, ethanol, leucine, isoleucine, and other unassigned aliphatic side chains, as shown in Fig. 2. Because solution-state NMR avoids the localization problem, characterization is significantly improved by combining solution-state NMR and HR-MAS techniques.
In the solution-state 2D 1 H-13 C-HSQC NMR spectrum of 13 C-poplar that was ball-milled and then dissolved in DMSO/pyridine solvent, we could detect a substantial number of peaks corresponding to polymeric components (e.g., polysaccharides). The signals detected in the polysaccharide anomeric region of the 1 H-13 C-HSQC spectrum were assigned on the basis of the results of a previous study 28 , in which the anomeric carbons of polysaccharides were identified. However, we were only able to identify anomeric carbons of polysaccharides by this approach; therefore, to identify other anomeric carbon signals, we combined 3D NMR experiments with 2D experiments, similar to the HR-MAS technique. By combining 2D and 3D measurements, we assigned additional polysaccharide peaks, except for the anomeric regions of the 2D 1 H-13 C-HSQC spectrum, on the basis of 13 C-13 C correlations by 3D NMR. Using this method, we completely assigned several polysaccharides in our sample. For example, using the 3D HCCH-TOCSY, 1 H-13 C-13 C-1 H correlations from C1 to C6 of (1, 4)-β -d-Glcp were confirmed (Fig. 5). However, some polysaccharides were only partially assigned due to the low intensities of the 3D NMR signals of these polymeric components. These results indicate that detection of small amounts of polymeric components will require different methods to improve signal sensitivity. The assignment of acetylated hemicelluloses cause the migration of acetyl groups during the preparation of the NMR sample 66 . We detected pectin-like polysaccharide components in the HR-MAS NMR spectrum of the cell-wall-rich sample of 13 C-poplar ( Figure S1). Therefore, this method is useful to detect gel-like, faster molecular motion polysaccharides such as pectin. However, cellulose and hemicellulose were not detected using this approach because of the slower molecular motion of these polysaccharides. Thus, we analyzed the macromolecules of 13 C-poplar using solid-state NMR ( Figure S2). Although we could detect polysaccharides, lipids, and protein-derived materials using two different τ delay times in the refocused INADEQUATE method, it was difficult to make detailed assignments for these compounds. Therefore, the combined use of higher resolution solution-state NMR with solid-state NMR might be a complementary approach to characterize cell-wall components.
We applied the assignments of polysaccharides to our analysis of the transgenic VND6 and VND7 strains of 13 C-poplar using samples of less than 5 mg, in which protoxylem and metaxylem vessel formation were introduced, respectively 58 . Metaxylem and protoxylem are formed respectively at the early and late stages of primary xylem formation. In the overexpression measurements of VND6 and VND7, polymeric components were anticipated to be observed at both late (i.e., protoxylem) and early (i.e., metaxylem) stages. Our poplar samples were grown for short durations, i.e., early stage, and thus, polymeric components of VND7 were longer than those in the wild-type, and in contrast, VND6 components were shorter than those in the wild-type (Fig. 6). The results of the HR-MAS analyses of VND6 and VND7 of 13 C-poplar supported the polymeric components data ( Figure S3). Sugars based on polysaccharides of VND7 were shorter than those in the wild-type, because many sugars would have been used to synthesize polysaccharides. In contrast, the sugars of VND6 were longer than those in the wild-type, because fewer sugars would have been used to synthesize polysaccharides. These data indicate that polymeric component analysis by the DMSO/pyridine system is useful for comparing transgenic organisms. Namely, the 13 C-labeling technique allows reduction of the required sample amounts, e.g., less than 5 mg in our case, to approximately 10-fold lower than the ordinary case. This study may, therefore, provide valuable and detailed information relating to the improvement of biomass production.  carried out as described previously 58,67 . For shoot amplification, Murashige and Skoog (MS) medium (Sigma-Aldrich, St. Louis, MO), which contains indole-3-butyric acid and 6-benzylaminopurine, was placed on a sterilized plate with stalks of cut poplar stems. After approximately 30 days of rooting, the shoots were transferred to plant culture test tubes (IWAKI, Chiba, Japan) containing MS medium. After approximately 30 additional days, the rooted poplars were transferred to a container containing MS medium for plant culture (Combiness, Nazareth, Belgium). Stable isotope labeling of poplars using the above growing system was conducted using previously described methods 38,42 . The poplars were grown in the plant culture until they reached a height of approximately 10 cm, i.e., 35 days. Extraction and solution NMR of poplar. The lyophilized 13 C-labeled poplar ( 13 C-poplar) was crushed and extracted using previously described methods 68 . Briefly, aqueous buffer (100 mM potassium phosphate, pH 7.0) was used for extraction. Solution NMR experiments were performed using a DRU-700 spectrometer (Bruker Biospin, Billerica, MA, USA) equipped with a Z-axis cryogenically cooled probe operating at 25 °C. For 2D 1 H-13 C-HSQC analysis, a total of 128 complex f1 ( 13 C) and 1,024 complex f2 ( 1 H) points were recorded using 80 scans per f1 increment. The spectral widths and offset frequencies were 7,042 Hz (40 ppm) and 9,328 Hz (13.3 ppm) for f1 and f2, respectively. The chemical shifts were referenced to the methyl group of the sodium 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) internal standard (0 ppm of 1 H and 0 ppm of 13 C).

Method
Preparation of insoluble cell-wall-rich sample from poplar. The extracted residue sample of 13 C-poplar was prepared using previously described methods 34 . Briefly, chloroform, methanol, and sodium dodecyl sulfate were used to remove low-molecular-weight metabolites, lipids, and proteins from the poplar sample.
HR-MAS measurements of poplar sample, insoluble cell-wall-rich poplar sample, lipids, and pectins. HR-MAS measurements of the samples and standards without extraction were conducted using DRX-400 and DRX-500 spectrometers (Bruker Biospin, Billerica, MA, USA) equipped with Z-axis high-resolution magic angle spinning probes. The measurement temperature was maintained at 25 °C. The MAS rotational speed was regulated at a constant 4,000, 10,000, and 6,000 Hz for the 3D analysis of the poplar sample, the cell-wall-rich sample, and others, respectively. For 2D 1 H-13 C-HSQC measurements, the DRX-500 spectrometer was used and a total of 160 complex f1 ( 13 C) and 1,024 complex f2 ( 1 H) points were recorded using 56 scans per f1 increment. The spectral widths and offset frequencies were 5,031 Hz (40 ppm) and 6,667 Hz (13.3 ppm) for f1 and f2, respectively. For 3D HCCH-COSY measurements, the DRX-400 spectrometer was used and a total of 120 complex f1 ( 1 H), 56 complex f2 ( 13 C), and 1,024 complex f3 ( 1 H) points were recorded using 24 scans per f1 and f2 increments. The spectral widths and offset frequencies were 4,802 Hz (12.0 ppm), 4,025 Hz (40 ppm), and 5,593 Hz (14.0 ppm) for f1, f2, and f3, respectively. The water phase, including the sample in the rotor, was retained under MAS conditions. The chemical shifts were referenced to the methyl group of the DSS internal standard.
Solid NMR of poplar. Freeze-dried 13 C poplar was inserted into 4-mm ø ZrO 2 rotor. Solid NMR experiments were performed using a DRX-400 spectrometer (Bruker Biospin, Billerica, MA, USA) equipped with a 4-mm MAS triple resonance probe. The MAS rotational speed was regulated at a constant 13,500 Hz. For MAS-J-HMQC, 72 complex f1 ( 1 H) and 768 complex f2 ( 13 C) points were recorded using 160 scans per f1 increment. The spectral widths and offset frequencies were 11,043 Hz (27 ppm) and 25,253 Hz (250 ppm) for f1 and f2, respectively. The cross-polarization contact time was set to 3.0 ms. For refocused INADEQUATE, 96 complex f1 and 768 complex f2 points were recorded using 896 and 1,200 scans per f1 increment with the τ delay set to 3.4 and 6.0 ms. The spectral widths and offset frequencies were 48,077 Hz (480 ppm) and 24,038 Hz (240 ppm) for f1 and f2, respectively. The cross-polarization contact time was set to 2.0 ms. DMSO-d 6 /pyridine-d 5 system. Freeze-dried 13 C poplar was crushed in the same way as previously described 68 . The crushed sample was ball-milled with a FRITSCH pulversette P5 vibratory ball mill (FRITSCH, Idar-Oberstein, Germany) vibrating at 400 rpm using zirconium dioxide (ZrO 2 ) vessels (50 mL) containing ZrO 2 ball bearings (5 × 5 mm) for 12 h (in cycles comprising 10-min grinding/10-min interval). The milled sample was extracted with ethanol (shaking, 50 °C, 5 min, 3 times) and distilled water (shaking, 50 °C, 5 min, 3 times). The sample was dissolved in DMSO-d 6 /pyridine-d 5 (4:1) (Cambridge Isotope Laboratories, Andover, MA), shaken at 50 °C for 30 min, and centrifuged at 20,000 g for 5 min. Solution NMR experiments were performed on the soluble matter from the sample using the DRU-700 spectrometer (Bruker Biospin, Billerica, MA, USA) equipped with a Z-axis cryogenically cooled probe operating at 45 °C. For 2D 1 H-13 C-HSQC measurements to assign polymeric components combined with 3D HCCH-COSY, a total of 512 complex f1 ( 13 C) and 1,024 complex f2 ( 1 H) points were recorded using 16 scans per f1 increment. The spectral widths and offset frequencies were 26,410 Hz (150 ppm) and 9,328 Hz (13.3 ppm) for f1 and f2, respectively. For 3D HCCH-TOCSY measurements, a total of 128 complex f1 ( 1 H), 64 complex f2 ( 13 C), and 1,024 complex f3 ( 1 H) points were recorded using 24 scans per f1 and f2 increments. The spectral widths and offset frequencies were 7,002 Hz (10.0 ppm), 7,042 Hz (40 ppm), and 9,803 Hz (14.0 ppm) for f1, f2, and f3, respectively. The mixing time was set to 14 ms. For 2D 1 H-13 C-HSQC measurements of wild-type and two transgenic 13 C-poplars, a total of 256 complex f1 ( 13 C) and 1,024 complex f2 ( 1 H) points were recorded using 16 scans per f1 increment. The spectral widths and offset frequencies were 26,410 Hz (150 ppm) and 9,328 Hz (13.3 ppm) for f1 and f2, respectively. The chemical shifts were referenced to a DMSO internal standard (2.49 ppm for 1 H and 39.5 ppm for 13 C). Data processing. Each NMR spectrum was processed using the NMRPipe software 69 with appropriate window functions, zero-filling, linear predictions, and polynomial baseline corrections. Metabolites were assigned using a recent NMR chemical database 50 (unpublished data), while polysaccharides were assigned, in part, on the basis of previously reported results 28 .