Exploiting the Complementarity between Dereplication and Computer

Article pubs.acs.org/jnp

Exploiting the Complementarity between Dereplication and Computer-Assisted Structure Elucidation for the Chemical Profiling of Natural Cosmetic Ingredients: Tephrosia purpurea as a Case Study Jane Hubert,*,† Sébastien Chollet,† Sylvain Purson,†,‡ Romain Reynaud,‡ Dominique Harakat,† Agathe Martinez,† Jean-Marc Nuzillard,† and Jean-Hugues Renault† †

Institut de Chimie Moléculaire de Reims (UMR CNRS 7312), SFR CAP’SANTE, UFR de Pharmacie, Université de Reims Champagne-Ardenne, Reims, France ‡ Soliance−Givaudan, Pomacle, France S Supporting Information *

ABSTRACT: The aqueous-ethanolic extract of Tephrosia purpurea seeds is currently exploited in the cosmetic industry as a natural ingredient of skin lotions. The aim of this study was to chemically characterize this ingredient by combining centrifugal partition extraction (CPE) as a fractionation tool with two complementary identification approaches involving dereplication and computer-assisted structure elucidation. Following two rapid fractionations of the crude extract (2 g), seven major compounds namely, caffeic acid, quercetin3-O-rutinoside, ethyl galactoside, ciceritol, stachyose, saccharose, and citric acid, were unambiguously identified within the CPE-generated simplified mixtures by a recently developed 13C NMR-based dereplication method. The structures of four additional compounds, patuletin-3-O-rutinoside, kaempferol-3-Orutinoside, guaiacylglycerol 8-vanillic acid ether, and 2-methyl-2-glucopyranosyloxypropanoic acid, were automatically elucidated by using the Logic for Structure Determination program based on the interpretation of 2D NMR (HSQC, HMBC, and COSY) connectivity data. As more than 80% of the crude extract mass was characterized without need for tedious and labor-intensive multistep purification procedures, the identification tools involved in this work constitute a promising strategy for an efficient and time-saving chemical profiling of natural extracts.

dereplication and computer-assisted structure elucidation (CASE).4 Previous phytochemical investigations have reported a diverse array of secondary metabolites in the genus Tephrosia including flavones, flavanones, isoflavones, chalcones, rotenoids, and prenylated flavonoids.5 In view of these chemical data, and considering the hydro-alcoholic nature of the extract, two independent CPE fractionation experiments were applied to separate metabolites in a polarity range from moderately polar to hydrophilic constituents. Following this fractionation step, a recently developed 13C NMR-based dereplication procedure was used to rapidly and unambiguously identify known compounds without purifying individual constituents.6 The principle of this dereplication approach is to generate simplified mixtures of metabolites from a crude extract by a rapid supportfree liquid−liquid fractionation process. The mixtures generated are analyzed by 13C NMR spectroscopy. Then, all detectable 13C NMR resonances are automatically collected and aligned across spectra of the fraction series, and hierarchical clustering analysis (HCA) is applied on the resulting data set.

Tephrosia purpurea (Fabaceae) is an Indian plant traditionally used in Ayurvedic medicine to treat ulcers, leprosy, rheumatism, asthma, and bronchitis.1 Over the past few years, several reports have confirmed these traditional uses by demonstrating anti-inflammatory, antiulcer, antimicrobial, or hepatoprotective activities of different parts of this species in in vitro and in vivo models.1 In 1995, a 30% aqueous-ethanol extract produced industrially from T. purpurea seeds was patented for its ability to stimulate the synthesis of β-endorphin by keratinocytes.2 This extract is still traded on the European market as a natural cosmetic ingredient of lotions. Two oligosaccharides, stachyose and ciceritol, were suggested to be involved in the above-mentioned activity and thus have been considered as the best “phytochemical markers” of this natural cosmetic ingredient.3 However, little is known about the distribution and identification of other potentially interesting metabolites in this extract. The objective of the present work was to perform a more extensive chemical characterization of the 30% aqueous-ethanol crude extract of T. purpurea seeds in view of further biological investigation. Thus, we exploited the combination of centrifugal partition extraction (CPE) as a fractionation tool with two different and complementary identification methods involving © XXXX American Chemical Society and American Society of Pharmacognosy

Received: February 23, 2015

A

DOI: 10.1021/acs.jnatprod.5b00174 J. Nat. Prod. XXXX, XXX, XXX−XXX

Journal of Natural Products

Article

The obtained 13C NMR chemical shift clusters are finally assigned to their corresponding molecular structures using a locally developed 13C NMR chemical shift database. This approach was successfully used to unambiguously identify seven major ellagic acid derivatives, triterpenoids, saponins, and flavonoids from a bark extract of the African tree Anogeissus leiocarpus,6 as well as several depsides in a crude lichen extract.7 A dereplication process remains by definition limited to the identification of known compounds. Therefore, for a more extensive chemical profiling we completed this dereplication approach by using the Logic for Structure Determination (LSD) program.8 LSD is based on the interpretation of connectivity data found in the 2D NMR spectra of a single chemical entity without need for spectroscopic database or chemical shift values.8,9 A range of applications describing how complex structures of natural compounds can be determined with the assistance of the LSD structure elucidation program are available.10−12 Herein we will demonstrate that the combination of both identification approaches enabled the characterization of diverse metabolites representing a major mass proportion of the starting crude extract.

■

RESULTS AND DISCUSSION A first CPE fractionation (experiment A) was performed by using a biphasic solvent system composed of methyl tert-butyl ether (MTBE), CH3CN, and H2O (3/3/4, v/v) in order to recover the moderately polar compounds from the crude aqueous-ethanol extract of T. purpurea seeds. CPE is a solid support-free liquid−liquid separation technique involving the distribution and the transfer of solutes between at least two immiscible liquid phases according to their distribution coefficient. The specific design of the CPE column partition cells enables the pumping of the mobile phase at flow rates ranging from 10 to more than 100 mL/min and the handling of samples on a multigram scale with a column volume of only 300 mL.13 When applied in the elution mode, CPE has been proven as an efficient and cost-effective alternative for the rapid and selective fractionation of natural extracts.14 After loading 2 g of the crude aqueous-ethanol extract of T. purpurea seeds into the CPE column, the different constituents of the extract were eluted in a decreasing order of polarity by pumping the organic phase of the biphasic solvent system for 100 min at 20 mL/min. After combining the collected fractions on the basis of TLC profile similarities, 11 adjacent fractions (F1A−F11A) containing simplified mixtures or even pure compounds were obtained. As illustrated in Figure 1, the total mass recovered from the elution process of experiment A represented 11% of the injected quantity, while 89% of the crude extract constituents the more polar oneswere strongly retained by the stationary phase. These remaining constituents were all recovered as a complex and concentrated mixture by extrusion, dried, and weighed to validate the mass balance. A second CPE fractionation experiment (experiment B) was performed by using a more polar solvent system composed of n-BuOH, HOAc, and H2O (4/1/5, v/v) in order to fractionate the more polar constituents compared to those that eluted over the course of experiment A. After loading 2 g of the crude extract, the organic mobile phase was pumped through the stationary phase at 20 mL/min for 130 min. At the end of this elution step, 85% of the crude extract sample mass was fractionated, and the remaining 15% was recovered in 15 min by extrusion. As in experiment A, the collected fractions were combined according to their TLC chemical profile similarities, yielding a

Figure 1. Mass balance of CPE fractionation experiments A and B performed on the crude aqueous-ethanol extract of Tephrosia purpurea seeds.

final series of 14 fractions (F1B−F14B) representative of the injected sample and including both the elution and extrusion steps (Figure 1). Fractions F1A to F11A and F1B to F14B produced in experiments A and B, respectively, were then analyzed by 13C NMR spectroscopy. Automatic peak picking and alignment of 13 C NMR signals across spectra of each fraction series resulted in a table with 11 columns (one per fraction) and 217 rows (one per chemical shift bin containing at least one 13C NMR signal in at least one fraction) for experiment A and in a table with 14 columns and 312 rows for experiment B. Each matrix was submitted to hierarchical clustering analysis on the rows. In this way, statistical correlations between 13C NMR resonances belonging to a single structure within the fraction series were readily visualized as “chemical shift clusters” in front of the corresponding dendrograms. The HCA correlation maps obtained from experiments A and B are presented in Figures 2 and 3, respectively. Several well-defined clusters were intensely colored in red on the two-dimensional HCA correlation map of experiment A (Figure 2). Cluster 1 corresponded to an intense cluster of eight 13C NMR chemical shifts. After entering these chemical shifts into the local database containing 13C NMR data of natural compounds including known metabolites from the genus Tephrosia, the structure of caffeic acid was proposed as the first hit of over 114 proposals. This structure was confirmed by checking all chemical shifts of caffeic acid in raw NMR data of fractions F1A and F2A, where the intensity of cluster 1 was predominant, and by MS analyses, revealing an ion at m/z 179 [M − H]− in the negative ionization mode. By applying the same approach, cluster 6 was identified as quercetin 3-Orutinoside. Its structure matched as the first hit of over 31 proposals from the database and was confirmed by checking all 13 C NMR chemical shifts in the raw data of fractions F8A and B



Article

Figure 2. 13C NMR chemical shift clusters obtained by applying HCA on CPE fractions of experiment A: Moderately polar compounds representing 11% of the crude extract mass.

[M + Na]+ in the positive ionization mode. By applying the same approach, four additional intense 13C NMR chemical shift clusters, 13, 14, 15, and 16, were identified as ciceritol, saccharose, stachyose, and citric acid, respectively. The intense cluster 12 did not correspond to any molecular structure stored in the database; however, the 13C NMR profile of fraction F6B revealed a set of intense signals in addition to those of ethyl galactoside. Additional 2D NMR analyses including HSQC, HMBC, and COSY were thus performed on this fraction, and the structure of 2-methyl-2-glucopyranosyloxypropanoic acid was readily elucidated (Figure 3). The molecular structure of this compound was further confirmed by MS analysis of F6B, revealing a molecular ion at m/z 265 [M − H]− in the negative ionization mode. This molecule is reported here for the first time as a natural product. We found only one reference reporting the synthesis of several linamarin and glycolonitrile cellulosides among which the same molecular structure was described.15

F9A, as well as by MS analyses revealing molecular ions at m/z 609 [M − H]− and m/z 633 [M + H]+ in the negative and positive ionization modes, respectively. The other chemical shift clusters obtained from experiment A did not match any of the compounds stored in the database, hence precluding identification of additional moderately polar compounds by dereplication. In Figure 3, the HCA correlation map of experiment B also revealed several intense chemical shift clusters corresponding this time to the major polar organic molecules present in the initial aqueous-ethanol extract of T. purpurea seeds. Cluster 11 corresponded to an intense cluster of five 13C NMR chemical shifts. After entering these chemical shifts into the database, the structure of ethyl galactoside was proposed as a first hit over 49 proposals. This structure was confirmed by checking all 13C NMR chemical shifts of ethyl galactoside in raw NMR data of fractions F6B and F7B, where the intensity of cluster 11 was predominant, and by MS analyses revealing an ion at m/z 231 C



Article

Figure 3. 13C NMR chemical shift clusters obtained by applying HCA on CPE fractions of experiment B.

compounds including ethyl galactoside, ciceritol, saccharose, stachyose, 2-methyl-2-glucopyranosyloxypropanoic acid, and citric acid. The presence of ciceritol and stachyose in T. purpurea seeds was reported previously,3 but this is the first time that saccharose, ethyl galactoside, 2-methyl-2-glucopyr-

It needs to be emphasized that the molecular structures identified from clusters 11 to 16 corresponded to the largely predominant compounds present in fractions F5B to F14B. Taking into consideration that in Figure 1 these fractions represented 84% of the initial crude extract mass, it is clear that nearly 84% of the crude extract mass was made of five major D



Article

Figure 4. 13C NMR profile of fractions F3A, F6A, and F10A obtained from the CPE fractionation of the crude aqueous-ethanol extract of Tephrosia purpurea seeds (experiment A).

B remained unassigned. To alleviate this problem, we considered the application of a computer-assisted structure elucidation program referred to as Logic for Structure Determination. The aim of LSD is to find all possible molecular structures of an organic compound that are compatible with its spectroscopic data without reference to a chemical shift database.16,17 The principle is to define all 1H−1H and 1 H−13C through-bond chemical shift correlations detected in COSY, HSQC, and HMBC NMR spectra of the target compound, to determine its molecular structure with the help of MS analysis, and to define the status of each carbon and oxygen atom (hybridization state, number of attached hydrogens, and electric charge). Of course, the major constraint that governs the performance of LSD is the necessity of having a

anosyloxypropanoic acid, and citric acid are identified in this species. Thus, dereplication based on HCA pattern recognition of 13 C NMR signals within simplified mixtures of compounds permitted the rapid identification of the main chemical constituents of the crude aqueous-ethanol extract of T. purpurea seeds without the need to purify individual constituents. This approach, however, focused only on the identification of known compounds, i.e., either compounds already identified in the genus Tephrosia or known natural metabolites recorded in our nonexhaustive and still developing database (in May 2015, n = 1200 metabolites are stored in the database). As a consequence, a range of compounds even corresponding to well-defined chemical shift clusters on the HCA maps of experiments A and E



Article

Figure 5. LSD solutions proposed for the automatic structure determination of the metabolites present in fractions F3, F6, and F10 (experiment A).

sufficiently “pure sample”. Within the fraction series of experiment A, 13C NMR spectra of F3A, F6A, and F10A each assumed the presence of a major compound (Figure 4). We thus performed additional 2D NMR analyses on these three

fractions. The LSD input files for automatic structure determination in F3A, F6A, and F10A are supplied as Supporting Information. For the compound present in F3A, LSD produced three solutions, all containing two aromatic rings linked by the F



Article

same −O−CH(CH2OH)−CH(OH)− moiety (Figure 5). The first solution (s1) was characterized by the presence of a 5hydroxy-2-methoxybenzoic acid moiety instead of a vanillic acid unit in the second solution (s2) or a 2-hydroxy-5-methoxybenzoic acid unit in the third solution (s3). All solutions were chemically plausible, but after running the ACD/Laboratories CNMR spectrum predictor to evaluate the validity of these solutions, we observed that s2 was the only molecular structure for which all experimental chemical shifts matched with the corresponding predicted values. Therefore, solution s2 was retained and the predominant compound of F3A was identified as a guaiacylglycerol 8-vanillic acid ether. The presence of this compound, corresponding to clusters 2 and 8 on the HCA map of experiment A and to cluster 2 on the HCA map of experiment B (Figure 2), was reported in Boreava orientalis (Cruciferae) and shown to possess radical scavenging activity.18 However, this is the first report of its presence in the genus Tephrosia. For the aglycone part of the predominant compound in fraction F6A, four solutions were proposed by the LSD program (Figure 5). The first solution (s1) corresponded to a kaempferol skeleton. Solutions s2, s3, and s4 contained a 10-, 8-, and 12-membered ring, respectively, and thus were considered unrealistic from a biosynthetic point of view. In addition, s1 was the only solution for which all experimental and predicted chemical shifts matched correctly. Regarding the sugar moiety, 1D and 2D NMR data as well as MS analyses revealed the presence of a 6-O-rhamnosyl-glucose (rutinosyl) moiety. Therefore, the predominant compound of fraction F6A, corresponding to cluster 9 on the HCA map of experiment A, was identified as kaempferol-3-O-rutinoside (Figure 2). Several kaempferol derivatives have been reported in Tephrosia species including T. vogelii, T. candida, and T. procumbens,5 but kaempferol-3-O-rutinoside has not been reported from the genus Tephrosia. For the compound present in F10A, four solutions were produced by LSD (Figure 5). All solutions suggested a flavonoid skeleton, but again the second molecular structure was the only one for which all 1H and 13C NMR predicted chemical shifts matched with experimental data. Therefore, the predominant compound of F10A was identified as patuletin-3-O-rutinoside. This compound corresponded to cluster 4 on the HCA map of experiment A (Figure 2). This compound has not been reported from the Tephrosia genus. The lack of data regarding the distribution of glycosylated flavonoids in T. purpurea results mainly from the challenge of separating and purifying this class of molecules.19 Using CPE as a fractionation tool facilitated the selective separation of the structurally similar quercetin-3-O-rutinoside, kaempferol-3-Orutinoside, and patuletin-3-O-rutinoside. However, complex mixtures of minor glycosylated flavonoids were present in several fractions, particularly in fractions F7A and F8A, and this hampered the identification of clusters 3 and 7, respectively. The well-defined cluster 10 exhibits resonances common to several glycosylated flavonoids such as patuletin-3-O-rutinoside (cluster 4) and quercetin-3-O-rutinoside (cluster 6). In summary, a total of 11 metabolites were identified in a crude aqueous-ethanol extract of T. purpurea seeds after two rapid CPE fractionation steps. The dereplication method based on HCA pattern recognition of 13C NMR resonances in the series of simplified mixtures obtained by CPE permitted the rapid identification of seven major compounds of the extract. Completing this approach by 2D NMR analyses on some CPE fractions and by using the LSD program facilitated the identification of four additional compounds. Some of the

identified molecular structures were structurally close (several glycosylated flavonoids), while others belonged to different chemical classes such as flavonoids, sugars, and organic acids. All compounds were identified without need for labor-intensive multistep purification procedures. Thus, exploiting the complementarity between dereplication and computer-assisted structure elucidation constitutes a promising and coherent analytical strategy for a global, time-saving, and cost-effective chemical profiling of natural extracts.

■

EXPERIMENTAL SECTION

Chemicals, Reagents, and Plant Material. Acetonitrile, toluene, EtOAc, MTBE, and n-BuOH were purchased from Carlo Erba Reactifs SDS (Val de Reuil, France). HOAc and HCO2H were purchased from Merck (Darmstadt, Germany). Methanol-d4 and DMSO-d6 were purchased from Eurisotop (Saint-Aubin, France). Deionized H2O was used to prepare all aqueous solutions. The 30% aqueous-ethanol extract of T. purpurea seeds was provided by the company Soliance (Pomacle, France). Centrifugal Partition Extraction. Fractionation experiments were developed on a lab-scale CPE column of 303 mL capacity (FCPE300, Rousselet Robatel Kromaton, Annonay, France) containing seven circular partition disks, engraved with a total of 231 oval partition twin-cells (∼1 mL per twin cell). The liquid phases were pumped with a Knauer Preparative 1800 V7115 pump (Berlin, Germany). The column was coupled online with a UVD 170S detector set at 210, 254, 280, and 366 nm (Dionex, Sunnyvale, CA, USA). Fractions were collected by a Pharmacia Superfrac collector (Uppsala, Sweden). Two independent CPE experiments were developed in the elution mode to recover either the moderately polar constituents (experiment A) or the most hydrophilic constituents (experiment B) of the initial aqueous-ethanol extract of T. purpurea seeds. Experiment A was performed by using a triphasic solvent system composed of MTBE, CH3CN, and H2O (3/3/4, v/v). Experiment B was performed by using a triphasic solvent system composed of n-BuOH, HOAc, and H2O (4/1/5, v/v). In both experiments the aqueous phase of the triphasic solvent system was used as the stationary phase. The sample solution was prepared by directly dissolving 2 g (experiments A and B) of the crude extract in a 2/1 (v/v) mixture of aqueous and organic phases of the biphasic solvent system, respectively. The organic phase was pumped through the stationary phase in the ascending mode at 20 mL/min. The rotation speed was set at 1000 rpm for experiment A and at 1200 rpm for experiment B to ensure a satisfying initial stationary phase retention of ∼60% in both experiments. At the end of the elution step, the constituents retained inside the CPE column were extruded by pumping the aqueous phase in the ascending mode while maintaining the same rotation speed. Fractions of 20 mL were collected over the whole experiments. The fractionation conditions are summarized in Table 1. All collected fractions were spotted on Merck TLC plates coated with silica gel 60 F254 and developed with toluene/ EtOAc/AA/FA (3/7/1/1, v/v). After inspection at 254 and 366 nm, the plates were sprayed with vanillin/H2SO4 and heated to ∼100 °C for 5 min. Fractions of similar composition were combined, resulting in a series of 11 fractions for experiment A (F1A to F11A) and 14 fractions for experiment B (F1B to F14B). Rapid Identification of Known Compounds: Dereplication. A recently developed dereplication method was used to identify metabolites directly within mixtures in the series of fractions obtained from experiments A and B. All fractions were dried under vacuum at 40 °C, and a maximum of 20 mg of each was dissolved in 500 μL of methanol-d4 (experiment A) or DMSO-d6 (experiment B). NMR analyses were performed at 298 K on a Bruker Avance AVIII-600 spectrometer (Karlsruhe, Germany) equipped with a TXI cryoprobe optimized for 1H detection and with cooled 1H and 13C coils and preamplifiers. 13C NMR spectra were acquired at 150.91 MHz. A standard zgpg pulse sequence was used with an acquisition time of 0.909 s and a relaxation delay of 3 s. For each sample, 1024 scans were coadded to obtain a satisfactory signal-to-noise ratio. The spectral G



Article

CASE approach using the LSD program. LSD is a molecular structure generator that mainly relies on homo- and heteronuclear 2D NMR data.8,16 2D NMR and MS analyses were performed on fractions containing well-defined 13C chemical shift clusters but not successfully assigned to a molecular structure by the dereplication strategy. EditedHSQC, HMBC, and COSY spectra were acquired by using standard Bruker pulse programs on the same Bruker Avance AVIII-600 spectrometer (Bruker, Karlsruhe, Germany). For MS analyses, fractions were solubilized in MeOH/H2O (1/1, v/v) and directly infused in a quadrupole time-of-flight hybrid mass spectrometer (QTOF Micro, Micromass, Manchester, UK) equipped with an electrospray source. The mass range of the instrument was set at m/z 100−2000 and scan duration was set at 1.5 s in both positive and negative ionization modes. The capillary voltage was 3000 V, the cone voltage was 30 V, and the temperature was 80 °C. Once a molecular formula was defined with the help of MS analysis and NMR data, an LSD input text file was created according to a specific protocol.9 The file contains LSD commands that define the status of all heavy (i.e., non-hydrogen) atoms, 2D NMR correlation data, atom properties, substructure information, and program execution options. The status of a heavy atom is defined by its numeric index, its chemical element symbol, its hybridization state, its multiplicity (i.e., the number of attached hydrogen atoms), and its formal electric charge. The indexes are usually assigned first to carbon atoms, in the decreasing order of their chemical shifts. Carbon multiplicities are readily deduced from the multiplicity-edited HSQC and 1H and 13C 1D NMR spectra. An atom status is sometimes difficult to assign to heteroatoms. In case of ambiguity it may be necessary to create more than one LSD input file, unless the pyLSD wrapper of LSD is used.8 In any case, the reliability of MS data is highly important for the writing of the set of atom statuses. Hydrogen atoms are also defined by a numeric index, so that a carbon and a hydrogen atom that are bound together have the same index. Pairs of atom indexes are needed to encode the one-bond 1H−13C HSQC data, the three-bond 1H−1H COSY data, and the two- or three-bond 1 H−13C HMBC data.20 The user may impose an upper limit to the number of longer range COSY and HMBC data. The user may also directly impose bonds between atoms, either from chemical shift values or from one-bond INADEQUATE,21 from H2BC,22 or from 1,1-ADEQUATE spectra.23 An atom property imposes restrictions to the neighboring atoms of an atom, so that elementary rules of chemical shift or coupling pattern can readily be taken into account. The presence of a substructure may be imposed to the structure generated by LSD. More precisely, any logical combination of substructure search results may be used to determine the acceptance of a solution. Chemical shifts are not taken into account by LSD. LSD finds all the structures that fit with the status, correlation, property, and substructure constraints. The computation times ranged from less than 1 s (F3 A and F6 A) to 4 s (F10A). The ACD/Laboratories CNMR spectrum predictor software (ACD/Laboratories, Ontario, Canada) was used in conjunction with LSD to check the validity of the structures proposed by LSD on the basis of predicted chemical shifts values in comparison to experimental data.

Table 1. Experimental Conditions Optimized for the Fractionation of a Crude Aqueous-Alcoholic Extract of Tephrosia purpurea Seeds experiment A: fractionation of moderately polar constituents solvent system flow rate (ascending mode) rotation speed initial stationary phase retention mass sample loading elution duration extrusion duration

experiment B: fractionation of polar constituents

MtBE/ACN/H2O (3/3/4, v/ v) 20 mL/minb

n-butanol/AAa/H2O (4/1/5, v/v) 20 mL/minb

1000 rpm 63%

1200 rpmc 60%

2g

2g

100 min 20 min

130 min 15 min

a

AA: acetic acid. bInitially from 0 to 20 mL/min in 4 min. cSlight increase as compared to experiment 1 to ensure an initial stationary phase retention of about 60%.

width was 238.9070 ppm, and the receiver gain was set to the highest possible value. A 1 Hz line broadening filter was applied to each FID prior to Fourier transformation. The c13cryo macro was used to correct the typical roll of the baseline due to the high Q of cryoprobes. The spectra were manually phased and baseline corrected using the TOPSPIN 3.2 software (Bruker) and calibrated on the central resonance (δ 47.60 for methanol-d4, δ 39.80 for DMSO-d6). The absolute intensities of all 13C NMR signals were automatically collected by using a minimum intensity threshold of 0.3% (relative to the most intense signal of each spectrum). Each peak list was stored as a text file. The collected peaks in each fraction series (experiments A and B independently) were subsequently binned using a locally developed computer script written in Python language. Its goal was to divide the 13C spectral width (from 0 to 220 ppm) into regular chemical shift intervals (Δδ = 0.2 ppm) and to associate the absolute intensity of each 13C NMR peak to the corresponding bin. The bins for which no signal was detected in any fraction were removed from the bin list. The resulting tables (one table for experiment A and one table for experiment B) were imported into the PermutMatrix version 1.9.3 software (LIRMM, Montpellier, France) for clustering analysis on raw peak intensity values. The classification was performed on the rows only, i.e., on the chemical shift bins. The Euclidian distance was used to measure the proximity between samples, and the Ward’s method was applied to agglomerate the data. The resulting 13C NMR chemical shift clusters were visualized as dendrograms on a twodimensional map. The higher the intensity of the 13C NMR peaks, the brighter the color on the map (Figures 2 and 3). In parallel, a literature survey was performed to obtain names and structures for known T. purpurea metabolites. In total, 231 compounds were found, mostly including flavonoids and to a lesser extent sterols, phenolic acids, and fatty acids. They were added to a locally built 13C NMR chemical shift database (ACD/NMR Workbook Suite 2012 software, ACD/ Laboratories, Ontario, Canada) comprising the chemical shifts and structures of other natural products. Structures were drawn with ACD/Laboratories ChemSketch, and their chemical shifts were assigned to the corresponding carbon positions. When 13C NMR chemical shifts were not available in the literature, a predicted spectrum was calculated with the ACD/Laboratories CNMR Predictor software and the resulting 13C NMR chemical shifts were supplied to the database. For metabolite identification, each 13C NMR chemical shift cluster obtained from HCA was manually submitted to the structure search engine of the database management software. A 13C NMR chemical shift tolerance of ±2 ppm was used, and the computation time for each search was less than 1 s. Computer-Assisted Structure Elucidation of T. purpurea Metabolites: LSD. Metabolite identification was completed by a

■

ASSOCIATED CONTENT

S Supporting Information *

LSD input files for automatic structure determination in F3A, F6A and F10A. 1H and 13C NMR chemical shifts of the identified compounds.The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jnatprod.5b00174.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel: (+33) 03 26 91 83 25. Notes

The authors declare no competing financial interest. H



■

Article

ACKNOWLEDGMENTS The authors thank the Soliance Society, the CNRS, the Ministry of Higher Education and Research, and the “Champagne-Ardenne DRRT” for financial support. The EUprogram FEDER for the PlAneT CPER project is also gratefully acknowledged.

■

REFERENCES

(1) Palbag, S.; Dey, B. K.; Singh, N. K. Chin. J. Nat. Med. 2014, 12, 1−7. (2) http://www.freepatentsonline.com/. (3) Amor, B. B.; Andre, P.; Allaf, K. J. Chromatogr. A 2008, 1213, 118−124. (4) Elyashberg, M. E.; Williams, A. J.; Martin, G. E. Prog. Nucl. Mag. Res. Sp. 2008, 53, 1−104. (5) Chen, Y.; Yan, T.; Gao, C.; Cao, W.; Huang, R. Molecules 2014, 19, 1432−1458. (6) Hubert, J.; Nuzillard, J. M.; Purson, S.; Hamzaoui, M.; Borie, N.; Reynaud, R.; Renault, J. H. Anal. Chem. 2014, 86, 2955−2962. (7) Oettl, S. K.; Hubert, J.; Nuzillard, J. M.; Stuppner, H.; Renault, J. H.; Rollinger, J. Anal. Chim. Acta 2014, 846, 60−67. (8) Nuzillard, J. M.; Massiot, G. Tetrahedron. 1991, 47, 3655−3664. (9) Plainchont, B.; Emerenciano, V. d. P.; Nuzillard, J. M. Magn. Reson. Chem. 2013, 51, 447−453. (10) Nuzillard, J. M.; Connolly, J. D.; Delaude, C.; Richard, B.; Zèches-Hanrot, M.; Le Men-Olivier, L.; Nuzillard, J. M. Tetrahedron 1999, 55, 11511−11518. (11) Nuzillard, J. M. Chin. J. Chem. 2003, 21, 1263−1267. (12) Mulholland, D. A.; Langlois, A.; Randrianarivelojosia, M.; Derat, E.; Nuzillard, J. M. Phytochem. Anal. 2006, 17, 87−90. (13) Hamzaoui, M.; Hubert, J.; Reynaud, R.; Marchal, L.; Foucault, A.; Renault, J. H. J. Chromatogr. A 2012, 1247, 18−25. (14) Hamzaoui, M.; Renault, J. H.; Nuzillard, J. M.; Reynaud, R.; Hubert, J. Phytochem. Anal. 2013, 24, 367−373. (15) Fischer, E. A.; Gerda, S. Chem. Ber. 1919, 52B, 854−68. (16) http://eos.univ-reims.fr/LSD/index_ENG.html. (17) Elyashberg, M.; Williams, A. J. Computer-Based Structure Elucidation from Spectral Data, Lecture Notes in Chemistry, 89; Springer-Verlag: Berlin, 2015. (18) Sakushima, A.; Ohno, K.; Maoka, T.; Coskun, M.; Guvenc, A.; Erdurak, C. S.; Ozkan, A. M.; Seki, K.; Ohkura, K. Phytochem. Anal. 2003, 14, 48−53. (19) De Rijke, E.; Out, P.; Niessen, W. M.; Ariese, F.; Gooijer, C.; Brinkman, U. A. J. Chromatogr. A 2006, 1112, 31−63. (20) Furrer, J. Concepts Magn. Reson. A 2012, 40A, 101−127. (21) Bain, A. D.; Hughes, D. W.; Anand, C. K.; Nie, Z.; Robertson, V. J. Magn. Reson. Chem. 2010, 48, 630−641. (22) Nyberg, N. T.; Duus, J.Ø.; Sørensen, O. W. J. Am. Chem. Soc. 2005, 127, 6154−6155. (23) Buevich, A. V.; Williamson, R. T.; Martin, G. E. J. Nat. Prod. 2014, 77, 1942−1947.

I


Exploiting the Complementarity between Dereplication and Computer

Recommend Documents