Chapter 4. Metabolic characterization of green pods from Vanilla planifolia accessions grown in La Réunion

Transcription

1 Chapter Metabolic characterization of green pods from Vanilla planifolia accessions grown in La Réunion Abstract Large phenotypic variation has been observed between the cultivated vanillas since a single genetic source of Vanilla planifolia was spread to the Indian Ocean and the Indonesia in the 9th century. In order to differentiate the cultivated vanilla plants, genetic studies have been conducted in the past on the plants grown in various regions such as the French island, La Réunion. However, the genetic difference was not big enough to differentiate diverse accessions of V. planifolia. In this study, metabolomics, in which genetic variation could be amplified, was employed to delve into the variation between the cultivated vanilla plants. To obtain a broad view of the metabolome, nuclear magnetic resonance (NMR) spectroscopy was applied to the analysis of V. planifolia green pods. Principal component analysis (PCA) and partial least squares-discriminant analysis (PLS- DA) of the data showed that the accessions could be differentiated according to their glucovanillin and glucosidase A and B contents. Furthermore, a correlation between the glucovanillin content and the pod length, number of flower and growth capacity of the accessions has been observed from the multivariate data analysis. Adapted from: T. L. Palama, A. Khatib, Y. H. Choi, B. Côme, I. Fock, R. Verpoorte, H. Kodja Metabolic characterization of green pods from Vanilla planifolia accessions grown in La Réunion Submitted for publication

2 Green pods metabolic characterization. Introduction The genus Vanilla belongs to the Orchidaceae family. Among the more than 00 species present in the genus (Purseglove et al., 9), Vanilla planifolia Andrews, V. tahitensis J.W. Moore and V. pompona Schiede are the only cultivated species. The major natural vanilla flavor used in food, beverages, cosmetics and tobacco comes from the pods of V. planifolia. This species has its origin in Mexico where the pods were used for medicinal purposes as well as flavoring for hot chocolate drinks at the court of Emperor Montezuma (Bruman, 9). In the mid-th century, the first cuttings of V. planifolia were brought to Europe from where they were disseminated to the Indian Ocean and the Indonesia in the 0s and 0s (Bouriquet, 9; Bory et al., 00c). The lack of pollinators in the areas of introduction prevented natural reproduction and pod production. In spite of some controversies in the history (Arditti et al., 009), a -years-old French-owned slave by the name of Edmond Albius is known to have discovered in, in the island of La Réunion, a practical method to hand-pollinate the plant, allowing global cultivation of vanilla. Moreover, events of seed germination are not common. For this reason, vanilla crops are established from stem cuttings collected from healthy and vigorous growing vines (Purseglove et al., 9). Nowadays, V. planifolia is cultivated in various tropical countries, i.e. Mexico, Uganda, Papua New Guinea, Indonesia, India and islands such as Comoros, Mayotte, Tahiti and La Réunion but Madagascar is predominant on the vanilla market with around,00 tons of vanilla pods, representing more than half of the world production (Gleason-Allured, 009). In the many areas of introduction, several phenotypic variations have been described. This is the case in La Réunion where two major types of V. planifolia were characterized: the so-called Classique with a light green colour, flat leaves and pods progressively narrow and the Mexique with darker and more bluish leaves, a central gutter and curved sides and cylindrical pods up to the stem. Five others minor types can also be observed (Bory et al., 00b). In spite of the phenotypic variation observed, a low level of diversity have been revealed by amplified fragment length polymorphism (AFLP) (D max = 0.06) and Random Amplified Polymorphic DNA (RAPD) markers (D = 0.0) (Besse et al., 00; Bory et al., 00c). Similar studies have been conducted in other areas (Lubinsky et al., 00; Minoo et al., 00; Verma et al., 009). Nevertheless, this limited genetic variation is in agreement with the vegetative mode of propagation of V. planifolia. And none of the phenotypes observed in La Réunion could genetically be differentiated from the others

3 Chapter (Bory et al., 00c). However, the phenotype of a plant can also be characterized by the whole set of its metabolites. The major compound responsible for vanilla flavor is the vanillin, -hydroxy--methoxybenzaldehyde, but phenolic compounds such as p- hydroxybenzaldehyde, p-hydroxybenzoic acid and vanillic acid are also present in vanilla cured pods. Thus, most studies performed on phytochemistry of V. planifolia have focused on the phenolic content of the pods which developed their flavor only after a curing process (Voisine et al., 99; Havkin-Frenkel and Dorn, 99; Dignum et al., 00; Pérez-Silva et al., 006; Sinha et al., 00; Schwarz and Hofmann, 009). Some phenolic glucosides have also been identified in the green pods as the precursors of the flavor compounds (Leong et al., 99b; Tokoro et al., 990; Kanisawa et al., 99; Negishi and Ozawa, 996; Dignum et al., 00; Odoux, 006; Palama et al., 009). In an organism, possible genetic variation can be viewed in amplified form in metabolomic pools. But conventional metabolic analysis covers a too limited number of metabolites to show a broad view of the biochemical status. Metabolomics has as ultimate objective both qualitative and quantitative analysis of all primary and secondary metabolites of an organism (Verpoorte et al., 00). Chemical analysis techniques to be applied to metabolite profiling should be rapid, reproducible and stable over time, while requiring only simple sample preparation. Nuclear magnetic resonance spectroscopy (NMR) potentially meets these requirements. Recently, the combination of NMR and multivariate data analysis has been applied to various plants to find metabolites involved in species or cultivar differentiation (Choi et al., 00a; Choi et al., 00; Abdel-Farid et al., 00), in fruit quality (Mounet et al., 00; Deborde et al., 009) and geographical region differentiation (Son et al., 009). In this study, we report a H-NMR-based metabolomic approach to evaluate the metabolic variation of green pods obtained from different accessions of V. planifolia grown in La Réunion, France.. Materials and Methods.. Flowering and vegetative growth Diversity of cultivated vanilla plants was prospected in 999 in the tropical French island, La Réunion. Accessions of Vanilla planifolia were collected from different farmers

4 Green pods metabolic characterization and grown in a collection shade house in Saint-André, La Réunion (France). This shade house was used to select elite accessions, thus an evaluation of the vegetative growth and the flowering capacity of each accession was performed. Flowering capacity was determined by the number of inflorescences on three growing plants per accession in the years 00, 00 and 00. The vegetative growth of each accession was evaluated in 00 by numbering to nodes-long cuttings obtained from three growing plants. Because of the apical dominance, inflorescences never appear before the 0th node and thus, the collected cuttings didn t possess any inflorescences... Plant material for metabolomics A second shade house with the same plants was made in 00 in Sainte-Marie, La Réunion. Vanilla planifolia flowers were hand-pollinated and labeled during the second week of November 00. Five eight-months-old pods were collected from flowering accessions of V. planifolia. Length and weight of the pods were recorded before their immersion in liquid nitrogen and storage in -0 C. Frozen pods were then freeze-dried to reduce possible enzymatic degradation of metabolites. The pods were ground to a fine powder with a mortar and pestle before storage of the powder in a cold room (-0 C) until analysis. For each accession, analysis was done on the five pods from a single plant... NMR Analysis Extraction Freeze-dried plant material (0 mg) was transferred to a ml microtube. A volume of. ml of a mixture of KH PO buffer (ph 6.0) in D O containing 0.0% trimethylsilylpropionic acid sodium salt (TMSP, w/w) and CH OH-d (:) was added to the plant samples. The mixture was vortexed at room temperature for min, ultrasonicated for 0 min, and centrifuged at 000 rpm for 0 min. An aliquot of 0. ml was used for NMR analysis. 6

5 Chapter Measurements and Data Analysis All of the NMR parameters were the same as those used by Abdel-Farid et al. (00) and Jahangir et al. (00b). H-NMR and D J-resolved spectra were recorded at C on a 00 MHz Bruker DMX-00 spectrometer (Bruker, Karlsruhe, Germany) operating at a proton NMR frequency of 00. MHz. CH OH-d was used as the internal lock. Each H-NMR spectrum consisted of scans requiring a 0 min and 6 s acquisition time with the following parameters: 0.6 Hz/point, pulse width (PW) = 0 (. μs), and relaxation delay (RD) =. s. A presaturation sequence was used to suppress the residual H O signal with low power selective irradiation at the H O frequency during the recycle delay. Free induction decays (FIDs) were Fourier-transformed with LB = 0. Hz. The resulting spectra were manually phased and baseline-corrected and calibrated to TMSP at 0.0 ppm, using XWIN NMR (v.., Bruker Biospin). D J-resolved NMR spectra were acquired using scans per increments for F and K for F using spectral widths of 000 Hz in F (chemical-shift axis) and 66 Hz in F (spin-spin coupling constant axis). A. s relaxation delay was employed, giving a total acquisition time of 6 min. Data sets were zero-filled to points in F, and both dimensions were multiplied by sine-bell functions (SSB = 0) prior to double complex FT. J-resolved spectra tilted by were symmetrized about F and then calibrated, using XWIN NMR (v.., Bruker Biospin). H- H correlated spectroscopy (COSY) and heteronuclear multiple bonds coherence (HMBC) spectra were recorded on a 600 MHz Bruker DMX-600 spectrometer (Bruker). The COSY spectra were acquired with a.0 s relaxation delay and 66 Hz spectral width in both dimensions. The window function for COSY spectra was sine-bell (SSB=0). The HMBC spectra were obtained with a.0 s relaxation delay and 0 Hz spectral width in F and 6 Hz in F. Qsine (SSB =.0) was used for the window function of the HMBC. The optimized coupling constant for HMBC was Hz. The H-NMR spectra were automatically reduced to the ASCII file. Spectral intensities were scaled to total intensity and reduced to integrated regions of equal width (0.0), corresponding to the region of δ The regions of δ.-.9 and.0-. were excluded from the analysis because of the residual signal of H O and CH OH-d, respectively. Bucketing was performed by AMIX software (v.., Bruker Biospin) with scaling on total intensity. Multivariate data analysis such as principal component analysis (PCA), partial least squares analysis (PLS) and partial least squares-discriminant analysis

6 Green pods metabolic characterization (PLS-DA) were performed with the SIMCA-P software (v..0, Umetrics, Umeå, Sweden) with scaling based on Pareto and the unit variance method.. Results and Discussion.. Morphological variation Fourteen accessions were analyzed to obtain some morphological data. All those accessions are from the type Classique of V. planifolia described in La Réunion. In spite of the narrow genetic variation observed within V. planifolia plants from this introduction area (Bory et al., 00c), a morphological variation has been observed within the Classique type. This morphological and physiological variation is highlighted by the number of inflorescences and cuttings obtained from different V. planifolia accessions (Table ). Although there was an increase of the number of inflorescences for most of the accessions from 00 to 00, a variation within the accessions has been observed (Table ). Indeed, the accessions A and A9 start to flower only on four-years-old plants. Furthermore, the number of inflorescences collected in 00 ranges between per plant. Thus, a variation in time and intensity of the flowering is observed within the different V. planifolia accessions of the type Classique. The number of cuttings that can be obtained from one accession was also determined in 00 (Table ). As observed for the flowering, we noticed an important variation between the accessions since the number of cuttings varies between. and. per plant. Nevertheless, no correlation was clearly established between the flowering and the growth capacity, based on our data. Indeed, the accession B has high number of inflorescences and few cuttings whereas the accession 6B6 shows the opposite, i.e.. and. inflorescences,. and 0. cuttings per plant, respectively (Table ). Those morphological parameters are important for accession selection because they are correlated with vanilla pod production and vegetative multiplication capacity. According to those parameters, the accession B seems to be one of the best accessions. But on top of the morphological study, a NMR-based metabolomic analysis was performed to identify possible marker for the selection of elite accessions.

7 Chapter Table : Number of inflorescences and cuttings obtained from different Vanilla planifolia accessions grown in La Réunion. Number of Inflorescences a Number of cuttings a # Accession code B.0 ±.6. ±. 6. ±..0 ±. B.0 ±.6. ± 0.. ±. 0. ±. B 6. ±.0 0. ±.6 0. ±6.9. ±. A0. ±. 0. ±.. ±.6 A ±. 9. ±. 6 A ±.. ±. A. ± 0.0. ±.. ±..0 ±.6 B. ±.6. ±..0 ±.6. ±. 9 B.0 ± ± ± B6. ±.. ±.. ±6. 0. ±. A. ±.. ±. 9.0 ±.9. ±. A0. ±.0 0. ± 0.6. ±.. ±. B 0. ±.6. ± 9.6. ±.9. ±. 9B6. ±.. ±.. ±0.. ±.0 a (data are shown as mean ± standard deviation with n=) 9

8 Green pods metabolic characterization TMSP MeOD 6 Figure : H-NMR spectra (CH OH-d KH PO in D O extract) of -months-old V. planifolia pods in the range of δ 0-0 and Assignments:, glucovanillin;, p- hydroxybenzaldehyde glucoside;, p-hydroxybenzylalcohol glucoside;, bis[-(β-dglucopyranosyloxy)-benzyl]--isopropyltartrate (glucoside A);, bis[-(β-d-glucopyranosyloxy)- benzyl]--(- butyl)tartrate (glucoside B); 6, sucrose;, glucose;, malic acid; 9, homocitric acid.. Visual inspection of H-NMR Spectra and Assignments of compounds Eight-months-old V. planifolia pods of the accessions were analyzed by NMR to compare the metabolic profile between accessions. Different classes of metabolites such as phenolic compounds, carbohydrates and organic acids were detected in the H-NMR spectra of the CH OH-d and D O (:) extract (Figure ). The differences observed from the H- NMR spectra of the accessions were only from a quantitative order. The H-NMR signals were assigned by comparison of H-NMR spectra of the reference compounds. Apart from the chemical shift data of H-NMR, HMBC (heteronuclear multiple bound correlation) spectra can provide evidence for the identification of glucosides. The signals of glucovanillin (Figure ) were clearly distinguished in the H-NMR spectra. Major signals corresponding to H-6 at δ.6 (dd, J =.,.0 Hz), H- at δ. (d, J =.0 Hz), H- at δ. (d, J =. Hz) were observed in the aromatic region of the green pods extract (Figure ). In addition to these aromatic signals, other characteristic signals of 0

9 Chapter glucovanillin, such as H- at δ 9. (s), O-CH of C- at δ.9 (s) and H- of the glucose moiety at δ.9 (d, J =. Hz), were clearly identifiable in the spectra. Signals characteristic of p-hydroxybenzaldehyde glucoside were also detected in the V. planifolia green pods. Resonances at δ.0 (m), 0.9 (d, J = Hz), 0.6 (d, J =.0 Hz) were assigned to the bis[-(β-d-glucopyranosyloxy)-benzyl]--isopropyltartrate (glucoside A). Signals at δ.90 (m),. (m),.0 (m), 0. (d, J =.0 Hz), 0. (t, J =. Hz) were assigned to the bis[-(β-d-glucopyranosyloxy)-benzyl]--(- butyl)tartrate (glucoside B) (Figure ). Glc Glc ' O O " 6' ' " " ' O ' O O HO 6" O " 6 OH R R= H, glucoside A R=CH, glucoside B Figure : Chemical structures of the glucoside A and glucoside B identified by H-NMR in V. planifolia green pods. As reported previously (Chapter ), signals of sucrose, α-glucose, β-glucose and fructose moiety of sucrose, malic acid and homocitric acid were detected in the green pods spectra. Table summarizes all compounds identified in H-NMR spectra with the chemical shifts and the coupling constants of the signals. Integration of the single signal at δ 9. compared to the internal standard TMSP was used for the quantification of the glucovanillin in the green pods. Glucovanillin content ranges between. and 0. mg per grams of dry weight of green pod.

10 Green pods metabolic characterization Table : H Chemical Shifts (δ), Coupling Constants (Hz) of V. planifolia Pods Metabolites Identified by References and Using D and D NMR Spectra (CH OH-d KH PO in D O, ph 6.0). Compound Chemical shifts and coupling constants glucovanillin δ 9. (s),.6 (dd, J =.,.0 Hz),. (d, J =.0 Hz),. (d, J =. Hz),.9 (d, J =. Hz),.9 (s) p-hydroxybenzaldehyde glucoside δ 9. (s),.9 (d, J = 9.0 Hz),.9 (d, J = 9.0 Hz),.0 (d, J =. Hz) glucoside A δ.0 (m), 0.9 (d, J =.0 Hz), 0.6 (d, J =.0 Hz) glucoside B δ.90 (m),. (m),.0 (m), 0. (d, J =.0 Hz), 0. (t, J =.0 Hz) sucrose δ.0 (d, J =.6 Hz), δ. (d, J =. Hz) glucose malic acid homocitric acid δ. (d, J =. Hz), δ. (d, J =.9 Hz) δ.9 (dd, J =.,. Hz),. (dd, J = 6.0,. Hz),.9 (dd, J = 6.0,.0 Hz) δ.0 (d, J = 6.0 Hz),. (d, J = 6.0 Hz),.6 (dd, J = 9.,. Hz),. (dd, J = 9.,. Hz) The metabolomic analysis of developing V. planifolia green pods (between and months after pollination) has been reported previously (Chapter ). Compared with previous results, free forms of vanillin and p-hydroxybenzaldehyde were found only in trace amounts in the H-NMR spectra of -months-old pods of the present study. This difference is in line with the past discussion on the accumulation of the vanillin in pods. Authors have obtained contradictory results, sometimes the vanillin was found in the green pods only in the glucoside form, glucovanillin, sometimes as both in the glucoside and free form (Arana, 9; Ranadive et al., 9; Sagrero-Nieves and Schwartz, 9; Leong et al., 99b; Brodelius, 99; Kanisawa et al., 99; Havkin-Frenkel et al., 999). Thus, our results suggest that the aglycone / glucoside ratios of vanillin and p-hydroxybenzaldehyde could be dependent on the growing conditions.

11 Chapter By using the same analytical method and the same accessions but grown in different shade houses, our results suggest that both hypotheses are right, in fact the total vanillin content of green pods from both shade houses is more or less the same (data not shown). The vanillin can be present in it glucoside form only or in both glucoside and free form. This could be due to environmental factors influencing the metabolome of V. planifolia green pods. As the distance between the two shade houses is only km and considering that the island is, km large, the effect of environmental factors such as air/soil temperature, pluviometry and photoperiod on the metabolome of V. planifolia should be considered for future studies. Such parameters has been reported to affect the phenolic content of olive explants (Roussos et al., 00)... Multivariate Data Analysis For a first overview of the metabolomic variation of green pods from different accessions of V. planifolia, principal component analysis (PCA) was applied to the H-NMR data. PCA forms the basis for multivariate data analysis and allows an unbiased grouping of the samples. The groupings in the data set can be displayed in the scores plot. If the data are mean-centered with no scaling, then a covariance matrix is produced, but if the data are mean-centered and the columns of the data matrix scaled to unit variance, a correlation matrix is produced. In this study, both methods were evaluated, but the unit variance method showed better grouping results (Figure B) compared to the mean-centered method (Figure A). According to the PCA-based discrimination between the samples by principal component (PC), the accessions B, B, B and B are different from the others (Figure B). Results obtained from the PCA using the unit variance method are displayed in a dendrogram (Figure 6). This graphic allowed an easier determination of groupings within the data set. Indeed, three groups can be observed in our data set resulting in three major branches in the dendrogram. One group is composed by the accessions B, B, B and B, whereas B and the rest of the accessions are in the second and third groups, respectively (Figure 6). The data set was rearranged in three classes according to the groups observed in the dendrogram. A partial least squares-discriminant analysis (PLS-DA) was applied to our data with a model built with those classes (Figure ). A good clustering of the samples based on the three classes was then observed from the PLS-DA score plot (Figure A). The

12 Green pods metabolic characterization result of the CV-ANOVA suggested that the model was significant, since a p-value of.e- was obtained (Eriksson et al., 00). Permutation test was also applied and resulted in validation of the model (Supplementary data). Analysis of the loading score plot of this PLS-DA model allowed identification of the metabolites involved in the discrimination of the samples (Figure B). Accessions B, B, B and B contain more glucovanillin, malic acid and homocitric acid, whereas accessions from class have higher glucosides A & B content. The accession B has a lower glucovanillin, malic acid, homocitric acid and glucosides A & B content compared to the other accessions. A 6 6 PC (.%) PC (.%) B 0 PC (6.%) PC (0.0%) Figure : Score plot (PC vs PC) of PCA results using mean centering (A) or unit variance (B) scaling method obtained from H-NMR spectra of green pods from accessions V. planifolia. Accession codes:, B;, B;, B;, A0;, A; 6, A9;, A;, B; 9, B; 0, 6B6;, A;, A0;, B;, 9A6.

13 Chapter Class Class Class Figure 6: Dendrogram of PCA results using unit variance scaling method obtained from H- NMR spectra of green pods from accessions V. planifolia. Accession codes:, B;, B;, B;, A0;, A; 6, A9;, A;, B; 9, B; 0, 6B6;, A;, A0;, B;, 9A6. Length and weight of green pods vary from to cm and from 9. to. grams, respectively. Both variables seem to be positively correlated (Figure ). A PLS analysis has been performed to estimate the correlation of the length, weight and inflorescence variables with the metabolomic data (Figure 9A).

14 Green pods metabolic characterization A 0 Class PLS-Component (.%) Class Class B Class PLS-Component (9.%) 0.0 w*c Glucose Sucrose p-hydroxybenzaldehyde glucoside Homocitric acid Malic acid Glucovanillin Class Class Glucoside A &B w*c Figure : Score plot (PLS component and ) (A) and loading plot (B) of PLS-DA results obtained from H-NMR spectra of green pods from accessions V. planifolia. Accession codes:, B;, B;, B;, A0;, A; 6, A9;, A;, B; 9, B; 0, 6B6;, A;, A0;, B;, 9A6. 6

15 Chapter 0 Weight (g) Length (cm) Figure : Correlation between the length and weight of -months-old green pods. R = 0. The result of the CV-ANOVA suggested a good correlation of the length, weight and inflorescence variables with the metabolomic data, giving p-values of.e-,.0e- and.e-, respectively. The PLS loading plot suggests that when the length and weight of a green pod increase, the amount of glucovanillin increases and high flowering capacity is correlated with an increase of malic acid and homocitric acid (Figure 9B). According to this PLS analysis, the accession B is a good compromise between long and heavy pods, high glucovanillin content and high flowering capacity. Since the length is an important factor to assess the grade of cured vanilla pods, this accession shows a good potential for enhancing vanilla quality. Interestingly, the PLS loading plot shows that sucrose content is negatively correlated to the length and glucovanillin content of the pods (Figure 9B). Some accessions like B contain a high content of sucrose and have shorter pods while others like B and B contain a high amount of glucovanillin and have longer pods. This result suggests that the difference within the accessions of V. planifolia could be due to a switch from primary metabolism, e.g. sugar storage, to growth and secondary metabolism in the larger pods.

16 Green pods metabolic characterization A PLS Component (,%) PLS Component (9.%) B 0.0 p-hydroxybenzaldehyde glucoside Weight Length Glucovanillin w*c Glucoside A &B Glucose Malic acid Homocitric acid Inflorescence -0.0 Sucrose w*c Figure 9: Score plot (PLS component and ) (A) and loading plot (B) of PLS results obtained from H-NMR spectra of green pods from accessions V. planifolia and using the length, weight and inflorescence data as Y-variables. Accession codes:, B;, B;, B;, A0;, A; 6, A9;, A;, B; 9, B; 0, 6B6;, A;, A0;, B;, 9A6.

17 Chapter. Conclusions NMR-based metabolic approach was successfully applied to determine the profile of V. planifolia green pods from different accessions. Apparently, glucovanillin contributes to the segregation of the accessions, i.e. the target compound for breeding by chance is also the marker for phenotypic variation. Our results suggest that the accession B could be a good candidate for the selection of a good elite accession. Indeed, this accession shows long pod and high glucovanillin content, flowering and growth capacity. In spite of the limited genetic variation observed (Bory et al., 00c), our data suggest also that difference in activity in primary and secondary metabolism could be an explanation of the phenotypic variation within the accession of V. planifolia. Further experiments will be conducted to perform proteomic analysis and study variation of enzymatic expression between the different accessions, using techniques such as two-dimensional differential gel electrophoresis (Croes et al., 009). Furthermore, possible implication of phytohormones in cellular growth patterns of the different vanilla accessions should not be neglected (Srivastava, 00). 9

18 Green pods metabolic characterization 60