Attenuated Total Reflectance–Fourier Transform Infrared

Sep 11, 2015 - Sage (Salvia officinalis L.) is cultivated worldwide for its aromatic leaves, which are used as herbal spice, and for phytopharmaceutic...
0 downloads 1 Views 1MB Size
Article pubs.acs.org/JAFC

Attenuated Total Reflectance−Fourier Transform Infrared Spectroscopy on Intact Dried Leaves of Sage (Salvia officinalis L.): Accelerated Chemotaxonomic Discrimination and Analysis of Essential Oil Composition Gennadi Gudi,†,‡ Andrea Kraḧ mer,*,† Hans Krüger,§ and Hartwig Schulz†,§ †

Institute for Ecological Chemistry, Plant Analysis and Stored Product Protection, Julius Kühn Institute, Königin-Luise-Straße 19, D-14195 Berlin, Germany ‡ Institute of Pharmacy, Freie Universität Berlin, Königin-Luise-Straße 2-4, D-14195 Berlin, Germany § Institute for Ecological Chemistry, Plant Analysis and Stored Product Protection, Julius Kühn Institute, Erwin-Baur-Straße 27, D-06484 Quedlinburg, Germany S Supporting Information *

ABSTRACT: Sage (Salvia officinalis L.) is cultivated worldwide for its aromatic leaves, which are used as herbal spice, and for phytopharmaceutical applications. Fast analytical strategies for essential oil analysis, performed directly on plant material, would reduce the delay between sampling and analytical results. This would enhance product quality by improving technical control of cultivation. The attenuated total reflectance−Fourier transform infrared spectroscopy (ATR−FTIR) method described here provides a reliable calibration model for quantification of essential oil components [EOCs; R2 = 0.96; root-mean-square error of cross-validation (RMSECV) = 0.249 mL 100 g−1 of dry matter (DM); and range = 1.115−5.280 mL 100 g−1 of DM] and main constituents [e.g., α-thujone/β-thujone; R2 = 0.97/0.86; RMSECV = 0.0581/0.0856 mL 100 g−1 of DM; and range = 0.010− 1.252/0.005−0.893 mL 100 g−1 of DM] directly on dried intact leaves of sage. Except for drying, no further sample preparation is required for ATR−FTIR, and the measurement time of less than 5 min per sample contrasts with the most common alternative of hydrodistillation followed by gas chromatography analysis, which can take several hours per sample. KEYWORDS: infrared spectroscopy, essential oil, thujone, quantification, PLS, nondestructive, ATR−FTIR, Salvia off icinalis L., chemometry



INTRODUCTION Garden sage (Salvia officinalis L.) is a popular aromatic and medicinal plant that is found on all continents, except Antarctica and Australia.1−4 Apart from its use as an aromatic herb, leaves and extracts are widely used in phytomedicine for multiple purposes.5−13 The pharmaceutical uses and aromatic taste of sage are largely derived from the essential oil (EO) that is rich in thujone and camphor with α- and β-pinene, 1,8-cineole, or camphene as minor constituents. Today, a large number of genetically different accessions of S. officinalis L. have been described that are characterized by varying contents of EO and a high diversity of the related chemical composition.4 The use of sage as a medicinal tea or drug in human nutrition and phytomedicine is restricted by the standards of the European Pharmacopoeia. Hence, for both plant breeding and cultivation, the exact content and composition of the EO is needed to ensure that the final products are of consistent and high quality.14 Furthermore, plant secondary metabolites, such as EOs, may vary in content and composition depending on climatic and cultivation conditions, developmental stage of the plant/leaf, origin of plant material, and even daytime of harvest. Therefore, fast, field-based analytical methods provide a means of achieving stringent control over the final product.15−19 Today, the standard method for EO analysis is gas chromatography (GC) coupled with different detection techniques.4,20−25 © 2015 American Chemical Society

Isolation of EOs from the plant matrix as distillate or extract requires a laborious effort prior to GC. Therefore, these traditional analyses based on GC are time-consuming and expensive and also require laboratory infrastructure as well as specially trained personnel. Furthermore, because of labile constituents in EOs, chemical transformations during hot extraction processes cannot be avoided. In recent years, numerous vibrational spectroscopy methods have been developed to allow for investigation of different plant secondary metabolites directly in fresh or dried plant material without prior isolation.17,18,26−29 However, we are only aware of a single application of near-infrared spectroscopy (NIRS) for the prediction of the EO content and composition in sage.30 The comparison of two different NIR spectrometers showed a large variation in the prediction quality between both instruments with correlation factors of at best R2 = 0.83 for EO content and 0.81 for α-thujone, respectively. Because the EO in S. officinalis L. is located in special glandular trichomes on the leaf surface, attenuated total reflectance− Fourier transform infrared spectroscopy (ATR−FTIR) offers a Received: Revised: Accepted: Published: 8743

June 2, 2015 September 11, 2015 September 11, 2015 September 11, 2015 DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry promising alternative to GC and even NIRS analyses.31 With the ATR technique, monolayers of analytes can be investigated through direct contact between the sample and the infrared (IR) crystal. Hence, the surface trichomes of sage are easily accessible to the ATR crystal with a low background from the plant matrix.32 Therefore, the aim of this study was to investigate the use of ATR−FTIR for the direct qualitative and quantitative analysis of the EO composition of dried intact leaves of sage, including the effect of different varieties, leaf age, and date of harvest. On the basis of reference analysis of solvent extracts from the plant material by gas chromatography coupled with mass spectrometry or flame ionization detection (GC−MS or GC−FID, respectively), appropriate calibration models for the EO content

and composition can be developed that include the chemical diversity of different sage accessions.



Table 1. Denotation and Botanical Description of the Considered S. officinalis L. Accessions denotation SALV13 SALV15 SALV25 SALV28 SALV29 SALV30 SALV31 SALV34 SALV36 SALV63 SALV65 SALV67 BG

accession name

geographical origin D: Botanical Garden University of Budapest D: Botanical Garden University of Mainz D: Botanical Garden University of Athens D: VEB Arzneimittel Dresden, Forschungsstelle Artern D: Botanical Garden Cluji-Napoca (Dr. T. A. Szabó) D: Botanical Garden Cluji-Napoca (Dr. T. A. Szabó) E: Landraces CSSR 1981 (Western Slovakia): 67 D: Institute of Applied Botany, Pyongyang University D: Redwood City Seed Co.: 6 D: Federal Centre for Breeding Research on Cultivated Plants, Braunschweig, Genetic Resources Centre: 61443 D: Federal Centre for Breeding Research on Cultivated Plants, Braunschweig Genetic Resources Centre: 61862 D: Federal Centre for Breeding Research on Cultivated Plants, Braunschweig Genetic Resources Centre: 55892 Bingenheimer Saatgut GmbH, Echzell, Germany

MATERIALS AND METHODS

Chemicals. Pure standard substances (α- and β-thujone, α- and β-pinene, 1,8-cineole, camphor, and carvacrol) were purchased as analytical standards (at least >96%) for GC reference analysis from Roth (Karlsruhe, Germany) and Sigma-Aldrich (Taufkirchen, Germany). Isooctane [>99%, high-performance liquid chromatography (HPLC) grade] for solvent extraction of essential oil components (EOCs) was obtained from Sigma-Aldrich. Plant Material. Seeds of 12 different sage (S. officinalis L.) accessions maintained in a gene bank were obtained from the Leibniz Institute of Plant Genetics and Crop Plant Research in Gatersleben (Germany) and cultivated at the experimental field of the Julius Kühn Institute in Berlin (Germany). For further botanical description, see Table 1. Five fully developed young (leaf size: maximum length, 30 mm; maximum width, 10 mm) and old (leaf size: minimum length, 50 mm; minimum width, 15 mm) leaves of 10 individual plants per accession were collected at two different harvests in 2013 (July 25, H1; September 25, H2). In the second harvest, withering meant that less than 10 plants were available for some accessions (SALV13, 7; SALV25, 9; SALV28, 9; SALV29, 6; SALV30, 8; and SALV34, 5). For external validation, 3 times 10 plants grown from seeds from a commercial sage accession (article number k58, S. officinalis, Bingenheimer Saatgut GmbH, Echzell, Germany) were cultivated and sampled under identical conditions as described for gene bank material. At both harvests, old and young leaves were sampled (H1, 30 plants; H2, 29 plants), and the spectra were collected by ATR−FTIR prior to extraction and GC analysis. All leaves were dried at 35 °C for 48 h prior to ATR−FTIR and GC analyses and were stored in paper bags at room temperature in desiccators. ATR−FTIR. The mid-infrared (MIR) spectra of intact dried sage leaves and pure standards were recorded from 4000 to 375 cm−1 with a portable ATR diamond crystal infrared spectrometer (Alpha, Bruker Optics GmbH, Ettlingen, Germany). Approximately 5−10 mg of the standards was placed on the surface of the diamond crystal and measured directly. Intact dried sage leaves were analyzed by gently pressing the adaxial leaf side in the region of the central lamina onto the ATR crystal. Use of the standard pressing mechanism of the instrument allowed for a constant pressure for all measurements. All spectra were composed of 32 scans and acquired with a spectral resolution of 4 cm−1. For each plant per accession and harvest, five old and five young leaves were measured once and the five repeated spectra per experiment were averaged and used for later chemometric analysis.

Extrakta

Garden Sage

Salbei Echter Salbei K58

Table 2. Statistical Results for GC Analysis of All 448 Investigated Sage Plants Averaged over All 12 Considered S. officinalis L. Accessions Including Both Harvests and Both Developmental Stages of Leaves

a

component

meanb

SDa

maximumb

minimumb

Q1b

Q3b

meanc

EOC content α-thujone β-thujone camphor 1,8-cineole α-pinene β-pinene camphene myrcene limonene borneol α-caryophyllene β-caryophyllene manool not identifiedd

3.109 0.610 0.236 0.511 0.254 0.094 0.048 0.121 0.024 0.037 0.039 0.251 0.129 0.283 0.156

1.215 0.383 0.255 0.268 0.151 0.089 0.042 0.086 0.013 0.018 0.034 0.169 0.120 0.166 0.106

7.330 2.136 1.311 1.782 0.946 0.634 0.355 0.546 0.079 0.109 0.185 0.943 0.911 1.037 0.633

0.575 0.008 0.000 0.005 0.017 0.008 0.007 0.014 0.003 0.006 0.000 0.025 0.008 0.041 0.000

2.188 0.328 0.050 0.325 0.140 0.039 0.020 0.066 0.016 0.025 0.015 0.123 0.051 0.162 0.078

3.961 0.849 0.343 0.644 0.337 0.111 0.062 0.145 0.029 0.043 0.051 0.338 0.161 0.372 0.213

19.6 7.6 16.4 8.2 3.0 1.5 3.9 0.8 1.3 1.3 8.1 4.2 9.1 5.2

SD, standard deviation; Q1 and Q3, quantiles 1 and 3, respectively. bIn units of mL 100 g−1 of DM. cIn units of %. dRetention time = 30.94 min. 8744

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry

Figure 1. ATR−FTIR spectra of old and young leaves of S. officinalis L. for both harvest dates (July 25, 2013/H1 and September 25, 2013/H2) as averaged spectra of all considered gene bank accessions. The highlighted ranges indicate appropriate signals resulting from EO terpenoids as shown by spectra of the appropriate standards α- and β-thujone and camphor, respectively.

Figure 3. (A) PCA scores plot constructed from ATR−FTIR spectra of dried intact leaves of S. officinalis L. for harvest 1 (each color indicates one of the 12 accessions) and (B) GC reference data for the EOC content (green) and amounts of α-thujone (red) and β-thujone (blue) for the 12 different accessions. flow rate of 1 mL min−1. (split of 1:20). Analyses of the sage extracts were performed using an Agilent MSD 5972/HP 5890 Series II, equipped with a 30 m × 0.25 mm i.d., see above, 0.5 μm HP-5MS column. The column temperature was increased from 60 to 120 °C at a rate of 3° min−1, finally increased to 220 °C at 8° min−1, and held for 10 min. Helium was used as carrier gas with a constant flow rate of 1.1 mL min−1 and a split of 1:10. Injector and detector temperatures were set at 250 and 280 °C, respectively. The ionization energy was set at 70 eV. Identification of the detected compounds was based on their relative retention time and their mass spectra in comparison to those observed for the above-mentioned pure standard substances. The other compounds were tentatively identified using the NBS75K and Wiley 138 spectral library databases of the GC−MS system. The percentage composition of the volatile fraction was computed from the GC peak areas using peak area normalization. Chemometrics. To estimate the variability of the samples, principal component analysis (PCA) was used based on the instrument software package OPUS 7.2 (Quant 2 module, Bruker Optik GmbH, Germany) and Unscrambler X (Camo, Norway). Relevant spectral ranges were determined by comparison of sample spectra to those of appropriate reference standards and PCA. Calibration models were validated by 10-fold cross-validation (“leave 10% out”) using a partial least-squares (PLS) algorithm. Appropriate models designated by cross-validation were further tested by test set validation with an external sample set. Therefore, spectra and GC data of all 30 plants of the commercial accession were averaged, resulting in 12 test samples related to the four treatments (old/young and harvest July 25, 2013/harvest September 25, 2013).

Figure 2. (A) PCA scores plot constructed from ATR−FTIR spectra of old (blue) and young (red) leaves of S. officinalis L., (B) corresponding loadings for PC1 and PC2, and (C) averaged EOC content for both harvests and all accessions as determined by GC−FID. Reference GC−FID and GC−MS Analyses. For reference analysis, 200 mg of air-dried sage leaves were homogenized in 1.0 mL of isooctane [containing 1:2000 (v/v) carvacrol as the internal standard] with a Retsch ball mill MM2000 (Retsch GmbH, Germany). After centrifugation (10 min at 2415g), extracts were analyzed by GC−FID using an Agilent gas chromatograph 6890N, fitted with a HP-5 column (50 m × 0.32 mm i.d., with a film thickness of 0.52 μm). This GC instrument is routinely used for EO analysis and regularly validated with external standards. Hence, no repeated extraction was performed, and each sample was investigated once. Detector and injector temperatures were set at 280 and 250 °C, respectively. The following oven temperature program was used: heating from 80 to 220 °C at a rate of 8° min−1. This final temperature was held for 10 min. Hydrogen was used as carrier gas with a constant



RESULTS AND DISCUSSION Reference Analysis for S. officinalis L. According to the European Pharmacopoeia, EO is analyzed by hydrodistillation and GC methods.14 Because of the comprehensive sample set and the small sample amounts (five individual leaves per sample) investigated in the present study, solvent extraction 8745

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry

Chemotaxonomic Variability of S. officinalis L. Sage is known to show a large variety in the EO content and composition.3,4,19 Therefore, the impact of genetic and geographic origin (12 gene bank accessions), age, and date of harvest on the spectroscopic and chemical variability of the EO composition was investigated, resulting in 448 individual samples (for lost plants, see the Plant Material section). Figure 1 presents the ATR−FTIR spectra of old and young leaves at both harvests (H1 and H2) as means of all gene bank accessions and the spectra for α- and β-thujone and camphor as main constituents of sage EO. The spectral ranges highlighted in Figure 1 indicate distinct EO terpenoid-derived signals in the spectra of the plant material compared to spectra of pure standards. Both camphor and α- and β-thujone contain keto groups substituted on a cyclopentane ring system, resulting in a characteristic stretch vibration for the CO bond at around 1740 cm−1. Furthermore, both terpenoids contain dimethyl-substituted methylene groups with a characteristic two-band deformation vibration pattern around 1390 and 1370 cm−1. Besides matrix signals from cellulose around 1100 cm−1, contributions of these diagnostic IR bands are also found in the spectra of sage leaves. Hence, evaluation of this fingerprint region offers a promising tool for quantification of those metabolites and differentiating sage samples based on their camphor and/or α- and β-thujone contents. Averaging of the spectra from the adaxial leaf surface of plants followed by PCA resulted in clustering of old and young leaves separated by principal component 1 (PC1), as shown in Figure 2a. The differentiation among the clusters is caused by

instead of hydrodistillation was performed to describe “essential oil components (EOCs)” following previously published research on various EO plants.17,21,30,33 Even if solvent extracts instead of hydrodistillates were investigated by GC, the term EOCs has been retained because solvent extraction and hydrodistillation of sage leaves both lead to isolates of comparable chemical composition.34 Table 2 presents the summarized results of GC analysis of EOC composition for all accessions and all treatments. The EOC extract was composed of about 85−90% terpenoids, which are known constituents of sage EO. As indicated by the statistical characteristics, there was a high degree of variability in the content of individual compounds in the accessions that were investigated.

Figure 4. GC reference data for the EOC content (green) and amounts of α-thujone (red) and β-thujone (blue) for the 10 plant repetitions and both harvests of accession SALV36.

Table 3. Data Preprocessing and Spectral Ranges Used for the Developed Calibration Models of the Individual EOCs n

component

448

48

EOC content α-thujone β-thujone camphor 1,8-cineole α-pinene β-pinene EOC content α-thujone β-thujone camphor 1,8-cineole α-pinene β-pinene

spectral ranges (cm−1)

data preprocessing a

min−max 1. derivative SNVc SNVc 2. derivative 1. derivative 1. derivative 1. derivative 1. derivative 2. derivative 1. derivative min−maxa SNVc 1. derivative

3037−2952; 3037−2952; 1472−1305; 1705−1661; 1781−1709; 3500−3175; 3500−3175; 3037−2952; 2938−2801; 3037−2952; 3037−2952; 1705−1661; 1705−1661; 2938−2801;

+ MSCb

+ MSCb + SNVc + MSCb

+ SLSd

1781−1709; 2938−2771; 1295−1133; 1472−1305 1705−1661; 3037−2952; 3037−2952; 2938−2801; 1705−1661; 1781−1709; 1781−1709; 1472−1305; 1472−1305; 1546−1489;

1705−1661; 1295−1133 1781−1709; 1119−905 1119−905 1546−1489; 2938−2771; 2938−2771; 1781−1709; 1295−1133; 1648−1552; 1472−1305; 927−639 1295−1133; 1472−1305;

1295−1133 1705−1661; 1705−1661; 1705−1661; 1117−954 1546−1489; 1295−1133;

1472−1305; 1295−1133; 1119−905 1472−1305; 1295−1133; 1119−905 1472−1305; 1295−1133; 1117−954 1295−1133; 1117−954 1117−954

927−639 1295−1133

a

min−max = minimum−maximum normalization. bMSC = multiple scatter correction. cSNV = standard normal variate. dSLS = straight line subtraction.

Table 4. Statistical Parameters for 10-Fold Cross-Validation of ATR−FTIR and GC Data by Correlation of Averaged Spectra for Each Plant (n = 448)

a

component

R2

RMSECVa

biasa

rangea/meana

LVb

EOC content α-thujone β-thujone camphor 1,8-cineole α-pinene β-pinene

0.67 0.75 0.75 0.17 0.76 0.55 0.58

0.743 0.19 0.125 0.275 0.0824 0.0594 0.0321

0.00135 −0.00109 −0.00164 0.00878 −0.00022 0.00306 0.00291

0.575−6.744/3.058 0.008−1.943/0.541 0.000−1.311/0.218 0.005−1.782/0.557 0.018−0.946/0.252 0.008−0.464/0.091 0.007−0.355/0.05

6 10 10 7 6 8 8

In units of mL 100 g−1 of DM. bLV = number of latent variables. 8746

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry the varying amounts of EOCs with different contents of thujone and camphor, as shown by the characteristic IR signals appearing in the corresponding loadings of PC1 given in Figure 2b. Young leaves generally contain up to 2-fold more EOCs compared to old leaves, as observed for all accessions and both harvests by GC reference analysis (Figure 2c). The strong differences in the amount of EOCs are already clear in the appropriate spectra of old and young leaves presented in Figure 1. Furthermore, PC2 separates samples from harvests 1 and 2 based on signals belonging mostly to the leaf matrix (Figure 2b; PC2 around 1000 cm−1). A similar approach for discrimination of different citrus oils by PCA based on characteristic key Raman bands was described in a previous study.35 In contrast, spectroscopic differentiation according to the origin/accession of the material was not possible, as shown in Figure 3a, nor by GC reference data (see Figure S1 of the Supporting Information). Subsequent analysis of the EOCs of each accession revealed a large range in content and relative proportions of α- and β-thujone, as presented in Figure 3b. These results show a group of accessions with either low overall thujone content (SALV29 and SALV30) or only a low ratio of α-/β-thujone (SALV13 and SALV36). In contrast, the majority of accessions is characterized by a high content of α-thujone (SALV15, SALV25, SALV28, SALV31, SALV34, SALV63, SALV65, SALV67, and BG). However, the different accessions could not be discriminated possibly because of the large degree of scattering in the data. This might be explained by the genetic variability and heterogeneity of the plant material within the accessions. Generally, gene bank accessions are defined as sample collections taken from nature, and are characterized by the location and time of sampling. Therefore, a certain genetic diversity even within the accessions has to be considered. As indicated by the GC data for the 10 individual plants of accession SALV36 shown in Figure 4, large variations in EOCs, overall amount of thujone, and ratio of α-/β-thujone can be observed for both harvests. Such a high variability in not only the yield of EOC or sum of thujone but also the ratio of both stereoisomers results in insufficient clustering. Hence, because of the clear genetic variation discrimination into individual accessions based on ATR−FTIR spectra of intact dry leaves can only be achieved for those few accessions with an unusual chemical composition containing low amounts of α-thujone and/or high concentrations of β-thujone (Figure 3a). Nevertheless, information about chemical heterogeneity of the accessions is highly relevant for the development of suitable calibration models with ATR−FTIR spectroscopy (where only a small amount of material is used with each measurement) because it determines an appropriate sampling (e.g., number of individual plants sampled and measurement replicates per leaf) and statistical evaluation. Therefore, a representative number of plants (repetitions per accession) has to be incorporated, and a greater degree of averaging ATR−FTIR spectra (summarizing variability by reduction of data points) has to be applied. Quantitative Analysis of EOCs by ATR−FTIR. On the basis of the results above, quantification models for the sum of EOCs and individual compounds were validated by 10-fold cross-validation. As a first step, averaged spectra for each age and harvest (resulting in 448 samples) were correlated with GC reference data for the EOC yield as well as α-thujone, β-thujone, camphor, 1,8-cineole, α-pinene, and β-pinene. In Table 3, data preprocessing and the appropriate spectral ranges

Figure 5. Results of 10-fold cross-validation of ATR−FTIR and GC data for the (A) EOC content, (B) camphor, (C) α-thujone, (D) βthujone, (E) α-pinene, (F) β-pinene, and (G) 1,8-cineole by correlation of averaged spectra for each accession (n = 48). Blue color, old leaves; red color, young leaves; circle, harvest 1 (July 25, 2013); and triangle, harvest 2 (September 25, 2013).

are summarized, whereas Table 4 provides an overview of the statistical parameters of the corresponding models. For all components, insufficiently good predictions were obtained, as indicated by low values for the coefficient of determination R2 and relatively high values for the root-meansquare error of cross-validation (RMSECV) compared to the individual concentration ranges. The low correlation between ATR−FTIR and GC analyses is likely caused by the different amounts of sample used in the two methods. Spectroscopic data were composed of five measurement points of approximately 1 mm2 in size (size of ATR diamond crystal) distributed 8747

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry

Table 5. Statistical Parameters for Test Set Validation of the Developed Prediction Models of Plant-wise Averaged Spectra (n = 48) with 12 Independent Samples from the Commercial Accession

a

component

R2

RMSEPa

biasa

rangea/meana

LVb

EOC content α-thujone β-thujone camphor 1,8-cineole α-pinene β-pinene

0.98 0.74 0.80 0.74 0.87 0.55 0.84

0.174 0.124 0.0778 0.09 0.0482 0.0471 0.0111

0.0371 0.061 0.0375 −0.0768 0.0241 0.0184 0.00267

2.223−5.514/3.869 0.547−1.268/0.908 0.085−0.593/0.339 0.253−0.895/0.574 0.142−0.598/0.37 0.056−0.312/0.184 0.02−0.103/0.0615

7 8 5 7 6 7 8

In units of mL 100 g−1 of DM. bLV = number of latent variables.

β-thujone and camphor), and even minor components (α- and β-pinene and 1,8-cineole). An initial screening by ATR−FTIR enabled a rapid assessment of the chemical variability of the samples, which allowed us to adjust sampling protocols (e.g., more measurement repetitions by ATR−FTIR for better representation of sample variability). This, in turn, permitted an improvement of quantification models by different statistical pretreatments (e.g., higher degree of averaging) with only minor additional time required. In the current ATR−FTIR quantification protocol, plant material only needs to be dried and spectral measurements take only a few minutes. This allows for rapid quality control of the material when it is first received for processing compared to the time-consuming analysis recommended in the Pharmacopoeia. Transfer of the current analytical protocol to fresh material and application in the field is currently under investigation. This will undoubtedly improve growing and quality control for both plant breeding and cultivation and assist in evaluating the chemical variation of breeding populations and identifying optimal points for harvest.

over five individual leaves per plant. In contrast, for GC, the entire five leaves were extracted (see the Reference GC−FID and GC−MS Analyses section), and hence, variability within one plant might not be comparable. For a better match between the amount of sample used for spectroscopy and GC, the number of ATR−FTIR measurement repetitions per plant should be increased. This will enlarge the amount of leaf area investigated by ATR−FTIR but will lead to longer measurement times. Another option could be averaging more comparable spectra, e.g., all of those belonging to the same accession and treatment (old/young and H1/H2), to make them more representative of a larger sample, e.g., a genuine batch of drug or the yield from an entire parcel of land. Figure 5 presents the graphical results for cross-validation regarding averaged ATR−FTIR spectra from all plants belonging to the same accession, age, and time of harvest, resulting in 48 individual samples. Data preprocessing and relevant spectral ranges are summarized in Table 3. The use of spectra averaged across accessions leads to an improved prediction for not only the EOC yield and major constituents (α-thujone, β-thujone, and camphor) but also components with lower concentrations (in per mill range), such as α- or β-pinene and 1,8-cineole. For all EOCs, coefficients of determination higher than 0.86 were obtained with low values for bias and RMSECV of less than 10% of the corresponding mean values. The robustness of the cross-validated models was investigated with an additional test set validation of 12 independent samples (3 sets of 10 plants, old/young and H1/H2) which came from the commercial sage accession. As shown by the values of R2, root-mean-square error of prediction (RMSEP), and bias given in Table 5, for EOCs, a robust and reliable prediction model could be achieved with R2 = 0.98. For the individual compounds, prediction quality was at best for 1,8cineole (R2 = 0.87), with R2 = 0.74 and 0.80 for α- and β-thujone, respectively. The relatively low prediction accuracy for individual metabolites could again be explained by the genetic and chemotaxonomic diversity of sage. The accession used for test set validation was not included in the calibration model. The impacts of the plant matrix or variation in the ratio of α- and β-thujone were not considered in the calibration. Hence, application of calibration models to unknown samples of different origin might be limited, and the genetic and chemotaxonomic diversity must be taken into account. The RMSEP values were calculated according to the definition of Fearn as SEP2.36 In conclusion, all cross-validated calibration models verified against an external test set resulted in reliable prediction models for the overall yield of EOCs, their major constituents (α- and



ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jafc.5b03852. Results of GC−FID data for the leaf material from all 12 accessions of S. officinalis L. (Figure S1) and graphical results of 10-fold cross-validation of ATR−FTIR for the individual EOCs for single-plant investigation (n = 448; Figure S2) (PDF)



AUTHOR INFORMATION

Corresponding Author

*Telephone: +49-30-8304-2210. Fax: +49-30-8304-2503. E-mail: [email protected]. Funding

The authors gratefully acknowledge the financial support obtained from the Deutsche Forschungsgemeinschaft (DFG) in Bonn, Germany (Project Schu 566/14-1). Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Bärbel Zeiger and Mario Harke for their support in sample preparation and spectroscopic and GC analyses as well as Dr. Andreas Börner and Dr. Ulrike Lohwasser from the Leibniz Institute of Plant Genetics and Crop Plant Research in Gatersleben (Germany) for providing the seed material. Furthermore, the authors thank Dr. William 8748

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry

(12) Bauer, J.; Kuehnl, S.; Rollinger, J. M.; Scherer, O.; Northoff, H.; Stuppner, H.; Werz, O.; Koeberle, A. Carnosol and Carnosic Acids from Salvia of f icinalis Inhibit Microsomal Prostaglandin E-2 Synthase1. J. Pharmacol. Exp. Ther. 2012, 342, 169−176. (13) Miroddi, M.; Navarra, M.; Quattropani, M. C.; Calapai, F.; Gangemi, S.; Calapai, G. Systematic Review of Clinical Trials Assessing Pharmacological Properties of Salvia Species on Memory, Cognitive Impairment and Alzheimer’s Disease. CNS Neurosci. Ther. 2014, 20, 485−495. (14) Salbeiblätter-Salvia officinalis folium. European Pharmacopoeia 8.1, Grundwerk 2014; Deutscher Apotheker Verlag: Stuttgart, Germany, 2014; pp 2050−2051. (15) Generalic, I.; Skroza, D.; Surjak, J.; Mozina, S. S.; Ljubenkov, I.; Katalinic, A.; Simat, V.; Katalinic, V. Seasonal Variations of Phenolic Compounds and Biological Properties in Sage (Salvia of f icinalis L.). Chem. Biodiversity 2012, 9, 441−457. (16) Gudi, G.; Schütze, W.; Schulz, H. Rapid analysis of rosmarinic acid in lemon balm (Melissa of f icinalis L.). Asian Chem. Lett. 2011, 15, 1−10. (17) Gudi, G.; Krähmer, A.; Krüger, H.; Hennig, L.; Schulz, H. Discrimination of fennel chemotypes applying IR and Raman spectroscopy - discovery of a new γ-asarone chemotype. J. Agric. Food Chem. 2014, 62, 3537−3547. (18) Krähmer, A.; Gudi, G.; Weiher, N.; Gierus, M.; Schütze, W.; Schulz, H. Characterization and quantification of secondary metabolite profiles in leaves of red and white clover species by NIR and ATR-IR spectroscopy. Vib. Spectrosc. 2013, 68, 96−103. (19) Lakusic, B. S.; Ristic, M. S.; Slavkovska, V. N.; Stojanovic, D. L.; Lakusic, D. V. Variations in essential oil yields and compositions of Salvia of f icinalis (Lamiaceae) at different developmental stages. Bot. Serb. 2013, 37, 127−139. (20) Krüger, H. Essential oilsVariability in medicinal and spice plants. Julius Kühn Arch. 2009, 70−76. (21) Krüger, H. Characterisation of Chamomile Volatiles by Simultaneous Distillation Solid-Phase Extraction in Comparison to Hydrodistillation and Simultaneous Distillation Extraction. Planta Med. 2010, 76, 843−846. (22) Miraldi, E. Comparison of the essential oils from ten Foeniculum vulgare Miller samples of fruits of different origin. Flavour Fragrance J. 1999, 14, 379−382. (23) Pank, F.; Krüger, H.; Quilitzsch, R. Selection of essential oil rich populations of bitter fennel (Foeniculum vulgare MILL. ssp vulgare var. vulgare) with reduced estragole content. Z. Arznei- Gewuerzpflanzen 2008, 13, 161−166. (24) Pfefferkorn, A.; Krüger, H.; Pank, F. Chemical composition of Satureja hortensis L essential oils depending on ontogenetic stage and season. J. Essent. Oil Res. 2008, 20, 303−305. (25) Thompson, J.; Chalchat, J. C.; Michet, A.; Linhart, Y.; Ehlers, B. Qualitative and Quantitative Variation in Monoterpene Co-Occurrence and Composition in the Essential Oil of Thymus vulgaris Chemotypes. J. Chem. Ecol. 2003, 29, 859−880. (26) Fiedler, A.; Baranska, M.; Schulz, H. FT-Raman spectroscopy - a rapid and reliable quantification protocol for the determination of natural indigo dye in Polygonum tinctorium. J. Raman Spectrosc. 2011, 42, 551−557. (27) Krähmer, A.; Engel, A.; Kadow, D.; Ali, N.; Umaharan, P.; Kroh, L. W.; Schulz, H. Fast and neat - Determination of biochemical quality parameters in cocoa using near infrared spectroscopy. Food Chem. 2015, 181, 152−159. (28) Naumann, A.; Kurtze, L.; Krähmer, A.; Hagels, H.; Schulz, H. Discrimination of Solanaceae Taxa and Quantification of Scopolamine and Hyoscyamine by ATR-FTIR Spectroscopy. Planta Med. 2014, 80, 1315−1320. (29) Schulz, H.; Baranska, M. Identification and quantification of valuable plant substances by IR and Raman spectroscopy. Vib. Spectrosc. 2007, 43, 13−25. (30) Elementi, S.; D’Antuono, L. F.; Schulz, H.; Krüger, H.; Schütze, W.; Steuer, B. Salvia of f icinalis L. essential oil and carnosic acid analysis by means of NIR spectroscopy. Acta Hortic. 2006, 243−248.

Foley from the School of Botany and Zoology, Australian National University, Canberra (Australia), for his valuable comments on the manuscript.



ABBREVIATIONS USED ATR−FTIR, attenuated total reflectance−Fourier transform infrared spectroscopy; GC, gas chromatography; PLS, partial least squares; PCA, principal component analysis; RMSECV, root-mean-square error of cross-validation; RMSEP, rootmean-square error of prediction; SEP, standard error of prediction; SNV, standard normal variate; MSC, multiplicative scattering correction; min−max, minimum−maximum normalization; SLS, straight line subtraction; LV, latent variable; EOC, essential oil component; DM, dry matter; PC, principal component; GC−MS, gas chromatography−mass spectrometry; GC−FID, gas chromatography−flame ionization detection; NIRS, near-infrared spectroscopy; HPLC, high-performance liquid chromatography; SALV, accession name for gene bank material of S. officinalis L.



REFERENCES

(1) Al-Qudah, M. A.; Al-Jaber, H. I.; Abu Zarga, M. H.; Abu Orabi, S. T. Flavonoid and phenolic compounds from Salvia palaestina L. growing wild in Jordan and their antioxidant activities. Phytochemistry 2014, 99, 115−120. (2) Ben Farhat, M.; Jordan, M. J.; Chaouech-Hamada, R.; Landoulsi, A.; Sotomayor, J. A. Variations in Essential Oil, Phenolic Compounds, and Antioxidant Activity of Tunisian Cultivated Salvia of f icinalis L. J. Agric. Food Chem. 2009, 57, 10349−10356. (3) Grausgruber-Gröger, S.; Schmiderer, C.; Steinborn, R.; Novak, J. Seasonal influence on gene expression of monoterpene synthases in Salvia of ficinalis (Lamiaceae). J. Plant Physiol. 2012, 169, 353−359. (4) Lamien-Meda, A.; Schmiderer, C.; Lohwasser, U.; Börner, A.; Franz, C.; Novak, J. Variability of the essential oil composition in the sage collection of the Genebank Gatersleben: a new viridiflorol chemotype. Flavour Fragrance J. 2010, 25, 75−82. (5) Martins, N.; Barros, L.; Santos-Buelga, C.; Henriques, M.; Silva, S.; Ferreira, I. C. Evaluation of bioactive properties and phenolic compounds in different extracts prepared from Salvia of f icinalis L. Food Chem. 2015, 170, 378−385. (6) Benkherara, S.; Bordjiba, O.; Djahra, A. In vitro evaluation of the antibacterial activity of essential oil of Salvia of f icinalis. Phytotherapie 2015, 13, 14−18. (7) Tosun, A.; Khan, S.; Kim, Y. S.; Calin-Sanchez, A.; Hysenaj, X.; Carbonell-Barrachina, A. Essential Oil Composition and AntiInflammatory Activity of Salvia of f icinalis L (Lamiaceae) in Murin Macrophages. Trop. J. Pharm. Res. 2014, 13, 937−942. (8) Kozics, K.; Klusova, V.; Srancikova, A.; Mucaji, P.; Slamenova, D.; Hunakova, L.; Kusznierewicz, B.; Horvathova, E. Effects of Salvia of f icinalis and Thymus vulgaris on oxidant-induced DNA damage and antioxidant status in HepG2 cells. Food Chem. 2013, 141, 2198−2206. (9) Moreira, M. R.; Souza, A. B.; Moreira, M. A.; Bianchi, T. C.; Carneiro, L. J.; Estrela, F. T.; dos Santos, R. A.; Januario, A. H.; Martins, C. H.; Ambrosio, S. R.; Veneziani, R. C. RP-HPLC analysis of manool-rich Salvia of f icinalis extract and its antimicrobial activity against bacteria associated with dental caries. Rev. Bras. Farmacogn. 2013, 23, 870−876. (10) Russo, A.; Formisano, C.; Rigano, D.; Senatore, F.; Delfine, S.; Cardile, V.; Rosselli, S.; Bruno, M. Chemical composition and anticancer activity of essential oils of Mediterranean sage (Salvia of f icinalis L.) grown in different environmental conditions. Food Chem. Toxicol. 2013, 55, 42−47. (11) Ehrnhöfer-Ressler, M. M.; Fricke, K.; Pignitter, M.; Walker, J. M.; Walker, J.; Rychlik, M.; Somoza, V. Identification of 1,8-Cineole, Borneol, Camphor, and Thujone as Anti-inflammatory Compounds in a Salvia of f icinalis L. Infusion Using Human Gingival Fibroblasts. J. Agric. Food Chem. 2013, 61, 3451−3459. 8749

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750

Article

Journal of Agricultural and Food Chemistry (31) Charchari, S.; Chafaa, I.; Kassoussi, K.; Zekri, M. M.; Benhalla, A.; Boudina, N. Glandular Trichomes, Secretory Cavities and Essential Oil of Sage (Salvia of ficinalis L.). J. Essent. Oil-Bear. Plants 2010, 13, 267−274. (32) Hind, A. R.; Bhargava, S. K.; McKinnon, A. At the solid/liquid interface: FTIR/ATR - the tool of choice. Adv. Colloid Interface Sci. 2001, 93, 91−114. (33) Schulz, H.; Quilitzsch, R.; Drews, H.; Krüger, H. Estimation of minor components in caraway, fennel and carrots by NIRScomparison of results from dispersive and Fourier-transform instruments. Int. Agrophys. 2000, 14, 249−253. (34) Grassi, P.; Novak, J.; Steinlesberger, H.; Franz, C. A direct liquid, non-equilibrium solid-phase micro-extraction application for analysing chemical variation of single peltate trichomes on leaves of Salvia of ficinalis. Phytochem. Anal. 2004, 15, 198−203. (35) Schulz, H.; Schrader, B.; Quilitzsch, R.; Steuer, B. Quantitative analysis of various citrus oils by ATR/FT-IR and NIR-FT Raman Spectroscopy. Appl. Spectrosc. 2002, 56, 117−124. (36) Fearn, T. Assessing Calibrations: SEP, RPD, RER and R2. NIR News 2002, 13, 12−14.

8750

DOI: 10.1021/acs.jafc.5b03852 J. Agric. Food Chem. 2015, 63, 8743−8750