Introduction

As an important and special cash crop, the value of tobacco is estimated by smoking through cigarette. Commercial cigarettes are blends of different tobacco types or classes and then cased with flavorants and other additives [1]. So the aroma type and quality of the final products are greatly influenced by the quality of the tobacco raw materials, and therefore, evaluation of different tobaccos is an important work for fabricating blends in tobacco industry. Nowadays, tobacco leaves in China were classified into 42 levels according to stalk positions (upper leaves B, the middle-part leaves C and the lower leaves L) and colors (lemon L, orange F and red R) [2]. This system of evaluating quality characteristics of the material is more suitable for agriculture rather than manufacturing, considering the fact that the approach could hardly estimate the already existing chemicals. The manufacturing focuses on the chemical substances that contribute to grades and quality of tobacco, especially the volatile aromatic compounds [3], rather than the external characteristics. Actually, the quality and fragrance style of tobaccos were determined by the content of routine chemical constituents, including the total sugar, reducing sugar, nicotine, the total nitrogen and the neutral volatiles [4–6].

Previous studies indicated that neutral volatiles were positively correlated with tobacco flavor and some specific volatile compounds could be used as references for tobacco quality [7]. However, these profiles are so complex that multivariate statistics is needed to obtain a definitive differentiation among different tobaccos. PCA was chosen to analyze multivariate data, since it is successful in investigating relationships among larger numbers of variables. PCA, with the aim to reduce variables dimensions, concentrate information, simplify index structure and make index system of analysis more simple and effective, is a mathematical procedure that is used to separate interrelationships into statistically independent basic components [8–10]. PCA has been widely applied in many fields including food, feed and agricultural industries [11–14]. In these cases, the objectives of reducing the number of variables in a data set by finding linear combination of variables that explain most of the variability or distinguishing samples with similar characteristics were all well fulfilled.

PCA has also been applied in tobacco industry, mainly for the purpose of distinguishing the tobacco products based on carbohydrates, smoke constituents and nonvolatile organic acids [15–17]. So far, the technique has not yet been expanded to include assessment of tobacco leaf quality based on neutral volatiles. Therefore, the objective of this work was to analyze the basic chemicals and neutral volatiles and to get a good method for discriminating the tobacco leaves in different stalk positions by using multivariate statistical analysis. This study will provide guidance for evaluating the quality of the tobacco leaves in different stalk positions, in order to make full use of raw materials while fabricating blended cigarettes formula and casing flavorants in tobacco industry.

Materials and methods

Materials and reagents

Thirty flue-cured tobacco samples sourced from 10 different origins (Table 1) and 3 different positions (letters B, C and X represent the upper, middle and lower locations of tobacco leaves on the plant) were gifted by Anhui Cigarette Manufactory of China. Eleven authentic standard compounds were used in this study: 2-methyltetrahydrofuran-3-one, 2-furaldehyde, 2-acetyl furan, benzaldehyde, 5-methylfurfural, isophorone, furfuralcohol, methyl cyclopentenolone, benzylalcohol, phenylethylalcohol and 5-hydroxymethylfurfural. And 2-phenylethyl acetate were used as the internal standard for it was not found in tobacco volatiles. All the reagents mentioned above were purchased from J&K Scientific Ltd (NJ, USA) and in analytical purity. The solvent used in the experiment was dichloromethane (in GC grade purity) and was supplied by Tedia Company (Fairfield, USA).

Table 1 Varieties of tobacco used in experiment
Full size table

Basic chemical analysis

The content of total sugars, nicotine and total nitrogen was determined using continuous flow analysis on Auto Analyzer 3 (SEAL Analytical GmbH, Germany) according to the methods of YC/T 159 ~ 161—2002.

Extraction of the neutral volatiles

All samples were placed in oven at 40 °C for 2 h to remove moisture and then milled to a 40 mesh powder. The neutral volatile components in tobacco were extracted with simultaneous distillation extraction (SDE). Ten gram of tobacco powder and 350 ml redistilled water were placed in a 1,000-ml flask; 40 ml dichloromethane in a 100-ml flask was kept in a 60 °C water bath, and both were distilled for 2.5 h [18]; 50 μl phenylethyl acetate (with the concentration of 1.162 mg/ml) was added as internal standard to the extract. The extract was collected and concentrated to 1 ml with a rotary evaporator at 40 °C and then filtered with 0.22-μm nylon membranes before GC/MS analysis.

GC–MS analysis of neutral volatiles

Separations were performed on a HP-inowax column (30 m × 0.25 mm × 0.25 μm). Ultrahigh-purity helium was used as carrier gas at a constant flow of 1.0 ml/min. Injector temperature was 240 °C. The injection mode was split with a ratio of 10:1. Injection volume was 1 μl. The oven temperature program was 50 °C–4 °C/min–200 °C (0 min)–8 °C/min–230 °C (15 min). Electron impact mass spectra were acquired at an ionization energy of 70 eV, and MS source temperature was 230 °C. Mass spectral analysis was performed in selected ion monitoring (SIM) mode. SIM parameters, including ions and dwell times, were listed in Table 2.

Table 2 Authentic standards and their retention time (tR), quantitative masses and calibration curve
Full size table

Qualitative and quantitative analysis of the neutral volatile compounds

For qualitative analysis, the compounds with authentic standards were identified by comparing the gotten mass spectrum with the standard mass spectrum of NIST 08 library and confirmed with their authentic compounds. The analytes without authentic compounds were tentatively identified based on the comparison of mass spectra with NIST 08 library and published literature. For quantitative analysis, a calibration with internal standard was used to quantify the components with the standards. And the provisional identified components without standards were quantified with respect to the internal standard, assuming that the relative response factor was 1 and the recovery ratio was 100 %. The quantitative formula was as follows:

$$ M_{{\text{i}}} = A_{{\text{i}}} /A_{{\text{s}}} \times F_{{\text{i}}} \times M_{{\text{s}}} $$
$$ C_{{\text{i}}} = M_{{\text{i}}} /{\text{m}} $$

where A i is the peak area of compound i, A s the peak area of internal standard, M i the mass of compound i, M s the mass of internal standard, F i relative correction factor, assumed to be 1, C i the concentration (μg/g) of compound i and m the mass of sample (tobacco leaves).

Stock, standard and internal standard solutions

A stock solution containing 11 compounds was prepared in dichloromethane in a 50-ml volumetric flask, and the concentrations of them are shown in Table 2. Five-point different calibration solutions were prepared by diluting 8, 20, 60, 150 and 400 μl of stock solution of standard compounds to 50 ml with dichloromethane, respectively. And the internal standard was added to the calibration solutions at the concentration of 58.1 μg/ml.

Statistical analysis

The experimental results obtained were expressed as mean ± SD of triplicates. Factorial analysis of variance (ANOVA) was performed using t-Tukey test, and p < 0.05 was regarded as significant and p < 0.01 as very significant. The neutral volatiles data were statistically evaluated by analysis of variance principal component analysis (PCA) with IBM SPSS Statistics Version 19.0 software.

Results and discussion

Basic chemical analysis of tobacco in different positions

Sugars are natural components in tobacco, which are formed via enzymatic hydrolysis of starch during the period after harvesting and in the early stages of the curing process [18]. The predominant sugars found in tobacco are glucose, sucrose and fructose [16]. Sugars generate flavorings in mainstream smoke and acids that neutralize the harsh taste that impacts on throat during smoking [19]. In this view, the overall sugar concentration in tobacco is one of the most important indexes in tobacco industry. The water-soluble sugars in tobacco leaf of all the 30 samples were checked. The results in Table 3 showed that the total sugars did not vary significantly between middle and lower leaves at 0.05 level. The content of total sugars in lower leaves was, however, much lower than that in the other two positions. The total nitrogen and nicotine varied significantly in tobacco leaves in different positions. A slighter tendency was observed: B > C > X in all of individual class of tobacco samples. The ratio of total sugars and total nitrogen (S/N) was often used to describe concordance of the tobacco compounds. Previous studies showed that the appropriate ratio of soluble sugar and nicotine was about 10:1 in tobacco and its products [20]. It is important to note that the soluble sugar content increased with the grade of tobacco, while the protein content decreased with that. In Table 3, the results indicated that the middle leaves have a more suitable S/Nic (the ratio of total sugars and total nicotine), which was around 10:1 and not too high or too low. Due to the high content of nicotine, the upper leaves have a low S/Nic, and the irritation of the upper part was also the most conspicuous among all the leaves. But we also observed that the disparity in basic chemical components was not enough to assess the different quality of leaves in different positions. As a matter of fact, neutral volatile compounds were identified as the possible main factors contributing to the tobacco flavor.

Table 3 Basic chemical analysis of tobacco in different positions
Full size table

Neutral volatiles analysis of tobaccos in different stalk positions

The volatile compounds, especially the neutral oxygenated volatiles in the tobacco leaves, can be partly or totally transferred into the smoke and therein contributed to the flavor and taste of cigarettes. So, investigating the neutral volatiles in tobacco leaves may help to control the quality of tobacco products. The neutral volatiles in tobacco were categorized into five groups according to the degradation of the different flavor precursors: phenylalanine degradation products (benzaldehyde, benzylalcohol, phenylethylalcohol, etc.), cembranoids degradation products (solanone), plastid pigments degradation products (megastigmatrienone, β-ionone, 6-methyl-5-hepten-2-one, isophorone, etc.), Maillard reactions products (2-methyltetrahydrofuran-3-one, 2-furaldehyde, 2-acetyl furan, 5-methylfurfural, furfuralcohol, methyl cyclopentenolone, 5-hydroxymethylfurfural, etc.) and terpenes derivates (linalool, citronellol, farnesylacetone, etc.) [20]. However, the neutral volatiles are so complex that it is difficult to detect all of them. So in this study, selected ion mode in GC–MS was applied to detect 19 target neutral flavorant components from the complex mixture of the volatiles. These 19 volatiles are common and representative flavor substances in Chinese flue-cured tobacco and can be used as target chemicals in the volatiles of tobacco leaves [21].

The results in Fig. 1 indicated that the upper (B) and middle (C) tobacco contained larger quantities of the neutral volatile flavor compounds than the lower parts (X) of tobacco. A critical analysis of the data according to areas of cultivation (Suiyang, Qiannan, Panxiandashan and Panxianzhudong in Guizhou Province) with respect to B and C exhibited significant difference in their total volatile compounds. In case of Hunan Province (Chenzhou Yun 87, Chenzhou Heping, Chenzhou Qiaoshi and Chenzhou Haotang), B, C and X were in the order of B > X and C > X according to the content of their total volatiles. It is worthy to mention that B and C parts did not vary significantly. The total neutral volatiles contents in upper parts were higher than those in middle and bottom parts, with the middle parts in a little higher level of total neutral volatiles than the bottom parts. Possible reasons for that result might be as follows: Leaves lying in different stalk positions have different conditions in sunshine, temperature and nutrient, which results in the discrepancy in their accumulated volatile precursor levels.

Fig. 1
figure 1

The content of total neutral volatile components of different samples; different letters indicate significant difference (p < 0.05)

Full size image

Principal component analysis (PCA)

Since the total neutral volatile flavor compounds concentration is a comprehensive index, which can mask the differences in each part, the PCA was applied to identify the main factors that contributed to the disparity in different parts. We performed PCA on 11 of the 19 elements measured. Eight elements were excluded from the analysis because they were traced and not detected in many samples.

To ensure suitability for conducting factor analysis, the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity were initially employed. The KMO test measures the adequacy of a sample in terms of the distribution of values for the execution of factor analysis. The acceptable values should be greater than 0.5. Bartlett’s test was used to make sure that the samples do not have equal variances [22, 23]. The KMO value (between 0 and 1; the value of measure closer to 1 indicates a pattern of correlation) of 0.655 found in the present study means that the observation was appropriate, suggesting a strong correlation. Since Bartlett’s test conducted on the data of the present study (Sig, 0.000) was proved to be highly significant (<0.01), it means that PCA technique was suitable and appropriately employed in the current study.

The result of statistical evaluation is shown in Table 4. For different tobacco samples, PCA extracted four principal components with eigenvalues greater than one (PC1 = 4.95, PC2 = 1.99, PC3 = 1.27 and PC4 = 1.04). The four selected components accounted for over 84.18 % of the total variation. The separation of first and second PC scores of tobacco samples of three different stalks was shown in Fig. 2. The results indicated that the samples were clustered into three groups. The tobacco leaves in the same stalk position appeared in the close zone of the PC plot, which readily revealed that a good differentiation of the tobaccos in different positions could be achieved via analysis of the neutral volatile compositions. This is useful for fabricating formula blends in tobacco industry, which means that the samples in same cluster may share the similar aroma characteristic and could be interchanged in some cases. Table 4 lists the eigenvalues associated with each component before and after extraction. The higher the loading of a variable, the more that variable contributes to the variation accounted for by that particular principal component. A principal component with an eigenvalue ≥1 is usually considered as being statistically significant [10]. The loading plot of PCA was carried out to further understand the influence of the neutral volatiles on differentiation of these three species. The components that explain maximum variance in the data are given higher loading values, while others that do not play an important role are given loading values lower than 0.5 (Fig. 3). Figure 3 showed that the first principal component explained 48.1 % of the total variation and was mainly defined by solanone, megastigmatrienone, 5-methylfurfural, furfuralcohol, phenylethylalcohol and benzylalcohol. The second principal component explained 18.1 % of the total variation and was mainly defined by β-ionone and farnesylacetone. The results obtained from Figs. 2 and 3 showed that the differentiation of the upper parts from the middle parts was due to the different value of PC2 (Fig. 2) that was characterized by the β-ionone and farnesylacetone concentration (Fig. 3). When the upper tobacco leaves were selected in fabricating formula blends, exogenous β-ionone and farnesylacetone or their flavor precursors could be added to improve the taste of the cigarettes.

Table 4 Total variance explained
Full size table
Fig. 2
figure 2

Score plot of first two principal components (PCs) for classification of tobacco samples of different positions circle upper leaves, square middle leaves, time symbol lower leaves

Full size image
Fig. 3
figure 3

Loading plot of first two principal components (PCs) for classification of tobacco samples of different positions

Full size image

Conclusion

The tobacco leaves in different stalk positions have quite different quality and therefore influenced their applicability in the blend cigarettes formula. The factors affecting the quality of tobaccos are multifaceted. But comprehensive review on all of these factors is time-consuming and not necessary. The application of multivariate statistical analysis can be used to reduce variables dimensions, concentrate information and simplify index structure. In this study, PCA was employed to reveal the differences among these three species based on neutral volatile compounds. PCA-extracted four principal components altogether accounted for 84.18 % of total variances. The important variables contributing positively to these principal components were solanone, megastigmatrienone, 5-methylfurfural, furfuralcohol, phenylethylalcohol, benzylalcohol, β-ionone and farnesylacetone. These results demonstrated that the neutral volatiles can be used as useful variables to characterize and classify the tobacco leaves of different positions. Therefore, our findings provide useful information on the evaluation of tobacco leaves and will help to reduce workload and improve production efficiency while fabricating cigarette blends in tobacco industry.