Classification of different pineapple varieties grown in Malaysia based on volatile fingerprinting and sensory analysis

Background Pineapple is highly relished for its attractive sweet flavour and it is widely consumed in both fresh and canned forms. Pineapple flavour is a blend of a number of volatile and non-volatile compounds that are present in small amounts and in complex mixtures. The aroma compounds composition may be used for purposes of quality control as well as for authentication and classification of pineapple varieties. Results The key volatile compounds and aroma profile of six pineapple varieties grown in Malaysia were investigated by gas chromatography–olfactometry (GC-O), gas-chromatography–mass spectrometry and qualitative descriptive sensory analysis. A total of 59 compounds were determined by GC-O and aroma extract dilution analysis. Among these compounds, methyl-2-methylbutanoate, methyl hexanoate, methyl-3-(methylthiol)-propanoate, methyl octanoate, 2,5-dimethyl-4-methoxy-3(2H)-furanone, δ-octalactone, 2-methoxy-4-vinyl phenol, and δ-undecalactone contributed greatly to the aroma quality of the pineapple varieties, due to their high flavour dilution factor. The aroma of the pineapples was described by seven sensory terms as sweet, floral, fruity, fresh, green, woody and apple-like. Conclusion Inter-relationship between the aroma-active compounds and the pineapples revealed that ‘Moris’ and ‘MD2’ covaried majorly with the fruity esters, and the other varieties correlated with lesser numbers of the fruity esters. Hierarchical cluster analysis (HCA) was used to establish similarities among the pineapples and the results revealed three main groups of pineapples.


Background
Pineapple (Ananas comosus L. Merr) which is one of the most popular exotic fruits in the world trade is widely distributed in tropical regions such as the Philippines, Thailand, Malaysia and Indonesia. In 2016, the global pineapple production was estimated at 24.78 million metric tons with Costa Rica (2930.66 metric tons), Brazil (2694.56 metric tons), Philippines (2612.47 metric tons), India (1964 metric tons),Thailand (1811.59 metric tons, and Nigeria (1591.28 metric tons) as the top five pineapple producers in the world [1]. Other important producers are: Indonesia, China, India, Mexico, and Colombia bio-catalytic synthesis or isolated from microbial fermentations [4]. There are many pathways involved in volatile biosynthesis starting from lipids [6], amino acids [7], terpenoids [8] and carotenoids [9]. Once the basic skeletons are produced via these pathways, the diversity of volatiles is achieved via additional modification reactions such as acylation, methylation, oxidation/reduction and cyclic ring closure [6]. As the content of aroma compounds in pineapple depends on many factors such as the climatic and geographical origin [10], varieties [11], different stages of ripening [12], and postharvest storage conditions [13], the aroma compounds composition may be used for purposes of quality control as well as for authentication and classification of pineapple varieties.
Fingerprinting techniques, based on chemical composition and multivariate statistical analysis have been used in characterising or classifying wines according to origin, quality, variety and type [14,15]. It was also used in the authentication of green-ripe sea-freighted and airfreighted pineapple fruits harvested at full maturity [16]. Application of untargeted fingerprinting techniques as a means of gaining insight into the reaction complexity of a food system has received tremendous interest among researchers [17]. Fingerprinting is defined as a more unbiased and hypothesis-free methodology that considers as many compounds as possible in a particular food fraction [18]. Fingerprinting doesn't concentrate on a specifically known compound, rather it allows for an initial fast screening to detect differences among samples. Meanwhile, chemometric techniques such as principal component analysis (PCA) and hierarchical cluster analysis (HCA) are employed in the analysis of generated data. PCA is often complemented with HCA to explore data sets obtained by gas chromatography. This method has been used in the classification of wines based on their volatile profiles [19]. Multivariate techniques of data analysis represent a useful statistical tool to differentiate between different fruit varieties [20]. Also, this chemometric approach has been used to classify muskmelon [21], tomato fruit [22], and citrus juice [20].
Although much work has been done on volatile fingerprinting in apple fruits [23], and grape fruits [24], there has been no systematic study on volatile fingerprinting of fresh pineapple fruits grown in Malaysia. The purpose of this study were: (1) to identify and quantify the volatile compounds in six different varieties of pineapples grown in Malaysia (Moris, Maspine, MD2, N36, Josapine and Sarawak) and (2) apply fingerprinting technique to determine which volatile compounds may be potential markers for pineapple varieties grown in Malaysia.

Sensory evaluation
The aroma qualities of the six different pineapple varieties were elucidated by ten trained panellists. The obtained relative standard deviation from the mean aroma quality intensities varied within the range of 1.2-5.9% depending on the pineapple variety and the aroma quality. The details of the aroma qualities of the pineapples are listed in Table 1. Results of the aroma qualities revealed significant differences (p < 0.05) among varieties for all attributes. For instance, while pineapple 'MD2' presented the highest intensities for sweetness (8.62), floral (6.88) and apple-like (8.31) attributes, 'Moris' produced the highest intensities for fruity (6.83) and fresh (7.31) attributes, respectively. On the other hand, 'Sarawak' had the strongest woody (7.46) and green (7.62) attributes. The other pineapple varieties ('Josapine' , 'N36' and 'Maspine') produced varied aroma responses. 'Josapine' had strong sweet and woody attributes with relatively low floral aroma. 'Maspine' exhibited strong sweet and green aroma notes. 'N36' had strong sweet and woody aroma, respectively.
To have an insight into the reasons behind this observation, the different pineapple varieties were subjected to AEDA and GC-O.
Meanwhile, methyl-2-methylbutanoate which exhibited the highest FD factor had a bigger influence on the aroma profile of pineapple 'Moris' . It was however, not detected in the other varieties. On the other hand, methyl hexanoate and DMHF contributed significantly to the aroma profiles of the different pineapple varieties. This observation was similar to those of Zheng et al. [3]. For instance, the FD factors of methyl hexanoate in the different pineapple varieties were 64, 128, 64, 32 and 16 corresponding to 'Moris' , 'MD2' , 'N36' , 'Josapine' and 'Sarawak' . 2,4-Dihydroxy-2,5-dimethyl-3(2H)-furanone had greater influence on the aroma profiles of "Moris' 'Maspine' and 'MD2' with a corresponding FD factors of 16, 64 and 128, respectively. In addition, aroma-active compounds with relatively high FD factors such as δ-octalactone, 2-methoxy-4-vinyl phenol, methyl octanoate and hexadecanoic acid had appreciable influence on the aroma profile of the pineapple varieties ( Table 2).

Table 2 Detected aroma compounds with retention index and mean concentration (µg/kg fresh fruit) found in each pineapple varieties grown in Malaysia
No Compound a Aroma-quality b Moris Maspine MD2 N36 Josapine Sarawak RI on TG-5 ms Surprisingly, δ-undecalactone was mainly detected in 'MD2' and 'Josapine' . Lactones which exhibited creamy and coconut-like aroma notes in the pineapple varieties have been identified as most potent odorants in pineapples [27]. The formation of lactones in fruits has been documented. There are two proposed pathways for the formation of lactones [28]. The first pathway is from unsaturated fatty acids to lactones via hydroperoxy fatty acids and monohydroxy fatty acids under the actions of lipoxygenase (LOX) and peroxygenase (PGX). The second pathway is from unsaturated fatty acids to lactones via epoxy fatty acids and dihydroxy fatty acids under the actions of PGX and epoxide hydrolase. 4-Hydroxy-2,5-dimethyl-3(2H)-furanone and its methyl ether 2,5-dimethyl-4-methoxy-3(2H)-furanone are important odorants of many fruits [29]. Whereas, 4-hydroxy-2,5-dimethyl-3(2H)-furanone and its derivatives are synthesized by a series of enzymatic steps in fruits, they are also products of Maillard reaction [30].

Relationship between pineapple varieties and odour-active compounds
In order to differentiate between the six different pineapples in terms of the aroma-active compounds associated with each variety, principal component analysis (PCA) was used. PCA provides a visual relationship between the pineapple varieties and their aroma-active compounds. This method makes the interpretation of the multivariate analysis easy. A first PCA was performed on the concentration of the 59 volatile compounds (Table 2) analysed in the pineapple varieties. Based on the samples grouping from PCA, a partial least square discriminant analysis (PLS-DA) was established (Fig. 1a). The scatter plot of scores of the first two components (in PLS-DA which explained 95% of the total variance in the data) showed the differences among the six pineapple varieties. The corresponding PLS weight plot (Fig. 1b) revealed the inter-relationship between the aroma compounds and the pineapple varieties.
In order to validate the results obtained by PCA analysis, a hierarchical cluster analysis (HCA) was carried out using Ward's method of agglomeration and Euclidean distances to evaluate similarity between varieties. The test was performed on the complete dataset, thus obtaining the dendrogram in Fig. 3. Three main groups of pineapple varieties were identified by HCA. The first group comprised pineapple 'Moris' and 'MD2' Fig. 3. This group was characterized by high numbers of aroma-compounds most especially the fruity esters. They contained some of the highly intense aroma-active compounds (FD ≥ 64) such as methyl-2-methyl butanoate, methyl hexanoate, methyl-3-(methylthiol)-propanoate and 2,4-dihydroxy-2,5-dimethyl-3 (2H)-furanone. The second group contained pineapple 'Maspine' . This group contained the least quantity of fruity esters. The third group included 'Sarawak' , 'Josapine' and 'N36' . This group contained more of the fatty acid methyl esters.

Conclusion
Sensory evaluation, GC-O and GC-MS analysis were employed to elucidate the characteristic aroma of six pineapples varieties grown in Malaysia. Application of qualitative descriptive sensory analysis on the six pineapple varieties revealed seven quality terms such as sweet, floral, fruity, fresh, green, woody and apple-like. In addition, 97 aroma-active compounds were identified by GC-O and AEDA in the pineapple varieties. Of this, pineapple 'Moris' had the highest numbers of aroma-active compounds with a total of 31 compounds and this was followed by 'MD2' with 27 compounds. The next were the 'N36' , 'Maspine' , and 'Sarawak' which produced 24, 20 and 18 aroma-active compounds, respectively. 'Josapine' had the least number of aroma-active compounds (16). In order to address the inter-relationship between the sensory attributes and the aroma compounds, the PLSR analysis was employed. Results of the analysis showed that 'Moris' and 'MD2' covaried majorly with the fruity esters with higher FD factors. 'Sarawak' , 'Josapine' and 'N36' were correlated with fewer fruity esters; they covaried majorly with the lactones. However, the variety 'Maspine' was correlated with 2-methoxy-4-vinyl-phenol (C33), (Z)-7-tetradecenal (C43), 3,5-dimethoxy-4-hydroxycinnamaldehyde (C45), pentadecanoic acid (C46), methyl hexadecanoate (C47) and octadecanoic acid (C55), respectively. In addition,  Table 2 Page 9 of 12 Lasekan and Hussein Chemistry Central Journal (2018) 12:140 hierarchical cluster analysis was used to establish similarities among the pineapples and the results revealed three main groups of pineapples.

Isolation of pineapple volatile compounds
The isolation of the pineapple volatile compounds was performed by extracting 300 mL of juice with dichloromethane (300 mL), followed by distillation in vacuum [34]. A similar workup procedure reported earlier [35] was carried out on juice to produce 400 µL extract.

GC-MS and GC-FID analyses
The extracts were injected into a QP-5050A (Shimadzu, Kyoto, Japan) gas chromatograph equipped with a GC-17A Ver.3, and a flame ionization detector (FID). Two microliters of the extract was vaporized in the injector port maintained at 220 °C in split less mode (1 min). The oven temperature was varied from 50 °C to 250 °C at 15 °C/min, and holding times of 3 and 10 min respectively [36]. A 30-300 m/z mass range was recorded in full-scan mode. The quadrupole ion source and transfer line temperatures were maintained at 150 and 250 °C. respectively and the ionisation energy was set at 70 eV. The column (30 m × 0.25 mm i.d., and 0.25 µm film thickness; 5% diphenyl/95% dimethylpolysiloxane phase; Thermo Scientific, Milan Italy) was a TG-5 ms [36]. The carrier gas was helium at 1.5 mL/min (column-head pressure of 13 psi).

GC-O analysis
A Trace Ultra 1300 gas chromatograph (Thermos Scientific, Waltham, MA, USA) fitted with a TG-5 ms column (30 m × 0.25 mm i.d., film thickness, 0.25 µm, Thermo Scientific, Milan Italy) and an ODP 3 olfactory Detector Port (Gerstel, Mulheim, Germany), with additional supply of humidified purge air, was operated as earlier reported by Lasekan [35]. The split ratio between the sniffing port and the FID detector was 1:1. Two replicate samples were sniffed by three trained panellists who presented normalized responses, with strong agreement with one another.

Identification and quantification
Kovats method which employs a mixture of normal paraffin C 7 -C 30 as external references was used to calculate the linear retention indices [36]. The identification of compounds was as described by Lasekan and Ng [34]. When it was not possible to find appropriate reference standard, a tentative identification was obtained by matching retention index with mass spectral libraries data (WILEY 275, NBS75K). Semi-quantitative data were obtained by relating the peak area of each compound to that of the corresponding standard and were expressed as µg/kg. For compounds tentatively identified, their semi-quantitative data were obtained by relating their peak area to that of octadecane and were expressed as µg/kg octadecane.

Aroma extracts dilution analysis (AEDA)
The flavor dilution (FD) factors of the aroma-active compounds were evaluated by GC-O using the AEDA approach earlier reported by Lasekan [35]. Each of the obtained dilution was injected into the GC-O. The highest dilution in which an aroma compound was observed is referred to as the flavor dilution (FD) factor of that compound [37].

Sensory analysis
Sensory analysis was carried out by ten trained panelists (6 females and 4 males) in a sensory laboratory according to the International Standard ISO 8589: [29]. All panelists who have passed screening test as described earlier [34] were recruited from the University Putra Malaysia. Prior to the test, the panelist were taken through 1 h training session with selected aroma compounds such as: ethyl hexanoate (fruity), 2,5-dimethyl-4-hydroxy-3(2H)-furanone (Strawberry), β-damascenone (floral), ethyl isohexanote (pineapple-like), etc. Descriptors used by panelists were determined after three preliminary sensory experiments. Finally, the panelists were asked to evaluate ortho-nasally fresh pineapple juice placed inside glass containers (7 cm × 3.5 cm). Seven aroma attributes (sweet, floral, fruity, fresh, green, woody and apple-like) were obtained. Panelists were asked to score each attribute on a 10-point interval scale with 9 = strong intensity, and 0 = weak with no perception. To aid the sensory analysis, the following reference compounds: ethyl hexanote (fruity), β-damascenone (floral), methyl-3(methylthiol)propanoate (apple-like), hexanal (green), germacrene (woody), p-anisaldehyde (sweet) and (E,Z)-3,5-undecatriene (fresh, pineapple-like) were dissolved in water at a concentration of 100 × above their respective threshold values. The fresh pineapple varieties were evaluated in triplicate and the results obtained were averaged.

Statistical analysis
Analysis of variance (ANOVA) and Duncan's multiple comparison tests were carried out to establish if statistical differences existed among individual pineapple variety for each sensory attribute at (p < 0.05). Partial least square discriminate analysis (PLS-DA) and PLS-regression coefficient were employed as an exploratory tool to describe and summarise the data by grouping variables that are correlated. The mean concentrations of the 59 aroma-active compounds and the six different pineapple varieties (Table 3) were the data set. The multivariate statistical analyses were performed using the SIMCA-P software (V. 10.0, Umetricus, Umea, Sweden). Principal Components Analysis (PCA) and Hierarchical Cluster Analysis (HCA) using the Software package SPSS Statistics 17.0 (SPSS Inc., Chicago, IL) were also employed.
Authors' contributions OL conceptualized this study and critically review the content of the manuscript. FKH carried out the experiments, data analysis and interpretations. OL has made intellectual contributions. All authors read and approved the final manuscript.