Volatilomics for halal and non-halal meatball authentication using solid-phase microextraction–gas chromatography–mass spectrometry

2.1 Meat sample collection

All samples were collected in December 2019. The wild boar meat samples used in this experiment were obtained from three male wild boars (40–50 kg) originating from Jambi Forest in Sumatera, Indonesia. The meat samples were chopped from the flanks, ribs, and shanks in equal amounts, tightly packed, and then sealed in plastic bags, put in an icebox, and transported to Bogor. The beef samples were obtained from three cattle (Brahman crossbred cattle, 400–550 kg) and taken from the flank only since this is most commonly used by meatball sellers in Indonesia. Three fresh broiler chicken meat samples were obtained from the local slaughterhouse in Bogor. All meat samples were immediately transported to the laboratory and kept in a freezer (−33 °C) prior to analysis.

2.2 Preparation of meatball samples

The meatballs were prepared according to the cooking method commonly used by meatball sellers in Indonesia, though we did not use any spices or taste enhancers, to avoid masking effects. The meat samples were homogenized using a Phillips ProMix hand blender (HR2533), and the meatballs were made by mixing 200 g of minced meat with 20 g of tapioca flour (Cap Pak Tani, Bogor, Indonesia) and 50 g of ice cubes. The dough was then formed by hand into balls with 3–4 cm diameters, put in water at a temperature of 80 °C–100 °C and boiled for 15 mins. The beef, chicken, and wild boar meatballs were each made in triplicate. Similar steps were conducted to prepare the mixed meatballs. The mixed meatballs also contained 200 g of chicken along with wild boar and with beef in the following ratios: 20:80, 40:60, 60:40, and 80:20. In addition, meatballs made of a combination of beef, chicken, and wild boar with a ratio of 40:40:20 were also prepared. Each mixed meatball was made in duplicate. A total of 31 meatball samples were used in this study (Table 1). For SPME analysis, the meatballs were crushed with a porcelain mortar and pestle. The crushed meatball samples (8 g) were put in a closed headspace vial with 5 mL of distilled water for exposure to SPME fiber.

2.3 Analysis of volatile compounds

3.1 Volatile profiling by gas chromatography–mass spectrometry

Table 2 presents the volatile compounds with known molecular formulas that were identified in the meatball samples. Only compounds detected in the control samples are listed, and data from the mixed meatballs are not presented. Overall, 150 volatile compounds were found in the chicken meatballs; 148, in the beef meatballs; and 141, in the wild boar meatballs. These volatiles consisted mainly of aldehydes, ketones, alcohols, acids, esters, aliphatic hydrocarbons, aromatic hydrocarbons, terpenes, and miscellaneous compounds (Table 2 and Fig. 1). They may have been formed as the result of lipid oxidation, the Maillard reaction, interactions between Maillard reaction products and lipid oxidation products, and/or the thermal degradation of thiamine that occurs during the cooking process (Kosowska et al., 2017). The oxidation of lipids produces a wide range of aliphatic compounds, including saturated and unsaturated hydrocarbons, alcohols, aldehydes, ketones, acids, and esters (Ayseli et al., 2014). At the same time, the Maillard reaction may include many heterocyclic compounds including sulfur and nitrogen compounds (Dashdorj et al., 2015). The data in Table 2 indicate that most of the volatiles detected in this study were the result of lipid oxidation (alcohols, hydrocarbons, aldehydes, and ketones), whereas a few were based on thiamin degradation (sulfur compounds) and the Maillard reaction (acetoin and aldehydes). Some compounds might also have resulted from an interaction of these three processes.

Close

Fig. 1

Composition of the volatile compounds detected in each type of control meatball.

Export to PPT

The volatile compounds found in each type of meatball were grouped based on their functional groups (Fig. 1). It can be seen that all the meatballs had a similar composition of volatile components. However, compounds from the ketone group were more common in the chicken meatballs, whereas aldehyde compounds were least present in the chicken. In fact, ketones and aldehydes are major contributors to the “chicken-like” scent (Kerler and Grosch, 1997). A number of compounds, including nonane; 3-methyl-, 2,3-octanedione, 4-nonanone, acetophenone, 6-dodecanone, 2-heptanone, 6-methyl-,2-methyl-3-octanone, 2,5-octanedione 5-hepten-2-one, 6-methyl-,2-nonanone, 6,7-dodecanedione, 2-undecanone, 11-dodecen-2-one, 3-tridecanone, and nona-3,5-dien-2-one, were detected in the chicken meatballs. A few of the aforementioned compounds, such as 4-nonanone and acetophenone, have previously been identified in boiled chicken (Kerler and Grosch, 1997), and 2,3-octanedione has been found in raw chicken breast (Ayseli et al., 2014).

Only a few previous reports on the composition of the volatile compounds in fresh or boiled wild boar meat were found. Sales and Kotrba (2013) reported that fried wild boar meat contained 48 volatile compounds, including 16 aldehydes, 5 ketones, 6 alcohols, 8 acids, 4 sulfur compounds, 8 pyrazines, 2 furanones, 1 pyrrole, and 3 aromatic compounds. In the present study, 141 volatile compounds were found in the wild boar meatballs, including 2 sulfur compounds, 6 heterocyclics, 20 terpenes, 13 ketones, 3 esters, 3 acids, 29 alcohols, 26 aldehydes, 17 aliphatic hydrocarbons, 17 cyclic aromatic hydrocarbons, and 5 miscellaneous compounds. Alcohol was the most abundant chemical family in the wild boar meatballs. One of the compounds, 1-octen-3-ol, is an important volatile compound and product related to meat fatty acid autoxidation (Mottram, 1998), leading to a mushroom-like scent (Lammers et al., 2009). Other alcohol compounds found in samples of fresh wild boar meat (Sales and Kotrba, 2013) and fried wild boar meat (Lammers et al., 2009) include 1-pentanol, 1-hexanol, 1-heptanol, and 1-octanol.

The volatile compounds detected in the raw and cooked beef were categorized into eight groups: hydrocarbons, alcohols, aldehydes, acids, esters, ketones, furans, and sulfur compounds. The alcohols, acids, and esters were less diverse in the cooked beef as compared with those in the raw beef. By contrast, the aldehydes and ketones were more diverse in the beef after cooking. In particular, the aldehyde diversity increased from four compounds before cooking to 20 compounds after cooking, with hexanal being one of the most abundant (Wang et al., 2018). Overall, 30 aldehydes were detected in the beef meatballs, with butanal being the most abundant (Table 2).

3.2 Volatilomics

Volatilomics is a term for volatilome analysis aimed at the detection, characterization, and quantification of volatile metabolites from organics (Lytou et al., 2019). The volatilome is defined as the group of all volatile organic compounds produced by a living organism (plants, animals, etc.), an ecosystem, or a substrate (such as food), and it includes exogenously derived compounds (organic and inorganic; Lytou et al., 2019). The volatilomic approach has recently been applied in various research fields, for applications such as plant analysis (Lytou et al., 2019) and the discrimination of beef and pork (Pavlidis et al., 2019). Most volatilomics studies have employed GC–MS and electronic nose methods. Multivariate data analysis has also been used to analyze the resulting high-dimensional data. Various multivariate methods can be used to extract information from large amounts of volatilomics data, with PCA, PLS-DA, and OPLS-DA being the most common (Worley and Powers, 2016).

To evaluate the meatball classification pattern based on the volatile compound composition, unsupervised PCA was performed on the GC–MS data (chromatographic relative peak area) using unit-variance (UV) scaling. In UV scaling, the scaling weight is 1/sk, where sk is the standard deviation of parameter k. Thus, in UV scaling, all variables have an equal opportunity to influence the data, making it more objective than other scaling methods (Eriksson et al., 2006). Multiplicative signal correction (MSC) filtering was applied to remove the signal noise. The main objective of using MSC filtering is to remove artifacts and interference that are not correlated to the presence of the target analytes (Eriksson et al., 2006). After excluding three sample outliers (one 100%-chicken meatball sample, one mixed chicken–wild boar meatball sample with a 4:6 ratio, and one mixed chicken–wild boar meatball sample with a 6:4 ratio), a PCA model with three principal components explaining 59.8% of the variation was obtained using Hotelling’s T² analysis with a 95% confidence interval. PC1, PC2, and PC3 of the PCA explained 43.8%, 9.14%, and 6.81% of the variation, respectively. Only the first two components are presented (Fig. 2). The aforementioned results are consistent with the minimum requirements for the model mentioned by Worley and Powers (2016), which includes Q² = 0.41. The score plot revealed three distinct groups: beef meatballs (B1–B3), chicken meatballs (C1–C2; C3 was excluded because it was located outside the 95% confidence interval), and meatballs made from 100% wild boar meat or a mixture of wild boar with beef and/or chicken (Fig. 2). Notably, all meatballs made from a mixture of wild boar with beef and/or chicken at different ratios were clustered together with the 100%-wild boar meatballs. This may have occurred because of the strong influence of the volatile components present in wild boar meat. The loading plot revealed several volatiles responsible for the three groupings (Fig. 2). For the beef meatball group, the discriminating compounds were 2-amino-5-methylbenzoic acid, 2-amino-6-methylbenzoic acid, benzaldehyde, dimethyl trisulfide, and 1-octanol. For the chicken meatball group, 5-hepten-2-ol, 6-methylnonane, 3-methyl, and 2-ethylbutanol were among those predominant discriminating compounds. Lauric acid, 2,4-di-tert-buthyl-phenol, and furan 2-pentyl were the discriminating compounds for the wild boar meatballs and wild boar-containing meatballs.

Close

Fig. 2

PCA loading biplot of meatballs made from 100% beef (B1–B3), 100% chicken (C1–C3), 100% wild boar (W1–W3), and mixtures thereof (WB: wild boar–beef, WC: wild boar–chicken, and WBC: wild boar–beef–chicken). The loading biplot illustrates several markers of the beef, chicken, and mixed meatballs. The numbers after the letters represent the percentages of the respective meats. The last number represents the number of replications.

Export to PPT

To obtain a clearer classification pattern, a supervised multivariate data analysis method (PLS-DA) was employed. Further analysis using a supervised method is only recommended when the PCA model for the same set of data has an acceptable predictive coefficient (Q² of at least 0.4; Worley and Powers, 2016), which was fulfilled by the aforementioned PCA model. The PLS-DA score plot exhibited better performance than the PCA. The PLS-DA model with three classes (class 1: chicken meatballs, class 2: beef meatballs, and class 3: wild boar and wild boar–beef–chicken meatballs) had good performance, with a cumulative explained variance of R²X = 0.69, R²Y = 0.99, and Q² = 0.84 (Zhang et al., 2020). Here, the beef and chicken samples were clustered separately from each other. Again, the wild boar meatballs and meatballs made from mixtures of wild boar with chicken and/or beef at different ratios were grouped separately from the beef and chicken samples (Fig. 3).

Close

Fig. 3

PLS-DA plot score of meatball samples (C: chicken, B: beef, W: wild boar, WB: wild boar–beef, CW: chicken–wild boar, and WBC: wild boar–beef–chicken). The number represents the ratio of each meat and replication number.

Export to PPT

Further validation with 100 random permutations was performed. As illustrated in Fig. 4, the values of R²Y (green circles) and Q²Y (blue squares) from the permuted analysis (bottom-left corner) were lower than the associated initial values (top-right corner), indicating the stability of the model and the absence of overfitting (Song et al., 2021). The p-value for the cross-validated analysis of variance (CV-ANOVA) was less than 0.005 (2.6 × 10⁻⁴), demonstrating the model validity (Eriksson et al., 2008).

Close

Fig. 4

Permutation test for the PLS-DA model for (A) chicken meatballs, (B) beef meatballs, and (C) wild boar and beef–wild boar–chicken meatballs.

Export to PPT

3.3 Potential volatile marker of wild boar meatballs

To elucidate the volatile compounds that serve as markers for each PLS-DA class, a correlation coefficient and the VIP values were used. The compounds that were positively or negatively correlated with the groupings could be determined using the coefficient, whereas the VIP value has only a positive value. Fifteen compounds with both positive and negative correlation values and the largest VIP values were selected from each PLS-DA class (Table 3).

The volatile with the highest VIP value in the chicken meatball class of the PLS-DA was β-cymene (Table 2). We could find no previous report on the occurrence of this compound in fresh or cooked chicken meat. However, a recent review indicated that cymene was found in essential oils, which are often added to poultry feed as natural antibiotics and immune-stimulants (Brenes and Roura, 2010). A previous study also reported that the second-strongest positive compound, 3-methylbutanal (an aldehyde), was detected in thermally processed chicken as a result of the Maillard reaction (Tian et al., 2007). This was also recently reported as one of the volatiles detected in Dezhou braised chicken (Duan et al., 2015) and grilled chicken (Ngamchuachit et al., 2015).

In the beef class, the most robust discriminator was 5-ethyl-m-xylene. Other discriminating volatiles, including benzaldehyde, octanol, 2-nonenal, and heptanal, have been found among the volatiles isolated from heat-treated beef, and 2-nonenal was also found in processed pork (Dwivedi and Brockmann, 1975). In this study, heptadecanal exhibited a significant contribution as a discriminating volatile in the beef, though it has been previously found in processed pork and ham (Dwivedi and Brockmann, 1975). Dimethyl trisulfide was also one of the potent odorants identified in stewed beef juice (Guth and Grosch, 1994).

In the wild boar and mixture meatballs group, xylene was identified as the strongest discriminator. This compound has previously been detected as a volatile in processed pork and ham (Dwivedi and Brockmann, 1975). The positive volatile with the second-highest VIP value in the wild boar and mixture meatball class was 2,4,5-trimethyl-thiazoles. Thiazoles have been reported as volatiles directly leading to a complex, meaty aroma (Piao et al., 2019). Pentanal has also been identified in cooked Iberian pigs (Estévez et al., 2003) and had a strong positive effect on the discrimination of minced pork from minced beef (Pavlidis et al., 2019).

As previously described, a strong discriminating volatile was assumed to be present in the 100% wild boar meatballs, which was responsible for clustering all the wild boar-containing meatballs in the same group. To address this, a PLS-DA model with three classes was created, including the meatballs made from 100% chicken, 100% beef, and 100% wild boar (Fig. 5). Although a CV-ANOVA indicated that the model was slightly overfitting, the R²Y and Q²Y values were 0.99 and 0.88, respectively. A further test with 200 random permutations also indicated an acceptable model (figure not shown). In this PLS-DA, 2-nonanone and pentanal were the two strongest positive discriminating volatiles from the wild boar group. Other discriminating compounds for 100% wild boar meatballs were summarized in Table 4. The compounds were compared with the 15 strongest positive discriminating volatiles when the wild boar meatballs were put in the same group with the meatballs made from a mixture of wild boar with beef and/or chicken at different ratios (Table 3; derived from the first PLS-DA model). As a result, six compounds (pentanal, 2,6-dimethylcyclohexanone, 1-undecanol, cyclobutanol, 2,4,5-trimethyl-thiazole, and 5-ethyl-3-(3-methyl-5-phenyl pyrazol-1-yl)-1,2,4-triazol-4-amine), were found as discriminating volatiles in mixture meatballs but with different VIP value (Table 3). However, the strongest pure wild boar meatballs discriminating volatiles (2-nonanone) was not found among volatiles that positively correlate with mixture meatballs grouping. This compound was reported as one of major ketones found in raw pork (Soncin et al., 2007), but there is no reports on its availability in wild boar.

Close

Fig. 5

PLS-DA score plot of meatball volatiles' data. Only samples of 100% chicken (class 1; C1, C2, and C3), 100% beef (class 2; B1, B2, and B3), and 100% wild boar (class 3; W1, W2, and W3) are included.

Export to PPT

In the further analysis, we excluded 100% wild boar meatballs data to obtain another PLS-DA with 3 classes (100% chicken, 100% beef, and mixture meatballs) Fig. 6. Fifteen compounds with both positive and negative correlation values and the largest VIP values were selected from each PLS-DA class and summarized in Table 5. Six discriminating volatiles found in pure wild boar meatballs (pentanal, 2,6-dimethylcyclohexanone, 1-undecanol, cyclobutanol, 2,4,5-trimethyl-thiazole, and 5-ethyl-3-(3-methyl-5-phenyl pyrazol-1-yl)-1,2,4-triazol-4-amine) were consistently found as discriminating volatiles of mixture meatballs class in this new PLS-DA model, although with different VIP value. Similarly, the strongest discriminating compound, 2-nonanone, was not found. These data may partially support the hypothesis that several strong wild boar-discriminating volatiles heavily influenced the clustering of all wild boar-containing meatballs in the same group.

Close

Fig. 6

PLS-DA score plot of meatball volatile compounds data. Only samples of 100% chicken (class 1; C1, C2, and C3), 100% beef (class 2; B1, B2, and B3), and mixture meatballs (class 3; WB with different compositions) are included. Pure wild boar meatballs are excluded. Only the first two PC is presented (PC1 = 46.4%, PC2 16.4%, R²Y = 0.984, Q²Y = 0835).

Export to PPT

Meat flavor formation during heating, especially when involving volatiles, is a complex process involving various reactions, including the Maillard and unsaturated lipid reactions. Compounds resulting from the Maillard reaction may also react with those from the unsaturated lipid degradation. The exact volatile composition of the meat flavor formed by these reactions depends on not only the types of precursors present in the meat but also the temperature and reaction time (Aaslyng and Meinert, 2017). Meatballs made from a mixture of different types of meat (chicken, beef, and wild boar) at different ratios may develop different volatiles, as concentrations of the precursors vary. This assumption might explain why the discriminatory volatiles of wild boar meatballs were not exactly the same with when clustered together with mixed meatballs as when separated from them.

This study did not include commercial meatballs in its analysis. Instead, the present work is a preliminary study with a very simple meatball formulation. Commercial meatballs typically have a much more complex formulation. Besides meat, flour, salt, and pepper, commercial meatballs may also contain garlic, beef flavor, or a taste enhancer, which could affect the selection of volatile markers. Further research using more complex meatball formulations resembling those of commercial meatballs and including samples of commercial meatballs themselves is required.

In addition, this study did not consider the effect of the animal feed, which can significantly contribute to meat's volatile composition. The volatile compounds in cooked meat can be directly diverted from animal feed into the tissue by the transformation of feed molecules through the action of ruminal microorganisms or by both the Maillard reaction and the oxidation of lipids during the heating process (Vasta and Priolo, 2006). An example of such a case is in a study by (Resconi et al., 2010), in which male Corriedale lambs that were only fed by pasture were found to have significantly lower levels of alkanals, alkadienals, and ketones compared with those of lambs fed by pasture and concentrate and concentrate-plus-lucerne hay.