Screening biogenic amines and fish-based food (keropok lekor) extracts in induction of inflammation using Principal Component Analysis

Background: Food-borne biogenic amines (BAs), namely, histamine, putrescine, cadaverine, tyramine, spermine and spermidine are known for their contributions as fish-based food freshness biomarkers to determine level of contamination. The remaining food–borne BAs (phenylethylamine, tryptamine and agmatine) effects on promoting inflammation are yet to be investigated. The effect of these compounds on induction of inflammation in macrophages was investigated using Principal Component Analysis (PCA) from independent (BAs at 1, 10 and 100 µg/ml, food extract and its standard mixture solution) and dependent variables cell viability, nitric oxide (NO) and tumor necrosis factor-α (TNF-α) secretion. Nine individual BAs and keropok lekor extracts were exposed to RAW 264.7 macrophages for 18-24 hr at 37oC with 5% carbon dioxide environment. Cell viability, NO and TNF-α secretion were determined using MTS assay kit, Greiss Reagent System and Enzyme-linked Immunosorbent Assay (ELISA) kits, respectively. Results: Q2V values were not equal to Q2 (an estimate of the predictive ability of the model) values for individual variables because the eigenvalues values were more than 0.5, indicating a good model. All variance (R2VX) values were > 0.9, suggesting goodness of fit. Conclusions: PCA thus as tool to discriminate between inflammogenic and non-inflammogenic food-borne BAs.


Abstract
Background: Food-borne biogenic amines (BAs), namely, histamine, putrescine, cadaverine, tyramine, spermine and spermidine are known for their contributions as fishbased food freshness biomarkers to determine level of contamination. The remaining food-borne BAs (phenylethylamine, tryptamine and agmatine) effects on promoting inflammation are yet to be investigated. The effect of these compounds on induction of inflammation in macrophages was investigated using Principal Component Analysis (PCA) from independent (BAs at 1, 10 and 100 µg/ml, food extract and its standard mixture solution) and dependent variables cell viability, nitric oxide (NO) and tumor necrosis factorα (TNF-α) secretion. Nine individual BAs and keropok lekor extracts were exposed to RAW 264.7 macrophages for 18-24 hr at 37oC with 5% carbon dioxide environment. Cell viability, NO and TNF-α secretion were determined using MTS assay kit, Greiss Reagent System and Enzyme-linked Immunosorbent Assay (ELISA) kits, respectively. Results: Q2V values were not equal to Q2 (an estimate of the predictive ability of the model) values for individual variables because the eigenvalues values were more than 0.5, indicating a good model. All variance (R2VX) values were > 0.9, suggesting goodness of fit. Conclusions: PCA is thus proven as an effective tool to discriminate between inflammogenic and noninflammogenic food-borne BAs.

Background
Inflammation has been associated with various chronic and acute diseases. As prevention, a method to screen foods which can cause inflammation before we eat is warranted. Foodborne biogenic amines (BAs) [histamine (HIM), putrescine (PUT), cadaverine (CAD), 2phenylethylamine (PHM), tyramine (TYM), tryptamine (TPM), spermine (SPM), spermidine (SPD) and agmatine (AGM)] had been shown to be present in various foods such as fishes, seafood-based products either raw, cooked or fermented, cheeses and beverages [12,13,15] and is associated to Scombroid fish poisoning (SFP) at high doses. SFP causes allergylike reactions and inflammation seems to be involved in the manifestation of symptoms such as urticaria.
Inflammation associated with food toxicity and food intolerance in particular had been described previously [14]. The toxicity of some BAs is determined through acute and subacute studies. Nonetheless, the toxicity of BAs on in vitro RAW 264.7 macrophages is yet to be determined. Currently, in vitro cell culture work is an alternative to in vivo toxicity study in animal models to minimizing animal usage in laboratory experiments. Mammalian cell culture has been used since they mimicked the cells of human at specific organs. BAs has been claimed to cause allergy-like reactions but their involvement in induction of proinflammatory mediator secretion is poorly understood. The relationship between cytotoxicity and potentiality to induce pro-inflammatory secretion remains to be determined.
A food model is needed to validate the PCA results, thus, keropok lekor had been chosen.
Keropok lekor or fish sausage is a traditional Malay fish snack popular among the locals in Malaysia. It is made from fish and sago flour and seasoned with salt and sugar. It comes in two main forms: lekor (long and chewy) and keping (thin and crispy). There are various BAs found in keropok lekor [10] which may cause the inflammation during oral itchiness, a problem common after eating keropok lekor.
In this study, a method to screen inflammation-causing food compounds and extracts was developed through an experimental flow in Figure 1. The cell viability and proinflammation mediator secretion effects of different biogenic amines individually, as well as in the keropok lekor extracts using the RAW 264.7 macrophages cell line was determined. BAs and keropok lekor extract were used to develop and verify the method to screen food potentially causing inflammation using RAW 264.7 macrophage cell line and multivariate statistics of Principal Component Analysis (PCA). There is no method yet as to screen food compounds possible to cause inflammation either orally or gastrointestinal. RAW 264.7 macrophages were used as a vehicle to respond to food compounds and their viability and pro-inflammatory mediator secretion data were analyzed using PCA. From the score plot, the pattern of distribution of data was studied and custom pattern for food compounds causing inflammation was determined along with patterns exhibiting apoptosis or necrosis presentation in the RAW 264.7 macrophages.

Results And Discussions
Data dimensions were reduced and information was retained by replacing the original correlated variables the systemic variation of the uncorrelated principal components (PCs). The link between the original correlated variables and the uncorrected PCs data were described by correlation circle. The correlation depends on the distance of a variable to the axis and circle in which closeness shows higher correlation to corresponding PC.
In interpreting the chemokine and cytokine data, adapted PCA segregated those combinations of actions leading to inflammatory and non-inflammatory consequences by using multivariate projection methods. The data on the score plot was separated into four quadrants with suggested four distinguished characteristics: (i) cell non-viability and inflammation suggests necrosis; (ii) cell non-viability without inflammation suggests apoptosis; (iii) cell viability without inflammation suggests cell proliferation; and (iv) cell viability with inflammation suggests cell repair upon injury ( Figure 2).
The projection of data showed that the further data from the graph origin are mostly from biogenic amines with concentrations of 10 and 100 µg/ml [ Figure 3 The loading plots for all data, 1 µg/ml, 10 µg/ml and 100 µg/ml biogenic amines were similar with NO and TNF-α on the same side of y-axis depicting positive correlation ( Figure   4). On the other hand, NO and TNF-α were on the opposite side of y-axis with cell proliferation (cell viability increment) depicting negative correlation with cell proliferation.
The results were in parallel with previous studies where both NO and TNF-α levels are reported to be in high concentration during inflammation while cell proliferation was low during inflammation. The positive control for inflammation in the score plot ( Figure 3 (Table 3). Q2VX showed the total variation of X which is predicted by the respective component estimated through cross-validation. Upon performing PCA with cross validation, Q2 and the limit are obtained ( Table 3). The limit, which was calculated depending on the number of components for PCA, showed the value of 0.502-0.505 for all data. The limit setting increases the account for loss in degrees of freedom by subsequent component. Q2V values are not equal to Q2 values for individual variables because the eigenvalues are < 1.5. Q2V (cum) values were more than 0.5, indicating that the models for the variables are good. All R2VX (Table 3) values are > 0.9, suggesting goodness of fit since it is correlated to the multiple correlation coefficient.
N3 significance level (Table 3) means that the component is not significant because it lacks degrees of freedom. Improvement can be made by addition of N numbers of data and K numbers of variables such as interleukin-12, prostaglandin (PGE 2 ) and phospholipase (PLA 2 ) which has their own roles in progression of inflammation. Huge variances may come from cell viability data which comprised of triplicates measurement from seven batches of cells. Differences of measurement exist between batches of macrophage cells but they are more representative by triangulation of data.
PCA correlation quadrants depicted positive correlation between the secretion of TNF-α and NO ( Figure 4). These results are found to be parallel with the findings which had identified them as mediators of inflammation [18]. In contrary, TNF-α and NO secretion were negatively correlated with cell proliferation, in which they are parallel with previous findings where NO was shown to inhibit T-cell proliferation [11] and TNF-α alpha showed in vitro anti-proliferative effects in normal and transformed cells [16].
These results showed similar positive correlations between NO and other cytokines such as TNF-α and PGE 2 [8,9]. In the score plots (Figure 3), the mapping of controls and biogenic amines and keropok lekor extracts were done in the axes spanned by the first two principal components, namely PC1 and PC2. PC1 includes variables with the largest variation while PC2 includes variables with second largest variation and it is orthogonal to PC1.
PCA for keropok lekor extracts (K) and standard mixture mimicked biogenic amines content in keropok lekor (BK) was plotted on the score plot in the same quadrant as the negative control for inflammation (Neg). These results showed that K and BK possessed strong correlation with L-NAME as negative control (Neg) and did not exhibit inflammationinducing properties in RAW 264.7 cell culture. The loading plot showed that NO and TNF-α were at the same side of y-axis exhibiting positive correlation while both NO and TNF-α showed negative correlations with cell viability increment. Although the toxicity of histamine, putrescine and cadaverine had been extensively studied, their combined effects remain unclear. From the PCA results, it was shown that the BA standard mixture (BK) and keropok lekor extract (K) did not induce pro-inflammatory mediator secretion which may be due to their total BA concentrations not exceeding 100 mg/kg food, which is the maximum permissible concentration of histamine in fish and fish products [1, 3].
Keropok lekor extract and its BA mixture solution did not induce inflammation in RAW 264.7 macrophages. NO and TNF-α showed positive correlations with each other while both NO and TNF-α showed negative correlations with cell viability.
Food intolerance and food toxicity caused by ingesting biogenic amines may pose risk to bowel diseases in susceptible consumers. In various bowel diseases such as IBD, IBS and Crohn's disease, mechanism is believed to be multifaceted including cytotoxicity, proinflammatory mediator secretion and free radical formation [2,4]. Various chemokines and cytokines have been associated with numerous pathophysiological conditions which are related to the induction of inflammation [5].
In this study, in interpreting the chemokine and cytokine data, PCA was adapted, segregating those combinations of actions leading to inflammatory and non-inflammatory consequences by using multivariate projection methods. The data was analyzed by determining the clusters of chemokine and cytokine that discriminate the potential inflammatory properties upon exposure to biogenic amines and keropok lekor extracts.

Sample aggregation
After thawing, the keropok lekor samples were weighed. The utensils and laboratory apparatus used for sample preparation were swabbed with 70% ethanol. RKL was cut into small pieces with scissors into a plastic basin and approximately 300 to 400 g of sample was minced to coarse powder using the heavy-duty blender (Waring; U.S. A.). Mincing was repeated for all cut samples. The minced sample was made into a circle-shaped heap and it was then divided into four equal parts. Of these, only two parts positioned diagonally were taken and mixed homogenously. It was shaped into a circle and divided again into four equal parts. Following this, two parts of the samples positioned diagonally were taken and mixed homogenously. This final mixture was divided into two parts and kept in two different polystyrene plastic containers. These homogenous samples were then used for further analysis. The same procedures were repeated for the SKL.

Preparation of keropok lekor extracts
Two types of keropok lekor, namely RKL and SKL were sampled. The differences between them are RKL is moulded as long small rod with diameter of 2 cm and length of approximately 30 cm and sold in rods and SKL is moulded as long big rod but kept chilled and sliced into 1-2 mm thick slices upon hardening and sold as slices. Ten grams of minced RKL and minced SKL were weighed and put inside a sterile stomacher bag. Six

Preparation of BA standard mixture
The standard mixture mimicking the biogenic amines contents in keropok lekor extracts were prepared according to the results obtained using LC-MS (Table 2).
Exposure study for determination of cell viability and pro-inflammatory mediator secretion The cells were treated with test compounds (BA individual solutions, keropok lekor extracts, and BA standard mixture solutions mimicking the BA contents in each keropok lekor extract (Table 2) at 37 o C with humidified 5% carbon dioxide overnight. Aliquots of 150 µl supernatant were collected for determination of nitric oxide and TNF-α secretion.
Cell viability test was done on the cells in the remaining media consecutively using MTS assay.
Aliquots of 20 µl of CellTiter 96® AQueousOne Solution Reagent were added into each well and incubated for 20 min at 37 o C. Absorbance was measured at 492 nm using a microplate reader. Cell viability was determined as percentage of viable cells.

Figure 2
Suggested derivation of summary from Principal Component Analysis (PCA) score plot from the study.  Loading plot for (A) 1 µg/ml single biogenic amines, (B) 10 µg/ml single biogenic amines and (C) 100 µg/ml single biogenic amines.