**Biogenic Amines and Food Safety**

Editors

**Maria Martuscelli Dino Mastrocola**

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin

*Editors* Maria Martuscelli University of the Studies of Teramo Italy

Dino Mastrocola University of the Studies of Teramo Italy

*Editorial Office* MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

This is a reprint of articles from the Special Issue published online in the open access journal *Foods* (ISSN 2304-8158) (available at: https://www.mdpi.com/journal/foods/special issues/biogenic amines).

For citation purposes, cite each article independently as indicated on the article page online and as indicated below:

LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. *Journal Name* **Year**, *Volume Number*, Page Range.

**ISBN 978-3-0365-0636-4 (Hbk) ISBN 978-3-0365-0637-1 (PDF)**

© 2021 by the authors. Articles in this book are Open Access and distributed under the Creative Commons Attribution (CC BY) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications.

The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

## **Contents**



## **About the Editors**

**Maria Martuscelli** was born in Muro Lucano (PZ) in 1970. She is married and has two daughters. Maria Martuscelli is Associate Professor of Food Technology in the Faculty of Bioscience and Technology for Food, Agriculture, and Environment of University of the Studies of Teramo (Italy). She is a member of the Italian Society of Food Technology and entered in the Register of Experts in Industrial Research. Many outputs of her research are carried out in scientific projects which are supported by public institutions or private companies. Her research activity has seen her contribution in foreign institutions such as the 'Unite Mixte de Recherche et Flaveur Vision du consommateu comportment, of the' Institut National de la Recherche Agronomique, in Dijon (France); moreover, she was involved in data collection for the European Food Safety Autority, on "biogenic amines in food". She is the author of 96 scientific works, published in international and national journals, proceedings of national and international Congresses and chapters of books.

**Dino Mastrocola** was born in Guardiagrele (CH) in 1958. He is married and has a son. Dino Mastrocola is full professor of Food Technology in the Faculty of Bioscience and Technology for Food, Agriculture, and Environment of University of the Studies of Teramo (Italy) and Rector of the same institution since November 2018. He carries out research activities in the field of Food Science and Technology and is involved in national and international research projects. He participated in the Food Improvement Project in China, coordinated by prof. Antonino Zichichi at the International Centre for Scientific Culture - World Laboratory in Lausanne. He carried out a strict research activity in the R&D of Unilever Research in Colworth House (England) and at the Academy of Sciences in Beijing. From 1996 to 2003, he took part in the European food science dissemination project Flair Flow Europe. He is the author of 20 essays and over 180 scientific publications, and is also a founding member of the Italian Society of Food Sciences and Technologies.

## *Editorial* **Biogenic Amines' Content in Safe and Quality Food**

**Maria Martuscelli \*, Luigi Esposito and Dino Mastrocola**

Faculty of Bioscience and Technology for Food, Agriculture and Environment, University of Teramo, Via R. Balzarini 1, 64100 Teramo, Italy; lesposito2@unite.it (L.E.); dmastrocola@unite.it (D.M.) **\*** Correspondence: mmartuscelli@unite.it

Biogenic amines (BAs) are low-molecular-weight, nitrogenous compounds (mainly polar bases) coming from the decarboxylation of free amino acids or by amination or transamination of aldehydes and ketones. To our knowledge, BAs are essential for cellular development and growth, are important regulators of several processes such as brain activity, regulation of body temperature, stomach pH, gastric acid secretion, the immune response, and the synthesis of hormones and alkaloids, among others [1]. Decarboxylation of free amino acids represents the primary way of BAs' obtention. Microorganisms involved in this process are positive to the decarboxylase enzyme, with the pathways that seem to be strain dependent rather than species specific [2]. At any rate, the presence of proteins (amino acids), favorable growing and fermenting conditions, and the possibility of external contaminations during food processing are important factors in BAs' increase. An important contribution is also given by several pro-technological strains, in particular lactic acid bacteria (LAB) from the genera *Lactobacillus*, *Leuconostoc*, *Lactococcus*, *Enterococcus*, and *Streptococcus,* were recently deeply reviewed as they are high tyramine producers. Del Rio et al. [3] clarified the harmful effect of this amine in boosting histamine toxicity besides being responsible for the so-called "cheese reaction". Although starters are generally considered secure and good for both food safety and the general health status of the human body, there does not exist any regulation looking at the decarboxylase positivity of bacteria. As a matter of fact, it is challenging to use BAs' content in food as a unit of measure to establish food safety. Evidence of strict correlations between personal sensitivity and genetical predisposition for BAs' intoxication was found. In particular, the compromising of the detoxification system was enacted by mono and di-amine oxidase (MAO and DAO) enzymes in the intestinal epithelium that change for every individual. Great attention should be reserved not only to those subjects consuming mono and di-amino oxidase inhibitors (MAOI and DAOI) drugs, as they may become particularly sensitive to BAs' action, but should also include those experiencing any impairment in the functioning of the small intestine or kidneys and so, even coeliac subjects, people who suffered surgery, or those who are in treatment for cancer and other pathologies [4,5]. The scientific research is giving growing insights into BAs' presence in all food matrices including fresh fruit and vegetables, pulses, baby foods, alcoholic beverage [6–8], and halal foods [9]. This scenario forces scientists to turn their attention to the fact that all the population is at risk for experiencing BAs' accumulation by their choices in meal composition, food sources, and of course specific sensitivity. This editorial has collected papers giving an interesting outlook on the content of BAs' in food and a possible strategy to reduce their occurrence, BAs' role in the promotion of aroma, and the specific capacity of selected bacteria in promoting their accumulation and/or degradation. All these papers actively contribute to creating a more complete frame on the theme keeping constant the fact that the presence of BAs' in food represents an essential part of food quality and food safety.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

**Citation:** Martuscelli, M.; Esposito, L.; Mastrocola, D. Biogenic Amines' Content in Safe and Quality Food. *Foods* **2021**, *10*, 100. https://doi.org/ 10.3390/foods10010100

Received: 21 December 2020 Accepted: 2 January 2021 Published: 6 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **References**


*Article*

## **Extensive Chemometric Investigations of Distinctive Patterns and Levels of Biogenic Amines in Fermented Foods: Human Health Implications**

## **Martin Grootveld 1,\*, Benita C. Percival <sup>1</sup> and Jie Zhang <sup>2</sup>**


Received: 19 October 2020; Accepted: 27 November 2020; Published: 5 December 2020

**Abstract:** Although biogenic amines (BAs) present in fermented foods exert important health-promoting and physiological function support roles, their excessive ingestion can give rise to deleterious toxicological effects. Therefore, here we have screened the BA contents and supporting food quality indices of a series of fermented food products using a multianalyte-chemometrics strategy. A liquid chromatographic triple quadrupole mass spectrometric (LC-MS/MS) technique was utilized for the simultaneous multicomponent analysis of 8 different BAs, and titratable acidity, pH, total lipid content, and thiobarbituric acid-reactive substances (TBARS) values were also determined. Rigorous univariate and multivariate (MV) chemometric data analysis strategies were employed to evaluate results acquired. Almost all foods analyzed had individual and total BA contents that were within recommended limits. The chemometrics methods applied were useful for recognizing characteristic patterns of BA analytes and food quality measures between some fermented food classes, and for assessing their inter-relationships and potential metabolic sources. MV analysis of constant sum-normalized BA profile data demonstrated characteristic signatures for cheese (cadaverine only), fermented cod liver oil (2-phenylethylamine, tyramine, and tryptamine), and wine/vinegar products (putrescine, spermidine, and spermine). In conclusion, this LC-MS/MS-linked chemometrics approach was valuable for (1) contrasting and distinguishing BA catabolite signatures between differing fermented foods, and (2) exploring and evaluating the health benefits and/or possible adverse public health risks of such products.

**Keywords:** biogenic amines (BAs); fermented foods; chemometrics; multivariate (MV) statistical analysis; liquid chromatographic triple quadrupole mass spectrometric (LC-MS/MS) analysis; public health; lipid peroxidation; antioxidants

#### **1. Introduction**

Biogenic amines (BAs) may be biosynthesized and degraded via normal metabolic activities in animals, plants, and micro-organisms. As such, these amines occur in a wide variety of foods, such as fish, meat, and cheese products, and especially in fermented foods such as wines, and yoghurts, etc. [1–3]. BA formation in foods usually occurs via the decarboxylation of amino acids [3], of which there are rich sources in these matrices; for example, amino acids are present at very high levels in grapes, and comprise ca. 30–40% of the total nitrogen content of wines [1–3].

Metabolic pathways available in lactic acid bacteria, which have the ability to grow and thrive in foods and beverages, generate significant levels of BAs. Routes available for this are the enzymatic production of putrescine from ornithine (catalyzed by ornithine decarboxylase) and/or from arginine via agmatine, a scheme involving prior conversion of the amino acid substrate to agmatine with arginine decarboxylase, followed by transformation of agmatine to N-carbamoylputrescine via the action of agmatine imino-hydroxylase, and then on to putrescine (a second route for its generation involves the conversion of arginine to ornithine and then to this product via the above ornithine decarboxylase-catalyzed route); putrescine to spermine, a process involving the enzyme spermine synthase, and then spermine to spermidine via the actions of spermidine synthase; cadaverine from lysine with lysine carboxylase and a pyridoxal phosphate co-factor; 2-phenylethylamine from phenylalanine catalyzed by aromatic amino acid carboxylases, including tyrosine decarboxylase; tyramine from tyrosine via tyrosine decarboxylase action; histamine from histidine with histidine decarboxylase; tryptamine from tryptophan with trypotophan decarboxylase, another pyridoxal phosphate-dependent enzyme; and trimethylamine from trimethylamine-N-oxide with a trimethylamine-N-oxide reductase (enzymes involved in the conversion of amino acids to BAs are classified as decarboxylase deaminases) [4,5]. BAs may also be biosynthesized from the amination and transamination of aldehydes and ketones [5], and this may be of some relevance to their detection in marine oil products which have been allowed to autoxidize. Indeed, a range of aldehyde species arise from the fragmentation of conjugated hydroperoxydienes, which are lipid oxidation products resulting from the peroxidation of polyunsaturated fatty acids (PUFAs) [6].

Overall, microbial sources of BAs include yeasts, as well as gram-positive and -negative bacteria [7]. The physiological activity of BA synthesis in prokaryotic cells predominantly appears to be associated with bacterial defense mechanisms employed to combat environmental acidity [8–10]. Hence, amino acid decarboxylation in this manner enhances survival under harsh acidic stress states [9] via proton consumption, and amine and CO2 excretion required to facilitate restorations of internal pH values [11].

As with their biosynthesis, the catabolism of BAs is extensively outlined and reviewed in [5]. In view of their potentially toxic nature, fortunately humans have detoxification enzyme systems which catabolically oxidize BAs in vivo. These enzymes principally comprise monoamine and diamine oxidases (MAOs and DAOs respectively). MAOs are flavoproteins acting by the oxidative deamination of BAs to their corresponding aldehydes, along with hydrogen peroxide (H2O2) and ammonia. Two different forms of MAO have been identified in humans [5]. DAOs are responsible for histamine catabolism, as is histamine-N-methyltransferase, the latter catalyzing a ring methylation process [5].

Evidence available indicates that BAs may confer a series of human health benefits, which involve their interactions with a wide variety of intracellular macromolecules such as proteins, DNA, and RNA. Indeed, monoamines are typically precursors of neuromodulators and neurotransmitters [12]. Moreover, evidence is accumulating that the polyamines spermine and spermidine are important for sexual function and fertility [13], and polyamines in general are associated with cell growth and differentiation, including protein biosynthesis [14]. Indeed, the generation of BAs in eukaryotic cells is essential, since they are required for the critical biosynthesis of hormones, alkaloids, proteins, and nucleic acids [15]. One further plausible health benefit offered by both monoamine and polyamine forms of BAs is their antioxidant potential [16], and recent studies have shown that they function efficiently in this context, and protect against adverse unsaturated fatty acid peroxidation reactions when present in or supplemented to culinary oils, and other foods rich in PUFAs [6] (details regarding the nature and mechanisms of these antioxidant actions are provided in Section S1 of the Supplementary Materials).

Notwithstanding, the availability of these amines in the diet has not been without its problems. Indeed, adverse toxicological events may be stimulated by the ingestion of foods which are known to provide high concentrations of these agents, and one notable example is the provocation of deleterious hypertensive events in patients receiving therapies with monoamine oxidase inhibitor (MAOI) drug treatments [17]. A further problem is the depression of histamine oxidation, a process which arises from the ingestion of putrescine and agmatine, which serve as potentiators of this process; this promotes histamine toxicity episodes in humans [18]. Moreover, it has been reported that BAs such as putrescine and agmatine give rise to their corresponding carcinogenic nitrosoamines from reactions with nitrite anion, dietary or in vivo [19].

Human sensitivity to BAs is contingent on the availability and activities of detoxifying enzymes featured in BA metabolism, i.e., specific ones such as histamine methyltransferase, and those less specific such as mono- and diamine oxidases. However, since these enzymes are inhibited by different classes of drugs, including neuromuscular blocking agents such as alcuronium, antidepressants [20], and ethanol [21], the accumulation of BAs by the consumption of selected foods and beverages can, at least in principal, give rise to clinical disorders, including the extremely hazardous serotonin syndrome [22]. Further details regarding the adverse health effects associated with the excessive intake of BAs are delineated in Section S2 of the Supplementary Materials.

Current consumer demands for safer and healthier foods has prompted a high level of research investigations focused on BAs, although it should be noted that further studies are required to expand this area. High levels of BAs can build up in fermented foods, including fish, fish sauce, and cheese products. Their biosynthesis and accumulation therein are critically dependent on the availability of bacteria with decarboxylase-deaminase enzyme activities, environmental conditions that are unrestrictive towards their growth and propagation, and the efficient functioning of BA-generating enzymes, together with the presence of sufficient amounts of the relevant amino acid substrates required.

Hence, supporting analytical methodologies for the identification and measurement of BAs are of much importance to the food industry, and also from a public health perspective. Such methods should ideally offer high levels of reliability in order to monitor the potential health benefits offered by fermented food products, and also to circumvent any toxicological risks to consumers arising from their excessive production therein; realistic estimates of their human consumption are also major factors for consideration. To date, BA determinations in foods have represented a major challenge for analytical chemists in view of their non-chromophoric nature, their natural occurrence in complex multicomponent food and biological matrices, and high polarities, factors which are further complicated by a requirement for high analytical sensitivity, potential interferences, and, where relevant, chromatographic separation/resolution issues arising from the presence of many structurally-related agents in samples requiring such analysis [23]. Methods previously available for this purpose, and those for the screening of BA-producing bacteria, are outlined in Section S3 of the Supplementary Materials.

Notwithstanding, in principle, the simultaneous and direct multicomponent determination of BAs by the LC-MS/MS method described here, or a newly-developed strategy focused on largely non-invasive high-resolution proton (1H) nuclear magnetic resonance (NMR) analysis [6], serve as valuable assets which, in combination with MV chemometrics strategies, may be employed for the recognition of patterns of these bacterial catabolites which are characteristic of differential bacterial sources of these agents.

Multivariate (MV) data analysis of multicomponent analytical datasets serves as an extremely powerful means of probing and tracking metabolic signatures that are characteristic of differential groups or classifications of samples, and when applied to explore the biochemical basis of human disease etiology, this technique is commonly known as metabolomics [24]. Indeed, to date this combination of multianalyte-MV analysis has been copiously utilized in many biomedical and clinical investigations, mainly for the identification of diagnostic or prognostic monitoring biomarkers for human diseases. However, when applied in a non-biomedical context, the technique can best be described as chemometrics, a technology which also commonly employs many of the MV data analysis strategies used in metabolomics experiments.

In view of the rich sources of BAs in fermented food products, in this study we determined the contents of a total of 8 different BAs in a series of commercially-available fermented fish, fish sauce/paste, vegetable sauce, cheese, wine/vinegar, and cod liver oil (FCLO) products. For this purpose, we employed both univariate and MV chemometrics analysis techniques in order to recognize differential patterns of these catabolites, which may be representative or characteristic of their food, bacterial, metabolic pathway, and/or food processing technology sources. Such analytical information also serves to furnish us with valuable information regarding the provision of these important nutrients in the human diet, and to evaluate the toxicological/adverse health risks presented by the ingestion of fermented foods containing portentously excessive levels of these agents. Currently, a total BA content of *ca*. 1000 ppm is linked to toxicity, and in recommended manufacturing practices, 100 ppm histamine, or a total BA content of 200 ppm, are considered acceptable levels which do not give rise to any associated adverse health effects [25].

These studies were supported by the consideration of further food quality determinations on these fermented food products, which consisted of pH values, titratable acidities (TAs), and total lipid contents, along with an adapted method for determining lipid peroxidation status (thiobarbituric acid-reactive substances (TBARS)).

With the exception of a small number of studies focused on BAs detectable in selected wine products, e.g., [26], to the best of our knowledge this is the first time that MV chemometrics techniques have been applied to explore potentially valuable "between-food classification: distinctions between the concentrations and patterns of BAs in a series of different food products, albeit fermented ones. Therefore, the aims of this investigation are to explore the abilities and reliabilities of LC-MS/MS-based chemometrics analysis techniques to: (1) evaluate the possible public health benefits and/or risks of BAs arising from the human consumption of fermented foods; and (2) effectively compare and distinguish between differing patterns of BA molecules in different classes of fermented food products.

#### **2. Materials and Methods**

#### *2.1. Fermented Food Products*

Fermented food products (cheese, fish, fish sauce/paste, vegetable sauce, and wine/vinegar classifications) were randomly selected and purchased from a variety of US retail outlets based in the state of Nebraska. These comprised *n* = 4 fish samples, *n* = 9 fish sauce/paste samples, *n* = 4 vegetable sauce samples, *n* = 5 cheeses, and *n* = 4 wine/vinegar samples (Table 1). Details of the fermentation processes employed by the manufacturers involved were unavailable. Prior to analysis, all samples were stored in a darkened freezer at a temperature of −20 ◦C for a maximal duration of 72 h.


**Table 1.** Details of fermented food products investigated for each classification.

Fermented cod liver oil (FCLO) was a natural product that was manufactured and kindly donated by Green Pastures LLC, 416 E. Fremont O'Neill, NE 68763, USA for this study. Separate batches (*n* = 10) of this FCLO product were randomly selected by independent visitors to its manufacturing site throughout a 6-month period, as noted in [6].

FCLO products were prepared from the fermentation of Pacific cod livers. Livers were frozen (−20 ◦C) within 40 min following their harvest from the Pacific Ocean, and then transported to a preparation facility whilst remaining in the frozen state. Fermented CLO was produced from these cod liver sources using a novel and proprietary fermentation technology. Briefly, cod livers were loaded into a fermentation tank, and both salt and the fermentation starter agent were added to induce the process. The tank was completely sealed during the fermentation and, following periods of 28–84 days, the raw FCLO product accumulated and was then isolated from the tank. Following fermentation, products were centrifuged, filtered to remove particulates, and then packed.

On arrival at the laboratory, FCLO product sample batches were de-identified through their transfer to coded but unlabeled universal storage containers. Each sample was subsequently stored in a darkened freezer at −80 ◦C until ready for analysis (predominantly within 24 h of their arrival).

#### *2.2. Analysis of BAs in Fermented Food Product Samples*

A liquid chromatographic triple quadrupole mass spectrometric (LC-MS/MS) technique was employed for the simultaneous analysis of up to 11 BAs in fermented food products using an adaption of the LC-MS/MS method reported in [27]. A Shimadzu 8045 LC-MS/MS facility was used for this purpose, the MS/MS detection system for the monitoring and molecular characterization of eluting BA analytes. Primarily, pre-set accurately weighed masses of food samples were shaken with a 20.0 mL volume of 70% (*v*/*v*) methanol/30% (*v*/*v*) water for 20 min, which were then centrifuged at 7000 rpm at 4 ◦C for another 20 min period. The clear supernatant was subsequently transferred to 1.7 mL volume amber auto-sampler vials for LC-MS/MS analysis. For wine/vinegar and FCLO samples, fixed aliquots were filtered using a 0.45 μm filter paper prior to the above methanol/water extraction stage.

The LC facility comprised a pump, vacuum degasser, auto-sampler, and column compartment, and finally a secondary variable wavelength spectrophotometric detection system was used for these analyses. This system could operate up to 800 bar. The internal standard (IS) utilized was tetra-deuterated histamine (histamine-α,α,β,β−d4, (2HCl)), which was purchased from C/D/N Isotopes Inc. (Pointe-Claire, Quebec, Canada). IS *m*/*z* values employed for quantification purposes were 116.1 and 99.0 for precursor and product ions, respectively (112.1 and 95.1 respectively for undeuterated histamine).

A 3-μm 50 <sup>×</sup> 2.1 mm Pinnacle® DB pentaflurophenyl (PFP) base with propyl spacer column was employed for optimal BA analysis. Mobile phase 1 contained water solutions of the ion-pair reagent trifluoroacetic acid (TFA) (either 0.05 or 0.10% (*w*/*v*)), and mobile phase 2 was acetonitrile containing equivalent TFA concentrations. BA analytes were monitored in positive ion mode for the MS/MS detection system. Reporting limit values for fermented food samples were 1 ppm for all BAs determined.

Authentic BA calibration standards were purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO, USA) (histamine, H7125; cadaverine, 33220; putrescine, D13208; 2-phenylethylamine, P6513; spermidine, 85578; tyramine, T2879; tryptamine, 193747), and Alfa Aesar Inc. (Heysham, UK) (spermine, J63060). BA contents were determined from calibration curves developed with standard solutions of concentrations 0.5, 1.0, 10.0, 50.00, 100.0, 200.0, and 400.0 ppb for each BA.

#### *2.3. Total Lipid Analysis*

Total lipid (fat) analysis was performed according to the AOAC 922.06 method. Briefly, homogenized samples were treated with HCl, and then washed at least two-fold with both petroleum ether and diethyl ether; solutions arising therefrom were then placed in pre-weighed beaker containers. Subsequently, the lipid-containing ether solutions were evaporated, and the (*w*/*w*) % content of lipid was determined directly from the weight gain of the container.

#### *2.4. Determination of Thiobarbituric Acid-Reactive Substances (TBARS) Values*

Primarily, accurately-weighed samples were digested with perchloric acid (HClO4), and subsequently the resulting clear filtered supernatant solution was reacted with thiobarbituric acid (TBA) for a period of 15–18 h at 27.5 ◦C according to the method outlined in [28]. The absorbance value at a wavelength of 532 nm was then determined, and TBA-reactive substance (TBARS) values were

reported as mg/kg (ppm) units following their quantification from a calibration curve developed with MDA standards.

#### *2.5. Titratable Acidity (TA) and pH Value Determinations*

Titratable acidity values were determined using the AOAC 947.05 method [29], and pH measurements were made using a modified FO PROC 31 protocol which is based on the USDA PHM method. The latter approach is based on the formation of a homogenized food/water slurry which was allowed to stand prior to pH determination with a probe.

#### *2.6. Experimental Design and Statistical Analysis*

#### 2.6.1. Univariate Statistical Analysis

The experimental design for univariate analysis of the individual BA, TA, pH, and further variable dataset involved an analysis-of-variance (ANOVA) model, which incorporated 1 prime factor and 2 sources of variation: (1) that "between-fermented food classifications", a qualitative fixed effect (FF*i*); and (2) experimental error (e*ij*). The mathematical model for this experimental design is shown in equation 1, in which y*ij* represents the (univariate) BA or alternative analyte dependent variable values observed, and μ their overall population mean values in the absence of any significant, influential sources of variation.

$$\mathbf{y}\_{\ddot{\imath}\ddot{\jmath}} = \mu + \mathbf{F} \mathbf{F}\_{\dot{\imath}} + \mathbf{e}\_{\dddot{\jmath}} \tag{1}$$

ANOVA was conducted with *XLSTAT2016* and *2020* software. Datasets were autoscaled (i.e., the mean value of each parameter monitored was subtracted from each entry, and the residual then divided by food class standard deviation, which was computed with an (*n* − 1) divisor) prior to analysis. In view of heterogeneities between the intra-sample variances of fermented food classifications, i.e., heteroscedasticities, the robust Welch test was employed to determine statistical significance of differences observed between the mean BA and other food quality variable values for each fermented food group. *post*-*hoc* ANOVA evaluations of the statistical significance of differences between the mean values of individual fermented food groups were performed using the Bonferroni test.

A similar ANOVA-based experimental design was applied to additional design models selected to determine the statistical significance and food class specificities of BA analytes only. For these purposes, the 8 BA dataset, which included those determined in the *n* = 10 batches of the FCLO product, was either constant sum (CS)-normalized or not, and then generalized logarithmically (glog)-transformed, and finally autoscaled prior to analysis. The CS normalization data preparation task was applied in order to evaluate the significance of fermented food classification-dependent BA profile patterns. The non-CS-normalized dataset also included total BA level as a further possible explanatory variable. *MetaboAnalyst 4.0* (University of Alberta and National Research Council, National Institute for Nanotechnology (NINT), Edmonton, AB, Canada) was utilized for the analysis of these data. Probability values obtained from *post*-*hoc* ANOVA comparisons of individual BA levels between fermented food classes were false discovery rate (FDR)-corrected.

Tests for the heteroscedasticity of ANOVA model residuals (Levene's test) were performed using *XLSTAT2020* (Addinsoft, Paris, France).

#### 2.6.2. Multivariate Chemometrics and Algorithmic Computational Intelligence (CI) Analyses

Principal component analysis (PCA), partial least squares-discriminatory analysis (PLS-DA), correlation, and agglomerative hierarchical clustering (AHC) analyses of the combined BA dataset were performed using *XLSTAT2016* and *2020* and *MetaboAnalyst 4.0* [30] software module options. The dataset was generalized glog-transformed, and autoscaled prior *to MetaboAnalyst 4.0* analysis, but only autoscaled for *XLSTAT2016* and *2020* analyses. All these MV analysis strategies were primarily performed on non-CS-normalized data. For the PCA and PLS-DA analyses, limits for significant explanatory variable loadings vectors/coefficients were set at ≤−0.40 or ≥0.40. Validation of PLS-DA

models was performed by determining component number-dependent Q2 values (predominantly for two classification comparisons), and permutation testing with 2000 permutations. The significance of variable contributions to these models was determined by the computation of variable importance parameter (VIP) values (values >0.90 were considered significant).

Additional PCA analysis was performed in order to explore associations or independencies of individual BAs and other active variables considered, e.g., pH and TA values, total lipid contents, etc. For this purpose, a maximal 5 PC limit was applied, and PCA was then conducted on autoscaled data using varimax rotation and Kaiser normalization. The loadings of each analytical variable on successive orthogonal PCs was then sequentially evaluated. Similarly, this form of PCA was employed to investigate possible inter-relationships and orthogonalities between BA variables analyzed in FCLO batches sampled from the same manufacturing source specified above.

A further PCA model involved its application to the 8 BA dataset alone, which was either CS-normalized or not, glog-transformed, and autoscaled prior to analysis. As noted above, the CS-normalization data preparation step was utilized in order to evaluate the significance of any differential patterns or distributions of BA analytes which may be characteristic of fermented food classifications. This analysis was performed using *MetaboAnalyst 4.0*.

The random forest (RF) machine-learning algorithm approach was also utilized for classification and discriminatory variable selection purposes (*MetaboAnalyst 4.0* Random Forest module), with 1000 trees (*ntree*) and 4 predictors selected at each node (mtry) subsequent to tuning. The dataset was randomly split into training and test sets containing approximately two-thirds and one-third of entries respectively. The training set was employed to construct the RFs model, and an out-of-the-bag (OOB) error value was determined to evaluate the classification performance of this. Again, this analysis was performed on the glog-transformed and autoscaled dataset, either with or without prior CS-normalization as specified in the manuscript.

Missing data, specifically total lipid and (TBARS):(total lipid) ratios for 2 × fish sauce/paste, 1 × vegetable sauce, 1 × wine/vinegar, and 1 × cheese samples, were estimated by the support vector machine (SVM) impute technique [31] (*MetaboAnalyst 4.0*), or supplementation with the explanatory variable column mean values, along with a corresponding reduction in degrees of freedom available for parametric univariate statistical testing (*XLSTAT2016 or 2020*).

#### **3. Results and Discussion**

#### *3.1. BA Levels and Food Quality Indices in Fermented Food Products, and Univariate Analysis of These Analytical Data*

Mean ± SEM values for the individual and total BA contents of the FF products investigated are provided in Table 2. The major contributors towards the relatively high BA levels observed in fermented cheese samples were cadaverine (mean 60% of total) and tyramine (mean 21.5% of total). Although three of the cheese products analyzed had total BA concentrations of 30–63 ppm, two of them were found to be as high as 666 and 780 ppm, which were markedly above the recommended 200 ppm content limit. The ANOVA Welch test demonstrated that there were highly significant differences between these total BA values (Table 3), as expected (*p* = 2.84 <sup>×</sup> 10−4); such differences were largely explicable by those observed between the cheese and wine/vinegar product classifications investigated.

Hence, characteristic "markers" of fermented cheese samples appeared to be cadaverine and tyramine, which had contents markedly elevated over those of the other fermented food products evaluated, although there were very high intra-fermented food classification variances for these estimates.

**Table 2.** Biogenic amines (BA) contents and quality indices of fermented foods investigated. Mean ± SEM BA levels, and titratable acidity (TA), pH, total lipid, thiobarbituric acid-reactive substances (TBARS) and (TBARS):(total lipid) ratio values, for five classes of fermented food products (cheese, fish, fish sauce/paste, vegetable sauce, and wine/vinegar) purchased at a range of U.S. retail outlets (bracketed numbers represent the number of different products analyzed for each classification).


nd: not determined.

**Table 3.** Statistical significance and nature of differences between the mean BA contents and other food quality indices for fermented food products. Both robust Welch and Bonferroni-corrected *post*-*hoc* ANOVA test significance (*p*) values are provided. Abbreviations: ns, not statistically significant. \* These values were close to statistical significance, but did not attain a *p* value of ≤0.05 with the robust Welch test.



**Table 3.** *Cont.*

ns: not statistically significant.

Univariate statistical analysis performed by ANOVA (robust Welch test derivative), and also *post*-*hoc* Bonferroni test values, demonstrated that the mean values of each food classification examined were significantly or highly significantly different for 7 and 9 of the marker index variables respectively (*p* values ranging from <0.0003 to 0.04 for the former test, Table 3). Figure 1 shows a heatmap of the mean BA contents, and further variables included in this analysis; this clearly displays significantly higher tyramine, cadaverine, putrescine, and tryptamine levels in the fermented cheese products; higher histamine concentrations in the fish sauces/pastes explored, as expected (although vegetable sauces also had quite high levels of this BA); and also greater spermine contents in the fish paste/sauce products (*ca*. 1.5-fold greater than the mean value found for the fish classification, the next highest concentration). The vegetable sauce products had the highest mean spermidine levels, whereas the fermented fish group contained the largest amounts of 2-phenylethylamine detectable.

As expected, mean TA values were significantly greater for the wine/vinegar products than they were for all the other fermented food classes investigated, and correspondingly the mean pH value for the former group was significantly lower than those of all the other fermented foods. Of course, the mean total lipid content of the cheese group (23.3%) was significantly greater than all other food classifications tested (*p ca*. 10<sup>−</sup>3), although no significant differences were found for the secondary lipid peroxidation TBARS marker. However, an examination of the mean ratio of TBARS index to total lipid content revealed that this value was markedly greater for the wine/vinegar group than that of all other food product types (Bonferroni-corrected *post*-*hoc* ANOVA tests), and significantly so over that of the cheese samples analyzed, as might be expected in view of the very low fat contents of fermented wine/vinegar samples (for example, it varies from 0.15–0.44% (*w*/*v*) in Zhenjiang aromatic vinegar samples [32]), and potentially substantially inflated TBARS levels resulting from quite high levels of TBA-reactive acetaldehyde and acrolein, amongst other aldehydes, present in such fermented products [33–36]. Indeed, many other aldehydes are reactive towards the TBA reagent, and also form chromophoric products on reaction with it [28]. Estimates for acetaldehyde in vinegar products can be as high as 1.0 g/kg respectively [33], but such levels are highly variable, with much lower levels being found, e.g., 2.6 mg/L (*ca*. 60 μmol/L) [37].

**Figure 1.** Heatmap diagram displaying the nature, extent and ANOVA-based significance of univariate differences between mean values of all 8 BA and further chemoanalytical food quality variables (near right-hand side *y*-axis) for the fermented cheese (red), fish (green), fish sauce/paste (dark blue), vegetable sauce (pale blue), and wine/vinegar (mauve) products. The complete dataset was glog-transformed and autoscaled prior to analysis, but not CS-normalized. Transformed analyte intensities are shown in the far right-hand side *y*-axis: deep blue and red colorations represent extremes of low and high contents respectively. The left-hand side of the plot shows results arising from an associated agglomerative hierarchical clustering (AHC) analysis of these variables, which reveals two major analyte clusterings, with three sub-clusterings for one of these. The top right-hand side major cluster comprises TBARS level, (TBARS):(total lipid) ratio and TA value, whereas the second contains all other analyte variables, including all BA contents. The first, second, and third sub-clusters within the bottom right-hand side major cluster feature spermine, spermidine, and histamine (the first two of these arising from the same putrescine and metabolically upstream ornithine and agmatine/arginine sources respectively); tyramine, cadaverine, and total lipid; and 2-phenylethylamine, putrescine, tryptamine, and pH respectively.

Acetaldehyde, a volatile flavor component of a variety of foods and beverages such as cheese, yoghurt, and wines [34], represents one of the most abundant carbonyl compounds detectable in wine, and typically accounts for *ca*. 90% of the total aldehydes present; its concentrations therein usually range from 10 to 200 mg/L (predominantly, it is generated as a yeast by-product during alcoholic fermentation processes [35], or from the chemical oxidation of ethanol [36]). However, very high levels of the unsaturated aldehyde acrolein are also present in red wine products [33]. Furthermore, a wide range of further aldehydes have been found to serve as major flavor constituents of traditional Chinese rose vinegar, and these include aliphatic *n*-alkanals such as heptanal, hexanal, nonanal, and dodecanal (ranging from 6–147 μg/kg), with larger amounts of benzaldehyde (851 μg/kg) [38].

Hence, overall these data clearly demonstrated that, in a univariate context, there were indeed significant differences between the mean contents of BAs and further parameters considered for the five classes of fermented food products studied.

Prior to the performance of MV statistical analysis of the dataset acquired, simple Pearson correlations were explored between all explanatory variables considered, and Figure 2 shows a correlation heatmap for these relationships. Clearly, there were moderate to strong positive correlations observed between all fermented food BAs present, the strongest observed between 2-phenylethylamine and tyramine (both aromatic BAs), tryptamine and spermine, and most notably, between cadaverine and histamine. Food pH values were found to have the strongest positive correlations with tyramine > putrescine > tryptamine, although spermidine was predominantly uncorrelated with this index. Moreover, as anticipated, TA was strongly negatively correlated with pH value > putrescine > tyramine ≈ histamine contents in that order. TBARS level, however, was largely independent of all BAs and their concentrations, with the exception of spermidine, which exhibited a weak positive relationship with this variable. Similarly, total lipid level was also mainly uncorrelated with all BA contents but was quite strongly anti-correlated with (TBARS):(total lipid) ratio and non-lipid-normalized TBARS value (both expected). The (TBARS):(total lipid) ratio was either strongly or moderately anti-correlated with all BA levels, and this may provide an indication of their potential antioxidant functions. In view of the complexity of these inter-relationships, the MV PCA and PLS-DA techniques were employed to explore them further.

**Figure 2.** Correlation heatmap displaying positive and negative inter-relationships between BA concentrations, pH and TA values, total lipid contents, TBARS indices and (TBARS):(total lipid) ratios (TBARS/total lipid). The left-hand ordinate and top abscissa axes show AHC analysis based on these Pearson correlations (as a similarity criterion). From the top abscissa axis, of the two major clusterings revealed, that on the right-hand side contains all BA variable levels with the exception of spermidine, together with positively-correlated pH values, whereas the left-hand side one consists of all lipid- and lipid peroxidation-based variables, spermidine concentrations, and TA values.

#### *3.2. Principal Component Analysis (PCA) of the Multivariate Fermented Food Dataset*

PCA was primarily conducted in order to acquire an overview of the degree of distinctiveness between, i.e., clustering of, the fermented food classifications investigated, and also to identify any potential data outliers. An examination of two-dimensional (2D) scores plots from this analysis demonstrated that no significant outliers were detectable, and that PCs 1, 2, and 3 accounted for 41.5, 16.4, and 11.1% of the total variance respectively for the complete dataset which was glog-transformed and autoscaled. 2D and three-dimensional (3D) scores plots featuring these two most important PCs revealed that there was a reasonable level of distinction between the wine/vinegar and all other food product groups, and also between the cheese and fish classifications (Figure 3a); however, distinctions between the fish, fish sauce/paste, and vegetable sauce groups were not found, there being a significant degree of overlap between them. Notwithstanding, the sample sizes of the fermented fish and vegetable sauce groups involved were quite limited. A corresponding preliminary correlation circle diagram is shown in Figure 3b. Clear observations from this diagram are that (1) 2-phenylethylamine, tyramine, and cadaverine, and to a lesser extent, putrescine and tryptamine, are all correlated with PC1, and this observation indicates their communality in this model; (2) food pH values are also strongly correlated to PC1, and this indicates that higher values of this parameter may arise from the basicity of the above BAs (gas-phase primary amine basicity values increase with the length of its carbon chain substituents in view of their electron-donating positive charge-stabilizing effects—such values also increase with progression from primary to secondary to tertiary alkylamines [39]); (3) an at least partial correlation of histamine contents with PC2, which indicates distinction of this BA from those aligned with PC1; (4) an inverse correlation (anti-correlation) of total lipid level with the (TBARS):(total lipid) ratio index, as might be expected; and (5) a strong anti-correlation of TA value with BA levels, particularly tryptamine and putrescine, and this suggests that these amines serve to offer neutralization potential against acidic fermented food products. Also notable from this Figure are very strong correlations between the fermented food supplementary variable cheese and total lipid content, and between wine/vinegar and TA value, as indeed expected.

<sup>(</sup>**a**)

**Figure 3.** *Cont*.

**Figure 3.** (**a**) 3D PCA scores plot of PC3 vs. PC2 vs. PC1, showing some degrees of distinction between different fermented food classes, i.e., those of cheese, fish, fish sauce/pastes, vegetable pastes, and wines/vinegars (particularly that between the wine/vinegar classification and all others). (**b**) Preliminary correlation circle diagram displaying correlations between all explanatory variables considered, and PCs 1 and 2 in a PCA model applied to the complete autoscaled (standardized) dataset. Active variables are depicted in red, whereas two of the supplementary variable classifications (cheese and wine/vinegar) are shown in blue. Variance contributions for PC1 and PC2 are indicated.

(**b**)

A more detailed analysis of these PCA loadings was made with the application of varimax rotation, Kaiser normalization, and a maximal number of 5 PCs considered. For this model, such variable loadings, and the percentage of total variance accounted for by each PC are available in Table 4. This analysis revealed that cadaverine, tryptamine, 2-phenylethylamine, and tyramine all strongly and positively loaded on PC1, spermidine and histamine strongly and positively loaded on PC3 (along with a more minor contribution from 2-phenylethylamine), and putrescine and spermidine loaded strongly and positively on PC5, albeit also with histamine to a much lesser extent. Interestingly, all aromatic BAs strongly loaded on PC1, as observed above (Figure 3b), whereas spermidine and its metabolic precursor putrescine both co-loaded onto the same PC (PC5).



The TBARS secondary lipid oxidation index, along with its value normalized to total food lipid content, both loaded strongly and positively on PC2, as might be expected, although histamine also contributed somewhat towards this PC. Moreover, TA and pH values powerfully loaded on PC4 negatively and positively respectively, as would be expected from their anticipated negative correlation in fermented food products (putrescine also made a moderate positive contribution towards this component). Total lipid content was found to load significantly on PCs 1 and 5, positively and negatively so, respectively.

In a related study focused on PCA of both BAs and polyphenolics in Hungarian wines, Cosmos et al. [26] found that PC scores successfully clustered differential groups of these product classes, and that PC loadings vectors displayed significant patterns of BA and polyphenol levels. However, it should be noted that for this analysis, spermidine, and tyramine strongly loaded on PC1 (positively and negatively, respectively), agmatine and the sum total BA concentration loaded strongly and positively on PC2, spermine and cadaverine both strongly and negatively loaded on PC3, and that histamine loaded strongly and positively on PC4 alone. These associations between the BA analytes tested did not correspond to those found in the present study, although in the above MV analyses we elected not to include the total summed BA concentration value. Furthermore, our study also included the determinations of 2-phenylethylamine and putrescine, and not agmatine, but that reported in [26] monitored the latter BA but not 2-phenylethylamine and putrescine. However, as noted by the authors of [26], these PC loadings are only applicable to one region of Hungarian wine production, and their results will not be readily transferable to others, let alone other classes of fermented foods, especially in consideration of the often highly variable methods of fermentation, sources of fermentative micro-organisms, and conditions employed for these purposes. Notwithstanding, these researchers also concluded that in view of the loading patterns of BAs observed, it was unnecessary to measure all BA variables for quality assessments, and that only one per orthogonal PC was sufficient to provide acceptable levels of distinction between different sub-classes of such wines.

From this analysis, the unambiguously strong loadings vectors of the aromatic BAs 2-phenylethylamine and tyramine on PC1 provide evidence that they may indeed arise from the same biological and/or metabolic sources; however, this observation may also be rationalized by the natural production of tyrosine from phenylalanine, i.e., that involving the possible hydroxylation of the latter substrate to the former catalyzed by the enzyme phenylalanine hydroxylase (PAH) potentially available in fermentative lactobacilli employed for the production of fermented food products, followed by enzymatic transformation of the tyrosine product to tyramine by fermentative bacteria. To date, PAH is the only known aromatic amino acid hydroxylase found in bacteria [40].

The loadings of spermine and spermidine on different orthogonal PCs (PC3 and PC5, respectively) is not simply explicable, although the co-loading of spermidine's metabolic precursor putrescine on PC5 is consistent with them being featured in the same metabolic pathway. However, the co-loadings of BAs on differential PCs, particularly PC1, may reflect their engenderment from identical or related bacterial sources.

Notably, PC2 was dominated by powerful loading contributions from TBARS level and (TBARS):(total lipid) ratio (both positive), and PC4 by strong loadings from TA and pH values (negative and positive loadings vectors, respectively). These inter-relationships are, of course, expected, and are consistent with the data presented in Figure 3b. PC5 was retained in the model since it was the only one available which had a strong loading contribution from spermidine.

#### *3.3. Distinction of Fermented Food Classifications Using PLS-DA*

Similarly, PLS-DA of the dataset revealed an effective discrimination between the cheese and wine/vinegar classifications, although the fish, fish sauce/paste and vegetable sauce sample PC score datapoints were again unresolved; however, a visualized combination of these three fermented food classifications was at least partially resolved from the fermented cheese group (Figure 4). Permutation testing of the PLS-DA model confirmed its ability to distinguish between all the differing fermented food classifications evaluated (*p* = 0.022). For this model, key discriminatory variables were selected on the basis of their variable importance parameters (VIPs), and these were total lipid content (1.81) > cadaverine content (1.61) > (TBARS):(total lipid) ratio (1.36) > TA value (1.24) > histamine content (1.14) > 2-phenylethylamine content (0.78); data were glog-transformed and autoscaled prior to analysis. The top three discriminators largely arise from differential levels of lipids, cadaverine, and (TBARS):(total lipid) ratio between each of the fermented food groups, e.g., for the total lipids and cadaverine variables, the cheese content was significantly greater than that of all other fermented food groups, and for the above ratio, its value was significantly greater in the wine/vinegar group than it was in all other groups.

The quite strong distinctions observed between the cheese, wine/vinegar, and fish-fish sauce/paste-vegetable sauce composite products is readily explicable by significant or even substantial differences between the higher contents of cadaverine, tyramine, and, to a lesser extent, tryptamine in cheese, than those of the four other fermented food product classes. Further key discriminators are TA, pH, and total lipid contents, the latter of which is, of course, much higher in the cheese group.

**Figure 4.** (**a**,**b**). 2D and 3D PLS-DA scores plots (PC2 vs. PC1, and PC3 vs. PC2 vs. PC1, respectively) revealing strong distinctions between the cheese, wine/vinegar, and a considered combination of fish, fish sauce/paste and vegetable sauce fermented food groups ((**a**) also shows 95% confidence ellipses for each fermented food classification). Little or no distinction between the latter three fermented food groups were discernable using this MV analysis approach.

#### *3.4. RF Modelling of Fermented Food Classifications*

Application of the RF CI classification technique was found to be only partially successful for the classification of the different fermented food groups investigated. Using the models described in Section 2.6.2, this approach correctly classified 4/4 wine vinegar, 6/9 fish sauce/paste, and 3/5 cheeses, but 0/4 for both fish and vegetable sauce products.

#### *3.5. PCA of FCLO BAs*

The FCLO product considered was primarily investigated separately since only BA contents, and not parameters such as pH and TA were available for it. Moreover, its total lipid content is, of course, not far removed from a value of 100%, and therefore it would be inappropriate to test this index in the above MV analysis models (similarly, total lipid level-normalized TBARS values would also be inappropriate to test in these systems). However, it was possible to explore inter-relationships between FCLO BA concentrations and/or their orthogonality status using a rigorous PCA approach featuring varimax rotation and Kaiser normaliszation in order to maximize success with the assignment of individual BA variables to PCs.

Table 5 lists the BA contents of *n* = 10 FCLO product batches. The total concentrations of BAs in these samples was higher than the recommended "limit" of 200 ppm in only two out of ten batches of the samples tested, albeit marginally so (only 14 and 20% higher). Similarly, bioactive histamine was completely undetectable in this product. As noted in [6], all BAs monitored were completely undetectable in three other natural, albeit unfermented, CLO products included for comparative purposes. All BAs tested were found to be reasonably soluble in FCLO lipidic matrices, and also in 1/3 (*v*/*v*) diluted solutions of this product in deuterochloroform (C2HCl3), presumably as the uncharged species with their amine functions deprotonated (solubility in these media is expected to increase with increasing amine function substituent chain length and hydrophobicity).


**Table 5.** (BA concentrations (ppm) of *n* = 10 separate batches of a FCLO product. Total BA and corresponding mean ± SEM values are also provided. Histamine and spermine were undetectable in all samples analyzed.

PCA performed on the FCLO BA dataset revealed that cadaverine, putrescine, and tryptamine all loaded strongly and positively on the first of two automatically-selected PCs (PC1), whereas the aromatic BAs 2-phenylethylamine and tyramine loaded strongly and positively on the second (PC2), along with spermidine (Table 6). These data displayed some consistency with PC loading values obtained on the full fermented food dataset (Table 4), which had 2-phenylethylamine and tyramine both strongly loading on one PC (PC1). However, such levels will, of course, be critically dependent on the microbial fermentation sources, parameters employed for fermented food production, and production conditions for these processes.



#### *3.6. MV Chemometric Analysis of BA Data Only: Recognition of Fermented Food Class-Distinctive BA Patterns Using CS-Normalization*

Additionally, we conducted univariate and MV analyses of datasets which were restricted to the BA profiles only, but also included the *n* = 10 FCLO samples reported above. Additionally, these analyses were performed with and without application of constant sum (CS) normalization. The CS-normalized data format was employed in order to facilitate the recognition of fermented food class-specific BA patterns. For the non-CS-normalized format, the total BA content value was also included as an explanatory variable, as indeed it was in [26].

Firstly, ANOVA performed on the CS-normalized, glog-transformed, and autoscaled dataset found very highly significant, albeit FDR-corrected *p* values for three of the sum-proportionate mean BA concentration differences observed between the fermented food classifications explored in this manner. Notably, these differences were observed for cadaverine, 2-phenylethylamine, and tryptamine (Table 7), and *post*-*hoc* testing revealed that for cadaverine, the cheese products had significantly greater proportionate levels than three others, and for both 2-phenylethylamine and tryptamine, FCLO had significantly higher ones than all other products examined. These differences in CS-normalized values are readily visualizable in the form of an ANOVA-based heatmap (Figure 5a), which revealed characteristic BA signatures for three of the fermented food product classifications. Clearly, the cheese, FCLO, and wine/vinegar sampling groups have high proportionate levels of cadaverine, 2-phenylethylamine/tyramine/tryptamine (all aromatic BAs), and metabolic pathway-associated putrescine/spermidine/spermine, respectively. However, when evaluated in this univariate system, "between-fermented food class" mean differences observed for putrescine, spermine, spermidine, histamine, and tyramine were not found to be statistically significant.

Secondly, both PCA and PLS-DA models were employed, and these approaches were successful in providing evidence for the MV distinctiveness of the FCLO, cheese, and wine/vinegar groups; however, as noted for the analyses conducted on the combined BA/further food quality parameter dataset, unfortunately no distinctions were observed between the fermented fish, fish sauce/paste, and vegetable sauce products (Figure 5b,c).

For the CS-normalized dataset (without total BA concentrations as an additional variable), PLS-DA variable importance parameter (VIP) values were in the order spermidine (1.48) > putrescine (1.34) > spermine (1.20) > histamine (1.06) > 2-phenylethylamine (0.94), whereas those for the non-CS-normalized dataset were spermidine (1.56) > spermine (1.35) > 2-phenylethylamine (1.32) > putrescine (0.84) (total BA level was a very poor predictor variable for the latter). As expected, there were significant differences between the sequential orders of these values when prior CS-normalization was implemented.

**Table 7.** Univariate statistical significance and nature of differences observed between the mean CS-normalized, glog-transformed, and autoscaled BA contents of fermented food samples (cheese, FCLO, fish, fish sauce/paste, vegetable sauce, and wine/vinegar products) in a completely randomized, one-way ANOVA model. The significance of FDR-corrected *post*-*hoc* ANOVA tests are also provided (significant differences are ranked in order of their decreasing statistical significance, i.e., increasing *p* value). The "between-fermented food class" source of variation was not statistically significant for putrescine, spermidine, spermine, histamine, or tyramine when tested in this model.


Moreover, for the PLS-DA model adopted without CS-normalization, histamine, spermidine, and spermine contents all loaded significantly on component 1 (loading vector coefficients 0.48, 0.57, and 0.47 respectively); 2-phenylethylamine, cadaverine, tyramine, and total BA levels on component 2 (loadings vector coefficients 0.42, −0.61, −0.57, and −0.61 respectively); 2-phenylethylamine and tryptamine levels on PC3 (loadings vector coefficients 0.50 and 0.57 respectively); and putrescine and spermine on PC4 (loadings vector coefficients 0.75 and −0.73 respectively). For this dataset, a four-component model was found to be most effective (permutation *p* value 0.0055)

Importantly, it should be noted that one now common issue in chemometrics/metabolomics experiments is the occurrence of a univariately-insignificant variable which remains multivariatelysignificant. Such observations are readily rationalized, firstly by the complementation (i.e., correlation) between explanatory variables, i.e., separately they do not, but when combined together as a MV composite (e.g., as a sufficiently-loading PC variable), they do serve to explain "between-classification" differences detected; secondly, consistency effects arising from the "masking" of potential univariately-significant differences by high levels of biological source sampling and/or measurement variation may be responsible (such variation may be averaged out via the conversion of datapoints to orthogonal component scores as in the PCA and PLS-DA models applied here); and thirdly, relatively small sample sizes for each classification involved (fermented foods in this case)—unfortunately, strategies applied to correct for FDRs promote the risk of statistical type II errors (i.e., false negatives) [24].

The PLS-DA evaluation was then extended and performed for pairwise comparisons of the differing fermented food classifications (CS-normalized dataset only). Firstly, as expected, Q2 values for the fish vs. fish sauce/paste, fish sauce/paste vs. vegetable sauce, and fish vs. vegetable sauce comparisons were all moderately negative, and *p* values for associated permutation tests were all >0.10. However, these values for the wine/vinegar vs. FCLO, and FCLO vs. cheese two classification model comparisons revealed that Q2 (permutation *p* values) indices for these comparisons were 0.71 (0.059) and 0.72 (0.090), but only 0.38 (0.16) for the wine/vinegar vs. cheese one (values were based on models containing two, five, and one components respectively). Hence, these results provide some evidence for the success of this strategy in distinguishing between the FCLO product, and both the cheese and wine/vinegar ones, although permutation test *p* values obtained for these models were a little higher than the 0.05 significance level, i.e., they were close to statistical significance.

**Figure 5.** (**a**) Heatmap diagram displaying the most univariately-significant differences between mean values of eight BA explanatory variables (near right-hand side *y*-axis) for the fermented cheese (red), FCLO (green), fish (dark blue), fish sauce/paste (pale blue), vegetable sauce (purple), and wine/vinegar (yellow) products. The complete BA dataset was CS-normalized, glog-transformed, and autoscaled

prior to analysis. AHC analysis shown on the left-hand side ordinate axis demonstrated two major analyte clusterings, the upper one consisting of putrescine, spermidine, and spermine pathway biomolecules (and histamine), whereas the lower one features all aromatic BAs, along with cadaverine. (**b**) 3D PCA PC3 vs. PC2 vs. PC1 scores plot for the same CS-normalized dataset shown in (**a**), showing reasonable or strong distinctions between the cheese, wine/vinegar and FCLO fermented food classes. (**c**) 3D PLS-DA PC3 vs. PC2 vs. PC1 scores plot for the corresponding non-CS-normalized dataset, which also incorporated total BA content as a potential explanatory variable (again, effective distinctions between the cheese, FCLO, and wine/vinegar classes were notable).

We then elected to statistically combine the fish, fish sauce/paste, and vegetable sauce groups, and repeated the PLS-DA modelling in order to compare the sauce/fish composite, cheese, FCLO, and wine/vinegar groups using the CS-normalized dataset. This analysis exhibited a quite high level of classification success (Figure 6a); Q<sup>2</sup> for this comparative four-classification analysis was 0.44, and a PLS-DA permutation test confirmed its significance (*p* = 0.031). The loadings of each BA variable on PLS-DA components 1 and 2 is shown in Figure 6b, and this demonstrates three groups of these predictors: the first with highly positive component 1 and highly negative component 2 loadings (all aromatic BAs, i.e., 2-phenylethylamine, tyramine, and tryptamine); the second with low to intermediate positive component 1 but highly positive component 2 loadings (metabolically-related putrescine, spermidine, and spermine, together with histamine); and the third with highly negative loadings on component 1, but negligible loadings on component 2 (cadaverine only). These grouped BA loadings vectors were very consistent with other observations made from the MV analysis of these data as a full six fermented food classification dataset. Specifically, they are completely reflective of the patterns of BA "markers" found in fermented FCLO, wine/vinegar, and cheese products respectively (Figure 5a).

Finally, RF analysis of this revised dataset showed that this approach had an at least reasonable level of classification success, with all (10/10) FCLO and 88% (15/17) of the fish/sauce combination samples being correctly classified; notwithstanding, only 60 and 50% of the cheese and wine/vinegar fermented food products, respectively, were.

**Figure 6.** *Cont*.

**Figure 6.** PLS-DA evaluation of revised dataset with combined fish, fish sauce/paste, and vegetable sauce classifications (abbreviated COMBO); CS-normalization was applied to the dataset prior to analysis. (**a**) 3D PLS-DA component 3 vs. component 2 vs. component 1 scores plot revealing some clustering of the fermented food classifications (i.e., cheese, wine/vinegar, FCLO, and COMBO). (**b**) Corresponding component 2 vs. component 1 loadings plot for this PLS-DA analysis.

#### *3.7. Scientific Significance and Human Health Implications of Results Acquired*

Results acquired from the combined applications of univariate and MV chemometrics techniques in this study clearly demonstrated that the latter strategy was valuable for distinguishing between fermented wine/vinegar products and cheeses, and the discrimination between both of these food classes from either fish, fish sauce/paste, or vegetable sauce products (or a statistical combination of them) was possible on the basis of their BA, total lipid, pH, and TA values; nevertheless, such techniques were not readily able to distinguish between the latter three fermented food classes. However, a rigorously-constrained univariate analysis method selected to overcome complications arising from intra-food classification heteroscedasticities and FDRs was able to successfully distinguish between the vegetable sauce and fish groups through significantly higher and lower levels of spermidine and 2-phenylethylamine, respectively, present in the former class. Moreover, experimental results indicated that cadaverine, tyramine, putrescine, and tryptamine concentrations may all contribute significantly towards food pH values in view of their strong positive correlations with this parameter found, together with corresponding negative ones with TA values (Figure 3b).

Moreover, BA-targeted univariate and multivariate analyses of CS-normalized data was found to be valuable for providing useful discriminatory information, which highlighted the characteristic patterns of BA biomolecules, which may be valuable for further investigations of the particular nature and/or geographic origins of fermented foods, and the mechanisms involved in their formation. Indeed, the present study found that such patterns comprised cadaverine only for cheese samples, three aromatic BAs (2-phenylethylamine, tyramine, and tryptamine), for FCLOs (sourced from fermented cod livers), and those from the sequential metabolic pathway which transforms the amino acid substrates ornithine or arginine to spermine (i.e., putrescine, spermidine, and spermine itself) for wine/vinegar products. Such idiosyncratic, fermented food product-dependent signatures for CS-normalized fermented food BA concentrations may serve to provide valuable information regarding the fermentative bacterial sources, routes involved in fermentation, and product manufacturing conditions employed for them.

For the putrescine → spermidine → spermine metabolic pathway, which was identified as representing a wine/vinegar-specific one from analysis of the CS-normalized dataset, and which accounted for >70% of total BAs in this fermented food class (Table 2), both positive or negative correlations could arise between a BA catabolite and its immediate upstream precursor, but not necessarily between the terminal spermine metabolite and that upstream of its spermidine substrate (i.e., putrescine).

With regard to toxic concentrations and health risk recommendations available in [25], it should be noted that all mean histamine levels determined in the fermented food samples tested here lie markedly below the recommended 100 ppm limit for it (with no single product exceeding this value—the highest level observed was 57 ppm in one of the fish sauce products assessed). Furthermore, with the exception of the cheese products evaluated, the mean total BA values all food groups were <200 ppm, the wine/vinegar classification substantially so (Table 2). However, although three of the cheese products tested had total BA contents of <200 ppm, two of them had levels ranging from 600–800 ppm, and therefore their dietary consumption may present a health risk for susceptible individuals.

Mean BA concentrations for the FCLO product examined ranged from 0 (histamine) to only 34 ppm (2-phenylethylamine), with the highest levels observed for the most predominant species, 2-phenylethylamine and tyramine, being 103 and 88 ppm. Since the United States of America's recommended dietary intake of health-friendly, highly unsaturated omega-3 (O-3) fatty acids (FAs) is a maximum of 1.0 g/day [41], and the oil explored here contains a mean of 29% (*w*/*w*) total O-3 FAs (predominantly the sum of eicosapentaenoic and docosahexaenoic acids) [6], then daily consumption of 100/29% × 1.0 g = 3.45 g of this FCLO product would provide estimated absolute maximal daily intake levels of 3.45 × 103 μg = 355 μg, and 3.45 × 88 μg = 304 μg of 2-phenylethylamine and tyramine, respectively. Based on the 10 samples of this product analyzed, estimated mean daily intakes of these BAs will be 111 and 95 μg only. Therefore, it appears that daily consumption of this product at the recommended U.S.-recommended dosage levels will certainly not provide any health risks to consumers, even if they are susceptible to the adverse effects experienced by their excessive intake (e.g., migraines induced by 2-phenylethylamine).

As noted above, one potentially important health benefit offered by the ingestion of dietary BAs is their novel antioxidant properties, both for the prevention of food spoilage during storage or transport episodes, but also in vivo following their ingestion. Indeed, our laboratory recently explored the powerful antioxidant capacities of BA-containing natural FCLO products, and their resistivities to thermally-mediated oxidative damage to unsaturated FAs therein, particularly O-3 PUFAs [6]. These marine oil products, which arise from the pre-fermentation of cod livers (Section 2), were indeed found to display a very high level of antioxidant activity, and PUFAs therein were also more resistant towards thermally-mediated peroxidation than other natural cod liver oil products evaluated. Resonances assignable to aromatic BAs, specifically those arising from 2-phenylethylamine and tyramine, were directly observable in the 1H NMR profiles of ca. 1/3 (*v*/*v*) diluted solutions of these products with C2HCl3. Additionally, corresponding spectra acquired on both 2H2O and C2H3O2H extracts of these oils confirmed the presence of both these BAs, together with a series of others, both aromatic and aliphatic. In the present study, mean concentrations of 2-phenylethylamine and tyramine detectable in these products were found to be 34 and 24 ppm, respectively. Antioxidant actions of the phenolic BA antioxidant tyramine found in this FCLO product may be explicable by its chain-breaking antioxidant effects, and this may offer contributions towards the potent resistance of PUFAs, particularly O-3 FAs, present therein. However, in view of the absence of a phenolic function in 2-phenylethylamine, its antioxidant potential is likely to involve an alternative radical-scavenging mechanism, presumably that involving O2-consuming carbon-centered pentadienyl radical species, as found in [22,23].

The TBARS method employed here to determine the lipid peroxidation status of fermented foods, which involved an extended low temperature equilibration process [28], successfully avoids the artefactual generation of TBA-reactive aldehydes, including malondialdehyde (MDA), during commonly-employed alternative protocols for this assay system, which generally involve a

short (ca. 10–15 min) heating stage at 95–98 ◦C in order to develop the monitored pink/red chromophore rapidly. However, from an analysis of TBARS and (TBARS):(total lipid) ratios determined on the preliminary FCLO-excluded fermented food samples, there appears to be only little evidence for the ability of BAs to offer any protection against lipid peroxidation in such products. Although Table 3 shows that the above ratio is significantly greater in the wine/vinegar group, this observation is perceived to be derived from their very low lipid contents, and the presence of a range of non-MDA TBA-reactive aldehydes present therein, including acetaldehyde and acrolein, for example, although these are also lipid oxidation products. Moreover, despite taking steps to avoid the artefact-generating heating stage of this assay, this test still remains poorly specific in view of the reactions of a variety of non-aldehydic substrates to react with it to form interfering chromophores, which also absorb at a monitoring wavelength of 532 nm. Nevertheless, TBARS level appeared to be positively correlated with fermented food spermidine concentration (Figure 2), and both this lipid peroxidation index and its lipid-normalized value appeared to be positively correlated with fermented food histamine content (loading on PC2, Table 4). However, in view of the many complications associated with this TBARS lipid peroxidation index, which offers only a very limited and still often erroneous viewpoint on the highly complex lipid peroxidation process [42], such observations cannot be rationally considered at this stage. As expected, the lipid-normalized TBARS value was negatively correlated with total lipid content (Figure 3b). The latter variable also appeared to be negatively correlated with histamine and putrescine levels (loading on PC5, Table 4). Unfortunately, results from unspecific TBARS assays are still widely employed as important quality indices throughout the food industry.

One quite surprising observation made in the current study was the detection of lipids, albeit at low levels, in wine and vinegar samples. Notwithstanding, as noted above, FAs have been detected in Zhenjiang aromatic vinegar products at similar contents to those found here [32]. Furthermore, Yunoki et al. [43] explored the FA constituents of some commercially-available red wine products, and found that lipid constituent concentrations varied from 27 to 96 mg/100 mL for *n* = 6 domestic (Japanese) wines, and 31 to 56 mg/100 mL for *n* = 6 foreign products, and that a total of 12 different FAs were detectable, mainly saturated ones. Although the extraction method described in the latter report was a 2:1 chloroform:methanol (Folch) one that targets non-polar triacylglycerols (TAGs) and more polar phospholipids, it is likely that the FAs detectable in the wine/vinegar products explored here, and also those present in Zhenjiang aromatic vinegars [32], are present as free non-glycerol-esterified species and their corresponding anions, and this would account for their higher levels detectable in these studies than those reported in [43]. Indeed, fermentation processes readily induce the hydrolysis of TAGs to free FAs, together with mono- and diacylglycerol adducts, and free glycerol [44]; such FAs will be expected to contribute towards the food pH values determined here. Similarly, Phan et al. [45] found a broad spectrum of lipidic species, specifically TAGs, polar lipids, free FAs, sterols, and cholesterol esters present in pinot noir wines.

The official AOAC gravimetric method for lipid determination employed in the current study involves an acid hydrolysis step involving HCl in any case, followed by extraction with mixed ethers, i.e., both diethyl and petroleum ethers. Hence, the HCl added will be sufficient to hydrolyze any residual TAGs present to free FAs and glycerol, and also fully protonate the former so that they are extractable as such into ether solvents. Indeed, it has been demonstrated that such free FAs are readily soluble and extractable into these ether solvent systems [46,47]. Hence, the passage of lipidic species from grapes and/or micro-organisms to finalized bottled wine and vinegar products has been confirmed in further investigations.

Interestingly, 1H NMR analysis of 2H2O extracts of the FCLO product investigated found proportionately high concentrations of free FAs and free glycerol therein (data not shown). These FAs were mainly present as PUFAs, as would be expected from the overall lipid composition of this product which contains high levels of omega-3 FAs as TAG species prior to fermentation induction. This observation is fully consistent with the ability of lactobacilli-mediated fermentation processes to partially hydrolyze TAGs in such a product. High levels of the short-chain organic acids propionic

and acetic acids (as their propionate and acetate anions in neutral solution media), both lactobacilli fermentation catabolites, the former arising from the metabolic reduction of lactate [48], were also detectable in these extracts. These results will be reported in detail elsewhere.

#### **4. Limitations of the Study**

One important limitation of this study is the limited sample sizes of some of the fermented food sampling classes incorporated into our primary experimental design. This was largely a consequence of only small numbers of differing fermented food products being available for purchase locally, for example vegetable sauce and fish products. However, it should be noted that the cheese and wine/vinegar classifications had BA contents and patterns which markedly contrasted with those of the other fermented food groups evaluated. These differences, along with those for other food quality markers observed (Table 3, Figures 1 and 3–5), were found to be very highly statistically significant, even with these limited sample sizes. Hence, this did not present a major constraining issue. Moreover, the performance of additional MV analyses on a revised model including a combined fish, fish sauce/paste, and vegetable sauce classification (on the basis of only a limited level of significant differences between them) with *n* = 17 overall served to overcome this problem (Figure 6), and this incentive did not distract from the main objectives and focus of the investigation in view of their predominant MV similarities in BA contents. However, univariate analysis found that the mean spermidine concentration was significantly higher in fermented vegetable sauces than it was in corresponding fish products (Table 3), and vice-versa for mean 2-phenylethylamine levels (Table 7). Further evidentiary support was provided by data analysis strategies applied, which were highly rigorous, and included the preliminary tracking of sample outliers. Furthermore, rigorous Welch tests were implemented for the ANOVA models employed, and either Bonferroni or FDR corrections were applied for *post*-*hoc* "between-fermented food classification" tests in order to circumvent potential problems with false positives (type I errors).

Another limitation of the current study was the unavailability of differing manufacturing sources of FCLO products, and therefore unlike other fermented food products assessed here, statistical evaluations involved an investigation of 10 separate, randomly-selected batches of a single product, both separately (Table 5) and jointly with all other classes involved in the primary statistical analysis conducted (Table 7, and Figures 5 and 6). However, the very wide between-batch variance of all FCLO samples explored facilitated this approach.

Finally, one further limitation is the poor specificity and interpretability of the TBARS method employed for the quality assessment of fermented food products here, specifically for assessments of their degrees of lipid peroxidation. However, one major precautionary step was taken in this study to minimize problems and potential interferences in this assay system, and this involved the avoidance of an aldehydic artefact-forming heating stage. Future investigations of the lipid oxidation status of fermented foods should therefore employ more reliable and specific methodologies such as those involving high-resolution 1H NMR analysis for the direct, simultaneous, multicomponent analysis of a series of both primary and secondary lipid oxidation products, e.g., conjugated hydroperoxydienes and their aldehydic fragmentation products, respectively. This protocol may be applied directly to solution-state products, or indirectly to either aqueous or lipid/deuterochloroform extracts of fermented food products.

#### **5. Conclusions**

This study demonstrated that almost all fermented foods tested had total BA levels which lay below the maximum recommended values for them. A composite application of univariate and MV chemometrics techniques clearly demonstrated that the MV approach applied was valuable for discriminating between fermented wine/vinegar products and cheeses, and the distinction between these two fermented food classes and a combination of fish, fish sauce/paste, and vegetable sauce products. Further MV analysis performed on CS-normalized BA profiles revealed

distinctive patterns for cheese (cadaverine only), FCLOs (the aromatic BAs 2-phenylethylamine, tyramine, and tryptamine), and wine/vinegar products (pathway-associated putrescine, spermidine, and spermine). Such distinctive signatures for fermented food BA contents may offer useful information regarding the nature of, and regulatory conditions employed for, fermentation processes utilized during their commercial production.

The simultaneous untargeted analysis of eight or more BAs using the LC-MS/MS analysis strategy employed here offers major advantages which are unachievable by alternative, more targeted techniques with the ability to determine only single or very small numbers of chemometrically-important analytes. Notably, the diagnostic potential of a series of n (for example, five or more) BA content analyte variables in a MV chemometrics investigation offers major advantages over the analytical acquisition of only a single possible marker. Indeed, food sample patterns of BAs and related food quality indices, which are characteristic of a particular fermented food product classification, will be expected to provide a much higher level of statistical power, reliability, and confidence concerning the accurate distinction between these classifications, and their accurate and selective assignment to one of them, than that discernable from a single BA analyte level only. Secondly, the patterns of BAs and associated food quality criteria determined, together with their correlations to particular factors or components (predominantly linear, but occasionally quadratic or higher combinations of predictor BA and supporting variables), may potentially serve to supply extensive information regarding the sources of such BAs, bacterial, commercial, or otherwise.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/9/12/1807/s1, S1: Summary of Antioxidant Activities of BAs, S2: Potential Adverse Health Effects of Dietary Bas, S3: Outline of Analytical Techniques Available for BA Determinations and the Screening of BA-Generating Bacteria in Foods.

**Author Contributions:** J.Z. was responsible for the manufacture of FCLO samples and the random distribution of these samples from different batches for analysis; he was also responsible for surveys of the availabilities, and purchases of all fermented food products from US retail outlets, together with their distribution for analysis. M.G. and B.C.P. monitored and validated all chemical analysis methods for fermented food products, involving those for BAs, TA and pH values, total lipid contents, and TBARS levels. M.G. was responsible for study experimental design, and also performed the univariate and MV chemometrics analyses of analytical datasets acquired, with assistance from B.C.P., M.G. also prepared, drafted, and finalized the manuscript for submission purposes. J.Z., B.C.P., and M.G. reviewed and edited manuscript drafts, and also contributed towards the interpretation of experimental results obtained. M.G. also fully supervised the complete study. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was part-funded by The Weston-Price Foundation, grant number WP1-MG3.

**Acknowledgments:** All authors are very grateful to the Weston A. Price Foundation (DC, USA) for part-funding the study, and to Midwest Laboratories (13611 B Street, Omaha, NE 68144-3693, USA) for performing the laboratory analysis of BAs. We are also grateful to Dave Wetzel of Green Pastures Products Inc. (NE, USA) for valuable discussions.

**Conflicts of Interest:** J.Z. is an employee of Green Pasture Products, 416 E. Fremont Street, O'Neill, NE 68763, USA. None of the other authors declare any conflicts of interest. The sponsoring body had no role in the design, execution, interpretation, or writing of the study.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Occurrence and Reduction of Biogenic Amines in Kimchi and Korean Fermented Seafood Products**

#### **Young Kyoung Park, Jae Hoan Lee and Jae-Hyung Mah \***

Department of Food and Biotechnology, Korea University, 2511 Sejong-ro, Sejong 30019, Korea; eskimo@korea.ac.kr (Y.K.P.); jae-lee@korea.ac.kr (J.H.L.)

**\*** Correspondence: nextbio@korea.ac.kr; Tel.: +82-44-860-1431

Received: 4 October 2019; Accepted: 24 October 2019; Published: 4 November 2019

**Abstract:** Biogenic amines produced during fermentation may be harmful when ingested in high concentrations. As current regulations remain insufficient to ensure the safety of fermented vegetable products, the current study determined the risks associated with the consumption of kimchi by evaluating the biogenic amine concentrations reported by various studies. Upon evaluation, some kimchi products were found to contain histamine and tyramine at potentially hazardous concentrations exceeding the recommended limit of 100 mg/kg for both histamine and tyramine. The biogenic amines may have originated primarily from metabolic activity by microorganisms during fermentation, as well as from *Jeotgal* (Korean fermented seafood) and *Aekjeot* (Korean fermented fish sauce) products commonly used as ingredients for kimchi production. Many studies have suggested that *Jeotgal* and *Aekjeot* may contribute to the histamine and tyramine content in kimchi. Microorganisms isolated from kimchi and *Jeotgal* have been reported to produce both histamine and tyramine. Despite the potential toxicological risks, limited research has been conducted on reducing the biogenic amine content of kimchi and *Jeotgal* products. The regulation and active monitoring of biogenic amine content during kimchi production appear to be necessary to ensure the safety of the fermented vegetable products.

**Keywords:** kimchi; *Jeotgal*; *Aekjeot*; *Myeolchi-jeot*; *Myeolchi-aekjeot*; biogenic amines; recommended limits; occurrence; reduction; starter cultures

### **1. Introduction**

Kimchi refers to a group of traditional Korean fermented vegetable products consumed worldwide [1]. Dating back to the 12th century during the Three Kingdoms period of ancient Korea, salted and fermented vegetable products represent the earliest form of kimchi, however, the addition of several ingredients such as the introduction of red peppers in the 16th century was eventually adopted for kimchi production [2]. The availability of local ingredients across different provinces in Korea led to the development of many regional kimchi varieties [3] (Figure 1). Currently, there are over 200 varieties of kimchi with over 100 different ingredients used for kimchi production [4]. Each kimchi variety is categorized according to the ingredients selected for production [5]. Kimchi in its current form has been recognized globally through international standardization as well [6]. Kimchi is prepared by trimming Napa cabbage, followed by salting, rinsing, and then draining excess water. The seasoning ingredients include red pepper powder, garlic, ginger, radish, glutinous rice paste, sugar, *Jeotgal*, and *Aekjeot*. The salted Napa cabbage is then mixed with the seasoning and stored at low temperatures (typically 0–10 ◦C in Korea [5]) to ferment until ripened [6]. While the production method described by the Codex only describes *Baechu* kimchi (Napa cabbage kimchi), slight variations are used to produce other kimchi varieties.

**Figure 1.** Kimchi varieties available across different provinces in Korea. *Baechu* kimchi: Napa cabbage kimchi; *Baek* kimchi: Napa cabbage kimchi prepared without red pepper powder; *Chonggak* kimchi: ponytail radish kimchi; *Gat* kimchi: mustard leaf kimchi; *Godeulppaegi*: Korean lettuce kimchi; *Kkakdugi*: diced radish kimchi; *Pa* kimchi: green onion kimchi; *Yeolmu* kimchi: young radish kimchi.

Nonetheless, nearly every kimchi variety benefits from preliminary brining, which inhibits the growth of pathogenic bacteria while selecting for lactic acid bacteria (LAB) known for promoting beneficial effects such as gastrointestinal regulation and prevention of colon cancer [7,8]. The LAB such as *Leuconostoc*, *Lactobacillus*, and *Weissella* species as well as the enzymes present in the ingredients are responsible for kimchi fermentation [9,10]. Consumption of kimchi is reported to provide numerous health benefits such as anti-oxidative, anti-carcinogenic, anti-mutagenic, and anti-aging effects [8,11,12].

Despite the numerous beneficial functional qualities, fermented foods such as kimchi may contain potentially harmful substances known as biogenic amines (BA). The nitrogenous compounds are mostly produced by microorganisms during fermentation through enzymatic decarboxylation of amino acids, as well as transamination of ketones and aldehydes [13]. BA are often categorized as aliphatic: putrescine, cadaverine, spermidine, spermine; aromatic: β-phenylethylamine, tyramine; heterocyclic: tryptamine, histamine [14,15]. The intake of BA at high concentrations as well as amine oxidase inhibition and deficiency may lead to toxic effects [16]. Recently, histamine, tyramine, putrescine, cadaverine, spermidine, and spermine were found to be cytotoxic toward human intestinal cells [17–19]. Furthermore, BA may also be converted to potentially carcinogenic N-nitrosamines in the presence of nitrites [20,21]. Excessive intake of foods containing high concentrations of histamine may potentially induce "scombroid poisoning" with symptoms such as headaches, hives, diarrhea, dyspnea, and hypotension [22]. Similarly, ingestion of foods with excessive tyramine content may cause a "cheese crisis" with symptoms that include severe headaches, hemorrhages, hypertensive effects or even heart failure [23]. As a result, many countries have implemented regulations on the production of histamine-rich seafood products, however many other food products are not currently regulated [24]. Several studies have suggested limits for BA content in food products of 100 mg/kg for histamine, 100–800 mg/kg for tyramine, 30 mg/kg for β-phenylethylamine, and 1000 mg/kg for total BA content [14,15]. The concentrations of BA in many fermented food products such as fermented meats and cheese have been widely reported to exceed limits for safe consumption. Similarly, BA have been detected in kimchi products, the most widely consumed traditional Korean food. High concentrations of BA have also been detected in kimchi ingredients *Jeotgal* (Korean fermented seafood) and *Aekjeot* (Korean fermented fish sauce), which contribute to the overall BA content in kimchi [25]. In addition, microorganisms isolated from kimchi as well as the fermented seafood products *Jeotgal* and *Aekjeot* have been reported to produce BA. Current regulations remain insufficient to address the potential health risks associated with the consumption of kimchi with high concentrations of BA. Therefore, the current article evaluated the risks associated with the BA content of kimchi products according to intake limits for β-phenylethylamine (30 mg/kg), histamine (100 mg/kg), and tyramine (100 mg/kg) as recommended by Ten Brink et al. [15], and reviewed potential sources of BA, and methods for reducing BA content.

#### **2. Biogenic Amine Content in Kimchi Products**

Table 1 displays the BA content of kimchi products as reported by various studies. The BA content of *Baechu* kimchi (Napa cabbage kimchi), the most popular kimchi variety consumed worldwide, has been reported by several studies. Cho et al. [25] reported histamine and tyramine concentrations in *Baechu* kimchi that exceeded recommended limits. Another study also showed that tyramine content in *Baechu* kimchi exceeded the recommended limit [26]. Tsai et al. [27] notably reported the highest histamine content which exceeded the recommended limit by a factor of 53. Tsai et al. [27] suggested that the high concentration of histamine in kimchi might be due to ingredients such as fish sauce or shrimp paste used in the kimchi production process. Shin et al. [28] reported β-phenylethylamine, histamine, and tyramine content at safe concentrations below 30 mg/kg. Similarly, Mah et al. [29] reported both histamine and tyramine content at safe concentrations below 30 mg/kg. In ripened *Baechu* kimchi, Kang et al. [26] reported tyramine content at concentrations that nearly reached the recommended limit.

Aside from *Baechu* kimchi, several studies have also reported the BA content of other kimchi varieties as well. *Chonggak* kimchi (ponytail radish kimchi) as reported by Jin et al. [30] contained histamine concentrations that exceeded the recommended limit. Tyramine content in *Chonggak* kimchi as reported by Kang et al. [26] were at safe concentrations, while Mah et al. [29] reported safe concentrations of both histamine and tyramine below recommended limits. As for *Gat* kimchi (mustard leaf kimchi), Lee et al. [31] reported histamine concentrations which exceeded the recommended limit by a factor of 2, while tyramine content slightly exceeded the limit. In contrast, Mah et al. [29] reported that *Gat* kimchi did not contain histamine and tyramine at detectable levels. *Kkakdugi* (diced radish kimchi) as reported by Jin et al. [30] contained tyramine at safe concentrations below the recommended limit, however, histamine concentrations exceeded the recommended limit. In contrast, Mah et al. [29] reported that histamine was not detected in *Kkakdugi*, and tyramine content was at safe concentrations below recommended limits. Similarly, Kang et al. [26] also reported tyramine concentrations in *Kkakdugi* below the recommended limit. As for *Pa* kimchi (green onion kimchi), Lee et al. [31] reported histamine and tyramine concentrations exceeded recommended limits by a factor of 4 and 2, respectively. In contrast, Mah et al. [29] reported histamine and tyramine in *Pa* kimchi at safe concentrations as tyramine content was not detected while histamine content remained below 30 mg/kg. Other kimchi varieties such as *Baek* kimchi (Napa cabbage kimchi prepared without red pepper powder), *Godeulppaegi* (Korean lettuce kimchi), and *Yeolmu* kimchi (young radish kimchi) were reported to contain histamine and tyramine at safe concentrations below 100 mg/kg [29].

Nonetheless, as the vast majority of studies are primarily focused upon *Baechu* kimchi, further research on the BA content of other kimchi varieties remains necessary. Currently, the severity of the risks associated with the BA content of kimchi remains difficult to thoroughly assess as limited research has been conducted. Though various BA have been detected in kimchi products, several studies have reported histamine and tyramine content at concentrations that exceeded the recommended intake limits of 100 mg/kg. Furthermore, the risk of nitrosamine formation entails the need for continuous monitoring of BA content during fermentation, especially as putrescine and cadaverine were detected at particularly high concentrations. Due to the toxicological risks associated with the consumption of BA, the content in kimchi necessitates regulation and control to ensure its safety.



samplestryptamine,β-phenylethylamine,putrescine,tyramine,spermidine,spermine;NT: not tested; 4 Values are the minimum and maximum concentrations reported. The same number of digits is used after the decimal point in the values, as was presented in the correspondingreferences;5 ND:notdetected;6 tr:trace.

#### **3. Biogenic Amine Content of Other Vegetable Products**

Research has also been conducted on the BA content of vegetable products originating from other countries (Table S1). The popular fermented food sauerkraut is produced through lactic acid fermentation of white cabbage [32,33]. Among European fermented food products, sauerkraut most closely resembles Korean kimchi [34]. Despite its popularity, Taylor et al. [35] reported that sauerkraut contained histamine concentrations that exceeded recommended limits. Ten Brink et al. [15] also reported that histamine and tyramine in sauerkraut exceeded recommended limits by a factor of 1 and 2, respectively. Many varieties of Japanese *Tsukemono* are preserved vegetables produced utilizing methods such as fermentation, salting, and pickling [36]. *Tsukemono* are differentiated based on ingredients, pickling method, and microorganisms responsible for fermentation [5]. Handa et al. [37] reported that histamine and tyramine in *Tsukemono* exceeded recommended limits by a factor of 3 and 4, respectively. As an important part of the Taiwanese diet, mustard pickle is prepared using mustard greens submerged in 14% NaCl brine for 4 months [38]. Kung et al. [38] reported that mustard pickles contained histamine and tyramine at safe concentrations below 100 mg/kg. Though fermented vegetable products are consumed worldwide, limited research has been conducted on the BA content of vegetable-based fermented foods. The few studies available had reported a wide range of BA content, including concentrations that exceeded recommended limits. Therefore, as the risks associated with the consumption of fermented vegetables remains largely undetermined, additional research is necessary to ensure the safe consumption of fermented foods.

#### **4. Determinants for Biogenic Amine Content in Kimchi**

#### *4.1. Biogenic Amine Content of Kimchi Ingredients: Jeotgal and Aekjeot*

Kimchi production involves the use of many ingredients including the fermented seafood products *Jeotgal* and *Aekjeot*. Used as seasoning ingredients during the production of kimchi [39], *Jeotgal* and *Aekjeot* contain flavor compounds that contribute greatly to the ripening process during kimchi fermentation [40]. Reports of the fermented seafood products as kimchi ingredients date back to the 16th century during the age of the *Chosun* dynasty of Korea [41]. Though *Jeotgal* and *Aekjeot* used during modern kimchi production vary by region, the most commonly used varieties include *Myeolchi*-*jeot* (salted and fermented anchovy), *Myeolchi*-*aekjeot* (salted and fermented anchovy sauce), *Saeu*-*jeot* (salted and fermented shrimp), and *Kkanari*-*aekjeot* (salted and fermented sand lance sauce) [42]. *Jeotgal* production typically involves submersion of seafood in brine with 20% salinity for 2–3 months at room temperature, and results in the final product resembling the initial seafood ingredient [43]. Some *Jeotgal* products undergo additional seasoning for consumption as side dishes rather than as ingredients during kimchi production [44,45]. Similarly, *Aekjeot* production involves the submersion of seafood in brine with salinity ranging from 20 to 30% for 1–2 years, however solid particles are removed through filtration for the final product [46]. In both *Jeotgal* and *Aekjeot*, the salt content inhibits putrefactive bacteria, and the enzymatic activity partially breaks down the proteins to develop a rich flavor [41]. Also, the addition of *Jeotgal* contributes to the protein, amino acid, and mineral content of kimchi, further reinforcing the nutritional value of kimchi products [5].

Despite the benefits described above, *Jeotgal* and *Aekjeot* have been reported to contain high concentrations of potentially hazardous BA such as histamine and tyramine [29]. Table 2 displays the BA content of the fermented seafood products. The reported BA content of *Aekjeot* and *Jeotgal* were evaluated according to recommended limits for intake. *Myeolchi*-*jeot* was reported to contain histamine and tyramine concentrations which exceeded recommended limits by a factor of approximately 6 and 2, respectively [47]. *Myeolchi*-*aekjeot* reportedly contained histamine and tyramine at concentrations that exceeded recommended limits by a factor of approximately 12 and 4, respectively [29]. The BA content of *Myeolchi*-*aekjeot* as studied by Cho et al. [25] showed β-phenylethylamine, histamine, and tyramine content at concentrations that exceeded recommended limits by a factor of about 2, 11, and 6, respectively. Moon et al. [48] also studied the BA content of *Myeolchi*-*aekjeot* by reporting

β-phenylethylamine, histamine, and tyramine content at concentrations that exceeded recommended limits by a factor of approximately 3, 12, and 4, respectively. Similarly, Shin et al. [28] reported that *Myeolchi*-*aekjeot* contained β-phenylethylamine, histamine, and tyramine at concentrations that exceeded recommended limits by a factor of approximately 1, 4, and 4, respectively. Cho et al. [49] and Joung and Min [50] reported histamine concentrations in *Myeolchi*-*aekjeot* which greatly exceeded recommended limits by a factor of about 21 and 11, respectively.

As for *Kkanari*-*aekjeot*, histamine and tyramine content were reported at concentrations that exceeded recommended limits by a factor of approximately 10 and 2, respectively [29]. Cho et al. [25] also reported the β-phenylethylamine, histamine, and tyramine content in *Kkanari*-*aekjeot* at concentrations which exceeded recommended limits by a factor of 2, 11, and 6, respectively. Moon et al. [48] reported histamine and tyramine content at concentrations that exceeded recommended limits by a factor of about 7 and 3, respectively. Similarly, Shin et al. [28] reported that *Kkanari*-*aekjeot* contained β-phenylethylamine, histamine, and tyramine at concentrations that exceeded recommended limits by a factor of approximately 1, 10, and 3, respectively. Notably, the highest histamine content in *Kkanari*-*aekjeot* was reported by Cho et al. [49] as concentrations greatly exceeded the recommended limit by a factor of approximately 18.

As for *Saeu*-*jeot*, Mah et al. [47], Cho et al. [25], Moon et al. [48], and Shin et al. [28] reported BA content at safe concentrations below recommended limits for β-phenylethylamine, histamine, and tyramine, respectively.

Overall, the considerably high BA concentrations, especially histamine, reported for both retail *Jeotgal* and *Aekjeot* products may be potentially hazardous. All *Kkanari*-*aekjeot* and *Myeolchi*-*aekjeot* products contained histamine concentrations which exceeded 100 mg/kg indicating that safety regulations are necessary. According to Mah et al. [29], the high BA content may be due to the considerably long fermentation duration for the production of the fermented seafood. Furthermore, the results of the research conducted by Moon et al. [48] suggested that total BA content increased alongside crude protein concentrations for both *Jeotgal* and *Aekjeot*. After all, the high concentrations of BA reported for kimchi appears to originate partly from fish sauce such as *Myeolchi*-*aekjeot* and *Kkanari*-*aekjeot* [25]. Given the high concentrations of BA detected in kimchi and fermented seafood products, safety regulation and standardization of the manufacturing process appears to be necessary.

High BA concentrations were not limited to *Jeotgal* and *Aekjeot* products as the similar observations were reported for fermented seafood products originating from other countries (Table S1). Saaid et al. [51] studied the BA content of Malaysian seafood. The study showed that *Cincalok* (salted and fermented shrimp) contained histamine and tyramine at high concentrations that exceeded the recommended limits by a factor of approximately 3 and 7, respectively. *Budu* (salted and fermented anchovy) also contained high histamine and tyramine concentrations that exceeded recommended limits by a factor of 4 and 9, respectively. Similarly, research conducted by Rosma et al. [52] revealed histamine concentrations in *Budu* exceeding the recommended limit by a factor of 11.

The reported results indicated that fermented seafood products tended to contain high concentrations of BA, especially histamine. As the BA content exceeded well beyond recommended limits, consumption of the fermented seafood products may lead to adverse effects on human health. Due to the potential toxicological risks, expansion of current regulations regarding the BA content of seafood appears to be necessary to cover the aforementioned fermented seafood products as well as to include other amines such as tyramine and β-phenylethylamine.



*Foods* **2019**, *8*, 547

corresponding

 references; 5 ND: not detected.

#### *4.2. Biogenic Amine Production by Bacterial Strains from Kimchi and Fermented Seafood Products*

Microorganisms play a major role in the production of BA during fermentation through the decarboxylation of free amino acids. LAB responsible for fermentation have been reported to produce putrescine, cadaverine, histamine, and tyramine [15]. Table 3 displays the BA production by bacterial strains isolated from various kimchi and fermented seafood products. Tsai et al. [27] reported that LAB strains isolated from kimchi products purchased from Taiwanese markets were capable of producing histamine and other BA. The isolated strains identified as *Lactobacillus paracasei* subsp. *paracasei*, *Lb*. *brevis*, and *Brevibacillus brevis* were tested for β-phenylethylamine, putrescine, cadaverine, histamine, and spermine production in assay media. The reported results showed that *Lb*. *paracasei* subsp. *paracasei*, *Lb*. *brevis*, and *Bb*. *brevis* produced histamine at concentrations of 15.1, 13.6, and 16.3–43.1 μg/mL, respectively. Other BA were detected at concentrations lower than 15 μg/mL. Kim and Kim [53] isolated LAB strains from kimchi identified as *Lb*. *brevis*, *Lb*. *curvatus*, *Leuconostoc mesenteroides*, and *Staphylococcus hominis* that demonstrated tyramine production capabilities at over 200 μg/mL in assay media. Jeong and Lee [54] reported on putrescine, cadaverine, histamine, and tyramine production in assay media by LAB isolated from kimchi including *Leu*. *citreum*, *Leu*. *lactis*, *Leu*. *mesenteroides*, *Weissella cibaria*, *W*. *confusa*, and *W*. *paramesenteroides*. The results revealed that *Leuconostoc* spp. did not produce histamine and tyramine, however, putrescine and cadaverine were produced at concentrations lower than 20 μg/mL. *Weissella* spp. also produced putrescine and cadaverine at concentrations lower than 20 μg/mL, however, some strains produced histamine and tyramine at concentrations higher than 50 μg/mL. Compared to *Leuconostoc* spp., *Weissella* spp. produced a wider variety of BA at higher concentrations, prompting recommendations for stricter safety guidelines for screening starter *Weissella* strains suitable for kimchi fermentation [54].

Other varieties of kimchi were also reported to contain microorganisms capable of BA production. While the majority of the LAB strains isolated from *Chonggak* kimchi and *Kkakdugi* did not produce BA at detectable levels, some isolated LAB strains reportedly produced tyramine in the ranges of 260.93–339.56 μg/mL and 287.23–386.17 μg/mL, respectively, in BA production assay media [30]. Aside from tyramine, other BA were not detected in the same assay media. Although the study did not specify the bacterial species capable of producing BA, *Lb*. *brevis* was suggested as a strong producer of BA. Lee et al. [31] reported the BA production in assay media by LAB strains isolated from *Gat* kimchi and *Pa* kimchi. From *Gat* kimchi, *Enterococcus faecium*, *Lb*. *brevis*, and *Leu*. *mesenteroides* produced the highest concentrations of tyramine in the ranges of 259.10–269.57 μg/mL, ND-365.96 μg/mL, and 145.14–301.67 μg/mL, respectively. *Lb*. *brevis* strains also produced putrescine ranging from ND to 320.42 μg/mL. From *Pa* kimchi, the isolated LAB strains identified as *Lb*. *brevis* and *Lb*. *sakei* produced the highest concentration of BA such as tyramine in the ranges of ND-301.52 μg/mL and 113.98–131.36 μg/mL, respectively. Also, a *Lb*. *brevis* strain produced putrescine at 362.44 μg/mL. Aside from putrescine and tyramine, other BA produced by LAB strains were reported at concentrations lower than 60 μg/mL. Based on the reported BA production capabilities of isolated strains, LAB appear to contribute to the BA content in kimchi, especially tyramine which were produced at the highest concentrations.

Aside from LAB, other bacterial species isolated from *Jeotgal* products were reported to have BA production capabilities. *S. equorum* strains isolated from *Saeu*-*jeot* and *Myeolchi*-*jeot* were reported to be capable of producing putrescine, cadaverine, histamine, and tyramine in assay media [55,56]. The reported results showed that all BA were detected at concentrations below 50 μg/mL. Lim [57] isolated bacterial strains from *Myeolchi*-*jeot* which were identified as *Bacillus licheniformis*, *Serratia marcescens*, *S. xylosus*, *Aeromonas hydrophila*, and *Morganella morganii*, and the strains were capable of producing high concentrations of histamine in assay media at 1699.3 ± 35.6 μg/mL, 1987.2 ± 27.8 μg/mL, 2257 ± 30.7 μg/mL, 1655.5 ± 41.2 μg/mL, and 2869.4 ± 49.0 μg/mL, respectively. Mah et al. [58] suggested that *Bacillus* species, especially *B*. *licheniformis*, contributed towards BA content as the isolated strains isolated from *Myeolchi*-*aekjeot* were capable of producing putrescine, cadaverine, histamine, and tyramine. Thus, the isolated bacterial strains appear to contribute to the high histamine content of fermented seafood products, which in turn contribute to the BA content of kimchi.



*Foods* **2019** , *8*, 547

corresponding

 references.

The aforementioned studies reported BA production by isolated strains at widely varying concentrations, even among the same species. Lee et al. [31] suggested that the BA production by LAB isolated from kimchi may be strain-dependent. Differences in BA production are widely considered to be strain-dependent, and not species-dependent [59]. The claim is further substantiated by the evidence for horizontal gene transfer for decarboxylase genes [60–62]. For example, as tyrosine decarboxylation was observed only for some strains, even belonging to the same species of LAB, tyramine production is considered strain-specific rather than species-specific [63]. Nonetheless, BA production by isolated strains indicates a risk for BA accumulation during *Jeotgal* and kimchi fermentation. Consequently, the control of BA accumulation during the production of fermented foods necessitates the reduction of microbial BA production by control of fermentation conditions, utilization of starter cultures, and sanitary practices to prevent contamination by BA-producing microorganisms.

#### **5. Strategies to Reduce Biogenic Amine Content in Kimchi Products**

Despite the risks associated with BA accumulation, limited research has been conducted on reducing the BA content of kimchi products. Instead of directly reducing BA content in kimchi, several studies have reported various methods to reduce BA concentrations in the kimchi ingredients *Jeotgal* and *Aekjeot*. Kim et al. [64] reported that kimchi produced using fermented seafood products contained BA at significantly higher concentrations. Lee et al. [65] suggested that the BA concentration of kimchi products may be reduced by limiting the quantity of the fermented seafood products used during kimchi production. For example, Kang [66] reported the histamine content of kimchi without *Myeolchi*-*aekjeot* at safe levels, however, the addition of *Myeolchi*-*aekjeot* raised histamine content to unsafe concentrations above the recommended limit by a factor of approximately 6. The study also described the effect of heat treatment of *Myeolchi*-*aekjeot* on the histamine content of kimchi. Histamine concentrations in kimchi produced using heat-treated *Myeolchi*-*aekjeot* were reported at 546.14 ± 1.33 mg/kg, while non-treated kimchi contained 592.78 ± 3.43 mg/kg. The reported results indicate that microorganisms from *Myeolchi*-*aekjeot* contributed towards the production of histamine during kimchi fermentation. Also, as research shows that histamine is heat-stable [67], the lower BA content in kimchi produced using the heat-treated *Myeolchi*-*aekjeot* may be due to the sterilization of histamine-producing microorganisms [66]. In addition to the contribution of BA content in kimchi by *Myeolchi*-*aekjeot*, Lee et al. [31] suggested that microorganisms from *Myeolchi*-*aekjeot* may produce BA during kimchi fermentation. Utilizing substitute ingredients in lieu of *Myeolchi*-*aekjeot* and *Kkanari*-*aekjeot* may also be effective in reducing BA content in kimchi. As other *Jeotgal* products including *Ojingeo*-*jeot* (salted and fermented sliced squid), *Toha*-*jeot* (salted and fermented *toha* shrimp), *Jogae*-*jeot* (salted and fermented clam), *Baendaengi*-*jeot* (salted and fermented big-eyed herring), and *Eorigul*-*jeot* (salted and fermented oysters) have been found to contain individual BA content below 100 mg/kg [47], utilization of the fermented seafood products with low BA content for kimchi production is expected to reduce the overall BA content of kimchi products [29].

Research on using additives to reduce the BA content of fermented seafood products has also been reported. Mah et al. [68] conducted research to reduce BA production by microorganisms isolated from *Myeolchi*-*jeot*, introducing additives into assay media and *Myeolchi*-*jeot*. The results confirmed that compared to the control, garlic extract was the most effective inhibitor of bacterial growth and BA production by yielding lower in vitro production of putrescine, cadaverine, histamine, tyramine, and spermidine by 11.2%, 18.4%, 11.7%, 30.9%, and 17.4%, respectively. Further results revealed that compared to *Myeolchi*-*jeot* samples treated with ethanol (control), the addition of 5% garlic extract to *Myeolchi*-*jeot* (treatment) inhibited bacterial growth and consequently reduced overall BA production by up to 8.7%. In another study by Mah and Hwang [69], other additives were also used for the reduction of BA production by *Myeolchi*-*jeot* microorganisms in assay media and *Myeolchi*-*jeot*. Among the additives tested in assay media, glycine most effectively inhibited in vitro BA production by bacterial strains. In comparison to the control without additives, the addition of 10% glycine in assay media resulted in reductions in putrescine, cadaverine, histamine, tyramine, and spermidine production by 32.6%, 78.4%, 93.2%, 100.0%, and 100.0%, respectively. Compared to the *Myeolchi*-*jeot* samples salted at 20% NaCl, additional supplementation of 5% glycine reportedly reduced overall BA content by 73.4%. The results suggest that the addition of glycine as well as salt may improve the safety of fermented seafood products. It is noteworthy that despite the results showing effective BA reduction, the use of garlic extract or glycine may affect the flavor of the final product.

Aside from additives, other studies have utilized starter cultures to reduce BA content in *Jeotgal*. In a study by Mah and Hwang [70], some bacterial strains isolated from *Myeolchi*-*jeot* were found to reduce BA content in *Myeolchi*-*jeot*. The reported results showed that, of the 7 starter candidate strains, *S*. *xylosus* exhibited the highest histamine degradation capability as well as the ability to slightly degrade tyramine in assay media. In comparison to the uninoculated *Myeolchi*-*jeot* control, the addition of the starter culture reduced the production of putrescine, cadaverine, histamine, tyramine, and spermidine by 16.5%, 10.8%, 18.0%, 38.9%, and 45.6%, respectively. Jeong et al. [56] isolated strains from *Jeotgal* for use as potential starters and found that *S. equorum* strain KS1039 did not produce putrescine, cadaverine, histamine, and tyramine in vitro.

A limited number of studies have even attempted to directly reduce the BA content of kimchi through the inoculation of bacterial strains. Kim et al. [71] reported reductions in tryptamine, putrescine, cadaverine, histamine, and tyramine levels in *Baechu* kimchi fortified with *Leu. carnosum*, *Leu. mesenteroides*, *Lb. plantarum*, and *Lb. sakei* strains. Similarly, Jin et al. [30] reported that *Kkakdugi* and *Chonggak* kimchi inoculated with *Lb*. *plantarum* strains incapable of producing BA contained lower level of tyramine (but not the other BA) than the uninoculated control. Therefore, utilizing LAB strains unable to produce (and/or able to degrade) BA as kimchi starter cultures may likely reduce the total BA content during kimchi fermentation.

Although the aforementioned studies have shown both direct and indirect methods of reducing BA content in kimchi, current commercial kimchi production processes do not appear to utilize the BA reduction techniques. This might be due to the application of BA reduction methods such as the use of additives, starter cultures, and adjusting the quantity of fermented seafood products have been reported to affect the flavor of kimchi products [69,72,73]. Consequently, inconsistent product quality is reflected in the wide range of BA content of kimchi products, including concentrations that exceed recommended limits for safe consumption. The high BA content reported for various kimchi products indicates that modern production methods require further preventative measures to ensure the safety of the fermented vegetable products, including practical application of research-based BA reduction techniques described above. Commercial kimchi production may greatly benefit from utilizing the aforementioned and novel strategies including control of fermentation conditions, utilizing starter cultures, alternative ingredients, and/or ingredients with low BA content. Furthermore, the establishment and expansion of regulations limiting BA content in fermented foods remain necessary to safeguard consumers against the potential BA intoxication.

#### **6. Conclusions**

The current study evaluated the BA content of kimchi, a term used to describe a group of Korean fermented vegetable products. Some kimchi samples have been reported to contain high concentrations of BA which exceeded recommended limits. Consumption of the fermented foods with high BA content may have detrimental effects on the body. Several factors contribute to the high BA concentrations in kimchi, which include BA production by microorganisms during fermentation and BA content of ingredients such as *Jeotgal* and *Aekjeot*. As variables such as ingredients, microorganisms, and initial BA content of *Jeotgal* that influence kimchi fermentation differed extensively, the reported BA concentrations of kimchi products also varied widely, even among the same varieties. Due to the large variations among kimchi products, standardization of kimchi production appears to be necessary to limit BA content. Furthermore, though several studies have described methods to indirectly reduce BA concentrations in kimchi by reducing the BA content of ingredients *Jeotgal* and *Aekjeot*, limited research has been conducted on the direct reduction of BA content in kimchi products. To ensure the safe consumption of kimchi products, further research on methods to reduce the BA concentrations below recommended limits appears to be necessary. In conjunction with BA reduction studies, implementation of regulations such as continuous monitoring during production remains necessary to control BA content in kimchi and *Jeotgal* products.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/8/11/547/s1, Table S1: Biogenic amine content of fermented vegetable and seafood products from various countries.

**Author Contributions:** Conceptualization, Y.K.P. and J.-H.M.; Literature data collection, Y.K.P.; Writing—original draft, Y.K.P. and J.H.L.; Writing—review and editing, Y.K.P., J.H.L. and J.-H.M.; Supervision: J.-H.M.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (no. 2019R1H1A2100972).

**Acknowledgments:** The authors thank Young Hun Jin, Junsu Lee, and Alixander Mattay Pawluk of Department of Food and Biotechnology at Korea University for technical assistance.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **E**ff**ect of Brine Concentrations on the Bacteriological and Chemical Quality and Histamine Content of Brined and Dried Milkfish**

## **Chiu-Chu Hwang 1,\*, Yi-Chen Lee 2, Chung-Yung Huang 2, Hsien-Feng Kung 3, Hung-Hui Cheng <sup>4</sup> and Yung-Hsiang Tsai 2,\***


Received: 12 October 2020; Accepted: 31 October 2020; Published: 3 November 2020

**Abstract:** In this research, the occurrence of hygienic quality and histamine in commercial brined and dried milkfish products, and the effects of brine concentrations on the quality of brined and dried milkfish, were studied. Brined and dried milkfish products (*n* = 20) collected from four retail stores in Taiwan were tested to investigate their histamine-related quality. Among them, five tested samples (25%, 5/20) had histamine contents of more than 5 mg/100 g, the United States Food and Drug Administration guidelines for scombroid fish, while two (10%, 2/20) contained 69 and 301 mg/100 g of histamine, exceeding the 50 mg/100 g potential hazard level. In addition, the effects of brine concentrations (0%, 3%, 6%, 9%, and 15%) on the chemical and bacteriological quality of brined and dried milkfish during sun-drying were evaluated. The results showed that the aerobic plate count (APC), coliform, water activity, total volatile basic nitrogen (TVBN), and histamine content values of the brined and dried milkfish samples decreased with increased brine concentrations, whereas those of salt content and thiobarbituric acid (TBA) increased with increasing brine concentrations. The milkfish samples prepared with 6% NaCl brine had better quality with respect to lower APC, TVBN, TBA, and histamine levels.

**Keywords:** histamine; dried milkfish; hygienic quality; brine-salting

#### **1. Introduction**

Histamine is a biogenic amine in charge of histamine fish poisoning (HFP) or scombroid poisoning. Histamine fish poisoning is a food outbreak with allergy-like symptoms arising from ingesting mishandled scombroid fish that have high levels of histamine in their flesh [1]. Histamine is formed mainly through the decarboxylation of free histidine in fish muscles by histidine decarboxylases produced by a number of histamine-forming bacteria present in seafood [2]. HFP has occasionally been associated with the consumption of milkfish, marlin, mackerel, and tuna in Taiwan [2–6]. However, there is compelling evidence to implicate that other factors, such as other biogenic amines, can potentiate histamine toxicity, as spoiled fish containing histamine tends to be more toxic than the equivalent amount of pure histamine that is ingested orally [1,2]. Putrescine and cadaverine were shown to enhance histamine toxicity when present in spoiled fish by inhibiting the intestinal histamine metabolizing enzyme, including diamine oxidase [1,2].

Milkfish (*Chanos chanos*) is widely distributed throughout the Indo-Pacific region and is the second most important inland aquaculture fish in Taiwan [7,8]. This fish has been cultivated in Taiwan for more than 350 years. Taiwan's total milkfish production is approximately 50,000–60,000 tons each year [8]. Chiou et al. [9] demonstrated that histidine at 441 mg/100 g is the most prominent free amino acid (FAA) in the white muscles of milkfish and accounts for 80% of the total FAAs in the fish. Therefore, milkfish products have become most often associated with HFP in Taiwan, including dried milkfish [6], milkfish sticks [10], and milkfish surimi [11]. In addition, our research team determined that 78% of commercial dry-salting and dried milkfish products have histamine contents greater than the 5 mg/100 g recommended value of the United States Food and Drug Administration's (USFDA) guidelines, while 43.7% of the fish samples were found to exceed 50 mg/100 g of histamine [12].

In general, there are two major salting methods for milkfish preservation, namely, dry-salting and brine-salting. In Taiwan, the traditional processes of dry-salting and dried milkfish include scaling, back-cutting, degutting, and dry-salting with 3–12% NaCl (*w*/*w*) followed by sun-drying for 5–7 days [12]. However, the consumption of high salt levels from seafood can result in several chronic diseases, such as hypertension and cardiovascular diseases [13]. Brine-salting for fish processing may be a better method to reduce salt uptake and water loss and, thus, to reach a higher weight yield and better quality in salted fish compared to dry-salting [14]. Therefore, in recent years, brine- and light-salting milkfish has gained popularity with Taiwanese people. However, the quality of brined and dried fish is influenced by the brine concentrations and dry methods used for drying the fish [15].

There is no information of the occurrence of hygienic quality and histamine in brined and dried milkfish products, and the formation of histamine and the quality of brined and dried milkfish produced with different brine concentrations. Therefore, the main aim of this study was to monitor the bacteriological and chemical quality, including histamine content, in 20 brined and dried milkfish samples sold in retail stores in southern Taiwan. This work also aimed to examine the effects of different brine concentrations (0%, 3%, 6%, 9%, and 15%) on the bacteriological and chemical quality and histamine contents in brined and dried milkfish products during sun-drying for five days.

#### **2. Materials and Methods**

#### *2.1. Materials*

Twenty brined and dried milkfish products were collected from four retail stores in southern Taiwan, including store A (six samples), store B (five samples), store C (five samples), and store D (four samples). All brined and dried milkfish products were home-made by the farmer or manufacturer and delivered to the store for sale. Trackback information indicated that the samples collected from store A and D were processed using higher brine concentration (>10%) and longer sun-drying days (5–7 days); on the other hand, the samples of store B and C were processed using lower brine concentration (<6%) and shorter sun-drying days (3–5 days). In general, the processing of brined and dried milkfish include scaling, back-cutting, degutting, and brine-salting with 3–15% NaCl concentrations at room temperature for 1-2 h, followed by sun-drying for 3–7 days. After the samples were purchased, they were wrapped in aseptic bags, placed in an ice box, and instantly delivered to the laboratory for analysis within 6 h. The dorsal part of the commercial dried milkfish samples were cut and taken for microbiological and chemical determinations.

Sixty fresh milkfish (weights of 546 ± 11.6 g, lengths of 31.9 ± 1.2 cm) were purchased from the fish market of the city of Kaohsiung in Taiwan and transported to our laboratory within half an hour in an ice box. Once the fish samples arrived at the laboratory, they were manually scaled, back-cut, gutted, washed with clean water, and then drained.

#### *2.2. Reagents*

Histamine dihydrochloride, trichloroacetic acid, 2-thiobarbituric acid, and butylated hydroxytoluene were purchased from Sigma-Aldrich (St. Louis, MO, USA). Acetonitrile (LC grade) and dansyl chloride (GR grade) were purchased from E. Merck (Darmstadt, Germany).

#### *2.3. Brine-Salting and Drying of Milkfish*

The back-cut milkfish were brine-salted with concentrations of 3%, 6%, 9%, or 15% NaCl with a fish-to-brine ratio of 1:2 for 60 min at 20 ◦C, and unsalted milkfish were used as controls. After brine-salting, all milkfish samples were placed under sun light at 30–33 ◦C for seven hours each day for five days. Sampling analyses were conducted at days 1, 3, and 5 for sun-drying. The experiments were conducted in triplicate for each brine concentration and sampling time. The dorsal part of the fish samples was used for analysis.

#### *2.4. Determination of pH Value, Moisture Content, Water Activity, and Salt Content*

Ten grams of the samples was weighted and homogenized with a mixer (FastPrep-24, MP Biomedicals, Solon, OH, USA) for 2 min with 20 mL of deionized water to make a thick slurry. The pH of this slurry was determined using a digital pH meter (Mettler Toledo FE20/EL20, Schwerzenbach, Switzerland). The moisture of each sample (1–3 g) was measured using the oven-dry method at 105.0 ± 1.0 ◦C for drying, followed by the determination of the sample weight until a constant weight was achieved. Water activity was determined by an Aqualab 4TE (Decagon Devices, Pullman, WA, USA) at 25 ◦C. The salt (NaCl) content was determined using Mohr's titration method [16].

#### *2.5. Determination of Total Volatile Basic Nitrogen (TVBN) and Thiobarbituric Acid (TBA)*

The TVBN values were measured using Conway's dish method as described by Cobb et al. [17]. Five grams of the minced samples was homogenized with 45 mL of 6% trichloroacetic acid (TCA; Sigma-Aldrich, St. Louis, MO, USA). After the extract was filtered, saturated K2CO3 was added to the filters. The released TVBN was absorbed by boric acid and then titrated with 0.02 N HCl, while the TVBN value was expressed in milligrams per 100 g fish sample. The TBA values were determined by the modified method of Faustman et al. [18]. Briefly, 20 g of dried milkfish sample was added into a tube containing 180 mL of deionized water and then homogenized with a mixer for 3 min. Twelve milliliters of 0.1 M TBA reagent in 0.2% HCl and 0.15 mL of 0.2% butylated hydroxytoluene (BHT) in 95% ethanol were added into 2 mL of the homogenate and then mixed well. The mixtures were heated in a water bath at 90 ◦C for 20 min and then filtered, and the absorbance of the filtrates was detected using a spectrophotometer (UV-1201, Shimazu, Tokyo, Japan) at 532 nm. The TBA values in the fish samples are expressed in milligrams of malondialdehyde (MDA) per kilogram.

#### *2.6. Microbiological Analysis*

Twenty-five grams of the minced samples was homogenized with 225 mL of sterile 0.85% (*w*/*v*) physiological saline in a sterile blender at a 1200 rpm speed for 2 min. The homogenate was serially diluted with a sterile physiological saline for 1:10 (*v*/*v*) dilutions. With regard to spread plate counting, 0.1 mL of the dilutes was spread on aerobic plate count (APC) agar (Difco, BD, Sparks, MD, USA) with 0.5% NaCl and then incubated at 30 ◦C for 24–48 h. After the bacterial colonies grown on the plate were counted, the data were expressed as log10 colony forming units (CFUs) per gram. The levels of coliform and *Escherichia coli* in the milkfish samples were performed according to the three-tube most probable number (MPN) method as described by the FDA [19].

#### *2.7. Histamine Analysis*

Histamine dihydrochloride (82.8 mg) was dissolved in 50 mL of 0.1 M HCl and used as the working solution, and the final concentration of histamine (free base) was 1.0 mg/mL. Five grams of the ground milkfish samples were homogenized with 20 mL of 6% cold trichloroacetic acid (TCA) using a Polytron PT-MR 3100 homogenizer for 3 min. The homogenates were collected via centrifugation at 4500× *g* for 8 min at 7 ◦C and filtered through Advantec Toyo No. 2 filter paper. The filtrates were diluted up to 50 mL with a 6% TCA solution. For the derivatization reaction of histamine, 1 mL aliquots of the TCA extract of each sample and histamine standard solution were derivatized with dansyl chloride using the method of Chen et al. [3] with some modifications. Briefly, 0.2 mL of 2 M sodium hydroxide and 0.3 mL of saturated sodium bicarbonate were added to 1 mL aliquots of the TCA extract of each sample and the histamine standard solution. The solution was added to 2 mL of 1% dansyl chloride solution dissolved in acetone, mixed by a vortex mixer, and left to stand at 40 ◦C for 45 min. After the reaction, 100 μL of ammonia was added to terminate the derivatization reaction. Acetonitrile was added to a final volume of 5 mL and the solution was centrifuged (10,000× *g*, 5 min, 4 ◦C). After the supernatants were filtered through 0.22 μm membrane filters, 20 μL of the filtrates were injected into high-performance liquid chromatography (HPLC). The histamine levels in each milkfish sample were analyzed by HPLC (Hitachi, Tokyo, Japan) equipped with a LiChrospher 100 RP-18 reversed-phase column (5 μm, 125 × 4.6 mm, E. Merck, Damstadt, Germany) and a UV-Vis detector (Model L-4000, Hitachi, wavelength at 254 nm). The mobile phase consisted of eluent A (acetonitrile) and eluent B (water). At the beginning, eluents A and B at a ratio of 50:50 (*v*/*v*) were applied for 19 min, followed by a linear gradient with an increase of eluent A up to 90% during the next minute. In the final 10 min, the eluent A and B mix was set to a linear decrease to 50:50 (*v*/*v*). The flow rate was 1.0 mL/min. Validation of the histamine analysis method including inter- and intra-day repeatability (expressed as % and relative standard deviation, RSD) was determined by fortifying homogenized dried milkfish meats with 1.0, 5.0, and 10 mg/100 g of standard histamine. Each spiked amount was extracted and derivatized with dansyl chloride using the above procedure in triplicate, including a blank test to evaluate the average recovery.

#### *2.8. Statistical Analysis*

One-way analysis of variance (ANOVA) and Tukey's pairwise comparison tests were performed within the 95% confidence interval. Pearson correlation was carried out to determine relationships between pH, moisture, water activity, salt content, TVBN, APC, coliform, and histamine contents in the brined and dried milkfish samples. All statistical analyses were carried out using the Statistical Package for Social Sciences (SPSS) Version 16.0 for Windows (SPSS Inc., Chicago, Il, USA), and *p* < 0.05 was used to consider significant deviation.

#### **3. Results and Discussion**

#### *3.1. Chemical and Bacteriological Quality of the Brined and Dried Milkfish Samples*

For all 20 brined and dried milkfish samples collected from the four retail stores, the pH, moisture, water activity, salt content, TVBN, APC, coliform, and histamine ranged from 5.67 to 6.05, 38.27% to 69.78%, 0.89 to 0.98, 0.16% to 4.37%, 8.86 to 19.88 mg/100 g, 3.51 to 8.25 log CFU/g, <3 to >2400 MPN/g, and 0.16 to 301 mg/100 g, respectively (Table 1). *E. coli* was not detected in any milkfish samples. Store A samples had significantly lower (*p* < 0.05) mean water activity (0.94) than did samples collected from the other three stores, while the mean salt content (3.23%) in store A samples was higher (*p* < 0.05) than the others (Table 1). Moreover, the mean TVBN and APC values in store B samples (16.06 mg/100 g and 6.62 log CFU/g, respectively) and store D samples (16.82 mg/100 g and 6.33 log CFU/g, respectively) were markedly higher (*p* < 0.05) than those samples obtained from the other two stores, while the mean coliform level (356 MPN/g) in store D samples were higher than that of the other stores (Table 1). The highest mean histamine content of 79 mg/100 g was obtained from five samples from store B, followed by store D with a mean of 4.9 mg/100 g of histamine.



 values in parentheses represent the mean ± standard deviation (SD). Values in the same column with different letters are statistically different (*<sup>p</sup>* < 0.05). CFU, colony forming unit; MPN, most probable number.

In this study, the proportion of the 20 brined and dried milkfish samples that did not meet the 6.47 log CFU/g Taiwanese regulatory standard for APC was 35% (7/20). Therefore, brined and dried milkfish manufacturers may need to be more careful with hygienic handling or processing in their preparation of brined and dried milkfish products. The distribution of histamine contents in the brined and dried samples is shown in Table 2. Five samples (25%, 5/20) failed to meet the 5 mg/100 g level of histamine, the allowable limit by the USFDA for scombroid fish and/or products, while two (10%) had 69 and 301 mg/100 g of histamine, greater than the potential toxicity level (50 mg/100 g). According to information by Bartholomev et al. [20], which showed that fish with histamine levels >100 mg/100 g could result in illness and health hazards if ingested by humans, one sample with 301 mg/100 g of histamine could have caused disease symptoms if consumed (Table 2). In contrast, our previous research showed that 78.1% (25 samples) and 43.7% (14 samples) of 32 dry-salted and dried milkfish products contained more than 5 mg/100 g and 50 mg/100 g of histamine, respectively [21].


**Table 2.** Distribution of the histamine content in the 20 brined and dried milkfish products.

High levels of histamine have been found in various types of milkfish implicated in HFP. Our research group detected 61.6 mg/100 g of histamine in dried milkfish products that were implicated in an incident of HFP [6]. Two fried milkfish sticks implicated in a poisoning incident contained 86.6 mg/100 g and 235.0 mg/100 g of histamine [10]. The high content of histamine (i.e., 91.0 mg/100 g) in a suspected milkfish surimi product could be the etiological factor for this fish-borne poisoning in Taiwan [11]. Therefore, it is also very important for people, especially those from the Indo-Pacific region, such as the Philippines, Indonesia, and Taiwan, to be aware that milkfish products could become a hazardous food item, causing histamine poisoning.

Pearson correlation was conducted to determine if there existed any relationship among the pH, moisture, water activity (aw), salt content, TVBN, APC, coliform, and histamine contents of the tested 20 samples. In general, positive correlations existed between moisture and aw (*r*, correlation coefficient = 0.81, *p* < 0.05), TVBN and APC (*r* = 0.76, *p* < 0.05), APC and histamine (*r* = 0.71, *p* < 0.05), and histamine and TVBN (*r* = 0.76, *p* < 0.05). However, negative correlations were noted between moisture and salt content (*r* = −0.73, *p* < 0.05), and aw and salt content (*r* = −0.76, *p* < 0.05).

#### *3.2. E*ff*ect of Brine Concentrations on the Quality of Brined and Dried Milkfish*

Changes in the moisture and water activity (aw) of the milkfish samples pre-immersed in different brine concentrations (i.e., 0%, 3%, 6%, 9%, and 15%) during a sun-drying period of five days are shown in Figure 1. The initial moisture of the fish samples was 70.3%, while the moisture of all fish samples rapidly decreased with increasing drying time. At the end of the drying period, the moisture content in all of the samples ranged from 44.2% to 46.9%, and no significant differences (*p* > 0.05) were observed among the samples of the various brine concentrations and control samples (Figure 1A). For all fish samples with an initial aw value of 0.985, the aw values gradually decreased with an increase in the drying time and reduced to 0.967 in the control sample, 0.959 in the 3% NaCl sample, 0.950 in the 6% NaCl sample, 0.945 in the 9% NaCl sample, and 0.942 in the 15% NaCl sample at the end of the sun-drying period (Figure 1B). It was found that the milkfish samples with higher brine concentrations had lower aw values (*p* < 0.05).

**Figure 1.** Changes in the moisture (**A**) and water activity (aw) (**B**) in the milkfish samples as a result of brine-salting with 0% (control), 3%, 6%, 9%, and 15% NaCl during sun-drying. Each value represents the mean ± SD of three replications. Different lower letters indicate significant differences (*p* < 0.05) within the data at the end of the sun-drying period.

Changes in the pH and salt content of the milkfish samples pre-immersed in different brine concentrations (i.e., 0%, 3%, 6%, 9%, and 15%) over a sun-drying period of five days are presented in Figure 2. The pH values of the milkfish samples slightly increased from the initial reading of 5.41 to 5.69 for the control sample, 5.70 for the 3% and 6% NaCl samples, 5.87 for the 9% NaCl sample, and 5.89 for the 15% NaCl sample at the end of the sun-drying period. The increase in the pH for all of the group samples may be due to the formation of basic components, including ammonia, trimethylamine, and other amines by bacterial spoilage [22]. Moreover, the final pH values of the 9% and 15% NaCl samples were higher (*p* < 0.05) than those of the control and the 3% and 6% NaCl samples (Figure 2A). As shown in Figure 2B, the salt content in the fish sample slightly increased from 0.05% to 0.13% in the control sample, 0.20% to 0.70% in the 3% NaCl sample, 0.51% to 1.17% in the

6% NaCl sample, 0.85% to 2.24% in the 9% NaCl sample, and 1.62% to 2.87% in the 15% NaCl sample after give days of sun-drying. The results also show that the milkfish samples pre-immersed in a higher brine concentration had a higher salt content (*p* < 0.05).

**Figure 2.** Changes in the pH (**A**) and salt content (**B**) of the milkfish samples as a result of brine-salting with 0% (control), 3%, 6%, 9%, and 15% NaCl during sun-drying. Each value represents the mean ± SD of three replications. Different lower letters indicate significant differences (*p* < 0.05) within the data at the end of the sun-drying period.

Figure 3 shows the changes in the TVBN and TBA values in the milkfish samples pre-immersed in different brine concentrations (i.e., 0%, 3%, 6%, 9%, and 15%) during a sun-drying period of five days. Initially, the milkfish samples had 13.7 mg/100 g of TVBN, and subsequently, the TVBN content in all fish samples increased gradually while drying, reaching 34.0 mg/100 g for the control sample, 30.5 mg/100 g for the 3% NaCl sample, 29.76 mg/100 g for the 6% NaCl sample, 27.0 mg/100 g for the 9% NaCl sample, and 26.9 mg/100 g for the 15% NaCl sample at the end of the sun-drying period. Thus, the highest TVBN level was detected in the control sample, followed by the 3% and 6% NaCl samples, and the lowest levels were observed for the 9% and 15% NaCl samples (*p* < 0.05) (Figure 3A). Connell [23] revealed that the increase in TVBN is due to the production of volatile basic compounds, including ammonia, trimethylamine and dimethylamine, via decomposition by autolytic enzymes and spoilage bacteria. Moreover, Nooralabettu [15] demonstrated that the addition of NaCl addition in Bombay duck can decrease autolytic enzyme activity in fish meat. An increase in salt content above 1% in fish can have an inhibitory effect on the bacteria associated with fish spoilage [24]. Consequently, the high content of TVBN in the unsalted samples (i.e., the control sample) obtained in this study was probably due to the increasing decomposition by enzymes and spoilage bacteria with the lack of salt's inhibitory effect.

**Figure 3.** Changes in total volatile basic nitrogen (TVBN) (**A**) and thiobarbituric acid (TBA) (**B**) in the milkfish samples as a result of brine-salting with 0% (control), 3%, 6%, 9%, and 15% NaCl during sun-drying. Each value represents the mean ± SD of three replications. Different lower letters indicate significant differences (*p* < 0.05) within the data at the end of the sun-drying period.

Thiobarbituric acid (TBA), a measure of MDA as a secondary lipid oxidation product, is one of the most widely used indicators for the assessment of food lipid oxidation [25]. Initially, the TBA values for the control and brined samples were 2.18 MDA mg/kg. The value of TBA in all of the samples increased during the sun-drying period, reaching 5.9 MDA mg/kg for the control sample, 6.5 MDA mg/kg for the 3% NaCl sample, 8.6 MDA mg/kg for the 6% NaCl sample, 11.5 MDA mg/kg for the 9% NaCl sample, and 11.4 MDA mg/kg for the 15% NaCl sample at the end of the sun-drying period. In contrast to TVBN, the highest levels of TBA were observed in the 9% and 15% NaCl samples, followed by the 6% NaCl sample, and the lowest TBA level was detected in the control and 3% NaCl samples (*p* < 0.05) (Figure 3B). Yanar et al. [26] also reported that hot-smoked tilapia samples treated with a 15% brine concentration contained very high levels of TBA. Sodium chloride can promote lipid oxidation, while sodium ions may replace iron from myoglobin, thereby resulting in free iron ions for the catalysis of lipid oxidation [26,27]. Therefore, the results in this study reveal that the high TBA values in the samples prepared with 9% and 15% brine concentrations may be attributed to the addition of NaCl by accelerating the rate of lipid oxidation. In addition, when seafood is dried by exposure to sunlight, lipids can be oxidized and low molecular weight carbonyl components can be produced [28]. The results of this study are in agreement with a previous study reporting that the TBA values of dried yellow corvina increased rapidly during sun-drying [28].

Figure 4 shows the changes in APC and coliform bacteria in the milkfish samples pre-immersed in different brine concentrations (i.e., 0%, 3%, 6%, 9%, and 15%) over a sun-drying period of five days. The APC numbers of the milkfish sample gradually increased from the initial population of 3.21 to 6.88 log CFU/g for the control sample, 6.81 log CFU/g for the 3% NaCl sample, 6.15 log CFU/g for the 6% NaCl sample, 6.0 log CFU/g for the 9% NaCl sample, and 5.86 log CFU/g for the 15% NaCl sample at the end of the sun-drying period. Thus, the APC bacteria detected in the control and 3% NaCl samples were markedly higher (*p* < 0.05) than those of other brine concentration samples (Figure 4A) and exceeded the 6.47 log CFU/g Taiwanese regulatory standard. Similar to the APC population, the growth of coliform in this fish samples was considerably faster in the unsalted (control) sample than in the other brined samples (*p* < 0.05). The coliform counts in the control, 3%, 6%, 9%, and 15% NaCl samples increased to 3.51, 2.87, 2.75, 2.70, and 2.41 log MPN/g, respectively, at the end of the sun-drying period (Figure 4B). These results are in agreement with our previous report, in which the APC and coliform levels of dry-salted and sun-dried milkfish samples decreased with increasing salt concentrations [21]. A similar finding was also reported by Yang et al. [14], who found that higher brine-salting could inhibit the growth of bacteria in grass carp. Moreover, higher brine concentrations (>6%) in the milkfish samples obviously had a repressive action on microbiological growth in this study, indicating that salt content is able to inactivate or inhibit bacteria.

**Figure 4.** Changes in aerobic plate count (APC) (**A**) and coliform (**B**) in the milkfish samples as a result of brine-salting with 0% (control), 3%, 6%, 9%, and 15% NaCl during sun-drying. Each value represents the mean ± SD of three replications. Different lower letters indicate significant differences (*p* < 0.05) within the data at the end of the sun-drying period.

Figure 5 shows that the histamine content in the control sample increased gradually during the sun-drying period, reaching 4.8 mg/100 g by the end. On the other hand, the histamine contents in the 3%, 6%, 9%, and 15% NaCl samples only slightly increased during the sun-drying period, reaching 2.8, 2.0, 0.79, and 0.27 mg/100 g, respectively, by the end. In conclusion, the histamine content observed in the control sample was markedly higher (*p* < 0.05) than that of the other brine concentrations samples (Figure 5). These results agree with the previous research of Hwang et al. [21], where high contents of histamine at 67 mg/100 g were found in unsalted dried milkfish samples via sun-drying. The low levels of histamine (<2.8 mg/100 g) detected in the salted samples (>3% NaCl) in this study may be due to the growth reduction of histamine-forming bacteria by the preservative effect of salt,

indicating that the addition of salt could be effective in reducing or inhibiting histamine accumulation. In our previous study, high levels of aw, moisture, TVBN, APC, and histamine were detected in unsalted dried milkfish samples produced by sun-drying; therefore, dried milkfish producers should be aware that dried milkfish with low salt and sun-drying periods could become a vehicle for histamine poisoning [21]. Similarly, since high levels of TVBN (>30 mg/100g), APC (>6.81 log CFU/g), and histamine (>2.8 mg/100 g) were observed in the unsalted and 3% NaCl samples during the sun-drying period, brined and dried milkfish manufacturers should pay attention to the fact that dried milkfish brined with a low amount of salt (<3% NaCl) and a sun-drying period could lead to worse hygienic quality and potential hazards, such as food poisoning. However, the samples with higher brine concentrations (>9% NaCl) had higher TBA values (>11.4 MDA mg/kg) (Figure 3B). With regard to an assessment of APC, TBA, TVBN, and histamine, this study suggests that dried milkfish brined with a 6% NaCl addition has better chemical and bacteriological quality.

**Figure 5.** Changes in the histamine of milkfish samples as a result of brine-salting with 0% (control), 3%, 6%, 9%, and 15% NaCl during sun-drying. Each value represents the mean ± SD of three replications. Different lower letters indicate significant differences (*p* < 0.05) within the data at the end of the sun-drying period.

Pearson correlation was conducted to determine if there existed any relationship among the moisture, aw, pH, salt content, TVBN, TBA, APC, coliform, and histamine contents of the samples at the end of the sun-drying period. In general, positive correlations existed between APC and aw (*r* = 0.95, *p* < 0.05), APC and histamine (*r* = 0.88, *p* < 0.05), coliform and aw (*r* = 0.93, *p* < 0.05), coliform and histamine (*r* = 0.90, *p* < 0.05), aw and TVBN (*r* = 0.88, *p* < 0.05), salt content and TBA (*r* = 0.85, *p* < 0.05), aw and histamine (*r* = 0.89, *p* < 0.05), and histamine and TVBN (*r*=0.86, *p* < 0.05). However, negative correlations were noted between salt content and APC (*r* = −0.89, *p* < 0.05), salt content and coliform (*r* = −0.90, *p* < 0.05), salt content and aw (*r* = −0.92, *p* < 0.05), aw and TBA (*r* = −0.90, *p* < 0.05), and salt content and histamine (*r* = −0.95, *p* < 0.05).

#### **4. Conclusions**

This study, aimed at investigating the hygienic quality of 20 brined and dried milkfish products, revealed that the APC numbers in seven samples (35%) exceeded the 6.47 log CFU/g Taiwanese regulatory standard. Moreover, 25% of the tested samples had histamine contents greater than the

5 mg/100 g recommended by the USFDA in their guideline levels, and 10% (2/20) of the fish samples had >50 mg/100 g of histamine. After the consumption of these samples, histamine fish poisoning could occur. In addition, the chemical and bacteriological quality of the brined and dried milkfish pre-immersed in various brine concentrations during a sun-drying period were observed in this study. Although the samples prepared with higher brine concentrations presented a retarded APC growth rate and a reduced formation of TVBN and histamine, as compared with the control sample, they produced higher TBA values. It is suggested that 6% NaCl for brined milkfish is the optimal condition for maintaining the quality of brined and dried milkfish. Our results could suggest that application of brine concentration information is effective in controlling quality and enhancing the safety of brined and dried milkfish products.

**Author Contributions:** Conceptualization, C.-C.H., Y.-C.L., and Y.-H.T.; methodology, C.-C.H., Y.-H.T., and H.-H.C.; analysis, C.-C.H., Y.-C.L., and H.-F.K.; data curation, Y.-C.L., H.-F.K., and C.-C.H.; writing—original draft preparation, C.-C.H. and Y.-H.T.; writing—review and editing, C.-C.H., Y.-H.T., and C.-Y.H.; supervision, Y.-C.L. and H.-F.K.; project administration, C.-C.H.; funding acquisition, C.-C.H. and Y.-H.T. All authors read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the Ministry of Science and Technology, R.O.C. (Contract No. MOST 107-2635-E-412-001).

**Acknowledgments:** The authors gratefully acknowledge Shinn-Lih Yeh, a director of Mariculture Research Center, Fisheries Research Institute with providing equipment, Su-Sing Liu with the operation and maintenance of the equipment, and partly financial support from higher education sprout project of National Kaohsiung University of Science and Technology.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Biogenic Amine Contents and Microbial Characteristics of Cambodian Fermented Foods**

## **Dalin Ly 1,2,\*, Sigrid Mayrhofer 1, Julia-Maria Schmidt 1, Ulrike Zitz <sup>1</sup> and Konrad J. Domig <sup>1</sup>**


Received: 30 January 2020; Accepted: 11 February 2020; Published: 15 February 2020

**Abstract:** Naturally fermented foods are an important part of the typical diet in Cambodia. However, the food safety status of these products has not been widely studied. The aim of this study was, therefore, to provide an overview of the quality of these foods in relation to microbiology and biogenic amines. Additionally, the obtained results were compared to the habits and practices of Cambodians in handling this type of food. A total of 57 fermented foods (42 fishery and 15 vegetable products) were collected from different retail markets in the capital of Cambodia. Pathogenic *Salmonella* spp., *Listeria* spp., and *Listeria monocytogenes* were not detected in 25 g samples. Generally, less than 10<sup>2</sup> cfu/g of *Staphylococcus aureus*, *Escherichia coli*, *Pseudomonas* spp., Enterobacteriaceae, and molds were present in the fermented foods. *Bacillus cereus* group members (<102 to 2.3 <sup>×</sup> 10<sup>4</sup> cfu/g), lactic acid bacteria (<10<sup>2</sup> to 1.1 <sup>×</sup> <sup>10</sup><sup>7</sup> cfu/g), halophilic and halotolerant bacteria (<10<sup>2</sup> to 8.9 <sup>×</sup> 106 cfu/g), sulfite-reducing *Clostridium* spp. (<102 to 3.5 <sup>×</sup> 10<sup>6</sup> cfu/g), and yeasts (<102 to 1.1 <sup>×</sup> 106 cfu/g) were detected in this study. Still, the presence of pathogenic and spoilage microorganisms in these fermented foods was within the acceptable ranges. Putrescine, cadaverine, tyramine, and histamine were detected in 100%, 89%, 81%, and 75% of the tested products, respectively. The concentrations of histamine (>500 ppm) and tyramine (>600 ppm) were higher than the recommended maximum levels in respectively four and one of 57 fermented foods, which represents a potential health risk. The results suggest that the production process, distribution, and domestic handling of fermented foods should be re-evaluated. Further research is needed for the establishment of applicable preservation techniques in Cambodia.

**Keywords:** Cambodian fermented foods; microbial characteristics; biogenic amines; food quality; food safety

#### **1. Introduction**

Cambodia is an agricultural country that has a tropical climate with two distinct monsoon seasons (dry and rainy seasons). Thus, the availability of certain products is not stable through the year, and food preservation and storage are required to maintain the food supply [1]. Since food spoilage is mainly caused by microorganisms, preventing their access to susceptible foods is one method of food preservation. Another one is the inhibition of microbial growth through fermentation, salting, drying, or smoking, as it is common in Cambodia [2,3]. The storage time of food depends on factors that affect the growth of spoilage microorganisms like intrinsic food characteristics (e.g., pH, *aw*, composition) and extrinsic parameters (e.g., temperature, relative humidity, atmospheric gases). Due to higher ambient temperatures and moisture, food spoils faster in the tropics. As a result, it is not surprising

that food security issues are reported in densely populated tropical cities [2]. However, the majority of the Cambodian population lives in rural areas where poverty is high and access to drinking water, electricity, and sanitation is limited [4].

Fermented foods are an important part of the typical diet in Cambodia [1]. Since Cambodia has an extensive network of waterways, freshwater fish, along with marine, fermented and preserved fish, is a major component of the diet of most Cambodians [5]. Fermented fishery products are consumed daily as main dishes, side dishes, or condiments/seasonings [5]. Additionally, they are applied as flavor enhancers due to their delicacy and high nutritional properties [6,7]. Vegetables also play an essential role in daily dishes for their nutrient content. The availability of certain fresh vegetables, however, does not last throughout the year. Depending on the varieties of domestic raw vegetables, many types of fermented vegetables have emerged in Cambodia with the popularity of traditional fermentation [8]. In the meantime, these foods have become a common part of the Cambodian diet [1,8].

Cambodian fermented foods are produced through knowledge that is passed on from generation to generation and from person to person [1]. The great majority of fermented products are locally produced by smallholders, many of them women, and sold in traditional wet markets where women also predominate as retailers [9]. Most of them are illiterate and have a poor knowledge about hygiene practices. Additionally, the awareness of food safety is limited in Cambodia. The quality of these foods is influenced by raw materials, processing methods, and climate, but there is no quality control of these determinants as well as of the finished products in Cambodia [1]. Fermented foods are generally not labelled with an appropriate shelf-life and usually stored at room temperature until they are completely consumed [1]. Fermented fishery products are usually cooked before consumption. However, fish paste and sauce can also be eaten raw and are often mixed with chili or lemon juice [1]. In contrast, fermented vegetables are normally considered as ready-to-eat (RTE) foods. As a result, it is not surprising that foodborne outbreaks are common in Cambodia [8]. But there is no coordinated food surveillance program and little analytical data regarding microbiological or chemical contamination of food are present [10]. Nevertheless, food safety is a key priority of the Cambodian government [11], and efforts to improve foodborne disease surveillance and food safety are being undertaken [10].

*Escherichia coli*, *Cronobacter sakazakii*, *Enterobacter* spp., opportunistic non-Enterobacteriaceae, *Staphylococcus* spp., and *Listeria* spp. have already been detected in fermented vegetables in Cambodia [8,12]. Furthermore, potentially pathogenic bacteria such as *Bacillus*, *Clostridium*, and *Staphylococcus* were found in traditional Cambodian fermented fish products [5]. Next to microbiological contamination, chemical substances can lead to acute poisoning or even long-term diseases such as cancer [13]. The most prevalent ones are biogenic amines (BAs) and biotoxins [14]. BAs are low molecular weight organic molecules, formed by microbial decarboxylation of their precursor amino acids or by transamination of aldehydes and ketones by amino acid transaminases [15]. Beside spoilage, preservative technological processes such as fermentation, salting, and ripening may increase BA formation in food. As BAs are thermostable, they cannot be inactivated by thermal treatment [13]. The most common BAs found in foods and beverages are histamine (HIS), tyramine (TYR), putrescine (PUT), and cadaverine (CAD) [16,17]. Low levels of BAs in food are not considered as a serious risk. However, when high amounts of BAs are consumed, various physiological effects may occur, namely, hypotension (in the case of HIS, PUT, and CAD) or hypertension (in the case of TYR), nausea, headache, rash, dizziness, cardiac palpitation, and even intracerebral hemorrhage and death in very severe cases [18]. BAs with more severe acute effects for human health are HIS and TYR [19]. PUT and CAD have low toxicological properties on their own, but they can act as precursor of carcinogenic N-nitrosamines when nitrite is present. These two BAs also potentiate the effects of HIS and TYR by inhibiting their metabolizing enzymes [19]. HIS is the only BA with regulatory limits [20]. In addition to their potential toxicity, BAs are also used to evaluate the hygienic quality of foods, as their levels in food can be an indirect indicator of excessive microbial proliferation [19].

Baseline surveillance data are essential to monitor the disease burden of fermented foods in Cambodia. To obtain such data, the physicochemical properties (pH, *aw*, and salt content) as well as the presence of certain microorganisms (spoilage and pathogenic bacteria) and the concentrations of the BAs HIS, TYR, PUT, and CAD were determined in 57 Cambodian fermented food samples within this study. The main purpose of this manuscript is to give an overview of the quality of Cambodian fermented foods, to correlate physicochemical parameters with BA contents, and to describe the prevailing habits and practices of Cambodians in dealing with this type of food.

#### **2. Materials and Methods**

#### *2.1. Sample Collection*

Fifty-seven samples of naturally fermented foods (42 raw fermented fish and 15 RTE fermented vegetable products) were randomly purchased from wet markets in Phnom Penh, the capital city of Cambodia. These products originated from various provinces of the country. Fermented fishery samples included fish sauce (teuktrey; *n* = 7), fish paste (prahok; *n* = 12), shrimp paste (kapi; *n* = 6), fermented fish (paork chav; *n* = 7; mam trey; *n* = 3)*,* sour fermented fish (paork chou; *n* = 3), and salted fish (trey proheum; *n* = 4). Fermented vegetables such as salty fermented radish (chaipov brey; *n* = 3), sweet fermented radish (chaipov paem; *n* = 3), fermented melon (trasork chav; *n* = 3), fermented mustard (spey chrourk; *n* = 3), and fermented papaya (mam lahong; *n* = 3) were bought on the next day. Detailed information about each fermented product is provided in Table 1. After purchasing, the samples were immediately packed into plastic boxes and stored at their storage temperature. One day later, the samples were transported to Vienna by airplane. At the Department of Food Science and Technology, Vienna, the samples were checked shortly after arrival and kept in their original containers at 4 ◦C until analysis. All samples were analyzed within the usual shelf-life of the products [21].

#### *2.2. Physicochemical Properties Analysis*

#### 2.2.1. Determination of pH and Water Activity (*aw*)

The pH value was determined by penetrating the spear tip of the Blueline 21 pH electrode (Schott AG, Mainz, Germany) into the samples. The pH values were then measured using a digital pH meter (Schott Lab 870, Mainz, Germany).

The *aw* value was measured using the digital water activity meter Rotronic Hygropalm HP23-AW-A (Rotronic, Zurich, Switzerland) after equilibration at room temperature (~25 ◦C).

#### 2.2.2. Determination of Salt Content (NaCl)

The salt content was analyzed by potentiometric precipitation titration of chloride-ions with the 877 Titrino plus-Titrator equipped with a calomel electrode (Metrohm AG, Herisau, Switzerland). The protocol was performed according to the producer with minor modifications. Depending on the expected salt content, 1 to 10 g ± 0.01 g (a) of the samples was weighed into a glass beaker and filled with distilled water to 200 g ± 0.01 g (b). Subsequently, the samples were homogenized for 2 min at 9500 rpm using an Ultra Turrax T25 (IKA, Germany). Fifty grams ± 0.01 g (c) of the homogenized samples was weighed into a new glass beaker, and 50 mL of distilled water was added. Afterwards, 2 mL (2 M) HNO3 was added. The samples were then titrated with 0.1 M AgNO3. With the obtained results, the salt content of the original samples was calculated as % (*w*/*w*) NaCl = *V* × *M* × 0.0584 × 100/m; where *V* and *M* are the volume and molarity of the AgNO3 standard solution used. The initial sample weight is *m*, which was calculated considering the sample preparation: *m* = *a* × *c*/*b*. The test was conducted in duplicate.



#### *2.3. Microbiological Analysis*

From each fermented product, a 10 g sample was aseptically collected, transferred to a stomacher bag, and homogenized (Stomacher 400 Circulator, Seward, UK) with 90 mL of buffered peptone water for 45 s at 230 rpm. Appropriate dilutions of the samples were prepared using the same diluent, and 0.1 or 1 mL aliquots of each dilution were applied on various selective media using the spread plate method or pour plate method. Lactic acid bacteria (LAB) were anaerobically grown on DeMan Rogosa Sharpe agar (MRS, Merck, Darmstadt, Germany) at 30 ◦C for 72 h according to ISO 15214 [22]. Enterobacteriaceae were enumerated using the pour plate method with an additional overlay on Violet Red Bile Dextrose agar (VRBD, Merck, Darmstadt, Germany) and an incubation at 37 ◦C for 24 h according to ISO 21528-2 [23]. *Pseudomonas* spp. were detected by plating appropriate dilutions on Cephalothin-Sodium Fusidate-Cetrimide agar (CFC, Oxoid, Hampshire, UK) and incubation at 25 ◦C for 44 h based on ISO 13720 [24]. Yeasts and molds were determined according to ISO 21527-2 [25] using the spread plate method on Dichloran Glycerol agar (DG18, Merck, Darmstadt, Germany). The plates were incubated at 25 ◦C for 5–7 d and counted on the 5th and 7th day of incubation. Selected yeast colonies were confirmed by methylene blue staining and microscopy [26]. Halophilic and halotolerant bacteria were counted after an incubation at 30 ◦C for 2–4 d on Tryptone Soya agar (TS, Oxoid, Hampshire, UK) supplemented with 10% NaCl (Roth, Karlsruhe, Germany) [27]. *Staphylococcus aureus* was enumerated on Baird Parker agar (BP, Merck, Darmstadt, Germany), which was incubated at 37 ◦C for 24 h based on ISO 6888-1/AMD 1 [28]. The plates were evaluated again after an additional 24 h incubation. The confirmation of colonies was performed using Gram-stain and DNase agar (Oxoid, Hampshire, UK) according to Kateete et al. [29]. The number of presumptive *Bacillus cereus* group members was investigated by spreading dilutions on Mannitol Yolk Polymyxin agar (MYP, Merck, Darmstadt, Germany) and incubating plates at 30 ◦C for 18–48 h. The evaluation was also performed after 18 h and 48 h of incubation. Colonies were confirmed by endospore staining [30]. *E. coli* was enumerated by pour plating on Tryptone Bile Glucuronic medium (TBX, Oxoid, Hampshire, UK) with an incubation at 44 ◦C for 18–24 h based on ISO 16649-2 [31]. The presence of sulfite-reducing *Clostridium* spp. (SRC) was analyzed by pour plating with an additional overlay on Sulfite-Cycloserin agar (SC, Oxoid, Hampshire, UK). Plates were anaerobically incubated at 37 ◦C for 20 h. Confirmation tests were performed using Lactose-Gelatine medium (Conda, Madrid, Spain) and Motility-Nitrate medium (Conda, Madrid, Spain) according to ISO 7937 [32]. The presence of *Salmonella* spp. was investigated using the VIDAS UP (BioMerieux, Crappone, France) *Salmonella* (SPT) system, whereas that of *Listeria* spp. and *L. monocytogenes* was tested by the VIDAS LDUO (BioMerieux, Crappone, France) system. VIDAS SPT and VIDAS LDUO are based on an enzyme-linked fluorescent immunoassay. The preparation of the samples was similar to the previous method, but 25 g of the sample was weighted in instead of 10 g. After pre-enrichment (*Listeria* spp.) and enrichment (*L. monocytogenes*, *Salmonella* spp.) steps, which were carried out according to the manufacturer's manual, the assay steps were performed automatically by the instrument.

#### *2.4. Determination of Biogenic Amines (BAs)*

The concentrations of BAs in the supernatant were analyzed by reverse-phase HPLC (Waters 2695 Separations Module, Waters, MA, USA) according to the method of Šimat et al. and Saarinen et al. [33,34]. Briefly, 1 g of the homogenized sample was extracted overnight with 5 mL of 0.4 M perchloric acid (Merk, Darmstadt, Germany). Then, the sample was centrifuged at 5000 rpm for 10 min and the supernatant was kept for further analysis. For derivatization, 80 μL of 2 M NaOH (Roth, Karlsruhe, Germany), 120 μL of saturated sodium bicarbonate solution (Merck, Darmstadt, Germany), and 400 μL of derivatization reagent (1% dansyl chloride in acetone; prepared daily; Fluka, Seelze, Germany) were added to 400 μL of sample solution. The sample was mixed and incubated for 45 min at 40 ◦C. Afterwards, 60 μL of 1 M ammonia solution (Roth, Karlsruhe, Germany) was added, mixed, and incubated in the dark for 60 min at room temperature. Finally, 940 μL acetonitrile (Roth, Karlsruhe, Germany) was added. The sample was mixed and centrifuged for 10 min at 13,400 rpm. A RP-18 column (Li Chro CART 250-4, 5 μm,

Merck, Darmstadt, Germany) with a LiChroCART 4-4 Guard Column (RP-18, 5 μm) (Merk, Darmstadt, Germany) and a manu-CART NT cartridge holder (Merck, Darmstadt, Germany) was used for separation. The flow rate was 1 mL/min, the column temperature was 40 ◦C, and the injection volume was 20 μL. The mobile phase A consisted of 0.1 M ammonium acetate (Roth, Karlsruhe, Germany) and the mobile phase B was 100% acetonitrile. The following gradient was used for the separation: time = 0 min, 50% A and 50% B; time = 19 min, 10% A and 90% B; time = 20 min, 50% A and 50% B; time = 28 min, 50% A and 50% B. The detection was performed by UV–vis (Waters 2489 UV-visible detector, Waters, MA, USA) at a wavelength of 254 nm. Heptylamine (Fluka, Seelze, Germany) was used as an internal standard that was well separated from other compounds. The specificity of the method was checked using standard mixtures of 12 BA chemicals, which included spermine tetrahydrochloride, spermidine trihydrochloride, ethanolamine, isopropylamine, histamine dihydrochloride, putrescine (1,4-diaminobutan dihydrochloride), methylamine hydrochloride, agmatine sulfate, cadaverine (1,5-diaminopentan dihydrochloride), tyramine hydrochloride, dimethylamine hydrochloride, and pyrrolidine. Except for the last four chemicals, which were from Sigma-Aldrich (St. Louis, MO, USA), all chemicals were from Fluka. Standard stock solutions of BAs were prepared at 500 mg/L in 0.01 M perchloric acid. The stock solutions were diluted with 0.4 M perchloric acid to obtain series of working standard solutions (0.25, 1, 5, 10, and 15 mg/L). The derivatization procedure was the same as for the samples. All compounds were separated and could be identified by their retention times. The linearity of the method was tested by analyzing the series of working standard solutions. The correlation coefficients for the linear regression lines were better than 0.99 for all compounds. The limit of detection (LOD = 3x standard deviation of y-residuals of low concentrations/slope of calibration curve) of all BAs ranged between 0.5 and 1.5 ppm, and the limit of quantification (LOQ = 10x standard deviation of y-residuals of low concentrations/slope of calibration curve) ranged between 1.5 and 4.8 ppm. PUT, CAD, HIS, and TYR were analyzed in duplicate.

#### *2.5. Statistical Analysis*

Result units of quantitative microbiological analyses were cfu/g. The physicochemical parameters and BA concentrations results were analyzed with statistical analyses using the Statistical Package for the Social Sciences (SPSS, Version 20.0.0 for Windows, 2011; IBM Co., Somers, NY, USA). Data were analyzed for the degree of variation by calculating the mean and standard deviations (SDs) of the results. The significance of differences was evaluated using analysis of variance (ANOVA). A *p* value of less than 0.05 was considered statistically significant. The least-squares difference (LSD) test was used to determine the significance of differences in the physicochemical parameters and BA contents among the samples. The relationship value was determined using the Pearson correlation coefficient.

#### **3. Results**

#### *3.1. Physicochemical Characteristics in Fermented Foods*

Physicochemical parameters such as pH, *aw*, and salt content were measured and compared to discuss possible causes for the different BA levels. Table 2 shows the physicochemical parameters of the tested fermented products. The pH values in fermented fishery samples were in the range of 4.4 to 7.6. Lower pH values were found in paork chav (fermented fish), while higher values were detected in kapi (shrimp paste) and trey proheum (salted fish). The *aw* values of the fermented fishery products ranged from 0.69 to 0.84. The lowest *aw* value (0.69) was detected in kapi (shrimp paste) and the highest (0.84) in paork chav (fermented fish). The salt contents were in the range of 6% to 34%, with the lowest value (6%) found in trey proheum (salted fish) and the highest (34%) in prahok (fish paste). In the fermented vegetables, the pH values were between 3.6 and 5.5. Lower pH values were found in spey chrourk (fermented mustard, 3.6–3.9) and mam lahong (fermented papaya, 3.7–3.8), while the highest value was detected in chaipov brey (salty fermented radish) (4.6). The *aw* values in these fermented products were between 0.75 and 0.97. The highest salt concentration (25%) was

found in chaipov brey (salty fermented radish), while the lowest (2%) was detected in spey chrourk (fermented mustard) (Table 2).


**Table 2.** Physicochemical characteristics (pH, *aw*, % NaCl) of fermented food products.


Based on the statistical analysis (ANOVA) of fermented fishery products, there were significant differences (*p* < 0.05) between the physicochemical parameters of teuktrey (fish sauce) and those of paork chav (fermented fish) and trey proheum (salted fish). There was no statistically significant difference (*p* > 0.05) among the samples of teuktrey (fish sauce) and prahok (fish paste), and of paork chav (fermented fish), paork chou (sour fermented fish), and mam trey (fermented fish). Statistical analysis of fermented fish and vegetables samples was conducted separately. Regarding the fermented vegetable products, the physicochemical values of chaipov brey (salty fermented radish) were significantly different from that of spey chrourk (fermented mustard) and mam lahong (fermented papaya) (*p* < 0.05), while no significant difference was found among chaipov paem (sweet fermented radish) and trasork chav (fermented melon) (*p* > 0.05) (Table 2).

#### *3.2. Presence of Microorganisms*

Counts of LAB (<10<sup>2</sup> to 1.1 <sup>×</sup> <sup>10</sup><sup>6</sup> cfu/g), halophilic and halotolerant bacteria (<10<sup>2</sup> to 8.9 <sup>×</sup> <sup>10</sup><sup>6</sup> cfu/g), Enterobacteriaceae (<10<sup>2</sup> cfu/g), *Pseudomonas* spp. (<10<sup>2</sup> cfu/g), yeasts (<102 to 1.1 <sup>×</sup> 10<sup>6</sup> cfu/g), and molds (<sup>&</sup>lt; <sup>10</sup><sup>2</sup> to 2.3 <sup>×</sup> 102 cfu/g) from the different types of fermented fish tested are indicated in Table 3. Table 3 also presents the results regarding the *B. cereus* group members (<102 to 2.3 <sup>×</sup> 104 cfu/g), SRC (<10<sup>2</sup> to 3.5 <sup>×</sup> 106 cfu/g), *S. aureus,* and *E. coli* (<102 cfu/g, respectively).

The microbial profiles found in fermented vegetables are displayed in Table 3 as well. The LAB counts were in the range of <102 to 1.1 <sup>×</sup> 107 cfu/g. The highest LAB counts were detected in spey chrourk (fermented mustard) and mam lahong (fermented papaya). Halophilic and halotolerant bacteria were found in numbers ranging from 2 <sup>×</sup> 102 to 5.5 <sup>×</sup> 104 cfu/g. The counts of *B. cereus* group members ranged from <102 to 1.2 <sup>×</sup> 10<sup>4</sup> cfu/g. SRC and yeasts were detected in the range of < 102 to 1.5 <sup>×</sup> 103 cfu/g and <10<sup>2</sup> to 2.6 <sup>×</sup> 105 cfu/g in the tested vegetable samples, respectively (Table 3). The counts of all other microorganisms were <102 cfu/g.


**3.**Microbialprofilesfoundinfermentedfoodproducts.

a Lactic acid bacteria; b

sulfite-reducing

 clostridia.

#### *Foods* **2020** , *9*, 198

#### *3.3. Quantification of Biogenic Amines (BAs) in Fermented Foods*

Table 4 shows the BA contents of 57 fermented product samples. The detection limits in this study were <0.5 ppm (PUT, CAD, and TYR) and <2 ppm for HIS. The results indicate that PUT was detected in quantifiable amounts in all tested fishery samples (100%), while CAD, TYR, and HIS concentrations were quantified in approximately 95%, 88%, and 86% of these products, respectively. PUT and CAD were the most frequently detected BAs in the tested samples. The highest concentrations of PUT (830 ppm), CAD (2035 ppm), HIS (840 ppm), and TYR (691 ppm) were detected in paork chav (fermented fish). PUT concentrations in 42 fishery samples were in the range between 23 to 830 ppm, with the lowest (23 ppm) presented in paork chou (sour fermented fish) and the highest (830 ppm) found in paork chav (fermented fish). The concentrations of HIS in the quantifiable fishery products (86%) ranged from 32 to 840 ppm (Table 4). The current results show that, overall, less than 50 ppm HIS was determined in all kapi (shrimp paste) samples. The concentrations of TYR in the quantifiable fishery samples (88%) ranged from 10 to 691 ppm (Table 4). In general, lower levels of TYR were detected in kapi (shrimp paste) and paork chou (sour fermented fish) than in other fermented fishery products in this study.

**Table 4.** Contents of biogenic amines in fermented food products.


\* PUT, putrescine; CAD, cadaverine; HIS, histamine; TYR, tyramine; - Ranged values (minimum to maximum) ¥ Mean ± SD; # ND, not detected (Limit of detection <sup>&</sup>lt; 0.5 ppm for PUT, CAD, and TYR; <sup>&</sup>lt;2 ppm for HIS). Values with different superscript letters in the same column indicate significant differences at *p* < 0.05 by LSD test. Statistical analysis of fermented fish and vegetable samples was conducted separately.

The four types of BAs were also analyzed for the safety evaluation of fermented vegetables from Cambodia. The BA levels varied among the collected RTE fermented vegetables (Table 4). PUT, CAD, TYR, and HIS were detected in 100%, 73%, 60%, and 47% of the fermented vegetables, respectively. The ranges of the quantifiable BAs were from 11 to 197 ppm for PUT, 10 to 118 ppm for CAD, 18 to 103 ppm for HIS, and 7 to 86 ppm for TYR (Table 4). The results clearly show that most BA concentrations in the five types of fermented vegetables were less than 100 ppm. Even no HIS and TYR could be detected in chaipov brey (salty fermented radish) and chaipov paem (sweet fermented radish) samples (Table 4).

According to one-way ANOVA and LSD tests of 42 fermented fisheries samples, statistically significant differences (*p* < 0.05) were found among the detected concentrations of PUT, CAD, HIS, and TYR in each product type. The statistical analysis of 15 fermented vegetable samples showed no

statistical difference (*p* > 0.05) among concentrations of PUT, CAD, and HIS, while TYR concentrations were statistically different (*p* < 0.05) (Table 4).

Analyzing the correlation between total BA contents and the physicochemical parameters pH, *aw*, and salt content (%) in 42 fermented fishery products, a weak positive relationship between total BAs and *aw* values (*r* = 0.22, *p* > 0.05; *n* = 42), and a weak negative with pH values (*r* = −0.22, *p* > 0.05; *n* = 42) were found. There was no correlation among total BAs and salt content *(r* = 0.00, *p* > 0.05; *n* = 42) (Figure S1A–C). Furthermore, the linear functions between total BA contents and parameters of pH (*r* = −0.57, *p* < 0.05; *n* = 15) and salt content (*r* = −0.81, *p* < 0.05; *n* = 15) were characterized by a moderate and strong negative correlation coefficient, respectively, while a strong positive correlation between total BAs and *aw* value (*r* = 0.79, *p* < 0.05; *n* = 15) were determined in fermented vegetable products (Figure S2A–C).

#### **4. Discussion**

#### *4.1. Physicochemical Characteristics in Fermented Foods*

Based on the physicochemical results, types of fermented fishery products were more different than those of fermented vegetables. Nevertheless, the pH values of this study are in good agreement with those of fermented fish products in Thailand, Vietnam, Laos, Myanmar, China, Korea, Japan, Malaysia, and Taiwan [5,35–38]. The results of the salt concentration analysis are also consistent with previous data for fermented fish products [5,38], shrimp paste [6], and fish sauce [37]. The aw values of fermented fish products were comparable to fermented fish products from other countries, for example, Thai shrimp paste (0.65–0.72) and Indonesian fermented fish (0.75–0.93) [39,40].

The pH values found in the fermented vegetables were between 3.6 and 5.5 (Table 2). This is in agreement with a previous study, which reported that the pH of Cambodian fermented vegetables ranged from 3.6 to 6.5, depending on the raw materials and processing techniques [8]. Chaipov brey (salty fermented radish) was found to have the highest salt value (25%) of the fermented vegetables. Salty fermented radish with high salt concentrations (20–25%) has also been reported elsewhere [8]. As salt reduces *aw*, the lowest *aw* values were also determined in these samples (0.75–0.76) (Table 2).

Growth of microorganisms in foods are mainly influenced by the *aw* and pH [41]. The addition of salt, in turn, has an inhibitory effect on the growth of microorganisms due to its impacts on the *aw* value [42].

#### *4.2. Microbiological Parameters in Fermented Foods*

Microorganisms associated with fermented foods are commonly present on the external surface and in the pre- and post-harvest environment of raw materials. Additionally, they exist in the gill and gut of seafood [14].

Regarding *Bacillus* spp. and *Clostridium* spp., our results are comparable to those of Chuon et al., who also analyzed Cambodian traditional fermented fish sauce, fish paste, and shrimp paste [5]. Such traditionally home-prepared salted or fermented products are often associated with foodborne botulism [43]. However, routine testing for *C. botulinum* to ensure food safety is not recommended [43]. Instead, SRC have been proposed to identify risks from *C. botulinum* [43]. In addition to *C. botulinum*, *C. perfringens*—the most important of the SRC—poses a frequent problem and challenge in fish industry [44]. It is estimated that 10<sup>5</sup> to 10<sup>8</sup> cfu/g *C. perfringens* are capable of generating toxinfection [44]. Foodborne diseases that have *C. perfringens* as causative agent are related to inadequate storage, processing, and food service operations [44]. Nevertheless, no *C. perfringens* could be confirmed within this study. It is known that >105 cfu/g *B. cereus* group members are potentially harmful for human consumption [45]. None of the fermented products exceeded this limit (Table 3). The survival of *B. cereus* in low numbers in several fermented products, including those based on fish and vegetables, has already been described [46]. The inactivation of this pathogen could be attributed to the presence of organic acids or higher salt concentrations [46]. Moreover, no *S. aureus* could be quantified (<10<sup>2</sup> cfu/g)

in the tested samples, which is in contradiction to a previous study [5]. In addition, it is reported that *S. aureus* is uniquely resistant to adverse conditions such as low *aw* values (0.83), high salt contents, and pH stress. Thus, most strains can grow over an *aw* and pH range of 0.83 to >0.99 and 4.5–9.3, respectively [47]. Although 11 of all 57 food products tested (19.3%) had an *aw* value in the range specified above, 10 of them (e.g., all trasork chav (fermented melon), spey chrourk (fermented mustard), and mam lahong (fermented papaya), and one chaipov paem (sweet fermented radish) product had a pH < 4.5 (Table 2). Overall, *S. aureus* could only have grown in one sample. Although *L. monocytogenes* appears to be relatively tolerant to acidic conditions, no representatives of this species as well as of other *Listeria* species were verified, which may be due to the low *aw* (<0.9) of most food samples tested (90%). Also, less than 102 cfu/g of *Pseudomonas* spp., Enterobacteriaceae, and *E. coli* were detected in all products examined (Table 3). Furthermore, no *Salmonella* spp. could be determined using the VIDAS system. These Gram-negative bacteria are often inhibited by a salt concentration >10%, an *aw* value <0.95, a pH value <3.8 or >9.0 (depending on the acidulant), and the fermentation process itself [48]. LAB are not only responsible for the fermentation, they also significantly contribute to the flavor, texture, and nutritional value of fermented products [48], produce effective antimicrobial agents, and are the primary preservation factor in fermented fish products [49]. However, LAB are generally only tolerant to moderate salt concentrations (10%–18%). Consequently, their counts decrease as the salt concentration increases [50]. Forty-three samples (~75%) of all fermented products tested in this study contained more than 10% salt (Table 3). LAB were only present in high numbers in samples with less than 10% salt (Table 2 and 3).

Since typical spoilage bacteria are generally non- or only slightly halotolerant (e.g., pseudomonads, enterobacteria), the extensive use of salt is another technological process for food preservation besides fermentation [51]. Up to 25% and 34% salinity was respectively determined for fermented vegetable and fishery products in this study. Classifying the various products according to their salt content (e.g., 0–10%, 11–20%, >20%, data not shown), the numbers of halophilic and halotolerant bacteria generally decreased with increasing salinity.

The unfavorable conditions for bacterial growth (high salt content, a low pH or *aw*) may result in higher yeasts and mold numbers. These microorganisms are quite salt-tolerant [51]. As recommended by the European Food Safety Authority (EFSA), the accepted limit for molds in foods is <10<sup>3</sup> cfu/g [52]. As shown in Table 3, all 57 fermented food products were acceptable regarding molds. It has been reported that <106 cfu/g of yeasts are acceptable in RTE foods placed on the market [53]. An unsatisfactorily higher yeasts count (>10<sup>6</sup> cfu/g) was only found in one paork chav (fermented fish) sample, which could lead to spoilage by acid and gas production [45,53]. However, the limit was just exceeded marginally (Table 3).

According to different organizations and previous studies [45,52–56], the detected counts of the investigated microorganisms in this study are satisfactory. Thus, the fermented foods tested are suitable for human consumption regarding the microbiological quality.

#### *4.3. Formation of BAs in Fermented Foods*

A deviation in BA concentrations within a specific food category is probably due to intrinsic food characteristics such as pH and *aw* values, nutrients, and microbiota, as well as extrinsic factors including storage time, temperature, and manufacturing processes [57–59]. This may explain the wide variation of BA concentrations between the fermented fishery products and even within the same tested product type. Shalaby (1996) stated that BA levels differ not only between different food varieties but also within the same variety [57]. However, no significant difference was observed in fermented vegetables within this study.

Fish species associated with a high amount of histidine belong to the families *Scombridae, Clupeidae, Engraulidae, Coryphenidae, Pomatomidae*, and *Scombreresosidae* [60]. As seen in Table 1, the fish species of some fermented fish products belong to the families *Engraulidae* and *Scombridae* for teuktrey (fish sauce)*, Engraulidae* for paork chou (fermented fish) and *Clupeidae* for mam trey (fermented fish). Hence, these products contained higher HIS amounts (Table 4). In contrast, low HIS and TYR contents are reported in crustaceans such as shrimp [61]. Corresponding values were determined for six kapi (shrimp paste) samples within this study. Fresh fruits and vegetables such as melon, cabbage, radishes, and cucumber contain lower HIS levels; however, papaya is considered as a HIS liberator [62]. Mustard is generally an allergen, and sometimes listed as moderately high in HIS [63]. Accordingly, HIS has been found in all fermented spey chrourk (fermented mustard) and mam lahong (fermented papaya) samples, but only in one trasork chav (fermented melon) and in no chaipov (fermented radish) sample within this study. TYR has been detected in more fermented vegetable samples than HIS, although in lower concentrations. TYR and CAD have been described in few vegetables in relatively low concentrations [64]. In contrast, it has been reported that PUT is the most common BA found in food of plant origin. It is particularly abundant in vegetables [64,65] and fermented products [60]. As seen in Table 4, this BA was the only one that was verified in all fermented vegetable samples with relatively high values.

The possible involvement of molds and yeasts in BA (especially CAD and PUT) accumulation is still discussed [19]. However, it is known that different genera, species, and strains of Gram-positive and Gram-negative bacteria are able to produce BAs by the action of microbial decarboxylases [66]. In particular, Enterobacteriaceae were identified as HIS-producing bacteria, but also halophilic and halotolerant bacteria (among other representatives of the families Enterobacteriaceae, Pseudomonadaceae, and the genera *Photobacterium*, *Vibrio*, and *Staphylococcus*), LAB, *Bacillus* spp., and *Clostridium* spp. were said to be capable of HIS formation [14,57,67]. According to Rodriguez-Jerez et al., microbial species with the capacity to form HIS and those with the capacity to form other BAs are similar [68]. Thus, Enterobacteriaceae were also reported to produce PUT, CAD, and to a lesser extent TYR. These BAs have also been detected when testing various *Bacillus* strains [67]. However, TYR should be mainly formed by LAB (*Lactobacillus*, *Enterococcus*) during fermentation [16]. Next to strains of the genera *Clostridium*, *Pseudomonas,* and *Staphylococcus*, LAB (*Enterococcus*, *Lactococcus*) are also involved in the production of PUT. It should be kept in mind that decarboxylase activities are often related to strains rather than to species or genera [69]. The capabilities of such strains, in turn, vary depending on the type and even batch of food product from which the strains are isolated [67].

The main factors affecting microbial activities in food are temperature, salt concentration, and pH [19]. Most fermented foods in this study had a pH value within the range of 3.0 to 6.0 (79%, Table 2), providing an acidic environment. The transcription of many decarboxylase genes is induced by a low pH value, which improves the fitness of the microbial cells subjected to acidic stress [19]. As the decarboxylation of amino acids is a mechanism of BA-forming bacteria to counteract acidic stress and to adapt to environmental conditions, their decarboxylase activity increases, resulting in higher BA concentrations [7,58]. Hence, it contributed to higher BA contents (Table 4). This effect could be confirmed within this study, as weak and moderate negative correlations (*r* = −0.22 and −0.57) were respectively found between the total BA contents and pH values for fermented fishery and fermented vegetable products. A strong negative linear fit (*r* = −0.81) could be detected between total BAs and salt content in fermented vegetables, whereas there was no correlation between these parameters in fermented fishery products (Figures S1 and S2). In general, increasing salt concentrations contribute to the reduction of BA accumulation in foods, mainly reducing the metabolic activities of decarboxylase-positive microorganisms [19] as it may have been the case for the fermented vegetables. However, a possible enhancing effect of NaCl on the BA production has also been described [19]. Thus, stressed cells seem to activate the decarboxylating pathways in the framework of more complex response systems [19] being probably more present in fermented fishery than vegetable products. In contrast, the rate of BA accumulation decreases with the decrease of *aw* values due to the water loss [19]. Correspondingly, a positive correlation should be observed between total BA contents and *aw* values. In fact, weak and strong positive relationships were determined for fermented fishery (*r* = 0.22) and vegetable (*r* = 0.79) products, respectively.

The ability of microorganisms to produce BAs is limited by low temperature [19]. Within this study, samples were stored at 4◦C until investigation. However, fermented fishery and vegetable products are usually stored at room temperature in Cambodia due to the given conditions. Thus, even higher BA amounts could be expected for these products in this country. Paork chav (fermented fish) and other fermented fish products are stored at room temperature up to a year [21]. In the case of fermented vegetables, the salt content seems to be particularly relevant for the storage time. Thus, vegetables with 5–6% salt should be sold as soon as possible, while chaipov brey (salty fermented radish) samples with high salt concentrations (20–25%) have a longer storage time [8]. In this regard, higher BA values were detected in fermented vegetables with lower salt contents (2–5%).

#### *4.4. BAs and Food Safety*

Table 5 shows the distribution of the tested fermented food products according to the different allowable limits. Several organizations have set legal maximum limits on HIS concentrations in fermented foods that should ensure safe human consumption if the limits are not exceeded. Such organizations are the US Food and Drug Administration (FDA) with 50 ppm, FAO/WHO with 200 ppm for fish and fishery products, respectively, and EFSA with 400 ppm for fishery products that have undergone enzyme maturation treatment in brine [60,70,71]. The HIS level in fish sauce has been regulated in particular by the Codex Alimentarius Commission (CAC) and EFSA, with a maximum allowable limit of 400 ppm [55,72]. Correspondingly, the contents of HIS in all teuktrey (fish sauce) products did not exceed 400 ppm (Table 5). Thus, the levels of HIS in teuktrey (fish sauce) products in the current study can be regarded as safe for human consumption according to EFSA and CAC. Due to numerous outbreaks with toxic HIS concentrations ≥500 ppm [60], one paork chou (sour fermented fish), one mam trey (fermented fish), and two paork chav (fermented fish) products, representing about 7% (4/57) of the fermented products (Table 5), could pose a health risk. Although a food safety criterion is only set for HIS, HIS is not the only BA responsible for health hazards. Healthy individuals should also not be exposed to TYR values of 600 ppm or more by meal as recommended by EFSA [60]. The concentrations of TYR in all tested products were less than 600 ppm (Table 4), except for one paork chav (fermented fish) sample, which may constitute a health hazard [60]. Nevertheless, this sample is still fine according to Prester et al., who suggested a dietary value of up to 800 ppm of TYR as acceptable. More than 1080 ppm are toxic for adults [61].


**Table 5.** Distribution of fermented foods with quantifiable histamine contents.

TYR concentrations from <0.4 to 270.6 ppm in commercially Chinese fish sauces [73], from 77.5 to 381.1 ppm in Korean anchovy sauces [74], and from 0 to 1178 ppm in commercial fish sauces the Far East sold at German markets [75] were reported. Hence, the concentrations of TYR in teuktrey (fish sauce) from retail markets in Cambodia were generally within or even below these concentrations (Table 4).

It has also been reported that the acute toxicity levels of TYR and CAD are respectively greater than 2000 ppm and the oral toxicity level of PUT is 2000 ppm [15]. It is known that TYR has a stronger and more rapid cytotoxic effect than HIS [76]. Unlike HIS and TYR, the pharmacological activities of PUT and CAD seem to be less potent. Nonetheless, both amines show in vitro cytotoxicity at concentrations easily reached in inherently BA-rich foods [77] and enhance the toxicity of HIS and TYR [19]. In the tested teuktrey (fish sauce) products, the highest levels of PUT (404 ppm) and CAD (766 ppm) were higher than those in fish sauce sold at Malaysian markets (242.8 ppm PUT and 704.7 ppm CAD) [78] and Chinese markets (276.6 ppm PUT and 606.3 ppm CAD) [73] but much lower than the levels in imported fish sauce products sold at German markets (1257 ppm PUT and 1429 ppm CAD) [75], Austrian markets (510 ppm PUT and 1540 ppm CAD) [65], and other European markets (1220 ppm PUT and 1150 ppm CAD) [60]. Extremely high PUT and CAD contents characterize inferior fish sauces, which may be due to the minor production hygiene, less salt content (<20%), the type of fish species, and storage condition [19,61]. Nevertheless, a health risk from consuming such a fish sauce is likely to be excluded due to the relatively small average intake [75]. Interestingly, the PUT concentrations of fish sauce samples were generally lower than the associated CAD concentrations (Table 4). The complexity of fish protein, which releases more lysine (precursor of CAD) during the fermentation of fish sauce, resulting in increased CAD concentrations could be the reason [78]. The current results also show that the highest PUT and CAD values in fish and shrimp pastes were higher than those in paste products in Taiwan [35] and in the Maldives [79]. Generally, BAs were detected in low levels in the tested fermented vegetables, which should not cause any risk when consumed. These results were consistent with previous studies [60,80], which reported that fermented vegetables should be considered as low-risk products in terms of BAs.

#### *4.5. BAs and Food Quality*

The HIS content alone may be a reliable indicator of food safety, but not of food quality. TYR and CAD are used as spoilage index [81]. Other authors have considered PUT and CAD as spoilage indicators [82]. Furthermore, PUT and CAD increase with longer storage times [19] and give strong unpleasant decaying odors at very low concentrations [75]. Therefore, the PUT and CAD concentration could be used as quality indicator [19], and their accumulation should be avoided [15,77]. These BAs are also included in the biogenic amine index (BAI) [83]. The BAI, the sum of HIS, TYR, PUT, and CAD, is more indicative of food quality, as these BAs are mostly produced at the end of shelf-life, indicating spoilage [83,84]. This index was also established to facilitate the evaluation and comparison of the BA concentrations in food. However, the usefulness of the BAI as quality index depends on many factors, mainly concerning the nature of the product (e.g., fresh or fermented food). Owing to the number of different factors (e.g., fermentation, maturation, starters), BA amounts vary much more in fermented products [13]. Thus, the BAI has proven to be more satisfactory for fresh products, and there is a BAI for freshwater fish of 50 ppm [85], while it is missing for fermented fishery products. The only BAI for a fermented food product was given by Wortberg and Woller [84] for Bologna sausages at 500 ppm. The higher BAI mainly results from the fermentation process and/or ripening. Using this limit, about one-third (31%) of the fermented fishery products in this study had a BAI of less than 500 ppm, while two-thirds (69%) had a higher BAI (>500 ppm), indicating a poor hygienic quality (Figure S3). Of the fermented vegetables, about 13% (2/15) had a BAI higher than 300 ppm (Figure S4). This value corresponds to the sum of HIS, TYR, PUT, and CAD, which should not be exceeded by acceptable sauerkraut [57].

In view of these results, the production process, distribution, and domestic handling of fermented products should be re-evaluated under strict hygienic practices together with the hazard analysis critical control point (HACCP) approach in order to minimize the content of BAs and microbiological contamination. The storage of food by cooling or freezing requires electricity that is not available to all Cambodians. Therefore, preservation techniques, such as the use of antimicrobial substances and/or autochthonous starter cultures, which are characterized by the absence of any BA formation ability or the presence of a BA detoxification activity, should be tested for their possible application on an industrial and small scale.

#### **5. Conclusions**

The presence of microorganisms in the examined fermented samples presented no health risk since pathogenic and spoilage microorganisms were in acceptable ranges. Nevertheless, one paork chou (sour fermented fish), one mam trey (fermented fish), and two paork chav (fermented fish) products represent a health risk because of the high level of HIS (>500 ppm). One of the paork chav (fermented fish) samples additionally exceeded the recommended TYR maximum (>600 ppm) per meal. The totals of all BAs tested were higher than the recommended corresponding BAI values in about 69% of the tested fermented fishery and 13% of the vegetable products. This may indicate a poor hygienic quality for these products. Hence, the production process, distribution, and domestic handling of fermented products in Cambodia should be re-evaluated in order to minimize the content of BAs and microbiological contamination. Further research is required to establish preservation techniques that could be applied on an industrial and small-scale in Cambodia.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/9/2/198/s1, Figure S1: Linear fitting between total biogenic amines and physicochemical parameters, including pH (A), water activity (B), and salt content (C), in fermented fishery products (*n* = 42). Each dot indicates a data set obtained from a single sample, Figure S2. Linear fitting between total biogenic amines and physicochemical parameters, including pH (A), water activity (B), and salt content (C), in fermented vegetable products (*n* = 15). Each dot indicates a data set obtained from a single sample, Figure S3. Biogenic amine index (BAI) evaluated for 42 samples of fermented fishery products. The bold horizontal line describes the limit value of 500 ppm, which is used to distinguish between fermented fishery products of good and poor hygienic quality, Figure S4. Biogenic amine index (BAI) evaluated for 15 samples of fermented vegetable products. The bold horizontal line describes the limit value of 300 ppm, which is used to distinguish between fermented vegetable products of good and poor hygienic quality.

**Author Contributions:** Conceptualization, D.L., K.J.D., and S.M.; Analysis, D.L.; Investigation, J.-M.S.; Resources, U.Z. and K.J.D.; Writing—original draft, D.L.; Writing—Review and Editing, D.L., S.M., and K.J.D.; Supervision, K.J.D. and S.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed by the European Commission for the Erasmus Mundus Action 2 under the ALFABET project (the reference number: 552071) and partially funded by the Schlumberger Foundation Faculty for the Future Program for supporting the first author to pursue a Ph.D.

**Acknowledgments:** Special thanks is given to Vibol San for his valuable contribution. This work was supported by the European Commission for the Erasmus Mundus scholarship under the ALFABET project Reference number: 552071 and partially funded by the Schlumberger Foundation, Faculty for the Future Program for supporting the first author to pursue a Ph.D.

**Conflicts of Interest:** The authors confirm that they have no conflict of interest with respect to the work described in this manuscript.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Identification of a Lactic Acid Bacteria to Degrade Biogenic Amines in Chinese Rice Wine and Its Enzymatic Mechanism**

#### **Tianjiao Niu 1,2, Xing Li 1, Yongjie Guo <sup>2</sup> and Ying Ma 1,\***

<sup>1</sup> School of Chemistry and Chemical Engineering, Harbin Institute of Technology, Harbin 150090, China

<sup>2</sup> Mengniu Hi-tech Dairy (Beijing) Co., Ltd., Beijing 101107, China

**\*** Correspondence: maying@hit.edu.cn

Received: 28 June 2019; Accepted: 31 July 2019; Published: 2 August 2019

**Abstract:** A *L. plantarum*, CAU 3823, which can degrade 40% of biogenic amines (BAs) content in Chinese rice wine (CRW) at the end of post-fermentation, was selected and characterized in this work. It would be an optimal choice to add 106 cfu/mL of selected strain into the fermentation broth to decrease the BAs while keeping the character and quality of CRW. Nine amine oxidases were identified from the strain and separated using Sephadex column followed by LC-MS/MS analysis. The purified amine oxidase mixture showed a high monoamine oxidase activity of 19.8 U/mg, and more than 40% of BAs could be degraded. The biochemical characters of the amine oxidases were also studied. This work seeks to provide a better solution to degrade BAs in CRW prior to keeping the character and quality of CRW and a better understanding of the degradability of the strain to the BAs.

**Keywords:** biogenic amines; *L. plantarum*; amines oxidase; Chinese rice wine; industrial fermentation

#### **1. Introduction**

Biogenic amines (BAs) are low molecular weight organic compounds that have been identified as toxicological agents in various foods, such as fishery products, dairy, meat, wine, and so on [1,2]. The ingestion of foods containing relatively high concentrations of BAs could lead to several health hazards, such as headaches, hypotension, respiratory distress, heart palpitations and digestive problems, particularly when alcohol is present [3,4]. Histamine, which is well-known because of its implication in many food poisoning cases, has a potent vasodilatory action that could cause important drops in blood pressure [5]. Tyramine, as one of the vasoconstrictor amines, can provoke a release of noradrenaline resulting in an increase of arterial pressure [5]. Even though there are no accurate regulations for BAs, several countries including France, Germany and Australia have set regulations and limits for histamine and many wine importers in the EU require a BA analysis [4,6]. The presence of BAs is considered a marker of poor wine quality and bad winemaking practices [4,7].

BAs are synthesized in fermented food by decarboxylation of corresponding amino acids by microorganisms [1]. According to the previous studies, BAs could be formed by lactic acid bacteria in wine [8,9], Chinese rice wine [10] and Korean rice wine [1]. As a traditional alcoholic beverage, Chinese rice wine (CRW), which has been popular in China for thousands of years [11], has high nutritional values, and thus, it has been used as an ingredient in traditional Chinese medicine [12]. Since the brewing process of CRW is the typical open semisolid-state fermentation, lots of microorganisms (molds, yeast, bacteria) are brought in the glutinous rice with the addition of Chinese koji [3,13], and the system is favorable to BAs generation combining with the high amount of free amino acids [2]. The abundant bacteria in CRW, mainly originating from Chinese koji, the surroundings and the surfaces of the equipment, could be one of the main reasons for the formation of BAs [10].

Histamine, tyramine, putrescine, cadaverine and phenylethylamine are the most representative BAs detected in the wine [6]. Histamine and tyramine have been considered as the most toxic products in wine, and putrescine and cadaverine could potentiate these effects [4]. The formation of BAs was traditionally controlled by avoiding the growth of spoilage bacteria, decreasing the amino acid precursors and inoculating starter cultures with negative decarboxylase activity [6,7]. Driven by greater awareness of the importance of food quality and safety by consumers, the methods for degradation of BAs in fermented foods have been explored. Biological enzymatic degradation of BAs would be a safe and economic way while avoiding the production difficulties. Two *Lactobacillus plantarum* strains (named NDT 09 and NDT 16) isolated from red wine were able to degrade 22% of tyramine and 31% of putrescine, respectively [14]. Three different strains of *Brevibacterium linens* were utilized to eliminate tyramine and histamine in cheese [6], and the strain *K. varians* LTH 1540, it was also found, could degrade tyramine during sausage ripening [15]. Two lactic acid bacteria were used to degrade 50%–54% of histamine in fish silage [16]. However, the relationship between BAs degradation and microbiological enzymes of the strains has not been explored yet.

In this work, a *Lactobacillus plantarum* was obtained from CRW which could degrade BAs. The optimal industrial conditions of the selected strain were analyzed, and the microbiological amine oxidase enzymes were identified and biochemically characterized. Our results could receive considerable interest by providing a green industrial strategy to control the BAs contents in the rice wine and improve the safety consumption of the fermented foodstuffs.

#### **2. Materials and Methods**

#### *2.1. Materials*

Man Rogosa Sharpe agar (MRS) medium was obtained from Oxoid. Ltd. (Basingstoke, Hants, UK). The BA standards were purchased from Sigma-Aldrich (St. Louis, MO, USA). Bacterial genomic DNA extraction kit was obtained from Tiangen (Beijing, China). Ultra-pure water was obtained from a Millipore purification system (>18.3 MΩ·cm). Formic acid, methanol and acetonitrile used in the preparation of the mobile phase were of LC-MS grade. All other chemicals used were of analytical grade.

#### *2.2. Strains Screening and Identification*

Fermentation broths were collected at the later stage from a typical rice wine production process in Shaoxing (Zhejiang, China). The suspension was filtered through four layers of sterile gauze to remove the unliquefied rice and sealed in a sterile plastic bottle. One gram of fermentation broths was diluted 10-fold by a 0.85% NaCl solution and routinely subcultured 5 to 10 times on MRS medium to obtain purified clones. The screening medium designed was based on the method of Landete [17] to obtain the bacteria that could decrease biogenic amine content. These strains isolated were kept frozen at −20 ◦C in a sterilized mixture of culture medium and glycerol (50:50, *v*/*v*) according to the methods described by García-Ruiz [18], and further identified by 16S rRNA gene sequencing.

#### *2.3. HPLC Determination of Biogenic Amines*

Eight biogenic amines of Histamine (HIS), tyramine (TYR), putrescine (PUT), cadaverine (CAD), phenylethylamine (PHE), tryptamine (TRY), spermine (SPM) and spermidine (SPD) were analyzed according to the method of Callejon, Sendra [13] with slight modifications. The individual strains were cultured on MRS, and 10<sup>7</sup> cfu/mL were inoculated with the MRS liquid medium contaminated with 50 mg/L of each amine at pH 5.5. After 48 h incubation at 30 ◦C, the reaction was stopped by adding HCl. Samples were centrifuged at 8000 rpm for 15 min and the supernatant was pipetted into a screw-capped vial. The pre-column derivatization procedure using dansyl chloride as derivatization reagent was performed according to the report of Yongmei, Xin [12]. The samples were filtered through 0.22 μm millipore syringe filters and analyzed by RP-HPLC using on LC-20A HPLC system (Shimadzu, Kyoto, Japan) with an Agilent C18 column (250 mm × 4.6 mm, 300 A pores, 5 μm particles, Agilent Technologies, Inc., Santa Clara, CA, USA). The column temperature was kept at 30 ◦C and the detection wavelength was 254 nm with a flow rate of 1.0 mL/min by using water (A) and methanol (B) as eluents. The gradient elution program consisting of a linear gradient from 65% to 70% B in 7 min followed by from 70% to 80% B in 13 min and 3 min isocratic elution.

The percentage of BAs degradation was calculated based on the HPLC data as following,

BAs degradation (%) = (Ccontrol − Cstrain)/Ccontrol

where Ccontrol was the concentration of the BAs in the control medium and Cstrain was the concentration of the BAs in the medium incubated with the strain.

#### *2.4. Bacterial Growth Analysis*

The bacterial growth was measured according to the methods described by Cui [19]. Briefly, the isolated lactic acid bacteria (LAB) strains were diluted to 10<sup>5</sup> cfu/mL in MRS liquid medium, and the pH and optical density (OD600 nm) of medium was checked at 28 ◦C, 33 ◦C and 37 ◦C for 36 h, respectively.

#### *2.5. The Bacterial Starter Application in Pilot Scale Fermentation*

A pilot fermentation was performed according to the methods described by Zhang, Xue [10] with modifications (Figure 1). Glutinous rice (12 kg) was soaked at 18 ◦C for 20 h and steamed for 30 min. After naturally cooling to room temperature (about 25 ◦C), the steamed rice was transferred into a 33 L wide-mouth bottle to which 14.5 kg water, 1.5 kg Chinese koji (unique saccharifying agent including molds, yeasts and bacteria, obtained from COFCO Shaoxin wine Co., Ltd., Shaoxin, China) were added. The main fermentation was carried out at 33 ◦C for 4 days with intermittent oxygen filling, and post-fermentation was then carried out at 28 ◦C for 20 days. The isolated strain with 10<sup>5</sup> (low level), 10<sup>6</sup> (middle level) and 10<sup>7</sup> (high level) cfu/mL was added into the CRW at the main fermentation and post-fermentation stage, respectively. After filter pressing, clarification, wine frying and sterilization (90 ◦C for 3 min), finished Chinese rice wines were obtained. Ten milliliters of fermentation broths were taken from different fermentation stages, including addition of starter (AS); main fermentation (MF); post-fermentation 5d (PF5d); post-fermentation 10d (PF10d); and post-fermentation 20d (PF20d)), to analysis the changes in the BAs contents by using the HPLC method. According to the previous studies [20,21], pH, alcohol content, total sugar, total acid, non-sugar solid and amino acid nitrogen of CRW were analyzed by using official methods (Chinese National Standard GB/T 13662-2008). Sensory evaluation of CRW was conducted by 30 panelists (15 males and 15 females) who have professional training certificates. The procedure was conducted in a sensory laboratory following GB/T 13662-2008 and ISO 4121. A total of 11 sensory attributes of appearance (color and turbidity), aroma (alcohol, fruit and cereal), taste (sweet, sour and bitter), mouthfeel (astringency, continuation and full body) and harmony were chosen to characterize the sensory properties using quantitative descriptive analysis involving a 0–9 ten-point linear scale (0: none; 1–2: very weak; 3–4: ordinary; 5–6: moderate; 7–8: strong; 9: very strong).

**Figure 1.** Diagram of the Chinese rice wine production process.

#### *2.6. Separation of the Amine Oxidases*

Cell-free extracts were obtained by using the method of Callejon [22]. The bacterial cells froma1L culture were collected by centrifugation at 10,000 rpm, 20 min at 4 ◦C and washed twice with 50 mM sodium phosphate buffer (PBS), pH 7.4. The samples were resuspended in PBS buffer containing 1 mM of phenyl methylsulfonyl fluoride (PMSF) as protease inhibitor. Cell-free extracts were obtained by disrupting the bacterial cells with 1 g of 106 μM diameter glass beads in a Mikro-dismenbrator® Sartorius: 10 cycles of 40 s, alternating 5 cycles of disruption with a cooling step of 5 min in ice. The samples were centrifuged at 13,000 rpm for 15 min (PrismR, Labnet, USA), and supernatants were saved at −20 ◦C until use. The protein content was determined by using the bicinchoninic acid assay kit (BCA, Solarbio, Beijing, China). Monoamine oxidase (MAO) assay kit and diamine oxidase (DAO) assay kit (Jiancheng Institute, Nanjing, China) were both used to determine the amine oxidase activity. The MAO assay kit was based on the ability of MAO to form H2O2 substrate, which could be determined by a fluorimetric method. The DAO assay kit was based on the oxidation of PUT to pyrroline plus NH3 and H2O2, which can be determined by the fluorimetric method.

The cell-free extracts were further ultracentrifuged at 47,000 rpm for 1 h, and the supernatant was precipitated by 75% saturation of ammonium sulfate precipitation [22]. The protein was redissolved with 50 mM PBS and were loaded onto a Sephadex G-100 column (1.6 cm × 70 cm) followed by a linear gradient elution with a flow rate of 1 mL/min. The protein fraction was collected and measured at 280 nm by using a HD-93-1 spectrophotometer (Purkinje General Instrument Co. Ltd., Beijing, China). There fractions were collected (P1, P2 and P3, Supplement Figure S1), and were then concentrated and freeze-dried. The degradation ability of the fractions was further evaluated by incubation with 50 mg/L eight biogenic amines at pH 4.0, 33 ◦C for 2 h.

#### *2.7. Identification of the Amine Oxidases*

The fractions separated from the cell-free extracts were digested with trypsin (Promega, Madison, WI, USA) overnight at 37 ◦C and were identified by LC-MS/MS using the Easy nLC-1000 nano ultra-high-pressure system (Thermo Fisher Scientific, San Jose, CA, USA) coupling with a Q Exactive mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). The peptide mixture was loaded onto a Zorbax 300SB-C18 peptide traps (Agilent Technologies, Wilmington, DE, USA) in buffer A (0.1% Formic acid) and separated with a linear gradient of 4%–50% buffer B (80% acetonitrile and 0.1% formic acid) for 50 min, 50%–100% B for 4 min, and held at 100% B for 6 min at a flow rate of 250 nL/min. The mass spectrometer was operated in positive ion mode. MS data was acquired using a data-dependent top10 method dynamically choosing the most abundant precursor ions from the survey scan for high-energy collisional dissociation (HCD) fragmentation and was searched by using MASCOT engine and Proteome Discoverer 1.3 against the local uniport\_lactobocilluspiantarum database.

#### *2.8. Enzymatic Properties of the Amine Oxidases*

Effects of temperatures (15, 20, 25, 28, 30, 35, 40, 80 ◦C at pH 4.0 for 2 h), pH (3.0–5.0) at 30 ◦C for 2 h, and metal ions (0.2 mol/L, copper ion, ferrous ion, zinc ion, calcium ion and magnesium ion) at 30 ◦C for 2 h (pH 4.0) on the amine oxidase degradation activity were further investigated.

#### *2.9. Statistical Analysis*

All samples were prepared in three independent and each was analyzed in triplicate by the analysis of variance (ANOVA). The results were considered significant at *p* ≤ 0.05 by the Duncan test.

#### **3. Result**

#### *3.1. Strains Screening and Identification*

A total of 61 strains were isolated from the five major stages (soaking rice, steamed rice, addition of starter, main fermentation and post-fermentation, Figure 1) of CRW fermentation. After screening their potentials to degrade/eliminate the contents of BAs, about 30% of strains were able to degrade BAs even though most of them degraded BAs to less than 10% extents (results not known). Only one strain drew attentions for more than 40% degradation efficiency of the BAs (Table 1). 16S rDNA sequencing identified that the strain had 100% similarity in 16S rDNA sequences to *Lactobacillus plantarum* CAU 3823 (GenBank accession no. MF424991.1). In the details, *Lactobacillus plantarum* CAU 3823 was a *L. plantarum* that exhibited the greatest potential for BAs degradation, as 56% degradation, for TRY, 41% for PHE, 42% for PUT, 43% for CAD, 40% for TYR, 45% for HIS, 44% for SPD and 43% for SPM, which should be considered in the further analysis.

**Table 1.** Percentage (%) of degradation of the biogenic amines by *Lactobacillus plantarum* CAU 3823 from Chinese rice wine a.


<sup>a</sup> 107 cfu/mL of *Lactobacillus plantarum* CAU 3823 was incubated in the Man Rogosa Sharpe agar (MRS) liquid medium contaminated with 50 mg/L of each amine at pH 5.5 for 48 h.

#### *3.2. The Bacterial Growth Ability*

The growth ability of *L. plantarum* CAU 3823 at different temperatures (28 ◦C, 33 ◦C and 37 ◦C) was shown in Figure 2. *L. plantarum* CAU 3823 was able to grow at different temperatures, showing OD600 > 1 at main fermentation temperature (33 ◦C) for 9–25 h and post-fermentation temperature (28 ◦C) for 12~25 h. The maximum OD600 value of 1.4 was found at different temperatures at 25 h of growth, suggesting the good growth trends indicated that *L. plantarum* CAU 3823 could be used in industry producing CRW fermentation.

**Figure 2.** The growth ability of *L. plantarum* CAU 3823 at different temperatures (28 ◦C, 33 ◦C and 37 ◦C). Growth curves are representative of all determinations.

#### *3.3. Changes in the BAs Induced by L. plantarum in Pilot Scale Fermentation*

To investigate the capability to degrade BAs of *L. plantarum* CAU 3823 to the BAs in pilot scale fermentation of CRW, RP-HPLC was applied to quantify the contents of BAs in CRW incubation with various levels (105, 106 and 107 cfu/mL) of *L. plantarum* CAU 3823 as extra starter during fermentation, and the results are shown in Figure 3. Compared to control group, the total contents of BAs in CRW with *L. plantarum* CAU 3823 were significantly lower (*p* < 0.05) during the entire fermentation period (Figure 3A). The degradation percentages of BAs were 32%, 54% and 58%, respectively, at low, middle and high level of *L. plantarum* CAU 3823 at the main fermentation stage, suggesting the dose dependent manner. Total content of BAs was significantly reduced to 34%, 60% and 61% at low, middle and high levels of *L. plantarum* CAU 3823, respectively, at 5th day of post-fermentation, and similar degradation efficiency was obtained in the 10th day of post-fermentation and 20th day of post-fermentation, respectively.

The degrading abilities of *L. plantarum* CAU 3823 to TRY, PUT, HIS, CAD, PHE, SPD and SPM were also studied in Figure 3B–H, respectively. A marked decrease in the contents of BAs was observed during fermentation with the increasing of the strain content. As the most content of BAs detected in Chinese rice wine, TRY was degraded by *L. plantarum* CAU 3823 with the degradation rate of 39% at low level, 56% at middle level and 58% at high level strain at main fermentation; 41% at low level, 60% at middle level and 62% at high level strain at post-fermentation 5d; 51% at low level, 63% at middle level and 66% at high level strain at post-fermentation 10d; and 49% at low level, 57% at middle level and 61% at high level strain at post-fermentation 20d (Figure 3B). Similar degradation efficiency to PUT, HIS, CAD, PHE, SPD and SPM was also found as follows: PUT with 13% reduction at low level, 39% at middle level and 43% at high level strain; HIS with 4% at low level, 42% at middle level and 55% at high level; PHE with 45% at low level, 74% at middle level and 82% at high level; CAD with 38% at low level, 55% at middle level and 55% at high level; SPD with 23% at low level, 46% at middle level and 89% at high level; SPM with 25% at low level, 50% at middle level and 75%, respectively, at high level at the end of post-fermentation. Overall, more than 40% contents of BAs could be degraded incubation with *L. plantarum* CAU 3823 at the middle and high levels than the one at low level during fermentation.

**Figure 3.** The contents of the total biogenic amines (BAs, **A**), tyramine (TYR, **B**), putrescine (PUT, **C**), Histamine (HIS, **D**) phenylethylamine (PHE, **E**), cadaverine (CAD, **F**), spermidine (SPD, **G**) and spermine (SPM, **H**) in Chinese rice wine adding different level of *L. plantarum* CAU 3823 at the post-fermentation and main fermentation stage during different fermentation stages (addition of starter (AS); main fermentation (MF); post-fermentation 5d (PF5d); post-fermentation 10d (PF10d); post-fermentation 20d (PF20d)).

#### *3.4. Total Acid and pH in Pilot Scale Fermentation*

As shown in Table 2, the changes in total acid and pH value of Chinese rice wine when different levels of *L. plantarum* CAU 3823 were added during fermentation were investigated, to evaluate the effect of this strain on the quality of CRW. At the initial stage of starter addition, there was no difference (*P* > 0.05) in lactic acid content and pH value among the four CRW samples. The total acid content of the CRW showed a slightly increase from 6.53 at low level to 6.86 g/L at middle level strain incubated with *L. plantarum* CAU 3823 at the end of post-fermentation, compared to the control group of 5.94 g/L. However, the total acid of CRW of 9.14 g/L incubated with high level of *L. plantarum* CAU 3823 indicated the over-acidification.


**Table 2.** Changes in the total acid and pH in the Chinese rice wine adding different level of *L. plantarum* CAU 3823 at the post-fermentation and main fermentation stage during different fermentation stages (addition of starter; main fermentation; post-fermentation 5d; post-fermentation 10d; post-fermentation 20d).

Presented data (mean ± standard deviation) are the mean values of three independent samples and each analyzed in triplicate. Values in a column with different superscripts differ significantly (*p* < 0.05).

#### *3.5. Alcohol Content, Total Sugar, Non-Sugar Solid and Amino Acid Nitrogen in Pilot Scale Fermentation*

The effects of *L. plantarum* CAU 3823 on the alcohol content, total sugar, non-sugar solid and amino acid nitrogen in the Chinese rice wine were analyzed after production process. As presented in Table 3, there was no notable change in alcohol, amino acid nitrogen and total sugar contents among the CRWs incubated with low and middle level of *L. plantarum* CAU 3823. The non-sugar solid was markedly higher (*p* < 0.05) when CRW was fermented involving with the selected strain.

**Table 3.** The alcohol content, amino acid nitrogen, total sugar and non-sugar solid in the Chinese rice wine after production process.


Presented data (mean ± standard deviation) are the mean values of three independent samples and each analyzed in triplicate. Values in a column with different superscripts differ significantly (*p* < 0.05).

#### *3.6. Sensory Evaluation*

The sensory characteristics of CRW adding with different levels of the isolated strain were described by the 30 sensory panelists. As presented in Figure 4, CRW with high level of strain exhibited the lowest score (appearance 6, aroma 7, taste 6, mouthfeel 6 and harmony 6.2) among the four CRW samples. No significant difference was observed between the CRW incubated with middle level and low level strain compared to the control CRW (*p* > 0.05), indicating *L. plantarum* CAU 3823 with low and middle level would not have an influence on the sensory behaviors of the Chinese rice wine.

**Figure 4.** Average radar map of the Chinese rice wine including different level of biogenic amine-reduced *Lactobacillus plantarum* based on sensory scores.

#### *3.7. Purification and Identification of the Amine Oxidases*

To gain a deeper insight into the amine-degrading activity exhibited by *L. plantarum* CAU 3823, LC-MS/MS experiments were designed to show whether the amine oxidases existed in the strain. Cell-free extracts were obtained at a protein concentration of 5.5 mg/mL (Table 4). The MAO activity was 36.9 U/mg and the DAO activity was 128 U/L at 37 ◦C, pH = 7 in the cell-free extracts **(**Table 4**)**. Three fractions were collected from a Sephadex G-100 column (Supplement Figure S1), and little DAO activity was detected in all three fractions, but only fraction 1 showed a good MAO activity of 19.8 U/mg compared to fraction 2 of 2.4 U/mg, and no amine oxidase activity was determined in fraction 3, which might be due to the low protein concentration.

To further investigate the amine degradation ability, the BA degradation rate (%) was calculated by incubating the three fractions with the eight BAs at pH 4.0, 33 ◦C for 2 h (Table 5). The BAs contents in fraction 1 significantly declined with the degradation rate of 41.9% for TYR, 41.1% for HIS, 40.3% for PUT, 44.3% for PHE, 41.1% for CAD, 41% for SPD, 43.5% for SPM and 47.9% for TRY. However, there were slight or little changes observed in the BA contents in the Fractions 2 and 3.

The fraction 1 was further identified by using LC-MSMS. Ten proteins including 9 amine oxidase proteins were identified in fraction 1, and hereinto, 8 amine oxidase proteins were monoamine oxidases, including 4 amine oxidase [flavin-containing] A (accession: P58027, P21396, Q5NU32 and A0A011QTL0), 2 amine oxidase [flavin-containing] B (accession: Q5RE98 and A0QU10), 1 monoamine oxidase [flavin-containing] (accession: A0A375EQX7) and 1 monoamine oxidase (accession: U2EF11) (Supplement Table S1). The MWs of the amine oxidases were closer and range from 46 to 60 kDa.


**Table 4.** The protein concentration, monoamine oxidase activity and diamine oxidase activity of the cell-free extracts (37 ◦C, pH = 7).

ND = Not determined.

**Table 5.** Degradation percentages (%) of the eight biogenic amines in the three fractions by Sephadex separation incubation with 50 mg/L of the eight biogenic amines at pH 4.0, 33 ◦C for 2 h.


ND = Not determined.

#### *3.8. Amine Oxidases Assays*

As shown in Figure 5, the purified amine oxidases mixture (fraction 1) retained its activity in a wide temperature range from 15 to 80 ◦C and was shown to maintain the 50% MAO activity after on heat treatment at 80 ◦C for 2 h. The optimal temperature for the amine oxidase activity was 28 ◦C and the MAO activity was 36.9 U/mg (Figure 5A). The MAO activities increased from 22.3 U/mg to 35.9 U/mg accompanied by the pH value from 3.0 to 5.0 while the amine oxidases were incubated at 30 ◦C for 2 h (Figure 5B). All the ions could inhibit the MAO activity, as 73%, 31%, 58%, 64% and 79% activity retained when adding 0.2 mol/L Zn2<sup>+</sup>, Cu2<sup>+</sup>, Fe2<sup>+</sup>, Ca2<sup>+</sup> and Mg2<sup>+</sup>, respectively (Figure 5C).

**Figure 5.** The monoamine oxidase (MAO) activity in the amine oxidase mixture at different temperatures (**A**) at pH 4.0 for 2 h; different pH (**B**) at 30 ◦C for 2 h and different metal ions (**C**) at 30 ◦C for 2 h (pH 4.0).

#### **4. Discussion**

Biogenic amines are considered as potential health risks since high amounts of them can lead to a series of health problems. The intake of foods with high level of BAs could induce the release of adrenaline and noradrenaline, provoking gastric acid secretion, increased cardiac output, migraine, tachycardia, increased blood sugar levels, and higher blood pressure [23]. Several researches supported the view that the BAs were formed in winemaking mainly by lactic acid bacteria carrying specific metabolic pathways that convert precursor amino acids into BAs [24]. In contrast, there is a lack of studies concerning BAs degradation by food sourced micro-organisms in wine, especially in Chinese rice wine.

In this paper, a *L. plantarum* CAU 3823, isolated from Chinese rice wine, can degrade more than 40% of the BAs, especially the five major BAs of TYR, PUT, HIS, PHE and CAD in Chinese rice wine. A similar research in grade wine showed that only one strain, *L. casei* IFI-CA 52, showed a strong ability to degrade the BAs (54% HIS, 55% TRY and 65% PUT) isolated from wine/ grape cell cultures of 85 strains [18]. However, the histamine-degrading ability of *L. casei* IFI-CA 52 was only 17% when addition of 12% ethanol, suggesting that the ability of *L. casei* IFI-CA 52 to reduce amine concentrations in wines would be rare. Regrettably, the ability of this strain to degrade other BAs was not analyzed. Moreover, a pilot scale fermentation, rather than addition of ethanol, would be a better choice to simulate accurately the complicated wine matrix.

In our experiment, pilot scale fermentation tests had proved that *L. plantarum* CAU 3823 was competent to be used as an extra starter in CRW industrial producing. Chinese koji was added at the beginning of brewing, which could bring in lots of bacteria, thus the BAs accumulated significantly at the beginning [13]. The BAs concentration showed a notably increase in the common CRW (the control group) from the starter addition stage to 10-days post-fermentation, indicating the proliferation of bacteria [13]. The concentration of BAs decreased at the end of post-fermentation, which might be due to the bacteria growth inhibition as the total acid increased during fermentation. According to our results, *L. plantarum* CAU 3823 could degrade the BAs in the CRW brewing process, and the formation of biogenic amines was further degraded by increasing the dose of strain. In this study, HIS, TYR, PUT and CAD were degraded significantly during the pilot scale fermentation, especially TYR, which indicated *L. plantarum* CAU 3823 could provide a more safety traditional fermented beverage for consumers.

Identification of functional microorganisms in CRW to reduce the formation of BAs has received more interest. Liu, Yu [13] utilized an in vivo screening process based on the next-generation sequencing technology to find BA-decreasing microorganism in CRW, and three *Lactobacillus* strains were detected that would not form biogenic amines, but only *L. plantarum* JN01 could grow under 15% ethanol, and the wine could form an unpleasant rancidity taste and more than 8 g/L total acid when the *L. plantarum* JN01 was more than 0.01 gDCW/t. Indeed, high level of functional bacteria could bring about unsatisfactory changes in CRW. A similar trend found in the current study showed that the total acid increased, and alcohol content decreased when 10<sup>7</sup> cfu/mL (high level) of *L. plantarum* CAU 3823 was added into the fermentation mash. Although the sensory scores were also decreased, the whole CRW was within the acceptable range for consumers at high level of the strain. Therefore, *L. plantarum* CAU 3823 could be the best choice to date to decrease BAs in CRW.

As a traditional alcoholic beverage, total sugar, alcoholic degree, pH value, total acid, amino acid nitrogen and non-sugar solid would play important roles in the flavor, taste and nutrition of Chinese rice wine [22]. Although high level (10<sup>7</sup> cfu/mL) of *L. plantarum* CAU 3823 could degrade the BAs maximally, undesirable influence on the acceptability was also noteworthy. Low level (105 cfu/mL) and middle level (10<sup>6</sup> cfu/mL) of *L. plantarum* CAU 3823 could eliminate the negative effect on the qualities of the wine, and what's more important, similar sensory characteristics were obtained in CRW. Thus, to degrade the content of BAs in CRW to the highest extent, middle level (106 cfu/mL) of the *L. plantarum* could be chosen in the CRW fermentation process.

Non-sugar solids, a major nutrition indicator to evaluate the quality grade of CRW, are mainly composed of dextrin, glycerin, non-volatile acid, protein and hydrolysates [25]. Interestingly, the content of non-sugar solids was increased remarkably when *L. plantarum* CAU 3823 was used, especially at middle level (106 cfu/mL), which provided a novel insight that the *L. plantarum* CAU 3823 could produce more non-sugar solids in CRW and thus have potential nutritional values.

BA can be converted into products via oxidation by microorganisms which can be used as a carbon and/or energy source or as a nitrogen source [26]. Limited studies attributed these transformations to amine oxidase activity derived from microorganisms. Yagodina [27] reported that flavoprotein oxidases existing in some microorganisms could catalyze the oxidation of BAs. Sekiguchi [28] found a histamine oxidase in the actinobacteria *Arthrobacter crystallopoietes* KAIT-B-007 isolated from soil. In this study, the amine oxidases from *L. plantarum* CAU 3823 were purified and characterized. Nine amine oxidase proteins, a mixture from *L. plantarum* CAU 3823, contributed the most of amine-degrading ability of *L. plantarum* CAU 3823. Eight MAOs were identified and thus confirmed a good monoamine oxidase activity shown in fraction 1. Amine oxidases can be divided into two subfamilies based on the cofactor they contain. MAO (EC 1.4.3.4) are a family of enzymes containing flavin that catalyze the oxidation of monoamines, employing oxygen to clip off their amine group [29]. The amine oxidases containing copper as cofactor (EC 1.4.3.6) are homodimers, which contain three subclass, namely, diamine oxidase, primary-amine oxidase and diamine oxidase [30]. Amine oxidase [flavin-containing] A and B can catalyze the oxidative deamination of biogenic amines [31]. Amine oxidase [flavin-containing] B that in humans was encoded by the MAOB gene could preferentially degrade PHE [32], which confirmed 44.3% PHE degradation in fraction 1. An "aromatic cage" has been found to play a steric role in substrate binding and in flavin accessibility and helps to increase the substrate amine nucleophilicity [33], which might enhance BA degradation. It is noted that no diamine oxidase was identified although cell-free extracts showed diamine oxidase activity.

To provide a seemingly feasible solution to degrade the BAs in foodstuffs, the biochemical character assays of the amine oxidases mixture from *L. plantarum* CAU 3823 were designed. The enzymes were very thermostable, as the activity remained stable at 80 ◦C, and were fully stable over the pH range of 3–5. Similar results were reported that a putrescine oxidase from *Rhodococcus erythropolis* NCIMB 11540 could be stable at 50 ◦C for 2 h [34] and a thermostable histamine oxidase was found in *Arthrobacter crystallopoietes* KAIT-B-007 [29]. These results indicated that the amine oxidases could be stable to use in fermented food processing.

#### **5. Conclusions**

In this paper, *Lactobacillus plantarum* CAU 3823 was a *L. plantarum* originating from Chinese rice wine which could effectively degrade the BAs. Middle level (106 cfu/mL) of *L. plantarum* could be an optimal choice to decrease the BAs maximally while keeping the CRW character and quality in the pilot scale fermentation. Nine amine oxidase proteins were identified from *L. plantarum* using Sephadex separation followed by LC-MS/MS analysis. The enzymes were very thermostable and fully stable at pH 3–5. All the ions can inhibit the amine oxidase to an extent. *L. plantarum* seemed to be an interesting species displaying BAS degradation, both in culture media conditions and in CRW fermentation, suggesting its suitability as a commercial malolactic starter. This paper provided an efficient method to decrease the biogenic amine contents in the traditional fermented food made by multiple microbes like wine, rice wine, sausages, vinegar, cheese, kimchi and so on.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2304-8158/8/8/312/s1. Figure S1: The three fractions separated by a Sephadex G-100 column, Table S1: A group of proteins by Sephadex separation followed by LC-MS/MS analysis in fraction 1.

**Author Contributions:** T.N. and Y.M. conceived and designed the experiments; T.N. and Y.G. performed the experiments; T.N. and Y.G. analyzed the data; T.N. and X.L. wrote the paper.

**Funding:** This work was supported by the National Key Research and Development Program of China (No. 2018YFC1604303-04).

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
