Analytical strategies for the determination of biogenic amines in dairy products

Biogenic amines (BA) are mainly produced by the decarboxylation of amino acids by enzymes from microorganisms that emerge during food fermentation or due to incorrectly applied preservation processes. The presence of these compounds in food can lead to a series of negative effects on human health. To prevent the ingestion of high amounts of BA, their concentration in certain foods needs to be controlled. Although maximum legal levels have not yet been established for dairy products, potential adverse effects have given rise to a substantial number of analytical and microbiological studies: they report concentrations ranging from a few mg/kg to several g/kg. This article provides an overview of the analytical methods for the determination of biogenic amines in Abbreviations


INTRODUCTION
Biogenic amines (BA) are low-molecular weight nitrogenous compounds that emerge through enzymatic decarboxylation of amino acids (Linares et al., 2011;Zhang et al., 2019) or by amination and transamination of aldehydes and ketones (McCabe et al., 2003;Pluta-Kubica et al., 2020). Their chemical structures can be classified as: aliphatic, aromatic, or heterocyclic (Linares et al., 2011;McCabe et al., 2003;Papageorgiou et al., 2018;Spano et al., 2010). Furthermore, according to number of amino groups, they can be classified as monoamines, diamines, or polyamines Spano et al., 2010). An extensive variety of BA is usually present in plants, animals (Farooqui, 2013), and foods (Naila et al., 2010), but the most important BA in certain foods and beverages are histamine, tyramine, putrescine, cadaverine, tryptamine, spermidine, spermine, and phenylethylamine, which are produced by decarboxylation of their respective amino acids (Brito et al., 2014;del Rio et al., 2017;Palermo et al., 2013;Restuccia et al., 2011;Spano et al., 2010), as can be seen in Table 1. BA act as precursors for the synthesis of hormones, alkaloids, nucleic acids, and proteins; some of them, such as histamine and tyramine, play an important role as neurotransmitters (del Rio et al., 2018). Despite their benefits, multiple negative effects on health derived from food intake with high amounts (ranged from 190 mg/kg to 500 mg/kg for histamine and from 301.8 mg/kg to 500 mg/kg for tyramine) of different amines have been described (del Rio et al., , 2018Fernández et al., 2006;Ladero et al., 2010;Lehane & Olley, 2000;Linares et al., 2016;McCabe et al., 2003;Ruiz-Capillas & Herrero, 2019;Spano et al., 2010). Synergistic effects among different BA present in food can increase these adverse effects, as is the case of tyramine, which has synergistic cytotoxic effects in combination with histamine (Palomino-Vasco et al., 2019). It should also be noted that BA are precursors to nitrosamines, which have been linked to carcinogenic and mutagenic activity (McCabe et al., 2003;Papageorgiou et al., 2018).
The toxicities of the various biogenic amines, alone or in combination, have for the most part not been established (Paulsen et al., 2017). Most regulations thus focus exclusively on the most dangerous ones, especially on histamine. While legal limits have been established to regulate histamine in certain foods such as fresh fish or enzyme-maturated fish product, allowing up 100 mg/kg and 200 mg/kg, respectively (Commission Regulation No. 2073/2005 UE, 2005), no regulation has been established for dairy products (Moniente et al., 2021). Only recommended maximum levels of histamine have been suggested for dairy products (Rauscher-Gabernig et al., 2009), despite the fact that high amounts of BA have been found in some of them. For example, 1000 to 2500 mg/kg of histamine have been detected in some varieties of cheese, as well as 1500 to 4000 mg/kg of tyramine in cheddar and Camembert cheese (Maintz & Novak, 2007).
Growing concern about the presence of BA in different foods is reflected in a growing quantity of published papers and reviews (Huang et al., 2011;Mayr & Schieberle, 2012;Papageorgiou et al., 2018;Paulo Vieira et al., 2020;Ramos et al., 2020). Given the variety of matrices in which BA can be found, recent reviews have covered the analytical methods developed for a variety of different foods and beverages (Biji et al., 2016;Guo et al., 2015;Önal, 2007;J. L. Ordóñez et al., 2016;Papageorgiou et al., 2018;Suzzi & Gardini, 2003;Zhang et al., 2019). Consumed in all cultures on a daily basis, dairy products encompass a wide range of products: milk, yogurt, cheese, kefir, butter, and cream, among others. This food group features considerable physicochemical variability with different characteristics of texture (liquid, solid, and gel), viscosity, or composition. Due to the global importance and variability of dairy products, we consider it necessary to review and compare the different strategies used to analyze BA in this group of foods, spotlighting the latest advances in the field.

SAMPLE PRETREATMENT METHODS FOR DAIRY PRODUCTS
The decision of using sample pretreatments is based on the selectivity and sensitivity of the determination step and the physicochemical characteristics of the sample. So, two of the main objectives of sample pretreatment methods are cleaning the sample and preconcentrating the analytes. Sometimes, samples can be analyzed without any treatment and the easiest is applying the "dilute and shoot" method; however, in dairy products, perhaps due to their complexity and to the concentration of BA, it has not been used for the determination these analytes. The high protein content in dairy products implies that most sample pretreatments focus on protein removal; solvent pretreatments have thus been extensively applied for the precipitation of proteins (Gianotti et al., 2008;Gloria et al., 2011;He et al., 2016;Korös et al., 2008;Latorre-Moratalla et al., 2009;Madejska et al., 2018;Molaei et al., 2019;Redruello et al., 2013;Ubaldo et al., 2015).
The application of solid-liquid (S-L) or liquid-liquid (L-L) extraction methods depends on the dairy product's texture, seeking not only to precipitate proteins but also a maximum solubility of BA in the extracting phases. A wide variety of such methods (Andiç et al., 2010;Gaya et al., 2005;Innocente et al., 2007;Lange & Wittmann, 2002;Mayer et al., 2010;Rabie et al., 2011;Shalaby et al., 2016;Yigit & Ersoy, 2003) have been used to extract BA from cheeses. Custódio et al. (2007) carried out a remarkable comparison, evaluating hydrochloric, trichloroacetic, perchloric, sulfosalicylic, and acetic acids, borate buffer, methanol, and ethanol as extracting agents of amines in grated Parmesan cheese. The level of amines present in the sample, the type of sample, the concentration and temperature of the extractor, as well as the method used, affected the degree of extraction efficiency. Each one of the studied BAs could be better recovered in a different way, although 1 M HCl was the most adequate solution for most of the amines.
Although solvent extractions are the most common methods for the analysis of BA in different dairy matrices such as milk, cheese, or yogurt (Mohammadi et al., 2017;Molaei et al., 2019;Wu et al., 2015) due to their high performance in removing undesirable compounds (Liu et al., 2018), they have several disadvantages, such as slowness, high amounts of harmful organic solvents, loss of objective analytes (Wu et al., 2015), low repeatability, complexity, laborious emulsion formation (Liu et al., 2018;Saaid et al., 2009), a large sample volume for trace analysis (Huang et al., 2011), and time-consuming procedures (J. L. Ordóñez et al., 2016). To overcome these disadvantages, modifications and improvements have been made based on liquidliquid extraction (LLE) techniques by using liquid phase microextraction methods such as dispersive liquid-liquid microextraction (DLLME) and salting-out assisted liquidliquid extraction (SALLE). Ultrasonic-assisted extraction (UAE) and microwave-assisted extraction (MAE) have also been used to avoid these drawbacks.
Some of the advantages of DLLME analysis are high enrichment capacity, simple technique, low consumption of organic solvents, time saving due to short extraction time, low cost, good repeatability, and high sensitivity (Kamankesh et al., 2013;Rezaee et al., 2006;Wu et al., 2015). Mohammadi et al. (2017) used this method in combination with the auxiliary application of MAE for the simultaneous determination of four BA in samples of Iranian Lighvan cheese, and demonstrated that MAE-DLLME-GC coupled with mass spectrometry (MS) is effective in extracting and determining small amounts of BA in cheese with 20% fat. The analysis procedure was as follows: 1 g of sample was homogenized with 5 ml of 0.1 M HCl and microwaved at 500 MHz for 3 min; after the precipitation of the proteins, BA derivatization and microextraction were carried out simultaneously by adding 25 μl of isobutyl chloroformate (IBCF), 2 g of NaCl, 600 μl of acetonitrile (ACN) (dispersive solvent), and 60 μl of 1-octanol (extractive solvent). Solution was centrifuged at 4000 rpm for 5 min, and 2 μl of the floated phase of 1octanol was injected directly into the GC-MS. DLLME has also been used a step prior to LC analysis. Wu et al. (2015) studied BA in cheese with UAE-DLLME, but added chloroform and acetone as extraction and dispersing solvents, respectively.
SALLE is used to prepare samples for organic extraction in chromatographic analysis (Tang & Weng, 2013). To increase the distribution ratio of a particular solute, the effect of salt is used by adding an electrolyte to an aqueous solution (Rice et al., 1993). This effect promotes the extraction of molecular species to an organic phase and allows for phase separation between the two solvents (Shishov et al., 2019). Ramos et al. (2020) recently used SALLE for the determination of BA in seven different types of cheese, including derivatization with dansyl chloride (DNS-Cl) and HPLC coupled to a fluorescence detector (FLD). An extraction of soluble BA was carried out with 1 M HCl prior to the use of SALLE, simultaneously with the derivatization step, and the extract was injected into a liquid chromatography system.
UAE is used to improve extraction efficiency and reduce analysis times, thanks to the effects it produces such as vibration, cavitation, and agitation generated by ultrasonic waves that induce the effective components to enter the solvent (Khaled, 2014;Redruello et al., 2013;Švarc-Gajic & Stojanovic, 2011).
Several authors have used UAE to extract BA in cheese (Fernández et al., 2007;Herrero-Fresno et al., 2012;Krause et al., 1995;Redruello et al., 2013;Švarc-Gajic & Stojanovic, 2011;Zotou & Notou, 2012). Švarc-Gajic and Stojanovic (2011) compared the efficiency of different extraction procedures: they tested three different solvents and carried out a comparison among classical reflux extraction, UAE, and MAE. They concluded that methanol was the best solvent and that UAE was the most efficient procedure.
Further alternative pretreatment techniques have been developed, such as SPE (Calbiani et al., 2005;Gianotti et al., 2008;Gosetti et al., 2007;Restuccia et al., 2011;Spizzirri et al., 2019;Yang et al., 2016) and solid phase microextraction (SPME) (Ali Awan et al., 2008). Advantages provided by these methods include low consumption of extraction solvents, ease of use, and low cost, as well as the possibility of automation and of online coupling with analytical instruments (Pena-Pereira et al., 2012). Gosetti et al. (2007) compared the recoveries obtained with LLE and SPE. In the SPE method, the BA were extracted from cheese with 0.1 M HCl and the solution was loaded onto the SPE cartridge (C18 sorbent and Strata X were compared and C18 was selected), which was washed with water and eluted with methanol. Better results were obtained with SPE than with LLE, with recovery yields greater than 90% for all amines through SPE. This procedure of extraction with 0.1 M HCl, loading onto C18 cartridges and cleaning with water, is the most common one when SPE is used to determine BA in cheeses (Gianotti et al., 2008;Restuccia et al., 2011); however, other kinds of sorbent material, such as cyanopropyl (CN) cartridges, and other washing steps have also been used (Calbiani et al., 2005).
One of the aforementioned advantages of SPE is the possibility of automatization. Yang et al. (2016) successfully developed a novel method based on online SPE coupled to capillary HPLC for the simultaneous separation and determination of 15 BA in cheese. Using the online SPE purification technique, it was possible to reduce the matrix effect of the samples while simplifying sample pretreatment, reducing manual error, and greatly improving analysis efficiency.
SPE has been also applied to determine BA in milk. Spizzirri et al. (2019) studied BA in reconstituted milk powder and ready-to-use liquid milk using SPE prior to derivatization and LC with ultraviolet (UV) analysis. The SPE cartridges (ENVI-18SpeTubes) were first preconditioned with a water/ammonia solution (70/30% v/v) and then with methanol. The sample was loaded, after which the washing step was performed with water and eluted with methanol.
Ali Awan et al. (2008) demonstrated the advantages of SPME as a technique to extract and derivatize BA for GC analysis. They used trifluoroacetylacetone (TFAA) as derivatizing agent; derivatization and extraction were carried out simultaneously into a SPME vial at 120 • C during 20 min. This method was applied for the determination of putrescine and cadaverine in a cheese sample.

ANALYTICAL METHODS FOR DAIRY PRODUCTS
One of the greatest challenges in food analysis remains BA determination, which is of considerable importance for two main reasons: on the one hand, the potential toxicity of BAs and, on the other hand, the possibility of using them as food quality markers. F I G U R E 1 The most common analytical techniques used for the determination of biogenic amines. Abbreviations: AQC, 6-aminoquinolyl-Nhydroxysuccinimidylcarbamate; Bnz-Cl, benzoyl chloride; CE, capillary electrophoresis; Dbs-Cl, dabsyl chloride; DEEMM, ethoxymethyl malonate diethyl ester; Dns-Cl, dansyle chloride; ELISA, enzyme-linked immunosorbent assay; FMOC-Cl, 9-fluorenylmethyloxycarbonyl chloroformate; GC, gas chromatography; LC, liquid chromatography; NBD-CL, 4-Chloro-7-nitrobenzofurazan; NDA, naphtalene-2,3-dicarboxyaldehyde; OPA, o-phthalaldehyde; TLC, thin layer chromatography One of the difficulties in BA determination is the polarity of these compounds, which makes them have a higher solubility in water than in the organic solvents used for analysis. Other difficulties are: the complexity of the sample; the presence of interfering compounds that cause the spread of chromatographic peaks and the appearance of shoulders; a limited concentration range within which analytes cannot be detected by the analytical detectors; and the absence of intrinsic properties of the BA for most of the usual detectors (Önal et al., 2013;Papageorgiou et al., 2018). To solve these analytical difficulties, analytical methods are generally combined with previous processes of extraction and derivatization prior to separation and detection.
A wide range of methods (Tables 2 and 3) have been developed for the analysis of BA in dairy products with different analytical techniques ( Figure 1): TLC, GC, CE, ELISA, chronopotentiometry, and biosensors, whereby LC is the most widespread analytical technique.

Analytical methods based on LC with derivatization
Many BA, for example putrescine, cadaverine, spermine, and spermidine, do not absorb in the UV region, while others, such as phenylethylamine, tyramine, and histamine, absorb in a shorter wavelength region (close to 200 nm) where matrix effects can appear (Jain & Verma, 2018;Zhang et al., 2019). Most of them require derivatization due to low volatility and lack of chromophores, and also in order to reduce their polarity with the purpose of improving chromatographic and detectability properties (Önal, 2007).
The derivatization process allows to improve the resolution of the analytes in reverse phase (RP) columns as well. Therefore, most analytical LC methods for the quantification of BA feature a pre-or postcolumn derivatization step (García-Villar et al., 2009). A broad array of derivatizing reagents have been used for the determination of BA in dairy products. The most common derivatization reactions are shown in Figure 2. Reproducibility, stability, and speed are the main reasons for choosing among the different options (Munir & Badri, 2020). The most common derivatizing reactions are the following:

Derivatization by DNS-Cl reagent
Sulfonyl chlorides work to derivatize primary and secondary amines, and DNS-Cl is the one most widely used. This derivatizing reagent offers a high fluorescence efficiency, especially in organic solvents (Chen, 1967). Fluorescence detection is therefore recommended, although other types of detectors can also be used.

F I G U R E 2
The most common derivatization reactions used for the determination of biogenic amines in dairy products with LC Eerola et al. (1993) for the determination of BA in dry sausages has been modified and applied by several authors (Andiç et al., 2010(Andiç et al., , 2011Contreras et al., 2007;Durlu-Özkaya, 2002;Ercan et al., 2019;Lanciotti et al., 2007;Marijan et al., 2014;Min et al., 2004;Tittarelli et al., 2019), in order to quantify BA in different cheeses. This method consists in a first step of acid extraction with 0.4 M HClO 4 , followed by derivatization of the acid extract with DNS-Cl. Separation of the analytes is carried out by RP-HPLC with a C18 column, and the peaks are detected at 254 nm. The obtained limit of detection (LOD) varies between 1 mg/kg and 5 mg/kg for the different amines.
Another study which has been a reference for many authors (Bonczar et al., 2018;Esatbeyoglu et al., 2016;Forzale et al., 2011;Galgano et al., 2001;Gennaro et al., 2003;Martuscelli et al., 2005;Pluta-Kubica et al., 2020;Schirone et al., 2013;Zazzu et al., 2019) was published by Innocente et al. (2007). Improving upon a previous study by the same author (Innocente & D'Agostin, 2002), this method features a step of extraction of BA from cheese with 0.1 M HCl, after the derivatization of this extract with DNS-Cl. The addition of an L-proline solution is necessary to remove the excess of derivatizing reagent, after which the sample is extracted with diethyl ether, dried, and re-dissolved with ACN for injection. Forzale et al. (2011) used this method for the quantification of BA in several pecorino cheeses, obtaining better sensitivity than with methods based on the work of Eerola et al. (1993); they obtained a LOD of 0.3 mg/kg and a limit of quantitation (LOQ) of 1 mg/kg. A slight modification was proposed by Zazzu et al. (2019): detection was carried out at 218 nm instead of 254 nm, obtaining some improvement in terms of sensitivity (LOD 0.07 to 0.23 mg/kg). Yang et al. (2016) developed an analytical method for the quantification of BA in cheese. It included a clean-up step using online SPE prior to separation. In spite of the fact that SPE is a cleaning and preconcentration tool, detection limits in that study ranged between 0.05 mg/kg and 0.25 mg/kg, which are not the best ones described in the literature for this derivatizing agent; still, the matrix effect of samples was reduced thanks to SPE. Some of the advantages of automatized analysis were the simplification of the sample pretreatment, as well as reduction of manual error, thereby resulting in a more efficient method.
BA have been also analyzed in commercial milk for infants and young children after derivatization with DNS-Cl and detection by UV-Vis (Spizzirri et al., 2019). Spizzirri used SPE (Supelclean ENVI-18Spe Tubes) for concentration and purification, but only for ready-to-use liquid milks, obtaining LOD one order of magnitude lower, in this case between 0.03 mg/L and 0.05 mg/L, than for powder milk samples.

Fluorescence detection
One example illustrating the better fluorometric properties of dansyl derivatives is the study by Standarová et al. (2010) carried out in 2010. They quantified seven different BA as dansyl derivatives in cheese, using FLD at 330 nm and 500 nm as excitation (ex.) and emission (em.) wavelengths, respectively, except for histamine, which was detected by UV. This method reached the lowest LOD (0.03 mg/kg for putrescine and cadaverine) published for these derivatives, to our knowledge. Excellent sensitivity was likewise reported in the case of Ramos et al. (2020) who used SALLE as sample preparation step. In spite of good results with FLD, both studies showed that an UV detector is more appropriate for the determination of histamine. Several studies also reported LOD lower than 1 mg/kg for different BA in cheese using this system of detection (Brito et al., 2014;Sawilska-Rautenstrauch et al., 2010), thus achieving better sensitivity than most studies conducted with UV-Vis detection.

MS detection
Derivatization reactions are not only conducted with the purpose improving the detectability of analytes in terms of sensitivity and selectivity. Mazzucco et al. (2010) used a HPLC-MS method to validate a technique with UV detection for the determination of BA and amino acids in cheese and other foods. They found that dansyl derivatives decreased the matrix effect in the case of MS detection. As was expected, better LOD were achieved with this method compared with UV. The authors also included a SPE procedure for cleaning the dansylated sample extract.
That same year, La Torre et al. (2010) also combined dansyl derivatives with MS detection. They developed an excellent method with good linearity, free of matrix effects, and featuring high levels of sensitivity (ranging from 0.5 μg/L to 15 μg/L) for the determination of BA in donkey milk samples.

Derivatization by dabsyl chloride
This derivatizing agent is a sulfonyl chloride analogous to DNS-Cl. It also acts on primary and secondary amino groups, obtaining compounds with good stability, although they are not detected by fluorescence (Toyo'oka, 1999).

UV-Vis detection
Several authors (Bockhardt et al., 1996;Calzada et al., 2013;Fernández et al., 2007;Valsamaki et al., 2000) have determined BA by extraction with 0.1 M HCl from dairy products, precolumn derivatization with DBS-Cl and UV-Vis detection (436 nm). Krause et al. (1995) pioneered this analytical method for the quantification of amino acids and BA in biological samples and food. LOD was between 0.12 pmol and 0.52 pmol, and repeatability was lower than 3.1%. Also, Bockhardt et al. (1996) developed a method of derivatization with DBS-Cl in combination with a HPLC analysis for the determination of BA in cheese. Although they used a sample clean-up procedure by ultrafiltration, sensitivity was not improved as compared to the method of Krause et al. (1995) and the LOD was between 0.34 pmol and 0.76 pmol.

3.1.3
Derivatization by benzoyl chloride BNZ-Cl is another acyl chloride, similar to DNS-Cl or DBS-Cl, used for derivatization of amines, and obtaining shorter elution times than dansyl derivatives (Toyo'oka, 1999). Özdestan and Üren (2010) studied BA in kefir samples. They were extracted with 0.2 M HCl, and derivatization was subsequently carried out with BNZ-Cl. Benzamide derivatives were extracted with diethyl ether and redissolved with methanol, injected, and detected with UV at 254 nm. LOD lay between 0.2 mg/L and 2.5 mg/L, and linearity ranged between 133.6 mg/L and 1433 mg/L. Other authors used similar procedures or HClO 4 in the extraction step to quantify BA in yogurt (Paulo Vieira et al., 2020;Silva et al., 2019). An important improvement with this derivatizing agent was the detection wavelength. Costa et al. (2015) quantified BA in fermented milks, and carried out detection at 198 nm. The detection limit thereby decreased to values between 0.03 mg/L and 1.30 mg/L, one order of magnitude lower than detection at 254 nm.

UV-Vis detection
A remarkable analytical method was developed by Molaei et al. (2019) to quantify BA in Kashk, a sour yogurt, milk, and buttermilk used in Persian and Turkish cuisine. Histamine, cadaverine, putrescine, and tyramine were extracted and derivatized using SPE, based on magnetic mesoporous silica nanosorbent. Detection was carried out at 198 nm after chromatographic analysis. An excellent detection limit was reached (8 μg/L for Put), and reproducibility was lower than 4.14% of relative standard deviation (RSD).

MS detection
Using isotopically labeled BA as internal standards, Mayr and Schieberle (2012) developed an analytical method to quantify thirteen BA in several foods, including cheese and yogurt. Analysis of benzamide derivatives was carried out by LC-MS/MS. High sensitivity was reached (LOQ 0.05 μg/kg) and, thanks to the labeled BA, %RSD lower than 5.1% and a matrix effect-free method were obtained.

Derivatization by 9-fluorenylmethoxycarbonyl chloride
This derivatizing agent has the considerable advantage of reacting not only with primary and secondary amines, but also with tertiary amines. The reaction product can be detected by FLD and UV (Toyo'oka, 1999). Kirschbaum et al. (1994) quantified BA in cheese after their extraction with 0.1 M HCl, precolumn derivatization with FMOC-Cl, and analysis by HPLC/FLD (ex. 265 nm, em. 315 nm). Only one step was required to remove the excess of reagent, thereby avoiding the interferences of hydrolysis products. Satisfactory detection limits were obtained for determination of BA in cheese, except for tryptamine and serotonine, for which the fluorescence quantum yield was too low.

Derivatization by o-phthalaldehyde
As in the case of DNS-Cl, this reagent is preferably combined with FLD. Although it has been widely used (probably due to its versatility for pre-, on-, or postcolumn derivatization), it has the disadvantage of not reacting with secondary amines, and its products are less stable than dansyl or benzoyl derivatives (Peña-Gallego et al., 2012;Toyo'oka, 1999).

UV-Vis detection
Determination of BA in foods by o-phthalaldehyde (OPA) derivatization and UV detection is not very common in the literature: for dairy products, only studies of tyramine in cheese by Komprda et al. (2008) and Renes et al. (2014) were found. Komprda et al. (2008) extracted BA with 0.1 M HCl, and the extract was precolumn derivatized by OPA in borate buffer in the presence of 2-sulfonylethan-1-ol. Separation, after derivatization, was carried out with a C18 column, and tyramine was detected using a UV-Vis detector at 254 nm. Although the authors did not include figures of merit for this method, they reported concentrations of tyramine ranging from 5 mg/kg to 392 mg/kg; therefore, it can be supposed that LOD was lower than 5 mg/kg. Vale and Glória (1998) conducted the first determinations of BA in cheese using OPA as postcolumn derivatizing agent and FLD (340 nm for excitation and 445 nm for emission) (Gençcelep et al., 2008). They optimized the extraction of BA from cheese and the purification of the extracts, whereby 0.1 N HCl and diethyl ether were the best options, respectively. Good precision values, lower than 10% of RSD (except for tryptamine and agmantine) were obtained, and LOD was also acceptable (between 0.6 mg/kg and 1.4 mg/kg). Custódio et al. (2007) also tried out different extraction methods and analyzed the extracts according to the method described by Vale and Gloria (Gençcelep et al., 2008). They likewise concluded that HCl allowed for adequate extraction of BA from cheese; how-ever, they stated that 1 M was more appropriate, resulting in better coefficients of variation. Santos et al. (2003) conducted a study of formation of BA by Lactococcus in milk. BA were extracted with sulphosalicylic acid and analyzed by ion-pair HPLC/FLD after postcolumn derivatization with OPA. The same method was used by Gloria et al. (2011) for the determination of BA in milk, and they reported a highly satisfactory LOQ of < 0.010 mg/L.

3.1.6
Derivatization by naphthalene-2,3-dicarboxaldehyde Naphthalene-2,3-dicarboxaldehyde (NDA) has a structure analogous to OPA, although its derivatives are more stable: it can therefore be more properly used as a precolumn derivatizing agent (Toyo'oka, 1999). For analysis of BA in dairy products, FLD (ex 424 nm; em 494 nm) has mainly been used. Zotou and Notou (2012) developed an analytical method based on the derivatization of BA with NDA combined with analysis by HPLC-FLD for the quantification of BA in cheese. Several parameters of the reaction were studied: time of reaction, concentration of NDA and potassium cyanide (KCN) (used as nucleophile), and derivative stability (18 h at −4 • C). The method's LOD lay between 0.6 mg/kg and 165 mg/kg. Some years later, the same authors developed an improvement based on zonefluidics/HPLC (ZF-HPLC) for the simultaneous analyses of BA and online derivatization for the determination of BA in milk (Fernández et al., 2007). They managed to reduce the amount of NDA, the reaction time (3 instead of 30 min), and detection limits (0.6-6 μg/L in aqueous solutions).

UV-Vis detection
A comparison between HPLC/FLD and UPLC/UV after derivatization of BA in cheese with AQC was carried out by Mayer et al. (2010). Results regarding differences in sensitivity were not shown, and only detection limits with UV detection were supplied, ranging from 0.4 mg/kg to 16.2 mg/kg. In spite of the fact that the determination of AQC derivatives with UV has certain disadvantages, such as their lower signal intensity or a pronounced peak due to the byproduct of the AQC derivatization reaction, they were compensated by a much higher resolution due to the use of UPLC. It is remarkable that dopamine and tryptamine could only be detected by UV because they did not yield fluorescent-active AQC derivatives.

3.1.8
Derivatization by diethylethoxymethylenemalonate Diethylethoxymethylenemalonate (DEEMM) reacts with primary and secondary amines, and their derivatives remain stable for several weeks at room temperature. They can be detected by UV-Vis spectroscopy at 280 nm (Toyo'oka, 1999). Redruello et al. (2013) extracted BA and amino acids from cheese with 0.1 M HCl and quantified them after derivatization with DEEMM and analysis by UPLC. They validated the method, obtaining LOD lower than 0.15 mg/L and intraday repeatability much lower than 1% for BA (values estimated in standard mixtures). No significant effect of the cheese matrix was observed.

3.1.9
Derivatization by ninhydrin The reaction between this derivatizing agent and amines (NINHYDRIN only reacts with primary amines) yields purple, blue-violet-colored compounds (Toyo'oka, 1999). Ninhydrin is used for postcolumn derivatization, and only when high sensitivity is not required. Joosten and Olieman (1986) adapted the analytical method developed by LePage and Rocha (1983) for the determination of BA in cheese and other food products. BA were extracted from cheese using a trisodium citrate solution and 0.6 M trichloroacetic acid (TCA), separated by HPLC and detected after postcolumn derivatization at 546 nm. Although a significant disadvantage of this method lay in the extended amount of time required for it (1.5 h), good repeatability values (less than 10%) were nevertheless achieved. The detection limit for BA in cheeses was 2 mg/kg for each amine.

UV-Vis detection
Despite the fluorescent properties of the NBD-Cl compound, in the case of dairy products it has only been used for UV detection. Yigit and Ersoy (2003) determined tyramine in cheese after extraction with 5% HClO 4 and derivatization with NBD-Cl. The reaction product was measured by UV at 458 nm after chromatographic separation on a C18 column. The LOD was 25 mg/kg, and the RSD was lower than 3%.

Analytical methods based on LC without derivatization
Although BA are frequently determined after derivatization reactions in order to be analyzed by LC, some studies have quantified them directly. Some of the advantages of analyzing them without derivatization are shorter analysis times, as well as avoiding the cost of derivatizing agents (some of which are expensive), along with the waste products thereof and further solvents used during the derivatization process. Several detection modes have been used, but the development of MS has probably played the greatest role in the growth of the number of studies without derivatization.
One of the earliest detection modes studied was by Arlorio et al. (1998). They determined histamine, tyramine, 2-phenylethylamine, and tryptamine in cheese with ionpair HPLC. The analytes were detected at 215 nm, and they compared sensitivity using an UV and a diode array detector (DAD), obtaining better results with DAD. One of their remarkable achievements was the addition of octylamine to the eluent in order to reduce peak asymmetry. Further authors (Andiç et al., 2011;Min et al., 2004;A. I. Ordóñez et al., 1997) have also used DAD as a detection method to analyze BA by HPLC.
One of the main reasons for using derivatization is the absence of chromophores, fluorophores or the very weak UV absorbance of some BA; however, the specificity of some detectors, such as a chemiluminescent nitrogen detector (CLND), could serve as a good solution, as was shown by J. Sun et al. (2011). They compared three different detection strategies: CLND, postcolumn derivatization with ninhydrin, and a charged aerosol detector (CAD) after the separation of 14 BA by ion-pair HPLC and applied them in a variety of samples, including dairy beverages. Narrower peaks, better baselines, and excellent linearity were achieved with the CLND. In terms of sensitivity, detection limits ranged from 0.1 μg/ml to 0.4 μg/ml (estimated in standard solutions), which is lower than with ninhydrin, but not as good as other derivatizing agents.
The evaporative light scattering detector (ELSD) is also another possibility for the quantification of BA without derivatization, although it is not the most adequate one if a high sensitivity is required, as was shown by Restuccia et al. (2011). They compared a method based on LC-ELSD with a method based on derivatization by DNS-Cl for the determination of BA in cheese. Better repeatability was achieved with the derivatization method. As in the case of the CLND, good detection limits were estimated (lower than 3 mg/kg); however, they were not as good as those estimated with DNS-Cl.
In order to obtain very low detection limits without using derivatizing agents, MS needs to be used as a detector. Calbiani et al. (2005) obtained a LOD of 0.05 mg/kg (estimated in cheese samples) for histamine, with a method based on matrix solid phase dispersion (MSPD) followed by LC-MS. This method's figures of merit were highly satisfactory: excellent intraday repeatability (RSD lower than 5%) and linearity over two orders of magnitude.
The comparison carried out by Gosetti et al. (2007) between the analysis of native BA by LC-MS/MS and by LC-UV after derivatization with DNS-Cl showed that, in spite of the higher sensitivity of MS (LOD ranging from 5.1 μg/L to 35 μg/L in ricotta cheese, 1.7 to 22.5 μg/L in standard solutions), some parameters such as the linearity range are better when a derivatizing agent is used. Draisci et al. (1998) integrated the pulsed amperometric detection method to IEC for the simultaneous determination of underivatized BA, obtaining recoveries ranging from 87.3% to 97.7% and a lower LOD of 1.25-2.50 ng in cheeses. Palermo et al. (2013) applied an analytical method based on IEC conductimetric detection for the determination of certain BA, and obtained recoveries of 82-103 % as well as good LOD, lying between 23 μg/kg and 65 μg/kg.
One of the difficulties of analyzing BA without derivatization is their bad chromatographic properties. Hydrophilic interaction chromatography (HILIC) offers a new possibility to avoid tedious derivatization processes. The hyphenation of HILIC and MS developed by Gianotti et al. (2008) to quantify seven BA in cheese is a perfect example of its possibilities (LOD ranging from 1.0 μg/L to 3.5 μg/L).

3.3
Analytical methods based on capillary electrophoresis CE is the second most commonly applied separation technique for the determination of BA in food. CE is sensitive, fast, uses small sample quantities, provides good precision, and does not require derivatization or sample cleaning (Karovičová & Kohajdová, 2005;Kvasnička & Voldřich, 2006). It is powered by high voltage electric fields and capillary separation. The separation is caused by the different mobility and distribution characteristics of each component in the sample (Shah et al., 2020). CE is efficient, but has a lower reproducibility of migration times than LC. Although it is a sensitive technique, its sensitivity is low compared to other methods (as results show below). Therefore, although derivatization is not compulsory, its use is generally recommended to solve the problem of determining BA in food samples. Detection can be carried out in two ways: either it is carried out directly with a UV detector or a DAD, or it is carried out after derivatization of BA with FLD (Etienne, 2006). Some of the reagents used are OPA, DNS-Cl, DBS-Cl, and β-naphthol (Karovičová & Kohajdová, 2005), with the same advantages and disadvantages previously explained for LC. Nouadje et al. (1995) developed an analytical method based on CE and detection with a ball-lens laser-induced fluorescence detector for the determination of BA in cheeses, previously derivatizing them with fluorescein isothiocyanate (FITC). Detection limits ranged from 0.5 x 10 −10 M to 1.5 × 10 −10 M.
CE without derivatization has also been used for the quantification of BA in cheeses. Fernández-García et al. (1999 studied the influence of proteinases on the formation of biogenic amines. They were analyzed by CE and detected at 214 nm. The lowest amount of histamine they found was 2.84 mg/kg. Pulsed amperometric detection (PAD) also offers a possibility for the detection of BA after separation by CE (X. Sun et al., 2003). Ten grams of fresh milk were homogenized with 6 ml of 5% HClO 4 for 30 min and centrifuged at 5000 rpm for 10 min. LOD ranged from 10 −7 M to 4 10 −7 M. Linearity was two orders of magnitude. Putrescine, cadaverine, spermidine, and spermine were searched for in six different samples, but only spermidine was found. Kvasnička and Voldřich (2006) developed a capillary zone electrophoresis (CZE) method with conductometric detection for the determination of BA in different foods. For the analysis of two different cheeses, 5 g were extracted with 100 ml of 0.1 M HCl for 30 min in ultrasonic bath. The mixture was then filtered and diluted with demineralized water before analysis. Linearity lay between 0 μmol/ml and 100 μmol/ml, and LOD ranged from 2 μmol/L to 5 μmol/L.
More recently, a faster method (less than 6 min) based on CE and conductivity detection was applied for the analysis of five different BA in yogurt, kefir, and cheese (Adımcılar et al., 2018). LOD ranged from 0.041 mg/L to 0.098 mg/L, and reproducibility was highly satisfactory (less than 4.25% RSD), as in the previous method.

3.4
Analytical methods based on thin layer chromatography TLC is used as a technique for the rapid separation and determination of small amounts of substances. In terms of advantages, TLC is a simple method, does not involve special equipment, and can analyze several samples at the same time (Hernández-Cassou & Saurina, 2011;Önal, 2007;Özdestan & Üren, 2009;Zhang et al., 2019). Nevertheless, in some cases, an excessive amount of time is required for analysis, and results are not sufficiently accurate (Bockhardt et al., 1996;Lázaro de la Torre & Conte-Júnior, 2014;Önal, 2007). The use of a variety of derivatizing agents to determine BA is a common practice.
The first TLC method for determining BA in cheese samples was the one developed by Shalaby (1999). They were extracted with TCA and then derivatized with DNS-Cl. Separation of the dansyl derivatives was achieved on silica gel TLC plates and quantified by densitometry at 254 nm. The limit of detection ranged from 5 ng to 10 ng, and the precision, expressed as %RSD, ranged from 0.15% to 4.41%. Pham and Nguyen (2016) carried out a study of BA in milk using HPLC and TLC. They demonstrated the use of the TLC technique as a qualitative method and HPLC as a quantitative method to determine histamine. Ninhydrin was sprayed in TLC technique, and a purple spot appeared when histamine was present. The quantitative result was obtained by HPLC.
Švarc-Gajic and Stojanovic (2011) determined histamine concentration in cheese using the chronopotentiometric technique as a detection method after purification of the extract by TLC. LOD was 1.31 mg/L. This is a simple, rapid cheese histamine analysis technique, cheaper than other chromatographic techniques, although its sensitivity is not as good.

Analytical methods based on gas chromatography
The volatility and polarity of BA is one of the reasons why the GC technique is not usually applied to determine them; derivatization is thus frequently used to improve those properties (Karovičová & Kohajdová, 2005; Lázaro de la Torre & Conte-Júnior, 2014;Papageorgiou et al., 2018). Frequently analyzed derivatives are trifluoroacetyl, trimethylsilyl or 2,4-dinitrophenyl (Kielwein et al., 1996). The most common detectors used to quantify them are flame ionization detectors (FID), electron capture detectors (ECD), and MS (Etienne, 2006;Kielwein et al., 1996;Silla Santos, 1996).
The number of studies that have analyzed BA in dairy products by GC is low. One featured technique is the analytical method developed by Ali Awan et al. (2008) for the determination of putrescine and cadaverine by SPME-GC-MS. This method is based on on-fiber derivatization using trifluoroacetylacetone (TFAA). Automatization is one of its main advantages, along with the fact that is matrix effect free. Although the authors generally applied the method by GC-MS, they applied it to cheese samples by GC-FID, whereby the concentration of Put and Cad was of 38 and 22 mg/kg, respectively.
Some authors have used GC-MS as a rapid and sensitive method for the analysis of BA in cheese samples (Kamankesh et al., 2021;Mohammadi et al., 2017). Mohammadi et al. (2017) also used simultaneous derivatization and microextraction to quantify BA in cheese samples. In this case, the derivatizing agent was isobutyl chloroformate (IBCF), and the microextraction technique was MAE-DLLME. Derivatization products were detected by GC-MS. The LOD for cheese samples ranged from 5.9 μg/kg to 14 μg/kg, which lies in the range of other works.

3.6
Rapid methods 3.6.1 ELISA ELISA-based methods started to be developed in the 1970s, and they have been expanded since then, mainly due to their speed and low cost (Vaz et al., 2020). The first time that a competitive direct-ELISA (CD-ELISA) was applied to the determination of histamine in cheese was in Aygün et al. (1999). They compared their results with a HPLC method and found a good agreement, although some differences were found in the low concentration range from 2 mg/kg to 10 mg/kg. The LOD of the ELISA method (2 mg/kg) enabled detection of histamine at levels well below those that have been considered safe. They also showed one of the main advantages of ELISA methods: their speed. While more than 30 samples per day could be evaluated by CD-ELISA, the maximum daily evaluation achieved by the LC method was 10 samples of cheese. Leszczyocha and Pytasz (2018) also compared an ELISA method with a spectrofluorimetric method, but in this case for the determination of histamine in yogurt and kefir. Their comparison evidenced the possibilities of the immunoenzymatic method, which is faster and, in this case, more sensitive than the reference method, although more expensive and less precise, while adequate for higher concentrations.

Biosensors
Development and use of sensors and biosensors have been growing exponentially in the last decades. Speed, low price, ease of use, and the minimum required amount of sample are some of the advantages that have contributed to their expansion in a wide number of sectors, including the food industry (Ahangari et al., 2021;Salleres et al., 2016).
Most biosensors used to quantify BA in dairy products are based on electrochemical responses (Compagnone et al., 2001;Rotariu et al., 2016) of the detection of hydrogen peroxide, generated as a byproduct in BA oxidation, catalyzed by amino oxidase enzymes (Calvo-Pérez et al., 2013;Oliver et al., 2021). Other systems also exist, such as the molecularly imprinted electrochemical sensor developed by Huang et al. (2011) for the determination of tyramine in yogurt samples. The LOD was 5.7 × 10 −8 M, and the sensor exhibited an excellent selectivity and repeatability. Compagnone et al. (2001) studied BA by amperometric detection in cheese using immobilized diamine oxidase (DAO). Two different systems were optimized to respond to six BA. The best system's sensitivity ranged from 5 × 10 −7 to 2 × 10 −6 M, with three orders of magnitude of linearity and good reproducibility (1-3% RSD). Operational lifetime was calculated to last for 300 samples.
The problem of using amine oxidases in biosensors to quantify BA is that most amine oxidases do not react to a single BA, but to several amines. Lange and Wittmann (2002) solved this problem for the determination of histamine, tyramine, and putrescine in different foods (two different cheeses were analyzed) by immobilizing three different amine oxidases and applying neural networks for pattern recognition. The LOD was 10 mg/kg for histamine and tyramine and 5 mg/kg for putrescine. This system was compared with a HPLC method, and good correlation between results was found. This comparison revealed one of the main drawbacks of some biosensors: false-positive and false-negative results, which reached 12% and 25%, respectively.
Selectivity problems were also shown by Carelli et al. (2007), who reported the total content of BA in cheese expressed as histamine equivalents, assuming that it was not possible to discriminate the contribution of each biogenic amine to the biosensor. In order to remove certain interferences, the authors developed a biosensor based on DAO, entrapped by glutaraldehyde, onto an electrosynthesized bilayer film, after having studied different electrodes and films. The sensitivity achieved with this system was high, with LOD between 6 μM and 12 μM. Calvo-Pérez et al. (2013) developed an amperometric biosensor for the determination of tyramine in cheese, and they used a redox mediator (hydroxymethylferrocene) to reduce the applied potential and thereby decrease the possibility of interferences in the electrochemical measurement. In this case, the detection is based on the signal produced during the reduction of the mediator, instead of the electrochemical detection of H 2 O 2 as in most of these systems. The enzyme used was plasma amine oxidase (PAO), and it was immobilized on a screen-printed carbon electrode. The LOD was 2 μM.
Recently, two different biosensors have been developed for the quantification of tyramine in cheese (Navarro et al., 2020;Oliver et al., 2021). One of them (Navarro et al., 2020) is based on the colorimetric determination (540 nm) of gold nanoparticles generated when Au (III) is reduced in the oxidase enzymatic reaction, using tyramine oxidase (TAO) as enzyme. A higher sensitivity is obtained after 30 min of reaction, although results can be achieved more rapidly (after 4 min) if reaction takes place at 50 • C. Reproducibility was high (5.6% RSD), and the LOD was 2.9 × 10 −6 M. Another biosensor (Oliver et al., 2021) (with similar analytical characteristics) has two important advantages: pretreatment of cheese is not necessary for determination, and measurements are conducted using a smartphone. This system is based on colorimetric test strips containing TAO, peroxidase, and 3,3′,5,5′tetramethylbenzidine, which are added to the sample for 2 min for the reaction to take place, after which the RGB color coordinates are measured.

CONCLUSION
The considerable number of studies that analyze BA in dairy products reflect general concern about the presence and concentration of BA in this kind of product, in view of the adverse health effects produced by their ingestion, especially in histamine-sensitive or allergic patients. The most common sample pretreatment previous to BA analysis is extraction with acids or solvents to precipitate the proteins and extract the analytes from the matrix. Some studies improve on cleanliness and sensitivity by using SPE; however, a greater number of steps is not always worth the effort. Without a doubt, the most widely used analytical technique for the determination of BA is HPLC with derivatization: sensitivity, the structure of the BA, and the detection technique are the three main reasons for choosing one of the numerous derivatizing agents. One of these techniques' most important drawbacks is their slowness; the development of rapid methods, such as biosensors, is thus slightly on the increase (less than in other matrices), with good results in terms of sensitivity, although specificity still needs to be improved.

A C K N O W L E D G M E N T S
The authors acknowledge the continuous support of Gobierno de Aragón (T29 and A03) and European Social Fund. Marta Moniente acknowledges the Gobierno de Aragón for her predoctoral fellowship. The authors thank S. Hanks for his collaboration on the English-language revision of this manuscript.

C O N F L I C T O F I N T E R E S T
The authors declare no conflict of interest.  (2008). Determination of biogenic diamines with a vaporisation derivatisation approach using solid-phase microextraction gas chromatography-mass spec-