**3. Results**

With the developed MS based methodologies we were able to quantify the amount of the amines dimethylamine (DMA), trimethylamine (TMA), trimethylamine-N-oxide (TMAO), and cadaverine (CAD) with the UHPLC-tandem mass analysis. The concentration of H2O2 residues was measured with the use of a HS SPME-GC-MS method.

#### *3.1. Results of UHPLC-Tandem Mass Analysis of NPN*

The chromatographic separation of the analyzed amines is shown in Figure 1. The ion pairing effect of the heptafluoro-butanoic acid present in the mobile phase allowed a valuable retention to obtain a satisfying separation of the analytes.

**Figure 1.** Chromatographic separation of dimethylamine (DMA, Rt 2.2 min), trimethylamine (TMA, Rt 2.25 min), trimethylamine-N-oxide (TMAO, Rt 2.5 min), and cadaverine (CAD Rt 4.3 min).

In order to quantify the amines in fresh fishery products samples or in fishery products subjected to an illicit treatment with H2O2, three calibration curves were prepared: (I) In pure pH 2.5, 0.1 M phosphate buffer, (II) In the extraction buffer fresh samples of Atlantic bonito, and (III) In the extraction buffer of fresh samples of European squid.

Each matrix material was weighted and extracted as previously described; since the matrix has a basal amount of amines, it was mandatory to prepare a matrix-blank without the addition of analyte (DMA, TMA, TMAO, and CAD) standards. Once obtained, the matrix-blank and the calibration curves were obtained by adding increasing amounts of amines as follows: 50, 100, 200, 400, 600, 800 ng/mL. A standard addition curve of TMAO in a real sample of European squid is shown in Figure 2. The curve had a positive intercept value because of the basal amounts of the analyte in fish.

**Figure 2.** A standard addition curve of TMAO in a real sample of fresh European squid. The basal concentration of TMAO was 1150 ng/mL.

A full validation of the UHPLC-MS method for amines determination was performed. We followed the Food and Drug Administration (FDA) guidelines to evaluate the protocol [30] and used our previous work as an example for validation parameter definition [32]. Validation parameters were listed in Table 2 for all the UHPLC-MS analytes. LOD and LOQ were evaluated on the standard calibration curve by the signal to noise values of 3 and 10, respectively. LLOQ is the lower limit of quantification determined on the basis of a simple LOQ. To obtain LLOQs, LOQ standard solutions were prepared, used as the first level of each calibration curve, and validated by comparison of blank solutions. Validation parameters definition is given in the following paragraph.

To test the possible interferences at the analyte of interest's known retention time and *<sup>m</sup>*/*<sup>z</sup>*, a standard solution that had every analyte at a known concentration except for the one of interest, was analyzed. The selectivity % (Sel% = (Area at analyte retention time/Average area analyte in LLOQ) × 100) had to be ≤30%. The statistical parameter related to the linearity of calibration curve is the percentage difference (Diff% = (slope − average slope)/ average slope) × 100). Diff% had to be ≤25%. Inaccuracy of lower limit of quantitation (LLOQ) had to be ≤20% and relative standard deviation % of accuracy of LLOQ ≤15%. Finally, to test recovery, matrices were spiked with a combined standard solution of amines at known concentration and processed as described in the Sample preparation section. The recovery (Rec = (Area at analyte retention time in spiked matrix/Average analyte area in spiked water solution) × 100) had to be between 85% and 120%. The validation parameters were respected in all cases.

**Table 2.** Validation parameters for calibration curves in (I) pH 2.5, 0.1 M phosphate buffer, (II) Atlantic bonito extraction solution, (III) European squid extraction solution. Diff% slope: Difference % of the slope of the calibration curve; RSD% LLOQ%: Relative standard deviation % of accuracy of lower limit of quantitation; BIAS% of LLOQ: Inaccuracy of lower limit of quantitation; LLOQ: Lower limit of quantitation.


The UHPLC-tandem mass method was then applied to real fishery product samples of Atlantic bonito (*Sarda sarda)* and European squid (*Loligo vulgaris*): A) Freshly caught; B) freshly purchased in a local market; C) aged (four days at room temperature); and D) H2O2 treated (as described before).

The obtained results are summarized in Table 3. In freshly caught and freshly purchased samples the measured amount of amine was similar; in Table 3 the quantity of amines in freshly caught samples only was reported. High levels of TMAO were found in freshly caught Atlantic bonito (1700 ± 238 mg/kg) and European squid (1200 ± 336 mg/kg) samples. Conversely, in these samples the TMA amount was low, 170 ± 24 and 210 ± 29 mg/kg, respectively.

When fishery products initiated to degrade due to temperature (4 h at room temperature) or bacterial activities, the TMAO and TMA amounts reversed. Finally, when hydrogen peroxide was used as an illicit treatment as whitening and refreshing agents, exploiting its oxidant properties, the balance was shifted again towards a higher level of TMAO. In *Sarda sarda* the values after H2O2 treatment were TMAO 410 ± 57, TMA 720 ± 180 mg/kg; and in *Loligo vulgaris* were TMAO 850 ± 212, TMA 200 ± 28 mg/kg.


**Table 3.** Concentration values of trimethylamine-N-oxide (TMAO) and trimethylamine (TMA) in real fish samples of Atlantic bonito (*Sarda sarda*) and European squid (*Loligo vulgaris*). The amount is expressed in mg/kg. Fresh referred to freshly caught samples; aged to samples left at room temperature for 4 h; H2O2 treatment to samples treated with hydrogen peroxide (see Material and Methods).

#### *3.2. Results of GC-MS Analysis of H2O2 Residues*

To confirm the illicit managemen<sup>t</sup> of fishery products with hydrogen peroxide, we implemented a previously published GC-ECD assay based on peroxide detection by indirect oxidation of anisole to guaiacol (2-hydroxyanisole) (Scheme 1) [31].

**Scheme 1.** Oxidation reaction with hydroxyl peroxide of anisole to guaiacol (2-hydroxyanisole) catalyzed by potassium ferricyanide.

We developed a headspace solid-phase microextraction (HS SPME) GC-MS methodology to improve sensitivity, easiness of operation, and reliability.

To quantify the residues of H2O2 we prepared a calibration curve with the standard addition method using fresh fish product samples of European squid and Atlantic bonito as matrices. The extracted solution of fresh fishery products was added with aliquots of hydrogen peroxide to give amounts of 0.0, 0.05, 0.1, 0.25, 0.5, and 1.0 μg/mL of H2O2. Then, 100 μL of potassium ferricyanide (K8Fe(CN)6) and 2 μL of anisole were added to the obtained solutions before fiber exposing for head space analysis. We measured the 2-hydroxyanisole peak area which was related to the H2O2 addition. Figure 3 shows a standard addition calibration curve of H2O2 in European squid samples. The obtained LLOQ was 50 ng/mL. When the concentration of hydrogen peroxide was zero, no 2-hydroxyanisole was detected.

**Figure 3.** A standard addition curve of H2O2 in a fresh sample of European squid.

We then applied the developed HS SPME-GCMS method to five real fishery products samples, in particular squid which were the subject of illicit treatment with hydrogen peroxide due to its properties of whitening agent.

QC and real samples were prepared as described in the Sample Preparation section: The amount of H2O2 added for the redox reaction of anisole to guaiacol was set at 0.5 μg/mL. In these conditions we tested a blank solution without fishery products (no *Sarda sarda* or *Loligo vulgaris*), fresh (blank matrix), fresh and treated in controlled conditions (QC) and illicit treated fishery products samples of Atlantic bonito and European squid. All of the five samples coming from legal controls showed H2O2 values higher than 100 ppm. Table 4 shows the results. The peak area of 2-hydroxyanisole in the buffer without matrices confirmed the added amount of H2O2 of 0.5 ± 0.07 ppb. In the presence of matrices, the peak area decreased due to the matrix effect. In the case of fresh fish product samples, this effect was quantifiable in a loss of 12% of H2O2 amount (0.44 ± 0.11 μg/mL for Atlantic bonito and 0.43 ± 0.10 μg/mL for European squid). In fresh samples treated with hydrogen peroxide in controlled conditions (complete immersion of samples in a 0.8% H2O2 solution followed by rinsing with fresh water in laboratory) the amount of H2O2 was much higher than 1 ppm, the highest calibration curve point. However, extrapolating the results, it seemed larger than 100 ppm. The same was for the samples, especially squid, illegally treated with hydrogen peroxide.

**Table 4.** Quantitation of H2O2 in a 0.1 M phosphate buffer pH 2.5 solution and in fishery products samples. QC were quality control samples treated with hydrogen peroxide in controlled conditions (0.5 μg/mL of H2O2); five real samples were illegally treated with peroxide.

