Sulphuric acid (95–97%, cat. no. 1.00731), 1-octanol (puriss., cat. no. 1.00991), sodium chloride (NaCl, p.a., cat. no.1.06404.1000), sodium nitrite (NaNO _2, p.a., cat. no. 1.06549.0100), sodium nitrite-N15 (Na ¹⁵NO _2, 99%, cat. no. 490814-1G), ascorbic acid (p.a., cat. no. 127) were all from Merck (Buchs, Switzerland) and sodium ascorbate (99%, cat. no. 352681000) from Acros (New Jersey, U.S.A.). Certified NO- gas (100 ppmv N ₂, 5 L, 150 bar, cat. no.1878) was purchased from Carbagas (Basel, Switzerland) and N ₂O- gas (100 ppmv N ₂, 2 L, 150 bar, cat. no. 7092325) was from Messer AG (Lenzburg, Switzerland).

Tuna samples (frozen loins, tuna saku, loose and pre-prepared packages) were taken from retailers and importers in Basel and stored at – 18°C until sample preparation. An aliquot of 15 g was weighed into a 70 mL centrifuge tube and 15 g of crushed ice as well as 15 mL of water were added. The mixture was homogenised at 7,000–9,000 rpm with a polytron homogeniser (Kinematica AG, Switzerland) and centrifuged for 5 min at 3,500 rpm. An aliquot of the resulting supernatant (5 mL) was transferred into a 20 mL headspace vial. After adding 5 µL of n-octanol and 2 mL of sulphuric acid (20 % in deionised water) the vial was immediately sealed and gently shaken by hand for 30 s. Certified NO- gas 100 ppmv (122.6 ng/mL) and N ₂O- gas 100 ppmv in N ₂ (180.6 ng/mL) served as standards for calibration: A headspace vial was filled with 5 mL of deionised water and 2 mL of sulphuric acid (20 %), the cap only loosely fitted and the reference gas introduced with a needle through the septum. The cap was then sealed tight.

The sealed vials were placed into the incubation unit of the CombiPAL autosampler (CTC Analytics, Zwingen, Switzerland) for 60 min at 25°C and analysed with a Trace 1310 GC / ISQ quadrupole system (Thermo Scientific, U.S.A.) under the following conditions: 500 μL of headspace was injected at 70°C (split 1:20) onto a Porous Layer Open Tubular (PLOT) column (30 m × 0.32 mm inner diameter, 12 µm film thickness, HP-PLOT Molsieve, Agilent Technologies, cat. no. 19091P-MS4), Palo Alto, CA, U.S.A.). Helium (99.999%, Carbagas, Basel, Switzerland, cat. no. P6201RG) was used as carrier gas at a constant flow rate of 1 mL/min. The oven temperature was programmed from 70°C (5 min) to 220°C (3 min) at 30°C/min. The quadrupole detector was set in the electronic ionisation mode (full scan 10–100 amu) with selected-ion monitoring (70 eV, ion source = 170°C, delay = 2.5 min) with a transfer-line to the GC (1 m fused silica deactivated, 0.25 mm i.d.) at 290°C. Specific masses for the detection were m/z 30 for NO (m/z 31 for ¹⁵NO) at a retention time of about 7.9 minutes and m/z 44 for N ₂O (m/z 46 for ¹⁵N ₂O) at 11.4 minutes.

For brining experiments, 15 g of nine untreated fish samples (certified as 100 % tuna from the Maldives, the Philippines, China, Vietnam) were each cut into pieces of about 1 cm and treated with 20 mL of brine solution (10% NaCl and 0.4% NaNO ₂ in deionised water) ⁸ at room temperature for 24 hours. After removing the solution, the now reddish looking tuna samples along with eight untreated control samples were each bathed twice in 15 mL of antioxidant solution (3.0 % ascorbic acid and sodium ascorbate in deionised water) for one hour at room temperature and subsequently washed twice with antioxidant solution. Thereafter, samples were prepared as described above and analysed by headspace-GC/MS.

Verification experiment: The whole procedure was also done with three tuna samples using Na ¹⁵NO ₂ for the brine solution instead.

In contrast to carbon monoxide, where the agent used for treatment, the ligand responsible for colour stabilisation and the target compound for sample screening is one and the same molecule, nitrite treatment identification has to consider the two-step reduction from nitrite to nitric oxide to nitrous oxide. This required changing the original GC temperature programme of the CO method: Instead of an isothermic run at 40°C, oven temperature was ramped up from 70°C to 220°C.

Both carbon monoxide and nitric oxide are bound in the sample as a ligand. Their release is enhanced by adding sulfuric acid. Being a free radical, however, nitric oxide is far more reactive than carbon monoxide: Instead of a sharp symmetrical peak, chromatograms show nitric oxide as a broad hump, barely distinguishable from baseline noise. The direct approach therefore is not feasible. In an acidic environment, however, two NO molecules, upon reduction, react to nitrous oxide (N ₂O), which shows up as a perfect peak in the chromatogram ( Figure 2), allowing for the indirect detection of NO. For the purpose of identifying nitrite treatment, a semi quantitative assessment of N ₂O proved to be sufficient, as any clear signal is alone due to a former nitrite treatment.

Verification of this mechanism was performed using isotope labelled nitrite (Na ¹⁵NO ₂) for brining ( Figure 1). During subsequent analysis, both the ¹⁵NO hump and ¹⁵N ₂O peak were detected ( Figure 2), which is consistent with the postulated reaction pathway. In addition, brining experiments with nine tuna samples using unlabelled NaNO ₂ showed sharp N ₂O signals in the range of 1000 to 12000 μg/kg ( Figure 3, underlying data ⁹ ). In contrast, untreated reference samples, even though they were exposed to antioxidants, were free of nitrous oxide (limit of detection about 20 μg/kg). Therefore, naturally occurring relevant concentrations of N ₂O can be excluded.

The method was validated with seven reference tuna samples of a trustworthy source from 2017 which were stored at -18°C for two years in our lab (see underlying data ¹⁰ ). Five of them were declared as untreated and consists of 100% tuna. The remaining two samples were declared as treated with rosemary extract (E392) and/or nitrite and labelled with additives such as salt and antioxidants (E300, E301, E331). The untreated samples were free of N ₂O whereas the treated ones showed stable N ₂O signals of 260 and 750 μg/kg. These results not only confirm our findings from the brining experiments, but are also consistent with the corresponding documents of the reference tuna samples.

The method was then used on a routine basis in a market survey of 13 samples of raw tuna collected in Basel in 2019 (see underlying data ¹¹ ). Seven samples (54%) were negative. Six samples (46%) were tested positive with N ₂O peaks corresponding to levels of between 90 and 1130 μg/kg. As a consequence, these six samples, all originating from Vietnam, were objected according to law.