Reagents and material

Liquid pepsin (660 u Ph. Eur./mL) was obtained from AppliChem GmbH (Darmstadt, Germany). Evans blue dye, NR dye, and n-hexane (C6H14) were obtained from Carl Roth GmbH & Co. KG (Karlsruhe, Germany). Carbon disulphide (CS2) and Tween20® were obtained from Honeywell International Inc. (Wabash, IN, USA). Calcofluor white staining agent, chloroform (CHCl3), dichloromethane (CH2Cl2, DCM), fuming hydrochloric acid (HCl), and potassium iodide (KI) were obtained from Merck KGaA (Darmstadt, Germany). Acetone (C3H6O), ethanol (C2H6O), hydrogen peroxide (H2O2), isopropanol (C3H8O), and potassium hydroxide (KOH) were obtained from Th. Geyer & Co. KG (Renningen, Germany). All chemicals were of analytical purity grade.

Reference particles of synthetic and natural polymers

Commercially relevant plastic particles and nurdles [33], UV-aged MP, coloured household MP, potential procedural contaminants (cotton fibres), and PNO potentially occurring in edible tissues of seafood were selected for establishing fluorescence threshold values for MP identification of NR-stained particles. Synthetic polymers were provided by the Bundesanstalt für Materialforschung und -prüfung (BAM, Berlin, Germany), referred to as BAM-MP. Further polymers were purchased from Goodfellow Cambridge Ltd. (Lille, France), and Alfa Aesar (Haverhill, MA, USA), referred to as in-house reference. PNO were obtained from fishbone, shrimp shells, mussel shells, and cotton. When necessary, small particles (≤ 500 µm) were obtained by cutting, precipitation, or ultra-centrifugal milling and consecutive sieving with stainless-steel sieves. A comprehensive list of materials is provided in the supplementary information (SI), Table S1. For sample spiking, the most commonly detected MP in food were selected, namely nylon 6 (PA6), polyethylene (PE), polyethylene terephthalate (PET), polypropylene (PP), polystyrene (PS), and polyvinylchloride (PVC) [34]. Furthermore, nylon 12 (PA12) was sieved with a 50 µm and 25 µm stainless-steel mesh for spiking with particles within a small size range. Particles were suspended in solutions of Tween20® and KI or isopropanol depending on the polymer density (Table 1). Particle counts of each spiking suspension were determined by pipetting 100 µL-aliquots (n = 5) onto glass fibre filters, consecutive NR staining, fluorescence microscopy, and image analysis as described in “Sample analysis”. For pipetting, a displacement pipette (Transferpettor™, Brand GmbH & Co. KG, Wertheim, Germany) equipped with a glass capillary (1.95 mm opening) was used. To avoid particle sedimentation or floatation, the particle suspension was shaken vigorously before each pipetting step.

Table 1 Composition of MP suspensions used for sample spiking; detergent–aqueous 0.5% Tween20®-solution; RSD relative standard deviation; rcv recovery of polymer mass estimate based on weighed particle massPrevention of procedural contamination

Experiments were conducted in a laboratory with restricted access wearing a white cotton lab coat and trousers. Filtration and filter treatment (oxidation, staining) took place within a laminar flow box. All liquids (reagents, solvents, water) were filtered with glass fibre filters (0.7 µm particle retention, Th. Geyer & Co. KG, Renningen, Germany) directly before use. Labware not suited for thermal cleaning (e.g. filtration apparatus, PTFE-coated stirring rods, heat-sensitive filter membranes) was rinsed three times with 10 mL deionised water (DI water, generated with a reverse osmosis system and additional mixed bed filter). Glass slides and flasks (covered with aluminium foil) and glass fibre filters (stored in Petri dishes) were heated at 500 °C for 5 h. Glass flasks were additionally rinsed with 10 mL DI water prior to use. Preliminary analysis of singular potential contamination sources (e.g. glassware, deposition from air, reagents) indicated a high randomness of each individual source. Therefore, three procedural blank samples were prepared for each sample series to account for the total contamination of the respective series. The procedural blanks were prepared and analysed like matrix samples but using the respective amount of DI water instead of seafood.

Sample preparationHomogenisation and spiking of seafood matrix

Whole herring (Clupea harengus), and fresh salmon fillets with skin (Salmo salar) were purchased from a local market and transported on ice in an expanded polystyrene box as supplied by the merchant. Frozen whitefish fillet (Theragra chalcogramma), shrimps (Penaeus longirostris), and fresh mussels (Mytilus edulis) were purchased pre-packaged from German retail stores. Non-edible tissues (skin, shells, fishbone, innards) were removed. Fish fillets, shrimp tails, and mussels’ tissues were then homogenised with a commercial stainless-steel hand blender. Samples were stored at − 20 °C in aluminium cups covered with aluminium foil.

A subset of samples was spiked with MP for recovery tests and method comparison. MP mass estimation was evaluated with weighed BAM-MP spiked to 1–2 kg aliquots homogenised herring fillet in different concentrations (0.75 mg/kg, 30.18 mg/kg, and 234.66 mg/kg). The spiked homogenates were mixed again with a hand blender. Subsequently, 1 g and 10 g aliquots were weighed into glass flasks (n = 5). Particle counting was evaluated by spiking pre-homogenised salmon fillet with 100 µL in-house reference MP suspensions (Table 1) in a mixture (PA6, PE, PET, PP, PS, PVC) and with PA12 (n = 5 each). Additionally, 10 mL filtered DI water was spiked with the same amount of each spiking suspension (or only PA12) and immediately filtered (n = 3).

Extraction of MP from edible seafood tissue

Aliquots of 10 g homogenised seafood were digested with a two-step procedure as described by Süssmann et al. [35]. First, the sample was digested with 90 mL of a 1% pepsin solution in 0.063 mol/L HCl (stirring for 2 h at 40 °C). Afterwards, 10 mL 50% KOH solution (50:50, w/w in water) was added (stirring for 4 h at 40 °C) [36]. MP was isolated from most digested samples by vacuum filtration using Ø 47 mm PTFE filters (pore size 1–2 µm; Pieper Filter GmbH, Bad Zwischenahn, Germany). Spiked herring fillet was filtered with silver filters (pore size 0.8 µm; Pieper Filter GmbH, Bad Zwischenahn, Germany). PA12-spiked salmon was filtered with glass fibre filters (particle retention 1.2 µm; Th. Geyer GmbH & Co. KG, Renningen, Germany). The glass flask and filtration apparatus were rinsed three times with 10 mL DI water and once with 10 mL isopropanol. The filters were then placed in glass Petri dishes, covered with 2 mL H2O2 solution (15% in DI water, v/v) and dried for 48 h at room temperature.

Staining of sample filters

Optimal conditions for NR staining of edible seafood samples were determined with preliminary tests (SI section 2.5). Seafood samples were stained with 1 mL NR in hexane (c = 50 µg/mL) for 30 min at 40 °C. Afterwards, 1 mL NR in ethanol:acetone (1:1, v/v; EtAc; c = 50 µg/µL) was added to the filters and incubation was repeated for 30 min at 40 °C. After cooling to room temperature, excess dye was removed from the filter surface by rinsing with 5–10 mL isopropanol.

A subset of seafood samples was additionally stained with 0.5–1 mL Calcofluor white staining (10 min incubation at room temperature) after optimised NR staining [36] in order to test the effects of counterstaining on MP detection in seafood.

Sample analysisAnalysis with fluorescence microscopy and semi-automated image processing

Particles on the filter were detected and analysed by fluorescence microscopy using the Axioscope 7 equipped with an Axiocam 503 colour camera (both Carl Zeiss AG, Germany), 565 nm LED illumination (5% intensity, 100 ms exposure), and an orange filter. Samples were observed with a 2.5 × or 5 × objective (10 µm or 5 µm resolution respectively). No colour correction was applied. A subset of samples was analysed with a 10 × objective (1 µm resolution). Due to the 10 × objective’s low depth of field, a z-stack of the sample was measured and a 2D image was generated by applying maximum projection.

Particle size, morphology, brightness, and colour were obtained with image analysis. Therefore, binary images of the scans were generated with Adobe Photoshop® (manual adjustment of brightness and contrast, separation of particles and background). Morphological attributes were obtained by analysing binary images with ImageJ. Isolated pixels (artefacts from image editing) were removed using the “open” function. Fluorescence colour and total particle brightness (TPB) were obtained with the original colour images.

MP estimation and procedural blank correction

Particle numbers were assessed in size classes of 5–10 µm, 10–50 µm, 50–100 µm, 100–500 µm, 500–1000 µm, and 1000–5000 µm [16]. Each size class was further separated by morphology and fluorescence. For particle mass estimation, the two-dimensional particle morphologies were approximated to three-dimensional objects, namely spheres (spheroids), cuboids (fragments,) and cylinders (fibres) for volume calculation. Details on the calculations are provided in SI section 2.2. The number and mass of MP suspect particles of each series were corrected by the respective procedural blanks [37]. Therefore, a limit of quantification (LOQ) was calculated based on the mean particle number of the respective procedural blanks plus ten times the standard deviation for each combination of size class, morphology, and fluorescence group. Results exceeding the LOQ were corrected by subtracting the mean particle number or mass estimate of the procedural blanks for the respective particle type or mass category.

Quantum cascade laser-based laser direct imaging analysis

The proposed method was compared with LDIR imaging for assessing the plausibility of MP analysis. Therefore, MP-spiked samples were first analysed with fluorescence microscopy and then with LDIR imaging (8700 LDIR Chemical Imaging System, Agilent Technologies Inc., Santa Clara, USA).

Aliquots of 1 g (n = 7) of spiked herring fillet homogenate were digested with 9 mL pepsin solution and 1 mL KOH. Five samples were filtered with glass fibre filters (1.2 µm particle retention, 25 mm diameter) from Th. Geyer & Co. KG (Renningen, Germany). Two samples were filtered with polyethylene terephthalate glycol (PETG) gold-coated membrane filters (0.2 µm pore size, 100/0 nm coating, 25 mm diameter) from Sterlitech Corp. (Auburn, WA, USA), as required for LDIR analysis. Due to the small pore size (0.2 µm), the filters clogged rapidly and filtration was aborted after 20 min, discarding the remaining liquid. The filters were rinsed once with 3 mL filtered DI water and placed onto GFF stored in Petri dishes for better soaking with H2O2. After drying, the samples were stained with NR (c.f. 2.3.3). The stained filters were mounted on specialised filter holders (Agilent Technologies Inc., Santa Clara, USA) and imaged with fluorescence microscopy (c.f. 2.4.1). The sample holder was then stored in a Petri dish for transport and analysed with LDIR imaging [15, 38]. Hereby, the size fractions 10–100 µm and 100–5000 µm were analysed separately using the Clarity Software (Version 1.5.58, Agilent Technologies Inc., Santa Clara, USA). Accordingly, the datasets were merged and evaluated applying a hit quality index of 0.85 (Pearson’s correlation coefficient of 1st derivatives of IR spectra) by means of a custom-written Excel© spreadsheet.