2.1 Materials
Pyrex glassware was used whenever possible during this study. Whatman (grade 4; 25 μm pore size) and nitrocellulose (pore size of 1 μm) filter papers (Sigma Aldrich) were used for Fourier-transform infrared spectroscopy (FTIR) and Raman spectroscopy analysis. Minnow traps were used for collecting Atlantic killifish.
2.2 Quality Control
We followed the quality control criteria established by Hermsen et al. (2018). All water samples and solutions were filtered through a 0.2 μm filter prior to use. All sample processing apparatus (glass Erlenmeyer flasks and ceramic Buchner funnels) were combusted in a muffle furnace (Fisher Scientific Isotemp Programmable Forced-Draft Furnace) at 500 ℃ for 5 h. Glassware that could not be combusted in the furnace was rinsed three times with both acetone and filtered water prior to use. All work surfaces were wiped down with ethanol prior to working with samples. Sample manipulation took place in a laminar flow hood (AirClean 600 PCR Workstation). Samples were not exposed to ambient air. When the samples were not being actively manipulated, they were covered with aluminum foil. While working with samples, a 100% cotton jumpsuit was worn to prevent contamination from polyester clothing. During sample processing, occupancy of the room was kept to a single person. Procedural blanks were used for every 5 samples.
2.3 Fish collection and tissue sampling
Atlantic killifish were collected using minnow traps from two locations in Buzzards Bay, Massachusetts (Fig. 1). The sampling dates, coordinates, number of fish collected, and the environmental conditions are shown in Table S1. Fish were immediately euthanized using MS222 (1 g/L) buffered with sodium bicarbonate. Fish were kept in an aluminum-foil lined bucket during transport. The total length and weight, sex, and tissue wet weights were recorded upon collection. Prior to dissection, fish were rinsed three times with 0.2 μm filtered water to remove any loose particles on the fish’s skin. The GI tract and a section of the dorsal muscle (without skin) were collected. The total GI tract, including contents, were used in our analysis. Tissues were stored in plastic-free aluminum foil at -80 °C. All collected fish were considered mature as they exceeded 3.2 cm or 3.8 cm for males and females, respectively, and they were in good overall health according to the calculated condition factors (Abraham, 1985). Based on length-age relationships (Abraham, 1985), the collected fish from Bourne, MA appear to be a couple of years younger than the collected fish from Falmouth, MA.
At both sampling sites, we collected water samples in 1 liter glass jars. The jars were rinsed three times in water from the collection site prior to sample collection. To prevent air contamination, the jars were immersed underwater prior to opening. Duplicate water samples were collected from each site. Samples were stored at room temperature (20 - 25 ℃).
2.4 Sample Digestion
A flow chart of the sample processing and analysis is shown in Figure 2. Tissue samples were digested with 10% KOH (3x sample wet weight) at 60 ℃ for 48 h (Fig. 2). KOH digestion at this temperature has previously been shown to not degrade MPs (Gulizia et al., 2022). Digests were neutralized with a combination of sodium bicarbonate (0.05 g/mL) and 10% HCl (0.54 mL HCl/mL KOH) prior to filtration. Particles in the samples were size fractionated by filtering initially through a 25 μm pore-size filter followed by a 1 μm filter (Fig. 2). Filters were stored in plastic-free aluminum tins until analysis. The 25 μm pore-size filter was used for FTIR spectroscopy analysis and the 1 μm filter was used for Raman spectroscopy (Fig. 2). Size fractionation was used to analyze samples more efficiently. Raman spectroscopy is more accurate at identifying particles smaller than 20 μm in length, and FTIR is a more efficient method for particles >20 μm in length (De Frond et al., 2023).
2.5 Particle Recovery experiments
Particle recovery experiments were conducted using polyvinyl chloride (PVC), polyethylene terephthalate (PET), and polyethylene (PE) particles (250 μm) dyed with 300 μL of Nile Red (10 μg/mL). Blue mussel (Mytilus edulis) tissue was used to determine the particle recovery from the MP isolation process from biological tissues. Fifty particles from each polymer type were manually added to whole mussel tissue (Mytilus edulis) in Erlenmeyer flasks. The mussels were then digested for 48 h as in section 2.4. The resulting digests were then filtered through 25 μm pore-size filters. Each flask was washed three times with filtered water. The number of particles on the filter were counted under a dissecting microscope. There was 80% recovery of particles regardless of the polymer type (n=2 experiments per polymer) (Fig. S1).
2.6. FTIR spectroscopy Sample Analysis
The 25 μm filters were scanned for particles under a dissecting microscope at 8.6x magnification. The whole filter was examined, and any particles found were imaged (ThorCam Imaging Software V.3.5.1.1). Particles were then transferred to a piece of double-sided tape in a glass petri dish and numbered. These particles were analyzed with a diamond attenuated total reflection (ATR) attachment on the Cary 630 FTIR (Agilent Technologies Inc). Particles were placed individually on the detection area. MicroLabPC software was used to collect the spectra.
2.7. Raman spectroscopy Sample Analysis
The 1 μm filter was used for Raman analysis. The particles were analyzed with a Renishaw inVia Raman microscope using Wire 3.4 software to collect images and spectra of the particles. Spectra were collected with a 532 nm excitation laser. The filter was scanned at 20x magnification. Eight transects were made in a straight line across the filter covering 8% of the filter (Fig. 2). Particles were imaged prior to spectra collection. If particles were smaller or larger than the 20x field of view, 50x or 5x magnification was used to image the particle.
2.8. Data Validation and Analysis
The collected spectra from FTIR and Raman spectroscopy were baseline corrected and smoothed prior to identification. OpenSpecy (Cowger et al., 2021) and a custom library database were used for identifying the particles based on their spectra. Pearson’s correlation coefficient (Pearson’s r) of greater than or equal to 0.8 was used as a statistical cutoff for a good fit between the reference and the particle spectrum (Giani et al., 2023; Zhu et al., 2023). All the plastic particle spectra (Pearson’s r > 0.8) were then manually checked to ensure that the spectral peaks were well matched with the reference peaks from both the libraries (Renner et al., 2019). Particles (Pearson’s r > 0.8) were not used in the subsequent analysis if they had peaks that either did not match the overall pattern in the reference spectrum or had a low enough signal-to-noise ratio that it was challenging to interpret the true signal. Examples of spectra that were included or excluded in the analysis are shown in Figure 3,
Spectral peaks were identified according to commonly reported Raman shifts in the literature. Plastic degradation or weathering peaks were identified in all of the samples. Non-plastic particles were compared to reference spectra for a variety of different materials, including fur, cellulose, cotton, sand, chitin, and plant material. The non-plastic particles were identified by a Pearson’s r of 0.8 or greater without manually comparing the spectra matches.
2.9. Statistics
Fulton’s condition index (K) (Equation 1) was calculated to determine the overall health of the fish (Ricker, 1975).
Equation 1: K=(W/TL3)* 100; W is the fish weight and TL is the total length.
MP abundance and occurrence data were non-normally distributed, so nonparametric analyses were used. To predict the occurrence of MPs in the muscle, a Random Forest model was generated using the R CARET package’s leave-one-subject-out train() function (Kuhn, 2008). Fish length, sex and plastic content in the GI tract were used as input variables. Spearman’s rank correlation was used to determine the correlation between fish total length and GI tract and muscle MP abundance. Graphpad Prism (10.0.2) was used to calculate the Spearman’s rank correlations and to generate the visualizations. Significant correlations were accepted if the p-value of the Spearman’s rank correlation was less than 0.05.
The Random Forest analysis used MP occurrence in muscle samples, and the Spearman’s rank correlation used the muscle MP abundance.
All concentrations are reported as the number of particles per gram of tissue wet weight. Data from the Raman analysis are based on the whole filter plastic count estimates extrapolated from the 8% of the filter that was analyzed. No corrections were made for the FTIR data since the whole filter was analyzed.