Comparison of pre-treatment methods and heavy density liquids to optimize microplastic extraction from natural marine sediments

The ubiquitous occurrence of anthropogenic particles, including microplastics in the marine environment, has, over the last years, gained worldwide attention. As a result, many methods have been developed to estimate the amount and type of microplastics in the marine environment. However, there are still no standardized protocols for how different marine matrices should be sampled or how to extract and identify these particles, making meaningful data comparison hard. Buoyant microplastics are influenced by winds and currents, and concentrations could hence be expected to be highly variable over time. However, since both high density and most of the initially buoyant microplastics are known to eventually sink and settle on the seafloor, marine sediments are proposed as a suitable matrix for microplastics monitoring. Several principles, apparatuses, and protocols for extracting microplastics from marine sediments have been presented, but extensive comparison of the different steps in the protocols using real environmental samples is lacking. Thus, in this study, different pre-treatment and subsequent density separation protocols for extraction of microplastics from replicate samples of marine sediment were compared. Two pre-treatment methods, one using inorganic chemicals (NaClO + KOH + Na4P2O7) and one using porcine pancreatic enzymes, as well as one with no pre-treatment of the sediment, were compared in combination with two commonly used high-density saline solutions used for density separation, sodium chloride (NaCl) and zinc chloride (ZnCl2). Both pre-treatment methods effectively removed organic matter, and both saline solutions extracted lighter plastic particles such as polyethylene (PE) and polypropylene (PP). The most efficient combination, chemical pre-treatment and density separation with ZnCl2, was found to extract > 15 times more particles (≥ 100 µm) from the sediment than other treatment combinations, which could largely be explained by the high presence and efficient extraction of PVC particles.

Approximately 400 mL of the chemical treatment (twice the volume of the sediment) was added to the glass containers with glass lids with the sediment. The samples were incubated at room temperature for 1 h on an oscillating stirring table at 160 rpm. An additional rinsing step was required after the chemical treatment to lower the pH and avoid precipitation of zinc hydroxides. Subsequently, the sediment was washed by adding Milli-Q water followed by vigorous shaking and centrifugation for 30 min at 1000 rpm. This washing procedure was repeated three times. The supernatant was removed between repetitions, and new Milli-Q water was added. The supernatant was filtered through a 50 μm filter, which was kept for further analysis. After centrifugation, the sediments were transferred back to the glass containers. Porcine pancreatic enzymes were used for the enzymatic digestion of organic matter 43 . A buffered enzyme working solution was prepared by dissolving one capsule of the pharmaceutical enzyme (Creon 40 000, Abbott Laboratories GmbH, Germany, Mylan) per 10 mL of Tris hydrochloride solution (Trizma, pH 8.0, 1 M, 0.2 μm filtered, Sigma-Aldrich, T3038, USA). Complete dissolution was achieved through gentle warming (30 °C) and shaking at 125 rpm for 30 min on a heated incubation shaker (New Brunswick Scientific, Innova 40). Ten mL of the working solution was added per g wet weight (WW) of sediment in a prewashed glass jar. The sediment-enzyme solution was mixed, and the pH was assessed using a pH indicator stick. The pH was adjusted to 8.0 by adding more Tris hydrochloride solution if necessary. The filled glass containers were lidded and placed on vigorous shaking (150 rpm) at 37.5 °C overnight on the incubation shaker. The level of degradation was quantified by TOC analysis using an elemental analyzer coupled to an isotope ratio mass spectrometer (20-22, Secron Ltd., Crewe, UK) before and after degradation treatment.
Density separation and filtration. To compare the efficacy of two different widely used density separation solutions, NaCl (density 1.2 g/cm 3 ) and ZnCl 2 (1.8 g/cm 3 ), the pre-treated samples were divided into two groups further processed with ZnCl 2 (n = 7) or NaCl (n = 8). The density separation was performed using the Kristineberg Microplastic Sediment Separator (KMSS). This separator has been designed in-house with inspiration from the Munich Plastic Separator 19 , but is smaller, both in height and width, with a steeper incline on the standpipe with a glass cylinder above the sediment container to monitor the sedimentation (Fig. 1).
After pre-treatment, the sediment samples were moved to the bottom sediment container of the KMSS, the standpipe part of the tower was mounted, and the rotor, positioned in the bottom of the tower, was turned on. The saline solution was introduced through the bottom valve and filled to 85% of the tower's volume. The rotor was turned off three hours later, and the sediment was allowed to settle. After 12 h of settling, the dividing chamber was mounted on top of the standpipe, and the tower was filled with the density separation fluid, either ZnCl 2 or NaCl. When filled, the ball valve was closed, the liquid level lowered, and the dividing chamber was removed. Next, the dividing chamber was turned upside down, and the solution was filtrated. The sediment exposed to the pre-treatment with inorganic chemicals and separated with ZnCl 2 was filtrated through a 300 μm polyamide (PA) mesh and the remaining top solution was collected, and a second separation with ZnCl 2 in a glass beaker was performed due to high mineral content. After the second-density separation (agitation and left to settle for 24 h), the solution was filtered through a PA filter with mesh sizes 100 μm. For the enzymatic and none treated samples as well as all samples extracted with NaCl, only one separation was performed since fewer mineral particles were extracted compared to pre-treated with inorganic chemicals and separation with ZnCl 2 . The solution was filtrated through two PA filters with mesh sizes 300 μm, and 100 μm. After filtration, the filters were rinsed with Milli-Q water to remove salt crystals. www.nature.com/scientificreports/ QA/QC. All equipment used during sampling was cleaned and rinsed in the laboratory with Milli-Q water before drying in a plastic-free flume hood. For the storage of sediment samples, glass containers with glass lids were used, and a stainless-steel pot and a stainless-steel spoon were used for homogenization of the sediment. All samples were stored in their glass bottles until processed in the lab. All equipment and lab surfaces were cleaned before laboratory work, and only cotton lab coats and clothes were worn. The ZnCl 2 was filtrated through three membranes 10 μm, 5 μm, and 1 μm. The NaCl was filtrated through 10 μm. Six blank samples without sediment were processed, one for each combination of pre-treatment and density separation solution. All blanks were treated as their corresponding sediment sample from cleaning the bottles before sampling, in the field, during storage, treatment, separation, and analysis.
Analysis. The analysis workflows followed the current consensus guidelines, either published or available as working drafts from regional seas conventions and the European Commission 10,14,32 . All filters were first visually inspected with a stereomicroscope (Leica M205 C 80-160× Wetzlar, Germany). Subsequently, all particles that were visually identified as suspected anthropogenic following Karlsson et al. 2020 44 were transferred with www.nature.com/scientificreports/ a tweezer to different aluminum-oxide filters with a pore size of 200 nm (Whatman Anodisk 25). One filter for each treatment and size fraction, i.e., the particles collected on the 300 μm mesh were moved to one aluminumoxide filter, and the particles from the 100 μm were moved to another. The particles collected from the centrifugation solution were also transferred to one filter. Two filters were used if there were more suspected particles than what could fit on one Anodisk filter. The entire area of the Anodisk filters with particles was imaged with a light microscope (Zeiss, AxioImager). All particles (collected on the 300 μm filters, 100 μm filters, and from the centrifugation solution, 50 μm) were characterized according to their visual appearance, following Karlsson et al. 2020 44 were their 2D shape, 3D shape, solidity, color, and classical visual identification were noted. Subsequently, all particles from the 300 μm filter and all particles on eight randomly chosen 100 μm samples were chemically identified with Raman microscopy (Witec, alpha 300R) using a 532 nm laser and 600 g/mm grating. The laser power was selected based on the polymer, signal intensity, size of the particles, and magnification. Spectra were measured in a wavenumber range of 200 to 3500 cm −1 and were compared to our in-house library for identification (HQI minimum 75, majorities above 80). The library consists of spectra we have obtained from known plastic particles, including weathered particles, and spectra from the RUFF 45 and ST Japan spectral databases. For comparison, 300 µm filters from two randomly chosen samples were also chemically identified with Fourier-transform infrared spectroscopy, FTIR (Thermo Scientific Nicolet iN10) using transmission mode (256 scans, resolution 4 cm −1 , spectral range 4000-675 cm −1 , detector cooled by liquid nitrogen and correlated against 256 background scans). The particles in the supernatant from the centrifugation solution (all inorganic pre-treatments and one of the enzymes pre-treated) were first visually characterized before being analyzed with Raman microscopy. The remaining particles on the PA filter (300 μm, 100 μm, and 50 μm from centrifugation solution) from two random samples, i.e., the particles that were not visually identified as anthropogenic particles, were analyzed with Raman microscopy to identify false negatives.

Results and discussion
Particle identification, sizes, and concentrations. The anthropogenic particles extracted from the sediment were first categorized using a stereomicroscope according to their visual and tactile appearance 44,46 . The categories used were semi-transparent microplastics, white microplastics, black firm elastomer, paint particles, synthetic fibers, and other anthropogenic microparticles (Fig. 2). On the 300 μm filter, most of the plastic particles were semi-transparent (59% of 277 particles) and visually divided into three subclasses based on apparent differences in morphology. Subclass 1 had a stripe pattern, subclass 2 was built up by spherical patterns, and subclass 3 was the semi-transparent particles that had neither stripes nor a spherical pattern. All particles within visual subclass 1 were confirmed to be PE (n = 106), and all particles within subclass 2 were confirmed to be PP (n = 15) (Fig. 3). In subclass 3, polymers such as PS and PMMA were identified. This implies that, in this case, visual identification could be used to separate the semi-transparent particles into subclasses with different polymer compositions. The specific surface patterns of the semi-transparent PP and PE particles indicate that they both derive from distinct sources. In total, 66% of the semi-transparent particles on the 300 μm filter were identified as PE and 10% as PP (Fig. 4A). As expected, more anthropogenic particles were collected on the 100 μm filter, in total, 4329 particles (Fig. 4B, Table 1). On the 100 µm filters, there were two main categories of plastic particles, semi-transparent and white, representing 46% and 38% of the total number, respectively. All spectroscopy analyzed particles (n = 335) were confirmed to be PE, PP, and PVC, as identified visually. From the supernatant of the centrifugation samples, the majority of the particles (64%) were categorized as semitransparent and spectroscopically identified as PE (n = 47) and PP (n = 9). The highest concentration, 2.82 microplastic particles per g dry weight (DW) of sediment, was found in the samples exposed to the inorganic chemical pre-treatment and extracted with ZnCl 2 . Most of these particles (74%) were identified as PVC (Fig. 4). The number of PVC particles was distinctly higher in the combination of inorganic chemical pre-treatment and ZnCl 2 compared to the other treatments, 1.81 to 2.29 PVC particles/g DW sediment compared to 0-0.12 PVC particles/g DW sediment (Fig. 4C).  www.nature.com/scientificreports/ The concentration of semi-transparent plastic particles, i.e., mainly PP and PE particles, ranged from 0.25 to 0.76 particles per g DW sediment. The highest concentrations of semi-transparent microplastics were found in samples pre-treated with enzymes and extracted with NaCl. The lowest concentration was identified in samples which had not been treated to remove labile organic matter, independent of density separation solution. However, a certain difference in particle distribution could be expected, regardless of treatment or density separation solution, since it is not possible to obtain sediment samples that are perfectly identical.
Blank samples. From the blanks, only between 5 and 13 particles per sample were extracted, including only the categories semi-transparent microplastics, synthetic fibers, and other anthropogenic microparticles. In the blank samples extracted with ZnCl 2 , most anthropogenic particles belonged to the category of other anthropogenic microparticles, while in those extracted with NaCl, semi-transparent plastic particles dominated, followed by synthetic fibers and other anthropogenic microparticles. White microplastic particles were not extracted from any of the blank samples. The blank sample with the highest number of extracted microplastics was from the pre-treatment with enzymes and extracted with NaCl, where 13 particles were identified, and 10 of these were categorized as semi-transparent microplastics, subclass 3. All particles visually identified as anthropogenic in the blank samples are shown in Fig. 4.  www.nature.com/scientificreports/ Comparing the extracted particles from the blank with the particles in the sediment samples show a significant difference in plastic concentrations between the blanks and the sediment samples, p < 0.001 (2 sample T-test) (Fig. 4).

False negatives using visual identification.
There is a scientific concern regarding the visual identification of microplastic particles because of the risk of bias between operators and the risk of both false positive and false negative identifications 20,[33][34][35] . The particles remaining on the PA membranes after removal of all particles identified with stereomicroscopy as being composed of plastic, i.e., all particles visually considered to be non-anthropogenic, were analyzed with Raman microscopy. On the membranes from the pre-treatment with enzymes and extracted with NaCl, 3 particles on the 300 μm and 11 particles on the 100 μm were identified as plastics, representing 2% and 13%, respectively. On membranes from the pre-treatment with inorganic chemicals and extracted with ZnCl 2 , there were 5 particles remaining and later identified as plastic on each of the membranes, representing 1% of the total amount of particles. This shows that, at least in the present study, a skilled operator was able to identify plastic particles down to 100 μm with few false negatives, using visual and tactile identification techniques.
Effects of pre-treatment steps. The two pre-treatments (inorganic chemicals and enzymatic) that served to reduce the organic matters stickiness in the sample, thus breaking microplastics-matrix adhesion, was functionally evaluated with TOC analysis. The inorganic chemical pre-treatment had a TOC content of 1.55%, whereas the untreated sample had 2.89%. On the other hand, the enzymatic treatment showed a higher TOC level, 4.02%, most likely since enzymes add carbons to the sample. However, from the recovery rates, it is clear that for particles > 300 μm, the identified concentrations of particles from the different pre-treatment methods were similar to the untreated samples showing that if only > 300 μm particles are of interest, a pre-treatment is not necessary. However, for smaller lighter particles (PVC excluded) between 100 and 300 μm, a significant difference in concentrations between pre-treated and untreated samples was found (p = 0.00116). Moreover, there is a significant difference (p < 0.001) in plastic concentrations found in the sediment samples pre-treated with inorganic chemicals and extracted with ZnCl 2 , mainly due to this combination's high extraction of PVC particles.
Our results are in agreement with Enders et al. (2017), who tested KOH in combination with NaClO for digestion of fish stomach and showed that this treatment effectively digested the tissue. Moreover, they tested if the treatment affected the Raman spectra of 12 common polymers and only found a low peak alternation for acrylonitrile butadiene styrene (ABS) particles; however, they were still able to identify this polymer 30 . The enzymatic treatment has also been shown to effectively digest tissue 43 without harming common plastics 47 .

Density separation solution.
There was a higher extraction efficiency of plastic particles with ZnCl 2 than with NaCl. Heavier polymers such as PVC were recovered with ZnCl 2 (Fig. 4) but not with NaCl, as is expected based on the lower density of the NaCl saline solution. From a monitoring perspective, NaCl is an attractive alternative since it is non-toxic and has a lower cost than ZnCl 2 (technical grade ZnCl 2 is ~ 4 euros per kg). However, NaCl is not recommended when aiming to target particles with a higher density than 1.2 g/cm 3 , e.g., Polyethylene terephthalate (PET) or PVC. For a comprehensive polymer extraction, a higher density solution (> 1.5 g/cm 3 ) should be recommended since it can extract the majority of polymers also including PET and PVC. A density of at least 1.6 g/cm 3 for saline density separation solutions was recently recommended for Arctic microplastic monitoring programs 14 .

Conclusions
Higher concentrations of microplastic particles were identified in samples pre-treated with either inorganic chemicals or enzymes to remove labile organic matter, as compared to non-pre-treated samples, showing that a pre-treatment step improves the extraction of microplastics. However, if only particles > 300 μm are considered, the pre-treatment step was not necessary. Both density separation solutions, NaCl and ZnCl 2, successfully extracted lighter microplastic particles such as PE and PP. In addition, the combination of the inorganic pretreatment with ZnCl 2 successfully extracted PVC particles. From a monitoring perspective, our results are in agreement with the Arctic Monitoring and Assessment Program (AMAP) 14 , which recommends a density of at least 1.6 g/cm 3 to achieve a comprehensive polymer extraction which also includes PET and PVC particles.

Data availability
The datasets used and analyzed during the current study are available from the corresponding author on reasonable requests. www.nature.com/scientificreports/