Title: Evaluation of existing methods to extract microplastics from bivalve tissue: adapted KOH digestion protocol improves filtration at single-digit pore size

Methods standardisation in microplastics research is needed. Apart from reagent-dependent effects on microplastics, varying target particle sizes can hinder result comparison between studies. Human health concerns warrant recovery of small microplastics. We compared existing techniques using Hydrogen peroxide, Proteinase-K, Trypsin and Potassium hydroxide to digest bivalve tissue. Filterability, digestion efficacy, recoverability of microplastics and subsequent polymer identification using Raman spectroscopy and a matching software were assessed. Only KOH allowed filtration at ≤ 25 µm. When adding a neutralisation step prior to filtration, KOH digestates were filterable using 1.2-µm filters. Digestion efficacies were >95.0 % for oysters, but lower for clams. KOH destroyed rayon at 60° C but not at 40° C. Acrylic fibre identification was affected due to changes in Raman spectra peaks. Despite those effects, we recommend KOH as the most viable extraction method for exposure risk studies, due to microplastics recovery from bivalve tissues of single-digit micrometre size. Graphical abstract


Introduction
Research into microplastic pollution is a relatively recent topic. Microplastics are generally defined as particles of 0.1 to 5,000 µm (Alexander et al., 2016;Arthur et al., 2009) and are a diverse group of particles in terms of shape and type (Hidalgo-Ruz et al., 2012). Plastics is an umbrella term for synthetic and semi-synthetic materials, such as polypropylene, polyethylene and acrylic. Small plastics emerged as a contaminant in the marine environment in the early 1970s (Buchanan, 1971) and can be present in the environment as fragments, microfibres, microsheets and sphere-like particles (Hidalgo-Ruz et al., 2012). Thompson et al. (2004) coined the term 'microplastics' in an analysis of microplastic concentrations in the marine environment. Since then, microplastics have become a globally recognised contaminant (Andrady, 2011) and research output has increased greatly. With this, a variety of methodologies has been used to extract microplastic particles from marine-based samples. These methodologies vary according to the matrix of interest, such as water, There is a need to standardise methodologies to obtain comparable results, if this type of work is to be policy-relevant.
One matrix of interest is marine organisms. Demand for fish and shellfish is constantly rising; global per capita consumption was at 19.7 kg in 2013 (FAO, 2016), making microplastics in seafood a potential concern for human health. Within marine fisheries, bivalves deserve a special focus due to their value but also their exposure risk. Almost 13.6 million tonnes of bivalves were cultured globally in 2014 (FAO, 2016). As filter-feeders, many bivalves are at a greater exposure risk to microplastics (Wright et al., 2013). Mean concentrations of microplastics in such organisms have been reported as 2 particles per gram of the wet weight of the organism (g -1 ) (e.g.

Sampling and sample preparation 36
Magallana gigas (n =-33, tissue wet weight (w.w.) 14.31 ±5.84 g), Ostrea edulis (n =-12, 7.15 ±4.63 g 37 w.w.) and Ruditapes philippinarum (n =-4, 2.46 ±0.14 g w.w.) were collected from coastal locations in 38 southern England and stored at -20° C (Oaten et al., 2015). Defrosted soft tissues were removed 39 from shells, placed in glass containers and re-frozen to -20° C. Two digestion methods were excluded from further experiments: H 2 O 2 due to excessive foaming 99 and therefore possible sample losses and Proteinase-K due to its cost. The trypsin protocol could not 100 be optimised to 25-µm filtration. However, its efficacy was assessed on M. gigas, filtering over 63-101 µm mesh. This was done because of the negligible damage of enzymes to microplastics in the 102 digestion process (see references in Table 1). Tissue digestions were performed in triplicate. Reagent 103 controls were run alongside to establish effects on filters. The KOH protocol was applied to M. gigas, 104 O. edulis and R. philippinarum. Digestates were filtered over pre-weighed filters ( Table 2), rinsed 105 with ultrapure water, dried overnight at 60° C and re-weighed at room temperature. Microplastics were created from ten post-consumer items (Table SI.3). Fragments were produced 115 with an electrical coffee bean grinder and dry-sieved through 63 and 600 µm stainless steel sieves. 116 Particles <600 µm were retained. Fibres were obtained by plugging and cutting. Film/sheet was also 117 cut. Thirty particles of each material were added to M. gigas tissue. Containers used for storing 118 those particles prior to spiking were subsequently inspected for potentially left behind items. Spiked 119 samples were re-frozen and treated like efficacy experiment specimens. Since small pore size for 120 filtration is paramount when assessing human exposure rates (Wright and Kelly, 2017), only the 123 Handling controls were prepared by dosing ultrapure water (60 ml, n = 4) and kept at room 124 temperature for 48 hours. This was done to assess potential losses unrelated to the reagent, 125 temperature or interaction with biological tissue. An additional water blank was run after filtering 126 the samples to assess if any particles from previous samples could have contaminated subsequent 127 ones. Samples were vacuum filtered (25-µm filters) and rinsed with 20 ml of ultrapure water. This 128 amount was chosen because initial trials with tissue samples indicated that larger amounts of water 129 were not filterable. During such trials, particles were observed stuck to the sample containers and 130 filtration funnels. To enumerate potential losses, after sample filtration a new filter was placed in 131 the unit and the sampling jars and funnel flushed with approximately 400 ml of ultrapure water. 132 Filters were placed in lidded petri dishes and dried at 60° C overnight. Lids were secured with tape. 133 Particle enumeration was conducted with a microscope (magnification 10x -60x). See 134 supplementary information for additional quantification of losses. 135 136 Rayon was affected by 10 % KOH at 60° C during the dosing experiment. Therefore, the effects of 137 further treatments on rayon were investigated. Ten fibres (n = 4) were treated as follows: 5 % KOH, 138 10 % KOH (both incubated at 40° C for 48 hours) and Trypsin (according to the optimised protocol). 139 Samples were filtered over 11-µm and fibres counted. 140 141

Effect of KOH on Raman spectra 142
To establish the effect of KOH on the Raman spectra, microplastics from the following treatments 143 were compared: untreated, exposed to ultrapure water at room temperature and to 10 % KOH at 144 60°C for 48 hours. Further spectra were obtained from rayon and acrylic fibres exposed to 10 % KOH 145 at 40°C for 48 hours. Particles were transferred from filters onto quartz slides. Raman spectroscopy 146 was performed using a 785 nm Renishaw inVia with Leica DM 2500 M microscope, 50x magnification 147 lens. Particles were manually selected. Extended spectra were obtained at 1-5 % laser power, 10 148 seconds exposure time and five accumulations. WiRE 4.1 software was used to process the data. 149 BioRad KnowItAll was used to identify Raman spectra by individual and multi-components as well as 150 agreement of spectral peaks. 151

Statistical analysis 153
Data are reported as means with one standard deviation. Statistical analyses were conducted with 154 RStudio 1.0.153 using an alpha value of 0.05. Independence of data was assumed. Normality was 155 assessed with Shapiro-Wilk tests and homogeneity of variance with Levene's tests. Failing those 156 tests, filter weight changes were assessed with Kruskal-Wallis tests. Recovery rates after KOH 157 digestion were assessed with a two-way ANOVA (factors: polymer type and microplastic category, 158 i.e. fragment, fibre or microsheet). A non-parametric χ-squared test was performed to establish if 159 recovery rates were significantly different from 100 %. 160 161

162
Digestion efficacy and competitiveness. Treatments with H 2 O 2 and Proteinase-K were abandoned 163 due to excessive foaming and subsequent sample loss, and high cost respectively. Digestion efficacy 164 for oysters treated with Trypsin and KOH were ≥95.0 % ( Table 2). Trypsin digestates were only 165 filterable using 63-µm mesh. KOH digestates were filterable with 25-µm and smaller filters. The 166 lowest variability was achieved using 25-µm filter paper. Digestion efficacy of Ostrea edulis was 167 comparable to Magallana gigas at 5-µm pore size: 97.0 ±2.7 % and 96.4 ±4.0 % respectively. The 168 slightly lower efficacy and increased variation for M. gigas tissue is likely to be due to the presence 169 of small pearls. Further optimisation by neutralising the KOH digestate allowed for filtration over 170 1.2-µm borosilicate filters (see Figure SI.1 for results on effects of KOH on filter papers). Reduced 171 incubation temperature to 40° C did not affect the digestion efficacy but decreased variability for 172 digestates of M. gigas tissue ( Table 2). Digestion efficacy for the clam species Ruditapes 173 philippinarum was 91.2 ±0.5 % at 1.2 µm pore size (  (Table SI.2). Overall, trypsin did score once, the H 2 O 2 method twice and the other methods 187 five to 13 times in the health hazard class. 188 189 Recovery rates. The recovery rates (RR) of 10 types of microplastic (Table SI.

206
Note that rayon was recovered when exposed to 10 % at 40° C (see paragraph 3 in The effect of KOH and Trypsin on rayon fibres at 40° C was assessed. This reduction in incubation 222 temperature from 60° led to a recovery of rayon similar to the other microplastic types (Figure 1). 223 Treating rayon fibres (10 fibres per treatment, n = 4) with 5 % KOH for 48 hours led to a recovery 224 rate of 87.5 ±37.7 %. Recovery was 80.0 ±8.2 % for exposure to 10 % KOH for 48 hours. Exposing 225 rayon to trypsin (4 hours, without using a magnetic stirrer) led to a recovery of 82.5 ±23.6 %. The 226 increased variation for the trypsin exposure can be traced to a recovery of 50 % for one of the 227 samples, the remaining samples yielded 80 -100 % recovery. There is no significant difference in 228 recovery between treatments (Kruskal-Wallis, H(2) = 0.128, p = 0.938). 229 230 Comparison of Raman spectra of microplastics after KOH exposure. Raman spectra were obtained 231 from the microplastics used in this study prior to exposure and after 48-hour exposure to ultrapure 232 water at room temperature and to 10 % KOH at 60° C ( Figure 2). In general, fluorescence and 233 subsequent baseline adjustments hindered comparison below a Raman shift of 700 cm -1 . Spectra of 234 polypropylene fibres (A) and fragments (B) exhibited a decreased intensity at peak locations of 840 235 and 1150 cm -1 after exposure to water and KOH. At 997 cm -1 , peak intensity increased with water 236 and KOH exposure for fibres but was diminished for those treatments in PP fragments. No change in 237 number of peaks or peak intensity was observed for PS fragments (C) across treatments. Changes 238 were observed in spectra of acrylic fibres (D). Fluorescence hindered comparison of spectra between 239 treatments, especially <1500 cm -1 . The number of peaks and their intensities varied between 240 treatments. The peak at 1000 cm -1 , was most pronounced for untreated acrylic, reduced for the 241 water-exposed and absent for the KOH-exposed fibres. Peaks observed at 1603 cm -3 in untreated 242 and KOH-exposed acrylic shifted slightly to 1590 cm -1 . Furthermore, a peak observed at 2240 cm -1 for 243 KOH-exposed fibres was not present in water-exposed and untreated acrylic. PET fibres were not 244 evaluated due to issues with fluorescence and the small fibre diameter. Spectral acquisition of PET 245 fragments was also hampered by fluorescence, but the number of peaks and their intensity was 246 similar across treatments (E). A similar number of peaks was observed for LDPE microsheets across 247 treatments (F). Peak intensities were reduced in untreated compared to water and KOH-exposed 248 LDPE between 1250-1480 and 2830-2900 cm -1 . Conversely, at 1094 cm -1 , the peak was most 249 pronounced for untreated LDPE. The intensity of peaks for PA fibre were generally similar (G) with 250 the exception of the peak at 994 cm -1 . Here, a peak was observed for the untreated fibre, which was 251 reduced for the KOH-treated and absent for the water-treated fibre. PVC sheet (H) only exhibited a difference at 1003 cm -1 ; a peak was observed in the untreated but not in the two treated 253 microsheets. The intensity and location of peaks for rayon fibres (I) were generally similar. untreated, blue (middle) exposed to ultrapure water at room temperature for 48 hours 262 and red (bottom) exposed to 10 % KOH at 60° C for 48 hours, except for rayon fibre 263 where top fibres were exposed to 10 % KOH at 40° C instead.

265
Using a matching software to compare microplastics of unknown composition to a spectral library, 266 most particles exposed to KOH were correctly identified ( Table 3). Acrylic could circumstantially be 267 identified as an anthropogenic fibre due to the identification of dye. A number of untreated and 268 water-exposed microplastics were misidentified by the software. Rayon was only identified when 269 exposed to KOH at 40° C. 270 271   Table 3 -Results of using a matching software to identify polymer types using Raman spectra.

272
Spectral peak locations as well as individual and multi-component results were taken into 273 account. "Likely to be" signifies that a choice had to be made by the researcher.

Microplastic
Untreated Water-exposed KOH-exposed LDPE microfilm Likely to be rayon

277
A number of studies already set out to compare different digestion methods (Table 1) 80-µm mesh. Depending on the research aim, such minimum particle size may be suitable for some 309 studies, especially due to its relatively low cost (Table SI. LDPE. The digestion temperature needs to be carefully chosen. At 60° C, rayon was not recovered. It 412 seems that rayon was dissolved by the alkaline reagent, similar to the process of producing rayon 413 where cellulose is dissolved in NaOH and fibres subsequently recovered from the solution (Jarvis, 414 2003). However, decreasing the incubation temperature to 40° C allows for recovery of rayon fibres 415 in the same range as the other tested microplastic polymers. In line with previous research (Table 1) degraded. This degradation was observed at 60° C. An incubation temperature of 40° C using 10 % 420 KOH is therefore recommended for the recovery of all microplastics types assessed here. 421 The degree of polymer damage through KOH exposure seems acceptable as most obtained Raman 423 spectra were correctly identified by a matching to a spectral database. Acrylic fibre was only 424 identified circumstantially by the presence of the dye 'Eriochrom Blue' that is used in the textile 425 industry (Ding and Freeman, 2017). Acrylic fibre may be affected on a molecular level as revealed by 426 Raman spectroscopy. It appears that acrylic fibres are easily hydrolysed by alkalis. Gupta et al. temperature is important and effects on semi-synthetic polymers (e.g. rayon) can be minimised by 449 choosing an incubation temperature of 40° C or lower. The lower temperature also delivered more 450 consistent results. The use of a matching software to assess the Raman spectra of the microplastics 451 used in this study showed that all types were identifiable after KOH exposure. KOH is also the most 452 economical and least time-consuming method. Its hazardousness to human health is similar to or 453 less than other reagents used for tissue digestions. Since KOH has also successfully been used for other tissues, we suggest that this method is adapted as a standard method for biological tissue 455 digestions to extract microplastics. Such a harmonised approach will allow for comparability 456 between studies and make such results more policy-relevant. o Section on 'Quantification of microplastic particle losses during the spiking experiment'. Digestate was not filterable with vacuum filtration.

Costs, time needed and hazardousness of reagents
Costs of digestion per sample were based on a mean oyster tissue volume of 20 ml (11.8 ±6.4 g w.w.). When optimisation was not possible, calculations were performed with the most promising protocol (e.g. observation of highest digestion rates). Prices were taken of the website from Fisher Scientific, or Sigma Aldrich if not available at the former, on 29 th December 2017. The cost of ultrapure water-which will be laboratory-dependant -was estimated at £0.232/litre based on the following estimates: • DI pack for ultrapure -£556.00 (replaced approximately every 9 months on average) • UV lamp for ultrapure -£141.00 (replaced approximately every 18 months) • Delivery speed of pump -2L/min • Estimated daily usage -10 L • Average month: 30 days • Consumables that provide the water in the tank to get it to a certain grade to feed the ultrapure water are not included. Effect on filter papers. Initial trials revealed that KOH affected the filtering capacity of filter papers, e.g. the longer the digestate took to filter the more difficult it was to flush the filter with pure water afterwards. This effect was tested by subjecting pre-weighed cellulose (Whatman Grade 4), cellulose nitrate (5.0 µm Whatman) and borosilicate (Whatman GF/C) filters to three different neutralisation treatments in triplicate. Samples of 25 ml of 10 % KOH solution were heated to 60° C. The following treatments were used: 1) no neutralisation, 2) part-neutralisation with 15 ml of 1 M citric acid and 3) complete neutralisation with 22 ml of 1 M citric acid. Mixtures were transferred immediately to the vacuum filtration unit. To mimic slow filtration, the pump was not switched on for 90 seconds, during which time part of the mixture passed through the filter gravitationally. The remainder (c.10 ml) was filtered using the vacuum. Filters were rinsed with 20 ml ultrapure water, transferred to petri dishes, dried at 60° C overnight and re-weighed at room temperature. The pH of the each mixture was established using paper indicator strips. The relative weight change of the filters was used as a measure of effect (Equation 4, supplement information).
The effect of the solution on the filter papers can be seen in Figure 3   Incorrectly counted out microplastics in preparation for the spiking experiment. Prior to spiking, known amounts of microplastics were placed into glass petri dishes. Spot checks of closed dishes showed that quantities were often below 30 particles per type of microplastic. In four containers, between six and all ten reference material types were recounted. All fragments and microsheets were counted. Estimation of number of fibres proved difficult when fibres were present in agglomerations. Dishes were not opened during this process to prevent additional losses through particles jumping or sticking to tools. In no case were there more than 30 particles. The minimum number of fragments was 24 for PP, 27 for LDPE microsheet and 28 PP fibres. Based on these findings, it can be assumed that a mean of 28.75 ±1.91 fragments, 29.00 ±0.76 fibres and 29.25 ±1.04 microsheets was prepared for each sample. Since only a fraction of prepared reference materials was recounted, results were not adjusted for those estimated losses. Otherwise, recovery rates would have increased between 2.1 and 3.9 %, except for rayon.
Lack of transfer of counted out reference materials from their storage container to the sample.
Counted out microplastics were transferred from petri dishes to samples. Petri dishes were sealed with tape and inspected for non-transferred particles. In 16 out of 21 containers some particles remained. Mean particle content in those 16 was 5.4 ±6.1. The maximum number of particles that remained in one dish was 21. Out of the microplastics created, PS and PET fragments, PA fibres and LDPE sheet all transferred to the samples. Between two and five particles were not transferred for PP fragments, PP, rayon and acrylic fibres respectively. Forty-four pieces of PVC microsheet and 29 PET fibres did not transfer overall. The results shown in Figure 1 (main text) were adjusted for those losses.

Microplastics recovered through additional rinsing
As mentioned in the results section, Figure 1 was adjusted for microplastics that were only recovered through additional rinsing. Those additional recovery rates can be found in Table SI.4.