Burial of microplastics in freshwater sediments facilitated by iron-organo flocs

Microplastics are ubiquitous in standing freshwater bodies, consequently lakes and reservoirs may be important sinks for these contaminants. However, the mechanisms governing the deposition of microplastics and their interactions with the sediments are understudied. We demonstrate how aggregation-based transport facilitates the sinking and infiltration of buoyant microplastics into freshwater reservoir sediments by employing experiments with intact sediment cores. Buoyant polyethylene microplastics were rapidly (1–4 h) incorporated into sinking iron-organic aggregates, followed by swift deposition into sediments. Ingression of microplastic bearing flocs into sediments was completed within 6 days and led to stable deposition of the incorporated particles for at least 2 months. Most microplastics were deposited in the top 2 cm of the sediments and few particles (5–15%) were re-released into the water. Our results show at least 85% burial of microplastics, indicating the significant role of freshwaters with low flow velocities in reducing microplastic loads to the oceans.

in which the hypolimnion is oxygenated 30 . Therefore iron floc stability and microplastic release might depend on the prevailing redox condition in the hypolimnion.
In this study, we explored how the aggregation of PE microplastics into sinking iron-organo flocs affects their long-term deposition in freshwater reservoir sediments, which is a crucial step for the complete understanding of this globally important microplastic sink. We hypothesized that iron-organo floc formation can be simulated in the laboratory and that the incorporation of PE microplastics into such aggregates leads to sinking of this buoyant polymer. In addition, we assumed that the size and shape of the particles govern their enclosure rate into iron flocs. This was tested by amending surface water from a eutrophic reservoir with the iron flocculent Fe(II) SO 4 and PE microplastics of three different shapes (fragments of four different size classes, fibers and spheres). Furthermore we tried to elucidate the fate of such microplastic bearing iron-organo flocs once reaching the bottom sediments of the water body. The first few mm of sediments and the overlying water column of lakes and reservoirs might be anoxic or oxic, depending on the season. With iron being a redox-sensitive element, we hypothesized that iron-organo-flocs containing PE microplastics lying on top of sediments will be stable under oxic conditions, while being disintegrated under anoxic conditions leading to microplastic release. This was tested with sediment cores from the eutrophic Bautzen reservoir (see Materials and methods section), which allow lab simulation of natural sediment processes due to their intact sediment surfaces. By addition of microplastic bearing iron-organo flocs to these sediment cores, we aimed to reconstruct the route of microplastics initially floating in the water column, into the sediments via an aggregation based transport mechanism.

Results
Iron-organo flocs formed by FeSO 4 oxidation and their characteristics. The addition of 100 or 300 µM FeSO 4 to filtered Bautzen reservoir water (Supplementary Table 1) led to the formation of large and sinking iron-containing flocs within < 1 h (300 µM Fe) to 3-4 h (100 µM Fe) ( Supplementary Fig. 1). The flocs formed by 100 µM Fe were generally fewer, smaller and had a lower density (Table 1) than the 300 µM Fe flocs. They were of reddish color emphasizing the high content of Fe (oxy)hydroxides ( Supplementary Fig. 1). The main component was water (> 90%), while dry mass consisted of similar ratios of organic to inorganic components. The 300 µM Fe flocs had a significantly lower content of organics compared to the 100 µM flocs (ANOVA, F value: 52.56, p < 0.05). The inorganic content of the flocs was dominated by Fe, the organics consisted primarily of extracellular polymeric substances (EPS) enclosing microbial cells and minerals (Fig. 1). The sticky EPS can be considered as a binding agent gluing the cells and iron minerals together, thereby shaping the flocs' gellike appearance (Fig. 1a). Most cells within the flocs were identified as bacteria using Confocal Laser Scanning Microscopy (CLSM), but also small numbers of eukaryotic algae and cyanobacteria were present (Fig. 1b).
The floc sizes showed high variability with means ranging from 2632 ± 1666 µm (median 2263 µm, n: 30) for spheres to 6270 ± 1666 µm (median 5650 µm, n: 30) for fibers (Fig. 2b). Flocs formed with fibers had a pronounced elongated shape and contained macroscopic structures of entangled fibers ( Supplementary Fig. 2). Beside this there were no clear differences in size between flocs with or without microplastics (Fig. 2b). By contrast, the presence of microplastics significantly reduced the sinking velocity of the flocs (non-parametric bootstrapping, 95% (CI), Fig. 2c). The mean sinking velocities of the flocs ranged from 0.006 ± 0.0018 m s −1 (median 0.0059 m s −1 , n: 30) for fragments 10-100 µm to 0.01 ± 0.0017 m s −1 (median 0.0101 m s −1 , n: 30) for no microplastics. However, given the high variability of the data, the absolute difference between the sinking velocities can be considered as minor, although being statistically significant.
Flocs formed by 100 µM Fe also aggregated PE spheres but showed lower precipitation of microplastics (28% of added spheres, mean, n: 3) compared to the 300 µM Fe flocs (99%, mean, n: 3). The lower aggregation efficiency was linked to the lower amount of flocs which were precipitated by this treatment (Supplementary  Table 2). Normalized to the total floc mass the aggregation efficiency was similar for the 100 µM Fe compared to the 300 µM Fe treatment (7.4 vs 7.3 spheres per mg floc, mean, n: 3).
To distinguish microplastics released or retained from small flocs generated at 100 µM Fe from that of large flocs formed at 300 µM Fe, initially buoyant spheres with different fluorescence labels were used (yellow: small flocs, red: large flocs). Experiments were started by adding iron flocs with PE to the overlying water of the sediment cores. Approximately 11,775 red spheres inside of 300 µM Fe flocs and 3231 yellow spheres inside of 100 µM Fe flocs were added to each core (Supplementary Table 2). The flocs settled through the water column and accumulated shortly at the sediment surface ( Supplementary Fig. 3). Then they continued sinking through the sediment surface. Flocs were completely buried in the sediments and no longer visible at the surface after 24 h in the anoxic and 6 days in the oxic treatments ( Supplementary Fig. 3). The anoxic cores showed extensive www.nature.com/scientificreports/ gas formation, which resulted in bubble release from the sediment. No obvious gas formation took place in the oxic cores, but bioturbation by burrowing chironomid larvae down to a depth of ~ 24 cm was observed. Iron reduction took place in the anoxic cores, while it was less pronounced in the oxic cores (Fig. 3). Microplastic release from the sediment was low throughout the whole experiment and not correlated to the iron release (Spearman's rank correlation, red spheres rho: 0.45, yellow spheres rho: -0.28; Fig. 3). At the end of the experiment, 85% of the recovered yellow spheres and > 95% of the red spheres were found within the sediment ( Fig. 4a and Supplementary Table 5 for absolute values). Hence the majority of microplastics were retained within the sediments and not released into the water phase during the experiment (Fig. 4a). Significantly more spheres (yellow and red; non-parametric bootstrapping, 95% CI) were recovered from the oxic compared to anoxic cores, indicating a sampling bias (Supplementary Table 5).
Most spheres were found within the uppermost 2 cm of the sediments in October cores (Fig. 4b). However, spheres were also detected in deeper layers. In the anoxic cores low numbers of spheres were present in the 2-5 cm layer. Their abundance decreased sharply with depth with only very few spheres recovered from 5 to 8 cm depth and no spheres recovered from the 8-11 cm layer of anoxic cores. In the oxic cores, spheres were found even in the deepest layer of 8 -11 cm depth, indicating a deeper burial compared to the anoxic cores. Comparatively high numbers of spheres were recovered from the 2-5 cm and 5-8 cm layer of the oxic cores. Similar patterns were also observed for July cores ( Supplementary Fig. 4). Chironomid burrows were found throughout all layers of the oxic cores, averaging on 2-5 visible burrows per layer ( Supplementary Fig. 5).   www.nature.com/scientificreports/

Discussion
In this study we have presented evidence that PE microplastics are aggregated into sinking flocs formed by iron precipitates and organic material, irrespective of their shape. These flocs rapidly transported initially buoyant PE micro-spheres deep into freshwater sediments, leading to stable deposition given the long incubation time of 63 days. The addition of ferrous sulfate to surface water of Bautzen Reservoir induced the formation of sinking flocs. This iron flocculation is a well-known process described in nature 23 or employed in the context of water treatment technologies 32 . Flocs formed in our study were comparatively large, ranging from approximately 500 to > 3000 µm, whereas iron-organo lake aggregates or riverine composite suspended sediment particles 33 are typically smaller with sizes from 60 µm 16 to 457 µm 34 . However, organic and iron contents were comparable  www.nature.com/scientificreports/ to natural flocs 16,35 . Iron-floc formation is usually considered a complex process involving the interaction of different microbial consortia forming complex 3-dimensional structures of EPS 36 which enclose biogenic iron mineral grains and bacterial cells 28 . However, floc formation might also occur rapidly during seasonal lake mixing events without the involvement of complex microbial consortia 21 . This mode of formation might be more similar to the water treatment iron coagulation processes 24 and to the experimental procedure we used. Even though we attempted to mimic natural conditions by using reservoir water, natural pH and partly natural iron concentrations 28 , the formed flocs will differ from natural ones. This should be considered when discussing the potential impact of iron flocculation in lakes on the fate of microplastics.
In addition, we could show that the iron-organo flocs readily incorporated buoyant microplastics. The incorporation depended on the size of the plastic particles. This is in line with other studies showing an increasing aggregation potential with decreasing microplastic particle size 37 , which might be explained by the higher collision frequency of smaller particles 38 . The overall aggregation efficiency of 300 µM Fe was higher for spheres compared to fragments or fibers, which might be explained by differences in size or polymer density. In another study iron concentrations of 370 µM removed up to 90% of small polystyrene microplastics from wastewatereffluents 39 , emphasizing the high potential of iron-organo flocs to trap microplastic.
The iron-organo flocs investigated in our study can be considered as a relevant type of aggregate capable of incorporating and sinking buoyant microplastics in freshwater. Previous studies showed that iron flocs formed during lake mixing could precipitate buoyant cyanobacteria colonies (size: 63 µm) 21 and large PE microplastics (size: 4 × 4 × 0.15 mm) 10 . Although not proven in field experiments, it is likely that comparable aggregates forming in freshwater systems such as stratified lakes, reservoirs or flow-calmed zones of rivers are able to precipitate buoyant PE microplastics 10 and thereby transport microplastics to the sediments or channel beds 40 . This is in line with many studies showing the removal of microplastics from the water column via aggregation mechanisms 41 . Iron-organic flocs are similar to other microplastic bearing aggregates such as marine 42 or lake snow 15 , phytoplankton-43 , EPS-44 , and TEP-8 based flocs or riverine composite suspended sediment particles 33,34 regarding their sizes, sinking velocities and densities.
Using sediment core experiments we could show that iron-organic aggregates containing initially buoyant microplastics rapidly subside into sediments and are not re-mobilized within 63 days. Hence we conclude that aggregation followed by sedimentation onto muddy sediments might lead to relatively stable deposition of microplastics in stagnant water bodies. Given that comparable flocs can also be found in fluvial environments, this deposition process is likely to be relevant for flow-calmed zones and floodplains of rivers or even estuaries.
Contrary to our initial hypothesis, anoxic conditions and iron reduction did not lead to a significant remobilization of microplastics from the sediments. Although more microplastics were released from anoxic compared to oxic cores, the majority of PE spheres remained in the sediments. This might be explained by the rapid downward transport of the iron flocs from the sediment surface into the sediments, by which microplastics were most likely trapped within the sediment matrix. Hence unlike iron oxide bound solutes such as pollutants or nutrients 25 , microplastics are not re-mobilized by anoxic conditions. Sediments from the deepest part (~ 11 m) of Bautzen reservoir were used for our study. They were finegrained, organic rich and rather unconsolidated. These are common properties of muddy sediments found in low current zones such as profundal zones of lakes 45 and reservoirs 46 , oceanic basins (Baltic Sea 47 ) or river estuaries. Recent studies showed that microplastics are often deposited in such low flow zones, making muddy sediments a likely depository of these anthropogenic particles 48 . In comparable deep-sea sediments from the Rockall Trough, microplastics were found in undisturbed layers with an approximate age of > 150 years 49 . This can only be explained by re-distribution of microplastics in the sediments after their deposition. Our results showed that aggregates containing microplastics can indeed easily penetrate the first cm of sediments. This offers a possible explanation for the unexpected appearance of microplastics in sediment layers deposited before the industrial production of the detected polymer types, which cannot be explained by bioturbation alone. The penetration of large particles through muddy sediments has been attributed to their gravitational force 50 overcoming the cohesive force of the sediment particles 51 . The cohesiveness and density of sediments increase with depth, which will stop the downward movement of the flocs. It has been reported that large flocs formed by aluminum flocculation will accumulate to a depth of 10 cm in muddy lake sediments 52 . However in our study most PE spheres, which can be considered as proxy for the flocs, accumulated within the first 2 cm of the sediment. This is in line with findings indicating that phytoplankton aggregates accumulate and degrade within the upper few cm of lake sediments 53 . Still, we showed that a minor fraction has been transported deeper into the sediment reaching a depth of at least 11 cm. Interestingly, the smaller 100 µM Fe flocs reached deeper layers than the larger 300 µM Fe flocs, which is contradictory to the settling model driven by gravitational force. This might be explained by the higher potential of small flocs to migrate through small channels or cavities 54 . In conclusion, our study is in line with previous reports of microplastics 49 or metal-organo flocs' subsidence into sediments 52,55 . We showed that the rapid transport of buoyant PE into freshwater sediments can be facilitated by low density iron-rich organic aggregates. As the fate of aggregates strongly depends on the properties of the underlying sediment, the wider significance of our results is limited. Still, our findings should be applicable to most reservoirs and lakes because the type of sediment we investigated is broadly typical of these environments. More studies are needed to determine the response of other sediment types.
There is some evidence that PE spheres were transported deeper into the sediments of oxic cores compared to anoxic cores. This might have been caused by the presence of chironomid larvae in the oxic cores building long burrows ( Supplementary Fig. 5). It has been previously reported that bioturbation by invertebrates will transport microplastics in fine-grained muddy sediments down-rather than upwards 47,56 . Therefore the burrowing and biodiffusive 57 activity of the chironomids might explain the deeper distribution of spheres in the oxic cores. Our experimental setup excluded bioturbation by larger animals such as macroinvertebrates and benthivorous fish www.nature.com/scientificreports/ which exert complex and diverse modes of re-mobilization 58 . Based on our results the possible effect of bioturbation on the microplastic distribution in Bautzen reservoir cannot be evaluated completely. Different limitations of the sediment core experiments need to be mentioned. Firstly, lower numbers of spheres were recovered from the anoxic compared to the oxic cores. This is particularly striking for the yellow spheres bound to 100 µM Fe flocs. It cannot be ruled out that the spheres or flocs were transported even deeper than the 11 cm used as lowest boundary. Extensive gas bubbles formed within the anoxic cores, which produced large voids in the deeper layers of the sediment. PE spheres might have fallen through these voids which brought them deeper into the cores than the sampled 11 cm layer. Considering that the missing spheres were not liberated from the sediments, but rather incorporated deeper than expected, this does not affect most of our statements. However, the assumption that spheres were transported deeper into the sediments of the oxic cores compared to the anoxic might not be justified.
Flocculation processes might contribute to the retention of microplastics transported from land to sea, as indicated by the presented results. Once initially buoyant microplastic is incorporated into sinking aggregates and reaches the sediment, it will rapidly be deposited inside the sediment matrix. Given the similiarties of ironorgano flocs compared to other lacustrine floc types or riverine composite suspended sediment particles, this process might play a role for the transport of microplastics in various freshwater environments. The observed accumulation of initially buoyant microplastics in fine sediments of riverbeds might also be influenced by their incorporation into larger and sinking aggregates. Excluding other processes such as current driven sediment re-suspension or bioturbation by larger organisms, this deposition might be stable even for longer time periods. This may not be valid for riverine environments were flooding events will re-suspend and transport microplastic particles downstream 59 . In lakes, deposition of microplastics during summer stratification could lead to permanent deposition, as no bioturbation takes place under anoxia, while several mm of sediments are settling over the microplastics during this time. Comparable conditions have been described leading to excellent fossilization of organic tissue or carcasses in stratified lakes 4 , for which undisturbed deposition over geological time-scale is required. This indicates that permanent and undisturbed deposition of microplastics in freshwater sediments is possible under certain conditions. However, lakes and reservoirs are typically located upstream of large urban centres which act as main source for environmental microplastics. This might diminish the revervoirs' potential to act as important sink for these particulate contaminants. Still, the findings might draw more attention to the role of aggregation processes in reducing the plastic loads of aquatic systems in general and of rivers in particular. Comparable mechanisms might be relevant in the estuaries of large rivers with their flocculation zones and muddy sediments. Furthermore, factors leading either to re-suspension or to permanent burial of aggregated microplastics deposited in sediments should be investigated in further studies to improve the understanding of microplastic fate in the environment.

Material and methods
Study site and sampling. Bautzen Reservoir is located in Germany (size: 5.3 km 2 , mean depth: 7.4 m 60 ) and shows labile summer stratification (June-September), with an anoxic hypolimnion frequently disrupted by strong winds 61 . Sediment cores and surface water samples were taken at the deepest point (~ 11 m) of the reservoir. Sediment cores were retrieved on 30th of July and 19th of October 2020 using a gravity corer (UWITEC, Austria, and PVC liners of 60 × 9 cm). Profiles of water parameters (Supplementary Fig. 6) were recorded using a multiparameter probe (Sea & Sun Technologies, Germany).
Floc formation and investigation of microplastic aggregation potential. Bautzen reservoir surface water was stored at 20° C in the dark and used within 2 weeks after sampling. Prior to use, the water was filtered through 10 µm stainless steel sieves and the filtrate was adjusted to pH 9.5 by 1 M NaOH (1 M) to reflect the alkaline conditions of the surface water during summer stratification ( Supplementary Fig. 6). Experiments were conducted in triplicates per PE shape by amending 500 mL filtrate with 300 µL or 100 µL (final concentration: 300 or 100 µM) of a FeSO 4 × 7 H 2 O stock solution (500 mM, pH 1.8) in airtight 1-L bottles. Thereafter 20 mg L −1 of microplastic fibers (9.0 × 10 3 particles L −1 ), spheres (2.4 × 10 4 particles L −1 ) or fragments (either fragments 10-100 µm: 6.6 × 10 4 particles L −1 , fragments 100-250 µm: 3.2 × 10 3 particles L −1 , fragments 250-500 µm: 2.2 × 10 2 particles L −1 or fragments > 500 µm: 1.1 × 10 2 particles L −1 ) were added separately to the respective bottles. Additionally, triplicates without added microplastics were used as control. The bottles were incubated on tumbling roller incubators (3)(4)  www.nature.com/scientificreports/ The resulting suspension was filtered onto stainless steel filters and examined for their total plastic content by light microscopy. Sinking velocities of flocs formed by 100 µM Fe could not be assessed due to their small size. Therefore solely total plastic content and average floc size (n: 30) were recorded.
Floc characterization. The properties of iron flocs produced by 100 µM and 300 µM Fe without additional microplastics were further characterized by different methods. Densities of six individual flocs were determined by titration with NaCl solution (20% m/v) until neutral buoyancy, followed by pycnometer measurement of the resulting solution at 20 °C 12 . Water content, dry mass (60 °C, 24 h), loss on ignition (550 °C, 24 h) and total mass of flocs (n: 6) were measured after centrifugation (3000 rpm, 20 min) in pre-weighted conical centrifuge tubes. Fe content of the flocs' wet mass was determined by dissolving defined, centrifuged (3000 rpm, 20 min) fractions (n: 6) in hydroxylamine hydrochloride-HCl (0.5/1 M) followed by measurement via ferrozine assay 63 . Confocal laser scanning microscopy (CLSM) was used to examine 5 randomly chosen spots each on 10 individual flocs produced with 300 µM Fe. The biovolumes of algae, bacteria, cyanobacteria and EPS were calculated from the resulting CLSM imaging datasets (n: 50).
Experimental set-up of sediment incubation experiment. Cores were grouped into oxic and anoxic treatment with 3 replicates each. The anoxic triplicates were bubbled with N 2 until depletion of O 2 as measured via an internal oxygen optode (Pyroscience, Germany). Afterwards the anoxic cores were closed with custommade covers preventing intrusion of oxygen and allowing anoxic sampling 64 . The oxic triplicates were closed with the same covers and bubbled constantly by air. The O 2 levels in the cores were permanently recorded and adjusted on daily basis by N 2 or O 2 bubbling if necessary. Flocs formed by 300 µM Fe and containing red-fluorescent Rhodamine B PE spheres were produced as already described. In addition, for the October experiment, flocs formed by 100 µM Fe aggregating yellow-fluorescent PE spheres were prepared. The flocs were gently washed with tap water (three times) to remove non-aggregated microplastic spheres. The October cores were first supplemented with flocs containing red spheres and afterwards with flocs containing yellow spheres. The overlying water was exchanged by bottom to top through-flow of approximately 2 L of Bautzen surface water to remove spheres released by physical breakage of the flocs. Cores were photographed in 24 h intervals in the first week after floc application and in weekly intervals thereafter.
Sampling procedure of sediment incubation experiment. Water samples for Fe(II), Mn, pH and released microplastics were taken in weekly intervals using syringes. Fe(II) was measured using ferrozine assay 63 , while Mn was measured using formaldoxime 65 . Microplastics were sampled by removing 120 mL of the uppermost part of the water-column using a syringe. The water was filtered over stainless steel filters (10 µm) and retained spheres were counted under a light microscope. The experiments were run for 63 days at 16 °C in the dark, after which the remaining water column was removed and examined for their plastic content. Afterwards the cores were sliced into sections of 0-2 cm, 2-5 cm, 5-8 cm and 8-11 cm using a sediment core cutter (Uwitec, Austria). The sections were transferred into centrifugation tubes and extracted by sonication followed by density separation with NaCl (20%, ρ: 1.56 g cm −3 ). The resulting suspension was centrifuged (3000 rpm, 15 min) and frozen by immersion in dry ice. The top layer of the frozen solution was transferred into conical centrifuge tubes by flushing with water and then filtered onto stainless steel grids. Microplastic contents of the layers were counted under the light microscope. The recovery rate was assessed by the addition of 1.2 × 10 4 fluorescent PE spheres to sediment sections (0-2 cm, 2-5 cm, 5-8 cm, 8-11 cm) of a control core, followed by the already described extraction method. 1.06 × 10 4 ± 3100 particles (mean ± sd, n: 4) were recovered from the sediments leading to recovery rate of 89.51 ± 2.61% (mean ± sd) for this method (data not shown). One anoxic core of the July experiment was lost due to inappropriate handling, resulting in lower sample numbers for the anoxic treatments. The bulk density, porosity, water content, dry mass, organic content and grain size distributions of two control cores were determined by standard methods 60 . Furthermore the Fe(II)/Fe(III) and Mn contents of the sediments were determined after extraction with HCl (1 M) and hydroxylamine hydrochloride-HCl (0.5/1 M) using ferrozine or formaldoxime assay, respectively. CLSM imaging. Flocs were visualized using CLSM in combination with different fluorescent dyes 12 . In brief, flocs were mounted in microscope chamber slides (Thermo Fisher Scientific) and stained. Aleuria aurantia lectin (Vector Laboratories, USA) labeled with Alexa Fluor 633 (Thermo Fisher Scientific, USA) was used to visualize the extracellular polymeric substances (EPS) of the flocs 66,67 . Bacteria were detected via SybrGreen staining, while algae and cyanobacteria were identified by the autofluorescence of their chlorophyll a or phycobilins, respectively. Imaging was done by a TCS SP5X upright microscope equipped with white laser and waterimmersible lens (25x/0.95), controlled by LAS AF version 2.4.1 (Leica, Germany). The filter configurations used for excitation and emission are listed in Supplementary Table 7. Imaris (Bitplane) was used to visualize the images, which were printed by Photoshop (Adobe). Biovolumes of algae, bacteria, cyanobacteria and EPS were semi-quantitatively calculated employing an adaption of ImageJ 68 .

Statistical analysis.
Statistical testing was only conducted for datasets with a minimum sample size of 5 individual replicates. Q-Q plots were used to check for data normality. Bartlett's test was used to test variance homogeneity prior to one-sided ANOVA (Type II) which was used to compare group means. Residual plots were examined to verify the reliability of the ANOVA. Group means were assumed to be significantly different from each other for p < 0.05. Tukey's post-hoc test was computed for pair-wise comparison. Samples not meeting assumptions of the ANOVA were tested by non-parametric bootstrapping 69  www.nature.com/scientificreports/ of median differences of 10,000 bootstrapped samples were reported. Differences in median CI higher or lower than zero were defined as significantly different from each other by 95% chance. Spearman's rank correlation was used to calculate correlation coefficients. Software R 70 was used for all statistical analysis and data visualizations.