Biofilms on plastic litter in an urban river: Community composition and activity vary by substrate type

In aquatic ecosystems, plastic litter is a substrate for biofilms. Biofilms on plastic and natural surfaces share similar composition and activity, with some differences due to factors such as porosity. In freshwaters, most studies have examined biofilms on benthic substrates, while little research has compared the activity and composition of biofilms on buoyant plastic and natural surfaces. Additionally, the influence of substrate size and successional stage on biofilm composition has not been commonly assessed. We incubated three plastics of distinct textures that are buoyant in rivers, low‐density polyethylene (rigid; 1.7 mm thick), low‐density polyethylene film (flexible; 0.0254 mm thick), and foamed polystyrene (brittle; 6.5 mm thick), as well as wood substrates (untreated oak veneer; 0.6 mm thick) in the Chicago River. Each material was incubated at three sizes (1, 7.5, and 15 cm2). Substrates were incubated at 2–10 cm depths and removed weekly for 6 weeks. On each substrate we measured chlorophyll concentration, biofilm biomass, respiration, and flux of nitrogen gas. We sequenced 16S and 23S rRNA genes at Weeks 1, 3, and 6 to capture biofilm community composition across successional stages. Chlorophyll, biomass, and N2 flux were similar across substrates, but respiration was greater on wood than plastics. Bacterial and algal richness and diversity were highest on foam and wood compared to polyethylene substrates. Bacterial biofilm community composition was distinct between wood and plastic substrates, while the algal community was distinct on wood and foam, which were different from each other and polyethylene substrates. These results indicate that polymer properties influence biofilm alpha and beta diversity, which may affect transport and distribution of plastic pollution and associated microbes, as well as biogeochemical processes in urban rivers. This study provides valuable insights into the effects of substrate on biofilm characteristics, and the ecological impacts of plastic pollution on urban rivers.


INTRODUCTION
Since the onset of large-scale plastic production in 1950, plastic production has increased at a compound rate of 8.4% per year (Geyer et al., 2017).Of all plastic waste generated from 1950 to 2015, about 79% was discarded in landfills or the environment (Geyer et al., 2017).In addition, plastic's chemical and physical properties facilitate environmental persistence (Chamas et al., 2020).Each year, rivers export approximately 1.15-2.41 million tons of plastic to oceans (Lebreton et al., 2017).Rivers are also key sites of plastic retention, leading to interactions with freshwater organisms (Windsor et al., 2019).
In aquatic environments, plastic litter is rapidly colonized by biofilms (Wright et al., 2020;Zobell, 1943).Biofilms are aggregates of microbial cells (i.e., bacteria, algae, fungi, and protozoa) in a matrix of extracellular polymeric substances that usually exist in a solid-liquid interface (Battin et al., 2016;Flemming & Wingender, 2010).In rivers, biofilms form the base of food webs and are sites for nutrient cycling processes (Battin et al., 2016).In addition to being centers of metabolic activity, biofilms are also highly diverse and dynamic (Battin et al., 2016).
Plastic litter provides a novel substrate that is well colonized by biofilms in freshwater ecosystems (Harrison et al., 2018;McCormick et al., 2016;Vincent et al., 2022).Biofilm colonization and growth is influenced by the chemical and physical properties of surfaces, including plastic (Cazzaniga et al., 2015;Donlan, 2002;Kim et al., 2021).The individual physical and chemical properties of plastic particles may act as forces of selection that lead to taxonomically distinct biofilm communities and differences in activity (Rummel et al., 2021), although evidence in the literature for this is mixed (Amaral-Zettler et al., 2015;Coons et al., 2021;Debroas et al., 2017;Oberbeckmann et al., 2018).More research is needed to understand the impact of plastic on biofilm community composition, activity, and biofilm-mediated ecosystem processes in freshwaters (Amaral-Zettler et al., 2020;Harrison et al., 2018;Wright et al., 2021).
Plastic litter spans a gradient of material types and sizes (Rochman et al., 2019), which could impact the composition and activity of biofilm communities.A central tenant of community ecology is that larger habitats sustain greater species richness and diversity (Wilson & MacArthur, 1967).Larger plastic substrates might therefore support greater microbial diversity than smaller substrates.However, the influence of plastic substrate size on biofilm community composition is not well known.
In this study, we asked the following question: do biofilm activity and composition differ across three common plastic polymers (with different chemical and physical properties) relative to a natural surface?We incubated three types of plastics, along with wood substrates, for 6 weeks in the Chicago River, Chicago, IL, USA, during the summer of 2021.Each substrate was deployed in three sizes.We measured biofilm amount (i.e., biomass and chlorophyll concentration) and activity (i.e., respiration and flux of nitrogen gas) weekly.We used highthroughput sequencing to assess bacterial and algal community composition at Weeks 1, 3, and 6.We predicted that the activity of biofilms would be similar across different plastic types, whereas biofilms on wood would be more active (e.g., higher rates of respiration) and have more biomass.We also predicted that microbial community assemblages would differ across all substrate types, with the community on wood exhibiting the highest diversity at the largest substrate size for all material types.

Study site
The site for this project was the Chicago River near North Ave in Chicago, IL, USA (41.905730, À87.651231).At this location, the river is turbid, approximately 4 m deep, and accessible only by boat.The plastic substrates we selected for this study are commonly found in this watershed (Hoellein et al., 2017;McCormick et al., 2014McCormick et al., , 2016)).Previous work on plastic litter and microbial biofilms has been conducted within the Chicago River at several shallower sites upstream of this location (Chaudhary et al., 2022;Hoellein et al., 2014;McCormick & Hoellein, 2016;Vincent et al., 2022;Vincent & Hoellein, 2021).

Experimental design
We set up an experiment to incubate plastic and natural substrates at the water surface.We selected three common types of buoyant plastic with different physical and chemical properties: foamed polystyrene (Density Virgin Expanded Polystyrene [EPS] Sheets, Master451, Indianapolis, IN, USA; thickness = 6.5 mm), low-density polyethylene film (Plastic Drop Cloth, VicMore, Zhengzhou, China; thickness = 0.0254 mm), and rigid low-density polyethylene (opaque off-white LDPE sheet, Small Parts, Logansport, IN, USA; thickness = 1.7 mm).We used untreated oak veneer (Oak White Flat Sawn Veneer Pack, Woodworkers, Scottsdale, AZ, USA; thickness = 0.6 mm) as the control surface because it is analogous to the floating woody debris found at the site.We compared the foamed polystyrene and polyethylene plastic substrates to wood because all represent a buoyant and persistent microbial habitat in the river and are subject to similar environmental drivers of biofilm growth (i.e., light, temperature, and movement) at the water surface.
Each substrate was deployed in three sizes: 1, 7.5, and 15 cm 2 (N = 54 pieces/material type/size class) (Table SA1).One replicate of each material type and size (N = 3 individual substrates) was arranged on a mesh wire rectangle (dimensions = 13 cm Â 6 cm with 0.64 cm mesh; 308247B Hardware Cloth, YARDGARD, Long Grove, IL, USA) that was folded into a square so that the substrates were trapped in a wire mesh "sandwich."Plastic zip ties (HS2515007, HUASU International, Zhenjiang, China) were used to attach the folded mesh, with zip ties placed between substrates to prevent contact with one another (Figure SA1).We constructed a raft (size = 1.52 Â 0.76 m) made of 1.27 cm polyvinylchloride (PVC) pipe (Charlotte Pipe and Foundry, Charlotte, NC, USA) to hold the substrates for incubation near the water surface, as occurs in situ.
We arranged substrates on a floating raft for deployment.Twelve wires (24 Gauge-30.5 m Steel Galvanized Wire, OOK, Pompano Beach, FL, USA) were wrapped around the PVC parallel to the raft's length.On every two wires, 36 substrate "sandwiches" (N = 9 of each size and material type combination) were attached to a wire in randomized order (Figure SA2).Two floats (i.e., Deluxe Party Noodle, CONNELLY, Lynnwood, WA, USA) were attached by zip ties (46-315, Gardner Bender, Milwaukee, WI, USA) to two sides of the raft.The substrate 'sandwiches' sank 2-10 cm below the surface.However, "sandwiches" with foamed polystyrene floated, so we added a metal nut (9.5-mm diameter) using a zip tie to submerge them.The raft was attached to the seawall using metal chains.We note that some plastic materials were used in the raft's construction, specifically the zip ties, floats, and the PVC frame, and that the substrates were in direct physical contact with the wire mesh.We would not expect biofilms associated with the raft's structural components to have any significant effect on the substrate-associated biofilms, and any effect that might have occurred was uniform across all substrates since they were all enclosed in the same apparatus.
The study took place from June 14 to July 26, 2021.We attached data loggers (Model UA-002-08, Onset HOBO, Bourne, MA, USA) to record temperature and light hourly.We collected a subset of substrates weekly, each time removing nine replicates of each size for each substrate type.The mesh "sandwiches" were placed individually in clean, wide mouth, glass mason jars (0.95 L, Ball Corporation, Broomfield, CO, USA), which were filled with river water off the dock, approximately 20 m from the raft.Mason jars were placed in coolers with ice packs and transported to the laboratory within 1 h.Each week we collected three 20 L carboys of river water for use in the respiration and nitrogen gas (N 2 ) flux incubations.Three of the nine replicates were used to determine the biofilm community composition and chlorophyll concentration, three replicates were used to measure respiration and biofilm biomass, and three replicates were used to measure N 2 flux.
We measured several environmental parameters on the deployment date and on each collection date (Table SA2).We collected triplicate water samples by filtering river water (C2225-NN, Thermo Fisher Scientific, Waltham, MA, USA) into 20-mL scintillation vials.The water samples were kept cold and dark during transit to the lab, where they were frozen until analysis.On each collection date, we measured Secchi depth as an estimate of light penetration.We measured dissolved oxygen (DO), water temperature, (HQ40d portable meter, Hach, Loveland, CO, USA), and conductivity (30-10FT, YSI, Yellow Spring, OH, USA).

Sequence of measurements
On each date, substrates were collected in the morning, and brought to the lab within 1 h.We placed the jars holding substrates for flux of N 2 in a refrigerator (4 C) until processing the next day.Substrates for respiration were removed from their mesh enclosure and gently rinsed with deionized (DI) water to remove debris and invertebrates.Respiration measurements began immediately, while a separate team removed and gently rinsed the substrates designated for chlorophyll and DNA extraction, initiating the sample processing.

Respiration and biomass
We poured 15 L of river water into a clean bucket and measured the DO and temperature.After rinsing the substrates, each was placed in a 160-mL specimen cup, submerged in the bucket, and capped underwater making sure no air bubbles were present.Three cups with only river water were incubated to account for respiration in the water and abiotic changes in DO.We placed cups in a dark container to prohibit photosynthesis.After 3 h, we recorded DO, temperature, and time elapsed (Hoellein et al., 2014;Vincent & Hoellein, 2021).Respiration was calculated as the change in DO (final-initial) while correcting for changes in DO of the river water alone (units: mg O 2 cm À2 h À1 ).Each substrate was individually wrapped in aluminum foil and frozen (À20 C) until later measurement of biofilm biomass.
Biofilm biomass was quantified by staining substrates with 1% aqueous crystal violet (hexamethyl pararosaniline chloride; C 25 N 3 H 30 Cl) and measuring absorbance at 595 nm on a spectrophotometer (Spectronic 20 Genyses Spectrophotometer, Thermo Fisher Scientific, Waltham MA) (Burton et al., 2007).Prior to quantifying biomass, substrates were thawed overnight at 4 C. Large, medium, and small substrates were placed in 150 mL aluminum pans and stained with 1, 0.5, and 0.25 mL of 1% crystal violet respectively.Crystal violet stained the entire surface area of substrates for 45 min, and substrates were then rinsed with DI water three times to remove excess stain.Water was added to submerge substrates for 1 h to remove any remaining crystal violet unadhered to biofilm.A sterile metal nut was placed atop foam substrates to keep them submerged.Substrates were removed from the water and dried for 24 h at room temperature.We transferred substrates to plastic weight boats, added 15 mL of 95% ethanol, and swirled for 30 s to elute crystal violet adhered to biofilm.The elution continued for 10 min with occasional swirling.This solution was poured into 5-mL test tubes, and 3 mL was measured on a spectrophotometer at 595 nm in a 10-mm quartz cuvette.A 1:5 dilution was applied if the optical density was above the spectrophotometer's detection limit.Each time the crystal violet assay was performed (N = 6), we included one of each substrate and size that was fresh (not deployed) to account for background absorbance, and the background absorbance was subtracted from the values measured for the deployed substrates of the corresponding substrate type.Optical density was calculated according to substrate surface (cm À2 ) (Vincent et al., 2022).

Biofilm community composition and chlorophyll-a
Biofilm DNA was extracted from substrates from Weeks 1, 3, and 6 for bacterial and algal community composition as indicators of the early, intermediate, and late stages of incubation.After carefully rinsing each substrate, they were placed on a 150 ml aluminum pan and cut in half with a sterile razor blade or scissors.Each half was placed in an individual 15 ml centrifuge tube for either extracting DNA or measuring chlorophyll concentration.Tubes were frozen (À20 C) until each assay was performed.For DNA extraction, we first removed frozen substrates from the 15-mL centrifuge tubes using sterile forceps.For film substrates, the entire substrate was added directly to Powerbead tubes (i.e., no cutting), because their flexible nature allowed the entire substrate to fit in the tube and still receive sufficient bead beating to extract DNA.The same was true for the small substrates for wood, foam, and rigid polyethylene.However, we cut the medium and large wood, foam, and rigid substrates with a razor blade into fragments and added a random portion to the Powerbead tubes to fill $25% of the tube volume.This was needed given their thickness and lack of flexibility, which inhibited bead beating.
DNA extractions were performed using the Qiagen Power Soil DNA extraction kit (Qiagen, Hilden, Germany) following the manufacturer's instructions.Two "kit" controls (i.e., no substrate) were also included with DNA extractions to control for contamination.After DNA was extracted, PCR was performed using the CS1_515F and CS2_806R primers to amplify the V4 hypervariable region of the16S rRNA gene (Caporaso et al., 2012) for bacterial taxonomic identification.PCR was also performed using the CS1_p23SrV_f1 and CS2_p23SrV_r1 primers to amplify the plastid 23S rRNA gene in eukaryotic algae and cyanobacteria (Sherwood & Presting, 2007).PCR was confirmed by agarose gel electrophoresis.PCR products were sequenced by the Rush University Medical Center Genomics and Microbiome Core Facility in Chicago, IL on an Illumina MiniSeq (2 Â 150 paired-end sequencing) for 16S PCR products and on an Illumina MiSeq (2 Â 250 paired-end sequencing) for 23S PCR products.
Paired end reads were cleaned, assembled, and analyzed in DADA2 (Callahan et al., 2016) after demultiplexing.Raw bacterial 16S reads were 154 bp and algal 23S reads were 251 bp.Forward and reverse reads were trimmed to remove primer sequences with trimLeft = c (19,20) for 16S reads and trimLeft = c(20,20) for 23S reads.Algal 23S reads were also trimmed with trimRight = c(20,20) to remove low-quality bases at the ends of reads.We also trimmed 16S reads at the first occurrence of Q = 13 with TruncQ = 13 and 23S reads at the first occurrence of Q = 11 with TruncQ = 11.Filtering was done with maxEE = c(1,1) for 16S reads and with maxEE = c(2,2) for 23S reads.These filtering and trimming steps were performed as a quality control measure.Any 16S and 23S reads with ambiguous bases were also removed.A minimum read length filter minLen = c (135,134) was applied to 16S reads and a minimum read length filter minLen = 210 was applied to 23S reads to keep reads long enough for proper merging.For both 16S and 23S reads, the DADA2 error learning model was run with nbases = 10 10 .Bacterial ASVs were assigned and merged if they overlapped by 13 or more bases (minoverlap = 13).Algal ASVs were assigned and merged if they overlapped by 48 or more bases (minoverlap = 48).Bacterial reads that merged outside the range of the V4 region (250-256 bp) and algal reads that merged outside 365-374 bp were discarded.Lastly, for both 16S and 23S reads chimeras were removed with the default parameters.Bacterial 16S reads were aligned against the SILVA SSU 138.1 database (Quast et al., 2012) and were assigned taxonomy using the IDtaxa algorithm with DECIPHER (Davis et al., 2018).Algal 23S reads were aligned against a custom algal database based off the SILVA LSU 132 database (Quast et al., 2012) using the AssignTaxonomy algorithm.Any 16S reads unclassified at the domain level or classified as chloroplast, archaea, or mitochondria were removed.Any 23S reads unclassified at the kingdom level were removed.For both 16S and 23S reads contaminant sequences were identified and removed with the Decontam R package (Davis et al., 2018) with a threshold of 0.5.Before downstream analysis, we rarefied 16S reads to 11,502 reads and 23S reads to 7289 reads without replacement with Phyloseq (McMurdie & Holmes, 2013).
Chlorophyll-a was measured using the hot ethanol method (Sartory & Grobbelaar, 1984).A day prior to measuring chlorophyll, tubes with frozen substrates were moved from the freezer into a refrigerator (4 C) to thaw overnight.Ethanol (95%) was added to each tube to fully submerge the substrate, and the volume of ethanol added to each tube was recorded.A sterile glass weight was added to tubes containing foam substrates to keep them submerged.Tubes were placed in a rack and submerged in a 75 C water bath.We removed the rack from the water bath after 15 min and placed it in the dark for 2 h.Then, each tube was inverted and placed back in the rack for 15 minutes to allow any loose particles to settle.Once the extraction was complete, chlorophyll concentration was measured on a Turner Designs Trilogy Laboratory Fluorometer (Trilogy Laboratory Fluorometer, Turner Designs, San Jose, CA, USA) using the chlorophyll-a acidification module.The fluorometer was calibrated at five known concentrations of pure chlorophyll (Chlorophyll-a analytical standard, Millipor-eSigma, Burlington, MA, USA) in 95% ethanol the same week chlorophyll concentration was measured (Vincent et al., 2022).

Nitrogen flux
Flux of N 2 was measured the day after substrates were removed from the river.Substrates were removed from the refrigerator, separated from the mesh enclosure, and gently rinsed with DI water to remove any loose debris.River water was also removed from the refrigerator and allowed to come to room temperature.The river water was enriched with sodium nitrate (final concentration added = 3 mg N L À1 ) and dextrose (final concentration added = 15 mg C L À1 ) with the objective of measuring denitrification potential (Hoellein & Zarnoch, 2014).Individual substrates were placed in 160-mL specimen containers, submerged in enriched river water, and placed in the dark at room temperature for about 3 h as described for respiration.Three controls were also incubated with river water alone.To collect dissolved gas samples, we used a 60-mL syringe with attached rubber tubing to slowly draw water from the specimen container (using care not to introduce turbulent flow or bubbles).We filled a 12-mL exetainer with the water by placing the tube at the bottom and allowing it to overflow for several volumes.We then sterilized samples with 200 μL of 50% zinc chloride and stored the exetainers under water in a refrigerator until analysis of dissolved gas.Triplicate samples of starting conditions in the enriched water were also taken using the same approach.Ratio of N 2 : argon (Ar) was measured on a Membrane Inlet Mass Spectrometer (MIMS Bay Instruments, Easton, MD, USA) with ultra-pure water as the standard (18 M Ω resistance; E-Pure, Barnstead International, Dubuque, IA, USA).The standard temperature was set at 21.0 C using a circulating water bath (VWR International, Radnor, PA, USA) equilibrated to the atmosphere with low-speed stirring for 24 h (Lab Egg RW11 Basic, IKA Works, Inc., Wilmington, NC, USA) (Kana et al., 1998).During each run, we analyzed a standard every 3-9 samples to account for instrument drift.Concentrations of dissolved N 2 gas were calculated by multiplying the N 2 :Ar ratio by the equilibrium concentration of Ar (Kana et al., 1994).Flux in N 2 was calculated by subtracting initial N 2 from the final concentration and correcting for the flux observed in water alone.We did not quantify intermediate products (e.g.nitrous oxide) or measure N-fixation or denitrification separately.

Data analysis
We used a non-parametric, aligned rank transformation (ART) test (Wobbrock et al., 2011) to compare differences in respiration, biomass, chlorophyll, and flux of N 2 by two factors: material type and week.We conducted the 2 factor ART tests for each substrate size separately.We used the ART approach as the data were not normally distributed and could not be transformed to meet assumptions of parametric (i.e., ANOVA) analyses.In cases where there was a significant interaction, we performed a multifactor contrast test (ART-C) (Elkin et al., 2021) with Tukey's method for correcting p-values following multiple comparisons.All pairwise comparisons were generated but only comparisons where week was the same are presented to show differences among material types within individual weeks.All statistics were performed in R (R version 4.1.1,R Core Team 2021).
Biofilm 16S and 23S α-diversity was quantified by ASV richness and ASV Shannon diversity through Phyloseq (McMurdie & Holmes, 2013).We tested the assumptions of normality and equal variance with the Shapiro-Wilk test and Levene's test respectively and excluded outliers, which we identified as any points more than 1.5 IQR below Q1 or more than 1.5 IQR above Q3, when they generated different results or caused a violation of the assumptions of equal variance or normality.We compared 16S and 23S ASV richness and Shannon diversity for each size separately with material type and week as our two factors by two-way ANOVA for all sizes but used an ART test in four cases where ANOVA assumptions were not met.Following a significant effect, we used post-hoc test (i.e., ART-C while adjusting for multiple comparisons with Tukey's method or Tukey's post-hoc test following ANOVA).β-diversity was visualized by non-metric multidimensional scaling (NMDS) with Bray-Curtis dissimilarity index through the R package Vegan (Oksanen et al., 2020).We used Bray-Curtis distances and PERMANOVA (Oksanen et al., 2020) to compare community composition by material type, size, and week.This was followed with pairwise PERMANOVA to compare the effects of material type alone, week alone and size within each material type, while adjusting for multiple comparisons by the false discovery rate (FDR) method.We next calculated the mean relative abundance at the family level, grouping by material and week.We did not include size as an additional variable because PERMANOVA results revealed there were no differences among size for each material type.We visualized mean relative abundance by a stacked bar plot where we pooled families with a maximum relative abundance <2.85% for bacteria and <2.70% for algae across our grouping variables as "Other."We performed differential abundance analysis at the family level comparing plastic to wood within each week.We tested differential abundance (Wilcoxon rank-sum test and FDR approach) within each week because our NMDS also showed separation by week.We calculated Wilcoxon effect size for each differentially abundant family.

Physical and chemical conditions
The chemical and physical parameters of the Chicago River at the incubation site were typical of urban, eutrophic waterways (Table SA2).Conductivity was 782-995 μS cm À2 , DO was 4.95-6.37mg O 2 L À1 and 56.6%-82.7%.Secchi depth showed relatively low light penetration (0.5-0.75 m) and nitrate concentrations were 2109-6198 μg NO 3 À -N L À1 .Water temperature and illuminance at the surface throughout the study were 19.95-30.26C and 0-187378.15lx, following a diel pattern (Figure SA3).

Respiration, biofilm biomass, chlorophyll, and N 2 flux
The study was conducted for 6 weeks.Because each response metric (respiration, biofilm biomass, chlorophyll, and N 2 flux) showed a significant substrate Â week interaction, we compared among substrates (film, foam, rigid, and wood) for each size class (large, medium, and small) separately on each week.We used a multifactor contrast test (ART-C) with Tukey's method for correcting p-values following multiple comparisons to compare substrates within each week.Thus, N = 18 comparisons where the response variable was compared among the four substrates (N = 6 weeks each for large, medium, and small, e.g., Figure 1).Biofilm respiration showed a significant interaction between material type and week for the large (ART F 15 = 2.96, P = 0.002), medium (ART F 15 = 2.31, P = 0.015), and small (ART F 15 = 6.40,P < 0.001) substrates (Table 1).For 9 of the comparisons across substrates on a given week (here after "weekly comparisons"), respiration on wood was significantly higher than all plastic substrates (Figure 1).For the remaining nine of the weekly comparisons, wood was not significantly different from one or more of the 3 plastic substrates (Figure 1).For 16 of the weekly comparisons, respiration on the three plastic substrates showed no differences among one another (Figure 1).
Biofilm biomass showed a significant interaction between material type and week for all three substrate sizes (ART, large F 15 = 5.40, P < 0.001, medium ART F 15 = 5.25, P < 0.001, small ART F 15 = 2.17, P = 0.022; Table 1).For all substrates, biomass increased in a linear fashion from Weeks 1-3, with no differences among substrate types (Figure 2).On Week 4, film had significantly less biomass than either wood (large; Figure 2a) or foam (medium and small; Figure 2b,c).On Week 5, biomass showed no difference among substrate types.On Week 6, biomass was significantly lower on large wood than all other large substrates (Figure 2a), and on medium-sized substrates, wood had significantly less biomass than foam and rigid plastic (Figure 2b).
Patterns for chlorophyll resembled biomass, with a significant interaction between week and material type for the large (ART, F 15 = 1.97,P = 0.039) and medium (ART, F 15 = 2.18, P = 0.021) substrates, although there was no interaction for small substrates (ART, F 15 = 1.41,P = 0.182) (Table 1).For large substrates, there was no difference among substrate types for Weeks 1-5, (Figure 3a) but on Week 6, wood had significantly less chlorophyll than rigid polyethylene (Figure 3a).On medium and small substrates, chlorophyll concentration increased over time with no differences among substrates within individual weeks (Figure 3b,c).
Flux of N 2 was highly variable across sizes and weeks, with a significant interaction between week and material type for large (ART F 15 = 3.57P < 0.001) and medium (ART F 15 = 2.97 P = 0.002) substrates, while small substrates showed only a significant week effect (ART F 15 = 3.32 P = 0.012; Table 1).For large substrates, ART-C pairwise comparisons showed N 2 flux on rigid plastic was significantly higher than foam on week 4, with F I G U R E 1 Mean (±SE) respiration rates (mg O 2 cm À2 h À1 ) over 6-week period on (a) large, (b) medium, and (c) small substrates.Small letters correspond to ART-C pairwise comparison results after a significant substrate Â week interaction; "*" is used when wood was significantly different than all other substrate types.More negative values indicate higher rates of respiration.
the opposite pattern on Week 5 (Figure A4a).For medium substrates, film had significantly higher N 2 flux than rigid substrates on Week 5 (Figure A4b).

Biofilm community composition: Bacteria
Biofilm community metrics were only measured for Weeks 1, 3, and 6, but we used the same process for comparing patterns among substrates and time as described above.Bacterial ASV richness showed a significant interaction between week and material type on large (twoway ANOVA, F 6 = 2.87, P = 0.030) and small (ART, F 6 = 7.13, P < 0.001; Table 2) substrates.On large substrates, Weeks 1 and 6 showed no differences among substrate types, while on Week 3, richness was higher on foam and wood than on film and rigid substrates (Figure 4a).On small substrates in Weeks 3 and 5, foam had the highest richness (Figure 4c).On medium substrates, week (two-way ANOVA, F 2 = 10.46,P < 0.001) and material (two-way ANOVA, F 3 = 5.46, P = 0.006; Table 2) had significant effects, where Week 6 was different from Weeks 1 and 3, and rigid substrates were different than foam (Tukey's post hoc test, P = 0.022) and wood (Tukey's post hoc test, P = 0.009) (Figure 4b).
Bacterial ASV Shannon diversity patterns were similar to richness, with an interaction between material type and week on large (two-way ANOVA, F 6 = 4.42, P = 0.004) and small substrates (two-way ANOVA, F 6 = 5.15, P = 0.002; Table 2).For large substrates, there were no differences among substrates Week 1, but on Weeks 3 and 6 diversity was higher on foam and wood relative to rigid and film substrates (Figure A5a).For small substrates, there were no differences on Week 1, in Weeks 3 and 6 foam was significantly more diverse than other substrates (Figure SA5c).ASV Shannon diversity on medium substrates was different by material type (two-way ANOVA, F 3 = 5.22, P = 0.006) and week (twoway ANOVA, F 2 = 6.26,P = 0.006; Table 2) without an interaction.For medium substrates, diversity on foam was higher than rigid substrates (Figure SA5b).
Non-metric multidimensional scaling (NMDS) by Bray-Curtis dissimilarity distances showed bacterial communities on wood differed from communities across plastic types (Figure 5a).Bacterial communities also grouped separately according to week, with no grouping by substrate size (Figure 5a).Pairwise PERMANOVA comparing substrate types showed that communities on wood were different than all three plastic types, and that communities on foam were different from rigid substrates (Table SA3).For each material type, there was no difference based on substrate size (Table SA3).Bacterial communities also differed between each week (Table SA3).
We identified 396 unique bacterial families, 113 of which were differentially abundant between plastic and wood (Tables SA4-SA6).On all plastic types Comamonadaceae was the most abundant family on Week 1, and it was significantly more abundant on plastic than wood (Figure 6a and Table SA4).On wood, the most abundant family Week 1 was Sphingomonadaceae which was significantly more abundant than on plastic (Figure 6a and Table SA4).In Week 3, Methylomonadaceae was the most abundant family on all plastic and was significantly more abundant than on wood (Figure 6a and Table SA5).On wood, Sphingomonadaceae remained the most abundant family for Week 3 (Figure 6a), significantly higher than plastic (Table SA5).By Week 6, the most abundant family on rigid substrates was Comamonadaceae, on film it was Methylomonadaceae, and on foam it was Chitinophagaceae (Figure 6a).Of these families only Methylomonadaceae was significantly more abundant on plastic than wood (Table SA6).On wood the most abundant family Week 6 was Methylophilaceae (Figure 6a) and was significantly more abundant than on plastic (Table SA6).

Biofilm community composition: algae
We also quantified algal community composition on Weeks 1, 3, and 6 and used the same process for comparing patterns among substrates and time as described for bacteria above.Algal ASV richness was significantly different by material type on large (two-way ANOVA, F 3 = 24.75,P < 0.001), medium (two-way ANOVA F 3 = 17.33,P < 0.001), and small (ART, F 3 = 26.31,P < 0.001; Table 2) substrates.For all, ASV richness on F I G U R E 2 Mean (±SE) optical density (cm À2 ) of crystal violet (biomass) over 6-week period on (a) large, (b) medium, and (C) small substrates.Small letters correspond to ART-C pairwise results after a significant substrate Â week interaction; "*" is used when wood was significantly different than all other substrate types.
film and rigid substrates were the same and lower than ASV richness on wood and foam, which were not different from each other (Figure 4).
Algal ASV Shannon diversity showed an interaction between material type and week on large (ART, F 6 = 6.97,P < 0.001), medium (two-way ANOVA, F 6 = 3.79, P = 0.009), and small (ART, F 6 = 4.10, P = 0.006; Table 2) substrates.For all sizes, there were no differences among substrates in Week 1 (Figure SA5).On Weeks 3 and 6, algal ASV diversity on foam and wood were higher than and rigid substrates on large, medium, and small substrates (Figure SA5).
Non-metric multidimensional scaling (NMDS) by Bray-Curtis dissimilarity distances showed algal communities on wood differed from communities across plastic types (Figure 5a).PERMANOVA showed algal communities were different with a material and week interaction (PERMANOVA F 6 = 7.62, P < 0.001; Table SA7).
Pairwise PERMANOVA comparing substrate types showed that communities on wood and foam differed from each other and were different than film and rigid substrates (Table SA7).Communities on film and rigid substrates did not differ from each other (Table SA7).Algal communities also differed between each week, while there was no difference based on substrate size (Figure 5b and Table SA7).
We recorded 114 unique algal families, 53 of which were differentially abundant between plastic and wood (Tables SA8-SA10).On film and rigid plastic types, Unclassified Bacillariophyta was the most abundant family Week 1, on foam it was Unclassified Chlamydomonadales and on wood it was Chlamydomonadaceae (Figure 6b).After pooling plastic together and testing for differentially abundant algal families, neither Unclassified Bacillariophyta nor Unclassified Chlamydomonadales were differentially abundant; however, F I G U R E 3 Mean (±SE) chlorophyll concentration (μg cm À2 ) over 6-week period on (a) large, (b) medium, and (C) small substrates.Small letters correspond to ART-C pairwise comparison results after a significant substrate Â week interaction; "*" is used when wood was significantly different than all other substrate types.
T  Chlamydomonadaceae was significantly more abundant on wood in Week 1 (Table SA8).On Week 3, Unclassified Bacillariophyta was the most abundant family on all plastic (Figure 6b) and was significantly more abundant than on wood (Table SA9).On wood, Chlamydomonadaceae was the most abundant family on Week 3 (Figure 6b) and was significantly more abundant than on plastic (Table SA9).By Week 6, the most abundant family on film and rigid substrates was Unclassified Bacillariophyta, on foam it was Eustigmataceae, and on wood it was Chlamydomonadaceae (Figure 6b).Unclassified Bacillariophyta was significantly more abundant on plastic in Week 6 and Chlamydomonadaceae was significantly more abundant on wood (Table SA10).

DISCUSSION
Our objective was to determine if plastic selects for distinct biofilms relative to wood by measuring amount, activity, and community composition of biofilms.
Because the experimental design also incorporated time Mean relative abundance (%) of the top 20 (a) bacterial and (b) algal families on each material type for Weeks 1, 3, and 6.Families are organized from bottom to top in order of greatest overall mean relative abundance and the legend follows the same order.The "Other" category includes all taxa where the maximum mean relative abundance was <2.85 for bacterial families and <2.70% for algal families.
and substrate size as factors, a variety of patterns emerged across the response variables.Our hypothesis that wood would show distinct patterns from all three plastic types was supported by data on respiration and biofilm assemblage, but not for biofilm biomass, chlorophyll, and alpha diversity.Our hypothesis that the three plastic types would show distinct patterns from one another was not supported, as biofilms on foamed polystyrene were distinct from the two polyethene types, which were similar to each other.Finally, our prediction that larger substrates would have increased taxonomic diversity relative to smaller ones was not supported, as size had no impact on community composition.

Biofilm activity: Respiration and N 2 flux
The clearest trend in rates for ecosystem processes was that wood had higher rates of respiration compared to all three plastic types, which showed low and similar rates for all substrate sizes.The likely explanation is wood is a growth surface and energy source for microbes, while plastic was a growth surface but not an energy source (Oberbeckmann et al., 2021).Anecdotally, wood degradation was visible after 6 weeks.From an ecological perspective, wood and buoyant plastic provide a similar habitat for biofilm growth and move in a comparable way in large rivers.However, the durability of wood as habitat is short-lived relative to plastic, suggesting plastics can sustain biofilm organisms further than wood (Hoellein et al., 2019;Webster et al., 1999).
Previous studies where biofilm respiration was compared on natural and plastic surfaces found that biofilm respiration rates on tile (i.e., a surrogate for rocks) were not overall consistently higher compared to biofilm respiration rates on plastic surfaces (Chaudhary et al., 2022;Vincent et al., 2022).This provides additional support for our conclusion that the consistently higher rates of respiration we observed on wood were a result of its metabolism by biofilm.
Unlike respiration, there were no significant trends among substrates for N 2 flux.We added nitrate and dextrose to the incubation assays with the objective of measuring potential denitrification (i.e., N 2 production when N and C are not limiting; DNP).However, N 2 flux was highly variable and distinct from other biofilm metrics.This suggests that the chemical and physical properties of the substrates did not affect DNP.Alternatively, different approaches to DNP (e.g., acetylene block), measurements of N cycling genes, or longer incubation times may offer more insight into potential impacts of plastic on N 2 flux.
Relatively few studies have examined the impacts of plastic litter on N 2 dynamics and have focused on microplastics.In a mesocosms study by Chen et al. (2020), DNP was initially higher on microplastic compared to the water column, but the difference did not persist.Increased denitrification was reported in sediments with microplastic due to the greater porosity provided by microplastic (Seeley et al., 2020).The collective data provide evidence that plastic litter may impact N dynamics under some circumstances, via microbial colonization or effects on sediment porosity, and merits more research attention.

Biofilm structure: Biomass and chlorophyll
Biofilm biomass and chlorophyll were similar on all substrates throughout the incubation, suggesting that biofilms have similar structural characteristics over time.Other studies have found similar chlorophyll concentrations among biofilms across plastic polymers (Chaudhary et al., 2022); plastic polymers and tile (Vincent et al., 2022); and plastics, steel, and wood (Muthukrishnan et al., 2019).However, the same studies noted different patterns for biofilm biomass across substrates.This suggests that chlorophyll density is more similar across substrates while biofilm biomass shows variable dynamics, perhaps because biomass measurements are broader and may include living and dead cells as well the EPS.However, based on our respiration results and its noticeable degradation, we expected higher biomass on wood.Biomass on wood may have been underestimated because the retention and variability in retention of crystal violet by wood control substrates was considerably greater than for plastic control substrates (Table SA11).Perhaps longer soaking periods in water prior to eluting the bound stain could enhance removal of excess stain retained by wood.It is also possible that the lower-than-expected biomass on wood was due to biofilm maturation and detachment but because we did not measure detachment we are unable to attribute biomass results on wood to this.

Biofilm assemblage: Richness and diversity
We predicted richness and diversity of bacterial and algal taxa would be highest on wood relative to all three plastic types, but the results did not align clearly with our expectation.By substrate type, richness and Shannon diversity for bacteria and algae were highest on wood and foam compared to film and rigid polyethylene.We attributed the results to the impact of physical surface properties, as foam and wood have more complex and rougher surfaces than the film and rigid polyethylene, which were smoother and more homogenous.
Surface roughness promotes biofilm development because it offers greater surface area for colonization, protection from sheer stress, and increased asperity reduces energy requirement for cell adhesion (Ammar et al., 2015;Cazzaniga et al., 2015;Donlan, 2002;Katsikogianni & Missirlis, 2004;Renner & Weibel, 2011).Foam and wood offered a porous texture and thereby greater heterogeneity of conditions for biofilm organisms, which likely enhanced ASV richness and diversity.In contrast, the homogenous, smooth surface of film and rigid substrates may have diminished potential microhabitats, which limited richness and diversity.
Increased metrics of alpha diversity for biofilms on wood also emerged when compared to polyethylene (PE) and polyethylene terephthalate (PET) (Muthukrishnan et al., 2019), high density polyethylene (HDPE) and hard polystyrene (Kesy et al., 2019;Oberbeckmann et al., 2021), linear polyethylene (LPE) and polypropylene (PP) (Miao et al., 2019), and PET and polymethyl methacrylate (PMMA) (Shen et al., 2021).The consistent pattern across studies shows greater richness and diversity of bacterial taxa on wood than plastic, which is a logical consequence of its biodegradable nature and the long evolutionary history of the wood-degrading microbes.
Increased roughness and textural heterogeneity also likely explained the patterns of diversity and richness on our foamed polystyrene substrates.Published studies have reported increased metrics of alpha diversity on polystyrene (PS) compared to other plastic types in some cases (Chen et al., 2021;Frère et al., 2018).While other studies that have assessed alpha diversity metrics across plastic types have found either no differences among PS and other plastics (Li et al., 2019) or lower alpha diversity on PS compared to other plastics (Vincent et al., 2022).Not all studies report the specific type of PS used, which limits our comparisons.Physical and chemical differences within foamed PS varieties may explain why patterns of richness and diversity differ across studies, which may impact interpretations across studies, but more work is needed to quantify the factors affecting biofilm patterns on foamed PS, a common type of plastic pollutant in ecosystems worldwide.

Biofilm assemblage: Bacterial communities among substrates
We predicted bacterial community composition would be different across the four substrate types, but the NMDS and PERMANOVA results offered mixed support for this hypothesis.The bacterial community on wood differentiated the most compared to communities on plastic.But the communities on plastic showed nuanced differences: no difference between film polyethylene and foamed polystyrene, and no difference between film and rigid polyethylene.Our results were consistent with studies that have demonstrated differences in bacterial biofilm assemblage on wood relative to plastic (Debroas et al., 2017;Kesy et al., 2019;Miao et al., 2019;Muthukrishnan et al., 2019;Oberbeckmann et al., 2018Oberbeckmann et al., , 2021;;Shen et al., 2021).The collective body of work offers a consistent statement that distinct bacterial communities occur on plastic relative to wood and persist across a wide variety of ecosystem types and conditions.
Several bacterial families that were dominant members of the overall community showed differences between plastic and wood.Those abundant on wood may suggest selection for organisms that metabolize it.For example, Rhizobiaceae were abundant overall and significantly more common on wood than plastic in Weeks 1, 3, and 6.Rhizobiaceae are mostly aerobic, chemoorganotrophs that metabolize carbohydrates and organic acids (Carareto Alves et al., 2014), including wood (Pettersen, 1984).Brii41 was also significantly more abundant on wood than plastic throughout the study, and decomposes organic matter (Cai et al., 2018;Dai et al., 2021).Finally, Methylophilaceae was significantly more abundant on wood than plastic in Weeks 3 and 6.Release of methanol from the degradation of the pectin and lignin in wood substrates (Galbally & Kirstine, 2002) could stimulate Methylophilacae, which oxidize singlecarbon compounds (Doronina et al., 2014;Kalyuhznaya et al., 2009).Our phylogenetic approach does not identify the metabolic capacity, however, the patterns among substrates, combined with other metrics (e.g., respiration), suggest some substrate-mediated selection occurred.
Several families of bacteria were common overall and showed preferential abundance on plastic relative to wood, including Methylomonadaceae, Beijerinckiaceae and Unclassified Gammaproteobacteria.Previous studies have also found Unclassified Gammaproteobacteria on plastic biofilms (Hoellein et al., 2014;McCormick et al., 2016;Vincent et al., 2022).Each family contains methane oxidizing bacteria (MOB) (Cabrol et al., 2020), which could be connected to the oxidation of plastic polymers into CH 4 when exposed to solar radiation and water (Royer et al., 2018).Amoebophilaceae was also significantly more abundant on plastic throughout the study.Elsewhere, this taxon increased in the gut of a soil oligochaete that was fed nanoplastics (Zhu et al., 2018).Finally, the families Alteromonadaceae, Xanthomonadaceae and Spirosomaceae were differentially abundant on plastic in one or more weeks.These families were noted in a recent review (Wright et al., 2021) as potential hydrocarbon degrading bacteria that are consistently more abundant on plastic in situ.

Biofilm assemblage: Algal communities among substrates
The differentially abundant algal families on plastic vs. wood offers some insight into substrate-based selection of biofilm composition.Overall, green algae were abundant on all substrates, so patterns between wood and plastic were relegated to other algal groups.The algal taxa with greater abundance on wood than plastic included two families of red algae (Compsopogonaceae on Weeks 1 and 3 and Batrachospermaceae on Week 6), one golden algae family (Chromulinaceae on Weeks 1 and 3), one yellow-green algal family (Vaucheriaceae on Week 1), and one brown algal family (Unclassified Ectocarpales on Week 6).In contrast, there were few taxa that were more abundant on plastic than wood.Of the 85 entries of differentially abundant algae (Tables SA8-SA10), in only 17 cases was the algal family significantly more abundant on plastic than on wood.Our data cannot determine a definitive mechanism but offers some support for the preferential colonization of wood, which is a component of the evolutionary history of algae, rather than the novel surface of plastic.
Although we did not perform differential abundance analysis across the three plastic substrates, the mean relative abundance results show differences between the two LDPE substrates and the foamed PS.We note that the diatom community structure was more similar between wood and foam than foam compared to the LDPE substrates, which had greater abundance of diatoms.A possible mechanism for this might be stronger adherence by diatoms to hydrophobic surfaces (Finlay et al., 2002;Holland et al., 2004) since PE was more hydrophobic than foamed PS (Min et al., 2020).However, we did not measure the hydrophobicity of our substrates directly and we note that biofilm colonization decreases hydrophobicity (Lobelle & Cunliffe, 2011;Wright et al., 2020).To our knowledge, there are no published studies that compare algal community composition on plastic and wood substrates, which limits our ability to make comparisons to previous work.However, previous assessments of overall eukaryote community composition (which includes most algae), show that eukaryote communities on wood were distinct from the those on HDPE and hard PS (Kettner et al., 2019).Distinct eukaryote communities were also identified between wood and plastic debris in the north Atlantic (Debroas et al., 2017).Overall, the unique properties of plastic surfaces can select for distinct algal communities relative to natural habitats, but this merits greater investigation to quantify the mechanisms of action and the ecological implications.

Biofilm assemblage: No impact of substrate size
We predicted a positive relationship between substrate size and biofilm diversity for all substrate types, but our data did not support this hypothesis.Neither the bacterial NMDS (Figure 5a) nor the algal NMDS (Figure 5b) showed any grouping by substrate size, and our PERMA-NOVA analysis further confirmed that there was no difference in community structure based on size for any material type (Tables SA3 and SA7).At least within the range of particle sizes we selected, there was no difference in community composition depending on size.
Previous work has found similar patterns when comparing bacterial biofilm communities by particle size (Frère et al., 2018;Parrish & Fahrenfeld, 2019).In contrast, Debroas et al. (2017) noted lower bacterial and eukaryote community composition on microplastics compared to macroplastics.Based on our data and those of others, for a particular polymer type, size is not likely to affect biofilm richness and diversity unless compared across large size ranges (macro-and microplastics).Further, the differences within a much broader size range might also be associated with the methodological challenges of recovering DNA from microplastic samples.Thus, microbial community differences between microplastics and macroplastics could simply be a result of those methodological challenges.

Long-term ecological implications of biofilm growth on plastic relative to wood surfaces
The degradation of wood relative to the recalcitrance of plastic, combined with the robust biofilm growth on both surfaces, has implications for the fate of plastic and associated biofilms.The durability of plastic surfaces allows more generations of biofilms to grow on the same surface relative to wood, which will degrade.In addition, the relatively new (i.e., 75 years) and accelerating input of plastic litter to aquatic ecosystems provides more floating surface area for biofilms to colonize than would be available otherwise.Plastic retention in rivers can also be enhanced by biofilm growth which adds mass and "stickiness" to plastics resulting in retention rather than transport (Chen et al., 2019;Fazey & Ryan, 2016;Hoellein et al., 2019;Kaiser et al., 2017;Lobelle & Cunliffe, 2011).At larger spatial scales, the implications of plastic-associated biofilms may have an impact on biogeochemistry (e.g., respiration, production, and N cycling).However, the overall impact of plastic pollution on biogeochemical cycles has not been established thus far.Future studies can improve our understating of the biofilm communities on plastic litter by focusing on higher temporal resolution, including fungi and protozoa in the analysis of the biofilm community, and measuring its effect on biogeochemical processes.In addition, "-omics" approaches including metagenomics, metatranscriptomics, metaproteomics as well as machine learning will help answer questions regarding the ongoing genetic and physiological processes of plastic on biofilm communities and biofilm-mediated ecosystem processes.

F
I G U R E 5 Nonmetric multidimensional scaling (NMDS) ordination of (a) bacterial communities (Bray-Curtis dissimilarity, stress = 0.0927) and (b) algal communities (Bray-Curtis dissimilarity, stress = 0.158) from substrates of four different material types and three different sizes in Weeks 1, 3, and 6.Ellipses represent 95% CI.Solid ellipses correspond to material colors in the legend.Dashed line ellipses correspond from left to right to Weeks 1, 3, and 6.Ordinations created using 100 iterations.
Aligned-rank-transform comparisons for each measurement type for each size class.Non-significant P-values are in bold.
T A B L E 1 A B L E 2 Results from two-way ANOVA for bacterial ASV richness and ASV Shannon diversity and from ART for small ASV richness and results from two-way ANOVA on algal large and medium ASV richness and medium ASV Shannon diversity and ART on algal small ASV richness and large and small ASV Shannon diversity among material type and week for each size class among material type and week for each size class.Values in bold indicate significant P-values.