The authors have declared that no competing interests exist.
Conceived and designed the experiments: SO AMO. Performed the experiments: SO. Analyzed the data: SO MBD. Contributed reagents/materials/analysis tools: MBD AMO. Wrote the paper: SO MBD AMO.
Plastic debris pervades in our oceans and freshwater systems and the potential ecosystem-level impacts of this anthropogenic litter require urgent evaluation. Microbes readily colonize aquatic plastic debris and members of these biofilm communities are speculated to include pathogenic, toxic, invasive or plastic degrading-species. The influence of plastic-colonizing microorganisms on the fate of plastic debris is largely unknown, as is the role of plastic in selecting for unique microbial communities. This work aimed to characterize microbial biofilm communities colonizing single-use poly(ethylene terephthalate) (PET) drinking bottles, determine their plastic-specificity in contrast with seawater and glass-colonizing communities, and identify seasonal and geographical influences on the communities. A substrate recruitment experiment was established in which PET bottles were deployed for 5–6 weeks at three stations in the North Sea in three different seasons. The structure and composition of the PET-colonizing bacterial/archaeal and eukaryotic communities varied with season and station. Abundant PET-colonizing taxa belonged to the phylum Bacteroidetes (e.g. Flavobacteriaceae, Cryomorphaceae, Saprospiraceae—all known to degrade complex carbon substrates) and diatoms (e.g. Coscinodiscophytina, Bacillariophytina). The PET-colonizing microbial communities differed significantly from free-living communities, but from particle-associated (>3 μm) communities or those inhabiting glass substrates. These data suggest that microbial community assembly on plastics is driven by conventional marine biofilm processes, with the plastic surface serving as raft for attachment, rather than selecting for recruitment of plastic-specific microbial colonizers. A small proportion of taxa, notably, members of the Cryomorphaceae and Alcanivoraceae, were significantly discriminant of PET but not glass surfaces, conjuring the possibility that these groups may directly interact with the PET substrate. Future research is required to investigate microscale functional interactions at the plastic surface.
Plastic pollution was first reported in remote, offshore basins of the north Atlantic ocean over forty years ago [
As the spatial distribution of marine plastic debris continues to be better resolved, research is increasingly focused on assessing its effects on environmental and public health. The impacts of plastic pollution range from organismal, such as morbidity and mortality due to entanglement and intestinal blockage [
Plastic surfaces in seawater can form microbial biofilms visible by eye within one week and cause a physical change, with a significant increase in plastic hydrophilicity and a shift from positive towards neutral buoyancy after 2 weeks [
The first two studies to use high-throughput sequence analysis of 16S ribosomal RNA genes to describe microbial biofilm communities from open ocean [
The present study examines the influence of season, geographic location, seawater, and substrate material type on the microbial colonization of single-use plastic drinking bottles composed of poly(ethylene terephthalate) (PET) deployed at multiple stations in the North Sea. PET is a semicrystalline thermoplastic polyester that is commonly used in textiles and food and beverage packaging and comprises 50% of synthetic fiber production worldwide [
An exposure experiment was designed to investigate spatial and seasonal dynamics of microbial biofilm communities attached to plastic fragments in marine waters. Six poly(ethylene terephthalate) (PET) drinking water bottles (Evian, 500 ml) were attached to three SmartBuoys [
(a) Map of SmartBuoy locations Dowsing (purple), Warp (green) and Gabbard (red), including dominant regional current systems (current flow information modified after [
Upon sample collection from the buoys, the PET samples were cut from the bottles using sterilized scissors and glass slides were retrieved whole. PET and glass were rinsed with sterile seawater. Plastic fragments of approximately 10 cm2 were cut from the bottles using sterilize scissors for DNA extraction. The plastic and glass samples were treated to remove ambient seawater, residual debris and non-attached organisms. The plastic samples were cut with sterilize scissors into pieces to fit in 25 ml tubes and centrifuged for 30 sec at 5,000 rpm. Glass samples were manually agitated in sterile seawater and briefly shaken off after washing. The plastic and glass samples were stored at -20°C until further analyses.
Seawater samples were collected at each station in summer (Sep 2012; due to sampling instrument access, water collection was possible in summer only). Niskin bottles attached to a sampling rosette were used to collect surface water (1 m) at each station. Collection bottles were rinsed three times with station water from the Niskins before holding water to be filtered. Water was then serially filtered in triplicate through 3 μm (1 liter) and 0.22 μm (500 ml) filters (47 mm, cellulose acetate). Filters were stored at -20°C until further analyses.
No privately owned or protected land requiring specific permits was accessed and no protected or endangered species were collected in this field sampling.
For qualitative assessment of biofilm structure, a random collection of 8 samples across all seasons and sites was chosen for scanning electron microscopy (SEM) and prepared for microscopy, modified after [
DNA was extracted from exposed PET bottles, glass slides, and filters. PET (0.5 g per sample) and filters (cut into several pieces) were transferred to a 2 ml tube containing TSE buffer (50 mM Tris, 6.7% sucrose, 1 mM EDTA; 700 μl total), incubated in lysozyme (0.3 mg/ml final; 37°C for 30–60 mins), then incubated at 50°C for 60 mins in Tris-EDTA (50 mM Tris, 250 mM EDTA, pH 8; 74 μl) and SDS buffer (20% [w/v] SDS, 20 mM EDTA, 50 mM Tris, pH 8; 44 μl). Tubes were centrifuged (8,000 g, 10 mins) to remove lysed biofilm and cellular material, the supernatant (approx. 920 μl) transferred to a new tube and 1/10 volume of NaCl and 1 volume phenol:chloroform added. Phenol:chloroform extraction and isopropanol precipitation were performed as described previously [
Dual-indexing was used to generate a barcoded MiSeq library of tag sequences (V4 region of 16S and V9 region of 18S rRNA genes, for bacteria and eukaryotes, respectively (see
PCR amplification was carried out in 96 well plates, with each well containing 17 μl Accuprime Pfx Supermix (Life Technologies, NY, USA), 1 μL template DNA (20–100 ng; samples >100 ng were diluted 1:10, as the high yield inhibited PCR amplification) and 2 μl of each paired set of index primers [10 μM]. Sterile PCR grade water served as negative control and known DNA as positive control (16S:
Quality filtering of reads (permitted length = 225–275 bp for 16S, 100–180 bp for 18S sequences, maximum number of ambiguous bases per sequence = 0, maximum number of homopolymers per sequence = 8), taxonomy assignment (Bayesian classifier, reference database SSURef_119_SILVA, required bootstrap value ≥ 85%) and picking of operational taxonomic units, OTUs, (label = 0.03) were carried out using Mothur [
For beta diversity analysis and related hypothesis testing, samples with less than the 5th percentile of reads (<1110 reads for 16S, <800 reads for 18S) were removed from analyses, proportions of OTUs per sample were calculated (OTU count/total OTUs in sample), proportions were square root-transformed, and scaled by the minimum library size per subset. This work was performed in R [
Permutational Multivariate Analysis of Variance, PERMANOVA [
Total community OTU representations across treatments were depicted in a taxonomic framework by importing OTU tables with their Silva [
To identify OTUs that discriminate the treatments, stations, and seasons, we used a linear discriminant analysis effect size method (LEfSe; [
For the metadata analyses, three environmental parameters collected from the SmartBuoys (temperature, salinity, chlorophyll fluorescence) were collated across each season for each station. The average values were calculated for each station-season combination; salinity data for Dowsing-Summer were not available. A Mantel test (R; mantel, vegan package [
To determine whether microbes colonizing plastic are distinct from those in seawater and on other inert hard substrates, microbial community structure of PET biofilms was compared to free-living (0.22–3 μm) and particle-associated (>3 μm) communities in seawater and those colonizing glass. The 16S rRNA gene sequence comparisons showed a significant difference between the PET-colonizing and free-living bacterial/archaeal seawater communities (p = 0.009; pairwise PERMANOVA,
PCOs representing similarity of biofilm communities based on counts of OTUs across samples (16S/18S rRNA gene data, see
Subset: Factors | marker | df | SS | Pseudo F | p(perm) | Unique perm | p(MC) | p(PERMDISP) |
---|---|---|---|---|---|---|---|---|
PET: Station | 16S | 2 | 10987 | 2.5436 | 999 | 0.542 | ||
18S | 2 | 10258 | 2.8614 | 999 | 0.632 | |||
PET: Season | 16S | 2 | 14855 | 3.7968 | 997 | 0.281 | ||
18S | 2 | 9542.3 | 2.5881 | 999 | 0.794 | |||
Summer: Treatment (PET-3-0.2) | 16S | 2 | 9435.2 | 3.0775 | 991 | 0.125 | ||
18S | 2 | 12006 | 3.4363 | 905 | 0.956 | |||
Spring: Treatment (PET-glass) | 16S | 1 | 3746.9 | 1.9241 | 0.058 | 980 | 0.057 | 0.841 |
18S | 1 | 1968.7 | 1.1056 | 0.29 | 905 | 0.344 | 0.894 | |
Summer: Station (Warp-Gabbard-Dowsing) | 16S | 2 | 10097 | 3.4277 | 994 | |||
18S | 2 | 9426.7 | 2.2778 | 937 | 0.38 | |||
Spring: Station (Warp-Gabbard-Dowsing) | 16S | 2 | 9373.4 | 2.7822 | 997 | 0.848 | ||
18S | 2 | 12674 | 6.111 | 997 | 0.573 | |||
Summer: “Attached” versus Free-living | 16S | 1 | 6692.7 | 4.0966 | 843 | 0.124 | ||
18S | 1 | 5358.4 | 2.3385 | 416 | 0.084 |
PERMANOVA main tests compare both bacterial/archaeal and eukaryotic (16S and 18S rRNA gene, respectively, denoted by ‘marker’) community structure across seasons, stations, and treatments. Tests are displayed for three data subsets (PET, spring, summer). Significant results (p < 0.05) highlighted in bold and marked with *. P-values were obtained using type III sums of squares and 999 permutations [‘p(perm)’] or calculating Monte-Carlo tests [‘p(MC)’]. Pseudo F, PERMANOVA F statistic; d.f., degrees of freedom; SS, sums of squares; Unique perm, unique permutations. p(PERMDISP) are p-values of PERMDISP tests, calculated to centroids.
As with the bacterial and archaeal communities, the eukaryotic communities colonizing the PET significantly differed from the free-living seawater communities (pairwise PERMANOVA, p = 0.031,
The PET communities contained several hundreds of different operational taxonomic units (OTUs) per sample. OTUs assigned to the families Flavobacteriaceae, Cryomorphaceae, Saprospiraceae, and Rhodobacteraceae (Figs
Most abundant (top 25) bacterial and archaeal families present in PET-attached biofilm communities after deployment in the North Sea, grouped per deployment site/station and season (based on 16S rRNA gene analysis). * Gabbard represents data for winter and spring only; ** summer represents Warp and Dowsing data only; Gabbard summer was removed from analysis due to insufficient sequencing effort.
Phylogenetic representation (based on 16S rRNA gene-based taxonomy assignment) of abundant OTUs (>0.5% of at least one community) and their relative abundances (pie charts based on log-scaled OTU counts) across treatments in summer.
Multiple OTUs and taxonomic groups significantly discriminated the plastic biofilm communities from the seawater communities. The orders Sphingobacterales (Bacteriodetes) and Myxococcales (Deltaproteobacteria) significantly discriminated the PET-attached communities from either fraction of seawater across all stations (
Despite the lack of significant difference between PET and glass total communities, differences were observed at the OTU level. OTUs assigned to the genera
Representation of taxa significantly discriminant of either PET- or glass-attached communities across all stations after 5–6 weeks incubation in the North Sea. See
Eukaryotes were readily visible living attached to the PET surface (
Most abundant (top 25) eukaryotic families present in PET-attached biofilm communities after deployment in the North Sea for 6 weeks, grouped per deployment site/station and season (based on 18S rRNA gene analysis).
Fungal taxa were noticeably more prevalent in the hard surface-colonizing (plastic and glass) biofilms than in either fraction of seawater (
Discriminant of the PET communities as compared to seawater communities was an OTU belonging to the Bacillariophytina (super-group Stramenopiles-Alveolates-Rhizaria, SAR). The ciliate class Spirotrichea (SAR) was discriminant of the free-living seawater community. Several OTUs belonging to the Dinoflagellata (SAR) and
Although the overall variation between glass- and PET-attached eukaryotic communities was not significant (p = 0.29,
A Mantel test between the 16S and 18S OTU dissimilarity matrices indicate that the composition of the bacterial/archaeal and eukaryotic community assemblages were significantly positively correlated (r-statistic 0.901; p = 0.0001). A Mantel test between the Euclidean distance matrices of water properties (T, S, F;
The community structure of PET-colonizing biofilms was influenced by both spatial and seasonal factors. PET-colonizing bacterial/archaeal (16S) communities were significantly different between stations and seasons. Between-station comparisons indicated that PET communities sampled at the geographically distant Dowsing were significantly distinct from those at both Warp and Gabbard (p = 0.001 and 0.002, respectively). The difference between these neighboring stations, Warp and Gabbard, was not significant (p = 0.095). Significant differences were found between all seasons (all p = 0.001). These relationships were visualized in the sample clustering patterns in principle coordinate (PCO) ordination (
Eukaryotic community assemblages followed similar patterns as was seen for bacterial/archaeal communities. The PET-attached communities varied significantly with season (p = 0.002;
Plastic-attached communities were different from ‘free-living’ seawater communities, but not distinct from particle-associated or glass-attached communities. This is the first time this observation has been reported, as previous studies have restricted their analyses to comparisons between marine plastic biofilms and free-living seawater communities only [
In contrast with our second hypothesis, the PET-colonizing biofilms were not significantly distinct from those found on glass biofilms. Previous studies support these findings, in that no significant differences were found between communities that formed on “hard” substrates in freshwater [
Despite the similarities between PET and particle associated communities, a few notable OTUs uniquely discriminated PET from the other “attached” (3 μm and glass) communities across all stations (
Two genera and one family were present on PET (>0.5%;
Members of the Verrucomicrobia phylum were identified in both the attached (plastic and >3 μm) and free-living microbial communities and their distribution indicated that ecologically distinct lineages within this diverse phylum occupy unique niches in the marine environment. In this study, the Verrucomicrobiae family (Verrucomicrobia subdivision 1) was significantly discriminant of PET-attached communities and was also abundant on plastic collected from the Northern Atlantic [
Bacterial families that comprised >0.5% of the communities on our deployed North Sea plastics and that were described previously on North Atlantic plastic fragments [
Notably, members of the Rhodobacteraceae and Alteromonadaceae are known hydrocarbon degraders [
Roughly one third of the plastic-associated bacterial or archaeal sequences identified in this study were assigned to the genus
Diatoms have been found attached to surfaces of ocean plastics [
Fungi represent an unexplored component of the aquatic plastic microbiome. We found a high prevalence of fungal OTUs on plastic and glass relative to either fraction of seawater, though the minimal information in the short 18S rRNA V9 region used in this study offers little taxonomic resolution and insight into the fungal populations (
Consistent with our hypotheses, biofilms formed on plastics from different sites had significantly different community structures. However, the microbial community similarity followed a trend in the physical proximity of the stations. The biofilm and seawater communities at the proximal Warp and Gabbard stations were not significantly different from each other, yet they both differed significantly from those at the more distant Dowsing station (
To specifically determine the extent to which the seasonal and spatial factors uniquely influence the PET biofilm communities relative to the background seawater and glass-associated communities, one would need to compare these three communities across all seasons and at all study sites. Due to technical challenges that impeded our study design (
Taken together, the seasonal and spatial patterns suggest that the plastic surface environment at the polymer-water interface does not exert strong enough selection to drive species sorting to overcome other niche-defining factors. If the plastic selects for some unique microbial constituents (e.g., polymer-consuming microbes), the approach used here to study the overall community may not provide sufficient resolution, as plastic-influenced organisms may be minor community members. As biofilms matured over a 6-week incubation, only the initial recruits would have direct contact with the polymer surface; later recruits are more likely to interact with existing biofilm members and the abiotic components of the biofilm matrix or surrounding seawater. Such a scenario would result in a generalized marine particle/surface-associated community when the community members are studied en masse, as was done and found here.
Understanding plastic-dwelling biofilms at increased spatial resolution will improve our knowledge of biofilm community assembly processes, as well as plastic-microbe interactions, such as the potential for members of the plastic microbiome to degrade plastic hydrocarbons. Future studies should include other major plastic polymers in addition to PET, e.g., polyethylene, polypropylene, polystyrene or polyamide and may benefit from longer incubation periods (6–12 months minimum) to allow for development of possible degrading populations. Three-dimensional confocal fluorescence
The rate of plastic debris accumulation in aquatic ecosystems is increasing and will persist for long time scales [
Identified bacterial/archaeal genera (16S rRNA gene) comprising PET-attached biofilms across all stations and seasons sampled. OTU counts have been square root transformed.
(PDF)
Phylogenetic representation (based on 16S rRNA gene-based taxonomy assignment) of abundant OTUs (>0.5% of at least one community) and their relative abundances (pie charts based on log-scaled OTU counts) attached onto PET and glass substrates in spring.
(PDF)
Eukaryotic biofilm members living at the surface of a PET plastic bottle after incubation for 5–6 weeks in the coastal North Sea. (a) Diatom members of PET-colonizing community. (b) A mass of interacting eukaryotes (diatoms, algae, possible ciliates) within the PET-colonizing biofilm community.
(PDF)
Bar graph representing the abundance of reads assigned to fungal OTUs across all treatments (PET-attached, glass-attached, 0.2–3 μm seawater, >3 μm seawater). OTU counts are normalized to the number of samples of each treatment to account for unbalanced representation of each sample type.
(PDF)
PCOs representing similarity of biofilm communities based on counts of OTUs across samples (16S/18S rRNA gene data, see
(PDF)
Average temperature, salinity, and chlorophyll fluorescence data for each season-station pair used in the dissimilarity matrix for Mantel test. Data were continuously collected by the SmartBuoy system at each station the substrates were deployed and the water was collected. Dowsing-Summer dropped from Mantel analysis, as no data were available for salinity.
(XLSX)
See Kozich et al, 2013 for dual indexing strategy.
(XLSX)
PERMANOVA was performed on Bray-Curtis dissimilarity matrices based on OTU counts across microbial communities (a: bacterial/archaeal, 16S rRNA gene; b: eukaryotic, 18S rRNA gene). Significant results (P < 0.05) highlighted in bold and noted by **. Factors include station (Warp, Gabbard, Dowsing), treatment (PET, >3 μm, 3–0.2 μm seawater fractions in summer) and season (spring, winter, summer). P-values were obtained using type III sums of squares and 999 permutations [‘p(perm)’] or calculating Monte-Carlo tests [‘p(MC)’]. Unique perm, unique permutations. The results (p values) of the PERMDISP tests calculated to centroids are also provided.
(XLSX)
Discriminant OTUs (based on16S rRNA gene analysis) identified, using class and subclass distinctions, in comparisons of: (a) PET-associated, particle-attached (>3 μm) and free-living (0.2 μm-3 μm) seawater communities in summer (lefse class: treatment) across all stations (lefse subclass: station), (b) PET- and glass-attached communities in spring (lefse class: treatment) across all stations (lefse subclass: station), (c) PET-associated communities from summer, spring, winter (lefse class: season) across all stations (lefse subclass: station), (d) PET-associated communities from Dowsing, Warp, Gabbard (lefse class: station) across all seasons (lefse subclass: season), (e) attached (PET and >3μm seawater fraction) and free-living (3–0.2 μm seawater fraction) communities (PET and 3.0 μm; lefse class: treatment) across all stations (lefse subclass: station).
(XLSX)
Discriminant OTUs (based on18S rRNA gene analysis) identified, using class and subclass distinctions, in comparisons of: (a) PET-associated, particle-attached (>3 μm) and free-living (0.2 μm-3 μm) seawater communities in summer (lefse class: treatment) across all stations (lefse subclass: station), (b) PET- and glass-attached communities in spring (lefse class: treatment) across all stations (lefse subclass: station), (c) PET-associated communities from summer, spring, winter (lefse class: season) across all stations (lefse subclass: station), (d) PET-associated communities from Dowsing, Warp, Gabbard (lefse class: station) across all seasons (lefse subclass: season), (e) attached (PET and >3μm seawater fraction) and free-living (3–0.2 μm seawater fraction) communities (PET and 3.0 μm; lefse class: treatment) across all stations (lefse subclass: station).
(XLSX)
We acknowledge the Centre for Environment, Fisheries & Aquaculture Science (CEFAS) for operating and allowing access to the SmartBuoys and their collected environmental data, in particular Naomi Greenwood, the SmartBuoy team and the crew of RV Endeavour. The contribution of CEFAS took place within the SmartBuoy and Wavenet programs funded by the Department for Environment, Food and Rural Affairs (DEFRA). Gunnar Gerdts for advice on concept and experimental design, Leonie Elie for guidance on the scanning electron microscope, the Denef, Dick, and Schloss Labs at the University of Michigan for technical support regarding sequence library prep and analysis, and Thomas Jenkinson for providing