Marine biofilms on different fouling control coating types reveal differences in microbial community composition and abundance

Abstract Marine biofouling imposes serious environmental and economic impacts on marine applications, especially in the shipping industry. To combat biofouling, protective coatings are applied on vessel hulls which are divided into two major groups: biocidal and non‐toxic fouling release. The current study aimed to explore the effect of coating type on microbial biofilm community profiles to better understand the differences between the communities developed on fouling control biocidal antifouling and biocidal‐free coatings. Biocidal (Intersmooth® 7460HS SPC), fouling release (Intersleek® 900), and inert surfaces were deployed in the marine environment for 4 months, and the biofilms that developed on these surfaces were investigated using Illumina NGS sequencing, targeting the prokaryotic 16S rRNA gene. The results confirmed differences in the community profiles between coating types. The biocidal coating supported communities dominated by Alphaproteobacteria (Loktanella, Sphingorhabdus, Erythrobacter) and Bacteroidetes (Gilvibacter), while other taxa, such as Portibacter and Sva0996 marine group, proliferated on the fouling‐release surface. Knowledge of these marine biofilm components on fouling control coatings will serve as a guide for future investigations of marine microfouling as well as informing the coatings industry of potential microbial targets for robust coating formulations.

biofouling in the shipping industry, fouling control coatings are applied to ships' hulls where the biofilms first attach (Finnie & Williams, 2010). These commercial fouling control coatings are either biocidal antifouling or non-biocidal fouling-release coatings.
Biocidal antifouling coatings function through the release of certain toxic chemicals (biocides) to deter the settlement and growth of organisms. Biocidal coatings remain the most popular choice and still dominate the market reportedly accounting for more than 90% of coatings sales (Lejars et al., 2012;Winfield et al., 2018), although concerns over the potential environmental impact of biocides have led to increased attention being paid to the development of biocide-free approaches to fouling control (Lejars et al., 2012).
Non-biocidal fouling-release coatings function based on low surface energy, smooth and non-porous, free of reactive functional groups (Finnie & Williams, 2010) which reduces an organism's ability to generate a strong interfacial bond with the surface (Chambers et al., 2006;Lejars et al., 2012). Thus, such coatings minimize the adhesion strength of organisms and facilitate their removal either by hydrodynamic forces (water flow) as the vessel moves or through their organisms' weight (Swain, 1999). Both coating types have been shown to bear similar low levels of fouling after extended immersion if exposed to flow levels comparable to slow ship speeds, with the fouling-release coating losing most accumulated biomass in flow, consistent with its low adhesion mechanism of action (Davidson et al., 2020). Fouling-release coatings have a smaller market share when compared to biocidal since they generally require flow to be effective against biofouling (Briand et al., 2012;Molino et al., 2009).
Although the coating industry has an increasing interest in the development of biocide-free (micro)fouling control solutions that rely on surface physico-chemical properties, the development of a successful marine coating that is simultaneously effective against biofouling while being substantially environmentally benign is very challenging.
Biofilm research is important to the marine coating industry as it directly provides insights into the response of biofilm communities on coating surfaces and consequently may inform the development of new paint technologies. Several studies investigated the effect of fouling control coatings on in situ biofilm community composition either by employing light and epifluorescent microscopy (Cassé & Swain, 2006), or molecular fingerprinting and microscopic observations (Briand et al., 2012), or flow cytometry coupled with denaturing gradient gel electrophoresis and light microscopy (Camps et al., 2014); all of which have reported that the observed biofilm community compositions were influenced by coating type.
To date, only a handful of studies have reported the application of next-generation sequencing (NGS) techniques to investigate the response of marine biofilm community profiles developed on marine fouling control coatings (von Ammon et al., 2018;Briand et al., 2017;Ding et al., 2019;Dobretsov et al., 2019;Flach et al., 2017;Hunsucker et al., 2018;Leary et al., 2014;Muthukrishnan et al., 2014;Winfield et al., 2018). NGS technology based on sequencing ribosomal RNA genes is appropriate for a range of applications including highly diverse community analysis while offering a large volume of data that allow for statistical testing (Fukuda et al., 2016).
Despite the current knowledge, certain aspects of biofilm research on fouling control coatings remain elusive. Differences in biofilm profiles between biocidal and fouling control coatings can help to highlight potential targets of importance for effective antibiofilm control, as well as identifying potential biocidal-tolerant biofilm components at low taxonomic levels.
The aims of the present study are (1) to explore and characterize marine biofilm communities isolated from commercial fouling control coatings using 16S rRNA gene amplicon sequencing and (2) to compare the biofilm profiles developed between fouling release and biocidal coating types. To reflect the biofilm formation based on state-of-the-art analyses and study design, a combination of biocidal antifouling coating, fouling-release coating, and reference surfaces were used, testing in situ four biological replicates of biofilms using Illumina MiSeq sequencing targeting the V4-V5 region of the 16S rRNA gene to examine bacterial composition.
The purpose of this work is to elucidate biofilm components at the genus level that are selectively attached on biocidal and/or fouling-release surfaces. The study findings will contribute knowledge into the growing body of NGS studies of biofilms on fouling control paints and subsequently inform the future design of fouling control surfaces.

| Commercial fouling control coatings
Three treatments were exposed during the immersion study including (1) a commercial biocidal antifouling coating which will be termed as "BAC" (Intersmooth® 7460HS SPC, self-polishing copolymer coating that contains cuprous oxide and copper pyrithione biocides), (2) a commercial non-biocidal fouling-release coating which will be termed as "FRC" (Intersleek® 900, fluoropolymer), and (3) a nonbiocidal inert surface termed as "PDMS" (silicone paint film incorporating a generic unmodified polydimethylsiloxane matrix). A red pigmentation was incorporated in all coated panels to minimize the potential influence of surface color on community variation. Details of all surfaces are presented in Appendix Table A1.
Experimental panels were prepared by brush application at the International Paint Laboratories in Gateshead UK following the correct scheme for each coating type (BAC: anticorrosive primer plus finish coat; FRC, PDMS: anticorrosive primer, silicone tie coat, finish coat).
The panels were double-side-coated with dimensions 8.5 × 8.5 cm 2 .

| Panel deployment and study site
Experimental panels were attached to a metal frame using cable ties and deployed to the anchored University of Portsmouth (UoP) raft (50°48'23.4"N 1°01'20.1"W) in Langstone Harbour, UK. Frames were immersed vertically to the seawater surface at 0.5 -1 m depth for 119 days from April 6 until August 3, 2018 (Appendix Figure A1).
The sampling location is characterized as a semi-diurnal system, where two high and two low tides take place every 24 hours. It has a temperate climate moderated by prevailing southwest winds and significant rainfall. Langstone Harbour entrance is characterized by maximum spring tide flood velocities of 0.9 ms −1 (1.7 knots), ebb velocities 1.8 ms −1 (3.5 knots), and mean flood tidal stream velocity of 0.7m s−1 (1.4 knots) (www.scopac.org.uk, www.eocea nic.com).

| Biofilm sample collection and storage
The biofilms samples were collected (n = 4 per coating) from panels using sterile swabs (Appendix Figure A2). Macrofoulers were removed from heavily fouled panels using sterile forceps. The swab was passed 10 times over the panel with circular movements for biofilm collection. During sampling, the frames were manually removed from the seawater and exposed to air during collection, for approximately 5-15 min. Between sampling, all panels were hydrated with surrounding seawater. After biofilm collection, each swab was placed into a sterile Eppendorf tube and the breakpoint was cut out using sterile scissors. Samples were then immediately snap-frozen in liquid nitrogen (in the field), transferred to the laboratory, and stored at −80 ˚C within 4 hours. DNA extraction took place within 2 months of sampling.

| DNA extraction and quantification
Genomic DNA (gDNA) was extracted using the DNeasy PowerBiofilm Kit (QIAGEN). The samples were transferred from −80 ˚C to room temperature. In a laminar flow hood, each biofilm swab sample was placed into a PowerBiofilm Bead Tube using sterile forceps. Qiagen's protocol (DNeasy® November 2016) was followed according to the manufacturer's instructions, except the first step was omitted, since the saturated biofilm material was attached to the swab; therefore, no weighing and centrifugation was applicable. To beadbeat, the sample, a PowerLyzer 24 Homogenizer (MP Biomedical, FastPrep-24™ 5G) was used. At the final step, extracted DNA was eluted following the manufacturer's instructions and stored in −80 ˚C. The quantity and partial quality of nucleic acid samples were assessed based on absorbance spectrums using a spectrophotometer (Thermo Scientific, NanoDrop 1000).
The generated OTU table was then analyzed using the R programming language (version 4.0.2) (R Core Team, 2020). The phylogenetic analysis was implemented using the phyloseq package (McMurdie & Holmes, 2013) available as part of the Bioconductor project (Gentleman et al., 2004), which supports OTU-clustering formats and provides ecology and phylogenetic tools. Sequences detected with high similarity to chloroplast and mitochondria from the eukaryotic component of the community were removed from the analysis. Plots were generated using the ggplot2 library (Wickham, 2016).

| Statistical analyses
Statistical tests were performed in R. The significance of coating type on the resulting diversity indices (Chao1, Shannon) was assessed by ANOVA (sum of squares type II), followed by the estimated marginal means (EMMs) to identify significant differences between pairwise comparisons.
Biofilm community structure (relative abundance) of phyla, classes, families, and genera was evaluated for changes between coating types using analysis of similarities (ANOSIM) (Clarke, 1993) in the vegan R package (Oksanen et al., 2019) with Bray-Curtis of 9999 permutations. To determine finer resolution taxa (genus level) that significantly contribute to differences between coating samples diversity (shown in ANOSIM), similarity percentages analysis (SIMPER) (Clarke, 1993) in vegan was performed using kruskal.pretty function (Steinberger, 2016)

| Alpha diversity
The alpha diversity indices were calculated after rarefication to 15,000 OTU depth (per sample) ( Table 2  The alpha diversity indices calculated at lower sub-sampling depths, that is, 10,000 and 1,000 displayed consistent patterns with the TA B L E 1 Characteristics of replicated biofilm samples including the sample type where biofilms were collected from, DNA concentration and quality ratios, the number of reads retrieved and assigned OTUs  (Table 1).
These contrasting results confirm that the low diversity of BAC samples in the present dataset is not a result of potential low sequence coverage, but rather the presence of few very abundant biofilm taxa.
A significant difference between Chao1 estimates in different treatments at the 15,000 OTU depth was shown (p = 0.0003***,

| Beta diversity
The principal coordinates analysis (PCoA) plot of the relative abundance of OTUs across the dataset revealed distinct communities in BAC samples, while FRC and PDMS biofilm communities showed significant overlap (Figure 2). This PCoA plot captures 45.8% of the variation in relative abundance across the dataset, with differences between BAC and both FRC and PDMS samples accounting for the majority (34.6%).

| Core biofilm microbiome
Particular groups that contribute to similarities (shared) and differences (distinct) between treatments were quantified and illustrated at the genus level. More specifically, OTU genera are shown with a 0% threshold regardless of their abundance in the dataset  In terms of abundant genera (representing ≥1% of the community), only 4 taxa were seen in common between all treatments, while the BAC samples showed the greatest number of surfacespecific genera with 9 ( Figure 3b). Therefore, the biofilm community present in BAC samples potentially contributed to the total dataset with less diverse but highly abundant genera (9 out of 18), as highlighted by the low alpha diversity measures (Table 2).
Overall, the core community of unique OTUs shared between all samples consisted of diverse genera (60) (Figure 3a), with only F I G U R E 1 Alpha diversity estimates including the observed (unique OTUs), Chao1, and Shannon indices. Alpha diversity scores are plotted for the four replicates of each coating type. Samples are colored by coating type, each of the four replicates is indicated by a different symbol F I G U R E 2 Principal coordinates analysis (PCoA) plot of the relative abundance across OTUs for each coating sample including PDMS, FRC, and BAC from 16S rRNA amplicon sequencing analysis. Variations in the dataset are explained by 34.6% with the first principal coordinate axis (PCoA 1) and 11.2% with the second axis (PCoA 2) F I G U R E 3 Venn diagrams representing the number of unique genera identified across OTUs identified with a relative abundance greater than (a) 0% or (b) 1% on each coating type from 16S rRNA amplicon sequencing. The overlap represents genera seen amongst the community of multiple surfaces a small fraction of them (4) contributing with 1% abundance to the core community of abundant OTUs (Figure 3b). These results signify that the differences between surface types are defined from a few taxa that are abundant in this biofilm community.
To confirm that shared genera between surfaces were not the result of contamination between samples, a similar plot was generated for all three surfaces at the OTU level with greater than 1% relative abundance (Appendix Figure A4)
The genera that significantly contribute to these differences in beta-diversity among coating types were determined by SIMPER analysis (Table 3). In total, 24 OTU biofilm genera changed with coating type (SIMPER contribution >1.5%, Kruskal p value <0.05).
Statistical differences were significantly driven by Loktanella

| Reported marine biofilm taxonomic profiles on fouling control coatings
The dominant phyla of the examined marine biofilms on the panels coated with two commercial fouling control coatings (BAC, FRC) and one inert surface (PDMS) were Proteobacteria, Bacteroidetes, Planctomycetes, Actinobacteria, Cyanobacteria, and Verrucomicrobia (Appendix Figure A5). Bacteria belonging to the classes of Alphaproteobacteria (33-47%), Bacteroidia (19-25%), and Gammaproteobacteria (16-20%), were the most dominant across all samples (Figure 4), individually contributing to more than 16% of the total biofilm community for each coating type. Deltaproteobacteria Although not frequently reported, Verrucomicrobia has also been found in fouling control studies (Leary et al., 2014;Winfield et al., 2018) and was confirmed to be abundant in all three coating treatments (>1.8%) in this study. The high abundance of other taxa at the genus level could either be attributed to the presence of diverse rare taxa or to the lack of alignment of certain taxa in the database, however, that was not observed at a higher taxonomic level ( Figure 4).  (Dobretsov et al., 2019).
Sequences belonging to the genus Erythrobacter, which were found in high abundance in this study (1.8% -5%), were previously identified on biofilms from two moving ships traveling from Norfolk North and Baltic Seas (7.7%), and Norfolk to Rota, Spain (21.3%) (Leary et al., 2014), as well as in biofilms on panels coated with cuprous oxide-containing antifouling paints (Muthukrishnan et al., 2014) and biofilms on a coated ocean glider off the coast of Muscat, Oman (Dobretsov et al., 2019). Sphingorhabdus (class Alphaproteobacteria, family Sphingomonadaceae) was also abundant in BAC samples (3.6%); nevertheless, it was absent from the PDMS or FRC samples.

| Differences between the biofilm communities on BAC and FRC coatings
The biofilm community profiles in the present study revealed major differences in OTU relative abundance and richness between the two fouling control coating treatments. Biofilm community structure was found significantly different between BAC and FRC samples for all taxonomic levels tested with ANOSIM. The differences between sample communities on the two fouling control coatings that resulted from SIMPER analysis (Table 3) were mainly driven by Loktanella, Gilvibacter, Sphingorhabdus, and Erythrobacter; sequences with high similarity to these taxa were found abundant in BAC samples, as shown in Figure 5. Additionally, SIMPER analysis illustrated TA B L E 3 The significant contribution (SIMPER % >1.5%, Kruskal p-value <0.05) of biofilm genera to the total similarity percentages between the different coatings revealed with SIMPER analysis that Portibacter and Sva0996 marine groups which were abundant on the FRC surface ( Figure 5), constituted key components defining the different community profiles between the two fouling control coatings.
Biofilm communities found on FRC panels were similar to those on PDMS surfaces, while BAC biofilms exhibited a distinct response, as indicated by sample clustering in the PCoA plot ( Figure 2). The highest biofilm diversity indicated by all diversity indices (Table 2) was found on the PDMS and FRC surfaces. The biofilm profile of BAC panels was characterized by a lower diversity and a higher relative abundance of the present taxa.
In terms of the BAC biofilm community profile, the higher relative abundance may be due to the relative proliferation of a few biocide-tolerant taxa or may be a result of species competition which shifted the community composition. The observed relatively high abundance of few taxa in BAC biofilms is consistent with earlier (microscopic) investigations of biofilm composition and relative abundance in samples from fouling release and biocidal antifoulingcoated surfaces that revealed lower abundance and higher diversity in samples from fouling-release surfaces (Cassé & Swain, 2006  antifouling-coated surfaces (Leary et al., 2014;Muthukrishnan et al., 2014). The present study shows the opposite since Cyanobacteria (class Oxyphotobacteria) detected sequences dominated PDMS (4.6%) and FRC (1.39%) surfaces, contrary to BAC (0.1%). It has to be noted that high dominance of Cyanobacteria on BAC coatings has been suggested after 1 year of immersion in Oman (Muthukrishnan et al., 2014) and after 7 months on two moving vessels crossing the North and Baltic Seas, and North-East Atlantic Ocean, respectively (Leary et al., 2014). Here, Cyanobacteria were not abundant on BAC that was exposed for 4 months in Langstone Harbour, UK.
Certain bacterial genera such as Loktanella and Gilvibacter, which possibly exhibit tolerance to biocides contained in BAC, potentially reduced the settlement or growth of other organisms on BAC that were abundant in the other two surfaces (e.g., Portibacter) ( Figure 5).
In comparison with the PDMS, Portibacter was the only bacterial genus where the relative abundance was reduced in both coatings, BAC and FRC. On the BAC panels, two factors that are possibly involved in shaping the shifted community are the performance of the biocidal paint and the interplay between biofilm components at certain conditions (e.g., biocidal release rate, environmental conditions, antagonistic relationships).

| Study design suggestions for biofilm research on fouling control surfaces
The current study has carefully implemented the most relevant design (four biological replicates were tested, with immediate biofilm storage in liquid nitrogen, targeting the V4-V5 region of 16S rRNA gene, using Illumina MiSeq NGS technology, sequence annotation against the SILVA SSU 132 database, etc.) to support the purpose of the study, as many factors during experimental design and data analysis could significantly impact the results-especially in a complex microbial community.
The V4-V5 region of the 16S rRNA gene has been one of the most broadly used variable regions in studies examining environmental biofilms on artificial surfaces (e.g., Bakal et al., 2018;Li et al., 2017;Pereira et al., 2017), while 515F/926R has been suggested as a primer set that increases percentage detection of various prokaryotic taxa (Pollet et al., 2018) as well as been the most effective region in minimizing overestimation due to intragenomic heterogeneity (Sun et al., 2013).
For microbial community analyses, Illumina is the most widely used NGS platform, due to the large output and cost performance (van Dijk et al., 2018;Fukuda et al., 2016) which are indispensable in complex and diverse study systems. Illumina produces high throughput and short read length with a low error rate (de Sá et al., 2018).
The selection of 16S rRNA sequence reference database is an important element for taxonomic classifications; therefore, it is worth mentioning that only Briand et al., (2017) Dobretsov et al., 2019;Muthukrishnan et al., 2014) or Greengenes (Hunsucker et al., 2018;Winfield et al., 2018) databases. The SILVA database constitutes one of the most actively maintained and largest databases which includes curated 16S rRNA gene sequences (Quast et al., 2013;Yilmaz et al., 2014), while it has been suggested that it provides the lowest error rates compared to Greengenes and RDP (Lu & Salzberg, 2020).
It is worth highlighting that in biofilm studies on fouling control it is difficult to examine a "true" control to enable understanding the effect of specific coatings to the already-existing communities due to the extent of macrofouling. Moreover, the free-living microorganisms in the surrounding seawater at the time of sample collection could not serve as an indicator sample for comparison with mature biofilms developed on fouling control surfaces. The limited number of studies employed to date has examined biofilm composition on different types of fouling control coatings without testing a reference surface (Hunsucker et al., 2018;Winfield et al., 2018). It is also worth noting that a lack of negative controls in the present study represents a limitation of the study design since the presence of systematic contamination from the extraction and PCR stages cannot be identified and accounted for. However, the lack of systematic OTU abundance across the surfaces (Appendix Figure A4) suggests that contamination was not present at levels likely to affect the differential abundance analysis between the three major surfaces considered in this study.
In the present study, the generic unmodified PDMS coating was included as an inert surface to reflect the representative biofilm communities under the given conditions (e.g., location, season).
Unmodified PDMS is not suitable for commercial use as a fouling control product. However, it shares some surface characteristics with fouling-release coatings as an elastomeric material with a very smooth surface profile, and it demonstrates greater resistance to macrofouling compared to other unprotected artificial surfaces which is a useful pragmatic property for field studies. It is therefore advantageous to incorporate a non-toxic, inert, and macrofoulingresistant surface in fouling control research studies to (1) improve understanding of the microfouling communities that form with respect to coating properties, (2) better contextualize similarities and differences that arise between the complex biofilm communities that develop on different surfaces, and (3) discover the potential interplay between biofilm taxonomic components.
It is important to highlight that the fouling control coatings used in this study are designed primarily for use on the world's commercial shipping fleet, whose operational profiles typically involve alternating static periods in and around port and periods of active movement at sea. As expressed by Davidson et al., (2020), lay-up periods are common, inevitable, unavoidable, and of high significance because of the potential for fouling establishment. In their study of simulated lay-up periods followed by resumption of service, panels coated in biocidal and fouling release coatings showed similar levels of low fouling after exposure to flow at the lower range of ship speeds.
The final biofouling loads were attributable to lower initial fouling on the biocidal surface and the loss of fouling that had accrued on the fouling release surfaces, consistent with the modes of actions of the coating types (Davidson et al., 2020 Excellence in England (E3) grant.

CO N FLI C T S O F I NTE R E S T
None declared.

E TH I C S S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
The dataset generated and/or analyzed during this study are avail-   Figure A1 Environmental parameters throughout the 4-month coatings deployment, including (a) salinity (PSU), (b) pH, (c) temperature (˚C), (d) dissolved oxygen (mg/L). Records were concluded using sensor YSI 6820V2 sonde located in the Institute of Marine Science, University of Portsmouth Figure A2 Visual observation of endpoint biofilm samples before collection of fouling control coatings; PDMS, FRC, and BAC for DNA extraction. The gray squares indicate the individual replicate sample of each surface type Figure A3 Rarefaction curves of observed alpha diversity across a range of sub-sequencing depths for all replicates of PDMS, FRC, and BAC samples from 16S rRNA amplicon sequencing. In the present dataset, rarefaction curves for all replicates of PDMS, FRC, and BAC samples reached saturation level ( Figure S3), which indicated that the current sequencing depth (Table 1) was sufficient to provide a representative diversity for these biofilm samples. The results show that BAC curves reached a horizontal asymptote at a smaller depth (~15,000 OTUs); hence, it can be inferred that a smaller number of reads could potentially reflect a good representation of the total biofilm community diversity for BAC samples, compared with PDMS or FRC that require a higher number of reads Figure A4 Venn diagram showing the overlap between PDMS, FRC, and BAC surfaces at the OTU level with greater than 1% relative abundance from 16S rRNA amplicon sequencing