not-yet-known not-yet-known not-yet-known unknown [Figure 1] Protein extraction After incubation, protein was extracted from plastics and digested according to Messer et al. (2024). Plastics were removed from media and allowed to dry under laminar flow. Once dry, plastics were bead beaten (1mm glass beads) in a 2% sodium dodecyl sulphate (SDS) solution 3 times in 10-minute intervals. Cell media was then sonicated (1s pulse, 1s gap, 40% Amp.) twice, and centrifuged (8000 g, 10 minutes). The pellet was then sonicated and centrifuged again (as above). A quarter of the supernatant was preserved at -80℃ for metagenomic analyses (Section 2.4 ), and the remaining supernatant (5-6 ml) was transferred to 3K centricons and centrifuged at 7000 g until ≤ 250μl sample remained. Proteins were precipitated with cold acetone overnight at −80°C, with an acetone/aqueous protein solution ratio of 4:1. The protein pellet was resuspended in 6M Guanidine HCl in dipotassium phosphate buffer, sonicated (1s pulse, 1s gap, 40% Amp.), then centrifuged (13,000 g, 15 minutes). The supernatant was then diluted with LC-MS grade water, with a water/aqueous protein solution ratio of 1:1 and preserved at −80°C. Protein quantification revealed 35.13mg (±0.67; n = 4), and 6.68mg (±2.38; n = 4) of protein within the samples extracted after three (D3) and seven (D7) days incubation respectively. For protein digestion, 100μg of protein per sample was reduced (10mM 1,4-Dithioerythritol), alkylated (25mM iodoacetamide), and precipitated with acetone (acetone/aqueous protein solution ratio of 4:1) before being digested at 37°C overnight in 20μl of sequencing grade modified trypsin (EC 232-650-8) as described previously [16]. Protein samples were analysed on an ultra-high-performance liquid chromatography–high-resolution tandem mass spectrometer (UHPLC-HRMS/MS) with a Eksigent NanoLC 400 and AB Sciex TripleTOF 6600 system. Two micrograms of peptides were analysed using acquisition parameters previously reported [21] in data dependent acquisition (DDA) mode. Mass spectrometry (MS/MS) runs were conducted with micro injection (75 min LC separation) modes. Shotgun metagenomics DNA was co-precipitated from the plastisphere samples for metagenomic analysis by adding protein precipitation solution (PPS; Promega, A7951) with a sample /PPS solution ratio of 2:1, followed by centrifugation (20,379 g, 15mins). Filter-sterilised 3M sodium acetate (300μl) was added to the supernatant, followed by >95% cold ethanol at a sample/ethanol solution ratio of 1:2, and DNA was precipitated overnight (−20°C). The precipitated DNA was pelleted by centrifugation (20,379 g, 15mins), and washed twice using 70% ethanol, centrifuged again (16,000 g, 15mins), then briefly dried under laminar flow. DNA pellets were resuspended in 20μl UltraPure water, heated to 55°C (5 mins) to facilitate dissolution, vortexed, and preserved at −80 °C. Metagenomic analysis was performed on the MinION (Mk1c; Oxford Nanopore Technologies; ONT), using the NEBNext Oxford Nanopore companion module and associated protocols to prepare the DNA sequencing library. Replicates were pooled to reach 800ng, resulting in one sample per condition (D3 and D7) for input into the Native Barcoding Kit 24 (SQK-NBD112.24). The resulting library was loaded onto the flowcell (FLO-MIN106) according to the manufacturer instructions, and samples were sequenced for 24 hours. Read lengths below 200 bp, and a quality score of 8 were discarded during subsequent real-time basecalling in MinKnow (v22.10.5). This resulted in 6.72 Gb of estimated bases, and 5.66 million reads which were trimmed (minimum quality score 7) and normalised using BBDuk and BBNorm respectively. These reads were taxonomically annotated using Kaiju [22], assembled (2955 total contigs) using MetaFlye [23], and functionally annotated with DRAM, leveraging KBase, as previously described [16]. Shotgun metaproteomics Protein searches were performed within ProteinPilot (v5.0.3.1029, 9521aa4603a; Paragon Algorithm: 5.0.3.1029, 1029; AB SCIEX) using the AB SCIEX OneOmics software package, and databases created from our metagenomic data (Section 2.4 ), and other publicly available databases for plastic metabolism (PlasticDB) [24], virulence factors (VFDB) [25], and antimicrobial resistance (CARD) [26]. To create a database of our metagenomic data, the genus-level taxonomic (.txt), and functional (.faa) annotations (Section 2.4 ) were used in the mPies (v1.0) [27] database creation workflow (https://github.com/johanneswerner/mPies). Each database was then used to match spectral (DDA) data to taxonomic and proteomic data in ProteinPilot with a global false discovery rate (FDR) of 1%. The resulting files were then reprocessed with mPies using parameters described in Messer et al., (2024) for protein grouping. This produced functionally annotated proteins, and corresponding taxonomic annotations after protein alignment with Diamond, within the mPies workflow [27]. All taxonomic data was subsequently filtered according to confidence score (≥80%), and proteins with < 2 associated peptides were removed. Before annotation, relative quantification between the conditions was also performed in Skyline (v. 22.2) using the following parameters: mass analyser TOF, MS1 tolerance 0.05Da, MS2 tolerance 0.1Da, maximum missed cleavages 2, and minimum peptides per protein 2. To identify significantly up- or down-regulated proteins between D3 and D7, the MSstats package was leveraged within Skyline. An adjusted P-value of < 0.05 and fold change of 1.5 was considered significant. Missing annotations caused by protein inference issues were manually curated using BLAST (https://www.uniprot.org/blast) if the same protein was matched repeatedly (>50% matches), the highest-rated peptide score equalled ≥100, there was a match at the species level, or a combination of these criteria. Metagenomic data, database search exports, relative quantification data, peptide and protein data are provided in Supplementary files S1: S5. Subsequent data wrangling, and the production of figures was performed using RStudio (v. 4.2.2). Results Metaproteomic analysis of early plastisphere communities resulted in the identification of 2037 and 1886 proteins at D3 and D7 respectively, with corresponding spectral coverages of 33.85% and 28.1%. The majority of proteins in each dataset (D3 = 70.2%; D7 = 72.6%) were successfully annotated using the consensus approach employed by mPies, and the remaining 29.8% (D3) and 27.4% (D7) of proteins were annotated manually. Of the identified proteins, 1499 were non-redundant at D3 and 1449 at D7, of which only 1.85% (±0.05%) were taxonomically unknown or functionally uncharacterised after manual annotation. Overall, 97.4% (±1.2%) of proteins were taxonomically annotated to the phylum level, 96.5% (±0.8%) were annotated to class level, 93.4% (±0.5%) to the order level, 92.2% (±0.6%) to the family level, 87.5% (±3.7%) were annotated to the genus level, and 27.1% (±8.1%) of proteins were annotated to species level. However, a difference in the coverage of these annotations was observed between D3 and D7 metaproteomes, such that 90.9% of proteins were annotated to genus level at D3, compared to 84.1% at D7, and 35.2% of proteins were annotated to species level at D3, relative to 19% at D7. The relative quantification of identified proteins revealed that 711 non-redundant proteins were shared across D3 and D7, with 387 of these found to be upregulated on D3 (representing 120 unique protein groups), and 323 were upregulated on D7 (comprised of 292 unique). These results allowed detailed characterisation of the active taxa and their expressed functions during the early colonisation of plastic. A heterotrophic marine plastisphere At both time-points, the majority of the bacterial community belonged to the class Gammaproteobacteria (>85%), with an increased relative abundance of Alphaproteobacteria (1.1% to 10.3%) and Bacteroidetes (0.05% to 0.35%) between D3 and D7 (Figure 2). Within Gammaproteobacteria, two high quality metagenome assembled genomes (MAGs) were recovered for the genera Pseudomonas and Marinomonas , with a completeness of 98.08% and 96.54% respectively. At the genera level, both the composition and activity of the community exhibited a marked shift between D3 and D7. Initially dominated by active Pseudomonas at D3, the plastisphere transitioned into a more diverse community by D7, where Marinomonas , Pseudomonas , Acinetobacter , Paracoccus , Psychromonas , Vibrio , Rhodobacter , Pseudoalteromonas , and other minor genera were present in greater abundance (Figure 2). All active bacterial genera identified in this study were mainly facultative heterotrophs. Four bacterial autotrophs, photoheterotrophs, and mixotrophs (Rhodobacter , Photobacterium , Sinirhodobacter , and Sinorhizobium ) were active on both D3 and D7, though no proteins expressed by these minority genera were associated with autotrophic processes. This heterotrophic nature of the biofilm was evidenced by the expression of key metabolic enzymes, including succinate dehydrogenase, citrate synthase, malate dehydrogenase, and isocitrate dehydrogenase— all integral to the tricarboxylic acid (TCA) cycle— as well as glyceraldehyde-3-phosphate dehydrogenase and enolase, which are involved in carbohydrate metabolism, and ATP synthase, which is involved in energy production (Supplementary File S3). These metabolic proteins were among the most ubiquitously expressed across both time points, encompassing 10 of the 20 active organisms in this study.