Microbial community succession on PET
The microbial community obtained from beached plastics diverged over time in all treatments when incubated with (a) no additional carbon (control), (b) amorphous PET films, (c) PET powder, (d) weathered PET powder or (e) PET monomer BHET (bis(2-hydroxy ethyl) terephthalate; Fig. 1). While PET powder had a highly crystalline conformation, amorphous PET and BHET monomers were expected to be more accessible for microbial biodegradation. Interestingly, although the assessment of microbial growth is problematic in such settings , DNA extraction yields significantly increased over time for the weathered and non-weathered PET powder as well as the amorphous PET film treatments when compared with the no-carbon control (Fig. S1). Bacterial community structure was assessed via 16S rRNA gene sequencing obtained from the inoculum as well as from all five treatments across six weeks of incubation (i.e. days 1, 3, 7, 14, 21, 30, 42), including separate analysis of planktonic and biofilm communities grown with the amorphous PET films. Only two samples (replicate 2 from day 42 amorphous PET biofilm and replicate 1 from day 42 PET powder) as well as procedural controls (i.e. DNA extraction and PCR negative controls) were removed due to low numbers of reads (<1000). All other samples had a minimum of 4,000 reads and a mean of 19,000 reads per sample.
Microbial community differences between treatments
Gammaproteobacteria dominated all samples (>60% relative abundance; Fig. S2), largely due to the dominance of Vibrionaceae at all time points along with Alteromonadaceae (e.g. ASV2, maximum abundance 25%) and Thalassospiraceae (e.g. ASV18, maximum abundance 9%) during early stages and Alcanivoraceae (e.g. ASV8, maximum abundance 15%) at later stages in all treatments but BHET (Fig. 1 and Fig. S2). All communities were both significantly different from (Fig. 1A and Table S1; ANOSIM R=0.885, p=0.001) and less diverse than (Fig. S3) the inoculum, but the amorphous PET biofilm and BHET treatments showed the most remarkable differences from all other communities (ANOSIM R=0.709, p=0.001 and R=0.446; p=0.001, respectively; Fig. 1A and Table S1). This may come as a consequence of a higher availability of substrate when compared with the other treatments. The principal response curve redundancy analysis (PRC; Brink and Besten, 2009) was used to identify the ASVs that drove these community differences (i.e. had weights higher or lower than 1) or contributed to similarities (i.e. had weights close to one) between treatments (Fig. 1B). Interestingly, many of the ASVs that were contributing towards the differences between the amorphous PET biofilm and no carbon control treatments were not present in the BHET treatment or were present only in very low relative abundances and vice versa (Fig. 1C; Table S2). The planktonic communities surrounding the amorphous PET films became more similar to the no carbon control communities over time, possibly due to a reduction in their access to the PET substrate as mature biofilms developed on the material. On the other hand, the PET and weathered PET powder communities slowly diverged from the no carbon control communities over time, suggesting a possible increase in crystalline-PET degradation and substrate availability as also supported by the increase of PETases within the community as shown below.
Community succession induced by PET-like substrates
The clear distinctness of the microbial communities growing in the presence of BHET and amorphous PET films suggests the presence of an available substrate (i.e. the BHET monomer and available PET chains from low crystallinity/amorphous PET films) that may have selected for distinct biodegrading microbes. To further identify these microbial groups, ASVs were defined as early, middle or late colonisers depending on whether they peaked in abundance on days 1-7, 14-30 or 42, respectively (Fig. 2). This was carried out separately for each treatment and ASVs were only included if they were above 0.5% abundance in at least one time point for that treatment. Overall, this analysis identified 77 ASVs, of which some showed a clear early (n=24), middle (n=15) or late colonisation pattern (n=2), and also revealed a number of ASVs (n=42) that were prevalent in only one condition (highlighted in bold in Fig. 2). Interestingly, the ASVs that drove community divergence in amorphous PET biofilm (PRC analysis contribution to effect >2) or BHET treatments (contribution to effect <0.9; see Figure 1B) compared with the no carbon control were predominantly early and middle colonisers, although these ASVs were also present in other treatments (Fig. 2; Table S3). Other ASVs that were abundant but that did not contribute to community divergence in the PRC analysis in Fig. 1B (i.e. with weights close to one), were generally middle or late colonisers, or varied between treatments (Fig. 2). This may suggest that the readily available substrates to both the amorphous PET biofilm and BHET microbial communities, that initially exert a selection for organisms that are capable of degrading them, may be depleted after one week of incubation. After substrate depletion, the community experiences a succession similar to other treatments in which the substrate is less available. Curiously, the thermal weathering of PET did not produce an apparent increase in PET-derived substrate availability. PET is a highly thermostable polymeric material and, therefore, only small (but significant; two independent samples T-test p<0.05) chemical variations were observed by FTIR after nine months of thermal weathering (incubation at 80˚C; Fig. S4). Specifically, there were small increases in the ratios between the reference wavenumber 1410 cm-1 (I1410) and wavenumbers corresponding to C=O (I1711/I1410) and C-O (I1240/I1410) carboxylic acid, C-O (I1090/I1410) ester and C-H (I725/I1410) aromatic bonds (Fig. S4). This low thermal oxidation and generation of oligomeric PET by weathering may explain the high similarity observed between weathered and non-weathered PET powder exposed communities (Table S1 and Fig. 1) and colonisation dynamics in these treatments (Fig. 2). This contrasts with other plastic materials, e.g. polyethylene, that release large amounts of carbon when thermo-oxidised, which is known to induce the growth of a distinct microbial community during early stages of colonisation [40,42].
Isolation of PET-degrading microbes from the marine Plastisphere
The isolation of PET-degrading microbes from Plastispheres was carried out to further confirm the biodegrading potential extant on marine plastic debris and characterise the metabolic pathways involved. Microbial enrichments using a mix of PET powder and BHET led to the isolation of two bacteria that grew on agar plates with BHET as the sole carbon source. These were identified through partial sequencing of their 16S rRNA genes as Thioclava dalianensis (99% identity) and Bacillus aquimaris (99% identity) (named hereafter as Thioclava sp. BHET1 and Bacillus sp. BHET2) and were selected for further proteogenomic and metabolomic characterisation. Their genome sequences revealed that Thioclava sp. BHET1 had a genome size of 7.66 Mb, with 7,568 coding sequences and a GC content of 63.26%, while Bacillus sp. BHET2 had a genome size of 4.23 Mb with 4,368 coding sequences and a GC content of 40.97% (Table S4).
PET is known to be degraded through an initial hydrolysis (by a PET hydrolase, or PETase) to PET oligomers, BHET, MHET, terephthalic acid and ethylene glycol . MHET may then be acted upon by a MHET hydrolase, producing ethylene glycol and terephthalic acid, although PETases may also exhibit hydrolytic activity on BHET and MHET [15,17]. Ethylene glycol metabolism usually takes place either via conversion to acetaldehyde and acetate [43,44] or via the formation of glyoxylate, while terephthalic acid is usually metabolised to protocatechuate via dioxygenases . A search of the genomes of both isolates was carried out for potential PETases involved in PET hydrolysis using a HMM constructed from known PETase sequences (Table S5), as in Danso et al. , revealing seven enzymes that were above the inclusion threshold in Thioclava sp. BHET1 and four in Bacillus sp. BHET2 (Table S6). Interestingly, though, while the genome of Thioclava sp. BHET1 encoded canonical catabolic pathways for PET intermediate degradation (e.g. terephthalic acid degradation), these were not found in Bacillus sp. BHET2 (Table S4).
A comprehensive proteomic (i.e. of the bacterial isolates; Tables S7 and S8) and metabolomic analysis (i.e. of both the bacterial isolates and microbial communities; Tables S9 and S10) was performed to further identify the mechanisms used by marine microbes to breakdown PET and its intermediates.
Proteomic analysis of PET degradation by two marine isolates
The two new marine isolates, Thioclava sp. BHET1 and Bacillus sp. BHET2, were incubated with the labile substrate fructose (Figure S5), amorphous PET films, BHET and terephthalic acid for a full cellular- and extracellular-proteomic analysis that would shed light on the pathways and enzymes induced by PET substrates.
In Thioclava sp. BHET1, the proteomic data suggests the esterase CDS 0051 as the enzyme involved in PET depolymerisation, which was slightly increased in the presence of PET and its derivatives (i.e. BHET and terephthalic acid) relative to the fructose positive control (Fig. 3). This enzyme was highlighted as a possible PETase by our HMM analysis (Table S6), is predicted to be secreted, and contains an alpha/beta hydrolase fold domain which is characteristic of this kind of esterase. No enzymes that were similar to the MHETase of I. sakaiensis could be identified in the Thioclava sp. BHET1 genome, however, the carboxylesterase (CDS 1741) is likely also capable of this hydrolysis and was 2.1-, 3.5- and 2.3-fold more abundant in the PET, BHET and terephthalic cellular proteomes, respectively. There were also several Tripartite ATP-independent periplasmic (TRAP) transporters that were upregulated in all treatments when compared with the positive control (Table S11), and that could be involved in the transport of PET degradation products into the cell i.e. ethylene glycol and terephthalic acid.
As expected, enzymes involved in ethylene glycol catabolism (encoded by the CDSs 0884, 0999 and 0538; Fig. 3) were more abundantly detected in Thioclava sp. BHET1 in the BHET treatment. Particularly, the acetaldehyde dehydrogenase (encoded by CDS 0999) – necessary for the conversion of acetaldehyde to acetate – was highly abundant in the cellular proteomes, representing 3.33 and 1.42% relative abundance in the BHET and PET treatments.
The other PET biodegradation product, terephthalic acid, is usually converted to protocatechuate for its degradation. In this case, the Thioclava sp. BHET1 proteins that were annotated with these functions (i.e. proteins tphA1, tphA2, tphA3, and tphB encoded by the gene cluster 2867-2870; Table S4) were not detected in the proteomes (Table S8). The conversion of terephthalic acid to protocatechuate is more likely catalysed by other terephthalate dioxygenase orthologues, i.e. 1142-1143, that were particularly induced by the presence of terephthalate (Fig. 3). Protocatechuate is expected to be funnelled into the -ketoadipate pathway via 3-oxoadipate-enol-lactone although, again, the expected enzyme (i.e. protocatechuate 3,4-dioxygenase made by subunits pcaG and pcaH and encoded by genes 1124-1125) was not detected or not particularly induced by the presence of PET degradation products. Interestingly, though, the incredibly high induction of a catechol 1,2-dioxygenase (i.e.catA encoded by 1817; over 8,000x increased in the terephthalate treatment versus the control) and a muconate cycloisomerase (i.e.catB encoded by 1816; 57x increased) may suggest that protocatechuate may be degraded via catechol, as previously hypothesised by Hara et al.  (Fig. 3). Nevertheless, the lack of a protocatechuate decarboxylase in Thioclava sp. BHET1 raises the question of whether the catA and catB-like enzymes may directly attack protocatechuate. Curiously, PET and PET sub-products BHET and terephthalic acid also seemed to co-induce all enzymes involved in the anaerobic degradation of phenylacetate (i.e.via phenylacetyl-CoA and other intermediates to acetyl-CoA ; encoded by the gene cluster 1572-1593; Table S8) as well as some of the enzymes for aerobic phenylacetate degradation (i.e.via homogentisate and other intermediates to fumarate and acetoacetate ; gene cluster 2777-2782).
The pathway used by Bacillus sp. BHET2 to metabolise PET products could not be determined by our proteogenomic analysis. For Bacillus sp. BHET2 the genome annotations by Prokka and KEGG KOALA as well as subsequent local BLAST searches with known terephthalic acid and protocatechuate degradation proteins did not reveal any proteins with significant homology. However, PET treatments did seem to induce a large number of proteins that are usually involved in the degradation of xenobiotics, such as Cytochrome C oxidases and monooxygenases (Table S12) that were upregulated in the PET, BHET and terephthalic acid treatments compared with the control (i.e. with fructose). This, along with the metabolomic analyses shown below (Fig. 3) suggests that these PET compounds are being degraded, but possibly using enzymes that share little homology with those previously described or via a pathway that is currently unknown and needs further characterization.
PET biodegrading potential in the predicted metagenomes from the communities
Having identified canonical and alternative pathways for PET degradation in Thioclava sp. BHET1, we used PICRUSt2  to determine their predicted abundance in the communities (Fig. 4). The default database used for PICRUSt2  uses the KEGG ortholog annotations for 20,000 genomes contained within the JGI genome database . The version used did not contain the KEGG ortholog for PETase (i.e. K21104) and, hence, we used the PETase HMM constructed above to determine the abundance of PETases in the JGI genomes (full details are given in the methods section). Weighted nearest sequenced taxon indices (NSTI) ranged between 0.03 (in planktonic amorphous PET samples) and 0.1 (in the inoculum) and had a median of 0.05 (Table S13), indicating that these predictions are expected to be of acceptable accuracy , although we do note that these are not real metagenomes and the results need to be taken with caution. The predicted metagenome revealed that 156 ASVs (of the 18,114 total ASVs) potentially encoded for a PETase-like enzyme, three of which belonged to ASVs with abundances above 0.5% (all belonging to the order Pseudomonadales): ASV20 (Pseudomonas, 2 copies, maximum abundance 7.4%, NSTI 0.002), ASV43 (Azomonas, 2 copies, maximum abundance 3.6%, NSTI 0.035) and ASV98 (Pseudomonas sp., 2 copies, maximum abundance 0.9%, NSTI 0.025; taxonomically classified by NCBI BLAST), all of which were middle or late colonizers and, curiously, were only abundant in the PET and weathered PET powder treatments (Figs. 2 and 4). Hence, while known PETases were identified in <0.5% of microbial community members in all other treatments, both PET and weathered PET powders showed a remarkable abundance of bacteria that encode a PETase-like enzyme, reaching 20-25% of the microbial community by the end of the incubation (i.e. 1 gene copy per every 4 or 5 bacteria; Fig. 4). Only three confirmed MHETases (i.e. responsible for the conversion of MHET to terephthalic acid) are currently known  and the initial conversion of ethylene glycol to glyoxylate is catalysed by dehydrogenases with broad specificity and, hence, these genes were not included in this analysis.
Enzymes involved in the conversion of terephthalic acid to protocatechuate (i.e. terephthalate 1,2-dioxygenases) were predicted by using a HMM of known terephthalate degradation genes as done above for PETases. The genes tphA2 and tphA3 showed a general decrease in abundance over time in the PET and weathered PET powder treatments, as well as in the no carbon control treatment (Fig. 4). These enzymes are only useful after an initial conversion of the PET, or BHET, to terephthalic acid, and we had therefore expected that in the PET treatments the pattern of their abundance would follow that of the PETases, i.e. would increase in abundance over time. It is possible that the rate at which PET is being hydrolysed is too slow to exert an effect on the abundance of genes for terephthalic acid degradation. It is interesting, though, to note the high abundance of tphB in the biofilm on amorphous PET and in the presence of BHET, possibly because these were the treatments where terephthalic acid was most available (Fig. 4). Interestingly, the alternative pathway detected by proteomics in Thioclava sp. BHET1 for terephthalic acid degradation, i.e. genes 1142-1143 (Fig. 3), followed a similar abundance pattern as tphB (Fig. S6) and may well be worth further biochemical characterization to confirm the hypothesised function given in this study.
The abundance of the genes involved in catalysing protocatechuate towards the -ketoadipate pathway, i.e. the genes pcaG and pcaH, were remarkably abundant in the BHET treatment (Fig. 4), as this substrate may be more readily available than PET. We also explored the abundance within the communities of the alternative pathway suggested for protocatechuate degradation via catechol, i.e. catechol dioxygenase genes catA and catB, because of its strong induction in the proteome of Thioclava sp. BHET1. In this case, the abundance of both catA and catB decreased over time in almost all treatments when compared with their abundance in the community inoculum (Fig. S6). We also analysed the abundance of phenylacetate degradation genes, a pathway that seemed to be co-regulated by the presence of PET sub-products in Thioclava sp. BHET1, observing a consistent increase in abundance of all genes paaABCDEGJKZ in the amorphous PET biofilm treatment (Fig. S6).
While Bacillus sp. BHET2 did not encode for any of the known enzymes for terephthalate biodegradation, there were a considerable number of oxidases and monooxygenases upregulated in its proteome when exposed to PET substrates. Despite that these are very generic enzymes, we analysed the abundance of mono- and dioxygenases in each one of the communities and found, on average, more than one dioxygenase and monooxygenase gene copy per bacterium in the predicted metagenomes (Fig. 7). Hence, no analysis of the distribution of the biodegradation pathway of Bacillus sp. BHET2 could be made within the communities.
Metabolomic assessment of the degradation of PET by isolates and communities
The detection of PET degradation intermediates and the build-up of these metabolites in the culture supernatant are the clearest evidence of PET breakdown and, furthermore, flags bottlenecks where the biodegrading potential of the bacteria may be less efficient (Fig. 3). Non-targeted metabolomics were carried out on the supernatants of our newly isolated marine strains Thioclava sp. BHET1 and Bacillus sp. BHET2 as well as the characterised terrestrial PET-degrader I. sakaiensis [15,18,53], when incubated with amorphous PET films, BHET, terephthalic acid and fructose. Substrates with no inoculum were included as negative controls. We also performed a non-targeted metabolomics analysis on all of the community incubations on day 42 in order to identify products of PET degradation.
The metabolomic analyses confirmed that all three bacterial isolates as well as the community were able to hydrolyse BHET as they significantly accumulated MHET (Thioclava sp. BHET1 and the community), terephthalic acid (Bacillus sp. BHET2) or both (I. sakaiensis), when compared with control incubations with no bacteria (Fig. 3 and Tables S9 and S10). Furthermore, two potential oxidised derivatives of BHET (C12H12O6 and C12H12O7) also accumulated significantly in the Thioclava sp. BHET1 and I. sakaiensis incubations i.e. C12H12O6 and C12H12O7, the first of which was also identified in the incubations with microbial communities (Fig. 3, Tables S9 and S10). The generation of these oxidised derivatives of BHET seems to occur only when this substrate is in excess as they were not detected in the PET treatments, and could be carried out by oxidases, oxidoreductases or dehydrogenases, of which there are many detected in the Thioclava sp. BHET1 proteome. Most interestingly, during the incubations with PET, BHET accumulated in the supernatants of Thioclava sp. BHET1 and Bacillus sp. BHET2 (i.e. almost 22-fold and 2.6-fold higher than in controls, respectively; Fig. 3 and Table S9). In incubations with I. sakaiensis, BHET also accumulated in the PET treatments (1.88-fold higher than in controls) although this accumulation was not statistically significant.
Curiously, no PET degradation sub-products were observed in the community incubations where polymeric PET was present. PET sub-products may have not been observed because the rate at which PET hydrolysis occurs is lower than the assimilation of these oligomers by the microbial community and, therefore, there is no measurable build-up of these metabolic intermediates. This may also explain the strong differences between the biofilm and planktonic microbial community in the amorphous PET film condition (Figs. 1 and 2), where degradation intermediates may be rapidly consumed close to the surface of the plastics and are not accessible to the planktonic community. Hence, due to the lack of PET degradation intermediates, it is not surprising that nly the supernatants of BHET treatments grouped separately from all other supernatants from other conditions on the metabolomic nMDS plot (Fig. S7). The high accumulation of this degradation intermediate in cultures of Thioclava sp. BHET1 exposed to PET suggests that, while it is capable of PET hydrolysis, it is not as efficient at using the BHET as the community is. No degradation intermediates were detected when each of the three bacteria were incubated with terephthalic acid suggesting: i) that the toxicity of terephthalic acid limits degradation of this compound when no other carbon source is present, as we previously found for phthalic acid ; or ii) the degradation pathway of terephthalic acid has no bottleneck that produces an accumulation of detectable levels of the intermediate in the culture supernatant.
PET surface oxidation
Given the distinct proteomic response in the presence of PET and the metabolomic detection of PET hydrolysis products (Fig. 3), we performed an additional experiment to determine the modifications to the amorphous PET surface after incubation with both our marine isolates (i.e.Thioclava sp. BHET1 and Bacillus sp. BHET2), the microbial community and control incubations with no microbial inoculum. The increase in absorbance measured by FTIR at key oxidation peaks after five months of incubation is indicative of PET polymer chain hydrolysis, thus exposing more functional groups and specifically leading to: (1) an increase in the number of C=O and C-O carboxylic acid end groups (increased I1711 and I1240, respectively); (2) C-H bending of the aromatic ring (increased I725); and (3) an increase in the number of C-O ester end groups (increased I1090) [55,56]. Indexes of peak variation were normalised using the invariable peak I1410 that corresponds to ring C-H bending and ring C-C stretching, as in Donelli et al. . Both isolates significantly oxidised the amorphous PET surface (Fig. 5), but the truly remarkable increase in the oxidation produced by Bacillus sp. BHET2 would be in accordance with a non-specific oxidation carried out by the large number of cytochrome C oxidases and oxygenases detected by proteomics. This may also explain that, while Thioclava sp. BHET1 generated a large accumulation of the degradation intermediate BHET, Bacillus sp. BHET2 may produce a diversity of oligomeric intermediates other than BHET. The incubation with the community produced a slight increase in PET surface oxidation although this was not statistically significant (Fig. 5).