Morphological and proteomic analysis of biofilms from the Antarctic archaeon, Halorubrum lacusprofundi

Biofilms enhance rates of gene exchange, access to specific nutrients, and cell survivability. Haloarchaea in Deep Lake, Antarctica, are characterized by high rates of intergenera gene exchange, metabolic specialization that promotes niche adaptation, and are exposed to high levels of UV-irradiation in summer. Halorubrum lacusprofundi from Deep Lake has previously been reported to form biofilms. Here we defined growth conditions that promoted the formation of biofilms and used microscopy and enzymatic digestion of extracellular material to characterize biofilm structures. Extracellular DNA was found to be critical to biofilms, with cell surface proteins and quorum sensing also implicated in biofilm formation. Quantitative proteomics was used to define pathways and cellular processes involved in forming biofilms; these included enhanced purine synthesis and specific cell surface proteins involved in DNA metabolism; post-translational modification of cell surface proteins; specific pathways of carbon metabolism involving acetyl-CoA; and specific responses to oxidative stress. The study provides a new level of understanding about the molecular mechanisms involved in biofilm formation of this important member of the Deep Lake community.

Other than these Antarctic haloarchaea, the best characterized biofilm structures for cold-adapted Archaea are for SM1 Euryarchaeon that grows in sulfurous marsh waters at ~10 °C 14,15 . Forming macroscopic structures (e.g. 3 mm in diameter), the archaeon synthesizes unique appendages (hami) and appears to synthesize a polysaccharide matrix in which it also encases a specific species of Thiothrix sipK4 or IMB1 Epsilonproteobacteria [15][16][17] . The biofilm formed is thought to facilitate nutrient exchange between the two species, enabling syntrophic anaerobic sulfur metabolism. In general for Archaea, biofilm development has not been well studied, and the composition of extracellular material present in biofilms has been reported to be variable 8,18 . The process of cell signaling, or quorum sensing, is often important in biofilm development, but while it has been linked to bacterial biofilms [19][20][21] , few reports exist for equivalent analyses in Archaea; the presence of quorum sensing molecules has been described in haloarchaea 22,23 and a methanogen 24 . A limited number of proteome-based studies have been used to assess archaeal biofilm development, with studies of Ferroplasma acidarmanus 25 and Sulfolobus sp. 26 identifying specific metabolic and morphological characteristics of cells in biofilms.
In view of Hrr. lacusprofundi producing extracellular material and forming biofilms, and the potential ecological importance of this capacity, here we used strain ACAM34 to study cell morphology, the composition of extracellular material and quorum sensing associated with biofilms, and used quantitative iTRAQ proteomics to assess the cellular pathways and processes involved. These analyses complement other ongoing studies of this species and collectively serve to expand our understanding of the ecophysiology of cold adapted Archaea 27 .

Results
Growth conditions leading to biofilm formation. The ability of media composition to promote biofilm formation of Hrr. lacusprofundi ACAM34 was identified during studies aimed at assessing the ability of the strain to utilize urea. Hrr. lacusprofundi ACAM34 was previously reported to have weak growth in DBCM2 minimal medium supplemented with low concentrations of peptone (0.025% w/v) and yeast extract (0.005% w/v) plus pyruvate (10 mM) as a carbon source and urea (10 mM) as a nitrogen source 9 . However, the strain was also found to be urease negative with the genome lacking identifiable genes for urea transport or catabolism 9 . To gain further understanding about the capacity of Hrr. lacusprofundi to utilize urea, cells were examined throughout the growth phase in medium containing 5 mM NH 4 Cl (medium A) or 5 mM urea (medium B), plus varying concentrations of peptone and yeast extract (Fig. 1, Fig. S1). Growth in the presence of urea, and absence or low concentrations of peptone (2.5 × 10 −4 % w/v) and yeast extract (5 × 10 −5 % w/v), produced essentially no growth over 30 d indicating that urea cannot be used as a sole source of nitrogen by Hrr. lacusprofundi ACAM34 (Fig. S1A). In contrast, growth in the presence of urea but with higher concentrations of peptone (0.025% w/v) and yeast extract (0.005% w/v) (i.e. 1 x concentration of medium B) enabled cells to reach OD 600 > 0.2 (Fig. S1A). When ammonium was substituted for urea (1 x concentration of medium A), a higher growth rate and final OD 600 (> 0.3) was achieved (Fig. S1B), illustrating that, unlike urea, ammonium supports growth of this strain.
These growth studies revealed that the cell aggregation state developed differently in the two types of media. With ammonium added to the medium (medium A), cells remained planktonic throughout the 14 d growth phase, but without the addition of ammonium (medium B), aggregates (floating clumps) formed and then attached to flask walls late in the growth phase (Fig. 1). Aggregates were first visible around day 6, increased in size by day 8, and despite cultures being shaken at 120 rpm, aggregates attached to the walls of the flasks by day 10. The attached aggregates could only be removed by being physically pried off the wall of the flask (e.g. using a wire loop). Between day 10 and 14, planktonic cells, floating aggregates and attached aggregates all remained at similar levels. The floating or attached aggregates are herein referred to as biofilms, based on the definition of biofilms by Costerton and colleagues 28 . Cell morphology, extracellular material, and quorum sensing associated with biofilms. To learn about the mechanisms involved in forming biofilms, cell morphology and the presence of extracellular material were assessed using scanning electron microscopy, differential interference contrast microscopy and fluorescence microscopy of cells grown in the absence of added ammonium. Scanning electron microscopy revealed that Hrr. lacusprofundi cell shape was pleomorphic, with rods and cocci present ( Fig. 2A), consistent with the original description of the organism 29 . Scanning electron microscopy performed on cells from 5, 7, 10 and 12 d revealed an increasing extent of extracellular structures connecting cells (Fig. 2). Cell morphology changed from discrete, relatively smooth cellular units ( Fig. 2A,B), to include stringy protuberances (Fig. 2D,F,H) and heavy or rough cell surfaces (Fig. 2H), with sheets or rafts of extracellular material also present (Fig. 2C,E,G).
To further assess the roles of DNA and protein in extracellular material and biofilm formation, cultures were treated with DNase I or proteinase K. DNase I 31-33 and proteinase K [34][35][36] have previously been used to study the structure of biofilms. In order to assess the effects of the enzymatic activity on the formation or disruption of biofilms, the nuclease or protease was added to media before inoculation with cells, or to preformed biofilms, respectively. Quantitation of biofilms was performed by crystal violet staining or bicinchoninic acid assay, with total DNA quantitated by acridine orange staining (Table 1), and cells and extracellular material visualized by scanning electron microscopy (Fig. 4).
Treatment with DNase I showed a concentration-dependent (0, 10 and 100 μ g mL −1 ) reduction in the development of biofilms and total DNA (Table 1, Fig. 4). Scanning electron microscopy revealed a reduction in the extent of lattice-like extracellular material surrounding cells, while the morphology of individual cells remained largely unchanged (Fig. 4). Higher concentrations of DNase I (1 mg mL −1 ) completely inhibited biofilm formation, and led to a large increase in final OD 600 , consistent with the growth of only planktonic cells (Fig. 5A). The high concentration of DNase I (1 mg mL −1 ) produced an early spike in OD 600 (for cultures and uninoculated controls) ( Fig. 5A) consistent with some of the enzyme aggregating in the high salt, and this coincided with a period of no increase in colony forming units (Fig. 5B). However, DNase I activity (assessed by in vitro digestion of Hrr. lacusprofundi DNA) was retained throughout the 14 d incubation period (Fig. S2). Moreover, the replacement of DNase I with an equivalent concentration of bovine serum albumin (up to 1 mg mL −1 ) led to a spike in OD 600 (later in the incubation period) (Fig. S3), but had no effect on biofilm formation throughout the 20 d period (data not shown). Collectively the data for DNase I treatments illustrate that the extracellular material contains DNA that is important for biofilm formation.
Proteinase K treatment (0.1 and 1 μ g mL −1 ) disrupted lattice-like extracellular material and led to changes in cellular morphology (small spherical cells), with 10 μ g mL −1 greatly reducing the rate of cell growth, viability, biomass formation and the production of biofilms (Table 1, Fig. 4). Increasing the proteinase K concentration to 100 μ g mL −1 completely prevented cell growth (data not shown). Similar to DNase I, activity of proteinase K was retained in the growth medium throughout 14 d of incubation (data not shown). While the structure of the extracellular material lost its integrity with relatively low concentrations (0.1 and 1 μ g mL −1 ) of proteinase K (Fig. 4), the yield of filtered biofilm biomass and total DNA increased relative to the control (Table 1). These findings are consistent with proteinase K causing the release of DNA from cells, with DNA accumulating in extracellular material and biomass, and all of it being captured on filters.
The biomass of biofilms was unchanged after 10 d-old cultures with preformed biofilms were incubated for a further 4 d with a high concentration of DNase I (1 mg mL −1 ) or proteinase K (10 μ g mL −1 ) (data not shown), indicating enzyme activity primarily affected growing cells. The findings are overall consistent with cell growth leading to DNA being released from the cell by active export or cell lysis, with the DNA fulfilling a structural role in the extracellular material involved in biofilm formation. Because proteinase K appeared to cause release of DNA and destroy extracellular material lattice structure, it is likely that cell surface proteins were proteolytically degraded during growth of the cells thereby compromising cell structural integrity and possibly protein structures involved in linkages to extracellular material.
The presence of N-acyl homoserine lactone-like quorum sensing molecules was assessed using an Escherichia coli green fluorescent protein reporter assay 37 with the supernatant fraction of planktonic cells (4 d growth) or biofilms (14 d growth) (Fig. 6). Both quantitative measurements that were normalized to protein concentration Proteomics of biofilm formation. Proteomics was performed on whole cells and the extracellular fraction (supernatant) of Hrr. lacusprofundi ACAM34 to learn about global gene expression during biofilm formation and from that data infer pathways and cellular processes involved in forming biofilms. A total of 1996 proteins were detected, accounting for ~54% of the 3665 protein-coding genes in Hrr. lacusprofundi ACAM34. An 8-plex iTRAQ labelling approach 38 (Table S1) was used for assessments of differential abundance, enabling multiple growth conditions to be simultaneously compared. A total of 109 proteins had significant abundance differences (≥ 1.5-fold) between log phase (4 d growth) and stationary phase (14 d growth) in medium lacking ammonium (Table S2); 165 were differentially abundant between stationary phase cultures from media containing or lacking   (Table S2); and 56 proteins were common to both assessments and represented proteins important for biofilms under both the growth phase and growth medium conditions that were tested ( Table 2). The core set of 56 proteins included 36 with higher abundance and 20 with lower abundance (Table 2), and described specific cellular processes ( Table 2, Fig. 7). A total of 13 additional proteins were added to the core set because they had ≥ 1.5-fold differential abundance in one assessment (e.g. log vs stationary) and 1.2-1.5 fold differential abundance in the other assessment (e.g. plus or minus ammonium) and were considered relevant to biofilm formation because they belonged to the functional processes already defined by the 56 core proteins ( Table 2, Fig. 7). All original iTRAQ datasets for two iTRAQ labelling experiments, which included four biological replicates for each growth condition, are provided in Tables S3-S10.
Proteins that decreased in biofilms were mainly involved in protein synthesis and folding, including a large number of ribosomal proteins, and protein chaperones ( Table 2, Fig. 7). Additionally, biofilms had lower levels of a proteasome protein (Hlac_0185), a protein translocase (Hlac_2426), and a cell division protein (Hlac_1716) ( Table 2).

Discussion
The major metabolic pathways and cellular processes involved in biofilm formation were inferred by integrating the core proteomic data with cell morphology, extracellular material, and quorum sensing data (Fig. 8). The abundance data that were most informative about biofilm formation were the relatively large number of proteins that had elevated levels (see below). However, the proteins that decreased in biofilms, particularly those involved in protein synthesis (Table 2), informed about the overall reduced demand for cell growth; this finding is consistent with bacterial biofilm cells growing more slowly than planktonic cells 39-41 . Modification of the cell envelope. In Archaea, the surface layer (S-layer) is composed of identical protein subunits arranged into a monolayer, forming a highly porous lattice structure 42 that maintains the structural integrity of the cell envelope 43,44 . Biofilms appear to require post-translational modification of the S-layer. There is evidence for N-glycosylation 44,45 by a nucleoside-diphosphate-sugar epimerase, and peptidolysis by Hlac_1583, which is homologous to an extracellular endopeptidase I from Lysinibacillus sphaericus that cleaves spore cortex peptidoglycan 46,47 . As Hlac_1583 is a zinc-dependent carboxypeptidase, the elevated levels of the zinc ABC transporter lipoprotein (Hlac_1191) may derive from an increased demand for zinc. The inferred importance of cell envelope proteins is consistent with the potent inhibitory effect of proteinase K on Hrr. lacusprofundi biofilm formation (Fig. 4).

Biofilm biomass
Total intra-and extra-cellular DNA Bicinchoninic acid assay (total protein μg)

Acridine orange assay [(LAU-B) mm −2 ]
Control 1400 ± 30 1.0 ± 0.1 101000 ± 1100 DNase I 10 μ g mL −1 1100 ± 60* 0.75 ± 0.08** 76000 ± 710** DNase I 100 μ g mL −1 600 ± 40** 0.50 ± 0.05*** 26800 ± 270*** Proteinase K 0.1 μ g mL −1 1900 ± 50* 1.7 ± 0.1*** 181000 ± 9700* Proteinase K 1 μ g mL −1 1500 ± 60 1.1 ± 0.09 257000 ± 12000** Proteinase K 10 μ g mL −1 17 ± 6*** 0.03 ± 0.01**** 3190 ± 130*** Table 1. Effect of DNase I and proteinase K on biofilm formation of Hrr. lacusprofundi. Bicinchoninic acid, crystal violet and acridine orange assays were performed at 10 d. The results are the means ± standard errors from three biological replicates. Asterisks indicate the significance of the difference compared to the value for the control, calculated using a paired t-test: *p < 0.05, **p < 0.01, ***p < 0.001; ****p < 0.0001. Metabolic networks in biofilm development. Biofilms are associated with changes in carbon metabolism, particularly via acetyl-CoA (Fig. 8). Acetyl-CoA is proposed to function as a metabolic sensor of the availability of nutrients in E. coli, and accumulation of acetyl-CoA promotes biofilm production 48,49 . The dual action of the glyoxylate cycle (which bypasses the decarboxylation steps of the tricarboxylic acid cycle) and phosphoenolpyruvate carboxylase (which performs anaplerotic fixation of CO 2 ) allows conservation of carbon for biosynthetic purposes (such as biosynthesis of carbohydrates). The increased expression of phosphoenolpyruvate carboxylase may relate to sugar synthesis during growth in the media used, which lacks carbohydrates. Phosphoenolpyruvate carboxylase is reported to play a critical role for extracellular material production in biofilm development in Salmonella 50 . Thus, it appears that carbon sources that are metabolized to acetyl-CoA and further to specific carbohydrates contribute to biofilm formation in Hrr. lacusprofundi. The increased levels of enzymes involved in glycerol catabolism also appear to represent a route for carbon utilization. However, in this case, cells in biofilms may utilize the glycerophosphate backbones of lipids 51 from neighboring lysed cells, as has been suggested for phospholipids from Desulfovibrio vulgaris biofilms 52 . The data are consistent with evidence of increased cell lysis within biofilms (see eDNA and regulation of biofilm formation). More broadly, these proteomic data highlight the relevance of carbohydrate metabolism in biofilm formation, and may reflect a role for carbohydrates as components of extracellular material involved in forming biofilms.  Responses to oxidative damage. The increased abundance of a YggE homolog (Hlac_2298) in biofilms was an indicator of oxidative stress as it is homologous to a periplasmic protein from E. coli that is inferred to protect cells against oxidative stress 53,54 . Intracellularly, cysteine serves as a reductant that drives the Fenton reaction, which generates hydroxyl radicals from iron (II) and hydrogen peroxide, thereby causing damage to DNA 55 . The predicted increase in cysteine biosynthesis (CysK) may therefore enhance oxidative stress. The elevated levels of proteins involved in the assembly of FeS clusters may also be in response to oxidative stress, as has been reported for some bacteria 56,57 . Moreover, the increased abundance of a predicted DNA helicase, and chromosomal protein MC1 which has been reported to protect DNA in Methanosarcina sp. CHTI 55 58 , may reflect an increased demand to protect and repair DNA damage (Fig. 8).
However, the specific enzymes directly involved in catalyzing the decomposition of hydrogen peroxide indicate that levels of this oxidant may be lower in biofilms than in planktonic cells. Ahp is associated with scavenging low levels of hydrogen peroxide, whereas HPI is associated with high levels 59 . The increase in abundance of Ahp and decrease of HPI in biofilms suggests that under the growth conditions that were used, hydrogen peroxide is present, but at relatively low levels. This would be consistent with biofilms growing more slowly than planktonic cells and thus generating lower levels of hydrogen peroxide. An increased abundance of Ahp was previously identified in biofilms of Sulfolobus 26 , several species of Bacteria 60-62 , and the fungus Candida albicans 63 .

Extracellular DNA and regulation of biofilm formation.
Hlac_1867 is homologous to ComA from Halobacterium sp. NRC-1, with both possessing a metallo-β-lactamase domain and Lamin Tail Domain (Fig. S5A). In common with Hlac_1867, the Bacillus subtilis competence protein ComEC and Neisseria gonorrhoeae ComA 64,65 possesses a metallo-β-lactamase domain (Fig. S5B), and the Haloferax volcanii extracellular DNA metabolism protein Hvo_1477, possesses a Lamin Tail Domain 66 (Fig. S5A). A Lamin Tail Domain is present in some membrane associated hydrolases and likely affords membrane association 67 . Metallo-β-lactamase domains are often associated with nucleases 68 and have been speculated to directly catalyse hydrolysis of nucleic acids 69 . In Hvo_1477, the metallo-β-lactamase domain is replaced by a thermonuclease domain, and the protein has been speculated to function in the Hfx. volcanii biofilm lifecycle 66 . Hlac_1867, Hvo_1477 and Halobacterium sp. NRC-1 ComA also possess N-terminal lipobox motifs, indicating they may be lipoproteins. The overall functional similarity of these archaeal and bacterial proteins suggests that Hlac_1867 fulfils a role in DNA metabolism in Hrr. lacusprofundi, and its elevated abundance in biofilms is consistent with the important role that extracellular DNA plays in biofilm formation (Fig. 3).
The elevated levels of proteins involved in purine biosynthesis may also be a response to the demand for extracellular DNA synthesis during biofilm formation. The requirement for purine synthesis is a characteristic of bacterial biofilms [70][71][72] ; for example, Bacillus cereus produces extracellular material containing extracellular DNA, but mutants with defects in purine biosynthesis fail to synthesize biofilms 70 . For some bacterial biofilms, cell lysis and/or secretion of DNA are the main sources of extracellular DNA [73][74][75] . Both mechanisms seem possible for Hrr. lacusprofundi, and if active DNA export does occur, it may be facilitated by Hlac_1867. Given the implications for purine biosynthesis, the role of Hlac_1867, and the effects of DNase I on preventing the formation of biofilms (Figs 4 and 5, Table 1), extracellular DNA stands out as a particularly important component of biofilms in Hrr. lacusprofundi.
Cell signaling via N-acyl homoserine lactones has been directly linked to biofilm development, including the processes of initial surface attachment, biofilm maturation and biofilm detachment 20,76 . The enhanced fluorescence of quorum-sensing induced biosensors observed with Hrr. lacusprofundi biofilms (Figs 6 and S4) suggests quorum sensing may signal biofilm development, perhaps in response to changes in nutrient availability (Fig. 8). The Hrr. lacusprofundi genome does not possess genes known to be involved in N-acyl homoserine lactone quorum sensing (e.g. luxI or luxR homologues), indicating the effector compounds are likely to be mimics of N-acyl homoserine lactones.

Current knowledge of biofilm development in Archaea.
A total of three global gene expression studies examining biofilms have been performed on Archaea: F. acidarmanus 25 , Sulfolobus spp. 26 , and Hrr. lacusprofundi (this study). The majority of F. acidarmanus biofilm associated proteins related to anaerobic growth, consistent with anaerobic zones occurring in some bacterial biofilms [77][78][79] , and the ability of F. acidarmanus to grow anaerobically 25 ; a characteristic not shared by Sulfolobus spp. or Hrr. lacusprofundi. The only class of enzyme with increased abundance from all three studies was Ahp/peroxiredoxin (refs 25, 26 and this study), indicating that a common feature of archaeal biofilms appears to be cells mounting a specific response to relatively low levels of hydrogen peroxide. In Sulfolobus acidocaldarius biofilms, an increased abundance of extracellular material, a putative glycosyl transferase and polysaccharides on cell surfaces were reported 26 , which accords with the proposed role of post-translational modification of the S-layer in Hrr. lacusprofundi biofilms. In terms of quorum sensing, Sulfolobus biofilms were reported to have a decreased abundance of a FabG homolog which is proposed to be involved in the production of a quorum sensing autoinducer 26 , and no evidence for quorum sensing was reported for F. acidarmanus 25 ; findings that differ from Hrr. lacusprofundi. Other than oxidative stress and the role of extracellular material, collectively, these studies do not identify general principles involved in archaeal biofilm formation. This is perhaps not surprising as while the three organisms are members of the Euryarchaeota, they represent three very different ecophysiologies: F. acidarmanus, iron-dependent chemolithotrophic acidophile 25 ; Sulfolobus spp., sulfur-dependent heterotrophic/autotrophic thermoacidophile 26 Table 2. Proteins from proteomics linked to Hrr. lacusprofundi biofilms. Proteins associated with biofilms (higher differential abundance, normal font) or planktonic cells (lower differential abundance, italic font) under both the growth phase and growth medium conditions that were tested. Protein functions assigned based on manual annotation. Differential abundance values are expressed as ratios for two fractions: whole cell fraction (WCF) or extracellular supernatant fraction (SF). Core proteins (total of 56) with significant differential abundance (≥ 1.5-fold) in both assessments (medium B, stationary vs log phase; stationary phase, medium B vs medium A) are shown in bold font; proteins (total of 13) from functional categories represented by core proteins with ≥ 1.5-fold differential abundance in one assessment and 1.2-0.5 fold differential abundance in the other assessment are shown in plain font. Abbreviations: predicted transmembrane domain, TMD; twin-arginine translocation signal, TAT signal; detected in the expressed proteome, but not with significant differential abundance, ns; not detected in the expressed proteome, -.

Figure 7. Protein functional categories associated with biofilms or planktonic cells.
Proteins were identified using an 8-plex iTRAQ proteomics protocol with higher abundance in biofilms (black bars) or lower abundance in biofilms (white bars) under both the growth phase and growth medium conditions that were tested, and assigned to functional categories (also see Table 2).

Figure 8. Depiction of the major metabolic pathways and cellular processes in
Hrr. lacusprofundi involved in forming biofilms. The depiction represents findings for cell morphology, extracellular material, and quorum sensing integrated with the core proteomic data ( Fig. 7 and Table 2 stationary phase, no ammonium), we obtained strong evidence for specific proteins involved in Hrr. lacusprofundi biofilm development. To the best of our knowledge, this is the first quantitative proteomics report to assess biofilms from haloarchaea or Antarctic archaea. The ultrastructural analyses visually characterized the development and extent of extracellular material associated with biofilms, and demonstrated the important role of extracellular DNA in forming biofilms. Hrr. lacusprofundi is an important member of the Deep Lake community representing ~10% of the cellular population, and was identified through network analyses as supporting the highest level of intergenera exchange of high identity regions between haloarchaea in the lake 1 . The establishment of proteomic methods for Hrr. lacusprofundi coincides well with the development of a system for genetic manipulation of Hrr. lacusprofundi 80 , enabling future work to better define the mechanisms involved in formation and regulation of Hrr. lacusprofundi biofilms, and the role of biofilms in promoting gene transfer events. Microscopy and DNase I and Proteinase K treatments. Cells, including biofilms were examined by fluorescence and differential interference contrast microscopy using an Olympus BX61 microscope with DP71 camera (Olympus, Tokyo, Japan) and cellSens Standard 1.11 (Olympus, Tokyo, Japan), and scanning electron microscopy using a JEOL 7001 F field emission scanning electron microscope (JEOL, Freising, Germany), based on previously described methods 8,82 . Acridine orange (Ex 505 nm/Em 525 nm) was used at 10 μ g mL −1 (final concentration) to stain extracellular and intracellular DNA, and 4′ ,6-diamidino-2-phenylindole (Ex 359 nm/Em 461 nm) was used at 5 μ g mL −1 (final concentration) to stain extracellular DNA 8,30 . To prepare samples for imaging by scanning electron microscopy, 1 mL of cells was directly fixed with 2% glutaraldehyde at 4 °C overnight, cells were washed three times in DBCM2 salt solution to remove glutaraldehyde, and dehydrated sequentially in 30, 50, 70, 90, 100, 100, 100% ethanol, 2:1 ethanol/hexamethyldsilazane, 1:1 ethanol/hexamethyldsilazane, and 100, 100, 100% hexamethyldsilazane for 10 min each. Samples were mounted on scanning electron microscopy sample stubs and chromium coated. DNase I 31-33 and proteinase K 34-36 treatments have previously been used to disrupt biofilms. Here, DNase I (Sigma-Aldrich, MO, USA) or proteinase K from Tritirachium album (Sigma-Aldrich) at specific concentrations was added to cultures (25 ml in 100 mL flasks) grown in medium B at the commencement of growth (DNase I: 10 μ g mL −1 , 100 μ g mL −1 , 1 mg mL −1 ; proteinase K: 0.1 μ g mL −1 , 1 μ g mL −1 , 10 μ g mL −1 , 100 μ g mL −1 ), or after biofilms had formed (DNase I: 1 mg mL −1 ; proteinase K: 10 μ g mL −1 ).

Methods
To quantify biofilm biomass, biofilms attached to the inner surfaces of flasks were scraped into solution, and the solution was passed through Whatman no. 54 filter paper using vacuum filtration to collect total biofilms on the filter papers and planktonic cells in the filtrate. The biomass captured on filters was washed three times with a 30% salt water solution 83 . Filter papers were placed in tubes in 25 mL of 30% salt water solution at 30 °C and 120 rpm overnight to release cells into the liquid phase, filters removed, cells pelleted by centrifugation for 5 min at 4,500 × g and resuspended in 2 ml 30% salt water solution. For assessments of the retention of DNase I (1 mg mL −1 ) activity in cultures, aliquots (100 μ L) were withdrawn daily and incubated with 500 μ g Hrr. lacusprofundi genomic DNA at 37 °C for 1 h and the sample analysed by agarose (1%) gel electrophoresis. Quantitative assessments of total biofilm biomass using crystal violet (Sigma-Aldrich) was performed based on previously described methods 84,85 and bicinchoninic acid assays were performed using a Thermo Scientific Pierce BCA Protein Assay Kit. The crystal violet method 84,85 detects bacterial peptidoglycan in cell walls, so the method was modified to make it suitable for Hrr. lacusprofundi. Additional details of methods are provided in Supplementary Information. Quorum sensing. N-acyl homoserine lactone-like compounds were prepared by an ethyl acetate extraction procedure and quorum sensing-like responses determined using an E. coli reporter system, based upon published methods 37,86,87 . Hrr. lacusprofundi cells were centrifuged at 4,500 × g for 20 min at room temperature and the supernatant recovered. Ethyl acetate extractions 86 were performed using 200 mL of cell free supernatant and 200 mL of medium B as a control. Extracts were redissolved in 300 μ L of methanol acidified with 0.1% (v/v) formic acid, and 10 μ L of the redissolved extract and 10 μ L of 200 nM oxo-hexanoyl homoserine lactone (Sigma Aldrich) control were evaporated in a 96 well plate (all in triplicate). The reporter strain, E. coli MT 102 harbouring the plasmid pJBA132, was grown aerobically in LB broth supplemented with ampicillin (100 μ g mL −1 ) at 30 °C with shaking at 150 rpm. Overnight cultures of E. coli MT102 (pJBA132) were diluted 10 times in AB minimal medium 88 . A volume of 200 μ L of diluted E. coli MT102 (pJBA132) was added to each well in the microtiter plate that contained extract, or oxo-hexanoyl homoserine lactone (as a control). Plates were incubated for 5 h at 30 °C at 100 rpm, and GFP fluorescence (485 nm excitation, 535 nm emission) was determined using a CLARIOstar plate reader (BMG Labtech, Ortenberg, Germany) as previously described 37 . Relative fluorescence units were determined by subtracting the background fluorescence (media blank), and fluorescent intensities were normalized to mg of protein present in whole cell extracts. Protein concentration was determined by the bicinchoninic acid method (see Microscopy and DNase I and Proteinase K treatments). Slides for microscopy were prepared using biomass from the wells of the microtitre plates immediately following fluorescence determinations.

Proteomics.
Proteomics was performed based on previously described methods 38,89 . Cells were grown in medium A (with ammonium) or medium B (without ammonium) that contained peptone (0.025% w/v) and yeast extract (0.005% w/v). After cells reached mid-logarithmic phase (4 d), half of the culture volume was removed and cells harvested to obtain a whole cell pellet and an extracellular fraction (supernatant). The remaining half of the culture was grown until 14 d, and cells were harvested and processed as for log phase cells. Quantitative proteomics was performed using 8-plex iTRAQ labelling 90 according to manufacturer's instructions (Sciex, Framingham, MA, USA) using a specific labelling design (Table S1). In brief, a total of four biological replicates were prepared for each growth condition. The growth conditions were: medium A, log phase (day 4); medium A, stationary phase (day 14); medium B, log phase (day 4); medium B, stationary phase (day 14). For each 8-plex iTRAQ labelling run, two of the biological replicates was used from each growth condition. The approach was used for proteins from whole cell extracts and the extracellular fraction, and two labeling experiments were performed for each fraction. This resulted in a total of four 8-plex iTRAQ labelling experiments and a total of 32 protein samples analyzed. Samples were analyzed using a TripleTOF 5600+ hybrid tandem mass spectrometer (ABSciex, Foster City, USA), and data acquired in information-dependent acquisition mode with Analyst TF 1.7 software (AB Sciex, Foster City, USA). Each 8-plex iTRAQ labelling experiment was run twice to provide two technical replicates. By running technical replicates of each 8-plex iTRAQ labelling experiment, a total of four datasets were generated for whole cell extracts and extracellular fractions of each growth condition. Each of the four datasets for the specific growth condition and fraction were combined and searched with ProteinPilot software 4.5 (AB Sciex, Foster City, USA) against the local Hrr. lacusprofundi ACAM34 protein FASTA database to identify proteins. A minimum unused score of 1.3 was accepted for protein identification, representing 95% confidence in correct sequence identification.
For quantitative analysis of relative abundance level changes, data were considered statistically significant when p was less than 0.05 and the error factor was less than 2. The weighted average mean and standard deviation of differential abundance between iTRAQ reporter ion ratios were calculated 91 . In addition, an average weighted abundance ratio of 1.5-fold or more was used as the cutoff for differential abundance for assessing proteins involved in biofilm formation. Typically the iTRAQ proteomics literature cites a minimum fold change of 1.2 92,93 . In this study, we adopted a more conservative approach in selecting a fold change of 1.5, in addition to a minimum p value of 0.05. Pearson's correlation analysis between biological replicates, technical replicates and labelling experiment replicates was performed using SPSS 22.0 software (Fig. S6). By comparing differential abundance between log phase and stationary phase in medium B, and differential abundance between stationary phase cultures from medium A and medium B, the core set of proteins common to both comparisons represented proteins specific to biofilm development under both the growth phase and growth medium conditions that were tested. In some cases, proteins with 1.2-1.5 fold change were considered if they were from functional categories represented by the core set of proteins ( Table 2). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository 94 with the dataset identifier PXD004202. Additional methodological details are provided in Supplementary Information.
All proteins of relevance were manually functionally annotated based on a previously described approach 95 . InterPro platform (http://www.ebi.ac.uk/interpro/) was used to predict protein domains or signal peptides, and features diagnostic of protein function. Transmembrane domains were predicted by TMHMM Server 2.0 (http://www. cbs.dtu.dk/services/TMHMM/) and annotation as a membrane protein required a positive result from TMHMM. Proteins with N-terminal signal peptides and/or a single C-terminal transmembrane helix and/or homology to experimentally characterized cell surface proteins (e.g. S-layer proteins) were annotated as cell surface/envelope proteins according to a previously described approach 5 .