Transporter genes expressed by coastal bacterioplankton in response to dissolved organic carbon

1Coastal ocean ecosystems process dissolved organic carbon (DOC) from both terrestrial and marine sources, making these complex and productive systems critical to our understanding of the global carbon cycle. Bacterial processing of coastal DOC remains a significant conceptual and analytical challenge, however. Thousands of compounds make up the DOC pool, each with different biological turnover rates and many of which no longer resemble the parent biomolecules from which they were formed. This complex DOC pool is processed by a diverse community of heterotrophic bacterioplankton composed of hundreds of different taxa with varying ecological strategies for the uptake and metabolism of organic carbon.

2Several methodologies can provide insights into biological turnover of DOC in marine waters. Approaches that track changes in substrate concentrations over time, or use radiotracers to estimate turnover rates (i.e. the fraction of a compound transformed per unit time) can measure fluxes of individual components of DOC into bacterioplankton cells. These studies show important roles for amino acids and monosaccharides (usually glucose) in bacterially mediated DOC turnover; together, these two compound classes can account for ∼40% of total carbon assimilation by bacterial cells. Approaches measuring DOC uptake at the single-cell level (e.g.;; ) show that most marine bacterioplankton are capable of transporting amino acids and glucose and that phylum- or class-level taxonomic groupings exhibit differential uptake of DOC components, including amino acids, glucose, dimethylsulfoniopropionate, glycine betaine and vanillic acid. These methodologies are typically limited, however, to a small number of compounds whose importance as a bacterial substrate must be assumed a priori. Thus further work is needed to assemble a comprehensive understanding of the compounds that act as primary conduits for carbon flow in coastal waters, and how these pipelines for DOC processing fluctuate on a spatial and temporal basis.

3Here we use a complementary methodology based on functional metagenomics that surveys the DOC pool to identify potentially bioreactive components as well as the taxa that may be responsible for their turnover. We assembled large-scale libraries of mRNA sequences from a coastal bacterioplankton community, identified transcripts involved in the transport of DOC, and analysed the predicted substrates and taxonomic origin of these transporter sequences. Although it cannot furnish quantitative measures of flux, transporter gene expression analysis provides insight into potential uptake of organic compounds by bacterioplankton, including substrates that are in low concentration but high demand (and therefore difficult to assess by direct chemical measures of DOC) and those not presupposed to be important conduits for DOC turnover. Using this functional metagenomics approach, we generate hypotheses about the bioreactive components of south-eastern US coastal DOC and the taxonomic identities and substrate preferences of the bacterial taxa controlling DOC flux.

4Transcript sequences were obtained in May 2007 from duplicate samples of a bacterioplankton community from south-eastern US coastal seawater (31°22′57.22″N, 81°16′51.19″W) with a DOC concentration of 603 μM. Reverse transcription of mRNA-enriched community RNA followed by amplification and pyrosequencing yielded > 140 000 potential protein-encoding sequences from each sample (Table 1). Using bioinformatic criteria established by in silico analyses of known bacterial genes (see Experimental procedures for details), ∼40% of the potential protein-encoding sequences had hits to the NCBI RefSeq database and ∼17% could be assigned to a COG (clusters of orthologous groups) functional category (Table 1). The remaining sequences likely represent transcripts from unknown genes or poorly conserved regions of known genes, or small RNAs.

5Transporter-related sequences accounted for 13% of COG annotations, approximately half of which were assigned to categories relevant to DOC consumption (Table S1). Thus about 1 in 20 COG-annotated mRNAs were involved in the uptake of organic molecules by bacterioplankton cells. Predicted substrates for these DOC-related transporters were amino acids (25%; including branched chain and polar amino acids), oligopeptides (11%), carbohydrates (34%), carboxylic acids (16%), compatible solutes such as glycine betaine and proline (5%), polyamines (2%) and lipids (1%) (Fig. 1). Most DOC-related transporter genes expressed by the bacterioplankton community were for components of ATP-binding cassette (ABC) transporters, followed by tripartite ATP-independent periplasmic (TRAP) transporters, and Na+ symporters (Fig. 1); transcripts for phosphotransferase system (PTS) transporters were not common. The independently collected and processed duplicate samples produced highly similar transcript profiles (Fig. 1).

6The taxonomic origin of expressed DOC transporter sequences was conservatively inferred from a blastx analysis against the NCBI RefSeq database. Forty-seven per cent of all DOC transporter sequences were assigned to just eight order-level marine bacterial taxa: Rhodobacterales (30%; α-Proteobacteria, Primarily Roseobacter), Rickettsiales (10%; α-Proteobacteria, primarily SAR11), Flavobacteriales (1%), and five orders of γ-Proteobacteria (6%; sum of Alteromonadales, Oceanospirallales, Pseudomonadales, Vibrionales, and an uncharacterized taxon related to sulfur-oxidizing symbionts). The remaining DOC transporter sequences (53%) were either assigned to one of 12 orders that each accounted for < 1% of sequences or could not be confidently assigned at the order level. The inferred taxonomy of expressed DOC transporters is consistent with a companion PCR-amplified 16S rRNA clone library in which 51% of the sequences were affiliated with these same eight groups: SAR11 (19%), Roseobacter (12%), Flavobacteriales (11%) and five γ-Proteobacteria orders (9%) (Fig. S1). About 20% of the 16S rRNA sequences were classified into taxa without a reference genome sequence (e.g. SAR86, SAR116 and SAR432). In our bioinformatic pipeline, transcripts from these groups would likely be classified only to phylum level and thus are without order-level taxonomic assignments.

7Expression profiles indicated that Roseobacter (the taxon with the most DOC-related transporter sequences) and γ-Proteobacteria expressed transporter genes for all major classes of organic compounds and may therefore be DOC generalists (Fig. 2). This is in agreement with a previous study in this same coastal ocean showing a diversity of carbon-processing capabilities in bacterial assemblages selected for their ability to metabolize only a single component of the DOC pool. Yet while the absence of strict specialization agrees with the idea that heterogeneity in the supply rate and composition of DOC to coastal oceans may favour generalist bacteria, some order-level taxonomic groupings appeared to have substrate preferences. The Roseobacter and γ-Proteobacteria groups together accounted for 44% of all carbohydrate-related transporter sequences (and 83% of those that could be taxonomically assigned) and may have dominated carbohydrate turnover in this system at the time of sampling. SAR11 members expressed numerous transporters for carboxylic acids (34% of SAR11-like transporter sequences) and amino acids (37%), but very few for carbohydrates (6%) (Fig. 2). Flavobacteriales transporters for the uptake of inorganic compounds (sodium, sulfate, metals) were abundant (Table S1), but few Flavobacteriales-like sequences for organic monomer uptake could be recognized in our annotation pipeline. All eight major bacterioplankton taxa expressed genes for amino acid uptake (Fig. 2).

8The abundance of bacterioplankton transcripts that mediate uptake of amino acids (36% of DOC-related transporter sequences) and sugars (34%) (Table 2) contrasts with typical measured concentrations of hydrolysable amino acids (3%) and carbohydrates (7%) in marine DOC. Carboxylic acids, polyamines and compatible solutes were likewise considerably better represented in the bacterioplankton transcriptome (17%, 2% and 5% of DOC-related transporters) than in the marine DOC reservoir (5%, < 0.1%, < 1%). These discrepancies are expected, since rapid consumption of freshly produced DOC by heterotrophic bacterioplankton results in a DOC reservoir that is biased towards biologically recalcitrant constituents. Transporter expression analysis is therefore likely to be a better proxy for the flux of DOC components into bacterioplankton cells than for the chemical composition of seawater DOC (assuming mRNAs generally lead to transporter synthesis and transporter abundance is roughly proportional to uptake). Indeed, transcript abundance correlates considerably better with the chemical composition of the biologically labile DOC pool (e.g. > 20% for amino acids, 10–30% for glucose; ) than with the composition of the total DOC pool.

9Significant mismatches are nonetheless expected between measures of transporter gene expression and the relative fluxes of organic molecules from the DOC pool into heterotrophic bacterioplankton cells. Constitutively expressed transporter genes yield transcripts regardless of substrate availability (although many bacterioplankton transporters were inducible by changes in the DOC pool; see below). Effects of post-transcriptional regulatory mechanisms, transporter affinity, transporter half-life (typically several hours;; ) and substrate concentration similarly complicate efforts to infer transport from transcription. For high-molecular-weight polysaccharides and proteins, bacterial consumption will be evident only indirectly, through transporter expression for constituent monomers or oligomers following extracellular cleavage. Finally, inference of bioreactive DOC from metatranscriptomic data is constrained by the accuracy of bacterial transporter annotation. Sequences assigned to ‘general’ or ‘hypothesized’ transporter categories are uninformative (17% of transporter sequences), and some of the more specific annotations may be misleading for protein families that have diverse functions. If the potential protein-encoding sequences that could not be annotated based on similarity to known genes (∼50%; Table 1) contain DOC-related transporters in the same proportion as the annotated sequences, only half the relevant transporter sequences were detected here.

10Compositionally different bioreactive DOC should induce different transcriptional responses by a bacterioplankton community. If so, transporter gene expression analysis offers a sensitive bioassay for changes in sources and composition of biologically labile DOC over highly resolved temporal scales (i.e. minutes to hours) useful for addressing dynamics in bacterially mediated carbon flux through the coastal DOC pool. We tested this using coastal seawater samples collected in parallel with the unmanipulated samples described above but amended for 1 h with model coastal DOC derived from phytoplankton or vascular plants prior to RNA extraction. The addition of ∼250 μM model DOC to a ∼600 μM DOC background increased the relative abundance of transcripts mediating protein synthesis, as evidenced by upregulation of genes required for synthesis of ribosomes, t-RNAs, and associated initiation, elongation and termination factors (COG J; Fig. 3). This functional category, which typically represents the first macromolecular pool to respond to a shift-up in bacterial activity, accounted for 14% and 15% of COG assignments for unmanipulated coastal seawater, but 23% and 19% following DOC amendments (Fig. 3). COG assignments for DNA replication and cell division did not increase (COGs L and D; Fig. 3). These data suggest that a transcriptional response to the DOC additions took place by the end of the 1 h incubation while increases in bacterial cell division rates, an indication of possible changes in the composition of the community gene pool that might complicate data interpretation, had not.

11Relative abundance of transporter-related sequences expressed in experimentally modified DOC compared with unmanipulated coastal DOC was assessed using an M–A (magnitude–amplitude) plot. This plot was developed for microarray analysis, and is used here to evaluate the abundance ratio for a COG between the two samples (log abundance ratio; M) against the overall abundance of that COG in both samples (mean log abundance; A). A statistical resampling technique was used to distinguish the transporter COGs for which there were significant differences in relative abundance following amendment (P< 0.05). For the two replicate unmanipulated coastal samples (independently collected, processed and sequenced), the pair-wise analysis found that only 2% of populated transporter COGs (5 out of 234) were significantly different (Fig. S2).

12In contrast, 15% of the transporter COGs were statistically different in pair-wise comparisons between the pooled coastal samples versus the phytoplankton-derived DOC amendment (39 out of 253 transporter COGs; Fig. 4A) and 16% in pair-wise comparisons between the pooled coastal samples versus the vascular plant-derived DOC amendment (39 out of 250; Fig. 4B). Since coastal samples and experimentally manipulated treatments share a common DOC background, significantly enriched transporters in the experimental treatments signal a substrate-induced response to labile compounds introduced with the DOC amendments. Transcript annotations suggested that bacterioplankton exposed to fresh phytoplankton-derived labile DOC (defined here as the portion consumed within a 1 h period) expressed considerably more genes for the transport of carboxylic acids and fewer for the transport of some classes of carbohydrates relative to coastal DOC (Table 2, Fig. 4A). Cells exposed to fresh vascular plant-derived labile DOC expressed more genes for the transport of carbohydrates (possibly as maltose, fructose, pentoses and mannitol) and fewer for amino acids, peptides and compatible solutes (proline, glycine betaine) (Table 2, Fig. 4B). Time-series analyses in future studies could provide additional information on the order in which components of the model DOC pools invoke the cells' transporter expression response. These bioreactive compound profiles should be informative about typical inputs of fresh DOC at this coastal site, which is largely derived from phytoplankton, benthic algae and Spartina alterniflora-dominated marshes. Further, they agree with previous research showing that organic acids, sugars and amino acids are rapidly consumed from plankton-derived DOC, while sugars (particularly glucose and xylose) are the major bioreactive component of S. alterniflora leachate.

13To determine which bacterioplankton taxa were most responsive to changes in the DOC pool, we inferred the taxonomic origin of sequences assigned to significantly enriched DOC-related transporter COGs (see above). Roseobacter-like sequences accounted for most of the significantly enriched transporter sequences to which an order-level taxonomy could be assigned (80% in phytoplankton exudate and 72% in vascular plant leachate). SAR11 sequences were relatively more important among transporter sequences from fresh phytoplankton-derived organic matter than from vascular plant leachate (12% versus 2% of enriched DOC-related transporter sequences), while Flavobacteriales showed the opposite pattern (0% versus 14%). Since substrate transport is the first step of many downstream cellular processes, this taxonomic pattern of activity should be mirrored in other (i.e. non-transport-related) functional gene categories; this was found to be the case (Figs S3 and S4). Transporter expression profiles for individual bacterioplankton taxa in experimental treatments showed specialization on DOC compound classes that was similar to that in unmanipulated coastal DOC (Fig. 2).

14Determining which components of the marine DOC pool serve as conduits for bacterially mediated carbon flux, and how they vary over time and space, has long been a central challenge for microbial ecology. The analysis of coastal bacterioplankton transporters provides a new perspective on this important question by identifying labile compounds being detected or anticipated by bacteria. These include many low-concentration/high-flux compounds that could challenge the detection limits of chemical analyses in seawater and would be difficult to measure simultaneously (Fig. 1), as well as compounds that are not typically evaluated for their role in DOC turnover (e.g. carboxylic acids, nucleic acid constituents and polyamines) (Table 2). For example, one out of six COG-annotated DOC transporter sequences encoded proteins for carboxylic acid uptake (Table 2), indicating a major role in carbon flux for a poorly studied compound class with both planktonic and photochemical sources. Improved understanding of the specific substrates targeted by bacterial transporter proteins will allow more information to be extracted from community transcript analysis. Further, this approach can work synergistically with developing methods for high-resolution chemical characterization of DOC (including mass spectrometry and nuclear magnetic resonance spectroscopy;; ) to simultaneously identify and assess the bioavailability of thousands of distinct compounds.

15To the extent that transporter mRNAs can be considered proxies for DOC flux across bacterioplankton cell membranes, our analysis suggests that two taxa dominated bacterial uptake of DOC in this system at the time it was sampled. Members of the Roseobacter group produced 30% of DOC-related transporter sequences and may have been responsible for the bulk of carbohydrate transport, while SAR11 members produced about 10% and showed a preference for carboxylic acids, amino acids and polyamines (Fig. 5); both appeared active in amino acid turnover. Since about half the DOC-related transporter sequences were not assigned to an order-level taxon, there may be other important groups involved or there may be additional sequences from these taxa that did not meet the annotation cut-offs set here. Future sampling efforts that are temporally resolved will provide much-needed insights into the dynamics of bioreactive DOC in this coastal ecosystem, exploring changes in transporter gene expression patterns over daily, monthly and annual cycles. Despite constraints imposed by annotation accuracy and sequencing depth, this analysis of community gene expression provides a novel perspective on rapidly cycling components of coastal DOC, detailed information on the identity and ecological strategies of the bacterial taxa that drive it, and biological detail at a level crucial for future predictive carbon cycle models.

16Coastal water samples were collected in May 2007 at night and high tide from Sapelo Island, GA. Surface water (10–15 l) was passed in succession through a 5.0-μm-pore-size polypropylene cartridge filter (USFilter, Warrendale, PA), a 3.0-μm-pore-size Poretics polycarbonate membrane (Osmonics, Livermore, CA) and a 0.22-μm-pore-size Poretics polycarbonate membrane. Total filtration time was ∼30 min. Two replicate samples were obtained in succession.

17The 0.22-μm-pore-size filters were vortexed for 10 min in the presence of RNase-free beads (PowerSoil Total RNA Isolation Kit, MoBio, Carlsbad, CA) and lysis/binding solution (RNAqueous-Midi kit; Ambion, Austin, TX). Samples were centrifuged (10 000 r.p.m., 10 min), and supernatants mixed with ethanol solution (RNAqueous-Midi kit). The mixture was passed repeatedly through an 18-gauge needle, filtered through a glass-fibre filter unit (RNAqueous-Midi kit), and washed and eluted according to the manufacturer's instructions. On average, 10 μg of RNA was obtained from each filter. RNA was frozen immediately in liquid nitrogen.

18RNA was treated with DNase (TURBO DNA-free kit; Ambion). To minimize rRNA, samples were treated with the mRNA-ONLY Prokaryotic mRNA Isolation Kit (Epicentre Biotechnologies, Madison, WI) followed by the MICROBExpress kit (Ambion). Approximately 500 ng of RNA was linearly amplified (MessageAmp II-Bacteria Kit; Ambion), and amplified antisense RNA (aRNA) was converted to double-stranded cDNA with random hexamers (Universal RiboClone cDNA Synthesis System; Promega, Madison, WI) and purified (Wizard DNA Clean-up System; Promega). cDNAs were sequenced at the Joint Genome Institute using the 454 FLX sequencing system. The sequences have been deposited in the NCBI Short Read Archive with the Genome Project ID #33823.

19DNA was extracted from filters (PowerMax Soil Mega Prep DNA Isolation Kit; MoBio) and small-submit rRNA libraries were constructed following the JGI recommended protocol (http://my.jgi.doe.gov/general/index.html). The sequences have been deposited at NCBI under the Accession No. FJ744762–FJ745272.

20Ribosomal RNA sequences were eliminated following identification by blastn queries against the GenBank nucleotide database (nt) as described previously. Criteria for annotation of unassembled potential protein-encoding sequences by blastx analysis against the NCBI non-redundant reference sequence database (RefSeq) were established with in silico tests of randomly fragmented full-length known genes with high similarity to metagenomic sequences from Sapelo Island coastal water. Based on these analyses, the cut-off criteria were set at E-value ≤ 0.01, similarity ≥ 40%, and overlapping length ≥ 65 bp to the corresponding best hit. blastx analysis was also carried out against the COG database using similar cut-off criteria, except the E-value was set at ≤ 0.1. COG categories annotated with functions involved in the uptake of organic compounds were identified for further analysis of DOC-related transporters. Taxonomic binning of sequences was carried out using MEGAN. Further assignment into higher-resolution marine taxa (e.g. Roseobacter and SAR11) was based on the NCBI taxonomy of closest blast hits.

21A resampling program for metagenomic data sets was used to compare coastal and experimental sequences categorized based on COG assignments, both for all COG categories and for the subset of COGs representing DOC transport-related functions. The significance level (P) was set at < 0.05.

22Two additional samples, collected in parallel with those described above, were processed 1 h after amendment with model DOC preparations from coastal phytoplankton and vascular plants. Axenic cultures of marine phytoplankton (Skeletonema costatum, CCMP1332; Chaetoceros calcitrans, CCMP1315; Alexandrium tamarense, CCMP1771; and Synechococcus strain WH8101) were grown on a 2- to 3-week schedule at 20°C with a 13:11 light: dark regime of approximately 100 μmol photons m−2 s−1. To obtain concentrated organic matter for experimental manipulations, stationary-phase cells were filtered onto combusted GF/F filters, homogenized, and filtered again to remove cell and filter debris. The chemical composition of intracellular organic compounds extracted from living phytoplankton has been shown previously to be similar to material released extracellularly by healthy cells. Senescent (standing dead) S. alterniflora culms were collected from Sapelo Island, GA, rinsed with sterile deionized water and ground into fragments. Leachate was prepared by incubation in 10 l of sterile artificial seawater in the dark for 5 days followed by removal of cell debris. DOC concentration was measured by high-temperature catalytic oxidation. Additions of phytoplankton- and vascular plant-derived DOC were made to 10 l of coastal water (250 μM final concentration). After 1 h in situ incubation, water samples were filtered as described above.