Abstract
Osmotic stress can accompany increases in solute concentrations because of freezing or high-salt environments. Consequently, microorganisms from environments with a high-osmotic potential may exhibit cross-tolerance to freeze stress. To test this hypothesis, enrichments derived from the sediment and water of temperate lakes with a range of salt concentrations were subjected to multiple freeze-thaw cycles. Surviving isolates were identified and metagenomes were sampled prior to and following selection. Enrichments from alkali lakes were typically the most freeze-thaw resistant with only 100-fold losses in cell viability, and those from freshwater lakes were most susceptible, with cell numbers reduced at least 100 000-fold. Metagenomic analysis suggested that selection reduced assemblage diversity more in freshwater samples than in those from saline lakes. Survivors included known psychro-, halo- and alkali-tolerant bacteria. Characterization of freeze-thaw-resistant isolates from brine and alkali lakes showed that few isolates had ice-associating activities such as antifreeze or ice nucleation properties. However, all brine- and alkali-derived isolates had high intracellular levels of osmolytes and/or appeared more likely to form biofilms. Conversely, these phenotypes were infrequent amongst the freshwater-derived isolates. These observations are consistent with microbial cross-tolerance between osmotic and freeze-thaw stresses.
Introduction
Subzero environments are challenging for microorganisms in a number of ways, not the least of which is physical damage subsequent to ice formation. To cope, microorganisms produce cryoprotectants, cold shock proteins (CSPs), antifreeze and/or ice nucleation proteins (Jones et al., 1987; Xu et al., 1998). As a result, such stress is mitigated by the protection of intracellular processes, a change in the morphology of ice crystals, or by influencing the temperature at which freezing is initiated. Ice formation is also known to generate osmotic stress because solutes and cells become concentrated in brine pockets (Mader et al., 2006; Amato et al., 2009). Indeed, survival under either hyperosmotic or low-temperature conditions is associated with alterations in gene expression (Li et al., 2006) and membrane composition (Allakhverdiev et al., 1999), as well as the accumulation of osmoprotectants (e.g. compatible solutes or inorganic salts; Yancey et al., 1982; Shahjee et al., 2002). Biofilms may also offer protection from cold and freeze stress (Williams et al., 2009).
Cross-tolerance resulting in a similar response to distinct environmental stresses has been previously observed with respect to low temperature and osmotic stresses. For example, salt pretreatments of bacterial strains resulted in the production of CSPs or heat shock proteins, which correlated well with subsequent low-temperature survival (Schmidt & Zink, 2000). Schmid et al. (2009) further showed that CSP genes could be induced either by low temperature or by salt exposure, and such products were necessary for optimal resistance to osmotic stress. Thus, we hypothesized that bacterial assemblages from hyperosmotic environments would show more resistance to freeze-thaw stress than those from freshwater sites. To test this hypothesis, we subjected the enriched bacterial assemblages to repetitive freeze-thaw cycling and compared the post-freeze-thaw survival of the consortia derived from temperate lakes (water and sediment) of varying salinities.
Materials and methods
Sample sites and culture conditions
Samples were obtained from six temperate lakes all located in the same geographic region in south-east British Columbia, Canada (Table 1). Samples were collected in late summer from two freshwater (Fly and Leeches), two brine (East of 83 and Liberty), and two alkali (Clinton and Spotted) lakes. Both water and sediment (top 2–3 cm) samples were obtained, with the exception of Clinton, from which no water sample could be gathered because of seasonal drying. Samples were kept at 4 °C during shipping and before processing (≤ 6 months) and were mixed thoroughly prior to use. Major ions were determined (Table 1), and appropriate growth media chosen or formulated based on the water analysis. Previous data obtained from Clinton Lake water were used analogously. Freshwater and alkali-lake samples were cultured in 10% tryptic soy broth (TSB; Bacto, Dickinson and Company, Sparks, MD) consisting of 3 g TSB, 0.1 g KNO3, 0.1 g (NH4)2SO4 and 0.1 g K2HPO4 per liter of deionized water; pH 7) or 50% Marine Broth (pH 7; Difco, Becton, Dickinson and Company, Sparks, MD), respectively. Media suitable for the brine lakes were formulated as follows: East of 83 medium (3 g TSB, 0.01 g CaCl2·2H2O, 0.34 g K2HPO4·3H2O, 0.03 g MgSO4, 1.15 g NaCl, 2.15 g Na2CO3, 0.78 g NaHCO3, per liter of deionized water; pH 10) and Liberty medium (3 g TSB, 0.01 g CaCl2·2H2O, 1.36 g K2HPO4·3H2O, 0.05 g MgSO4, 3.90 g NaCl, 44.17 g Na2CO3, 16.06 g NaHCO3, per liter of deionized water; pH 10). In all cases, 1.5% agar was added for semi-solid media.
Sample location, ionic characteristics, and culture medium for each lake sample
, S2−Sample location, ionic characteristics, and culture medium for each lake sample
Consortia derived from the water samples were obtained by inoculation (1 mL) into the appropriate medium (6 mL), cultured for 72 h at 22 °C, and subsequent subculture (1 mL culture in 6 mL of fresh medium) as described. Sediments were similarly enriched by inoculation of the sediment (0.5 g) into suitable medium (10 mL), culture for 72 h at 22 °C, and subculture as described. Control bacterial strains (Escherichia coli TG-2 and Chryseobacterium sp. C14) were cultured in 10% TSB. All subcultures and controls were incubated for 48 h at 22 °C and transferred to 4 °C for approximately 18 h prior to freeze-thaw treatments. After treatments, isolates recovered as monocultures were again cultured in the respective media prior to sequencing, except for one isolate (see ), which was cultured in peptone, yeast extract, alkaline buffer broth (PYA, pH 10; Yumoto et al., 2004). Isolates subjected to further characterization were cultured as described, but after growth at 22 °C for ≥ 48 h they were transferred to 4 °C (48 or 72 h) prior to assay (Walker et al., 2006). All cultures were shaken at ~ 100 r.p.m. during growth at 22 °C, while the 4 °C cultures were incubated statically.
Freeze-thaw selection
Aliquots (2 mL) of the enriched samples (in the culture medium, in triplicate) were subjected to 48 consecutive freeze-thaw cycles in an automated apparatus dubbed a cryocycler. Each cycle (120 min per cycle) consisted of cooling to −18 °C (over 60 min) and warming to +5 °C (over 60 min), with the samples at or below 0 °C for ~ 95 min per cycle (Walker et al., 2006). Colony forming units (CFU) per mL were determined after the water and sediment enrichments were serially diluted with dilute ‘lake water’ (1 : 1 sterilized lake water from the respective freshwater or brine lake and sterile distilled water). Spotted and Clinton Lake samples were diluted 1 : 4 with sterilized water from Spotted Lake. These freeze-thaw selected samples (undiluted) were used for all subsequent analysis, including survivor isolation and metagenomic sequencing (following recovery). To facilitate comparison, each sample was normalized to a starting density of 1 × 108 CFU mL−1. One-way analysis of variance (anova; α = 0.05), followed by Tukey–Kramer honestly (least) significant difference (HSD; http://udel.edu/~mcdonald/anova.xls) statistical tests, was used.
Isolate recovery and identification
Morphologically distinct colonies were isolated as monocultures, a subset of which were putatively identified by their 16S rRNA gene sequence after polymerase chain reaction (PCR) amplification with the universal primers 8F and r1406 (Lane et al., 1985; Hicks et al., 1992). Standard reaction conditions were used, and cycling conditions were performed according to a manufacturer's protocol (Novagen T7 Select phage display protocol; Madison, WI). Amplified products (~ 1.4 kb) were visualized on 1% agarose gels, purified (Qiagen, Mississauga, ON, Canada), and sequenced in both directions with the same primers. Sequences were analyzed using ‘Manipulate and Display a DNA Sequence’ from Molecular Toolkit (www.vivo.colostate.edu/molkit/) and CodonCode Aligner (http://www.codoncode.com/aligner/trial.htm). Sequences derived from the forward and reverse primers were manually aligned by overlapping the sequences, typically by ~ 100 bases. Potential chimeric sequences were not specifically investigated; however, all sequences with corresponding chromatogram traces that indicated multiple sequences were omitted or repeated to ensure the isolates were not mixed cultures. Putative identities were determined based on the nearest phylogenetic relative in the blastn (NCBI; http://www.ncbi.nlm.nih.gov/blast; default settings; Altschul et al., 1997) or Ribosomal Database Project II (RDP II; http://rdp.cme.msu.edu; default settings; Cole et al., 2009) databases. The RDP II database was used for sequences that could not be identified at the genus level following the blastn searches.
Metagenomic analysis
DNA was isolated from Leeches, Liberty and Spotted Lake sediment-derived enrichments prior to and following freeze-thaw selection (following recovery) with the DNeasy kit (Qiagen; Gram positive protocol). Metagenomic sequencing using a Roche genome sequencer (GS), FLX titanium chemistry, was performed at Genome Quebec (QC, Canada). The six (one per sample) 454 DNA sequencing files (extension SFF) were clipped, as recommended by the Roche 454 software and converted to FASTA format. Individual reads were compared with a curated database of bacterial 16S ribosomal subunit sequences (http://greengenes.lbl.gov) using the BLASTn program from the Blast 2.2.24 suite (NCBI). Reads found to contain 16S rRNA gene sequence were extracted from the.SFF file (using the sfffile program in the GSassembler Suite; Roche) and compared with a database of all NCBI bacterial sequences (retrieved September 2010). The use of blastn enabled the GenInfo Identifier (GI) numbers to be used to obtain taxonomic information and to facilitate the identification of any flanking sequence on the reads. The filtered 16S rRNA gene-containing reads yielded taxon assignments to a high degree of significance, and the highest scoring ‘hit’ from the tabular blast output for each was than concatenated into a single text file. The taxonomic identity of the top blast ‘hit’ from each putative 16S rRNA gene sequence-containing read was determined using Galaxy Genomics suite (http://galaxy.psu.edu). The proportion of reads belonging to each taxonomic group was determined at the class, phylum, and genus levels.
To identify sequences encoding the B protein of DNA gyrase (gyrB), the recombination protein recA, the β-subunit of RNA polymerase (rpoB), and the RNA polymerase sigma D factor (rpoD), reads were compared with a database of all bacterial sequences in NCBI (retrieved September 2010). Using blastx, each nucleotide read was then conceptually translated into each of three reading frames in both directions and compared with the peptide database (expectation cutoff of E = 0.001). Taxonomic data were then generated using the Galaxy Genomics suite as described above. The sequences in the FASTA files (above) were compared with the NCBI nonredundant database using blastx (with searches limited to Bacteria, Archaea and Fungi). The final blast results, concatenated into a single file, were used as the input for megan (MEtaGenome ANalyzer; Huson et al., 2011). All subsequent analysis was carried out with megan [lowest common ancestor (LCA) parameters: Min Support (1), Min Score (50.0), Top Percent (10) and Win Score (0)], normalized to 100 000 reads per sample.
Isolate characterization
A subset of the freeze-thaw selected isolates was assayed for: ice recrystallization inhibition (IRI), ice nucleation activity (INA), osmolyte content, and biofilm formation. All assays were performed at least in duplicate. In all cases, media controls were included to help ensure that the varying composition and conduciveness of the media did not influence the results.
IRI assays were conducted as previously described (Tomczak et al., 2003; Wilson & Walker, 2010). Briefly, after snap freezing (~ −35 °C), capillary tubes loaded with whole-cell cultures were annealed at −6 °C overnight, with the images obtained prior to and following incubation used to assess ice crystal growth. INA was estimated by cooling whole-cell droplets (~ +2 °C to −12 °C) and recording the nucleation temperature (modified from Vali, 1971 and Maki et al., 1974; as described in Wilson & Walker, 2010). Prepared, recombinant antifreeze proteins (Gordienko et al., 2010) and an ice nucleation protein preparation (Wards Natural Science Establishment, Rochester, NY) were used as the positive controls for the IRI and INA assays, respectively. Escherichia coli TG-2 and culture media were used as negative controls for both assays.
A vapour pressure osmometer (model 3MOplus; Advanced Instruments Inc., Norwood, MA) was used to estimate the intracellular osmolyte content of the isolates. After transfer of isolate cultures to 4 °C for 72 h, cells were harvested via centrifugation (15 min at 10 060 g). Pellets were resuspended in distilled water (500 μL). Cells were subsequently disrupted by sonication (Belgrader et al., 1999). Osmolyte content, in milliosmoles (mOsm), was determined. Escherichia coli TG-2 and appropriate culture media were used as negative controls. Cell numbers were normalized using optical density, to an OD600nm of 0.24 (the postresuspension OD600 of E. coli TG-2). One-way anovas (α = 0.05) followed by Tukey–Kramer HSD statistical tests were employed.
Biofilm assays were performed as previously described (O'Toole & Kolter, 1998; Balestrino et al., 2008). Briefly, isolates were cultured for 48 or 72 h at 22 °C in 96-well plates (4 μL culture in 100 μL suitable media). After overnight incubation (4 °C), samples were stained with gentian violet (50 μL; 0.5% w/v) for 15 min, the wells washed at least five times with water, and the dye released upon addition of 95% ethanol (200 μL). After the alcohol extracts were transferred to new 96-well plates, the OD570 was determined and the mean control level (which varied depending on the media) subtracted. Escherichia coli TG-2 and the culture media were used as negative controls. T-tests (α = 0.05) were used to assess the significance of the differences between the isolates and E. coli TG-2.
Results
Water chemistry analysis
Chemical analysis (Table 1) indicated that the lakes were distinct with Fly and Leeches Lakes classified as freshwater, with a salt content below the detectable limit of the vapour pressure osmometer (0 mOsm), and with slight differences in H+ ion concentration (pH 8.2–8.9). East of 83 and Liberty Lakes were classified as brine lakes because they were high in Na+ or Na+ and Cl- ions (885 mOsm and 2074 mOsm, respectively; pH 10). Clinton and Spotted Lakes were classified as alkali lakes because they were high in alkali metals, Mg2+ and 
, both with salinity values > 360 g L−1 (Wilson et al., 1994, 1996) varying at 2389 mOsm; pH 7 and 6016 mOsm, pH 8.2, respectively.
Freeze-thaw selection
Highly freeze-resistant members of the original lake water or sediment assemblages were selected for with 48 consecutive freeze and thaw cycles. Overall, the postselection reduction in consortia abundance (Fig. 1a) correlated with the ion content of the ‘home’ lake [Fig. 1b; linear correlation coefficient (R) of 0.99 for the water enrichments excluding Liberty Lake, or 0.78 including Liberty Lake; 0.99 for sediment enrichments excluding Liberty Lake, or 0.94 for the sediment enrichments including Liberty Lake; 0.92 for the average resistance of the water and sediment samples including Liberty Lake]. The viability of the consortia derived from the two freshwater lakes decreased by approximately 10 000-fold in the water-derived enrichments, and 100 000–1 000 000-fold in the sediment-derived samples (Fig. 1a). The microbial abundance in the brine lakes decreased 100–10 000-fold in the water-derived samples, and approximately 10 000-fold in the sediment-derived consortia. Enrichments from the alkali lakes decreased about 100-fold for both the water and the sediment-derived consortia. One-way anova (α = 0.05) and Tukey–Kramer HSD tests indicated that the water and sediment samples from the two freshwater lakes and Lake East of 83 had statistically similar CFUs per mL and thus similar freeze-thaw resistance levels. All were less resistant (P ≤ 0.05) than those samples derived from Liberty Lake water and the alkali-lake-derived samples (Fig. 1a). Sediment cultures from Liberty Lake, which had a lower salinity than Spotted Lake, were less resistant (P ≤ 0.05) than the Spotted Lake water and sediment cultures. Sediment-derived cultures from both of the alkali lakes were similarly resistant and had slightly less viable cell numbers after selection (P ≤ 0.05) than the Spotted Lake water consortia. Such high viability after 48 freeze-thaw cycles was only surpassed by Chryseobacterium sp. C14, a known highly freeze-thaw-resistant strain (Fig. 1a; Walker et al., 2006; Wilson & Walker, 2010).
Average CFU per mL for the water and sediment-derived consortia after 48 freeze-thaw cycles (a). These are compared with single isolate control cultures, the highly freeze-thaw-resistant Chryseobacterium sp. C14 and the relatively freeze-thaw-susceptible Escherichia coli TG-2. ‘All’ indicates the normalized starting density of 1 × 108 CFU mL−1 for all cultures. Significant differences (α = 0.05) amongst each sample type are indicated by different letters (a–d) over the bars and freshwater (F), brine (B) and alkali (A) samples are indicated along the top axis. Linear correlation curves (b) demonstrating a positive correlation between freeze-thaw resistance and lake water osmolyte concentration (R of 0.78 for all water enrichments; R of 0.94 for all sediment enrichments; R of 0.92 for the average of the water and sediment enrichments).
Isolate recovery and identification
The 16S rRNA gene sequence of a subset of the morphologically distinct microbial isolates (with redundancy) post-freeze-thaw selection were sequenced to gain insight into the community composition following selection. The putative identity (based on NCBI blastn or RDP II search results), accession number, sequence length, query coverage, and percent similarity to the closest database match are shown in Supporting Information, Table S1. Notwithstanding the orders of magnitude reduction in overall cell viability, a variety of genera were identified in the cultures including known halophiles from the brine and alkali lakes.
Metagenomic analysis
Metagenomic analysis was performed on the DNA extracted from sediment-derived enrichments from Leeches, Liberty, and Spotted Lakes, prior to and following selection (postrecovery). The six metagenomes (NCBI accession numbers SRX147897-SRX147902) were sequenced in a partial plate run; an average of 564 reads containing 16S rRNA gene sequence (242–967) and 212 reads containing gyrB, recA, rpoB, or rpoD gene sequence (21–823) were identified within the metagenomes. Table 2 summarizes the number of reads per sample, as well as an overview of the assigned protein and phylogenetic content within these metagenomes. This analysis was carried out to better determine the effects of freeze-thaw stress on sample richness and diversity, using the 16S rRNA gene sequence, gyrB, recA, rpoB, and rpoD (Table 3, Figs 2 and 3) in the absence of possible PCR-induced bias. The metagenomes also provided a database for mining the genetic content within the enrichments.
Phylogenetic composition (class-level) within the lake sediment consortia that were classified as freshwater (Leeches), brine (Liberty) or alkali (Spotted) before and after freeze-thaw selection (as determined by 454 sequence reads containing portions of the 16S rRNA gene). Less abundant classes include Actinobacteria (Liberty Lake before and after freeze-thaw; 0.4% each) and Bacteroidia (Leeches Lake before freeze-thaw; 0.4%), Mollicutes (Liberty Lake before freeze-thaw; 0.4%) and Negativicutes (Leeches Lake before and after freeze-thaw; 0.4 and 0.1, respectively).
Phyla-level consortia composition within sediment-derived enrichments from lakes classified as freshwater (Leeches; a), brine (Liberty; b) or alkali (Spotted; c). Composition was determined, before and after freeze-thaw selection, by 454 sequence reads containing portions of the gyrB, recA, rpoB and rpoD genes.
Overview of the metagenomic analysis (megan), before and after freeze-thaw selection
Number of raw reads based on the individual data sets, not normalized.
Number of classes identified by SEED analysis (functional analysis) per lake, within the merged, normalized metagenomes.
Number of KEGG (Kyoto Encyclopedia of Genes and Genomes; functional analysis) classes identified per lake, within the merged, normalized metagenomes.
Percent of total 16S or 18S rRNA gene sequence-containing reads within the normalized metagenomes assigned to Bacteria, Archaea and Fungi.
Overview of the metagenomic analysis (megan), before and after freeze-thaw selection
Number of raw reads based on the individual data sets, not normalized.
Number of classes identified by SEED analysis (functional analysis) per lake, within the merged, normalized metagenomes.
Number of KEGG (Kyoto Encyclopedia of Genes and Genomes; functional analysis) classes identified per lake, within the merged, normalized metagenomes.
Percent of total 16S or 18S rRNA gene sequence-containing reads within the normalized metagenomes assigned to Bacteria, Archaea and Fungi.
Microbial community richness and diversity within the metagenomes derived from the freshwater (Leeches), brine (Liberty) or alkali (Spotted) lake consortia before and after freeze-thaw selection
A measure of community evenness based on the Shannon index.
Microbial community richness and diversity within the metagenomes derived from the freshwater (Leeches), brine (Liberty) or alkali (Spotted) lake consortia before and after freeze-thaw selection
A measure of community evenness based on the Shannon index.
Table 3, showing the richness and diversity within the metagenomes based on 16S rRNA gene sequence (genera level), indicates that freeze-thaw selection reduced the microbial diversity and richness within the freshwater sample, as well as the richness within the brine sample. In contrast, there was no reduction in the diversity of the brine or alkali samples, or in the richness within the alkali sample. In fact, the detected diversity and richness increased within the alkali samples; however, differences in sequencing depth amongst the samples may affect these measures.
Prior to selection, the freshwater lake metagenome was dominated by β- and γ-Proteobacteria (36.0% and 55.2%, respectively), and Firmicutes (Bacilli and Clostridia; 4.0% each), while after 48 freeze-thaw cycles, the relative proportions shifted to predominately Firmicutes (63.5% Bacilli; 0.2% Clostridia) and γ-Proteobacteria (36.2%; Fig. 2). The brine-lake consortium was also dominated by γ-Proteobacteria (50.2%) and Firmicutes (34.3% Bacilli; 14.7% Clostridia) prior to selection and changed modestly in response to freeze-thaw selection [ γ-Proteobacteria (60.9%) and Firmicutes (24.9% Bacilli; 13.9% Clostridia; Fig. 2)]. Conversely, the alkali-lake-derived metagenome, also dominated by Firmicutes (52.5% Bacilli; 39.8% Clostridia) and γ-Proteobacteria (7.6%) prior to selection, was entirely dominated by Firmicutes (100%; Fig. 2) postselection. Within the Firmicutes, there was an increase in Bacilli (62%) postselection, while Clostridia remained consistent (38.1%). Minor phyla are indicated in Fig. 2.
The metagenomes were mined for other coding sequences that may confer stress resistance, the preliminary results of which are presented in Fig. 4. The total number of SEED and KEGG classes identified per lake sample is reported in Table 2. There was a slight overall increase in the number of normalized sequence reads encoding proteins involved in stress response within the Liberty Lake metagenome, including heat shock (dnaK gene cluster extended, heat shock protein GrpE, and the heat-inducible transcription repressor HrcA), universal stress protein family, ectoine biosynthesis (osmotic stress), an arginine/agmatine antiporter (acid stress) and signal transduction (NarL) pathways. There was, however, a decrease in the SigmaB stress response regulation pathway (chaperones and proteins). There was an overall decrease in the normalized number of reads after selection, relative to the preselected consortia, corresponding to general stress responses within the Leeches and Spotted Lake sediment-derived metagenomes. Within the Leeches and Spotted Lake postselection metagenomes, there was an apparent increase in reads containing genes involved in the OmpR and NarL families (signal transduction), most notably proteins involved in osmotic or low-temperature response, sporulation, and biofilm formation. Sequences encoding ice nucleation and antifreeze proteins have not yet been found within any of the metagenomes.
Summary of some stress resistance genes within the normalized metagenomic data sets, before and after freeze-thaw selection. The number of reads related to global stress response (SEED analysis) and signal transduction (KEGG pathway) are shown. Heat shock refers to the heat shock dnaK gene cluster extended. The OmpR family includes proteins involved in osmotic and low-temperature response as well as sporulation and biofilm formation. The NarL family includes proteins involved in osmotic and low temperature response. Reads relating to acid stress were not detected within the Leeches Lake metagenome after freeze-thaw selection.
Isolate characterization
A subset (21) of the microorganisms recovered following freeze-thaw selection were assayed for known hyperosmotic and/or low-temperature-resistant phenotypes (Table S2), including ice-association properties, biofilm characteristics, and osmolyte content.
None of the recovered bacteria demonstrated IRI activity. Only two isolates, both from brine lakes (Altermonadales EW4 and Roseinatronobacter monicus LLW20), exhibited INA, and this activity was only apparent when the bacteria were cultured in media similar to their ‘home’ lakes and subsequently dialyzed (overnight at 4 °C) in 50% Marine Broth. This INA activity was lower than that observed in a highly active commercial preparation (approximately −6 °C for Alteromonadales EW4 and −8 °C for R. monicus compared with −2 °C for the P. syringae positive controls).
The intracellular osmolyte content was determined with a vapour pressure osmometer. Only 22% (2/9) of the freeze-thaw-resistant bacteria derived from freshwater lakes had a statistically significantly higher intracellular osmolarity than the E. coli TG-2 control (Table S2). As anticipated, a greater proportion of isolates derived from the higher saline environments had a higher osmolyte content than the E. coli TG-2 control. Indeed, 100% (6/6) and 83% (5/6) of those obtained after selection of the brine- and alkali-lake samples showed a significant elevation in osmolyte content, relative to E. coli TG-2.
Lastly, the ability of the isolates to form biofilms was assessed and values were normalized against the respective culture media and compared with E. coli TG-2, because this species typically has low levels of biofilm production (Reisner et al., 2006). Of the tested freeze-thaw selected bacteria, none (0/9) derived from the freshwater lake samples produced significantly more biofilm than control samples (Table S2). In contrast, at least 50% (3/6 or 5/6, depending on the media) of the freeze-thaw selected brine-lake isolates and 67% (4/6) of the alkali-lake samples were positive for biofilm formation.
Discussion
Freezing and high extracellular ionic concentrations are similarly challenging for microorganisms in that they both can result in osmotic stress, and thus, there is potential for cross-tolerance between freeze-thaw and osmotic stresses, in whole assemblages, as has previously been explored in certain bacteria (Schmidt & Zink, 2000; Schmid et al., 2009). When microbial consortia from six lakes within the same geographic region, but of varying ionic concentrations, were subjected to repetitive freeze-thaw cycles, distinct resistance profiles were seen. Sediment- and water-derived cultures from the two freshwater lakes were the most freeze-thaw susceptible and cell viabilities were reduced at least 100 000-fold (Fig. 1a). Remarkably, in comparison, the alkali-lake water- and sediment-derived samples generally showed the most resistance with only 100-fold losses in viability. For the most part, culturable consortia derived from the brine-lake samples showed an intermediate resistance between these two extremes. With the exception of the Liberty Lake water- and sediment-derived enrichments, there was a strong positive correlation between resistance to freeze-thaw stress and the osmolyte content. The Liberty Lake enrichments seemed to demonstrate a level of freeze resistance which was disproportionately high relative to the high-osmotic potential of the lake water (Fig. 1b). Together, these results support a hypothesis of a cross-protection phenotype at the assemblage level.
Generally, enrichments of water vs. sediment assemblages were concordant in their loss of viability after freeze-thaw selection. However, when they differed, those derived from the water column proved to be more resistant than their sediment-derived counterparts (Fig. 1a). This may result from seasonal freezing of the upper portions of the water column (as reported by local residents), and not the lake sediments. Although collections were made in late summer, prior freeze-exposure may have preselected the water samples to be more resistant in a few lakes. Alternatively, any modest differences may reflect the assemblage diversity of the water and sediment niches in the same lake.
The sequencing results of a subset of isolates (Table S1) indicated that freeze-thaw survivors from freshwater lakes were largely spore formers such as Bacillus and Paenibacillus, in addition to the adaptable Pseudomonas. All three of these genera, as well as others recovered from both water and sediment samples, including Acinetobacter, Arthrobacter, and Sphingomonas have previously been identified after freeze-thaw or ice-affinity selection from soils (Walker et al., 2006; Wilson et al., 2006; Wilson & Walker, 2010). Many of these, as well as Rhodococcus, have also been previously reported from permanently frozen environments (e.g. Christner et al., 2000; Gilbert et al., 2004; Miteva et al., 2004).
Likewise, some of the brine-lake survivors were also spore formers (Bacillus) and Pseudomonas. Additionally, however, because these lakes were alkaline (pH 10; Table 1), it is not surprising that alkaliphiles and halotolerant microorganisms such as Alkalibacterium, Halomonas, Idiomarina, Nesterenkonia, and Roseinatronobacter were also identified. Genera with similar environmental preferences were even more apparent amongst the survivors from the alkali lakes (pH 7 and 8.2; Table 1), which included Gillisia, Halomonas, Marinobacter, Nesterenkonia, and Salegentibacter, in addition to spore formers (Bacillus and Lysinibacillus) and the near ubiquitous Pseudomonas. The vast majority of these freeze-thaw-resistant bacteria have also previously been reported in cold (or frozen), hypersaline and/or alkaline environments (e.g. Stougaard et al., 2002; Brinkmeyer et al., 2003; Steven et al., 2007, 2008; Niederberger et al., 2010). Indeed, the recovery of such microorganisms underscores the utility of the rigorous freeze-thaw selective regime.
A number of morphologically distinct microorganisms survived selection, so much so that Table S1 necessarily represents a limited list. Admittedly, sequencing morphologically distinct, hand-picked, resistant isolates does not necessarily reflect the most abundant bacteria recovered. However, by supplementing isolate sequencing with metagenomic analysis, performed both initially and following selection, the most abundant microorganisms could be revealed. A phylogentic analysis indicated that each of the six metagenomes were predominately bacterial (> 99%), with a small portion of the phylogenetic reads annotated as archaeal and fungal (fungal sequences were not detected within the Spotted Lake metagenomes; Table 2). The number of reads (postquality control) obtained per sample is indicated in Table 2; normalized data sets were used for all subsequent analyses. More genera were identified, based on 16S rRNA gene sequence, within the Liberty and Spotted Lake postselection metagenomes relative to the respective preselection metagenomes. This is likely due to the disparity in sequencing depth, resulting in an underestimation of richness in the preselected consortia (smaller) metagenomes. There was little change in the diversity and evenness of the communities following selection. Overall, community richness decreased following selection within the Leeches and Liberty Lake metagenomes (Chao1). The converse was observed for Spotted Lake; however, this is likely an artifact of unequal sequencing depth (Tables 2 and 3).
Several phylogenetic markers (16S rRNA gene sequence, gyrB, recA, rpoB and rpoD) were mined from the six metagenomes, and each yielded a different richness profile (Figs 2 and 3), likely due to disparities within the publically available databases. Although the proportions varied with the gene under study, the microbial assemblages were dominated by Proteobacteria and Firmicutes. Overall, as indicated in the Results section, the relative proportions of the taxa appeared to more obviously shift after selection in the fresh water-derived samples as compared to the brine- and alkali-lake consortia (Fig. 2, Table 3). Within the fresh water sediment-derived assemblages, freeze-thaw treatment selected for Bacillus (Bacilli; 0.6–63.3% of total 16S rRNA gene sequences), largely at the expense of Pseudomonas (γ-Proteobacteria; 21.4–1.2%). In the alkali-lake sediment-derived enrichments, selection was not as extensive because the frequency of Bacillus increased from 52.5% to 61.5% (of the total 16S rRNA gene sequences). Similarly, the Chromohalobacter (γ-Proteobacteria) abundance increased modestly (32.6–46.8%) within the brine-lake-derived consortia, with slight decreases in a number of other genera. All these observations are consistent with the view that assemblages in high-osmotic environments are somehow preselected for freeze-thaw tolerance, and thus, changes in taxa profiles are less dramatic after selection when compared with the freshwater samples.
Preliminary analysis of the metagenomes also indicated changes in sequences associated with stress response. The number of metagenomic reads encoding proteins involved in general stress response was either constant (Liberty Lake) or decreased (Leeches and Spotted Lakes) within the postselection metagenomes. This was also observed for proteins involved in response to specific stresses such as acid, osmotic, and oxidative stresses, universal stress protein family, heat shock dnaK gene cluster extended, and SigmaB stress response regulation (Fig. 4).
However, our analysis of the KEGG pathways revealed an increase in metagenomic reads encoding proteins involved in signal transduction within all three of the post-freeze-thaw selection derived metagenomes. These included proteins in the OmpR family that are involved in osmotic and low-temperature stress response, biofilm formation, and sporulation, as well as proteins in the NarL family, which are also involved in osmotic and low-temperature stress response. Such sequences were more frequently identified within the postselection metagenomic reads relative to the metagenomes obtained from the preselected enrichments (Fig. 4). Therefore, proteins involved in signal transduction may help confer resistance to repetitive freeze-thaw stress. Neither antifreeze proteins (AFP) nor ice nucleation proteins (INP) sequences have been identified thus far, although it must be acknowledged that AFPs have highly diverse sequences.
Freeze-thaw resistance in soil enrichments has previously been shown to be affiliated with ice-association phenotypes (Walker et al., 2006; Wilson et al., 2006; Wilson & Walker, 2010), which would presumably increase the likelihood of survival. As a consequence, a subset of the isolates recovered from each lake were assayed to determine whether any phenotypes typical of freeze-thaw resistance in soils could also be found in isolates from the lakes of different chemistries. None of the isolates demonstrated IRI, but Alteromonadales EW4 and R. monicus LLW20, both from the brine-lake samples had some INA (Table S2). As might have been anticipated (Shahjee et al., 2002), a number of the genera recovered from the brine and alkali lakes have been previously associated with osmoprotectants or hyperosmotic environments, including Acinetobacter, Arthrobacter, Bacillus, Halomonas, Idiomarina, and Marinobacter, Nesterenkonia, and Paenibacillus (e.g. Mogilnaya et al., 2005; Naganuma et al., 2005). Isolates derived from the brine- and alkali-lake samples were five- and four-fold, respectively, more likely than isolates derived from freshwater lakes, to have an osmolyte content that was statistically greater than the E. coli TG-2 control. We suggest that this phenotype may partially explain the observed cross-tolerance.
Biofilms have also been implicated in both osmotic and low-temperature stress resistance (Williams et al., 2009), and again, the results were striking as all of the recovered isolates shown to form biofilms originated from the brine or alkali lakes (Table S2). Overall, 50–67% of the tested isolates from those lakes were associated with biofilms, while none of the isolates from the freshwater lakes produced biofilms under the conditions assayed. Moreover, two brine-lake isolates demonstrated multiple adaptations (INA, elevated osmolarity, and biofilm formation). Numerous adaptations would presumably serve to ensure survival under the most extreme conditions, and in fact, Alteromonadales EW4 is related to the extreme psychrophile Psychromonas ingrahamii, with multiple membrane, protein, and metabolite adaptations (Riley et al., 2008).
These experiments show that the consortium potential for increased freeze-thaw resistance was correlated with elevated solute concentration in the ‘home lake’, indicating that consortium adaptation to osmotic stress increases freeze-thaw survival. Freeze-resistant phenotypes in brine- or alkali-lake assemblages did not frequently have ice-association activities, but rather seemed to be dominated by high intracellular osmolyte concentrations, biofilm formation, and potentially spore formation, all of which are known to facilitate survival against freeze-thaw and osmotic stresses (Shahjee et al., 2002; Williams et al., 2009). Taken together, the data presented here indicate that there is cross-tolerance between freeze-thaw and osmotic stresses at the assemblage level, as has been hitherto shown in individual isolates (Schmidt & Zink, 2000; Schmid et al., 2009).
Acknowledgements
An NSERC (Canada) grant to V.K.W. and NSERC scholarship to S.L.W. financially supported this work. The Analytical Services Unit, Queen's University conducted the water chemistry analysis, and A. Kanawaty and S. Franchuk are thanked for their assistance with this, and for help with a pilot study. B. Momciu, R. Murray, and A. Stanczak are also acknowledged for their work on that pilot study, and M. Chalifoux and T. Vanderveer are acknowledged for technical assistance. Dr. B. Tufts is thanked for the use of his vapour pressure osmometer, Dr. G. Palmer is greatly thanked for technical knowledge and for his assistance with sample collection. As well, we are very grateful to the tenants and landowners, including the Syilx Okanagan Nation Alliance, for giving us access to the sites.
References
Supporting Information
Additional Supporting Information may be found in the online version of the article:
Author notes
Present address: Sandra L. Wilson, Department of Biological Sciences, University of Calgary, Calgary, AB, T2N 1N4, Canada
Corey Frazer, Department of Biomedical and Molecular Sciences, Queen's University, Kingston, ON, K7L 3N6, Canada
Editor: Max Haggblom
