Correspondence: Julia M. Foght, Department of Biological Sciences, University of Alberta, CW-405 Biological Sciences Building, Edmonton AB, Canada T6G 2E9. Tel.: +1 780 4923279; fax: +1 780 4929234; e-mail: email@example.com
Viable microorganisms are present in subglacial waters and sediment-laden ice beneath John Evans glacier in the Canadian high Arctic. The Bacterial communities resident in three subglacial samples were examined by amplifying 16S rRNA genes extracted from community DNA and from axenic isolates. Restriction fragment length polymorphism analysis of 341 clones produced 153 operational taxonomic units (OTUS), of which 25 dominant OTUS were sequenced. A subglacial water sample yielded Betaproteobacteria (25% of clones, particularly Comamonadaceae), Bacteroidetes (23%, particularly Flavobacterium) and Actinobacteria (14%). A second water sample had 51%Betaproteobacteria, 5%Bacteroidetes and no Actinobacteria, and a sediment sample was dominated by Betaproteobacteria (15%) and Bacteroidetes (38%). A collection of 158 morphologically distinct isolates was obtained on R2A agar using three incubation conditions: fully aerobic at 20°C or 4°C, or microaerobic at 20°C. A total of 52 isolate OTUs were defined, comprising Bacteroidetes (predominantly Flavobacterium isolated at 4°C), Betaproteobacteria (particularly Comamonadaceae), plus Actinobacteria and Alpha- and Gammaproteobacteria not detected as clones. Otherwise, the clone library and isolate collection results were quite comparable and supported earlier molecular studies at this site. Although additional undescribed diversity likely exists in these samples, combining culture-based results with molecular analysis increased the observed bacterial diversity and confirmed previous observations at this glacier and others.
Contrary to previous views that geochemical processes beneath glaciers and ice sheets are exclusively abiotic (e.g. Raiswell, 1984), there is recent and growing evidence of widespread microbial life in permanently cold and dark subglacial environments (e.g. Sharp et al., 1999; Skidmore et al., 2000, Foght et al., 2004). Subglacial systems of warm-based glaciers have liquid water in at least some regions at the glacier base, potentially permitting relatively rapid biochemical activity compared with microorganisms cryopreserved in overlying glacial ice at temperatures below freezing. The subglacial drainage systems are associated with sediments at the glacier bed that can provide organic and inorganic nutrients for microbial activity, again contrasting with overlying glacial ice. These physical and chemical characteristics have implications for the distribution of contemporary microbial life in permanently cold environments such as beneath glaciers and vast ice sheets in Antarctica and Greenland, as well as for archaic global carbon budgets during glaciation (Skidmore et al., 2000).
John Evans glacier on Ellesmere Island, Nunavut, Canada, has been studied because it has unfrozen waters and sediments at its base, and has been shown to harbour viable microorganisms including aerobic heterotrophs, nitrate- and sulphate-reducers, and methanogens (Skidmore et al., 2000). Dividing cells associated with sediment particles were also observed, indicating that bacteria within these samples were active (Skidmore et al., 2000). Previous studies at this site used only molecular methods, namely hybridization, clone libraries (Skidmore et al., 2005) and terminal restriction fragment length polymorphism analysis (T-RFLP; Bhatia et al., 2006) to estimate bacterial diversity in subglacial samples. The former study suggested that microbial community composition was influenced by the subglacial geochemistry. The latter analysis showed that the microbiota of John Evans subglacial environments (i.e. water, sediment and basal ice) were distinct from microorganisms in the supraglacial and proglacial environments (on top of and in front of the glacier, respectively), suggesting that the subglacial environment selects for a distinct subset of microorganisms adapted to in situ conditions (Bhatia et al., 2006). Therefore, to study the subsurface properly, a variety of subglacial samples must be examined and neither proglacial nor supraglacial samples, despite their easy accessibility, can substitute for true subglacial material. Hence, for the current study we selected three types of samples to represent the subglacial environment, including two liquid water samples presumably with a currently active microbiota and one frozen basal ice sample with entrained sediment, presumably with a formerly active microbiota.
Cultivation-based analysis of John Evans glacier samples to date has been limited to determining activities in mixed enrichment cultures (e.g. nitrate reduction, acetate mineralization; Skidmore et al., 2000) and detailed study of axenic isolates has not yet been conducted. Our objective for this study was to use cultivation-dependent examination of bacterial isolates to complement parallel and previous molecular analyses, recognizing that both approaches have biases and limitations and that combining these methods would broaden the current understanding of microbial diversity beneath this glacier. This work has also generated an isolate collection on which to perform future physiological tests to study the effects of in situ conditions on growth and, conversely, to examine the possible effects of the microbiota on subglacial biogeochemistry.
Materials and methods
Sample sites and processing
John Evans glacier is a polythermal glacier located at 79° 40′ N, 74° 00′ W in the Canadian high Arctic on Ellesmere Island, Nunavut (Skidmore et al., 2000). A subglacial reservoir near the glacier toe stores liquid water in contact with fine sediments overwinter (Skidmore & Sharp, 1999). When seasonal meltwater enters the reservoir via a large crevasse field, it pressurizes the reservoir until an ice barrier bursts, creating drainage outlets at the glacier surface and snout. This allows the stored subglacial waters, which are progressively diluted by the surface meltwaters, to exit first through a pressurized stream and an artesian fountain, then by a channelized outflow from the front of the glacier. These outflow events permit aseptic collection of subglacial materials (Bhatia et al., 2006) that otherwise would only be accessible by drilling through tens or hundreds of meters of glacial ice. In summer 2002, samples were collected from the subglacial initial burst (SIB) representing sediment-rich water in contact with the glacier sediments overwinter, and from the subglacial outflow channel (SOC), representing overwintered subglacial water diluted to an unknown degree by contemporary surface meltwaters. We also collected frozen sediment-rich basal ice from the junction of John Evans glacier and the tributary Fox glacier (basal Fox ice; BF) where basal ice and sediment from beneath the glacier had been transported to the glacier margin, making it accessible for aseptic surface sampling. These sample sites, methods for aseptic collection of materials and in situ sample handling have been described (Bhatia, 2004; Bhatia et al., 2006).
All samples were thawed in the laboratory at 4°C and processed aseptically in a UV-sterilized biohazard hood under HEPA-filtered airflow as previously described (Bhatia et al., 2006). For SIB and SOC waters, 1 L of melted sample was passed through a disposable, sterile analytical filter unit containing a removable 0.2 μm cellulose nitrate membrane (09-740-21A, Nalgene). The filter unit was disassembled and the filter was aseptically divided into quarters using a sterile scalpel blade and placed in sterile bead-beating tubes (Biospec Products, Bartlesville, OK). Fractions of the same filter (typically quarters) were used for genomic DNA extraction to construct clone libraries, and for cultivation-based study (described below) and T-RFLP studies (Bhatia et al., 2006).
BF basal ice was transferred to and thawed in twice-autoclaved covered glass beakers, resulting in a clear layer of water overlying fine silty sediment and coarser debris from the glacier bed. The supernatant was filtered for DNA extraction and clone library construction, and the sediment was used directly for cultivation of isolates.
Clone library construction and RFLP analysis
Total genomic DNA was extracted from the filter sections or from 0.3 g (wet weight) of BF sediment using a Model B110BX mini bead-beater (BioSpec Products) with equal masses of 0.1-mm and 2.5-mm diameter zirconium beads (Fisher Scientific, Canada) using methods and precautions against contamination described by Bhatia et al. (2006). Bead beating conditions were based on preliminary tests to determine optimum lysis times (Cheng, 2005), specifically, 40 s at 5000 reciprocations per minute at room temperature using the lysis buffer previously described (Foght et al., 2004). Trace soil contaminants were removed from the extract using GENECLEAN® (Bio 101, Carlsbad, CA) according to the manufacturer's directions, with final DNA recovery in 40 μL of sterile Milli-Q water. Parallel negative control reactions with reagents were performed routinely.
Near full-length 16S rRNA genes were amplified from extracted genomic DNA using the bacterial primers PB36 and PB38 (Table 1) in a final PCR volume of 50 μL containing 5 μL of extracted DNA (∼25 ng), 0.5 μM of each primer, 50 mM Tris HCl (pH 9.0), 1.5 mM magnesium chloride, 0.4 mM β-mercaptoethanol (BDH Laboratory Supplies), 5 μg nuclease-free bovine serum albumin (Roche Diagnostics), 10 mM ammonium sulphate, 0.2 mM each of deoxynucleotide, 5% dimethylsulfoxide (DMSO) and 0.5 U Taq (Roche Diagnostics). PCR was performed, consisting of an initial denaturation of 4 min at 95°C followed by 30 cycles of 39 s at 93°C, 60 s at 54°C, 120 s at 73°C, and followed by a final extension at 73°C for 10 min. PCR product (100 μL) was purified using a High Pure PCR Product Purification Kit (Roche Diagnostics) and cloned into Escherichia coli JM109 using a pGEM-T Easy Vector cloning kit (Promega) with blue-white colony selection.
Table 1. Primers used to obtain near full-length 16S rRNA gene sequences
The cloned 16S rRNA gene fragments were re-amplified from clones for RFLP analysis using M13 forward and reverse primers (Table 1) as for initial amplification except that 5 μL of bacterial clone suspension and 1.2 μM each of forward and reverse M13 primer were used. Reactions were performed for two cycles (90 s at 94°C, 45 s at 56°C, 90 s at 72°C), followed by 22 cycles (30 s at 90°C, 30 s at 56°C, 60 s at 72°C) and a final extension at 72°C for 10 min. The amplified PCR product was digested using either HaeIII or CfoI (Roche Diagnostics) according to manufacturer's directions, the fragments separated on a 2% agarose gel and the fragment sizes calculated by comparison with 100 bp molecular weight markers (Roche Diagnostics) included as two or three lanes per gel. RFLP patterns were visually inspected after digital image capture and categorized into operational taxonomic units (OTUs) based on the combination of HaeIII and CfoI RFLP patterns, taking into account inter- and intragel variations. Calculated RFLP molecular weights for each OTU were based on mean band sizes from replicate lanes within the same gel and different gels. Standard deviations from the mean fragment size were calculated and were typically about 20 bp. In most cases the total length of the fragments equalled the expected amplicon length, but for some RFLP patterns the low molecular weight bands (<100 bp) were not visible on the gel, so the total size was less than ∼1500 bp.
Cultivating bacterial isolates
Filter sections from SOC or SIB water samples were placed into 9 mL of sterile 3 mM potassium phosphate buffer (pH 7) at 4°C in test tubes containing c. 0.5 g of 3-mm diameter glass beads. Each tube was mixed vigorously by vortex at high speed for 1 min, after which no material was visible on the membrane. Five 10-fold serial dilutions were performed immediately using the same cold buffer. Three sets of triplicate plates of R2A agar (Difco) at 4°C were each inoculated with 0.1 mL of dilution and spread with chilled sterile glass spreaders. Plates were incubated at one of three conditions in the dark to obtain viable numbers: aerobically at room temperature (c. 20°C, designated A20 isolates); aerobically at 4°C (A4); or microaerobically at 20°C (M20) in sealed chambers (BBL GasPak Systems; BD, Franklin Lakes, NJ) incubated with Anaerocult C sachets (Merck, Darmstadt, Germany) that produce a headspace with 5–6% O2 and 8–10% CO2. Each time the chambers were opened to inspect the plates, new sachets were installed. After incubation for 27 days, morphologically distinct colonies (based on colour, texture, size, margin, opacity, etc.) were purified by transfer onto fresh plates. The isolate collection was preserved in 25% glycerol at −80°C.
BF sediment 1 g (wet weight) was diluted in 9.0 mL of 0.1% sodium pyrophosphate buffer (pH 7.0; 4°C), serially diluted in the same cold buffer and plated as above in two parallel sets. One set was incubated aerobically at 4°C (A4) and the other aerobically at 20°C (A20). Morphologically distinct isolates were purified and preserved as described above.
PCR amplification and analysis of 16S rRNA genes from isolates
Well-isolated colonies on R2A agar were used directly for PCR by preparing a cell suspension in TE buffer (10 mM Tris HCl, pH 9.0; 1 mM EDTA, pH 8.0) and amplifying 16S rRNA genes using the PB36 and PB38 primers as described above. Occasionally, when this direct template procedure was unsuccessful, genomic DNA was extracted from the pure culture using bead-beating as described above, and the extracted DNA subsequently used as template for 16S rRNA gene amplification. Amplified DNA was digested and categorized into OTUs as for the clone libraries, except that the 5′ and 3′ fragments lacked the M13 primer sequences present in the clone RFLP patterns.
Sequencing and phylogenetic analysis
Typically, OTUs containing two or more representatives were chosen for sequence analysis. 16S rRNA genes amplified from clones or isolates were cleaned using either the Qiaquick PCR purification kit (Qiagen) or High Pure PCR purification kit (Roche Diagnostics). Sequencing reactions were carried out using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems Instruments, Foster City, CA; ABI) in a final volume of 20 μL using 90 ng of purified PCR product, 0.25 μM of each sequencing primer, 1X sequencing dilution buffer [200 mM Tris HCl (pH 9.0), 5 mM magnesium chloride] and 2 μL of Big Dye terminator sequencing premix (ABI). To confirm that clones or isolates assigned to an OTU were indeed phylogenetically related, multiple representatives of several OTUs were sequenced. Sequences showed 97% or greater similarity to each other after pairwise analysis (data not shown). Based on these results, clones or isolates within the other OTUs were also assumed to be phylogenetically coherent and representatives of OTUs were selected at random for sequencing.
Nearly full-length bi-directional 16S rRNA gene sequences were obtained using seven sequencing oligonucleotide primers (Table 1). Sequencing reactions were performed for 25 cycles (20 s at 95°C, 15 s at 58°C, 60 s at 68°C) and the product was cleaned by ethanol precipitation. Capillary electrophoresis was performed in the Molecular Biology Service Unit (Department of Biological Sciences, University of Alberta) using an ABI 3100 genetic analyzer (ABI). Sequences have been deposited in GenBank under accession numbers DQ628916-DQ628970 and DQ530258.
Sequences were aligned using the pregap and gap4 v3.0 programs from the Staden Package software. Sequence data processing and analysis were conducted using the emboss suite, and phylip software package v 3.6a3 (Felsenstein, 1989). Bioinformatics tools were accessed through the Canadian Bioinformatics Resource (http://cbr-rbc.nrc-cnrc.gc.ca/). The program chimera_check (http://rdp.cme.msu.edu/) was used to confirm that none of assembled clone sequences was chimeric. The best matches were selected from GenBank by blastn search (http://www.ncbi.nlm.nih.gov/blast/) and these known sequences were used to construct phylogenetic trees using the Kimura 2-parameter model for nucleotide change (Kimura, 1980). Bootstrap values (100 replications) were generated by the neighbour joining method.
Culture-independent RFLP analysis
Culture-independent RFLP analysis was performed on 341 clones obtained from the SOC, SIB, and BF samples. These represented a total of 153 OTUs, 32 (21%) of which contained two or more clones, with the rest comprising a single clone (‘singletons’). One or more representatives from 25 abundant OTUs were selected for near-full-length (c. 1450 bp) bi-directional 16S rRNA gene sequencing, whereas the singletons were not sequenced.
The SOC clone library contained the largest number of clones (159) and OTUs (73). The 13 OTUs selected for sequencing represented 62% of the SOC clone library, and were most similar to members of the Betaproteobacteria, Bacteroidetes and Actinobacteria, constituting 25, 23 and 14% of the clone library, respectively (Table 2). Clones belonging to the Betaproteobacteria typically had ∼99% identity to bacteria isolated from aquifers, from sediment beneath two New Zealand glaciers and to uncultured clones from a high mountain lake in Austria and contaminated sediments (Table 2), and clustered with psychrotolerant members of the Comamonadaceae (Fig. 1). Two Betaproteobacteria OTUs with a total of 20 clones clustered with Rhodoferax spp. and other glacier bacteria (Fig. 1), whereas two others (total 21 clones) clustered with Polaromonas spp. SOC clones belonging to the Bacteroidetes phylum fell into four clades (Fig. 2), the majority of which (total 20 clones) were affiliated with uncultured Flectobacillus clones. Clones from all four Actinobacteria-affiliated OTUs fell within two poorly-defined clusters of unnamed and uncultured clones (Fig. 3) derived from freshwater and estuary communities. Sixty singletons representing 38% of the library were not sequenced. Interestingly, using blastn, none of the sequences deposited in GenBank from previous clone libraries of John Evans material (Skidmore et al., 2005) was the best match for the current sequences; therefore those sequences do not appear in the phylogenetic trees.
Table 2. Identity and distribution of sequenced representative clones from the most abundant OTUs within the SOC, SIB and BF clone libraries, and hypothetical T-RF sizes corresponding to T-RFs detected in the same sample material
The SIB clone library of 92 clones yielded 39 OTUs, of which 8 OTUs containing multiple clones were selected for sequencing. More than half of the clone library (51%) was related to members of the Betaproteobacteria, with the dominant OTU (represented by SIB2 1E, affiliated with Rhodoferax spp.) alone accounting for 20% of the SIB clone library (Table 2). Other SIB Betaproteobacteria OTUs fell in the genus Polaromonas and matched either glacier isolates or uncultured clone sequences, often clustering with other OTUs from SOC and BF (Fig. 1). The single sequenced SIB OTU affiliated with the Bacteroidetes represented 5% of the clone library, had 99% homology to Flavobacterium omnivorum isolated from a frozen glacier soil (Table 2) and clustered with SOC and BF clones (Fig. 2).
The BF clone library contained 90 clones comprising 50 OTUs of which nine were sequenced. As with the SIB clone library, the Betaproteobacteria (15%) and Bacteroidetes (38%) dominated the library. One prevalent Bacteroidetes OTU (represented by clone BFA 7C and affiliated with Flavobacterium spp.; Fig. 2) alone represented 26% of the BFA clones (Table 2). This OTU was also detected as 4–5% of the other two libraries (Table 2) and appears to be a widespread phylotype in the subglacial samples. Most BF OTUs grouped phylogenetically with other clones and isolates from the BF, SIB and SOC samples (Figs 1–3) or with psychrophilic or psychrotolerant isolates from frozen glacial soil, Antarctic cryoconite sediment, or members of Antarctic microbial mats, indicating that these may also be phylotypes that are widely distributed in permanently cold environments.
Culture-dependent RFLP analysis
Viable counts ranged from 101 to 107 CFU mL−1 of subglacial water or melted ice after 27 days incubation on R2A agar, with the greatest viable numbers arising from the sediment-rich BF sample (106–107 CFU mL−1 melted ice) and lower numbers from the water samples. These ranges correspond to the viable numbers reported for aerobic heterotrophs beneath New Zealand glaciers, i.e. 101–102 CFU mL−1 of ice and 105–106 CFU g−1 dry weight of subglacial sediment (Foght et al., 2004). Plates from the SOC sample yielded more CFU at 4°C than at 20°C, whereas the SIB sample showed the opposite. Generally, fewer CFU were recovered under microaerobic conditions than under fully aerobic conditions. A higher proportion of colonies from the SOC and SIB water samples were pigmented yellow, orange or red (90–95%) compared to the BF basal ice (11%), despite the generally higher viable numbers cultivated in the latter sample.
Colonies with visibly different morphologies were selected from plates, purified by passage on fresh plates, and preserved as an isolate collection for RFLP analysis. Some isolates, initially isolated only under one condition (e.g. aerobic at 4°C), were able to grow under additional incubation conditions (e.g. at higher temperature) upon transfer, as observed by Christner et al. (2003). In many cases, morphologically distinct colonies subsequently yielded the same RFLP pattern and therefore contributed to some OTUs having more than one sequenced representative. Unfortunately, many axenic isolates from the SIB and BF samples failed to transfer or were contaminated with fungi during purification and preservation, leaving a total of 158 morphologically distinct isolates representing 52 OTUs, mostly from the SOC sample.
The three different incubation conditions (A20, A4 and M20) generally cultivated isolates within different OTUs (Fig. 4) from the SOC sample. Only two OTUs arose under all three incubation conditions: OTU1, represented by isolate SOC A20(40) and most closely related to Flavobacterium spp. in the Bacteroidetes (Fig. 2); and OTU4, represented by SOC A20(17) in the Polaromonas group of the Betaproteobacteria (Fig. 1). These two OTUs were abundant in the SOC culture collection (Table 3), suggesting that they are versatile and easily cultured. RFLP analysis of 134 morphologically distinct colony types in the SOC isolate collection defined a total of 39 OTUs, comprising 21 OTUs (71 isolates) isolated aerobically at 20°C, 12 OTUs (39 isolates) aerobically at 4°C and 12 OTUs (23 isolates) microaerobically at 20°C (Table 3 and Fig. 4). Eleven OTUs, comprising 106 colony types and representing 79% of the SOC isolate collection, were chosen for sequencing. The SOC isolate collection comprised 46%Bacteroidetes, 17%Betaproteobacteria, 11%Alphaproteobacteria, 3%Actinobacteria, 2%Gammaproteobacteria and 21% unsequenced singletons. The most commonly isolated OTU (27% of the SOC isolate collection, represented by isolate SOC A20 (40)) clustered with Flavobacterium spp., some of which appear to be misnamed (e.g. Chryseobacterium sp. TB4-8-II; Fig. 2). Another well-represented OTU [SOC A4(12); comprising 16% of the SOC isolates] was isolated only at 4°C and was affiliated most closely with Flavobacterium psychrophilum (Fig. 2; Table 3). The prevalence of these Flavobacterium OTUs in the SOC isolate collection indicates that they produce a variety of morphologically different colony types, resulting in multiple depositions in the isolate collection. The second most abundant group of SOC isolates were members of the Comamonadaceae family of Betaproteobacteria (Fig. 1; Table 3), all of which were most closely related to bacteria found in aquifers or contaminated groundwater. The remainder of the sequenced SOC representatives belonged to the Alpha- and Gammaproteobacteria and Actinobacteria, clustering with genera such as Brevundimonas and Arthrobacter found in samples from freshwater, polar or oligotrophic environments.
Table 3. Identity and distribution of sequenced isolates from the most abundant OTUs within the SOC isolate collection, and hypothetical T-RF sizes corresponding to T-RFs detected in the same sample material
The isolate collections from SIB water and BF basal ice samples were very limited because of loss and contamination during transfer and preservation. Only 20 morphologically distinct isolates cultivated from SIB survived for RFLP analysis. A total of 11 OTUs were detected, of which four sequenced representatives belonged to the Betaproteobacteria, Bacteroidetes, Actinobacteria and Staphylococcaceae. The two most common SIB OTUs clustered with Flavobacterium spp. and isolates from SOC (Fig. 2) and an Arthrobacter sp. from Antarctica (Fig. 3). Only a single OTU, represented by Clone SIB2 1C and Isolate SIBA4(8), overlapped between the SIB clone library and culture collection.
Only four morphologically distinct isolates were maintained from the BF sample: all four were sequenced and each represented a different OTU. Two were most similar to Kocuria and Arthrobacter spp. (Fig. 3), one was related to the Comamonadaceae, clustering with other subglacial clones and isolates (Fig. 1), and one was most similar to an uncultured Alphaproteobacterium clone (not shown). In contrast, the BF clone library was dominated by a Flavobacterium clone. The discrepancy between BF clones and isolates may be exacerbated by the fact that the BF ice separated into a supernatant and a sediment-rich slurry during melting; the clone library was prepared from the upper layer, whereas the sediment slurry was used for cultivation.
It is not valid to compare the microbiota of ancient glacial ice cores (e.g. Christner et al., 2000; Miteva et al., 2004) directly with those of John Evans subglacial waters and sediment because the former are permanently frozen and have limited contact with sediment, whereas the latter may be unfrozen, seasonally influenced by meltwater input, and have intimate, prolonged contact with mineral surfaces to support active metabolism in situ and generally higher numbers of viable microorganisms than ice cores. As well, microorganisms in glacial ice likely partition into solute-rich, discontinuous veins of liquid water surrounding ice crystals (Mader et al., 2006), whereas subglacial microorganisms live in low-nutrient bulk water subject to chemical flux. This may explain why the microbiota of ice cores differ in gross composition from subglacial waters and sediments. For example, Miteva et al. (2004) found that direct plating of Greenland deep ice core samples onto R2A agar yielded predominantly Proteobacteria and high-GC Gram-positive bacteria. Ancient ice cores are often dominated by sporulators such as Bacillus and Actinomycetes (Christner et al. 2000), likely reflecting the survival potential of their spores, but are minor components of John Evans subglacial communities, which presumably are active in situ. Nor is there a direct parallel between cryoconite holes and the subglacial system, because even though heterotrophs are present in cryoconite holes (e.g. Margesin et al., 2002; Christner et al. 2003) and may seasonally inoculate the subsurface, the cryoconite microbiota are dominated by contemporary photosynthetic productivity, whereas the subglacial environments are dark. Therefore, there is a limited set of data against which to compare the current results.
Subglacial clone library diversity
This study focussed on Bacteria and did not examine Archaea for two reasons: first, no parallel cultivation of Archaea (specifically, methanogens detected in earlier studies; Skidmore et al., 2000) was attempted because all samples were exposed to oxygen during processing; and secondly, previous attempts at amplifying Archaeal 16S rRNA genes from John Evans subglacial material were unsuccessful (Skidmore et al., 2005).
Only one DNA extraction from each sample was used to create the clone libraries: the amount of sample was limited because each filter was subdivided for T-RFLP analysis, clone libraries and viable count experiments. Had we used multiple filters to prepare clone libraries, pooled multiple PCR reactions before cloning, or picked and sequenced more clones, additional diversity may have been revealed. Finally, we extracted DNA only from the material retained on the filter, whereas ultrasmall bacteria present in ice core material pass through 0.2-μm filters (Miteva & Brenchley, 2005); our study would have missed such organisms.
Despite these potential limitations, accumulation curves (not shown) approached a horizontal asymptote for all three libraries, at ∼75 clones for SOC, 40 clones for SIB and 50 clones for BF, suggesting that the libraries (all containing ≥90 clones) sampled the species richness in the extracted DNA reasonably well. The Shannon–Weaver diversity indices (Shannon & Weaver, 1949) were 3.713, 3.143 and 3.317, respectively. Good's estimate of coverage (Good, 1953) was 62, 74 and 54% for the SOC, SIB and BF clone libraries, respectively, compared to 97% for the subglacial sediment clone library prepared by Skidmore et al. (2005).
The sequenced clones comprised only Bacteroidetes, Betaproteobacteria and Actinobacteria, agreeing with previous results from John Evans subglacial water and ice (Skidmore et al., 2005; Table 4), and indicating the probable importance of these phyla subglacially. Although Skidmore et al. (2005) detected a few clones affiliated with Alpha- and Gammaproteobacteria, Holophaga/Acidobacteria, Planctomycetales and Verrucomicrobium, none was detected in the current study. This may be because (1) we did not sequence singletons, whereas Skidmore et al. (2005) sequenced representatives of all OTUs; (2) different universal Bacterial 16S rRNA gene primers were used in the two studies; and (3) different subglacial materials from John Evans glacier were examined. Despite these differences, there is reasonable overlap between the previous and current clone libraries regarding dominant and absent phylotypes (Table 4) and correlation with previous hybridization studies (Skidmore et al., 2005), confirming the subglacial abundance of Comamonadaceae at John Evans.
Table 4. Comparison of isolate collection and clone library compositions from the current study and previous studies
Isolate collection from unfrozen sediment beneath Fox (FX) and Franz Josef (FJ) glaciers, New Zealand (Foght et al., 2004).
Clone library prepared from frozen sediment in basal ice beneath John Evans glacier, Canada (Skidmore et al., 2005).
‡ Percentage not calculated because of small number of isolates in collection due to loss and contamination (see text).
§ Basal ice and sediment from BF region but not identical to current study material.
In silico analysis of the 5′ fragments of sequenced clones and isolates was used to compare the current results with previous T-RFLP analysis (Bhatia et al., 2006) of the same sample material (i.e. subsections of the same filters). OTUs with hypothetical 5′ fragments corresponding to T-RFs (minus the T-RF primer sequence) are indicated in Tables 2 and 3. Good correlation between the three methods (T-RFLP, clone library and isolate collection) was obtained, with 24 OTUs having a corresponding T-RF.
Subglacial isolate collections
To reduce the likelihood of reisolating species with minor differences in colony morphologies on different media, we used only R2A agar and varied the incubation conditions. R2A was selected because in previous experiments we found it to be equal or superior to other low-nutrient media, as did Miteva et al. (2004) with Greenland ice core samples. Using two temperatures enabled isolation of both psychrotolerant and psychrophilic Bacteria. Microaerobic conditions were included because previous studies (Foght et al., 2004) showed that microaerophiles represented a considerable proportion of the viable subglacial aerobes. This decision was later justified by the isolation from SOC of OTUs that were not observed under fully aerobic conditions (Fig. 4), although upon subsequent transfer many of these isolates grew under fully aerobic conditions, suggesting that they were sensitive to oxygen only during initial isolation. The current isolate collection therefore is limited to aerobes, microaerophiles or facultative aerobes able to grow on R2A agar. Additional Bacterial (and possibly Archaeal) species surely would have been isolated if additional selective and differential media and incubation conditions had been employed.
Many of the colonies were pigmented yellow or orange, agreeing with previous results from subglacial samples (Foght et al., 2004) and ice cores (Miteva et al., 2004). Pigmentation has been proposed as a physiological adaptation to cold temperatures and carotenoids may be used to regulate membrane fluidity at low temperatures (Chattopadhyay & Jagannadham, 2001; Fong et al., 2001). However, RFLP analysis and sequencing did not reveal any correlation between pigmentation and the assigned OTU. For example, four very differently coloured isolates [SOC A4(12), SOC A4(40), SOC A4(44), and SOC A4(51)] had identical RFLP patterns with single and double digestions with CfoI and/or HaeIII, but after sequence analysis all four clustered with Flavobacterium spp. (data not shown). Therefore, the strategy of describing bacterial diversity by selecting colonial variants can be misleading. These biases, combined with incomplete cultivation from samples and loss of viability during purification and archiving, contribute to the widely-held opinion that culture-based studies are limited in the diversity that they can describe. Nevertheless, the isolate collections were more diverse than the clone libraries, detecting OTUs that were not apparent from the culture-independent analysis (i.e. Alpha- and Gammaproteobacteria; Table 4) and demonstrating that culture-based study did augment molecular analysis.
The John Evans isolate collection resembles a suite of 37 aerobic isolates cultivated from unfrozen sediment beneath two New Zealand glaciers (Foght et al., 2004; Table 4), and in fact several John Evans glacier isolates clustered phylogenetically with Fox and Franz Josef isolates (Figs 1 and 2; glacier isolates with prefixes FX or FJ). In that study, Betaproteobacteria predominated, with many phylogenetically affiliated to Polaromonas, Rhodoferax and uncultivated clones from cold environments, as in the current study. Fewer Bacteroidetes, predominantly Flavobacterium spp., and only five Actinobacteria were isolated, again in parallel with John Evans glacier. Although Skidmore et al. (2005) showed that glaciers with different lithologies can harbour distinct microbiota, the overlap of isolates in these two geographically distant glaciers is remarkable and may indicate important trends in subglacial communities.
Although physiology cannot be inferred from 16S rRNA gene phylogeny with confidence, a few cautious observations can be made regarding trends in the detection of subglacial phylotypes. The abundance of OTUs affiliated with phyla commonly detected in oligotrophic freshwaters, polar regions and glacial environments suggests that low nutrient availability is as much a selective condition beneath John Evans glacier as low temperature. Growth temperature ranges were determined for some isolates by spotting standard inoculum suspensions on R2A agar and incubating at temperatures of 4–35°C (data not shown), but only a few putative psychrophiles were isolated; the majority were psychrotolerant (capable of growth at ≤4°C but having optima ≥20°C). Similarly, few clones were affiliated with obligate psychrophiles, although these two observations could be due to the bias of cultivation or limited numbers of psychrophilic sequences available for comparison in GenBank.
The diversity of aerobic isolates and detection of abundant clones affiliated with strictly aerobic genera (e.g. Arthrobacter, Polaromonas, Flavobacterium) suggests that sufficient oxygen is available in situ to sustain aerobic or microaerobic subglacial communities, perhaps supplied through release of dissolved oxygen from melting glacier ice, seasonal input of oxygen-saturated supraglacial melt waters or through chemical weathering reactions (Bhatia et al., 2006).
Bacteroidetes were abundant in the isolate collection and in both the current and previous clone libraries (Table 4). Members of this phylum, particularly Cytophaga and Flavobacterium, are often well-adapted to low temperature environments (Noble et al., 1990), tend to form aggregates and surface biofilms on small particles (DeLong et al., 1993) and also efficiently metabolize dissolved organic carbon (Cottrell & Kirchman, 2000; Kirchman, 2002). It would be interesting to test the ability of the subglacial isolates to adhere to and form biofilms on sediment particles, as other cold-tolerant organisms have been shown to produce extracellular polymers (Miteva et al., 2004) that may function as cryoprotectants, or in biofilm formation or transient nutrient sequestration. Testing the relative and synergistic effects of nutrient composition and concentration vs. low temperature may be revealing. The existence of the subglacial isolate collection now provides an opportunity to perform such physiological studies.
Betaproteobacteria were also abundant in the subglacial community. This phylum is commonly detected in oligotrophic freshwater systems such as glacial stream runoff (Battin et al., 2001), and several genera within the Betaproteobacteria family Comamonadaceae have been found in polar and alpine environments (Skidmore et al., 2005), suggesting that they are important members of glacial microbial communities. Within that clade, several clones (but no isolates) clustered closely to Rhodoferax spp. (Fig. 1), which are facultative phototrophs that respire anaerobically with nitrate and Fe(III). Such activities could be an asset in surviving transfer from the glacier surface with seasonal meltwaters into nitrate-containing subglacial waters in contact with sediments (Bhatia et al., 2006).
Neither the current nor previous clone libraries (Table 4) detected strictly anaerobic Bacteria such as sulphate-reducers (i.e. Deltaproteobacteria), even though sulphate-reducing activity previously was observed in analogous cultivated subglacial samples (Skidmore et al., 2000). This lack of correlation between sulphate-reducing activity and molecular detection of appropriate phylotypes may reflect low numbers in situ, and anaerobic enrichment cultures will be required to detect these bacteria by cultivation. We did, however, detect some potential facultative nitrate-reducers (e.g. clones affiliated with Rhodoferax spp.), supporting the previous observation of nitrate-reducing activity in John Evans subglacial samples (Skidmore et al., 2000).
A multiphasic approach combining mixed culture enrichments (e.g. Skidmore et al., 2000), DNA hybridization (Skidmore et al., 2005), T-RFLP analysis (Bhatia et al., 2006), clone libraries (Skidmore et al., 2005) and isolate collections (current study) has broadened our view of the diversity of Bacteria living beneath John Evans glacier and provided confirmation of previous observations. It also provides an expanded 16S rRNA gene database for comparison with microbiota in other permanently cold and dark environments. The current results will help direct future studies towards detection of metabolic classes, including sulphate-reducers and methanogens, that are predicted by geochemistry to exist subglacially but that to date have evaded molecular detection. Although the combined selective pressures of limited carbon, nutrients, light and oxygen, and near freezing temperatures could potentially limit microbial abundance and diversity, the subglacial environment at John Evans glacier holds considerable diversity, revealed by combining culture-independent and -dependent analyses.
We gratefully acknowledge funding from NSERC (JF) and Sigma Xi The Research Society (SMC). Thanks to M. Bhatia for sample collection and assistance with sample processing, to E. Marques, L. Rear, L. Sullivan and S. Lou for assistance with the isolate collections and to J. Klassen for advice on phylogenetic analysis. Sequencing was conducted in the Molecular Biology Services Unit (Biological Sciences Department, University of Alberta).