Water mass and depth determine the distribution and diversity of Rhodobacterales in an Arctic marine system


Correspondence: Andrew S. Lang, Department of Biology, Memorial University of Newfoundland, St. John's, NL A1B 3X9, Canada. Tel.: +1 709 864 7517; fax: +1 709 864 3018; e-mail: aslang@mun.ca


Marine Rhodobacterales are recognized as a widespread, abundant, and metabolically versatile bacterial group in the world's oceans. They also show a nearly universal conservation of the genes for production of gene transfer agents (GTAs), virus-like particles that mediate genetic exchange between cells. It is not yet clear what factors determine the distribution of the various taxonomic subgroups of this order. To address this question, we analyzed the Rhodobacterales communities in 10 seawater samples from northern Baffin Bay collected during September 2008. A conserved gene from the GTA gene cluster was used to characterize the Rhodobacterales community structure. A total of 320 clones from 10 clone libraries were sequenced, and 22 operational taxonomic units representing putative species and 13 clusters representing putative genera were identified. A cluster related to Octadecabacter comprised 59% of total clones from the northern Baffin Bay. Phylogenetic analysis of the clones showed that the Rhodobacterales communities had distinct compositions in the different water masses that were sampled. A change in community structure related to depth was also observed. Therefore, in northern Baffin Bay where two ocean currents meet and mix, the Rhodobacterales community structures were primarily determined by water mass and depth.


Marine heterotrophic picoplankton, such as bacteria, display biogeography patterns over global and local scales. Nutrients, temperature, and mortality are the main determinants in the global patterns of distribution of marine bacteria (Pommier et al., 2007; Fuhrman et al., 2008; Taniguchi & Hamasaki, 2008), while at the local scale, the primary factors controlling bacterial diversity and distribution are more complex and variable and include physical factors such as temperature, salinity (Bouvier & del Giorgio, 2002), depth (Brown et al., 2009), and overall hydrography (Pinhassi et al., 2003), as well as biotic factors such as nutrient availability, rates of mortality, and the phytoplankton community (González et al., 2000). Some bacterial lineages are cosmopolitan, such as SAR11 (García-Martínez & Rodríguez-Valera, 2000; Morris et al., 2002), but most taxa are more restricted (Pommier et al., 2007; Ghiglione et al., 2012). Although the biogeography of marine microorganisms is actively studied [e.g. (Galand et al., 2010; Ghiglione et al., 2012; Morris et al., 2012)], both the deterministic factors as well as the associated environmental correlates are not fully understood (Nemergut et al., 2011; Hanson et al., 2012).

Bacteria in the order Rhodobacterales, especially the Roseobacter clade, are widespread and abundant in marine environments (Buchan et al., 2005; Brinkhoff et al., 2008; Giebel et al., 2009). The characterized and cultured representatives from marine systems show a large diversity of physiological attributes that allow them to exploit their environment (Buchan et al., 2005; Wagner-Döbler & Biebl, 2006; Newton et al., 2010). These include aerobic anoxygenic photosynthesis (Shiba, 1991; Allgaier et al., 2003), carbon monoxide oxidation (Moran et al., 2007; Cunliffe, 2011), dimethylsulfoniopropionate (DMSP) metabolism and transformation (González et al., 2000; Vila-Costa et al., 2010), symbiosis with and pathogenesis of eukaryotes (Green et al., 2004; Webster et al., 2004; Seyedsayamdost et al., 2011), degradation of hydrocarbons (Buchan & González, 2010), and secondary metabolite production (Brinkhoff et al., 2004; Martens et al., 2007).

The Rhodobacterales additionally show a nearly universal presence of the gene cluster for producing gene transfer agents (GTAs) (Lang & Beatty, 2007; Biers et al., 2008; Paul, 2008; Newton et al., 2010; Lang et al., 2012; McDaniel et al., 2012). GTAs are bacteriophage-like particles that transfer cellular genomic DNA between cells (Stanton, 2007; Lang et al., 2012). There is evidence that production of GTAs is widespread in this group of bacteria, with GTA-mediated gene transfer demonstrated for Ruegeria pomeroyi DSS-3 (Biers et al., 2008) and release of GTA structural protein to the extracellular environment observed in multiple lineages within the order (Fu et al., 2010). An additional three Rhodobacterales isolates, Ruegeria mobilis 45A6, Roseovarius nubinhibens ISM, and Nitratireductor sp. 44B9s, make GTAs capable of gene transfer to different genera of bacteria (McDaniel et al., 2010, 2012). The abundance of Rhodobacterales in marine environments coupled with their high conservation of the GTA genes has contributed to the speculation that presence of GTAs in marine viral communities could explain the high proportion of cellular genes that are detected in viral metagenomic datasets (Kristensen et al., 2010; Lang & Beatty, 2010). The possible role of Rhodobacterales in promoting genetic exchange in natural environments makes them even more compelling model microorganisms.

Despite the assumed ecological importance of Rhodobacterales, their distribution patterns and their response to environmental drivers are poorly understood. However, it is known that distinct distribution patterns exist. For example, the RCA cluster [also referred to as DC5-80-3 (Buchan et al., 2005)] is abundant in polar oceans (Giebel et al., 2009, 2011), but has not been reported from tropical/subtropical environments (Selje et al., 2004). Although such distribution patterns are known for selected representatives of the Roseobacter clade, the patterns for most Rhodobacterales members have not been described. Several studies have used the 16S rRNA gene marker to track changes in the relative abundance of total Rhodobacterales communities in the South Atlantic basin (Morris et al., 2012) and during a coastal six-year time series (Gilbert et al., 2012), but such an approach fails to capture the finer detail expected within this physiologically diverse order. The g5 gene in GTA gene cluster encodes the major capsid protein of GTA particles. It has been used as a higher resolution phylogenetic marker to document the seasonal and spatial patterns in the structure of Rhodobacterales communities in both temperate and sub-Arctic locations (Zhao et al., 2009; Fu et al., 2010). This sequence generally produces a similar phylogeny with 16S rRNA gene (Lang & Beatty, 2007; Biers et al., 2008; Zhao et al., 2009) but provides a more detailed view of Rhodobacterales diversity (Zhao et al., 2009; Fu et al., 2010).

In this study, we used g5 sequences to examine Rhodobacterales communities in northern Baffin Bay in the Canadian Arctic. During the observation period, this region was characterized by distinct water masses where colder and lower salinity Arctic water merged with the warmer and higher salinity water of the western Greenland Current (Melling et al., 2001; Motard-Côté et al., 2012). These water masses have distinct physicochemical and biological characteristics, and we examined whether the diversity and distribution patterns of Rhodobacterales communities differed as a function of water mass characteristics and environmental parameters.

Materials and methods

Sample sites, measurements of physicochemical parameters and bacterial community characteristics

Samples were collected during the Surface Ocean-Lower Atmosphere Study (SOLAS) expedition on the CCGS Amundsen in September 2008. A total of 17 stations were sampled (Motard-Côté et al., 2012), among which 10 samples from five stations were included in this study (Fig. 1, Table 1). Samples were analyzed from the depths of the 50% surface light intensity and from the depth of the maximum chlorophyll a (chl a) concentration at stations 111, 126, 141, and 202. An additional deep sample was analyzed from 71 m at station 111, and one sample was analyzed from station 134 from the depth of 50% surface light intensity. Water samples were collected using a rosette sampler equipped with 12-L Niskin bottles.

Table 1. Physicochemical and biological characteristics at the stations studied
StationLatitude, longitudeSampling dateDepth (m)Sample featureaTemperature (°C)Chl ab (μg L−1)Salinity (psu)O2 concentration (mL L−1)PARc (quanta m−2 s−1)HPAd (cells L−1)BPe (μg C L−1 day−1)SGRf (day −1)Water mass classificationg
  1. a

    50% L: 50% surface light intensity; chl max: highest chlorophyll a concentration in the water column; 0.2% L: 0.2% surface light intensity.

  2. b

    Chlorophyll a concentration.

  3. c

    Photosynthesis active radiation.

  4. d

    Heterotrophic picoplankton abundance.

  5. e

    Bacterial production.

  6. f

    Specific growth rate.

  7. g

    ASW, Arctic surface water; BBSW, Baffin Bay surface water; INT, intermediate water mass; NA, sample was excluded from water mass segregation.

20276°48′ N, 76°48′ W16-9-2008650% L−0.960.0928.368.617.14 × 10183.79 × 1082.430.29ASW
35chl max−0.360.1530.28.670.14 × 10183.11 × 1081.500.19ASW
13474°25′ N, 78°25′ W9-9-2008650% L3.310.0732.137.5523.65 × 101816.6 × 1080.360.01BBSW
11176°18′ N, 73°14′ W13-9-2008850% L3.810.1331.517.6111.87 × 10184.23 × 1080.840.09BBSW
26chl max3.340.4632.027.912.19 × 10185.41 × 1080.580.06BBSW
710.2% L−0.850.1433.47.450.07 × 10182.72 × 1080.200.04NA
12677°20′ N, 73°25′ W18-9-2008650% L3.410.1331.527.5855.25 × 10189.97 × 1081.100.05BBSW
14chl max1.170.2632.087.8517.01 × 10188.44 × 1081.030.07BBSW
14173°52′ N, 74°17′ W23-9-2008750% L1.30.1131.557.768.39 × 10185.38 × 1080.680.06INT
38chl max1.020.1732.347.750.45 × 10184.93 × 1080.340.03INT
Figure 1.

Sampling stations for this study, which were part of the Arctic Surface Ocean-Lower Atmosphere Study (SOLAS) during the CCGS Amundsen cruise in September 2008.

Vertical profiles of physicochemical, biological, and hydrographical parameters including depth, temperature, in vivo chlorophyll fluorescence, salinity, water density, dissolved O2 concentration, pH, nitrate concentration, and photosynthetic active radiance (PAR) were measured at each sampling station by a Seabird 911plus conductivity–temperature–depth (CTD) probe, equipped with a fluorometer (SeaPoint 2465) and an irradiance sensor (QCP-2300, Biospherical Instruments), mounted on the rosette. In situ chlorophyll fluorescence was calibrated against measurements of extracted chl a from water samples collected at five depths between 5 and 100 m, following the acidification method (Parsons et al., 1984). Extracted chl a concentrations were determined with a Turner Designs Model 10-AU fluorometer.

Heterotrophic picoplankton abundance (HPA) was determined by flow cytometry using a FACSort™ (Becton Dickinson) equipped with a 488-nm argon-ion laser (Li, 1995). Samples for flow cytometry were fixed in a final concentration of 1% paraformaldehyde for 10 min at room temperature, rapidly frozen in liquid nitrogen within 1 h of sample collection, and stored at −80 °C until analysis. Samples were stained with SYBR Green 1 and analyzed using standard protocols (Marie et al., 1999; Li & Dickie, 2001). Flow cytometry counts were verified by comparison with acridine orange direct counts (Hobbie et al., 1977). Bacterial production (BP) was calculated from [14C]-leucine incorporation rates during 6-h dark incubations using the final concentration of 10 nM leucine and a standard conversation factor of 3.1 kg C mol−1 [14C]-leucine incorporated (Simon & Azam, 1989). Bacterial biomass (BB) was estimated from bacterial abundance and cellular carbon content, computed from the geometric mean cell volume of each sample (Hale et al., 2006). The bacterial growth rate (SGR) was calculated assuming exponential growth of bacteria during the bacterial production incubations: SGR = loge [(BPΔ+ BB)/BB]/Δt, where BPΔt is bacterial production obtained during the incubation and Δt is the incubation time in days. The average volumetric HPA, BP, and SGR were calculated for the surface mixed layer (from the bottom of the mixed layer to the pycnocline) and the deep layer (below the pycnocline to 200 meters) at the included sampling stations.

DNA extractions and PCR amplification, cloning, and sequencing of GTA g5 genes

Water samples (500 mL) were filtered onto 0.22-μm polycarbonate filters (Millipore), which were then stored at −20 °C until DNA was extracted. Total DNA was extracted from individual filters in 2-mL lysis buffer (10 mM Tris-HCl buffer with 1 mM EDTA and 0.8 M sucrose, pH 8) with 1 μg mL−1 lysozyme at 37 °C for 30 min. This was followed by addition of 200 μL of 10% (w/v) cetyltrimethylammonium bromide (CTAB) in 1.4 M NaCl and incubation at 65 °C for 30 min. Samples were sequentially extracted with phenol–chloroform–isoamyl alcohol (25 : 24 : 1) and chloroform–isoamyl alcohol (24 : 1), followed by isopropanol precipitation of DNA. The DNA pellets were washed with 70% ethanol, air-dried, and dissolved in 200-μL Tris-EDTA buffer (pH 8).

The g5 gene fragments (approximately 780 bp) were amplified from the extracted DNA using primers MCP-3F (5′-GGCTAYCTGGTSGATCCSCARAC-3′) and MCP-4R (5′-TAGAACAGSACRTGSGGYTTKGC-3′) described previously (Zhao et al., 2009). The target gene fragments were amplified by PCR in 25 μL volumes containing 0.5 μM of each primer, 3% DMSO (v/v), 200 μM dNTPs, 0.5 U Phusion Hot Start polymerase (Finnzymes), and 1× Phusion GC buffer (Finnzymes). Thermocycling conditions were as follows: 30 s at 98 °C followed by 30 cycles of 98 °C for 10 s, 60 °C for 10 s, and 72 °C for 30 s, and a final incubation at 72 °C for 5 min. The reactions were then incubated with 1 U of Taq polymerase (NEB) at 72 °C for 10 min in 1× Taq buffer (NEB) to add 3′ adenine overhangs, and PCR products were purified using a MinElute PCR purification kit (Qiagen). Amplicons were cloned into pGEM-T Easy (Promega). The clones containing appropriately sized inserts were sequenced at the Center for Applied Genomics (Toronto, Canada) using the M13 primers that bind to the cloning vector.

Sequence and phylogenetic analyses

The g5 sequences from this study have been deposited in the NCBI GenBank database, accession numbers KC111451KC111767. After removal of the primer sequences, the g5 nucleotide sequences were aligned according to the corresponding aligned amino acid sequences with ClustalW (Thompson et al., 1994) within MEGA 4 (Tamura et al., 2007). The evolutionary divergence was calculated based on the nucleotide alignments. Sequences with ≥ 95% identity were assigned to operational taxonomic units (OTUs). This criterion represented a species-level distance as inferred by comparing the currently available g5 sequences from the genome-sequenced Rhodobacterales representatives. The coverage (C) of each clone library was estimated using the formula C = 1(N/n), where N is the number of unique sequences in the sample and n is the total number of sequences in the library (Good, 1953). Rarefaction analysis was conducted using Analytical Rarefaction v1.3 (http://strata.uga.edu/software/index.html) as another estimate of the proportion of the population diversity sampled. The evolutionary distances of g5 sequences were calculated based on the alignment using the maximum composite likelihood method (Tamura et al., 2004), and the phylogenetic trees were constructed by the neighbor-joining method using the pairwise deletion of gaps option (Saitou & Nei, 1987).

Community composition comparisons

The Rhodobacterales sequences were grouped at the species level using ≥ 95% nucleotide identity and at the genus level using ≥ 80% identity criteria for the g5 sequences. These values were inferred by comparing the g5 sequences from available Rhodobacterales genome sequences. The grouping of OTUs and clusters was subsequently revised according to the phylogenetic analysis. The community compositions were compared at both the individual sample and water mass scales using UniFrac (Lozupone & Knight, 2005) based on the phylogenetic trees of the g5 clones. The communities were clustered using the UniFrac jackknife analysis, and the robustness of nodes was estimated with 1000 iterations.

The Shannon index (Shannon & Weaver, 1949) was calculated to quantify the Rhodobacterales community diversity with the 95% nucleotide identity as the criterion for demarcation of species/OTUs. The diversity of Rhodobacterales communities in different water masses was compared by analysis of variance (anova), with Tukey's post hoc tests to identify significant differences in diversity in the water masses. The relationships between Rhodobacterales diversity and environmental parameters were assessed by the Pearson correlation coefficient.

The relative distribution of a cluster or OTU was estimated by its percent proportional occurrence, N/n × 100, where N is the number of clones in the OTU or cluster and n is the total number of clones sequenced from the sample. The relationships between the proportional occurrences for abundant phylotypes (defined as a cluster or OTU, which accounted for > 5% of total clones) and physicochemical parameters were examined using the Pearson correlation coefficient.


Field site and physicochemical characteristics

The complete description for all 17 sampling stations in this SOLAS expedition is reported elsewhere (Motard-Côté et al., 2012). Water temperatures ranged from − 0.96 to 3.81 °C for the samples analyzed in this study, with the coldest waters at station 202 and the warmest at station 111 (Fig. 1, Table 1). Salinity ranged from 28.35 to 33.40 psu (Table 1). A temperature–salinity plot (Supporting Information, Figure S1) showed the segregation of the SOLAS sampling stations into three water masses, which were classified as Arctic surface water (ASW), Baffin Bay surface water (BBSW), and an intermediate zone (INT) between the distinct BBSW and ASW masses. Lower salinity accompanied with lower temperatures was the defining feature of the ASW. Detailed characterization of the Rhodobacterales was performed on samples collected in all three water masses (Table 1), where station 202 represents ASW, stations 111, 126, and 134 represent BBSW, and station 141 represents INT. Based on the temperature–salinity characteristics above the depth of the pycnocline, the three water masses were distinct and were each hydrographically coherent. Below the depth of the pycnocline, there were no significant differences in the temperature–salinity relationships, suggesting that segregation of these water masses is restricted to the surface layer and depths above the pycnocline. Therefore, the sole deep sample, collected at 71 m from station 111, was excluded from analyses comparing the different water masses but it is still discussed in other contexts below.

The ASW had a higher O2 concentration than BBSW and INT, but the differences in chl a concentration among water masses were not significant (anova, F = 0.4343, P = 0.667). The sea surface PAR varied with stations, but also did not show significant interwater mass variation. The HPA ranged from 3.11 to 16.64 × 108 cells L−1, with the highest abundance in the BBSW and the lowest in the ASW. The BP showed the reverse pattern, where ASW had significantly higher BP than BBSW and INT (anova, F = 8.84, P = 0.016, d.f. = 8; Tukey, P < 0.05). Consistent with BP, ASW also showed a higher SGR (anova, F = 18.28, P = 0.003, d.f. = 8; Tukey, P < 0.01).

Description of GTA g5 clone libraries

We created 10 clone libraries of GTA g5 amplicons representing the 10 samples collected at the five stations with a total of 320 g5 clones being sequenced from the 10 libraries. All libraries were sequenced to ≥ 90% coverage (Table 2), although all the rarefaction curves did not reach saturation (Figure S2) indicating there was undetected diversity for some of the samples. Twenty-two OTUs were identified with the criterion of 95% nucleotide sequence identity, representing 22 putative species, and 13 different clusters were found with the 80% identity criterion, representing 13 putative genera. Of the 22 OTUs, 12 (54.5%) were singletons, which were only detected once in all samples. These singletons represented only 3.4% of all clones. The dominant OTU (i) accounted for 48.4% of all clones and was detected in all 10 samples. At the putative genus level, five singletons were found, representing 1.5% of all clones. Cluster A, mainly comprised of OTU i, was the most abundant cluster in all water masses.

Table 2. Characteristics of the 10 g5 clone libraries
Sampling stationWater massaSample featurebNumber of clone sequencesNumber of OTUsNumber of clustersCoverage (%)Shannon index (H')OTUs in each sample
  1. a

    ASW, Arctic surface water; BBSW, Baffin Bay surface water; INT, intermediate water mass; NA, sample was excluded from water mass segregation.

  2. b

    50% L: 50% surface light intensity; chl max: highest chlorophyll a concentration in the water column; 0.2% L: 0.2% surface light intensity.

202ASW50% L3454910.5547i, iii, xxix, xxx, xxxi
202ASWchl max3633940.2253i, iii, x
134BBSW50% L2676961.6852i, ii, v, vi, vii, viii, xi
111BBSW50% L4886900.9695i, ii, iii, v, vi, x, xii, xiii
111BBSWchl max26651001.6179i, ii, iii, iv, v, vi
111NA0.2% L17331000.8351i, iv, v
126BBSW50% L4797941.7781i, ii, v, vi, vii, ix, xiv, xv, xvi
126BBSWchl max3386911.5037i, ii, iii, iv, v, vi, ix, x
141INT50% L3077901.3911i, ii, iii, iv, vii, viii, xvii
141INTchl max2342910.8934i, iii, xviii, xix

The phylogenetic relationships of the clones from this study with reference strains are shown in Fig. 2. The clone sequences are very diverse within the Rhodobacterales. Some of the OTUs and clusters have well-supported close relationships to cultured strains with available g5 sequences, but most do not (Fig. 2).

Figure 2.

Phylogenetic relationships of GTA g5 sequences. The phylogenetic tree presents clone sequences detected and reference g5 sequences from available genomes in the NCBI database. Branches for OTUs were collapsed, and the numbers in parentheses report the number of clones detected for each OTU. The assigned clusters are indicated on the right. The unrooted neighbor-joining tree is based on the aligned nucleotide sequences corresponding to the aligned translated amino acids sequences. Bootstrap values (percentage based on 1000 iterations) are shown for the branch points supported > 50%. The scale bar indicates the number of substitutions per site.

Assigning taxonomic labels to the environmental Rhodobacterales OTUs and clusters

Representative GTA g5 sequences from known isolates that are available in the public sequence databases can be affiliated with only two of the 22 the OTUs (ii and vii) and three of the 13 clusters (A, B, and G) identified in the clone libraries (Table 4; Fig. 2). At the species level, Oceanibulbus indolifex HEL-45 clearly grouped with OTU ii and Sulfitobacter sp. EE-36 and NAS-14.1 grouped with OTU vii. The Sulfitobacter sequences formed a rare group (2%) of the detected sequences, and they were not detected in the ASW. The three samples where these clones were detected were all from the surface mixed layer at stations 134, 126, and 141 where water temperatures were relatively high (Table 1).

At the putative genus level, cluster A, which was the most abundant group of sequences in all three water masses (Fig. 4), is distantly related to the genus Octadecabacter (Fig. 2). Cluster A accounted for 91%, 53%, and 33% clones in the ASW, INT, and BBSW, respectively.

A number of other OTUs and clusters appeared affiliated with known Loktanella representatives, but none of these relationships were within our criteria for species or genera relationships or supported by bootstrapping of the phylogenetic tree (Fig. 2).

Correlation of Rhodobacterales diversity with environmental parameters

We calculated the Pearson correlation coefficients between the Rhodobacterales diversity (as estimated by Shannon index) and the physicochemical and biological parameters measured for the 10 samples (Table 3). The HPA and temperature showed significant positive correlations with Shannon index, whereas O2 concentration and SGR showed negative correlations with Shannon index.

Table 3. Pearson correlation coefficients of environmental parameters with the Shannon index from the 10 samples and the proportional occurrence of the three most abundant sequence clusters
ParameterShannon indexClustera
  1. a

    Clusters are defined in Fig. 2.

  2. b

    Numbers in bold font indicate significant correlations (P < 0.05).

Temperature 0.7387 b −0.40090.6052−0.0664
Chl a0.2917−0.31830.23620.1499
O20.6624 0.7789 −0.5701−0.0275
Photosynthesis active radiation0.6285−0.3823 0.7007 −0.3859
Heterotrophic picoplankton abundance 0.7112 −0.4787 0.7251 −0.2508
Bacterial production−0.4470 0.6816 −0.3124−0.2681
Specific growth rate0.6811 0.7893 −0.5559−0.1980

Comparison of the Rhodobacterales communities in different water masses

Rhodobacterales community characteristics, including diversity and community composition of the three water masses, were compared. The Shannon indices were significantly different among water masses (anova, P-value = 0.0148, F-ratio = 9.23, d.f. = 8) and were highest in the BBSW and lowest in the ASW, and the differences were significant (Tukey, P = 0.012). The INT diversity showed no significant difference (Tukey, P > 0.05) with either of the other two water masses (Fig. 3). However, the Shannon index quantifies the degree of complexity and does not indicate differences in composition or community structure. We therefore used UniFrac analyses (Lozupone & Knight, 2005) to compare the community compositions based on phylogenetic relationships of the g5 sequences. This showed that samples from different water masses form separate clusters, with exception of the surface sample from station 111 (Fig. 4). Therefore, the different water masses show distinct Rhodobacterales community compositions. The differences are the most distinct and robust between ASW and BBSW samples, and INT samples were more related to ASW samples.

Figure 3.

Comparison of Rhodobacterales diversity in different water masses as quantified by Shannon index. The Rhodobacterales diversity was significantly lower in ASW (a) than BBSW (b) but there was no significant difference between the diversities of INT (ab) and ASW or BBSW. The horizontal line in the box is the mean Shannon index, and the bottom and top edges of the box are 25% and 75% quartiles, respectively. The vertical bars are the 95% confidence interval, when outside the quartiles demarcated by the box.

Figure 4.

Comparison of Rhodobacterales community structure in northern Baffin Bay. The top tree was constructed with the UniFrac jackknife clustering, based on the phylogeny of all g5 clones in the 10 samples. The support value at each node is a percentage calculated from 1000 iterations, and the scale bar represents the UniFrac distance. The bar chart shows the proportional occurrence of each identified cluster in all samples, as indicated in the color legend. The sample identifiers and water mass are at the bottom of the figure (see Table 1 for details); *, not assigned to any water mass.

Different Rhodobacterales phylotypes showed different distribution patterns (Figs 4 and 5). Only three clusters (A, C, and F) were found in all three water masses, and only cluster A was found in all samples. Pearson correlation coefficient analysis was used to investigate the relationship between the proportional occurrence of the three most abundant clusters and the environmental parameters (Table 3). Cluster A showed significant positive correlation to O2 concentration, BP, and SGR and a negative correlation to salinity. Cluster B showed significant positive correlation to PAR and HPA. Cluster C did not show significant correlations to any of the measured parameters.

Figure 5.

Proportional occurrence of clusters in samples collected at three depths from station 111. The 50% surface light intensity depth was 8 m, the depth where chl a concentration was the highest was 26 m, and the 0.2% surface light intensity depth was 71 m. The diameter of each circle represents the relative proportional occurrence of each cluster in each sample.

Influence of depth on Rhodobacterales communities

Three samples were collected at station 111 in BBSW, and these showed that there is stratification of Rhodobacterales in the water column (Fig. 5). The chl a maximum depth had the highest diversity (Table 2). Clusters A and D had obvious and opposite differences in proportional occurrence according to depth (Fig. 5). Cluster A was mainly found in the surface layer and comprised 73% of the clones from this layer, while cluster D was more abundant in the deep samples. In contrast, cluster B was fairly evenly distributed throughout the water column.


We characterized the diversity of bacteria in the order Rhodobacterales in northern Baffin Bay in the eastern Canadian Arctic. The diversity of these bacteria in this location is lower than previously reported in Newfoundland coastal water (Fu et al., 2010) and in the Chesapeake Bay (Zhao et al., 2009). This is consistent with the global trend of decreasing biodiversity from the equator to the poles (Martiny et al., 2006). Furthermore, the composition of the Rhodobacterales community in northern Baffin Bay differs compared to temperate locations. For example, the dominant cluster we identified, which is related to Octadecabacter, was not found in the Chesapeake Bay. The Rhodobacterales sequences recovered in this study affiliate with sequences from the Roseobacter clade, as opposed to Rhodobacter, consistent with their previously recognized abundance in marine systems based on 16S rRNA gene (González & Moran, 1997; Buchan et al., 2005; Gilbert et al., 2012). Overall, however, identifiable isolates closely affiliated with the taxa we have identified here are few, and more efforts are needed to characterize cultured representatives to further explain Rhodobacterales biogeography. Quantification of specific bacterial groups using fluorescence in situ hybridization during this SOLAS expedition (Motard-Côté et al., 2012) indicated that Roseobacters accounted for 12% and 6% of prokaryotes in the ASW and BBSW masses, respectively. However, these percentages should be treated with caution as current tools are likely unable to accurately quantify the Roseobacter clade as a whole (Giebel et al., 2011), and subgroup-specific probes for lineages within the clade appear more reliable (Buchan et al., 2009; Giebel et al., 2009; Yao et al., 2011) with the remaining problem that they do not provide data for the entire group.

We found water mass to be the most important factor for the distribution of marine Rhodobacterales communities in northern Baffin Bay, which agrees with previous studies looking at different levels of bacterial community composition (Baldwin et al., 2005; Pommier et al., 2005; Giebel et al., 2009, 2011; Schattenhofer et al., 2009). There are some general patterns observed for the distributions of the three predominant clusters in the different water masses (Fig. 4). Cluster A was the only group found in all samples in all water masses. Cluster B was not found in the ASW mass. Cluster C was found in all three water masses, but had its highest proportional occurrence in the INT samples. A comparison of the diversity among the three water masses in northern Baffin Bay shows that low diversity in ASW corresponded with high bacterial growth rate, as indicated by SGR, and the situation was reversed in BBSW. Overall diversity showed an inverse correlation with SGR, whereas the abundance of cluster A showed a strong positive correlation with SGR. This presumably reflects the dominance of this group, particularly in conditions favoring higher growth rates. In contrast, cluster B is correlated with light intensity and HPA, whereas cluster C did not show any significant correlations. Therefore, although some of the taxa distribution patterns correlate with some environmental factors, these environmental factors do not sufficiently explain all of the observed patterns. Marine Rhodobacterales have been shown to show a strong seasonal pattern in abundance at a specific temperate coastal location with peaks in abundance in concert with primary production (Gilbert et al., 2012), which again reflects a combination of environmental factors. We do not think the presence of only one station from each of the ASW and INT in our analyses has artificially caused the observed water mass differences in diversity (Fig. 3) because the 50% light and chl a maximum samples group together for the stations within the INT and ASW, while the same is not seen for the BBSW where there are three stations (Fig. 4). Using different criteria to assign all 17 SOLAS stations to either ASW or BBSW (Motard-Côté et al., 2012), our INT station fell within the ASW, and the UniFrac community analysis does cluster our INT and ASW stations together (Fig. 4). However, the Shannon index measure of overall diversity (Fig. 3) supports the INT designation we decided upon from the temperature–salinity profiles (Figure S1). It would be ideal to have more samples from the INT and ASW masses, but samples for Rhodobacterales characterizations were only collected at a subset of the 17 SOLAS stations.

Depth affects the Rhodobacterales community structure, as is generally observed in studies of bacterial communities (Baltar et al., 2007; Schattenhofer et al., 2009; Morris et al., 2012). Although it was not a statistically supported correlation (Table 3), perhaps because there were generally only two depths sampled at most of the stations, the depth profile from the three samples collected at station 111 shows distinct distributions for different sequence clusters (Fig. 5). The additional deep sample from this station came from below the pycnocline. Although the pycnocline is usually considered as a physical barrier of vertical mixing, we find it is not likely acting as a physical barrier in Rhodobacterales community distribution because all the taxa found in the 71-m sample are also found at the chl a maximum depth and the UniFrac analysis clustered these two samples (Fig. 4). One possibility for these bacteria to cross the pycnocline mixing barrier is by adhering to sinking partials, such as marine snow, and Rhodobacterales are known to form aggregates and adhere to particles (Gram et al., 2002; Dang et al., 2008). When reaching the deeper location, the environmental factors could further tune the community structure based on the physiological properties of each group. Sequences representing cluster D contributed 65% of clones to the deep sample but were only recovered in one surface sample, and there represented only 3% of the clones (Fig. 4), suggesting this group is adapted to deeper water conditions.

The dominant Rhodobacterales sequences found in this study are related to Octadecabacter antarcticus (based on > 80% nucleotide identity of g5 fragments for genus-level relationships). This group was previously found to be abundant in polar oceans and sea ice (Staley & Gosink, 1999; Brinkmeyer et al., 2003). All cultured Octadecabacter representatives have been obtained from polar environments and display low optimal growth temperatures of 4–15 °C (Gosink et al., 1997). In our study, the occurrence of the Octadecabacter-related cluster is positively correlated to the SGR, which suggests that it is one of the major active bacterial groups in northern Baffin Bay. This cluster is dominant in the ASW samples (Fig. 4), which are the coldest samples. The abundance and phylogeny of this cluster suggest that it is related to the RCA cluster that is often detected in large numbers from polar and subpolar oceans (Selje et al., 2004; Buchan et al., 2005; Wagner-Döbler & Biebl, 2006; Giebel et al., 2009). The relative amounts of the RCA cluster have been shown to correlate with a number of different environmental variables such as prokaryote abundance, particulate matter, temperature, and phytoplankton pigments (Giebel et al., 2009, 2011; Sperling et al., 2012).

The sequences in OTU ii (cluster B) are most closely related to Oceanibulbus indolifex. Strains from this genus are described as mesophiles, with a documented lowest growth temperature of 8 °C (Wagner-Döbler et al., 2004). This OTU was absent from ASW clone library sequences and most of the clones (92%) representing this OTU were obtained from BBSW samples, which have the higher water temperatures compared to the other water masses. However, the proportional occurrence of this cluster was not significantly correlated with temperature (Table 3). OTU vii grouped with two Sulfitobacter strains at the species level (≥ 95% identity). This is the highest similarity we found between clones and cultured representatives. However, OTU vii only contributed 2% of total clones and they were all found in surface water at relatively warm stations. This pattern could be explained by the temperature preferences of Sulfitobacter isolates, where the observed optimal growth temperature of cultured Sulfitobacter strains is ~20 °C (Sorokin, 1995), and perhaps, the low temperature of northern Baffin Bay limits their distribution. Understanding these distribution patterns is important in light of ongoing climate change, which will likely alter bacterial distributions in the oceans and thereby alter the represented microbial physiological activities.

The genome sequences available for isolates that affiliate with our dominant sequence groups (Fig. 2) show conservation of the GTA gene cluster (Newton et al., 2010). In the limited cases where Rhodobacterales genomes do not contain contiguous GTA clusters, they appear to lack the entire set of genes and not just a portion of the cluster (Lang & Beatty, 2007; Biers et al., 2008; Newton et al., 2010; Lang et al., 2012). Therefore, we predict that the g5 amplicons represent strains with the genetic potential to produce functional GTA particles and our measure of Rhodobacterales diversity in this work also represents a measure of potential GTA diversity. However, lack of cultured representatives of the actual phylotypes identified in studies such as ours remains a problem in studying Rhodobacterales diversity and understanding the potential for GTA production in natural environments. For example, our cluster A is loosely related to the type strain of Octadecabacter antarcticus (Fig. 2; Table 4). The cluster A g5 sequences share c. 89% sequence identity with this strain, but the cluster A clones share c. 98% identity with each other. Therefore, how well this type strain represents the most abundant Rhodobacterales in our samples and to what extent we can apply this strain's physiological characteristics to explain the observed biogeography are uncertain. Comparison of the genome contents of cultured strains and uncultivated marine Roseobacters has indicated there are important differences in their genomic repertoires (Luo et al., 2012), further highlighting the need to better characterize these currently uncultured groups. As discussed above, cluster A likely corresponds to the RCA cluster, and a complete genome sequence for an RCA cluster representative, Planktomarina temperata, is reported to be in progress [(Giebel et al., 2011); NCBI project ID 42833, accession PRJNA42833], which will help inform future studies. The overall lack of cultured representatives related to obtained Rhodobacterales sequences, as illustrated in Fig. 2, is a barrier to further explain the biogeography of this important group of bacteria.

Table 4. Cultured organisms with g5 sequences available that can be associated with g5 library clone sequences
Phylogenetic groupsaReference cultured representatives
  1. a

    Defined in Fig. 2.

Cluster AOctadecabacter antarticus 307
Cluster B, OTU iiOceanibulbus indolifex HEL-45
Cluster GSulfitobacter sp. EE-36, Sulfitobacter sp. NAS-14.1, Roseobacter sp. GAI101
OTU viiSulfitobacter sp. EE-36, Sulfitobacter sp. NAS-14.1


We thank Alison Buchan for comments on the manuscript. YF and KFK were supported by fellowships from Memorial University School of Graduate Studies. This research was supported by funding from the Natural Sciences and Engineering Research Council (NSERC) to ASL and from NSERC and International Polar Year to RBR. We thank Steven Holland for provision of the rarefaction analysis tool through http://strata.uga.edu/software/index.html.