The winter in the Arctic Ocean is characterized by an almost complete lack of sunlight and water temperatures below zero degrees Celsius. These extreme conditions are expected to have a large impact on the composition and activity of marine microbes in surface waters down to the nitracline (i.e., the depth interval where nitrate concentration rapidly increase from depleted surface concentrations), as carbon and nutrient availability drastically changes when primary production is reduced or stops during darkness. Indeed, changes in the relative abundance of major bacterial groups over the course of the winter (Alonso-Sáez et al., 2008), as well as in their use of dissolved organic substrates (Nikrad et al., 2012), have been previously reported. Interestingly, despite the harsh environmental conditions, heterotrophic bacterial production is sustained at a baseline level during the entire winter (Garneau et al., 2008), and the single-cell activity of some abundant groups of higher taxonomic rank has been shown to be relatively high (Alonso-Sáez et al., 2008).

Identifying metabolically active ribotypes in low productive marine environments is challenging and information on the specific phylotypes that can actively grow in polar winter waters is scarce. Here, we address this challenge by using Bromodeoxyuridine (BrdU) incubations. Labeling of DNA with BrdU was first used in microbial ecology as an alternative to radioactive thymidine (TdR) to measure bacterial growth (Steward and Azam, 1999). BrdU is an analog of TdR that is incorporated into newly synthetized DNA. This feature was also used to identify the pool of growing bacteria within a community (Borneman, 1999; Urbach et al., 1999) and later coupled to microscopy to identify DNA-synthetizing cells by fluorescence in situ hybridization (FISH; Pernthaler et al., 2002; Pernthaler and Pernthaler, 2005). More recently BrdU incubations have been combined with 16S rDNA-based analysis to study the spatial distribution of active communities (Taniguchi and Hamasaki, 2007) or identify active bacteria degrading dissolved organic compounds (Mou et al., 2007).

A significant advantage of the BrdU approach is that this method can potentially show differences in the activity of closely related phylotypes. In previous studies of Arctic waters, a high degree of diversity was found within the most abundant phylogenetic groups of marine bacteria (Kirchman et al., 2010; Lovejoy et al., 2011), implying that many closely related ribotypes can co-exist when conditions are favorable. To what extent the specific metabolic characteristics of such related but ecologically distinct phylotypes differ has not yet been studied extensively. Ecological differences based on spatial and temporal distribution patterns have been described for phylotypes of major marine prokaryotic groups such as unicellular cyanobacteria (Rocap et al., 2003; Johnson et al., 2006) or the SAR11 cluster (Brown et al., 2012). However, differences in the specific activity of related clusters in the environment have seldom been reported. By using the BrdU approach, we aimed at identifying bacterial ribotypes with the capacity to grow in Arctic winter waters, potentially including closely related phylotypes that may constitute ecotypes in this marine environment.

BrdU-labeled DNA (BrdU), representing the actively dividing bacterial fraction, was successfully separated from the total DNA fraction (Tot) by immunocapture in incubations of winter Arctic seawater. The BrdU method had been successfully applied in different freshwater and marine environments, but its application in the extremely low productive winter Arctic water was challenging. Therefore, we increased the concentration of BrdU and the incubation time as compared to previous studies. While short incubation times are generally desirable in order to obtain results representative of in situ conditions, we found that only BrdU-labeled DNA incubated for at least 10 days was successfully amplified by PCR in the winter Arctic seawater incubations. According to estimates based on the different phases of cell cycles, a BrdU incubation time of 65% of the doubling time would label 100% of all growing cells (Pernthaler and Pernthaler, 2005). In winter waters of the Amundsen Gulf, a prokaryotic growth rate of 0.1 d^-1 was measured in dilution cultures (Alonso-Sáez et al., 2010), which indicates a doubling time of ca. 7 days. Therefore, theoretically, the labeling of actively growing Arctic communities would take ca. 4.5 days. However, growth rates in our incubations were probably lower as we incubated the water without dilution. That may explain that insufficient BrdU-labeled DNA could be extracted from samples collected during the first week of incubation. In the experiment carried out in February, additional BrdU samples were analyzed after 22 and 28 days of incubation. For all the samples, the active and total community DNA was amplified and analyzed by TRFLP (Figure 1).

As shown by the TRFLP analysis, BrdU-labeled communities were different from the total communities, and NMDS grouped all labeled samples away from non-labeled samples (Figure 1; ANOSIM, p < 0.01). This indicated that the actively growing bacterial fraction (BrdU) represented only a subset of the total community. Interestingly, during the February experiment the composition of the total bacterial community changed little from 13 to 28 days of incubation, while differences in the active communities were detected over the course of the experiment, particularly from 22 to 28 days (Figure 1). The active bacterial communities after ca. 10 days of incubation in November and January showed higher similarity, but were different to the active community in the experiment of February after 13 days. As incubation conditions were strictly similar for the three experiments, our results suggest that Arctic communities contained different bacterial phylotypes that varied in their potential for growth during the course of the winter.

It should be emphasized that the studied communities do not directly represent in situ active bacterial communities, as incubations may promote or limit the growth of some bacteria from the natural assemblages, particularly when long incubation periods are required and high concentrations of potentially inhibiting compounds such as BrdU are applied. Other potential pitfalls in our approach are that the fingerprinting techniques such as TRFLP may miss some bacterial taxa, and that the BrdU substrate could potentially be toxic for some bacteria. For example, early studies have shown that a strain of Bacillus subtilis had reduced sporulation and growth under BrdU incubation (Binnie and Coote, 1986). We were, however, able to detect different bacteria in the BrdU fraction, such as SAR11, Polaribacter, Arctic96B -16 and Colwellia, which are known to be environmentally relevant in Arctic waters (Bano and Hollibaugh, 2002; Alonso-Sáez et al., 2010). Our results thus show that even though some Arctic microorganisms may have been inhibited, a large diversity of them take up BrdU at the concentration added.

The major components of the bacterial communities were identified by cloning and sequencing bacterial 16S rRNA genes from the three experiments, and we used these data to identify the TRFLP peaks by using the program TRiFLe. We obtained 41, 57, and 61 sequences (ca. 950 bp long) from the libraries Tot_Nov_6d, Tot Jan_10d, and Tot_Feb_13d, respectively. Diversity was low in all three libraries, which were dominated by the genus Colwellia that represented >88% of the sequences. Colwellia is a psychrophilic marine bacteria, which is well adapted to life in cold environments (Methé et al., 2005). Members of Colwellia have been detected in a wide range of marine habitats, including polar regions (Deming and Junge, 2005), and appear to be important to carbon and nutrient cycling in the cold marine environments (Methé et al., 2005). Colwellia was also abundant in previous Arctic winter water incubations, particularly in samples taken from the nitracline (i.e., 38% of clones; Alonso-Sáez et al., 2010). Clone libraries from in situ samples collected at the same location in winter showed communities dominated by Gammaproteobacteria and Alphaproteobacteria, a composition that is typical for the Arctic Ocean (Lovejoy et al., 2011). The Arctic 96B cluster and Polaribacter were also present in situ while Colwellia represented 4% of the sequences (Alonso-Sáez et al., 2010).

Colwellia were earlier shown to be selectively enriched in North Sea seawater incubations (Eilers et al., 2000) and able to outcompete other bacteria during bottle incubations. Thus, Colwellia probably dominated in our experiment because they were able to take advantage of the incubation conditions. However, in contrast to previous studies, it is remarkable that this phylotype was able to grow in winter Arctic waters, as the availability of fresh organic substrates is very low. Colwellia species have typically been cultured on heterotrophic media and grow from -1 to 10°C (Huston et al., 2000). An analysis of the genome of the psychrophile Colwellia psychrerythraea 34H has revealed the potential for this strain to degrade fatty acids and polyamides, and suggests a possible C1 metabolism (Methé et al., 2005). This type of metabolism could be related to the fact that Colwellia cells showed active uptake of bicarbonate during incubations in Arctic winter seawater, and may be interpreted as a strategy to deal with starvation under resource depleted conditions (Alonso-Sáez et al., 2010).

Colwellia previously isolated from a variety of polar habitats (Bowman et al., 1998; Junge et al., 2002) are very diverse, including the presence of closely related ribotypes. Yet, the ecological significance of their diversity, and the ecological role of different Colwellia clusters remain unclear. In our Arctic seawater incubations, Colwellia showed intra-genus diversity and sequences fell into three separate clusters (Figure 2). Most of the cloned sequences belonged to cluster 3 (123 sequences), represented by the Colwellia strain ZS4-15, while fewer fell into cluster 2 and 1 (6 and 12 sequences, respectively). The least abundant cluster 2 was closest to C. psychrerythraea 34H. Sequences from clusters 2 and 3 had a minimum of 97% identity and sequences from cluster 2 and 1, a minimum of 93% identity. The three Colwellia clusters were detected in clone libraries from seawater dilution cultures with Arctic winter waters in an earlier study (11 sequences from cluster1, 1 from cluster 3 and 1 from cluster 2, Alonso-Sáez et al., 2010), and were successfully resolved based on their TRFLP profiles using the software TRiFle

Additional analyses identified other abundant TRFLP peaks, including the Gammaproteobacteria Arctic 96B-16 cluster, Polaribacter, and SAR11 (Figure 2). Colwellia cluster 1, Polaribacter, and Arctic 96B-16 were more abundant in the active community represented by the BrdU-labeled DNA, indicative of a high potential for growth in Arctic winter waters. In contrast, Colwellia cluster 3, SAR11, and Colwellia cluster 2 were underrepresented in the BrdU-labeled community compared to total communities (significant difference between means between total and active communities (p < 0.05) as calculated with a Wilcoxon test for paired values). Our observation that Polaribacter was more frequently found in the active fraction is in agreement with a recent in situ study which found that Polaribacter had higher leucine uptake rates during winter compared to summer in the Beaufort Sea (Nikrad et al., 2012). It should be noted that other microorganisms present in the incubations might have been overlooked by the TRFP fingerprinting techniques.

The abundant Colwellia clusters 3 and 1 showed marked differences in activity in the experiments based on BrdU labeling (Figure 3). A SIMPER analysis conducted on TRFLP peaks confirmed that Colwellia cluster 1 contributed more than cluster 3 to the active communities across all incubations (Table 2). A comparison between months showed that the November and January BrdU samples (BrdU_Nov_11d and BrdU_Jan_10d) contained more Colwellia cluster 1 sequences compared to the February sample (BrdU_Feb_13d), characterized by a higher contribution of Colwellia cluster 3 (SIMPER analysis, not shown). Such results are consistent with different activity among clusters, but could also reflect differential ability to assimilate BrdU. BrdU uptake capacity and efficiency varies among bacterial genera (Hellman et al., 2011) and could also differ within a single genus, given the high intra-genus diversity of marine microbes. Overall, we conclude that related phylotypes of Colwellia showed marked differences in DNA synthesis under similar incubation conditions, and thus hypothesize that they also have consistent differences in their metabolic traits. While these results do not represent changes of the activity of Colwellia in situ, our data suggest that the two most abundant Colwellia clusters present in the incubations likely represent ecologically distinct groups.

In an environment like the winter Arctic Ocean where the bacterial community grows slowly, longer incubation times are needed to label dividing cells. However, such long incubations may also cause changes in the natural communities and is an inherent limitation with the technique. Despite the likelihood that the incubation contents did not reflect the true natural community, the BrdU approach was useful in detecting environmental relevant Arctic bacterial phylotypes with the potential to grow in Arctic winter waters, and highlighting differences among Colwellia clades. These results emphasizes that caution is needed when inferring functional roles of taxonomically defined bacterial groups identified by slowly evolving marker genes such as the 16S rRNA. Activity-based studies remain a valuable complement to gene surveys and may reveal functional coherence and incoherence among populations.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.