The predicted increase in surface temperatures of 2–5° by the end of this century (IPCC, 2013) will lead to warmer oceans (Levitus, 2000), with important implications for pelagic communities and biotic interactions (Petchey et al., 1999; Burrows et al., 2011). For example, the differential impact of warming on autotrophic production and the consumption of dissolved organic matter (DOM) by heterotrophs induces a higher rate of bacterial degradation of phytoplankton-derived organic matter (Wohlers et al., 2009). At low temperatures, phytoplankton spring blooms can be temporally decoupled from bacterial degradation (Pomeroy and Deibel, 1986; Lignell et al., 1993; Bird and Karl, 1999), whereas rising water temperatures intensify phytoplankton-bacteria coupling and thereby enhance carbon flow into the microbial food web (Müren et al., 2005; Morán et al., 2006; Hoppe et al., 2008; Degerman et al., 2013). The mechanisms underlying these processes are poorly understood, although they probably depend on the ability of the bacterial assemblage to respond rapidly to the input of labile dissolved organic carbon (DOC) supplied by blooming phytoplankton (Pinhassi et al., 2004; Grossart et al., 2005; Sapp et al., 2007; Teeling et al., 2012). This has indeed been observed for bacteria belonging to Alteromonadales (Gammaproteobacteria), in which both cell abundance and transcriptional activities increase rapidly in response to DOC, particularly phytoplankton-derived organic carbon (Eilers et al., 2000; McCarren et al., 2010; Sarmento and Gasol, 2012; Beier et al., 2015). Although Alteromonadales are strongly grazed by bacterivorous protists (e.g., Beardsley et al., 2003), high proportional abundances have nonetheless been detected during marine phytoplankton spring blooms (Tada et al., 2011; Teeling et al., 2012), indicating the central role of these bacteria in the utilization of phytoplankton-derived DOC (Pedler et al., 2014).

The succession of dominant, free-living bacterial lineages during the proliferation of a phytoplankton spring bloom has not been well-studied at high taxonomic resolution, with a few exceptions (Teeling et al., 2012). Most studies quantitatively examined the bacterial succession during the phytoplankton blooms, using relatively broad phylogenetic probes such as those for Bacteroidetes, Gammaproteobacteria, and Roseobacter in marine environments (Alderkamp et al., 2006; Tada et al., 2011; Sintes et al., 2013) and for Actinobacteria, Bacteroidetes, and Betaproteobacteria in limnic systems (Zeder et al., 2009; Eckert et al., 2012). To our knowledge, no study has quantified the species- or genus-level dynamics of the dominant bacteria during the development of a phytoplankton spring bloom. The identification of these bacterial taxa would enable a more detailed investigation of their role during early phytoplankton spring blooms, when low water temperatures are thought to otherwise suppress most bacterial activities (Pomeroy and Deibel, 1986).

Mesocosm studies have shown that warming results in a tighter phytoplankton-bacterial coupling and an enhanced carbon flow into the microbial food web (Hoppe et al., 2008; Wohlers et al., 2009; Wohlers-Zöllner et al., 2012; von Scheibner et al., 2014). Reduced bacterial development in the colder treatments is due to a combination of lower growth rates and relatively temperature-insensitive protist grazing pressure (von Scheibner et al., 2014). In this study, we were particularly interested in following the growth dynamics of Glaciecola spp. (Alteromonadaceae, Gammaproteobacteria), identified as a dominant taxa in a previous mesocosm study carried out during early phytoplankton bloom conditions. We also examined how the growth of these bacteria is affected by an increase in temperature. The population dynamics of Glaciecola spp., were investigated by using a newly developed, specific catalyzed reporter deposition fluorescence in situ hybridization (CARD-FISH) probe to analyze the experimental phytoplankton blooms that developed at low (in situ: ∼2°C) and high (Δ+6°C ∼8°C) temperatures in the mesocosm experiment reported in von Scheibner et al. (2014). The results showed that Glaciecola spp. dominated the free-living bacterial assemblages in the mesocosms, despite intense grazing pressure, and thus presumably play an important role in carbon processing during diatom blooms, both at low and at elevated temperatures.

During the peak of the phytoplankton bloom, bacterial community composition, analyzed by 16S rRNA Sanger-sequenced clone libraries, was clearly dominated by Gammaproteobacteria under the low (∼2.0°C) and high (∼8.0°C) temperature regimes, as previously reported (von Scheibner et al., 2014). The most abundant OTU was that of a Glaciecola sp. closely related to Glaciecola sp. HTCC2999 (98%) and to Glaciecola pallidula (96%), which accounted for 57% and 44% of the total clones in the colder and warmer treatments, respectively (von Scheibner et al., 2014). A precise cellular quantification of Glaciecola spp. using the newly designed probe GC1252 was now possible. However, while GC1252 was primarily designed to detect the abundant Glaciecola sequence, it is also able to target a narrow clade containing 30 taxa within the genus Glaciecola (Figure 1; Supplementary Tables S1 and S2), including two OTUs previously found in the clone libraries (von Scheibner et al., 2014). In our mesocosm experiment, GC1252-positive cells were present as uniform free-living, relatively large rod-shaped cells (Supplementary Figure S1). The measured area of Glaciecola spp. cells was on average 34% larger than that of all bacterial cells, resulting in a ∼43% higher biomass per cell.

Phytoplankton spring blooms with similar growth patterns developed in all 12 mesocosms, but significant differences were consistently observed between the two temperature scenarios (Δ0°C and Δ+6°C). These included the lower Chl a maximum and its advanced onset (4–6 days) with warming (Figure 2) (Lewandowska and Sommer, 2010; Biermann et al., 2014). The phytoplankton bloom was dominated by diatoms, with Skeletonema costatum as the dominant species under warmer conditions (55 ± 8% of total phytoplankton biomass vs. 19 ± 5% under colder conditions) and Thalassiosira rotula as the most abundant species under colder conditions (39 ± 6% of total phytoplankton biomass vs. to 3 ± 1% under warmer conditions) (Lewandowska and Sommer, 2010). Bacterial abundance was strongly enhanced during the phytoplankton bloom at the higher temperature, reaching 2.15 ± 0.08 × 10⁶ cells ml^-1 at its peak (Figure 2B) compared to 0.64 ± 0.10 × 10⁶ cells ml^-1 in the colder treatment (Figure 2A).

The population dynamics of Glaciecola spp. (GC1252-positive cells), especially during the exponential growth phase, were followed by analyzing samples from all 12 mesocosms over the entire course of the phytoplankton spring bloom (Figure 2). Glaciecola spp. were of low abundance before the bloom, constituting only 1–2% of the total free-living bacterial cells (DAPI-stained cells), except in mesocosm IV (5.7%). A strong increase in Glaciecola spp. was first observed in parallel with the developing phytoplankton bloom, although it was somewhat slower in the colder treatments (Figure 2). In the warmer treatments, the relative abundance of Glaciecola spp. developed in parallel with that of the whole bacterial assemblage, increasing within 2 days from 2.2 ± 1.3% (0.13 ± 0.08 10⁵ cells ml^-1) on day 8 to 24.0 ± 8.5% (1.85 ± 0.51 10⁵ cells ml^-1) on day 10 (Figure 2). In the colder treatments, Glaciecola spp. increased even though total bacterial numbers stayed constant or in some cases decreased. The observed increase occurred over 5 days, from 1.5 ± 0.4% (0.16 ± 0.04 10⁵ cells ml^-1) on day 8 to 15.2 ± 7.3% (0.77 ± 0.31 10⁵ cells ml^-1) on day 13 (Figure 2).

The resulting doubling times of Glaciecola spp., calculated from the increase in cell numbers, were significantly (t-test: p = 0.019) lower in the warmer than in the colder treatments (9.6 ± 1.5 vs. 23.0 ± 8.0 h), with corresponding maximal growth rates (μ) of 1.77 day^-1 and 1.10 day^-1, respectively. Based on the appraised doubling times of the different temperature treatments, the Q₁₀ value for Glaciecola spp. was 2.2.

Heterotrophic nanoflagellate biomass increased in synchrony with the growth of the phytoplankton bloom and with Glaciecola spp. abundance (Figure 2). In the warmer treatments, HNF biomass rose from 26 ± 8 μg C L^-1 on day 8 to 116 ± 56 μg C L^-1 on day 10, whereby the Glaciecola spp. population increased from 0.13 ± 0.08 × 10⁵ to 1.85 ± 0.50 × 10⁵ cells ml^-1, respectively. In the colder treatments, HNF biomass increased from 22 ± 1 μg C L^-1 on day 8 to 54 ± 9 μg C L^-1 on day 13, whereby the Glaciecola spp. population increased from 0.16 ± 0.04 × 10⁵ to 0.77 ± 0.31 × 10⁵ cells ml^-1. Hence, the maximum calculated grazing pressure occurred during the highest Glaciecola spp. abundance in the warmer treatments and shortly before the Glaciecola spp. peak in the colder treatments.

During the phytoplankton bloom peak, the average number of total free-living bacterial cells was clearly higher in the warmer (2.15 ± 0.08 × 10⁶, at day 13) than in the colder (0.64 ± 0.10 × 10⁶, at day 17) mesocosms, with 1.74 ± 0.49 × 10⁶ cells (85 ± 7%) and 0.36 ± 0.05 × 10⁶ cells (60 ± 10%) attributable to the Bacteria probe (EUB I-III), respectively. At the same time, Gammaproteobacteria predominated, with a relative abundance in the warmer and colder treatments of 32 ± 9 and 33 ± 6% of total free-living cells, respectively. Within this bacterial group, Glaciecola spp. accounted for 65 ± 13 and 69 ± 30% of total Gammaproteobacteria in the warmer and colder treatments, respectively (Figures 2 and 3). The maximum abundances of Glaciecola spp. were detected in the warmer treatments at the phytoplankton peak (4.20 ± 1.73 × 10⁵ cells, day 13) and in the colder treatments 1 day earlier (1.45 ± 0.82 × 10⁵ cells, day 16). Although Glaciecola spp. achieved higher growth rates at warmer temperatures, its proportions at the phytoplankton peak were similar such that the difference between the two temperature scenarios was not significant (t-test: p = 0.578).

Micro-CARD-FISH analysis revealed that during the phytoplankton bloom peak, 65% of Bacteria (EUB I-III) showed activity, based on their incorporation of [³H]-leucine (data not shown). However, no [³H]-leucine uptake was detected for Glaciecola spp. despite the fact that at least 200 cells that hybridized with probe GC1252 were microscopically examined. The inability of Glaciecola spp. to incorporate [³H]-leucine implies that cell production by these bacteria was not included in the overall calculations of bacterial production. The free-living life style of Glaciecola spp. during the phytoplankton bloom peak was confirmed by micro CARD-FISH analysis with non-pre-filtered water, as none of the cells attached to algae or detritus particles.

According to the measured cell sizes of Glaciecola spp., the mean cell volume of these bacteria was ∼40% higher than the average bacterial cell volume. Glaciecola spp. reached its highest biomass in the warmer treatments on day 10 and in the colder treatments at the phytoplankton peak, which occurred on day 16 (34 ± 12 and 29 ± 19% of total bacterial biomass, respectively). During the proliferation phase of the phytoplankton bloom, total primary production was clearly lower in the warmer than in the colder treatment. By contrast, bacterial production was strongly enhanced at the warmer temperature, as evidenced by the ∼60% higher measured bacterial production (BPm) and the nearly two-fold higher estimated production rate of Glaciecola spp. (BP_GC) (Table 2). This enhanced bacterial activity under warmer conditions in combination with the reduced autotrophic production resulted in clearly higher ratios of BPm and BP_GC to PP (Table 2).

The collapse of the phytoplankton spring bloom was initiated by nutrient depletion (Biermann et al., 2014). This was paralleled by a sharp decline in the abundance of Glaciecola spp. within 2 days, to 0.38 ± 0.67 × 10⁵ cell ml^-1 (2.7 ± 4.9%) and 0.30 ± 0.20 × 10⁵ cell ml^-1 (4.5 ± 3.0%) on day 18 in the warmer and colder treatments, respectively. In both, Glaciecola abundances remained low until the end of the experiment (Figure 2). Interestingly, the dynamics of Glaciecola spp. were much more pronounced than those of total Gammaproteobacteria, whose abundance remained high or even increased further (in the colder) until the end of the experiment. The pattern was indicative of a succession within this bacterial group during the degradation of the phytoplankton bloom.

The occurrence of Glaciecola spp. was also determined in spring bloom mesocosm experiments performed in 2006 (Sommer et al., 2007; Hoppe et al., 2008; Wohlers et al., 2009) by again applying probe GC1252 to samples filtered and stored in the same way as in the above-described experiment (Table 3). Thus, pre-bloom, bloom, and post-bloom samples from mesocosms in which water temperatures were maintained at Δ0°C +Δ2°C +Δ4°C +Δ6°C were examined for Glaciecola spp. The results confirmed the high proportion of Glaciecola spp. in all treatments, as evidenced by abundances of 2–12% (5.0 ± 3.5%) of all bacterial cells close to the phytoplankton bloom peak and independent of the water temperature (Table 3).

To examine the abundance of Glaciecola spp. under comparable in situ conditions in the Baltic Sea, surface water samples were collected in brackish areas from the Gulf of Finland to the southern Baltic Sea during early spring (March), when water temperatures are low (-0.3 to 3.1°C). In the northern and central Baltic Sea, phytoplankton biomass was very low (<50 μg C L^-1) and Glaciecola spp. was not detected. At the more southern stations, sites of an ongoing diatom-dominated phytoplankton bloom, the maximum phytoplankton biomass detected was 347 μg C L^-1 and Skeletonema costatum was the dominant phytoplankton taxa (von Scheibner et al., in preparation). At these southern stations, Glaciecola spp. was detectable but only in low numbers, accounting for 0.01–0.3% of the total DAPI-stained prokaryotic cells.

The successful enumeration of Glaciecola cells with the CARD-FISH probe GC1252 revealed a positive correlation between Glaciecola spp. abundance and Chl a during the exponential growth phase of the phytoplankton spring blooms in all treatments. Glaciecola spp. represented a large part of the bacterial community under both colder and warmer temperatures and dominated the free-living bacterial community. Probe GC1252 was primarily designed to detect a dominant OTU belonging to Glaciecola, as among the tested probes the GC1252 sequence showed the highest specificity within the smallest phylogenetic range in combination with the lowest outgroup hits (Supplementary Table S2). However, GC1252 not only targets the dominant Glaciecola OTU (up to 55% of total clones) but also another Glaciecola OTU present within the clone libraries at low abundance (∼1% of total clones). Nonetheless, the highly abundant Glaciecola species probably largely accounts for the reported results.

A temperature increase of 6°C enhanced the growth of Glaciecola spp., but the general growth pattern at warmer and colder temperatures were remarkably similar. It consisted of a rapid buildup of the populations that paralleled phytoplankton bloom development and the rapid collapse of the Glaciecola population after phytoplankton numbers had peaked (Figure 2). Adaptation by the genus Glaciecola to low temperatures and cold environments is suggested by a number of studies. A recent metagenomic analysis of the genus Glaciecola identified several cold-adapted mechanisms in the pan-genome that would allow the survival and growth of this genus at low temperatures (Qin et al., 2014). Although cold environments (e.g., in polar regions) seem to be the favored habitat of Glaciecola (Bowman et al., 1998; Nichols et al., 1999; van Trappen et al., 2004; Zhang et al., 2006; Prabagaran et al., 2007), its members have also been found in Mediterranean coastal waters (Alonso-Sáez et al., 2007) and during a phytoplankton spring bloom in the North Sea (Teeling et al., 2012).

Glaciecola, and the family Alteromonadaceae as a whole, is generally not an abundant member of bacterioplankton communities, but their abundance can rapidly increase following the input of labile phytoplankton-derived DOC (Eilers et al., 2000; Grossart et al., 2005; Mou et al., 2008; McCarren et al., 2010; Gómez-Consarnau et al., 2012). This is probably due to an up-regulation of metabolic genes associated with DOM utilization at optimal substrate concentrations (Tada et al., 2011). This conclusion is also supported by a metatranscriptomic study, in which, after the addition of diatom-derived DOM, Glaciecola sp. (HTCC2999) dominated carbohydrate metabolism pathways (Beier et al., 2015). Furthermore, growth experiments with specific DOC compounds revealed the generalist nature of some phylotypes in the genus Glaciecola, based on their ability to respond to a broad range of carbon compounds (e.g., Glaciecola pallidula), and the specialist nature of others, as demonstrated by their use of a highly restricted number of carbon compounds (e.g., pyruvate) (Gómez-Consarnau et al., 2012). Although the DOM composition in our mesocosms was not determined, the two dominant diatom species, Skeletonema costatum and Thalassiosira rotula, most likely provided exudated DOC compounds during their active growth phase and thereby effectively generated an ecological niche for Glaciecola. In a previous mesocosm experiment with a similar experimental setup and phytoplankton composition, DOC dynamics and composition were itemized in detail (Engel et al., 2011). In that study, large increases in the percentages of deoxysugars (mainly rhamnose and fucose) were registered before the peak of the bloom and decreased thereafter (Engel et al., 2011). This pattern is remarkably similar to that of the growth dynamics of Glaciecola and indicative of specific DOC-compounds as regulating factors for these bacteria. A limited labile organic carbon supply does not account for the sharp decline in Glaciecola spp. after the phytoplankton peak, since during this phase of the bloom both total DOC and mono-/polysaccharide concentrations strongly increased (Biermann et al., 2014). Rather, the changes in the DOC composition from the active phytoplankton growth phase to the nutrient-limited phytoplankton decline were likely to have regulated the growth of Glaciecola spp. in our experiment and suggest an ecological niche of this group. This conclusion is supported by batch-culture experiments in which, among the free-living bacterial community, the presence of Gammaproteobacteria (Glaciecola and Pseudoalteromonas) was associated exclusively with the growth phases of the diatoms Thalassiosira rotula and Skeletonema costatum (Sapp et al., 2007).

In addition to the assumed changes in DOC composition after the phytoplankton bloom peak, reinforced grazing pressure exerted by small protists may have contributed to the decline in Glaciecola spp. abundance (Figures 2E,F). Gammaproteobacteria, and especially Alteromonas, are particularly vulnerable to protist grazing, as indicated in previous studies that demonstrated selective flagellate grazing on these large, metabolically active bacteria (Beardsley et al., 2003; Worden et al., 2006; Allers et al., 2007). The high rate of phytoplankton-derived DOC utilization by Glaciecola and subsequent grazing by bacterivores may constitute a characteristic link in the microbial food web during phytoplankton blooms.

Considering the above-average cell size of Glaciecola spp., its estimated biomass was about one-third of the total bacterial biomass, indicating an enormous impact of one narrow bacterial clade (or even one species) on carbon processing during the proliferative stage of the phytoplankton bloom. Interestingly, given that Glaciecola spp. proved to be incapable of [³H]-leucine uptake, we no doubt highly underestimated bacterial production (von Scheibner et al., 2014). The inability of certain bacterial taxa to incorporate leucine or thymidine, both of which are used to assess bacterial productivity, is not unusual (Salcher et al., 2013). For this experiment, we estimate that total BP was at least 30% higher during the bloom proliferation phase than during other phases (Table 2) since Glaciecola spp. was not covered by the leucine incorporation approach.

In situ growth rates were calculated based on the net increase in Glaciecola spp. abundance and therefore can be considered as lower-end estimates, because mortality due to grazing or viral lysis was not considered. Further, a more frequent sampling interval might have provided slightly higher growth rate estimates. However, the maximum calculated growth rates were similar to those determined in earlier batch culture experiments at low temperatures. For example, the generation time of the isolate Glaciecola punicea was ∼13 h at 8°C and ∼26 h at 2°C (Nichols et al., 1999). A comparable net growth rate (μ ≈ 2.3 day^-1) was determined for the family Alteromonadaceae in incubation experiments, and it increased after the reduction of predators (μ ≈ 4.0 day^-1) and viruses (μ ≈ 5.8 day^-1) (Ferrera et al., 2011). The high growth rates at cold temperatures may explain the ability of Alteromonadaceae to rapidly proliferate such that it dominates microbial communities exposed to a substrate pulse (Eilers et al., 2000; Beardsley et al., 2003; Alonso-Sáez et al., 2007; Mou et al., 2008). Substrate pulses, such as occur during phytoplankton blooms, can promote the expansion of specific bacterial populations (Pernthaler and Amann, 2005), which in turn can briefly but strongly impact carbon processing and transfer to higher trophic levels and thereby disrupt the general seasonal bacterial succession (Lindh et al., 2015).

The detection of a substantial abundance of Glaciecola spp. also in a previous mesocosm experiment that included a diatom bloom at low water temperature (Table 3) suggests that our results exemplify a general pattern. The relative low abundance of Glaciecola spp. in the southern Baltic Sea during an early spring bloom is a deviation from this pattern (Table 3) but might have resulted from the relatively low phytoplankton biomass or differences in phytoplankton community composition. Thus, only intensive and detailed studies can reveal natural seasonal bacterial succession involving transiently expanding bacterial taxa. For example, Teeling et al. (2012) carried out a detailed investigation of the bacterioplankton response to a diatom bloom in the North Sea. Their results suggested that algal substrate availability provides a series of ecological niches in which specialized populations, such as the Gammaproteobacteria Glaciecola spp. and Reinekea spp., can increase (Teeling et al., 2012).

Since the original description of the genus Glaciecola as comprising gram-negative, aerobic, psychrophilic slightly halophilic bacteria (Bowman et al., 1998), there have been many extensions and modifications (van Trappen et al., 2004; Baik et al., 2006; Zhang et al., 2006; Shivaji and Reddy, 2014). However, most species seem to be psychrophilic, including those responsible for the results of this study, which achieved high in situ net growth rates at a low temperature of ∼2°C. This is in accordance with the lack of a significant temperature effects on the relative abundance of Glaciecola spp. Also the relatively low Q₁₀ value of 2.2 is within the normal range of temperature-accelerated bacterial growth (Kirchman et al., 2009). Remarkably, despite higher grazing pressure from HNF, the growth rate of Glaciecola spp. at the low temperature was high (∼1.10 day^-1), in contrast to that of the total bacterial community (∼0.4 day^-1).

Yet, besides direct temperature effects, the growth dynamic of Glaciecola spp. determined in this experiment could also have been influenced by biotic interactions within a complex food-web as well as by differences in phytoplankton community composition or grazer abundance (Lewandowska and Sommer, 2010; Aberle et al., 2012; Winder et al., 2012). Overall, the coincidence of the high growth rate and abundance of Glaciecola spp. with the onset of the phytoplankton bloom indicates a tight coupling between blooming diatoms and Glaciecola spp. Moreover, this relationship provides evidence of a key role of cold-adapted Glaciecola spp. in the processing of phytoplankton-derived DOC during the proliferation of phytoplankton at low water temperatures. Increasing the experimental temperature did not substantially change this relationship, as Glaciecola spp. remained the dominant bacterial taxa also after warming, although other bacterial groups gained in importance as well (von Scheibner et al., 2014). Whether Glaciecola is replaced by other taxa at temperatures higher than that tested here and how this dynamic is regulated by the interplay of substrate composition, grazing, and viral lysis remain to be determined. It will also be interesting to compare our results with those obtained in other marine environments (e.g., those with higher salinity), where different bacterial taxa dominate the respective carbon flow during early phytoplankton blooms.

MvS, US, KJ designed the experiment. MvS performed the sampling and analysis. MvS, US, KJ wrote the manuscript.

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.