The Algicidal Bacterium Kordia algicida Shapes a Natural Plankton Community

Plankton communities change on a seasonal basis in temperate systems, with distinct succession patterns; this is mainly due to algal species that have their optimal timing relative to environmental conditions. We know that bacterial populations are also instrumental in the decay and termination of phytoplankton blooms. Here, we describe algicidal bacteria as modulators of this important species succession. Upon treatment of a natural plankton consortium with an algicidal bacterium, we observed a strong shift in the phytoplankton community structure, compared to controls, resulting in formation of a succeeding Phaeocystis bloom. Blooms of this alga have a substantial impact on global biogeochemical and ecological cycles, as they are responsible for a substantial proportion of primary production during spring in the North Sea. We propose that one of the key factors influencing such community shifts may be algicidal bacteria.

induced forcing, rather than phytoplankton host specificity (4). There is an association between a large abundance of bacteria and phytoplankton blooms after their peak and during the declining phase. Substrates released from degrading algae thereby support bacterial growth (6)(7)(8). However, bacteria often also play active roles in the shaping of plankton communities. Symbiotic, mutualistic, competitive, and predatory interactions can directly affect communities of primary producers (9), especially with algicidal bacteria, which have the potential to modulate entire phytoplankton communities. Algicidal bacteria can inhibit microalgal growth and use algal exudates as a resource or they can actively lyse the algae and utilize the released nutrients (10). Research on this class of bacteria is often motivated by potential applications in biotechnology, environmental engineering, and aquaculture. In particular, the possibility of being able to control harmful algal blooms by using algicidal bacteria that act against specific hosts is under discussion (11). As a consequence, knowledge has been accumulated mostly in bilateral alga-bacterium studies, and studies mostly exclude the dynamics of algicidal bacterial growth and the influence of the bacteria on the complex phytoplankton community in nature.
The mode of action of algicidal bacteria can be categorized in local interactions relying on direct contact of bacteria with their target algae or phycosphere interactions within the immediate vicinity of the cells. However, free-living bacteria can also become harmful to phytoplankton when they reach high cell concentrations. A recent review summarized the diversity of bacterial strategies and their ecological relevance in negatively affecting algal growth (10). The specificity of algicidal bacteria varies profoundly, from highly specialized bacteria that are associated with one host species to nearly universally active bacteria that affect most phytoplankton community members. The algicidal flavobacterium Kordia algicida, which is the subject of the current study, was isolated during a bloom of the diatom Skeletonema costatum (12). It is capable of lysing a broad range of algal species, including the diatoms Skeletonema costatum, Thalassiosira weissflogii, and Phaeodactylum tricornutum (13). The lysis is dependent on bacterial protease activity, regulated in a quorum-sensing-dependent manner (13). While most tested algae were lysed by K. algicida or its extracts, the diatom Chaetoceros didymus was resistant. Resistance was associated with upregulation of proteases from the algae, and these proteases are suspected to counteract the enzymes of the bacteria (14). In a second line of defense, oxylipins from the diatom can contribute to the resistance (15).
The specific resistance of Chaetoceros didymus prompted us to ask the question of how the bacteria mediate natural plankton populations with mixed assemblages of resistant and susceptible species. Considering the fact that microalgae compete for light and nutrients during bloom formation, removing one or more species from the assemblage might lead to dramatic shifts within the phytoplankton consortium.
In this study, we hypothesized that parts of the resistant phytoplankton community would benefit from the lysis of susceptible species. Here we tested whether and how this might lead to a population shift, thereby contributing to the understanding of the specificity and complexity of alga-bacterium interactions. We selected the especially well-characterized plankton community of Helgoland, which has been closely monitored for more than 50 years, and manipulated it with the algicidal bacterium K. algicida, which is globally distributed and lyses diatoms with high efficiency (3). Knowledge about cascading effects within microbial communities in the environment will help to explain natural patterns in plankton development.

RESULTS
Dominant phytoplankton members at Helgoland Roads. Plankton community enclosures were set up during the time of a naturally occurring bloom of Chaetoceros spp. just after the peak of a bloom of Phaeocystis spp., using water from the site where samples for species enumeration were taken (Fig. 1). Diatoms at Helgoland Roads at the end of April 2016 were dominated by Chaetoceros spp. (Fig. 1A and B). Chaetoceros spp. in outside waters decreased slightly during the time of our experiment (Fig. 1B).
Between sampling and inoculation of the enclosures, the bloom of Phaeocystis spp. outside declined further and then stayed comparatively stable over the course of the experiment (Fig. 1C). After the experiment, Chaetoceros spp. continued to grow in the outside waters until collapsing 1 month later (Fig. 1E). Phaeocystis spp. in the outside waters increased before peaking again approximately 3 months later (Fig. 1F).
Phytoplankton community patterns in enclosures. We monitored the phytoplankton community by light microscopy at a species or genus level. The survey included 12 diatoms, 7 dinoflagellates, 3 cryptophytes, 1 chlorophyte, 1 haptophyte, and 1 raphidophyte, thereby covering the most abundant species (Table 1).
At the beginning of our experiment, the community was dominated by the diatom Chaetoceros socialis (ϳ2 ϫ 10 6 cells per liter) (Fig. 2) and the haptophyte Phaeocystis sp. (3 ϫ 10 5 to 4 ϫ 10 5 cells per liter) (Fig. 2). The species composition in the enclosures at the start of the experiment was thus in accordance with the cell abundances from the outside waters (Fig. 1). Microscopic evaluation of plankton samples before Lugol staining revealed that nearly all Phaeocystis sp. cells were found in colonies.
C. socialis abundance was stable in control enclosures throughout the experiment, as were Phaeocystis sp. concentrations. A significant increase (P value of Ͻ0.001, compared to day 0) was seen only on day 8 ( Fig. 2A; also see Table S1 and Table S2 in the supplemental material). The stable cell counts of the two most dominant species were reflected in the uniform total chlorophyll a contents of the samples throughout the experiment. This stability was a prerequisite for evaluation of the effects of the phytoplankton assemblage manipulation by addition of K. algicida, which was previously isolated during a bloom of the diatom Skeletonema costatum (12). The effects of the bacteria on total chlorophyll a concentrations, on individual phytoplankton species, on the bacterial community, and on nutrient levels were monitored and revealed significant changes induced by the introduced bacteria.
Infecting the enclosures with K. algicida substantially changed the phytoplankton community composition over the following 8 days. In both infection scenarios, i.e., with high bacterial cell densities (final optical density at 550 nm [OD 550 ] of 0.02) and low cell densities (final OD 550 of 0.01), the total chlorophyll a contents decreased significantly (P values of Ͻ0.001), compared to the control, within 24 h ( Fig. 2B and C; also see Table  S3). In both infection scenarios, chlorophyll contents remained significantly lower (P values of Ͻ0.001) than the control levels throughout the experiment (except in the low infection scenario versus control on day 6, with data not normally distributed). From day 5 to day 7, there was significantly lower chlorophyll a fluorescence in the low infection scenario than in the high infection scenario.
In both infection scenarios, the abundance of C. socialis decreased to below 10% of the initial cell counts over the course of the experiment ( Fig. 2B and C). C. socialis abundance was significantly lower than the control with both treatments from day 4 onward. There was no significant difference in C. socialis cell density between the two infection scenarios at any time point (Table S4).
Phaeocystis sp. cell counts doubled within 2 days after infection (significantly different from the control only for the high infection scenario), and then cell counts stayed constant until there was a drastic increase at day 8 to reach ϳ10-fold greater concentrations, compared to day 0 ( Fig. 2B and C). From day 6 onward, Phaeocystis sp. cell densities were significantly higher than the control levels in both infection scenarios, while there was no statistically significant difference between the two infection scenarios at any time point (Table S4). The increase in Phaeocystis sp. levels at day 8 was not reflected in total in situ chlorophyll a levels, which is in agreement with previous findings (16). The sum of all other phytoplankton species did not exceed 7 ϫ 10 5 cells per liter in control or infected enclosures at any time of sampling. Due to mainly episodic occurrences of these minor phytoplankton species, no obvious trend in cell counts that was dependent on the treatment was observed. Total bacterial abundance. Bacterial abundance was monitored as a group parameter throughout the experiment. Total bacterial abundances in control enclosures remained constant throughout the 8 days of incubation (P ϭ 0.075, analysis of variance  Table S4 in the supplemental material). The number sign marks a data point without normal distribution. Further statistical analyses of cell counts and chlorophyll a levels for the comparison between treatments (B and C) and for assessment of control stability (A) are given in Tables S1 to S4.
[ANOVA]) (Table S5), compared to day 0 (Fig. 3). With low and high K. algicida inoculations, the bacterial densities increased 100-and 200-fold, respectively, compared to control levels (Fig. 3), thus rendering K. algicida the dominant bacterial species after infection. Bacterial densities in the high infection scenario were similar to those in the control in which K. algicida was added to sterile filtered seawater. Bacterial abundance in K. algicida-treated enclosures with the natural plankton community increased slightly but significantly during the experiment ( Fig. 3; also see Table S5).
The viability of K. algicida was confirmed at the end of the experiment by plating of bacteria from high and low infection scenarios. In this nonquantitative survey, K. algicida colonies were clearly identifiable by their yellow pigmentation and colony morphology. No such colonies were found when a sample from the natural community was plated. The 16S rRNA sequencing of five randomly picked yellow colonies and five morphologically different colonies confirmed the identification. We found the majority of the colonies in the treatments to be K. algicida, confirming the survival of the bacteria throughout the experiment.
Nutrients. Levels of dissolved organic carbon (DOC) and inorganic nitrogen and phosphate in control enclosures containing the natural phytoplankton community did not change significantly over the course of the experiment (Table 2; also see Table S6 and Table S7). Elevated DOC contents were found in infected natural communities and controls in which K. algicida was introduced into sterile filtered seawater. DOC levels with these treatments decreased over the course of the experiment (day 4, P values of Յ0.01 for both; day 8, P values of Յ0.004 for both) ( Table 2; also see Table S6). Initial concentrations of inorganic nutrients were higher in treatments with K. algicida inoculation than in the seawater control ( Table 2). The inoculation with K. algicida in sterile filtered seawater resulted in elevated nitrite, ammonium, and phosphate concentrations, presumably due to carryover of nutrients associated with the bacterial cells. Lysis of the phytoplankton community contributed to a further substantial and significant increase in nutrients. Comparison of the infection scenario to K. algicida inoculation in sterile filtered seawater revealed that ϳ40% of the additional phosphate could be attributed to nutrient release during infection. For ammonium, ϳ20% of the surplus could be attributed to release during infection. Ammonium concentrations in infected plankton consortia were significantly increased at day 4 (P ϭ 0.022) and day 8 (P ϭ 0.011), compared to initial values (Table S8). Nitrate concentrations did not exceed the lower working limit of 7.14 M in any treatment. . Error bars denote SDs between biological triplicates except for K. algicida in seawater, which was technically replicated (means Ϯ SDs of three replicates; no statistical test was applied for this treatment). Significant differences from one-way ANOVA, compared to day 0 for the respective treatment, are indicated (*, P Ͻ 0.05; **, P Ͻ 0.01) (see Table S5 in the supplemental material).
Confirming laboratory experiments. Following the field experiment, C. socialis was isolated from the control enclosures. The identity of C. socialis was confirmed by genetic analysis. According to a BLAST search using the D1-D3 large subunit (LSU) rDNA sequence (GenBank accession no. MH992142) as a query, the most similar sequence (GenBank accession no. JQ217339.1) (17), with 99% identity, originated from C. socialis. Subsequent laboratory experiments with the monoclonal culture supported the findings of the enclosure experiment. C. socialis was susceptible to K. algicida under conditions similar to those in the field and was lysed rapidly after inoculation with the bacterium (Fig. S1).

DISCUSSION
Algicidal bacteria are important players in marine ecosystems. To date, however, they have mainly been studied in simplified bioassays with isolated species. Investigations thus focus on rather artificial interactions between two partners, without taking into account potential cascading effects that might occur in the natural surroundings (10). This is also the case for the flavobacterium K. algicida, which has been tested with regard to its activity on different isolated phytoplankton species. Those previous investigations revealed that, of the four species tested, only Chaetoceros didymus was resistant (13). The approach of community enclosures and manipulation presented in this study overcame the experimental limitations of few interaction partners in previous infection experiments. This enabled an evaluation of whether species-specific resistance is also observed in nature and how such resistant species respond to the lysis of surrounding cells. We could observe direct and cascading effects after infection of a natural Chaetoceros bloom with the algicidal bacterium. C. socialis cell counts decreased dramatically after inoculation with K. algicida, compared to the stable counts in the control. The treatment caused an increase in Phaeocystis sp. cell counts; this species obviously benefited from lysis of the competing dominant C. socialis. Interestingly, Phaeocystis sp. levels were already declining in the outside waters at the start of the experiment. Removal of the competing C. socialis by the algicidal bacteria apparently restored favorable conditions. The algicidal bacteria thus have the potential to shift the entire plankton population, since resistant species rapidly take over the liberated resources and benefit from the removal of competitors.
The long-term phytoplankton survey at Helgoland Roads conducted by the Biologische Anstalt Helgoland (18), with its daily screening, permitted us to time the experiment during a diatom bloom. Chaetoceros species often occur in the spring around March to May. The peak of the bloom in 2016 defined the beginning of our experiment. The bloom in 2016 was the most intensive one recorded within the monitoring period (from 2010 to 2017) (Fig. 1). The blooming species was confirmed to be C. socialis and was the most abundant algal species, in terms of cell numbers, in the sea and in the enclosure experiments at the start of our experiment. Total diatom cell counts and Chaetoceros cell counts were virtually superimposable during the course of the experiment, indicating that the diatom community was consistently dominated by this species (Fig. 1A and B). The Chaetoceros genus is particularly important at Helgoland Roads and the habitat hosts different Chaetoceros species, with each showing an individual growth pattern throughout the year (3). The second dominant species was Phaeocystis sp., which was already in the declining phase of a bloom that peaked before the start of our experiment (Fig. 1). To study specifically the effect of the introduced algicidal bacterium in our experiment, we limited grazers by filtering the water through 100-m filters, to remove the majority of herbivores. The limited effect of grazing was confirmed by the stable cell counts in the control enclosures ( Fig. 2A). We also added the bacteria at a high optical density, to ensure that immediate bloom termination was achieved and to exclude overlaying effects of competing bacteria. The success of this strategy was documented by the fact that the two bacterial treatments (high and low infection scenarios) caused similar effects (Fig. 2).
The susceptibility of C. socialis was initially unexpected, since the only tested diatom of the same genus, C. didymus, was resistant to K. algicida in a laboratory screening (13). However, such variability in algal host susceptibility toward algicidal bacteria, even within isolates from the same species, has been already documented in the literature (19). Apart from K. algicida, other bacteria are known to affect Chaetoceros negatively; thus, this genus has no outstanding resistance traits (20)(21)(22). In contrast, Phaeocystis sp. was not lysed by the bacterium and cell counts increased immediately on day 1 after lysis of the diatoms. The observed host specificity is in agreement with laboratory experiments on the algicidal bacterium Brevibacterium sp., which is active against a broad selection of microalgae, including a Chaetoceros sp., but does not affect the Phaeocystis sp. (23). Phaeocystis resistance has been connected to the ability to form large colonies. In colonies, cells produce transparent exopolymer particles (TEPs) that protect the colony from infections (24,25). TEPs have been directly connected to physical protection by scavenging bacteria and viruses (26,27). The algicidal agent excreted by K. algicida is a protease (13), and the saccharide-rich exopolymeric matrix might prevent the protease from reaching cellular structures by limiting diffusion (28).
The kinetics of the C. socialis decline and the rise of Phaeocystis were strikingly similar for the two infected treatments, and we conclude that the inoculation density was already sufficiently high in the low inoculation scenario to trigger a maximum effect. Phaeocystis bloomed twice in 2016, once in parallel with C. socialis and once ϳ3 months after the experiment, when the diatom spring bloom was over. The algicidal bacterium thereby accelerated the natural succession of Phaeocystis, inducing the second bloom already a few days after the decline of the diatoms that was triggered by the bacterium. Such community manipulations by algicidal bacteria might have to be generally considered in Helgoland and northern waters, and future studies should include the monitoring of algicidal bacteria during bloom decline. In the period from 2010 to 2016, Phaeocystis often followed the diatom blooms in Helgoland (with the exception of 2012 and 2016, when they cooccurred) (Fig. 1) (3, 4, 29). This seasonal succession was confirmed at seven long-term monitoring stations along the North Sea coast, where a diatom-dominated bloom was often followed by a Phaeocystis bloom and then by mixed blooms in summer (29). Such patterns also hold true for other phytoplankton field studies apart from North Sea waters (30)(31)(32)(33).
Compared to diatoms, Phaeocystis species have greater light requirements and therefore should be outcompeted during the low-light conditions of a dense spring bloom of other species (34). The addition of K. algicida removes the most dominant competitor of the Phaeocystis sp. from our enclosure experiments and we assume allows the Phaeocystis sp. to thrive due to improved light availability. This effect would be relevant in our experiments, where we simulated low-light conditions in the turbid coastal environment at Helgoland Roads (35). This idea is supported by the previous evaluation of 50 years of high-resolution data, demonstrating that light is the main driver for phytoplankton dynamics at Helgoland Roads (3).
A second factor supporting the Phaeocystis bloom can be the availability of nutrients. We detected higher nutrient concentrations after bacterial infections. Portions of the nutrients stem from carryover during infection. The additional increases of nutrients and DOC in the infected enclosures indicate contributions of cellular nutrients released during the rapid lysis of C. socialis (Table 1). Already within the first day, a 38% decrease in total chlorophyll a fluorescence was observed in the high infection scenario. The associated cell lysis liberated substantial amounts of intracellularly stored nutrients and DOC. Diatoms are known to efficiently store nutrients such as nitrate, which can be liberated upon lysis (36). Adding to the complexity of the system, however, even ammonium has been discussed recently as an algicide (37). Excreted bacterial proteases that mediate phytoplankton lysis can also contribute to higher initial DOC concentrations (13). DOC levels were lower at the end of the experiment (64% of initial levels) in the treatments in which the algal community was exposed to K. algicida, compared to the control in which K. algicida was added to sterile filtered seawater (82% of initial levels). The observed bacterial growth and/or the blooming Phaeocystis likely contributed to the consumption. A long-standing hypothesis claims that growth of Phaeocystis may be dependent on dissolved organic matter from decaying diatom blooms (38,39). Haptophytes in general have been linked to a mixotrophic lifestyle (40,41). Our data are in accordance with this hypothesis, which would explain the onset of Phaeocystis growth right after the decline of the diatoms. Signaling metabolites such as quorumsensing molecules or allelopathic chemicals also might contribute to the observed succession, but these multiple factors cannot be fully untangled with the enclosure experiments.
In conclusion, we document that algicidal bacteria can shift natural plankton populations and accelerate plankton succession. In Helgoland waters, the introduction of K. algicida led to an accelerated bloom decline of the dominant diatom C. socialis and the earlier onset of Phaeocystis sp. Resistance to the bacteria thus provides a competitive advantage in the multispecies communities of the plankton.

MATERIALS AND METHODS
Experimental design. To start the enclosure experiments, a natural seawater community was taken from subsurface water during the phytoplankton spring bloom period at the research site Kabeltonne at Helgoland Roads in the German North Sea (54˚11.3=N, 7˚54.0=E) on 19 April 2016. The collection procedure and sampling site were the same as described for routine sampling during the long-term plankton monitoring at the Biologische Anstalt Helgoland (42,43). Polycarbonate vessels (25 liters; Nalgene, Rochester, NY, USA) were acid washed (10% hydrochloric acid), rinsed with ultrapure water, and autoclaved. Before filling, the vessels were additionally washed with natural seawater, and natural plankton samples were filtered (100-m mesh size) to exclude macro-zooplankton. Nine culturing vessels were set up, each containing 20 liters of the resulting natural community. Additionally, two control vessels with sterile filtered seawater (20 liters each) were prepared by further filtration through 1.2-m filters (GF-C; Whatman, Kent, UK), followed by filtration through 0.2-m filters (Filtropur S; Sarstedt, Nümbrecht, Germany) under reduced pressure. In sum, 11 vessels containing either the natural community or sterile filtered seawater were set up in a climate chamber (7.6°C, light/dark cycle of 15 h/9 h, and light intensity of 3 mol ϫ m Ϫ2 ϫ s Ϫ1 ). Treatments were acclimated for 2 days until the experiment was started by inoculation with K. algicida. K. algicida was plated 4 days in advance on petri dishes with solid marine broth agar and was cultured at room temperature. Colonies were washed off with sterile filtered seawater on the day of inoculation. This procedure was applied for approximately 200 plates (10-cm diameter) with bacterial lawns, resulting in a dense bacterial solution that was subsequently diluted (1:20 [vol/vol]) to reach a final OD 550 of 0.43. This bacterial suspension was added to bottles filled with the natural community to a final OD 550 of 0.01 (23 ml; low treatment [n ϭ 3]) or 0.02 (46 ml; high treatment [n ϭ 3]). A vessel containing sterile filtered seawater was inoculated with bacterial solution to a final OD 550 of 0.02 (K. algicida control [n ϭ 1]). Three vessels containing the natural community and one containing sterile filtered seawater served as controls and were not treated with K. algicida. Prior to sampling, all vessels were briefly and vigorously shaken (horizontally) manually every day, and their position in the climate chamber was subsequently exchanged. Total chlorophyll a levels, levels of individual phytoplankton species, total bacterial abundance, the presence of K. algicida, inorganic nitrogen and phosphate levels, and dissolved organic matter (DOC) levels were followed over the course of the experiment (8 days). Sampling was performed every second day except for chlorophyll a fluorescence, which was determined daily.
were manually edited in BioEdit (48). The final sequence was subjected to a BLAST search (BLASTn and Megablast) against the nucleotide collection (nonredundant/nucleotide) (49) on 1 December 2016.
Field data. Total diatom abundance and cell counts for Chaetoceros spp. and Phaeocystis spp. in Fig.  1 were obtained as described previously (3). Data are deposited and updated regularly on the Pangea platform (18).
Statistics. Statistical analysis was conducted using SigmaPlot software (version 11 or higher; Systat Software Inc., London, UK). An unpaired two-sided t test was used to analyze the significance of differences in enclosure chlorophyll a levels and cell counts in infection scenarios, compared to the uninfected control, for each time point and species individually. Normal distribution (Shapiro-Wilk test) and equal variance of the data were tested in advance. If the normality test failed, then a Mann-Whitney rank sum test was performed instead. A one-way ANOVA was carried out for the analysis of control stability and for the comparisons of total bacterial abundances and DOC and ammonium concentrations over the course of the experiment; it was followed by either a Bonferroni or Holm-Sidak post hoc test when there were significant differences within the data set. Phosphate concentrations were tested using one-way ANOVA on ranks. Levels of significance are indicated, and P values of Ͼ0.05 were considered not significant. Data points in the figures without normal distribution are indicated.
Accession number(s). Sequence data were deposited in GenBank and are available under accession no. MH992142.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AEM .02779-18.

ACKNOWLEDGMENTS
This work was funded by the German Research Foundation within the CRC Chem-BioSys (grant CRC1127) and the Jena School for Microbial Communication. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Ines Hilke (Max Planck Institute for Biogeochemistry, Jena, Germany) is acknowledged for nutrient analysis. The team of the Biological Research Station Helgoland, especially Eva-Maria Brodte, Alexandra Kraberg, and the research vessel crew, is acknowledged.