Consistency in microbiomes in cultures of Alexandrium species isolated from brackish and marine waters

Summary Phytoplankton and bacteria interactions have a significant role in aquatic ecosystem functioning. Associations can range from mutualistic to parasitic, shaping biogeochemical cycles and having a direct influence on phytoplankton growth. How variations in phenotype and sampling location, affect the phytoplankton microbiome is largely unknown. A high‐resolution characterization of the bacterial community in cultures of the dinoflagellate Alexandrium was performed on strains isolated from different geographical locations and at varying anthropogenic impact levels. Microbiomes of Baltic Sea Alexandrium ostenfeldii isolates were dominated by Betaproteobacteria and were consistent over phenotypic and genotypic Alexandrium strain variation, resulting in identification of an A. ostenfeldii core microbiome. Comparisons with in situ bacterial communities showed that taxa found in this A. ostenfeldii core were specifically associated to dinoflagellate dynamics in the Baltic Sea. Microbiomes of Alexandrium tamarense and minutum, isolated from the Mediterranean Sea, differed from those of A. ostenfeldii in bacterial diversity and composition but displayed high consistency, and a core set of bacterial taxa was identified. This indicates that Alexandrium isolates with diverse phenotypes host predictable, species‐specific, core microbiomes reflecting the abiotic conditions from which they were isolated. These findings enable in‐depth studies of potential interactions occurring between Alexandrium and specific bacterial taxa.


Introduction
Interactions between phytoplankton and bacteria in aquatic environments are intricate and have large impacts on ecosystem functioning (Azam and Malfatti, 2007;Buchan et al., 2014). Specifically, phytoplankton bacteria interactions, ranging from mutualism (Amin et al., 2009) to resource competition (Risgaard-Petersen et al., 2004), have effects on the whole ecosystem in terms of the amount of carbon that is recycled or stored (Cole, 1982;Jiao et al., 2010), the level of regenerated nitrogen (Eppley and Peterson, 1979) and utilization of captured energy (Buchan et al., 2014). The diversity, functioning and efficiency of bacteria associating with phytoplankton cells (the microbiome) are thought to be determined by the biochemical composition of phytoplankton and their released organic matter (Azam and Malfatti, 2007;Buchan et al., 2014;Seymour et al., 2017). The composition of the microbiome has been shown to be host specific in association with, for example, diatoms and dinoflagellates (Schäfer et al., 2002;Guannel et al., 2011;Lawson et al., 2017;Behringer et al., 2018).
Dinoflagellates release organic matter continuously during their life cycle (Villacorte et al., 2015) affecting both composition and function of their microbiome (Sarmento et al., 2013). The role of bacteria-dinoflagellate interactions may include growth stimuli (Sakami et al., 1999;Ferrier et al., 2002;Bolch et al., 2011), life cycle cues (Adachi et al., 2003) and regulation of toxin production (Doucette and Powell, 1997;Gallacher and Smith, 1999). Lawson et al. (2017) showed the existence of a consistent core microbiome among dinoflagellate Symbiodinium clades, pointing at taxonomically consistent interactions with significance to dinoflagellate functions. Few studies have characterized these complex interactions (Hattenrath-Lehmann and Gobler, 2017), which is a prerequisite for understanding consequences for large scale biogeochemical processes.
Dinoflagellates of the genus Alexandrium are reported from many geographical areas in the world's oceans (Anderson et al., 2012) and are known to release bioactive compounds, potent toxins (Anderson et al., 2012) and allelochemicals that are poorly chemically characterized (Ma et al., 2009), which can interact with microbial food web dynamics (Weissbach et al., 2011). Alexandrium populations display a large intraspecific phenotypic diversity, in terms of growth rate (Suikkanen et al., 2013;Brandenburg et al., 2018), toxin production Suikkanen et al., 2013;Martens et al., 2017;Brandenburg et al., 2018), allelochemical activity (Hakanen et al., 2014;Brandenburg et al., 2018), nutritional strategies (Glibert and Legrand, 2006) and ability to bioluminesce (Valiadi et al., 2012). Due to this large phenotypic diversity Alexandrium is an ideal model organism for investigating the consistency of microbiomes over host specific, intraspecific or environmental variations (Tahvanainen et al., 2012;Suikkanen et al., 2013;Hakanen et al., 2014). We hypothesize that the high diversity among Alexandrium isolates would result in diverse, isolate specific, microbiomes in culture.
Several earlier studies have taxonomically characterized bacteria associated with marine Alexandrium in culture, identifying Proteobacteria [mainly Roseobacter (Alphaproteobacteria) and Alteromonas (Gammaproteobacteria) clades] and Bacteroidetes as dominating bacterial phyla (Kopp, 1997;Hold et al., 2001;Biegala et al., 2002;Jasti et al., 2005). However, these studies are limited to marine conditions (salinity~31) and low sequencing resolution and coverage, which preclude conclusions of the microbiome consistency and specific associations of individual bacterial species and Alexandrium isolates. A recent study of an Alexandrium bloom at marine conditions, using 16S amplicon sequencing, found associations with Flavobacteriia (Owenweeksia and the NS5 marine group), Alpha-(Rhodobacterales) and Gammaprotebacteria (Alteromonadales) (Hattenrath-Lehmann and Gobler, 2017), showing concordance, at phylum and order level, between findings in culture with those from the environment.
In this study, we define microbiome as consisting of the whole bacterial assemblage found in the Alexandrium cultures. Our aim was to explore how host specificity, intraspecies variations and sampling location influence the diversity and consistency of the microbiome of A. ostenfeldii, A. minutum and A. tamarense isolates, and investigate the existence of a species-specific core microbiome. In total 28 strains were included, of which 20 belong to A. ostenfeldii, isolated from four distinct locations in the brackish Baltic Sea Proper. These strains have been shown to have phenotypic (Suikkanen et al., 2013;Hakanen et al., 2014) and genotypic (Tahvanainen et al., 2012) differences. The remaining strains belonged to A. minutum (n = 7) and A. tamarense (n = 1), isolated from eight locations in the northwest Mediterranean Sea (marine conditions) with different anthropogenic impact levels, classified using the Land Uses Simplified Index, LUSI (E. Flo, unpubl.). Two in situ datasets from the Baltic Sea were used to investigate if members from the identified core microbiome of A. ostenfeldii in culture were linked to natural dinoflagellate dynamics.

Identification of dinoflagellates and diversity of microbiomes
The dinoflagellates were taxonomically identified by amplification and sequencing of their ITS-regions which corroborated identifications previously made using microscopy (S. Fraga and I. Bravo, unpubl.; Kremp et al., 2009) except for the strain AL10C, which was previously identified as A. minutum, but showed a high level of identity (99%) to an A. tamarense strain (AJ005048.1; Supporting Information Fig. S1). The Alexandrium strains were cultured using sterile methods at conditions mimicking those of their respective origins, brackish (salinity 6.5 or 7) for A. ostenfeldii and marine (salinity 31) for A. minutum/tamarense (Supporting Information Table S1). Nucleic acid samples were obtained by filtering cultures (density of A. ostenfeldii: 0.03-0.15 × 10 5 cells/ml and A. minutum/tamarense: 2.10-3.50 x 10 5 cells/ml; Supporting Information Table S1) using 0.2 μm filters, capturing the whole bacterial assemblage of the cultures at late exponential growth phase. The microbiomes were characterized using high-throughput 16S rRNA gene amplicon sequencing, which resulted in a total of 9 776 193 sequences and 3649 operational taxonomic units (OTUs; 97% similarity clustering; Supporting Information Table S2). In this first high resolution characterization of the associated bacterial community of multiple species of Alexandrium, large differences in richness and diversity of the microbiomes were observed between the species (Supporting Information Fig. S2). At the OTU-level, the A. ostenfeldii microbiomes showed a higher level of richness [average chao1-index: 1081, AE 252 (SD)] and a higher diversity [average alpha Shannon index: 4.22, AE 0.67 (SD)] compared to the A. minutum/tamarense richness [average chao1-index: 481, AE 152 (SD)] and diversity [average alpha Shannon index: 3.15, AE 0.56 (SD)] (Supporting Information Fig. S2). The differences in richness and diversity between A. ostenfeldii and A. minutum/tamarense microbiomes might reflect the longer time the latter species were kept in culture, 2 years compared to 8-16 years respectively (Supporting Information Table S1).

Similarity of microbiome composition
To study if the composition of the microbiomes associated with Alexandrium could be related to intraspecies variations of Alexandrium strains or specific conditions at the location of isolation an nMDS analysis, using Bray-Curtis dissimilarity, was performed. This analysis showed that there was a clear separation between A. ostenfeldii and the A. minutum/tamarense strains regarding frequently occurring taxa ( Fig. 1 and Supporting Information Fig. S3). The A. ostenfeldii microbiomes were very homogeneous (permAnova R 2 = 0.13, P value: 0.73; Fig. 1) considering that the strains were isolated from four geographically distant locations of the Baltic Sea and the inherent intraspecific variations previously demonstrated for A. ostenfeldii (Kremp et al., 2016;Martens et al., 2017;Supporting Information Fig. S4). Consequently, our results indicate that intraspecific variations or sampling location in the Baltic Sea did not have a significant impact on the composition of the microbiomes. These findings indicate that the interactions that occur between the dinoflagellate and their microbiomes, are consistent within a species-specific functional range of Alexandrium, disregarding phenotypic variations of individual dinoflagellate strains. In the nMDS, the A. minutum/tamarense associated microbiomes were dispersed with a spread similar to that of the A. ostenfeldii microbiomes (Fig. 1), and they did not cluster according to the anthropogenic impact level at each location (estimated using LUSI; E. Flo, unpubl.; Supporting Information Fig. S4). The results indicate a high microbiome similarity also across the A. minutum/tamarense strains. Taken together, our results show that both the environment at the sampling location, brackish or marine, and the species of Alexandrium, had important impacts on the composition of the microbiomes.

Microbiome of A. ostenfeldii isolates
The microbiomes of the A. ostenfeldii isolates were highly similar, forming four clusters (based on Bray-Curtis dissimilarity) of 3-7 samples each, originating from different locations (Supporting Information Fig. S3). These microbiomes included sequences affiliated with a broad range of bacterial classes: Beta-(58% of total sequences), Alpha-(14% of total sequences), Gamma-(3% of total sequences) and Deltaproteobacteria (6% of total sequences), Flavobacteriia (9% of total sequences) and Cytophagia (1.5% of total sequences). Represented, at family level, by Methylophilaceae (28% of total sequences), Hydrogenophilaceae (10% of total sequences), Flavobacteriaceae (7% of total sequences), Comamonadaceae (4% of total sequences), Desulfobulbaceae (4% of total sequences) and Rhodobacteraceae (3% of total sequences; Fig. 2). The high frequency of Betaproteobacteria in association with A. ostenfeldii, may be a reflection of the in situ Fig. 1. Dissimilarity analysis, nMDS, using a distance matrix with Bray Curtis dissimilarity of all strains, using R version 3.3.2, with packages Vegan (Oksanen et al., 2008) and Cluster (Maechler et al., 2016). Inherent stress: 0.09. Samples with > 162 000 read-pairs were included in the analysis (excluding samples 18 (AOKAL), 20 (AOPL), 25 (LUSI1) and 26 (LUSI6), having < 32 000 read-pairs; Supporting Information Table S2). The colours represent location of sampling: A. ostenfeldiistrains were isolated as resting cysts from four locations in the Baltic Sea Proper (AOF -Föglö Archipelago, Åland, AOKAL -Kalmar strait, Sweden, AOPL -Hel, Poland, AOVA -Valleviken, Sweden) (Tahvanainen et al., 2012); or level of anthropogenic impact (LUSI index): A. minutum/tamarensestrains were isolated from the north west Mediterranean Sea as vegetative cells, from locations with varying levels of anthropogenic impacts estimated using the LUSI index, 1-6 (E. Flo, unpubl.; Supporting Information Fig. S4); and sample ID for each strain (Supporting Information Fig. S4 and Table S2). The confidence limit was set at 0.80. brackish bacterial community composition of the Baltic Sea (Herlemann et al., 2011), which hosts more Betaproteobacteria compared to environments with higher salinity. The betaproteobacterial sequences, were primarily affiliated with Methylophilaceae, Burkholderiaceae/Limnobacter and Comamonadaceae (Supporting Information Fig. S5). Although Betaproteobacteria are not frequently found to dominate bacterial communities associated with phytoplankton, both Methylophilaceae and Limnobacter have previously been found in association with diatoms and coccolithophores (Amin et al., 2012;Green et al., 2015). Members of the family Methylophilaceae, are known to produce the growth promoting hormone indole-3-acetic acid (IAA) in association with plants (Doronina et al., 2014). Bacterial IAA has been observed to be exchanged for tryptophan in diatom cultures (Amin et al., 2015). Thus, a similar interaction may also occur between A. ostenfeldii and Methylophilaceae. Methylophilaceae prefer C1-compounds, methanol or methylamine, as a source of carbon (Doronina et al., 2002) and they are known to degrade dimethylsulfide (DMS) (Eyice et al., 2015). The consistency of the dominating groups of the microbiome suggest that associations between these groups and A. ostenfeldii may be of importance for dinoflagellate metabolism, at brackish conditions, and are not an artefact of culturing.
(16% of total sequences), Alteromonadaceae (10% of total sequences) and Flavobacteriaceae (6% of total sequences; Fig. 2). The microbiomes were highly similar to what has previously been reported from cultures of marine Alexandrium species including bacterial taxa such as Rhodobacteraceae, Alteromonadaceae/Marinobacter and Cryomophaceae (Sala et al., 2005;Jasti et al., 2005). Flavobacteriia as a group are known as degraders of complex organic carbon (Kirchman, 2002), and are commonly found in close association to phytoplankton cells in culture (Grossart et al., 2005). Alphaproteobacteria belonging to the Roseobacter group can be found both attached and free-living and exhibit a broad potential for nutrient uptake (Grossart et al., 2005;Moran et al., 2007). Gammaproteobacteria are associated with phytoplankton bloom conditions, with specialized populations responding to phytoplankton decay (Teeling et al., 2012). Taken together the results show that the microbiomes of and A. tamarense strains included in the study are representative of Alexandrium cultured at marine conditions (salinity 31).
Alexandrium core microbiome Core microbiomes were calculated for A. ostenfeldii (Supporting Information Fig. S6) and A. minutum/tamarense (Supporting Information Fig. S7) strains, respectively, according to Lawson et al. (2017). The A. ostenfeldii core microbiome, contained five OTUs present in all 20 samples with an abundance > 0.0001%, affiliated to Limnobacter and Hoeflea (Supporting Information Fig. S6 and Table S3A). Within each group (Supporting Information Fig. S3), additional ubiquitous OTUs were identified (50), among which all four groups had OTUs that were affiliated to Hydrogenophaga, Methylothenera, Flavobacteriaceae or Comamonadaceae (Supporting Information Table S4A). These OTUs, though not identical, had the same taxonomical affiliations and were therefore considered to belong to the core A. ostenfeldii microbiome, in total 55 OTUs. For the A. minutum/ tamarense isolates, 13 OTUs were present in all samples (excluding samples 25 and 26 with < 162 000 reads) (Supporting Information Fig. S7 and Table S3B). These were affiliated to Limnobacter, Pyruvatibacter, Tepidamorphus, Hoeflea, Hyphomonas, Marivita, Marinobacter, Methylophaga and Salinispirillum. Within the two groups of strains (Supporting Information Fig. S3), additional OTUs with the same taxonomical annotations were identified (39), which were affiliated to Taesokella, Fabibacter, Sphingopyxis, Spongibacter, Salinirepens and Ahrensia (Supporting Information Table S4B). Together, these 52 OTUs were considered the core microbiome of A. minutum/tamarense. The consistency of the core microbiomes indicates that associations between Alexandrium and bacteria in culture are stable and predictable. These findings enable further identification and characterization of specific Alexandrium-bacterial interactions.

Culture conditions
There are several considerations that should be made before drawing conclusions about in situ associations from in vitro conditions (Sala et al., 2005;Garcés et al., 2007;Hattenrath-Lehmann and Gobler, 2017;Behringer et al., 2018). For example, culture conditions and time in culture can have significant impacts as culturing can alter the bacterial community composition within hours (Sapp et al., 2007;Weissbach et al., 2010). To investigate the effect of long term culturing on the associated bacterial communities, A. ostenfeldii (strains 13-20) and A. minutum/ tamarense (strains 21-28) isolates were monitored after 6 and 4 months of culturing respectively. Similar to studies of diatoms in culture (Schäfer et al., 2002;Behringer et al., 2018), the bacterial communities of the Alexandrium isolates is this study were shown to be consistent over time (Supporting Information Table S1 and Fig. S8). However, in the high-resolution sequencing data, shifts in the relative abundance among some of the most frequent betaproteobacterial OTUs were observed in A. ostenfeldii strains 1-12, compared to strains 13-20 (Supporting Information Fig. S5). Strains 13-20 had previously been kept at the same conditions as 1-12, but were moved to another lab and sampled after 6-9 weeks of exposure to slightly different conditions with regards to light, salinity and temperature (Supporting Information Table S1). Taken together these results show that, given stable conditions, the members of the core microbiome of Alexandrium are consistent, but shifts in the relative abundance of the microbiome, including members of the core, can be expected as a reflection of changed environmental conditions.

Connections with in situ dinoflagellate populations in the Baltic Sea
In contrast to the Mediterranean Sea isolates, which were isolated as vegetative cells, the A. ostenfeldii isolates were isolated as cysts (Tahvanainen et al., 2012). To investigate if OTUs from the A. ostenfeldii core microbiome were present in the pelagic zone, and if their frequency could be related to natural dinoflagellate dynamics, two Baltic Sea microbial datasets were explored (Fig. 3A). The Prodiversa cruise took place during April 2013, representing a transect of the Baltic Sea Proper, capturing a mixed dinoflagellate spring bloom (Bunse et al., 2016;Godhe et al., 2016). The Planfish dataset was sampled during the productive season (April-October) of 2010-2011, covering a smaller geographic area west of the island of Gotland, capturing both spring and summer blooms of different dinoflagellate species (Legrand et al., 2015) (Fig. 3B). OTUs corresponding to members of the A. ostenfeldii core microbiome (OTU_000043 and OTU_000260), were found to follow the biomass dynamics of the dinoflagellate spring bloom, captured by the Prodiversa cruise (Fig. 3A). These OTUs were affiliated to Betaproteobacteria (Oxalobacteriaceae and Comamonadaceae; Supporting Information Table S4A). In the Planfish dataset, one of the same core OTUs (OTU_000260), increased in relative abundance during spring and summer dinoflagellate blooms (Fig. 3B). These results indicate that members of the core microbiome of A. ostenfeldii in culture followed in situ dinoflagellate occurrences, suggesting that they may be closely associated with Baltic Sea dinoflagellates, regardless of species.

Conclusions
This study, the first high resolution molecular characterization of the microbiome of A. ostenfeldii isolates, showed that neither intraspecific variations nor the location of sampling in the brackish Baltic Sea Proper had a significant impact on the isolate microbiome composition. Similarly, the level of anthropogenic impact at the different sites of isolation of A. minutum/tamarense strains could not be linked to the composition of the different microbiomes. Instead, the microbiomes were highly consistent among strains of A. ostenfeldii and A. minutum/tamarense, respectively, and core microbiomes were identified. These core microbiomes were likely shaped by abiotic conditions, like salinity, light and temperature as shifts in relative abundances could be seen upon changed culture conditions, and selected by the respective dinoflagellate host in a species-specific manner. The identification of an Alexandrium core microbiome makes this model system ideal for in-depth studies of specific interactions likely occurring between the phytoplankton and key bacterial taxaprincipal for understanding the impact of interactions on the ecosystem. In future studies, core microbiomes could be identified from natural phytoplankton blooms using culture independent methods such as single cell genomics. Fig. 3. Dynamics of OTUs with a relative abundance > 0.2% and a similarity of ≥ 95% to members from the core microbiome of A. ostenfeldii, in (A) a naturally occurring spring bloom, Prodiversa, 2013 (Bunse et al., 2016;Godhe et al., 2016) and (B) a time series covering April-October in a coastal to off-shore transect, Planfish, 2010-2011 (Legrand et al., 2015), both from the Baltic Sea Proper. Relative abundance (%) of OTUs on the left y-axis, the dinoflagellate biomass dynamics from the corresponding datasets on the right y-axis (%). The legend lists Planfish/Prodiversa-OTUs named by corresponding OUTs from the A. ostenfeldii dataset, including their taxonomic affiliation at family level. The relative abundances of dinoflagellate biomass were calculated based on the carbon content of dinoflagellates per total phytoplankton carbon, including the species: Dinobryon spp, Dinophysis sp, Gymnodiniales spp, Gymnodinium spp, Gyrodinium spp, Heterocapsa rotundata, Katodinium glaucum, Peridinella catenata, Peridinella (single cell), Protoperidinium spp and Scrippsiella. Plots were made using ggplot2 (v. 2.2.1) (Wickham, 2009) in R 3.4.0.

Supporting Information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Appendix S1: Supporting Information Table S3A -Core microbiome of Alexandrium ostenfeldii, with an OTU-abundance > 0.0001% (Lawson et al. ) . Specified by the Class and Family/Genus of the closest relative in GenBank (as given in Table S4A), the number of OTUs per Family and the relative abundance of those OTUs per group. Groups are specified in Fig. S4; AOall A. ostenfeldii strains; AOAstrains 1-4, 6, 8, 10; AOBstrains 5, 11, 12; AOCstrains 7, 14, 16; AODstrains 13, 15, 17, 19. Note that samples 18 and 20 are excluded (< 162 000 reads) and samples 9 is considered and outgroup of the A. ostenfeldii cluster (Fig. S4) and are therefore not included in any group. Table S3B -Core microbiome of Alexandrium minutum/tamarenes, with an OTU-abundance > 0.0001% (Lawson et al. 2017). Specified by the Class and Family/Genus of the closest relative in GenBank (as given in table S4B), the number of OTUs per Family and the relative abundance of those OTUs per group. Groups are specified in Figure S4; AMall A. minutum/tamarense strains (excluding samples 25, 26 (< 162 000 reads); AMAstrains 21, 23, 28; AMBstrains 22, 27. Note that strain 24 is considered and outgroup of the A. minutum/tamarense cluster in Figure S4 and therefore not included in any group. Table S4A Core microbiome (55 OTUs) of Alexandrium ostenfeldii, specifying: OTU, accession number of closest relative in GenBank, % identity, genus and family of that strain and the sequence of the OTU. Table S4B. Core microbiome (52 OTUs) of Alexandrium minutum/tamarense, specifying: OTU, accession number of closest relative in GenBank, % identity, genus and family of that strain and the sequence of the OTU.