Recurring patterns in bacterioplankton dynamics during coastal spring algae blooms

Correlation values were considered only when p-value <0.05. Spearman rank correlations measure the strength of association between two ranked variables. A precondition (assumption) for this nonparametric statistical analysis is that the two variables share a monotonic relationship (i.e. when one variable increases, the other either consistently increases as well or consistently decreases). This monotonic relationship does, however, not need to be linear. While no particularly strong correlations were found between major algae groups and bacterioplankton clades, some noteworthy trends were detected. For example, Ulvibacter abundances were positively correlated with diatoms (in particular centrales) and negatively correlated with silicoflagellates, whereas an opposite trend was observed for the VIS1 clade of the NS5 marine group (i.e. negative correlation with diatoms and a positive correlation with silicoflagellates). While the explanatory power of such pairwise rank-based correlations is limited and correlation does not necessarily imply causation, these data suggest at least for Ulvibacter sp. (and to a lesser extent for Formosa clade B) that they are associated with diatoms, while the NS3a, NS5 and NS9 marine groups seemed to be rather positively correlated with flagellates such as dino- and silicoflagellates. The positive correlation between Ulvibacter sp. and diatoms was supported by Spearman rank correlation analysis between the abundances of prominent bacterioplankton clades and distinct phytoplankton groups. Ulvibacter abundances were positively correlated with the diatoms Mediopyxis helysia, Chaetoceros debilis, Chaetoceros minimus, Thalassiosira nordenskioeldii and the Phaeocystis spp. haptophytes. Correlations of other bacterioplankton groups with distinct phytoplankton groups were less pronounced. Noteworthy positive correlations were found between Chattonella spp. and Bacteroidetes abundances (probe CF319a), in particular Marinoscillum spp. (probe CYT-734), Polaribacter spp. (probe POL740) and the NS5/VIS1 and NS3a marine groups. The NS5/VIS1 marine group was negatively correlated with most diatom groups and the Phaeocystis sp. haptophytes, and only positively correlated with Chattonella species. This might indicate a preference for Chattonella spp. or simply reflect that the members of the VIS1 clade of the NS5 marine group were less competitive than other Flavobacteriia under the conditions where other algae dominated. Spearman rank correlations between major bacterioplankton clades and physicochemical parameters were most conclusive for Ulvibacter species. Ulvibacter abundances were positively correlated with chlorophyll a, consistent with the previously detected positive correlation with in particular diatoms. Ulvibacter abundances were also negatively correlated with silicate, phosphate, nitrate and ammonium. In particular the strong negative correlation with silicate, which is the limiting factor for diatom frustule formation, supports a particularly strong association of Ulvibacter spp. and diatoms. Similar but weaker trends were observed also for other bacterioplankton clades, such as Polaribacter, Formosa clade B, or VIS6 clade Cryomorphaceae. Of all genus-level clades, Polaribacter and Formosa clade A showed the strongest associations with temperature.

For all tested clades of Flavobacteriia, multiple abiotic factors and multiple algae groups were obtained as explanatory variables. The strongest abiotic predictors were temperature, salinity, silicate and nitrate. The strongest biotic predictors were Phaeocystis spp. haptophytes, Rhizosolenia spp., Chaetoceros debilis, and Chaetoceros minimus diatoms and the silicoflagellate Chattonella. It should be noted though that these regressions were computed based on log-transformed abundance data and not algae biovolumes (which were not measured). Insofar the in influence of the rather small cell-sized algae such as Chaetoceros miniumus is likely overestimated. Such limitations notwithstanding it is noteworthy that in no case a simple one-to-one relationship between specific algae and specific bacterioplankton groups was obtained so far.

Two additional metagenomes are listed in italics for completeness, one from a test run sampled on August 21^th, 2008, and a metagenome of particle-attached (3 - 10 µm) bacterioplankton sampled on April 14^th, 2009.

Taxonomic bins with too small sizes for a sound analysis were excluded (red text).

Searches were performed against the CAZy database, the dbCAN database and the Pfam database using E-value thresholds that were adjusted for each family by extensive manual annotations. CAZymes were only annotated when at least two of the three database searches yielded positive results (GH = glycoside hydrolase; CBM = carbohydrate- binding module; CE = carbohydrate esterase; PL = polysaccharide lyase; GT = glycoside transferase; AA = auxiliary activities).