Bacterioplankton Biogeography of the Atlantic Ocean: A Case Study of the Distance-Decay Relationship

Abstract

In order to determine the influence of geographical distance, depth, and Longhurstian province on bacterial community composition and compare it with the composition of photosynthetic micro-eukaryote communities, 382 samples from a depth-resolved latitudinal transect (51°S-47°N) from the epipelagic zone of the Atlantic ocean were analyzed by Illumina amplicon sequencing. In the upper 100 m of the ocean, community similarity decreased toward the equator for 6000 km, but subsequently increased again, reaching similarity values of 40-60% for samples that were separated by ~12,000 km, resulting in a U-shaped distance-decay curve. We conclude that adaptation to local conditions can override the linear distance-decay relationship in the upper epipelagial of the Atlantic Ocean which is apparently not restrained by barriers to dispersal, since the same taxa were shared between the most distant communities. The six Longhurstian provinces covered by the transect were comprised of distinct microbial communities; ~30% of variation in community composition could be explained by province. Bacterial communities belonging to the deeper layer of the epipelagic zone (140-200 m) lacked a distance-decay relationship altogether and showed little provincialism. Interestingly, those biogeographical patterns were consistently found for bacteria from three different size fractions of the plankton with different taxonomic composition, indicating conserved underlying mechanisms. Analysis of the chloroplast 16S rRNA gene sequences revealed that phytoplankton composition was strongly correlated with both free-living and particle associated bacterial community composition (R between 0.51 and 0.62, p < 0.002). The data show that biogeographical patterns commonly found in macroecology do not hold for marine bacterioplankton, most likely because dispersal and evolution occur at drastically different rates in bacteria.

Keywords: Particle associated bacteria; bacterioplankton; biogeography; distance-decay relationship; macroecology; marine bacteria; microalgae; oceanographic province.

Figure 1

Sampling sites and oceanographic provinces: (A) Sampling sites. Color codes display the six oceanographic provinces identified in the temperature-salinity (TS) diagram (B). The six oceanographic provinces were defined on the basis of similar properties (temperature and salinity) of the first 200 m of the water column.

Figure 2

Similarities between bacterial communities. Bray-Curtis similarity was calculated for standardized abundance data at the OTU level. Panels (A,B) show similarities between the three size fractions of the bacterioplankton FL (free-living), SPA (small particle associated) and LPA (large particle associated). Panels (C,D,E) show similarities for each of the three communities separately. Bacterial abundances were standardized by the total for (A,B,C,D,E) and root transformed for (A,B). Color codes discriminate the size fractions of the communities for (A,B), while they distinguish provinces for (C,D,E). Depth layers are shown by different symbols.

Figure 3

Distance-decay relationship for free living (FL) communities along the water column. Bray-Curtis similarity was calculated on standardized abundance data, and plotted against the geographical distance expressed in Km. The transect was analyzed entirely and for the Southern and Northern hemisphere separately, which are displayed from left to right. Samples were divided into five depth layers: (A) 20 m, 40 m (B), 50–80 m (C), 85–120 m (D), and 140–200 m (E). Adjusted R² value are displayed for each of the linear regressions and second order polynoms on the chart area, as well as the significance level (^**p < 0.01 and ns p > 0.05). For both hemispheres the marginal provinces FKLD and NADR are marked in red. For the Southern hemisphere, due to high variation in community similarity driven by the FKLD province, this province was left out of the calculation for the linear regression.

Figure 4

Co-occurrence analysis of OTUs across provinces and along depth. Sample size was resampled to obtain an equal number of reads in each sample. Bray-Curtis similarity was calculated and used for the clustering of samples. For a better visualization of the relative abundances the data were scaled by row with values ranging from −2 to 10. The color reflects province and depth layer to visualize clustering of samples. The analysis was performed separately for each of the three size fractions of the plankton FL (A), SPA (B), and LPA (C). On the upper part of the heatmaps sample clustering is reported, while on the left side of the heatmaps the OTU clustering is shown. Every horizontal line therefore indicates an OTU, while every vertical line represents a sample.

Figure 5

Province similarity along the water column. Bray-Curtis similarity was calculated on standardized abundances data. The data were combined into three depth layers: 20–80, 85–120, and 140–200 m for the four intermediate provinces: BRAZ, SATL, WTRA, and NAG. The two marginal provinces FKLD and NADR are showed separately (Figure S4) because of the small number of samples for the 140–200 m depth layer. The average of the Bray-Curtis similarity was calculated for each depth layer of the provinces, and compared against all six provinces. From left to right: BRAZ, SATL, WTRA, and NAG are displayed while from top to bottom the three size fractions of the plankton are shown. The color key shows different depth layers.

Figure 6

Biogeography of photosynthetic micro-eukaryotes. Biogeographical patterns were investigated for the small subunit of RNA sequences from chloroplasts. Bray-Curtis similarity was calculated for standardized abundance data at the OTU level for the SP (small particles) community (left panel, A) and the LP (large particles) community (right panel, B). Color code discriminate the six oceanographic provinces, while depth layers are shown with different symbols. In panels (C,D) the distance-decay relationship for SP and LP communities is displayed. For those two panels the color key indicates the distance between samples (pairwise) express of oceanographic provinces.

Figure 7

Beta diversity patterns are shaped by environmental parameters and phytoplankton communities. Principal coordinate analysis was used to show similarities among samples: (A) (FL), (B) (SPA), and (C) (LPA). In order to assess the effect of environmental parameters on the structure of the plot, Pearson correlation was calculated with the principal coordinates, after Bray-Curtis similarity was calculated on standardized abundance data at the OTU level. Only correlation higher than 0.4, with the first two axes were displayed on the chart area. The total amount of samples was reduced to match the biological data (bacterial abundances) with the environmental data (temperature, salinity, depth, and chlorophyll a) and with the relative abundance of phytoplankton (SP and LP). Color codes discriminate oceanographic provinces, while depth layers are shown with different symbols.