Diversity and Biogeography of Bathyal and Abyssal Seafloor Bacteria

Abstract

The deep ocean floor covers more than 60% of the Earth's surface, and hosts diverse bacterial communities with important functions in carbon and nutrient cycles. The identification of key bacterial members remains a challenge and their patterns of distribution in seafloor sediment yet remain poorly described. Previous studies were either regionally restricted or included few deep-sea sediments, and did not specifically test biogeographic patterns across the vast oligotrophic bathyal and abyssal seafloor. Here we define the composition of this deep seafloor microbiome by describing those bacterial operational taxonomic units (OTU) that are specifically associated with deep-sea surface sediments at water depths ranging from 1000-5300 m. We show that the microbiome of the surface seafloor is distinct from the subsurface seafloor. The cosmopolitan bacterial OTU were affiliated with the clades JTB255 (class Gammaproteobacteria, order Xanthomonadales) and OM1 (Actinobacteria, order Acidimicrobiales), comprising 21% and 7% of their respective clades, and about 1% of all sequences in the study. Overall, few sequence-abundant bacterial types were globally dispersed and displayed positive range-abundance relationships. Most bacterial populations were rare and exhibited a high degree of endemism, explaining the substantial differences in community composition observed over large spatial scales. Despite the relative physicochemical uniformity of deep-sea sediments, we identified indicators of productivity regimes, especially sediment organic matter content, as factors significantly associated with changes in bacterial community structure across the globe.

Fig 1. Community composition of bacterial communities in deep-sea sediment (water depth ≥ 1000 m), at the class level (89 classes).

The large pie chart (top left) summarizes the findings based on all samples (N = 27 samples), and indicates the average relative abundances (only when ≥ 2%) of each class and the associated ranges in individual samples. Small pie charts on the map give the average community compositions in nine different oceanic regions. The numbers of samples as well as the number of sequences (n) are indicated. For comparison, the average community composition in subsurface sediments (2.5–90 mbsf, N = 5 samples, 98 classes) (http://icomm.mbl.edu, projects ICM_CFU and KCK_ODP) is displayed (top right). All sequence data were denoised and analysed using the standard operating procedure in mothur.

Fig 2. Proportions of unique and cosmopolitan OTU between oceanic regions and individual samples at the class (a, b) and OTU_0.03 (c, d) level, after averaging of 100 sequence random resampling results (n sequences = 6883, standard deviations are indicated).

Fig 3. Range-abundance relationships.

a) Log-transformed relative OTU_0.03 sequence abundance (filled orange squares) as a function of the number of samples an OTU_0.03 was detected in, and the fraction of OTU_0.03 from the total number of OTU_0.03 (filled blue circles) that fall into the different categories. b) Log-transformed relative OTU_0.03 sequence abundance (filled orange squares) as a function of the maximum distance an OTU_0.03 was detected at, and the fraction of OTU_0.03 from the total number of OTU_0.03 (filled blue circles) that fall into the different range classes. Dashed lines indicate linear models for range-abundance relationships: a) Adj. R² = 0.66, p<0.0001, b) Adj. R² = 0.30, p<0.0001.

Fig 4. Distance-decay and geographic patterns of bacterial deep-sea sediment communities.

The proportion of shared OTU_0.03 between samples significantly decreased with geographic (earth surface) distance (a) and with distance through water (b). The proportion of shared OTU_0.03 decreased with longitudinal distance (c), showed no correlation with latitudinal distance (d), and correlated with water depth (e). Dotted lines are linear model fits. Linear model’s R², Spearman’s rho correlations, and their significance (Mantel tests with 1000 permutations) are reported in each panel (n.s., not significant). The dotted line in d displays a LOESS curve to indicate the trend with latitudinal distance.