Diversity and Biogeography of Bathyal and Abyssal Seafloor Bacteria.

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.

Introduction

The deep ocean floor comprising the lower continental margin and abyssal plains at >1000 m water depth covers about half of Earth’s surface. Deep-sea surface sediments of the top 2 cm consist mostly of clay minerals, shells of planktonic organisms and organic matter; the benthic communities inhabiting the deep-sea floor are dominated by bacteria in terms of total organism abundance and biomass [1, 2], as well as in carbon and nutrient recycling and oxygen fluxes [3, 4]. Hence, the characterization of the composition and structure of bacterial communities, as well as their patterns of distribution, can provide important insights into the ecological and biogeochemical functioning of this vast ecosystem [5, 6]. In particular, the identification of relevant bacterial groups and of their functions in deep-sea sediments are key to understanding matter fluxes in deep-sea ecosystems and feedback mechanisms to environmental change and impacts [5, 7–9].

In recent years, the concept of microbiomes has emerged, whereby the collective entities of microorganisms and their genes typical for a specific host, habitat or ecosystem are identified. In this regard, a core microbiome represents those genomes or genetic markers common to the majority of samples considered. These are generally abundant in a given sample and have been hypothesized to mark important genetic functions and members of the microbial community ([10] and references therein). Such core microbiomes have been identified for different parts of the human body [11], plants [12], terrestrial systems [13], and marine ecosystems such as the pelagic realm [14] and methane seeps [15]. Using 454 tag sequencing, globally distributed marine sediment samples were also found to consist of characteristic bacterial classes, which were distinct from pelagic communities [6].

The surface deep-sea floor represents a rather uniform, specific environment, characterized by low temperatures (-1° to 4°C), high pressures (several hundred bars) as well as the absence of light and hence photosynthesis. Other key characteristics are a generally low supply of organic matter (1–10 mmol C m-2 yr-1), and fine-grained oxygenated sediments (250–300 μM oxygen) forming a dense sediment matrix of low permeability [3]. It is therefore likely that environmental selection has led to the establishment of a core deep-sea sediment microbiome that is distinct from those of other deep-sea environments. In support of this hypothesis, substantial differences in bacterial and archaeal communities of subsurface sediment, seep and vent ecosystems have been detected [15, 16]. But little is known about the community similarity of these seafloor systems to that of typical deep-sea sediments and of the spatial turnover of bacterial communities in deep-sea sediments. It may be assumed that the dimensions of the deep-sea realm are too large to support a global dispersal of sediment microorganisms, especially given the sluggish deep ocean currents.

Animal communities in the deep sea have been studied for a much longer time than microbial communities, and previous studies have shown a high degree of endemism for deep-sea animals, with most species only recorded as one or two individuals from one or two sampling sites [17]. Also for microbial eukaryotes, most taxa seem to be regionally restricted, with only few maintaining cosmopolitan distributions, and indicating positive range-abundance relationships (i.e. the size of the species geographic range increases with species abundance) [18]. Positive range-abundance relationships for bacterial populations have been observed in a variety of other microbial realms, including soil and the pelagic realm [19-24] and have originally been described for macroorganisms [25, 26]. But for deep-sea bacterial communities, biogeographic patterns, such as endemism vs. cosmopolitanism or species range-abundance relationships, remain largely unknown. In a first analysis at the global scale, indications of high levels of provincialism were found for bacterial communities in marine sediments [6], suggesting a limited dispersal of marine benthic bacterial communities in the deep sea. Also previous regional studies of deep-sea sediment bacterial communities (e.g. [27-32]) found a high degree of endemism and a high turnover of bacterial communities on the scale of meters to kilometres. However, all of these studies were regionally restricted or only marginally touched upon deep-sea sediments, and did not specifically test biogeographic patterns across the vast oligotrophic bathyal and abyssal seafloor.

Here, we used a dataset of 27 deep-sea surface sediment samples from all major ocean regions (top 1–2 centimetres of sediment), which focused on the zone between 1000–5300 m water depth, representative of 70% of the depth distribution of the global deep-sea floor [33]. We investigated four ecological rules, namely that I) The deep-sea surface sediment microbiome is distinct from subsurface communities. II) It is composed of a few sequence-abundant types, which form a core microbiome, and of many rare, endemic types. III) There is a positive relationship between the sequence-abundance of a taxon and the size of its geographic range (i.e. positive range-abundance relationship). IV) Deep-sea sediment bacterial communities exhibit a distance-decay relationship, i.e. community similarity decreases with increasing geographic distance due to isolation by distance and lack of population mixing.

Material and Methods

Datasets and 454 massively parallel tag sequencing (454 MPTS)

Deep-sea sediment samples analysed here mainly originated from our own sample repository (n = 20; S1 Table), but also included data from the ICoMM initiative (n = 7) [34], i.e. from the South Pacific (NZS) and the North Pacific (SMS). The 27 samples were obtained from water depths between 1025 m and 5347 m by winch-operated coring with a multiple corer, and consisted of 0.5–1 g of the top layer (0–2 cm) of deep-sea sediments composed of fine clays and biogenic particles. Sediment subsamples were removed directly after retrieval with a clean spatula and stored into cryovials at -20°C until DNA extraction. A list of all samples used in this study, their corresponding project names and geographic locations can be found in the Supporting Information (S1 Table). Permits for coring seafloor sediments within the Exclusive Economic Zones of coastal states were acquired before each seagoing expedition (S1 Table) from the legal authorities where necessary. The locations sampled are not privately-owned or protected in any way and the field studies did not involve endangered or protected species.

In all cases, sequencing data of the V6 region of the bacterial 16S rRNA gene were obtained according to the standardized sequencing pipeline of the ICoMM project (see S2 Table for the primer cocktail used; https://vamps.mbl.edu/index.php) [35, 36]. Fragments were sequenced by pyrosequencing on a Genome Sequencer FLX system (Roche, Basel, Switzerland) at the Marine Biological Laboratory in Woods Hole, MA, USA. Standard flowgram format files (sff) have been deposited in the GenBank Sequence Read Archives (www.ncbi.nlm.nih.gov/sra) and their accession numbers are provided in the Supporting Information (S1 Table). Flowgrams were processed and converted into an OTU-by-sample table with mothur (Version 1.29.2) [37] according to the standard operating procedure [38], including the implemented denoising algorithm [39]. Sequences were clustered into operational taxonomic units at a 3% nucleotide difference (hereafter referred to as OTU0.03). Alignment of sequences and taxonomic classifications were carried out using the SILVA reference database (release 119) [40] and the mothur standard operating procedure.

The global deep-sea surface sediment dataset comprised in total 501,480 sequences, corresponding to 88,247 OTU0.03. Absolute singletons (SSOabs), i.e. OTU0.03 consisting of a sequence occurring only once in the full dataset [41], accounted for 63% of all OTU0.03 (11% of all sequences). A reduced dataset with absolute singletons excluded hence comprised 455,822 sequences and 32,589 OTU0.03. We also defined a group of relative singletons (SSOrel) as those OTU0.03, that were found in several samples of this global study, but occurred at least once as singleton, i.e. with one sequence in at least one sample [41]. This group accounted additionally for 19% of all OTU0.03 (3% of all sequences). The relevance of such rare types in bacterial communities is not well resolved in general [42-44], and may indicate either that bacterial diversity is still under-sampled, or that the observed diversity is the result of technical artefacts (i.e. PCR or sequencing errors). With the systematic noise-removal that we applied here to all datasets using the mothur-implemented denoising algorithm, technical (i.e. sequencing) errors are likely to be greatly reduced in our study. Earlier studies have also suggested that the removal of singletons or rare types neither affect the overall patterns of bacterial communities nor their ecological interpretation [6, 45, 46]. However, to account for any uncertainties related to the presence of SSOabs, we report results based on a reduced dataset with SSOabs excluded, unless indicated otherwise.

Potential contaminants in sequencing data

Betaproteobacteria, especially affiliated to Burkholderiales and Ralstonia, have been reported from deep-sea sediments and also for the terrestrial deep subsurface ([16] and references therein). However, there are also indications that these sequences may originate from contamination of clean laboratory water or reagents [47-50]. In our current dataset, OTU0.03 affiliated with Burkholderiales accounted for 5% of Betaproteobacteria OTU0.03 (36% of Betaproteobacteria sequences). None of these OTU0.03 occurred in all 27 samples (e.g. only 3 OTU0.03 occurred in 21 and 20 samples), while contaminant sequences would have been expected to occur in all samples. A contamination with sequences from laboratory reagents would need to be tested individually for each protocol, sample analysis pipeline and reagents used, and would need to be conducted in parallel with sample handling. We therefore could not categorize specific sequences as contaminants based on sequence similarities to previously found contaminants, and refrained from excluding sequences a priori from this dataset.

Statistical analyses

Observed richness (i.e. number of OTU0.03 per sample) and richness estimates (Chao1) were calculated with 100 sequence re-samplings per sample based on the smallest dataset (n = 7,922 and 6,883 sequences with and without SSOabs respectively), to account for differences in sequencing depth between samples. Overall differences in bacterial community composition were visualized with non-metric multidimensional scaling plots. A corresponding analysis of similarity was used to assess significant differences between samples grouped by oceanic regions. Shared or endemic OTU0.03 (i.e. found in only one sample or oceanic region) were also calculated with 100 sequence re-samplings per sample based on the smallest dataset. Geographic distances between stations were calculated in two different ways: i) as the surface distance between samples (function geodist in R package ‘gmt’), and ii) as the shortest path by sea between samples, only allowing connecting routes through water (function lc.dist in R package ‘marmap’). To test whether community similarity was significantly correlated with different spatial components, non-parametric Mantel tests [51, 52] based on the Spearman correlation coefficient were applied and significance assessed based on 1000 Monte Carlo permutations.

We further tested how biogeochemical provinces defined by Longhurst et al. [53] (http://www.vliz.be/vmdcdata/vlimar/downloads.php), and oceanographic regions based on total organic carbon measurements [54], as proxies for surface productivity and sediment total organic carbon content, accounted for changes in bacterial community similarity. A partitioning of the variation in bacterial community composition between spatial distance, water depth, surface productivity and total organic carbon was conducted according to Legendre [55]. All statistical analyses were performed in R (v. 3.1.1) (R Development Core Team 2009, http://www.R-project.org) using packages vegan [56], gplots [57], gmt [58], marmap [59], gdistance [60] and with custom R scripts that are provided in the supplementary information (S1 Script).

Results & Discussion

Recent studies found deep-sea sediments to be extremely rich and diverse in bacterial types despite the low quantities of organic matter available, rivalling even the diversity of much more organic rich soils on land [61]. The deep-sea floor samples analysed here exhibited an average Chao1 richness estimate of 4,599 ± 1572, with absolute singletons included (for Chao1 richness estimates with and without SSOabs see S1 Fig). However, since the Chao1 estimator corrects the observed richness by adding a term based on the number of singletons and doubletons, we also report here the observed richness. Observed richness was on average 2,623 ± 554, ranging from 2,245 ± 267 (average for South Atlantic) to 3,499 ± 416 (average for South Pacific) (S2 Fig). These estimates are within the range of, or higher than values reported from regional studies of deep-sea sediments [31, 32]. They are substantially higher than estimates from the ocean pelagic realm [14, 62], seeps and vents [15]. Accumulation curves (S3 Fig) indicated that at coarse taxonomic resolution (i.e. phylum to family), the diversity of most taxa was captured with this global sample set, but a considerable part of the global bacterial diversity in deep-sea sediments at the OTU0.03 level was still missing.

The microbiome of deep-sea surface sediments

Marine sediments characteristically show a dominance of Proteobacteria [6, 34]. Also in this global study of deep-sea sediments, half of the sequences belonged to the phylum Proteobacteria (50%), most of which were affiliated with the classes Gammaproteobacteria (20 ± 5%), Alphaproteobacteria (12 ± 4%), and Deltaproteobacteria (10 ± 4%) (Fig 1). The phylum Actinobacteria was second in sequence abundance (13 ± 6%). Gammaproteobacteria sequences were the most abundant at the majority of sampling sites. The dominance of these taxonomic groups is in agreement with both global [6] and regional studies, e.g. in the Eastern Mediterranean [63], Arctic [28, 29], East Pacific [27], South Pacific [64], and South Atlantic Ocean [30], the latter being based on 16S rRNA gene clone libraries. The deep-sea surface sediment community differs from surface and deep-water communities, which are usually dominated by Alphaproteobacteria, Gammaproteobacteria, Cyanobacteria and Flavobacteria (e.g. [6, 14, 65]).

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.

We further compared the average composition of deep-sea surface sediment communities to an average community from subsurface sediment (5 samples between 2.5 and 90 m below seafloor from the Peru Margin), which were analysed with the same sequencing and bioinformatics methods (454 pyrosequencing of the V6 region; see http://icomm.mbl.edu) (Fig 1, S2 Table). The deep-sea surface sediment community differed clearly from subsurface sediments in the dominance of Gammaproteobacteria, and in the rarity of Chloroflexi (vadinBA26), Bacilli and candidate division OP9/JS1 that are typical of subsurface environments [16, 66–69]. This demonstrates that pelagic, benthic surface and deep-subsurface environments exhibit distinct bacterial community signatures already at broad taxonomic resolution levels. At a finer taxonomic resolution, even less overlap was detected, with none of the twenty most abundant OTU0.03 from each environment shared between the two realms (S3 Table), probably reflecting differences in life styles and environmental pressures between these habitats, even though this comparison is yet based on a limited number of samples.

The deep-sea sediment core microbiome, here defined as OTU0.03 that occurred in more than 90% of the deep-sea surface sediment samples (i.e. in ≥ 25 of the 27 samples) and in all oceanic regions, consisted of only 18 OTU0.03 (0.1% of all OTU0.03), comprising 6.2% of all sequences (Table 1). They included many taxa comprising heterotrophic polymer degraders, as is expected for deep-sea sediment communities, where the main source of energy and nutrients is marine detritus [3]. Only three highly abundant OTU0.03 were truly cosmopolitan (i.e. found in all 27 samples). Among these, two OTU0.03 were affiliated with the JTB255 clade (order Xanthomonadales), and one with the OM1 clade (order Acidimicrobiales) (Table 1, S3 Table); these OTU0.03 made up 21% of the JTB255 and 7% of the OM1 clade. Nevertheless, the truly cosmopolitan OTU0.03 of the JTB255 and the OM1 clade also showed variations in relative sequence abundance (in relation to the total number of sequences per sample) between oceanic regions. They ranged from 0.1% in the Eastern Mediterranean to 3.6% in the Antarctic for JTB255, and from 0.1% in the Eastern Mediterranean to 1.4% in the Arctic Ocean for OM1 (S4 Fig). Thus, both types appear to be more sequence-abundant in polar, cold regions (e.g. Antarctic, Arctic), and less abundant in warmer regions (e.g. Mediterranean). Overall, sequences of the JTB255 clade have been reported in a range of local and regional marine benthic studies (e.g. [28, 30, 70–72]), but the function of this group remains unknown, as no relative has yet been cultivated. Previously, members of the OM1 clade have been predominantly described from seawater [73, 74], and further investigations are needed to address their functional relevance in deep-sea sediments.

Table 1

Most common OTU0.03 occurring in 90% of the samples (i.e. in ≥25 out of 27 samples), their taxonomic affiliation, and their sequence abundance in this global set of deep-sea surface sediment samples.

Class	Order	Family	Genus	Number of samples in which present	Absolute sequence abundance in dataset	% sequence abundance in whole dataset	% sequence abundance across oceanic regions
							Average	Min.	Max.
Gammaproteobacteria	Xanthomonadales	JTB255_marine_benthic_group	unclassified	27	4513	1.01	1.16	0.07	2.76
Acidimicrobiia	Acidimicrobiales	OM1_clade	unclassified	27	2399	0.54	0.58	0.05	1.19
Gammaproteobacteria	Xanthomonadales	JTB255_marine_benthic_group	unclassified	27	1963	0.44	0.36	0.08	0.90
Acidimicrobiia	Acidimicrobiales	OM1_clade	unclassified	26	4226	0.95	1.00	0.06	2.66
Gammaproteobacteria	Xanthomonadales	JTB255_marine_benthic_group	unclassified	26	3611	0.81	0.68	0.12	1.94
Gammaproteobacteria	Xanthomonadales	JTB255_marine_benthic_group	unclassified	26	2316	0.52	0.60	0.06	1.48
Gemmatimonadetes	BD2-11_terrestrial_group	unclassified	unclassified	26	1084	0.24	0.24	0.07	0.47
Gemmatimonadetes	BD2-11_terrestrial_group	unclassified	unclassified	26	594	0.13	0.14	0.06	0.23
JTB23	unclassified	unclassified	unclassified	26	903	0.20	0.24	0.05	0.58
SPOTSOCT00m83	unclassified	unclassified	unclassified	26	624	0.14	0.11	0.02	0.26
Alphaproteobacteria	4-Org1-14	unclassified	unclassified	26	270	0.06	0.08	0.02	0.19
Gammaproteobacteria	Gammaproteobacteria_Incertae_Sedis	Unknown_Family	Methylonatrum	25	973	0.22	0.22	0.03	0.40
Flavobacteriia	Flavobacteriales	Flavobacteriaceae	Aestuariibaculum	25	1596	0.36	0.60	0.07	3.28
Gammaproteobacteria	BD7-8_marine_group	unclassified	unclassified	25	719	0.16	0.19	0.01	0.43
Gammaproteobacteria	Gammaproteobacteria_Incertae_Sedis	Unknown_Family	Methylonatrum	25	520	0.12	0.11	0.01	0.19
Gammaproteobacteria	Order_Incertae_Sedis	Family_Incertae_Sedis	Marinicella	25	761	0.17	0.17	0.02	0.43
Gammaproteobacteria	Methylococcales	pItb-vmat-59	unclassified	25	404	0.09	0.12	0.03	0.26
Gammaproteobacteria	Methylococcales	pItb-vmat-59	unclassified	25	248	0.06	0.07	0.02	0.16

Our study supports the hypothesis of a distinct core microbiome in global deep-sea sediments with yet unknown adaptations, as no close relative has been cultivated yet or has had its genomic composition established. This core bacterial community may consist of generalists highly adapted to life in the deep sea, e.g. with a high flexibility in the use of resources, similar to what has been suggested for the few cosmopolitan types in soil microbial communities [75]. In addition, we tested whether the diversification of this core microbiome accounts for a substantial fraction of the observed diversity of surface deep-sea sediments, as shown for the microbiome of cold seeps [15]. In cold seep communities, endemic taxa closely related to the members of the core microbiome make up a substantial proportion of total richness. This trend could not be confirmed for deep-sea sediments, as those families that contained the 18 most abundant OTU0.03 contributed only a small proportion of the endemic types. Our data suggest that a substantial fraction of the global diversity of bacteria in deep-sea sediments is endemic.

Deep-sea sediment bacteria endemism, cosmopolitanism and positive range-abundance relationship

Since deep-sea sediments can be considered as a relatively stable and uniform environment, forming a matrix of fine particles that immobilizes their bacterial inhabitants, dispersal of benthic bacteria in the deep sea is probably limited. Microbial dispersal may, however, occur via the resuspension of sediments by water currents and faunal activity (i.e. bioturbation or by bentho-pelagic organisms feeding on sediments and migrating (e.g. [76, 77]). Yet, deep-water currents above the seafloor are usually weak, hence long-distance passive transport of deep-sea sediments probably occurs rarely [78, 79].

Comparing community composition at the broad taxonomic levels of phylum to class, a rather uniform distribution of the sediment microbiota was detected across all oceans (Fig 2a–2b, S5 Fig). Differences in community composition appeared at the family and higher taxonomic resolution levels (data not shown), as reported from other global microbiome studies of permafrost soils [80], cryosphere habitats [81], or other environments (see also [46]). Previous studies of bacterial OTU-distribution have shown that average spatial ranges of OTU can change with environment, latitude and sequence abundance [23, 82]. Here, a high degree of endemism was detected; higher than in water column environments [82]. At the resolution of OTU0.03, up to 70% of all bacterial taxa were unique to one sample (Fig 2c and 2d), and the proportion of pairwise shared OTU0.03 between oceanic regions was 11% on average (ranging from 3–19%; S6 Fig). Absolute singletons (SSOabs) logically increased the levels of endemism at the OTU0.03 level (to ca. 80%, S7 Fig). This observation supports previous findings on global bacterial distribution in other ocean realms [23, 82]. In comparison, the Census of the Diversity of Abyssal Marine Life also reported that the majority of deep-sea animals occurred at only one or two sampling sites, at a similar proportion as bacterial taxa in this study [17]. However, it is not known whether this recurring observation points to severe undersampling of the deep sea, or indeed to rarity and limited dispersal [17, 83].

Fig 2

Proportions of unique and cosmopolitan OTU between oceanic regions and individual samples at the class (a, b) and OTU0.03 (c, d) level, after averaging of 100 sequence random resampling results (n sequences = 6883, standard deviations are indicated).

We looked at another group of rare OTU0.03, i.e. those that occurred in three samples or less (≤10% of all samples). These OTU0.03 were of low sequence-abundance (<0.01% of all sequences) and accounted for 80% of all OTU0.03, therefore they displayed the typical long tail of rare types observed in microbial rank-abundance curves [36, 84] (Fig 3a). The majority of these OTU0.03 had an average geographic range of 3,781 km (ranging from 0–18,700 km), suggesting that most low abundance OTU0.03 are limited in their range to within ocean basins (Fig 3b). Such a high turnover of OTU0.03 between samples has also been evidenced in other studies at smaller spatial scales (tens to hundreds of kilometres) in Arctic deep-sea sediments [31], for marine bacterioplankton communities [21], or when considering different habitat types [20]. In contrast, sequence-abundant OTU0.03 (defined as OTU0.03 comprising >0.1% of all sequences) were found across average distances of 18,029 km (ranging from 8,000–18,700 km). Consequently, the more sequence-abundant an OTU is, the more likely it is to be found in samples located much further away. These results indicate that, despite our assumption of very slow rates of dispersal in the deep-sea environment, it is still possible to observe truly ubiquitous, cosmopolitan taxa, at the level of their 16S (V6) gene signature. Future studies should direct effort to the question of their identities, traits and functions, to better understand the evolution of core microbiota in the deep-sea realm and on Earth in general.

Fig 3

Range-abundance relationships.

a) Log-transformed relative OTU0.03 sequence abundance (filled orange squares) as a function of the number of samples an OTU0.03 was detected in, and the fraction of OTU0.03 from the total number of OTU0.03 (filled blue circles) that fall into the different categories. b) Log-transformed relative OTU0.03 sequence abundance (filled orange squares) as a function of the maximum distance an OTU0.03 was detected at, and the fraction of OTU0.03 from the total number of OTU0.03 (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.

The analysis of how relative sequence abundance changes with geographic range (either defined as the number of common samples or the maximum distance an OTU0.03 is observed at) supported the presence of a positive range-abundance relationship (Fig 3). Comparable patterns have been reported for microbial eukaryotes in the deep sea, where the majority of taxa were regionally restricted, and only a small percentage maintained cosmopolitan distributions [18]. Positive range-abundance relationships for bacterial types have also been observed in a variety of other microbial realms, including soil and the pelagic realm [19-24].

While methods based on sequencing of 16S rRNA genes do not fully reflect the true abundance of organisms in the environment [85], a plausible ecological explanation for the observed positive range-abundance relationship would be that higher local population sizes—as approximated by high sequence abundances here—enable a larger organismal pool to be further passively dispersed, higher colonization and lower extinction rates (mass-effect of metapopulation dynamics as described in e.g. [86]). In addition other mechanisms may generate positive range-abundance relationships, such as resource breadth and availability, also proposed previously [86, 87]. As aforementioned, undersampling is most likely one reason for the pattern observed here, as we still miss—despite the use of high-throughput sequencing—very low abundant types in some samples (S3 Fig) and therefore underestimate their distribution ranges.

Distance-decay and predictors for variation in bacterial surface sediment communities

Significant distance-decay relationships for bacterial communities have been reported in a global study of pelagic and seafloor environments [82], in soil [88-90], woodland [91], and saltmarsh sediments [92], suggesting this relationship to hold true across different ecosystems. Here we focused on deep-sea sediments at >1000 m water depth. We found that community similarity (based on the proportion of shared OTU0.03) between samples decreased significantly with increasing geographic distance (Fig 4a), with a slope coefficient for the distance-decay relationship |ß| in our dataset of 0.066 (calculated accordingly to [82]).

Fig 4

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

The proportion of shared OTU0.03 between samples significantly decreased with geographic (earth surface) distance (a) and with distance through water (b). The proportion of shared OTU0.03 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.

We also tested whether the distance-decay relationship holds true when considering water paths around continents instead of straight distances (earth surface) between sampling locations (Fig 4b). The explained variation and slope of the relationship (|ß| = 0.088) were higher than when considering direct connecting lines, reinforcing the idea of spatial isolation in deep-sea bacterial communities, when using an appropriate distance metric. A connectivity of microbial populations via deep-water currents has been suggested for sediment and deep-water communities, and for benthic thermophilic endospores [93-95]. This dispersal mechanism would be similar to what has been proposed for larval stages of benthic deep-sea fauna [1]. In the future, more advanced sampling schemes and models [96] should be applied to test for the effect of deep-water transport (speed, direction) on bacterial deep-sea communities.

The distance-decay relationship observed for bacterial communities may arise from multiple mechanisms, involving environmental filtering, neutral processes, and isolation by distance, which is a complex product of limited dispersal, ecological drift, and speciation processes [97]. To shed more light on the mechanisms generating the observed distance-decay relationship, we also considered changes in community similarity as a function of differences in water depth, latitudinal distance, and longitudinal distance (Fig 4c–4e). Increasing water depth is a general indicator of decreasing particle flux as key energy source for deep-sea bacteria [3]. A relatively weak, but significant relationship was observed for community changes along water depth (Fig 4). This confirmed that even below 1000 m, bacterial communities are structured by changes in biological or physical parameters that are correlated with water depth, especially the dynamics in particulate organic matter flux that represent the main source of energy and carbon [31, 98]. The range of investigated water depths itself did not explain a significant fraction of community variation when other variables, such as geographic distance or organic carbon content, were considered (S8 Fig).

Latitudinal distance correlates with climatic regions of the surface ocean, and previous studies have reported correlations between bacterial community richness and latitude for communities from the pelagic [21, 99], and from terrestrial realms [100, 101] (for controversial findings see [93, 102]). But, according to the physical stability of the deep sea, latitudinal distances were neither a good predictor of community similarity (Fig 4d), nor of richness in deep-sea surface sediment communities (S9 Fig). However, a trend analysis based on LOESS curve fitting (Fig 4d) indicated that community similarity increased towards both polar regions, as detected already for epipelagic marine bacteria [24]. This may be related to latitudinal changes in ocean productivity and particle flux, which increase both southwards and northwards from the equator to about 70° latitude in both hemispheres [53]. Distances along latitude and longitude were also correlated with water depth (Spearman’s rho = 0.2 and 0.28, p = 0.01 and 0.006 for latitude and longitude, respectively.) But interestingly, changes in community similarity with geographic distance appeared to be mainly due to changes with longitude (Fig 4c). On the one hand, geographic features like mid-ocean ridges, and deep-water currents [83], but also land masses, may present barriers to dispersal along longitudinal axes. However, this pattern may also result from changes in productivity regimes with proximity to the productive ocean margins.

Effects of spatial and environmental parameters on seafloor bacterial community composition

We further tested how other environmental parameters may account for changes in bacterial community composition (based on relative sequence abundances) with geographic distance. For example, the role of surface productivity, particle flux, and of other biological factors in the structuring of benthic communities have previously been suggested [6, 98]. Productivity indices based on biogeochemical provinces defined by Longhurst et al. [53] (http://www.vliz.be/vmdcdata/vlimar/downloads.php) and oceanographic regions based on total organic carbon measurements [54] were used as proxies to estimate environmental differences at the global scale. A partitioning of the biological variation between geographic distance, water depth, surface productivity and total organic carbon content confirmed a significant effect of geographic distance (4% of variation explained, p = 0.04) on bacterial community structure, even when taking other environmental parameters into account (S8 Fig). However, differences in total organic carbon at the seafloor also played a significant role in shaping bacterial community structure (10% of variation explained, p = 0.003), and in addition there was a noticeable level of co-variation between geographic distance and total organic carbon categories (3%). Also surface productivity explained some of seafloor bacterial community variation, but this was not statistically significant (3% of variation explained, p = 0.058). Discrepancies between surface productivity and total organic carbon availability at the seafloor may be explained by biological processes or hydrographic features altering vertical particle flux, or by a lateral input of organic material. The effect of organic matter availability on benthic communities is in agreement with general trends reported for different benthic size classes in the deep sea [1, 98, 103–105]. The effects of these factors on bacterial community structure and distribution will need to be further explored for the deep seafloor at the global scale. Future studies should aim at integrating different spatial scales and at measuring a large range of environmental parameters, e.g. total organic carbon, particle flux, nutrients, chlorophyll pigments, as well as biological factors such as the presence of fauna, to provide potential descriptors of microbial community patterns in the deep sea.

Conclusion

By investigating the composition and distribution of benthic deep-sea bacterial communities at the global scale, we show that bacterial communities of deep-sea surface sediments are distinct from those of the pelagic or the subsurface seafloor biosphere, and this already at the class level. Deep-sea sediments are inhabited by a core community of few cosmopolitan, sequence-abundant bacterial OTU which are affiliated with the JTB255 marine benthic group (class Gammaproteobacteria, order Xanthomonadales), and the OM1 clade (class Actinobacteria, order Acidimicrobiales), but which still lack representative genomes and cultured organisms. At the same time, our study revealed a high degree of endemism and isolation, hence a significant part of bacterial communities in deep-sea surface sediments appears to be geographically restricted. We found evidence that the relative sequence-abundance of a taxon and the size of its geographic range are positively related to each other. We also detected that deep-sea sediment bacterial community similarity decreases with increasing geographic distance, most likely due to isolation-by-distance processes (especially along longitudes). Bacterial communities mostly changed with indicators of productivity regimes, such as TOC content of sediments.

Chao1 richness estimates (blue, left y axis) calculated with 100 sequence re-samplings for data without (a) and with (b) SSOabs (n resampling = 6,883 and 7,922 sequences for a and b, respectively).

Standard deviations for richness are indicated in black. Water depth of each sample is displayed in red (right y axis). No significant relationship was found between richness and water depth (Spearman’s ρ = -0.33 and -0.37 for a and b, respectively, P > 0.05 in both cases).

(PDF)

Click here for additional data file.

Observed richness (blue, left y axis) calculated with 100 sequence re-samplings for data without (a) and with (b) SSOabs (n resampling = 6883 and 7922 sequences for a and b respectively).

Standard deviations for richness are indicated in black. Water depth of each sample is displayed in red (right y axis). The relationship between richness and water depth was significant but weak (Spearman’s ρ = -0.44 and -0.43; P = 0.02 in both cases).

(PDF)

Click here for additional data file.

Species accumulation curves based on different bacterial taxonomic categories: phylum to family (a), genera (b), and OTU0.03 level (c).

Colors in a) mark the taxonomic categories phylum: white, class: red, order: orange, family: yellow. The boxplots show a summary of 100 permutations, calculated with random subsampling, including absolute singletons for comparison.

(PDF)

Click here for additional data file.

Variations of truly cosmopolitan OTUs affiliated with the clades JTB255 (a, class Gammaproteobacteria, n = 2) and OM1 (b, class Actinobacteria, n = 1) between oceanic regions.

Relative abundances were averaged across samples and oceans. Error bars indicate standard deviations when considering samples from one oceanic region.

(PDF)

Click here for additional data file.

Differences in bacterial community composition between oceanic regions.

Non-metric multidimensional scaling plots for community composition at the class (a-b) and OTU0.03 (c-d) levels in terms of presence/absence (a,c; using the Jaccard index) and relative abundance (b,d; using the Bray-Curtis index). Samples originating from a same oceanic region are connected by a coloured line, as follows: black: South Pacific; red: North Pacific (St. M); green: Indian Ocean; blue: NE-Atlantic; light-blue: E-Mediterranean; pink: Arctic; yellow: North Pacific (Japan); brown: Antarctic; orange: South Atlantic. Differences in community composition were weak at the class level (ANOSIM R = 0.2 and 0.51, p = 0.03 and 0.001 for a and b, respectively), but significant at the OTU0.03 level (ANOSIM R = 0.7 and 0.66, p = 0.001 for c and d, respectively).

(PDF)

Click here for additional data file.

Percentage of shared OTU0.03 between oceanic regions.

(PDF)

Click here for additional data file.

Proportions of unique and cosmopolitan OTU0.03 between oceanic regions and individual samples, including SSOabs, and after averaging of 100 sequence random resampling results (n sequences = 7,922, Standard deviations are indicated).

(PDF)

Click here for additional data file.

Partitioning of the biological variation in bacterial community structure at the OTU0.03 level (with absolute singletons excluded) between the following contextual parameters: geographic distance between samples, water depth, TOC availability (TOC regions based on Seiter et al. 2004, Deep Sea Res., Part I 51: 2001–2026), and surface productivity (Longhurst productivity index based on Longhurst et al. 1995, J Plankton Res 17: 1245–1271).

** p = 0.01, * p = 0.05, (*) p = 0.07, as tested with 100 permutations.

(PDF)

Click here for additional data file.

Bacterial OTU0.03 richness as a function of latitude.

Linear model is not significant.

(PDF)

Click here for additional data file.

Contextual data for all deep-sea samples: VAMPS (http://vamps.mbl.edu) sample ID, geographic origin, water depth, oceanic region, and sequence archive accession numbers for GenBank Sequence Read Archives (www.ncbi.nlm.nih.gov).

(PDF)

Click here for additional data file.

Relative sequence abundance of the most abundant bacterial classes in deep water (>1000 m water depth), deep-sea surface sediment (>1000 m water depth), and deep subsurface sediment (2.5–90 m below seafloor) samples.

(PDF)

Click here for additional data file.

a) Twenty most abundant bacterial OTU0.03 for deep-sea surface sediment (> 1000 m water depth). Total number of samples considered is 27. Total number of sequences in the dataset is 501,480. b) Twenty most abundant bacterial OTU0.03 for deep subsurface samples (between 2.5 and 90 m below seafloor from the Peru Margin; http://icomm.mbl.edu). Total number of samples considered is 5. Total number of sequences in the dataset is 72,294.

(PDF)

Click here for additional data file.

R script including all relevant analyses for this study.

(R)

Click here for additional data file.