High-throughput sequencing results
A total of 26 million COI reads, 19 million raw 18S V1-V2 reads,, 14 million 18S V4 reads, and 17 million 16S V4–V5 reads were obtained from three Illumina HiSeq runs of amplicon libraries built from pooled triplicate PCRs of 22 environmental samples, 2 extraction blanks, and 4–6 PCR blanks (Supplementary Table S4 online). The in situ pump yielded less raw reads for COI and 16S (Supplementary Fig. S1 online, F = 4.02–14.4, p = 0.0003–0.03), while more raw reads were recovered from both water sampling methods with 18S V4 (F = 6.5, p = 0.007). Water samples generally yielded fewer raw clusters (F = 5.1–35.1, p = 3.2 × 10−6–0.02), except for 18S V4 where numbers were comparable across sample types (Supplementary Fig. S1 online).
Bioinformatic processing (quality filtering, error correction, chimera removal, and clustering for metazoans) reduced read numbers to 20 million for COI, 12 million for 18S V1–V2, 11 million for 18S V4, and 10 million for 16S V4–V5, resulting in 10,351 and 17,608 raw OTUs for COI and 18S V1–V2 respectively; 35,538 raw 18S V4 ASVs, and 62,646 raw 16S ASVs (Supplementary Table S4 online). For eukaryote markers, 17–55% of the raw reads remained in PCR blanks after bioinformatic processing, while 50–75% remained in extraction blanks and 52–87% in true samples. In contrast, with 16S, these values were at 87% for PCR blanks, 67% for extraction blanks, and 29–73% for true samples. Thus, negative control samples accounted for 7–13% of bioinformatically processed reads with eukaryotes, compared to 27% with prokaryotes. The vast majority of 16S reads generated by negative controls belonged to a common contaminant of Phusion polymerase kits, which is well amplified in low concentration samples such as negative controls. These reads however accounted for < 1% of 16S ASVs. After data refining (decontamination, removal of all control samples and of all unassigned or non-target clusters), rarefaction curves showed a plateau was reached for all samples except in situ pump samples with microbial loci and sediment samples with 18S V4, suggesting that not all protist and prokaryote diversity was captured at this sequencing depth in these samples (Supplementary Fig. S1 online). Refined datasets contained 7 million reads for prokaryotes and between 4.8 and 8.5 million target reads for eukaryotes, delivering 38,816 prokaryote ASVs (16S V4–V5), 8031 protist ASVs (18S V4), and 2,319 (COI) and 1,460 (18S V1–V2) LULU curated metazoan OTUs (Table S4). For COI, while only 1.2% of metazoan OTUs were unassigned at phylum-level, 57% had BLAST hit identities < 80%, i.e. unreliable at phylum-level. For 18S, 7% (18S V4) to 16% (18S V1-V2) of final ASVs/OTUs were unassigned at phylum-level, but only 12% (18S V4) and 13% (18S V1-V2) had BLAST hit identities < 86%. For 16S, 0.9% and 3% of ASVs had no or unreliable phylum-level assignments, respectively.
Alpha diversity between sampling methods
The number of recovered molecular clusters significantly differed with sampling method, water samples detecting less diversity than sediment samples for metazoans and prokaryotes (COI: F = 20.1, p = 4.4 × 10−5, 18S V1–V2: F = 6.5, p = 0.01, 16S: F = 56.0, p = 3 × 10−7), but both sample types recovering similar levels of protist diversity (18S V4: F = 2.9, p = 0.07, Fig. 1). Sieved and unsieved sediment resulted in statistically comparable total cluster numbers in all loci investigated (Table 1) although, for metazoans detected with 18S V1–V2, this lack of significance was likely due to the very low yield observed in one sieved sample (PL11), as the other sieved samples detected considerably more OTUs (Fig. 1). For metazoans detected with COI, sieved samples tended to detect less OTUs although being based on larger sediment volumes (pool of 5 DNA extracts). The number of total recovered OTUs did not differ significantly between the water sampling box and the in situ pump for metazoans (COI, 18S V1–V2), but differences were observed for unicellular eukaryotes (18S V4) and prokaryotes (16S V4–V5) depending on the sampling box size fraction (Table 1), with the smallest fraction (0.2–2 µm) detecting more ASVs (Fig. 1).
Numbers of metazoan OTUs (COI, 18S V1–V2), unicellular eukaryote (18S V4) and prokaryote (16S V4–V5) ASVs recovered by deep-sea sediment (brown) and aboveground water (blue), using for each sample type two sampling methods based on varying amounts of starting material. Sediment was either sieved through 5 mesh sizes to size-sort organisms prior DNA extraction, or DNA was extracted directly from crude sediment samples. Water was collected with a 7.5 L sampling box, allowing recovery of up to two size classes per taxonomic compartment, or sampled in large volumes with an in situ pump. Cluster abundances were calculated on rarefied datasets.
Differences in total recovered diversity were not solely a result of differences in sample volume. Indeed, sieved samples, based on 3–6 times more sediment, did not consistently detect more diversity (Fig. 1). Similarly, the in situ pump, although sampling ~ 800 times more water than the sampling box, did not consistently detect more diversity than any size fraction of the sampling box (Fig. 1).
Recovered diversity among sample types strongly differed depending on phylum (Fig. 2). For metazoans, water samples led to the detection of significantly higher numbers of OTUs than sediment samples for Arthropoda, Chordata (COI, t-tests, p = 0.006–0.01) and Mollusca (18S V1–V2, t-test, p = 0.03), and some phyla like Brachiopoda, Ctenophora, Echinodermata, or Gastrotricha, were only detected in water samples (Fig. 2). In contrast, phyla such as Platyhelminthes, Porifera (COI, 18S V1–V2, t-tests, p = 0.001–0.04), Kinorhyncha, Nematoda, Tardigrada, or Xenacoelomorpha (18S V1–V2, t-tests, p = 0.001–0.04) produced significantly more OTUs in sediment than water samples (Fig. 2). Similarly, some protistan groups, such as the Acantharea, Chlorophyta, Dinophyceae, or Syndiniales (t-tests, p = 0.002–0.03) were predominant in water samples (Supplementary Fig. S2 online), while others were significantly more diverse in sediment, e.g., Apicomplexa, Filosa groups, Ciliophora, Labyrinthulea, RAD-B (t-tests, p = 0.001–0.04). For prokaryotes, most lineages were predominant in sediment (t-tests, p = 2.2 × 10−7–0.02, e.g., Archaea, Acidobacteria, Actinobacteria, Bacteroidetes, Chloroflexi, Delta-, Gamma-proteobacteria, Gemmatimonadetes, Latescibacteria, Hydrogenedentes, Nitrospirae, Planctomycetes), and only Cyanobacteria were significantly more diverse in water samples (t-test, p = 0.001).
Mean numbers (± SE) of metazoan COI and 18S V1-V2 OTUs detected in target phyla for sediment (brown) and water (blue), using two sampling methods for both sample types. Sediment was either sieved to size-sort organisms prior DNA extraction, or DNA was extracted directly from crude sediment samples. Water was collected with a 7.5 L sampling box, allowing recovery of two size classes, or sampled in large volumes with an in situ pump. OTU numbers were calculated on rarefied datasets.
For sediment, recovered levels of alpha diversity among sampling methods also varied by phyla and organism size class (Fig. 2). For meiofauna phyla, best detected with 18S V1–V2, more OTUs were detected from sieved than from unsieved sediment (Kinorhyncha, Nematoda, Platyhelminthes, Rotifera, Tardigrada, Xenacoelomorpha), although this difference was not significant, likely due to the low sample size. However, differences in alpha diversity among sampling methods were not only a result of differences in sample volume, as some unsieved samples yielded similar or greater numbers of OTUs than sieved samples for many phyla (Supplementary Fig. S3 online). Sieved and unsieved sediment detected comparable ASV numbers in most microbial groups, except the Chrysophyceae, Actinobacteria, Cyanobacteria, Gammaproteobacteria, Nanoarchaeaeota (Supplementary Fig. S2 online, paired t-tests, p = 0.02–0.04).
Water sampling methods strongly differed in terms of recovered alpha diversity depending on taxonomic compartment. The in situ pump generally detected more metazoan diversity than the sampling box, and phyla such as Brachiopoda, Ctenophora, and Echinodermata were only detected by the pump (Fig. 2). However, the in situ pump detected significantly more ASVs than the sampling box only in some taxonomic groups for protists (Bacillariophyta, Oomycota, Phaeodarea) and prokaryotes (e.g., Bacteroidetes, Chlamydiae, Firmicutes, Tenericutes, Lentisphaerae, and Delta-, Gammaproteobacteria, see Supplementary Fig. S2 online, t-tests, p = 9 × 10−5–0.003). Other clades were significantly more diverse in the sampling box (e.g., the protist groups Haptophyta, Picozoa, Telonemia, and the Cyanobacteria, t-tests, p = 0.002–0.02). With the sampling box, the smallest size fraction (0.2–2 µm) allowed recovering more alpha diversity in all microbial groups than the larger size fraction (2–20 µm). This difference was significant only for Chlorophyta, Labyrinthulea, Chloroflexi, and Verrucomicrobia (paired t-tests, p = 0.01–0.03), although non-significant comparisons may result from the limited number of samples available. The two size fractions available with the sampling box for COI (2–20 µm, > 20 µm) did not reveal differences in diversity recovery with size class, as most phyla were detected equally well in both (Fig. 2).
Effect of sampling method on community structures
Community compositions significantly differed among sampling methods for all investigated loci (COI: pseudo-F = 2.3, p = 0.001; 18S V1–V2: pseudo-F = 2.3, p = 0.001, 18S V4: pseudo-F = 4.1, p = 0.001, 16S: pseudo-F = 18.3, p = 0.001) and sampling method accounted for 41–45% of variation among samples for metazoans (COI, 18S V1–V2), 60% for protists (18S V4), and 87% for prokaryotes (16S).
Pairwise PERMANOVAs showed that community structures differed most strongly among sample types (water or sediment, R2 = 0.28–0.89), although not all pairwise comparisons were significant, likely due to the limited number of samples available for the sampling box (Table 1). Relative taxonomic compositions revealed by aboveground water samples differed from sediment samples, with higher proportions of arthropods, chordates, annelids, and tunicates in the water samples, while nematodes, poriferans, platyhelminths, and xenacoelomorphs were predominant in sediment samples (Fig. 3 COI and 18S V1–V2). Similarly, protist diversity in aboveground water samples was dominated by Dinophyceae, Haptophyta, Phaeodarea, Syndiniales, and to a lesser extent Bacillariophyta and Telonemia, while apicomplexans, ciliates, filosans, and labyrinthuleans represented higher proportions of diversity in sediment samples (Fig. 3 18S V4). For prokaryotes, aboveground water communities were characterised by Alphaproteobacteria, Cyanobacteria, and Gammaproteobacteria, while Acidobacteria, Deltaproteobacteria, Archeae, Latescibacteria, and Planctomycetes showed higher diversity in sediment (Fig. 3 16S V4–V5).
Patterns of relative cluster abundance resolved by eDNA metabarcoding of deep-sea sediment (brown) and aboveground water (blue), using two sampling methods for both sample types, and using four barcode markers targeting metazoans (COI, 18S V1–V2), micro-eukaryotes (18S V4), and prokaryotes (16S V4–V5). Sediment was either sieved to size-sort organisms prior DNA extraction, or DNA was extracted directly from crude sediment samples. Water was collected with a 7.5 L sampling box, allowing recovery of up to two size classes per taxonomic compartment, or sampled in large volumes with an in situ pump. Top 20 most abundant taxa are displayed for microbial groups.
Only 3% (COI), 5% (18S V1–V2), 8% (18S V4), and 5% (16S) of clusters were shared between sediment and water samples, resulting in strong segregation in ordinations (Fig. 4). For metazoans, taxa shared among water and sediment samples were mostly assigned to hydrozoans (COI, 28%, 18S, 7%), copepods (COI, 6%, 18S, 20%), gastropods (COI, 31%), demosponges (COI, 6%), or polychaetes (18S, 10%), and chromadorean nematodes (18S, 11%). For protists, ASVs shared among sample types primarily belonged to the Syndiniales (39%), but other taxa included dinophyceans (11%), filosans (9%), labyrinthuleans (5%), and bacillariophytes (6%). For prokaryotes, ASVs shared across sample types were predominantly belonging to the Proteobacteria (Gamma, 19%, Alpha, 10%, Delta, 8%), Bacteroidetes (15%), or Planctomycetes (16%).
Detrended Correspondence Analyses (DCA) ordinations (left) and Venn diagrams (right), showing differences in community compositions detected by deep-sea sediment (brown) and aboveground water (blue) for metazoans (COI and 18S V1–V2), micro-eukaryotes (18S V4), and prokaryotes (16S V4–V5). Community segregation is strongest between sample types, but also among target size class in the water samples. Sediment was either sieved to size-sort organisms prior DNA extraction, or DNA was extracted directly from crude sediment samples. Water was collected with a 7.5 L sampling box, allowing recovery of two size classes in each taxonomic compartment, or sampled in large volumes with an in situ pump.
Sediment processing did not significantly affect detected community structures (Table 1), and sieved and unsieved sediment resolved comparable communities at the phylum-level (Fig. 3), although community segregation was observed in ordinations of metazoans resolved with 18S V1–V2 and protists resolved with 18S V4 (Fig. 4). Between 21 and 36% of sediment OTUs/ASVs were shared among sieved and unsieved sediment samples. Shared metazoan OTUs primarily belonged to Hydrozoa (18S, 2.5%, COI, 32%, Siphonophorae, Anthoathecata, Leptothecata), Demospongiae (COI, 9%), Gastropoda (COI, 32%), Nematoda (18S, 61% Chromadorea, 11% Enoplea), Polychaeta (18S, 2.5%), or Copepoda (18S, 4.5%). Microbial ASVs shared among sieved and unsieved sediment mostly belonged to Syndiniales (17%), Filosa (19%), Ciliophora (11%), Dinophyceae (9%), Planctomycetes (22%), Acidobacteria (10%), or Proteobacteria (Gamma, 9%, Alpha, 8%, Delta, 11%).
In contrast, sampling method significantly affected resolved community structure for water, as for all size fractions, the sampling box detected significantly different communities than the in situ pump (Table 1). Both sampling methods resolved different communities at the phylum-level (Fig. 3), and water samples always clustered apart in ordinations (Fig. 4) for all taxonomic compartments investigated.. Between 8 and 11% of ASVs/OTUs detected in water were shared between the in situ pump and the sampling box. Taxonomic structures resolved by both sampling methods clearly changed due to targeted size fraction. The sampling box’s 2–20 µm size fraction did not detect the same metazoan community assemblage as the > 20 µm assemblage detect by the pump (Fig. 3 COI and 18S V1–V2). Similarly, for microbial data, the in situ pump targeting the > 20 µm size class, and the sampling box targeting both the 2–20 µm and the 0.2–2 µm size classes, detected different community assemblages. For protists, the in situ pump detected higher proportions of ASVs for Bacillariophyta, Ciliophora, Labyrinthulea, or Phaeodarea, while the sampling box detected more cryptophytes, haptophytes, MAST, and telonemians (Fig. 3 18S V4). For prokaryotes, the sampling box detected more diversity in the Alphaproteobacteria, Chloroflexi, or Marinimicrobia (Fig. 3 16S V4–V5).
Source: Ecology - nature.com
