Content uploaded by Chris Greening
Author content
All content in this area was uploaded by Chris Greening on Sep 28, 2021
Content may be subject to copyright.
ARTICLE OPEN
Hydrodynamic disturbance controls microbial community
assembly and biogeochemical processes in coastal sediments
Ya-Jou Chen
1,2,5
, Pok Man Leung
1,2,5
, Perran L. M. Cook
3
, Wei Wen Wong
3
, Tess Hutchinson
3
, Vera Eate
3
, Adam J. Kessler
4
and
Chris Greening
1,2
✉
© The Author(s) 2021
The microbial community composition and biogeochemical dynamics of coastal permeable (sand) sediments differs from cohesive
(mud) sediments. Tide- and wave-driven hydrodynamic disturbance causes spatiotemporal variations in oxygen levels, which select
for microbial generalists and disrupt redox cascades. In this work, we profiled microbial communities and biogeochemical dynamics
in sediment profiles from three sites varying in their exposure to hydrodynamic disturbance. Strong variations in sediment
geochemistry, biogeochemical activities, and microbial abundance, composition, and capabilities were observed between the sites.
Most of these variations, except for microbial abundance and diversity, significantly correlated with the relative disturbance level of
each sample. In line with previous findings, metabolically flexible habitat generalists (e.g., Flavobacteriaceae, Woeseaiceae,
Rhodobacteraceae) dominated in all samples. However, we present evidence that aerobic specialists such as ammonia-oxidizing
archaea (Nitrosopumilaceae) were more abundant and active in more disturbed samples, whereas bacteria capable of sulfate
reduction (e.g., uncultured Desulfobacterales), dissimilatory nitrate reduction to ammonium (DNRA; e.g., Ignavibacteriaceae), and
sulfide-dependent chemolithoautotrophy (e.g., Sulfurovaceae) were enriched and active in less disturbed samples. These findings
are supported by insights from nine deeply sequenced metagenomes and 169 derived metagenome-assembled genomes.
Altogether, these findings suggest that hydrodynamic disturbance is a critical factor controlling microbial community assembly and
biogeochemical processes in coastal sediments. Moreover, they strengthen our understanding of the relationships between
microbial composition and biogeochemical processes in these unique environments.
The ISME Journal; https://doi.org/10.1038/s41396-021-01111-9
INTRODUCTION
The defining feature of intertidal sediments is their exposure to
regular hydrodynamic disturbance due to tidal flows and waves.
The resultant frequent variations in physicochemical conditions
exert major selective pressures on the microorganisms that
control biogeochemical cycling in these environments [1,2]. The
low permeability of cohesive (mud/silt) sediments buffers micro-
organisms from disturbance; as a result, these systems become
depth-stratified in redox state, community composition, and
biogeochemical reactions [3]. The scenario is very different for
the permeable (sand/gravel) sediments that span half of
continental shelves [4,5]. Pressure gradients form at these sites
due to interaction of wave action with sediment topography,
bottom currents, and bioirrigation. These gradients force water to
flow through sediment through the process of advective transport
[2,6,7], resulting in rapid exchange of dissolved particles, solutes,
gases, and microorganisms between porewater and sediment
grains [8–10]. In turn, these physical processes cause large
variations in the levels of hydration, oxygen, light, and nutrients
available to grain-associated microorganisms across short spatial
and temporal scales [2,11,12]. Various factors, including the
degree of tide- and wave-driven hydrodynamic force on
sediments, control the extent of porewater advection and in turn
the spatiotemporal variability of these systems [13,14]. This
disturbance is predicted to profoundly influence microbial
community assembly and biogeochemical processes.
Permeable sediments host microbial communities that are
distinct from those of cohesive sediments [15–20]. Variations in
resource availability and oxygen exposure select for flexible
habitat generalists rather than niche-restricted specialists [21].
Consistently, many of the most abundant and prevalent bacterial
lineages in permeable sediments, most notably Woeseiaceae and
Flavobacteriaceae, are highly metabolically versatile [21–25].
Based on metagenome-assembled genomes (MAGs), many of
these taxa are capable of simultaneously or alternately using
multiple energy sources (e.g., organic carbon, sulfide, hydrogen,
sunlight), carbon sources (organic carbon, carbon dioxide), and
metabolic strategies (e.g., aerobic respiration, denitrification,
fumarate reduction, fermentation) [21–23]. Continual variations
in oxygen levels in these sediments select for facultative
anaerobes; in situ evidence suggests that some bacteria can even
perform aerobic and anaerobic respiration simultaneously, for
example aerobic denitrifiers [26,27]. In contrast, obligate anae-
robes such as sulfate reducers and methanogens are thought to
Received: 30 April 2020 Revised: 6 September 2021 Accepted: 8 September 2021
1
Department of Microbiology, Biomedicine Discovery Institute, Clayton, VIC 3800, Australia.
2
School of Biological Sciences, Monash University, Clayton, VIC 3800, Australia.
3
Water
Studies Centre, School of Chemistry, Monash University, Clayton, VIC 3800, Australia.
4
School of Earth, Atmosphere and Environment, Monash University, Clayton, VIC 3800,
Australia.
5
These authors contributed equally: Ya-Jou Chen, Pok Man Leung. ✉email: chris.greening@monash.edu
www.nature.com/ismej
be inhibited by transient oxygenation, despite their preferred
electron donors and acceptors being available [21,23].
The physical features and microbial communities of permeable
sediments in turn influence biogeochemical processes. Permeable
sediments exposed to high tidal disturbance are minimally
stratified in geochemistry and thus carbon mineralization does
not follow the classical ‘redox cascade’established for cohesive
sediments [3]. In sediments from Port Philip Bay, Australia,
fermentation is the dominant pathway of carbon mineralization
under anoxic conditions and is largely uncoupled from anaerobic
respiration [23,28]. This reflects that facultative fermenters are the
dominant community members and switch to hydrogenogenic
fermentation when preferred electron acceptors such as oxygen
and nitrate are limiting. In turn, fermentation products accumulate
in situ and ex situ due to low levels of sulfate reducers and other
obligate anaerobes [23]. Likewise, multiple studies have inferred
that rates of denitrification exceed those of dissimilatory nitrate
reduction to ammonium (DNRA), again suggesting the predomi-
nance of processes associated with facultative anaerobes rather
than obligate anaerobes [29–31]. Nevertheless, there is evidence
of some variation in the anaerobic respiratory processes between
sediments, with high levels of sulfate reducers reported in some
environments [13,32,33]. For example, Probandt et al. observed
that the abundance of sulfate-reducing Desulfobulbaceae and
Desulfobacteraceae in surface sands decreases with permeability
[18]. Thus, differences in mixing levels between different
sediments, for example due to variations in hydrodynamic forcing
or sediment topography, likely influence microbial community
assembly and in turn biogeochemical processes.
In this work, we build on this conceptual framework to
investigate how the microbiology and biogeochemistry of perme-
able sediments varies across a disturbance gradient. To do so, we
sampled sediment cores from three sites along a 2.1 km stretch of
beach in Port Philip Bay, Australia, which differed in levels of
hydrodynamic forcing: one site was fully exposed to wave
disturbance (site A), whereas the others were either moderately
(site B) or highly (site C) buffered by a breakwater (Fig. S1). Based
on the above framework, we developed three key testable
hypotheses for how these sites may differ: (1) Less disturbed
sites will be more stratified in geochemistry and microbial
community structure; to test this, we combined geochemical
analysis with 16S rRNA gene-based community profiling of
sediment core subsections. (2) More microbial specialists, includ-
ing obligate anaerobes, will be present in less disturbed sites; to
test this, we combined deep metagenomic sequencing of each
site, yielding 169 MAGs, with microbial community analysis and
biogeochemical assays. (3) Carbon mineralization processes will be
more tightly coupled to anaerobic respiration in less disturbed
sites; we tested this by performing microcosm experiments to
measure rates of hydrogenogenic fermentation, sulfate reduction,
denitrification, and DNRA across the sediments. Our investigations
suggest that these predictions are partially correct, though some
unexplained patterns were observed. Community composition
and metabolic genes showed strong correlations with the
disturbance level of a given sample within a depth profile (as
inferred by distance to sulfidic layer), as well as by site or depth
alone. These findings also enhance knowledge of the processes
and microorganisms controlling marine biogeochemical cycling.
MATERIALS AND METHODS
Sediment sampling
Permeable sediments were sampled across a 2.1 km stretch of Port Phillip
Bay, Australia. Three different sampling sites, site A (Middle Park Beach;
37.851342°S, 144.954377°E), site B (Cummings Reserve; 37.856283°S,
144.964258°E), and site C (St. Kilda Pier; 37.863159°S, 144.971026°E), were
selected based on their different levels of hydrodynamic exposure due to
the St. Kilda Breakwater providing shelter from the prevailing westerly and
southerly wind directions (Fig. S1). The three sites were sampled on eight
different dates for different purposes: preliminary community and
geochemical profiling (28/02/2018); complete community and geochem-
ical profiling (14/06/2018); measurement of mixing layer depth and grain
size distribution (14/06/2018, 26/02/2019, 11/03/2019); and microcosm
experiments to analyze carbon fixation (07/08/2018), H
2
metabolism (20/
06/2018), sulfide production (26/02/2019), denitrification (16/03/2020),
DNRA (16/03/2020), and nitrification (21/03/2021). We confirmed that the
sediments varied in hydrodynamic forcing by measuring mixing layer
depth and grain size distribution at three dates (Table S1). To measure
mixing layer depth, cores of 30 cm were collected and photographed, and
the depth of the darker sulfidic layer was quantified using ImageJ [34]. To
determine grain size distribution, 100 g sediments (dry weight) were
collected at each of the three sampling dates. Sand was progressively
separated using eight sieves of different sizes (4, 2, 1, 0.85, 0.5, 0.25, 0.125,
and 0.063 mm). Median grain size (D
50
) was calculated based on the
equation of Ferguson & Church [35,36]. Five individual sediment cores of
30 cm depth were collected from each site at two dates. Cores were kept
on ice until delivery to the laboratory and then immediately sliced every 2
cm, with the eight sections from the top 0 to 16 cm used for downstream
analysis. For each site, two cores were used for DNA extraction and
chlorophyll ameasurement, and three other cores were used for sulfide
and ammonium measurements.
Geochemical measurements
Sulfide and ammonium content were measured immediately after
sediment sectioning. Approximately 30 g from each sediment slice was
transferred into 30 ml of N
2
-purged artificial seawater. Following stirring for
10 s, the supernatant was extracted using a syringe for further analysis.
Free sulfide concentrations were quantified using the methylene blue
method with a GBC UV-Visible 918 Spectrophotometer at 670 nm as
previously described [37]. For acid-volatile sulfide (AVS) measurements,
sediments were stored frozen and then analyzed as previously described
[38]. Briefly, after samples were thawed and homogenized, 0.1 g of
sediment was treated with an acidified methylene blue reagent,
centrifuged, and stored in the dark for 90 min before analysis. This process
results in the conversion of AVS to free sulfide, which is then quantified in
the same way as free sulfide. Ammonium concentrations were determined
by the phenate method using a Lachat Quickchem 8000 Flow Injection
Analyzer at 630 nm as previously described [39]. Chlorophyll awas
extracted and quantified as previously described using a Hitachi U-2800
spectrophotometer (Hitachi High-Technologies Corporation, Tokyo, Japan)
[21,40].
DNA extraction and microbial community analysis
DNA was extracted from 0.3 g of each 2 cm sediment slice using the MoBio
PowerSoil Isolation kit according to the manufacturer’s instructions
(https://www.qiagen.com/au/resources/download.aspx?id=5c00f8e4-c9f5-
4544-94fa-653a5b2a6373&lang=en). In total, 48 samples were sequenced
and analyzed (3 sites × 2 cores × 8 depths). Samples were eluted in DNase-
and RNase-free UltraPure Water (Thermo Fisher Scientific). A sample-free
negative control, containing only UltraPure Water, was also extracted.
Nucleic acid purity and yield were confirmed using a Nanodrop
1000 spectrophotometer and a Qubit 2 fluorometer. To estimate the
number of bacteria and archaea present in each sample, quantitative PCR
(qPCR) of the 16S rRNA gene was performed using universal primer pairs
F515 and R806 [41]; assays were performed using a 96-well plate in a pre-
heated LightCycler 480 Instrument II (Roche, Basel, Switzerland) and 16S
rRNA gene copy number was quantified against a serially diluted pMA
plasmid containing the Escherichia coli 16S rRNA gene as previously
described [21]. For amplicon sequencing, the V4 hypervariable region for
16S rRNA gene was amplified using the primer pairs F515 and R806 [41].
Amplicons were subject to 2 × 300 bp sequencing on a MiSeq platform
(Illumina) at the Australian Centre for Ecogenomics (ACE), the University of
Queensland. Amplicon sequences were then processed using the pipeline
provided by ACE (https://wiki.ecogenomic.org/doku.php?id=amplicon_
pipeline_readme). For sequencing runs did not attain requested depth,
the same library was re-sequenced and combined in QIIME2. The forward
reads were trimmed to 250 base pairs and low quality reads were removed
using Trimmomatic [42]. All reads were then subjected to de-noising using
the DADA2 pipeline [43] in QIIME2 [44]. A total of 2,969,857 reads from
48 samples were obtained from the dataset (Table S2), with reads removed
by the DADA2 pipeline provided in Table S3. The negative control did not
yield quantifiable DNA or detectable amplicons on agarose gels,
Y.-J. Chen et al.
2
The ISME Journal
1234567890();,:
suggesting minimal contamination during sample processing, and was not
sequenced. For taxonomic assignment, all reference reads that matched
the F515/R806 primer pair were extracted from the Genome Taxonomy
Database (GTDB) release 04-RS89 [45] and used to train a naïve Bayes
classifier by using the fit-classifier-naive-Bayes function with default
parameters.
Biodiversity analysis
All statistical analysis and visualizations were performed with R software
version 3.6.2 (December 2019) using the packages phyloseq ggplot2 [46]
and microbiome [47]. The outputs from QIIME2, without rarefaction, were
used to analyze community composition at phylum, order, family, and
genus levels. The relative abundance of each assigned order, family, and
genus was compared by site (categories: site A, site B, site C) and depth
(categories: 0–4 cm (shallow), 4–10 cm (medium), 10–16 cm (deep)) using
one-way ANOVAs. In addition, linear regressions were performed to test
how the relative abundance of each assigned order, family, and genus
varied relative to the disturbance level of each individual sample (as
inferred by their distance in cm relative to the average depth of the dark
sulfidic layer). To analyze alpha diversity and beta diversity, all sequences
were rarefied at 10,000 sequences per sample using the phyloseq function
rarefy_even_depth() to account for the difference in library sizes [48]. The
rarefied dataset retained 907,977 reads (30%) across 43 of the 48 original
samples (90%). Alpha diversity was calculated using several metrics,
including Shannon index and Simpson index. Significant differences in
Shannon index between sites and depths were tested using an ANOVA
(one-way analysis of variance) with Tukey’s post hoc tests. Beta diversity
was calculated using weighted UniFrac distances [49] of log-transformed
data and visualized using a principal coordinate analysis (PCoA) plot. A
pairwise analysis of similarities (ANOSIM) was also used to test for
significant differences in community composition. First, a nested
permutational multivariate analysis of variance (PERMANOVA) was
performed using 999 permutations to test for significant differences.
Second, a beta dispersion test (PERMDISP) was used to ascertain if
observed differences were influenced by dispersion. An analysis of
composition of microbiomes (ANCOM) [50] was also performed in QIIME2
to detect differential abundance of genera between sites A and C. This
analysis compares the centered log-ratio (clr) transformed data of a specific
taxon to the rest of taxa between two distinct environments.
Shotgun metagenomic assembly and binning
DNA samples extracted from shallow (0–2 cm), intermediate (6–8 cm), and
deep (14–16 cm) sediment slices from each of the three sites were subject
to shotgun metagenomic sequencing. Metagenomic shotgun libraries
were prepared for each sample using the Nextera XT DNA Sample
Preparation Kit (Illumina Inc., San Diego, CA, USA) and sequencing was
performed on an NextSeq500 platform (Illumina) with a 2 × 150 bp High
Output run. Sequencing yielded 523,268,618 read pairs across the nine
metagenomes. The BBDuk function of the BBTools v38.51 (https://
sourceforge.net/projects/bbmap/) was used to clip contaminating adapters
(k-mer size of 23 and hamming distance of 1), filter PhiX sequences (k-mer
size of 31 and hamming distance of 1), and trim bases with a Phred score
below 20 from the raw metagenomes. After removing resultant reads with
lengths shorter than 50 bp, 450,359,308 high-quality read pairs were
retained for downstream analysis. Reads were assembled individually with
MEGAHIT v1.2.9 [51] (-k-min 27, -k-max 127, -k-step 10, -min-contig-len
500). To improve recovery of metagenomic bins, metagenomes from
samples of high hydrodynamic disturbance (A0-2, A6-8, B0-2), intermediate
hydrodynamic disturbance (A14-16, B6-8, C0-2), and low hydrodynamic
disturbance (B14-16, C6-8, C14-16) were also assembled collectively using
MEGAHIT with same parameters as above. Bowtie2 v2.3.5 [52] was used to
map short reads back to assembled contigs using default parameters to
generate coverage profiles. Subsequently, genomic binning was per-
formed using Autometa (09/2019) [53], CONCOCT v1.1.0 [54], MaxBin2
v2.2.6 [55], and MetaBAT2 v2.15 [56] on contigs with lengths over 2000 bp.
Resulting bins from the same assembly were then dereplicated using
DAS_Tool v1.1.2 [57]. RefineM v0.0.25 was used to remove potentially
contaminating contigs with incongruent genomic and taxonomic proper-
ties in the bins [58]. Applying a threshold average nucleotide identity of
99%, bins from different assemblies of each hydrodynamic disturbance
category were consolidated to a non-redundant set of MAGs using dRep
v2.5 [59]. Completeness and contamination of MAGs were assessed using
CheckM v1.1.2 [60]. In total, 30 high quality (completeness > 90% and
contamination < 5%) and 139 medium quality (completeness > 50% and
contamination < 10%) [61] MAGs were recovered. Their corresponding
taxonomy was assigned based on GTDB release 05-RS95 by GTDB-Tk v1.0.2
[45]. Open reading frames (ORFs) in MAGs were predicted using Prodigal
v2.6.3 [62].
Shotgun metagenome community and functional analysis
Bacterial, archaeal, and eukaryotic community composition of the
metagenomes was profiled using phyloFlash v.3.4 [63]. All quality-filtered
reads were screened for the small subunit ribosomal RNA gene (SSU rRNA)
sequences and assembled with the command phyloFlash.pl and the option
–almosteverything. The SSU Ref NR99 database from the SILVA release
138 served as the reference for the sequence searching and taxonomy
assignment of SSU reads to Nearest Taxonomic Units (NTUs) [64]. To
estimate the metabolic capability of the sediment communities, meta-
genomes and derived genomes were searched against custom protein
databases (https://doi.org/10.26180/c.5230745) of representative meta-
bolic marker genes [65] using DIAMOND v.0.9.31 (query cover > 80%) [66].
Searches were carried out using all quality-filtered unassembled reads with
lengths over 140 bp and the ORFs of the 169 MAGs. These genes are
involved in respiration (AtpA, NuoF, SdhA, CoxA, CcoN, CyoA, CydA), sulfur
cycling (AsrA, FCC, Sqr, DsrA, Sor, SoxB), nitrogen cycling (AmoA, HzsA,
NifH, NarG, NapA, NirS, NirK, NrfA, NosZ, NxrA, NorB), iron cycling (Cyc2,
MtrB, OmcB), reductive dehalogenation (RdhA), photophosphorylation
(PsaA, PsbA, energy-converting microbial rhodopsin), methane cycling
(McrA, MmoA, PmoA), hydrogen cycling (large subunit of NiFe-, FeFe-, and
Fe-hydrogenases), formate oxidation (FdhA), carbon monoxide oxidation
(CoxL, CooS), fumarate reduction (FrdA), arsenic cycling (ARO, ArsC),
selenium cycling (YgfK), and carbon fixation (RbcL, AcsB, AclB, Mcr, HbsT,
HbsC) [67–69]. Results were further filtered based on an identity threshold
of either 80% (PsaA), 75% (HbsT), 70% (AtpA, PsbA, ARO, YgfK), 60% (NuoF,
RbcL, CoxL, AmoA, NxrA, MmoA, FeFe-hydrogenase, group 4 NiFe-
hydrogenase), or 50% (all other databases). Subgroup classification of
reads was based on the closest match to the sequences in databases. MtrB
in MAGs was screened additionally using hidden Markov models (HMM)
[70], with search cutoff scores as described previously [71]. Read counts to
each gene were normalized to reads per kilobase million (RPKM) by
dividing the actual read count by the total number of reads (in millions)
and then dividing by the gene length (in kilobases). In order to estimate
the gene abundance in the microbial community, high-quality unas-
sembled reads were also screened for the 14 universal single copy
ribosomal marker genes used in SingleM v.0.12.1 and PhyloSift [72]by
DIAMOND (query cover > 80%, bitscore > 40) and normalized as above.
Subsequently, the average gene copy number of each gene in the
community was inferred by dividing the read count for the gene (in RPKM)
by the mean of the read count of the 14 universal single copy ribosomal
marker genes (in RPKM). Linear regressions were performed to test how
the relative abundance of each gene varied relative to the disturbance
level of each individual sample (as inferred by their distance in cm relative
to the average depth of the dark sulfidic layer).
Sulfate reduction assays
Anoxic slurry experiments were performed to compare the rates of H
2
metabolism and sulfate reduction between the three sites. Sediments of
0–10 cm depth were collected from site A, B, and C for H
2
measurements and
sulfide measurements. Each slurry comprised a 160 mL serum vial containing
30 g of sieved sand (wet weight) and 70 mL of seawater (filtered on 0.45 µm
Whatman membrane filters). The serum vials were sealed with butyl rubber
stoppers and Wheaton closed-top seals. An autoclaved vial was used as the
control group. All vials were purged with high-purity helium gas and covered
with aluminium foil to create dark anoxic conditions. The headspace of the
vial was amended with 100 ppmv H
2
and, for the glucose treatment, 1 mM
glucose. All vials were incubated on a shaker (100rpm) at room temperature.
For H
2
measurements, a 2 mL subsample was collected from headspace
every 24h and analyzed using a VICI Trace Gas Analyzer Model 6K (Valco
Instruments Co. Inc., USA) fitted with a pulsed discharge helium ionization
detector that was configured and calibrated as previously described [73].
Three independent slurries were performed per treatment. Headspace H
2
mixing ratios were converted to dissolved H
2
concentrations in the slurries by
applying Henry’s law. After two weeks of incubation, DNA was extracted from
the sediments and subject to 16S rRNA gene amplicon sequencing as
described above. For free sulfide measurements, a total of 8 mL of seawater
was extracted from each slurry and filtered for spectrophotometric analysis
using the methylene blue method [37]. AVS measurements were performed
as described above.
Y.-J. Chen et al.
3
The ISME Journal
Carbon fixation assays
Shallow (0–10 cm) and deep (10–20 cm) sediments were collected from sites
A, B, and C for comparison of rates of dark carbon fixation. 30 g sediment
(wet weight) and 70 mL seawater in 160 mL serum vials were sealed (ambient
air headspace) with butyl rubber stoppers and Wheaton closed-top seals.
Slurries either remained unamended (native electron donors or were
supplemented with electron donors 200 µM sodium sulfide (Na
2
S·9H
2
O) or
200 µM ammonium chloride (NH
4
Cl). Vials containing autoclaved sediments
were used as controls. Radiolabelled sodium bicarbonate solution (NaH
14
CO
3
,
Perkin Elmer, 53 .1 mCi nmol
−1
) was administered to a concentration of 300
µM to each slurry. Slurries were incubated for 18 h in a light proof box (175
rpm, room temperature). After this time, the slurry was adjusted to pH 2 with
5 mL 1 M HCl to stop carbon fixation and acidify unfixed bicarbonate, before
centrifugation at 1000 × gfor 10 min. Overlying seawater was discarded
before sediment was left to dry in oven (80 °C) and then the acidification was
repeated. Sediment was weighed into scintillation vials, combined with
scintillation cocktail (EcoLume), and radioisotope analysis conducted using a
liquid scintillation spectrometer (Tri-Carb 2810 TR, Perkin Elmer). The
scintillation counts from autoclaved controls were subtracted from all
samples. Initial seawater (preserved with 6 % w/v HgCl
2
) was analyzed for
dissolved inorganic carbon (DIC analyzer, Apollo SciTech) and used in
combination with the specific radioactivity of the bicarbonate solution to
calculate the amount of
14
Cfixed per vial.
Denitrification, DNRA, and nitrification assays
Shallow (0–5 cm) and deep (20–25 cm) sediment samples were collected
from sites A, B, and C. Slurries containing 30 g sediment and 100 mL filtered
seawater were prepared in 160 mL serum vials, which were then crimp-
sealed with a butyl rubber septum. For denitrification and DNRA assays, each
slurry was amended with Na
15
NO
3
(>98%
15
N) to a final concentration of
1 mM and purged with argon to create anoxic conditions. The slurries were
continuously mixed on a shaker table at 125 rpm for the duration of the
incubation period. To determine denitrification rates, at each time point, 3 mL
headspace samples (containing
15
N-N
2
) were removed and replaced with
3mL argon.
15
N-N
2
was analyzed using a Sercon 20–22 continuous flow
isotope ratio mass spectrometer coupled to a gas chromatograph as
previously described [74]. To determine DNRA rates and for geochemical
analyses, 15 mL filtered seawater (containing
15
N-NH
4
+
) was removed and
replaced with 15 mL Ar-purged filtered seawater amended with 1 mM
Na
15
NO
3
. For iron analysis, 2 mL filtered samples were added to 0.5 mL of
10 mM ferrozine and analyzed as described [75]. For sulfide analy sis , 2 mL
samples were preserved with 10% v/v 28 mM Zn acetate and analyzed as
described above. 7.5 mL samples for
15
N-NH
4
+
analysis were transferred to a
12.5 mL gas tight exetainer and preserved with 250 μL ZnCl
2
.Thesamples
were purged with He to remove background N
2
before 200 μL alkaline
hypobromite was added to each sample to convert
15
N-NH
4
+
to
15
N-N
2
as
described [76]. To ensure quantitative conversion of NH
4
+
to N
2
,samples
were shaken at 130 rpm for 16h prior to instrumental analysis using the gas
chromatograph-isotope ratio mass spectrometer. Denitrification and DNRA
from the slurry experiments were estimated based on the accumulation of
15
N-N
2
and
15
NH
4
+
, respectively, over eight days after
15
N-NO
3
−
addition. To
measure nitrification, oxic slurries were amended with 50 µM NH
4
Cl. NO
2
−
and NO
3
−
concentrations in the nitrification assay were determined by the
Griess method [77] using a Lachat Quickchem 8000 Flow Injection Analyzer.
Rates of nitrification were calculated from linear regression of NO
2
−
and
NO
3
−
increase over time. The significance of differences in denitrification,
DNRA, and nitrification were tested by ANCOVA using R (v 4.0.3).
RESULTS
Geochemical stratification and microbial abundance are
higher in less disturbed permeable sediments
We confirmed that the three sediment sites (Fig. S1) differed in
levels of hydrodynamic disturbance by measuring mixing layer
depth and grain size distribution at three different sampling dates.
The depth and variability of the mixing zone were greatest for the
highly disturbed site as expected (Fig. 1a); the average depth of
the mixing zone (i.e., depth to the black sulfidic layer) shifted from
13.9 ± 5.2 cm for site A compared to 7.8 ± 3.5 cm for site B and
5.4 ± 1.1 cm for site C (p< 0.001, one-way ANOVA). This supports
previous inferences that greater hydrodynamically-driven pore-
water flow results in stronger sediment mixing and deeper oxygen
penetration overall, though in a spatiotemporally heterogeneous
manner [12,13]. In addition, grain size distribution varied as
anticipated. All three sediments were mainly comprised of sand
and gravel grains, with median grain size larger in site A (D
50
=
443 µm) than sites B (295 µm) and C (217 µm) (Table S1).
Altogether, these findings suggest that the highly exposed site
A is the most permeable, disturbed, and aerated of the sediments,
whereas the breakwater-protected site C is far less so and site B
has intermediate characteristics.
Consistent with our hypotheses, geochemical stratification was
more pronounced for the less disturbed sites. Reflecting this,
significant differences in the concentrations of sulfide (p< 0.05,
one-way ANOVA) and ammonium (p< 0.001, one-way ANOVA)
were detected between the sites. Free sulfide increased with
depth to high concentrations in site C (av. 321 μM) and moderate
concentrations in site B (av. 309 nM) (Fig. 1b), indicating activity of
dissimilatory sulfate reducers. In contrast, sulfide was below
detection limits at all sampled depths for site A (Fig. 1b), in
agreement with previous observations that hydrogenotrophic
sulfate reduction is inhibited and aerobic sulfide oxidation occurs
at rapid rates in highly disturbed sediments [21,23,28]. Acid-
volatile sulfide (AVS) measurements followed similar patterns
(Fig. 1c). In addition, ammonium accumulated in site C but to a
lesser extent at the other sites (Fig. 1d), suggesting ammonia
production from organic matter mineralization and DNRA
predominates over nitrification in more anoxic sediments. Linear
regression analysis confirmed concentrations of free sulfide (p<
0.0001), AVS (p=0.038), and ammonium (p< 0.0001) were
significantly negatively correlated with disturbance level, as
inferred by average distance of each sample (in cm) to the
sulfidic layer (Fig. S2).
Chlorophyll acontent, which indicates cyanobacterial and
eukaryotic photosynthesis, decreased with depth as expected
given variations in light exposure (Fig. 1e). Somewhat surprisingly,
site A contained twofold lower chlorophyll acontent than the
other sites across the depth profile (p< 0.001, one-way ANOVA),
suggesting regular temporal transitions from light oxic to dark
anoxic conditions in these sediments exclude photoautotrophs.
However, trends in the abundance and diversity of bacterial and
archaeal communities were complex and not clearly related to
differences in disturbance level. Microbial abundance (inferred
from 16S rRNA gene copy number; Table S1) was relatively high
across all sites (av. 4.1 × 10
8
copies per gram of sediment),
suggesting all sediment subsections harbor abundant commu-
nities well-adapted to their respective environmental conditions
and disturbance regimes. For unclear reasons, abundance was the
highest and most variable at site B, and the lowest and least
variable at site A (Fig. 1f). In contrast, bacterial and archaeal
diversity of each sample (Shannon index and estimated richness
based on 16S rRNA gene amplicon sequencing; Tables S2 & S4)
was highest for site C, lowest for site B, and considerably varied
within the depth profiles (Fig. 1g). Likewise, fewer amplicon
sequence variants (ASVs) were observed across the entire
sediment profiles for site B (13651, 7515, and 18288 ASVs
detected at sites A, B, and C respectively). No correlations were
observed between sample disturbance level with microbial
abundance (p=0.31) or Shannon index (p=0.36) based on linear
correlation analysis (Fig. S2). One explanation for these observa-
tions is that the intermediate level of disturbance in site B
increases carrying capacity by enabling extensive aerobic and
anaerobic growth of different community members on the various
electron donors available, in contrast to the relatively oxic site A
and anoxic site C, but in turn reduces diversity by intensifying
competition.
Community composition is highly differentiated by sediment
disturbance level
Bacterial and archaeal community composition was analyzed in
duplicate sediment cores from each site by 16S rRNA gene
Y.-J. Chen et al.
4
The ISME Journal
amplicon sequencing (Fig. 2). Across the 48 samples, 30,830 ASVs
were detected from 82 cultured or candidate phyla (Table S2).
Beta diversity analysis (weighted UniFrac) confirmed community
composition differed between sites (p< 0.01, PERMANOVA)
(Fig. 2c). Based on a PCoA visualization (Fig. 1d), samples most
strongly clustered by disturbance level along axis 1 (explaining
20.9% of the variation): community composition was similar
between the most disturbed layers of site B (0–4 cm) with those of
site A, and between the least disturbed layers of site B (10–16 cm)
with those of site C. For site A, communities exhibited weak
minimal depth stratification (Fig. 2b, 2d), consistent with the
observation hydrodynamic mixing selects for habitat generalists
[21]. However, contrary to our original hypothesis, depth
stratification was greatest for the moderately disturbed site B
(Fig. 2b, d); this reflects that large community shifts occur between
mixing and sulfidic zones, but there are many shared taxa within
each of these zones. Overall, these findings suggest differences in
hydrodynamic forcing between samples, due to the combination
of site and depth, controls microbial community assembly.
Disturbance level also strongly correlated with variations in
the relative abundance of most phyla, orders, families, and
genera between the sites (Fig. 2; Table S5). The most abundant
families were Flavobacteriaceae (Bacteroidota; av. 8.9 ± 4.8%
relative abundance) and Woeseiaceae (Proteobacteria; av. 6.9 ±
3.5%), consistent with reports that they are metabolically
flexible habitat generalists [19,21–23,25]; their abundance
was higher than all other families in all samples except six
from the deeper depths of site C (Fig. 2a, b; Table S5). Several
other families were also abundant and prevalent, especially in
more disturbed sites, including within phyla Bacteroidota
(Saprospiraceae, Cyclobacteriaceae), Proteobacteria (Rhodobac-
teraceae, Sedimenticolaceae, GCA-1735895, Helieaceae), and
Planctomycetota (Pirellulaceae, Phycisphaerae families SG8-4
and SM1A02) (Fig. 2a, b; Table S5). Conversely, lineages
reported to be obligate anaerobes were enriched in the least
disturbed sites and deepest sediment depths. These include
both named and novel families within the sulfate-reducing class
Desulfobacteria [78], as well as putative fermenters within
Bacteroidia, Anaerolineae, and Spirochaetia [79,80](Fig.2a, b;
TableS5).Althoughweobservedconsiderable differentiation in
the relative abundance of families, most families were still
present across multiple sites (average occupancy of 19 out of
48 samples) and 50 of them were shared across all sites,
including all the previously named families (Table S5). While
occupancy was lower at the ASV level (average occupancy of 3.9
out of 48 samples), 15 abundant ASVs were shared across all
samples (including several ASVs each within the families
Flavobacteriaceae, Woeseiaceae, and Desulfocapsaceae)
(Table S2). Overall, these findings support our hypothesis that
decreased disturbance select for more obligately anaerobic taxa
more typical of cohesive sediments, though suggest that
stratification is relatively modest given many bacteria can
adapt to a wide range of disturbance levels.
We performed statistical tests to analyze variations in the
relative abundance of the 35 most abundant named families
(Fig. 2a). 31 families significantly varied in relative abundance
between sites, and 20 significantly varied between depths (p<
0.05, one-way ANOVA) (Fig. S3; Table S5). Reflecting that
disturbance level is a nested variable, we also used linear
regressions to determine whether the relative abundance of each
Fig. 1 Differences in geochemical stratification and microbial abundance between sites. a Depth of the sediment mixing layer, based on
sampling across three dates. The average median grain size (D
50
; in mm) is tabulated, with full details in Table S1. Concentrations of (b) free
sulfide, (c) acid-volatile sulfide, and (d) ammonium are shown relative to sediment depth for each of the three sites. Also shown are (e)
chlorophyll acontent, (f) 16S rRNA gene copy number, and (g) alpha diversity based on Shannon index relative to sediment depth for each of
three sites. Dot points show averages and error bars show standard deviations from either three (b,d,e) or two (f,g) sediment cores. For AVS
measurements, one replicate was performed per slurry and hence error bars are not shown. One-way ANOVAs were used to test significant
differences in parameters between sites.
Y.-J. Chen et al.
5
The ISME Journal
family varies continuously with disturbance level, as inferred from
the distance of each sediment sample from the sulfidic layer
(Fig. 2a; Fig. S4). 18 families significantly increased (e.g.,
Flavobacteriaceae, Woeseiaceae) and 11 families significantly
decreased (e.g., Desulfosarcinaceae, Anaerolineaceae) with dis-
turbance level (p< 0.0001 for 24 families). However, it should be
noted that the coefficients of determination (R
2
) considerably
varied between these families (Fig. S4; Table S5), suggesting
disturbance level interacts with other factors to control their
relative abundance. Two families (Sulfurovaceae, Thiovulaceae)
were only abundant at the oxic-sulfidic interface of site C (Fig. S3),
in line with their reported for aerobic chemolithoautotrophic
growth on high levels of sulfide such as those present at this site
(Fig. 1b) [81]. Two other families (Acidobacteriota Mor1, Acid-
imicrobiia UBA5794) exhibited no clear patterns with disturbance
level (Fig. 2a). We additionally performed an analysis of composi-
tion of microbiomes (ANCOM) [50] to detect significant differences
in the relative abundance of different genera, including members
of the rare biosphere (Fig. 2d). Interestingly, archaea were among
the strongly differentiated: putative aerobic ammonium-oxidizing
Nitrosopumilaceae [82,83] were most enriched in the more
disturbed sites, whereas members of the putatively anaerobic
orders Thorarchaeaceae [84], Bathyarchaeia BA1 [85], and
Thermoplasmatota DHVEG-1 were enriched in the anoxic sites
(Fig. 2d; Fig. S3). Methanogens (e.g., Methanosarcinaceae) were
also enriched in less disturbed sediments (relative abundance of
0.56% at site C, 0.83% at 10–16 cm depth; Table S5). Consistent
with these observations, the average occupancy of archaeal taxa
(15.8 at family level, 2.6 at ASV level) was lower than for bacteria
(19.0 at family level, 4.0 at ASV level) (Tables S2 & S5).
Finally, we profiled composition of the entire microbial
community by sequencing metagenomes of each site at three
depths (0–2 cm, 6–8 cm, 14–16 cm; Table S6) and taxonomically
assigning ribosomal small subunit genes using phyloFlash
(Tables S7 & S8). Bacterial and archaeal community composition
was similar based on metagenomic and amplicon sequencing
(Fig. 2b, c; Table S8). Diverse eukaryotes were also detected in the
metagenomes, including putatively photoautotrophic diatoms,
dinoflagellates, charophytes, and euglenids; in support of the
chlorophyll adata (Fig. 1e; Table S1), their abundance relative to
bacteria and archaea decreased with sediment depth (Fig. S5;
Table S7).
Fig. 2 Differences in bacterial and archaeal community composition between sites. a Comparison of the relative abundance of the 35 most
abundant bacterial and archaeal families, based on 16S rRNA gene amplicon sequencing, of sediment cores sectioned into 2 cm slices.
Samples are arranged by disturbance level, based on their average distance (in cm) from the sulfidic layer. pvalues denote whether a linear
regression line of family relative abundance versus sample disturbance sample is significantly non-zero. * p< 0.05, ** p< 0.01, *** p< 0.001,
**** p< 0.0001. bPhylum-level bacterial and archaeal composition, based on 16S rRNA gene amplicon sequencing, of sediment cores
sectioned into 2 cm slices. Results are averaged based on two independent sediment cores. cPhylum-level bacterial and archaeal
composition, based on 16S rRNA gene reads in shotgun metagenomes, of a representative subset of samples from the sediment cores. dBeta
diversity of communities based on weighted Unifrac analysis of 16S rRNA gene amplicon sequencing data. Samples are visualized by principal
coordinates analysis (PCoA) with colors used to denote site and shapes used to denote sediment depth. The ellipse represents 95%
confidence intervals. eDifferential abundance of rare and common genera between sites A and C based on analysis of composition of
microbiomes (ANCOM). In the volcano plot, the F-score represents the log-fold change of the centered log ratio (clr) transformation, with
positive values representing taxa more abundant at site A and negative values indicating taxa more abundant in C. The W-statistic determines
whether differential abundance is significant. The false discovery rate (FDR) of 0.05 was controlled by Benjamini–Hochberg correction.
Y.-J. Chen et al.
6
The ISME Journal
Capacity for aerobic respiration, anaerobic respiration, and
fermentation varies with disturbance level
We analyzed the nine metagenomes to gain mechanistic insights
into observed differences in community structure and geochemical
parameters (Tables S6 & S9). The distribution of 50 metabolic
marker genes involved in energy acquisition, electron acceptor
utilization, and carbon fixation was determined across the
metagenomic short reads (Table S9). In line with previous studies
[21,23], the high abundance of various marker genes (Fig. 3)
suggests bacteria within the sediments can switch between
aerobic respiration (using terminal oxidases), anaerobic respiration
(via denitrification steps), and fermentation (using evolving
hydrogenases). There is also wide capacity for organic carbon,
sulfide, hydrogen, carbon monoxide, and formate oxidation (Fig. 3).
The relative abundance of several genes significantly varied with
disturbance level (p< 0.05, linear regression), as inferred from the
distance of each sediment sample from the sulfidic layer (Table S9).
These include various genes associated with anaerobic metabolism
that decreased with disturbance level. Notably, there was a
significant increase of at least fivefold in the marker genes for
sulfate reduction (dsrA,asrA), fermentation (FeFe-hydrogenases),
the Wood-Ljungdahl pathway (acsB,cooS), and methanogenesis
(mcrA) from the most to least disturbed samples (p< 0.05, linear
regression). In contrast, marker genes associated with aerobic
growth (coxA), archaeal nitrification (hbsT), and surprisingly
stepwise nitrite reduction (nirK,nirS) increased in abundance by
at least twofold from the least to most disturbed samples (p< 0.05,
linear regression) (Fig. 3; Fig. S6). Alongside the community analysis
(Fig. 2), these findings support our second hypothesis that
metabolically flexible habitat generalists are abundant throughout
the sediments, though metabolically constrained anaerobic
specialists are relatively enriched in less disturbed sediments.
Fig. 3 Metabolic capacity of microbial communities. Homology-based searches were used to detect key metabolic genes in nine
metagenomes and 169 derived metagenome-assembled genomes. The left heatmap shows the percentage of community members in each
metagenome predicted to encode each gene based on the short reads. Hits were normalized to gene length and single-copy ribosomal
marker genes. Samples are arranged by their disturbance level (bottom left panel), as inferred from their average distance (in cm) to the
sulfidic layer (more disturbed sites have more negative distances and less disturbed sites have more positive distances relative to sulfidic
layer). pvalues denote whether a linear regression line of gene community abundance versus sample disturbance sample is significantly non-
zero. * p< 0.05, ** p< 0.01, *** p< 0.001. The right-hand heatmap show the proportion of metagenome-assembled genomes from each family
that are predicted to encode each metabolic marker gene. The histogram (bottom right panel) shows the number of MAGs per family on a
logarithmic scale.
Y.-J. Chen et al.
7
The ISME Journal
In order to link community members to metabolic processes,
the metagenomes were individually assembled and co-assembled
based on sediment disturbance level. Binning yielded 169 high- or
medium-quality metagenome-assembly genomes (MAGs) [61]
from ten phyla and 46 families (Fig. 3; Table S10), including 18
of the 35 most abundant families (Fig. 2). Given the MAGs vary in
completeness, we inferred the potential lifestyles of the families
present based on what metabolic genes were present rather than
absent, in conjunction with referencing previous literature
regarding each lineage. Abundant habitat generalists, such as
Flavobacteriaceae, Woeseiaceae, Rhodobacteraceae, and Sapros-
piraceae, encode diverse metabolic genes; in line with previous
findings [21,23], they’re variably capable of shifting between
aerobic organotrophic respiration, and sulfide-dependent auto-
trophic growth under light oxic conditions to performing
hydrogenogenic fermentation and denitrification steps when
sediments become dark and anoxic (Fig. 3; Table S10). The MAGs
also support the divergent distributions of two archaeal families
highlighted by the ANCOM analysis (Fig. 2e): whereas the
Nitrosopumilaceae MAGs are likely to be aerobic specialists that
oxidize ammonia, Bathyarchaeia BA1 are predicted to be
anaerobic acetogens, in support of previous reports [21,82,86]
(Fig. 3; Table S10).
The MAG-level analysis rationalizes the strong differentiation in
sulfur metabolism between samples. In support of recent findings
[21,23,32,87], the MAGs suggest multiple families from the
phylum Desulfobacterota are the dominant sulfate reducers in
permeable sediments (Fig. 3b). Genes encoding enzymes mediat-
ing sulfite reduction (dissimilatory sulfite reductase; dsrA) were
detected in seven MAGs, usually in conjunction with those for
the oxidation of H
2
(group 1b [NiFe]-hydrogenases) and acetate
(acetyl-CoA decarbonylase/synthase; cooS/acsB). In support of the
high abundance of uncultured Desulfobacterales in the commu-
nity analysis (Fig. 2a), some Desulfobacterales MAGs are affiliated
with previously unreported families (Fig. 3; Table S10). The
terminal oxidases encoded by the MAGs differ in a manner
consistent with the contrasting distributions of Desulfobacterota
families (Fig. 2a; Table S5); families significantly enriched in the
least disturbed samples (e.g., Desulfosarcinaceae) encode cyto-
chrome bd and cbb
3
oxidases known to detoxify O
2
in sulfate-
reducing bacteria [88], whereas a family enriched in moderately
disturbed samples (Desulfocapsaceae) also encodes a potentially
growth-supporting cytochrome aa
3
oxidase. Dissimilatory sulfite
reductases were also encoded by two other phyla recently
inferred to be sulfate reducers, Gemmatimonadota (family
UBA6960) [89] and Bacteroidota (Ignavibacteriaceae) [90] (Fig. 3);
both families appear to be highly metabolically flexible and may
also mediate aerobic growth (Table S10). The capacity for aerobic
and anaerobic sulfide oxidation was more widespread, with 96
MAGs from 29 different families encoding sulfide-quinone
oxidoreductases, flavocytochrome coxidoreductases, or reverse
dissimilatory sulfite reductases (Table S10). All genes associated
with sulfide and thiosulfate oxidation were most abundant at site
C(p< 0.05) (Fig. 3), in line these differences are compatible with
the increased sulfide availability at this site (Fig. 1b, c) and the
enrichment of chemolithoautotrophic sulfide oxidizers such as
Sulfurovaceae (Fig. 2a); the two Sulfurovum MAGs confirm these
bacteria encode genes for aerobic sulfide oxidation (sqr),
thiosulfate oxidation (soxB), and carbon fixation via the reverse
tricarboxylic acid cycle (aclB).
Our results also add to growing evidence that inorganic
nitrogen metabolism in permeable sediments depends on
complex interspecies interactions. Genes associated with nitrifica-
tion (amoA,nxrA,hbsT) were in low abundance and, in agreement
with the geochemistry (Fig. 1d) and ANCOM (Fig. 2d) results,
largely confined to the relatively aerated site A (Fig. 3a). In
contrast, genes for denitrification and DNRA were abundant in the
metagenomic short reads (Fig. 3a) and encoded by 99 of the 169
MAGs (Fig. 3b), suggesting oxidized nitrogen compounds are
preferred electron acceptors in permeable sediments. Overall, the
genes for denitrification were abundant across the metagenomic
reads (av. 32% of community), whereas those associated with
DNRA were significantly lower (av. 8.8%). However, the ratio of
genes encoding denitrification-associated nitrite reductases (nirS,
nirK) compared to DNRA-associated nitrate reductases (nrfA)
exhibited a strong decrease relative to disturbance level (R
2
=
0.85, p=0.0005, linear regression) from the most disturbed
sample (site A 0–2 cm, ratio 5.9) to least disturbed sample (site
C14–16 cm, ratio 2.0). Based on the MAGs (Fig. 3; Table S10), nirS
and nirK genes were primarily associated with facultatively
anaerobic habitat generalists such as Woeseiaceae, Flavobacter-
iaceae, and Rhodobacteraceae, whereas nrfA is encoded by few
MAGs (Fig. 3), including relatively rare families Ignavibacteriaceae
(Bacteroidota) and Pontiellaceae (Verrucomicrobiota) [91]. Also
notable is the patchwork distribution of nitrate, nitrite, nitric oxide,
and nitrous oxide reductases between families, with no MAGs
(even those with >95% completeness) encoding complete
denitrification pathways (Table S6). The ecophysiological advan-
tages of such specialization in sediments that otherwise select for
metabolic versatility remain unclear, but these observations are
compatible with recent findings in permeable sediments and
other systems [21,25,92].
Fermentation and respiration are more coupled in less
disturbed sites
We performed a series of microcosm-based activities studies to
validate the above metagenome-based inferences. First, we
compared the coupling between fermentation and respiration
processes in surface sediments by comparing rates of H
2
production or consumption following a transition to anoxic
conditions. For site A, fermentation rates initially exceeded
respiration rates, resulting in a fourfold increase in H
2
concentra-
tions after 48 h (Fig. 4a; Table S11). In these sediments, net
hydrogenotrophic respiration was observed only after prolonged
anoxia, though sulfide levels remained below limits of detection
(Table S11; Fig. 4b). In contrast, fermentation and respiration
processes were tightly coupled in the less disturbed sites B and C,
and hence H
2
consumption was observed immediately following
the onset of anoxia (Fig. 4a). Reflecting these differences, high
levels of sulfide were detectable in sites B (free sulfide 1.2 µM, AVS
2.7 µM) and C (free sulfide 14.1 µM, AVS 3.9 µM) following the
incubations (Table S11; Fig. 4b), suggesting efficient coupling of
H
2
oxidation to sulfate reduction. Based on 16S rRNA gene
amplicon sequencing (Table S12), community composition of
sediments from site A also diverged from those of sites B and C
during the incubations (Fig. 4e). Notably, Desulfobacteraceae and
Desulfocapsaceae grew to become the dominant sulfate-reducing
bacteria in site A, compared to Desulfosarcinaceae in sites B and C
(Fig. 4d).
Addition of the fermentable carbon source glucose accentuated
differences between the three sites. H
2
accumulated to mixing
ratios above 1% in site A, though unexpectedly also reached high
levels in site C and remained low for the stably coupled site B.
Strong coupling between fermentation and respiration was only
observed after four and eight days of prolonged anoxia for sites C
and A respectively (Fig. 4a; Table S11). AVS measurements
confirmed all sites eventually mediated high rates of hydrogeno-
trophic sulfate reduction (Fig. 4b). Community composition
differences provide some rationale for these divergent responses.
Strong community shifts occurred following glucose spiking for
sites A and C (Fig. 4d, e), including rapid growth of facultative
fermenters (Flavobacteriaceae harboring group 3 NiFe-hydroge-
nases; Table S12) and obligate fermenters (e.g., Spirochaetaceae
likely encoding FeFe-hydrogenases [93]) relative to the three
above-mentioned hydrogenotrophic sulfate-reducing families
(encoding group 1 NiFe-hydrogenases; Table S12). In contrast, a
Y.-J. Chen et al.
8
The ISME Journal
much milder response was observed for site B (Fig. 4d, e). Overall,
these findings extend previous observations that hydrodynamic
disturbance causes sediments to become uncoupled, by selecting
for facultatively anaerobic fermentative bacteria and excluding
obligately anaerobic sulfate reducers [21,23]. Moreover, they
clearly confirm that differences in hydrodynamic gradients impact
not only community structure, but also biogeochemical reactions.
In addition, we tested the inferences from community and
functional profiling that sulfide-oxidizing chemolithoautotrophic
bacteria (e.g., Sulfurovaceae) are enriched in site C (Figs. 2&3)by
measuring rates of dark carbon fixation and assimilation under oxic
conditions. Consistent with predictions, rates in site C were on
average seven-fold higher compared to sites B and C (p< 0.0001,
one-way ANOVA) (Fig. 4c; Table S11). Rates were minimally affected
by supplementation with additional sulfide or ammonium (Fig. S7),
suggesting the high levels of sulfide and potentially other electron
donors already present in site C drive most fixation (Fig. 1b, c),
though some sulfide-dependent stimulation of carbon fixation was
observed in the sulfide-depleted surface sands of sites A and B
(Fig. S7). Overall, these assays provide supporting evidence that
sulfide accumulation due to activities of sulfate-reducing bacteria in
less disturbed sediments stimulates chemolithoautotrophic growth
when oxygen becomes available. However, tracing studies (e.g.,
stable isotope probing; SIP) would be required to confirm which
microorganisms are differentially active between the sites.
Ratios of denitrification to dissimilatory nitrate reduction to
ammonium are lower in less disturbed sites
Finally, we measured rates of nitrification, denitrification, and
DNRA rates between sites. As predicted from the metagenomic
analysis (Fig. 3) and in support of some previous findings [29,94],
nitrification rates were slower than denitrification and DNRA rates
(Fig. 5; Table S13). Nitrification rates in oxic microcosms containing
surface sands were highest for site A and negligible for site C (p=
Fig. 4 Differences in sulfate reduction and associated processes between sites. a Dissolved H
2
concentrations in anoxic slurries amended
with a headspace 100 ppmv H
2
. Samples either contained native organic carbon content or were spiked with 1 mM glucose. H
2
production
suggests hydrogenogenic fermentation, whereas H
2
consumption indicates hydrogenotrophic sulfate reduction. Symbols show means and
error bars show standard deviations from three independent replicates used per site. bAcid-volatile sulfide concentrations in sediments
before and after anoxic incubations in slurries in the presence and absence of 1 mM glucose (one replicate per site for 0 h, three replicates per
site for 336 h with error bars showing standard deviations). Free sulfide concentrations are shown in Table S11. cComparison of rates of dark
carbon fixation rates under oxic conditions and without electron donor spiking between surface and deep sands sampled from each site. Bars
show means and error bars show standard deviations from three independent slurries. Significant differences were measured by one-way
ANOVA. dHeatmap showing relative abundance of abundant microbial families, based on 16S rRNA gene amplicon sequencing, before and
after anoxic incubations in the presence and absence of 1 mM glucose. eBeta diversity of samples from d based on weighted Unifrac analysis
of 16S rRNA gene amplicon sequencing data and visualized by principal coordinates analysis (PCoA).
Y.-J. Chen et al.
9
The ISME Journal
0.002, ANCOVA; Fig. 5a) in agreement with differences in the
relative abundance of ammonia-oxidizing microorganisms (Nitro-
sopumilaceae; Fig. 2a) and genes (amoA; Fig. 3). NH
4
+
was
converted primarily to NO
2
−
in sites A and B, and some conversion
to NO
3
−
was observed after prolonged incubation at site A (Table
S13). This suggests ammonia oxidation exceeds nitrite oxidation.
Following oxic-anoxic transitions, it is likely that much of the
nitrite produced would be reduced through denitrification, in line
with previous reports of nitrifier denitrification in permeable
sediments [30,94]. Ammonium supplementation did not result in
a detectable enrichment in dark carbon fixation above back-
ground levels (Fig. S7).
Rates of denitrification and DNRA were measured by comparing
15
N
2
and
15
NH
4
+
production in anoxic microcosms spiked with
Na
15
NO
3
. As also predicted from the metagenomic analyses (Fig. 3),
denitrification occurred at rapid rates throughout the 216 h time-
course (Fig. 4c). Cumulative denitrification rates significantly varied
by site (p=0.0002, ANCOVA), with highest activities at sites B and C
and lowest activities in site A, in line with expectations based on
their relative disturbance levels. Rates were also higher in surface
compared to deeper sediments (p=0.002, ANCOVA), likely reflect-
ing their higher labile carbon and microbial abundance (Fig. 4c;
Table S13). In contrast, DNRA rates were one to two orders of
magnitude lower than denitrification (Fig. 4d) and again occurred at
Fig. 5 Differences in nitrogen-cycling processes between sites. a Cumulative nitrification in oxic slurries amended with 50 µM NH
4
Cl, as
measured by NO
2
−
and NO
3
−
production. bCumulative denitrification in anoxic slurries amended with 1 mM Na
15
NO
3
, as measured by
15
N-N
2
production. cCumulative dissimilator y nitrate reduction to ammonium (DNRA) in anoxic slurries amended with 1 mM Na
15
NO
3
, as measured
by
15
N-NH
4
+
production. Results are shown for sediments of 0–5 cm depth were used for nitrification measurements. Results are shown for
sediments of both 0–5 cm (shallow) and 20–25 cm (deep) depth for denitrification and DNRA measurements. Symbols show means and error
bars show standard deviations from three independent slurries. ANCOVA tests of linear regressions were used to test significant differences by
site (a,b,c) and by depth (b,c). dCorrelation of DNRA rate with dissolved iron content based on linear regression analysis.
Y.-J. Chen et al.
10
The ISME Journal
higher rates at less disturbed sites (p< 0.0001, ANCOVA) and
deeper sediment depths (p=0.036, ANCOVA) Table S13).
Reflecting shifts in the ratios of denitrification (nirK,nirS)to
DNRA (nrfA) genes (Fig. 3), the ratio of denitrification to DNRA was
approximately twofold lower for the less disturbed sites (p<
0.0001, ANCOVA) and increased during the time-courses (Fig. S8;
Table S13). It should be noted that these ratios may be skewed
given nitrate concentrations used in these assays are much higher
than typical in situ concentrations, though this is unlikely to affect
comparisons between sites. DNRA was correlated with dissolved
Fe concentrations (p=0.0007, ANCOVA), suggesting that it may
be mediated by lithotrophs in these sediments (Fig. 5d). No
accumulation of sulfide was detected in these slurries. This may
reflect two factors. First, given nitrate is a stronger electron
acceptor than sulfate, those sulfate reducers that harbor nitrate
reduction genes (Fig. 3) may preferentially use nitrate when it is
abundant. Second, in line with the widespread co-occurrence of
sulfide oxidation and nitrate reduction genes (Fig. 3), any sulfide
produced by sulfate reducers may be immediate reoxidised by
sulfide-oxidizing nitrate reducers. Together with the sulfate
reduction assays (Fig. 4), these findings further support the third
hypothesis that respiration processes associated with obligate
anaerobes are more active in less disturbed sediments.
DISCUSSION
Here we provide multifaceted evidence that hydrodynamic
disturbance is a major driver of microbial composition and
biogeochemical pathways in permeable sediments. In support of
our three original hypotheses, we show that geochemical
stratification, microbial specialism, and biogeochemical coupling
increases in less disturbed sites. Thus, as hydrodynamic forcing
decreases, conditions sufficiently stabilize for obligately anaerobic
specialists to increasingly outcompete the facultatively anaerobic
generalists that dominate at more disturbed sites. For example, we
provide multifaceted evidence that the presence and activity of
sulfate-reducing bacteria are differentiated by disturbance level: (i)
sulfide concentrations are higher in less disturbed sediments, (ii)
relative abundance of sulfate-reducing bacteria and their marker
genes is inversely correlated with sediment disturbance level, (iii)
MAGs of known and novel families of sulfate-reducing bacteria
were recovered from sediments with low and intermediate
disturbance, and (iv) anoxic carbon mineralization processes are
strongly coupled to sulfate reduction only in sites with low or
intermediate disturbance. Such findings confirm previous reports
that sediment disturbance and permeability influences commu-
nity composition [18,21], and supports our previously published
conceptual frameworks on the distributions and traits of habitat
generalists versus specialists in these environments [21–23,32].
Some findings were nevertheless contrary to our original
expectations. Most notably, though we observed strong differentia-
tion between the mixing and sulfidic zones, we observed that depth
stratification in microbial composition and activities was otherwise
modest even in sites with low disturbance. Many of the families and
even some individual ASVs were shared across all samples, albeit at
varying relative abundances. Also in contrast to highly stratified
cohesive sediments, common biogeochemical activities were
observed in samples from different sites and depths, albeit at
distinct rates. One explanation is that, unlike cohesive sediments,
there is still considerable mixing of compounds and microorganisms
through porewater advection even in less frequently disrupted
sands [2]. Indeed, single sand grains harbor diverse and abundant
communities that can be readily dispersed [19]. These observations
also likely reflect that many of the bacteria present can grow or
survive despite changes in resource availability and other physico-
chemical factors. In agreement with previous work [21–23,25], we
observed taxa such as Flavobacteriaceae, Woeseiaceae, and
Rhodobacteraceae possess sufficient metabolic flexibility to grow
in both light oxic and dark anoxic conditions, explaining their
abundance in even the least disturbed sands. However, even some
apparent anaerobes appear to possess considerable flexibility. For
example, we predict that the Desulfocapsaceae present across all
samples and enriched in sediments with intermediate disturbance
are either highly aerotolerant or potentially even capable aerobic
growth, especially given the MAG-level analysis shows they encode
cytochrome aa
3
oxidases.
In turn, our analyses suggest that disturbance operates as a
continuous variable in permeable sediments. Reflecting this, we
observed strong linear correlations between sediment disturbance
level with sulfide and ammonium concentrations, relative abun-
dance of microbial families as inferred from 16S rRNA gene
amplicon sequencing, and abundance of metabolic genes as
inferred from metagenomic short reads. The most abundant
families were present in all samples, though the habitat generalists
increased in relative abundance with disturbance level and the
relative anaerobic specialists showed the opposite trend, in a
manner compatible with their metabolic capabilities. Only a few
families, for example the key aerobic specialist Nitrosopumilaceae
[21] or the sulfide-oxidizing chemolithoautotroph Sulfurovaceae
[81], displayed contrary trends. Assuming high levels of dispersal in
these ecosystems, we predict that disturbance level interacts with
other deterministic factors to shape growth and survival dynamics
of each taxon, thereby controlling their relative abundance in a
potentially predictable way. However, it should be noted that
disturbance level did not explain the abundance and diversity of
the overall microbial community, suggesting other as-yet-
unrecognized overarching controls also operate.
By combining genome-resolved metagenomics with microcosm-
based biogeochemical assays, we additionally developed under-
standing of the mediators of biogeochemical processes in permeable
sediments. Collectively, the data suggests that fermentation is the
dominant process of anoxic carbon mineralization in highly disturbed
sedimentsaspreviouslypredicted[23,28],butthattheclassicalsuite
of heterotrophic mineralization reactions (e.g., denitrification, DNRA,
and sulfate reduction) occur at higher rates in less disturbed
sediments. These differences likely reflect the complex effects of
hydrodynamic disturbance on the relative distributions, gene
expression, and enzymatic activities of habitat generalists and
anaerobic specialists. Denitrification is a more active process than
DNRA and appears to be mediated by more community members, in
support of earlier findings [27,74]. However, given the MAGs only
encode partial denitrification pathways, this process is likely to
depend on extensive metabolic interactions between species for
unclear reasons. It is proposed that organism specialization for each
step in the denitrification pathway is more thermodynamically
efficient than one organism mediating the multiple steps [95]. In
combination, the stepwise pathways and diverse mediators of
nitrification, denitrification, and DNRA suggest complex ecological
interactions control nitrogen cycling in permeable sediments. Our
findings also provide a deeper understanding of the diversity and
capabilities of sulfate-reducing bacteria present in these environ-
ments; marker genes for sulfate reduction were detected in
differentially distributed lineages of Desulfobacterota, together with
surprisingly Gemmatimonadota and Bacteroidota, suggesting com-
plex controls on sulfate reduction. Low levels of acetogenic and
methanogenic archaea were also observed in the least disturbed
sediments, though it remains to be confirmed if they are active. Our
analyses also highlight the key roles of lithoheterotrophy and
lithoautotrophy in permeable sediment function, as suggested by
the wide distribution of genes for sulfide, hydrogen, and carbon
monoxide oxidation and the diverse pathways for carbon fixation.
Future studies should focus on integrating culture-based and
culture-independent approaches to gain a deeper perspective on
the ecophysiology of the major taxa in these ecosystems. We
recommend targeted tracing studies (e.g., DNA-SIP) and metatran-
scriptomics to better link biogeochemical activities to mediators;
Y.-J. Chen et al.
11
The ISME Journal
similar approaches have previously helped to confirm taxa such as
Woeseiaceae, Desulfocapsaceae, and uncultured Desulfobacterales
are key mediators of sulfur and carbon cycling in permeable
sediments [22,24,32,87,96]. In addition, more extensive cultivation-
based studies are needed to validate metabolic predictions, for
example to confirm the metabolic versatility of Woeseiaceae, test
the potential for aerobic growth of Desulfocapsaceae, and extend
the capacity for sulfate reduction and DNRA to other phyla. Such
approaches are also important to understand physiological
responses to changes in resource availability (e.g., oxic-anoxic
transitions). Further work is needed to understand the dynamics and
drivers of temporal turnover in permeable sediments, including
seasonality, and to what extent this is affected by disturbance level.
Wider spatial and temporal sampling is also important to under-
stand what drives the unexplained differences in microbial
abundance and biodiversity between samples, and resolve which
other factors interact with disturbance level to control microbial
composition and biogeochemical activities.
DATA AVAILABILITY
All amplicon sequencing data, raw metagenomes, and metagenome-assembled
genomes will be deposited to the NCBI Sequence Read Archive under the BioProject
accession number PRJNA623061.
REFERENCES
1. Raffaell i D, Hawkins SJ. Intertidal ecology. Berlin, Germany: Springer Science &
Business Media; 2012.
2. Huettel M, Berg P, Kostka JE. Benthic exchange and biogeochemical cycling in
permeable sediments. Ann Rev Mar Sci. 2014;6:23–51.
3. Canfield D, Kristensen E, Thamdrup B. Aquatic geomicrobiology. Amsterdam,
Netherlands: Elsevier; 2005.
4. Hall SJ. The continental shelf benthic ecosystem: current status, agents for
change and future prospects. Environ Conserv. 2002;29:350–74.
5. Boudreau BP, Huettel M, Forster S, Jahnke RA, McLachlan A, Middelburg JJ, et al.
Permeable marine sediments: overturning an old paradigm. EOS, Trans Am
Geophys Union. 2001;82:133–6.
6. Santos IR, Eyre BD, Huettel M. The driving forces of porewater and groundwater
flow in permeable coastal sediments: a review. Estuar Coast Shelf Sci. 2012;98:1–15.
7. Meysman FJR, Galaktionov OS, Cook PLM, Janssen F, Huettel M, Middelburg JJ.
Quantifying biologically and physically induced flow and tracer dynamics in
permeable sediments. Biogeosciences. 2007;4:627–46.
8. Huettel M, Ziebis W, Forster S. Flow‐induced uptake of particulate matter in
permeable sediments. Limnol Oceanogr. 1996;41:309–22.
9. Huettel M, Ziebis W, Forster S, Luther, Iii GW. Advective transport affecting metal
and nutrient distributions and interfacial fluxes in permeable sediments. Geo-
chim Cosmochim Acta. 1998;62:613–31.
10. Huettel M, Rusch A. Transport and degradation of phytoplankton in permeable
sediment. Limnol Oceanogr. 2000;45:534–49.
11. Precht E, Franke U, Polerecky L, Huettel M. Oxygen dynamics in permeable sedi-
ments with wave‐driven pore water exchange. Limnol Oceanogr. 2004;49:693–705.
12. Glud RN. Oxygen dynamics of marine sedimen ts. Mar Biol Res. 2008;4:243–89.
13. Cook PL, Frank W, Glud R, Felix J, Markus H. Benthic solute exchange and carbon
mineralization in two shallow subtidal sandy sediments: Effect of advective pore‐
water exchange. Limnol Oceanogr. 2007;52:1943–63.
14. Huettel M, Røy H, Precht E, Ehrenhauss S. Hydrodynamical impact on biogeo-
chemical processes in aquatic sediments. The Interactions Between Sediments
and Water. Berlin, Germany: Springer; 2003. p. 231–6.
15. Gobet A, Böer SI, Huse SM, Van Beusekom JEE, Quince C, Sogin ML, et al. Diversity
and dynamics of rare and of resident bacterial populations in coastal sands. ISME
J. 2012;6:542–53.
16. Böer SI, Arnosti C, Van Beusekom JEE, Boetius A. Temporal variations in microbial
activities and carbon turnover in subtidal sandy sediments. Biogeosciences.
2009;6:1149–65.
17. Hunter EM, Mills HJ, Kostka JE. Microbial community diversity associated with
carbon and nitrogen cycling in permeable shelf sediments. Appl Environ
Microbiol. 2006;72:5689–701.
18. Probandt D, Knittel K, Tegetmeyer HE, Ahmerkamp S, Holtappels M, Amann R.
Permeability shapes bacterial communities in sublittoral surface sediments.
Environ Microbiol. 2017;19:1584–99.
19. Prob andt D, Eickhorst T, Ellrott A, Amann R, Knittel K. Microbial life on a sand
grain: from bulk sediment to single grains. ISME J. 2017;12:623–33.
20. Rodriguez-r LM, Overholt WA, Hagan C, Huettel M, Kostka JE, Konstantinidis KT.
Microbial community successional patterns in beach sands impacted by the
Deepwater Horizon oil spill. ISME J. 2015;9:1928–40.
21. Chen YJ, Leung PM, Wood JL, Bay SK, Hugenholtz P, Kessler AJ, et al. Metabolic
flexibility allows bacterial habitat generalists to become dominant in a frequently
disturbed ecosystem. ISME J. 2021. https://doi.org/10.1038/s41396-021-00988-w.
22. Mußmann M, Pjevac P, Krüger K, Dyksma S. Genomic repertoire of the Woe-
seiaceae/JTB255, cosmopolitan and abundant core members of microbial com-
munities in marine sediments. ISME J. 2017;11:1276–81.
23. Kessler AJ, Chen Y-J, Waite DW, Hutchinson T, Koh S, Popa ME, et al. Bacterial
fermentation and respiration processes are uncoupled in permeable sediments.
Nat Microbiol. 2019;4:1014–23.
24. Lenk S, Arnds J, Zerjatke K, Musat N, Amann R, Mußmann M. Novel groups of
Gammaproteobacteria catalyse sulfur oxidation and carbon fixation in a coastal,
intertidal sediment. Environ Microbiol. 2011;13:758–74.
25. Marchant HK, Tegetmeyer HE, Ahmerkamp S, Holtappels M, Lavik G, Graf J, et al.
Metabolic specialization of denitrifiers in permeable sediments controls N
2
O
emissions. Environ Microbiol. 2018;20:4486–502.
26. Gao H, Schreiber F, Collins G, Jensen MM, Kostka JE, Lavik G, et al. Aerobic
denitrification in permeable Wadden Sea sediments. ISME J. 2010;4:417–26.
27. Marchant HK, Ahmerkamp S, Lavik G, Tegetmeyer HE, Graf J, Klatt JM, et al.
Denitrifying community in coastal sediments performs aerobic and anaerobic
respiration simultaneously. ISME J. 2017;11:1799.
28. Bourke MF, Marriott PJ, Glud RN, Hasler-Sheetal H, Kamalanathan M, Beardall J,
et al. Metabolism in anoxic permeable sediments is dominated by eukaryotic
dark fermentation. Nat Geosci. 2017;10:30–35.
29. Marchant HK, Lavik G, Holtappels M, Kuypers MMM. The fate of nitrate in inter-
tidal permeable sediments. PLoS ONE. 2014;9:e104517.
30. Marchant HK, Holtappels M, Lavik G, Ahmerkamp S, Winter C, Kuypers MMM.
Coupled nitrification–denitrification leads to extensive N loss in subtidal
permeable sediments. Limnol Oceanogr. 2016;61:1033–48.
31. Evrard V, Glud RN, Cook PLM. The kinetics of denitrification in permeable sedi-
ments. Biogeochemistry. 2013;113:563–72.
32. Dyksma S, Pjevac P, Ovanesov K, Mussmann M. Evidence for H2 consumption by
uncultured Desulfobacterales in coastal sediments. Environ Microbiol. 2018;20:450–61.
33. Saad S, Bhatnagar S, Tegetmeyer HE, Geelhoed JS, Strous M, Ruff SE. Transient
exposure to oxygen or nitrate reveals ecophysiology of fermentative and sulfate‐
reducing benthic microbial populations. Environ Microbiol. 2017;19:4866–81.
34. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 year s of image
analysis. Nat Methods. 2012;9:671–5.
35. Prodger S, Russell P, Davidson M. Grain‐size distributions on high‐energy sandy
beaches and their relation to wave dissipation. Sedimentology.
2017;64:1289–302.
36. Ferguson RI, Church M. A simple universal equation for grain settling velocity. J
Sediment Res. 2004;74:933–7.
37. Fonselius S, Dyrssen D, Yhlen B. Determination of hydrogen sulphide. Methods of
Seawater Analysis. Third Ed. Hoboken, New Jersey, USA: Wiley; 2007. p. 91–100.
38. Simpson SL. A rapid screening method for acid‐volatile sulfide in sediments.
Environ Toxicol Chem Int J. 2001;20:2657–61.
39. Federation WE. Association APH Standard methods for the examination of water
and wastewater. Am Public Heal Assoc Washington, DC, USA 2005.
40. Lorenzen CJ. Determination of chlorophyll and pheo‐pigments: spectro-
photometric equations 1. Limnol Oceanogr. 1967;12:343–6.
41. Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone CA, Turnbaugh PJ,
et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences
per sample. Proc Natl Acad Sci USA. 2011;108:4516–22.
42. Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina
sequence data. Bioinformatics. 2014;30:2114–20.
43. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. DADA2:
high-resolution sample inference from Illumina amplicon data. Nat Methods.
2016;13:581–3.
44. Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, Al-Ghalith GA, et al.
Reproducible, interactive, scalable and extensible microbiome data science using
QIIME 2. Nat Biotechnol. 2019;37:852–7.
45. Parks DH, Chuvochina M, Waite DW, Rinke C, Skarshewski A, Chaumeil P-A, et al. A
standardized bacterial taxonomy based on genome phylogeny substantially
revises the tree of life. Nat Biotechnol. 2018;36:996–1004.
46. Scanlon DO, Morgan BJ, Watson GW. Modeling the polaronic nature of p -type
defects in Cu2 O: The failure of GGA and GGA+U. J Chem Phys. 2009;131:221.
47. Shetty SA, Lahti L. Microbiome data science. J Biosci. 2019;44:11 5.
48. Weiss S, Xu ZZ, Peddada S, Amir A, Bittinger K, Gonzalez A, et al. Normalization
and microbial differential abundance strategies depend upon data character-
istics. Microbiome. 2017;5:1–18.
49. Lozupone C, Lladser ME, Knights D, Stombau gh J, Knight R. UniFrac: an effective
distance metric for microbial community comparison. ISME J. 2011;5:169–72.
Y.-J. Chen et al.
12
The ISME Journal
50. Mandal S, Van Treuren W, White RA, Eggesbø M, Knight R, Peddada SD. Analysis
of composition of microbiomes: a novel method for studying microbial com-
position. Micro Ecol Health Dis. 2015;26:27663.
51. Li DH, Luo RB, Liu CM, Leung CM, Ting HF, Sadakane K, et al. MEGAHITv1.0: A fast
and scalable metagenome assembler driven by advanced methodologies and
community practices. Methods. 2016;102:3–11.
52. Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nat
Methods. 2012;9:357.
53. Miller IJ, Rees ER, Ross J, Miller I, Baxa J, Lopera J, et al. Autometa: Automated
extraction of microbial genomes from individual shotgun metagenomes. Nucleic
Acids Res. 2019;47:e57.
54. Alneberg J, Bjarnason BS, De Bruijn I, Schirmer M, Quick J, Ijaz UZ, et al. Binning
metagenomic contigs by coverage and composition. Nat Methods. 2014;11:1144.
55. Wu Y-W, Simmons BA, Singer SW. MaxBin 2.0: an automated binning algorithm to
recover genomes from multiple metagenomic datasets. Bioinformatics.
2015;32:605–7.
56. Kang D, Li F, Kirton ES, Thomas A, Egan RS, An H, et al. MetaBAT 2: an adaptive
binning algorithm for robust and efficient genome reconstruction from meta-
genome assemblies. PeerJ. 2019;7:e7359.
57. Sieber CMK, Probst AJ, Sharrar A, Thomas BC, Hess M, Tringe SG, et al. Recovery of
genomes from metagenomes via a dereplication, aggregation and scoring
strategy. Nat Microbiol. 2018;3:836–43.
58. Parks DH, Rinke C, Chuvochina M, Chaumeil P-A, Woodcroft BJ, Evans PN, et al.
Recovery of nearly 8,000 metagenome-assembled genomes substantially
expands the tree of life. Nat Microbiol. 2017;2:1533–42.
59. Olm MR, Brown CT, Brooks B, Banfield JF. dRep: a tool for fast and accurate
genomic comparisons that enables improved genome recovery from metagen-
omes through de-replication. ISME J. 2017;11:2864–68.
60. Parks DH, Imelfort M, Skennerton CT, Hugenholtz P, Tyson GW. CheckM: assessing
the quality of microbial genomes recovered from isolates, single cells, and
metagenomes. Genome Res. 2015;25:1043–55.
61. Bowers RM, Kyrpides NC, Stepanauskas R, Harmon-Smith M, Doud D, Reddy TBK,
et al. Minimum information about a single amplified genome (MISAG) and a
metagenome-assembled genome (MIMAG) of bacteria and archaea. Nat Bio-
technol. 2017;35:725–31.
62. Hyatt D, Chen G-L, LoCascio PF, Land ML, Larimer FW, Hauser LJ. Prodigal: pro-
karyotic gene recognition and translation initiation site identification. BMC
Bioinforma. 2010;11:119.
63. Gruber-Vodicka HR, Seah BKB, Pruesse E. PhyloFlash: rapid small-subunit rRNA
profiling and targeted assembly from metagenomes. mSystems. 2019;5:
e00920–20.
64. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA
ribosomal RNA gene database project: improved data processing and web-based
tools. Nucleic Acids Res. 2012;41:D590–D596.
65. Bay SK, Dong X, Bradley JA, Leung PM, Grinter R, Jirapanjawat T, et al. Trace gas
oxidizers are widespread and active members of soil microbial communities. Nat
Microbiol. 2021;6:246–56.
66. Buchfink B, Xie C, Huson DH. Fast and sensitive protein alignment using DIA-
MOND. Nat Methods. 2014;12:59–60.
67. Greening C, Geier R, Wang C, Woods LC, Morales SE, McDonald MJ, et al. Diverse
hydrogen production and consumption pathways influence methane production
in ruminants. ISME J. 2019;13:2617–32.
68. Søndergaard D, Pedersen CNS, Greening C. HydDB: a web tool for hydrogenase
classification and analysis. Sci Rep. 2016;6:34212.
69. Cordero PRF, Bayly K, Leung PM, Huang C, Islam ZF, Schitt enhelm RB, et al.
Atmospheric carbon monoxide oxidation is a widespread mechanism supporting
microbial survival. ISME J. 2019;13:2868–81.
70. Eddy SR. Accelerated profile HMM searches. PLoS Comput Biol. 2011;7:e1002195.
71. Anantharaman K, Brown CT, Hug LA, Sharon I, Castelle CJ, Probst AJ, et al.
Thousands of microbial genomes shed light on interconnected biogeochemical
processes in an aquifer system. Nat Commun. 2016;7:13219.
72. Darling AE, Jospin G, Lowe E, Matsen FA IV, Bik HM, et al. PhyloSift: phylogenetic
analysis of genomes and metagenomes. PeerJ. 2014;2:e243.
73. Islam ZF, Cordero PRF, Feng J, Chen Y-J, Bay SK, Gleadow RM, et al. Two Chlor-
oflexi classes independently evolved the ability to persist on atmospheric
hydrogen and carbon monoxide. ISME J. 2019;13:1801–13.
74. Kessler AJ, Roberts KL, Bissett A, Cook PLM. Biogeochemical controls on the
relative importance of denitrification and dissimilatory nitrate reduction to
ammonium in estuaries. Glob Biogeochem Cycles. 2018;32:1045–57.
75. Viollier E, Inglett PW, Hunter K, Roychoudhury AN, Van Cappellen P. The ferrozine
method revisited: Fe (II)/Fe (III) determination in natural waters. Appl Geochem.
2000;15:785–90.
76. Risgaard-Petersen N, Rysgaard S. Nitrate reduction in sediments and waterlogged
soil measured by
15
N techniques. Methods in applied soil microbiology and
biochemistry. Cambridge, Massachusetts, USA: Academic Press; 1995. p. 287–95.
77. American Public Health Association. Standard methods for the examination of
water and wastewater. Washington, DC: American Public Health Association; 1992.
78. Landwig MV, De Anda V, Dombrowski N, Seitz KW, Rambo IM, Greening C, et al.
Large scale protein level comparison of Deltaproteobacteria reveals cohesive
metabolic groups. ISME J. https://doi.org/10.1038/s41396-021-01057-y.
79. Thauer RK, Jungermann K, Decker K. Energy conservation in chemotrophic
anaerobic bacteria. Bacteriol Rev. 1977;41:100–80.
80. Dong X, Greening C, Rattray JE, Chakraborty A, Chuvochina M, Mayumi D, et al.
Metabolic potential of uncultured bacteria and archaea associated with petro-
leum seepage in deep-sea sediments. Nat Commun. 2019;10:1816.
81. Waite DW, Vanwonterghem I, Rinke C, Parks DH, Zhang Y, Takai K, et al. Com-
parative genomic analysis of the class Epsilonproteobacteria and proposed
reclassification to Epsilonbacteraeota (phyl. nov.). Front Microbiol. 2017;8:682.
82. Walker CB, De La Torre JR, Klotz MG, Urakawa H, Pinel N, Arp DJ, et al. Nitroso-
pumilus maritimus genome reveals unique mechanisms for nitrification and
autotrophy in globally distributed marine crenarchaea. Proc Natl Acad Sci.
2010;107:8818–23.
83. Kitzinger K, Marchant HK, Bristow LA, Herbold CW, Padilla CC, Kidane AT, et al.
Single cell analyses reveal contrasting life strategies of the two main nitrifiers in
the ocean. Nat Commun. 2020;11:767.
84. Spang A, Stairs CW, Dombrowski N, Eme L, Lombard J, Caceres EF, et al. Proposal
of the reverse flow model for the origin of the eukaryotic cell based on com-
parative analyses of Asgard archaeal metabolism. Nat Microbiol. 2019;4:1138–48.
85. Castelle CJ, Wrighton KC, Thomas BC, Hug LA, Brown CT, Wilkins MJ, et al.
Genomic expansion of domain archaea highlights roles for organisms from new
phyla in anaerobic carbon cycling. Curr Biol. 2015;25:690–701.
86. He Y, Li M, Perumal V, Feng X, Fang J, Xie J, et al. Genomic and enzymatic
evidence for acetogenesis among multiple lineages of the archaeal phylum
Bathyarchaeota widespread in marine sediments. Nat Microbiol. 2016;1:16035.
87. Dyksma S, Lenk S, Sawicka JE, Mußmann M. Uncultured Gammaproteobacteria
and Desulfobacteraceae account for major acetate assimilation in a coastal
marine sediment. Front Microbiol. 2018;9:3124.
88. Ramel F, Amrani A, Pieulle L, Lamrabet O, Voordouw G, Seddiki N, et al.
Membrane-bound oxygen reductases of the anaerobic sulfate-reducing Desul-
fovibrio vulgaris Hildenborough: roles in oxygen defence and electron link with
periplasmic hydrogen oxidation. Microbiology. 2013;159:2663–73.
89. Baker BJ, Lazar CS, Teske AP, Dick GJ. Genomic resolution of linkages in carbon,
nitrogen, and sulfur cycling among widespread estuary sediment bacteria.
Microbiome. 2015;3:14.
90. Anantharaman K, Hausmann B, Jungbluth SP, Kantor RS, Lavy A, Warren LA, et al.
Expanded diversity of microbial groups that shape the dissimilatory sulfur cycle.
ISME J. 2018;12:1715–28.
91. van Vliet DM, Lin Y, Bale NJ, Koenen M, Villanueva L, Stams AJM, et al. Pontiella
desulfatans gen. nov., sp. nov., and Pontiella sulfatireligans sp. nov., two marine
anaerobes of the Pontiellaceae fam. nov. producing sulfated glycosaminoglycan-
like exopolymers. Microorganisms. 2020;8:920.
92. Kuypers MMM, Marchant HK, Kartal B. The microbial nitrogen-cycling network.
Nat Rev Microbiol. 2018;16:263–76.
93. Greening C, Biswas A, Carere CR, Jackson CJ, Taylor MC, Stott MB, et al. Genomic
and metagenomic surveys of hydrogenase distribution indicate H
2
is a widely
utilised energy source for microbial growth and survival. ISME J. 2016;10:761–77.
94. Kessler AJ, Glud RN, Cardenas MB, Cook PLM. Transport zonation limits coupled
nitrification-denitrification in permeable sediments. Environ Sci Technol.
2013;47:13404–11.
95. Zumft WG. Cell biology and molecular basis of denitrification. Microbiol Mol Biol
Rev. 1997;61:533–616.
96. Dyksma S, Bischof K, Fuchs BM, Hoffmann K, Meier D, Meyerdierks A, et al.
Ubiquitous Gammaproteobacteria dominate dark carbon fixation in coastal
sediments. ISME J. 2016;10:1939–53.
ACKNOWLEDGEMENTS
This study was supported by ARC Discovery Project grants (DP180101762 awarded to
PLMC and CG; DP210101595 awarded to PLMC, CG, and WWW), an ARC DECRA
Fellowship (DE170100310; salary for CG), an NHMRC EL2 Fellowship (APP1178715;
salary for CG), PhD scholarships from Monash University and the Taiwan Ministry of
Education (Y-JC), and an Australian Government Research Training Scholarship
(awarded to PML). We thank S. Kessler for technical assistance and S. Bay for
analytical advice.
AUTHOR CONTRIBUTIONS
CG and PLMC conceived and supervised this study. CG, PLMC and Y-JC designed
experiments. Y-JC, PLMC, AJK and CG conducted fieldwork.VEandY-JCconducted
Y.-J. Chen et al.
13
The ISME Journal
geochemical analysis. Y-JC performed DNA extractions. Y-JC, CG and PML analyzed
community composition. PML assembled and analyzed metagenomes. Y-JC and VE
measured H
2
and sulfide metabolism. TH and AJK measured carbon fixation. AJK, WWW
andY-JCmeasureddenitrification, DNRA, and nitrification. YJC and PML prepared figures,
Y-JC and CG wrote the paper, and CG edited the paper with input from all authors.
COMPETING INTERESTS
The authors declare no competing interests.
ADDITIONAL INFORMATION
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41396-021-01111-9.
Correspondence and requests for materials should be addressed to Chris Greening.
Reprints and permission information is available at http://www.nature.com/
reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder. To view a copy of this license, visit http://creativecommons.
org/licenses/by/4.0/.
© The Author(s) 2021
Y.-J. Chen et al.
14
The ISME Journal
