Experimentally-validated correlation analysis reveals new anaerobic methane oxidation partnerships with consortium-level heterogeneity in diazotrophy

The ISME Journal (2021) 15:377

–

396

https://doi.org/10.1038/s41396-020-00757-1

ARTICLE

Experimentally-validated correlation analysis reveals new anaerobic

methane oxidation partnerships with consortium-level

heterogeneity in diazotrophy

Kyle S. Metcalfe

●

Ranjani Murali

●

Sean W. Mullin

●

Stephanie A. Connon

●

Victoria J. Orphan

Received: 10 April 2020 / Revised: 28 July 2020 / Accepted: 21 August 2020 / Published online: 15 October 2020

Abstract

Archaeal anaerobic methanotrophs (

“

ANME

”

) and sulfate-reducing Deltaproteobacteria (

“

SRB

”

) form symbiotic

multicellular consortia capable of anaerobic methane oxidation (AOM), and in so doing modulate methane

fl

ux from

marine sediments. The speci

fi

city with which ANME associate with particular SRB partners in situ, however, is poorly

understood. To characterize partnership speci

fi

city in ANME-SRB consortia, we applied the correlation inference technique

SparCC to 310 16S rRNA amplicon libraries prepared from Costa Rica seep sediment samples, uncovering a strong positive

correlation between ANME-2b and members of a clade of Deltaproteobacteria we termed SEEP-SRB1g. We con

fi

rmed this

association by examining 16S rRNA diversity in individual ANME-SRB consortia sorted using

fl

ow cytometry and by

imaging ANME-SRB consortia with

fl

uorescence in situ hybridization (FISH) microscopy using newly-designed probes

targeting the SEEP-SRB1g clade. Analysis of genome bins belonging to SEEP-SRB1g revealed the presence of a complete

nifHDK

operon required for diazotrophy, unusual in published genomes of ANME-associated SRB. Active expression of

nifH

in SEEP-SRB1g within ANME-2b

—

SEEP-SRB1g consortia was then demonstrated by microscopy using

hybridization chain reaction (HCR-) FISH targeting

nifH

transcripts and diazotrophic activity was documented by FISH-

nanoSIMS experiments. NanoSIMS analysis of ANME-2b

—

SEEP-SRB1g consortia incubated with a headspace containing

and

revealed differences in cellular

N-enrichment between the two partners that varied between individual

consortia, with SEEP-SRB1g cells enriched in

N relative to ANME-2b in one consortium and the opposite pattern

observed in others, indicating both ANME-2b and SEEP-SRB1g are capable of nitrogen

fi

xation, but with consortium-

speci

fi

c variation in whether the archaea or bacterial partner is the dominant diazotroph.

Introduction

The partnership between anaerobic, methanotrophic Archaea

(ANME) and their associated sulfate-reducing bacteria

(SRB) is one of the most biogeochemically-important sym-

bioses in the deep-sea methane cycle [

]. As a critical

component of methane seep ecosystems, multicellular con-

sortia of ANME and associated SRB consume a signi

fi

cant

fraction of the methane produced in marine sediments, using

sulfate as a terminal electron acceptor to perform the anae-

robic oxidation of methane (AOM) [

–

]. ANME-SRB

consortia are thought to perform AOM through the direct

extracellular transfer of electrons between ANME and SRB

[

–

]. Along with symbiotic extracellular electron transfer,

ANME-SRB consortia also exhibit other traits of mutualism

such as the sharing of nutrients. For example, members of

the ANME-2 clade have been reported to

fi

x and share N

with partner bacteria [

–

], but the extent to which diazo-

trophic capability might vary across the diverse clades of

ANME and associated SRB is the focus of ongoing research.

Comparative studies of ANME [

] and associated SRB

[

] genomes from multiple ANME-SRB consortia

These authors contributed equally: Kyle S. Metcalfe, Ranjani Murali

*

Kyle S. Metcalfe

ksmetcalfe@gmail.com

*

Victoria J. Orphan

vorphan@gps.caltech.edu

Division of Geological and Planetary Sciences, California Institute

of Technology, 1200 E California Blvd, Mail Code 170-25,

Pasadena, CA 91125, USA

Supplementary information

The online version of this article (

https://

doi.org/10.1038/s41396-020-00757-1

) contains supplementary

material, which is available to authorized users.

1234567890();,:

have revealed signi

fi

cant diversity across clades, particu-

larly for SRB genomes falling within subclades of the

Desulfobacteraceae SEEP-SRB1a [

], common SRB

partners to ANME [

]. However, the implications of

symbiont diversity for metabolic adaptation in ANME-SRB

consortia are obscured by the absence of clearly-established

ANME-SRB pairings in the environment. A framework

fi

ning these pairings would address this gap in knowl-

edge. Establishing this framework for partnership speci

fi

city

in ANME-SRB consortia

—

being the preference that certain

ANME exhibit for speci

fi

c SRB partners

—

would shed light

on the extent to which ANME or SRB physiology may

differ in consortia constituted of different ANME-

SRB pairs.

As an aspect of ANME or SRB physiology that may

differ in different ANME-SRB pairings, nitrogen anabolism

has been observed to be involved in the symbiotic rela-

tionship between partners [

] and has been shown to

fl

uence niche differentiation of different ANME-SRB

consortia via nitrate assimilation ability [

]. Previous

evidence documenting active diazotrophy by AOM con-

sortia from cDNA libraries of

nifH

[

] and

stable

isotope probing with FISH-nanoSIMS, indicated that the

methanotrophic ANME-2 archaea

fi

xed more nitrogen than

SRB in consortia and may supply

fi

xed nitrogen to their

syntrophic partners [

–

]. The diazotrophic potential of

syntrophic SRB, however, and their role in nitrogen

fi

xation

within consortia is poorly understood. Evidence from SRB

genomes [

] and the expression of unidenti

fi

ed nitrogenase

sequences in methane seep sediments [

] suggested that

some seep-associated SRB may also

fi

x nitrogen, opening

up the possibility of variation in diazotrophic activity

among taxonomically-distinct ANME-SRB consortia.

Previous research characterizing the diversity of part-

nerships in ANME-SRB consortia have employed

fl

uores-

cence microscopy, magnetic separation by magneto-FISH,

and single-cell sorting techniques (e.g., BONCAT-FACS)

that are robust against false positives, but are often limited

in statistical power. Fluorescence in situ hybridization

(FISH) has helped to establish the diversity of ANME-

bacterial associations, with ANME constituting four diverse

polyphyletic clades within the Methanomicrobia: ANME-

1a/b [

–

], ANME-2a,b,c [

–

], ANME-2d

[

], and ANME-3 [

]. ANME-associated SRB

have also observed by FISH to be diverse, representing

several clades of Deltaproteobacteria including the

Desul-

fococcus/Desulfosarcina

(DSS) clade [

–

], two separate subclades within the Desulfobulba-

ceae [

], a deeply-branching group termed the

SEEP-SRB2 [

], and a thermophilic clade of Desulfo-

bacteraceae known as HotSeep-1 [

]. These FISH

studies documented associations for different ANME-SRB

consortia, including partnerships between members of

ANME-1 and SEEP-SRB2 [

] or HotSeep-1 [

ANME-2a and SEEP-SRB1a [

], ANME-2c and SEEP-

SRB1a [

], SEEP-SRB2 [

], or Desulfobulbaceae [

and ANME-3 and SEEP-SRB1a [

] or Desulfobulbaceae

[

]. Conspicuously, SRB found in consortia with

ANME-2b have only been identi

fi

ed broadly as members of

the Deltaproteobacteria targeted by the probe S-C-dProt-

0495-a-A-18 (often referred to as

Δ

495) [

], leaving

little known about the speci

fi

c identity of this SRB partner.

Visualizing ANME-SRB partnerships by FISH has been a

valuable aspect of AOM research, but FISH requires the

design of probes with suf

fi

cient speci

fi

city to identify

partner organisms and thus will only detect partnerships

consisting of taxa for which phylogenetic information is

known [

]. Magneto-FISH [

] or BONCAT-

enabled

fl

uorescence-activated cell sorting (BONCAT-

FACS) of single ANME-SRB consortia [

] complement

FISH experiments by physical capture (via magnetic beads

fl

ow cytometry, respectively) and sequencing of ANME

and associated SRB from sediment samples. These studies

corroborated some of the patterns observed from FISH

experiments, showing associations between ANME-2 and

diverse members of the DSS [

]. Magneto-FISH and

BONCAT-FACS observations of ANME-SRB pairings are

also highly robust against false positives but can lack the

statistical power conferred by more high-throughput

approaches that is necessary to establish a general frame-

work for partnership speci

fi

city.

Recently, a number of correlation analysis techniques

have been introduced in molecular microbial ecology stu-

dies, providing information about patterns of co-occurrence

between 16S rRNA operational taxonomic units (OTUs) or

amplicon sequence variants (ASVs) recovered from envir-

onmental diversity surveys [

–

]. Correlation analysis

performed on 16S rRNA amplicon surveys provides a

complementary method to Magneto-FISH and/or

BONCAT-FACS that can be used to develop hypotheses

about potential microbial interactions. While predictions of

co-occurrence between phylotypes from these correlation

analysis techniques have been reported in a number of

diverse environments, they are rarely validated through

independent approaches, with a few notable exceptions

(e.g., [

]).

Here, we present a framework for ANME-SRB partner-

ship speci

fi

city, using correlation analysis of 16S rRNA

amplicon sequences from a large-scale survey of sea

fl

oor

methane seep sediments near Costa Rica to predict potential

ANME-SRB partnerships. A partnership between ANME-

2b and members of an SRB group previously not known to

associate with ANME (SEEP-SRB1g) was hypothesized by

correlation analysis and independently assessed by FISH

and by analysis of amplicon data from Hatzenpichler et al.

[

] of BONCAT-FACS-sorted ANME-SRB consortia.

378

K. S. Metcalfe et al.

With this new framework, we were able to identify a novel

partnership between ANME-2b and SEEP-SRB1g and map

predicted physiological traits of SEEP-SRB1g genomes

onto partnership speci

fi

city with ANME-2b. Our approach

led us to formulate new hypotheses regarding how SEEP-

SRB1g physiology may complement ANME-2b physiol-

ogy, focusing on nitrogen

fi

xation in SEEP-SRB1g. We

demonstrate in this study that the symbiotic relationship

between ANME and associated SRB can vary depending on

the nature of the partner taxa and af

fi

rm the importance of

characterizing individual symbiont pairings in under-

standing AOM symbiosis.

Materials and methods

Here, we present an abridged description of the methods

used in this study. A full description can be found in the

Supplemental Materials and Methods.

Sample origin and processing

Pushcore samples of sea

fl

oor sediment were collected by

DSV

Alvin

during the May 20

–

June 11, 2017 ROC HITS

Expedition (AT37-13) aboard R/V

Atlantis

to methane seep

sites southwest of Costa Rica [

–

]. After retrieval from

the sea

fl

oor, sediment pushcores were extruded aboard R/V

Atlantis

and sectioned at 1

–

3 cm intervals for geochemistry

and microbiological sampling using published protocols

[

]. Samples for DNA extraction were immediately

frozen in liquid N

and stored at

−

80 °C. Samples for

microscopy were

fi

xed in 2% paraformaldehyde for 24 h at

4 °C. A full list of samples used in this study can be found

in Supplementary Table 1 and additional location and

geochemical data can be found at

https://www.bco-dmo.

org/dataset/715706

DNA extraction and Illumina 16S rRNA amplicon

sequencing

DNA was extracted from 310 samples of Costa Rican

methane seep sediments and seep carbonates (Supplemen-

tary Table 1) using the Qiagen PowerSoil DNA Isolation Kit

12888 following manufacturer directions modi

fi

ed for sedi-

ment and carbonate samples [

]. The V4-V5 region of

the 16S rRNA gene was ampli

fi

ed using archaeal/bacterial

primers, 515 F (5

′

-GTGYCAGCMGCCGCGGTAA-3

′

) and

926 R (5

′

-CCGYCAATTYMTTTRAGTTT-3

′

) with Illu-

mina adapters [

]. PCR reaction mix was set up in dupli-

cate for each sample with New England Biolabs Q5 Hot

Start High-Fidelity 2x Master Mix in a 15 μL reaction

volume with annealing conditions of 54 °C for 30 cycles.

Duplicate PCR samples were then pooled and 2.5 μL of each

product was barcoded with Illumina NexteraXT index 2

Primers that include unique 8-bp barcodes. Ampli

fi

cation

with barcoded primers used annealing conditions of 66 °C

and 10 cycles. Barcoded samples were combined into a

single tube and puri

fi

ed with Qiagen PCR Puri

fi

cation Kit

28104 before submission to Laragen (Culver City, CA,

USA) for 2 × 250 bp paired-end analysis on Illumina

’

MiSeq platform. Sequence data were submitted to the NCBI

Sequence Read Archive as Bioproject PRJNA623020.

Sequence data were processed in QIIME version 1.8.0 [

]

following Mason et al. [

]. Sequences were clustered into

de novo operational taxonomic units (OTUs) with 99%

similarity [

], and taxonomy was assigned using the

SILVA 119 database [

], which uses NCBI rather than

GTDB taxonomy. Known contaminants in PCR reagents as

determined by analysis of negative controls run with each

MiSeq set were also removed (see Supplementary Materials

and Methods) along with rare OTUs not present in any given

library at a level of at least 10 reads. The produced table of

OTUs detected in the 310 methane seep sediment and seep

carbonate amplicon libraries was analyzed using the corre-

lation algorithm SparCC [

To examine phylogenetic placement of SRB 16S rRNA

gene amplicon sequences predicted by network analysis to

associate with particular ANME subgroup amplicon

sequences, a phylogeny was constructed using RAxML-

HPC [

] on XSEDE [

] using the CIPRES Science

Gateway [

] from full-length 16S rRNA sequences of

Deltaproteobacteria aligned by MUSCLE [

]. Genomes

downloaded from the IMG/M database were searched using

tblastn. Chlorophyllide reductase BchX (WP011566468)

was used as a query sequence for a tblastn

nifH

search using

BLAST

+

. BchX was used as the query sequence to recover

divergent

nifH

sequences covering the diversity of all

nifH

clades, following the approach of Dekas et al. [

]. Genome

trees were constructed using the Anvi

’

o platform [

] using

HMM pro

fi

les from a subset [

] of ribosomal protein

sequences and visualized in iTOL [

FISH probe design and microscopy

A new FISH probe was designed in ARB [

]. This probe,

hereafter referred to as Seep1g-1443 (Supplementary

Table 2), was designed to complement and target 16S rRNA

sequences in a monophyletic

“

Desulfococcus

sp.

”

clade.

Based on phylogenetic analysis (see below), this clade was

renamed SEEP-SRB1g, following the naming scheme of

Schreiber et al. [

]. Seep1g-1443 was ordered from Inte-

grated DNA Technologies (Coralville, IA, USA). FISH

reaction conditions were optimized for Seep1g-1443, with

optimal formamide stringency found to be 35% (Supple-

mentary Fig. 1). FISH and hybridization chain reaction

(HCR-) FISH was performed on

fi

xed ANME-SRB

Experimentally-validated correlation analysis reveals new anaerobic methane oxidation partnerships with. . .

379

consortia using previously published density separation and

FISH protocols [

], using a selection of following FISH

probes: Seep1g (Alexa488; this work), Seep1a-1441 (cy5;

[

]), ANME-2a-828 (cy3(3

′

); M. Aoki, personal commu-

nication), ANME-2b-729 (cy3; [

]), and ANME-2c-760

(cy3; [

]). FISH was performed overnight (18 h) using

modi

fi

cations (G. Chadwick, personal communication) to

previously-published protocols [

]. Structured-

illumination microscopy (SIM) was performed on FISH and

HCR-FISH (see below) experiments to image ANME-SRB

consortia using the Elyra PS.1 SIM platform (Zeiss, Ger-

many) and an alpha Plan-APOCHROMAT 100X/1.46 Oil

DIC M27 objective. Zen Black software (Zeiss) was used to

construct

fi

nal images from the structured-illumination data.

mRNA imaging using HCR-FISH

Hybridization chain reaction FISH (HCR-FISH) is a pow-

erful technique to amplify signal from FISH probes

[

]. The protocol used here was modi

fi

ed from

Yamaguchi and coworkers [

nifH

initiators, purchased

from Molecular Technologies (Pasadena, CA, USA; probe

identi

fi

“

nifH 3793/D933

”

) or designed in-house (Sup-

plementary Table 2) and ordered from Integrated DNA

Technologies, were hybridized to

fi

xed ANME-SRB con-

sortia. Hairpins B1H1 and B1H2 with attached Alexa647

fl

uorophores (Molecular Technologies) were added sepa-

rately to two 45 μL volumes of ampli

fi

cation buffer in PCR

tubes and snap cooled by placement in a C1000 Touch

Thermal Cycler (BioRad, Hercules, CA, USA) for 3 min at

95 °C. After 30 min at room temperature, hairpins were

mixed and placed in PCR tubes along with hybridized

ANME-SRB consortia. Ampli

fi

cation was performed for

15 min at 35 °C. Similar results were observed when the

HCR-FISH v3.0 protocol established by Choi et al. [

]

was used. ANME-SRB consortia subjected to HCR-FISH

experiments were imaged using the Elyra PS.1 SIM plat-

form (Zeiss, Germany) as mentioned above. In all cases, the

FITC channel was subject to a 500 ms exposure time,

TRITC to 200 ms, and cy5 to 1000 ms. Colocalization of

signal was analyzed in ImageJ using the Colocalization

Finder and JaCoP plugin [

]. These plugins were used to

compute the Pearson

’

s cross-correlation coef

fi

cient (PC)

and Manders

’

colocalization coef

fi

cients (M1, M2). In

addition, pairwise correlations between channels were

visualized using scatterplots of pixel intensity.

Stable isotope probing and nanoSIMS

Methane seep sediments containing abundant ANME-2b

and SEEP-SRB1g consortia (Supplementary Fig. 2) were

used in stable isotope probing (SIP) experiments to test for

diazotrophic activity by SEEP-SRB1g. SIP incubations

(Supplementary Table 3) were prepared by sparging source

bottles and 30 mL serum bottles with N

and mixing 5 mL

of sediment with 5 mL N

-sparged arti

fi

cial seawater with-

out a N source. N sources were removed from the sediment

slurry by washing with arti

fi

cial seawater without an N

source (see Supplementary Materials and Methods). Two

anoxic incubations were pressurized with 2.8 bar CH

with

1.2 mL

(Cambridge Isotopes, Tewksbury, MA, part #

NLM-363-PK, lot # l-21065) at 1 bar, approximately

equivalent to 2% headspace in 20 mL CH

at 2.8 bar

(Supplementary Table 3). Potential

+

contamination

stocks have been previously reported and can lead to

spurious results in nitrogen

fi

xation experiments. We did

not test for

fi

xed N in the speci

fi

c reagent bottle used in

these experiments. However, previous comparisons of

stocks identify those from Cambridge Isotopes as among

the least-contaminated

stocks available [

]. Positive

control incubations (

=

2) were amended with 500 μM

Cl and were pressurized with 2.8 bar CH

and 1.2 mL

natural-abundance N

at 1 bar. Incubations were periodi-

cally checked for AOM activity via sul

fi

de production using

the Cline assay [

] and were chemically

fi

xed for FISH-

nanoSIMS analysis [

] after 9 months. Samples of slurry

fl

uid were collected,

fi

ltered using a 0.2 μm

fi

lter, and

measured for dissolved ammonium concentrations using a

Dionex ICS-2000 ion chromatography system (Thermo

Scienti

fi

c) housed at the Environmental Analysis Center at

Caltech. Fixed ANME-SRB consortia were separated from

the sediment matrix and concentrated following published

protocols [

]. Samples were then embedded in Technovit

H8100 (Kulzer GmbH, Germany) resin according to pub-

lished protocols [

] and semi-thin sections (2 μm

thickness) were prepared using an Ultracut E microtome

(Reichert AG, Austria) which were mounted on Te

fl

on/

poly-L-lysine slides (Tekdon Inc., USA). FISH reactions

were performed on serial sections (

=

30) using Seep1g-

1443 and ANME-2b-729 probes as described above, with

the omission of 10% SDS to prevent detachment of section

from slide (G. Chadwick, personal communication), and

slides were imaged and mapped for subsequent nanoSIMS

analysis using a Zeiss Elyra PS.1 platform. Sequential

sections of each sample were imaged and mapped to iden-

tify the section most representative of a section through the

center of ANME-SRB consortia. This allowed for the

interpretation of spatial patterns correlated with distance

from the exterior of the ANME-SRB consortium on the x

–

plane as representative of those correlated with the unob-

served x

–

z and y

–

z planes. After removal of DAPI-Citifuor

mounting medium by washing in DI water following pub-

lished protocols [

], individual wells on the slides were

scored with a diamond scribe and cut to

fi

t into the nano-

SIMS sample holder (~1 cm diameter) and sputter-coated

with 40 nm Au using a Cressington sputter coater. Brie

fl

380

K. S. Metcalfe et al.

nanoSIMS analyses were performed using a Cameca

NanoSIMS 50 L housed in Caltech

’

s Microanalysis Center:

512 × 512 pixel raster images of 20 μm

were collected for

–

, and

–

ions by sput-

tering with a ~1 pA primary Cs

+

ion beam current with a

dwell time of 12

–

48 ms/pixel. Data were analyzed using

look@nanoSIMS [

Results

16S rRNA correlation analysis predicts a speci

fi

association between ANME-2b and SEEP-SRB1g

Correlation analysis applied to 16S rRNA gene amplicon

libraries has been frequently used to identify interactions

between microorganisms based on the co-occurrence of

their 16 S rRNA sequences in different environments or

conditions [

–

]. Here, we applied correlation analysis to

Illumina 16S rRNA amplicon sequences recovered from

Costa Rican methane seep sediments (Supplementary

Table 1) to explore partnership speci

fi

city between ANME

and associated SRB. QIIME processing of amplicon

sequences prepared from 310 Costa Rican methane seep

sediment and seep carbonate samples yielded 3,052 OTUs

after

fi

ltering in R. A table of read abundances for these

OTUs across the 310 samples was analyzed by SparCC to

calculate correlation coef

fi

cients and signi

fi

cance for all

possible 4,658,878 OTU pairs using 100 bootstraps

(Fig.

a). Of these pairs, 9.7% (452,377) had pseudo-

values < 0.01, indicating the coef

fi

cients for each of these

correlations exceeded that calculated for that same OTU

pair in any of the 100 bootstrapped datasets [

]. The

taxonomic assignment of the constituent OTUs of correla-

tions with pseudo-

values < 0.01 were then inspected,

where 18% (81,459) of correlations with pseudo-

values <

0.01 describe those involving ANME (Fig.

b). Of these,

32% occur between ANME and OTUs assigned to three

main taxa:

Desulfococcus

sp. (renamed SEEP-SRB1g, see

discussion below), SEEP-SRB1a, and SEEP-SRB2

Fig. 1

Analysis of SparCC-calculated correlations between 16S

rRNA amplicon sequences (OTUs clustered at 99% similarity)

from an ecological survey of 310 methane seep sediment samples

from sea

fl

oor sites off of Costa Rica.

A stacked histogram (

)

illustrates the proportion of correlations deemed signi

fi

cant on the

basis of pseudo-

values < 0.01 calculated by comparison with 100

bootstrapped correlation tables (see Materials and Methods). Of the

correlations with pseudo-

values < 0.01, 18% include ANME with a

non-ANME taxon (

). Signi

fi

cant correlations between OTUs with

taxonomy assignments that are identical at the genus level (e.g., two

Anaerolinea OTUs) are indicated by identical taxonomy assignment.

32% of correlations between ANME and non-ANME taxa are repre-

sented by OTUs assigned to three groups of sulfate-reducing bacteria:

SEEP-SRB1g, SEEP-SRB1a, and SEEP-SRB2 (

). Stacked histo-

grams of correlations between OTUs assigned to SEEP-SRB1g, SEEP-

SRB1a, or SEEP-SRB2 and ANME OTUs, parsed by ANME subtype

(

), highlights speci

fi

c associations predicted between ANME-1 and

either SEEP-SRB1a or SEEP-SRB2, ANME-2a and SEEP-SRB1a,

ANME-2c and SEEP-SRB1a, and between ANME-2b and SEEP-

SRB1g.

Experimentally-validated correlation analysis reveals new anaerobic methane oxidation partnerships with. . .

381

(Fig.

c). A complete list of signi

fi

cant correlations, their

coef

fi

cient values, OTU identi

fi

ers, and accompanying

taxonomy assignments can be found in Supplementary

Table 4.

16S rRNA phylogenetic analysis revealed the SILVA-

assigned

“

Desulfococcus

sp.

”

OTUs comprise a sister clade

to the SEEP-SRB1a that is distinct from cultured

Desulfo-

coccus

sp. (e.g.,

D. oleovorans

and

D. multivorans

, see

below). We therefore reassigned the

Desulfococcus

OTUs

to a new clade we termed SEEP-SRB1g following the

naming scheme outlined for seep-associated SRB in

Schreiber et al. (e.g., SEEP-SRB1a through -SRB1f) [

Furthermore, statistically-signi

fi

cant correlations between

OTUs of ANME and SRB taxa suggested that ANME-SRB

partnerships in the Costa Rica seep samples could be clas-

fi

ed into the following types: ANME-1 with SEEP-SRB1a

or SEEP-SRB2, ANME-2a with SEEP-SRB1a, ANME-2b

with SEEP-SRB1g, ANME-2c with SEEP-SRB1a or SEEP-

SRB2, and ANME-3 with SEEP-SRB1a (Fig.

d). While

physical association between different ANME lineages and

Deltaproteobacterial clades SEEP-SRB1a and SEEP-SRB2

had been well-documented [

], members of

the SEEP-SRB1g had not previously been identi

fi

ed as a

potential syntrophic partner with methanotrophic ANME.

These associations were further examined by detailed

network analysis in which the table of correlations with

pseudo-

values < 0.01 was further

fi

ltered to contain only

those correlations with coef

fi

cients (a measure of correlation

strength) in the 99th percentile of all signi

fi

cant correlations.

A network diagram in which nodes represent OTUs and

edges between nodes represent correlations was constructed

with force-directed methods [

], where edge length varied

in inverse proportion to correlation strength. A subregion of

this network focused on ANME-associated OTUs is pre-

sented in Fig.

. Cohesive blocks, subsets of the graph with

greater connectivity to other nodes in the block than to nodes

outside [

], were calculated and revealed 3 primary blocks

of ANME and SRB OTUs. Visualization of these 3 blocks

by a chord diagram [

] further highlighted the taxonomic

identity of ANME-SRB OTU pairs in these blocks: ANME-

1 or ANME-2c (one OTU with mean read count < 10) and

SEEP-SRB2, ANME-2a or ANME-2c and SEEP-SRB1a,

and ANME-2b or ANME-2a and SEEP-SRB1g (Fig.

b).

The predicted associations between ANME-2c and SEEP-

SRB2 and between ANME-2a and SEEP-SRB1g were

relatively more rare than the other associations; only one rare

ANME-2c OTU (mean read count < 10) and four uncom-

mon ANME-2a OTUs (mean read count < 100) were pre-

dicted between SEEP-SRB2 and SEEP-SRB1g,

respectively. Inferred partnership speci

fi

city in two of the

blocks has been previously corroborated by FISH studies,

namely associations between ANME-1 with SEEP-SRB2

[

], ANME-2c with SEEP-SRB1a [

], and ANME-2a

with SEEP-SRB1a [

]. The partnership between SEEP-

SRB1g and ANME-2b, however, had no precedent, as the

only previous FISH descriptions of ANME-2b had placed it

with a partner Deltaproteobacterium with taxonomy not

known beyond the class level [

Common patterns of association observed in

network analysis and in single ANME-SRB consortia

To test if ANME-SRB partnership speci

fi

city observed in

our correlation analysis of 16S rRNA amplicon sequences

from seep samples (Figs.

) was consistent with data

collected from individually-sorted ANME-SRB consortia

after BONCAT-FACS [

], we constructed a phylogeny

with full-length and amplicon 16S rRNA sequences from

ANME-associated SRB including SEEP-SRB1g (Fig.

;

Supplementary Fig. 3). These individual ANME-SRB

consortia sorted by BONCAT-FACS were sourced from

methane seep sediment samples recovered from Hydrate

Ridge off the coast of Oregon and sea

fl

oor sites in Santa

Monica Basin, California, allowing us to further test whe-

ther the ANME-2b

—

SEEP-SRB1g partnership can be

detected in sea

fl

oor sites beyond Costa Rica. 16S rRNA

amplicon sequences from network analysis of Costa Rica

seep sediment samples (Fig.

) and from BONCAT-FACS

sorted consortia from Hydrate Ridge and Santa Monica

Basin (Fig.

;[

]) were then annotated by ANME subtype

and identity of associated phylotypes. In the BONCAT-

FACS dataset, 8 out of 11 (72%) of the consortia with

ANME-2b OTUs had corresponding deltaproteobacterial

OTUs that belonged to the SEEP-SRB1g clade (Fig.

Similarly, of the Deltaproteobacteria OTU sequences from

the BONCAT-FACS sorted consortia af

fi

liated with SEEP-

SRB1g 89% (8/9) had ANME-2b as the archaeal partner

(Fig.

). Notably, we found that these SEEP-SRB1g

sequences were also highly similar to published full-

length 16S rRNA clone library sequences (e.g., NCBI

accession AF354159) from seep sediments where ANME-

2b phylotypes were also recovered [

]. A

χ

-test for

independence was performed on 16S rRNA OTUs recov-

ered from (39) to test the null hypothesis that the presence

of a given SRB taxon in a FACS sort is independent of the

type of ANME present in the sort. This test demonstrated

that the SRB taxon found in a given sort was dependent

on the ANME also present in the sort,

χ

=

30.6 (

d.f

=

=

30),

< 0.001. The pattern of association between

ANME and SRB OTUs in individual BONCAT-FACS-

sorted ANME-SRB consortia thus corroborated the infer-

ence from network analysis that ANME-2b and SEEP-

SRB1g OTUs exhibit signi

fi

cant partnership speci

fi

city. On

the basis of amplicon sequence associations found in the

BONCAT-FACS sorting dataset collected from sediment

samples of Oregon and California seeps as well as those

382

K. S. Metcalfe et al.

displayed by correlation analysis of amplicons from Costa

Rica methane seeps, we designed a set of independent

experiments to directly test the hypothesis that ANME-2b

form syntrophic partnerships with the previously-

undescribed SEEP-SRB1g deltaproteobacteria.

FISH experiments show SEEP-SRB1g in association

with ANME-2b

Speci

fi

c oligonucleotide probes were designed and tested

for the SEEP-SRB1g clade (Supplementary Fig. 1) and

FISH experiments were used to validate the predicted

ANME-2b

—

SEEP-SRB1g partnership. Simultaneous

application of FISH probes targeting SEEP-SRB1a, the

dominant deltaproteobacterial partner of ANME (Seep1a-

1441 [

]), the newly designed SEEP-SRB1g probe

(Seep1g-1443, this work), and a probe targeting ANME-2b

(ANME-2b-729 [

]) demonstrated that ANME-2b form

consortia with SEEP-SRB1g, appearing as large multi-

cellular consortia in seep sediment samples from different

localities at Costa Rica methane seep sites (see Supple-

mentary Materials and Methods for site details) that also

Fig. 2

Network analysis of the subset of correlations between

OTUs calculated by SparCC [

] that are both signi

fi

cant (pseudo-

values < 0.01, 100 bootstraps) and strong (

≥

99th percentile).

Edge length is inversely proportional to correlation strength and is

used to visualize the network (top panel) using force-directed methods

[

]. Edges are black where they belong to a set of cohesive blocks of

nodes [

] and gray otherwise. Chord diagram [

] visualizing

ANME-SRB partnership speci

fi

city (bottom panel), with band thick-

ness between SRB (left) and ANME (right) proportional to the number

of edges between ANME and SRB OTUs within cohesive blocks.

Network analysis supports (cf. Fig.

) previously-undescribed asso-

ciation between ANME-2b and SEEP-SRB1g.

Experimentally-validated correlation analysis reveals new anaerobic methane oxidation partnerships with. . .

383