Specimen Collection.

Over the course of two field seasons, five specimens of Kuphus polythalamia were collected in Mindanao, Philippines, at a depth of ∼3 m in a marine bay formerly used as a log storage pond. Specimens were transported alive to the Marine Science Institute, University of the Philippines, Diliman, Quezon City, Philippines, where they were dissected, and tissues were preserved for microscopy or were frozen at −80 °C until use (SI Appendix, Table S3).

Bacterial Cultivation.

Approximately 10- to 30-mg slices of gill tissue were rinsed with sterile seawater and homogenized manually in 1 mL of shipworm homogenization buffer (filter-sterilized natural seawater diluted to 75% final concentration with sterile distilled deionized water (ddH₂O), 0.025% cysteine hydrochloride, 20 mM Hepes buffer, pH 8.0). The homogenate was then streaked on STBD agar plates [66% (vol/vol) natural filter-sterilized seawater, 0.025% (wt/vol) NH₄Cl, 20 mM Hepes buffer, pH 8.0, 1.5% (vol/vol) metals and mineral mix (8), 20 mM Na₂S₂O₃, 10 mM NaHCO₃, 10% (vol/vol) DMEM/Nutrient F-12 Ham (Sigma; D6421), and 1.0% (wt/vol) Bacto agar, final concentrations]. Antimycotics [10 µg/mL cycloheximide and 125 units/mL of nystatin (Sigma; N1638)] were added to the agar medium to suppress fungal growth. Plates were incubated in microaerobic conditions using GasPaK EZ Campy container at 25 °C until small opaque white colonies appeared on the surface of the agar. Colonies were subsequently transferred to aerobic STB50 medium, which is similar to STBD medium but without the DMEM/Nutrient F-12 Ham component; and with concentrations of sodium thiosulfate and sodium bicarbonate increased to 30 and 15 mM, respectively. For long-term storage, freezer stocks of the isolates were prepared by adding 40% glycerol to broth culture at 1:1 ratio and storing at −80 °C. To produce bacterial cells for use in subsequent experiments, a single colony of K. polythalamia symbiont strain 2141T, grown from frozen stock on an STB50 agar plate, was used to inoculate a 6-mL starter culture in STB50 broth. After 3 d, 1 L of STB50 broth was inoculated with 0.1% (vol/vol) of the starter culture. The broth cultures were then incubated at 30 °C in a shaker incubator (Series 26; New Brunswick Scientific Co., Inc.) at 100–150 rpm for 3–5 d. To harvest cells, the culture flask was placed in an ice bath for 1 h, and then the bacterial pellets were harvested by centrifugation (3,500–10,000 × g, 4 °C, 30 min).

Transmission Electron Microscopy.

Bacterial pellets or tissue samples were fixed for 2 h using 2.5% glutaraldehyde in a 0.1 M sodium cacodylate buffer, followed by two 15-min washes in a 0.1 M sodium cacodylate buffer at 4 °C. Samples were postfixed for 2 h in 1% osmium tetroxide in a 0.1 M sodium cacodylate buffer, again followed by two 15-min washes in a 0.1 M sodium cacodylate buffer at 4 °C. Samples were then dehydrated through an ethanol series (30, 50, 70, 85, and 95% ethanol for 15 min at each stage), with a final dehydration stage in absolute ethanol for 1 h. Dehydrated samples were embedded in Spurr resin overnight and polymerized the following day at 60 °C for 24 h. Embedded samples were sectioned (50- to 200-nm thickness), mounted on TEM square mesh copper grids, and stained with uranyl acetate and lead citrate for ∼10–15 min each. Resulting grids were visualized on a JEOL JEM 1010 transmission electron microscope.

SEM and Energy-Dispersive X-Ray Analysis.

For SEM imaging of tissues, samples were frozen in liquid nitrogen, fractured with a razor blade, critical point dried using the SAMDRI-PVT-3D Critical Point Dryer (Tousumis), and mounted on a standard aluminum SEM stub and coated with platinum to a thickness of 5 nm using the Cressington 208 HR High Resolution Sputter Coater (Cressington Scientific Instruments). For SEM imaging of bacteria, cells of K. polythalamia symbiont strain 2141T or K. polythalamia gill homogenates were placed in 70% ethanol, pipetted directly onto a TEM square mesh copper grid, and were allowed to air dry before mounting onto a standard aluminum SEM stub. Samples were imaged on the Hitachi S-4800 field emission scanning electron microscope. Elemental analysis was conducted using the EDAX Element EDS Analysis System and accompanying Element EDS Analysis Software Suite.

DNA Preparation.

Cross-sectional tissue samples were taken from multiple locations (anterior, middle, and posterior) along the gills of three adult specimens of Kuphus polythalamia (SI Appendix, Table S3). Approximately 25 mg of tissue was excised from each section, rinsed thoroughly with filter-sterilized seawater, and homogenized in 4.5 M guanidinium thiocyanate, 2% N-lauroylsarcosine (wt/vol), 50 mM EDTA, pH 8.0, 25 mM Tris⋅HCl. Genomic DNA was extracted from each tissue homogenate using a QiaCube DNA extractor (Qiagen) and Qiagen DNeasy Blood and Tissue Kit followed by the Zymo DNA Clean and Concentrate Kit, according to the respective manufacturer’s recommended protocols.

Phylogenetic Characterization of Bacteria.

To establish the phylogenetic identity of bacterial isolates in culture, and uncultivated bacteria in gill tissue, genes encoding 16S rRNA (SSU), a multifunctional cellulase (celAB), and a rhodanese-like thiotransferase (sseA) were PCR amplified from genomic DNA extracts and sequenced using primers and PCR conditions described in SI Appendix, Table S1. PCR products were purified using GenCatch PCR Purification Kit and sequenced using an ABI 3130xl Genetic Analyzer (Macrogen). Similarity to previously published sequences in GenBank (National Center for Biotechnology Information) was determined using BLASTn or BLASTp as appropriate. Phylogenetic analyses of 16S rRNA sequences was performed using programs implemented in Geneious, version 8.1.2. Sequences were aligned using MAFFT (version 7.017) by using the auto algorithm. The aligned sequences were trimmed manually, resulting in a final aligned dataset of 1,172 nucleotide positions. Phylogenetic analysis was performed using MrBayes (version 3.2.6) using the GTR + I + Γ substitution model. Chain length was set to 2 million, subsampling every 2,000 generations, and discarding the first 20% of analytical results (burn-in).

FISH.

FISH was performed as in ref. 34. Briefly, all samples were fixed in 4% paraformaldehyde in filter-sterilized seawater at 4 °C overnight. Samples were transferred to 70% ethanol and stored at −20 °C. Fixed samples were embedded in paraffin and sectioned at 10-μm thickness. Sections were deparaffinized with a 5-min treatment in xylenes, followed by a 5-min treatment in ethanol, and then rinsed in ddH₂O. Slides were air-dried. Sections were hybridized with 5 ng/μL probe (final concentration) in hybridization buffer [35% formamide, 0.9 M NaCl, 20 mM Tris⋅HCl (pH 7.4), 0.01% SDS] for 2 h at 46 °C. After hybridization, the slides were incubated at 48 °C in wash buffer [35% formamide, 0.7 M NaCl, 20 mM Tris⋅HCl (pH 7.4), 50 mM EDTA, 0.01% SDS] for 20 min, rinsed with ddH₂O, and air-dried. After drying, 30 µL of 500 mM Sytox Green was added and incubated for 15 min, and then washed with ddH₂O. Slides were then mounted in a 4:1 Citifluor (Citifluor, Ltd.)/Vectashield (Vector Labs) mounting medium. Slides were visualized on a Zeiss AxioImager laser-scanning confocal microscope with LSM 5 Pascal, version 4.0, imaging software, with imaging parameters held constant between specific and control probes.

Genome Sequencing of K. polythalamia Symbiont Strain 2141T.

Genomic DNA was extracted from the bacterial pellets using CTAB/phenol/chloroform as detailed at www.pacb.com/wp-content/uploads/2015/09/DNA-extraction-chlamy-CTAB-JGI.pdf, except that Phase-Lock gel tubes were not used. The purity of the genomic DNA was assessed spectrophotometrically using NanoDrop ND-1000; size and quantity were estimated by agarose gel electrophoresis. A high-quality draft genome sequence was determined for K. polythalamia symbiont strain 2141T at the US Department of Energy (DOE) Joint Genome Institute (JGI) using the Pacific Biosciences (PacBio) RS platform, which generated 201,156 filtered subreads totaling 827.2 Mbp. All general aspects of library construction and sequencing performed at the JGI can be found at jgi.doe.gov/. The raw reads were assembled using HGAP (version: 2.0.0) (35). The final draft assembly contained 1 contig in 1 scaffold, totaling 4.8 Mbp. The input read coverage was 204×.

Metagenome Sequencing and Assembly.

Illumina libraries were prepared as ∼350-bp inserts and sequenced using an Illumina HiSeq 2000 sequencer with 125-bp paired-end runs at the Huntsman Cancer Institute’s High-Throughput Genomics Center at the University of Utah. Illumina fastq reads were trimmed using Sickle (36) with the parameters (pe sanger -q 30 –l 125). The trimmed FASTQ files were converted to FASTA files and merged using the Perl script ‘fq2fq’ in IBDA_ud package (37). Merged FASTA files were assembled using IDBA_ud with standard parameters in Futuregrid (https://portal.futuresystems.org/) or the Center for High Performance Computing at the University of Utah.

Identification of Bacterial Sequences in Metagenome Data.

Assembly-assisted binning was used to sort and analyze trimmed reads and assembled contigs into clusters putatively representing single genomes using MetaAnnotator (38). Each binned cluster was retrieved using Samtool (39, 40). To identify bacterial clusters, a BLASTp search was performed using as a query the longest protein sequence coded in the longest contig of each cluster. The longest contig in each cluster was extracted using the command ‘cat cluster.fasta | awk -v RS='>' 'NR>1*i++∼i>=0&&i<=1*print “>” $0^' | sed '/^>$/d'’. The protein coded in each longest contig was recalled using webAUGUSTUS (bioinf.uni-greifswald.de/webaugustus/). The length of each protein sequence was calculated using Samtool (39, 40). Command ‘awk 'BEGIN *max = 0∼ *if ($2+0>max+0) *max=$2; content=$0∼∼ END *print content∼'’ was used to extract the sequence head corresponding to the longest protein in each contig. Each of the longest protein sequences was extracted using Samtool (39, 40). A BLASTp search was performed using the command ‘blastp -max_target_ seqs 1 -word_size 6 -evalue 10 -matrix BLOSUM62 -query longest.faa -outfmt '7 qseqid sseqid pident qlen slen length mismatch gapopen evalue bitscore staxids sscinames scomnames sskingdoms sblastnames stitle' -db nr -remote -out kuphus1_longest_blastp_remote’. The bacterial hits were extracted with a threshold of e-value < 2.00E-19. The identified bacterial clusters were checked manually using BLASTx search to remove false positives originating in the animal genome. BLASTn search was used to identify clusters representing 2141T-like genomes. Contigs from each cluster were used as query to do BLASTn search against the genome of sulfur-oxidizing strain 2141T. Clusters were considered to be related to sulfur-oxidizing strain 2141T if >70% of the contained contigs aligned to the 2141T genome with a threshold of identity of >85%.

Metagenome and Genome Sequence Comparison.

Contigs from all 2141T-like bacterial clusters from each sample were combined in a single FASTA file and converted to a pseudomolecule. The BLASTn search comparison was performed by BLAST ring image generator (BRIG) (41) using the genome of sulfur-oxidizing strain 2141T against the pseudomolecule with the standard BLASTn search options. Synteny comparison (SI Appendix, Fig. S5) was done by MUMmer 3.0 (42), using the NUCmer alignment tool with parameters (-maxmatch -c 400). The alignment result was plotted by ‘mummerplot’ with a layout option (−l). A single FASTA file containing all bacterial contigs from 2141T-like clusters from each metagenome sample was used in the synteny comparison. The identity distributions of the average nucleotide identity (ANI) between the genome of strain 2141T and the pooled metagenome sequences (SI Appendix, Fig. S4) was calculated by online tool (enve-omics.ce.gatech.edu/ani/) with the following parameters: minimum length, 700 bp; minimum identity, 70%; minimum alignments, 50; window size, 1,000 bp; step size, 200 bp.

SNP Analysis.

A blast database was produced from the merged FASTA file used in IDBA_ud assembly with the command “makeblastdb -input_type fasta -dbtype nucl -parse_seqids”. A BLASTn search (-perc_identity 90 -max_target_seqs 100000000 -outfmt “6 qseqid sseqid pident length evalue qcovs qlen slen”) was performed to find Illumina reads that are homologs of the gyrB gene sequence in the genome of sulfur-oxidizing strain 2141T. Hits were defined as “length/slen > 0.95”; the sequence heads of the hit reads were extracted using Samtool (39, 40) and were aligned to the reference gyraseB gene using the CLC genomics workbench (version 9.0). The resulting alignment was equally split into 29 fragments (starting from the 85th base pair, 85 bp for each) by shell script (alignment_split.sh; for code, see Dataset S1). The alignment in each fragment were also removed by the shell script. SNP categories of the reads in each small alignment were retrieved by unambiguous assembly (Word Length = 4, Word Length = 4, Maximum Gaps Per Reads = 0, Maximum Mismatches = 0, Maximum Ambiguity = 1) using Geneious (R9).

Cell and Tissue Preparation for RuBisCO Assays.

Analyses on bacterial samples were performed on previously frozen cell pellets. To confirm that cells were intact, a small amount of the frozen pellet was resuspended in STB₅₀ and examined by light microscope. A 10× volume of RuBisCO Assay Buffer (RAB) [50 mM bicine, pH 8, 5 mM DTT (freshly added from a frozen stock), 20 mM MgCl₂, 10 mM NaHCO₃] was added to the remaining frozen cell pellet, and then sonicated on ice with a BRANSON Digital Sonifier 250 at 50% amplitude for 10 s twice, with 10 s between pulses, and the extract was examined microscopically to confirm lysis. Gill and siphon (symbiont-free control) tissues were dissected, kept at −80 °C, transported on dry ice, and stored at −80 °C until assayed. Tissues were chopped with a razor blade, a 10× volume of RAB added, and sonicated as above.

RuBisCO Assay.

Ribulose-1,5-bisphosphate (RuBP)-stimulated ¹⁴CO₂ incorporation was measured as an indicator of RuBisCO activity as in refs. 43⇓–45. Assays were performed in triplicate (three technical replicates, one biological replicate; only one specimen of this rare species was available for assay), with and without addition of RuBP over a 30-min time course. The rate of RuBP-stimulated incorporation was normalized to total protein (SI Appendix, Table S5). Positive controls contained purified spinach RuBisCO (Sigma; R8000-1UN). Additional controls were performed with no isotope and no extract (100 μL of RAB). Before assay, 50 µL of NaH¹⁴CO₃ (0.75 µCi per reaction) was added to 100 μL of sample and incubated for 20 min. The assay was then initiated by addition of 50 µL of 5 mM RuBP. Reactions were stopped at 10, 20, and 30 min by addition of 200 µL of glacial acetic acid. Unincorporated ¹⁴CO₂ was evaporated by drying in uncapped tubes for 4 h at 90 °C. Acid-stable products were resuspended in 400 µL of water and 7 mL of scintillation fluid, and label incorporation was measured using a Beckman LS6500 scintillation counter. Counts were converted to total micromoles of C incorporated and normalized using the negative controls at the corresponding time point (SI Appendix, Table S5).

Protein Assay.

The Thermo Scientific Pierce 660-nm Protein Assay was chosen to avoid interference from bicine and DTT present in the RAB. Sonicated extracts were assayed according to manufacturer’s instructions and compared with an albumin standard curve. All samples fell within the range of the standard curve.

Data Analysis.

All data analysis was carried out in GraphPad Prism 6. For the activity assays, each triplicate was plotted versus time, and a line was fitted to the data (SI Appendix, Table S5). The slope of the line corresponds to the activity of the sample. For the protein concentration assays, triplicate samples were interpolated to the standard curve, and the mean and SE were computed for the resulting protein concentration values. The specific activity of the samples was calculated by dividing the activity of the sample by the protein concentration in the sample. The error of the specific activity was calculated by propagating the SEs for the activity and the protein concentration according to standard techniques (46):

For multiplication and division: xy=z,Δzz=Δxx+Δyy.