The discovery of deep-sea hydrothermal vents in 1977 revolutionized our understanding of the energy sources that fuel primary productivity on Earth. Hydrothermal vent ecosystems are dominated by animals that live in symbiosis with chemosynthetic bacteria. So far, only two energy sources have been shown to power chemosynthetic symbioses: reduced sulphur compounds and methane. Using metagenome sequencing, single-gene fluorescence in situ hybridization, immunohistochemistry, shipboard incubations and in situ mass spectrometry, we show here that the symbionts of the hydrothermal vent mussel Bathymodiolus from the Mid-Atlantic Ridge use hydrogen to power primary production. In addition, we show that the symbionts of Bathymodiolus mussels from Pacific vents have hupL, the key gene for hydrogen oxidation. Furthermore, the symbionts of other vent animals such as the tubeworm Riftia pachyptila and the shrimp Rimicaris exoculata also have hupL. We propose that the ability to use hydrogen as an energy source is widespread in hydrothermal vent symbioses, particularly at sites where hydrogen is abundant.

Hydrogen consumption incubationsSampling site: For on-board physiological experiments specimens belonging to undescribed Bathymodiolus species (B. spp.) were collected at the Comfortless Cove and Lilliput hydrothermal vent fields on the southern Mid-Atlantic Ridge. Mussels were sampled using nets (40 cm length, 20 cm diameter opening, mesh size 335 µm, Hydrobios, Kiel, Germany) handled by the manipulator arm of the remotely operated vehicles (ROVs) Quest 4000 m (MARUM, Bremen, Germany). Mussel dissection for shipboard incubations: Mussels were opened with a scalpel by cutting through the posterior and anterior adductor muscles. Viability was tested by prodding the foot with a dissection needle and only mussels whose foot contracted were used. The foot and both gills were separated from the remaining tissue using dissection scissors. Tissue pieces of 6 mm in diameter were cut out of the gill and the foot tissues using a steel hole-puncher. One tissue piece from each individual was frozen in liquid nitrogen for weight determination in the home laboratory. For negative controls, foot tissues were used which do not contain endosymbiotic bacteria. Experimental setup: For hydrogen partial pressures up to 1,000 ppm (in helium) glass serum vials (58 ml) were fully filled with chilled (4°C) sterile-filtered (0.22 µm) bottom seawater retrieved from 3,000 m water depth. One piece of gill tissue was placed in the vial, the vial closed with a gas-tight butyl rubber stopper, and crimped with aluminium seals. Control vials contained foot tissue or no tissue at all. 20 ml of hydrogen gas (100 ppm, 250 ppm or 1,000 ppm in helium, Linde) were injected through the rubber stopper using a gas-tight syringe (SGE Analytical Science and Hamilton) with a second syringe allowing pressure compensation through the outflow of the same volume of seawater. For hydrogen partial pressures above 1,000 ppm (in air) glass serum vials were processed as described above with the exception that the vials were filled with only 38 ml of seawater. Pure hydrogen gas (100%; Air Liquide) was then injected in the air headspace through the rubber stopper to the desired final partial pressure (1,000-3,000 ppm) using a gas-tight syringe. All vials were placed upside down to minimize possible gas exchange via the rubber stopper and incubated at 4°C on a slowly rotating table. At given time points a subsample was taken with a gas-tight syringe from the headspace with the pressure decrease compensated through the inflow of oxygen-saturated, chilled sterile-filtered seawater from a second syringe. Analysis of the headspace hydrogen content: The hydrogen concentration in the headspace was determined using a gas chromatograph (Thermo Trace GC ultra) equipped with a packed stainless steel column (Molecular Sieve 5A, carrier gas: He) and a pulse discharge detector (PDD). Recording and calculation of results was performed using a PC operated integration system (Thermo Chrom Card A/D). Analytical procedures were calibrated daily with commercial gas standards (Linde). Incubation conditions and rate calculations: The concentrations of dissolved hydrogen and oxygen were calculated from Crozier and Yamamoto (1974) [doi:10.1021/je60062a007] and Weiss (1970) [doi:10.1016/0011-7471(70)90037-9] under the assumption of Henry's law i.e. that the concentration of a dissolved gas in a solution is directly proportional to the partial pressure of that gas above the solution. The molar volume of an ideal gas (22.414 l mol-1) was used to convert between the partial pressure of a gas (ppm) and the amount of the gas (in moles) in the headspace. The rates of hydrogen removal from the headspace [in nmol h-1 (piece tissue)-1] were calculated for the first 60 minutes after addition of the electron source performing non-linear regression through the data points obtained during the complete incubation period of up to 60 hours. Exponential and hyperbolic regression described by the equations f=aexp(b/(x+c)) and f=y0+(ab)/(b+x) gave well-fitting regression curves (R2 = 0.920 and 0.921, respectively; standard error of estimate 6.434 and 6.400, respectively). The effect of methodological hydrogen removal from the headspace through subsampling was considered. The rates at which hydrogen disappeared from incubation vials containing only seawater but no tissues (chemical oxidation, hydrogen loss by diffusion) were subtracted from the tissue rates. Resulting rates were then normalized to gram wet weight (in nmol h-1 [g wet weight]-1).