Chemosynthetic origin of 14 C-depleted dissolved organic matter in a ridge-ﬂank hydrothermal system

Hydrothermal ﬂuids circulate through extensive areas of the upper oceanic crust. Most hydrothermal circulation occurs on ridge ﬂanks 1,2 , where low-temperature ﬂuids ﬂow through porous basalts. These ﬂuids contain variable levels of dissolved organic carbon, but the source and composition of this carbon are uncertain. Here, we report (cid:49) 14 C and δ 13 C measurements of dissolved organic carbon in ridge-ﬂank and on-axis hydrothermal ﬂuids sampled from the Juan de Fuca Ridge. Dissolved organic carbon from two independent ridge-ﬂank sites was characterized by low δ 13 C and (cid:49) 14 C values. The δ 13 C values ranged from − 26 to − 35 (cid:104) , and were consistent with a chemoautotrophic origin. The 14 C ages of the dissolved organic carbon ranged from 11,800 to 14,400 years before present, revealing that the carbon was around three times older than dissolved organics in the deep ocean. The (cid:49) 14 C values of the ridge-ﬂank dissolved organic matter also corresponded closely to those of dissolved inorganic carbon in the same ﬂuid samples. Taken together, the data suggest that chemosynthetic crustal microbial communities synthesize dissolved organic carbon from inorganic carbon in ridge-ﬂank ﬂuids. We suggest that ridge-ﬂank circulation may support an indigenous biosphere extensive enough to export substantial ﬁxed carbon, with distinct isotopic and probably compositional character, to the overlying ocean. fraction of DOC (UDOC; 0 . 1 µ M—1,000D). UDOC subsamples were diluted to ∼ 1 litre with pre-irradiated purified Milli-Q water, and quantitatively oxidized to CO 2 with a mercury arc lamp photochemical reactor system for 14 C analysis 30 . Accelerator mass spectrometry (AMS) measurements were made at UCI Keck Carbon Cycle Accelerator Mass Spectrometry facility. All isotopic measurements were blank-corrected for graphitization and AMS analysis, and also corrected with error propagation for the ultraviolet-extraction procedural blank. Further details on sampling, DOC isolation and AMS measurements are provided in Supplementary Methods.

Although high-temperature systems at the mid-ocean ridges are better studied, growing evidence indicates that, globally, most hydrothermal circulation occurs instead in lower-temperature fluids on ridge flanks 1,2 . In the Juan de Fuca Ridge (JDFR) flank region that we sampled, such fluids contain low total DOC concentrations (∼10-15 µM; ref. 6). These values suggest that much oceanic DOC must be removed during the long crustal fluid residence times (∼10,000 yr; ref. 7), either by microbial utilization or abiotic adsorption 6 . However, concentration data carry no information regarding the source or composition of the DOC in vented fluids. If vented material were residual deepocean DOC, this would indicate a heterotrophic system with respect to dissolved organics, and potential impacts on the oceanic reservoir could be limited. In contrast, if a substantial portion of vented ridge-flank DOC were autochthonous, with distinct isotopic or biochemical composition, then even small inputs could represent an important unrecognized source of fixed carbon to the overlying ocean DOC pool.
We measured coupled stable carbon (δ 13 C) and radiocarbon ( 14 C) isotopic compositions of ultrafiltered DOC (UDOC) isolated from both ridge-flank and selected on-axis low-temperature fluid sources on the JDFR system ( Fig. 1; Supplementary Methods). Our ridge-flank sites are located approximately 100 km east of the ridge axis on 3.5-Myr-old crust, and include a venting outcrop (Baby Bare) sampled by stainless-steel probes driven directly into rock 8 , and also an Ocean Drilling Program borehole (Hole 1026B, freely venting at the time of sampling) accessing basement fluid. Hydrologic and chemical data indicate ridge-flank fluid circulation here is independent from on-axis circulation 9 ; thus, DOC in these fluids is not directly linked to high-temperature circulation or axial hydrology or processes.
Ridge-flank UDOC concentrations (1.2-1.4 µM) were low, consistent with total DOC data 6 , whereas UDOC concentrations at lower-temperature axial sites (6-11 µM) were more similar to background sea water (6.8 µM; Table 1). The offset between on-axis and ridge-flank values is consistent with total DOC concentrations 6 , and also with the large background seawater admixture expected at on-axis lower-temperature sites (Supplementary Methods). Background seawater UDOC properties, including overall UDOC recovery (6.8 µM), and isotopic values (δ 13   the deep Pacific Ocean 5,10 , and indicate no substantial effects from our isolation protocols on isotopic values. Both stable and radiocarbon values from the on-axis fluids showed clear, but small, offsets versus background seawater UDOC ( Table 1). The δ 13 C values for on-axis fluids are slightly higher relative to seawater UDOC, whereas 14 C values are lower (Fig. 2a,b; Table 1). This suggests the addition of locally produced DOC, however superimposed on a much larger background of deep-ocean material. This is consistent with total DOC data 6 and also expected hydrodynamics: on-axis lower-temperature fluids are variable mixtures of entrained local sea water and high-temperature fluid 7,11 . Broader sampling of on-axis lower-temperature fluids at the JDFR has indicated that although DOC concentrations are variable, on average these fluids represent a net DOC source 6 . Our results are consistent with this picture, because on-axis lower-temperature DOC composition at any specific site would be expected to reflect mixtures between background seawater organics (always quantitatively dominant) and variable amounts of added material.
In contrast, at ridge-flank sites, δ 13 C values are sharply offset from seawater values (Fig. 2). The δ 13 C values are both much lower (−26 and −35 ) relative to either seawater DOC or surfacederived marine organic carbon (−21 to −22 ). These values are similar to those characteristic of chemoautotrophic organic matter in modern and ancient environments 12,13 , and also many groups of chemoautotrophic microorganisms 14 . Autotrophic microbial δ 13 C values should derive from both fluid dissolved inorganic carbon (DIC) δ 13 C and subsequent autotrophic fractionation. When depleted DIC δ 13 C values in these fluids are taken into account  Furthermore, the more depleted UDOC δ 13 C at Baby Bare is also consistent with the offset in DIC δ 13 C between the two sites 7 , as would be expected for chemosynthetic production. However, significant variation in chemosynthetic fractionations 14 , coupled with probably diverse microbial sources and crustal habitats associated with these systems 8,15 , suggests caution in attempting to interpret δ 13 C values in terms of any specific organisms. Ridge-flank UDOC 14 C values for Baby Bare and Hole 1026B samples are also much lower (−772 and −835 respectively), relative to both local deep-seawater UDOC (−443 ; Table 1), and total DOC anywhere in the world ocean 5,16 . If the dominant UDOC source were chemoautotrophy, then 14 C values for organic matter would also be expected to be similar to 14 C of the DIC in the same fluids. In fact, UDOC 14 C values correspond very closely with fluid DIC 14 C (ref. 7; Fig. 2b). Although radioactive decay of 14 C in entrained seawater DOC could also produce older DOC, this would not influence stable carbon isotopes, and thus could not explain the depleted δ 13 C values. Taken together, the chemoautotrophic δ 13 C values and close correspondence between 14 C values of UDOC and local DIC strongly support the conclusion that dissolved organics vented are autochthonous, synthesized de novo by microbial communities in the crust.
There are also alternative environmental interpretations that might be considered for these data, such as selective degradation of oceanic DOC, and sedimentary or terrestrial influences. For example, as total DOC concentrations are low, highly selective degradation of ocean DOC might potentially alter the isotopic value of remaining material. However, isotopic change with DOC degradation is not observed in the overlying ocean 5 , suggesting this as unlikely. Furthermore, the operational lipid fraction is the only component having strongly offset 13 C and 14 C values versus total UDOC 17 , but this is present in only trace amounts (0.3% or less 17 ), and so could not explain our results through selective concentration. UDOC samples in fresh water can also have low δ 13 C values 18 ; however, terrestrial contribution to deep-Pacific dissolved organic matter is also trace 19 , and background Cascadia Basin UDOC (−21.1 ) bulk isotopic values suggest no significant terrestrial component. Sediments are another potential organic-matter source, because inorganic fluid composition has suggested some sedimentary influence in JDFR 20,21 . Radioactive decay of organic 14 C in sediments also would result in lower 14 C organic matter; however, in this case δ 13 C values would remain similar to overlying primary production that created them (∼−21 to −22 in this region 22 ). In particular, nearby sediment porewater DOC maintains very similar values down to near basement (−20 to −22 ; ref. 23), inconsistent with a sedimentary explanation. Finally, abiotic organic-carbon synthesis has also been suggested in the JDFR system 15 , representing an alternative carbon source possibly consistent with both δ 13 C and 14 C values. Although a major abiotic contribution seems unlikely (and any low-molecular-weight products from Fischer-Tropsch synthesis would not be retained in our high-molecular-weight samples), such compounds could comprise an electron and carbon source for chemolithotrophic bacteria 15,24,25 . This pathway would not fundamentally change our interpretation (because the abiotic organic matter would also have the 14 C of its source DIC pool), but it would add an important mechanistic step between fluid DIC, microbial biomass and high-molecular-weight DOC, and could also contribute to low δ 13 C values 26 . Further detail on these alternative scenarios is given in the Supplementary Discussion.
Together, these data indicate that DOC vented from lowertemperature ridge-flank regions can represent an unrecognized source of DOC to the subsurface ocean. In particular, the material we measured (at 11,400-14,000 14 C yr bp) is 2.5-3 times older than deep-ocean DOC. Although DOC concentrations across different crustal environments may vary widely 24 , any organic carbon ultimately derived from crustal DIC would necessarily represent an analogous source of 14 C-depleted DOC to the deep sea. These observations pose the question of possible influence on the deep ocean's DOC reservoir, in particular because low 14 C values are a central observation for interpreting its character and cycling 5,16 . Globally, fluid flux from lower-temperature ridge-flank hydrothermal systems is far higher than axial venting, and may approach total riverine input to the world ocean 1,2 . However, as few DOC studies are available, it is difficult to quantitatively assess possible carbon fluxes. Nevertheless, if one assumes a total ridge-flank fluid input to the world ocean between 4.8 ×  Although such contributions may seem relatively small, it is the very low 14 C signature, in combination with recent recognition of the varied and dynamic nature of the deep-ocean DOC reservoir, that may be most important. Recent work showing widely different 14 C values for specific DOC biochemical classes has been interpreted to indicate a spectrum of cycling rates 17 , but also suggests that vented low-14 C material with distinct composition could disproportionately influence individual DOC subcomponents in the deep sea. However, an important overall caveat is that the ultimate contribution to the deep-ocean reservoir will depend strongly on the relative residence time of vented organics. This is well illustrated by considering riverine DOC: its contribution to the total deep-ocean DOC pool is minor (probably owing to relatively rapid photochemical/microbial degradation 27 ), despite having a global input sufficient to support all subsurface DOC turnover 4 . If the vented material is more refractory than average deep-ocean dissolved organic matter, it could in fact build up to comprise a greater proportion than the 1-5% input estimate would imply, potentially biasing our view of deep DOC recalcitrance; however, if degraded rapidly once in the deep ocean, then the ultimate influence might be small or negligible. At present, as neither composition nor relative persistence of vented material is known, it is not possible to clearly assess quantitative significance.
Ultimately, the central implication of these data may be that ridge-flank circulation can not only support an indigenous biosphere, but one expansive enough to export substantial fixed carbon to the overlying ocean. Together, our data suggest that vented DOC from this system is almost entirely synthesized by autotrophic bacteria. Earlier work indicated the presence of thermophilic chemoautotrophic microbial strains at these sites 8,15 ; however, the low DOC concentrations alone suggested that ridge-flank circulation acts mainly as a carbon sink 6 . In contrast, our isotopic and concentration data indicate that over the multi-millennial timescales of ridge-flank circulation 7 , deep-ocean DOC is largely removed, and replaced with new material having very different isotopic, and probably compositional, character. This interpretation is strongly supported by a growing amount of literature demonstrating a widely distributed deep microbial biosphere with surprisingly diverse metabolic capabilities 23,28,29 , as well as the expected efficient degradation of DOC during long-term circulation over microbialcovered surfaces (Supplementary Discussion). Recent data demonstrating chemoautotrophic bacterial populations supported directly by basalt-seawater chemical reactions 28 may be particularly relevant to this system, because the energy potential from this mechanism alone may be able to support a biosphere on par with the entire overlying ocean 29 . The overall impact of many ridge-flank systems may thus be to simultaneously act as a 'scrubber' for surface-derived organic carbon, and a source of new, chemosynthetic material. Understanding the ultimate impacts on oceanic biogeochemical cycles will require determining the chemical composition and reactivity of crustal fluid DOC, coupled with broader global sampling.

Methods
Fluids were obtained from a variety of hydrothermal sources using methods adapted for each site, including stainless-steel probes driven into the exposed rock outcrop at Baby Bare, direct sampling of the over-pressured Hole 1026B, and specialized sampling devices for on-axis diffuse fluid vents. Oceanic water samples were also collected using conductivity-temperature-depth casts in Cascadia Basin (2,540 m depth) far from ridge influence, and isolated UDOC was used as our background seawater reference. Large volume crustal fluid samples were recovered from the sea floor using a custom-built large-volume elevator sampler, and inorganic composition data indicated that the off-axis samples we obtained were highly pure crustal fluid (Supplementary Methods). Furthermore, fluids in the passive-intake sampler design contacted only Tedlar bags and Teflon tubing, which were acid-cleaned and rinsed with Milli-Q water before each deployment. Tedlar sample bags treated with this cleaning protocol were tested before cruises and found to contribute no detectable DOC (ref. 6). Samples for total DOC measurements were also collected by collaborators from the identical fluid sources, using a small-volume sampler employing the same sample bags 6 . To isolate sufficient DOC for 14 C analyses, tangential flow ultrafiltration was used to concentrate the ultrafiltered higher-molecular-weight fraction of DOC (UDOC; 0.1 µM-1,000 D). UDOC subsamples were diluted to ∼1 litre with pre-irradiated purified Milli-Q water, and quantitatively oxidized to CO 2 with a mercury arc lamp photochemical reactor system for 14 C analysis 30 . Accelerator mass spectrometry (AMS) measurements were made at UCI Keck Carbon Cycle Accelerator Mass Spectrometry facility. All isotopic measurements were blank-corrected for graphitization and AMS analysis, and also corrected with error propagation for the ultraviolet-extraction procedural blank. Further details on sampling, DOC isolation and AMS measurements are provided in Supplementary Methods.