Introduction

Volcanic eruptions have played a key role in the evolutionary history of Earth, greatly impacting the Earth system. Most of the volcanic activity on Earth occurs in or close to the ocean, at subduction zones, mid-ocean ridges, and hot spots1,2. During explosive eruptions, volcanic ash is produced and can rise tens of kilometers into the atmosphere3, resulting in the deposition of >1km3 of material per year worldwide2. In the surface ocean, the iron fertilization effect of volcanic ash in oligotrophic seawater has been confirmed to promote the marine primary productivity via the alleviation of nutrient deficiencies4,5,6,7,8. However, the effect of volcanic ash input on carbon cycle in the deep ocean remains poorly understood9.

Volcanic ash eventually settles into the deep ocean and comprises an high proportion of seabed sediments10, but its impact on carbon cycling in these sediments is not well known. It has been shown that volcanic ash layers harbor a unique microbial community when compared to surrounding clay sediments11,12, but other than these preliminary investigations, it is unclear the extent to which this phenomenon occurs. Other work has indicated the ability of volcanic ash to enhance the burial of organic carbon in marine sediments9, indicating the possibility of a link between carbon cycling and ash deposition. A study of the organic matter associated with volcanic ash showed it was isotopically more negative than sedimentary carbon13, suggesting indigenous microbial activity and organic carbon production.

Such a suggestion is logical, as reduced iron and sulfur atoms in the mafic minerals comprising basalts (and therefore basaltic ashes) can act as electron donors, thus supplying energy for chemolithotrophs to transfer inorganic carbon into cellular organic carbon14,15,16. Evidence for such microbial activity may be seen in tubular or granular structures in basalts, interpreted as biogenic alteration of sub-seafloor volcanic glass17,18. Given the similarity in mineralogy and geochemistry between extrusive basalts and volcanic ash, we hypothesize that the ash alteration could feed similar chemolithotrophic activity in the seabed, which may contribute to the production of biomass during diagenesis. To test this theory, we investigated the distribution of isoprenoid and branched glycerol dialkyl glycerol tetraethers (isoGDGTs and brGDGTs, for structures see Supplementary Fig. 1), they are distributed in a wide variety of environments and are used as proxies for temperature, pH and salinity19,20. IsoGDGTs dominate the membrane lipids of marine Thaumarchaeota21, while brGDGTs can be produced by Acidobacteria which are living in terrestrial environments22,23,24. However, there are relatively high abundance of brGDGTs in typical chemolithotrophic systems, such as hydrothermal field, cold seeps and ocean crust rocks25,26,27,28. These environments are characterized by active geological processes that produce sufficient chemical sources for chemoautotrophs to be the primary producers in ecosystems29,30,31. Therefore, the composition of brGDGTs proxies has recently been suggested to be used as indicators of chemolithotrophic systems and to trace the in situ production of organic carbon. We performed these analyses on Mariana Trough tuffs and adjacent background sediments (the locations are shown in Supplementary Fig. 2).

Results

Distribution of TOC and GDGTs in tuffs and sediments

In this study, the abundance and δ13C of the bulk total organic carbon (TOC) in tuff samples ranged from 0.02% to 0.73% (Supplementary Data 1, mean = 0.13 ± 0.16%, n = 18) and from −27.21‰ to −21.17‰ (Supplementary Data 1, mean = −24.83 ± 1.54‰, n = 12), respectively. The concentration of brGDGTs and isoGDGTs in tuff samples varied from 0.002 μg/g TOC to 54.690 μg/g TOC and 0.003 μg/g TOC to 119.756 μg/g TOC, respectively. Generally, the ether lipid composition in tuffs was remarkably different from that in background sediments, which was characterized by the enrichment of brGDGTs, especially in TVG02-2, with the highest proportion of brGDGTs up to 87% (Supplementary Data 1). The average relative abundance of brGDGTs in the tuffs was 30%, compared with 5% in the background sediments. IsoGDGTs of tuffs and background sediments had similar distribution characteristics to that of normal marine sediments, which were dominated by GDGT-0 and Crenarchaeol19, which accounted for 31-60% and 29–47% of the total isoGDGTs in tuffs, while they were 26–30% and 48-51% in the background sediments (Supplementary Data 1).

The distribution of brGDGTs was more different between tuffs and background sediments compared with isoGDGTs. The tuffs were dominated by hexamethylated brGDGTs (25–73%), followed by pentamethylated brGDGTs (15–57%) and tetramethylated brGDGTs (11–22%), while their abundances were 59–64%, 15–21% and 20–22%, respectively, in the background sediments. The compound IIIa’ (acyclic hexamethylated 6-methyl brGDGTs) had the highest abundance in brGDGTs among all samples, with averages of 36% and 56% in tuffs and background sediments, respectively (Supplementary Data 1). The domination of IIIa’ in background sediments was coherent with Mariana Trench sediments where marine production of brGDGTs was considerable41. The results indicate that tuffs display unique brGDGT distribution characteristics, potentially produced under low temperature and slightly high pH conditions.

Fig. 5: Multivariate statistical analysis diagram based on relative abundance of 15 brGDGTs.
figure 5

a Principal component analysis; b linear discriminant analysis. Oval shaded areas indicates 95% confidence intervals. Tuffs and background sediments from this study, red and yellow, respectively; green, orange, gray, violet circles represent Mariana Trench (MT), Kermadec Trench (KT), New Britain Trench (NT), Atacama Trench (AT) sediments, respectively, and the data from refs. 32,39; and indigo circles represent Berau delta sediments, data from ref. 72.

Full size image

The brGDGTs share many characteristics with those of chemolithotrophic ecosystems, such as hydrothermal fields28,36,42,43,44,45, cold seeps27, terrestrial serpentinites37, and altered rocks from the Mariana Trench25,26. Specifically, all tuffs have higher BIT values and lower Ri/b ratios than normal marine sediments (see Fig. 6a, b). Notably, tuffs are most similar to those from hydrothermal deposits (Fig. 6c–f), with the enriched brGDGTs characterized by low MBT (degree of methylation of brGDGTs) and high DC (degree of cyclization of brGDGTs), #ringstetra and #ringspenta (the weighted average number of cyclopentane moieties for tetra- and penta-methylated brGDGTs, respectively). Therefore, we suggest the high enrichment of brGDGTs in tuffs were associated with chemolithotrophic systems.

Fig. 6: Distribution range of brGDGTs proxies in different environments.
figure 6

The box plots of a BIT index, branched isoprenoid tetraether; b Ri/b, abundance ratio of isoGDGTs and brGDGTs; c MBT, degree of methylation of brGDGTs; d DC, degree of cyclization of brGDGTs; e, f #ringstetra and #ringspenta, the weighted average number of cyclopentane moieties for tetra- and penta-methylated brGDGTs, respectively. Error bars indicate the range within 1.5 IQR. The dark and light orange color represent tuffs and marine samples, respectively; while samples from chemolithotrophic systems and land are represented by yellow and gray, respectively. Data sources: tuffs and background sediments from this study, Mariana Trench sediments32, altered rocks and sediments25,26,35, hydrothermal deposits28,42,43,44,45, cold seep carbonates27, serpentinites37, and the data source of global soil and peats datasets please see Supplementary References 9–24.

Full size image

Mineral alteration in tuffs and implications for tracking chemolithotrophic activities

Water-rock reactions can provide substrates for microorganisms, who in turn accelerate the rate of mineral alteration by changing their solubility and element valence31,46. For instance, oxidation of reduced iron (or other metals) in olivine and pyroxene by seawater can produce H247,48, which is the major electron donor in chemolithotrophic metabolic pathways49. In this study, the presence of the Fe-rich clay mineral chlorite in some tuffs indicates the occurrence of low-temperature water-rock reaction of mafic minerals50. Furthermore, the positive La anomaly in these tuff samples likely indicates the processes of element exchange between tuffs and seawater51, while their different degrees of positive Ce anomaly suggests the various states of oxidation in micro-environments during tuff alteration52. As a result, we consider that the minerals in tuffs have undergone water-rock reaction, releasing nutrients to the surrounding environment. This suggests volcanic ash input provides favorable environmental conditions for chemolithotrophic ecosystems to thrive.

We observe apatite growing around altered heamatite (Fig. 1b), possibly showing a life-induced enrichment of phosphorus in these microregions53. Moreover, as shown in Supplementary Figs. 6a, b and 7, the fractures of altered olivine ((Mg, Fe)2SiO4) filled with hematite, formed via the reaction between seawater and olivine54. Raman spectrum results also showed the co-occurrence of peaks of CM and the characteristic peak of hematite at the altered regions of these minerals (Supplementary Fig. 6m). The spatial distribution of CM was closely related to the vein-like structure, linking the production of organic carbon to the formation of ferric oxide minerals. This phenomenon does not appear to be limited to the Mariana subduction zone. Previous work has shown the proportion of Fe-bound organic carbon in ash layers is roughly twice as high as in background sediments in the Bering Sea13. All this points toward CM in deep sea rocks being the product of chemolithotrophic communities26,55,56, and/or produced via mineral-catalyzed Fischer–Tropsch reactions57,58. This biogenic and/or abiogenic organic matter may also provide support for the generation and maintenance of other microbial life in rocks59,60. Reference 61 confirmed the existence of CO2-fixing autotrophic bacteria in volcanic ash that oxidize sulfide to produce energy. As such, the sulfur-containing minerals found around amorphous CM in tuff are likely accessory substances of chemolithotrophy by sulfur oxidizing bacteria.

Summarily, the co-occurrence of essential elements for life such as C, N and S in amorphous CM of these altered mafic minerals, as well as enrichment of authigenic brGDGTs in tuff samples, indicate that the organic matter has been synthesized in situ by microbial activity during tuff alteration. Simultaneously, numerous biogenic tubular structures were found on the surface layers of some tuffs (TVG02 and TVG08) ranging in length from several hundred micrometers to several millimeters (Supplementary Fig. 8). This result further indicates that the organic matter produced by tuffs alteration is likely to serve as food for other higher trophic organisms and participate in the deep-sea carbon cycle through the energy transfer of the food chain62.

Biomass estimation of chemolithotrophic system fueled by volcanic ash alteration

In order to assess the potential scale of chemolithotrophic biosynthesis of organic carbon resulting from volcanic material alteration in the deep ocean, we performed calculations using the annual global volcanic ash fluxes into the ocean from 1960 to present (data from ref. 2). This means that all our calculations are based on levels of fresh volcanic ash deposited to the seafloor. It has been widely reported that biogeochemical reactions, including iron, sulfur, H2 and, CH4 oxidation by dissolved oxygen and nitrate12,14,63,64, in basalt can provide chemical energy for microbial metabolism. Although the rate of H2 production and the rate of Fe and S oxidation during ash alteration is still unknown, the chemolithotrophic activities in volcanic ash fed by mineral alteration as well as the alteration rates are likely to be similar to those in basalt bedrock. As such, we use assessments of Fe and S oxidation from basalt minerals to make our estimate of total carbon production.

The average Fe content in global ashes is 6 ± 1 wt.%, and iron flux released during volcanic ash diagenesis ranges from 91 to 493 Gmol yr−1, with a median value of 180 Gmol yr−1, representing the estimates under the low, medium and high ash volume scenarios, respectively (ref. 2). Based on the empirical relationship between total Fe and S content established in basaltic glasses65, the calculated S content is 0.079 ± 0.019 wt. %, and global input of volcanic ash into the ocean supplies from 2.1 to 11.3 Gmol yr−1 of sulfur. Meanwhile, the calculated Fe2+ flux in ash is between 66.1 and 356.6 Gmol yr−1 based on the measured Fe2+/Fe3+ ratios from ref. 66. According to ref. 14, half of Fe3+ was produced by consuming dissolved oxygen and nitrate to oxidize Fe2+, while the other half of Fe2+ is oxidized to Fe3+ by reacting with water to generate H2. This is based on the assumption that 1 mol Fe2+ can produce 0.5 mol H2. Therefore, Fe involved in aerobic and anaerobic oxidation in volcanic ash can also be calculated (see Supplementary Table 2).

We assume that all the reductive reactants released from volcanic ash alteration and the resulting chemical energy have been fully utilized by microorganisms. The Gibbs free energy was calculated based on the pressure (P = 250 bar) and temperature (T = 10 °C) at the average sampling depth (2500 m) of the tuffs. The calculation results showed that the chemical energy released by water rock reactions in tuffs could provide 1.1 × 1013 to 6.2 × 1013 kJ yr−1 chemical energy, and produce 0.7 × 1011 to 3.7 × 1011 g yr−1 biogenic carbon, with a median value of 1.3 × 1011. This estimates are one order of magnitude lower than the maximum primary production of organic carbon in ocean crust basalts and aquifers (1012 g yr−1, refs. 14,67,68), and subseafloor hydrothermal vents (1.4 × 1012 g yr−1, ref. 69), but are not insignificant.

In summary, our work investigating the alteration of tuffs suggests they are one of the potential locations for in situ production of organic carbon in marine sediments. However, it should be noted that our estimates of the size of this carbon source produced in surface ash layers may be overstated, as they are based on the assumption that microbial activities are involved in all volcanic ash oxidation reactions. In addition, due to the heterogeneity of the seafloor environment (e.g., protolith composition and ash proportion variability), the potential to produce reducing substrates likely varies widely in different regions. The more negative δ13C values of organic carbon found in ash layers than normal sediments from hundred meters below seafloor may indicate that deeper volcanic ash may also contribute to the survival of microorganisms13. Although it is unclear how sediment depth affects microbial communities in volcanic ash, the estimates in this paper are likely to be underestimated when considering volcanic ash in deeper sediments. Taken together, these uncertainties likely lead to unconsidered variability in the bioavailability of volcanic ash. Although it is known that large amounts of volcanic materials have been produced and eventually deposited into the seafloor during the geological history, quantifying absolute volumes of ash and the proportions of alteration remain a difficult task, which makes it difficult to precisely evaluate the global implications of our findings. To better limit these uncertainties, more seabed (and subseafloor) ash from other locations must be investigated to aid global quantification in the future.

Materials and methods

Sample collection

In October 2018, 18 tuff samples were collected from six sites in the Mariana Trough using the TV-grabber. The water depths of sampled sites were distributed between 1240 and 3535 m, with an average depth of 2500 m. In October 2019, three sediment push cores were collected by the manned submersible Shen-Hai-Yong-Shi at three sites with little influence of volcanic ash input in the water depths of 3247, 4010 and 2272 m, respectively. The 0–2 cm surface layers of each push core were taken as background sediment. After the samples were collected, they were divided on board and then immediately stored in −80 °C freezer for preservation.

Geochemical and mineralogical analyses

The freeze-dried tuff samples were ground to 200 mesh with an agate mortar, and each sample was analyzed by Bruker D8 Advance X-ray diffractometer to infer the mineral compositions at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences. The working conditions were 30 mA and 40 kV with CuK-alpha radiation, ranging from 3 to 85° 2θ. Phase analysis was completed using Jada 5.0 software. The sample was dissolved in HNO3 + HF solution on a hot plate. Then the eluted samples were spiked with internal standard In (1 ppm) and diluted with 2% HNO3. Quantitative analysis of trace elements was done via Agilent 7700X Inductively Coupled Plasma Mass Spectrometer at the State Key Laboratory of Marine Geology, Tongji University, China.

After subsampling of fresh tuff samples over drying, the surface of the tuffs was observed under different magnifications using a stereoscope. Thin sections were made, rinsed with Mili-Q water, dried and then blown with nitrogen to remove the contamination on the slices surface before microscopic observation at the Institute of Deep Sea Science and Engineering, CAS. The sections were first observed by polarizing microscope with mainly transmitted light and supplemented by reflected light. Next, the scanning electron microscope (Thermo ScientificTM Apreo) was used to observe the morphology and structure of minerals under voltage of 15 kV. Element distributions of thin sections were analyzed using energy dispersive spectrometer with 20 kV and high vacuum. Raman spectrometry (WITec alpha 300R) was used to scan the surfaces on tuff thin sections with cumulative time of 1 s × 3. The Project FIVE 5.2 software was used for subsequent integration and analysis of all Raman spectrum data. The tuff thin sections were finally made into 1-inch tin targets, and was analyzed by nano secondary ion mass spectrometry (Nano SIMS, CAMECA 50L) with a Cs+ ion beam intensity of 1.2 pA. The target area was pre-denuded prior to formal testing to remove contaminants and the conductive coating (Au) produced during pretreatment. The acquisition time of each image was approximately 12 min, targeting the elements C, O, N, P, and S.

The content and δ13C of total organic carbon in tuff samples were analyzed by Sercon Integra2 Elemental Analyzer coupled to an isotope ratio mass spectrometer after removing carbonate with 2 mol/l HCl overnight at the Third Institute of Oceanography, Ministry of Natural Resources, China. The δ13C result is calculated relative to Vienna PeeDee Belemnite as the reference standard. Duplicate samples of acetanilide and IAEA-600 standard (Thermo Scientific) were analyzed along with samples for quality control, which indicated an analytical error of <0.2‰ for δ13C.

Lipid extraction and analysis

After all tuffs and background sediments were freeze-dried, 30 ml mixed solution of dichloromethane and methanol (9:1, V/V) was added to each Teflon tube containing ca. 10 g sample. The total lipid extract was extracted ultrasonically (repeated 5 times) and dried with nitrogen. The GDGTs were separated from total lipids by eluting silica gel column with mixture of dichloromethane and methanol (1:1, V/V). The GDGTs fractions were filtered with 0.45 μm PTFE membrane and then transferred to 250 μl insert vial.

The GDGTs were measured at the State Key Laboratory of Biogeology and Environmental Geology, China University of Geosciences (Wuhan), China. The internal standard C46 GDGT was added before the GDGTs were analyzed using an Agilent 1200 series liquid chromatography connected to an Agilent 6460A triple-quadruple mass spectrometer. During the test, two silica columns (150 mm × 2.1 mm, 1.9 μm, Thermo Finnigan) were connected in series to separate the 5- and 6-methyl isomers of brGDGTs70. Single ion monitoring (SIM) was used to identify specific m/z ions, including 1302, 1300, 1298, 1296, and 1292 for isoGDGTs compounds, and 1050, 1048, 1046, 1036, 1034, 1032 1022, 1020, and 1018 for brGDGTs compounds. The concentrations of GDGT were normalized by TOC content of the samples.

Calculation of proxies related to GDGT

The abundances of 5- and 6-methyl isomers were both considered when calculating GDGT-related indicators. The BIT, Ri/b, and ΣIIIa/ΣIIa indexes were calculated as an indicator to distinguish the origin of brGDGTs between land and ocean according to refs. 33,34,38.

$${{{{{\rm{BIT}}}}}}=\frac{{{{{{\rm{Ia}}}}}}+{{{{{\rm{IIa}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{a}}}}}}^{\prime} +{{{{{\rm{IIIa}}}}}}+{{{{{\rm{III}}}}}}{{{{{\rm{a}}}}}}^{\prime} }{{{{{{\rm{Ia}}}}}}+{{{{{\rm{IIa}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{a}}}}}}{\prime} +{{{{{\rm{IIIa}}}}}}+{{{{{\rm{III}}}}}}{{{{{\rm{a}}}}}}{\prime} +{{{{{\rm{Crenarchaeol}}}}}}}$$
(1)
$${{{{{\rm{R}}}}}}_{{{{{{\rm{i}}}}}}/{{{{{\rm{b}}}}}}}=\frac{\sum {{{{{\rm{isoGDGTs}}}}}}}{\sum {{{{{\rm{brGDGTs}}}}}}}$$
(2)
$$\frac{\sum {{{{{\rm{IIIa}}}}}}}{\sum {{{{{\rm{IIa}}}}}}}=\frac{{{{{{\rm{IIIa}}}}}}+{{{{{\rm{III}}}}}}{{{{{\rm{a}}}}}}^{\prime} }{{{{{{\rm{IIa}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{a}}}}}}^{\prime} }$$
(3)

To evaluate the degree of methylation of brGDGTs, MBT index71 was calculated based on the fractional abundances of brGDGTs. While DC, #ringstetra and #ringspenta were calculated to appraise the degree of cyclization of whole, tetra- and penta-methylated brGDGTs, respectively72,73. Ringspenta-hexa represents the sum of the fractional abundance of cyclic penta- and hexa-methylated brGDGTs25.

$${{{{{\rm{MBT}}}}}}=\frac{{{{{{\rm{Ia}}}}}}+{{{{{\rm{Ib}}}}}}+{{{{{\rm{Ic}}}}}}}{\sum {{{{{\rm{brGDGTs}}}}}}}$$
(4)
$${{{{{\rm{DC}}}}}}=\frac{{{{{{\rm{Ib}}}}}}+{{{{{\rm{IIb}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{b}}}}}}{\prime} }{{{{{{\rm{Ia}}}}}}+{{{{{\rm{Ib}}}}}}+{{{{{\rm{IIa}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{a}}}}}}{\prime} +{{{{{\rm{IIb}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{b}}}}}}{\prime} }$$
(5)
$$\#{{{{{{\rm{rings}}}}}}}_{{{{{{\rm{tetra}}}}}}}=\frac{{{{{{\rm{Ib}}}}}}+2\times {{{{{\rm{Ic}}}}}}}{{{{{{\rm{Ia}}}}}}+{{{{{\rm{Ib}}}}}}+{{{{{\rm{Ic}}}}}}}$$
(6)
$$\#{{{{{{\rm{rings}}}}}}}_{{{{{{\rm{penta}}}}}}}=\frac{{{{{{\rm{IIb}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{b}}}}}}{\prime} +2\times {{{{{\rm{IIc}}}}}}+2\times {{{{{\rm{II}}}}}}{{{{{\rm{C}}}}}}{\prime} }{{{{{{\rm{IIa}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{a}}}}}}{\prime} +{{{{{\rm{IIb}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{b}}}}}}{\prime} +{{{{{\rm{IIc}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{c}}}}}}{\prime} }$$
(7)
$${{{{{{\rm{Rings}}}}}}}_{{{{{{\rm{penta}}}}}}-{{{{{\rm{hexa}}}}}}}={{{{{\rm{IIb}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{b}}}}}}{\prime} +{{{{{\rm{IIc}}}}}}+{{{{{\rm{II}}}}}}{{{{{\rm{c}}}}}}{\prime} +{{{{{\rm{IIIb}}}}}}+{{{{{\rm{III}}}}}}{{{{{\rm{b}}}}}}{\prime} +{{{{{\rm{IIIc}}}}}}+{{{{{\rm{III}}}}}}{{{{{\rm{c}}}}}}{\prime}$$
(8)

Principal component analysis was conducted using Canoco 5.0 software, and linear discriminant analysis was performed with the SPSS software to provide a general view of the distribution pattern of brGDGTs.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.