Global carbon fixation in Earth's subsurface biosphere
The subsurface of our planet hosts 15% of Earths biomass and plays a key role in mediating the exchange of volatiles and elements between deep, long-residence-time geological reservoirs and rapidly cycling surface environments, influencing planetary climate and habitability. While a significant fraction of subsurface microorganisms rely on surface-derived organic carbon, an unknown portion is sustained through chemolithotrophic carbon fixation. Despite its importance, the global diversity and distribution of microbial carbon fixation pathways in the subsurface, and the environmental drivers shaping them, remain poorly constrained. Here we systematically characterise carbon fixation pathways for 412 subsurface metagenomes, including 242 new metagenomes, and compare them to surface-derived datasets. We find that subsurface environments span a broader physicochemical space than surface systems and support a higher abundance and diversity of carbon fixation strategies. Using colocated geochemical data spanning >50 variables, we show that the reductive tricarboxylic acid cycle and the reductive acetyl-CoA pathway are enriched in reducing, geochemically evolved fluids. We use the metagenomic results together with previously published carbon fixation rates in the subsurface to derive a global continental subsurface carbon fixation rate of [~]2.65 Pg C yr-1 (range: 0.31-2.99). This represents [~]2% of terrestrial photosynthetic primary production, and is an order of magnitude higher than geological fluxes between the surface and the subsurface. These results identify the subsurface as a reservoir of autotrophic strategies organized along geochemical gradients, contributing substantially to the global carbon cycle. One Sentence SummaryThe subsurface is a widespread, environmentally and functionally diverse reservoir of autotrophic carbon fixation pathways that can contribute substantially to the global carbon cycle, fixing [~]2.65 Pg C yr-1 in continental settings alone.