Ecosystem Composition Controls the Fate of Rare Earth Elements during Incipient Soil Genesis

The rare earth elements (REE) are increasingly important in a variety of science and economic fields, including (bio)geosciences, paleoecology, astrobiology, and mining. However, REE distribution in early rock-microbe-plant systems has remained elusive. We tested the hypothesis that REE mass-partitioning during incipient weathering of basalt, rhyolite, granite and schist depends on the activity of microbes, vascular plants (Buffalo grass), and arbuscular mycorrhiza. Pore-water element abundances revealed a rapid transition from abiotic to biotic signatures of weathering, the latter associated with smaller aqueous loss and larger plant uptake. Abiotic dissolution was 39% of total denudation in plant-microbes-mycorrhiza treatment. Microbes incremented denudation, particularly in rhyolite, and this resulted in decreased bioavailable solid pools in this rock. Total mobilization (aqueous + uptake) was ten times greater in planted compared to abiotic treatments, REE masses in plant generally exceeding those in water. Larger plants increased bioavailable solid pools, consistent with enhanced soil genesis. Mycorrhiza generally had a positive effect on total mobilization. The main mechanism behind incipient REE weathering was carbonation enhanced by biotic respiration, the denudation patterns being largely dictated by mineralogy. A consistent biotic signature was observed in La:phosphate and mobilization: solid pool ratios, and in the pattern of denudation and uptake.


SI 1 Supplementary Methodology
A model ecosystem experiment was setup in the Desert Biome at University of Arizona's Biosphere 2 based on a design detailed in ( 1,2 ). Briefly, 6 enclosed chambers connected in parallel to a double air purification system (using 2 high-efficiency particulate absorption HEPA filters and 2 UV-B air sterilization light sources, capable of delivering about 1L air sec -1 per module; Germguardian, AC4850CAPT Digital 3-in-1 Hepa Air Purifier System) hold 288 experimental columns (30 x 5 cm internal diameter; Figure s1). Except for control columns, which were placed first in the direction of air flow, the columns were grouped by rock type and randomly distributed in each module. The temperature in the modules/experiment followed the one in the Desert Biome, which was kept at a mean temperature of 19±4°C, relative humidity of 48±19%, and natural O 2 /CO 2 saturation conditions. Modules' aerial chambers experienced additional solar radiative heating of about 5°C above the Biome average during the day. Belowground (soil) compartments were light and thermally shielded, which prevented overheating.

SI 1.2 Biological signature index
To infer a biological signature index, the following lines of evidence were considered: (a) REE sources in phosphate minerals, REE-oxides, and minor minerals (ilmenite, titanite, zircon, allanite) in the used rocks; (b) REE mobilization under biotic treatment was radius dependent, L-REE exhibiting increased mobilization under biota; and, (c) P is the mineral constituent of principal biotic relevance. Based on these premises we propose using an abiotic control-normalized ratio of La : phosphate water concentrations as biotic signature index, following the equation s1.

SI 1.3 Global denudation estimates
Estimated values for global (G) REE denudation rates (moles * year -1 ) were inferred by stoichiometric adjustment of Na-normalized total REE in our experiment (i) to global Na values from river data 3 , according to equation s2.
To infer REE contribution by different rocks, the global Na value from rivers was adjusted (multiplied) to the relative contribution (%) of the rock to the global exposed lithology described in ( 4 ). Abiotic and biotic contribution to the global cycle was estimated from their ratio in the experiment.   . REE mineral source. Back scatter images of studied rocks and high current electron microprobe maps of representative L-, M-and H-REE in the studied materials, and the limited number of minerals that host them. Scattered X-rays appear brighter relative to background (increased contrast) with increasing REE density. Ce-Nd-La-Pr oxide in basalt, and allanite and xenotime in schist have been identified as main REE mineral hosts. Pixel size in multi-element maps has been exaggerated 4X to improve perception of otherwise very low levels.  ]). Mineral formulas were calculated from electron microprobe elemental analyses (a mean of several point analyses), and mineral abundances were estimated quantitatively by Rietveld analysis of X-ray diffraction data.  (Table s1). It is higher in Ca, Mg and Fe, and lower in Si than other studied rocks. The vesicular structure of basalt suggested comparatively faster weathering potential. Rhyolite was rich in feldspars and quartz (Table s1). The geochemistry of rhyolite was similar to granite, but it was richer in Si and Na and had less Ca ( 2 ). Schist contained a high proportion of Mg-rich phengite, a transitional phase between muscovite and caledonite. Table s2. Total REE, Sr (major weathering indicator, for comparison), total organic carbon (TOC), water balance -representing the water used by biota (transpired and tissue-stored) + column evaporation (expressed as difference between input and output volumes) and mean measured pH and electrical conductivity (EC) over the 20-months experiment. Mycorrhiza infection rates are also presented for each substrate.

. Water
There was a strong correlation among concentrations of different REE in pore water (Table s3) consistent to their group behavior.