Geochemical constraints on the Hadean environment from mineral fingerprints of prokaryotes

The environmental conditions on the Earth before 4 billion years ago are highly uncertain, largely because of the lack of a substantial rock record from this period. During this time interval, known as the Hadean, the young planet transformed from an uninhabited world to the one capable of supporting, and inhabited by the first living cells. These cells formed in a fluid environment they could not at first control, with homeostatic mechanisms developing only later. It is therefore possible that present-day organisms retain some record of the primordial fluid in which the first cells formed. Here we present new data on the elemental compositions and mineral fingerprints of both Bacteria and Archaea, using these data to constrain the environment in which life formed. The cradle solution that produced this elemental signature was saturated in barite, sphene, chalcedony, apatite, and clay minerals. The presence of these minerals, as well as other chemical features, suggests that the cradle environment of life may have been a weathering fluid interacting with dry-land silicate rocks. The specific mineral assemblage provides evidence for a moderate Hadean climate with dry and wet seasons and a lower atmospheric abundance of CO2 than is present today.


Results and Discussion
The studied microorganisms show chemical similarities in their inorganic components, possibly inherited from LUCA and its surrounding environment (Fig. 1a). An alternative explanation for this similarity is that the cells incorporated their trace element patterns from growth media with similar trace element signatures. To avoid this possibility, the sets of corresponding pairs of observations on element contents in grown cells and nutrient media were compared using the statistical tests for two related samples (Supplementary Discussion). The implemented tests, as well as direct comparison with compositions of growth media (Supplementary Table 3), provide evidence that prokaryotes tend to control the content of most measured trace elements. In order to accelerate the cultivation of living cells, the standard laboratory cultivation media used in our experiments share similarities with prokaryotes with respect to their main cations, such as Mg 2+ , K + and Ca 2+ . The variation in the elemental composition measured for E. coli cultivated at different conditions is comparable with fluctuations among all tested prokaryotes (Fig. 1a, Supplementary Discussion). In fact, growth medium composition, temperature, pH, redox conditions, and the life stage of cells play a role in the chemical makeup of microorganisms 15,16 . Experiments 16 with A. violaceus cultivated in media with differing Mn content, one of the most flexible constituents in living cells, revealed that this element could be accumulated by those bacteria at wide interval (4·10 −5 -0.2 mol kg −1 of wet weight), which is correlated with medium concentration. However, the consistent growth of A. violaceus can continue only at the favorable interval of 4·10 −5 -4·10 −4 mol kg −1 of wet weight 16 . All prokaryotes considered here were cultivated in media without enforcing their elemental supply and, hence, observed compositions reveal those favorable intervals. The commonality of the chemical composition of prokaryotes we report here was recently confirmed for species living at extreme hydrothermal conditions 17 , demonstrating that the major trends of prokaryotic inorganic composition are insensitive to environmental settings. Our measurements also agree with the limited measurements available in the literature (Supplementary Table 1). We propose that a few outliers in literature reports shown in Fig. 1a and in Supplementary Table 1 for Li, Ti, Zr, Sn, Ba, Th and U stem from experiments with an excess of those elements in the nutrient media, or due to deprecated analytical techniques. Nevertheless, the bulk mineralization, pH and pe − of prokaryotic cytoplasm can vary with environmental parameters 18 , so that any elemental signature of the cradle environment is best discerned not from the absolute concentrations of elements or solution species, but in ratios of their chemical activities (Fig. 1b). These ratios can be interpreted using approaches traditionally employed in analyzing fluid geochemistry. These methods screen against effects caused by distinct ionic strength and type of speciation in living cells. Changes in the bulk elemental content may differ by factors of 100-1000 while retaining the same activity value of the essential solution species. The chemical composition of any natural fluid is established as a balance between dissolution of primary minerals and precipitation of alteration phases. The precipitating minerals are often in equilibrium with the  Table 1). All calculations were performed independently for each species. We found that the fluids corresponding to the composition of each prokaryote are saturated in barite (BaSO 4 ), sphene (titanite, CaTiSiO 5 ), and chalcedony or amorphous silica (SiO 2 ) (Fig. 2a). The saturation index of sphene is pH-sensitive (see Methods). Thus, the saturation indices calculated for species cultivated at strongly acidic or alkaline conditions (A. aceti, A. acidoterrestris, E. coli at pH = 10, E. coli in BS, N. magadii, and V. cholerae) deviate from equilibrium. This mineral association can precipitate from fluids at wide T-P conditions in various rock-dominated environments with low water-rock ratio. Ti-bearing mineral (sphene) is particularly relevant for determining formation conditions because Ti is often considered as an immobile element and does not migrate for long distances 19 . We found those minerals in the weathering profile of Paraná basalts (22.892764°S, 47.089161°W, Fig. 2b). Recently, the same mineral association was reported for the oldest known basalt paleosols (2.76 Ga) observed in Mount Roe 20 . However, weathering specifics in the Hadean is yet unclear and its geochemical features might be distinct even in Neoarchean regoliths.
Mineral fingerprints are encoded in cellular metabolisms as the needs for certain solution species (i.e. Ba 2+ , Ca 2+ , SO 4 2− , SiO 2 ,aq, Ti(OH) 4 ,aq). Ca 2+ and SO 4 2− are the main cytosolic constituents involved in numerous biochemical processes 21 (see Supplementary Discussion). Barite crystals can be used as a gravitational sensor 21 and to control the cell density 22 . Therefore, to maintain its saturation level the product of Ba 2+ and SO 4 2− contents needs to be constant. SiO 2 ,aq or Si(OH) 4 ,aq play a role in glycoprotein stabilization 21 . In addition, silica encapsulation is used by bacteria to aid survival under extreme pH conditions 23 . Titanium is a potential reducing agent responsible for the protection of cells from excessive oxidation. It can also participate in photosynthesis and in the fixation of molecular nitrogen 24 . At the stage of prebiotic evolution, SiO 2 ,aq and Ti(OH) 4 ,aq might catalyze the nucleobase and acyclo nucleoside synthesis 11,[25][26][27][28] . Under acidic and alkaline conditions (pH < 4 and > 8) the total content of Si and Ti in cytoplasm can be significantly higher than the concentrations of essential solution species involved in metabolic reactions (Fig. 3). This forces the cells to accumulate barren solution species, providing additional energetic consumption and shifts in their chemical makeup.
Assuming the obtained prokaryotic compositions sample the composition of natural fluid in which LUCA survived, mineral stability fields and various chemical features of these fluids can be used to estimate the temperature, pH and pe − ranges of the Hadean environment in which life formed. Significant concentrations of redox sensitive trace metals coupled with high iron and sulfur contents impose a lower limit on pe − (Fig. 4a). At values of the reduction potential corresponding to the sulfide-sulfate buffer, As, Cd, Cu, Mo, Pb, Se, V, and Zn cannot be retained in solution and deposit as sulfides, selenides minerals or are captured as impurities by Fe-smectites and pyrite. Because the rare earth elements (REE) patterns do not reveal the well-defined Ce anomaly for all studied prokaryotes (Supplementary Table 2), the upper redox boundary is constrained by precipitation of cerianite (CeO 2 ), approximating the ferrous-ferric iron buffer. The precipitation of titanite strongly limits the pCO 2 in the water-rock system 29 ; hence, carbon would have existed in reduced form, i.e. CO, CH 4 and/or hydrocarbons. This fluid likely interacted with Hadean igneous rocks, whose redox conditions were controlled by the fayalite-magnetite-quartz buffer 2 . Given that the fluid contained reduced C species, igneous magnetite containing ferric iron was probably the main oxidant. To prevent the resulting solution from reduction, the masses of dissolved Fe 3+ and reduced carbon should be comparable. In order to accomplish this, the water-rock ratio had to be < 1 and the reduction potential variable with time, constrained with stability fields of sulfides and Fe 2+ / Fe 3+ buffer. A variable reduction potential is easily reached with alternating wet and dry periods: during wet periods the fluids become more reduced promoting the precipitation of sulfides and during droughts the solution becomes oxidized and could remobilize them. The stability fields of Mn minerals and zincite (ZnO) restricts the activity of H + in solution to acidic and neutral values.
Because barite is thermodynamically stable in contact with organic molecules, the upper temperature limit can be established at ~100-140 °C, where thermochemical sulfate reduction by hydrocarbons starts 30 . This upper limit is compatible with the most thermophilic contemporary terrestrial microorganisms, which can grow at temperatures up to 122 °C 31 . A more stringent constraint on temperature can be obtained with the use of empirical geothermometers (see Methods). The Li geothermometer 32 provides T = 13 ± 6 °C. We can also estimate the temperature and pH of the cradle environment using a mineral equilibrium approach 33 . For the mineral concentrations measured, equilibrium is achieved at 14 ± 9 °C at weakly acidic pH = 6.3 ± 0.3, in good agreement with the Li geothermometer (Fig. 4b,c). The low ambient temperatures suggested by this analysis support the concept of a mesophilic LUCA, as do analyses of ribosomal RNAs and protein sequences 8 . At these parameters, the measured prokaryotic compositions are also in equilibrium with apatite (Ca 5 (OH,Cl)(PO 4 ) 3 ), montmorillonite ((Ca,Mg) 0.3 Mg 0.6 Al 1.4 Si 4 O 10 (OH) 2 ) and berthierine (Fe 2 Al 2 SiO 5 (OH) 4 ) -the latter a common Fe-rich clay mineral in rocks exposed to hydrocarbons 34 , which has been found in the Archean Mount Roe basalt paleosols 20 . The metabolic functions of chemical elements linked with those minerals are considered in the Supplementary Discussion.
Prokaryotes share a common ratio in the concentration of Cl and Br, with Cl/Br in A. acidoterrestris near 2800 and in E. coli in HVCM around 3500. The only literature data available for Cl and Br content shows a Cl/Br ratio of 3600 in phytoplankton 35 . These Cl/Br ratios are well above the oceanic ratio (~650) and the bulk Earth value (420). Solutions or mineral deposits with such high Cl/Br ratios can be formed only in contact with halite (NaCl) or evaporated saline waters 36 .
Cytosolic REE concentrations may further constrain the formation environment of LUCA. All tested prokaryotes concentrate REE by up to a factor of 1000 relative to their concentrations in the cultivation media, with individual variations among elements of this group varying by a factor of 100. The only exception is E. coli in BS, which initially has a well-fitted pattern in its culture medium (Fig. 5a). Selective accumulation and common REE patterns have been observed for various bacterial species both in laboratory and natural environments, and can be used as a biosignature 37,38 . To accumulate REE, cells use a variety of mechanisms such as surface adsorption, adsorption on extracellular biopolymeric substances and biominerals, and accumulation on carboxylate and phosphate binding sites. However, only a fraction of selectively captured REE can be transported in cytoplasm 39 . Our data support this concept. The patterns of prokaryotes did not change during washing, whereas some species lost up to 87% of initially adsorbed REE (Fig. 5a). There is no significant difference in REE patterns between the bacteria and archaea examined here. This evidences that prokaryotes retain a pattern derived from LUCA. Considering potential sources of the REE signature, the bulk REE contents are higher than in river and sea waters (Fig. 5b). Although Ce anomalies may be negative or positive (0.7-1.6), in all species there is a well-defined negative Eu anomaly (0.5-0.8) typical for weathered rocks 40 . La N /Yb N depicts the enrichment of light REE relative to heavy REE with median value 10.7. Those specific patterns resemble the REE distributions in Paraná basalts and especially in Mount Roe paleosols (Table 1). However, the affinity of prokaryotic pattern to the upper continental crust 41 is also well expressed. By contrast, the characteristic feature of oceanic hydrothermal fluids is a remarkable positive Eu anomaly 42,43 . Hence, the signature of those solutions is dissimilar to living cells. It should be noted that REE compositions of both oceanic and continental hydrothermal fluids are very flexible, depending on the composition of interacting rocks.
The concept of a mineral fingerprint in the chemical composition of prokaryotic cells is based on the constant activities of certain solution species of Mg 2+ , Al 3+ , SiO 2 ,aq, HPO 4 2− , SO 4 2− , Ca 2+ , Ti(OH) 4 ,aq, Fe 2+ , and Ba 2+ (Fig. 1b). Applied statistical tests (Supplementary Discussion) evidenced that the contents of corresponding chemical elements are highly distinct in the growth media and prokaryotic cytoplasm. Also, we demonstrated on the example of SiO 2 ,aq and Ti(OH) 4 ,aq that living cells tend to control the content of those solution species, whereas the total contents of Si and Ti vary in wide ranges depending on the pH value in the growth media (Fig. 3). All those components are abundant in cytoplasm and well-known to be essential for biochemical reactions (Supplementary Discussion and references therein). Other considered geochemical features (Li, Cl − /Br − , REE pattern) are attributed to minor components whose metabolic functions and behavior may be less understood, but they also support the concept.
The commonalities in the chemical composition of prokaryotes presented here suggest that the final stage of prebiotic evolution that preceded the emergence of LUCA (or its early evolution) took place in a dry-land environment. Hence, cavities in surface rocks may have been the first ecological niches to be occupied by early microorganisms 44 . Weathered basalts or basaltic komatiites are the first-order candidates to being those host rocks. They were exposed on the Hadean Earth 45 . However, other rock types bearing clay minerals, minor barite, sphene, silica minerals, apatite and enriched in metals and REE might provide similar fluids. Through modeling and comparison of the results with the chemical and mineralogical makeup of prokaryotes, we suggest that during the Hadean there were climatic conditions with alternating dry and wet periods and moderate ambient temperatures. Atmospheric pCO 2 must have been below the present atmospheric level (Fig. 4c), and the cradle fluid was enriched in reduced carbon. The weakly acidic pH suggested by this analysis also implies that the atmosphere was not necessarily CH 4 -dominated 46 , but may instead have contained a hydrocarbon haze 47 that provided C and N-bearing compounds. Alternatively, the cradle environment could have been at a cooling periphery of a dry-land hydrothermal system 11 . However, all other elements necessary for life besides C and N could be mobilized from host rocks. (The works described in references 44   SP, Brazil). Haloferax volcanii (DS70) and Natrialba magadii (ATCC43099) strains were provided by Dr. X. C. Abrevaya (IAFE -UBA -CONICET, Argentina). All prokaryotes were cultivated in the André Tosello Institute. The major challenge in the observation of chemical composition of bacterial and archaeal cells was to obtain cultures without contaminations from the culture medium or any other substance. Therefore, the washing from the nutrient media was made with ultrapure water 17,48 . First, unfrozen strains were placed in activation Stock Culture Agar. Then, E. coli was transferred to Luria Bertani broth (LB) 49 , A. acidoterrestris -to Bacillus Acidocaldarius Medium (BAM) 50 , N. lacusekhoensis and V. cholerae -to Tryptic Soy Broth (TSB) 51 , and A. aceti -to Mannitol medium 52 . BAM was adjusted to pH = 3.5 at 25 °C, the Mannitol medium -to pH = 5, V. cholerae was cultivated in medium with pH = 9, and other bacteria -with pH = 7. The active culture incubation proceeded in those media during 72 h. E. coli, V. cholerae, A. aceti and N. lacusekhoensis were cultivated at 32 °C with shaking at 130 rpm. A. acidoterrestris was incubated at 45 °C without shaking. Haloarchaeal strains were grown in specialized media 53  We aimed to investigate species dwelling in diverse environments. Hence, the cultivated prokaryotes required chemically distinct growth media. To investigate the impact of growth medium composition on the composition of cells, E. coli was grown in the media of other species; i.e., in the culture medium of H. volcanii and in the TSB medium of N. lacusekhoensis. In the culture media of A. acidoterrestris and N. magadii the production of E. coli was depressed and we could not collect a cell volume adequate for measurements. However, despite the increased growth time (120 h), we cultivated E. coli in the LB nutrient medium at pH = 10 attributed to growing conditions of N. magadii. E. coli was also cultivated in medium prepared with aqueous solution interacted with Paraná basalt (E. coli in BS). For this, the pure water acidified with nitric acid to pH = 3.5 was mixed with the basaltic powder in proportion 10 to 1 by weight and heated in an autoclave at 110 °C during 10 days. The elevated temperature and low pH provided a faster extraction of cations. This solution was then filtered and used to prepare the LB medium of E.coli. All culture media were sterilized by autoclaving at 120 °C for 20 minutes.
After incubation, cells were separated by centrifugation at 1000 × g for 10 min at 10 °C. In order to avoid contamination from the culture medium, collected live cells were suspended in sterilized ultrapure water and centrifuged again, as above. We found that further washing iterations did not impact the chemistry of collected samples. The washing with ultrapure water provided a risk of lysis of cells. Therefore, after each wash procedure, the viability of the cultures was controlled through the hanging drop technique 56 . Experimentally, we found that prokaryotes survived in pure water during ~15 minutes. Therefore, all operations were conducted over that time-frame. All samples were analyzed using the gram stain method to ensure no contamination of the culture by other microorganisms 57 . The optical density of the cultures was measured using a spectrophotometer (Thermo Electron) at λ = 600 nm, and 10 µl aliquots were taken for direct cell counting under the microscope.
The twice washed and centrifuged cultures gave up to 2 g of live biomass. They were collected into PFA vials (Savillex). Weighed portions of ~0.5 g intended for ICP-QMS analyses were separated. Then, 1 ml of distilled HNO 3 was added and the combination heated on a hot plate at 120 °C until almost complete evaporation of the samples. Next, 0.5 ml of distilled HNO 3 was added and the heating procedure was repeated. HNO 3 (100 μl) and ultra pure water (1 ml) were used to take up the residue. The vessels were heated for 5 minutes to obtain a clear solution. The solutions were gravimetrically diluted with ultrapure water to 10 g. All mass losses and gains attributed to evaporation of volatile constituents and addition of ultrapure water and HNO 3 were taken into account. Therefore, all yielded observations can be considered as the wet weight of the live cells. The rest of samples were used for Karl Fischer titration and TXRF (Total Reflection X-Ray Fluorescence) analyses.
To control for washing losses and gains, all species were sampled 2 or 3 times before washing, after the first and second washings. The culture media before and after the cultivation were also sampled. All samples were run in duplicate. In bulk, 160 samples were analyzed. The revealed uncertainties demonstrate the standard deviation between 4-6 independent observations. Analytical methods. Measurements of prokaryotic compositions were performed using an ICP-QMS (Inductively coupled plasma-mass spectrometry with quadrupole mass) equipped with collision cell technology (Xseries II , Thermo, Bremen, Germany) in the Isotope Geology Laboratory (IG, Unicamp, Brazil). To provide concentrations in P, S, Cl and Br, TXRF analyses for concentrations of those elements were yielded using the S2 PICOFOX in the Bruker Research Center (Berlin, Germany). The water content in Prokaryotes was observed with Karl Fischer titration in the Institute of Chemistry (Unicamp). Compositions of minerals, elemental maps and BSE images of Paraná basalts were acquired with MEV/EDS Link/Isis Oxford (IG, Unicamp). The bulk chemical composition of basalts was obtained with a Philips PW2404 X-Ray fluorescence (XRF) spectrometer (IG, Unicamp). Also, the types of silica and clay minerals were determined using the EDS/MEV and XploRA (Horiba) Raman spectroscopy in the Institute of Physics (Unicamp). The specific surface areas of minerals were obtained using the BET method with nitrogen gas and the Micromeritics Accelerated Surface Area and Porosimetry (ASAP) 2020 System (Faculty of Chemical Engineering, Unicamp).
Statistical handling. The obtained compositions of prokaryotes and their growth media were statistically compared using the IBM SPSS Statistics 20 software. We implemented the Wilcoxon Signed Ranks test, the Sign test and the two-sample Kolmogorov-Smirnov test (KS test) (see Supplementary Discussion). The Wilcoxon and Sign tests are used to outline the consistent difference between pairs of observations at the assumption that living cells and nutrient media represent the dependent samples. In contrast, the KS test considers the compositions of growth media and prokaryotes as independent samples and compares their empirical distributions. The rejection of the null hypothesis in all tests provides evidence that the elemental abundances in prokaryotes are distinct from their content in nutrient media. The samples' size is 11 and p is 0.05. We also compared E. coli compositions with other prokaryotes using the KS test (p = 0.05). The first sample with size of 5 comprised trace element contents in E. coli and the second sample with size of 6 in other bacteria and archaea.
The interquartile ratio (IQR) was used to compare the dispersion of abundances of chemical elements in prokaryotes (n = 11) and is calculated as follows: where Q 1 and Q 3 are first and third quartiles, respectively. In its turn, the quartiles were estimated for the whole set of IQRs. The chemical elements characterizing by IQR below the first quartile can be characterized by the most stable content in living cells, whereas the highest IQRs exceeding third quartile depict the most variable elements. Also the extent of dispersion was estimated for solution species (i.e. Mg 2+ , Al 3+ , SiO 2 ,aq, HPO 4 2− , SO 4 2− , Ca 2+ , Ti(OH) 4 ,aq, Fe 2+ , Ba 2+ ) constraining the saturation level of minerals considered in the present research.
Geochemical calculations. The inverse modeling, speciation calculations and estimation of saturation indices were performed with the software PHREEQC version 3 58 . The uncertainties in equilibrium constants and dissolution/precipitation rates of minerals were calculated using the CRONO software 59,60 . In the course of geochemical simulations to provide reliable results, we applied different databases of thermodynamic parameters. To calculate data shown in Fig. 2a the llnl.dat 61 was used. This dataset is a part of the standard PHREEQC download. We enforced the database (Database S1) with solubility constants for ilmenite, rutile, and titanite from Supcrt slot07 62 4a,b and for inverse modeling, we used the Thermoddem database because it includes more mineral species than analogical datasets. This tabulation lacks a temperature dependence on solubility constant for chlorapatite. Its values have been calculated using literature data 65 . Ti(OH) 4 (HPO 4 ) 2− was also added. At the preliminary stage of the research, we utilized the modified pitzer. dat 66 . However, because at the found range of pe − and pCO 2 the ionic strength of all simulated solutions was less than 1 M, the extended Debye-Hückel model was used to emulate the activities of aqueous species. The kinetic parameters stem from the tabulation composed by Palandri and Kharaka 67 . All considered databases lack the constants for high molecular weight organic species, which are assumed to be abundant in both cytoplasm and in the hypothetical Hadean environment. To outline the possible impact of organics to the speciation model, we implemented a series of additional calculations with different portions of fulvate 2− and discussed this issue in details in the Supplementary Discussion.
The deposition of sphene (titanite) from aqueous solution was described with the reaction: where K eq is the equilibrium constant and a Ca2+ , a SiO2,aq , a Ti(OH)4,aq are activities of solution species (mol kgw −1molal units, kgw means 1 kg of water). The plots shown in Fig. 3 were calculated in the Microsoft Excel spreadsheets (Database S2) using equilibrium constants at T = 32 °C for Si solution species from Supcrt slot07 62 and for Ti species from refs 63 and 68. To calculate the concentration of NaHSiO 3 ,aq, the content of Na observed in N. magadii was used. The concentrations of Ti species were estimated with values for a Ca2+ , a SiO2,aq , a Ti(OH)4,aq , a HPO4-2 taken the same as predicted for E. coli.
The saturation of prokaryotes with respect to sphene (titanite) is an important finding. Unlike rutile, titanite precipitates at low content of HCO 3 − in fluids. Therefore, its presence can be used as an indicator of CO 2 content in the atmosphere 20,29 . The upper limit for pCO 2 coexisting with titanite in subaerial sediments can be established with the following equation: where K eq represents an equilibrium constant for reaction 2CaTiSiO 5 (sphene) + 2H + + CO 2 ,g = Ca 2+ + CaCO 3 (calcite) + 2TiO 2 (rutile) + SiO 2 ,aq + SiO 2 (quartz) + H 2 O, and a Ca2+ , a SiO2,aq , a H+ are activities of solution species (mol kgw −1 ). Empirical geothermometers are often used to determine the in situ temperature of natural fluids. However, most geothermometers are based on certain contents of Na, Mg, K, Ca 69 . These are important nutrients and their abundance in prokaryotes might be likely considerably modified in the course of biological evolution. According to our observations, the silica (chalcedony) geothermometer 70 reflects the temperature of cultivation of cells. However, whereas their predictions are rather uncertain, Ca/Na, Mg/Li and Na/Li geothermometers 32,70,71 suggest low ambient temperatures for the cradle environment. The Li geothermometer 32 uses the following expression:°= predicted, for A. acidoterrestris, -T = 0 °C, for E. coli in BS, -T = 38 °C and for N. magadii, -T = 73 °C. Probably the supply on Li in the cells can be modified at extreme pH conditions. Also the temperature and pH in the cradle environment were reconstructed applying the mineral equilibria approach 33,72 . The technique relies on the use of the composition of a natural fluid to constrain a temperature and pH where a set of alteration minerals are computed to be in equilibrium with the aqueous phase. For all studied prokaryotic species, a series of calculation was yielded at the temperature range of 0.1-150 °C and pH varying between 5 and 9. The redox potential was fixed at the sphalerite saturation level. Pressure was equal to 1 bar at T < 100 °C and corresponds to the saturated steam pressure at T ≥ 100 °C. All minerals tabulated in the Thermoddem database were available for the calculations. The equilibrium temperature was computed as a median of saturation temperatures for the following phases: berthierine, ilmenite, Ca-and Mg-montmorillonite (Database S3). At the same temperature range, compositions of some species also revealed equilibrium with chlorapatite (Ca 5 (PO 4 ) 3 Cl), chlorite (Mg 5 Al 2 Si 3 O 10 (OH) 8 ) and goethite (FeOOH). When chlorapatite did not fit the equilibrium, the studied fluids were undersaturated with this mineral and supersaturated with hydroxyapatite (Ca 5 (PO 4 ) 3 OH). Hence, a solid solution of chlorapatite and hydroxyapatite, like verified in the Paraná basalts and other rocks (Ca 5 (OH,Cl)(PO 4 ) 3 ), could provide those solution compositions. The composition of N. magadii does not provide any equilibrium temperature at the tested range.
In the course of inverse modeling (Database S4), we used the compositions of primary minerals observed in Paraná basalts 73 . Apatite, barite, berthierine, halite, gibbsite, goethite, kaolinite, Ca-and Mg-montmorillonites, pyrite, and sphene were also available in the modeling system. The calculations were yielded at the previously found T-pH-pe − conditions. The uncertainties in solution components were equal to 0.2 that in most cases correspond to 2σ uncertainties specified in the Supplementary Table 2.
The formation time for this solution (Δt i ) can be calculated as follows: where Δx i represents a dissolved or precipitated amount of mineral species (mol), S i is a surface area of mineral exposed to solution (m 2 ) and r i is a dissolution/precipitation rate (mol m −2 sec −1 ) calculated at given T, P, pH and pCO 2 (Database S5). The BET surface areas of apatite, chalcedony, goethite, ilmenite, magnetite, and pyrite correspond to 1 m 2 g −1 ; K-feldspar, plagioclase and sphene -to 12 m 2 g −1 ; berthierine and barite -to 30 m 2 g −1 ; montmorillonites -to 100 m 2 g −1 .
Ce anomalies (Ce/Ce*) were calculated here using the expression: The affinity parameter was calculated as a normalized vector distance between all observed prokaryotic REE patterns and a given solution or rock pattern (Database S6). The logarithms of REE contents (C La , … C Lu , ppm) were considered as coordinates of vector z{log C La , …log C Lu }. They were approximated by a linear trend (y i = a x i + b), which is subtracted from the vector z{log C La -a x La -b, …log C Lu -a x Lu -b}, where x La = 1,… x Lu = 14. Based on the yielded normalized vectors z 1 and z 2 , the vector distance (Affinity) was calculated as: