Mechanisms of Pyrite Formation Promoted by Sulfate-Reducing Bacteria in Pure Culture

Pyrite, or iron disulfide, is the most common sulfide mineral on the Earth’s surface and is widespread through the geological record. Because sulfides are mainly produced by sulfate-reducing bacteria (SRB) in modern sedimentary environments, microorganisms are assumed to drive the formation of iron sulfides, in particular, pyrite. However, the exact role played by microorganisms in pyrite formation remains unclear and, to date, the precipitation of pyrite in microbial cultures has only rarely been achieved. The present work relies on chemical monitoring, electron microscopy, X-ray diffraction, and synchrotron-based spectroscopy to evaluate the formation of iron sulfides by the sulfate-reducing bacteria Desulfovibrio desulfuricans as a function of the source of iron, either provided as dissolved Fe 2 + or as FeIII-phosphate nanoparticles. Dissolved ferrous iron led to the formation of increasingly crystalline mackinawite (FeS) with time, encrusting bacterial cell walls, hence preventing further sulfate reduction upon day 5 and any evolution of iron sulfides into more stable phases, e.g., pyrite. In contrast, ferric phosphate was transformed into a mixture of large flattened crystals of well-crystallized vivianite ( Fe 3 ( PO 4 ) 2 ⋅ 8 H 2 O ) and a biofilm-like thin film of poorly crystallized mackinawite. Although being hosted in the iron sulfide biofilm, most cells were not encrusted. Excess sulfide delivered by the bacteria and oxidants (such as polysulfides) promoted the evolution of mackinawite into greigite ( Fe 3 S 4 ) and the nucleation of pyrite spherules. These spherules were several hundreds of nanometers wide and occurred within the extracellular polymeric substance (EPS) of the biofilm after only 1 month. Altogether, the present study demonstrates that the mineral assemblage induced by the metabolic activity of sulfate-reducing bacteria strongly depends on the source of iron, which has strong implications for the interpretation of the presence of pyrite and vivianite in natural environments.

Pyrite (FeS 2 ) is the most thermodynamically stable iron sulfide mineral in anoxic low-temperature conditions (Schoonen, 2004) and is ubiquitous in both modern environments and the sedimentary record (Rickard et al., 2017). Although surface sulfide minerals can have diverse origins, sedimentary processes predominate over magmatic, hydrothermal, and volcanic ones (Rickard et al., 2017). In modern environments, sulfide formation is mainly driven by the metabolism of sulfatereducing microorganisms (SRM), which are prevalent in anoxic environments, e.g., marine sediments or water column of permanently stratified euxinic waters such as the Black Sea (Vetriani et al., 2003) or Lake Cadagno (Tonolla et al., 2004) or even in ferruginous environments such as Lake Pavin (Lehours et al., 2005;Berg et al., 2016;Berg et al., 2019). Understanding the mechanisms of a possible biogenic pyrite formation pathway is crucial as pyrites are used as paleoenvironmental proxies and might also be good candidates for the search of biosignatures of early life (Shen and Buick, 2004). Sedimentary sulfides are generally enriched in light sulfur isotopes, suggesting their microbial origin (Thode et al., 1953). In addition, sedimentary pyrites generally display a specific texture called framboids (Rust, 1935), for which biogenic origin was proposed based on morphological criteria (Love, 1957;Folk, 2005). However, many reports of pyrite framboids formed under strictly abiotic conditions undermined their possible use as biosignatures (Berner, 1969;Farrand, 1970;Butler and Rickard, 2000;Ohfuji and Rickard, 2005). Later studies of sedimentary framboids using scanning transmission X-ray microscopy (Maclean et al., 2008) or nano-SIMS (Wacey et al., 2015) and revealing the presence of organic matter associated with framboids revived the biogenic interpretation. Since experimental biomineralization of framboidal pyrites has never been obtained, their biologic origin remains unfounded. Ultimately, while pyrite formation has been well constrained in abiotic systems, thanks to a set of experimental syntheses (Rickard and Luther, 2007), pathways of pyrite formation upon SRM activities remain largely unexplored.
Laboratory experiments are essential to understand the role of SRM in pyrite formation. Several studies of sulfate-reducing bacteria enrichments reported the formation of mackinawite (FeS) and greigite (Fe 3 S 4 ), both being potential precursors of pyrite (Fortin et al., 1994;Herbert et al., 1998;Watson et al., 2000;Gramp et al., 2009). Pyrites were obtained in some enrichments (Donald and Southam, 1999;Thiel et al., 2019;Berg et al., 2020), supporting the role of biology in sedimentary pyrite formation. However, due to the huge diversity of bacteria in these enrichments, especially iron-and sulfur-cycling bacteria (Lehours et al., 2009;Sitte et al., 2013;Zeng et al., 2018;Berg et al., 2019), deciphering the specific role of sulfate-reducing bacteria is tricky. Despite several attempts at iron sulfide biomineralization in pure sulfate-reducing bacteria cultures, mackinawite was observed almost exclusively (Ivarson and Hallberg, 1976;Neal et al., 2001;Williams et al., 2005;Ikogou et al., 2017;Stanley and Southam, 2018) and sometimes in association with greigite (Zhou et al., 2014;Picard et al., 2018). Pyrite formation in pure sulfate-reducing bacteria cultures was reported in only a single instance (Rickard, 1969b).
In the present contribution, the sulfate-reducing bacterium Desulfovibrio desulfuricans was cultured with either dissolved Fe 2+ or amorphous Fe III -phosphate nanoparticles to investigate the effect of the iron source on the nature of iron sulfides formed. Sulfate-reducing bacteria promoted pyrite formation in the medium supplied with amorphous Fe III -phosphate, whereas with dissolved Fe 2+ , only mackinawite occurred. These results allow understanding of the evolution and mechanisms of pyrite formation in pure sulfate-reducing bacteria culture and discuss implications for natural environments.

Culture and Biomineralization Conditions
All solution-preparation methods and manipulations were performed in a Jacomex® glove box under Ar (Alphagaz 1, Air Liquide) free of O 2 (<5 ppm), and solutions were prepared with Milli-Q water deoxygenated by N 2 (Alphagaz 1, Air Liquide) bubbling at 80°C for 45 min.
Once cells achieved the logarithmic phase of growth (around 1 week), they were centrifuged at 7,000 g for 10 min, rinsed three times in Milli-Q water, and transferred into a biomineralization medium at 50 % (v/v), corresponding to a final cell concentration of approximately 5 × 10 7 cells mL-1. The biomineralization medium was prepared in the same way as pre-culture medium but contained only Na 2 SO 4 , trace element, selenite-tungstate, vitamins, sodium DL-lactate, and MOPS buffer. A volume of 75 mL of the inoculated biomineralization medium was put into 100 mL serum vials and supplied with iron either as dissolved Fe 2+ (20 mM) or nanoparticulate Fe III -phosphate (10 mM). Dissolved ferrous iron was added from a 1 M solution of FeCl 2 · 4H 2 O. Nanoparticulate ferric phosphate was synthesized by successive addition of 2.72 g L − 1 KH 2 PO 4 and 5.56 g L − 1 FeSO 4 · 7H 2 O in a 0.1 M Na-acetate buffer solution, of pH 4.6 (Mirvaux et al., 2016). Culture vials were sealed with butyl rubber stoppers and incubated in the dark at 30°C without stirring for 8 months. Chemical monitoring was performed in the first month. All biomineralization experiments were performed in triplicate. Abiotic controls were prepared with the complete mineralization medium and iron source but no inoculum of bacteria.

Chemical Monitoring
Biomineralization experiments and abiotic controls were sampled periodically to monitor the chemical evolution of the liquid medium during biomineralization. Aliquots of 3 mL were collected from the 75-mL serum vials with syringes and needles in the glove box. Samples for total and dissolved (0.2 µm-filtered; Merck Millipore) iron were fixed with HCl to a final concentration of 0.5 M. Samples for total and dissolved sulfide were fixed with zinc acetate to a 0.5 M concentration. A separate filtered aliquot was stored in the dark at 4°C for dissolved phosphate and/or organic acid analyses.
Ferrous iron was analyzed spectrophotometrically with the ferrozine method (Stookey, 1970). Total iron was quantified by both a modified ferrozine method (Viollier et al., 2000) and ICP-AES (Perkin Elmer Optima 3,000). Sulfide concentrations were determined spectrophotometrically using the methylene blue method (Cline, 1969). Sulfate analysis was performed by ion chromatography (Dionex DX-600 IC System). Dissolved phosphate was measured spectrophotometrically using Biomol® Green Reagent (Enzo Life Sciences). Organic acids were detected by high-performance liquid chromatography (HPLC) on a U3000 Thermo Scientific series using a Rezex organic acid column (250 × 4.6 mm, 8 μm) with 5 mM H 2 SO 4 as eluent.

Mineralogical Characterization
At selected time points, around 10 mL of suspended matter was collected by centrifugation at 7,000 g for 10 min and rinsed three times with deoxygenated Milli-Q water, and subsamples were deposited on a Si(111) wafer for X-ray diffraction (XRD), on a 200-mesh Formvar carbon copper grid (Agar Scientific, United Kingdom) for transmission electron microscopy (TEM), and on a silicon nitride window (Norcada, Canada) for scanning transmission X-ray microscopy (STXM). Scanning electron microscopy (SEM) samples were prepared by filtering 10-100 μL of the culture through a polycarbonate GTTP 0.2 μm filter (Merck Millipore, Darmstadt, Germany) and then rinsing with 10 mL of deoxygenated Milli-Q water. All samples were dried in the glove box and stored under anoxic conditions until analysis.

X-Ray Diffraction
Silicon wafers were mounted in an airtight cell equipped with a Kapton window designed for X-ray diffraction analyses under anoxic conditions. Diffraction patterns were acquired from 5°to 95°with a 2θ step of 0.033°, at 40 kV and 40 mA with Co Kα radiation on a PANalytical X'Pert Pro MPD diffractometer equipped with an X'celerator® detector mounted in Bragg-Brentano configuration.

Electron Microscopy
For scanning electron microscopy analyses, filters were mounted onto pin stubs with an adhesive carbon tape and then carbon coated before analysis in a Zeiss Ultra 55 scanning electron microscope equipped with a field emission gun (FEG) and a Brucker EDS Quantax detector (Brucker Corporation, Houston, TX, USA). Images were preferentially acquired under low-voltage conditions at 3 kV and 3 mm working distance with either a highperformance annular detector (secondary electrons) or an annular detector with filter grids (backscattered electrons). Higher-voltage images were acquired at 15 kV and 7.5 mm working distance with an Everhart Thornley detector (secondary electrons) or an angle selective backscattered detector. Elemental characterizations were performed by energy dispersive X-ray spectrometry (EDS) at 15 kV and 7.5 mm working distance.
For transmission electron microscopy analyses, copper grids were observed under a JEOL 2100F transmission electron microscope equipped with a field emission gun (FEG) operating at 200 kV. Selected-area electron diffraction (SAED) patterns were acquired on areas of interest. Scanning transmission electron microscopy (STEM) was performed in the high-angle annular dark field (HAADF) mode and coupled to energy dispersive X-ray spectrometry (EDS) mapping.

Scanning Transmission X-Ray Microscopy
Scanning transmission X-ray microscopy analyses at the C K-edge and Fe L 2,3 -edges were performed at the HERMES beamline at SOLEIL (Saint-Aubin, France) (Belkhou et al., 2015;Swaraj et al., 2017). Energy was calibrated by setting the wellresolved 3p Rydberg peak of gaseous CO 2 at 294.96 eV for the C Kedge and the major L 2 -edge peak of hematite at 708.5 eV for the Fe L 2,3 -edges. Recommended procedures for radiation-sensitive samples were followed . In essence, image mapstacks were recorded at 288.2 and 280 eV for C K-edge and, then, image mapstacks were obtained on the same area from 690 to 740 eV for Fe L 2,3 -edges. Data were processed using aXis2000 software (Hitchcock, 2019) following procedures described in previous studies (Miot et al., 2009b;Miot et al., 2009a). Nearedge X-ray absorption fine structure (NEXAFS) spectra for Fe L 2,3edges were normalized by the double arctan method (Bourdelle et al., 2013). Several iron minerals were synthesized following reported procedures to serve as reference model compounds for scanning transmission X-ray microscopy analyses of Fe L 2,3 -edges, including pyrite (Wei and Osseo-Asare, 1997), greigite (Li et al., 2014), mackinawite (Donald and Southam, 1999), vivianite (Miot et al., 2009a), amorphous ferric phosphate (Mirvaux et al., 2016), and hematite (Cornell and Schwertmann, 2003).

Chemical Evolution
Desulfovibrio desulfuricans was cultivated in two different biomineralization media containing 20 mM sulfate and 20 mM lactate, with initial iron provided either as 20 mM dissolved Fe 2+ or 10 mM Fe in the form of nanoparticulate ferric phosphate, hereafter referred to as Fe-diss and FP-nano experiments, respectively. Dissimilatory sulfate reduction was attested in the two conditions by the formation of black precipitates of iron sulfides, whereas no precipitation was observed in abiotic controls. After 30 days of biomineralization, only 4 mM of sulfate was consumed in the Fe-diss condition (i.e., 20 % of initial sulfate Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 4 concentration) ( Figure 1A), while up to 10 mM was used in the FP-nano condition (i.e., half of initial sulfate) ( Figure 1B). Simultaneously, 2 mM of solid sulfides (determined by the methylene blue method) were produced in the Fe-diss condition against 7 mM in the FPnano condition, but no dissolved sulfide was detectable in both conditions (Supplementary Table S1, S2; Figures  1A,B). In contrast, no sulfate conversion was evidenced in the abiotic controls. After 1 month, 3 mM of Fe 2+ was removed from the solution and precipitated as FeS in the Fe-diss condition ( Figure 1C). In the FP-nano condition, solid Fe III was rapidly reduced and precipitated as Fe-bearing minerals, so that dissolved iron trapped into or sorbed on minerals did not accumulate in the medium (Supplementary Table S2). This reduction-precipitation reaction was accompanied by the release of 9 mM dissolved phosphate in the FP-nano medium. Lactate was converted to acetate in both conditions, but in the Fe-diss condition, consumption of lactate (6 mM) was incomplete ( Figure 1E), whereas all lactate was converted to acetate in the FP-nano condition ( Figure 1F). No significant chemical changes were observed in abiotic controls for both conditions ( Figure 1) with the exception of a slight increase in the Fe III concentration. The apparent increase in solid Fe III concentration is attributed to random sampling of suspended FP particles.

Mineralogical Characterization at the Bulk Scale
Solid phases were analyzed by X-ray diffraction ( Figure 2). In the Fe-diss experiment, precipitates formed after 1 week of incubation were mostly amorphous with only a tiny peak at 20°2θ angle (Co Kα), characteristic of mackinawite (FeS). After 1 month, mackinawite increased in crystallinity, as illustrated by several well-resolved peaks (Fe-diss-1m in Figure 2). In the FPnano experiment, initial nanoparticulate Fe III -phosphate was amorphous. After 1 week, vivianite (Fe 3 (PO 4 ) 2 · 8H 2 O), as well as a poorly crystalline mackinawite phase, was detected, as suggested by the slight and broad band around 20°2θ angle. Both phases gained in crystallinity after 1 month of incubation. Notably, pyrite (FeS 2 ), associated with mackinawite and vivianite, was detected after 3 months in the FP-nano condition.  (I) Energy dispersive X-ray spectrometry spectra pinned from previous SEM images. Letters in energy dispersive X-ray spectrometry maps correspond to energy dispersive X-ray spectrometry spectra provided in (I). Except for (H), all images were acquired in low-voltage conditions.
Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 6 FIGURE 4 | Transmission electron microscopy observation of precipitates for both conditions, after 1 week and 1 month. Transmission electron microscopy, selected-area electron diffraction (or scanning transmission electron microscopy for J panel), high-resolution images and energy dispersive X-ray spectrometry map for: (A-D) Fe-diss condition at 1 week, (E-H) Fe-diss condition at 8 months, (I-L) FP-nano condition at 1 week, and (M-T) FP-nano condition at 1 month. (U) Energy dispersive X-ray spectrometry spectra pinned from previous energy dispersive X-ray spectrometry maps. Letters in energy dispersive X-ray spectrometry maps correspond to energy dispersive X-ray spectrometry spectra provided in (U).
Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 resolution images obtained by transmission electron microscopy, showing the 5 Å d-spacing characteristic of mackinawite ( Figure 4C). Moreover, bacteria could be observed by scanning electron microscopy and transmission electron microscopy in association with mackinawite ( Figures 3C, 4A,D). After 1 month, the increase in mackinawite crystallinity was confirmed by single-crystal diffraction patterns ( Figure 4F). In these 1month old samples, the size and shape of mackinawite aggregates suggested the presence of encrusted cells ( Figure 4E).

FP-Nano Condition
Iron sulfides formed in the FP-nano experiment exhibited a morphology different from that obtained in the Fe-diss condition, specifically forming a thin film spreading over large areas (Figures 3A,B,D,I, 4I,J,L,M,P,U). After 1 week, this iron sulfide film was mostly amorphous, comprising only nano-domains of crystallized mackinawite ( Figure 4K). Residual amorphous FP-nano was still present at both 1 week and 1 month ( Figures 3A,B,D,I, 4I,J,L,P,U, 5A,G), but additional wide well-crystallized vivianite sheets (Figures 3D,I) were noticeable after 1 month. Due to mineral encrustation, bacteria could not be observed by scanning electron microscopy with secondary electrons ( Figure 3A). However, scanning electron microscopy with backscattered electrons ( Figure 3B), transmission electron microscopy ( Figure 4L), and scanning transmission X-ray microscopy C K-edge analyses ( Figure 5B) revealed that cells were intimately associated with the iron sulfide film. Scanning transmission X-ray microscopy analyses showed that this iron sulfide film had a structure close to that of mackinawite ( Figures 5C,G). Mackinawite crystallinity increased from mostly amorphous after 1 week to large crystals ( Figure 4O) attested by a polycrystalline mackinawite pattern ( Figure 4N) after 1 month, similar to that observed in the Fe-diss condition after 1 week. Greigite domains of several hundreds of nanometers were also observed in the mackinawite film ( Figures 4S,T). After 1 month, bacteria were still embedded within the mackinawite film ( Figures 3E, 4M,P, 5D,E,F,G) and iron sulfide spherules of several hundreds of nanometers in size were observed within the iron sulfide film (Figures 3G,H,  4Q). These spherules had a higher S/Fe ratio than in mackinawite ( Figure 3I, spot h). Transmission electron microscopy selected-area electron diffraction patterns indicated that they comprised pyrite particles nucleating within the FeS film ( Figure 4R). The S/(Fe + S) ratio of these spherules was estimated to 66 % which is consistent with that of FeS 2 pyrite and significantly differs from the values, ranging from 45 to 55 % estimated for the iron sulfide film, consistent with FeS mackinawite. Scanning transmission X-ray microscopy data also confirmed the presence of pyrite, given the similarity of the spherules and reference pyrite spectra sharing a wider L 3 -edge peak than the reference mackinawite spectrum and exhibiting a specific shoulder around 713.5 eV ( Figure 5G).

First Stages of Iron Sulfide Formation (1 Week)
Although it is the most stable sulfide phase in sediments, pyrite is unlikely formed by the simple precipitation of Fe 2+ and S 2− 2 as the latter is not stable in low-temperature aqueous solutions (Kamyshny et al., 2004). Pyrite is the thermodynamically stable end product of multi-step pathways, some of which include metastable intermediates such as amorphous iron monosulfide, mackinawite, or greigite which are barely found in sediments (Berner, 1962;Pye, 1981). In aqueous lowtemperature systems, mackinawite formation can be described as (Wei and Osseo-Asare, 1995) This reaction is kinetically fast (Rickard, 1995), explaining the presence of amorphous iron sulfide (or disordered mackinawite) as the first product of abiotic pyrite synthesis in both ferrous and ferric-sulfide systems (Schoonen and Barnes, 1991;Wei and Osseo-Asare, 1997). Similarly, laboratory sulfate-reducing bacteria pure cultures have mainly yielded the formation of amorphous FeS (Fortin et al., 1994;Williams et al., 2005;Ntarlagiannis et al., 2005;Peltier et al., 2011;Stanley and Southam, 2018) or of well-crystallized mackinawite (Rickard, 1969b;Ivarson and Hallberg, 1976;Zhou et al., 2014;Ikogou et al., 2017;Picard et al., 2018). Some of these studies also detected greigite in long-term experiments (Rickard, 1969b;Picard et al., 2018) with excess of electron donor (Zhou et al., 2014).
In the two types of cultures performed in the present study, bacteria converted lactate and sulfate into acetate and sulfide, respectively, through sulfate respiration. Chemical monitoring ( Figures 1E,F) showed a lactate/sulfate consumption ratio of ≈ two in agreement with sulfate respiration through incomplete oxidation of lactate (Muyzer and Stams, 2008): Hydrogen sulfide quickly reacted with the iron source to form and precipitate black iron sulfide preventing dissolved sulfide accumulation. After 1 week, iron sulfides were characterized as poorly crystallized mackinawite in both conditions ( Figure 2). However, significant differences prevailed between Fe-diss and FP-nano experiments. In Fe-diss condition, the flake-like morphology of iron sulfide aggregates ( Figures 3C, 4A) was consistent with that in previous observations of iron sulfides formed in sulfate-reducing bacteria enrichments (Herbert et al., 1998;Watson et al., 2000;Sitte et al., 2013;Berg et al., 2019) or in pure cultures of sulfate-reducing bacteria (Picard et al., 2018;Stanley and Southam, 2018) in the presence of dissolved ferrous iron. In the present study, microbial activity has likely enhanced mackinawite crystallinity produced in one week as shown by the selected-area electron diffraction polycrystalline pattern ( Figure 4B) and by high-resolution transmission electron microscopy that revealed large wellcrystallized particles with typical (001) lattice fringes of mackinawite ( Figure 4C). In contrast, abiotic precipitates were Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 usually reported as amorphous FeS (Ohfuji and Rickard, 2006;Csákberényi-Malasics et al., 2012;Picard et al., 2018). Transmission electron microscopy observations revealed that ≈ 10 µm-wide iron sulfide aggregates were in fact composed of elongated 1-µm long subunits, suggesting the presence of encrusted cells ( Figure 4A). Cell wall surfaces are usually negatively charged and can thus offer binding sites for cations such as dissolved Fe 2+ . In addition, the periplasm is also a dedicated site for Fe 2+ -mineral precipitation as shown for iron-oxidizing bacteria (Miot et al., 2009c;Miot and Etique, 2016) and sulfate-reducing bacteria (Watson et al., 2000;Picard et al., 2018;Stanley and Southam, 2018). Sulfide released by sulfate-reducing bacteria would have promoted iron sulfide precipitation on the cell surface and/or within their periplasm, resulting in bacteria encrustation. Nonetheless, some studies reported thinner crusts (Fortin et al., 1994;Donald and Southam, 1999) to no encrustation (Stanley and Southam, 2018), questioning the mechanisms of cell encrustation in sulfatereducing bacteria. As shown in Fe-oxidizing bacteria, cell wall encrustation can limit nutrient uptake eventually leading to cell death (Miot et al., 2015). Similarly, high sulfate-reducing bacteria encrustation levels in Fe-diss condition would explain the partial consumption of lactate and the strong slowdown of sulfate reduction after only 1 week ( Figures 1A,E). In contrast to previous studies, these data show that encrustation apparently disturbs the metabolism of sulfate-reducing bacteria at least under millimolar dissolved Fe 2+ concentrations (Watson et al., 2000;Picard et al., 2018). Such conditions differ from the low iron content of the modern ocean (nanomolar) or sediment porewater (hundreds of micromolars) but are expected in ferruginous . Dotted line showed bacteria revealed by scanning transmission X-ray microscopy C K-edge data. (G) Scanning transmission X-ray microscopy spectra pinned from previous composite maps. Letters in composite maps correspond to scanning transmission X-ray microscopy spectra provided in (G).
Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 environments, e.g., in some meromictic lakes Llirós et al., 2015), where Fe-encrusted cells have been reported , as well as in extreme environments such as hydrothermal vents, acid mine drainage sites, and acidsulfate systems in Yellowstone (Templeton, 2011). Encrustation patterns may vary across these environments depending on local Fe solubility. While the mineralogy results of Fe-diss experiment after 1 week support previous findings reported in sulfate-reducing bacteria laboratory cultures, the biomineral products obtained in the FP-nano experiment after 1 week exhibit specific characteristics which were never described in previous studies. First of all, an iron sulfide thin film containing significant amounts of organic matter ( Figure 4L) and hosting bacterial cells ( Figures 3B, 5B) could be interpreted as a mineralized biofilm. Iron sulfide precipitation occurred in this thin layer rather than at the surface of iron phosphate nanoparticles, which were the source of this iron. This implies that HS produced by sulfate-reducing bacteria first reduced Fe IIIphosphate following the reaction Coupling microbial sulfate-reduction (Eq. 2) and Fe III -phosphate reduction (Eq. 3) allows one to explain that the 10 mM of Fe III provided initially in the system was fully reduced by the consumption of 10 mM of lactate which is consistent with the results obtained on day 4 ( Figure 1B,F). Dissolved Fe 2+ released from Fe III -phosphate reduction may be adsorbed on cell surfaces or onto extracellular polymeric substances (EPS) whose negative charges provide binding sites for cations (Ferris et al., 1987;Beveridge, 1989;Picard et al., 2018). Further HS − produced by bacteria likely reacted with this adsorbed Fe 2+ and precipitated as mackinawite according to Eq. 1 to form an FeS film. In addition, dissolved Fe 2+ and phosphate released by Fe III -phosphate reduction (Eq. 3) precipitated as vivianite (Figure 2), due to its low solubility at the pH and temperature explored here (pKs ≈ 36 (Al-Borno and Tomson, 1994)), according to The difference in mackinawite crystallinity in the two different biomineralization conditions could be explained by the local concentration of iron bound at the cell surface. In the Fe-diss condition, high amounts of Fe 2+ were supplied in the starting biomineralization medium accounting for the significant precipitation of FeS in contact with the cell surface ( Figure 4A). In contrast, Fe 2+ in the FP-nano condition was gradually supplied by the reduction of Fe III -phosphate leading to a lower concentration of cell-bound iron and thus a more diffuse precipitation of poorly crystalline mackinawite ( Figure 4K) within a wide amorphous FeS film. Moreover, part of HS − produced by sulfate-reducing bacteria was consumed by the reduction of ferric phosphate, thus decreasing its availability for iron sulfide precipitation and crystallization. This suggests that the consumption of sulfide through ferric iron reduction may delay the increase in mackinawite crystallinity. Phosphate might also be responsible for mackinawite's lack of crystallinity as it has been shown that adsorbed phosphate prevents the evolution of iron oxyhydroxides into well-crystallized phases (Borch et al., 2007;Voegelin et al., 2013;Schoepfer et al., 2019). Future studies aimed at evaluating the competition between phosphate and sulfide for reaction with ferrous iron would be of particular interest to disentangle the mechanisms at play.

Long-Term Iron Sulfide Evolution ( ≥ 1 Month)
Aging of disordered mackinawite leads to its conversion into more stable iron sulfide phases such as crystalline mackinawite, greigite, or pyrite depending on environmental conditions. The increase in mackinawite crystallinity with aging has been interpreted as the expulsion of water molecules entrapped between lattice sheets during rapid mackinawite precipitation (Wolthers et al., 2003). In abiotic experiments, ordering of mackinawite is quite slow and could take several months to a year (Rickard, 1969a). In the present work, the high degree of mackinawite crystallinity in the Fe-diss condition is evidenced after only 1 month by X-ray diffraction (Figure 2), singlecrystalline selected-area electron diffraction patterns obtained by transmission electron microscopy ( Figure 4F), and highresolution images ( Figure 4G). Mackinawite increase in crystallinity was not as pronounced in the FP-nano experiment, but it showed relatively well-crystallized polycrystalline selected-area electron diffraction patterns after 1 month ( Figure 4N), and the small nanometric domains present after 1 week ( Figure 4K) evolved into domains of hundreds of nanometers long ( Figure 4O). Interestingly, the morphology of mackinawite after 1 month in the FP-nano condition was similar to the one observed in the Fe-diss condition after 1 week.
In the absence of oxidants, the lack of other stable iron sulfides than well-crystallized mackinawite in the Fe-diss experiment could be explained by the low level of sulfide produced, around 2 mM, compared to the high dissolved Fe 2+ concentration (over 15 mM, Figure 1C, Supplementary Table  S1). In addition, both greigite and pyrite are more oxidized than mackinawite. However, iron and sulfur are as Fe 2+ and S 2− in FeS mackinawite, two iron atoms out of three are Fe 3+ in Fe 3 S 4 greigite, and in pyrite, sulfur atoms are at the formal oxidation degree −1 in the disulfide ion S 2− . Oxidants are thus necessary for the conversion of mackinawite into pyrite. In aqueous solutions, protons might be the appropriate oxidants for the conversion of FeS into FeS 2 at low temperature as, for instance, in the H 2 S pathway (Rickard, 1997;Rickard and Luther, 1997): The rate of Eq. 5 has been well constrained (Rickard, 1997), and in Fe-diss condition, the formation of 1 mM pyrite after 1 month would require more than 1 mM of aqueous H 2 S, which would have been detected with the Cline method, if present. As sulfide produced by sulfate-reducing bacteria first precipitates with dissolved Fe 2+ to form FeS, the limited production of sulfide in the Fe-diss experiment may have hindered pyrite formation. This hypothesis is in agreement with thermodynamic models predicting a H 2 S concentration of around 10 nM in the Fe-diss Frontiers in Earth Science | www.frontiersin.org condition which implies that thousands of years are needed to form 1 mM of pyrite (Supplementary Table S3). Although H 2 concentration was not measured, it would be interesting to investigate in future studies if the accumulation of H 2 could also have limited the formation of pyrite through this reaction (Thiel et al., 2019).
In the FP-nano experiment, the pyrite formation pathway assumed in the Fe-diss condition (Eq. 5) was shifted to a more efficient pathway involving zero-valent sulfur formed through the reduction of ferric phosphate (Eq. 3). Indeed, pyrite formation might be strongly accelerated if other oxidants than protons are available for oxidation of FeS into FeS 2 . The only reported case where pyrite was indeed detected in pure sulfate-reducing bacteria cultures was in the presence of goethite (FeO(OH)) (Rickard, 1969b). In the present study, we report a second example of rapid pyrite formation in a pure culture of sulfatereducing bacteria by using Fe III -phosphate as an iron source. Thus, the presence of ferric iron (Fe III -phosphate or goethite) in pure cultures appears to be a crucial parameter for the formation of pyrite. Reaction of sulfide with ferric iron described in Eq. 3 likely provides zero-valent sulfur that can act as an oxidant promoting the conversion of mackinawite into pyrite as follows (Berner, 1970): The occurrence of greigite within the iron sulfide film ( Figures  4S,T) is consistent with the possible solid-state conversion of mackinawite into greigite (Lennie et al., 1997;Pósfai, 1998). Greigite may originate from the reaction of FeS with zerovalent sulfur: 3FeS + S 0 Fe 3 S 4 (7) Greigite possibly represents an intermediate phase in the global mechanism of pyrite formation, but it did not accumulate to sufficient levels to be detectable by X-ray diffraction analyses in the FP-nano condition. Pyrite formation through a greigite pathway has been suggested based on the occurrence of sedimentary magnetic pyrites and has been well studied Benning et al., 2000;Hunger and Benning, 2007;Lan and Butler, 2014). Although structurally complex (Rickard and Luther, 2007), the conversion of greigite to pyrite through a solid-state reaction may still be possible, as shown by hydrothermal experiments (Hunger and Benning, 2007) and, more recently, by the co-occurrence of greigite and nanocrystalline pyrite domains in micrometer-large pyrite spheres formed in an sulfate-reducing bacteria enrichment culture (Berg et al., 2020). Pyrite could thus have been produced by the reaction of greigite with zero-valent sulfur as follows: Fe 3 S 4 + 2S 0 3FeS 2 (8) Since FeS and S 0 (in the form of cyclo-octasulfur S 8 ) are both solid phases, a direct reaction between them in solution is improbable. Instead, S 0 may occur as polysulfides which have been shown to result from the reaction of HS − with ferric phases, especially Fe III -(hydr)oxides such as goethite or lepidocrocite (Hellige et al., 2012;Wan et al., 2014). In addition, polysulfides can be produced by the reaction of HS − with elemental sulfur and exist in equilibrium depending on pH: Then, polysulfides may react with FeS to form pyrite as follows (Rickard, 1975): Interestingly, experiments with millimolar concentrations of dissolved Fe 3+ did not yield pyrite formation even after almost a year (Ikogou et al., 2017). This suggests that the progressive supply of low amounts of dissolved Fe 2+ , possibly by preventing cell encrustation and eventually cell death, is crucial. Low Fe 2+ concentrations could be maintained by the presence of ferric minerals, organic-Fe 3+ complexes, or low dissolved Fe 3+ concentrations, rather than high concentrations of highly reactive dissolved Fe 2+ . Such conditions are prevalent in many anoxic sedimentary environments and may thus best represent natural conditions of pyrite formation. Further experiments using more realistic (micromolar) concentrations of dissolved Fe 2+ have to be conducted to evaluate the feasibility of biogenic pyrite formation in highly reduced environments. Moreover, the quantification of the evolution of Fe-bearing phases, for instance, by X-ray absorption spectroscopy, might help distinguish the types of iron sulfides in these biominerals over time.

Environmental Implications
Pyrite and vivianite are barely found together in natural environments and instead are usually reported to form separately (Manning et al., 1999). While pyrite is very abundant in anoxic sediments, in particular, in marine ones, due to the high level of sulfate in modern ocean, vivianite mainly forms in anoxic non-euxinic environments, such as some ferruginous meromictic lakes (Fagel et al., 2005;Rothe et al., 2014;Cosmidis et al., 2014). Indeed, highly reactive sulfide competes with phosphate for reaction with Fe 2+ , hence preventing vivianite formation at high sulfide concentrations (Nriagu, 1972;Rothe et al., 2015). However, recent studies reported the formation of vivianite in euxinic environments in association with pyrite, e.g., in the Baltic Sea Fjords, the Black Sea, or Lake Cadagno, leading to sulfide depletion in sedimentary microenvironments or close to the water-sediment interface (Jilbert and Slomp, 2013;Dijkstra et al., 2014;Xiong et al., 2019). Here, we report the formation of both vivianite and pyrite in the presence of D. desulfuricans, starting from nanoparticulate Fe III -phosphate, at high sulfate levels. These results confirm experimentally the possibility to form concomitantly vivianite and pyrite under euxinic conditions. Biogenic vivianite was often assumed to result from the activity of iron-reducing bacteria as it preferably forms in anoxic non-sulfidic environments rather than euxinic environments. However, enrichment cultures from the meromictic phosphate-rich Lake Pavin suggested that iron reduction and vivianite formation resulted from the activity of Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 both iron-reducing (Pseudomonas or Clostridium) and sulfatereducing bacteria (Lehours et al., 2009;Berg et al., 2019). The results of the present study demonstrate that Desulfovibrio desulfuricans alone had the ability to induce vivianite formation in the FP-nano condition, suggesting that sulfatereducing bacteria could play a significant role in vivianite formation in natural environments. Despite a low sulfate concentration (<20 µM), Lake Pavin hosts a plethora of sulfate-reducing bacteria in its water column (Lehours et al., 2005;Berg et al., 2019;Berg et al., 2020) and pyrite was found abundantly in the first 12 cm of its sediments, in association with abundant vivianite (Viollier et al., 1997;Busigny et al., 2014;Busigny et al., 2016). Although pyrite was only detected in the sediments and not in the water column, sulfate-rich enrichment cultures from this lake led to rapid pyrite formation after 21 days of culture (Berg et al., 2020). Here, our results demonstrate that the sole activity of Desulfovibrio (representing 80 % of the microbial consortium in the enrichment (Berg et al., 2020)) was able to promote pyrite formation. However, while, in the present study, pyrites occurred as infra micrometric spherules undetectable by X-ray diffraction after one month, pyrites produced within the Lake Pavin consortium were visible on diffractograms after only 21 days and occurred as beads of 1 µm in diameter (Berg et al., 2020). The additional presence of sulfide-oxidizing bacteria (e.g., Sulfuricurvum and Arcobacter) might have strongly enhanced the rate of pyrite formation in the enrichment through an increased delivery of polysulfides (Berg et al., 2020). Hence, kinetics of pyrite formation induced by sulfate-reducing bacteria activity, such as described in our present study, would be accelerated under environmental conditions due to the concomitant activity of sulfide-oxidizing bacteria. In the future, it would be interesting to measure and compare the levels of polysulfides produced in pure sulfate-reducing bacteria vs mixed enrichment cultures in order to evaluate how S 0 impacts the formation of pyrite in both systems.
Sedimentary pyrites are commonly found as framboids of several micrometers in size Busigny et al., 2016). Framboids are sparsely reported in anoxic water column and mainly in euxinic basins (Skei, 1988;Muramoto et al., 1991;Perry and Pedersen, 1993). Such framboids strongly deviate from the biogenic submicrometric pyrite spherules formed in the present FP-nano condition as well as from pyrites obtained in sulfate-reducing bacteria enrichments in previous studies (Donald and Southam, 1999;Thiel et al., 2019;Berg et al., 2020). This dichotomy suggests that framboids in natural environments originate from diagenetic processes rather than from purely biogenic pathways. We suggest that submicrometric pyrite spherules such as those obtained in our pure sulfatereducing bacteria cultures (FP-nano condition) might be precursors for larger pyrite framboids. Framboidal evolution from biogenic precursors might explain the intra-grain δS 34 variability in pyrite framboids (Bryant et al., 2020). In anoxic stratified waters, submicrometric pyrite spherules could form in the water column, below the oxycline, promoted by sulfate-reducing bacteria and other sulfur-and iron-cycling bacteria activities. Due to their small size and their dilution within the water column and the sediments, they may have been overlooked so far in euxinic and non-sulfidic water bodies. Interestingly, X-ray absorption spectroscopy analyses of Lake Pavin samples revealed that including pyrite significantly improved the fits of X-ray absorption spectroscopy spectra of samples collected just below the sulfate-reduction zone in this lake, while pyrites were not detected by X-ray diffraction (Cosmidis et al., 2014). Then, it is possible that pyrite precursors produced by sulfate-reducing bacteria in the water column would evolve into framboidal pyrites deeper in the sediments following diagenetic processes. This is consistent with the occurrence of pyrite microcrystals at the top of modern sediments which give way to pyrite framboids deeper in the sediment (Raven et al., 2016).
To conclude, our results attest that vivianite and pyrite formation can result from the single activity of sulfatereducing bacteria in the presence of ferric phosphate nanoparticles. Early stages lead to the formation of submicrometric spherules nucleated within an FeS-rich film. Under environmental conditions, this reaction may be accelerated by the contribution of polysulfide-producing microorganisms. Evolution into micrometric framboidal pyrites may require longer timescales, hence occurring preferentially in the sediments upon diagenesis. Further investigations are needed to explore the effects of diagenesis on these pyrite spherules and understand the origin of pyrite framboids.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
AD performed the laboratory experiments, AD, JSB, and JM achieved the chemical and mineralogical analyses. AD and FG realized the thermodynamic modeling. All the authors wrote and reviewed the manuscript.

ACKNOWLEDGMENTS
This work was funded by the ANR SRB project, under grant ANR-14-CE33-0003-01, of the French Agence Nationale de la Recherche to JM. The authors would like to thank the staff of the IMPMC (Sorbonne Universités) for their support and training on instruments: Benoît Baptiste and Ludovic Delbès for the X-ray diffraction facility, Jean-Michel Guigner for the transmission electron microscopy facility, Imène Estéve, Béatrice Doisneau, and Stéphanie Delbrel for the scanning electron microscopy facility, and Fériel Skouri-Panet and Mélanie Poinsot for the Biology Lab. The scanning electron microscopy facility at the IMPMC is funded by Région Ile de France, under grant SESAME 2006 N°I-07-593/R, INSU/CNRS, UPMC-Paris 6, and by the Agence Nationale de la Recherche (grant N°ANR-07-BLAN-Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 588310 0124-01). The transmission electron microscopy facility at IMPMC is supported by Région Ile de France, under grant SESAME 2000 E 1435. Parts of this work were supported by IPGP multidisciplinary program PARI and by Paris-IdF region SESAME, under Grant no. 12015908. The authors thank Laure Cordier (LGE and IPGP) for ICP-AES and iron-chromatography analyses. The authors also thank Séverine Zirah (MCAM and MNHN) for providing access to high-performance liquid chromatography. The authors acknowledge SOLEIL (HERMES beamline, Saint-Aubin, France) for having provided beamtime for scanning transmission X-ray microscopy experiments. The authors also especially thank Sufal Swaraj, Stefan Stanescu, and Rachid Belkhou (SOLEIL) for their support at the beamlines.