Iron minerals within specific microfossil morphospecies of the 1.88 Ga Gunflint Formation

Problematic microfossils dominate the palaeontological record between the Great Oxidation Event 2.4 billion years ago (Ga) and the last Palaeoproterozoic iron formations, deposited 500–600 million years later. These fossils are often associated with iron-rich sedimentary rocks, but their affinities, metabolism, and, hence, their contributions to Earth surface oxidation and Fe deposition remain unknown. Here we show that specific microfossil populations of the 1.88 Ga Gunflint Iron Formation contain Fe-silicate and Fe-carbonate nanocrystal concentrations in cell interiors. Fe minerals are absent in/on all organically preserved cell walls. These features are consistent with in vivo intracellular Fe biomineralization, with subsequent in situ recrystallization, but contrast with known patterns of post-mortem Fe mineralization. The Gunflint populations that display relatively large cells (thick-walled spheres, filament-forming rods) and intra-microfossil Fe minerals are consistent with oxygenic photosynthesizers but not with other Fe-mineralizing microorganisms studied so far. Fe biomineralization may have protected oxygenic photosynthesizers against Fe2+ toxicity during the Palaeoproterozoic.


Supplementary
The discontinuities in the cell walls of these cells can be explained by the growth of quartz crystals through the cell walls (arrows). We interpret these discontinuities as post-mortem rather than primary features of the cells such as wall reticulation, budding or opening of cells. Note that the platinum grain observed on the FIB section in a is a particle that was redeposited after the tearing of the Pt-coating in the top-right part in the FIB section. The length of cell-shaped domains, bound by filament-perpendicular organic segments was measured using imageJ as the maximum length of the quartz crystals in FIB sections, or more commonly in photomicrographs from the middle of each organic segment to the next (e.g. pink line in the zoom inset). b, Ratio of the average length of segment to filament diameter in each microfossil, reported against filament diameter. Vertical error bars represents one standard deviation, and horizontal error represents +/-100 nm error linked with measurement reproducibility. The length/width ratio (elongation) parameter has been chosen because it can relate to cellular ultrastructure and filament stiffness parameters 1 . Filaments can be sorted as three groups: Gunflintia minuta with 1.4<diameter<2µm and 1.7<average length/diameter<2.3 (blue), Gunflintia minuta with 2<diameter<2.5µm and 1.3<average length/diameter<1.5 (green), and Gunflintia grandis with diameter>4µm and 1<average length/diameter<1.2 (orange). c, Histogram of all segment length measurements. The narrower Type 2 G. minutia (blue) show a distribution of segment lengths with sharp maxima at 2.75, 3.5 and 4 µm. The slightly broader Type 2 G. minutia (green) are mainly characterized by a maximum at 3.5 µm with shoulders at lower lengths. G. grandis (orange) display a much higher variability in segment length. Remains of a previous gold coating (now mostly removed through re-polishing of the thin section) appear in white. c and f, SEM EDXS maps recorded at high electron beam power (30kV and spotsize#6 for ca. 1 hour) in order to detect Fe-bearing crystallites in the first micrometer inside the outcropping quartz grains (this is necessary as outcropping organic matter and Feminerals were largely removed through polishing). In contrast to the more energetic X-rays generated by iron, carbon within the quartz crystals cannot be mapped using this technique because of X-ray reabsorption by the matrix.

Supplementary Figure 15 | Identification of greenalite. a, SAED of greenalite crystal in
Gunflintia Grandis E55-1t2 (circled in Supplementary Fig. 7c). Arrows indicate lattice spacing diagnostic of greenalite 2 : 7.2 Å (0,0,1) planes (red), and 23 Å superlattice (green). b. STEM dark-field image of greenalite. The crystals in the green circle traverse the FIB section, allowing analyses without quartz interference. c, STEM EDXS spectra of greenalite in the green circle in b (1: same zone as in Fig. 3a), of the circle (3)  The global Fe-mineral paragenesis in the Gunflint Iron Formation can be summarized as follows. Stage (1a) of Supplementary Fig. 18 is indicated by Fe isotope compositions depleted in light Fe-isotopes throughout the Gunflint Iron Formation 5 . In Stage (1b) initial precipitation as Fe 2+ -minerals could only have enriched the precipitates in light Fe-isotopes 6 -this was rarely observed 5 , indicating that initial Fe 2+ -precipitation was minor compared to Fe 3+ -precipitation. Fe 2+ -deposition followed by oxidation by groundwater O 2 (Ref. 7 ) should preserve light Fe-isotope enrichments through quantitative transfer of iron from precursor mineral to product 8 . Light Fe-isotope enrichments are scarce in the Gunflint, indicating that a large fraction of the bulk-rock iron precipitated initially as Fe 3+ -minerals from a large (buffered isotope composition) reservoir, i.e. ferruginous seawater rather than groundwater 5 . Indeed, precipitation of Fe 3+ -minerals by groundwater should have rapidly enriched remaining groundwater Fe 2+ in the light isotopes 9 , allowing for the subsequent formation of deposits with light Fe-isotope enrichments. In stage (1c) Some of this primary haematite may have been preserved 5 , escaping reduction by e.g. organic matter in stage (2). Stage (2) could have proceeded through microbial and/or thermal oxidation of organic matter 10,11 . The depletion in light Fe-isotopes associated with Fe 3+precipitation of stage (1a) could have been preserved through quantitative transfer of iron from precursor to products as observed in Gunflint rocks preserving stage (2) Fe 2+ -minerals and consistent with the general absence of rocks enriched in light Fe-isotopes through Fe-reduction 5 . All Fe 3+ -rich microfossil-rich rocks of the Gunflint should have formed some Fe 2+ -minerals through thermal oxidation of the available organic matter as they suffered sufficient burial temperatures >150-170°C 10 Fig. 17 and Ref. 13  Our stromatolite sample has been affected by stages in blue boxes and, possibly, by stages in discontinuous boxes of Supplementary Fig. 18. Stages (1b) and/or (1a±1c+2) formed the observed greenalite+siderite+pyrite. Fe 3+ -oxidation stage (1a) is indicated by Fe-isotopes in haematitic stromatolites of the Gunflint as well as in deeper-water facies, suggesting that Fe 3+ -oxidation could have proceeded in all facies of the Gunflint. Our sample was unaffected by stage (3) oxidation (indicated by preservation of organic matter, nanoscale Fe 2+ -minerals, absence of haematite). At the sampled locality, lateral transitions between haematitic (stage 3) and preserved (organic-rich) stromatolites occur on the sub-meter scale 7,13 . This argues that the preserved stromatolite we studied (and other organic-rich stromatolites) could have accreted in similar Fe-oxidizing conditions than its haematitic counterparts at the Schrieber locality where Fe-isotope signatures of stage (1) are recorded. In our sample, stage (4) ankerite is absent in microfossils that preserve stage (2) Fe-minerals, indicating that it formed from distinct, Ca-and Mg-enriched fluids.
The following section presents five schematics ( Supplementary Figs. 19-23) of the scenario discussed in the main text for the origin and fate of Fe-minerals in order to explain the presence and crystal chemistry of the assemblage of greenalite + siderite + Fe-sulphides in Gunflint Iron Formation microfossils.
Supplementary Figure 19 | Cells, encrusted only externally by amorphous SiO 2 , have a decaying cytoplasm favoring post-mortem intramicrofossil mineralization through diffusion of externally-sourced Fe. a, Illustration of the case where Fe 2+ diffusion is associated with convection, which allows aqueous Fe 2+ (or colloidal Fe 3+ ) to reach the decaying cell in an isotropic fashion. In this case, the wall of the dead cell would act as a double diffusion boundary where externallysourced Fe 2+ -fluids meet the organic-rich conditions of the cell wall and/or of the decaying intracellular content; both conditions could favor precipitation of iron. Precipitation should thus start at this diffusion boundary, and post-mortem Fe 2+ or Fe 3+ mineralization should encrust the inner and/or outer surface of cell walls first 15,16 and not only scatter crystals at the center of cells. This is supported by experimental evidence and observation of younger fossils, as listed in the main text. b, Illustration of the case of an anisotropic, convection-free diffusion of Fe 2+ , which belongs to Liesegang-band formation processes. Similar to the diffusion+convection process, Fe-precipitation should occur at the diffusion boundary and/or at regular distances from this boundary. Fe-precipitation is expected to preferentially occur on the parts of the microfossils that are oriented toward the Fe 2+ -diffusion front. Moreover, Fe-minerals of distinct compositions are expected to form in distinct bands (e.g. bands of greenalite separated from bands of siderite) due to contrasting saturation states created at the limits of the diffusion boundary 17 . Such banded Fe-precipitates have been observed in petrified wood, where bands cross-cut cell walls and are associated with dark Femineral linings on the cell walls (Fig. 31 in Ref. 18 ). However, no mineralogical control is observed on the spatial distribution of Fe-minerals in Gunflint microfossils: greenalite and siderite are homogeneously mixed at the center of microfossils. Moreover, no spatial correlation appears between Fe-minerals and the main diffusion boundary that is the cell wall. Therefore, these models of abiotic Fe-mineralization associated are difficult to reconcile with the microfossils mineralized by greenalite+siderite with only scattered intra-microfossil Fe-minerals and no minerals near/at the wall boundary.  19 to add possible DIR as the source of Fe 2+ for pyritization. In these fossils, the organic matter is replaced by pyrite (Py), the product of heterotrophic SRB (possibly coupled with heterotrophic DIR) metabolism. SRB and DIR both oxidize organic matter (into CO 2 ) as a means to reduce sulfate (to HS -) and Fe 3+ -minerals (to Fe 2+ ), respectively. SRB and DIR have exoenzymes (carriers) that allow extraction and transport of organic "fuel", and DIR have carrier molecules (siderophores) for extraction and transport of Fe 3+ (Ref. 20 ). The products of these metabolisms (CO 2 , HSand Fe 2+ ) can form pyrite and/or siderite, as well as greenalite in presence of silicic acid (Ref. 11 ). In this model, almost all the organic molecules (black chains) of the fossils, including the resilient sheath/wall polysaccharides, are available for heterotrophic consumption, as indicated by the low abundance of organic matter in these fossils [Ref. 19 and Supplementary Fig. 17a-c]. The localization of pyrite replacing the organic walls of the fossil can be explained e.g. by Fe 2+ adsorption onto remnants of polysaccharide wall 21,22 during pyritization. Pyrite is abundant outside the microfossils, as scattered crystals and overgrowths of pyritized microfossils (pink arrows in Supplementary Fig. 17b) and encrustation of the heterotrophic SRB (Ref. 19 ). Supplementary Figs. 17bc show the absence of intracellular pyrite and the general absence of Fe-minerals inside the pyritized cells, suggesting that intracellular contents did not favor pyritization. Altogether, the mineralization pattern and the replacement of organic matter in pyritized microfossils strongly contrast with organically-preserved microfossils with intracellular greenalite+siderite, indicating that the latter could not form from oxidation (stage 3 or 5) of the former.

Dissimilatory Iron Reduction (DIR) of intra and/or extracellular Fe 3+ -mineralscase of a fossil cell wall that is not breached by DIR-bacteria.
Heterotrophic DIR bacteria can form greenalite, siderite and Fe-sulphides (Ref. 11 ) from remote Fe 3+ -minerals and organic molecules (Ref. 20 ) extracted for example from decaying cells. First, the possibility that the microfossils themselves could represent DIR bacteria (which can precipitate intracellularly: Ref. 23 ) is ruled out by their morphology (large, thickwalled spheres and filaments unknown in DIR bacteria). Second, organic matter is preserved in fossil sheaths, cell walls and inside microfossils. DIR-bacteria can produce catalytic exoenzymes (carrier) to reach intracellular content of the decaying cells and carry fuel for Fe 3+ -respiration (Ref. 20 ). Third, Fe 3+ , being highly insoluble, must be sourced in minerals outside or inside microfossils; siderophores (carriers, Ref. 20 ) can also reach and carry Fe 3+ from inside and outside the dead cells. In this model, both Fe 2+ and CO 2 , that can drive the precipitation of greenalite and/or siderite (Sid), are produced by the DIR bacteria. In this model, the DIR bacteria do not breach the cell wall and penetrate physically the microfossils, which is consistent with the wall preservation observed in the microfossils. Therefore, irrespective of the localization of the source of Fe 3+ , Fe 2+ precipitation should be favored outside the microfossils where DIR bacteria produce Fe 2+ and CO 2 . However, outside microfossils, possible greenalite is extremely small and scarce and siderite has not been observed. Assuming that the Fe 2+ produced outside microfossils by DIR diffuses and finds better conditions to precipitate inside than outside microfossils recalls the model of Supplementary Fig. 19, which has been discussed above. Figure   22 | Heterotrophic dissimilatory iron reduction (DIR) of intra and/or extracellular Fe 3+ -minerals -case of a fossil cell wall that is physically breached by DIR-bacteria.

Supplementary
Here, DIR-bacteria that breach the wall respire labile (e.g. amino acids, proteins, nucleic acids,…) organics inside the dead cell where it can form post-mortem intra-microfossil Feminerals (e.g. greenalite and siderite) even if Fe 3+ is sourced outside microfossils. However, the physical breach that allowed penetration of DIR bacteria inside the microfossil cell must leak the intracellular organic compounds that fuel DIR. In turn, extracellular DIR bacteria thriving outside the decaying cells are fed by leaked compounds outside the cells, and if the Fe 3+ -mineral source is only outside microfossil, then Fe 2+ -mineralization is expected to occur as much outside as inside microfossils. The assemblage of greenalite and siderite would dominantly have formed outside and/or on the surface ( Supplementary  Fig. 19) of microfossils, which is not the case in our sample, hence refuting this model as the origin of the large concentrations of iron in thick-walled Hurosniospora and Type 2 Gunflintia. In addition, physical breaching of the cell wall by heterotrophs prior to encapsulation by silica is difficult to reconcile with the 3D preservation of the microfossils. Indeed, in modern siliceous environments, 3D preservation is allowed by a first step of external (and/or external + internal) silicification of unbreached cells followed by cytoplasm degradation and secondary silica infilling (Ref. 24 ). Encapsulation by silica would drastically reduce physical access to the decaying cells by DIR and other heterotrophic bacteria, consistent with the molecular preservation of Gunflint microfossils (Ref. 12 ).

Supplementary Figure 23 | In situ,
thermal reduction of intra-microfossil Fe 3+ -minerals. In this model, oxidation of organic matter (ultimately, catagenesis to CO 2 ) is coupled with the reduction of Fe 3+ -bearing minerals (here simplified as FeOOH) and catalyzed by thermal energy 10 . Siderite can form from the produced Fe 2+ and the produced CO 2 10 . Greenalite can similarly form from the reaction of the produced Fe 2+ with the silica matrix 25 and H 2 O from fluids and/or organic matter catagenesis. The extremely scarce nanoscale Fe 2+ -sulphides can form by catagenetic desulphurization of organic matter producing H 2 S to react with Fe 2+ (Ref. 26 ), consistent with systematic embedding of nanoscale sulphides in organic matter (Fig. 4). Because reactive organic molecules are mobile during this stage, whereas precursor Fe 3+ -minerals are not until they are reduced (by their reaction with organic matter), localized Fe 2+ -minerals can form by in situ replacement of Fe 3+ -minerals. At this stage, silicification already occurred, amorphous SiO 2 was largely recrystallized to quartz, which limited the diffusion of OM and metamorphic products, favoring in situ replacement and explaining the scarcity of greenalite and absence of siderite outside microfossils. This contrasts with pyritization ( Supplementary Fig. 20) and bacterial (heterotrophic) iron reduction  that require chemical exchange between microfossils and heterotrophic bacteria through a more permeable matrix. Some of the internal organic matter now observed inside the microfossils (Figs. 1-2, Supplementary Fig. 6) could correspond to relics of the cytoplasmic contents that could have fueled iron reduction. Finally, these thermal diagenetic reactions require a localized, mostly intracellular source of iron to explain the localization of greenalite+siderite+sulphides. Some siderites display rod shapes (Figs. 4a-f, Supplementary Fig. 16d) instead of their normal rhombohedral shapes (Fig. 4g), consistent with in situ recrystallization of rod-shaped Fe 3+ -(oxyhydr)oxides. Moreover, similar to Supplementary Fig. 19, externally sourced Fe 2+ (e.g. from pore fluids) or colloids of Fe 3+ and organic molecules would lead to preferential precipitation of Fe-minerals in the matrix surrounding microfossils and/or at the cell wall interface, but not only at the center of microfossils.