Desulfovulcanus ferrireducens strain Ax17^T (DSM 111878) was used for this study and was obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (Braunschweig, Germany). The medium used for growth was modified DSM 981 medium (Kashefi et al., 2002). The base medium was composed of the following (per liter): 19.0 g of NaCl, 9.0 g of MgCl₂ · 6H₂O, 0.3 g of CaCl₂ · 2H₂O, 0.5 g of KCl, 0.14 g of KH₂PO₄, 0.05 g of NaBr, 0.02 g of SrCl₂ · 6H₂O, 0.15 g of MgSO₄ · 7H₂O, 0.1 g of (NH₄)₂SO₄, 1 g of NaHCO₃, 10 ml of DSM141 Trace Mineral solution, and 10 ml of DSM141 Vitamin solution. Synthetic nanophase ferrihydrite (2-line), akaganeite, and lepidocrocite were synthesized as previously described (Sklute et al., 2018). Then, 100 mmol L⁻¹ of each mineral was used separately as terminal electron acceptors. The pH of each medium was adjusted to 6.80 ± 0.05 using 1 N NaOH. All incubations were carried out in 160 ml serum bottles containing 50 ml of medium and sealed with butyl rubber stoppers. The media-containing serum bottles were degassed and flushed with H₂:CO₂ (80%:20%), reduced with 0.5 mM cysteine-HCl, supplemented with 1.3 mM FeCl₂ · 2H₂O, and the headspace of each bottle was filled with 200 kPa of H₂:CO₂. Each growth medium was inoculated in triplicate with a 1% (vol vol⁻¹, ~10⁵ cells ml⁻¹) logarithmic growth-phase culture that had been transferred at least three times on that terminal electron acceptor. Cultures were incubated at 55°C until they reached the late logarithmic-to-early stationary growth phase (~5 days). For cell counts, samples were preserved using 2% (vol vol⁻¹) formaldehyde and mixed with an oxalate solution (28 g L⁻¹ ammonium oxalate and 15 g L⁻¹ oxalic acid) to dissolve the particulate iron. Cells were filtered using 0.2-μm pore size black polycarbonate membrane filters (Whatman, Little Chalfont, UK), stained with 0.1% (wt vol⁻¹) acridine orange for 3 min, and counted using epifluorescence microscopy (Hobbie et al., 1977).

In addition to the biotic samples (heat + cells), there was an uninoculated control (heat/no cells) for each Fe(III) (oxyhydr)oxide mineral that was incubated at the same temperature and duration as the biotic samples and an uninoculated control that was incubated at room temperature (no heat/no cells). The composition of the controls was otherwise identical to the biotic samples. These two controls served to determine the abiotic effect of growth medium alone and growth medium plus incubation temperature on the minerals present and thus to differentiate microbially induced mineral transformations from abiotic mineral transformations. Fe(II) concentrations were measured spectrophotometrically for all samples following dissolution in oxalate solution using the ferrozine assay (Phillips and Lovley, 1987). Each sample was unfiltered and filtered through a 0.22-μm pore size polyethersulfone (PES) membrane syringe filter (Corning Inc., Corning, New York, USA) prior to dissolution in the oxalate solution to determine the proportion of Fe(II) that was in the microparticulate form.

For spectroscopy and XRD analyses, each sample was filtered through a 0.2-μm pore size polycarbonate filter (Whatman, Little Chalfont, UK) in an anoxic chamber, dried for at least 12 h, and stored in an anoxic chamber to minimize oxygen exposure and mineral alteration.

Visible-near infrared (VNIR) spectra were collected using an ASD Fieldspec 4 Max spectrometer in bidirectional geometry (input = 30°, emission = 0°) with an Ocean Optics HL-2000 light source directed down toward a 1,000-μm Si optical fiber and an 8° foreoptic for collection. The samples were loaded into a matte black sample cup in the air and leveled without packing. The samples were in the air for ~5 min. Each spectrum had an average of 3 × 240 136 ms integrations and was referenced to a Spectralon. Continuum removal was performed manually using OPUS software (Bruker Optics). Fourier transform infrared attenuated total reflectance (FTIR-ATR) spectra were obtained on a Bruker ALPHA FTIR spectrometer with a platinum diamond ATR attachment placed inside an anoxic chamber. Spectra were collected from 360–4,000 cm⁻¹ at 8 cm⁻¹ resolution. Each spectrum had an average of 128 scans. Continuum removal was performed with OPUS software using a concave rubberband correction.

Mössbauer spectroscopy was performed on anoxically mounted dried powder samples that were ground lightly with sugar and secured in plastic washers backed with Kapton tapes. Spectra were collected at Mount Holyoke College on a Web Research (now See-Co) W302 Mössbauer spectrometer at 295, 220, 150, 80, and 4 K (starting with the lowest temperature) employing a Janus closed cycle He compressor at <295 K. Samples were transported under an inert atmosphere to Mount Holyoke College and loaded into the sample tube, which was immediately purged and filled with He gas. Each 1024-channel spectrum was folded about the midpoint and calibrated to an α-Fe foil using the WMOSS4 program. Spectra were fit using the Mexfield program, provided by Eddie DeGrave and Toon Van Alboom of the University of Ghent, Belgium. The Mexfield program uses Lorentzian line shapes and solves the full hyperfine interaction Hamiltonian to minimize the chi-squared deviation between the data and the model using center shift, quadrupole splitting, linewidth, ±hyperfine field, and field broadening as free parameters (Vandenberghe et al., 1994).

X-ray diffraction (XRD) analysis was performed using a Rigaku Smartlab II SE XRD (Cu Kα radiation; Bragg-Brentano geometry). Anoxically dried samples were transported in an inert atmosphere to the XRD facility. XRD patterns were acquired in the air (42-min scans) on zero-background sample holders from 5 to 80° 2θ with a 0.02° step size at a rate of 2° per minute. Baseline corrections and mineral identifications used the Rigaku Smartlab II software (ICSD PDF2 database; 2019 version).

Transmission electron microscopy (TEM) was performed in the Materials Characterization and Processing Center at Johns Hopkins University. It was performed on a Thermo Fisher TF30 instrument operating at 300 kV with an EDAX windowless silicon drift energy dispersive X-ray (EDX) detector. Conventional imaging, high-resolution lattice imaging (HRTEM), selected area electron diffraction (SAED), scanning transmission electron microscopy (STEM), bright-field (BF) and high-angle annular dark-field (HAADF) imaging, and STEM EDX maps were collected for each bioreacted and heated control mineral sample. Samples analyzed by TEM were shipped cold to Johns Hopkins in their original media, and a grid was loaded with samples directly from the bottles onto lacey carbon grids.

Desulfovulcanus ferrireducens grew to ~10⁷ cells ml⁻¹ when grown on ferrihydrite with 24 mM of Fe²⁺ in the medium. A separate filtration step through a 0.2-μm pore size filter prior to the addition of oxalate solution showed that 84% of the biogenic iron product was retained by the filter. The iron produced was visibly black and magnetic (Figure 1A). The concentrations of Fe²⁺ in the heated and unheated abiotic controls were 2–3 mM, there was no visible color change to the iron (Figure 1A), and neither sample was magnetic.

Ferrihydrite that was bioreduced by D. ferrireducens is spectrally flat and minimally reflective in its VNIR spectrum, indicating the presence of a highly absorbing and opaque phase like magnetite (Figure 2A). When the spectrum is vertically stretched, a maximum wavelength of 0.743 μm is observable, which could indicate a cation-deficient magnetite or the mixing of multiple iron-bearing phases. The VNIR spectra for the heated and unheated controls are muted relative to the parent ferrihydrite spectrum, but they retain the visible wavelength maximum of 0.78 μm, almost identical to that in ferrihydrite. The hydration-related absorption features are visible but partially shifted to 0.915 and 1.937 μm for the unheated control and 0.979 μm, 1.937 μm, and 2.196 μm for the heated control in uncorrected spectra. Continuum-removed peak positions, while dependent upon the subjective choice of continuum anchor points, can often offer more direct comparisons between absorption features. Supplementary Figure 1, Supplementary Table 2 examine the standard and continuum-removed VNIR spectra more closely, and both uncorrected and continuum-removed VNIR data and references are included in the Supplementary material.

The FTIR-ATR spectra of ferrihydrite bioreduced by D. ferrireducens completely lose the ferrihydrite-like broad doublet at 672 and 603 cm⁻¹ and broad absorption at 401 cm⁻¹ (Figure 2B). These are replaced by a large, symmetric absorption at 553 cm⁻¹ and a broad, asymmetric absorption at 336 cm⁻¹. The shape and position of these features are consistent with a cation-deficient magnetite. The heated control spectrum also loses most of the ferrihydrite-like absorptions but instead indicates a transformation to lepidocrocite, which is identifiable by a sharp absorption at 1,022 cm⁻¹ (Sklute et al., 2018). The unheated control spectrum resembles ferrihydrite at frequencies of <800 cm⁻¹.

At higher frequencies, a broad triplet is visible, spanning from ~970 to 1,117 cm⁻¹, likely due to the absorption of various media-related compounds, such as phosphate, sulfate, or organics, to the mineral surface (c.f., Arai and Sparks, 2001). These features are in a similar location but more pronounced in the heated control spectrum. In the bioreduced sample spectrum, the complex feature is shifted to lower wavenumbers and includes a sharp embedded feature at ~852 cm⁻¹. Combined with a broad absorption centered at ~1,356 cm⁻¹, an absorption of 852 cm⁻¹ is a good indicator of carbonate, although both features are shifted to lower frequencies than those typically observed for siderite. Instead, these features are similar to those reported for carbonate green rust (Antony et al., 2008), which can form in circumneutral solutions of similar composition (Schwertmann and Fetcher, 1994). The shift in the underlying broad feature centered near 968 cm⁻¹ may point to the absorption of cell-associated phosphate or aqueous phosphate. Iron phosphate minerals contain additional absorptions at low wavenumbers that are not observed in this spectrum. Whatever the source of these absorptions, their shape and position in the bioreduced sample spectrum are clearly distinct from those in the control spectra. At higher wavenumbers (Figure 2C), the hydration feature position is closer to that of maghemite than magnetite, confirming that the spinel is likely some intermediate or cation-deficient magnetite, even under the reducing conditions of the study.

The 295K Mössbauer spectrum of the bioreduced sample is a mixture of sextets and doublets, including a prominent ferrous doublet (Figure 3A). The sextet structure is similar to that of cation-deficient nanophase magnetite, but admixing with other phases cannot be ruled out. The 4K spectrum confirms that the sample is not pure magnetite (Figure 3B). The lingering doublet structure centered near 0.2 mm/s, increased the depth to the shoulder at 9 mm/s, and decreased the absorption depth near −3.3 mm/s and 2.7 mm/s all point to a mixture of phases or a modification to the magnetite structure. There was too much overlap in the central region to identify minor green rust. The 295K spectra of the control samples were doublets with parameters consistent with ferrihydrite or lepidocrocite (Figure 3A). Mössbauer spectroscopy cannot distinguish these two minerals at 295K. At 4K, the heated control spectrum shows a clear mixture of ferrihydrite-like and lepidocrocite-like sextets, while the unheated control spectrum is almost identical to ferrihydrite (Figure 3B). Additional Mössbauer spectra for analyses at other temperatures are found in Supplementary Figures S2–S4.

Unlike the hyperthermophilic bioreduction products reported in Kashyap et al. (2023), the bulk XRD for ferrihydrite bioreduced by D. ferrireducens shows clear evidence of magnetite (or maghemite or cation-deficient magnetite). The broad peaks indicate small scattering domains, similar to the 11 nm magnetite reported in Sklute et al. (2018) (Supplementary Figure S5). Additional reflections are consistent with lepidocrocite, NaCl, and a small amount of non-iron carbonate, such as rhodocrocite (MnCO₃), although rhodocrocite is unlikely given the media composition. The carbonate identified is unlikely to have been sourced from impurities in media components, is more likely a structurally similar material with a different composition, and does not appear in the library. Interestingly, the phase does not appear in the XRD of the control samples. The XRD pattern of the heated control shows clear evidence of lepidocrocite observed by FTIR-ATR. Interestingly, there is a broad swell in the baseline in the position of the magnetite/maghemite reflection of ~35.5 2θ; magnetite may be a minor product in the heated abiotic sample. The major difference between the heated abiotic sample and the bioreduced sample is in the ratio of magnetite to lepidocrocite.

TEM with SAED shows that ferrihydrite bioreduced by D. ferrireducens comprises clumps of dark, nanocrystalline but well-ordered irregular spheres, many with amorphous rims (Figures 4A, E). SAED patterns (Figures 4B–D) are consistent with a mixture of ferrihydrite, magnetite/maghemite, and halite, and EDX analysis confirms the presence of the oxide, salt, and some media components (S, P) either in or on the grains (Figure 4F). The heated abiotic control is also largely crystalline (Figures 4G–N), with clear sheets of lepidocrocite (Figures 4L–N, seen in other techniques) and laths of goethite (Figures 4J, K, not seen by any other spectroscopy). This sample also appears to have a wealth of Cu-rich spheres that cannot be explained by the experimental conditions and may be copper mobilized from the grid. These spheres are not observed in the bioreduced sample for this mineral.

Desulfovulcanus ferrireducens reached a density of 2–3 × 10⁷ cells ml⁻¹ when cultured separately with akaganeite and lepidocrocite. For growth on lepidocrocite, the mineral product was visibly greenish black but not magnetic (Figure 1B). The mineral product following growth on akaganeite was visibly black but not magnetic (Figure 1C). There was no visible color change to either akaganeite or lepidocrocite in either the heated or unheated abiotic controls (Figures 1B, C). The concentration of oxalate-soluble Fe²⁺ in the biotic samples was not significantly different than that of the abiotic controls, suggesting that oxalate was ineffective at dissolving the Fe²⁺ in these samples. A separate filtration step through a 0.2-μm pore size filter prior to the addition of oxalate solution showed that 82% of the biogenic iron product was retained by the filter. Oxalate is less effective at dissolving more highly crystalline phases (Schwertmann and Fetcher, 1994). The lack of oxalate-soluble Fe²⁺ does not necessarily indicate that Fe²⁺-bearing mineral phases were not produced. Mössbauer spectroscopy clarified the extent of mineral reduction in these samples.

Like ferrihydrite, the akaganeite bioreduced by D. ferrireducens is spectrally flat and minimally reflective in its VNIR spectrum, indicating the presence of a highly absorbing and opaque phase (Figure 5A). However, unlike ferrihydrite, when the VNIR spectrum is stretched vertically, it does not display a maghemite-like VIS maximum but rather a red slope. It also lacks the distinctive akaganeite absorption at 2.460 μm. The heated and unheated abiotic control spectra are nearly identical to the parent mineral spectrum. Lepidocrocite bioreduced by D. ferrireducens is also spectrally flat and minimally reflective in its VNIR spectrum (Figure 5B). When the VNIR spectrum is stretched vertically, it has a clear VIS maximum at 0.551 μm, which is in the same position as the VIS maximum in a natural green rust sample (USGS “green slime”). Retroactive examination of the VNIR spectrum reported for the bioreduction of lepidocrocite by the marine hyperthermophile P. delaneyi (Kashyap et al., 2023) revealed a similar VIS maximum upon spectrum expansion. Supplementary Figures S6, S7 and Supplementary Table S2 examine the standard and continuum-removed VNIR spectra more closely, and both uncorrected and continuum-removed VNIR data and references are included in the Supplementary material.

The FTIR-ATR spectra of akaganeite that was bioreduced by D. ferrireducens and the two control spectra (Figure 6A) are similar in the ~900–500 cm⁻¹ range typically used to identify that mineral (Sklute et al., 2018). There is a slight shoulder in the bioreduced spectrum near 564 cm⁻¹ and a decrease in the feature near 420 cm⁻¹ that could indicate trace magnetite. While continuum removal can affect the latter, changes in this feature shape can be observed in additional experiments where lower wavenumber range coverage is available. All three experimental spectra display separation of the peaks that make up the main akaganeite asymmetric absorption at ~626 cm⁻¹. The bioreduced sample, however, has increased area in the higher wavenumber peak of ~681 cm⁻¹ with respect to the peak ~624 cm⁻¹. Although digital reference spectra for sulfate green rust are not available, this increase in intensity is consistent with the peak position for sulfate green rust reported by Misawa et al. (1969). Admixing with the other major green rust feature shown in the abovementioned study near 788 cm⁻¹ could be responsible for the slight shift to lower wavenumbers of the akaganeite peak near 805 cm⁻¹ (typically a doublet centered near 827 cm⁻¹). However, more recent studies of green rust and green rust oxidation products by Antony et al. (2008) report many absorptions not observed in our samples. Their chloride green rust matches our sample and is more realistic for the solution chemistry. Both the bioreduced and heated control samples show absorptions in the phosphate/sulfate region near 1,020 cm⁻¹; the shape and location of these features differ between the two samples and are much more intense in the heated control. Both control samples also have features near 2,976 and 2,889 cm⁻¹, which are consistent with organics that are added to the media. Of note is the absence of these features in the bioreduced sample, indicating these elements are consumed during bioreduction.

FTIR-ATR spectra of lepidocrocite bioreduced by D. ferrireducens and the two control spectra (Figure 6B) are almost identical to the reference lepidocrocite is at <1,200 cm⁻¹. Above 1,200 cm⁻¹, all three experimental spectra differ from the reference, primarily due to the loss of hydration features approximately at 1,623 cm⁻¹ and >3,250 cm⁻¹. All three spectra share a muted, broad feature from ~2,590 to 3,290 cm⁻¹. While this range overlaps with the lipid region reported in biofilm spectral studies (c.f., Gieroba et al., 2020), this seems more likely due to tightly bound water that survived anaerobic drying. A small feature in the two control spectra near 1,089 cm⁻¹ could indicate a carbonate, but it is too small to match any specific phase and is likely an adsorption product. This feature is, interestingly, less intense in the bioreduced spectrum. No spectral evidence in this wavelength range explains the color change observed in the VNIR.

Mössbauer spectra of akaganeite bioreacted by D. ferrireducens and control spectra (Supplementary Figures S8A, B) show subtle changes in the iron in the sample. Temperature series and fit parameters (Supplementary Figures S9–S11) show that the heated control sample has a new ferric phase and that the bioreacted sample has new ferric and ferrous phases, both of which persist below the magnetic ordering temperature of the ferric phases of the unheated control sample. The fit parameters for the new phase in the bioreacted sample are not unique due to overlap with other phases, but the general spectral shape suggests that admixing with green rust or a structurally and energetically similar phase is possible (Supplementary Figure S8). The SAED of the TEM sample is consistent with a mixture of akaganeite and lepidocrocite (Figure 7H). While the lepidocrocite would explain a new ferric doublet, the Mössbauer doublet for lepidocrocite should persist below 80K, which is not observed for this new phase.

Mössbauer spectra of lepidocrocite bioreacted by ferrireducens and control spectra (Supplementary Figures S12A, B) are almost identical to the reference lepidocrocite spectrum. Temperature series and fit parameters (Supplementary Figures S13–S15) also fail to identify any changes to the iron in these samples that would account for the color changes recorded in the VNIR. However, it is possible that small amounts of green rust could be hidden within the broad lepidocrocite features (Supplementary Figure S12).

Bulk XRD spectra of akaganeite bioreduced by D. ferrireducens and control spectra are almost identical (Supplementary Figure S16) and display a mixture of akaganeite with NaCl. An additional reflection near 13.97 2θ in the heated control spectrum is consistent with a reflection for nanophase lepidocrocite, as reported by Sklute et al. (2018). In that sample, the remaining lepidocrocite reflections align with slight baseline swells or are obscured by other peaks. For the bioreduced sample, there is a unique reflection of ~13.16 2θ. This is the position of the main reflection in the iron phosphate vivianite. While a single reflection is insufficient for phase identification, when added to the fit, the program Match! determined that it comprises >14 mol% of the sample. This could account for the color change observed in the VNIR. However, at such concentrations, the feature should also be visible in both FTIR-ATR and Mössbauer spectra. Vivianite features in Mössbauer and FTIR spectra could potentially be obscured by absorptions of other phases, but a more likely scenario is that the XRD match percentage is driven by the XRD baseline swell due to the fine-grained nature of the sample.

TEM with SAED of akaganeite samples shows that the heated control comprises 100–200 × 25 nm akaganeite laths, whose SAED indicates a structural shift toward lepidocrocite. The bioreduced sample, on the other hand, is comprised of poorly crystalline 50–200 nm balls (Figure 7) that we hereafter refer to as amorphous iron-rich nanospheres (AFNS). This sample also contains a film not observed in the heated control that could be the result of biological activity.

Bulk XRD was not possible for the lepidocrocite experiments, but TEM with SAED (Figure 8) shows that the heat-treated control sample is largely crystalline, with the laths displaying amorphous rims not uncommon in synthetic nanophase iron (hydro)oxides. The bioreduced sample, however, is largely amorphous, similar to the akaganeite experiments. However, unlike the akaganeite bioreduced sample, these spheres have jagged rather than smooth edges (Figure 8B). While the AFNS on both akaganeite and lepidocrocite are amorphous, assuming they are the same material or metamaterial would be a mistake. The AFNS in this sample is also accompanied by Cu spheres of unknown origin.

The presence of Cu spheres under different experimental conditions (heated control in the ferrihydrite experiments vs. bioreduced in the lepidocrocite experiments) suggests that they are analytical contaminants rather than experimental products and that the mobilization of Cu is related to specific solution chemistry macroenvironments or microenvironments in these two experiments.