Spectral Detection of Nanophase Iron Minerals Produced by Fe(III)-Reducing Hyperthermophilic Crenarchaea

Mineral transformations by two hyperthermophilic Fe(III)-reducing crenarchaea, Pyrodictium delaneyi and Pyrobaculum islandicum, were examined using synthetic nanophase ferrihydrite, lepidocrocite, and akaganeite separately as terminal electron acceptors and compared with abiotic mineral transformations under similar conditions. Spectral analyses using visible–near-infrared, Fourier-transform infrared attenuated total reflectance (FTIR-ATR), Raman, and Mössbauer spectroscopies were complementary and revealed formation of various biomineral assemblages distinguishable from abiotic phases. The most extensive biogenic mineral transformation occurred with ferrihydrite, which formed primarily magnetite with spectral features similar to biomagnetite relative to a synthetic magnetite standard. The FTIR-ATR spectra of ferrihydrite bioreduced by P. delaneyi also showed possible cell-associated organics such as exopolysaccharides. Such combined detections of biomineral assemblages and organics might serve as biomarkers for hyperthermophilic Fe(III) reduction. With lepidocrocite, P. delaneyi produced primarily a ferrous carbonate phase reminiscent of siderite, and with akaganeite, magnetite and a ferrous phosphate phase similar to vivianite were formed. P. islandicum showed minor biogenic production of a ferrous phosphate similar to vivianite when grown on lepidocrocite, and a mixed valent phosphate or sulfate mineral when grown on akaganeite. These results expand the range of biogenic mineral transformations at high temperatures and identify spacecraft-relevant spectroscopies suitable for discriminating mineral biogenicity.

Spectral data were collected as an average of 128 scans with spectral resolution set at 4 cm -1 using Norton Beer strong apodization function. Background spectra were collected at regular intervals to compensate for changing atmospheric and instrumental conditions. For analysis, continuum removal was performed using a rubberband baseline removal algorithm (Wartewig, 2003) and absorption positions determined using local minima in the OPUS 7.7 software package. For broad bands, peak positions were identified by Gaussian and Lorentzian curve fitting algorithms in MagicPlot v.2.9 (MagicPlot Systems, LLC). Multiple peaks were fit iteratively by varying the parameters of the fit function to minimize the residual sum of squares. All spectral data were compared with reference iron (oxyhydr)oxides taken from Sklute et al. (2018), iron phosphates, and iron carbonates from M.D. Dyar's mineral collection examined as part of this study to identify mineral transformation products.
Raman spectroscopy was performed using a Bruker Senterra II Raman microscope with a 532 nm excitation laser and a 20× objective at Bruker Optics. Spectral data were collected using either 0.25 mW or 2.5 mW laser power depending on the sensitivity of the sample to phase transformation. Each spectrum was collected using 30-120 2-4s integrations. Multiple individual mineral grains/regions (3-10) per sample were chosen for data acquisition. Darker grains were preferentially analyzed because these likely represented mineral transformations. Spectral data were baseline-corrected with an Asymmetric Least Squares (ALS) algorithm (Eilers and Boelens, 2005) and min-max normalized prior to spectral fitting for feature identification. They were smoothed using a Savitzky-Golay algorithm for visual representation, but no features were assigned based on smoothed data. Band positions were identified using Gaussian and Lorentzian curve fitting algorithms in MagicPlot v.2.9. Multiple peaks were fit iteratively by varying the parameters of the fit function to minimize the residual sum of squares. Phase identifications were confirmed by comparing samples to reference spectra for various iron minerals collected either as part of this study or obtained from the RRUFF database (Lafuente et al., 2015).
Mössbauer spectroscopy was performed using a WEB Research Co. (now SEE Co.) model WT302 spectrometer equipped with a source of ~40 mCi 57 Co in Rh and a Janis closed cycle He cooling system (Edina, MN). Spectral data for each sample were acquired over a ± 4 mm s -1 velocity range at 295 K and over a ± 10 mm s -1 velocity range at 220 K, 150 K, 80 K, and 4 K. They were acquired in 1024 channels and acquisition times varied from 4 h to 5 days depending on the iron content of the samples. All data were corrected for the fraction of the baseline due to Compton scattering of 122 keV gamma rays by electrons inside the detector and non-linearity via interpolation to a linear velocity scale using the 25 m α-Fe foil calibration spectrum collected at room temperature. The Mössbauer data were fitted using either Mex_disd or Mex_field (De Grave and Van Alboom, 1991). Both programs solve the full hyperfine interaction Hamiltonian for multiple distributions and minimize the chi-squared deviation between the fitted and experimental spectrum using center shift, quadrupole shift, full width at half maximum, and distribution area as adjustable parameters. Mex_field specifically uses Lorentzian-shaped sextets for each distribution but allows for a broadening effect in magnetically ordering samples (Dormann et al., 1985). Mex_disd uses hyperfine field distributions of lineshape independent sextets (Vandenberghe et al., 1994).

Additional VNIR spectroscopy results
The ferrihydrite used in this study displayed a VIS maximum at 0.801 μm, one Fe 3+ spinforbidden crystal field transition for 6 A1→ 4 T1 at 0.966 μm, and hydration associated absorption features at 1.403 μm and 1.933 μm. Abiotic ferrihydrite heated controls for both organisms, but especially for P. delaneyi, were significantly muted in overall reflectance and depth of spectral features relative to the ferrihydrite reference spectrum. This decrease in spectral contrast was apparent even in the unheated control for P. delaneyi but not for P. islandicum, suggesting a possibly strong effect of marine relative freshwater growth medium on ferrihydrite. Despite the decrease in overall reflectance and spectral contrast, all abiotic unheated and heated controls showed the expected VIS maximum feature for ferrihydrite (0.801 μm), although this position was slightly shifted compared to the reference ferrihydrite. Additionally, all abiotic controls except the unheated control for P. islandicum were missing the characteristic inflection at 0.966 μm and sharp OH absorption at 1.4 μm typical for ferrihydrite (Scheinost et al., 1998;. In the unheated abiotic control for P. islandicum, the OH absorption was shifted towards higher wavelengths ~1.43 μm, similar to that seen for hematite. The lepidocrocite used in this study displayed four Fe 3+ spin-forbidden crystal field transitions at 0.414 μm ( 6 A1→ 4 E), 0.482 μm (2( 6 A1→ 4 T1)), 0.726 μm ( 6 A1→ 4 T2), and 0.978 μm ( 6 A1→ 4 T1), as well as two local VIS maxima at 0.623 μm and 0.794 μm . A similar decrease in spectral contrast relative to the lepidocrocite reference spectrum was noted in the heated abiotic control for P. delaneyi. In this sample, positions of the two VIS maxima, as well as the 6 A1→ 4 T1 transition, were shifted to 0.593 μm, 0.802 μm and 0.966 μm, respectively, relative to the lepidocrocite spectrum. These features were unchanged in the heated abiotic control for P. islandicum, and the unheated abiotic control for both P. delaneyi and P. islandicum. This suggests that when lepidocrocite was the terminal electron acceptor, heated marine but not freshwater media, abiotically primed the starting mineral for transformation.
The akaganéite used in this study showed five Fe 3+ spin-forbidden crystal field absorptions at 0.387 μm, 0.428 μm, 0.512 μm (2( 6 A1→ 4 T1)), 0.650 μm, and 0.970 μm ( 6 A1→ 4 T1), a VIS maximum at 0.728 μm, and hydration bands at 1.456 μm, 1.966 μm, 2.324 μm, and 2.452 μm . The heated abiotic control for P. delaneyi and P. islandicum also showed a decline in spectral contrast compared to the unheated abiotic controls and the akaganéite reference. Both samples retained the positions of the 2( 6 A1 → 4 T1) absorption at ~0.5 μm; however, the 6 A1 → 4 T1 transition typically at 0.97 μm was shifted to shorter wavelengths. Additionally, only the heated abiotic control for P. islandicum, but not P. delaneyi, lost the characteristic sharpness of the VIS maximum at ~0.728 μm for akaganéite. These differences indicate that structural changes in the starting mineral were growth medium specific and affected by heat in separate ways. Neither the typical absorptions nor the VIS maximum for akaganéite were affected in the unheated abiotic controls for either organism.

VNIR slope analysis
Spectral slopes in the 0.475-0.550 μm and 0.86-1.3 μm (Fig. S1, Table 1) ranges are often associated with Fe 3+ and Fe 2+ electronic transitions and are distinctive based on mineral and redox state Dyar et al., 2019). Here, slopes among sample spectra were useful in distinguishing not only redox states, but also the extents of transformation, highlighting subtle differences for the two organisms Positive slopes typically corresponded with Fe 3+ phases, while negative or neutral slopes with Fe 2+ or mixed Fe 3+ -Fe 2+ phases.
In the 0.475-0.550 μm region, bioreduced ferrihydrite for P. delaneyi had a distinctively negative slope, similar to magnetite, while bioreduced ferrihydrite for P. islandicum had an intermediate slope between magnetite and ferrihydrite (Fig. S1A). In contrast, all abiotic controls in this region had positive slopes as expected for Fe 3+ phases. These subtle differences in the bioreduced spectra were consistent with the extent of ferrihydrite reduced by each organism , with P. delaneyi reducing more than P. islandicum. Such distinctions were not as apparent in the 0.86-1.3 μm region, because only the unheated control for P. islandicum showed a positive spectral slope similar to ferrihydrite (Fig. S1D). All other spectra showed neutral or negative slopes similar to those of magnetite or maghemite, making it difficult to distinguish bioreduced and abiotic spectra using this region alone.
Comparison of spectral slopes in the 0.475-0.550 μm (Fig. S1B) and 0.86-1.3 μm (Fig. S1E) regions for lepidocrocite transformations confirmed subtle differences in these spectra. Only the P. delaneyi bioreduced spectrum showed a drastic decrease in slope in the 0.475-0.550 μm region, which appeared similar to ferrous phases, particularly siderite (Fig. S1B). Bioreduced lepidocrocite for P. islandicum showed an intermediate slope, indicating less phase transformation, compared to P. delaneyi. In contrast, all the heated, unheated abiotic controls, and the lepidocrocite reference spectrum showed sharp positive slopes in this region. Slopes in the 0.86-1.3 μm (Fig. S1E) region in the bioreduced spectra displayed a similar trend. They exhibited a sharp decline becoming neutral, similar to siderite, indicating the formation of a ferrous phase for both organisms. This region also displayed a less positive and more neutral slope, suggesting a minor phase transformation or coating by a dark phase in the heated abiotic control for P. delaneyi. All the other heated and unheated abiotic controls showed sharp positive slopes, consistent with the lepidocrocite reference spectrum.
Spectral slopes in the 0.475-0.550 μm ( Fig. S1C) and 0.86-1.3 μm ( Fig. S1F) regions for akaganéite transformations suggested that the bioreduced spectrum for P. delaneyi had the most distinctive Fe 2+ contributions, similar to magnetite or maghemite. Bioreduced akaganéite for P. islandicum had an intermediate or neutral slope, that was indistinct from the heated abiotic control spectrum, particularly in the 0.86-1.3 μm region. The slope for the P. delaneyi heated abiotic control in this region was less positive than the unheated abiotic controls and the akaganéite reference, but not negative or neutral as noted in the bioreduced spectrum (Fig. S1F). This suggests that akaganéite, like lepidocrocite, was abiotically primed for reduction in the presence of heated growth medium.

Additional FTIR-ATR spectroscopy results
For the ferrihydrite samples, heated and unheated abiotic controls for both organisms showed subtle changes maintaining similar absorption positions but presenting different band shapes and relative contributions. These changes were particularly prominent in the 500-750 cm -1 region ( Fig. 2A). For this region, ferrihydrite can be distinguished from other iron oxides by absorption features at 682 cm -1 , 630 cm -1 , and 588 cm -1 in decreasing order of relative contribution. For P. delaneyi, both abiotic controls displayed a deeper absorption at ~588 cm -1 relative to the absorptions at ~680 cm -1 and ~630 cm -1 for ferrihydrite. The heated abiotic control spectrum lacked the 630 cm -1 feature altogether and displayed the ~680 cm -1 feature but shifted to higher wavenumbers. The unheated abiotic control for P. islandicum showed change in band shapes and absorption positions for this region similar to those noted in the unheated abiotic control for P. delaneyi. In contrast, the heated abiotic control for P. islandicum showed a slightly deeper absorption at 588 cm -1 , lacked the 630 cm -1 feature, but maintained the position of the absorption at ~680 cm -1 relative to ferrihydrite. Below 750 cm -1 , the heated abiotic control for P. islandicum showed many similarities in band position, but not band shape and contribution to bioreduced ferrihydrite for P. islandicum. This distinction was crucial for identifying the more subtle biogenic mineral transformations that occurred in the case of P. islandicum.
The features in the phosphate region (900-1200 cm -1 ) were expanded to lower wavenumbers in the bioreduced spectrum for P. delaneyi compared to the unheated and heated abiotic control spectra. This shifted position was also accompanied with a change in shape of these features, making them more consistent with iron-phosphate minerals like beraunite (Fe 2+ Fe 3+ 5(OH)5(PO4)4•4H2O), vivianite (Fe3(PO4)2•8H2O), or chalcosiderite (CuFe 3+ 6(PO4)4(OH)8•4H2O). However, all iron phosphate minerals also have a phosphate bending mode at ~450 cm -1 (Frost et al., 2002;Frost et al., 2014), which was absent in the bioreduced ferrihydrite spectrum for P. delaneyi. Therefore, it is possible that either 1) a different iron phosphate mineral formed, 2) no iron phosphate mineral formed but rather phosphate ions in solution changed between biotic and abiotic conditions, or 3) some of these distinctive features were cell associated and cell specific. Bioreduced ferrihydrite for P. islandicum also showed a similar shift and expansion of these phosphate vibrations towards lower wavenumbers, but unlike P. delaneyi, this was limited to only one of the three absorptions in this region at ~950 cm -1 . While sulfate vibrations are also found in this region, they typically extend above 1150 cm -1 (Lane, 2007;Sklute et al., 2015), making them unlikely contributions in sample spectra. In all three spectra, the ~1465 cm -1 feature for HCO3sorbed on ferrihydrite was absent, and the ~1360 cm -1 feature was shifted to higher wavenumbers.
For the lepidocrocite samples, the unheated and heated abiotic controls showed shifted feature positions and contributions relative to the lepidocrocite reference, suggesting the possible formation of new phases. For instance, diagnostic lepidocrocite features, associated with OH vibrations at ~1150 cm -1 , ~750 cm -1 , and ~460 cm -1 had shifted positions. Additionally, features at ~635 cm -1 and 585 cm -1 changed in depth and position, with the former peak increased in depth, and latter shifted to ~600 cm -1 . A new shoulder also appeared at ~168 cm -1 in these abiotic control spectra relative to the lepidocrocite reference. However, the most striking change was the broadened feature that spanned from ~850-1260 cm -1 in both abiotic control spectra for P. delaneyi. Its position was consistent with symmetric and antisymmetric phosphate stretching vibrations, either as a mineral or solution or sorbed phase. The absence of features at ~780 cm -1 or at ~500 cm -1 suggested that chalcosiderite, vivianite, beraunite or phosphosiderite (FePO4•2H2O) did not form. However, it is plausible that a different phosphate mineral phase not considered here formed in the abiotic conditions. These changes were distinct from those in the bioreduced spectrum for P. delaneyi, in that this entire phosphate region band envelope disappeared, and instead, carbonate features appeared. Such distinct spectral differences between experimental conditions suggest that any Fe 2+ (and/or Fe 3+ ) iron involved in the formation of phosphate phases in the abiotic conditions, got consumed in forming siderite upon lepidocrocite bioreduction by P. delaneyi.
Typical akaganéite features were also affected in all three conditions for P. islandicum as well as the heated abiotic control and bioreduced spectrum for P. delaneyi. The most significant changes were noted in the absorption features of a combined asymmetric band at ~630 cm -1 and ~686 cm -1 . The 630 cm -1 feature is attributed to a δOH out-of-plane bend, and the 686 cm -1 represents the combination of an OH-bending and a ν4 sulfate vibration (Murad and Bishop, 2000;Bishop et al., 2015). Additionally, contributions at 650 cm -1 for this asymmetric band represent OH…Cl in akaganéite (Bishop et al., 2015). In all but the unheated abiotic control for P. delaneyi, mid-peak contributions at ~650 cm -1 decreased in area and intensity, with the lowest intensities in both the bioreduced spectra.

Additional Raman spectroscopy results
Growth media differences and heat without cells caused subtle mineral changes and band position shifts towards lower wavenumbers for the starting oxides (Table 3).
With lepidocrocite, the P. delaneyi unheated and heated abiotic controls displayed all the typical bands for lepidocrocite, although these were slightly shifted to lower wavenumbers (Table 3). In addition, only in the heated control, the feature at 1056 cm -1 , expected for lepidocrocite, was shifted to 1073 cm -1 , indicating possible contributions from the symmetric stretching mode of a carbonate phase, similar to an iron hydroxycarbonate, such as chukanovite (Saheb et al., 2011). Other modes for chukanovite appeared to be lacking, so it is possible that this feature is instead due to aqueous carbonate sorbed on mineral oxides. For the P. islandicum lepidocrocite samples, the feature at 382 cm -1 was found at 377 cm -1 in the unheated control, 375 cm -1 in the heated control, and 373 cm -1 in bioreduced lepidocrocite. Similarly, the band at 1057 cm -1 was present at 1053 cm -1 in the unheated control, 1046 cm -1 in the heated control, and 1036 cm -1 in bioreduced lepidocrocite (Table 3, Fig. 3B). The latter band lies in the region (~950-1100 cm -1 ) where symmetric and antisymmetric stretching modes of PO4 3ions are expected (Frost et al., 2002;Litasov and Podgornykh, 2017). Within this range, features in ferrous phosphates typically are shifted towards lower wavenumbers and features in ferric or mixed ferrous-ferric phosphates towards higher wavenumbers (Fig. S2). Furthermore, the P. islandicum bioreduced spectrum also showed a slight increase in intensity at 310 cm -1 (Fig. 3B). This feature was likely due to a FeO stretching vibration overlapped with the lattice mode of PO4 3- (Frost et al., 2002;Litasov and Podgornykh, 2017;Sklute et al., 2018).
P. delaneyi and P. islandicum heated and unheated abiotic controls with akaganéite displayed all the expected major bands for akaganéite at 318 cm -1 , 389 cm -1 , 545 cm -1 , and 725 cm -1 , and additionally when fitted showed a minor feature in the 1040-1060 cm -1 range, possibly from abiotic phosphate in the growth medium (Fig. 3C, Table 3). While the magnetite reference (12 nm) reported in this study showed an increase in the relative band intensity at 330 cm -1 , Dar and Shivashankar (2013) noted that a 30 nm-synthesized magnetite could also have the same position as our reference for the T2g mode. However, polycrystalline films of 40 nm and 200 nm magnetite synthesized by Jubb and Allen (2010) showed this feature at 310 cm -1 , as did the magnetite (grain size unknown) reported by Hanesch (2009). The magnetite that formed from akaganéite bioreduction in this study had a broader range of larger grained spheres (10-250 nm) . It is possible that the position of the T2g mode in magnetite is grain size dependent, with larger grain sizes showing a shift towards lower wavenumbers, but further work would be needed to confirm this hypothesis.

Additional Mössbauer spectroscopy results
Synthetic ferrihydrite Mössbauer spectra were fit with two doublets (Fig. 4A) or three sextets (Fig. S3D) of Lorentzian lineshape as they allow for more consistent fits and clearer data interpretation. With P. delaneyi, the unheated, abiotic control spectra were largely consistent with the reference ferrihydrite (Fig. 4A, Fig S3A), but showed a small ferrous contribution (1% spectral area, within the error of the technique) resolvable at >80K. At 4K, the spectrum resembled that of the reference. The heated, abiotic control appeared indistinguishable from the reference ferrihydrite at all temperatures (Fig. 4A). For P. islandicum, the unheated, abiotic control appeared similar to the reference ferrihydrite, other than a slight increase in magnetic ordering temperature (apparent at 80K; Fig. 4A, Fig. S3). The heated abiotic control similarly magnetically ordered by 80K (Fig. 4A, Fig. S3) but also showed changes in the combined ferrihydrite peak shape at 4K, specifically in the separation of sextets in the lowest velocity peak (Fig. 3SD). Additionally, minor hematite (3% spectral area) was resolvable at all temperatures.
Reference Mössbauer spectra for lepidocrocite were fit with two Lorentzian doublets ≥80K ( Fig. 4B; Fig. S4) and with two sextets at 4K. Lepidocrocite doublets displayed moderately increasing CS (0.38-0.49 mm/s) and minimally increasing QS (0.52/0.62 -0.54/0.66 mm/s). Lepidocrocite sextets had QS values near zero and fields below 460 kOe (Fig. S4D). With P. delaneyi, Mössbauer spectra for both unheated and heated abiotic controls also showed a small Fe 2+ phase (3-4% spectral area) ( Fig. 4B; Fig. S4). Fit parameters for this phase vary between abiotic controls but the peak is not sufficiently resolved to conclude these changes indicate mineralogical differences. Below 220K, unheated and heated abiotic controls both displayed a small Fe 3+ sextet (5-8% spectral area) (Fig. 4B, Fig. S4B-D). This phase, which is likely superparamagnetic and overlapped with lepidocrocite at 295K, appeared best resolved at 80K, with CS = 0.48 mm/s, QS = -0.26 mm/s, and hyperfine field = 484 kOe. This is most consistent with nanophase goethite (Fig. 4B). In the P. delaneyi bioreduced spectra, one distribution (CS = 1.23 mm/s, QS = 2.56 mm/s) displayed the same parameters as the small Fe 2+ doublet in the abiotic controls. In addition to Fe 2+ distributions and lepidocrocite, the bioreduced spectrum also contained a small Fe 3+ sextet below 220K similar to the one observed in abiotic controls (Fig. 4B, Fig. S4B-D). With P. islandicum, Mössbauer spectra and fit parameters of the unheated and heated abiotic controls were consistent with those of reference lepidocrocite (Fig.  4B).
At 295K, reference nanophase akaganéite was fit with two doublets and one disordered sextet (Fig. S5A). While akaganéite is often adequately fit with two doublets (García et al., 2004;Barrero et al., 2006;Klimas et al., 2015), increased H2O (correlated with decreased Cl -) has been shown to affect ordering temperature, leading to such a sextet (Chambaere et al., 1978). This distribution is unsurprising in our data since all minerals were stored and used as fluid suspensions. At 220K, three sextets with similar CS but different QS and hyperfine fields were resolvable ( Fig. 4C; Fig. S5B-D). The sextet with the most negative QS was associated with the smallest hyperfine field and decreased in area to 8% by 4K. This is consistent with literature reports that conclude this distribution results from temperature effects, not a particular Fe site (Rezel and Genin, 1990). For P. delaneyi, the unheated abiotic control was similar to the reference but also contained a small, unresolvable Fe 2+ doublet (see baseline near 0 mm/s at 220K and 150K) ( Fig. 4C; Fig. S5B-C). The heated abiotic control had altered sextet structure at 295K (Fig. S5A), which can be seen at lower temperatures to be due, in part, to Fe 2+ contributions. This Fe 2+ doublet is similar to that in the P. islandicum heated control at 220K, but the 150K and 80K fit parameters show this is a different phase (Fig. 4C, Fig. S5C). For P. islandicum, the unheated abiotic control appeared similar to reference akaganéite over all temperatures. The heated abiotic control additionally displayed a Fe 2+ doublet at 220K (~7% spectral area) (Fig. S5B) and also a Fe 3+ component that persisted far below the splitting temperature for akaganéite. If the Fe 3+ and Fe 2+ components are from the same mineral, it could be a mixed valent phosphate or sulfate (Dyar et al., 2013;Dyar et al., 2014). Adsorbed Fe(II) can display parameters similar to those for the Fe 2+ doublet (ThomasArrigo et al., 2014), but all Fe 2+ components in this sample magnetically order at 4K, making adsorbed Fe 2+ an unlikely explanation.

Geochemical modeling: Methods and Results
To predict mineral phases that precipitate from the growth media, components of each growth medium were speciated and analyzed using the REACT module in The Geochemist's Workbench v.10.0 software package (Aqueous Solutions, LLC). For ferrihydrite analysis, Fe 3+ was swapped with precipitated Fe(OH)3. For lepidocrocite and akaganéite analysis, Fe 3+ was swapped with goethite. All minerals were allowed to precipitate and dissolve in the model during mixing. C, Fe, and NH4 + were decoupled from the reactions but all other reactions were allowed. To account for the hydrogenotrophic conditions for P. delaneyi, H2(aq) with a fugacity of 1.92 was swapped for O2. The module was run at pH 6.8 and 90°C for P. delaneyi and pH 6.8 and 95°C for P. islandicum. Mineral saturation states based on log Q/K values at equilibrium were used to assess which of the supersaturated minerals were expected to precipitate. Table S1 and S2 show the mineral saturation states for initial and final conditions of the growth medium for P. delaneyi and P. islandicum, respectively, with either ferrihydrite, lepidocrocite, or akaganéite as the terminal electron acceptor. Initial conditions were simulated to be indicative of the unheated and heated abiotic controls, and final conditions were simulated to represent an equivalent amount of Fe 2+ as present in bioreduced conditions.
With ferrihydrite, lepidocrocite, and akaganéite, saturation index values indicated that the P. delaneyi growth medium initially was favorable to abiotic precipitation of hydroxyapatite, magnesite, vivianite, hematite, and pyrrhotite (Table S1). Geochemical modeling of final conditions indicated that the growth medium for P. delaneyi provided conditions suitable for abiotic precipitation of all these phases, as well as siderite with ferrihydrite, lepidocrocite, and akaganéite as the starting minerals (Table S1). Magnetite and possibly vivianite were identified for ferrihydrite bioreduced by P. delaneyi in our dataset. Bioreduced lepidocrocite, on the other hand, resulted in the formation of siderite, some ferrous phosphate, and possibly nanophase goethite, while bioreduced akaganéite showed formation of magnetite, and vivianite. Given that some but not all the identified phases were predicted in the geochemical modeling, it was apparent that metabolic activity and presence of P. delaneyi controlled mineral precipitation processes.
For P. islandicum, initial conditions with ferrihydrite were favorable for the precipitation of vivianite, hematite, and hydroxyapatite (Table S2). Final conditions with ferrihydrite predicted that only vivianite and hematite of these phases were saturated. With lepidocrocite and akaganéite, only hematite and hydroxyapatite were saturated in the growth medium initially, but whitlockite, hematite and vivianite precipitated under final conditions. Magnetite, which was produced from bioreduction of ferrihydrite by P. islandicum, was not predicted in geochemical modeling. However, for lepidocrocite and akaganéite transformations, modeled results matched identified iron mineral phases, except for hematite, suggesting that abiotic mineral precipitation processes largely controlled phase formation in the case of P. islandicum. Fig. S2. Raman spectra of iron phosphates obtained from the RRUFF databasevivianite (050076), ludlamite (040152), sarcopside (070553), gladuisite (070432), barbosalite (070358), beraunite (060696), heterosite (070148), allanpringite (080147). Measured chemistry is described where known for each mineral. Yellow shaded region is the narrow range noted for stretching modes of phosphate in Fe 2+ phosphates and purple shaded region for mixed valent (Fe 2+ -Fe 3+ ) as well as Fe 3+ phosphates. Fig. S3. Mössbauer spectra of P. delaneyi (P. del) and P. islandicum (P. isl) grown on ferrihydrite. The raw and fitted spectra and all fitted parameters are shown for the unheated abiotic control ('no heat, no cells'), heated abiotic control ('heat, no cells'), and biotic ('heat, cells') conditions of each organism along with reference ferrihydrite and magnetite (at 4K). Each spectrum is scaled and offset for clarity. The temperature of acquisition is shown at the base of the x axis. Figure S4. Mössbauer spectra of P. delaneyi (P. del) and P. islandicum (P. isl) grown on lepidocrocite. The raw and fitted spectra and all fitted parameters are shown for the unheated abiotic control ('no heat, no cells'), heated abiotic control ('heat, no cells'), and biotic ('heat, cells') conditions of each organism along with reference lepidocrocite. Each spectrum is scaled and offset for clarity. The temperature of acquisition is shown at the base of the x-axis. Fig. S5. Mössbauer spectra of P. delaneyi (P. del) and P. islandicum (P. isl) grown on akaganeite. The raw and fitted spectra and all fitted parameters are shown for the unheated abiotic control ('no heat, no cells'), heated abiotic control ('heat, no cells'), and biotic ('heat, cells') conditions of each organism along with reference akaganeite. Each spectrum is scaled and offset for clarity. The temperature of acquisition is shown at the base of the x-axis.