Investigating Nanoscale Electron Transfer Processes at the Cell-Mineral Interface in Cobalt-Doped Ferrihydrite Using Geobacter sulfurreducens: A Multi-Technique Approach

Cobalt is an essential element for life and plays a crucial role in supporting the drive to clean energy, due to its importance in rechargeable batteries. Co is often associated with Fe in the environment, but the fate of Co in Fe-rich biogeochemically-active environments is poorly understood. To address this, synchrotron-based scanning X-ray microscopy (SXM) was used investigate the behaviour of cobalt at the nanoscale in Co-Fe(III)-oxyhydroxides undergoing microbial reduction. SXM can assess spatial changes in metal speciation and organic compounds helping to elucidate the electron transfer processes occurring at the cell-mineral interface and inform on the fate of cobalt in redox horizons. G. sulfurreducens was used to reduce synthetic Co-ferrihydrite as an analogue of natural cobalt-iron-oxides. Magnetite [Fe(II)/Fe(III)3O4] production was confirmed by powder X-ray diffraction (XRD), SXM and X-ray magnetic circular dichroism (XMCD) data, where best fits of the latter suggested Co-bearing magnetite. Macro-scale XAS techniques suggested Co(III) reduction occurred and complementary SXM at the nanoscale, coupled with imaging, found localised biogenic Co(III) reduction at the cell-mineral interface via direct contact with outer membrane cytochromes. No discernible localised changes in Fe speciation were detected in the reordered cobalt-iron-oxides that were formed and at the end point of the experiment only 11% Co and 1.5% Fe had been solubilised. The solid phase retention, alongside the highly localised and preferential cobalt bioreduction observed at the nanoscale is consistent with retention of Co in redox zones. This work improves our fundamental molecular-scale understanding of the fate of Co in complex environmental systems and supports the development of biogenic Co-doped magnetite for industrial applications from drug delivery systems to magnetic recording media.


INTRODUCTION
Cobalt is designated a critical metal by the EU (European Commission, 2017) with increasing demand coming from industrial applications including superalloys for the aerospace industry (Shedd, 2015;Pollock, 2016;Slack et al., 2017), 'permanent' magnets for military and industrial applications (Tkaczyk et al., 2018) as well as its central role in rechargeable battery technology (Hein & Koschinsky, 2013;Hein et al., 2015;Shedd, 2015). Typically, cobalt is mined from minerals such as sulphosalt cobaltite [(Co 2 Fe)As], heterogenite [HCoO 2 ] and Corich asbolane [(Ni, Co) x Mn(IV) (O,OH) 4 •nH 2 O] (Smith, 2001;Vanbrabant et al., 2013;Burlet and Vanbrabant, 2015;Dehaine et al., 2021); however, cobalt can also be found associated with ferrihydrite in soil sediments (Woodward et al., 2018) and polymetallic nodules (Mizell and Hein, 2018) and these have been identified as potential new sources of cobalt. In addition, anthropogenic activities, such as mining (Sharma and Agrawal, 2005), have led to increases in heavy metal pollution in soils and sediments. Cobalt is an essential nutrient in trace quantities, however, at elevated concentrations cobalt is known to cause physiological problems, such as insufficient chlorophyll production in plants and respiratory/cardiac diseases in mammals (Agency for Toxic Substances and Disease Registry, 2004;Slack et al., 2017;Woodward et al., 2018).
Cobalt is typically released to the environment as either Co(II) or Co(III) (Kosiorek and Wyszkowski, 2019), however, it is most commonly found as Co(II), the more thermodynamically stable species (Faucon et al., 2007;Newsome et al., 2020;Dehaine et al., 2021). Co(II) can be oxidised to Co(III) by Fe(III)-bearing minerals (Cornell and Giovanoli, 1989;Zachara et al., 2001;Pozas et al., 2004;He et al., 2020) and substituted into the crystal lattice of minerals such as goethite (Bousserrhine et al., 1999); as such, the fate of Co in the environment and the behaviour of Fe are intertwined Zachara et al., 2001;Grybos et al., 2007;Poggenburg et al., 2016). Ferrihydrite, an Fe(III)-oxyhydroxide ubiquitous in the environment, is characterised by its nanoparticulate nature, poor crystallinity, low stability and high dispersion (Dritus et al., 1992;Yee et al., 2006;Cismasu et al., 2016). Ferrihydrite has a high surface area and can readily sorb to the surface, or substitute small quantities of trace metals, such as Co, within the mineral structure (Cornell and Giovanoli, 1989;Kukkadapu et al., 2004). Co(III) [0.0545 nm] has a similar ionic radius to Fe(III) [0.0645 nm]; this, along with the valency of the cobalt ions and their electronic properties, in comparison to high-spin Fe(III), allows isomorphic substitution thus explaining the environmental occurrence of cobalt-iron-oxides .
Polymetallic nodules, also known as ferromanganese nodules, are formed hydrogenetically, where seawater is the exclusive source of the metals (Valsangkar et al., 1992;Atkins et al., 2014), or diagenetically, where metals are supplied from pore waters of seafloor sediments (Nitahara et al., 2017;Maciąg et al., 2019). Hydrogenetic nodules are typically enriched in Co, Te, platinum group metals (PGM) plus rare earth elements (REEs), while diagenetic nodules are enriched in Ni, Cu, Mn, and Li (Hein and Koschinsky, 2013). Most nodules, however, have a hybrid origin (Mohwinkel et al., 2014;Mizell and Hein 2018) and Co enrichment in these has been documented (Banakar, 2010;Marsh et al., 2013;Mohwinkel et al., 2014;Josso et al., 2017;Petavratzi et al., 2019), hosted within the structure of poorly crystalline Fe minerals such as ferrihydrite (Mizell and Hein, 2018). Polymetallic nodules contain an untapped cobalt resource yet there is currently no economically viable industrial scale cobalt extraction process (Petavratzi et al., 2019). The demand for cobalt, combined with concerns regarding future supply, has encouraged governments to investigate alternative sources of this critical raw metal and to improve our understanding of the biogeochemical cycling of cobalt (Naden, 2017). Furthermore, the fate and mobility of cobalt from Co-ferrihydrite warrants further investigation to aid in the development of novel remediation practices (Woodward et al., 2018).
In this study, we investigated the behaviour of cobalt during the enzymatic reduction of a synthetic Co-ferrihydrite using G. sulfurreducens, as an analogue to bioreduction processes in the natural environment. Using a combination of geochemical measurements, electron microscopy, powder X-ray diffraction (PXRD) and X-ray absorption spectroscopy (XAS), changes in metal speciation and mineral phases were identified. Synchrotron-based scanning X-ray microscopy (SXM) was used to map elemental and chemical information for both bacteria and mineral phases by directly measuring at the C K-edge and the transition metal L 2,3 -edges, with a high spatial (<40 nm; (Chao et al., 2012;Karunakaran et al., 2015)) and spectral resolution (0.1 eV; (Xia et al., 2013;Karunakaran et al., 2015)). SXM suggested highly localised Co(III) reduction in regions co-located with bacterial cells; however, no spatial changes in Fe speciation were observed. Analyses also suggested a high degree of solid phase cobalt and iron retention, despite a change to a more reduced mineral phase. These data can be extrapolated to understand the natural biogeochemical cycling of cobalt in Fe-rich environments undergoing microbially-mediated redox transformations.

Synthesis and Characterisation of Cobalt-Doped Ferrihydrite
Cobalt-doped 2-line ferrihydrite was prepared according to a method adapted from Cornell and Schwertmann (2003). Briefly, using a stirring plate, 20 g of iron (III) chloride hexahydrate (FeCl 3. 6H 2 O) and 3.1 g of cobalt (II) chloride hexahydrate (CoCl 2 .6H 2 O) was dissolved in 500 ml of deionised water. 10 M NaOH was used to increase the pH from 1.7 to 6.9 and the subsequent mineral precipitate was washed five times using de-ionised water. The mineral precipitates were then resuspended in 200 ml deionised water and stored at 4°C.
The mineral composition of the final solids was confirmed using (powder) X-ray diffraction (PXRD). A Bruker D8Advance instrument with anti-scatter slits opened to 3°and a scanning range of 10°-70°(2θ), using a step size of 0.02°(2θ) was used to characterise the mineralogy of key time points in the biogenic reduction reaction. The International Centre for Diffraction Data database; PDF4 Mineral 2021 package was used to analyse the data.
An aliquot of the final synthesized mineral was acidified in 2 w/v % HNO 3 to determine the concentration of Fe and Co using inductively coupled plasma atomic emission spectroscopy (ICP-AES, PerkinElmer Optima 5,300 dual view), located in a clean laboratory environment (Class 1,000) at the Manchester Analytical Geochemistry Unit.
Culturing Geobacter sulfurreducens G. sulfurreducens was obtained from the laboratory culture collection of the Geomicrobiology Group in the Williamson Centre for Molecular Environmental Science at the University of Manchester. Standard methods regarding culturing and strict anaerobic preparation of cells of G. sulfurreducens were followed as described previously (Lloyd 2003). The cultures were grown at 30°C in modified fresh water medium with sodium acetate (10 mM) and fumarate (40 mM) as the electron donor and acceptor, respectively. All manipulations were carried out under an atmosphere of N 2 -CO 2 (80:20). Late logarithmic phase cultures of G. sulfurreducens were harvested by centrifugation at 4°C (5,000 g; 20 min) and washed three times in sodium bicarbonate buffer (NaHCO 3 ; 30 mM, pH 7.1) under N 2 -CO 2 (80:20) prior to use.

Experimental Procedure
An anaerobic stock solution of Co-ferrihydrite in 10 mM sodium bicarbonate buffer was prepared, yielding a slurry of 2 mmol/L Fe at circum-neutral pH. The experimental samples contained 5 ml of the anaerobic stock solution, 10 mM sodium acetate (the electron donor) and an aliquot of the cell suspension, washed to achieve a final optical density (OD 600 ) of 0.6 (Jenway 6,715 series UV-vis spectrophotometer). Experiments were conducted in triplicate unless stated otherwise. All sample bottles were incubated in the dark at 30°C.
Experiments were monitored at 1, 9, 20, and 72 h. The ferrozine assay (Stookey, 1970) was used to measure the concentration of Fe(II) and total bioavailable Fe [Fe (tot)] by digestion of sample aliquots (suspension samples taken from well shaken bottles) in 0.5 M HCl for 1 h to measure Fe(II) and the addition of 0.25 M hydroxylamine hydrochloride for total Fe. Aliquots of sample suspensions were centrifuged (5 min; 16,000 g) and the supernatant was acidified for ICP-AES and inductively coupled plasma mass spectrometry (ICP-MS, VG Elemental Plasmaquad 2 STE) to monitor the concentration of Fe and Co in the aqueous phase, respectively.
At corresponding timepoints, samples were prepared for characterisation by spectroscopy and microscopy. One sample bottle was opened under a nitrogen atmosphere and 2 μl aliquots were mounted on carbon-coated Cu grids for transmission electron microscopy (TEM) and SXM analyses. The remaining sample slurry was dried under nitrogen, where a portion was attached to carbon tape on a copper holder and maintained in an anaerobic environment for bulk L 2,3 edge XAS analysis while the remainder was used for PXRD analysis.

Transmission Electron Microscopy (TEM)
A FEI Tecnai T20 (200 kV LaB6) instrument fitted with a X-Max 80 SSD Energy Dispersive X-ray analysis (EDS) system (Oxford Instruments; Aztec software) using a Gatan Orius SC200 CCD camera (Gatan GMS 2 software) was used to observe the interactions between the microbial cells and new mineral phases produced by G. sulfurreducens. The resultant TEM images are bright-field images.

X-ray Magnetic Circular Dichroism (XMCD)
Fe L 2,3 edge X-ray magnetic circular dichroism (XMCD) spectra were collected on beamline 6.3.1 at the Advanced Light Source (ALS), Berkeley, United States, using a spot size of 180 μm × 80 µm (Lawrence Berkeley National Laboratory, 2021) in totalelectron yield (TEY) mode with an analysis depth of~4.5 nm (van der Laan and Thole, 1991;Coker et al., 2007). The samples were orientated central to the magnetic pole pieces to allow the X-ray beam to enter through a hole parallel and anti-parallel to the magnetic field and perpendicular to the sample surface (Coker et al., 2007). Scanning between 705-730 eV, the Fe L 2,3 edge XAS data were collected for the two opposite magnetisation directions, and then normalised to the incident beam intensity and subtracted from each other to give the XMCD spectrum (Pattrick et al., 2002).
XMCD was used to determine the site occupancy of Fe cations in the biogenic mineral phases (Coker et al., 2007). To obtain the relative amounts of the three Fe sites: Fe d 6 O h , d 5 T d and d 5 O h ; the experimental spectra were fitted by means of a non-linear least-squares analysis as reported in Coker et al. (2007), using calculated spectra for each of the Fe sites provided by van der Laan and Thole (1991).
Bulk L 2,3 Edge X-ray Absorption Spectroscopy (XAS) Bulk Mn and Co L 2,3 edge X-ray absorption spectroscopy (XAS) data were collected on beamline 6.3.1 at the Advanced Light Source (ALS), Berkeley, United States using TEY giving a typical analysis depth of~4.5 nm into the sample (Coker et al., 2009). The experimental spectra were normalised to the incident beam.

Scanning X-ray Microscopy (SXM)
Scanning X-ray microscopy (SXM) data were collected at the Diamond Light Source (DLS), UK, on beamline I08 using a 25 nm Fresnel zone plate lens. Experimental spectra were calibrated to the C 1s-π* transition associated with carbonyl carbon in amide moieties present in proteins (Lawrence et al., 2003;Cody et al., 2008;Coker et al., 2012) and at the Fe and Co L 2,3 -edges using the Fe(II) peak at 708.7 eV (Bourdelle et al., 2013;Hawkings et al., 2018) and the Co(III) peak at 780.6 eV (Istomin et al., 2015). Image sequences (stacks), were collected at prescribed intervals over the relevant absorption edges, that is, C K edge: 280-310 eV, Fe L 2,3 edge: 700-735 eV and Co L 2,3 edge: 770-805 eV, with smaller increment steps at the L 3 edge to increase spectral sensitivity over this quantitative edge (Pattrick et al., 2002). To minimise beam damage, SXM data were obtained in order of increasing photon energies. Furthermore, dwell times # 5 ms were used and the shutter remained closed when data were not being gathered. No radiation damage was observed under these experiment conditions.
The 'Jacobson stack analyse' function (Jacobsen et al., 2000) from the "aXis2000" software (Hitchcock, 2011) was used to align the transmission images in the stacks. Sample free areas of the carbon film Cu grids were used to normalize the transmission signal obtained from analysed regions of interest. Transmission images from the stacks were converted to optical density (OD) images according to Beer-Lambert's law (Eq. 1), where I 0 is the incident photon intensity and I the transmitted intensity (Moffet et al., 2011).
To monitor spatial and temporal changes in metal speciation, Fe and Co L 2,3 -edge SXM spectral measurements were fitted by means of a non-linear least-squares analysis using calculations for Fe(II), Fe(III), Co(II) and Co(III) in octahedral coordination (Supplementary Figure S1) to determine the proportion of each metal species as described previously by van der Laan and Thole (1991). Spectral modelling of this nature has been shown to give quantitative results at the L 3 peak while the L 2 peak notes a qualitative agreement (Pattrick et al., 2002). Interpretation of spectra via this method has its limitations as the experimental spectra can be altered (e.g., Jahn-Teller distortion (van der Laan and Kirkman, 1992; Glatzel et al., 2005), consequently, the values of each metal species have a 10% error (Coker et al., 2009). Fe L 2,3 -edge images stacks collected at the 9 and 20 h timepoints had multiple regions of interest that were too thick to facilitate transmission, producing broad spectra due to oversaturation. Masks comprised of pixels from regions of appropriate thickness were prepared using the 'aXis2000' software (Hitchcock, 2011) to correct for this. The spectra from these masks were used alongside the aggregate SXM spectra, prepared from averages from the regions analysed.

Bulk Characterisation
Cobalt-ferrihydrite, with a final molar ratio of 1:0.12 Fe:Co, was subjected to bioreduction by G. sulfurreducens Elevated Fe(II) was detected in experimental samples where up to 0.8 mM Fe(II) (40% of total Fe) was found in the cell-mineral suspension by the ferrozine assay ( Figure 1A). Minor Co release was associated with the addition of G. sulfurreducens with 17 μM released in the experiment samples compared to 5.2 μM in the control at the final timepoint ( Figure 1B). At the final timepoint, 28 μM Fe had been released to solution in the experiment samples versus a negligible release (1.1 μM) in the no electron donor control ( Figure 1C).
Fe L 2,3 edge XAS data were collected in opposing magnetic fields and subtracted to give XMCD spectra for the 9 h sample shown here in Figure 2B with best fits derived from the calculated spectra (van der Laan and Thole, 1991). These data evaluated the magnetic component of the samples, therefore only the magnetite, and noted an Fe(II):Fe(III) ratio of 0.46, similar to stoichiometric magnetite (0.50) and slightly less Fe(II) than previous samples (Coker et al., 2007) of pure biogenic magnetite (0.55) ( Table 1). Normalising the data to Fe(III) O h equal to 1 (Table 1), an estimate of the chemical formula for the Co-substituted ferrite can be obtained of Co 0.38 Fe 2.73 O 4, which is the difference between the fit for the experimental data compared to that for biogenic magnetite from previous work (Coker et al., 2009).

Transmission Electron Microscopy
Transmission electron microscopy (TEM) images were taken at 1, 9, 20 and 72 h. At 1 h G. sulfurreducens cells (1.5 μm) were observed closely associated with electron dense clumps of small nano-particles (~2 nm) ( Figure 3A) and feather-like structures (1.5 μm) ( Figure 3B). At 9 h the feather-like structures (0.75-1 μm) ( Figure 3C) remained present, as well as dense aggregated particulates (~1 μm) ( Figure 3D) which dominated. Bacterial rods remained in contact with the mineral phases; these two structures were likely the hydrated sodium bicarbonate from the buffer solution and the magnetite, respectively, as identified by the PXRD in Figure 2A(II)). At 20 h, dense aggregated particulates (~1 μm) (Figures 2E,F) and feather-like structures 2 μm in length were again observed ( Figure 2F). At the final time point at 72 h, local agglomerations of rounded granules~0.3 μm in diameter were prominent ( Figure 2G), with residual laths (100 nm) ( Figure 2H). Here, the bacteria were no longer closely associated with the mineral phase.

Macro-scale Spectroscopy
Surface sensitive Fe and Co L 2,3 edge XAS ( Figure 4) measurements were taken on a bulk homogenised powder of dried sample slurry measured in TEY mode; results showed a shift in the position of the main L 3 peak for both elements to lower photon energies over the course of the experiment, suggesting a change to a more reduced species (Glatzel et al., 2005). Modelling the Fe and Co L 2,3 edge, using calculated spectra for 2+ and 3+ species, indicated that the synthetic Co-doped ferrihydrite initially contained 76% Co(II) and 24% Co(III) and 100% Fe(III) (Tables 2, 3). A marked difference was observed after 1 h where 88% Co(II) was detected as well as 21% Fe(II). The maximum proportion of Fe(II), 36%, occurred at 9 h and was maintained at 20 h. No bulk Fe data were available for the 72 h timepoint. At the final timepoint at 72 h, the proportion of Co(II) (90%) remained similar to that observed after 1 h (Table 2; Figure 4). Co and Fe L 2,3 -edge data from the areas analysed using SXM in transmission mode, called "aggregate SXM" data (see Supplementary Table S1 for sample area and pixels used), are spatially constrained averages comparable to the surface sensitive bulk data ( Figure 4; Table 2, Table 3). Aggregate Co L 2,3 -edge SXM data showed excellent agreement with the bulk data suggesting a Co(II)-rich end point to the experiment ( Table 2) and at corresponding timepoints, the proportions of Co(II) indicated by both XAS techniques were within 10% error (Coker et al., 2009). The aggregate Fe L 2,3 -edge data, alike with the bulk data, showed a significant change in Fe speciation at 1 h where 46% Fe(II) was detected by the former technique (Table 3). Fe L 2,3 -edge aggregate SXM data noted a Fe(II):Fe(III) ratio of 50:50 at 9 h, again indicating a more reduced Fe-bearing mineral, such as magnetite (Coker et al., 2007).

Bacterial Cells
To spatially resolve the location of the bacterial cells relative to the mineral phases, C K-edge stack maps were produced at 288.2 eV, corresponding to a 1s-π* transition associated with carbonyl carbon in amide moieties associated with proteins (Lawrence et al., 2003;Cody et al., 2008;Coker et al., 2012). Images mapping the location of Fe and Co were produced at L 3 -edge peak intensities for Fe (~710 eV) and Co (780.6 eV) and combined with the carbon map to create RGB composite images ( Figures  5-8). The RGB images serve to identify mineral regions co-  Figure 2B). * From Coker et al. (2007).

Sample
Fe (  located (X or EX if both Co and Fe were not measurable at the same location) with bacterial cells and regions not co-located (O) with bacterial cells. Fe and Co L 2,3 -edge spectra from co-located and non-co-located points were modelled as above for the bulk data (Supplementary Figure S2 and Supplementary Figure S3). Spectra from multiple regions of interest were analysed and the indicated proportions of metal speciation from the best fits were averaged (mean) ( Supplementary Table S2 and Supplementary  Table S3). Using this nanoscale approach, significant changes in Co(II) speciation were observed (Figure 9).

Cobalt SXM
On average, at 1 h 82 ± 3.1% Co(II) was detected in non colocated regions and 89 ± 6.5% was detected in co-located regions ( Figure 9; Table 2); therefore within experimental error (Coker et al., 2009). Analysis of the 9 h sample however detected 82 ± 2.1% Co(II) in the non co-located regions and a significant increase in the Co(II) proportion to 95 ± 7.4% in regions colocated with the bacterial cells ( Figure 9; Table 2). In common with the 9 h sample, the 20 h sample also noted significantly more Co(II) in the mineral regions in close proximity to the cells with 74 ± 1.5% Co(II) detected in the non co-located regions and 85 ± 23.5% detected in the co-located regions ( Figure 9; Table 2). Finally, at 72 h in regions co-located with the cells, the proportion of Co(II) increased significantly from 80 ± 1.8% to 97 ± 5% ( Figure 9; Table 2).

Iron SXM
In contrast to the Co SXM data, a comparison of the Fe species between the non co-located and co-located regions at 1, 20 and 72 h showed no appreciable difference in Fe speciation either spatially or temporally after the 1 h timepoint, with 50 ± 2.7% Fe(II) found in co-located regions at the end point of the experiment ( Figure 10; Table 3).
Frontiers in Earth Science | www.frontiersin.org May 2022 | Volume 10 | Article 799328 experiment, Fe(II) initially increased ( Figure 1A), which is indicative of reductive dissolution Hansel et al., 2003;Weber et al., 2006;Yang et al., 2010;Li et al., 2012;Newsome et al., 2018;Kappler et al., 2021). The aqueous Fe(II) then adsorbed to the surface of remaining ferrihydrite and induced the formation of magnetite, confirmed by PXRD, XMCD ( Figure 2) and aggregate SXM ( Table 3) and imaged using TEM. The removal of Fe(II) from solution during magnetite formation explained the limited extent of Fe in solution at the final timepoint ( Figure 1C). Co(III/II) can be substituted into the crystal lattice of Fe(III) minerals, such as goethite (Bousserrhine et al., 1999). During the enzymatic reduction and recrystallisation of the Co-bearing ferrihydrite, reductive dissolution of Fe(III) would prompt the release of incorporated Co(III/II), consistent with aqueous geochemistry observations ( Figure 1B) (Lee et al., 2001;Lloyd, 2003;Zhang and Cheng, 2007;Joshi et al., 2018). Co(III) is a strong oxidising agent, and no stable dissociated Co(III) compounds exist in aqueous solution (Ma et al., 2008). It is therefore likely that the minor amount of solubilised cobalt at the end point is Co(II), which is the more thermodynamically stable species (Faucon et al., 2007;Newsome et al., 2020;Dehaine et al., 2021), highly soluble in water and potentially very mobile (Woodward et al., 2018). However, the majority of the Co remained associated with the solid phase during microbial reduction suggesting that the solubilised Co was reincorporated into the magnetite during Fe mineral recrystallisation, explaining the fluctuations in Co concentration ( Figure 1B). Co(III) can be biotically reduced to 3 | Summary of Fe species proportions from key time points according to best fits using octahedral metal calculation for bulk L 2,3 -edge data and average SXM data points. Aggregate SXM, spectra of the entire area imaged and analysed.  . The corresponding composite RGB (red, green, blue) image (D) shows the spatial relationship between the bacterial cells (C K-edge) and the minerals (Fe and Co L 2,3 -edges). Spectra were taken from regions colocated with the bacterial cells (X or EX) and regions not co-located to the bacterial cell (O). The centre of the "X" and "O" denote the precise location from which the spectra were collected. All scale bars are 7 μm.
FIGURE 6 | Elemental maps for the 9 h timepoint for (A) C K-edge (288.2 eV; red), (B) Fe L 2,3 -edge (710.9 eV; green) and (C) Co L 2,3 -edge (780.6 eV; blue). The corresponding composite RGB (red, green, blue) image (D) shows the spatial relationship between the bacterial cells (C K-edge) and the minerals (Fe and Co L 2,3 -edges). Spectra were taken from regions colocated with the bacterial cells (X) and regions not co-located to the bacterial cell (O). The centre of the 'X' and 'O' denote the precise location from which the spectra were collected. All scale bars are 2 μm.
Frontiers in Earth Science | www.frontiersin.org May 2022 | Volume 10 | Article 799328 Co(II) as first noted by Gorby, Caccavo and Bolton (1998) for Co(III)-EDTA, therefore the increased detection of Co(II) may have been due to the enzymatic reduction of Co(III) by G. sulfurreducens; however, from the macro scale spectroscopy it was not possible to determine the precise mechanism that was responsible for the increase in Co(II) detected. Borch et al. (2010) and Zachara et al. (2001) suggested that during reductive dissolution processes, Fe(II) mobilisation may be interrupted by the adsorption or coprecipitation of trace metals with the newly formed minerals. Therefore, Fe(II) mobilisation enabled the cobalt to become incorporated into the structure of the magnetite explaining the deficit at the Fe(II)O h site, compared with biogenic magnetite (Table 1). This suggested that the solubilised Co(II) was reincorporated here, the expected site for Co substitution as seen previously (Wu et al., 2007;López et al., 2008;Coker et al., 2009).

Changes in Bulk Speciation of Co and Fe Over 72 h
During mineral synthesis, 24% (  , due to similar ionic radii Pozas et al., 2004), forming the Coferrihydrite. The bulk Co L 2,3 -edge XAS data noted an early significant increase in the proportion of Co(II), suggesting a rapid reduction of Co(III) producing a Co(II) rich mineral ( Table 2). The spatially constrained aggregate SXM data showed excellent agreement with the bulk data, but noted a significant fluctuation in the proportion of Co(II) at 20 h ( Table 2), highlighting  . The corresponding composite RGB (red, green, blue) image (D) shows the spatial relationship between the bacterial cells (C K-edge) and the minerals (Fe and Co L 2,3 -edges). Spectra were taken from regions co-located with the bacterial cells (X) and regions not co-located to the bacterial cell (O). The centre of the 'X' and 'O' denote the precise location from which the spectra were collected. All scale bars are 3 μm.
Frontiers in Earth Science | www.frontiersin.org May 2022 | Volume 10 | Article 799328 heterogeneity within the sample that the bulk data was unable to detect. Both the bulk and aggregate Fe L 2,3 edge data showed a significant early increase in Fe(II) ( Table 3) suggesting rapid Fe(III) reduction (Roden and Zachara, 1996;Thamdrup, 2000;Hansel et al., 2003;Han et al., 2018;Newsome et al., 2018). However, the aggregate SXM data detected more Fe(II) compared to the bulk data again demonstrating the ability of the aggregate SXM data to highlight heterogeneities within the samples. Given the limited extent of metal solubilisation and therefore the high degree of metal retention in the solid phase, the trends observed in the XAS data are likely representative of the behaviour of the majority of the Fe and Co present in the experiment.

A Greater Proportion of Cobalt Reduction Occurred in the Localised Cell Environment
Between 1 and 20 h ( Figures 3A-F), mineral phases were closely associated with the cell surface, where outer membrane c-type cytochromes, previously implicated in metal reduction, are located (Lovley et al., 2004;MacDonald et al., 2011;Malvankar et al., 2011;Levar et al., 2016;Newsome et al., 2018). Furthermore, in regions where bacteria and mineral were co-located, significantly more Co(II) was detected suggesting the possibility of highly localised enzymatic Co(III) reduction. The proximity between the cells and mineral phases, and the localised increase in Co(II) is consistent with outer membrane cytochromes mediating the reductive transformation of insoluble metal phases directly. At 72 h the reduced mineral phase was distanced from the cell surface ( Figures 3G,H), indicating the reducibility of the mineral phases had been exhausted with time. The nanoscale Co L 2,3 edge SXM data highlighted higher standard deviations and broader spreads of data in co-located regions (Figure 9, Supplementary Table S2) indicating that cells had a varying metabolic potential and therefore identifying heterogeneity within the system. Fe L 2,3 edge SXM data at the nano scale indicated no spatial changes in Fe speciation from 1 h as the Fe(II):Fe(III) ratio in the system had spatially stabilised whilst the Co(II):Co(III) ratio was still evolving.

CONCLUSION
This work improves our fundamental molecular-scale understanding of the fate of Co in complex environmental systems and supports the development of Co bioprocessing and remediation options in the future. The solid phase retention, alongside the highly localised and preferential cobalt bioreduction observed at the nanoscale is consistent with retention of Co in sediment redox zones, while the Comagnetite end product of bioreduction in these laboratory experiments has potential industrial applications from drug delivery systems to magnetic recording media.

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

AUTHOR CONTRIBUTIONS
DB, LN, JL and VC contributed to conception and design of the study. DB performed with bulk of the data collection with MK, FIGURE 10 | Box and whisker plot diagrams showing the changes in Fe speciation over the course of the experiment. The box and whisker plots for the starting material have a dotted outline. The box and whisker plots from the non co-located regions have a solid black outline while those from the co-located regions have coloured outline. In both spatial regions the Fe(III) and Fe(II) data is represented with blue and red, respectively. FIGURE 9 | Box and whisker plot diagrams showing the changes in Co speciation over the course of the experiment. The box and whisker plots for the starting material have a dotted outline. The box and whisker plots from the non co-located regions have a solid black outline while those from the co-located regions have coloured outline. In both spatial regions the Co(III) and Co(II) data is represented with red and blue, respectively.
Frontiers in Earth Science | www.frontiersin.org May 2022 | Volume 10 | Article 799328 BK, TA, VC (SXM), JW (XRD), HB (TEM), AN and VC (XAS and XMCD). DB performed the bulk of the data analysis with assistance from MK, BK, TA (SXM), JW (XRD), and GL (XAS and XMCD). DB wrote the first draft of the manuscript. JL, LN, and VC assisted with extensive manuscript review. All authors contributed to manuscript revision, read, and approved the submitted version.