NanoSIMS imaging of extracellular electron transport processes during microbial iron(III) reduction

ABSTRACT Microbial iron(III) reduction can have a profound effect on the fate of contaminants in natural and engineered environments. Different mechanisms of extracellular electron transport are used by Geobacter and Shewanella spp. to reduce insoluble Fe(III) minerals. Here we prepared a thin film of iron(III)-(oxyhydr)oxide doped with arsenic, and allowed the mineral coating to be colonised by Geobacter sulfurreducens or Shewanella ANA3 labelled with 13C from organic electron donors. This preserved the spatial relationship between metabolically active Fe(III)-reducing bacteria and the iron(III)-(oxyhydr)oxide that they were respiring. NanoSIMS imaging showed cells of G. sulfurreducens were co-located with the iron(III)-(oxyhydr)oxide surface and were significantly more 13C-enriched compared to cells located away from the mineral, consistent with Geobacter species requiring direct contact with an extracellular electron acceptor to support growth. There was no such intimate relationship between 13C-enriched S. ANA3 and the iron(III)-(oxyhydr)oxide surface, consistent with Shewanella species being able to reduce Fe(III) indirectly using a secreted endogenous mediator. Some differences were observed in the amount of As relative to Fe in the local environment of G. sulfurreducens compared to the bulk mineral, highlighting the usefulness of this type of analysis for probing interactions between microbial cells and Fe-trace metal distributions in biogeochemical experiments.


INTRODUCTION
The reduction of Fe(III) in minerals can have profound consequences on the mineral composition and aqueous geochemistry of the subsurface environment. One of the outcomes of the bioreduction of ferric minerals, such as ferrihydrite, is the reductive dissolution of Fe(III), which forms soluble Fe(II) and is likely to cause the release of other incorporated or sorbed elements such as arsenic (Coker et al. 2008;Cutting et al. 2009). Soluble Fe(II) can in turn recrystallise to form new Fe(II)-bearing minerals such as magnetite, siderite or vivianite depending on the pH, temperature and the other geochemical species present (Hansel et al. 2003). These processes can also affect the mobility of arsenic and other toxic metals and radionuclides (Lloyd et al. 2000;Smedley and Kinniburgh 2002;Newsome, Morris and Lloyd 2014;Watts et al. 2015). Focusing on the release of arsenic to groundwaters as a high profile example, this is likely caused by the development of reducing biogeochemical conditions linked to subsurface microbial activities (Islam et al. 2004a,b). However, As can also be re-sequestered by newly forming minerals such as vivianite and siderite (Islam et al. 2005;Muehe et al. 2016), and magnetite also was found to incorporate As(V) and create surface complexes with As(III) (Islam et al. 2005;Coker et al. 2006;Wang et al. 2014). Such interactions with As during the mineralisation of Fe(II)/Fe(III) minerals could potentially alter the distribution of As on the micro scale and delay its release into aquifers used to supply drinking waters on a macro scale. It is therefore crucial to understand the mechanism of microbial Fe(III) reduction in order to assess its impact on the release (and capture) of metals and metalloids. Given that the contamination of drinking water with arsenic is a major health concern that threatens the lives of more than 140 million people worldwide (Singh et al. 2015), this area warrants particular attention. Fe(III) is an electron acceptor commonly used in anaerobic microbial respiration. A diverse range of bacteria and archaea covering a wide range of phylogenetic groups has been identified that enzymatically reduce Fe(III) (Lovley, Holmes and Nevin 2004). Fe(III) mostly exists in the natural environment in the solid form as oxides such as ferrihydrite, goethite and haematite, except in strongly acidic conditions where it is soluble. This means that unlike the more energetically favourable electron acceptors oxygen and nitrate, insoluble Fe(III) is unable to diffuse or be transported into the cell to be respired. As the lipidrich cell membrane is impervious to electrons, Fe(III)-reducing microorganisms have had to evolve mechanisms for extracellular electron transfer from the surface of the cell, in order to be able to generate a proton motive force and hence conserve energy for growth. Most of the studies that underpin our knowledge of extracellular electron transport have been performed on two model organisms, the Gram-negative bacteria Geobacter sulfurreducens and Shewanella oneidensis MR1 (formerly Alteromonas putrefaciens), that both gain energy from reducing Fe(III) to Fe(II). These genera of bacteria were first shown to respire Fe(III) in the late 1980s (Lovley et al. 1987;Myers and Nealson 1988). Both Geobacter and Shewanella species use outer membrane porin-cytochrome complexes to transport electrons to the outer membrane, but the proteins involved are phylogenetically distinct and are considered to have evolved independently (Lovley, Holmes and Nevin 2004;White et al. 2016). They also use different mechanisms of transporting electrons from the surface of the outer membrane to the extracellular Fe(III) mineral electron acceptor, as described below.
Geobacter species are obligate anaerobes that couple the reduction of Fe(III) to the oxidation of acetate and other organic substrates (Lovley et al. 1987;Lovley and Phillips 1988). They are common in the Fe(III)-reducing zone of sediments (Lovley, Holmes and Nevin 2004), and have been found to dominate the subsurface microbial community following biostimulation of the subsurface by acetate additions during targeted uranium bioremediation programmes (Anderson et al. 2003;Wilkins et al. 2009;Kerkhof et al. 2011;Williams et al. 2011). The current consensus is that Geobacter species require direct physical contact with an insoluble electron acceptor in order to transfer electrons (Nevin and Lovley 2000;Lovley, Holmes and Nevin 2004), and this occurs via an electron transport chain that involves both ctype cytochromes and pili. Geobacter sulfurreducens uses multiple and parallel electron transport pathways from the inner membrane quinone pool to the OmcB cytochrome, which is partially exposed to the extracellular environment (Leang, Coppi and Lovley 2003;Malvankar et al. 2011;Liu et al. 2015;Levar et al. 2016). From OmcB the electrons are transported to OmcE, then to pili, and then to OmcS cytochromes that pass the electrons to the electron acceptor; without each of these cytochromes and pili, electron transport to a solid electron acceptor is compromised (Childers, Ciufo and Lovley 2002;Mehta et al. 2005;Reguera et al. 2005;Leang et al. 2010;Malvankar et al. 2011). Electron transport along the pili of Geobacter species has been proposed to occur via metallic-like conductivity (Malvankar et al. 2011;Lovley 2012;Malvankar, Tuominen and Lovley 2012;2017). An early study did suggest that G. sulfurreducens could produce extracellular cytochromes (Seeliger, Cord-Ruwisch and Schink 1998), but this was later proven to be incorrect (Lloyd, Blunt-Harris and Lovley 1999;Straub and Schink 2003). More recently it appears that a pili-deficient mutant was able to restore its ability to reduce electron acceptors by releasing an extracellular c-type cytochrome (PgcA), but the authors note that this adaptation is likely a result of the highly controlled laboratory conditions and is unlikely to occur in the environment (Smith et al. 2014).
Shewanella species are facultative anaerobes that can couple reduction of Fe(III) to the oxidation of organics such as lactate, but not acetate (Myers and Nealson 1988;Lovley, Phillips and Lonergan 1989;Lovley, Holmes and Nevin 2004;Brutinel and Gralnick 2012). They are widespread in the environment in sediments where organic matter is being degraded and are thought to be adapted to the oxic/anoxic transition zone (Fredrickson et al. 2008;Edwards et al. 2015). It is likely that electron transport in S. oneidensis occurs primarily via an electron transport chain that includes flavin-cytochrome complexes on the outer membrane. Transport of electrons from the inner membrane quinone pool to the outer membrane occurs via the MtrCAB porin-cytochrome complex, the MtrC cytochrome is partially exposed at the cell surface and from there the electrons can be transferred to the extracellular electron acceptor or to the outer membrane cytochrome OmcA (White et al. 2016). Long-range electron transport is possible in S. oneidensis by the extension of the outer membrane into micrometre long nanowires (Pirbadian et al. 2014), along which electron transport may occur via an electron hopping mechanism (Okamoto, Hashimoto and Nakamura 2012;White et al. 2016), or by direct electron transfer accompanied by intermediate diffusive events (Subramanian et al. 2017). As well as being able to transport electrons via direct contact with an electron acceptor, Shewanella species can also secrete soluble extracellular electron shuttles to support an indirect reduction mechanism (Nevin and Lovley 2000;Marsili et al. 2008;von Canstein et al. 2008). Indeed, flavin-mediated electron transport has been shown to be responsible for up to 75% of electronacceptor reduction in S. oneidensis (Kotloski and Gralnick 2013). The secreted flavins can also form flavo-cytochrome complexes with MtrC under anaerobic conditions (Okamoto et al. 2014;Edwards et al. 2015), and it has been argued that these bound complexes increase the rate of extracellular electron transport at the cell surface, rather than through electron shuttling via the free soluble flavin shuttles (Xu, Jangir and El-Naggar 2016). An in vivo study with Mn(IV)-oxides found that many S. oneidensis cells were planktonic and made occasional contact with the mineral surface, which could suggest electron transport via direct contact was more important than the secretion of extracellular flavins (Harris, El-Naggar and Nealson 2012). Nevertheless the role of secreted flavins in extracellular metal reduction remains controversial. Although Shewanella species can secrete extracellular flavins to reduce electron acceptors that were inaccessible to cells (Lies et al. 2005;Jiang et al. 2010), the likelihood of this mechanism occurring outside carefully controlled laboratory experiments has been questioned (Lovley, Holmes and Nevin 2004;Smith et al. 2014). Potential loss of the flavin mediator to planktonic cultures is clearly a key challenge, but the role of extracellular flavins in mediating Fe(III) and Mn(IV) reduction in biofilms remains poorly constrained.
A range of metals and metalloids can be associated with Fe(III) minerals, and one of the most intensively studied is arsenic. Like iron, arsenic is also redox active, with As(V) sorbing more strongly to Fe(III) oxyhydroxides present in aquifers than As(III) (Islam et al. 2004a). Arsenic can be released to solution when bacteria reduce the Fe(III) oxyhydroxides hosting the arsenic, to aqueous Fe(II), or it can be released via microbial reduction of As(V) to As(III), via dissimilatory As(V) reduction that is mediated by a periplasmic arsenate reductase encoded by the arr operon (Saltikov and Newman 2003;. Although certain species of Geobacter and Shewanella are able to respire As(V), such as G. uraniireducens and S. ANA3 (Saltikov, Wildman and Newman 2005;Giloteaux et al. 2013), neither G.sulfurreducens or S.oneidensis MR1 are capable of catalysing this process as they do not possess the arr operon (Islam et al. 2005;Lloyd et al. 2011;Jiang et al. 2013), although both strains have been reported to have putative arsenic resistance genes (which can reduce As(V) within the cytoplasm prior to efflux of the As(III) as part of detoxification process) (Islam et al. 2005;Jiang et al. 2009;Dang et al. 2017).
Most previous studies have focussed on the bulk chemical effects of microbial activity, but do not provide the critical spatial information on the effect of microbial activity that is essential for determining mechanistic information. NanoSIMS is a surface based secondary ion mass spectrometry (SIMS) technique that is able to image up to seven masses and achieve a spatial resolution down to 50 nm (Herrmann et al. 2007b). It can be used in conjunction with stable isotope labelling (for example 13 C and 15 N) and is therefore particularly useful for imaging metabolically active cells that have incorporated labelled substrates. NanoSIMS, often in combination with stable isotope labelling, has been used previously to study a range of microorganisms and microbial processes, including benzene degrading communities (Schurig et al. 2015), periplasmic encrustation in nitratereducing Fe(II)-oxidising bacteria (Miot et al. 2015), anaerobic phototrophy (Musat et al. 2008), nitrogen fixation in cyanobacteria (Ploug et al. 2010;Woebken et al. 2012) and more recently syntrophic interactions between species (Wu et al. 2013;Gieg, Fowler and Berdugo-Clavijo 2014;Green-Saxena et al. 2014), as well as the location of active microorganisms in soil (Herrmann et al. 2007a) and the nature of organic matter in soil microenvironments (Vogel et al. 2014).
The aim of this study was to observe the colonisation of an As(V)-doped Fe(III)-(oxyhydr)oxide surface by different model Fe(III)-reducing bacteria. Geobacter sulfurreducens was selected as a model Fe(III)-reducing bacterium that requires direct contact with an extracellular electron acceptor. In contrast, Shewanella ANA3 was selected as a model Fe(III)-reducing bacterium that does not require direct contact with an extracellular electron acceptor. The cells were supplied with 13 C-labelled substrates to enable the localisation of metabolically active cells in relation to the mineral surface by NanoSIMS, and therefore explore mechanisms of iron reduction. Finally, the role of microbial Fe(III) reduction (and enzymatic As(V) reduction) in controlling the fate of As(V) incorporated within Fe(III)(oxyhydr)oxide was also studied using these two organisms. As G. sulfurreducens is unable to reduce As(V), in this experiment the As(V) served as a tracer to allow observations of the effect of Fe(III)-reduction in the local environment of the cells. This contrasts with Shewanella ANA3, which is also able to reduce As(V) enzymatically.

Thin film preparation
To enable visualisation of cell-mineral interfaces by microscopy with limited disturbances, supported films of Fe(III) minerals were prepared. These thin films were made using synthetic ferrihydrite, a model amorphous iron(III)-(oxyhydr)oxide mineral relevant to environmental processes and known to be bioavailable for Fe(III)-reducers. Ferrihydrite was synthesised (Schwertmann and Cornell 2000;Cornell and Schwertmann 2006) with As incorporated via the addition of Na 2 HAsO 4 to achieve a 12% mol/mol As/Fe ratio. All reagents were analytical grade. The ferrihydrite-As suspension was added to glass slides (for optical microscopy) or boron doped silicon wafers (for SEM and NanoSIMS) by dropping a suspension onto each wafer and leaving it to dry in air.

Cell culturing
Geobacter sulfurreducens was obtained from the laboratory culture collection of the University of Manchester Geomicrobiology group, and grown anaerobically in a modified freshwater enrichment medium (after Lovley and Phillips 1988) at pH 7, with 15 mM acetate as the electron donor and 40 mM fumarate as the electron acceptor. Cultures were grown in the dark at 30 • C until the late-logarithmic phase, when they were harvested by centrifugation (5000g, 20 min), washed twice in an anaerobic 30 mM bicarbonate buffer, then suspended into the experimental medium as appropriate.
Shewanella ANA3 was also obtained from our laboratory culture collection and grown aerobically on LB agar plates and then inoculated into an anaerobic modified minimal medium ) at pH 7 with 20 mM lactate as the electron donor and 40 mM sodium fumarate as the electron acceptor. Cultures were grown in the dark at 30 • C until the late-logarithmic phase, when they were harvested by centrifugation (5000g, 20 min), washed twice in an anaerobic 30 mM bicarbonate buffer, then suspended into the experimental medium as appropriate.

Thin film colonisation preparatory experiments
To investigate the colonisation of Fe(III) mineral coatings and associated As mobilisation by a model Fe(III)-reducing bacterium, experiments were set up comprising a glass slide coated with the iron-(oxyhydr)oxide mineral held vertically in a sterile glass serum bottle with the headspace degassed with N 2 . This was submerged in a modified anaerobic freshwater minimal medium (after Lovley et al. 1991) at pH 7 containing 30 mM bicarbonate, 4.3 mM phosphate and with 10 mM acetate as the electron donor and the Fe(III) thin film coating as the sole electron acceptor. Cells of G. sulfurreducens were added to achieve an optical density OD 600 of 0.3. The bottles were incubated in the dark at 30 • C. To assess the rate of Fe(III)-reduction and cell colonisation, samples of the medium were obtained using a degassed needle and syringe and monitored for OD 600 . At selected time points the glass slides were harvested and either analysed for cell colonisation by staining with DAPI and counting via optical microscopy (Zeiss Axio Imager A1), or for Fe(III)-reduction by dissolving in 0.5 N HCl to assess Fe(II) content via the Ferrozine assay and then 0.5 N hydroxylamine-HCl for total bioavailable Fe (Lovley and Phillips 1986).

Thin film colonisation experiments for NanoSIMS
The methodology described above for the preparatory experiments was used to prepare samples for NanoSIMS analysis, except that conducting Si wafers were used instead of glass slides and for G. sulfurreducens 10 mM 13 C-acetate (99% CH 3 13 CO 2 Na, ISOTEC INC) or for S. ANA3 10 mM 13 C-lactate (99% 13 CH 3 CH(OH)CO 2 Na, Sigma Aldrich) were used as electron donors to label the metabolically active cells. Cells of G. sulfurreducens were added to an OD 600 of 0.3 and S. ANA3 to an OD 600 of 0.5. Negative controls contained no added electron donor, and controls with no bacteria (in triplicate) were used to assess the abiotic solubility of Fe and As under the experimental conditions used. Changes in aqueous geochemistry were monitored at selected time points by removing aliquots of the medium using a degassed needle and syringe, taking care not to disturb the wafer. Aliquots were diluted in 2% HNO 3 and monitored for Fe and As by ICP-AES (Perkin-Elmer Optima 5300 DV).
The wafers were harvested after 11 days (as this incubation time was found to be most suitable for obtaining optimal cell colonisation by DAPI staining and cell counting) and one replicate wafer was preserved for SEM and NanoSIMS analysis. The remaining aqueous phase was diluted in 2% HNO 3 and monitored for Fe and As by ICP-AES, or diluted in deionised water and analysed for As(V) and As(III) by ion chromatography inductively coupled plasma-mass spectrometry (IC-ICP-MS) . The coatings on the remaining two wafers were dissolved in concentrated HNO 3 , and the digest was diluted in deionised water and monitored for total Fe and As content via ICP-AES in order to calculate a mass balance.
To investigate whether flavins were secreted by S. ANA3, the supernatants of these experiments were analysed for the presence of flavins using HPLC with a fluorescence detector (Thermo Scientific Dionex BioLC, GP50 pump, Dionex Utimate 3000 fluorescence detector set at 450 nm (excitation) and 520 nm (emission), using a C18 column with a mobile phase of 40% v/v methanol at a flow rate of 1 mL min −1 ).

Preservation of samples for SEM and NanoSIMS
To preserve the cell structure for electron microscopy and NanoSIMS, and to maintain the stability of redox sensitive mineral phases, the wafer was removed from the experimental bottle in an anaerobic cabinet and placed immediately into an anoxic 2.5% glutaraldehyde solution in phosphate buffered saline (PBS in g/l NaCl 8, KCl 0.2, Na 2 HPO 4 1.42 and KH 2 PO 4 0.27) and left overnight (Schurig et al. 2013). The wafer was then placed into an anaerobic 1.5% glutaraldehyde PBS solution for 1 hour, then a 0.75% glutaraldehyde PBS solution for 1 hour, then dehydrated sequentially with 25%, 40%, 50%, 60%, 70%, 80%, 90% and 100% solutions of acetone or ethanol (Schurig et al. 2013), all under anoxic conditions. The wafer was dried in the anaerobic cabinet then immediately coated with gold using a sputter coater. Electron microscopy was performed after the NanoSIMS analysis using a Phillips XL30 ESEM-FEG operated in high vacuum conditions.

NanoSIMS analysis and data processing
The samples were analysed using a Cameca NanoSIMS 50L at the University of Manchester. A 16 keV Cs + primary ion beam with a current of 0.7-0.8pA was scanned over the sample surface to generate negative secondary ions. The instrument was aligned to detect 12 C − , 13 C − , 16 O − , 12 C 14 N − , 13 C 14 N − , 56 Fe 16 O − and 75 As − taking care to avoid isobaric interferences. Prior to imaging, each area was implanted with a high current, defocused beam to remove the gold coating and achieve a dose of 1 × 10 17 Cs + ions cm −2 . The areas for imaging were selected by identifying flat areas of mineral using the ion induced secondary electron image prior to imaging, in order to avoid topographical effects. A dwell time of 5000 μs/pixel was used with an aperture size of 200 μm (D1 = 3) and multiple image planes were acquired to increase the number of counts.
The OpenMIMS plugin (National Resource for Imaging Mass Spectrometry, Harvard, US) for ImageJ (Schneider, Rasband and Eliceiri 2012) was used to correct each image for dead time (44 ns) and then regions of interest (ROIs) were defined by hand using the summed 12 C 14 N images to identify the boundary of every cell (0.41 μm 2 ± 0.15, N = 1873). An example showing the cell ROIs for an area of the G. sulfurreducens sample is provided as Fig.  S1 (Supporting Information). Areas of the sample that appeared particularly bright in the secondary electron images, such as along the edges of flat areas of mineral, were considered likely to have a significant contribution to the signal from topography, therefore were not included in ROIs. The cell ROIs were manually assigned to groups based on whether the cells were located on the iron-(oxyhydr)oxide mineral surface 'cells on mineral', next to the mineral 'cells adjoining mineral' or located on the wafer background where no mineral was present 'cells on wafer'. ROIs were also defined where no cells were present on areas of iron-(oxyhydr)oxide mineral 'mineral' and the background of the wafer 'wafer background'.
Average counts per pixel of 12 C, 13 C, 16 O, 12 C 14 N, 13 C 14 N, 56 Fe 16 O and 75 As were extracted for each ROI for the G. sulfurreducens experiment, and 12 C, 13 C, 16 O, 12 C 14 N, 28 Si, 56 Fe 16 O and 75 As for each ROI for the S. ANA3 experiment. All values were tabulated in Microsoft Excel and normalised for the total dose implanted during imaging (Table S1, Supporting Information). Normalising by the dose implanted during imaging accounts for differences in signal intensity that may result from differences in primary beam current, number of image planes and area. For G. sulfurreducens the 13 C enrichment was calculated by taking a ratio of 12 C 14 N and 13 C 14 N to identify cells that had become enriched in 13 C compared to natural background; the 12 C 14 N and 13 C 14 N signals were used as these gave higher counts and brighter images with better contrast compared to using 13 C and 12 C signals. As 13 C 14 N was not measured for the S. ANA3 experiment, 13 C enrichment was calculated as the ratio of 13 C to 12 C. The ratio of 75 As relative to 56 Fe 16 O was calculated to indicate where As may be relatively enriched or depleted compared to Fe. Box and whisker plots were generated to visualise the large dataset, with boxes representing the interquartile range and whiskers the 5th and 95th percentile.
It should be noted that after analysing the G. sulfurreducens sample, an interference from 56 Fe 19 F − was observed in the region of mass 75. Fluorine was not present in the experimental set up, and there are no known biological interactions between G. sulfurreducens and fluorine. However, fluorine is usually present in the analysis chamber of NanoSIMS. Although the detector was positioned on the centre of the peak of the As standard it is possible that both 56 Fe 19 F − and 75 As − contributed to the signal at this mass for the G. sulfurreducens sample. Given the source of the fluorine was the analysis chamber, it would be expected that the contribution from 56 Fe 19 F − would be homogeneous across the surface of the iron-(oxyhydr)oxide mineral and therefore any changes in the local environment of the cells would be caused by microbially mediated processes. Hereafter we refer to the mass 75 signal as 75 As but we note that some of the counts are likely to have been from 56 Fe 19 F − as well as 75 As. After the NanoSIMS analysis, wavelength dispersive scanning was performed to measure the K a edges using an electron microprobe (JEOL JXA-8530F FEG-EPMA). The results showed conclusively that As was present but there was no F in the sample. The S. ANA sample was analysed after the identification of this interference and the detectors were aligned to avoid any issues with mass interference.

Colonisation of Fe(III) by Geobacter sulfurreducens
After 11 days incubation with G.sulfurreducens and electron donor, SEM imaging showed the wafer surface to be colonised by cells (Fig. 1a-c). Far fewer cells were visible on the wafer from the no electron donor control experiment (Fig.1d-e). The texture of the iron-(oxyhydr)oxide mineral in the no cell control was similar to that observed in the experiments with cells (Fig. 1f). Different cell morphologies were visible; some cells were present as well preserved individuals (Fig. 1a-b) while some appeared to be coated in extracellular filamentous material, which may have been involved in biofilm formation (Fig. 1c).
Geochemical measurements confirmed that soluble iron was released to solution during the 11 day experiment, presumably as Fe(II), given that Fe(III) is very poorly soluble under these experimental conditions (Fig. 2). After 11 days, up to 10% of the iron remaining in the thin film was present as Fe(II) as measured by the Ferrozine assay, after extraction with 0.5 N HCl. Minimal amounts of iron were released to solution in the controls, confirming that microbial Fe(III)-reduction was required for its release to solution.
NanoSIMS analysis was used to map the distribution of metabolically active cells in relation to the mineral substrate. Nine areas of the samples were analysed (Fig. S2, Supporting Information). Cell ROIs were assigned manually into groups based on whether the cells were located on the iron-(oxyhydr)oxide mineral surface 'cells on mineral' (N = 1174), next to the mineral 'cells adjoining mineral' (N = 117) or located on the wafer background where no mineral was present 'cells on wafer' (N = 582). ROIs were also defined where no cells were present on areas of iron-(oxyhydr)oxide mineral 'mineral' (0.25 ± 0.09 μm 2 , N = 861) and the background of the wafer 'wafer background' (9.55 ± 0.15 μm 2 , N = 53).
Cells enriched in 13 C were present in all of the areas analysed, with some enriched to more than 10 times the natural background levels. Artefacts associated with sample preparation, including dilution of 13 C from chemical fixation (Musat et al. 2014) and potential interference from 11 B 16 O − meant that it was not possible to measure absolute 13 C enrichment values. However, this was not necessary for this experiment as the 13 C was used to locate the metabolically active cells rather than measure absolute values of 13 C enrichment. To see where the enriched cells were located in relation to the mineral substrate, composite images were created using ImageJ for the iron-(oxyhydr)oxide mineral, 13 C 14 N and 12 C 14 N. These clearly showed that overall, the cells that were metabolically active and relatively enriched in 13 C tended to be co-located with the iron-(oxyhydr)oxide mineral, whereas the cells that were less enriched in 13 C tended to be located on the wafer background (Fig. 3a). There were some exceptions to this, such as in Area 6 where there was a cluster of cells that were highly enriched in 13 C to the left of the iron-(oxyhydr)oxide mineral, which may have been redistributed during fixation.
Box and whisker plots were made of all cell ROIs to further display the distribution of 13 C enrichment (Fig. 3b). The results showed that cells located on or adjoining the iron-(oxyhydr)oxide mineral were significantly more enriched in 13 C (P < 0.001) and therefore metabolically active, compared to cells located on the wafer background away from the iron-(oxyhydr)oxide mineral. The 13 C enrichment values for the wafer background and the mineral surface were slightly higher than natural background; this is likely due to sorption of some of the 13 C-acetate. Cells on the wafer were not significantly different in 13 C enrichment compared to the wafer background (P = 0.89).

Colonisation of Fe(III) by Shewanella ANA3
After 11 days incubation, SEM imaging confirmed that the wafer surface was colonised by cells ( Fig. 4a-d); some were in contact with (adjoining) the iron-(oxyhydr)oxide coating while others were located on the wafer background. It is noteworthy that unlike with G. sulfurreducens, almost no S. ANA3 cells were located directly on the iron-(oxyhydr)oxide mineral surface. Very few cells could be observed in the no electron donor control experiment ( Fig. 4e-f). The texture of the iron-(oxyhydr)oxide mineral in the no cell control was similar to that observed in the experiments with cells (Fig. 1f). Again, geochemical measurements confirmed Fe(III) reduction occurred during the 11 day experiment (Fig. 2a). A greater quantity of Fe was released to solution with S. ANA3 compared to G. sulfurreducens (Fig. 2b), perhaps due to the higher cell density of the S. ANA3 experiment.
Again, NanoSIMS analysis was used to locate metabolically active cells in relation to the mineral. Three areas of the sample were analysed (Fig. S3, Supporting Information). Cell ROIs were manually assigned to the groups, 'cells adjoining mineral' (N = 11), and 'cells on wafer' (N = 43). No cells were visible on the mineral surface therefore there were no ROIs for this group. ROIs were also defined for 'mineral' (1.92 ± 0.96 μm 2 , N = 31) and the 'wafer background' (3.35 ± 3.17 μm 2 , N = 32).
Cells enriched in 13 C were present in all of the areas analysed (Fig. 5a-c), with some cells enriched up to seven times the natural background levels (Fig. 5d). The lack of cells directly located on the mineral surface suggests that the metabolic activity of S. ANA3 was not correlated with the presence of the iron-(oxyhydr)oxide mineral. The S. ANA3 cells located on the wafer background were highly enriched in 13 C, in stark contrast with the G. sulfurreducens cultures, where the equivalent cells were O in red, 13 C 14 N in green, 12 C 14 N blue). Metabolically active cells relatively enriched in 13 C (in green) tend to be co-located with the iron oxide mineral (red), while less enriched cells tend to be located on the wafer background. (B) Box and whisker plots of % 13 C calculated from 13 C 14 N and 12 C 14 N counts ( 13 C 14 N/ 12 C 14 N × 100), showing that cells located on or adjoining the iron oxide mineral were significantly more enriched in 13 C than cells located on the wafer background (P < 0.001, Tukey-Kramer test). The 13 C enrichment on the wafer background and mineral surface, were not significantly different (P > 0.99, Tukey-Kramer test), nor was the 13 C enrichment of the cells on the wafer compared to the wafer background (P = 0.89, Tukey-Kramer test). 'Nat' shows known value of naturally occurring 13 C of 1.11%. not significantly enriched with 13 C compared with the natural background. This clearly demonstrates the different mechanisms of electron transport to extracellular Fe(III) minerals used to support the metabolism of these two bacteria.

Use of NanoSIMS to infer mechanisms of extracellular electron transport
Our experimental set up was successful in supporting the growth of cells of G. sulfurreducens on a thin film of iron(III)-(oxyhydr)oxide. Cells located on or adjoining the mineral were significantly more enriched in 13 C and therefore metabolically active compared to those located on the wafer surface. This is consistent with G. sulfurreducens requiring direct contact with an extracellular electron acceptor (i.e. the Fe(III)-(oxyhydr)oxide mineral surface) to transfer the electrons required for respiration, via outer membrane cytochromes or conductive pili extending from the cells surface (Lovley 2012). We believe this to be the first time that this relationship has been visualised using 13 C labelling on an in situ sample.
In contrast, there were almost no cells of S. ANA3 located on the iron-(oxyhydr)oxide mineral surface, and the cells adjoining the mineral or on the wafer background were similarly enriched in 13 C. This is consistent with previous results showing that Shewanella species do not require direct contact with an extracellular electron acceptor for electron transport to occur (Lies et al. 2005). Instead, electron transport may occur via flavocytochrome complexes on the outer surface of the cells, which make intermittent contact with the mineral surfaces via chemotaxis (Harris et al. 2010;Harris, El-Naggar and Nealson 2012) or alternatively by secretion of extracellular flavins that acted as an electron shuttle (von Canstein et al. 2008). Compared to the no cell control where riboflavin was not detected, when S. ANA3 cells were present riboflavin concentrations between 60 O in red, 13 C in green, 12 C 14 N blue). No metabolically active cells relatively enriched in 13 C (in green) were co-located with the iron oxide mineral (red). (D) Box and whisker plots of % 13 C calculated from 13 C and 12 C counts ( 13 C/ 12 C × 100), showing that cells located on the wafer background or adjoining the iron oxide mineral were significantly more enriched in 13 C than the wafer background and mineral surface (P < 0.001 for all combinations, Tukey-Kramer test). The 13 C enrichment on the wafer background and mineral surface were not significantly different (P > 0.98, Tukey-Kramer test). 'Nat' shows known value of naturally occurring 13 C of 1.11%. and 91 nM were measured, similar to those previously observed for S. oneidensis MR1 (von Canstein et al. 2008), although unlike in this previous study no extracellular flavin mononucleotide was detected in this experiment. This suggests that the secretion of extracellular flavins by S. ANA3 may have contributed to electron transport in this system.

Impact of microbial Fe(III)-reduction on arsenic solubility and speciation
To further explore the impact of extracellular electron transfer on redox-active trace elements associated with Fe(III) minerals during bioreduction, the fate of the incorporated As(V) was determined. After 11 days of incubation, As was released to solution (Fig. 6a). Some arsenic was solubilised as As(V) in the no electron donor controls and the no cell control, which indicates that As was partially soluble under these experimental conditions. This may have been due to the experimental set up that comprised a relatively large volume of minimal medium to a small quantity of mineral, with a large surface area presented as a thin film. Slightly more arsenic was released to solution in the microbially active incubations suggesting that microbial Fe(III)reduction had somewhat enhanced the release of arsenic under these experimental conditions. Speciation of aqueous As (IC-ICP-MS) showed that 99.9% was present as As(V) with G. sulfurreducens, and 21% As(V) and 79% As(III) (± 1.6%) with S. ANA3, as expected given that G. sulfurreducens lacks the operon that encodes for respiratory As(V) reduction. It is noteworthy that after incubation with the model As(V) reducer S. ANA3, the majority of As had been reduced to As(III), which suggests that at least some of the aqueous-phase As(V) that was released from the mineral abiotically must have been respired, as well as the As(V) that was likely released during microbial Fe(III)-reduction. Previous work has also noted the codependency of the rate of As(V) release from ferrihydrite and the rate of formation of As(III) by Shewanella species (Revesz, Fortin and Paktunc 2016).
The amount of iron and arsenic present on the wafers after 11 days was measured and compared to the amount that was solubilised to assess whether arsenic had been preferentially released from the mineral matrix to solution compared to Fe(II). Analyses were performed on the two replicate wafers that were At the end of the experiment there was more As in solution in the microbially active incubations. After subtracting the no cell control data there was 30 ± 18 nmol of As in solution after incubation with G. sulfurreducens and 19 ± 17 nmol of As after incubation with S. ANA3. G. sulfurreducens is shown with circle symbols (• average 4 replicates and • no electron donor control), S. ANA3 with triangles ( average 3 replicates and no electron donor control), and the no cell control with squares ( ). Controls are shown with dotted lines and open symbols. Error bars are ± 1 standard deviation. (B) Bulk geochemical monitoring of As on duplicate wafers after 11 days incubation with G. sulfurreducens or S.ANA3 supplied with electron donor compared to a no cell control. Annotations represent the % As in solution.
not preserved for SEM and NanoSIMS, and the data were calculated as nanomoles of Fe or As per experiment (Fig. 6b). These bulk measurements were used to estimate that post-incubation with G. sulfurreducens, there was 3.4 ± 0.1 nmol As per 100 nmol Fe present on the wafer, compared to 4.4 ± 0.4 nmol As per 100 nmol Fe with S. ANA3 and 5.3 ± 0.03 nmol As per 100 nmol Fe in the no cell control. All values were substantially lower than the 12 mol/mol starting value, showing that more than 50% of the As (relative to Fe) in the thin films was solubilised abiotically under these experimental conditions, also reflected in the aqueous geochemistry results (Fig. 6a). The values for the G. sulfurreducens wafers were significantly different to the no cell controls at the 95 % confidence level, confirming that the presence of metabolising G. sulfurreducens cells led to more As being released to solution relative to Fe, over and above the abiotic As solubilisation. The values for the S. ANA3 wafers were not significantly different to the no cell control wafers or the G. sulfurreducens wafers at the 95 % confidence level, but this was affected by the higher Figure 7. Box and whisker plot of 75 As counts as a percentage of 56 Fe 16 O counts illustrated for the entire G. sulfurreducens dataset. The proportion of 75 As relative to 56 Fe 16 O was significantly different in the local environment of the cells compared to the un-colonised mineral surface (P < 0.001, one way ANOVA). Similar results were observed when each individual area of the sample was analysed separately, see Figure S4(Supporting Information). variation in the duplicate S. ANA3 wafer measurements, suggesting more replicates would be required to conclusively show whether or not the presence of S. ANA3 caused increased mobilisation of As from the wafer relative to Fe.

Using the relationship between As and Fe to observe the effect of microbial Fe(III)-reduction by Geobacter sulfurreducens
Data were extracted from the NanoSIMS analysis (Fig. S2, Supporting Information) to assess whether the distribution of iron and arsenic was different in the local environment of the cells, and whether this was different to the bulk geochemical measurements. Given that S. ANA3 was not co-located with the mineral surface, the G. sulfurreducens data set was used for this analysis, which aimed to determine if NanoSIMS could be used to probe the direct impact of microbial metabolism on trace metals associated with Fe(III)-(oxyhydr)oxides.
The ratio of 75 As to 56 Fe 16 O was calculated for each individual cell ROI and for the ROIs on the mineral surface where no cells were present to assess the impact of the cells on the As/Fe distribution of the mineral (Fig. 7). The box and whisker plots clearly demonstrate that the distribution of 75 As relative to 56 Fe 16 O was significantly different in the local environment of the G. sulfurreducens cells compared to the mineral surface where no cells were present. There was some heterogeneity between individual areas of the sample (Fig. S4, Supporting Information), highlighting the importance of making multiple measurements even within the same sample.
To hypothesise the processes that could lead to differences in the relative amount of As in the local environment of G. sulfurreducens, we reason that the inoculated cells would first become associated with the wafer surface through diffusion, electrostatic attraction or chemotaxis. The cells that had direct contact with the iron(III)-(oxyhydr)oxide mineral would then begin to respire the Fe(III) coupled to the use of 13 C-acetate. This would lead to the release of aqueous Fe(II) and the consequent release of As(V) to solution, over and above what occurred in the no cell and no electron donor controls, probably due to the microbiallyinduced dissolution of the Fe(III) mineral via reduction and solubilisation as Fe(II). Here, in the local environment of the cells, there was proportionally more As present compared to where no cells were located, showing the preferential release of Fe(II) from the mineral, compared to As(V). Although the 75 As relative to 56 Fe 16 O results from the NanoSIMS analysis were supported by similar measurements from bulk geochemical monitoring, care should be exercised when interpreting these findings as the ionisation yields for 75 As − and 56 Fe 16 O − are not the same and will be affected by matrix effects.

Use of NanoSIMS to assess the impact of microbial metabolism on bulk and trace elements
Using NanoSIMS was an effective way to visualise the relationship between active Fe(III)-reducing bacteria and a solid Fe(III) electron acceptor substrate. The contrasting results for bacteria that use different mechanisms of extracellular electron transport illustrates the requirement for direct contact with the substrate and consequent mineral attachment with Geobacter sp., compared to the ability of Shewanella sp. to conduct extracellular electron transport without attaching to the mineral surface. This technique could be useful to study a wide range of Fe(III)reducing microorganisms, alongside biochemical and genetic analyses to fully understand the mechanisms of electron flow to Fe(III) and other insoluble electron acceptors.
Given that the environmental behaviour of many trace elements is controlled by Fe/Mn biogeochemistry, using NanoSIMS could provide new insights into how their fate could be controlled by microbial processes. Our findings using arsenic as an example of a trace element show that there appears to be differences in the proportion of arsenic relative to iron in the local environment surrounding metabolically active cells, compared to bulk geochemical measurements. Further work needs to be done to investigate this effect. It is likely that S. ANA3 reduced the bulk of As(V) to As(III) in the aqueous phase, and again additional work will be required to confirm this. In aquifer systems, the ability to respire As(V) that is mobilised during the reductive dissolution of Fe(III)-(oxyhydr)oxides, will result in the formation of poorly sorbing As(III), which will accumulate in groundwaters once available sorption sites in the aquifer have been saturated (Gnanaprakasam et al. 2017). This work opens the possibility for more detailed studies on Fe-As coupling using model systems and mixed community studies, using synthetic minerals and also aquifer materials. This type of analysis could be highly useful for probing Fe-trace metal distributions when studying a broader range of coupled processes, for example, Mn/Fe/S systems, and their impact on biogeochemical cycling of toxic metals and radionuclides such as Se, I and U.

SUPPLEMENTARY DATA
Supplementary data are available at FEMSEC online and the full dataset is available at Mendeley data (http://dx.doi.org/10.1763 2/gppn7stysx.1).