Elucidating heterogeneous iron biomineralization patterns in a denitrifying As(iii)-oxidizing bacterium: implications for arsenic immobilization

Anaerobic nitrate-dependent iron(ii) oxidation is a process common to many bacterial species, which promotes the formation of Fe(iii) minerals that can influence the fate of soil and groundwater pollutants, such as arsenic. Herein, we investigated simultaneous nitrate-dependent Fe(ii) and As(iii) oxidation by Acidovorax sp. strain ST3 with the aim of studying the Fe biominerals formed, their As immobilization capabilities and the metabolic effect on cells. X-ray powder diffraction (XRD) and scanning transmission electron microscopy (STEM) nanodiffraction were applied for biomineral characterization in bulk and at the nanoscale, respectively. NanoSIMS (nanoscale secondary ion mass spectrometry) was used to map the intra and extracellular As and Fe distribution at the single-cell level and to trace metabolically active cells, by incorporation of a 13C-labeled substrate (acetate). Metabolic heterogeneity among bacterial cells was detected, with periplasmic Fe mineral encrustation deleterious to cell metabolism. Interestingly, Fe and As were not co-localized in all cells, indicating delocalized sites of As(iii) and Fe(ii) oxidation. The Fe(iii) minerals lepidocrocite and goethite were identified in XRD, although only lepidocrocite was identified via STEM nanodiffraction. Extracellular amorphous nanoparticles were formed earlier and retained more As(iii/v) than crystalline “flakes” of lepidocrocite, indicating that longer incubation periods promote the formation of more crystalline minerals with lower As retention capabilities. Thus, the addition of nitrate promotes Fe(ii) oxidation and formation of Fe(iii) biominerals by ST3 cells which retain As(iii/v), and although this process was metabolically detrimental to some cells, it warrants further examination as a viable mechanism for As removal in anoxic environments by biostimulation with nitrate.


S2
Text S1. Supplementary materials and methods Text S1.1 Planktonic sample preparation Planktonic samples were collected at days 4 and 7 for STEM and on days 7 and 10 for XRD analysis.
Samples (1 mL) were centrifuged (2509 g/30 min) and washed twice in N2-degassed deionised water (dH2O). For STEM samples the washed precipitates were re-suspended in dH2O to an OD600≈0.1, this dilution (1.5 µL) was placed onto lacey carbon films on Cu TEM grids (Agar Scientific®) and further dehydrated using a turbo pump. For XRD analysis the concentrated precipitates (2 mL) were spread on glass slides (1 x 0.5 cm) and dried overnight in an anoxic cabinet. All these steps, including sample handling and loading into the instruments, were performed under anoxic conditions.

Text S1.2 STEM anoxic sample loading
Sample transportation to the TEM laboratory was done inside an airtight container in an anoxically sealed plastic bag. The TEM grids loading onto the instrument sample holder was done using a plastic box (~70 x 40 x 40 cm) filled with argon at the bottom layer, and the transfer of the TEM grid from the portable sample holder to the TEM instrument sample holder was done below the height of the argon layer filling inside the plastic box. This sample manipulation was challenging but with the aim to keep the anoxic conditions for as long as possible, although the presence of some ppm of O2 exposure during specimen loading cannot be ruled out.

Text S1.3 NanoSIMS analysis details
In NanoSIMS analysis the elemental ratio 13 C/ 12 C was used to identify cells that were metabolically active; 13 C 14 N/ 12 C 14 N was not chosen due to the unresolvable isobaric interferences 10 B 16 O and 11 B 16 O. The molecular ions 56 Fe 12 Cand 56 Fe 16 Owere used to monitor 56 Fe because 56 Fehas a low ionisation yield under the Cs + ion beam 1 .

Text S1.4 NanoSIMS depth profiles
Single cells were depth profiled using both primary ion beams in NanoSIMS. Isolated cells on the Si wafer were selected for this, and these cells were not implanted to reach steady state, to keep the cell mineralization coating intact. However, a quick implantation of Cs + or Oions (D1=1, 30 s) was done before the start of analysis to locate and focus single cells. In NanoSIMS, 56 Fe has a higher secondary ion yield under Obombardment than under Cs + ions bombardment 1 , where 56 Fe is normally collected as the molecular ion 56 Fe 16 O -, which enhances detection. Therefore, 56 Fe + S3 produced higher count intensities, typically one order of magnitude higher than 56 Fe 16 O -(Figures 3C,   3H, S5 C & S5 H). This is the opposite for 75 As, because 75 As has a lower ion yield under the Obeam   than under Cs + bombardment (Figures 3D, 3I, S5 D & S5 I). In these samples the Cs + beam was useful to infer metabolic activity by 13 C accumulation and to map As and Fe, while the Obeam was particularly useful to map Fe + , because it produced a stronger signal.
Text S1.4.1 NanoSIMS depth profiles using Cs + primary ions During Cs + depth profiling D1 apertures were reduced to 4 or 5, reducing the total current. Pixel sizes were 128 x 128 or 256 x 256, and dwell time varied from 5,000-20,000 µs px -1 . The number of planes collected were in the range of 50-170, and the scanning was stopped when the 12 C or 12 C 14 N signal disappeared, indicating that the bacterial cell had been entirely sputtered away and chemical information could be recovered from the whole cell.
Images were collected at a dwell time of 5000 µs px -1 . The positive secondary ions simultaneously detected were: 23 Na, 24 Mg, 28 Si, 39 K, 44 Ca, 56 Fe and 75 As. CAMECA MRP was improved to ≈3000 (ES=3 AS=2) to separate the mass interferences 12 C 16 O + at mass 28 and 23 Na 16 O + at mass 39. Mass 75 had no isobaric interferences ( 56 Fe 19 F + was not formed). In this case the disappearance of the 56 Fe + signal was monitored as an indicator of complete cell sputtering. The ionic salts (Na, Mg, K and Ca) were only used as a reference of cell sputtering and their mobility during the chemical fixation process is acknowledged 2, 3 .
Text S1.4.3 NanoSIMS data analysis L'image software was used to obtain stack NanoSIMS images (Larry Nittler, Carnegie Institution of Washington). ImageJ (https://imagej.nih.gov/ij/) with the plugin OpenMIMS (MIMS, Harvard University; www.nrims.harvard.edu) was used to create the hue saturation images (HSI) isotope ratio maps and to generate colour merge (overlay) images. Regions of interest (ROIs) were manually drawn around cells, the ion counts were normalized to primary ion doses.
Text S1.4.4 NanoSIMS single-cell 3D reconstructions 3D reconstructions of the depth profiles were created using the Thermo Scientific™ Avizo™ Software 9.7.0. Stack data of the negative secondary ions 12 C -, 56 Fe 16 Oand 75 As -, and the positive secondary S4 ions 23 Na + , 56 Fe + , and 75 As + , were first extracted with ImageJ and saved in the ".raw" format file.
Afterwards, these .raw files were loaded into Avizo™. The Z depth was compressed to 15-20 % to reduce the space between planes (generated by the software), bringing the volume of 3D visualisation to scale. The 12 Cand 23 Na + signals were used to generate the bacterial surface by smoothing (averaging) the signal over 2-3 pixels. The commands "generate surface" and "show surface" were used sequentially, and the "transparent" display with a transparency of 80 % was selected. 56 Fe + and 56 Fe 16 Osignals were the proxies of cellular Fe encrustation, smoothed to 2 pixels and displayed as "shaded". The 75 As +/ion counts were smoothed to 1 pixel, because of the lower ion counts, and displayed as "points" to enable their visualisation.

Text S2.1 Sample preservation effect imaged in SEM
Sample preservation is often overlooked but it is a key determining factor for accurate image analyses, including electron microscopy and SIMS techniques. For this reason, two preservation techniques were used (air-drying and chemical fixation-dehydration) for the biofilm samples after 4 and 7 days of incubation, and these were imaged in SEM to compare the effect on cells and biominerals. Abundant biomass colonising the Si wafer was observed in the samples by day 7, regardless of the preparation method used ( Figure S4 A & E), suggesting that both methods preserved biomass density, although substantial surface colonization was already present by day 4   Figure S1. Pictures of the experimental bottles with Acidovorax sp. strain ST3 cells after 1 and 7 days of incubation. (A) planktonic growth samples at day 1, (B) planktonic growth samples at day 7 and (C) biofilm growth sample at day 7. In (B) the bottle on the left is the no cells control, showing the thin red layer that formed abiotically at the surface in some bottles. After more than 1 month of incubation, some precipitates started to appear in the "no cells" control bottles.

No cells control
Acidovorax sp. total As by ICP-AES. Notice the abiotic removal of As from solution, probably through sorption to abiotically formed precipitates or minerals. Error bars are the standard deviation, N=3. There was no significant difference in aqueous As(III) between the samples and the no cells control, only in aqueous As(V) and total As. S9 Figure S3. Acetate and nitrate quantification in the aqueous phase of strain ST3 grown in biofilm and planktonic conditions (as well as no cell control). Error bars are the standard deviation, N=3. It is worth noting that even though nitrate was added in excess, its consumption was lower than what was stoichiometrically expected, as there were two suitable electron donors in the medium, acetate and arsenite. S10 Figure S4. As -S13 Figure S7. Diffractograms of the bulk precipitates of Acidovorax sp. strain ST3 at days 7 (anaerobic XRD analysis) and 10 of incubation (aerobic and anaerobic analysis). Aerobic analysis was collected without the air-tight dome, which produced higher intensity peaks. At day 10 lepidocrocite (L) and goethite (G) were the main mineral phases detected, more clearly noticed in the oxygen-exposed analysis, although peak matching indicates the presence of hematite, magnetite and vivianite. show O and Fe are the most abundant elements in both ROIs, whereas the As peak varied in intensity; this intensity was higher in ROI 1 and so was its abundance. The L3/L2 intensity ratio was used to estimate Fe 3+ abundance and the EELS spectra (P & Q) show the L3 and L2 peaks.