Remedial Treatment of Corroded Iron Objects by Environmental Aeromonas Isolates

Microbiology can greatly help in the quest for a sustainable solution to the problem of iron corrosion, which causes important economic losses in a wide range of fields, including the protection of cultural heritage and building materials. Using bacteria to transform reactive and unstable corrosion products into more-stable compounds represents a promising approach. The overall aim of this study was to develop a method for the conservation and restoration of corroded iron items, starting from the isolation of iron-reducing bacteria from natural environments. This resulted in the identification of a suitable candidate (Aeromonas sp. strain CU5) that mediates the formation of desirable minerals at the surfaces of the objects. This led to the proof of concept of an application method on real objects.

I ron corrosion represents a major problem in man-made ecosystems and causes large economic losses in fields as diverse as water supply, food industries, transport, and cultural heritage (1). While the first three fields are obviously of worldwide importance, the same is the case for the latter, as cultural heritage is a driver of tourism and sustainable development (2). The damage generated by uncontrolled corrosion of archeological objects is a clear example of the negative impact of iron corrosion. Reactive iron oxides and oxyhydroxides are the principal unstable iron corrosion products reported for these objects. Due to their molar volume (higher than that of native iron or iron alloys), these corrosion products generate cracks and irreversible damage of the objects if no intervention occurs (3).
In nature, the phenomenon of iron corrosion is an essential part of the biogeochemical cycling of iron. While aerobic corrosion is considered to be primarily a chemical process, anaerobic corrosion is often related to oxidative and/or reductive bacterial activity (1). Iron is a redox-sensitive element. It oxidizes easily in the presence of oxygen to form insoluble Fe(III) oxides. Iron oxides accumulate in sediments and soils (4), where iron can be remobilized through biotic and abiotic reactions (5,6). Some bacteria are able to catalyze transformations of iron minerals under oxic and anoxic conditions (4). These bacterial metabolisms can be exploited to transform reactive corrosion products (e.g., lepidocrocite) into stable compounds with low molar volumes (7). This represents a promising biotechnological approach to remediate the problem of iron corrosion.
In two recent studies (8,9), the feasibility of a biological treatment using bacteria to remediate iron corrosion has been demonstrated. The strategy consists of the reduction of reactive Fe(III) oxides and oxyhydroxides and the formation of biogenic minerals under anaerobic conditions. Initially, a strict anaerobe (Desulfitobacterium hafniense) was used to reduce reactive corrosion products during anaerobic growth. Although reductive dissolution was observed, this approach resulted in the formation of undesirable products. Indeed, in addition to bacterial iron reduction, abiotic reduction caused by the reductant in the medium also occurred and resulted in the formation of sulfur-containing Fe(II) minerals. In addition, the handling of a strict anaerobe and the medium used for reduction were not particularly suitable for application on real artifacts (8). In a follow-up study (9), the bacterium Shewanella loihica was used to stabilize corroded iron coupons, but it required the addition of 1% NaCl, as this bacterium could not reduce iron in the absence of salt. However, the addition of NaCl to the reductive medium is problematic because the presence of chlorides or the formation of chlorine-containing iron oxyhydroxides (e.g., akaganeite) enhances undesirable corrosion processes (10). Hence, there is a need for identifying more-suitable bacterial candidates by selecting facultative anaerobes that could reduce iron in the absence of added chlorinated salts.
Freshwater sediments are a potential source of metal reducers because of the presence of ferric ions as electron acceptors in the form of poorly crystalline Fe(III) oxides (6,11). In order to select new strains, a screening of bacterial isolates from two Swiss lakes was performed. As a result, two isolates belonging to the genus Aeromonas (Aeromonas sp. strain CA23 and Aeromonas sp. strain CU5) were identified. These facultative anaerobic bacteria were able to reduce soluble and solid Fe(III) phases (12). Aeromonas spp. are ubiquitous bacteria associated with aquatic environments (13). Some species are reported to be facultative dissimilatory iron reducers (13). Among the species known to reduce iron, Aeromonas hydrophila is the best-known example (14). However, one risk of an approach starting from the selection of environmental ironreducing strains is the unintentional enrichment of pathogens. Indeed, Aeromonas can be found in many environmental niches (15). Some Aeromonas species, such as Aeromonas sobria or Aeromonas salmonicida, are known to be pathogens of aquatic animals and especially of fish. Likewise, A. hydrophila and Aeromonas veronii have been related to human infection (16)(17)(18). The goal of this study consisted of improving the development of a biotechnological treatment for the stabilization of corroded iron heritage objects by testing the ability of the two Aeromonas isolates to convert reactive iron corrosion products into biogenic Fe(II) minerals. In addition, we investigated the iron reduction mechanism and the potential virulence of the isolates. Finally, we performed tests of an application procedure on real artifacts to demonstrate the validity of the bio-inspired solution proposed here for conservation and restoration of iron artifacts.

FIG 1
Iron reduction and biogenic mineral formation on corroded iron plates treated with strains CA23 and CU5. (A) Visual appearance (top) and scanning electron microscopy (SEM) images (bottom) of the iron coupons treated with strains CA23 and CU5 after 1, 2, and 4 weeks of incubation compared to an untreated coupon. The red letters f, s, and c indicate foil-, sphere-, and cube-like aggregates, respectively. (B) X-ray diffraction (XRD) diffractograms of corroded iron coupons treated with CU5 and CA23 for 4 weeks, the abiotic controls (Ac), and the untreated coupons (NT). (C) Confocal microscopy images of iron coupons. Left, plate treated with strain CU5 for 4 weeks. Right, untreated plate. average nucleotide identity (ANI) analysis. The ANI value calculated between the two strains was 85.61% (see Fig. S2A in the supplemental material). According to the analysis, strain CA23 could belong to either Aeromonas bestiarum CECT 4227 or A. salmonicida CBA100, given its placement in the phylogenomic tree ( Fig. 2A). This was supported by the ANI values, which were greater than 95% in both cases (Fig. S2B). Both A. bestiarum CECT 4227 and A. salmonicida CBA100 are reported to be fish pathogens. The latter was the first Chilean pathogenic isolate to be sequenced, and the analysis of its genome shows a diverse set of mechanisms involved in resistance to multiple antibiotics (21). However, none of these species has been reported to reduce iron. Concerning CU5, the results show that this bacterial strain is closely related to A. sobria CECT 4245 and Aeromonas sp. strain EERV15, suggesting that it could belong to the species A. sobria ( Fig. 2A and S2C). Although A. sobria is also considered an opportunistic fish pathogen, there is increasing evidence of a large phenotypic diversity within this species (22). Iron reduction for A. sobria has not yet been assessed.
Electron transport chain for solid iron reduction in CA23 and CU5. The iron reduction mechanism in Aeromonas is less well characterized than the mechanism described for the model iron reducers Shewanella and Geobacter. In Shewanella oneidensis, the mtrCAB operon, as well as the omcA and undA genes, encode the key proteins involved in iron reduction (23). In A. hydrophila, the mechanism for reduction of Fe(III)-bearing minerals involves a respiratory electron transport chain (ETC) in which reducing equivalents enter via a succinate dehydrogenase and are transferred from quinones and c-type cytochromes to a membrane-associated Fe(III) reductase (6). It has been proposed that in A. hydrophila the cluster involved in iron reduction is homologous to the mtrCAB operon of S. oneidensis, but no homologs of omcA, undA, or mtrDEF have been identified (24). We analyzed the genomes of CA23 and CU5 to propose a potential ETC for the reduction of iron minerals based on the model of the electron transport to solid iron reported for S. oneidensis (23) and A. hydrophila (24). Through using blastp (identities ranging from 60% to 80%), we proposed a model for CA23 including homologs of the mtrCAB complex (gene loci CK910_RS13405, CK910_RS13415, and CK910_RS13395, respectively) and a homolog of cymA (CK910_RS20025) (Fig. 2B). Interestingly, in CU5, only a homolog of cymA (CK911_02795) and no homologs of mtrCAB were identified using blastp (identities below 35%) and Needleman-Wunsch (identities below 20%) searches. This suggests that this bacterium has a different iron reduction pathway. As genes encoding pili have been found to be involved in iron reduction in other bacteria, such as Geobacter sulfurreducens, a search for genes encoding pili was also performed. However, only a homolog of the gene pilC in G. sulfurreducens was identified in the genomes of both strains (see Table S2 in the supplemental material).
a The letters f, s, and c refer to the foil-, sphere-, and cube-like aggregates, respectively, shown in Fig. 1. b The values from the EDS analysis for carbon (C) should be considered qualitative. c -, not detected. Because it was previously observed for S. loihica (9) that the production of Fe(II) phosphate minerals in a medium lacking phosphate can be related to polyphosphate accumulation (25), we investigated this trait in Aeromonas. Indeed, the genomes of CA23 and CU5, as well as those of A. hydrophila, A. salmonicida, and A. veronii, all possessed two genes coding for polyphosphate kinases and one gene coding for a polyphosphatase (Fig. S2D). These genes are known to be involved in polyphosphate accumulation (26). Interestingly, even though polyphosphate accumulation in Aeromonas has been reported (27), especially for A. hydrophila (28), no extensive information on this process could be found, which is surprising because genes involved in polyphosphate accumulation were predicted in several Aeromonas genomes.
Analysis of the pathogenic potential of the strains. The unintended selection of pathogenic bacteria is a risk when using environmental samples as a source of strains involved in iron cycling. Iron metabolism plays a pivotal role in competition and pathogenicity (29). Therefore, the pathogenic potential of environmental isolates has to be considered during the development of a biological approach to remediate iron corrosion. Aeromonas species are autochthonous inhabitants of aquatic environments such as freshwater, brackish water, sewage, and wastewater (16), and some studies have shown that some motile Aeromonas species are becoming food-and waterborne pathogens of increasing importance (30). Aeromonas species are known as causative agents of a wide spectrum of infections in humans and animals, predominantly amphibians, fish, and reptiles (31). The ability of A. hydrophila, A. caviae, and A. veronii to cause disease can be attributed to a broad range of virulence factors (16). These pathogens also display variable degrees of antimicrobial resistance.
Given the pathogenic lifestyle of the closest relatives to our environmental isolates, we performed several analyses to infer a potential pathogenic lifestyle. First, we considered the resistance of the strains to a series of antibiotics. CA23 and CU5 have the same antibiotic resistance profile (see Table S3 in the supplemental material). They are resistant to some ␤-lactam antibiotics, such as ampicillin, amoxicillin, and cefalotin, suggesting that they could have a cephalosporinase (a ␤-lactamase). Indeed, it has been demonstrated that almost all Aeromonas species are resistant to ampicillin and cefalotin (30).
Next, we investigated the potential pathogenic lifestyle of the two Aeromonas strains by analyzing the repertoire of virulence factors found to be encoded in their genomes. According to the literature, the main pathogenic factors associated with Aeromonas spp. include structural features such as pili and flagella, lipopolysaccharides (LPS), and an S-layer. Other extracellular factors, such as iron binding systems (e.g., siderophores), proteases, chitinases, amylases, secretion systems, ␤-lactamases, hemolysins, and toxins (mainly enterotoxins), also participate in pathogenicity. Indeed, hemolysin and aerolysin are considered the main virulence factors in A. hydrophila, being responsible for hemolytic, cytotoxic, and enterotoxic activities (13,31). The search for genes coding for potential virulence factors in the genomes of CA23 and CU5 showed that, indeed, these two Aeromonas strains possessed most of the genes reported for pathogenicity ( Table 2). CU5 appears to have a lesser pathogenic potential, as genes coding for toxins, enterotoxins, and aerolysin were not found in its genome. This tends to favor CU5 for the proposed bio-based treatment to be applied by conservators and restorers.
However, possession of this arsenal of pathogenicity factors does not imply a pathogenic lifestyle. For example, the ␤-lactamases encoded in our two strains were demonstrated to be functional and to confer resistance to three ␤-lactam antibiotics in vivo (Table S3). However, the same might not be the case for other pathogenicity factors detected in the genome. Therefore, and given the large range of potential virulence factors, we investigated the expression of these genes under the conditions used to convert reactive Fe(III) corrosion products to Fe(II) minerals (see Fig. S3 in the supplemental material). Under these conditions, none of the genes involved in pathogenicity were significantly expressed (https://figshare.com/s/dbf6be9c8286041db24f, Toxins and extracellular enzymes Toxins and enterotoxins CK910_RS00960 https://figshare.com/s/a5cdd9f0de2e04b612b3). Instead, the genes for the type II secretion system (epsE), multidrug transporters (norM), and antitoxin (yfjZ) were downregulated under iron-reducing conditions (the last was upregulated in CA23) ( Table 2). Although the results of the transcriptomic analysis would suggest that genes associated with pathogenicity are not relevant in the case of iron reduction, additional analyses are required to establish a potential pathogenic lifestyle for the strains. Therefore, in addition to the investigation of pathogenicity using genomic and transcriptomic approaches, the pathogenic potential of the two isolates was addressed more directly by analyzing the expression of the toxin hemolysin and by evaluating the cytotoxicity of the strains on the nematode Caenorhabditis elegans. Aeromonas spp., especially A. hydrophila, A. sobria, and A. salmonicida, are often reported as pathogens and associated with an hemorrhagic effect on warm-and cold-blooded animals (18,32,33). Many virulence factors, particularly toxins, have been reported to confer this phenotype (18,32). Hemolytic activity for strains CA23 and CU5 was tested on sheep blood agar. Strain CA23 was positive for ␤-hemolysin activity, with a minor presence of a hemolytic synergy with Staphylococcus aureus (Fig. 3A). This synergistic effect is probably due to the presence of CAMP factor (34,35). In contrast, no hemolysin or CAMP factor effect was observed for CU5 (Fig. 3A). Some studies on A. hydrophila have demonstrated that hemolytic activity in this bacterium requires the simultaneous presence of genes coding for hemolysin and aerolysin (32). Accordingly, Aeromonas sp. strain CA23 possesses several genes coding for hemolysins and one for aerolysin, which probably explains the hemolytic activity observed in vivo (Table 2). In the case of strain CU5, no genes coding for aerolysin were detected in the genome (Table 2), and this could explain the results obtained (Fig. 3A). In addition, strain CA23 has four genes coding for enterotoxins in its genome, while CU5 possesses only one. This could reinforce the higher pathogenic potential of CA23.
The second in vivo test performed measured the cytotoxic effect of both Aeromonas strains on C. elegans. The use of nematodes, especially the roundworm C. elegans, is considered an important emerging technique for studying host-pathogen interactions (36,37). Some studies have reported that the existence of universal virulence factors allows mammal-pathogenic bacteria to cause disease in nonvertebrate hosts (36).  Moreover, the presence of conserved signaling pathways could lead to the extrapolation of data from invertebrate studies to humans or other vertebrates (37). Working with the bacterivore C. elegans as a host model has many advantages. To test infection, its normal food source (Escherichia coli OP50) can be replaced (or combined) with the bacterium of interest (36)(37)(38). This model has been applied already to investigate the pathogenicity of some Aeromonas spp. (36,38,39). Adding Aeromonas sp. strain CA23 to a bacterial solution to feed C. elegans resulted in a reduced activity of the nematode as a function of the dose (Fig. 3B). At a low dose of CA23 (optical density [OD] of 0.3), C. elegans displayed higher activity (114%) than the control with only regular food bacteria (E. coli OP50). However, nematode activity decreased when CA23 biomass increased (50% at an OD of 0.7). This decrease suggests that CA23 has a deleterious effect on C. elegans. In contrast, the increase in the abundance of Aeromonas sp. strain CU5 added to the feeding medium resulted in higher activity of C. elegans (Fig. 3B) (42% at an OD of 0.3 and 116% at an OD of 0.7). This result indicates that the nematode is able to develop and feed on CU5 and suggests that this bacterium does not have a toxicological effect.
All the experiments suggest that CU5 might be a safer option for the development of a bio-based treatment, although entirely ruling out a pathogenic lifestyle by additional experiments is beyond the scope of this study. However, one aspect that must be addressed is the ease of decontamination of the material (for instance, iron artifacts and application matrix) after treatment. Therefore, a test to inactivate the strains by treatment with ethanol and heat was performed. The results showed that even a 10-min treatment with ethanol at a concentration as low as 23% inactivates both strains. The same can be obtained with higher concentrations, longer exposure times, or overnight incubation at 60°C (Fig. 3C), suggesting that a treatment with pure ethanol (to avoid the detrimental effect of water on the surface of an iron artifact) could be considered in actual practice. This indicates that the discarding of these strains can be achieved in a simple way, which is important for biotechnological applications. Treatment application on archaeological iron objects. In order to establish a prototype treatment that would convert the reactive corrosion layer into chemically stable iron minerals, we used a commercial gel to apply bacteria on archaeological iron nails. The preliminary results showed that the treatment was partially successful (see Fig. S4 in the supplemental material); the corrosion layer of the nails was not fully converted, in contrast to the case for the treated iron coupons (Fig. 1). With a real archaeological object, the formation of biogenic iron minerals was observed after 4 weeks with the CA23 bacterium and starting from 2 weeks with the CU5 bacterium. These minerals were composed mainly of O, C, and P according to EDS measurements. Fourier transform infrared spectroscopy (FTIR) analyses were performed to identify the minerals formed. The results showed typical vibration bands of proteins (amide I peak around 1,640 cm Ϫ1 and amide II peak around 1,420 cm Ϫ1 ) indicating the presence of bacteria. The peaks between 1,040 and 940 cm Ϫ1 correspond to the phosphate absorbance region (40,41), which confirmed the formation of iron phosphates. Absorbance bands corresponding to the coordinated water associated with iron phosphates could be observed at 3,161 and 795 cm Ϫ1 , which indicates that these minerals have a crystalline phase (40,41). According to the FTIR spectrum, the biogenic minerals belonged to the vivianite-related group (41).
To optimize the treatment, we performed a second test by increasing the number of bacterial cells and the incubation time and by treating archaeological iron objects both aerobically and anaerobically. The results are shown in Fig. 4. The increased concentration of bacterial cells had a positive effect on the performance of the treatment, especially compared with the first attempt (Fig. S4). In this case, the treatment using CA23 or CU5 was successful. The corroded archaeological objects were completely covered by mineral compounds, providing a velvety texture and gray-black color to their surfaces ( Fig. 4A and B; see Fig. S5A in the supplemental material). This result is not due to the chemical composition of the delivery system, as no change in the visual aspect was observed in the controls without bacteria (Fig. 4A). In order to determine the nature of these gray-black compounds, Raman spectroscopy was performed (Fig. 4C). Vivianite (shifts at 187, 192, 480, 954, and 1,001 cm Ϫ1 ) and siderite (shifts at 734, 736, 1,083, and 1,087 cm Ϫ1 ) were detected only in the samples treated with CU5 for 1 and 2 months, in comparison to the untreated archaeological iron nails, which have been previously characterized using Raman microscopy (8). The same was the case for the abiotic control (Fig. S5B). These Fe(II) minerals were also present in the treated iron coupons described above (Fig. 1B). Likewise, Raman analysis showed that vivianite was less abundant than siderite in the 2-month samples. Goethite and lepidocrocite (reactive corrosion products) were completely transformed, as lepidocrocite was detected only in the untreated and abiotic control samples, while weak peaks corresponding to goethite were detected mainly in the 1-month samples. Even though partial reoxidation of the treated surfaces was observed after 48 h of air exposure, incubation under aerobic or anaerobic conditions does not seem to influence the efficiency of the treatment, which is positive because an aerobic treatment is easier for application.
There is a growing interest in the synthesis of inorganic components by biological systems in processes that are respectful of the environment (3). As an example, bacterial iron reduction has been used as a low-cost and environmentally friendly biotechnological approach for iron extraction from ores (7). In the cultural heritage field, exploiting specific bacterial abilities for the transformation of unstable iron corrosion products into more stable iron minerals represents a promising bio-based approach to traditional conservation and restoration methods (8). Stepwise improvement of the treatment is leading to a biotechnological application. The first treatment employed a strict anaerobic bacterium growing in a complex medium, which renders optimization challenging (8). In particular, even though the visual characteristics and the composition of the corrosion layers were transformed, the production of undesirable products and abiotic reduction were also observed. Further optimization using the facultative anaerobe S. loihica facilitated the potential up-scaling of the process, but the presence of chloride ions in the iron-reducing medium, which are reported to enhance corrosion on archaeological iron objects (10), is incompatible with actual practice (9). In this study, we proposed a promising iron stabilization method and tested it on outdoor corroded surfaces, as well as on archaeological objects. This treatment was performed under aerobic conditions, without the addition of salt, and the macroscopic aspect of the objects improved due to the transformation of the unstable corrosion layers into mainly siderite and vivianite. The treated objects had a color that agrees with the aesthetic criteria that are conventionally applied in iron heritage conservation. Nevertheless, an improvement of the long-term stability of these Fe(II) minerals that are formed and an assessment of their effective protectiveness against further corrosion should be considered in the future. In addition, more stringent tests are needed to verify a nonpathogenic lifestyle of the strains selected. CU5 appears to be more promising than CA23, based on its gene content and the tests performed in this study. Further investigation of the long-term stability and additional pathogenicity tests with higher organisms are important in order to convert the proposed treatment into a real in situ practice for conservators and restorers.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Bacterial isolates from Lake Cadagno and Lake Neuchâtel (Switzerland) were used in this study. Regular cultivation was performed aerobically using Luria-Bertani (LB) medium (1% tryptone, 0.5% yeast extract, and 1% NaCl) at pH 7 and 30°C under agitation at 130 rpm.
Iron samples. The corroded iron coupons (10 by 10 by 2 to 3 mm) were cut from iron plates of 50 by 50 by 2 to 3 mm. The plates were exposed for 1 year to an outdoor marine environment (French Institute of Corrosion, Brest, France). The orientation and position of the plates during exposure were selected according to the ISO 9223 standard. A corrosion layer composed of lepidocrocite and goethite was formed and characterizes the iron coupons. The archaeological iron nails were excavated from a chalky soil in Champagne region, France, and dated from the late Roman period (3rd century AD).
Chemical matrix for solid Fe(III) reduction. Iron reduction and biogenic mineral formation with the coupons as a solid iron source were performed in a matrix composed of 20 mM piperazine-N,N=-bis(2ethanesulfonic acid) (PIPES), 5 mM sodium lactate, and 50 mM NaHCO 3 . For the sterilization of the iron coupons, autoclaving was avoided because further corrosion was observed under both oxic and anoxic conditions. Instead of autoclaving, coupons were washed with 70% (vol/vol) ethanol and dried under UV exposure for 1 h for each side. A control for this sterilization method was performed by incubating UV-sterilized plates on solid LB medium under oxic and anoxic conditions, which showed no bacterial growth after 1 week of incubation.
Iron reduction of solid Fe(III). In order to decouple growth from iron reduction, a culture of 50 ml of CA23 and CU5 was prepared by cultivating the bacteria under aerobic conditions in LB medium overnight at room temperature. Cell biomass was then collected by centrifugation (1,700 ϫ g for 10 min), and the pellet was washed with 20 mM PIPES solution. The pellet was resuspended in the same solution and added to the matrix containing iron. To evaluate iron reduction, samples were collected every 2 h for 24 h. The production of iron minerals was evaluated after 1, 2, and 4 weeks of incubation at room temperature under agitation (120 rpm). Experiments were performed in triplicates, including abiotic controls (without bacteria).

Measurements of Fe(II) concentration.
To measure the reduction of Fe(III) into Fe(II), the ferrozine assay was performed following a protocol modified from that of Bell et al. (42). First, 100 l of 5 M HCl was added to 900 l of collected samples and stored at 4°C. The samples were centrifuged for 1 min at 6,700 ϫ g, and then 100 l of the supernatant was taken and 900 l of ferrozine solution (0.1% ferrozine in a 100 mM HEPES solution at pH 7) was added. The Fe(II) concentration was determined by measuring the absorbance at 562 nm with a UV-visible spectrophotometer (Thermo Fisher Scientific Genesis 10s). A calibration curve was obtained using serial dilutions of 1 mM ferrous ammonium sulfate solution in acidic MilliQ water (pH 2).

Development of a user-friendly treatment of archaeological iron nails.
For this treatment, we tested the effect of the bacteria CA23 and CU5 on the stabilization of archaeological iron nails in a gel delivery system. Before being in contact with the bacteria, the nails were sterilized using 70% (vol/vol) ethanol and placed under UV light for 1 h. Phytagel (Sigma-Aldrich), was chosen as the delivery system. The gelling agent in its initial powder format was sterilized under UV light for 30 min before being mixed with the bacterial culture.
For each nail, 100 ml of overnight cultures of CA23 and CU5 (at room temperature) with ODs (measured at 600 nm) of 1.080 and 1.027, respectively, was centrifuged at 4,000 rpm for 10 min. The pellet was washed twice with 20 mM PIPES solution. The pellet was mixed with 1 g of the gelling agent, and 1 ml of 1 M sodium bicarbonate and 0.5 ml of 1 M sodium lactate were added to this gel mixture. The bacterium-gel system was applied to the nails, which were then covered with cellophane and stored at room temperature. Two conditions were tested: incubation under oxic conditions (the nails were placed in petri dishes) and under anoxic conditions (using Anaerocult A bags [Merck]). To test the treatment's efficiency and according to the results obtained with the iron coupons, the treatment was stopped after 1 and 2 months for each bacterium. The abiotic control consisted of 2 g of gelling agent mixed with 10 ml of 20 mM PIPES solution amended with 5 mM sodium lactate and 50 mM sodium bicarbonate. The sterility of the controls after treatment was not verified. However, gel without bacteria added did not change in color, while the gel mixed with bacteria turned black after the incubation period.

Analytical techniques. (i) SEM-EDS analyses.
For scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM-EDS) analyses, the iron coupons were washed twice with deionized water and 70% (vol/vol) ethanol before being mounted on stubs using carbon conductive tape without gold sputtering. To avoid humidification, the stubs were stored in desiccators with silica gel. A Philips ESEM XL30 FEG environmental scanning electron microscope equipped with an energy-dispersive X-ray analyzer was used. The samples and coupons were observed in secondary electron mode at an acceleration potential of 10 to 25 keV and with a 10-mm working distance.
(ii) XRD analyses. X-ray diffraction (XRD) analyses were performed on untreated, abiotic control iron coupons and on iron coupons treated for 1 month. The preparation of the iron coupons was the same as for SEM-EDS analyses. High-quality data q-2q (bulk) measurements were performed with a PANalytical X'Pert Pro MPD diffractometer equipped with a fast-linear detector and sample spinner. The measurement time was 60 h to acquire high statistics for better signal-to-noise ratio. The sensitivity of the XRD phase identification is about 3% by weight. The ICSD database was used for the identification of siderite (98-004-6078), vivianite (98-004-6142), and magnetite (98-001-1782).
(iii) Confocal microscopy. A three-dimensional (3D) configuration of the iron coupons was reconstructed from a scanning of the surface made at a magnification of ϫ50 with its corresponding color-coded height map. The images were taken using a Smartproof 5 Zeiss confocal microscope (aperture correlation microscopy) with Zeiss Efficient Navigation (ZEN) and Confomap software.
(iv) Raman spectroscopy analysis. Nondestructive Raman spectroscopy was performed directly on the samples to define the molecular compositions of the corrosion layer before and after bacterial treatment. The analysis was carried out with a Horiba-Jobin Yvon Labram Aramis microscope equipped with an Nd:YAG laser of 532 nm at a power lower than 1 mW. The spectral interval analyzed was between 100 and 1,600 cm Ϫ1 . Single-points analyses were carried out with the following conditions: 400-m hole, 100ϫ LWD objective, D2 filter, 100-m slit, and 5 accumulations of 100 s. The spectra recorded were corrected (automatic baseline correction) using LabSpec NGS spectral software. Reference spectra were used for identifying the compounds present.
(v) FTIR analyses. Fourier transform infrared spectroscopy (FTIR) measurements were performed on the minerals formed on the surfaces of the nails. An iS5 Thermo Fisher Scientific spectrometer with a diamond attenuated total reflectance (ATR) crystal plate (iD5 ATR accessory) was used. All spectra were acquired in the range 4,000 to 650 cm Ϫ1 at a spectral resolution of 4 cm Ϫ1 . A total of 32 scans were recorded and the resulting interferograms averaged. Data collection and postrun processing were carried out using Omnic software.
Molecular methods. (i) RNA extraction and rRNA removal. CA23 and CU5 were cultivated on LB medium and then transferred in the same matrix used for solid iron reduction as described above. To identify the genes involved in solid iron reduction, two conditions were tested: with or without iron coupons. The experiment was performed in triplicates. Sampling of 1 ml from each replicate was achieved before inoculating the cells into the matrices at 0, 12, and 24 h of incubation. After sampling, bacterial cells were stored at Ϫ80°C. Prior to RNA extraction, samples were thawed on ice and then briefly centrifuged (10,000 ϫ g for 1 min). RNA extraction was done using a Zymo Research Direct-zol kit with TRIzol reagent according to the manufacturer's instructions. A Ribo-Zero magnetic kit for Gram-negative bacteria (Illumina) was used for rRNA removal, according to the manufacturer's instructions. The concentration of RNA from the extraction was considered to calculate the amounts of reagents used for the preparation of the magnetic beads and for the treatment of total RNA with Ribo-Zero rRNA removal solution (4 l rRNA removal solution for 0.25 to 1 g total RNA per reaction). Samples in Ribo-Zero rRNA removal solution were incubated at 68°C for 10 min, followed by 5 min of incubation at 20°C. To remove rRNA, the Ribo-Zero mix incubated previously was mixed with the prepared magnetic beads and placed at room temperature for 5 min and then at 50°C for 5 min. The rRNA-depleted samples were purified using the RNeasy MinElute kit (Qiagen). The final mRNA eluted was quantified using the Qubit-IT RNA HS assay kit (Thermo Fisher Scientific) and Bioanalyzer RNA 6000 Pico kit (Agilent).
(ii) Illumina indexed library preparation. The preparation of indexed libraries from RNA and sequencing was performed using Illumina's ScriptSeq v2 sample preparation kit. As less than 40% of the RNA obtained was more than 600 nucleotides (nt) long, no fragmentation was done. Nevertheless, a maximum of 5 g of RNA was mixed with 2 l of cDNA primer (200 M), incubated at 65°C for 5 min on a thermal cycler, and then placed on ice (1 l of fragmentation solution was added after incubation according to the manufacturer's instructions). Three microliters of the cDNA synthesis premix obtained previously was mixed with 0.5 l dithiothreitol (DTT) (100 mM) and 0.5 l StarScript reverse transcriptase. The reaction mixture was incubated at 25°C for 5 min and at 42°C for 20 min and then held at 37°C. Next, 1 l of finishing solution was added, and the reaction mixture was incubated at 37°C for 10 min and 95°C for 3 min and held at 25°C. Eight microliters of tagging master mix (7.5 l terminal tagging premix and 0.5 l DNA polymerase) was added, and the tubes were incubated at 25°C for 15 min and 95°C for 3 min and held on ice. The purification of the cDNA was performed using Agencourt AMPure XP beads. PCR was performed using Failsafe PCR enzyme mix (Illumina), made by preparing a PCR master mix (25 l of PCR premix E, 1 l of forward PCR primer, and 0.5 l of Failsafe enzyme [2.5 U/l]) and adding 1 l of index primer (Script Seq index PCR primers; Illumina). The reaction mixture was incubated in the thermocycler as follows: 95°C for 1 min and then 15 cycles of 95°C for 30 s, 55°C for 30 s, and 68°C for 3 min. The samples were then incubated at 68°C for 7 min and held at 4°C. The PCR product cleanup was done using AMPure XP beads. The validation and the control of the quality of the library were done using the Qubit-IT double-stranded DNA (dsDNA) HS assay kit and BioAnalyzer DNA high-sensitivity kit (Agilent). Accurate quantification of the amplified library products is important for the sequencing results on the Illumina next-generation sequencing platforms. A quantitative PCR (qPCR) (using the Kapa SYBR Fast universal qPCR kit for Illumina libraries) was performed in order to obtain an accurate quantification of the amplifiable molecules in the DNA libraries.
(iii) Illumina sequencing and RNA data analysis pipeline. Libraries were sequenced on an Illumina NextSeq 500 instrument, using the NextSeq sequencing kit HO 300 cycles (Illumina) according to the manufacturer's instructions (paired-end sequencing mode). The raw reads of transcriptome sequencing (RNA-Seq) data were processed using Pipeline for Reference-Based Transcriptomics (PiReT) software (https://github.com/mshakya/PyPiReT). Briefly, the reads were aligned and assigned to the reference genomes using HISAT2 v2.05 (43) (version information is available in the github repository). Aligned reads were then analyzed and assigned to individual genes according to the genome annotations (GenBank accession no. GCF_002587065.1 and GCF_002586745.1 for CA23 and CU5, respectively) using featureCounts v1.5.2 (44). The normalized read counts for each gene (reads per kilobase per million [RPKM]) were calculated using the R software, package edgeR v3.14.0. Graphical representations of differential gene expression data obtained from pairwise comparison (volcano plots, principal-component analysis [PCA], and heat maps) were created using R software (45) including the ggplot 2 package (46).
(iv) Genomic analysis for the genes involved in solid iron reduction, polyphosphate accumulation, and virulence. Genomic analysis was performed in order to identify the potential genes involved in iron reduction in CA23 and CU5. To do so, we extracted the sequences of the proteins involved in solid iron reduction in the model bacterium S. oneidensis MR-1 (accession no. NC_004347. (v) Whole-genome sequencing of strains CA23 and CU5. Genomic DNA of the bacterial isolates was extracted using the Qiagen G20 instrument for bacteria according to the manufacturer's instructions but with a modification consisting of washing the bacterial pellet with PIV solution (10 ml 1 M Tris-HCl [pH 8] and 58.4 g NaCl in 990 ml of Nanopore water) before resuspension in B1 buffer. The two genomes were sequenced using a PacBio RSII sequencer (Microsynth AG, Switzerland). A PacBio long-read library was assembled with HGAP3 (51), version 2.3.0, and the genomes were annotated using the Prokka pipeline (52), version 1.11.
(vi) Phylogenetic analysis. Phylogenetic Analysis and Molecular Evolution (PhaME) software (53) was used to reconstruct the single nucleotide polymorphism (SNP) phylogeny of Aeromonas spp. Briefly, PhaME aligned the genomes of Aeromonas, removed regions that are not conserved across all genomes, and calculated SNPs from the conserved regions. The SNP alignment was then used to reconstruct maximum-likelihood phylogeny using RAxML v8.2.9 (54). Aeromonas genomes were downloaded from GenBank (last accessed on 27 July 2017). The resulting tree was visualized using the Newick utilities (55).
(vii) Genome comparison. With the Artemis comparison tool (ACT) (56), a tool for pairwise comparison between complete genome sequences and associated annotations, DNA sequences were used to identify and analyze regions of similarity and difference between genomes. An average nucleotide identity (ANI) calculator from the laboratory of Kostas Konstantinidis (57) (http://enve-omics .ce.gatech.edu/ani/) was used with default settings to measure the nucleotide-level genomic pairwise similarity between the coding regions of the genomes.
Physiological characterization of CA23 and CU5. The Vitek 2 Compact consists of biochemical tests for the physiological identification of bacteria. It is an automated microbiology system utilizing growth-based technology under different biochemical conditions contained in colorimetric reagent cards. From a fresh culture of pure bacterial strain, a 0.5 McFarland suspension was prepared using sterile physiological water. The bacterial suspension was then inoculated in wells contained in a Vitek card. Each well corresponds to an individual test substrate. Substrates measure some metabolic activities, such as carbon source utilization, enzymatic activities, and resistance. A transmittance optical system allows interpretation of test reactions using different wavelengths (660 nm, 568 nm, and 430 nm). During incubation at 35.5°C, each test reaction was read every 15 min to measure either turbidity or colored products of substrate metabolism.