Methyl red, methyl orange, and 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES) were purchased from Sigma–Aldrich. Goethite was synthesized by the procedure of Atkinson et al. (1967). Phosphate-adsorbed goethite was prepared by mixing goethite (final concentration of 1000 mg L^–1) with a KH₂PO₄ (100 mg L^–1) for 48 h and stored before use.

The n-octanol-water partition coefficient (log K_OW) is an important parameter of hydrophobicity, which can be used to assess the distribution trends of organic pollutants in both polar and nonpolar phases (Liu et al., 2015). Log K_OW of methyl red and methyl orange were obtained from Estimation Program Interface (EPI) Suite (USEPA, United States Environmental Protection Agency, Washington, 2017).

Shewanella decolorationis strain S12T (CCTCCM203093T = IAM 15094^T) and its mutant strains (△mtrC and △omcA) were preserved in our laboratory. All strains were grown aerobically in sterilized LB medium at 28°C. Then the bacterial suspensions were centrifuged at 6000 × g for 10 min at 28°C and washed three times with 30 mM HEPES buffer. The harvested cells were resuspended in HEPES buffer for subsequent use.

Experiments were conducted in 8 ml serum bottles capped with Teflon-coated rubber stoppers. Experimental preparations and measurements were performed in an anoxic chamber (Electrotek AEP) filled with a gas mix (N₂:H₂ 90:10). Each bottle was filled with 5 ml HEPES buffer containing various combinations of strain S12 (about 10⁷ cells ml^–1), goethite (0–1000 mg L^–1), methyl red (1 mM), or methyl orange (1 mM), and was deoxygenated by bubbling with N₂ for 30 min. Sodium formate was the electron donor, whose concentration (10 mM) was high enough to support regular bioreduction of both dyes. At predetermined time intervals (0.5, 2, 4, 6, 8, 10, 12, and 14 h), the bottles were taken out and analyzed for the concentration of methyl red or methyl orange. The concentrations of methyl orange and methyl red were determined using a UV–vis spectrophotometer by recording the absorption at 460 and 430 nm (Chen et al., 2010; Liu et al., 2017). Control reactors were prepared as described above but without bacterial inoculants. In addition, strain S12 mutants △mtrC and △omcA were used to reduce azo dye to assess the role of related proteins. All treatments were conducted in triplicates.

Competition for electrons between goethite and azo dyes was evaluated in reactors in the presence of strain S12, sodium formate, goethite, and either azo dye. As the colorimetric assay of Fe(II) can be disturbed by azo dyes, the incubation was extended to 14 h to reduce the interference. The 0.5 N HCl-extractable Fe(II) was analyzed using the ferrozine assay.

To characterize the effect of goethite surface Fe(III) on bioreduction of azo dyes, the reductive kinetics of two dyes was also evaluated in the presence of phosphate-adsorbed goethite, and data were collected and analyzed in the same manner as previously stated. Phosphate has been shown to effectively weaken the interaction strength between iron (hydr)oxides and bacterial cells (Appenzeller et al., 2002). The phosphate group can bind Fe(III) at the surface, displacing coordinated hydroxyl group and/or water molecules (Tejedortejedor and Anderson, 1990; Elzinga and Sparks, 2007), and thus reduce contact probability of Fe(III) with the bacterial outer membrane (Supplementary Figures S2, S3). The phosphate adsorption amount of goethite surface was about 8.09–39.46 mg g^–1 (Torrent et al., 1990), and thus excessive phosphate (100 mg L^–1) by reference was added. We measured the phosphate concentration of the supernatant of phosphate in phosphate-adsorbed goethite system after strain S12 spiked (about 10⁷ cells ml^–1), and there was no significant difference compared with bacteria-free bacteria treatment (p > 0.05).

Attenuated total reflectance Fourier transform infrared measurements were performed with a horizontal ATR cell (Pike Technologies, Inc.), equipped with a ZnSe crystal element, in a Bruker Tensor II FTIR spectrometer. All spectra were averaged with 256 scans at 4 cm^–1 resolution.

The spectra of bacterial cells were collected by subtracting the spectrum of background solution. A relatively high concentration of bacterial cells (∼10⁸ cells ml^–1) was used to clearly show FTIR signals of the bacterial cell. Background solution was passed through the flow cell at the rate of 2 ml min^–1 until the equilibrium was established and the spectrum was used as background spectrum. Then S12 cells were spiked and constantly stirred with a magnetic stirrer. The interfacial spectra of S12 cells were measured with goethite film on the surface of ZnSe crystal. The film was prepared by drying 1 ml of a 1 g L^–1 dispersed goethite suspension at 60°C for 6 h under an N₂-atmosphere in a similar way as the previous report (Elzinga et al., 2012; Elzinga and Kretzschmar, 2013; Wei et al., 2016). The coated film was pre-equilibrated with background solution at a flow rate of 2 ml min^–1 until there was no significant difference in the spectra. A background spectrum containing the absorbance of the solution, goethite, and ZnSe crystal was then obtained. Collection of interfacial spectra started as bacterial suspension entered the flow cell, for up to 4 h.

For the S12-goethite system, IR spectra collected with an increment of 20 min were analyzed using 2Dshige software version 1.3 (Kwansei Gakuin University, Japan). Synchronous and asynchronous maps in 2D-COS were created and the origin of band pair (υ₁, υ₂) can be known from the sign of synchronous and asynchronous correlation peak [Φ(υ₁,υ₂) and Ψ(υ₁,υ₂)] according to Noda’s rules (Noda and Ozaki, 2005). Briefly, if synchronous and asynchronous correlation peaks have the same signs, the spectral change at υ₁ precedes that at υ₂. As for the opposite sign, the order is reversed.

To determine the role of the terminal reductive proteins MtrC and OmcA in bacterial outer membrane–goethite interactions, we used CLSM to observe the behavior of S12 and its mutants on the surface of goethite through a parallel plate flow chamber. Goethite particles were attached to microscope glass slides (25 mm wide, 75 mm long, and 1 mm thick) and the coating method was similar to ATR-FTIR experiment. The glass slide was loaded onto the parallel plate flow chamber as a cover. A peristaltic pump was used to connect the flow chamber and circulate deoxygenated HEPES buffer containing sodium formate (10 mM) and bacterial cells (about 10⁷ cells ml^–1) at a rate of 2 ml min^–1 for 4 h. Then flow was switched to HEPES buffer for 5 min and the glass slide was removed and washed by HEPES to eliminate non-adhering bacteria cells. The slide was treated with LIVE/DEAD BacLight staining kit and imaged by CLSM (LSM 700, Zeiss). Twenty-four 319.53 μm² images were collected from three replicates. The number of adhering bacterial cells was analyzed using an image analysis software (Image J, NIH).

For TEM observation, bacterial cells were incubated with 100 mg L^–1 goethite for 4 h and then washed three times in HEPES buffer, fixed in 2% glutaraldehyde, dried in ethanol, and embedded in epoxy resin. Thin sections were cut using an ultramicrotome. Then samples were imaged on a Hitachi H-7650 with an accelerating voltage of 80 kV.

The molecular structure of methyl orange and methyl red differs in a high polar sulfonic acid group. Log K_OW of methyl orange and methyl red were −0.66 and 3.83, respectively. Based on the basic principle of “like dissolves like,” the difference of log K_OW demonstrated the predominantly high polarity and hydrophilic nature of methyl orange and the low polarity of methyl red.

As shown in Figure 1, 1 mM methyl orange and methyl red could be completely decolorized by strain S12 within 10 and 6 h, respectively, while no decolorization was detected in the abiotic reactors, suggesting that the azo dye reduction resulted from the biotic process. The first-order rate constant of methyl orange bioreduction (0.28 h^–1) was lower than that of methyl red (0.39 h^–1), indicating that the high polar sulfonic acid group impeded the bioreduction.

We tested the azo dye reduction abilities of mutant strains lacking MtrC or OmcA, which are the terminal reductive proteins on the outer membrane for the reduction of extracellular soluble iron (Ross et al., 2009). In the absence of MtrC or OmcA, 75 and 23% of methyl orange remained unreduced over 10 h (Supplementary Figure S1) and the ΔmtrC and ΔomcA complemented strains had a recovery of azo dye-reducing ability (data not shown). These results suggest that MtrC and OmcA are responsible for the reduction of methyl orange. It also indicates that methyl orange with a high polarity sulfonic acid group was mainly reduced extracellularly. In contrast, only 36 and 22% of methyl red remained after 6 h of cultivation of the MtrC- and OmcA-lack mutant strains, suggesting that the reduction of methyl red was less affected by the knock-out of MtrC.

The bioreduction of methyl orange and methyl red at different concentrations of goethite (0–1000 mg L^–1) is shown in Figure 1. No significant impacts were observed in the system with 10 mg L^–1 goethite. For methyl red, goethite significantly suppressed the bioreduction rate only at a concentration of 1000 mg L^–1. For methyl orange, the first-order reduction rate constant was 0.28 h^–1 in the absence of goethite and decreased to 0.18 and 0.11 h^–1 in the presence of 100 and 1000 mg L^–1 goethite. These results indicated that the inhibition of azo-reduction was proportional to the concentration of goethite added, and higher susceptibility was observed in the reduction of high polarity methyl orange.

To examine whether Fe(III) on goethite surface would compete with azo dye for electrons, we measured the levels of Fe(II) in various systems. Since the given electron donor (sodium formate) did not reduce Fe(III), Fe(II) in the system was solely attributed to biogenic production by strain S12. After 14 h of incubation, the total concentration of Fe(II) was 0.054 ± 0.006 and 0.077 ± 0.008 mM, respectively, in the 1000 mg L^–1 goethite-S12 systems with methyl orange or methyl red. Note that each molecule of sodium formate provides two electrons when oxidized (Hong et al., 2007) which is equivalent to two Fe(II) ions getting reduced. The electrons captured by Fe(III) accounted for merely 0.27 and 0.39% of the total electrons calculated by the consumption of sodium formate, respectively. These results strongly suggested that the reducing capacity of strain S12 for Fe(III) on goethite surface was weak under our experimental conditions.

We also tested the effect of goethite-surface Fe(III) on microbial azo dye reduction by masking the goethite surface with phosphate group binding. No significant differences were observed in the reduction rates between treatments with and without phosphate (Figure 1), suggesting that phosphate did not alter the electron flow to azo dyes. In the following assay, we replaced the 1000 mg L^–1 goethite with phosphate-masked goethite (1000 mg L^–1) and the inhibition of goethite on methyl orange and methyl red reduction disappeared. The first-order reduction rate constants of methyl orange and methyl red were 0.28 and 0.39 h^–1, respectively, showing no difference from the goethite-free treatments. Together, these results suggested that goethite-surface Fe(III) inhibited the bioreduction of azo dyes by the contact with bacterial cells, rather than serving as a competitive electron acceptor.

Transmission electron microscopy showed that the outer membrane of bacterial cells was coated with massive goethite particles (Figure 2a). In situ ATR-FTIR spectroscopy was further used to explore strain S12 interfacial behaviors on the goethite surface. As shown in Figure 2b, the ATR-FTIR spectrum of goethite-free S12 cells was similar to previous reports in the frequency range of 1800–950 cm^–1 (Jiang et al., 2004; Elzinga et al., 2012). The bands between 1800 and 1200 cm^–1 contain the amide groups [amide I (1653) and amide II (1546)] and carboxyl [e.g., deprotonated carboxyl groups (1400), carboxyl groups (1235)] (Jiang et al., 2004). The bands within ∼1150 and 950 cm^–1 are a complex composite of phosphate region, like υ(C-O-P, P-O-P) and υ_s(PO2-) (Parikh and Chorover, 2006). Compared to the case with ZnSe only, distinct changes were observed in the range of phosphate groups in the spectra of strain S12 on goethite surface and a prominent band at 1045 cm^–1 was found (Figure 2b). This band was attributed to the stretching vibrations of P-OFe in Fe2⁢PO4-x (Tejedortejedor and Anderson, 1990), which indicated that P-moieties of bacterial cells were involved in the interactions between strain S12 and Fe(III) at goethite surface through the information of inner-sphere complexes.

The synchronous and asynchronous maps for the S12-goethite system were shown in Figure 3. We chose the signal of amide II (1546) as a characteristic signal for bacterial surface proteins because the amide I band was often perturbed by water-related adsorptions (Shephard et al., 2010). In the synchronous map, all auto-peaks were positive (Figure 3A) at 1546/1045 cm^–1, indicating same sign of variation in peak intensity of amide II and υ(P-OFe). In the asynchronous map, the positive cross peak at 1546/1045 cm^–1 was found (Figure 3B). According to Noda’ rule, the sequence of the spectral change in intensity of amide II occurred before υ(P-OFe). Our result suggested that, when bacteria cells were introduced to the surface of goethite, the proteins at the outer cell surface firstly recognized or approached the interface. Then, phosphate group on bacterial envelope formed the inner-sphere surface complexes on goethite.

To understand whether proteins at the outer cell surface like MtrC and OmcA were in contact with Fe(III) on goethite surface, we collected CLSM images and the ATR-FTIR spectra of S12 mutants (△mtrC and △omcA) on goethite (Figures 4, 5). The surface cell density of △mtrC (0.077 ± 0.006 cells μm^–2) and △omcA (0.06 ± 0.005 cells μm^–2) on goethite-coated glass slides was significantly lower than strain S12 (0.176 ± 0.018 cells μm^–2) (p < 0.001), suggesting MtrC and OmcA are key binding sites for goethite surface. In addition, different from the ATR-FTIR spectrum of S12 wild-type on goethite, υ(P-OFe) at 1045 cm^–1 was not observed in the spectra of △mtrC or △omcA on goethite. It meant that, when MtrC and OmcA were not present on the S12 outer membrane, Fe(III) on goethite surface was not in contacted with phosphate moieties of the bacterial outer membrane to form a covalent bond. Furthermore, the number of bacteria on the mineral surface greatly reduced when surface Fe(III) was masked by phosphate, indicating that goethite-surface Fe(III) was the binding site of bacteria cell (Supplementary Figure S3). These results strongly suggested that MtrC and OmcA could participate directly in the interaction between the bacterial out membrane and Fe(III) on goethite surface by co-bounding to goethite surface.