To assess microbial mineral colonization in surface seawater under controlled conditions, we used the flow-through osmo colonization system (FLOCS) approach designed for use in the deep sea (Orcutt et al., 2010; Wheat et al., 2011). FLOCS are osmotically driven microbial colonization chambers that allow for long-term monitoring of microbiological and geochemical processes on mineral substrates. FLOCS typically consist of three primary components, in order of the direction of flow: colonization chambers made of plastic sleeves housing mineral substrates, a fluid collection coil, and an osmotic pump (Figure 2). In this experiment, colonization chambers consisted of a polycarbonate cylindrical sleeve that contained sterile (autoclaved) mild steel shavings (made by drilling into 0.125″ × 0.25″ 1018 Cold Finish Mild Steel Rectangle Bars)¹, and/or crushed pyrite (>250 μm size fraction, Ward’s Science, Catalog 466448) in sterile, nutrient deplete seawater from the Sargasso Sea (commercially available from Sigma-Aldrich, S9148-1L).

In this study, a single “free-flow” colonization chamber without attached fluid collection coil or osmotic pumps, and three independent FLOCS containing mild steel and pyrite (FeS₂) under various flow and nutrient scenarios, were deployed in aquarium tanks at the Wrigley Institute for Environmental Science (WIES) located in Big Fisherman’s Cove at Catalina Island, CA, USA (Figure 1B). The glass aquarium tank (189 L) was connected to a PVC piping system that pumps unfiltered seawater from the cove, from ∼10 m water depth, at a rate of 5 L per minute into and out of the tank. The tank volume replacement time is approximately 30 min. To preclude algal growth in the aquarium, the tank was completely covered in black plastic to keep the experiments in darkness. In these experiments, only the colonization chamber was submerged in the aquarium tank, while the Teflon tubing and osmotic pumps for the FLOCS were placed outside of the tank and secured on stands (Figure 2).

Four colonization conditions were used in this study (Figure 2; Table 1):

The HFMC experiment was deployed from July 20th to September 20th, 2012. FLOCS experiments were started on July 17, 2012, and terminated January 9, 2013, for a total of 176 days. Daily water temperatures in Fisherman’s Cove typically vary from 12 to 22°C, on average during these seasonal cycles (McAlary and McFarland, 1994). A record of the water temperature in the aquarium tank/delivery system during the deployment was not measured; however, aquarium tank temperatures measured from November 2014 to March 2015 show aquarium temperature ranges of ∼16–22°C (data not shown).

The HFMC was disassembled under sterile conditions for sample collection and discontinued after 60 days. FLOCS systems were disassembled after a 176-day deployment under sterile conditions. In addition, iron oxide precipitates at the bottom of the aquarium near the colonization experiments were also collected for cross-comparison to microbial communities in colonization samples (referred to as “microbial mat,” Figure 2B; Table 1). All collected samples were frozen immediately (-80°C) for subsequent nucleic acid extraction and microscopy analysis. At the end of the experiment, a total of 4-L of aquarium seawater was filtered onto a 0.2 μm mesh polycarbonate filter membrane and stored frozen (-80°C) for interrogation of the background microbial community from cove water (note that this community may have been influenced by aquarium conditions, although the relatively high flushing rate of the aquarium would minimize this). A decade-long study suggests that a stable euphotic zone core microbial community persists at the nearby San Pedro Ocean Time Series (Chow et al., 2013).

To examine the deposition of secondary minerals on the colonization materials, scanning electron microscopy (SEM) was performed using a Hitachi Model TM-1000 (WIES) at magnifications of 800–7000× immediately after collection of HFMC substrates and using a JSM-7001F analytical field emission scanning electron microscope at the Center for Electron Microscopy and Microanalysis at the University of Southern California. An ethanol dehydration series (70% for 12 h, 95% for 12 h, and 100% for 1 h), followed by oven drying (65°C) for 2 h, with no cell fixation procedure was followed prior to exposing samples to a vacuum.

Fluids (∼1.2 ml) contained within 1-m long sections of the chemical collection coil from the AF-FLOCS (Figure 2C) were expelled into acid-cleaned 1.5 ml plastic tubes following procedures outlined elsewhere (Wheat et al., 2011). Time stamps for individual samples are based on the number of samples, length of deployment and pumping rate. Temperature, salinity gradient, and surface area of the semipermeable membrane ultimately determine the pumping rate (Jannasch et al., 2004; Wheat et al., 2011). These pumps were exposed to air temperature fluctuations in the laboratory for the duration of the deployment, ranging across several degrees, which caused minor variations in the pumping rate over time. We recorded an average air temperature of 20.5°C in the laboratory throughout the course of the experiment, which would have resulted in an average pumping rate of ∼1 mL d^-1. Samples were then stored and shipped at 4°C for Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES) for major ions (Ca, Mg, K, S) and minor ions (Fe) following established protocols (Wheat et al., 2010).

Approximately 3 g of each substrate underwent total DNA extraction using the FastDNA^® Spin Kit for Soil (MP Biomedicals) following the manufacturer’s protocol (∼500 mg per extraction, six extraction tubes per sample). Each extraction was eluted in 75 μl of DES. Combined extracts were quantified in a NanoDrop 1000 Spectrophotometer and sent for library preparation and DNA sequencing by a commercial vendor (Molecular Research LP; MR DNA; Shallowater, TX, USA). Illumina MiSeq paired-end (2 × 250 bp) sequencing was performed targeting the V4 region of the 16S rRNA gene using the Earth Microbiome Project universal primers 515f (5′-GTG CCA GCM GCC GCG GTA A) and 806r (5′-GGA CTA CHV GGG TWT CTA AT) with 8-base barcodes on the forward primer (Caporaso et al., 2012). Briefly, amplification was carried out in a 30 cycle PCR using the HotStarTaq Plus Master Mix Kit (Qiagen, USA) under the following conditions as recommended elsewhere (Santiago-Rodriguez et al., 2015): 94°C for 3 min, followed by 28 cycles of 94°C for 30 s, 53°C for 40 s and 72°C for 1 min, after which a final elongation step at 72°C for 5 min was performed. After amplification, PCR products were analyzed on a 2% agarose gel via electrophoresis to determine the success of amplification and the relative intensity of bands. Multiple samples were pooled together in equal proportions and purified using calibrated Ampure XP beads. The pooled and purified PCR product was used to prepare the DNA library by following the Illumina TruSeq DNA library preparation protocol².

Sequence data curation and processing were performed with mothur v.1.34.4 (Schloss et al., 2009) following the mothur Illumina MiSeq Standard Operating Procedure (Kozich et al., 2013). In brief, paired reads were merged and any sequences with ambiguous base calls or homopolymers longer than 8 bp were culled. Merged reads were aligned to the mothur-recreated Silva SEED database from v119 (Yarza et al., 2010). Sequences were pre-clustered at a near 1% dissimilarity using the pre.cluster command with differences = 2. This process ranks sequences by abundance, then merges the most rare with the most abundant if they differ by only 2 bp as this has been shown to mitigate the generation of spurious sequences (Kozich et al., 2013). This pre-clustered dataset was, screened for chimeras, using the de novo mode of UCHIME (Edgar et al., 2011), which were then removed from any further processing and analysis. A distance matrix was generated for the remaining sequences and they were subsequently clustered into operational taxonomical units (OTUs) at 3% or less sequence dissimilarity using the average neighbor method. OTUs were taxonomically classified within mothur using the Ribosomal Database Project release 9 dataset (Cole et al., 2014).

High abundance OTU sequences were aligned against the NCBI Bacteria/Archaea database using the blastn algorithm. High similarity near-full length sequences retrieved from this database were aligned with selected high abundance OTUs using MUSCLE (Edgar, 2004). A phylogenetic tree was generated using the UPGMA method, using 1000 bootstraps for branching support, and branches cladogram transformed with the Geneious software package (Kearse et al., 2012). Pairwise community comparisons for shared community composition and species richness were performed on subsampled datasets standardized to equal sizes (n = 11,763 for each sample) using the following mothur calculators: J_est, Kulczynski, and Anderberg, for similarity in community membership analysis, in addition to the Bray–Curtis calculator, which analyzes community structure (Schloss, 2009). Results of each metric were compared to examine microbial community trends. As our data here were more driven by differences in abundances than by presence/absence, we utilized the Bray–Curtis dissimilarity in applicable downstream functions. Visualizations of the OTU abundance matrix were generated with RStudio version 0.98.1091 (Racine, 2012) using the packages vegan version 2.3-0 (Oksanen et al., 2015) and rgl version 0.95.1201. Principle components analysis was performed with the prcomp() function on a Bray–Curtis dissimilarity matrix calculated with the vegsdist() command. Sequence data were submitted to Genbank under Bioproject number PRJNA342057 and FASTA formatted sequences are available under the following accession numbers: KAHL01000001–KAHL01020009.

Scanning transmission X-ray microscopy (STXM) analysis was conducted at the 5.3.2.2 beamline, Advanced Light Source, Berkeley, CA, USA (Kilcoyne et al., 2003). This beamline is well-suited for investigations of nano-sized natural particles composed of organic and inorganic carbon and iron oxide and oxyhydroxide minerals. STXM was used to collect three types of data: (1) transmission images of 10 μm²–1 mm² areas, (2) elemental maps of C and O, and (3) C 1s and O 1s X-ray absorption near edge structure (XANES) spectra for points, lines, or areas (Toner et al., 2016). Elemental maps of C and O reveal morphology and the co-location of elements. Carbon 1s XANES spectra distinguish among organic and inorganic compounds while O 1s spectra are sensitive to several forms of Fe oxides and oxyhydroxides (Brandes et al., 2010; Chan et al., 2011; Bennett et al., 2014).

Incubated mineral chips from the NE-FLOCS-Pyrite and NE-FLOCS-Mild Steel samples (Table 1) were placed in clean tubes, gently rinsed with 18 MΩ grade deionized water (Milli-Q), and vortexed to suspend loosely adhered materials from the chips. Around 100 μL of the suspension was transferred to a new tube to dilute and wash any sea salt from the particles. Approximately 1 μL of suspension was deposited on a silicon nitride membrane and dried under ambient conditions. Once dry, samples were placed in the helium flushed STXM chamber. All data analysis—alignment of image stacks, principle component and cluster analysis, and normalization of spectra—was conducted using the freely available software axis2000³. Normalized spectra were compared to a reference database of C 1s and O 1s standards for species identification. Our C 1s reference database contains spectra of a lipid, PE lipid, BSA, agarose, and calcium carbonate. For O 1s we had access to reference spectra for various iron oxyhydroxides, but no iron oxide. A description of the reference materials for the C 1s and O 1s are published elsewhere (Toner et al., 2009, 2016; Chan et al., 2011). Distinct peak(s) and shape of O 1s iron oxides spectra were compared to data published elsewhere (Park et al., 2008).

The high flow (HFMC) experiment was subsampled and photographed periodically, revealing the steady build up of rust-colored particles in the colonization chamber over time (Figures 3A–D). The FLOCS colonization chambers exhibited visual changes related to metal corrosion (Figure 4A). The nutrient-enriched (NE-FLOCS) deployment was completely filled with black colored material and smelled strongly of hydrogen sulfide when opened, suggesting anoxic conditions inside the chamber at the time of collection. The active flow (AF-FLOCS) colonization chamber exhibited bands of rusty red and black zones that were more pronounced near the inlet on the mild steel substrate. The extent of rust accumulation was significantly lower in the active flow deployment, where active osmotic pumping of seawater was continuous, relative to stagnant condition (SC-FLOCS), where no osmotic pumping occurred. The stagnant condition exhibited profuse accumulations of rust-like material that was particularly accentuated over the mild steel cassettes located near the inlet of the colonization chamber. This rust-like precipitate exited the system via the inlet and accumulated on the aquarium bottom (see Figure 2B for schematic), creating an iron microbial mat-like deposit with iron oxides at the seawater interface and an apparent black anoxic zone a few millimeters below.

Scanning electron microscopy imaging of colonized mild steel substrates from the high flow experiment at 2 or 3 week intervals revealed an increasing abundance of twisted stalk particles (Figures 3A–D) that followed a steady increase in the DNA concentration recovered from nucleic acid extractions from ∼500 mg of sample (Figure 3E). Similar twisted stalk particles are known to be biogenic precipitates of neutrophilic FeOB (Chan et al., 2011). In contrast, the mild steel and pyrite substrates from the FLOCS experiments did not reveal such abundances of twisted stalk particles, but they did show evidence of amorphous alteration (Figure 5). Overall, the mild steel substrates from all three FLOCS treatments (Figures 5D,F,G) had more evidence of alteration than the pyrite substrates (Figures 5A–C). The stagnant condition pyrite substrate (Figure 5C) exhibited far less alteration as compared to the pyrite from the active-flow FLOCS (Figure 5B), whereas the mild steel substrate from the stagnant condition (Figure 5G) had similar levels of alteration to the mild steel substrates from the other treatments (Figures 5E,F). In some cases, surface pits in the colonized minerals from the FLOCS appeared to contain structures resembling degraded twisted stalks (Figures 4B–E).

A time series of the concentrations of major and minor ions in seawater was analyzed from the effluent from the AF-FLOCS only, as other experiments did not have the acid preservation of the fluid samples for analysis, as recommended (Wheat et al., 2011). All conservative, non-reactive major ions (i.e., Ca, Mg, K) exhibited steady concentrations throughout the experiment (data not shown). Dissolved sulfur, which could represent the sulfate ion from seawater or hydrogen sulfide oxidized to sulfate, exhibited an initial spike in concentration to 31.9 mmol kg^-1 as compared to background seawater concentrations of 28 mmol kg^-1, likely reflecting leaching of sulfide from the pyrite in the colonization chamber (Figure 9). For the remainder of the experiment, S remains fairly steady. Total iron also spiked at the beginning of the experiment, followed by a gradual decline to concentrations below the limit of detection after 3 months of incubation (Figure 9).

Scanning transmission X-ray microscopy analysis was performed on the NE-FLOCS only, due to constraints in beamline availability. The C1s XANES spectra are consistent with organic C, and the best spectral match with our reference database is a lipid compound [phosphatidylethanolamine (PE)]. It is unlikely that this organic matter was derived from the marine broth residue, as the samples were serially rinsed during preparation. In some instances, C was observed but the spectra lack identifying features (Figure 10A); this could be caused by a low concentration of C in an optically thick sample. Oxygen 1s XANES spectra for both the mild steel and pyrite did not match any spectra in our reference database. Instead, the spectral shapes and peaks best aligned with magnetite (Fe₃O₄) nanoparticles reported in Park et al. (2008) (Figure 10B). Two distinct peaks, ∼531 and ∼541 eV, were observed in magnetite nanoparticles that were distinct in shape and resolution from bulk magnetite. The first peak at ∼531 eV is broad in shape with only one absorption peak, with the second peak having a low energy shoulder with a main peak at ∼541 eV. These features are consistent with the electron configuration of Fe interacting with the oxygen in magnetite (Park et al., 2008). For both the pyrite and mild steel substrates, elemental maps and three-dimensional stacks were analyzed on at least three representative particles. C 1s and O 1s were measured to investigate speciation and understand particle morphology. Sample elemental maps indicated homogenous carbon-rich matrices surrounding pyrite particles and floccose distinct particles, rich in both carbon and oxygen, present on mild steel (Figure 11). For both the pyrite and mild steel incubations, ubiquitous carbon-rich matrices were observed.

During high Fe conditions in the FLOCS experiments (first month of the deployment, Figure 9), Fe and oxygen gradients at the interphase of both mild steel and, possibly, pyrite may have recruited FeOB as observed in our 2-month study (discussed below). Therefore, during this initial incubation stage, sulfate rises above its normal concentration in seawater (background sulfate + substrate S-species leachate, Figure 9), since microorganisms present in the colonization chamber, as suggested by microscopy results (Figures 3A–D), likely belong to Fe-rather than S-utilizing guilds. The decrease of Fe begins after approximately 4 weeks (Figure 9). This observation can be explained by extensive biological Fe-oxidation activity preventing further release of Fe²⁺ from substrates, ultimately disturbing the Fe and oxygen gradients necessary to favor the metabolism of FeOB (Edwards et al., 2003b; Emerson et al., 2010). We hypothesize that, in long-term FLOCS incubations, early FeOB proliferation leads to a relatively oxygen depleted niche (due to FeOB metabolism and slow FLOCS osmotic pump rates) that, as observed by co-stabilization of S within the timeframe of Fe depletion (Figure 9), begins to favor anaerobic S-cycling microbial cohorts, whose presence and predominance we confirmed on FLOCS substrates at the end of a 6-month deployment (Figure 6). Additionally, H₂S, the product of S-reduction potentially occurring in anoxic microniches early in FLOCS, can react with Fe²⁺ and form ferrous sulfide, effectively removing both S and Fe species from solution and explaining their co-maxima followed by stabilization to background seawater levels (Figure 9).

A full review of the complex ecological mechanisms responsible for marine MIC is beyond the scope of this work; however, this process may be briefly summarized as the microbiological influence of electrochemical or physical states at the interface of a metal surface and an aqueous layer by an exopolysaccharide-hosted, environmentally responsive and biologically dynamic biofilm (Von Wolzogen Kühr and van der Vlugt, 1934; Beech and Sunner, 2004; Beech and Campbell, 2008; Enning et al., 2012; Enning and Garrelfs, 2014). In this study, we aimed to assess the microbiological-chemical-physical interplay associated with ALWC, a phenomenon known to induce severe corrosion of steel pilings (Gubner and Beech, 1999).

Iron corrosion in anaerobic environments is understood to occur via two mechanisms: (1) an indirect mechanism where a chemical reaction occurs with hydrogen sulfide (Fe + H₂S → FeS + H₂); and (2) a direct mechanism where cathodic hydrogen reacts with iron directly (4Fe + SO₄^2- + 4H₂O → FeS + 3Fe²⁺ + 8OH^-) (Dihn et al., 2004; Kakooei et al., 2012). Desulfovibrio spp., dominant in NE- and AF-FLOCS substrates (Table 2), are known to employ the direct mechanism for iron corrosion (Von Wolzogen Kühr and van der Vlugt, 1934; Hamilton, 2003; Dihn et al., 2004). This mechanism may facilitate the presence of reduced iron in an increasingly anoxic environment that, due to the concomitant reduction of sulfate to sulfide by SRB, may result in a lower pH at the metal-biofilm interface. A statistically significant decrease in pH beneath corrosion products of ALWC sites was previously documented (Gubner and Beech, 1999) and potentially explains the presence of the Burkholderiales order, with acidophilic Fe-oxidizing members (Fukushima et al., 2015), and Desulfobacter spp., capable of sulfate reduction coupled to metallic iron oxidation (Dihn et al., 2004), in AF-FLOCS substrates (Figure 7).

Sulfide oxidizing bacteria have been reported in statistically higher proportions at ALWC sites relative to standard MIC (Gubner and Beech, 1999). In the current study, relatively higher proportions of Epsilonproteobacteria sequences were recovered from NE-FLOCS (Figure 6) corroborating a potential link between eutrophic conditions and ALWC microbial community profiles. The majority of these sequences are classified as Arcobacter spp., a genus recently enriched from electrochemical sulfide oxidation studies of Port of Los Angeles sediment (Li and Nealson, 2015). We propose that this genus may play a role in the global phenomenon of ALWC and suggest further investigation into the microbiological-electrochemical role played by Arcobacter spp. in the potential establishment of a localized sulfur cycle likely responsible for ALWC.

Long-term incubations under nutritional enhancement, intended to simulate high iron and nutrient conditions associated with ALWC, revealed different microbial communities relative to non-nutritionally enhanced deployments (Figure 8). The high prevalence of Delta and Epsilonproteobacteria, specifically the Desulfovibrio and Arcobacter genera (Figure 6; Table 2), at end of a 6-month nutritionally enhanced enrichment (NE-FLOCS), relative to a non-nutritionally enhanced (AF-FLOCS) system, suggests that the environmental conditions associated with ALWC, result in: (i) scant evidence of Zetaproteobacteria-like “twisted stalk” structures such as those observed during short-term incubations, (ii) anoxic conditions, (iii) microbial communities capable of Fe and S cycling, and (iv) a disproportionately high amount of Arcobacter spp.- a genus indirectly (De Gusseme et al., 2009) and directly (Roalkvam et al., 2015) implicated in MIC. Our 2-month incubation study provides microscopic evidence of colonization by stalk-forming microorganisms that are hitherto only represented by Zetaproteobacteria in the marine environment. This is consistent with previous studies that have shown early colonization of mild steel by Zetaproteobacteria (Dang et al., 2011; McBeth and Emerson, 2016). It is likely that FeOB communities colonized our NE-FLOCS experiment at earlier times and seceded to other communities at later times under nutritional enrichment and associated redox (oxic/anoxic) transition.