Biotic manganese oxidation coupled with methane oxidation using a continuous-flow bioreactor system under marine conditions

Mn(IV) oxyhydroxides (hereafter, Mn oxides) are ubiquitous absorbents of various elements in aquatic environments (Dixon & Skinner 1992; Manceau et al. 1992; Post 1999; Takahashi et al. 2007; Sherman & Peacock 2010; Kashiwabara et al. 2011, 2013, 2014). In particular, marine ferromanganese crusts and nodules, which are massive deposits of Mn oxides widely found on the deep seafloor, have received great attention as potential metal resources because of their strong enrichment of many valuable metals from seawater (Glasby 2006; Hein et al. 2013).

Mn oxidation in the natural environment is believed to be dominated by biological activity (Myers & Nealson 1988; Tebo et al. 2004). A diverse array of bacteria and fungi are known to be Mn oxidizers (Tebo et al. 2005; Templeton et al. 2005; Miyata et al. 2006; Santelli et al. 2011). Biological Mn oxidation is induced by Mn-oxidizing enzymes (i.e., multicopper oxidases) (van Waasbergen et al. 1996; Dick et al. 2008; Tebo et al. 2010) or by superoxides as byproducts of metabolisms (Hansel et al. 2012; Learman et al. 2013). Biogenic Mn oxides (hereafter BioMnOx) have a greater reactivity than abiotic forms due to the larger surface area accompanied by their smaller size, the more vacancies for cations associated with their crystal structures, and the greater chemical complexity with coexisting organics (Hochella et al. 2008). Because of their structural and chemical properties, BioMnOx absorb significant amounts of trace metals such as Ni and Co that are normally present at low levels in the environments and are economically valuable elements (Tebo et al. 2004, 2010; Miyata et al. 2007; Hennebel et al. 2009). Therefore, BioMnOx have been studied regarding engineering applications such as the recovery of trace metals from wastewater and remediation of groundwater contaminated with toxic metals (Hennebel et al. 2009).

Recently, the use of a continuous-flow system called down-flow hanging sponge (DHS) bioreactor for continuous Mn removal and BioMnOx production under freshwater conditions has been reported (Cao et al. 2015). In this report, BioMnOx production coupled with nitrification was successful with no supply of organic matter. In the presence of high concentrations of organic compounds, on the other hand, BioMnOx production was not observed (i.e., no Mn-oxidizing microbes could be cultivated), suggesting that the nitrifier-derived organic compounds seem to be essential for BioMnOx production and the development of potentially oligotrophic Mn-oxidizing microbial populations. In addition, Cao et al. (2015) reported the effective removal of trace metals such as Ni and Co with the Mn precipitation. These results indicate a great potential for BioMnOx production using the DHS bioreactor system over conventional batch incubation systems because this system can effectively combine the continuous precipitation of Mn and recovery of trace metals from the influent.

The mass of seawater is much greater (over 97% of the total water mass on the Earth) than that of freshwater. If BioMnOx production using a continuous-flow system can be achieved under marine conditions, there is the possibility of recovering trace metals (such as Ni and Co) continuously from seawater. Although numerous studies on recovery of metals including Ni and Co from seawater have been reported (reviewed in Diallo et al. (2015)), no continuous-flow bioreactor system for BioMnOx production and metal recovery has been established under marine conditions to date. Here, we report Mn removal and BioMnOx production during the long-term operation (over 2 years) of a DHS reactor under marine conditions. To this end, according to the cultivation strategy reported previously (Cao et al. 2015), we attempted to cultivate marine Mn-oxidizing microorganisms coupled with nitrification in a reactor system amended with ammonium, in which soluble microbial organic products were provided by nitrifiers to concurrent Mn oxidizers as the substrates. We also set up another DHS bioreactor to cultivate Mn oxidizers coupled with methane oxidation instead of nitrification. We were only able to successfully enrich a BioMnOx-producing microbial community from a deep-sea sediment sample with the methane-fed DHS bioreactor. In addition, the methanotroph-based microbial community capable of BioMnOx production during the enrichment was characterized.

We used a deep-sea sediment sample, which was collected from the Tarama Knoll, southern Okinawa Trough, Japan (25°5.576′N, 124°32.377′E), at a depth of 1,540 m, by the remotely operated vehicle Hyper-Dolphin (cruise NT10-06, dive no. 1108, April 10, 2010) (Makita et al. 2016), as the inoculum for the operation. Just before the operation of reactors A and M, the sponge cubes were soaked in a suspension of the sediment sample (Figure S1, available with the online version of this paper). It is noted that NH₄⁺ and CH₄ were detected in the seawater and shimmering fluid samples obtained from the Tarama Knoll (Yamanaka et al. 2015). Following the inoculation, the PVC columns were tightly closed and installed in an incubation chamber (M-600FN, Taitech, Koshigaya, Japan) in the dark at 15 °C.

The operational conditions used in this study are summarized in Table 1. The composition of the medium was as follows (L⁻¹, the values for x were adjusted at each operation phase; see Table 1): x g MnCl₂·4H₂O, x g (NH₄)₂SO₄, x g KNO₃, 0.04 g KH₂PO₄, 1.33 g CaCl₂·4H₂O, 3.4 g MgSO₄·7H₂O, 4.18 g MgCl₂·6H₂O, 0.33 g KCl, 24 g NaCl, 0.084 g NaHCO₃, 1 mL trace mineral element solution, and 1 mL vitamin solution. The trace mineral element solution contained the following (L⁻¹): 2.5 g EDTA·2Na, 5 g MnSO₄·2H₂O, 5 g CoSO₄·7H₂O, 1.8 g ZnSO₄·7H₂O, 0.1 g CuSO₄·5H₂O, 0.2 g KAl(SO₄)₂·12H₂O, 0.1 g H₃BO₃, 0.01 g Na₂MoO₄·2H₂O, 0.1 g SrCl₂·6H₂O, 0.1 g NaBr, 0.1 g KI, 1 g FeCl₂·4H₂O, 0.25 g NiCl₂·6H₂O, 0.0017 g Na₂SeO₃, and 0.0033 g Na₂WO₄·2H₂O. The vitamin solution was composed of the following vitamins (L⁻¹): 2 mg biotin, 2 mg folic acid, 10 mg pyridoxine hydrochloride, 5 mg thiamine hydrochloride, 5 mg riboflavin, 5 mg nicotinic acid, 5 mg dl-calcium pantothenate, 0.1 mg vitamin B₁₂, 5 mg p-aminobenzoic acid, and 5 mg lipoic acid. The salinity of the medium showed 3.55%, which is the same level of that of natural seawater. Although we did not measure the salinity of the original sediment samples, the salinity level is probably very similar to that of natural seawater. The pH of the medium was adjusted to 7 or 7.5. A total of 10 L of medium was made at a time. The 10 L medium bottle was stored at 15 °C in the sample incubation chamber. The influent medium was supplied into the reactor column from the top inlet port by a peristaltic pump (Masterflex L/S tubing pump 7550-50, Cole-Parmer, Vernon Hills, IL, USA). The medium then flowed down, passing through the sponge cubes by gravity, and was finally pumped out of the PVC column. The sponge cubes were not submerged, and hung freely in the atmosphere. Air (for reactor A) or a gas mixture of air and CH₄ (final 2% or 4% of CH₄, for reactor M) was packed into an aluminum bag (AAK-10, ASONE, Osaka, Japan) and supplied to the lower part of the bioreactor column via a peristaltic pump. The medium and gases were supplied intermittently at 1 min/9 min (on/off) regulated by an automatic on/off timer (FT-011, Tokyo Glass Kikai Co. Ltd, Tokyo, Japan) connected to the peristaltic pump. The HRT and GRT were changed by altering the flow rate of the peristaltic pump. The bioreactors were operated under atmospheric pressure.

The influent and effluent were routinely sampled to characterize the chemical composition of the medium related to Mn(II) oxidation. Prior to analysis, all the water samples were filtered using 0.22-μm pore size polyvinylidene difluoride membrane filter discs (Merck Millipore, Darmstadt, Germany). The dissolved Mn(II) was measured with the periodate oxidation method using a Hach water quality analyzer (DR-2500; Hach Co., Loveland, CO, USA). NH₄⁺-N concentration was determined by the Nessler method using the DR-2500 Hach water quality analyzer. Concentrations of NO₂⁻-N and NO₃⁻-N were measured by high-performance liquid chromatography using a TSKgel SAX column (Tosoh Co., Shunan, Japan; eluent, 1.0 N NaCl; column temperature was room temperature) and a UV-VIS detector (GL-7451; GL Science, Tokyo, Japan), as described previously (Maruo et al. 2006). Concentrations of CH₄ and CO₂ in the off-gas were determined by gas chromatography (GC3200G; GL Science) with a thermal conductivity detector. O₂ concentrations in the off-gas were monitored using an oxygen concentration meter (JKO-25LII; JIKCO, Tokyo, Japan) that was installed on the off-gas line. The pH of the influent and effluent water was measured using a portable pH meter (Five Go FG2; Mettler Toledo, Columbus, OH, USA).

Blackish materials produced on the sponges were collected, freeze-dried, and powdered using an agate mortar. The resultant powders were mounted on a glass holder and analyzed using an X-ray diffraction (XRD) analyzer (MiniflexII, Rigaku Co., Tokyo, Japan) with Cu Kα radiation operated at 30 kV, 15 mA. Signals were collected from 3° to 90° (2θ) with a scan speed of 4.0° min⁻¹ and a scan step of 0.02°.

Genomic DNA was extracted from several sponge cubes randomly selected from two or three portions of the upper, middle, and lower parts of the DHS reactors and the original inoculum. The 16S rRNA gene clone analysis was performed as previously described (Imachi et al. 2011). In brief, bacterial and archaeal 16S rRNA genes were amplified by polymerase chain reaction (PCR) with the EUB338F*/1492R and Arc21f/Ar912R primer sets, respectively. The PCR products were cloned, and randomly selected clones (approximately 100 clones for each library) were sequenced. Based on the 16S rRNA gene sequences, OTUs (operational taxonomic units) were defined at a 97% similarity level, and the coverage for each library and number of OTUs shared among the libraries were calculated using mothur 1.36.1 (Schloss et al. 2009). The potential microbial community structures based on the 16S rRNA gene phylotype compositions were determined using QIIME 1.9.1 (Caporaso et al. 2010). The sequences of representative clones for the OTUs, and relatives of environmental clones and cultured species were aligned using Muscle version 3.8.31 (Edgar 2004). Gap positions were removed from the alignment using TrimAl version 1.2 (Capella-Gutierrez et al. 2009). Using the alignment without gaps, phylogenetic trees were constructed using FastTree version 2.1.7 (Price et al. 2010) with the Jukes–Cantor nucleotide substitution model. The 16S rRNA gene sequences reported in this study have been deposited in the DNA Data Bank of Japan under accession numbers LC140542–LC140671.

Quantitative real-time PCR (Q-PCR) was performed using SYBR Green I fluorescent dye as previously described (Aoki et al. 2014). The numbers of cells were estimated from the Q-PCR results based on the averages of 4.12 copies cell⁻¹ for bacteria and 1.63 copies cell⁻¹ for archaea, respectively, reported in a curated database of ribosomal RNA operons (rrnDB) version 4.4.4 (Stoddard et al. 2015).

To obtain scanning electron microscopy (SEM) images of microbial cells attached to the sponges, a portion of a sponge was prefixed in 2.5% (w/v) glutaraldehyde in filtered artificial seawater (FASW) at 15 °C. After being cut with a cleaned blade, and washed in FASW, the sponges were stuck on a poly-l-lysine-coated glass slide for 30 min at room temperature. Following this treatment, the sponges were fixed in 2% (w/v) osmium tetroxide dissolved in FASW. After the samples were rinsed with distilled water, conductive staining was performed by incubation with 0.2% aqueous tannic acid (pH 6.8) for 30 min, and the samples were then washed with distilled water and treated with 1% aqueous osmium tetroxide for 1 hour. After that, the samples were dehydrated in a graded ethanol series and critical-point-dried in a JEOL JCPD-5 (Tokyo, Japan). The samples were coated with osmium using an osmium plasma coater (POC-3; Meiwafosis Co., Ltd, Tokyo, Japan) and observed with a JEOL JSM-6700F field emission scanning electron microscope operated at 5 kV.

The decrease in NH₄⁺ was stoichiometrically consistent with the increase in NO₂⁻ (Figure S2). It has been reported that a neutral or an alkaline pH is favorable for chemical and biological Mn(II) oxidation (Hallberg & Johnson 2005). Therefore, to prevent the decrease in pH in the medium by nitrification, a Good's buffer of 3-morpholinopropanesulfonic acid (MOPS) was added into the influent medium at phase 2. Active ammonium oxidation was observed continuously through phase 2, but the pH value in the effluent was maintained around 6.8. However, Mn(II) removal was not observed in phase 2. In addition, the Mn²⁺ concentrations in the effluent were higher than those in the influent. The reason for this difference is not clear, but the MOPS buffer might affect the variation of Mn²⁺ concentrations in the effluent. Therefore, the addition of MOPS was stopped at the end of phase 2.

The pH of the effluent was stable around 6.2 at the later period of phase 4. To increase the effluent pH value, the pH of the influent medium was changed to 7.5 in phase 5. As a result, the effluent pH slightly increased to 6.3–6.5. However, no apparent Mn removal was observed at the end of phase 5. Because it was also expected that a high concentration of Mn(II) could inhibit the enrichment of Mn(II)-oxidizing microbial populations, the Mn(II) concentration in the influent medium was reduced in phase 6. Although reactor A was operated for 161 days in phase 6, Mn removal and blackish precipitations indicative of Mn oxides were not observed. In phase 7, the HRT was returned to 4.5 hours to extend the reaction time with the microbial cells and Mn(II). The oxidation of NH₄⁺ to NO₃⁻ was continuously observed during phases 5 to 7. However, the production of BioMnOx was not observed during the operation of reactor A for 636 days. Therefore, we abandoned the production of Mn oxides using reactor A.

In fact, an elevated CO₂ concentration was observed in the off-gas (Figure 2). The produced CO₂ was likely partially dissolved into the liquid medium in the reactor, which is consistent with the pH decrease observed. To enhance the production of organic compounds from methane oxidation, the GRT was shortened, and the partial pressure of CH₄ in the input-gas was increased (from 2% to 4%) in the phases 3 to 5. Because NO₃⁻ did not seem to be utilized as the nitrogen source of the microbial community, the NO₃⁻ addition to the influent medium was stopped in phases 5 and 6. At the later period of phase 5, the pH was stable at 6.3. The decrease in the NH₄⁺ concentration in the effluent was probably due to microbial uptake and assimilation because there was very little production of NO₂⁻ and NO₃⁻. In phase 6, the pH of the influent medium was increased to 7.5. The pH of the effluent was slightly increased to 6.4–6.5. CH₄ consumption was continuously observed, but Mn(II) removal was not observed. It is unclear why a transient removal of Mn(II) was observed just after the start of phase 6 (Figure 2). Furthermore, when the GRT was more shortened in phase 7, Mn(II) removal was not observed.

Because the high concentration of Mn(II) was presumed to be inhibitory to the development of Mn(II)-oxidizing microbial populations, the Mn(II) concentration in the influent medium was reduced from 118 to 27 μM in phase 8. NO₃⁻ was added again to the medium as a nitrogen source, and the GRT was reduced in phase 8. Finally, Mn(II) removal was evident during phase 8 (Figure 2), and the color of the sponges gradually became blackish (Figure 1). These findings strongly suggested the occurrence of BioMnOx production in the bioreactor. The rate of Mn removal was continuously positive (average, 3.6 mg L⁻¹ day⁻¹; maximum, 7.6 mg L⁻¹ day⁻¹; Figures 2 and 3). In phase 9, the NH₄⁺ addition to the influent medium was stopped to assess the effect of NH₄⁺ on the Mn removal rate. As a result, the Mn removal rate slightly decreased (average, 3.3 mg L⁻¹ day⁻¹; maximum, 7.0 mg L⁻¹ day⁻¹), suggesting that NH₄⁺ is important for BioMnOx production (potentially as the nitrogen source for Mn oxidizers). In phase 10, the NH₄⁺ addition to the influent medium was resumed, and the Mn removal rate increased (average, 5.4 mg L⁻¹ day⁻¹; maximum, 11.4 mg L⁻¹ day⁻¹). The Mn removal rates observed in the reactor M are slower than that of the previous freshwater operation (constantly over 20 mg L⁻¹ day⁻¹; maximum, 48 mg L⁻¹ day⁻¹) (Cao et al. 2015).

The OTUs classified in Deltaproteobacteria (putative sulfate/metal reducers) and in Zetaproteobacteria (putative iron oxidizers) were the first and second most abundant in the bacterial clone library, respectively. The OTUs classified in Methylococcales of Gammaproteobacteria (putative methanotrophs) were also relatively abundant. The OTUs classified in Nitrospirae and Nitrospinae (both putative nitrite oxidizers) and Nitrosomonadales of Betaproteobacteria (putative ammonia oxidizers) were detected. OTUs classified in Planctomycetes, Alphaproteobacteria, and Gammaproteobacteria (putative heterotrophs) were also detected. An OTU related to Nitrosopumilus spp. in Thaumarchaeota (putative ammonia oxidizers) dominated in the archaeal clone library. The bacterial 16S rRNA gene phylotype composition in the original inoculum was quite different from those of the microbial communities colonizing sponges collected during the operation (Figure 5(a), as supported by the comparative analysis (Figure 5(b)). Remarkably, only one OTU (OTUb17 in Nitrospinae) in the original inoculum (OriB) was found in the bacterial phylotype compositions during the operation of reactor A (ATB, AMB and ABB), and no OTU of OriB was found in the bacterial phylotype composition in reactor M (MTB and MBB) (Figure 5(c)). The results indicate that the bacterial populations represented by the OTUs during the reactor operation were not just a remnant of the inoculum, but actively grew in the reactors.

At day 532, we collected the top, middle and bottom parts of the string of the sponges and determined the bacterial and archaeal community compositions (Figure 5(a)). An OTU classified in Nitrosomonadales of Betaproteobacteria, OTUb84, was the most abundant in the bacterial clone libraries for all parts (ATB, AMB, and ABB) (Figure S3). This OTU was closely related to ammonia-oxidizing autotrophic bacteria, such as Nitrosomonas spp. and Nitrosococcus spp. OTUb86 in Nitrospinae was also relatively abundant in the libraries (Figure S5). One OTU (OTUa1) related to an ammonia-oxidizing archaeon, Nitrosopumilus maritimus, was recovered from the archaeal clone libraries constructed from the middle and bottom parts (AMA and ABA) in reactor A. These OTUs related to Nitrosomonadales, Nitrospinae and Nitrosopumilus likely represent ammonia/nitrite oxidizers, and contribute to Reactions (1) and (2), respectively, observed in reactor A as described above. In addition, these putative autotrophs likely fix inorganic carbon into organic carbon and support the growth of heterotrophs in the reactor as the primary producers.

OTUs classified in Bacteroidetes, Phycisphaerales of Planctomycetes, and Kiloniellales and Rhodobacterales of Alphaproteobacteria (Figures S4–S6), which likely represent heterotrophs, were relatively abundant in the libraries (3.0–12.9% of the total number of clones) for all parts. These putative heterotrophs would use the organic carbon produced by the autotrophic ammonia/nitrite oxidizers as described above. To date, diverse heterotrophic Mn-oxidizing bacteria including members of Rhodobacterales and Pseudomonadales have been reported (Tebo et al. 2005; Templeton et al. 2005). However, all the sequences of Rhodobacterales and Pseudomonadales OTUs found in reactor A were distinct from those of previously known Mn-oxidizing bacteria (Figures S3 and S4). In addition, it is difficult to infer the ability of Mn oxidation from 16S rRNA gene sequence information because the ability is not common even among the same genus and/or species (for example, Pseudomonas species in Figure S3 (Francis & Tebo 2001)). Combined with the fact that no Mn removal was observed in reactor A after the long-term cultivation with various operational conditions, the host microorganisms of Rhodobacterales and Pseudomonadales OTUs are probably not Mn-oxidizing bacteria.

In the previous study, BioMnOx production was observed using a DHS reactor under freshwater conditions (total of 449 days of operation) (Cao et al. 2015), but in the present study, BioMnOx production was not observed in reactor A under the seawater conditions (total of 636 days), although NH₄⁺ was used as the sole energy source for both reactors. The microbial communities were clearly different between the previous freshwater and the present seawater conditions (Figure 5(a) and (b)). No OTU was common between the freshwater and seawater reactors (Figure 5(c)). The difference in the microbial community development is due to the different inoculum sources and enrichment conditions, which ultimately led to the unsuccessful BioMnOx production in our DHS reactor under the marine conditions.

An OTU classified in Thiotrichales of Gammaproteobacteria, OTUb87, was the most abundant (>50% of the total number of clones for each library) in the bacterial libraries (MTB and MBB) of reactor M. The closest cultured species to the OTUb87, with <95% of the 16S rRNA gene sequence similarity, was Methylophaga spp., which are methylotrophs that use one-carbon compounds (e.g., methanol, formaldehyde, and methylamine) except methane as their sole carbon and energy sources (Lidstrom 2006). Considering its relative abundance and the incubation conditions of reactor M, the host bacterium of OTUb87 can be regarded as a methanotroph. However, the same sequence was found in reactor A and no cultured methanotroph species have been reported in Thiotrichales. Therefore, the host bacterium may use methanol provided by other methanotrophs as described below.

An OTU classified in Methylococcaceae of Gammaproteobacteria, OTUb118, was relatively abundant in the libraries. The closest cultured species to the OTUb118, with 98% of the 16S rRNA gene sequence similarity, was Methylomonas spp., which are methanotrophs that use methane and methanol as the sole carbon and energy sources (Bowman 2015). Thus, the host bacterium of OTUb118 could be a methanotroph, which could grow in reactor M using methane and sustain the other concomitant other heterotrophs including Mn oxidizers as a primary producer. No sequences related to methanotrophs were detected in reactor A.

An OTU classified in Rhodospirillales of Alphaproteobacteria, OTUb115, was relatively abundant (5.5%–25.0% of the total number of clones) in the libraries of reactor M but was rarely detected in reactor A (Figure S4). The closest cultured species to the OTUb115, with 99% of the 16S rRNA gene sequence similarity, was Nisaea spp., which are heterotrophs including facultative anaerobes. As observed in reactor A, OTUs classified in Bacteroidetes, and Kiloniellales and Rhodobacterales of Alphaproteobacteria, which likely represent heterotrophs, were also detected in the libraries for reactor M (1.1%–6.4% of the total number of clones). The putative heterotrophs of these OTUs can grow using the organic compounds from the methanotrophic primary producers. In contrast, OTUs in Nitrosomonadales of Betaproteobacteria (putative ammonia oxidizers) and in Nitrospinae (putative nitrite oxidizers) were not detected in reactor M. These results indicate that in reactor M, the methanotrophic populations overcame the populations of ammonium oxidizers, and NH₄⁺ in the medium was preferentially consumed, probably by the nitrogen assimilation, by abundant methano/methylotrophic and heterotrophic populations rather than by the dissimilatory nitrogen metabolism of nitrifiers, which is consistent with the chemical data (Figure 2).

OTUa1 related to the ammonium-oxidizing archaeon Nitrosopumilus maritimus was the most abundant in the archaeal 16S rRNA gene clone libraries. The archaeal 16S rRNA gene abundance was estimated to account for 0.3–2.1% of the total number of cells in reactor M (Table 2), indicating that the detected archaeal members could actively grow in the reactor and represent a portion of the microbial community. Thus, it can be concluded that bacterial populations including methanotrophs predominated in reactor M.

Phylogenetically diverse microorganisms were detected from reactor M. However, we could not identify Mn-oxidizing bacteria because identical or very close to 16S rRNA gene sequences of previously known Mn-oxidizing bacteria were not detected from the reactor. Therefore, at present, it is still uncertain which of the OTUs and their host microbes are responsible for the BioMnOx production in reactor M. To date, phylogenetically diverse bacteria have been shown to be capable of Mn oxidation by means of indigenous Mn-oxidizing enzymes or metabolic superoxides. Enzymatic Mn oxidation has been demonstrated in several bacteria, such as Bacillus spp. (Firmicutes), Pseudomonas spp. (Gammaproteobacteria), Arthrobacter spp. (Actinobacteria), and Leptothrix spp. (Betaproteobacteria) (Tebo et al. 2005). Mn oxidation by superoxides has been shown in Roseobacter spp. (Alphaproteobacteria) (Learman et al. 2011a). A recent report showed that diverse bacteria in various environments, including marine environments, can produce extracellular superoxides (Diaz et al. 2013), which can oxidize circumambient Mn. In addition, phylogenetically diverse Mn-oxidizing bacteria, such as Roseobacter spp. and Sulfitobacter spp. (Alphaproteobacteria), as well as Pseudoalteromonas spp. and Alteromonas spp. (Gammaproteobacteria), have been isolated from deep-sea basaltic rocks, although the mechanisms of their Mn oxidation are unclear (Templeton et al. 2005). Considering the presence of the phylogenetically diverse Mn-oxidizing bacteria, it is not surprising that there are yet-unknown Mn oxidizers in reactor M. Indeed, we could not exclude the possibility that undetected microorganisms are involved in the BioMnOx production, as the coverage values of the 16S rRNA gene clone analyses (Table 2) indicate the presence of undetected microbial populations in the microbial community of reactor M.

The bacterial communities found in reactor M (and also reactor A) were totally different from those of the previously described freshwater reactor (Cao et al. 2015). No common OTUs were found between the microbial communities (Figure 5(c)). This result indicates that the Mn-oxidizing population enriched in our seawater reactor differ from that in the freshwater reactor. The difference in salinity between the marine and freshwater reactors, in addition to the differences in inoculum source (deep-sea sediment vs. activated sludge) and the availability of CH₄, likely results in the difference in the microbial community structures. It should be noted that microorganisms responsible for BioMnOx in the previously described freshwater reactor are also unclear (Cao et al. 2015) as in the present study.

In this study, we successfully achieved Mn removal and Mn oxides production under the marine conditions by coupling with methane oxidation during the operation of reactor M. Although the detailed mechanism of Mn oxides production in reactor M is still unclear, our data strongly indicated that Mn oxides were produced by microorganisms in reactor M as supported by the following reasons: (i) positive correlation was found between manganese removal rate and methane-oxidation rate (Figure 3) and (ii) XRD analysis showed that blackish material formed in reactor M has a very similar pattern to birnessite that is normally produced by previously known Mn-oxidizing bacteria (Figure 4). Based on comparisons of the operation of reactors A and M, certain organic compounds produced by methanotrophs may activate yet-unknown Mn-oxidizing microbial populations, whereas the compounds produced by nitrifiers from the deep-sea sediment sample in the seawater reactor may not activate them, even though a freshwater DHS reactor was previously successfully shown to produce the BioMnOx with nitrification (Cao et al. 2015). Otherwise, methanotrophs themselves may oxidize Mn in reactor M, although we obtained no clear evidence to support this hypothesis. Further attempt at cultivation of Mn oxidizers from the reactor and the biochemical analysis of isolates will clarify which microbial components are responsible for the observed Mn oxidation and how the Mn oxidation is linked to methane oxidation. In the future, we will develop more effective continuous-flow bioreactor systems, with higher Mn removal rates, containing various marine microbial communities and demonstrate their adsorptive properties for the recovery of trace metals and their possible applications in engineering.

Shingo Kato and Masayuki Miyazaki contributed equally to this work. The authors would like to thank Katsuyuki Uematsu and Akihiro Tame for assistance in preparing the electron micrographs. We are very grateful to the ROV Hyper-Dolphin operation team and the crew of the R/V Natsushima, and Dr Hiroko Makita for helping us to collect the deep-sea sample. This study was supported by the Japan Society for the Promotion of Science (JSPS) KAKENHI grants (no. 24561008 to author M. M., no. 24687011 and 15H02419 to author H. I.) and by the Cabinet Office, Government of Japan, through the ‘next-generation technology for ocean resources exploration’ (called as Zipangu-in-the-ocean project) in the Cross-ministerial Strategic Innovation Promotion Program. The authors have no conflict of interest to declare.