Manganese Reduction by Microbes from Oxic Regions of the Lake Vanda (Antarctica) Water Column

Abstract

Depth profiles of metals in Lake Vanda, a permanently ice-covered, stratified Antarctic lake, suggest the importance of particulate manganese oxides in the scavenging, transport, and release of metals. Since manganese oxides can be solubilized by manganese-reducing bacteria, microbially mediated manganese reduction was investigated in Lake Vanda. Microbes concentrated from oxic regions of the water column, encompassing a peak of soluble manganese [Mn(II)], reduced synthetic manganese oxides (MnO₂) when incubated aerobically. Pure cultures of manganese-reducing bacteria were readily isolated from waters collected near the oxic Mn(II) peak. Based on phylogenetic analysis of the 16S rRNA gene sequence, most of the isolated manganese reducers belong to the genus Carnobacterium. Cultures of a phylogenetically representative strain of Carnobacterium reduced synthetic MnO₂ in the presence of sodium azide, as was seen in field assays. Unlike anaerobes that utilize manganese oxides as terminal electron acceptors in respiration, isolates of the genus Carnobacterium reduced Mn(IV) via a diffusible compound under oxic conditions. The release of adsorbed trace metals accompanying the solubilization of manganese oxides may provide populations of Carnobacterium with a source of nutrients in this extremely oligotrophic environment.

Although manganese is a minor chemical constituent of marine water and freshwater, the oxides of manganese may play a pivotal role in aquatic geochemical cycles (28). The suggested central role for manganese oxides is due to their remarkable surface area and charge distribution, which make them a potentially rich reservoir of adsorbed metals. Hydrous manganese oxides are often present in amorphous or microcrystalline forms which have surface areas as high as 300 m² g⁻¹ (51). The binding of a variety of metals to manganese oxides has been observed in several environments, including sediments and freshwater systems (3, 8, 15, 43). Microorganisms capable of reducing manganese oxides, thereby solubilizing the particulate manganese oxides and releasing the adsorbed metals, may have a major impact on the geochemical cycles of manganese and associated metal cations.

Some temperate lakes contain small peaks of manganese in the oxic portion of the water column, which cannot be adequately explained by chemical equilibria (36, 49). The depth profiles of metals that adsorb to manganese oxides, including cobalt and lead, covary with profiles for soluble manganese in some freshwater sediments and water columns (13, 27, 33, 36). In addition to lakes, reduced manganous ions have been detected in the Earth’s major ocean basins and numerous groundwater systems.

The predominant oxidation state of manganese in an environment is influenced by the chemical and physical characteristics of that environment. In reducing environments with a low pH, the thermodynamically favored form of manganese is as soluble manganous ions [Mn(II)]. Water-insoluble manganese oxides are favored under high-pH and oxidizing conditions. However, even when environmental conditions are such that the oxidized or reduced states are thermodynamically favored, the strictly chemical conversions between the stable forms of manganese proceed slowly without catalysis (19, 44). Consequently, redox reactions of manganese in many environments have been linked to microbial metabolism. Microbial manganese oxidation has been observed in numerous aquatic environments, particularly just above oxic-anoxic interfaces (18, 21). In Lake Vanda, Antarctica, the potential for manganese(II) oxidation occurs in regions of the water column abutting the two peaks of reduced manganese (5). The direct reduction of manganese(IV) generally requires cell contact with manganese oxides which serve as terminal electron acceptors for respiration (26, 38, 39, 45). Manganese oxides can also be reduced indirectly by microbial metabolites such as sulfide, a product of the sulfate-reducing bacteria, and organic acids produced by fermentative bacteria and cyanobacteria (10, 26, 39, 50).

Given the potential role of manganese oxides in metal cycling and the capacity for microorganisms to solubilize manganese oxides, we were interested in determining the role of microbes in the manganese cycle of Lake Vanda. Lake Vanda is approximately 5.6 km long, 1.5 km wide, and capped by a smooth, permanent, 4-m-thick ice cover. Glacial meltwater from the Onyx River flows into the upper regions of the water column for approximately 4 to 6 weeks during the austral summer and is the major source of both dissolved and particulate matter in the lake. Lake Vanda can be viewed as a meromictic, two-layer system. The upper 55 m is freshwater with concentrations of dissolved oxygen near saturation; below 55 m the water becomes increasingly saline and anoxic with depth. The present stratification appears to be the result of flooding (1,200 years ago) of a shallow, saline brine, by freshwaters from the Onyx River. Chemical profiles have developed since then, largely through molecular diffusion (12). The unusually stable water column of the lake results in a well-defined vertical structure that provides an excellent opportunity for exploring the interactions between microbial metabolism and the cycling of manganese and other elements. The stability of the vertical structure of the water column is due to the 4-m-thick ice cover, which prevents wind-driven mixing, and the dense calcium chloride brine in the bottom waters. The relatively clear ice of Lake Vanda has been shown to reduce the incident photosynthetically available radiation by about 86% (52). Although reduced in intensity, the photosynthetically available radiation penetrates to depths below 65 m in the oligotrophic waters of Lake Vanda.

Microbial manganese reduction in Lake Vanda was assessed during two field seasons in Antarctica. Microbes were isolated from throughout the oxic zone of Lake Vanda and the Onyx river and tested for the capacity to reduce manganese oxides. Bacteria capable of aerobic manganese reduction were isolated from oxic waters surrounding a peak of soluble manganese in Lake Vanda and characterized.

MATERIALS AND METHODS

Water sampling and analyses.

Samples were collected during two field seasons: the first field season was during austral spring (27 October to 12 December 1994), before the Onyx River formed from glacial meltwater. The second season was during austral summer (January 1996), while the Onyx River was flowing. The smooth 4-m-thick ice cover provided a stable platform for work on Lake Vanda and eliminated uncertainties in establishing sampling depth that are associated with open-water sampling. An ice auger with a 10-in.-diameter, 36-in.-long bit and three 3-ft-long extensions was used to drill through the ice. The primary sampling site was above the lake’s deepest western depression (ca. 75 m deep).

Water samples of less than 10 liters were collected in Beta or Kemmerer bottles (Wildlife Supply, Saginaw, Mich.). Samples larger than 10 liters were collected with a Masterflex peristaltic pump (Barnstead-Thermolyne, Dubuque, Iowa) and plastic Tygon tubing. Trace element grade HCl (1 M; Fisher Scientific, Hampton, N.H.) was continuously circulated through the tubing for a day and was followed by a rinse with deionized, quartz-distilled water to ensure metal-free sampling. During sampling, the tube was purged with at least 5 liters of water at each new depth before sample collection. Collection of samples for manganese analysis began at the top of the water column and progressed downward due to the general increase in metal concentration with depth.

Samples for soluble-manganese analysis were filtered through polycarbonate filters (pore size, 0.2 μm) that had been cleaned by methods described previously (13) and were transported to the field in individual, acid-washed plastic petri dishes. The filters were rinsed on site with the water sample before the collections. Water samples were preserved by adding 2 ml of trace element grade nitric acid per liter upon collection. The chloroform extraction technique of Landing and Bruland (34) was used to prepare samples from the upper water column for analysis of manganese concentrations in a Perkin-Elmer 3030 graphite furnace atomic absorption spectrophotometer. Samples from the deep waters were diluted 10- to 100-fold with deionized, quartz-distilled water and analyzed directly for manganese by graphite furnace atomic absorption spectroscopy.

Oxygen concentrations and temperature were measured in situ with a dissolved oxygen meter and probe (YSI, Yellow Springs, Ohio). The oxygen measurements were corrected for salinity, which was calculated from conductivity and water density measurements taken in January 1991 (47). The pH of the water samples was determined immediately after collection by using a model 209A pH meter (Orion, Beverly, Mass.).

Direct cell counts.

Microbes in whole-water samples (1 to 100 ml) were stained with either acridine orange (AO) (32) or 4′,6-diamidino-2-phenylindole (DAPI) (31) before enumeration. DAPI was used when cells were not discernible on AO-stained filters due to high background fluorescence. Cells stained with DAPI were first fixed with formaldehyde (final concentration, 3.7%) to preserve the cell morphology and improve the staining efficiency. There was no significant difference in total cell counts between DAPI and AO staining methods (2.5 × 10⁴ and 2.7 × 10⁴ cells/ml, respectively) with fixed 10-m-deep lake water from the 1996 field season. Stained cells were then filtered onto black polycarbonate membranes (Poretics Corp., Livermore, Calif.) and counted by using a Zeiss epifluorescence microscope and Zeiss filter set 09 (AO) or 02 (DAPI). Photosynthetic microbes were enumerated in unstained samples by observing the natural autofluorescence of pigments when viewed with a blue-waveband filter set (Zeiss filter set 09), which excites chlorophyll a, divinyl-chlorophyll a, and type I phycoerythrins (41).

Nucleic acid extraction and hybridization.

Cells for nucleic acid extraction were collected from 25 to 70 liters of water on 142-mm-diameter Durapore filters by pressure filtration (ca. 30 lb/in²) with a stainless steel filtering apparatus (Millipore, Inc., Bedford, Mass.). The filters were stored at −70°C in separate Zip-Lok bags, each containing 5 ml of lysis buffer (20 mM EDTA, 400 mM NaCl, 750 mM sucrose, 50 mM Tris-HCl [pH 9.0]). Nucleic acids were extracted from the cells on the filters by the method of Gordon and Giovannoni (25). Filter hybridizations were performed as described previously (6) with ³²P-labeled oligodeoxynucleotide probes targeting the domains Bacteria (48), Eucarya (23), and Archaea (48). Hybridized probe was visualized by exposure of the membranes to scientific imaging film (X-Omat AR; Kodak, Rochester, N.Y.) and quantified by a radioanalytic image system (QuantProbe Software, version 4.3; AMBIS, Inc., San Diego, Calif.). The amount of probe bound per nanogram of nucleic acid was plotted for the environmental samples and control nucleic acids from Pseudomonas fluorescens (Bacteria), Saccharomyces cerevisiae (Eucarya), and Haloarcula marismortuii (Archaea). The slopes of these lines were used to determine the proportion of rRNA contributed by bacterial, archaeal, and eucaryal microbes in the environmental sample (22). In essence, the percentage of rRNA for each domain is the slope of the line for the specifically bound domain probe divided by the sum of the slopes of the lines for all three domain probes. Corrections were made for nonspecific binding, and each term was normalized to standards.

Manganese reduction assays.

The manganese reduction assays were performed with synthetic manganese oxides (MnO₂) prepared by the method of Lovley and Phillips (40).

(i) Field studies.

Water was pumped into a 20-liter carboy and concentrated with an Amicon hollow-fiber filter apparatus, either 100-fold (60- to 61-m-deep sample) or until the filtrate flow had become negligible (62- to 63-m-deep sample), resulting in a 55-fold concentration. For samples from each depth, 100 ml of concentrated water was transferred into each of five 1-liter polypropylene bottles and amended as follows: (i) no amendments; (ii) MnO₂ (final concentration, ca. 0.2 mg/ml); (iii) MnO₂ and sodium azide (final concentration, 10 mM); (iv) MnO₂ and nutrients (0.02% peptone, 0.01% yeast extract); and (v) MnO₂, nutrients, and sodium azide. The assay mixtures were incubated, stationary, in the dark at 15°C. Samples (1 ml) were taken at selected time points, and the particulates were removed with a syringe and syringe filter holder with a cellulose acetate membrane (Micron Separations, Inc.; pore size, 0.22 μm). Adsorbed Mn(II) was removed from the particulates by passing 0.5 ml of 0.5 N HCl (trace metal grade) through the filter, incubating it for 2 min, and then rinsing it with another 0.5 ml of 0.5 N HCl. Air (10 ml) was then passed through the filter to recover as much liquid as possible. The HCl rinses were combined with the initial filtrate, and the concentration of Mn(II) was determined by the formaldoxime assay (7).

(ii) Isolate studies.

A 10-ml volume of ribose/peptone (RP) medium (50 mM ribose, 100 mM NaCl, 1.8 mM KH₂PO₄, 2.9 mM K₂HPO₄, 0.25% peptone, 1× vitamin and trace element solution [2] [filter-sterilized ribose and vitamins were added aseptically after autoclaving]) was added to each of 12 18- by 100-mm glass test tubes. Four treatments were prepared as follows: no amendment, sodium azide (final concentration, 10 mM), MnO₂ (ca. 0.2 mg/ml), and MnO₂ plus sodium azide. Duplicate tubes from each treatment were inoculated with 100 μl of an overnight culture (optical density at 600 nm, ca. 0.2 to 0.4) of a phylogenetically representative manganese reducer (strain LV62:W1). An additional tube from each treatment was left uninoculated as a negative control. The cultures and controls were incubated in the dark at 15°C; samples (1 ml) were removed before inoculation and after 4 and 7 days to determine the concentration of accumulated Mn(II), as was done for the field samples.

Growth and identification of manganese-reducing bacteria.

Water samples from various depths in the lake and along the Onyx River were either spread directly onto R2A agar (Difco, Detroit, Mich.) or filtered through an HA filter (Millipore), which was then placed onto R2A agar medium. Incubations were carried out at 15°C in the dark for up to one month. The resultant colonies were streaked for isolation on R2A agar (two to four transfers). These isolates were screened for the ability to reduce manganese oxides in the presence of oxygen by spreading aliquots onto AMR (5 g of peptone per liter, 13.8 mM glucose, 1.8 mM KH₂PO₄, 2.9 mM K₂HPO₄) agar medium with a 1.5% agar overlay containing manganese oxides (ca. 17 g/liter); the manganese oxides were prepared as described by Burdige (9). Reduction of Mn(IV) results in solubilization of the manganese oxides, creating a clear zone in the overlay around colonies of manganese-reducing bacteria (see Fig. 4).

FIG. 4.

Manganese reduction by a Lake Vanda isolate (LV62:W1) is seen as a clearing in the black manganese oxide overlay around colonies in this backlit photograph of a typical streak plate.

Putative manganese-reducing bacteria were streaked onto tryptic soy agar medium (Difco), and single colonies were transferred repeatedly until colony and cell morphologies indicated the presence of a pure culture. Broth cultures were maintained on either AMR, tryptic soy broth (TSB; Difco) or TSB with a mineral oil overlay. Stock cultures were stored in 35% glycerol at −80°C. Isolates were named to reflect the site and water depth from which they were obtained.

Two methods were used to determine if a diffusible compound was responsible for manganese reduction, as was suggested by the clear zones surrounding the colonies on the manganese overlay plates. First, a 48-h culture of Carnobacterium strain LV62:W1, grown at room temperature (RT) in AMR supplemented with vitamins and trace elements (as described above for RP medium), was centrifuged and filter sterilized to remove all the cells. Manganese oxides (ca. 0.02 mg/ml) were then added to the filter-sterilized culture supernatants and uninoculated medium controls. The flasks were incubated at RT with shaking at 225 rpm and were observed visually for the disappearance of the manganese oxides. The second method used dialysis bags containing manganese oxides to prevent cell contact with the particulate MnO₂. After addition of the manganese oxides, the dialysis bags were autoclaved in a beaker of deionized water, rinsed with sterile deionized water, and then aseptically placed into flasks containing AMR supplemented as above. Manganese reduction during an RT incubation at 225 rpm was compared between control flasks of AMR and flasks inoculated with Carnobacterium strain LV62:W1.

Isolates were Gram stained and tested for anaerobic growth by using TSB tubes with mineral oil overlays and incubation in a BBL GasPak Jar (Becton Dickinson, Cockeysville, Md.) under a CO₂-H₂ atmosphere. The isolates were classified based on a phylogenetic analysis of their 16S rRNA gene (16S rDNA) sequences. The 16S rDNA was amplified by PCR from single colonies (37) by using oligodeoxynucleotide primers designed to anneal to conserved regions of the bacterial 16S rDNA. The forward primer (8F) corresponded to positions 8 to 27 of the 16S rDNA from Escherichia coli, and the reverse primer (1492R) corresponded to positions 1492 to 1510 (17). The amplified product was purified on Millipore 30,000-molecular-weight-cutoff columns (Ultrafree-MC; Millipore) as recommended by the manufacturer. Cycle sequencing was performed on a Perkin-Elmer 9600 thermal cycler with ABI dye terminator sequencing as recommended by the manufacturer, and the products were analyzed on an ABI 373a DNA sequencer.

A partial sequence of 16S rDNA (ca. 400 to 500 nucleotides) was obtained for each of the 22 isolates from primers 519R (35) and 8F (17). Based on the similarities of the partial sequence, representative isolates (LV62:W1, LV66, and LV65.5:W1) were selected for additional sequence determination. A collection of 12 primers provided an average redundancy of 3.0 per nucleotide for the nearly complete sequences and were assembled into consensus sequences.

The resultant sequences were aligned to the small-subunit rRNA sequences from the Ribosomal Database Project (42) based on regions of sequence conservation and secondary structure. Positions that were unambiguously aligned were used in the derivation of phylogenetic trees by using the maximum-likelihood method (46) or a least-squares distance method (16).

Nucleotide sequence accession numbers.

The consensus sequences for the 16S rDNA of the isolates described above have been deposited in GenBank as follows: Carnobacterium strain LV62:W1 (AFO76637), Carnobacterium strain LV66 (AFO76638), and Aerococcus strain LV65.5:W1 (AFO76639).

RESULTS

Depth profiles of the Lake Vanda water column were constructed from data collected from the west end of Lake Vanda, an area representing the deepest part of the lake (ca. 75 m). The water temperature increased with depth from 4°C at the surface to a maximum of 23°C, while the pH decreased from 8.5 to 6.5 between 48 and 60 m (Fig. 1A). The upper layer of the lake was at or near saturation for oxygen to a depth of approximately 65 m (Fig. 1B), which is the photic zone of the lake. Hydrogen sulfide was present in the anoxic waters below 67 m. The total number of microbes was enumerated in samples taken during the 1994 field season by using AO and DAPI (Fig. 2A). There was little variation in the size of the microbial community with depth. The total cell number did decrease slightly at 60 m, but this number appears to be an anomolous count, since the number of autofluorescent cells alone exceeds the total cell count. At least half of the microbes at all depths below 20 m appear to be photosynthetic cells. Of interest is the increase in both temperature and oxygen concentration near 30 m, the depth at which autofluorescent cells were first detected (Fig. 1A and 2A).

FIG. 1.

Physical and chemical profiles of the Lake Vanda water column during the 1994 field season. (A) Temperature and pH. (B) Concentrations of oxygen, hydrogen sulfide (▿; adapted from reference 4), and dissolved manganese.

FIG. 2.

Microbial profiles in Lake Vanda during the 1994 field season. (A) Direct total cell counts plotted against depth. The abundance of phototrophs as determined by counts of autofluorescent cells on unstained filters is shown. (B) Percentage of eucaryal rRNA (○) plotted against depth. The dotted line in both panels is at the depth of the oxic-anoxic interface.

The composition of the microbial community was examined by hybridization of extracted nucleic acids with oligonucleotide probes specific for the domains Bacteria, Archaea, and Eucarya. Eukaryotic nucleic acid levels increased to about 9.8% of total nucleic acid levels at 30 m, declined slightly, and then increased to 10.2% at 55 m (Fig. 2B). The remainder of the nucleic acids extracted from each depth examined were bacterial; no archaeal nucleic acids were detected. The yield of nucleic acids from samples collected below a depth of 55 m was insufficient for further analysis, perhaps due to difficulties resulting from the high concentration of calcium chloride in the samples.

The profile for Mn(II) contained a maximum in the anoxic, reducing zone of the lake (Fig. 1B), which is typical for many stratified lakes. There was also a submaximum Mn(II) peak at 61 m (approximately 5 m above the oxic-anoxic interface) which is not frequently observed in other lakes but which has been documented previously in Lake Vanda (12). Synthetic MnO₂ was reduced by microbes concentrated from the 60- to 61-m-deep water samples when incubated in the presence of oxygen and added nutrients (Fig. 3). The amount of soluble manganese [Mn(II)] accumulated by day 7 was equivalent to approximately 8 × 10⁻⁵ μmol per cell. Mn(IV) was also reduced by microbes concentrated from the 62- to 63-m-deep water samples but to a lesser extent (ca. 3 × 10⁻⁵ μmol per cell [data not shown]). Reduction was dependent upon nutrient addition but was not inhibited by the metabolic poison sodium azide. Mn(II) was not detected in the absence of synthetic manganese oxides and nutrients or when sterile water was added as a substitute for the concentrated cell suspension.

FIG. 3.

Reduction of synthetic manganese oxides to Mn(II) by microbes concentrated from a depth of 60 to 61 m. Assays were conducted in the presence or absence of nutrients as indicated and in the presence (▸, ■) or absence (○, ◊) of sodium azide. Open squares represent the control without added manganese oxides or nutrients.

The concentrated water samples used in the manganese reduction assay (60 to 61 m and 62 to 63 m deep) and microbes concentrated from depths of 65.5 and 66 m were used as inocula for direct isolation of manganese-reducing bacteria on AMR agar medium with an MnO₂-agar overlay. Manganese-reducing bacteria were identified by the development of a clear zone surrounding the colonies, resulting from the dissolution of the black, particulate MnO₂ (Fig. 4). Putative manganese-reducing strains were then transferred to fresh media to confirm the ability to reduce Mn(IV). Over 270 colonies from the spread plates were streaked onto MnO₂ overlay plates, and fewer than 6%, the majority of these from the 62- to 63-m-deep sample, were confirmed as manganese reducers. Bacteria were also isolated from throughout the water column and the Onyx River and tested for the capacity to reduce Mn(IV). Approximately 8% of the 107 isolates from the oxic zone of Lake Vanda and the Onyx River were able to solubilize manganese oxides (Table 1).

TABLE 1.

Survey of aerobic microorganisms from Lake Vanda and the Onyx River for the ability to reduce manganese oxides

Location and depth (m)	Total no. of isolates surveyed	No. able to reduce MnO2
Onyx River	9	0
Lake Vanda
5–30	6	0
40–45	19	0
50–54.5	7	0
54.5–59.5	34	1
59.5–64.5	22	1
64.5–67	10	7
Total	107	9

Twenty-two of the manganese-reducing strains were isolated and further characterized; most had a similar colony morphology. All the isolates grew under oxic and anoxic conditions and stained either gram positive or gram variable. Phylogenetic analysis of partial 16S rRNA sequences was used to differentiate and identify the manganese-reducing isolates (Fig. 5). The majority of the manganese reducers grouped within or close to the genus Carnobacterium (groups I and II [Fig. 5]). The largest group of manganese reducers, group I, consisted of 17 isolates. The sequence data for 13 of the group I isolates was virtually identical, and the remaining sequences were highly similar (≥98.4%). Group II consisted of four isolates whose partial sequences were indistinguishable (≥99.7% identical). On average, sequences from group I were 92.3% similar to those from group II. The remaining isolate, LV65.5:W1, was 81 to 86% similar to the others, grouping instead with sequences from the genus Aerococcus. The 16S rDNA PCR products were completely sequenced from LV62:W1, LV66, and LV65.5:W1, representatives of each of the three groups of isolates. Phylogenetic analysis (Fig. 6) indicated that the sequences for both LV62:W1 and LV66 were most closely related to Carnobacterium isolates from Ace Lake, Antarctica (96.5 and 92.2% similar, respectively) (20). The sequence for LV65.5:W1 was virtually identical (99.7%) to Aerococcus viridans.

FIG. 5.

Approximate phylogenetic relationships among 22 manganese-reducing isolates from the oxic portion of the Lake Vanda water column. These distance-based (16) relationships were derived by using approximately 420 positions of the 16S rDNA and were used to select representative isolates for complete sequencing.

FIG. 6.

Phylogenetic tree created with maximum likelihood and the nearly complete (ca. 1,400 bp) sequence of PCR-amplified 16S rDNA from isolates LV62:W1, LV66, and LV65.5:W1. LV62:W1 and LV66 were contained within the monophyletic group created by members of the genus Carnobacterium. This group includes Carnobacterium alterfunditum, an isolate from Ace Lake, Antarctica. Isolate LV65.5:W1, also a manganese reducer, was most closely related to Aerococcus viridans.

Isolate LV62:W1 (hereafter referred to as Carnobacterium strain LV62:W1) was tested for the ability to reduce manganese oxides in an oxic incubation with and without added sodium azide (Table 2). Manganese reduction occurred in the presence of sodium azide, as was seen in the Lake Vanda field assay. Similar results were obtained with 50 mM glucose, 0.5% peptone, and an incubation temperature of 25°C (data not shown).

TABLE 2.

Manganese reduction by a Lake Vanda isolate

Experimental treatment	Concn of accumulated Mn(II) (mM)
	Initial	Day 4	Day 7
LV62W1	2.67	3.32	4.79
Uninoculated control	2.63	—a	2.33
LV62W1 + azide	1.60	2.94	4.77
Uninoculated control + azide	1.80	—	1.73

^a—, not determined. 

Manganese oxide dissolution by the cell-free supernatant fraction from a 48-h culture of Carnobacterium strain LV62:W1 was apparent within 5 min of the addition of the particulate manganese oxides. Complete reduction of the added manganese oxides by cell-free supernatant occurred within 1 h, whereas the use of sterile AMR broth resulted only in minimal Mn(IV) reduction even after 24 h. Likewise, when the manganese oxides were contained within dialysis bags to prevent cell contact with the particles, Carnobacterium strain LV62:W1 cultures still reduced the manganese oxides.

DISCUSSION

Lake Vanda is one of the least productive lakes in the world (24). Our estimation of the size of the microbial community in Lake Vanda (ca. 1 × 10³ to 5 × 10³ microbes per ml) is indicative of extremely oligotrophic waters. There was no obvious variation in the vertical distribution of microbes in the water column that might be expected given the wide fluctuation in redox potential and vertical chemical profiles. Our enumeration of phototrophic microbes agrees well with previous reports of 10² to 10³ phytoplankton cells/ml (24, 53). Three communities of phytoplankton in Lake Vanda have been described previously (53). One of these communities, at depths of 25 to 30 m, contained small eukaryotic microflagellates. These microflagellates may be responsible for the observed increase in the amount of eukaryotic nucleic acids at this depth (Fig. 2B).

As in previous studies, two peaks of soluble manganese were detected in Lake Vanda. The highest concentration of soluble manganese was located in anoxic waters (Fig. 1) and has been attributed to chemical reduction of manganese oxides by sulfide (4). There was a second peak of soluble manganese at a depth of approximately 61 m. Despite the relatively high oxygen concentration at this depth, the E_h values calculated from the nitrate/nitrite couple and the low pH define an environment in which Mn(II) is thermodynamically favored (13). Although thermodynamic calculations reflect the stability of Mn(II) in this environment, manganese oxides are relatively stable in the absence of reducing agents or microbial interactions which greatly accelerate the kinetics of Mn(IV) reduction (8, 15, 19, 30).

Microbes concentrated from waters collected from a depth of 60 to 63 m reduced synthetic manganese oxides when supplied with an exogenous source of carbon and nutrients. The ability of nutrients to stimulate manganese reduction is consistent with microbially mediated manganese reduction, but the metabolic poison sodium azide did not inhibit the reduction. At a concentration of 10 mM, azide is an effective inhibitor of manganese reduction in enrichments of strictly anaerobic manganese reducers (11). However since azide acts primarily to disrupt membrane-associated electron transport processes, the possibility remained that fermentative bacteria were responsible for the observed reduction of manganese(IV).

Bacterial isolates from Lake Vanda and the Onyx River were screened for the ability to reduce Mn(IV) by using MnO₂ overlay plates. Although most isolates (ca. 94%) were unable to reduce manganese oxides, manganese reducers were found within 5 m of the water column above the oxic-anoxic interface. This is the region of the water column that contains the oxic Mn(II) peak. One of the isolates (Carnobacterium strain LV62:W1) reduced Mn(IV) while growing under oxic conditions and in the presence of sodium azide, as was seen in the Lake Vanda field assay.

A diffusible compound produced by Carnobacterium strain LV62:W1 was responsible for the manganese reduction. This is in contrast to the direct reduction seen for two well-known manganese reducers. Geobacter metallireducens is a gram-negative, obligate anaerobe that is able to use manganese and other metals as its terminal electron acceptor (39). Geobacter cells attach to the manganese oxide particles, coupling the reduction of Mn(IV) to the complete oxidation of an organic compound such as acetate to CO₂. Shewanella putrefaciens, on the other hand, is a facultative anaerobe which only incompletely oxidizes carbon compounds, such as formate or lactate, when coupled to the reduction of manganese. S. putrefaciens is also different because it is able to reduce manganese with the oxidation of H₂ (45).

Most of the manganese-reducing bacterial isolates from Lake Vanda belong to the genus Carnobacterium. This was a surprise since carnobacteria have complex nutritional requirements and have most frequently been isolated from meat, meat products, and fish (29), whereas the waters of Lake Vanda are strictly oligotrophic. The morphological and physiological traits of these isolates, including the Gram stain, facultatively aerobic growth, nutritional requirements, and azide resistance are all in agreement with the Carnobacterium classification (29). A phylogenetic analysis with nearly complete 16S rDNA sequences from two representatives of the major groups of isolates revealed that the isolates were most closely related to two Carnobacterium strains isolated from an Antarctic lake (20). This suggests that there is a group of bacteria within the genus Carnobacterium whose habitat is aquatic.

The genus Carnobacterium is a close evolutionary relative of the genus Lactobacillus. This is intriguing because many lactobacilli both require and contain high levels of Mn(II) (1). As a group, the lactic acid bacteria contain a wide variety of Mn(II)-requiring and Mn(II)-stimulated enzymes (1). Additionally, some lactobacilli contain millimolar concentrations of intracellular Mn(II) in lieu of superoxide dismutase to mitigate oxygen toxicity (1).

The biological roles of manganese in Lactobacillus suggest that the reduction of manganese oxides by the Carnobacterium isolates may be more than incidental reduction due to the production of lactic acid. Studies to investigate several possible benefits from manganese reduction, such as the aforementioned mitigation of oxygen toxicity and interactions with cell metabolism, are under way. It is also conceivable that there is a release of inorganic or organic nutrients concomitant with the reduction of manganese oxides that stimulates the growth of carnobacteria. An rRNA-based oligonucleotide probe specific for Carnobacterium has been developed to estimate the population size of Carnobacterium and will be used to determine if these fermentative microbes are dominant members of the microbial community in Lake Vanda.

Unlike many elements, heavy metals tend to bond reversibly with a number of naturally occurring solid phases, including manganese oxides. The interactions are often complex and depend upon the nature of the solid phase, the nature of the adsorbed metal, the characteristics of the environment, and interactions with microorganisms. Together, these factors are responsible for determining the cycling and ultimate fate of the heavy metals. The interactions between metals, microbes, and the environment in Lake Vanda may serve as a useful model for other aquatic systems that are less amenable to study due to the confounding effects of rapid mixing.