Nitrate-Utilizing Microorganisms Resistant to Multiple Metals from the Heavily Contaminated Oak Ridge Reservation

To determine if there were microorganisms at the ORR-contaminated site that could obtain energy for growth by nitrate reduction in the presence of high concentrations of metals, a multiple-metal mixture (COMM) containing eight metals (Al, Mn, Fe, Co, Ni, Cu, Cd, and U) was formulated that mimics the metals that are present in the groundwater in highly contaminated wells near the S-3 ponds at the ORR site (12). The composition of COMM is shown in Table 1. This is representative of the metal concentrations measured in a previous analysis of 80 different groundwater wells at the contaminated ORR site (12). The exceptions are concentrations of Mn and Al, which in COMM are significantly lower than those in the most highly contaminated wells in order to minimize precipitation issues with the growth medium used here. To isolate new strains, a series of over 2,000 high-throughput enrichments were set up in which groundwater and sediment from noncontaminated and contaminated areas at ORR were used as inocula. Noncontaminated water is defined as containing less than 126 nM uranium, which is the EPA limit allowed in drinking water, and this is found in ORR wells outside the contamination plumes from the S-3 ponds. The COMM was added to about 90% of the enrichments, and the remainder contained either a 10-fold lower concentration (0.1× COMM) or an elevated concentration of U (100 to 400 μM; the COMM contains 100 μM U) in the absence of other metals. Nitrate was used at concentrations of either 20 mM or 50 mM, which are also based on those in the contaminated wells (Table 1). Other parameters that were varied for enrichments include carbon source (ethanol, lactate, fumarate, pyruvate, glucose, cellulose, or compost), pH (3.0 to 7.0), oxygenation (aerobic, microaerophilic [3% O₂], or anaerobic), and the presence of sulfate (20 mM) rather than nitrate or with no terminal electron acceptor added. Enrichments were incubated at 22°C, and growth was monitored over the course of 3 months. Isolates were obtained from enrichments exhibiting growth after plating on solid medium.

By 16S rRNA gene sequencing, twenty-one new strains were obtained and all but six sequences were observed for multiple independent isolates. These are listed in Table S1 in the supplemental material with their closest relative species by 16S rRNA gene sequence. They represent a diversity of bacteria, including one Actinobacterium, eight Firmicutes of the Bacilli class, and ten Proteobacteria from the classes Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria. The most common isolates were various Rhodanobacter strains, one of which was isolated 17 separate times, and various Bacillus strains, of which there were four different strains, each isolated between 3 and 7 separate times.

Seven of these 21 ORR strains were selected for further in-depth study based on their ability to grow anaerobically by nitrate reduction (Table 2). These were designated MT049, MT058, MT066, MT086, MT094, MT123, and MT124. Their enrichment conditions along with their closest relatives by 16S rRNA gene sequence analysis are shown in Table 2. Of the seven strains, five were isolated from enrichments inoculated with highly contaminated groundwater from the area near the ORR S-3 ponds, while one (MT049) was from a noncontaminated ORR groundwater sample and the other (MT058) was from an uncontaminated ORR soil sample. They consist of two Bacillus (MT066 and MT094) and two Paenibacillus (MT086 and MT124) Gram-positive strains and one each of Pantoea (MT058), Castellaniella (MT123), and Serratia (MT049) Gram-negative strains. Six of the seven strains were isolated in the presence of at least 20 mM nitrate, while four were isolated in the presence of COMM, two in the presence of 0.1× COMM, and one in the presence of 200 μM U with no other metals added (Table 2). The carbon sources that the strains were isolated on included formate, ethanol, glucose, a lactate-fumarate mixture, and compost tea. The latter was prepared from compost obtained from a commercial composting facility and was used to enrich both Bacillus strains and one of the Paenibacillus strains.

The pH profiles for the seven ORR strains grown anaerobically in the presence of 20 mM nitrate and the indicated carbon source (Table S2) are shown in Fig. 1 (standard deviations are given in Table S3). In most cases, the pH values of the ORR groundwater samples that the strains were isolated from were lower than the pH of the enrichment cultures used for isolation (Table 2) or the pH optima observed for the isolates (Fig. 1 and Table 2). For example, both Paenibacillus species (MT086 and MT124) were isolated from groundwater (well FW126) that had the lowest pH (3.0) of all of the source wells and the highest concentrations of most of the contaminating metals (Table 1). However, MT086 was enriched at pH 5.5 and grows optimally from pH 6.0 to 8.0 (Fig. 1), while MT124 was enriched at pH 7.0 and has a growth pH range from pH 5.5 to 8.0 with an optimum at pH 7.5 (Fig. 1 and Table 2). None of the isolates grew at pH 3.0; however, all three Gram-negative isolates (MT049, MT058, and MT123) as well as one of the Bacillus Gram-positive isolates (MT094) had measurable growth at pH 4.0 (Fig. 1), which in each case is lower than the pH of the groundwater from which they were isolated (Table 2). The Castellaniella species strain MT123 was the most acidophilic, with an optimum at pH 5.5, which is lower than the pH of the enrichment (pH 6.5). All of the other strains had significantly higher pH optima than the media used for isolation. For example, the Bacillus sp. strain MT066 had a broad pH optimum surrounding 7.0 and was enriched at pH 5.5.

The pH optimum for each of the seven ORR strains was used for carbon source and metal resistance experiments, and these parameters are summarized in Table S2. Growth was measured using six different carbon sources (formate, ethanol, fumarate, lactate, pyruvate, and glucose) under anaerobic, nitrate-reducing conditions with 20 mM nitrate. This is within the range of nitrate concentrations in the contaminated wells near the S-3 ponds, which vary over the range of 9.2 to 188 mM (Table 1). The results are shown in Fig. 2 (the standard deviations are given in Table S4). Three of the strains (MT066, MT086, and MT123) were unable to grow on any carbon source unless nitrate was present, while the other four strains exhibited increased growth in the presence of nitrate on at least one of the carbon sources. Among the strains that could grow in the absence of nitrate, growth was observed only on pyruvate and glucose. For example, the Paenibacillus sp. strain MT124 only grew on glucose and the Pantoea sp. strain MT058 only grew on pyruvate and glucose in the absence of nitrate. Most of the strains, including the two Bacillus strains (MT066 and MT094) as well as the Castellaniella strain (MT123), showed significant growth in the presence of nitrate on all or most of the carbon sources tested. The carbon sources that supported the highest growth rates or the most increased growth rate in the presence of nitrate were used in designing the growth media for determining their responses to metals and in measuring nitrate and nitrite reductase activities. The parameters for these studies are summarized in Table S2.

One of the defining characteristics of the contaminated groundwater at ORR near the S-3 ponds is the elevated concentrations of a combination of metals that exist in three different ionic forms depending on the metal and the redox state. The groundwater contains elevated concentrations of the cations Al³⁺, Fe^2,3+, Cd²⁺, Cu²⁺, Co²⁺, Ni²⁺, and Mn²⁺, the oxyanion CrO₄²⁻, and the oxycation UO₂²⁺ (12). Dose-response curves for all seven ORR strains were obtained using all of these metals, either individually or with COMM. Al and Fe were not included in this study due to problems of precipitation (Al³⁺ and Fe³⁺ precipitate out of our growth medium at pH values of >5.5 and >3.0, respectively). The sensitivities of all seven strains to nitrate and nitrite were also determined. The dose-response curves were used to calculate half-maximal effective concentration (EC₅₀) values for each ion or for COMM, and these are shown in Table 3. The 95% confidence intervals (95% CI) for the EC₅₀ values are given in Table S5.

As expected, all seven strains were able to grow in the presence of high concentrations of nitrate, with EC₅₀ values ranging from 167 mM to a remarkable 711 mM. Growth at these concentrations of nitrate exceed concentrations observed in the ORR environment, where the most contaminated well (FW126) contained 188 mM nitrate (Table 1). EC₅₀ values for nitrite were lower than those for nitrate and ranged from 9.0 to 77 mM. Three of the four Gram-positive strains (MT066, MT094, and MT124) had significantly higher nitrite EC₅₀ values than the Gram-negative strains (MT049, MT058, and MT123). However, there does not appear to be a correlation between nitrate and nitrite resistance.

In the ORR contaminated groundwater, concentrations of Mn and U as high as 3 mM and 0.6 mM, respectively, were measured (Table 1). In general, EC₅₀ values were high for Mn and U, with five of the seven strains having EC₅₀ values for Mn of >150 μM (with some of >350 μM), and six of the seven strains have EC₅₀ values for U of >200 μM. The exception was the Castellaniella strain MT123, which has an EC₅₀ for U of only 6.0 μM. The optimal pH for growth of this strain (used to measure EC₅₀ values) is lower than that for any other strain (Table S2). Uranium speciation in growth media can dramatically change with pH, thereby affecting toxicity (26).

Although Cr is not a major contaminant in all of the ORR contaminated groundwater wells (Table 1), the seven strains exhibited a wide range of EC₅₀ values for chromate, from 7.0 μM to >0.5 mM (Table 3). Once again, Castellaniella strain MT123 had by far the lowest EC₅₀ value, perhaps due to increased toxicity of the metal at the lower growth pH of 5.5. In the ORR environment, Cr was present in only some of the contaminated groundwater samples tested (FW126, FW106, and FW109) (Table 1), and with the exception of MT123, all of the strains had EC₅₀ values greater than the highest Cr concentration measured in the ORR groundwater. Interestingly, the higher EC₅₀ values did not correspond to exposure of that strain to Cr in the environment. For example, the Serratia strain (MT049) had an EC₅₀ of 0.3 mM (Table 3), even though it was isolated from a well (GW066) that contained no detectable Cr (Table 1). With only a few counterexamples, the EC₅₀ values for each of the strains to one of the three classes of contaminating metals, divalent metal cations, decreased in the order Ni > Co > Cd > Cu (Table 3). This is generally the same pattern as that seen for the concentrations of these elements in the contaminated groundwater (Table 1). EC₅₀ values for Ni, Co, Cd, and Cu ranged from 79 to 160 μM, 31 to 120 μM, 13 to 157 μM, and 6 to 66 μM, respectively.

Each ORR strain was also investigated for its response to a combination of metal ions (Cd²⁺, Cu²⁺, Fe²⁺, Co²⁺, Ni²⁺, Mn²⁺, and UO₂²⁺). This was the same as those in COMM that were used in the enrichments, except that Al was omitted to avoid issues with precipitation. The strains are ordered in Table 3 in terms of their resistance to COMM, with EC₅₀ values ranging from 0.9 to 0.2 times the COMM enrichment concentration. The strain most sensitive to COMM was the Castellaniella strain MT123 (EC₅₀ of 0.24×), and it also had the lowest EC₅₀ values for Co, U, Cr (not present in COMM), and nitrite. (Table 3). Interestingly, the Pantoea sp. strain MT058, which was one of two strains that were isolated from a noncontaminated source and was the only strain isolated from sediment (the other six were from groundwater), had the second highest EC₅₀ value to COMM (0.7×).

Clearly, the properties of the groundwater from which the strains were isolated are not a good indicator of their metal resistance. For example, the two Paenibacillus sp. strains, MT086 and MT124, were both isolated from the most highly contaminated well (FW126) but had very different EC₅₀ values for COMM (0.3× versus 0.9×). This result is also supported by their resistance to individual metals, where MT124 was shown to be more resistant than MT086 to Cr, Cd, Cu, Ni, and Mn (Table 3). It is curious that MT124 is highly resistant to these metal cations, as it was enriched in the presence of U (200 μM) rather than COMM. In any event, it is obvious that there is large variation in the metal resistance properties of the seven ORR strains and that this is independent of the source groundwater.

A key issue to address was whether a mixture of metals reflecting the contaminated ORR environment affected the ability of any of the ORR strains to reduce nitrate. The seven ORR strains were grown in the presence and absence of COMM under denitrifying conditions with a carbon source that supported increased growth for each strain when nitrate was present (Table S2). Cells were harvested and whole-cell nitrate and nitrite reductase levels, measured by the production or disappearance of nitrite, respectively, were determined. As shown in Fig. 3, there was a 60-fold variation in the specific nitrate-reducing activities of the seven strains (in the absence of COMM). The values ranged from 5.4 ± 0.7 units/OD₆₆₀ for MT124 to 305 ± 32 units/OD₆₆₀ for MT086, both of which are Paenibacillus strains. There does not appear to be any relationship between nitrate reduction rate and nitrate resistance (Table 3 and Fig. 3). When the ORR strains were grown in the presence of COMM, six of the seven strains had nitrate reductase activities similar to those measured when they were grown in its absence (Fig. 3). Only the Castellaniella sp. strain MT123, the strain that was most sensitive to COMM (Table 3), showed a significant decrease (from 6.0 ± 1.3 to 1.5 ± 0.6 units/OD₆₆₀). MT123 was also the only strain to have detectable nitrite reductase activity. All of the other six strains lacked detectable nitrite reductase activity, indicating that these strains reduce nitrate and excrete the nitrite that is generated. When grown in the absence of COMM, the nitrite reductase activity of MT123 was 22.0 ± 12.5 units/OD₆₆₀, and when grown with the metal mixture it was 38.5 ± 3.6 units/OD₆₆₀. Since nitrite production is used to determine nitrate reductase activity in this assay, the presence of nitrite reductase activity for MT123 likely causes the nitrate reductase activity of this strain to be underestimated. Conversely, with the exception of MT123, the nitrate-reducing activities of the ORR strains shown in Fig. 3 are likely a true reflection of their ability to reduce nitrate, and clearly there is huge variation between the strains. More importantly, these are largely unaffected by high concentrations of multiple metals.

An inverse linear relationship (r² > 0.99) between the specific nitrate reductase activity and nitrite resistance (up to 80 mM) for the four Gram-positive strains (MT066, MT086, MT094, and MT124) was observed (Fig. S1). Since none of these strains have nitrite reductase activity, one explanation for this is that the strains with higher rates of nitrite production from nitrate are tuned to survive in electron donor-limited environments where competition for both nitrate and nitrite is high and neither is expected to accumulate.

In a previous survey of the ORR environment, groundwater samples were taken from 93 contaminated and noncontaminated wells and 16S rRNA gene sequence data for the V4 region were obtained (27). We used these data to search for ESV matches using the V4 region of the 16S rRNA gene sequences of the seven ORR strains described here. Of the seven strains, the two Paenibacillus strains (MT086 and MT124) did not have ESV matches in the ORR survey but the other five did. However, there was great variation in the abundance of these five ESVs. For example, the ESV for Bacillus strain MT066 was found in only one of the 93 wells, whereas the Castellaniella strain MT123 ESV was present in 32 different wells and was the 177th most common ESV found in the survey (Table 4). In particular, MT123 constituted 7% of the ESV reads in well FW104, a highly contaminated well near the S-3 ponds from which MT123 was isolated. As shown in Fig. 4, the ESV matches of all five ORR strains were found in wells that spanned a wide range of both pH values (from 3.1 to 10.4) (Fig. 4A) and uranium concentrations (0.007 to 143 μM) (Fig. 4B). In fact, for all four strains found in multiple wells (MT049, MT058, MT094, and MT123), each was found in both noncontaminated (pH ∼7.0, U < 0.126 μM) and highly contaminated wells (pH <4.0, U > 70 μM).

The ESV abundance at each well location is shown for MT123 (Fig. 5), MT049 (Fig. S2), MT058 (Fig. S3), and MT094 (Fig. S3). As shown in Fig. 5, the wells span a 11-km-by-1-km area within the Bear Creek Valley containing the S-3 ponds. The wells from which the four ORR strains (for which ESV data are available) were isolated are marked with a white pointer, and the S-3 ponds are indicated with a red square (Fig. 5). Two of the strains were isolated from contaminated wells within 10 m of the S-3 ponds (MT123 and MT094), while MT049 was isolated from a noncontaminated well and MT058 was isolated from a pristine sediment sample that was taken 7 km from the S-3 ponds. Three of the strains (MT058, MT094, and MT123) had ESV matches in the samples they were isolated from, and the other (MT049) was matched to a well (GW537) with 0.65% of the ESV reads, which was less than 500 m from the isolation well. In terms of geography, based on the ESV data, all four strains were found in both noncontaminated and contaminated wells. In fact, the two strains isolated from contaminated wells (MT094 and MT123) were both found in at least one pristine well in a defined site with no known contamination sources that is located in the far south of the site, some 7 km from the S-3 ponds. All four strains were also located in a least one highly contaminated well next to the S-3 ponds. For example, the highly contaminated wells include FW106, FW109, and FW104, all within 10 m of the S-3 ponds. MT049 made up 0.026% of the ESV reads in FW106, MT058 made up 0.051% of the reads in FW106, MT094 made up 1.21% of the reads in FW109, and MT123 made up 7.0% of the reads in FW104. Mapping the location of the isolates by ESV back to the ORR environment showed that strains capable of surviving in the contaminated environment were not restricted to this specialized environment. Additionally, enrichment and isolation of nitrate-reducing strains under environmentally relevant contaminating conditions produces metal-resistant strains, even from the noncontaminated environment.

High-throughput enrichment cultures (1.75 ml in volume) were set up in autoclave-sterilized 2-ml Fisherbrand 96-well DeepWell polypropylene microplates (Waltham, MA). Base medium contained 0.25 g/liter NH₄Cl, 0.1 g/liter KCl, 0.5 g/liter MgSO₄·7H₂O, 0.37 g/liter CaCl₂·2H₂O, 0.25 g/liter NaCl, 2.5 g/liter NaHCO₃, and 0.7 g/liter NaH₂PO₄·H₂O, as well as vitamins and trace elements prepared as described by Widdel and Bak (41). The pH of the base medium was adjusted using HCl or NaOH to the indicated pH before filter sterilization. Medium that was adjusted to pH 5.5 also contained 10 mM morpholineethanesulfonic acid buffer. Where indicated, the base medium also contained 1 g/liter yeast extract. The various carbon sources used in the enrichment cultures were added as sterile stock solutions to the bottom of the microplate wells, resulting in final concentrations of 10 mM glucose, 10 mM cellulose, 20 mM lactate, 20 mM fumarate, 20 mM pyruvate, 30 mM ethanol, 30 mM acetate, 100 mM formate, or 0.5%, vol/vol, compost tea. A mixture of 10 mM lactate and 10 mM fumarate was also used for some enrichments. Electron acceptors NaNO₃ and MgSO₄·7H₂O were also added as sterile stock solutions to the bottom of the microplate wells where indicated, resulting in the given final concentrations. Environmentally relevant metals (COMM) were also added to the bottom of the microplate wells using a 10 mM sterile solution of uranyl acetate and a sterile 100× metal mix solution (described below). For enrichments containing COMM, AlK(SO₄)₂·12H₂O powder was suspended in the medium at the desired concentration before adding base medium to the microplate wells. Depending on the pH of the base medium, the Al did not always dissolve completely. Once the base medium was added to the microplate wells containing the various additions, each well was inoculated with 50 to 150 μl of the indicated ORR groundwater sample or 0.1 g of the indicated ORR sediment sample. The microplates were then covered with PHENIX sterile microporous sealing films (Candler, NC) and incubated in the dark at 22°C. Anaerobic enrichment incubations were carried out in sealed anaerobic jars under an atmosphere of 20% CO₂ and 80% N₂, while microaerophilic enrichments where similarly incubated but with approximately 3% O₂ in the gas phase, which was exchanged every 24 h. The enrichments were routinely checked for growth over the course of 3 months. Isolates were obtained from enrichments displaying growth by streak plating on solid media similar to the composition of the enrichment culture or using a rich LB medium and culturing isolated colonies.

The 1× COMM is defined as containing 1 mM AlK(SO₄)₂·12H₂O, 100 μM UO₂(CH₃COO)₂, 100 μM MnCl₂·4H₂O, 100 μM NiSO₄·6H₂O, 30 μM CoCl₂·6H₂O, 10 μM (NH₄)₂Fe(SO₄)₂·6H₂O, 10 μM CuCl₂·2H₂O, and 5 μM Cd(CH₃COO)₂·2H₂O. The Al and U in the COMM were added to cultures separately as a dry powder and a 10 mM filter-sterilized stock solution, respectively. The remaining metals were added to cultures using a 100× filter-sterilized stock solution. This solution was either prepared fresh or stored in one-use aliquots at −80°C. As indicated, permutations of the COMM were used in some experiments in which Al and/or U were removed from the mix or were added at lower concentrations.

The compost tea was prepared from finished compost obtained from the ACC Commercial Composting Facility (Athens, GA). Finished compost (1.0 kg) was mixed with 5 liters of distilled H₂O and allowed to steep for 4 days at room temperature in a sealed glass container. The resulting compost tea was filtered through cheesecloth before filter sterilization with a Millipore express 0.22-μm filter (Burlington, MA).

High-throughput growth curves (400 μl in volume) were conducted using a Thermolab Scientific Equipments Bioscreen C automated microbiology growth curve analysis system (Maharashtra, India). For anaerobic growth, the Bioscreen C was placed in a Plas Labs (Lansing, MI) hard glove box with an atmosphere of 5% H₂ and 95% Ar. Growth curves were conducted at 22°C with normal speed shaking for 15-s intervals with 10 s between shaking. Growth was monitored at an optical density of 600 nm (OD₆₀₀) every hour. The same base medium used in the enrichment experiments was used for the growth curves, with additions including carbon sources, electron acceptors, and metals being handled in the same fashion. The carbon source and pH conditions of the media used for each strain are described in Table S2 in the supplemental material. All growth curves containing COMM at the indicated concentrations lacked Al, as there were precipitation issues that interfered with OD readings. Growth data were analyzed using the grofit package (42). The model-free spline fit was used to determine growth rates for each curve, which were then used to fit dose-response curves and calculate EC₅₀ values when applicable. The 95% CI were determined using a bootstrap value of 100. No yeast extract was added to growth curve experiments.

Whole-cell nitrate and nitrite reductase assays were performed on cells grown in the presence and absence of 1× COMM (43). In the case of strain MT123 grown with COMM, the concentration of U was decreased to 10 μM instead of 100 μM due to the sensitivity of MT123 to U. For the assays, cells were grown anaerobically in 50-ml cultures at room temperature without shaking on base medium containing 20 mM nitrate and 1 g/liter yeast extract to late-log phase. The carbon source and pH conditions of the media used for each strain are given in Table S2. Cells were harvested at 10,500 × g for 10 min before washing three times with 5 ml of assay buffer (50 mM potassium phosphate buffer, pH 7.2) and suspending in assay buffer to give a final OD₆₀₀ value between 0.1 and 0.4. In a 15-ml Falcon tube, 200 μl of the suspended cells was combined using gentle mixing with 25 μl of freshly prepared 0.5 mg/ml methyl viologen and 75 μl of a solution containing 4 mg/ml sodium dithionite, 4 mg/ml sodium bicarbonate, and either 100 mM NaNO₃ for nitrate reductase assays or 1.5 mM NaNO₂ for nitrite reductase assays. The assay mixture was incubated at room temperature for 5 to 10 min before vortexing in air until the solution became clear to stop the reaction. One ml of 1% sulfanilic acid in 20% HCl was then added with vortexing, followed by 1 ml of a 1.3 mg/ml N-(1-naphthyl) ethylenediamine-HCl solution, forming a nitrite-dependent azo dye that is red in color. The OD₅₄₀ and OD₄₂₀ of the assay solution were taken to measure the azo dye and absorbance due to light scattering, respectively. Nitrate reductase activity is reported in units/OD₆₆₀, where units are calculated using the formula 100 × [OD₅₄₀ − (0.72 × OD₄₂₀)]/(T × V), where T is time in minutes and V is reaction volume in milliliters (43, 44). Nitrite reductase activity is also reported in arbitrary units using the same equation, except the change in OD₅₄₀ between the final assay solutions and a no-cell control is used in place of the OD₅₄₀. Errors are reported as the standard deviations between three biological replicates.

Genomic DNA was isolated from each ORR strain using the ZymoBead genomic DNA kit (Irvine, CA). The 16S rRNA gene for each strain was amplified using EmeraldAmp GT PCR master mix (Kusatsu, Shiga Prefecture, Japan) with primers 8F (AGAGTTTGATCCTGGCTCAG) and 1492R (ACGGCTACCTTGTTACGACTT). Amplicons were sequenced by GENEWIZ (South Plainfield, NJ), and the sequences were analyzed for closest 16S rRNA gene match using nucleotide BLAST.

The 16S rRNA gene sequences were taken from a previous groundwater survey of 93 different noncontaminated and contaminated ORR wells (27). Read-paired merged rRNA sequences were downloaded from MG-RAST (mgm4554509.3). Low-quality sequences (>1 expected error) were eliminated using usearch-fastq_filter, and the reads were trimmed to cover the region between the V4 primers with a custom Perl script. Usearch-fastx_uniques was used to count the number of times each ESV occurred in each sample. The ESVs were then ranked by total abundance across the samples, and only the top 5,000 ESVs were retained to reduce contamination. Usearch-search_exact was used to find exact matches between the isolate 16S rRNA sequences and the 100-well V4 regions. In many cases isolates matched to multiple ESVs in each sample, so the matching ESV with the most reads per sample from a given well was selected. The highest percent abundance calculation is based on the ESV reads per sample divided by the total number of sample reads. Only ESVs with >1 read per sample were used for calculations and plots; however, all ESVs were used for maps. These aggregated data were combined with the longitude and latitude of each well into a Google Fusion Table (Mountain View, CA) for geospatial maps. Abundances on maps below 0.01% are not quantitative and are shown for illustrative purposes.