Kisameet Glacial Clay: an Unexpected Source of Bacterial Diversity

ABSTRACT Widespread antibiotic resistance among bacterial pathogens is providing the impetus to explore novel sources of antimicrobial agents. Recently, the potent antibacterial activity of certain clay minerals has stimulated scientific interest in these materials. One such example is Kisameet glacial clay (KC), an antibacterial clay from a deposit on the central coast of British Columbia, Canada. However, our understanding of the active principles of these complex natural substances is incomplete. Like soils, clays may possess complex mixtures of bacterial taxa, including the Actinobacteria, a clade known to be rich in antibiotic-producing organisms. Here, we present the first characterization of both the microbial and geochemical characteristics of a glacial clay deposit. KC harbors surprising bacterial species richness, with at least three distinct community types. We show that the deposit has clines of inorganic elements that can be leached by pH, which may be drivers of community structure. We also note the prevalence of Gallionellaceae in samples recovered near the surface, as well as taxa that include medically or economically important bacteria such as Actinomycetes and Paenibacillus. These results provide insight into the microbial taxa that may be the source of KC antibacterial activity and suggest that natural clays may be rich sources of microbial and molecular diversity.

each community revealed differences in the distribution of putative iron-oxidizing Gallionella spp. Bacterial community structure did not correlate with antibacterial activity. However, we found evidence that the KC microbiome includes potentially valuable species, such as Actinobacteria. This initial microbiological survey of a novel environment provides a framework for understanding the basis for the unique properties of KC.
biome characterization (listed in Table S1 in the supplemental material). Two of the samples, Kis3-0 Org and Kis5-0, consisted primarily of organic material from overburden (Kis3-0 was separated into "organic" and "clay" fractions). We also analyzed KC35, a sample from the surface that has been previously shown to have strong antimicrobial activity (20).
Physicochemical characteristics of the deposit. We first performed a general characterization of the physicochemical properties of KC. Mineralogical analysis of selected core samples showed that KC contains phyllosilicates such as biotite (~3 to 15%) and relatively small amounts of clay minerals such as illite (~5%) and clinochlore (~5 to 13%) (Table S2). We did not observe any major differences between the selected core samples and KC35. Likewise, analysis of acid-digested elements in KC core samples and elemental analysis of bulk clay did not reveal major variations (Table S3) but confirmed previous observations of significant amounts of Fe (~10 4 mg/kg). Because speciation could presumably affect the antibacterial activity of toxic metals, including Fe, we also determined the redox potential of each sample (Table S3). The measured potential of the core samples varied relatively widely from ϩ19 mV to a maximum of ϩ373 mV, with the redox of KC35 being ϩ427 mV.
In contrast, comparison of the pH of core sample suspensions revealed more marked differences. While previous studies suggested that KC is mildly alkaline (pH~8) (15), the low pH of KC35 we measured previously (20) suggests that some samples could be markedly more acidic. Measurement of different preparations of KC core samples (fresh aqueous suspensions, aqueous leachates, or suspensions of clay that had been stored exposed to air for approximately 1 month) suggested that the clay pH can vary depending on (i) depth of the sample in the deposit and (ii) exposure to air. First, the pH of fresh KC suspensions revealed a strong trend with depth ( Fig. 2A and Table S4), with the pH of clay from greater depths significantly higher than for samples from depths of 0 to 8 ft (pH 9.5 versus pH 8.0; P Ͻ 0.0001). Also, while the pH of prepared aqueous leachates showed the same positive correlation between increasing pH and depth, we also noted that in general, aqueous leachates showed a lower pH than their suspension counterparts ( Fig. 2A). This may reflect a lowering of pH during the 24 h of incubation under aeration during leachate preparation. In support of this, remeasurement of suspension pHs a few months after core cutting resulted in markedly higher levels of acidity. In addition, the pH of KC35, which was harvested at least 2 years before the core samples and stored at 4°C, had an extremely low suspension pH of 4.3.  Together, this suggests that the pH of KC can vary markedly depending on depthrelated factors, as well as exposure to air.
KC microbiomes cluster into three distinct community types that correlate with depth and pH. DNA was extracted from each core sample, and community profiling was performed by sequencing 16S rRNA amplicons. We identified 5,032 unique operational taxonomic units (OTUs) (3% genetic distance definition), with a range of 281 (Kis1-28) to 2,410 (Kis3-0 Org). Principal-coordinate analysis (PCoA) of unweighted UniFrac distances (25) was performed to compare beta diversity of KC communities. The first two components accounted for 21.09% and 8.59% of the difference between the communities (Fig. 2B). The KC communities grouped into three clusters (Table S1). The PCoA plot was first overlaid with depth information (i.e., 0 ft, 4 ft, and so on). This suggested that two of the clusters included most of the shallower samples (0 to 8 ft), as well as KC35 (also from the surface) (Fig. 2B). We therefore refer to these "shallow" clusters as SA and SB. In contrast, the third cluster, termed D ("deep/interior"), included mostly samples from 12 ft and below, although some surface samples, such as Kis1-0 and Kis3-4, were in this cluster. For surface (0 to 8 ft) samples in clusters SA and SB, there may also be an effect of the horizontal location of the sample. For example, cluster SA samples were all from Kis1, Kis3, and Kis5 (edge of the deposit [ Fig. 1A]), whereas cluster SB samples were from Kis2 and Kis4 (closer to the interior). When the leachate pH was overlaid onto PCoA analyses, we saw a trend reminiscent of depth analysis, where more acidic samples were part of either cluster SA or SB (Fig. 2C). Statistical analyses also suggested that both cluster SA and cluster SB samples had a significantly lower pH than cluster D (P ϭ 0.047 and P ϭ 0.0073, respectively), while the pH of clusters SA and SB was not significantly different (P Ͼ 0.5). This suggests that a depth-driven cline in pH affected the microbiome of KC. The KC deposit has a significant organic overburden of macroflora (Fig. 1B), which might decrease the pH at the surface. The KC deposit therefore harbors several different bacterial communities, likely shaped by environmental parameters that change with depth.
KC harbors rich bacterial diversity that is dominated by Betaproteobacteria. As noted above, we identified more than 5,000 unique OTUs, with a range of 281 (Kis1-28) to 2,410 (Kis3-0 Org). Alpha rarefaction analysis of the sequence data showed that our survey is likely an underestimation of the diversity of some samples (Fig. S1). Nonetheless, we compared the bacterial diversity of each KC sample (observed species and Shannon index) (26) at the same sequencing depth, not including singleton and doubleton OTUs (Fig. 3A). The KC deposit has a layer of overburden (Fig. 1B), and samples that were composed of almost entirely organic material (Kis3-0 Org and Kis5-0; greater than 90% amorphous by X-ray diffraction [XRD]) had a higher Microbiome of Kisameet Clay ® mean observed species (mean 1,251 OTUs versus 431 for clay samples; P ϭ 0.0015) and Shannon index (6.2 versus 3.6 for clay samples; P ϭ 0.0003) metrics. Among the clay samples, there was also a minor trend toward lower diversity measures at greater depths (Fig. 3A). We next identified the major bacterial phyla in KC samples. Most of the reads (88.3%) were assigned to Proteobacteria OTUs. Other significant phyla identified included Bacteroidetes (3.2%), Actinobacteria (2.4%), Acidobacteria (1.8%), and Firmicutes (1.3%) (Fig. 3B). Further separation of the Proteobacteria into constituent classes revealed that approximately 63.5% of reads for this phylum could be assigned to Betaproteobacteria, followed by Gammaproteobacteria (21.3%), Alphaproteobacteria (4.0%), and Deltaproteobacteria (0.4%). These results are consistent with other surveys of bacteria in soil environments, which have shown that most are dominated by Proteobacteria (27). Acidobacteria and Actinobacteria are also common in soils. Thus, KC harbors a surprising number of different bacterial species, with a large contribution from Proteobacteria.
The core bacterial communities of KC. In general, Betaproteobacteria and Gammaproteobacteria species dominated the sequences obtained for all KC samples (Fig. 3B). However, three distinct communities may be present in KC ( Fig. 2B and C). To identify taxa that might reveal differences between these communities and/or their environments, we defined a core bacterial community (CBC) for each of the three PCoA clusters. We defined CBCs based on common membership (i.e., an OTU was part of the cluster CBC if at least one sequence read was present in all samples of that cluster). CBC OTUs were also limited to those that made up at least 0.1% of the total reads in the cluster being compared. We chose this approach because of limitations in sample harvesting and storage (see Materials and Methods), and thus, we did not take into account the proportion of the community made up by an OTU, as has been done previously (28).
Next, the CBCs of clusters SA, SB, and D were compared based on the percentage of CBC OTUs belonging to different phyla (Fig. 4A, right panel). This revealed that CBC-SA, with the majority of reads from Betaproteobacteria (53.8%), actually had a similar proportion of OTUs from this group (21/85 CBC OTUs [24.7%]) as from Alphaproteobacteria (18/85 [21.1%]) or Acidobacteria (17/85 [20%]). Likewise, CBC-SB, while having more than 90% of reads from Betaproteobacteria, had a comparatively large proportion of Firmicutes (Clostridiales accounted for 5/26, or 19.2%, of OTUs). Finally, CBC-D had more Gammaproteobacteria species (9/18 OTUs) compared to the CBCs of SA and SB. Definition of the core bacterial species of each KC community therefore suggested that the Betaproteobacteria dominated all three communities but with significant differences between the constituent families of species. We next identified shared or unique species between the three CBCs (Table S5). Only three OTUs were shared by all three CBCs (one Oxalobacteraceae and two pseudomonads). Both taxa include common soil bacteria with diverse metabolic capabilities (29,30). CBC-SA and -SB (both shallow/low pH) shared more OTUs with each other (10 OTUs) than either one shared with cluster D (5 OTUs each [ Fig. 4B]). Three of the OTUs shared by only the CBCs of SA and SB (but absent from CBC-D) were from Alphaproteobacteria lineages, including nitrogen-fixing Rhizobiales normally associated with plant roots, which suggested that the shallower communities may be influenced by the rhizosphere. A notable difference between CBC-SA and CBC-SB was the number of Acidobacteria OTUs (20% versus 4%, respectively). In contrast, we observed more Clostridiales OTUs in CBC-SB (20% versus 3.5% for CBC-SA). The five Clostridiales OTUs of CBC-SB included two potential Desulfosporosinus species and one Thermosinus species, both of which include sulfate-reducing bacteria (31)(32)(33). For the CBC of cluster D, there was greater representation from Gammaproteobacteria (such as xanthomonads and pseudomonads), Acinetobacter, and Serratia. Also uniquely present in CBC-D was a member of the Anaerolineae, a Chloroflexi whose cultivated members appear to be strictly anaerobic (34,35).
Iron bacteria mark the KC microbiome near the surface. Acid-soluble metal ions, such as iron species, are important for the activity of other antibacterial clays (13), and may be responsible for the activity of KC (S. Behroozian and J. Davies, unpublished data), which is relatively rich in iron (15,22). To gain insight into the ferruginous environment of KC, we focused on clay bacteria that may play roles in iron cycling. Interestingly, many iron bacteria (those that perform dissimilatory iron oxidization at near neutral pH) that have been characterized are Betaproteobacteria (36,37), which were dominant in KC. Numerous comamonads (142 OTUs), which may include ironoxidizing Leptothrix spp., were identified in the total sequence data set, and several were defined as part of CBC-SA, CBC-SB, and CBC-D (Table S5). We also identified genera of iron bacteria or bacteria that can tolerate heavy metals such as Delftia (38) and Acidivorax (36).
Notably, a Gallionellaceae OTU was shared only by CBC-SA and CBC-SB but was not present in the CBC we defined for cluster D. The type species of this family (Gallionella ferruginea) colonizes the transition zone between anaerobic and aerobic environments (redox potential of ϩ200 to ϩ320 mV and low oxygen tensions of 0.1 to 1 mg/liter) and participates in dissimilatory iron oxidation, requiring Fe(II) and oxygen for growth (39; reviewed in references 36 and 37). For KC samples, we observed a trend of decreasing redox potential from 0 to 12 ft (approximately ϩ300 to ϩ50 mV) and an increase in redox from 20 to 36 ft (ϩ50 to ϩ200 mV) (Fig. 5). Interior samples tended to have a lower potential (ϩ50 to ϩ150 mV). Samples with at least 0.1% of reads from Gallion-ellaceae OTUs (all 14 identified in our studies) were from 12 ft and above. The absence of Gallionella below 12 ft is consistent with redox measurements of these samples, which were below ϩ200 mV and the previously reported redox niche of these bacteria (39). When the pH of each clay sample was included, all samples with Gallionella had circumneutral, or even slightly acidic pH (Fig. 5). In addition, all samples that had at least 0.1% of reads as Gallionellaceae species were part of clusters SA and SB, except for Kis1-12 (Fig. S2).
Antibacterial and economic potential of KC. KC exhibits potent inhibitory activity against a variety of bacteria and other microbes, including multidrug-resistant ESKAPE pathogens (20) and pathogenic fungi (Behroozian et al., unpublished). To gain insight into what might contribute to the antibacterial activity of KC and how this activity might shape its resident communities, we also determined the antibacterial activity of each sample used for microbiome analysis against E. coli MG1655. The activity of KC core samples was highly variable (Fig. 6A). While KC35 caused an approximately 5-log-unit decrease in CFU during incubation, the KC core samples varied from bactericidal (purple squares show greater than 3-log-unit killing) to more weakly active (blue squares show less than 1,000-fold decrease in CFU, but higher than water alone) (Fig. 6A). Some samples actually enhanced viability of the test inoculum compared to the H 2 O control (black squares) (Fig. 6A). While we did not identify a correlation between community type and activity, even though depth appeared to affect microbiome composition, all samples from the interior (i.e., 4 to 12 ft) showed little or no activity. There was also no correlation between activity and core (i.e., Kis3 versus Kis1 [not shown]). Finally, we sought to determine whether activity correlated with the type of bacterial community type identified by PCoA analysis (Fig. 2B and C), as we hypothesized, for example, that communities marking a particular redox environment could affect the speciation of antibacterial Fe ions. However, there was an almost random distribution of high-activity samples (Fig. 6B). There was a very weak correlation between redox potential and antibacterial activity, with samples with higher redox potential tending to show greater reduction of CFU (Fig. S3), while the presence of Gallionella showed no correlation with activity.
Soils are rich sources of economically valuable bacterial species, such as those that produce novel antimicrobials or with plant growth-promoting activities. Up to 3% of reads for KC samples were assigned to the Actinomycetales (Fig. S4A). This group includes many well-known producers of bioactive and antimicrobial secondary metabolites (40). Several species of Paenibacillus, a genus that includes nitrogen fixers with plant growth-promoting ability such as Paenibacillus polymyxa (41), are also present in KC (Fig. S4A). Preliminary culturing studies isolated viable bacteria on Actinomyceteselective agar at relatively high CFU (Fig. S4B).

DISCUSSION
Characterizing resident microbial populations is essential for a complete understanding of terrestrial environments and systems and identification of microbes with economically useful capabilities. Here, we have characterized the bacteria that colonize the Kisameet Bay glacial clay deposit, which is the source of potently antibacterial KC. The deposit harbors surprising bacterial diversity, with greater than 300 species in most samples and thousands of species in samples mixed with the organic overburden. In fact, our estimates may be an underestimate due to sequencing depth. Previous analyses of the bacterial diversity present in Boom clay borehole water found that this environment contained only 100 to 150 OTUs (42). Diversity in the Boom clay microbiome correlated weakly with total organic carbon measures (42). While we did not measure organic carbon in this study, the organic overburden is likely responsible for the increase in diversity in some samples from the deposit. Soil pH is also a strong predictor of bacterial diversity, with neutral environments harboring more diversity and species richness than environments with either low or high pH (43). The high pH of the interior of the KC deposit is consistent with its slightly lower bacterial diversity. In iron-rich freshwater wetlands, bacteria such as Gallionella tend to be more abundant in the spring compared to summer or fall (44,45). Sampling of KC in different seasons may reveal variable temporal bacterial community dynamics. Profiling of resident Archaea, Eukarya, or bacteriophages will also further an understanding of the microbial diversity of KC.
To date, there have been few studies of the bacterial content of native clays. Bacteria enriched from a desiccated sample of commercially available Wyoming bentonite clay were found to be closely related to Desulfovibrio africanus, a sulfate-reducing deltaproteobacterium (46). We did not detect this species in our survey; however, the deeper samples of cluster D did share a deltaproteobacterium OTU. Two studies have addressed the microbiome of the deep subsurface Boom clay. A 16S rRNA clone library approach tentatively identified 11 different OTUs in the clay (47). Interestingly, Proteobacteria made up 76% of the species, with Betaproteobacteria, such as the comamonad Acidovorax, accounted for 46%. Our survey of KC bacteria likewise suggests a dominance of Betaproteobacteria. In addition, both the CBCs defined in our survey and the FIG 6 Antibacterial activity of KC core samples. Suspensions (10 mg/ml) were prepared from dried clay in water from each core sample and tested for antibacterial activity against E. coli MG1655. Bacteria in log phase were added to suspensions and incubated at 37°C with shaking. After 24 h, samples were serially diluted and plated for CFU on LB agar. (A) Higher antibacterial activity is detected at the edges of the deposit. The mean fold reduction (n ϭ 4) for each depth (short lines) and the value for each individual sample (symbols) are plotted. Organic samples and KC35 are plotted separately. Samples with weak activity (blue squares) were defined as those resulting in at least a 66-fold decrease in viable CFU after 24 h, the reduction observed for phosphate buffer (pH 4.3). "Bactericidal" samples (Purple squares) were defined as those reducing viable CFU by at least 1,000-fold. For comparison, the activity of KC35 and water alone are indicated. (B) KC community structure is not predictive of clay activity. The results of unweighted UniFrac PCoA analysis of core sample communities with respect to activity against E. coli MG1655 are indicated as follows: weak activity (Ͼ66-fold reduction in CFU); bactericidal (Ͼ1,000-fold reduction); not active (less activity than phosphate buffer). species identified in the Boom clay revealed Gammaproteobacteria such as Pseudomonas and Acinetobacter.
We note that KC harbors at least three major bacterial community types. Two different communities (SA and SB) were identified in more shallow/lower pH samples. Comparison of the CBCs of these clusters showed that cluster SA had more Acidobacteria (27,48,49) than cluster SB. In contrast, acidobacteria were absent from the CBC defined for the deeper cluster D. The Acidobacteria is a relatively new, poorly characterized, and diverse phylum, although members of this group have been identified in many different environments, where they may constitute up to 20% of bacteria (27,49,50). We note that the two samples composed of mostly organic material were part of cluster SA. Cluster SA may represent communities in soil like KC under higher organic material or oxygen tension. Acidobacteria were not reported in the Boom clay (42,47) but were found to be relatively abundant in 16S rRNA libraries prepared from low-pH, clay-rich Cerrado soil (51). The presence of this clade in CBC-SA may therefore reflect a more soil-like, rather than clay-like, environment. Certainly, the correlation between the horizontal location of the core and the PCoA cluster for shallow samples (i.e., the tendency of surface samples from cores extracted closer to the edge of the deposit, such as Kis1, Kis3, and Kis5, to be part of cluster SA) provides an impetus to explore correlations of these communities with their macroflora neighbors.
In contrast to CBC-SA, CBC-SB included four potential sulfate-reducing clostridiathree Desulfosporosinus species and a Thermosinus species. Boom clay likewise included numerous clostridia such as sulfate-reducing Desulfotomaculum (42). Sulfate-reducing bacteria participate in anaerobic dissimilatory reduction of sulfate to compounds such as hydrogen sulfide, and are often associated with decaying organic matter (52). Thus, cluster SB may reveal environments that are more anaerobic, while still in proximity to the overburden layer of the deposit. Interestingly, cyclic S 8 extracted from KC with organic solvents (J. Tan and J. Davies, unpublished observations), has antibacterial activity (53)(54)(55). Therefore, microbial sulfur cycling may contribute to the complexity of KC. Furthermore, Desulfosporosinus and Thermosinus strains have been identified that can reduce not only sulfate but also Fe(III) (56,57), supporting the presence of both these Clostridia and iron bacteria in KC.
KC is iron-rich (15,22), and potential iron bacteria were identified in our survey. In particular, Gallionella-like iron bacteria are found in shallower samples of KC. Gallionella spp. are microaerobic iron oxidizers that live at iron-rich oxic/anoxic interfaces (36). While they have been described as inhabiting neutral pH zones, acid-tolerant species of Gallionella have been identified, and Gallionella-like organisms are increasingly found in heavy metal-rich acid mine drainage (58)(59)(60)(61)(62)(63)(64). Bulk clay iron levels are relatively uniform across the deposit (data not shown), but the low pH of some regions may affect the solubility and availability of iron species. KC provides an interesting environment for studying iron bacteria, which are associated with biofouling and the deterioration of pipes and water distribution systems (36), and the 14 Gallionellaceae OTUs identified in KC may represent species with novel characteristics. Iron-oxidizing bacteria have been shown to have potential bioagents removing heavy metals (65).
What makes KC antibacterial? Previous characterization of density-separated KC fractions showed that antibacterial activity was higher in those fractions that had higher water-leachable metal content (W. Xu, S. Behroozian, W. Yenjaichon, J. Grace, J. Davies, and L. Li, unpublished data). Low pH also affects KC activity, which may contribute to the solubility of ions such as Fe(II), Fe(III), or Al(III) (Behroozian et al., unpublished). We found a weak relationship between redox condition, which might affect the speciation or solubility of antibacterial ions, such as iron, and antibacterial activity. Alternatively, specific redox regions may indicate regions of the clay where antibacterial activity resides. For example, Jordan's red clay harbors bacteria of the genus Lysobacter, xanthomonads that are important biocontrol agents and sources of exoenzymes and antibiotics (66), which appear to be responsible for its antibacterial activity (19). However, we found no strong correlation between bacterial community type and activity of the clay. Highly active KC35 had an increased number of Gallionella reads, but we found no correlation between Gallionella and activity in core samples. Acidobacteria are known to produce antibiotics such as polyketides (67). KC awaits more-focused culturing studies for the identification of useful bacteria. While we cannot rule out the possibility that particular bacterial species contribute to its inhibitory properties, the bioactivity of KC does not strongly shape community structure.
KC is a very complex natural material, and many different factors may contribute to its unique properties, including the resident bacterial communities. If clays and other natural products are to be used as therapeutic agents, their resident bacteria must be characterized, as they may alter clay structure and affect the efficacy and/or consistency of preparations. KC bacteria should be examined in more detail for the production of novel bioactive compounds or useful metabolic capabilities. In addition, future microbiome studies of natural environments, such as the KC deposit with the unique chemical, absorptive, diffusion, and microbial properties of KC, may uncover novel microbial interactions involving small bioactive molecules, antimicrobial metal ions, nutrients, and oxygen availability. The KC deposit may serve as an environment in which to observe novel chemical and genetic interactions between microbes, to study ecological succession, and to reveal new bacterial types with valuable economic properties.

MATERIALS AND METHODS
Harvesting and storage of core samples. Five vertical cores (Kis1 to Kis5) were harvested from the Kisameet Bay deposit (51°58=16Љ N/127°52=56Љ W) on 24 October 2012. Following transport to the University of British Columbia, cores were stored sealed at 4°C under normal atmospheric conditions in the dark. The cores were opened, and the top 10 cm from each 4 ft length of core was saved for analyses of physical, chemical, and biotic characteristics. To avoid cross-contamination between depths, only the interior undisturbed part of each core was used for analysis. To measure antibiotic activity, a portion of each core sample was dried in a desiccator at room temperature under vacuum for 2 to 4 days, ground to a fine powder with a mortar and pestle, autoclaved for 20 min, and stored at room temperature in the dark prior to testing. For metagenome analysis and culturing studies, unautoclaved, native clay was used. The sample activity, physicochemical characteristics, and bacterial community were compared to that of the previously characterized, highly active sample KC35 (20).
Measurement of KC mineralogical and physicochemical characteristics. Details of the analysis of the abiotic properties of KC core samples (mineralogy by X-ray diffraction, pH, redox, and elemental analyses by inductively coupled plasma optical emission spectroscopy) are presented in Text S1 in the supplemental material.
Extraction of DNA from KC. Total DNA was extracted from wet, unautoclaved clay from each core sample using the FastDNA kit for soil (MP Biomedicals, Santa Ana, CA) according to the manufacturer's instructions with minor modifications. Briefly,~400 mg of wet, unautoclaved clay was added to a tube with~500 mg of silica beads (1-mm diameter; Sigma). Samples were shaken on a FastPrep homogenizer (MP Biomedicals) for 40 s at a speed of 6.0. Following removal of particulate matter at 14,000 ϫ g for 10 min, the clarified extract was applied to a spin column and purified according to the manufacturer's recommendations. DNA was eluted with water and quantified using a Nanodrop spectrophotometer.
Next-generation sequencing of KC 16S rRNA amplicons and sequence data analysis. Sequencing of the 16S rRNA gene variable 3 (V3) region, amplified by PCR with primers 341F (F stands for forward) and 518R (R stands for reverse) from total DNA extracted from KC core samples, was performed as previously described (68,69) on an Illumina Miseq platform. Details of library preparation and sequencing can be found in Text S1. Details of data quality control and analysis using Qiime (26) are likewise outlined in detail in Text S1.
Isolation of KC-resident bacteria. Viable bacteria from KC core samples and KC35 were cultured from unautoclaved clay. To determine viable counts of Actinomycete-like bacteria, equal weights of wet unautoclaved clay from each core sample that had been stored under native atmospheric conditions at 4°C (approximately 100 mg) were suspended in 10-ml portions of sterile saline, vortexed, and serially diluted in sterile saline. Dilutions were plated on Actinomycete isolation agar (AIA) and incubated at 30°C for 4 to 6 days until colonies were visible.
Antibacterial activity of core samples. Quantification of antibacterial activity was performed essentially as previously described (20) using the wild-type Escherichia coli K-12 strain MG1655 (70), grown aerobically at 37°C in Luria-Bertani broth (LB). Each clay sample was tested at a concentration of 10 mg/ml, as described in Text S1.