Presque Isle, Lake Erie, in Northern Pennsylvania, possesses a unique diversity of native animal and plant species (Discover Presque Isle Park Info) but has a microbial population that is not well understood. Since this location is predominantly sand with some loam, coastal areas where the soil has not been supplemented with either humus or fertilizer would be expected to be fast-draining and nutrient-poor with only a few dominant microbial species. Such locations might yield bacteria secreting narrow-spectrum antibiotics directed against the few microbes able to survive these hostile environments. Conversely, inland trails and tourist attractions, which have been landscaped and fertilized, would be more likely to possess a diverse population of microbes, including bacteria producing broad- rather than narrow-spectrum antibiotics.

To test these hypotheses, we investigated the microbial flora at twenty sites in and around the Presque Isle area, which included coastal sites on the Presque Isle bay, inland trails, and points of scenic interest, at the end of June 2002. Although both fungal and bacterial species are known to produce antibiotics, fungi tend to produce mostly broad-spectrum activities, and, in addition, far more antibiotics are produced by bacteria (Salyers and Whitt 2001). Our study focused instead on the recovery and isolation of bacterial species able to produce narrow-spectrum antibiotics, which are primarily in the hardy, spore-producing Bacillus and Actinomyces genera (Atlas 1995). To characterize the type of microhabitat favored by both narrow- and broad-spectrum antibiotic-secreting bacteria, chemical analyses of the soil for nitrogen, phosphorus, potassium, humus, and pH were made in addition to enumeration of the total numbers of bacteria, fungi, yeasts, staphylococci, and coliforms (both environmental and fecal), using standard microbiological techniques.

Table 1. Soil and Location Characteristics of Presque Isle Samples Studied (Click to view enlarged table)

Bacterial and fungal levels
The microbial content of the soils was determined by plating 100 µL of 10^-2 suspensions in sterile saline, using nutrient, Sabouraud-dextrose MacConkey, and mannitol agars (Tortora 1998). Total counts of all non-fastidious bacteria and mycetes were determined from the nutrient agar plates, with colonial morphology, and microscopic appearance following staining being used to differentiate between bacteria, yeasts, and molds. Sabouraud-dextrose agar was used to determine total fungal numbers. The remaining two agars were differential: MacConkey agar allowed for differentiation between environmental coliforms and non-coliforms, and mannitol salt between pathogenic and non-pathogenic staphylococci. All plates were incubated at 30° C. The number of colonies appearing after 48 hours was determined by direct visual assessment. Each soil sample was replicated a minimum of three times to determine consistent bacterial and fungal numbers. Fecal coliforms were differentiated from environmental coliforms according to Geldreich (1965). To differentiate Staphylococcus aureus from other staphylococcal species, a coagulase slide test using rabbit plasma was employed according to Pierce (1999).

Isolation and analysis of activity of antibiotic-producing isolates
Actinomycete/Bacillus species were isolated by the embedded agar method described by Atlas (1995). A panel of four commercially obtained human bacterial pathogenic strains, Bacillus subtilis, Escherichia coli, Pseudomonas aeruginosa, and Staphylococcus aureus, was used as the test bank. Antibiotic production by a colony was defined as the inhibition of embedded bacterial growth by a ≥ 3 mm ring around the soil bacterial isolate on at least six separate occasions. Narrow-spectrum activity was putatively defined as that inhibiting one of the four human pathogens, with broad-spectrum inhibiting two or more. Soil isolates impairing the growth of one or more human pathogens were putatively identified as either Actinomyces or Bacillus species by standard microbiological techniques and the Bacilli typed using a commercial miniaturized API 50CHB test system (Biomerieux-Vitek, Hazelwood, MO) (Schraft et al. 1996).

Statistical analysis
Bivariate correlation statistics using the Pearson coefficient were performed (SPSS for Windows Version 6.13, 2001) to determine the relationship between microflora levels at the 20 Presque Isle sites. Comparison statistics of the distribution of narrow- and broad-spectrum antibiotic producers relative to other microflora, and of coastal versus inland microflora, were made using nonparametric Wilcoxon Signed Ranks Test and parametric paired T-tests, respectively (Minitab for Windows 2000 Version 5.0, 2002). 

Figure 1. Differential soil microflora at sites from which bacteria secreting narrow- and broad-spectrum antibiotic activity were recovered, compared to remaining sites. Values shown represent the mean of three experiments, where 100 µl of a 10>-2 dilution of soil in sterile saline was plated, and results are given as CFU/ml ± SEM. Data indicated with an asterisk (*) are significant at p < 0.001.

Figure 2. Spectra of antibiotic activity displayed by Actinomyces and Bacillus soil isolates from Presque Isle, PA. Data in the graph are representative of the inhibition of the human pathogenic bacterial strain tested by the soil isolate of > a 3 mm diameter on at least six separate occasions. “Narrow-spectrum” (five isolates) indicates that activity was directed against a single bacterial pathogen, and “broad-spectrum” represents inhibition of two or more test bacteria.

Figure 3. Soil nitrogen and phosphate levels of costal and inland sites at Presque Isle, PA. Data in the figure indicate the number of coastal and inland sites tested that showed either low (< 10 ppm) or adequate (20-40 ppm) levels of nitrogen (N) or phosphate (P) as assessed using Luster chemical kits. Combinational sites indicate those at which one nutrient was adequate and the other was low or depleted.

Nature of soil microflora
All sites sampled contained bacteria (54.9 ± 7.5 x 10³ CFU/ml) and yeasts (20.1± 4.9 x 10³ CFU/ml) with staphylococci present at 18 of the 20 sites (11.8 ± 6.1 x 10³ CFU/ml). At four sites (sites 3, 4, 5 and 7) Staphylococcus aureus was the predominant staphylococcus. Fungi were isolated from 13 of the 20 sites and were at comparable levels with the staphylococci in these areas (15.3 ± 4.3 x 10³ CFU/ml) (Figure 4). Coliforms were found at 10 sites; three of which were fecal coliforms present at elevated levels (sites 3, 4 and 5) (Figure 4).

Figure 4. Soil microflora of coastal and inland sites at Preque Isle, PA. Values shown represent the mean of three experiments, where 100 µL of a 10-2 dilution of soil in sterile saline was plated, and results are given as CFU/mL ± SEM.

Distribution of soil microflora
Recovery of fungal isolates was directly correlated with that of bacterial species per se (p < 0.02, r = 0.47) and staphylococci (p < 0.001, r = 0.91). In contrast, no correlation was observed between the number of yeast colonies or coliforms at the 20 sites or any other microbe investigated (p ≥ 0.3, r = 0.3). Fungi were absent from all four sites with low or depleted levels of nitrogen and phosphate and from seven of eight sites with low to depleted levels of nitrogen alone (adequate phosphate) (p < 0.001). Adequate to sufficient levels of nitrogen and phosphate (20-50 ppm) correlated with the presence of fungal species at 10 of the 11 sites (p < 0.001). No other correlations between nutrient availability and microbial species were observed.

Fungal numbers were the only biotic factor found to statistically distinguish sites with narrow-spectrum and broad-spectrum antibiotic-producing bacteria, being elevated at all sites from which broad-spectrum antibiotic-producing bacteria were recovered compared with sites from which they were absent (Figure 1). Most fungi generate a variety of factors with broad-spectrum antimicrobial-activity that limits bacterial numbers in a particular ecosystem (Salyers and Whitt 2001), so a similar distribution with broad-spectrum antibiotic-producing bacterial organisms is not entirely unexpected. Actinomycete, Bacillus, and fungal species are also able to survive in drier soils provided by a sandy ecosystem through the generation of spores, which allow for colonization of a greater range of habitats, especially sandy and/or dry locations (Mishustin 1975). In addition, fungi possess a variety of enzymes that are released from cells to externally break down otherwise inaccessible substrates such as chitin and lignin (Ljungdahl and Eriksson 1985), therefore providing a microenvironment that would be comparatively nutrient rich. A study by Griffin (1969) suggested that in drier, humus-poor soils, fungal mycelia extend further, pulling water back to the main body of the microbe and producing a microenvironment that allows more resilient species of bacteria to grow. In addition, the Presque Isle soil pH of 5.5-6.5 would be expected to favor colonization by cellulose-degrading fungi whose activities are most active at an acid-neutral pH (Atlas and Bartha 1998).

Narrow-spectrum antibiotic-producing bacteria were more likely to be isolated from sites where coliforms but not staphylococci were elevated (p = 0.06, Figure 1). Three of these locations (sites 3, 4 and 8) were only colonized by antibiotic-producing bacteria with narrow-spectrum activities and were also the only sites from which fecal coliforms were recovered. Shorebird migration is incremental from April onwards, peaking in August, with sites 3 and 8 constituting major bird-watching areas in the park (Discover Presque Isle Park Info.) Although coliforms are rapidly eliminated in marine habitats by predation and competition (Mitchell 1968), the associated biological wastes they inhabit may provide transient nutrient sources and alter the pH of the microenvironment to create a suitable habitat for Actinomyces species. Given the closeness to statistical significance of the data (p = 0.06), despite the fact that very few antibiotic-producing isolates per se were identified, it is anticipated that further sampling and analysis of coliform-rich coastal soils should allow us to identify the relationship between these two groups of bacteria.

The low levels of antibiotic-producing bacteria isolated in the current study may have several contributing factors. The sandiness of the soil allowed easy drying and several species could have been lost during transport. In addition, sampling of sandy-loam temperate soils during late spring (May) has been previously shown to recover a bacterial biomass only around half that of late summer (August) (Neher et al. 1999). Since antibiotic-producing organisms are often the least prevalent bacteria in these ecosystems (Mishustin 1975), this could also significantly affect our results. Actinomyces species are more prevalent in soils with a neutral to alkaline pH (Hattori and Hattori 1976), so the acidic environment (pH 5.5-6.5) of Presque Isle would restrict colonization to the more resilient members of this genus. This is also the case for Bacilli, whose endospores are generally well-distributed across a variety of soils with active bacteria, usually being recovered from soils with alkaline or extremely alkaline culture conditions (7.0-9.0) (Kroll 1990). Another issue is that some endospores present in the samples may also have been unable to be cultured under laboratory conditions using standard microbiological methods, similar to more than an estimated 99% of environmental organisms (Pace 1996).

Although the current study was unable to successfully separate the biotic or abiotic factors influencing narrow- or broad-spectrum antibiotic-producing bacteria to colonize a habitat, certain factors were identified that appeared to favor the presence of both these species. First, all were isolated from sites that were primarily coastal (6/7 locations) but with adequate levels of nitrogen (6/7 locations) and/ or phosphate (5/7). Secondly, strongly positive correlations were observed between antibiotic-producing bacterial isolates per se and both total fungal and staphylococcal counts (p < 0.001, r = 0.68 and 0.58, respectively).
Sessitsch et al. (2001) observed that amending sandy soils with organic components had no effect on the overall bacterial distribution, which was related instead to the overall size of soil particles that the microbes inhabited. These data seem to support our conclusions since coastal and inland sites showed little overall difference in either microbial numbers (Figure 4) or nutrient levels (Figure 3). Coastal and inland sites at Presque Isle do however differ physically from one another. Due to high levels of erosion, the lakeside of the Presque Isle peninsula is annually supplemented with tons of lake-dredged sand. These sandy soils are low in nutrient content and have sparse vegetation (Table 1) but are also fresher, more disturbed, and constantly changing ecosystems. Inland soils, though sandy, are older and more consistent, characterized by a surface-covering of moss, established grass, and upland hardwood tree ecosystems of oak or maple. These observations may therefore suggest that the antibiotic-producing bacteria are more likely to be recovered from disturbed habitats than established ecosystems.

Another approach to understanding the distribution of the Presque Isle antibiotic-producing organisms is to examine their spectrum of antibiotic activity. Eight of the 12 possessed activity against P. aeruginosa, and two produced narrow-spectrum antibiotic activity (Figure 2). The three remaining isolates had antimicrobial activities directed against B. megatherium (2/5 isolates) and S. aureus (1/5 isolates), with none being recovered that specifically impaired E. coli (Figure 2).

Previous studies have suggested that the distribution of bacterial species in sandy soils is limited with predominant species from the a-Proteobacteria group (Sessitsch 2001), which includes certain species of Pseudomonas (Krieg 1984). Pseudomonas distribution was not documented in this study, but these species are known to be present in most moist soil and water environments (Black 1996). The sandiness of the Presque Isle samples might not allow sufficient water retention inland, but, at coastal sites, which receive regular moisture, these species might be expected to compete with resident antibiotic-producing bacteria for nutrients. In contrast, the direct correlation of antibiotic-producing bacterial distribution with that of staphylococci was surprising, since only four of the 12 antibiotic-producing Actinomyces and Bacilli species impaired the growth of S. aureus in our study (Figure 2). These data suggest that, unlike Pseudomonas species, Actinomyces, Bacilli, and staphylococci share the ecosystem. No single isolate possessed narrow-spectrum activity against E. coli, although three were found that inhibited both the growth of our test human pathogenic strain, together with P. aeruginosa (Figure 2). Several factors might explain these data. Both E. coli and P. aeruginosa are gram-negative bacilli, which would make them equally susceptible to a cell-wall-directed antibiotic. Another factor that supports a “shared” target for E. coli comes from the finding that it is rapidly cleared from coastal habitats, often within 24 hours (Mitchell 1968). As it would not play a major role in these habitats, being a transient occupant, indigenous microbes would not target this bacterium as a potentially competitive species.

Our study has therefore demonstrated, for the first time, that both narrow- and broad-spectrum antibiotic-producing bacterial species may be recovered from the temperate, sandy Presque Isle ecosystem. We have also shown that the biotic factors involved in the generation of a hospitable ecosystem for Actinomyces and Bacillus species at this site are the presence of fungal and staphylococcal species together with adequate soil nitrogen levels, but were unable to effectively identify a set of environmental factors selecting for narrow-spectrum antibiotic activity. We are planning to re-visit the Presque Isle site again this summer, to extend the data from the current study by increasing the number of sites sampled, determining additional constituents of the microbial flora (including Pseudomonas species), and measuring soil levels of the microbial micronutrients iron, zinc, and calcium. To determine whether narrow-spectrum antibiotic production is favored by a more subtle interplay between several variables, we also plan to employ artificial intelligence programs to create a “model” of the Presque Isle environment. Although the antibiotic moieties described in this study have not yet been characterized, all single specificity Actinomyces isolates are now being further evaluated for their spectra of activity, using species identification and chemical purification of activity.