The site, an NSF-funded Long-Term Ecological Research site in the Bonanza Creek Experimetal Forest in central Alaska, is ideal for this project because it offers a pristine environment from which to sample microbial communities. Our research strives to describe and understand the microbial communities in the boreal forest ecosystem.

Soil naturally contains a rich variety of microorganisms, and the majority of these cannot be grown in culture.  Indeed, the unculturable microorganisms in soil outnumber the culturable organisms by a factor of 100 or 1000.  With this in mind, our research uses both culturing and culture-independent molecular methods to describe the microbial diversity of the soil.

First, we amplified and sequenced 2,000 16S rRNA genes from the Alaskan soil. We then focused on statistical methods to assess the completeness of our census and to determine whether libraries constructed from different samples differ significantly in composition.  We used the existing computer program, LIBSHUFF, developed by Singleton et al., to build a faster and more precise method to compare 16S rRNA gene libraries.  Next, we developed a program, designated DOTUR, which develops and analyzes collector's curves to assess the completeness of a sample (as represented in a 16S rRNA gene library) and produces estimates of species richness.  We used DOTUR to assess the completeness of the entire Ribosomal Database Project II by analyzing 56,000 publicly available sequences, and then applied the program to our datasets from Alaskan soil.  To estimate the sampling needed to complete a census of the soil microbial community we used closed datasets – the words in books – as an analogy with species richness (the total number of words is similar to the total number of individuals in a community; the number of different words is similar to the number of species; and the total number of letters is similar to the number of phyla).  Three books of varying complexity were used:  Darwin's On the Origin of Species (7,426 different words; 150,951 total words); James' A Portrait of a Lady (12,427 different words; 230,485 total words); and Brown's Goodnight Moon (55 different words; 131 total words).  The datasets of words provided the basis for building heuristic models to describe the frequency distribution of items in a collection.  We modified the models to obtain a fit with the 16S rRNA genes from soil and used this model to predict species richness and the number of sequences that would be needed to query every species with a 95% confidence level.  The model predicts that there are 5,000 species in a gram of Alaskan soil and 300,000 to 400,000 sequences would be needed for a complete census. 

To describe the culturable portion of the soil community, we isolated ~2000 microorganisms from soil by culturing under various regimes of rich and starvation media, cold temperatures, and long incubation times designed to coax cold-dwelling organisms into laboratory culture.  Among the cultured community, we found a high frequency of bacteria that can grow on highly reduced forms of phosphorus (Metcalf, Handelsman, et al., in preparation) and these isolates are being characterized. 

We constructed metagenomic libraries in E. coli with DNA from the Alaskan soil microbial community . We identified four clones that enable E. coli to grow on extremely low phosphorus and appear to solubilize apatite, which we have shown is the dominant form of inorganic phosphorus in the Alaskan soil.  We also developed a novel, high throughput screening method to identify clones that produce small metabolites.