Serial analysis of rRNA genes and the unexpected dominance of rare members of microbial communities

Abstract

The accurate description of a microbial community is an important first step in understanding the roles of its components in ecosystem function. A method for surveying microbial communities termed serial analysis of rRNA genes (SARD) is described here. Through a series of molecular cloning steps, short DNA sequence tags are recovered from the fifth variable (V5) region of the prokaryotic 16S rRNA genes from microbial communities. These tags are ligated to form concatemers comprised of 20 to 40 tags which are cloned and identified by DNA sequencing. Four agricultural soil samples were profiled with SARD to assess the method's utility. A total of 37,008 SARD tags comprising 3,127 unique sequences were identified. A comparison of duplicate profiles from one soil genomic DNA preparation revealed that the method was highly reproducible. The large numbers of singleton tags, together with nonparametric richness estimates, indicated that a significant amount of sequence tag diversity remained undetected with this level of sampling. The abundance classes of the observed tags were scale-free and conformed to a power law distribution. Numerically, the majority of the total tags observed belonged to abundance classes that were each present at less than 1% of the community. Over 99% of the unique tags individually made up less than 1% of the community. Therefore, from either a numerical or diversity standpoint, taxa with low abundance comprised a significant proportion of the microbial communities examined and could potentially make a large contribution to ecosystem function. SARD may provide a means to explore the ecological roles of these rare members of microbial communities in qualitative and quantitative terms.

FIG. 1.

Serial analysis of rRNA genes (SARD). A conserved AluI restriction enzyme recognition site is located immediately downstream of the fifth variable (V5) region of the bacterial 16S rRNA gene. Step I, “universal” bacterial PCR primers are employed to create amplicons flanking this region from environmental genomic DNA. The amplicons, containing 5′ biotin groups, are digested with AluI. Step II, the 5′-most AluI restriction fragments are immobilized on magnetic streptavidin-coated (SA) beads. Step III, the beads are split into two pools and a unique double-stranded adapter is ligated to each pool. Step IV, the adapters, including short sequence tags, are released from the beads by digestion with BpmI. Step V, the 3′ overhangs of the released fragments are removed with the Klenow fragment and the products are recombined. Step VI, DNA fragments are ligated head-to-head, PCR amplified with primers specific to the unique adapters, and cleaved with the restriction enzyme FokI. Step VII, the resulting 28-bp ditags, possessing 4-bp AGCT 5′ overhangs, are purified and ligated to form concatemers. Step VIII, concatemers comprising 30 or more SARD tags are purified and cloned into the HindIII site of pUC19.

FIG. 2.

Reproducibility of SARD profiles. Data from two separate SARD profiles of the same sample are plotted against one another. Values are shown only for tags that were observed in both profiles. The symbol sizes reflect the number of coincident tags that overlapped on the plot as a result of occurring at the same abundance levels. The largest symbol corresponds to 45 tags that were seen once in both profiles.

FIG. 3.

Rank abundance plot of SARD tag sequences from four soil samples. Boxed values indicate Simpson's reciprocal diversity index (1/D) values.

FIG. 4.

Abundance class distributions of SARD data. SARD tag abundance data plotted as histograms with logarithmic binning.

FIG. 5.

Observed and estimated SARD tag richness from four agricultural soil samples. The bars show the 95% confidence intervals provided for the Chao1 richness estimates. The plots were made with EstimateS (9) following 50 randomizations. For clarity, only every 300th data point is plotted.

FIG. 6.

Unrooted phylogenetic tree of 16S rRNA gene sequences from sample WP45. Bootstrap values of 50 percent or greater are indicated. Identification numbers for SARD tags found in the 16S rRNA sequences are indicated to the right. The names of 16S rRNA sequences that possess the same SARD tag sequence are shaded. The corresponding SARD tag identification numbers are grouped by vertical bars. GenBank accession numbers for reference sequences utilized to deduce bacterial division affiliations are given in italics. SARD tags that occur in sequences of different division affiliations are indicated by an asterisk. NIT, not informative tag. The scale bars represent distances of 10 percent.

FIG. 6.

Unrooted phylogenetic tree of 16S rRNA gene sequences from sample WP45. Bootstrap values of 50 percent or greater are indicated. Identification numbers for SARD tags found in the 16S rRNA sequences are indicated to the right. The names of 16S rRNA sequences that possess the same SARD tag sequence are shaded. The corresponding SARD tag identification numbers are grouped by vertical bars. GenBank accession numbers for reference sequences utilized to deduce bacterial division affiliations are given in italics. SARD tags that occur in sequences of different division affiliations are indicated by an asterisk. NIT, not informative tag. The scale bars represent distances of 10 percent.

FIG. 7.

Expected phylogenetic coverage of a SARD profile. Approximately 5,100 16S rRNA gene sequences (21) were examined to determine the predicted locations of SARD tags for each sequence. The fraction of sequences for each bacterial division where a tag would be expected to be recovered from the V5 region is indicated. A SARD tag will not be recovered from 16S rRNA genes where an AluI restriction site is not present in or near the V5 region.