Natural product biosynthetic gene diversity in geographically distinct soil microbiomes

Abstract

The number of bacterial species estimated to exist on Earth has increased dramatically in recent years. This newly recognized species diversity has raised the possibility that bacterial natural product biosynthetic diversity has also been significantly underestimated by previous culture-based studies. Here, we compare 454-pyrosequenced nonribosomal peptide adenylation domain, type I polyketide ketosynthase domain, and type II polyketide ketosynthase alpha gene fragments amplified from cosmid libraries constructed using DNA isolated from three different arid soils. While 16S rRNA gene sequence analysis indicates these cloned metagenomes contain DNA from similar distributions of major bacterial phyla, we found that they contain almost completely distinct collections of secondary metabolite biosynthetic gene sequences. When grouped at 85% identity, only 1.5% of the adenylation domain, 1.2% of the ketosynthase, and 9.3% of the ketosynthase alpha sequence clusters contained sequences from all three metagenomes. Although there is unlikely to be a simple correlation between biosynthetic gene sequence diversity and the diversity of metabolites encoded by the gene clusters in which these genes reside, our analysis further suggests that sequences in one soil metagenome are so distantly related to sequences in another metagenome that they are, in many cases, likely to arise from functionally distinct gene clusters. The marked differences observed among collections of biosynthetic genes found in even ecologically similar environments suggest that prokaryotic natural product biosynthesis diversity is, like bacterial species diversity, potentially much larger than appreciated from culture-based studies.

Fig 1

(A) Overview of the approach used to compare secondary metabolism in different soil microbiomes. Independent environmental DNA libraries were constructed from three different arid soils. DNA from each library was used as the template in PCRs with degenerate primers designed to recognize nonribosomal peptide synthetase AD domains, type I polyketide KS domains, and type II polyketide KSα genes. The resulting amplicons were 454 sequenced, processed, and compared to assess the similarity of the three gene sets derived from different microbiomes. KS, ketosynthase; ACP, acyl carrier protein; AT, acyltransferase; C, condensation domain; A or AD, adenylation domain; PCP, peptide carrier protein. (B) Bar graphs show the frequency at which 16S rRNA genes from different major phyla appeared in each library. As might be expected for metagenomic libraries constructed from ecologically similar soils, 16S rRNA gene analyses indicate that DNA from a very similar distribution of major bacterial phyla was captured in each library.

Fig 2

Sequence richness and diversity estimates. (A) The number of raw reads (reads), unique cleaned reads (unique), and OTUs when grouped at 97% identity are shown for AD, KS, and KSα sequences amplified from each cloned metagenome (AB, AZ, and UT). Chao1 sequence richness estimates are reported at a cutoff of 3%. Shan, Shannon diversity index. Sequences are deposited in the NCBI-SRA database under accession number SRA045798.2. (B) Rarefaction curves (using a 3% cutoff value) for AD, KS, and KSα sequences amplified from each eDNA library.

Fig 3

Comparison of secondary metabolite gene sequences found in three cloned metagenomes. (A) Sequences from all three metagenomes were clustered at various identities, and Venn diagrams were then made to show the percentage of clades containing sequences from each cloned metagenome. Venn diagrams are drawn to scale whenever possible. (B) Even when clustered at 85% identity, a large number of AD and KS clades are sparsely populated. Venn diagrams representing the clustering analysis of only the top 500 most populated AD and KS are shown. Each of these clades contains >20 unique pyrosequencing reads. (C) KSα phylogenetic tree. Functionally characterized KSα sequences (pink) and representative sequences from each OTU that formed when eDNA sequences clustered at 85% identity were aligned using ClustalW and plotted with iTOL. Branches are color coded according to the soil sample that had the most unique sequences in that branch. The KSα gene from resistomycin biosynthesis was used to root the tree. A key advantage of using large libraries rather than crude eDNA to compare microbiomes is that gene clusters associated with novel biosynthetic genes can be recovered and functionally studied. Marked in purple are KSα sequences associated with eDNA-derived gene clusters that have yielded novel secondary metabolites.

Fig 4

Relationship between KSα sequence identity and gene cluster function. (A) The number of distinct groups formed when functionally characterized KSα sequences are clustered at different percent identities is shown. On average, functionally characterized KSα gene sequences (50 in total) that are known to be involved in the biosynthesis of structurally distinct metabolites do not group together when clustered at above 80 to 85% identity. When clustered below 85% identity, this correlation between KSα sequence divergence and small molecule structural divergence is no longer observed. Clustering was carried out using 300-bp KSα gene fragments corresponding to the amplicons that would be produced by the KSα degenerate primers used to access metagenomic sequences. (B) Metagenomic sequences that clustered with functionally characterized KSα genes at 85% identity are shown. A representative eDNA clone containing a KSα gene from each clade was recovered and sequenced. In each case, these clones closely resembled in gene identity, gene complement, and gene organization in the functional characterized gene cluster. The percent identity between KSα genes is shown. Genes are color coded according to the predicted enzymatic function of their products. Red, minimal PKS; blue, regulation and resistance; orange, polyketide biosynthesis; pink, starter biosynthesis; green, sugar biosynthesis; gray, unknown pathway/unrelated enzymes.

Fig 5

Comparison of metagenome-derived KS and AD domain sequences with those found in functionally characterized gene clusters. Each metagenomic data set was searched by tBLASTx for KS and AD amplicons that show high sequence identity to similar domains found in functionally characterized PKSI and NRPS gene clusters. Individual spokes of the graph correspond to the collection of identity scores for library-derived amplicons that show greater than 75% identity to any AD or KS domain found in the designated biosynthetic gene cluster. All of the hits shown here have E values of ≤10e⁻⁴⁰, and each alignment spans at least 80 amino acid residues. The final image was constructed using R and ggplot2 (36).