Discovery of Novel Biosynthetic Gene Cluster Diversity From a Soil Metagenomic Library

Abstract

Soil microorganisms historically have been a rich resource for natural product discovery, yet the majority of these microbes remain uncultivated and their biosynthetic capacity is left underexplored. To identify the biosynthetic potential of soil microorganisms using a culture-independent approach, we constructed a large-insert metagenomic library in Escherichia coli from a topsoil sampled from the Cullars Rotation (Auburn, AL, United States), a long-term crop rotation experiment. Library clones were screened for biosynthetic gene clusters (BGCs) using either PCR or a NGS (next generation sequencing) multiplexed pooling strategy, coupled with bioinformatic analysis to identify contigs associated with each metagenomic clone. A total of 1,015 BGCs were detected from 19,200 clones, identifying 223 clones (1.2%) that carry a polyketide synthase (PKS) and/or a non-ribosomal peptide synthetase (NRPS) cluster, a dramatically improved hit rate compared to PCR screening that targeted type I polyketide ketosynthase (KS) domains. The NRPS and PKS clusters identified by NGS were distinct from known BGCs in the MIBiG database or those PKS clusters identified by PCR. Likewise, 16S rRNA gene sequences obtained by NGS of the library included many representatives that were not recovered by PCR, in concordance with the same bias observed in KS amplicon screening. This study provides novel resources for natural product discovery and circumvents amplification bias to allow annotation of a soil metagenomic library for a more complete picture of its functional and phylogenetic diversity.

Keywords: biases; biosynthetic ability; metagenome; next-generating sequencing; soil.

FIGURE 1

Library construction and 3D pooling strategies. (A) Library construction strategy. Superscript letters correspond to citations: a: (Liles et al., 2008); b: (Kakirde et al., 2011); c: (Nasrin et al., 2018). (B) 3D Library Pooling strategy: Plate, Column, and Row are barcoded before sequencing. (C) Contig to clone deconvolution and in silico mining strategy. Reads from each pool are assembled separately into pool contigs. Highly homologous contigs from each dimension (Plate, Column, and Row) form triplets. Since their pool of origin is known, they form a coordinate system that points to their clone of origin. Contigs were analyzed bioinformatically for rational identification of candidates for functional screening.

FIGURE 2

Maximum likelihood tree of KS domains recovered from the soil metagenomic library. Clades in magenta are formed uniquely by domains recovered by NGS. (A) KS domains recovered either by PCR or by NGS. (B) KS domains recovered from the soil metagenomic library and from the MIBiG database showing branch lengths. (C) KS domains recovered from the soil metagenomic library and from the MIBiG database showing topology only to facilitate visualization. Circles in the tree represent branch bootstrap support >70%.

FIGURE 3

Maximum likelihood tree of A-domains recovered from NRPS clusters identified from the soil metagenomic library and the MIBiG database. (A) ML Tree showing branch lengths. (B) ML Tree ignoring branch lengths to facilitate topography visualization. Circles in the tree represent branch bootstrap support >70%.

FIGURE 4

Clustering of PKS (A) and NRPS (B) pathways against the MIBiG database using BiG-SCAPE. Library pathways are shown in orange and database pathways in blue. Circle size is proportional to the number of connections, and strike width is proportional to the similarity between pathways.

FIGURE 5

Analysis of the clones containing viable PKS and NRPS pathways reconstructed and validated from the Cullars metagenomic library. (A) Viable PKS and/or NRPS clusters recovered from the library compared by size (number of domains) and divergence from known BGCs (mean branch length and domain mean divergence). Each plot point represents a clone containing a viable BGC. Mean branch length for each clone is calculated as the average branch length of all the clones’ KS and/or A domains, as extracted from the ML trees against the MIBiG database. Mean domain divergence for each clone is calculated as the average domain divergence of all clones’ KS and/or A domains. Domain divergence is the complementary percentage to the% identity of the domain to the NCBI nr database (if a domain has% identity of 60, then the domain divergence is 40%). (B) Density distribution of branch length, domain divergence and number of domains across the viable PKS and NRPS BGCs. (C) Annotation of three interesting representative BGCs, representing the longest BGC from each type (PKS, NRPS, PKS-NRPS) to be recovered successfully in silico from the metagenomic library NGS contigs and validated upon clone resequencing. Annotation from the library contig was compared to the annotation from their corresponding validated insert sequence. ORFs are depicted by arrows and the PKS and/or NRPS domains within each ORF are represented by colored stripes.

FIGURE 6

Taxonomic diversity of the Cullars metagenomic library and of the original soil sample used for library construction. Soil PCR: 16S rRNA gene PCR amplification of DNA isolated from the original Cullars soil sample used to construct the library; Library PCR: 16S rRNA gene PCR amplification of the DNA template from all metagenomic library clones pooled together; Library NGS: In sillico mining for 16S rRNA genes from the metagenomic library NGS contigs. The phyla contained in the section “Other phyla” are: RCP2-54, SAR324, Spirochaetes, Sumerlaeota, WPS-2; all with relative frequencies below 0.05%.