Intraspecies comparison of Streptomyces pratensis genomes reveals high levels of recombination and gene conservation between strains of disparate geographic origin

Streptomyces are ubiquitous bacteria with many uncommon features and important industrial uses. They produce over half of the clinically useful antibiotics and a host of other bioactive, pharmaceutically relevant compounds [1]. The name Streptomyces means twisted fungus, reflecting the morphological and life cycle traits that these bacteria share with fungi. While Streptomyces taxonomy is notably problematic, multi-locus sequence analysis (MLSA) approaches are helping to resolve species boundaries in the genus [2–4]. For example, measurements of MLSA divergence in relation to DNA-DNA hybridization (DDH) values indicate that 0.7-0.8% divergence of MLSA loci roughly delineates species boundaries in Streptomyces[3, 5, 6]. Though asexual, Streptomyces are capable of genetic exchange within and between species [7], and there is evidence of widespread horizontal gene transfer within and between species of the genus [8]. The implications of horizontal gene transfer for genome evolution within Streptomyces remains poorly described.

Streptomyces genomes deviate from those of other bacteria in several ways. There can be multiple genomes per cellular compartment, though asexual spores have a single genome copy. Hyphae elongate at the tip and form septa that define cellular compartments at regular intervals [9, 10]. Roughly ten to twelve genomes can coexist inside of a single compartment [7, 11]. Streptomyces genomes are linear and replicate from a bidirectional central ori, although they can exist as unstable, circular molecules [12–14]. The ends of their chromosomes consist of terminal inverted repeats (TIRs), and the length of these TIR regions can vary largely, from 167 bp to 1 Mb [15, 16]. Their plasmids can be circular or linear, and can mobilize chromosomal markers at high frequency during interspecies transfer [17]. The chromosome has been classified into two sections: the central, conserved core region and the more variable terminal chromosome arms [18].

One surprise revealed by genome sequencing of Streptomyces species is the presence of numerous cryptic secondary metabolite gene clusters [19, 20]. These cryptic gene clusters encode products that are either silent or not identified as natural products during growth in the laboratory. For example, four decades of genetic analysis had identified four secondary metabolite gene clusters in the model organism Streptomyces coelicolor, but the first genome sequence revealed a total of 22 secondary metabolite gene clusters [19]. The diversity of natural product gene clusters is very high in Streptomyces compared to other bacterial genera. There is very little overlap in terms of shared natural product biosynthetic gene clusters between the currently closed Streptomyces genomes, all of which are from different species [21]. The intraspecies variability of natural product biosynthetic genes has not yet been determined through comparative genomic analysis of Streptomyces species. However, recent studies of Salinispora, another actinomycete genus rich in natural product biosynthetic genes, show high overall conservation within species [22].

The species S. pratensis has been described recently [23] to include isolates from a wide region of North America spanning sites found in North Carolina, New York, Michigan, and Quebec [8]. Nucleotide divergence of MLSA loci from strains of S. pratensis did not exceed 0.4%, justifying their inclusion in a single species [8]. Very high levels of homologous recombination were detected in S. pratensis, sufficient to promote linkage equilibrium for alleles at MLSA loci [8]. Interspecies recombination is widespread among Streptomyces, although interspecies gene exchange occurs at a much lower rate than intraspecies gene exchange [8]. The genetic coherence of Streptomyces species is surprising given the potential for widespread gene exchange, the high level of nucleotide similarity between many different species, and the null expectations for highly recombining populations [24, 25]. The maintenance of coherent genetic clusters that correspond to Streptomyces species suggests some mechanism for constraining interspecies gene exchange.

We have sequenced the genomes of four strains of the newly described species S. pratensis, including the type strain Ch24^T (=NRRL B-24916^T). The four strains were isolated from edaphically similar sites separated by 740 km. These draft genome sequences were compared with the complete genome of S. flavogriseus IAF-45-CD (=ATCC 33331), which has been reassigned recently to S. pratensis[23].

The size of the core genome can vary widely between bacterial species. For example, core genes can represent anywhere from 20% to 93% of the total genome across diverse bacterial lineages such as: Escherichia coli, ~20% [48]; Streptococcus pneumoniae, 52% [49]; Salmonella enterica, 61% [50]; Actinobacillus pleuropneumoniae, 79% [51]; Listeria monocytogenes, ~80% [52]; Campylobacter coli, 82.4% [53]; Campylobacter jejuni, 83.5% [53]; and Chlamydia trachomatis, 93% [54]. We have used a very strict definition of a core genome in our analysis of S. pratensis, using only positionally orthologous genes, as opposed to gene families, unlike many of the studies listed. Even with this strict definition, we calculate that 85-88% of each genome is comprised of core genes. Because our metric is more stringent than bi-directional best BLAST hits, which is used in many publications, these estimates should be viewed as a conservative estimate in comparison. Since the vast majority of variable genes are unique to a single genome (Figures 1A) and the estimate of core genome size changed little between the second and fifth genomes added to the analysis (Figure 1B) it is unlikely that the addition of new genomes will reduce substantially the size of the core genome in S. pratensis. Among published bacterial genome comparisons, only Chlamydia trachomatis, an obligate intracellular pathogen that has experienced extreme genome reduction and is left with only ~900 genes in the average genome [54], has a larger core genome proportion than S. pratensis. In addition, our assessment of the core genome does not exclude genes of putative plasmid origin, and as such represents a lower bound of core genome content for the linear chromosome. These observations suggest that the genome of S. pratensis is overwhelmingly dominated by core genes.

Comparative genomics of different species of Streptomyces has indicated that the central portion of the chromosome is highly conserved both in gene content and synteny while chromosome termini are highly variable between species. Our intraspecies comparison reveals islands of diversity spread throughout the chromosome, including at the terminal variable regions but also throughout the central core. The terminal variable regions are only 112 kbp (the 5’ arm as annotated in ATCC 33331) and 36 kbp (Figure 1D). This is significantly smaller than the 753 to 1,393 kbp terminal arms found in interspecific comparisons [55]. Strain specific islands are composed primarily of mobile genetic elements and genes that are likely to have been acquired by horizontal gene transfer. Strain specific islands composed of mobile genetic elements are a common feature of many microbial genomes.

All five of these genomes have the same repertoire of natural product gene clusters which are unambiguously part of the core genome of S. pratensis. This means that efforts to mine genomes for novel gene clusters will be facilitated by accurate species classifications, which can eliminate the need to needlessly screen many strains of the same species. However, our results for the bldB and hpb gene families suggest that changes to regulatory genes can occur at a fast rate within a species and may affect the expression of gene clusters between strains of a species. Examining diverse isolates from the same species may allow researchers to find regulatory changes that activate gene clusters that are otherwise cryptic in other strains. We have also shown that lanthipeptide precursors and NRPS and PKS genes can be highly variable within a species.

S. pratensis possesses a new combination of putative beta-lactam (MM 4550-like) and putative beta-lactamase inhibitor (clavulanic acid-like) biosynthetic gene clusters. This observation, on one level, suggests that observing patterns of natural product gene-clusters within the genomes of actinomycetes may reveal new possible drug combinations that have been proven effective by the crucible of evolution, and such new combinations may prove useful therapeutically. This observation also provides evidence for the in situ use of these products as antibiotics rather than as signaling molecules [56]. While the production of these molecules in S. pratensis has not yet been verified experimentally, the observation that convergent evolution has produced divergent combinations of beta-lactam and beta-lactamase inhibitor gene clusters independently in both S. pratensis and S. clavuligerus (species that do not share a common ancestor within Streptomyces) is evidence for the presence of an evolutionary arms race in soil communities. That is, selection has on at least two occasions driven the independent assembly of systems designed to produce both beta-lactam antibiotics and overcome beta-lactam resistance. The selection pressure for this pattern of gene cluster co-occurrence can only be explained by the hypothesis that: i) there is a high frequency of beta-lactam resistance in soils, and ii) these streptomyces benefit from the use of beta-lactam antibiotics to inhibit or kill other microorganisms in the soil community.

We have found that there is no consensus phylogenetic signal among S. pratensis genomes; the genome represents a mosaic of recombination between strains of the species. In this way S. pratensis resembles the East Asian H. pylori population in that recombination scrambles patterns of polymorphism between strains. A difference between these two species is that phylogenetic signal decays over shorter distances in H. pylori than in S. pratensis (Figure 7). This result could be caused by differences both in recombination rate and in the tract length of recombination. For example, H. pylori is naturally competent and can incorporate short stretches of DNA into its genome through transformation [45]. In contrast, acquisition of DNA by Streptomyces proceeds through an unusual dsDNA dependent mechanism of conjugation that may result in transfer of the whole chromosomes and backcrossing with the parent [7].

The frequency of recombination in S. pratensis was such that there is no detectable association between geographic distance and genomic divergence between strains from NY and NC. The geographic range of S. pratensis is unknown and it is not possible to estimate rates of migration from our current sample size, or to estimate whether dispersal limitation impacts the biogeographic pattern of genetic diversity within the species at very large spatial scales (e.g. continental). What we have shown, however, is that the extent of recombination between stains isolated from soils 740 km apart is not discernably different from the recombination observed between strains that co-occur in the same soil sample. Such a pattern could result from contemporaneous gene exchange at regional spatial scales (i.e. recombination between sites exceeds the ability of mutation to promote divergence between sites). Alternatively, this pattern could also result from the evolutionary recent regional expansion of a population that was recently in equilibrium.

While the majority of positionally orthologous core genes were highly similar between genomes, we observed 59 positional orthologs that differed by more than 5% between the S. pratensis genomes. Of these, 19 are members of two gene families known to be involved in regulation of differentiation and development. These two co-occurring gene families are BldB and what we have termed Hpb (for helix-turn-helix partner of BldB). BldB is a transcriptional regulator previously shown to be required for development of aerial mycelia, e.g. bldB mutants have bald colonies [57]. The conserved domain found within BldB is in the conserved domain database as DUF397 [34]. OrfD (SCO0703) in the antibiotic regulatory locus abaA and the developmental regulator WhiJ also fall within this class [58]. BldB has been characterized as a DNA-binding repressor that down-regulates its own expression [59], and was also previously suggested to have a binding partner that modulates BldB activity [60]. Hpb is predicted to have a helix-turn-helix domain of the Xre class. Based on the presence of a Xre domain, a history of gene duplication and recent proliferation, and limitation to a subset of the Actinobacteria, this gene pair was predicted by Makarova et al. [61] to represent a novel toxin-antitoxin system. The correspondence in diversification of the bldb/hpb family gene pairs suggests that a mutation in one gene encourages a compensatory change in its partner at the same locus. These observations from S. pratensis provide further indirect evidence that BldB-Hpb functions in a manner that resembles a toxin-antitoxin system.

We have uncovered patterns of genome evolution within a Streptomyces species through comparison of S. pratensis genomes isolated from disparate geographic origins. Core genes make up a high percentage of the genome, and natural product genes are unambiguously included within the set of core genes. High rates of intraspecies recombination homogenize polymorphisms in core genes across the genome and abolish any phylogenetic signal present within species. Two genomic islands exhibited a breakdown in intraspecies recombination promoting localized divergence between sets of genomes. One of these islands was centered on a hybrid NRPS-PKS gene, and the other was centered on a regulatory gene associated with a type III polyketide synthase biosynthetic gene cluster. In addition, highly divergent core genes included a lanthipeptide biosynthetic gene cluster as well as co-occurring members of the bldB and hpb gene families, genes which have been associated with the regulation of development and antibiotic production in Streptomyces. The conservation of biosynthetic gene clusters between strains of the species suggests that antibiotic production capacity is conserved within a species while variation in bldB and hpb gene families suggest that changes to regulatory genes can occur at a fast rate within a species and may affect the expression of biosynthetic gene clusters between strains of a species.