Annotation of the modular polyketide synthase and nonribosomal peptide synthetase gene clusters in the genome of Streptomyces tsukubaensis NRRL18488

Abstract

The high G+C content and large genome size make the sequencing and assembly of Streptomyces genomes more difficult than for other bacteria. Many pharmaceutically important natural products are synthesized by modular polyketide synthases (PKSs) and nonribosomal peptide synthetases (NRPSs). The analysis of such gene clusters is difficult if the genome sequence is not of the highest quality, because clusters can be distributed over several contigs, and sequencing errors can introduce apparent frameshifts into the large PKS and NRPS proteins. An additional problem is that the modular nature of the clusters results in the presence of imperfect repeats, which may cause assembly errors. The genome sequence of Streptomyces tsukubaensis NRRL18488 was scanned for potential PKS and NRPS modular clusters. A phylogenetic approach was used to identify multiple contigs belonging to the same cluster. Four PKS clusters and six NRPS clusters were identified. Contigs containing cluster sequences were analyzed in detail by using the ClustScan program, which suggested the order and orientation of the contigs. The sequencing of the appropriate PCR products confirmed the ordering and allowed the correction of apparent frameshifts resulting from sequencing errors. The product chemistry of such correctly assembled clusters could also be predicted. The analysis of one PKS cluster showed that it should produce a bafilomycin-like compound, and reverse transcription (RT)-PCR was used to show that the cluster was transcribed.

Fig 1

Apparent inversion in the FK506 cluster. (A) Module organization as deduced by the ClustScan program using module numbering compatible with the annotation of the FK520 cluster. (B) Module organization after correction of the apparent inversion introduced by an assembly error. (C) Chemical structure of FK506.

Fig 2

Screen shots of the three MIRA 3 contigs c43, c213, and c288, which contain cluster St-PKS4. The modules in the contigs were identified as belonging to the same cluster by phylogenetic analysis. The positions of genes and domains are shown on the three reading frames of the forward strand (top part of the windows) and the three reading frames of the complementary strand (bottom part of the windows). Thus, in this case, most of the domains are on the complementary strand, distributed over all three reading frames. The PKS domains are shown as blue boxes. By moving the cursor over the boxes, it is possible to see what sort of domain has been detected and whether it is complete. As the order of the domains in modules is highly conserved, it is easy to see whether the sequence in another contig is a candidate for the continuation of a cluster; gaps between contigs are usually small compared to the module size. The postulated order and orientation of the three contigs are indicated. The four numbered boxes contain the apparent frameshifts that cause the reading frames to change within modules. For clarity, the left end of c43 and the right end of c288 are omitted. The green and red boxes show predicted proteins and NRPS domains, respectively.

Fig 3

Predicted chemical structures of linear products of modular clusters. (A) Cluster St-PKS4. (B) Cluster St-PKS2. This structure should cyclize to form a bafilomycin. The generic side chains represented by A indicate parts of the molecule for which ClustScan cannot make a specificity predication. (C) Cluster St-NRPS1. ??? corresponds to a module for which no substrate prediction was possible.

Fig 4

Transcription of cluster St-PKS2. Lanes 1, 3, and 5, RT-PCR with primers for the pks2, pks4, and hrdB genes; tracks 2, 4, and 6, controls without reverse transcriptase; track 7, GeneRule 100- to 10,000-bp DNA ladder (Fermentas).