High quality genome annotation and expression visualisation of a mupirocin-producing bacterium

Abstract

Pseudomonas strain NCIMB10586, in the P. fluorescens subgroup, produces the polyketide antibiotic mupirocin, and has potential as a host for industrial production of a range of valuable products. To underpin further studies on its genetics and physiology, we have used a combination of standard and atypical approaches to achieve a quality of the genome sequence and annotation, above current standards for automated pathways. Assembly of Illumina reads to a PacBio genome sequence created a retrospectively hybrid assembly, identifying and fixing 415 sequencing errors which would otherwise affect almost 5% of annotated coding regions. Our annotation pipeline combined automation based on related well-annotated genomes and stringent, partially manual, tests for functional features. The strain was close to P. synxantha and P. libaniensis and was found to be highly similar to a strain being developed as a weed-pest control agent in Canada. Since mupirocin is a secondary metabolite whose production is switched on late in exponential phase, we carried out RNAseq analysis over an 18 h growth period and have developed a method to normalise RNAseq samples as a group, rather than pair-wise. To review such data we have developed an easily interpreted way to present the expression profiles across a region, or the whole genome at a glance. At the 2-hour granularity of our time-course, the mupirocin cluster increases in expression as an essentially uniform bloc, although the mupirocin resistance gene stands out as being expressed at all the time points.

Fig 1. Summary of the pipeline used to achieve a high quality genome annotation.

Fig 2. Summary of non-conserved repeats in mmpD showing the sequence information available over the mupirocin cluster for several Pseudomonas strains.

The drawing is not to scale. A high-identity repeat is present covering methyltransferase domains (indicated in pink at top, labelled ‘MT’) of mmpD in NCIMB10586, 2P24 and 8B, and inferred to be present in BRG-100 and NS383, while sequence conservation is lower between the two regions in QS1027. The repeated sequence in each strain is shown as a box with coloured chevron; inferred repeats as a chevron alone. Neither the sequence, nor exact end-point location of these repeats is conserved between strains. Two apparent frameshifts in mmpD are indicated (‘!’) in 2P24, however the overall genome coverage is low and these may be artefacts.

Fig 3. Plot of tile expression intensity vs a measure of variation ([root-mean-square deviation]/mean expression) categorised as determined using the final normalisation factors.

For display in this graph only, expression intensities of zero have been substituted by one to permit calculation of the geometric mean.

Fig 4. Visualisation of expression over the NCIMB10586 genome.

A) Key. The colour of a feature depends on time-point of highest expression (hue), expression intensity (lightness; x axis) and ratio of maximum to minimum expression (saturation; y axis). So: white = consistent low expression; grey = high expression, little variation; intense colour = high expression and big variation between highest and lowest level. B) Features are shown above or below the DNA line dependent on the coding strand. Protein-coding regions are shown as coloured boxes slightly separated from the DNA line. RNA features are shown as coloured boxes adjacent to the DNA line. Putative clusters identified by antiSMASH are shown: grey bar–neighbourhood; black bar(s)—protocluster regions.

Fig 5. Enlarged view of expression over each cluster in the NCIMB10586 genome identified by antiSMASH.

A, B, C, E, G - NRPS; D - Betalactone; F - NAGGN; H - siderophore; I, L - bacteriocin; J – NRPS-like; K - Arylpolyene; M – mupirocin cluster.