Genome update of the dimorphic human pathogenic fungi causing paracoccidioidomycosis

Abstract

Paracoccidiodomycosis (PCM) is a clinically important fungal disease that can acquire serious systemic forms and is caused by the thermodimorphic fungal Paracoccidioides spp. PCM is a tropical disease that is endemic in Latin America, where up to ten million people are infected; 80% of reported cases occur in Brazil, followed by Colombia and Venezuela. To enable genomic studies and to better characterize the pathogenesis of this dimorphic fungus, two reference strains of P. brasiliensis (Pb03, Pb18) and one strain of P. lutzii (Pb01) were sequenced [1]. While the initial draft assemblies were accurate in large scale structure and had high overall base quality, the sequences had frequent small scale defects such as poor quality stretches, unknown bases (N's), and artifactual deletions or nucleotide duplications, all of which caused larger scale errors in predicted gene structures. Since assembly consensus errors can now be addressed using next generation sequencing (NGS) in combination with recent methods allowing systematic assembly improvement, we re-sequenced the three reference strains of Paracoccidioides spp. using Illumina technology. We utilized the high sequencing depth to re-evaluate and improve the original assemblies generated from Sanger sequence reads, and obtained more complete and accurate reference assemblies. The new assemblies led to improved transcript predictions for the vast majority of genes of these reference strains, and often substantially corrected gene structures. These include several genes that are central to virulence or expressed during the pathogenic yeast stage in Paracoccidioides and other fungi, such as HSP90, RYP1-3, BAD1, catalase B, alpha-1,3-glucan synthase and the beta glucan synthase target gene FKS1. The improvement and validation of these reference sequences will now allow more accurate genome-based analyses. To our knowledge, this is one of the first reports of a fully automated and quality-assessed upgrade of a genome assembly and annotation for a non-model fungus.

Figure 1. Overview of genome assembly and annotation improvement process.

Figure 2. Examples of an artifactual insertion and an artifactual deletion that were corrected during the update of the P. brasiliensis Pb03 genome sequence.

Screenshots of Pilon-generated genome browser tracks in GenomeView v1.0 show the evidence used by Pilon to recognize and correct an incorrect insertion in the gene PABG_00120 (left) and an incorrect deletion in the gene PABG_00790 (right). Tracks (top panels) depict paired-end reads (green) aligned to the corresponding region of the reference assembly v1, a subset of the total depth of ∼150X or ∼170X; these alignments were used by Pilon to refine the consensus sequence, generating the improved Pb03 assembly v2. Positions in the v1 assembly where aligned reads suggest a change due to either a gap (red box) or an insertion (black line) are indicated with dashed red boxes. The changes suggested by Pilon are also supported by conservation of the changed bases in a multiple alignment (bottom panels) with the corresponding region of P. brasiliensis Pb18 and P. lutzii Pb01.

Figure 3. Improved consistency of gene annotation in v2 genomes.

The final predicted gene sets of the three Paracoccidioides strains were clustered using OrthoMCL, in v1 and v2. The scatterplots (A) compare, for each clustered group, the maximum length versus the minimum length of the three Paracoccidioides genes in the same cluster, for each of the two versions. The scatterplot contrasts the maximum-minimum pairs from annotation v1 (red points) and those from annotation v2 (blue points). The location of blue points closer to the diagonal illustrates that the annotation v2 was more consistent across the three genomes with smaller differences in gene length. In the same sense, the rank plots (B) show the difference between maximum and minimum length for each clustered group, for each of the two versions; again annotation v2 (blue line) showed fewer (later increase) and smaller (more gradual increase) differences, corresponding to the improvement of the genome annotation in v2.

Figure 4. Diverse error correction for the 90 kDa heat shock protein (HSP90 gene) of Paracoccidioides spp.

(A) In this example different annotation errors were present in v1 of all three Paracoccidioides reference strains, all of which were fixed in v2 after Pilon improvement and re-annotation. The example also illustrates how one or more single-nucleotide errors, unknown single nucleotides (N's), or single nucleotides that were erroneously reported as absent or duplicated by a Sanger sequencer can amplify across annotations, generating radically different gene structure (intron/exon and/or gene boundary) predictions. (B) Five changes are shown at assembly (DNA sequence) level, one of which was a single nucleotide error in a stop codon; as a result, the gene-calling program did not recognize the end of an exon and it was not reported.