Comparisons of genome assembly tools for characterization of Mycobacterium tuberculosis genomes using hybrid sequencing technologies.

Background: Next-generation sequencing of Mycobacterium tuberculosis, the infectious agent causing tuberculosis, is improving the understanding of genomic diversity of circulating lineages and strain-types, and informing knowledge of drug resistance mutations. An increasingly popular approach to characterizing M. tuberculosis genomes (size: 4.4 Mbp) and variants (e.g., single nucleotide polymorphisms (SNPs)) involves the de novo assembly of sequence data. Methods: We compared the performance of genome assembly tools (Unicycler, RagOut, and RagTag) on sequence data from nine drug resistant M. tuberculosis isolates (multi-drug (MDR) n = 1; pre-extensively-drug (pre-XDR) n = 8) generated using Illumina HiSeq, Oxford Nanopore Technology (ONT) PromethION, and PacBio platforms. Results: Our investigation found that Unicycler-based assemblies had significantly higher genome completeness (~98.7%; p values = 0.01) compared to other assembler tools (RagOut = 98.6%, and RagTag = 98.6%). The genome assembly sizes (bp) across isolates and sequencers based on RagOut was significantly longer (p values < 0.001) (4,418,574 ± 8,824 bp) than Unicycler and RagTag assemblies (Unicycler = 4,377,642 ± 55,257 bp, and RagTag = 4,380,711 ± 51,164 bp). RagOut-based assemblies had the fewest contigs (~32) and the longest genome size (4,418,574 bp; vs. H37Rv reference size 4,411,532 bp) and therefore were chosen for downstream analysis. Pan-genome analysis of Illumina and PacBio hybrid assemblies revealed the greatest number of detected genes (4,639 genes; H37Rv reference contains 3,976 genes), while Illumina and ONT hybrid assemblies produced the highest number of SNPs. The number of genes from hybrid assemblies with ONT and PacBio long-reads (mean: 4,620 genes) was greater than short-read assembly alone (4,478 genes). All nine RagOut hybrid genome assemblies detected known mutations in genes associated with MDR-TB and pre-XDR-TB. Conclusions: Unicycler software performed the best in terms of achieving contiguous genomes, whereas RagOut improved the quality of Unicycler's genome assemblies by providing a longer genome size. Overall, our approach has demonstrated that short-read and long-read hybrid assembly can provide a more complete genome assembly than short-read assembly alone by detecting pan-genomes and more genes, including IS6110, and SNPs.

High clustering rate and genotypic drug-susceptibility screening for the newly recommended anti-tuberculosis drugs among global extensively drug-resistant Mycobacterium tuberculosis isolates.

Multidrug-resistant tuberculosis (MDR-TB) and extensively drug-resistant tuberculosis (XDR-TB) make TB difficult to control. Global susceptibility data for six newly recommended anti-TB drugs against M/XDR-TB are still limited. Using publicly available whole-genome sequences, we determined the proportion of 513 phenotypically XDR-TB isolates that carried mutations associated with resistance against these drugs (bedaquiline, clofazimine, linezolid, delamanid, pretomanid and cycloserine). Mutations of Rv0678 and Rv1979c were detected in 69/513 isolates (13.5%) for bedaquiline resistance and 79/513 isolates (15.4%) for clofazimine resistance with additional mmpL5 mutations. Mutations conferring resistance to delamanid were detected in fbiB and ddn genes for 11/513 isolates (2.1%). For pretomanid, a mutation was detected in the ddn gene for 3/513 isolates (0.6%). Nineteen mutations of pykA, cycA, ald, and alr genes, conferring resistance to cycloserine, were found in 153/513 isolates (29.8%). No known mutations associated with linezolid resistance were detected. Cluster analysis showed that 408/513 isolates fell within 99 clusters and that 354 of these isolates were possible primary drug-resistant TB (292 XDR-TB, 57 pre-XDR-TB and 5 MDR-TB). Clonal transmission of primary XDR isolates might contribute significantly to the high prevalence of DR-TB globally.