Identifying mixed mycobacterium tuberculosis infection and laboratory cross-contamination during mycobacterial sequencing programs

Abstract

The detection of laboratory cross-contamination and mixed tuberculosisinfections is an important goal of clinical mycobacteriology laboratories. The objective of this study was to develop a method to detect mixtures of different Mycobacterium tuberculosis lineages in laboratories performing mycobacterial next-generationsequencing (NGS). The setting was the Public Health England National Mycobacteriology Laboratory Birmingham, which performs Illumina sequencing on DNA extracted from positive mycobacterial growth indicator tubes. We analyzed 4,156 samples yielding M. tuberculosis from 663 MiSeq runs, which were obtained duringdevelopment and production use of a diagnostic process using NGS. The counts ofthe most common (major) variant and all other variants (nonmajor variants) weredetermined from reads mapping to positions defining M. tuberculosis lineages. Expected variation was estimated during process development. For each sample, wedetermined the nonmajor variant proportions at 55 sets of lineage-defining positions.The nonmajor variant proportion in the two most mixed lineage-defining sets (F2 metric) was compared with that of the 47 least-mixed lineage-defining sets (F47 metric). Thefollowing three patterns were observed: (i) not mixed by either metric; (ii) high F47 metric, suggesting mixtures of multiple lineages; and (iii) samples compatible with mixturesof two lineages, detected by differential F2 metric elevations relative to F47. Pattern iiwas observed in batches, with similar patterns in the M. tuberculosis H37Rv control present in each run, and is likely to reflect cross-contamination. During production, the proportions of samples in the patterns were 97%, 2.8%, and 0.001%, respectively. The F2and F47 metrics described could be used for laboratory process control in laboratoriessequencing M. tuberculosis genomes.