Use of whole-genome sequencing to distinguish relapse from reinfection in a completed tuberculosis clinical trial

Abstract

Background: RIFAQUIN was a tuberculosis chemotherapy trial in southern Africa including regimens with high-dose rifapentine with moxifloxacin. Here, the application of whole-genome sequencing (WGS) is evaluated within RIFAQUIN for identifying new infections in treated patients as either relapses or reinfections. WGS is further compared with mycobacterial interspersed repetitive units-variable number tandem repeats (MIRU-VNTR) typing. This is the first report of WGS being used to evaluate new infections in a completed clinical trial for which all treatment and epidemiological data are available for analysis.

Methods: DNA from 36 paired samples of Mycobacterium tuberculosis cultured from patients before and after treatment was typed using 24-loci MIRU-VNTR, in silico spoligotyping and WGS. Following WGS, the sequences were mapped against the reference strain H37Rv, the single-nucleotide polymorphism (SNP) differences between pairs were identified, and a phylogenetic reconstruction was performed.

Results: WGS indicated that 32 of the paired samples had a very low number of SNP differences (0-5; likely relapses). One pair had an intermediate number of SNP differences, and was likely the result of a mixed infection with a pre-treatment minor genotype that was highly related to the post-treatment genotype; this was reclassified as a relapse, in contrast to the MIRU-VNTR result. The remaining three pairs had very high SNP differences (>750; likely reinfections).

Conclusions: WGS and MIRU-VNTR both similarly differentiated relapses and reinfections, but WGS provided significant extra information. The low proportion of reinfections seen suggests that in standard chemotherapy trials with up to 24 months of follow-up, typing the strains brings little benefit to an analysis of the trial outcome in terms of differentiating relapse and reinfection. However, there is a benefit to using WGS as compared to MIRU-VNTR in terms of the additional genotype information obtained, in particular for defining the presence of mixed infections and the potential to identify known and novel drug-resistance markers.

Keywords: Clinical trial; Tuberculosis; Whole genome sequencing.

Fig. 1

Flowchart of pairs of samples studied. MIRU mycobacterial interspersed repetitive units, WGS whole-genome sequencing

Fig. 2

a Phylogenetic reconstruction of 36 pairs of isolates. These were inferred using 5132 high-quality single-nucleotide polymorphisms (SNPs) following the removal of 661,083 low-quality sites and the remaining invariant sites. The tree was rooted using the H37Rv reference strain sequence. Relapse, reinfection and mixed are denoted with black/blue, green and red tips respectively. Blue tip labels are further shown in panels b–e. b–e Branches have been amplified where unexpected similarity was seen; the numbers of SNPs between the most divergent samples are given

Fig. 3

Identification of mixed infection. a Counts of genome sites which were called as a reference base but showed a significant proportion of sequence reads also supporting a variant base call (035-1); b the equivalent plot for an isolate with no mixed infection (035-2). The presence of a second peak in a is suggestive of a mixture with a minority genotype

Fig. 4

Analysis of single-nucleotide polymorphism (SNP) and mycobacterial interspersed repetitive units-variable number tandem repeat (MIRU-VNTR) differences between pairs of isolates. Data are summarised from Tables 1 and 2. a Number of SNP differences detected between paired isolates; b number of MIRU-VNTR differences detected between paired isolates; c correlation between SNP and MIRU differences; d number of informative MIRU loci on which differences were based (for each pair of samples, the lower number is shown)