Extensively drug-resistant tuberculosis in South Africa: genomic evidence supporting transmission in communities

Abstract

Despite evidence that transmission is driving an extensively drug-resistant TB (XDR-TB) epidemic, our understanding of where and between whom transmission occurs is limited. We sought to determine whether there was genomic evidence of transmission between individuals without an epidemiologic connection.We conducted a prospective study of XDR-TB patients in KwaZulu-Natal, South Africa, during the 2011-2014 period. We collected sociodemographic and clinical data, and identified epidemiologic links based on person-to-person or hospital-based connections. We performed whole-genome sequencing (WGS) on the Mycobacterium tuberculosis isolates and determined pairwise single nucleotide polymorphism (SNP) differences.Among 404 participants, 123 (30%) had person-to-person or hospital-based links, leaving 281 (70%) epidemiologically unlinked. The median SNP difference between participants with person-to-person and hospital-based links was 10 (interquartile range (IQR) 8-24) and 16 (IQR 10-23), respectively. The median SNP difference between unlinked participants and their closest genomic link was 5 (IQR 3-9) and half of unlinked participants were within 7 SNPs of at least five participants.The majority of epidemiologically-unlinked XDR-TB patients had low pairwise SNP differences with at least one other participant, consistent with transmission. These data suggest that much of transmission may result from casual contact in community settings between individuals not known to one another.

Figure 1.

Flow diagram for participant enrollment.

Figure 2.

Flow diagram for availability of whole-genome sequencing data and median SNP differences for participants with and without epidemiologic links. SNP = single nucleotide polymorphism; WGS = whole-genome sequencing; RFLP = restriction fragment length polymorphism; IQR = interquartile range.

Figure 3.

Distribution of pairwise SNP differences for participants with a matching RFLP pattern with their name-based epidemiologic link (n = 29), a hospital-based epidemiologic link (n = 37), and without an epidemiologic link (i.e., unlinked participants) (n = 240).