Tuberculosis (TB), caused by Mycobacterium tuberculosis, is a significant public health concern in Japan, being the leading cause of death from a single bacterial infection [1]. Detecting and preventing the expansion of clusters has been the main approach to preventing the spread of TB, but epidemiological studies have difficulty in identifying the route of infection because of the long incubation period [1]. In particular, studies have shown that TB-infected individuals with identifiable epidemiological links are less common in large urban areas such as Tokyo and its suburbs than in rural areas [2]. Reasons for this include the high proportion of newly registered patients who are of working age in urban areas and the active population flux of those of working age within the urban employment area, which includes an economic zone that contains a city and its surrounding areas. To overcome the above shortcoming, multi-locus variable number tandem repeat analysis (MLVA) [3], [4], a form of molecular epidemiological analysis, is currently used to supplement epidemiological investigations.

However, MLVA has two limitations. First, it is associated with the risk that genetically unrelated strains may be mistakenly identified as identical because of homoplasy [5], [6], [7]. Second, human-to-human transmission is thought to cause single-locus variants (SLVs) [8], but tracing from the strain from which an SLV originally developed to a further SLV strain obscures the extent to which outbreaks can be identified as being caused by the same clone. Additionally, there are conflicting reports on whether SLVs occur because of the spread of infection [9]. Furthermore, it has been shown that MLVA patterns that persist over multiple years are not necessarily indicative of a single outbreak [10]. Therefore, no criteria have been established to reliably track the route of infection using MLVA. Meanwhile, molecular epidemiological analysis based on whole-genome sequencing (WGS) data has established criteria for determining that M. tuberculosis strains are derived from the same clone [11]. However, cost continues to be the primary obstacle to implementing WGS for the surveillance of pathogens such as M. tuberculosis, which infects a large number of people [12], [13]. Additionally, beyond cost, the experimental burden of processes such as library preparation and sequencing poses a significant challenge [14].

Population genetic analysis can be useful in addressing the limitations of molecular epidemiological analysis. Since bacteria are self-replicating, they generally retain ancestral traits [15]. M. tuberculosis strains can be subdivided into phylogenetic groups from MLVA data using maximum a posteriori (MAP) estimation based on Bayes’ theorem, taking advantage of the fact that each subfamily has a unique MLVA pattern [16]. Bayesian cluster analysis can also reveal the genetic structure of bacterial strain populations using genomic features [17]. In other words, even a population of strains belonging to the same subfamily can be subdivided into multiple subpopulations by cluster analysis. This can occur not only because of the presence of several genetically distinct subpopulations [18], but also because a large number of strains originating from the same clone may be present within the population [19]. In both cases, linkage disequilibrium (LD; i.e., bias of genomic features) would be observed within subpopulations [18], [19]. Therefore, a Bayesian population genetic analysis of MLVA data would help narrow down the strains that need to be subjected to WGS. However, there are no reports of the successful detection of clusters of M. tuberculosis in actual infection cases using a combination of Bayesian population genetics and molecular epidemiological analysis.

In recent years, the number of cases of M. tuberculosis infection newly registered in Japan has been declining [20], but it was suspected that an outbreak occurred in Chiba Prefecture because the incidence increased in 2016 (Supplementary Table 1) [21], [22]. As part of the tuberculosis prevention program, all medical institutions in Chiba Prefecture have been requested to submit strains since 2008, and all strains of submitted M. tuberculosis have been analyzed by MLVA. In this study, we attempted to identify the cause of the increase in Chiba Prefecture by performing a population genetic analysis of MLVA data to select strains for efficient WGS. First, the strains were subdivided into subpopulations within each subfamily using population genetic analysis of MLVA data, and then the validity of this subdivision was assessed using LD within the subpopulation. Second, the annual trends of subpopulation strains were examined, and WGS was performed on subpopulations that showed an increase to identify the cause of the rise.