Stable association between strains of Mycobacterium tuberculosis and their human host populations

Abstract

Mycobacterium tuberculosis is an important human pathogen in virtually every part of the world. Here we investigate whether distinct strains of M. tuberculosis infect different human populations and whether associations between host and pathogen populations are stable despite global traffic and the convergence of diverse strains of the pathogen in cosmopolitan urban centers. The recent global movement and transmission history of 100 M. tuberculosis isolates was inferred from a molecular epidemiologic study of tuberculosis that spans 12 years. Genetic relationships among these isolates were deduced from the distribution of large genomic deletions, which were identified by DNA microarray and confirmed by PCR and sequence analysis. Phylogenetic analysis of these deletions indicates that they are unique event polymorphisms and that horizontal gene transfer is extremely rare in M. tuberculosis. In conjunction with the epidemiological data, phylogenies reveal three large phylogeographic regions. A host's region of origin is predictive of the strain of tuberculosis he or she carries, and this association remains strong even when transmission takes place in a cosmopolitan urban center outside of the region of origin. Approximate dating of the time since divergence of East Asian and Philippine clades of M. tuberculosis suggests that these lineages diverged centuries ago. Thus, associations between host and pathogen populations appear to be highly stable.

Fig. 4.

Host place of birth mapped onto the maximum parsimony phylogeny for both reactivated and recently transmitted cases. As in Fig. 3, isolates for which deletions were not phylogenetically informative are not shown.

Fig. 1.

The distribution of a phage-associated deletion (deletion 149) on the strict consensus tree. Lineages in which deletion 149 was observed or reconstructed by maximum parsimony reconstruction are gray. The other phage-associated deletion (deletion 198a) also exhibited more than one occurrence on the consensus tree. By contrast, deletions that were not associated with phage occurred only once. For example, deletions 145, 178, and 182 exhibited the same pattern of occurrence, appearing in lineages shown in black.

Fig. 2.

(a) The temporal intervals in which nonclustered isolates were contracted by their patients. Each nonclustered patient's place of birth and, thus, the strain's likely place of origin is indicated by a color corresponding to the points on the world map in b. Each horizontal bar represents a single, nonclustered patient. The bar spans the time interval over which the M. tuberculosis isolate is likely to have been contracted by the patient, starting from date of birth and ending at immigration to the U.S. The point to the right of each bar indicates the date of diagnosis of active tuberculosis in the U.S. For the five nonclustered patients born in the U.S., the bar spans the interval from date of birth to date of diagnosis. (b) The places of birth of the patients whose isolates were studied. Each point represents a patient, and each color corresponds to a country; both clustered and nonclustered cases are indicated on the map.

Fig. 3.

(a) Phylogeny of M. tuberculosis isolates from cases that were nonclustered and, therefore, very likely to be caused by reemergence of latent infection. Patient place of birth, the likely location of origin of the M. tuberculosis isolate, is mapped onto the maximum parsimony tree based on deletions. Four large geographic regions are indicated by colors: blue corresponds to East Asia; pink to the Philippines; black to the Americas; and orange to Africa, Europe, and the Middle East. Here and in b and Fig. 4, country of origin is indicated by the abbreviations at the top of the tree: CB, Cambodia; CH, China; CO, Colombia; ET, Ethiopia; HK, Hong Kong; IR, Iran; MC, Macau; MX, Mexico; RP, Philippines; RQ, Puerto Rico; US, United States; VM, Vietnam; YO, Yugoslavia. Isolates for which deletions are not phylogenetically informative (i.e., they do not reveal membership in any clade with more than one nonclustered member) are not shown. (b) Phylogeny of M. tuberculosis isolates from cases that were clustered and, therefore, very likely to be from recent transmission in the city of San Francisco. As in a, patient place of birth is mapped onto the maximum parsimony tree based on deletions. Note that genetic groups of M. tuberculosis remain associated with patients from certain regions, despite the occurrence of infection in San Francisco.