Background
Tuberculosis (TB) has been traditionally assumed to result from a single infection with a single Mycobacterium tuberculosis strain, and this infection is thought to confer immunity to additional infections. Therefore, a recurrence of disease has been most often considered to be caused by endogenous reactivation of the strain that caused the original infection (relapse). Consequently, almost all current analytical schemes of clinical or research relevance are still based on examination of single isolates of given disease episodes, with implicit assumption that this isolate is representative of an homogeneous bacillary population.
This model of homogeneous infection has been revised by several studies using strain typing methods, which have demonstrated the occurrence of infection with clonally distinct strains, especially in high-incidence settings [
1‐
3]. Both human immunodeficiency virus (HIV)-negative and HIV-positive individuals can be infected with more than one strain during a given disease episode (mixed infection), or re-infected by a second
M. tuberculosis strain during a recurrent episode (exogenous re-infection). Such findings have important implications for control programs, vaccine development, evaluation of treatment regimens [
4], and for epidemiological interpretation [
3,
5].
However, these studies of re-infection and mixed infection have so far been conducted by analysing the genotypes of the isolate from one sputum specimen from the initial and recurrent episodes or from one given episode, respectively [
6‐
8]. Such approaches discount the old postulate that bacilli sequestered at different pulmonary infection sites are not necessarily released in the sample provided. Therefore, analysis of a single isolate might underestimate the actual heterogeneity of the bacillary population in the host. Conversely, the consequences of clonal heterogeneity on the representativeness of a single isolate have remained unknown hitherto.
Here, we have prospectively evaluated both the frequency of mixed infections and the clonal heterogeneity among clinical isolates from the same patient by analysing at least two pre-treatment isolates from each of 199 TB patients from a prison TB hospital in Georgia, consecutively enrolled over a period of three years. These isolates were analysed by using standard IS6110-restriction fragment length polymorphism (RFLP) genotyping as a first-line screening, followed by typing based on PCR amplification of 15 different loci containing mycobacterial interspersed repetitive unit-variable number of tandem repeats (MIRU-VNTRs) for independent confirmation of simultaneous presence of multiple strains. The implications of the results for current analytical schemes of drug susceptibility testing (DST) and for evaluation of the contribution of re-infection to the epidemiology and pathogenesis of this disease are discussed.
Discussion
This report simultaneously assessed the validity of two interdependent postulates on which standard analytical schemes rely: (i) that a TB patient can only be infected with a single homogeneous M. tuberculosis strain at any given time, and (ii) that an isolate from a single sputum specimen is representative of the total bacillary population in a patient. Therefore, we systematically compared the genetic relatedness of M. tuberculosis isolates from multiple sputum samples collected prior to the initiation of anti-TB therapy from each of 199 smear-positive inmates admitted to a prison TB hospital. By using two independent genotyping methods to differentiate strains, we detected infection with two or even three distinct M. tuberculosis strains in 13.1% of the samples analysed. There was no significant difference in the proportion of retreatment cases between the excluded and included patients.
The mixed infection rate observed in this prison population can not be extrapolated to the general population because of overcrowding and higher incidence of TB in the prisons compared to the general population (5,995/100 000 vs. 155/100 000 population, respectively) [
26]. However, we believe that the so called "cheating" (
i.e. prisoners attempt to submit sputa mixed with that of other prisoners suspected of having smear-positive TB, so that they can be diagnosed with TB and transferred to the prison TB hospital with better living conditions than in other detention centres) had a low influence on our estimation of this phenomenon, if any, due to the strict and active surveillance by an aware staff at the sputum collection step inside the TB hospital. Likewise, laboratory cross-contamination is an unlikely explanation for the high frequency of mixed infection detected because of the specific precautions taken.
The use of IS
6110-RFLP-typing as an initial screening method may have led to some underestimation of mixed infection because this method has inherent limitations to detect mixed infections within a single isolate since various bands in a given profile can represent one or more strains. Moreover, it remains unclear to what extent low ratios of one of the strains present in a mixture are reflected in low-intensity bands [
27] or not detected at all. The latter is evidenced by the detection of mixed infections by MIRU-VNTR among 3 (10%) of 30 patients with isolates that showed identical IS
6110-RFLP patterns but double alleles in multiple loci within one isolate by MIRU-VNTR. By extrapolation, this suggests that up to 14 (7.0%) mixed infections might have been additionally detected among the other 139 patients with isolates that had identical IS
6110-RFLP patterns if they were also tested by MIRU-VNTR. There was no statistically significant difference in retreatment type between new and previously treated cases among the patients whose isolates were only typed by IS
6110-RFLP and those whose isolates were additionally typed by MIRU-VNTR (χ
2, p = 0.15). Finally, some mixed infections could have remained undetected by MIRU-VNTR typing itself although this PCR-based method is able to detect ratios of a given strain as low as 1:99 [
28].
Detection of genetically distinct strains among multiple pre-treatment sputum samples, as well as within a single sputum specimen might reflect separate lesions in the lungs containing different
M. tuberculosis strains and opening simultaneously or consecutively as also suggested from a previous study [
29]. Regardless of the explanation, it is crucial to note that if only the first pre-treatment sample was analysed by standard IS
6110 fingerprinting or by MIRU-VNTR typing, none or only about half of the mixed infection cases detected by analysis of multiple pre-treatment samples (14 cases vs. 26 cases) would have been identified. These observations imply that analysis of a single isolate, especially in high incidence settings may underestimate the actual heterogeneity of the bacillary population in the host.
It is relevant to observe that the 13.1% of mixed infections detected in this prison population is relatively close to the frequency of 19% recently reported in the study of Warren
et al. in a general population of a setting with an incidence 1000/100 000 population [
3]. Although the two values are not directly comparable as this latter evaluation was limited to the detection of patients simultaneously infected with strains of both the Beijing and non-Beijing lineages, we predict that their value is likely an underestimation as only single isolates per patient were analysed in that case. Similarly, previous studies in high incidence settings might have overestimated the contribution of reinfection vs. relapse due to undetected initial mixed infection [
7,
29]. Our observations imply that for specific research studies analyses of several isolates from different sputum samples at each disease episode (before and after treatment), especially in high TB incidence settings might be helpful in distinguishing true reinfection vs. relapse and/or mixed infection, preferably using a PCR-based typing method like MIRU-VNTR.
From the 26 proven mixed infection cases in our study, 30% harboured only pan-susceptible strains, whereas 70% showed any resistance in at least one of the isolates. In only 6 cases was mixed infection reflected in a variant DST profile. Remarkably, the remaining patients were infected with two or three strains seemingly showing identical resistance profiles. Although most of these patients were new cases according to WHO definitions [
9], we can not completely exclude the possibility that they might have taken TB drugs for less than one month, and therefore both strains might have acquired resistance as a result of the same drug pressure. On the other hand, independent infection with two or three strains showing exactly the same resistance profile for each of so many patients is very doubtful as well, even in a setting with a high rate of drug-resistant TB. Most probably, such frequent observations of identical resistance profiles among respective isolates from these mixed infections reflect the systematic presence of both a susceptible and a resistant strain in the corresponding specimens. This systematic duality was evidenced by the systematic detection of double alleles in the MIRU-VNTR patterns in the isolates from all these cases but one. In such conditions, the simultaneous growth of a (more) resistant strain will mask that of susceptible (or partly resistant) strains in DST assays either completely or partly resulting in either resistant isolates or isolates with borderline results (patient 9). In such situations, culturing and DST of single pre-treatment sputum had generally no predictable adverse consequences for the appropriateness of the treatment regimen of the respective patients.
As mentioned above, variant DST profiles were detected as a result of mixed infections in only 6 (3.0%) of the 199 patients. This finding lends support to previous reports that initial mixed infections may actually be responsible for changes in DST patterns in isolates of some patients [
29‐
31].
In general, our findings suggest that single-isolate analyses can be used for routine DST in most settings, except for some high drug resistant-TB-prevalent settings. However, for specific research studies like treatment evaluation and clinical trials, testing multiple isolates from different sputum samples at each disease episode could help in determining the respective contribution of mixed infection and reinfection versus relapse with gradual development of drug resistance, especially by PCR-based typing methods such as MIRU-VNTR.
Although the high rates of mixed infection in this prison setting can not be extrapolated to the general population with a lower risk of TB transmission, these findings nonetheless indicate that an initial infection is unable to provide protection against a subsequent infection in these populations, which have implications for the development and trials of new vaccines [
3]. Because higher rates of mixed infection imply possible higher rates of super infection, the protective effect of an initial infection against a subsequent infection may be even lower than expected. This parameter needs to be taken into account in the development of new prophylactic approaches.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ICS: Performed DNA fingerprinting, evaluated the data, drafted and reviewed the manuscript. LJ: Co-ordinated the study and helped in drafting the manuscript. NS: Supervised sample collection, culture and identification of mycobacteria. EW: Participated in MIRU-VNTR typing. FP: Conceived of the study, participated in its design and reviewed the manuscript. PS: Supervised MIRU-VNTR typing, evaluated the data, helped in drafting and revising the manuscript. LR: Conceived of the study, participated in its design, co-ordination and evaluation of data, and helped in drafting the manuscript. All authors read and approved the final manuscript.