Abstract
Despite the almost 50 yrs since the introduction of curative antituberculosis drugs, Mycobacterium tuberculosis continues to exert an enormous toll on world health, and tuberculosis remains the world's leading cause of death due to a single infectious agent. This has stimulated research efforts into finding new tools to tackle the continuing tuberculosis pandemic.
One of the few successes to date has been the development of a new discipline, molecular epidemiology. This has added a further dimension to the classical epidemiology of tuberculosis and enhanced understanding of how M. tuberculosis continues to be successfully transmitted within populations. In the process, inadequacies in tuberculosis control programmes have been identified, helping accumulate resources for their improvement.
Other technologies, based on knowledge of the complete genome sequence of M. tuberculosis, which will provide newer tools for probing the epidemiology of tuberculosis, are now emerging. In spite of these advances, tuberculosis continues to remain a devastating infectious disease, disproportionately impacting on the world's poorest countries.
The future challenge for molecular epidemiology is to provide better understanding of the transmission dynamics of tuberculosis in these settings and to stimulate the implementation of control measures on a more global scale.
- drug resistance
- fitness
- genomics
- molecular epidemiology
- Mycobacterium tuberculosis
- restriction fragment length polymorphism
Mycobacterium tuberculosis is one of the most successful bacterial pathogens in the history of mankind. Despite antituberculosis drugs having been available for almost 50 years, M. tuberculosis continues to exert an enormous toll on world health (fig. 1a⇓). Between a third and a half of the world's population is infected with M. tuberculosis. Each year, there are ∼2 million deaths due to tuberculosis, making tuberculosis the world's leading cause of mortality due to a single infectious agent 2. Tuberculosis is the number one cause of death among human immunodeficiency virus (HIV)-infected individuals (fig. 1b⇓) 2. In 1999, there were 8.4 million cases reported worldwide, and, for the year 2005, the World Health Organization projects an incidence of 10.2 million new cases 1. This increased incidence will occur mostly in countries in Africa and Asia, where the highest prevalence of coinfection with HIV and M. tuberculosis occurs. The economic impact of this pathogenic synergy is particularly great because HIV disproportionately affects persons during the most productive years of their lives.
The resurgence of tuberculosis around the world has renewed interest in understanding the epidemiology and pathogenesis of this disease. One important advance in the field of tuberculosis research has been the development of molecular techniques that allow the identification and tracking of individual strains of M. tuberculosis. This new discipline, the molecular epidemiology of tuberculosis, began with the identification of IS6110, a novel mycobacterial insertion sequence which formed the basis of a reproducible genotyping technique for M. tuberculosis. This method is now firmly established, but is still expensive, labour-intensive and only applicable using viable culture material. Although other typing methods, at varying stages of development, appear to offer certain advantages in terms of reproducibility, cost, ease of execution and general applicability to clinical settings, IS6110-based typing remains the internationally accepted standard and continues to provide new insights into the epidemiology of M. tuberculosis.
A newer research approach, initiated by determination of the complete genome sequence of M. tuberculosis, is to combine these conventional molecular epidemiological techniques with developments in mycobacterial genomics 3. The goal is to employ the array of typing techniques now available for the identification of individual strains or clonal groupings of strains with specific phenotypic characteristics, such as transmissibility, antigenicity or resistance to antimicrobial agents. These strains can then be subjected to genome-wide analysis, using techniques such as microarrays for expression profiling or detection of genomic deletions, to determine the genetic basis of these important phenotypic traits. This multidisciplinary approach could lead to important advances in understanding the pathogenesis of human tuberculosis as well as mechanisms of drug resistance.
Molecular epidemiological markers
Before the introduction of molecular typing methods, there was little to aid distinction between individual strains of M. tuberculosis. Drug susceptibility patterns had been used but were of limited utility because patterns can change and resistance is rare in most populations. Typing by variability in susceptibility to infection with mycobacterial phages has also been used, but was found to be difficult to reproduce and limited by the number of phages available. The development and application of molecular typing methods, such as restriction fragment length polymorphism (RFLP) analysis based on the insertion sequence IS6110, in the 1990s has brought a new dimension to the study of tuberculosis and with it a new appreciation of the ecological complexities that classical epidemiology could not provide. IS6110 is the most extensively studied and widely used of the known insertion sequences, although the availability of the M. tuberculosis genome sequence has led to the identification of >30 additional repetitive elements which promise to be useful typing markers 4. This article reviews the most commonly used typing methods, newer approaches and the application of these techniques in the study of tuberculosis.
IS6110 restriction fragment length polymorphism
IS6110 was initially described in 1989 5. This marker comprises 1,355 base pairs and belongs to the IS3 family of insertion sequences. IS6110 has two open reading frames encoding proteins required for transposition. The number of copies of IS6110 ranges 0–25, and their positions in the M. tuberculosis chromosome are highly variable between different isolates (fig. 2⇓) 7. This variability is sufficient to generate RFLP and for it to be used in fingerprinting. IS6110 is exclusively present in the M. tuberculosis complex species, although strains of M. tuberculosis lacking this element have also been described 8–11. IS6110 sequences are nonrandomly distributed, suggesting insertional hot spots. One of these is an area flanked by a direct repeat (DR) sequence 12. This nonrandom distribution of IS6110 within the genome is now recognised as a limitation to the discriminative power of typing based on the copy number and position of this sequence.
IS6110-based typing is the most widely applied genotyping method in the molecular epidemiology of M. tuberculosis and is the gold standard to which other methods are currently compared 13. Various important factors have been identified for the standardisation of M. tuberculosis genetic fingerprinting, including the use of the restriction enzyme Proteus vulgaris II for deoxyribonucleic acid (DNA) digestion and the incorporation of molecular weight standards for the estimation of band sizes. This standardisation has facilitated the comparison of fingerprints obtained in different laboratories around the world, allowing the global dispersal of strains to be tracked. For example a HIV-seropositive patient developed primary tuberculosis from a multidrug-resistant tuberculosis (MDR-TB) strain in San Francisco, CA, USA, which was found to be a unique fingerprint in the local database. On further questioning, the patient revealed that he had been hospitalised in Buenos Aires, Argentina, the previous year, during the time of a reported large outbreak of MDR-TB 14. The San Francisco strain was found to have an identical IS6110 RFLP pattern and antibiotic resistance profile to the Buenos Aires strains, demonstrating the geographical extent to which individual strains of M. tuberculosis can be disseminated. A similar situation occurred with a HIV-seropositive patient that developed tuberculosis in Holland who was shown to have acquired tuberculosis after exposure in a hospital in Spain 15. The standardised methodology and centralised databases made it possible to confirm that these patients acquired their infections while abroad and not in the area in which they lived.
Although the number of copies of IS6110 can range 0–25, population-based molecular epidemiological studies report that most strains contain 8–18 copies, a number sufficient to enable discrimination between the majority of strains. For example data from San Francisco show that, of 1,800 patients over a period of 9 yrs, 1,117 (62.2%) had distinct DNA fingerprints and 683 (37.9%) were in 171 clusters sharing identical patterns. Less than 1% of the strains identified had two or less bands 11. However, there are geographical areas in Asia and Africa in which the diversity of IS6110 is considerably reduced 8–10, 16, 17. In addition, greater numbers of IS6110 bandless strains have been detected in Asia 7, 8. The lack of polymorphism associated with low copy numbers limits the discriminatory power and the epidemiological inferences that can be drawn with this typing method. However, additional or secondary typing systems can be used to discriminate between strains with few copies 10, 18–22.
An important characteristic of a genetic marker is its stability over time. Markers which change too rapidly obscure epidemiological links, whereas those that are too stable infer direct links where they do not exist. Understanding the rate and determinants of IS6110 pattern change is therefore important for optimal interpretation of fingerprint patterns. One study found that the banding patterns of isolates collected ≥90 days apart from the same patient were different in 29% of cases 23. In the Netherlands, the half-life of change in IS6110 was estimated to be 3.2 yrs 24. Interestingly, extensive serial passage of strains in vitro, such as for bacille Calmette-Guérin vaccine (attenuated tuberculosis vaccine) or of M. tuberculosis strains, is not associated with such rapid change in RFLP fingerprint profiles 10, 25. It is possible, therefore, that either the studies involving serial secretors overestimate the rate of change or that replication within a host can provoke IS6110 transposition.
Secondary markers
Empirical evidence demonstrates that the certainty with which epidemiological links can be inferred between patients infected with M. tuberculosis is markedly reduced if the strains involved yield less than five IS6110 hybridizing bands, but can be improved using an additional marker 18–20, 26–29. Thus genotyping with a secondary marker is essential for determining whether transmission has occurred in these cases. Two techniques, polymorphic guanine-cytosine-rich repetitive sequence (PGRS) profiling and spoligotyping, have been the most extensively used for this purpose.
Polymorphic guanine-cytosine-rich repetitive sequence restriction fragment length polymorphism typing
The identification of a PGRS present in multiple chromosomal clusters 30–32 enabled the development of a second RFLP typing system. This has been shown to have a discriminatory power close to that of IS6110 typing, even in isolates with low copy numbers of IS6110, making it an ideal secondary typing system 21, 26, 27, 33. However, the PGRS regions comprise many nonperfect repeats, making the RFLP patterns complex and sometimes difficult to interpret. Initially, these repetitive sequences were thought to be noncoding, but analysis of the M. tuberculosis genome has established that they code for the C‐termini of a novel class of proteins of unknown function.
Spoligotyping
The use of DNA RFLP analysis to distinguish between strains of M. tuberculosis is hindered by the need to culture this organism. IS6110 RFLP analysis is also technically demanding and costly, requiring expensive computer software for interpretation and comparison of fingerprints. Conversely, polymerase chain reaction (PCR)-based methods require very small amounts of DNA, which can be obtained without resorting to bacterial culture. In addition, PCR can be carried out directly on clinical specimens, enabling simultaneous identification and strain typing of M. tuberculosis in sputum from patients 34–44.
Spoligotyping, which interrogates a DR sequence comprising a repetitive 36-base-pair element separated by short nonrepetitive sequences, is one such PCR-based technique 30, 39. Using one set of primers, it is possible to simultaneously amplify all of the unique nonrepetitive sequences, or spacers, between the direct repeats. The presence or absence of spacers is then determined via Southern hybridisation. Individual strains are distinguished by the number of spacers missing from the complete spacer set defined by sequencing this region from a large number of M. tuberculosis strains.
Spoligotyping has been shown to be helpful when discriminating between isolates of M. tuberculosis with few IS6110 bands 40, 42. Atypical strains of mycobacteria have been analysed using spoligotyping and do not give a signal, indicating the specificity of this technique for M. tuberculosis 42. The ability to perform spoligotyping directly on sputum samples makes it applicable in acute clinical situations 39, 45. Another advantage of this secondary marker technique is that it is economical, easy to perform and a rapid means of typing the M. tuberculosis complex. These characteristics make it a candidate for use in resource-poor situations. Nonetheless, the discriminatory power of this method is less than that of IS6110 typing 40, 46–48. Since M. tuberculosis isolates with a different spoligotype invariably have distinguishable IS6110 profiles, spoligotyping could conceivably be used as an initial screening step before applying a secondary technique of greater discriminatory power 42, 47, 48.
Mycobacterial interspersed repetitive unit variable number of tandem repeats and other polymerase chain reaction-based techniques
A variety of other PCR-based techniques for typing M. tuberculosis isolates have been developed. In a recent interlaboratory comparative study, six such techniques were compared for their discriminatory power and reproducibility 46. These techniques were easy and quick to perform but only two of them, mixed-linker PCR and the variable number of tandem repeats (VNTR) method were sufficiently reproducible 44, 49. However, of these two techniques, only mixed-linker PCR was comparable to IS6110 typing in terms of discriminatory power. Unfortunately, mixed-linker PCR is also a technique that depends on IS6110-generated polymorphisms, and so is of limited use amongst low-band-number isolates. This is also the case with another promising IS6110-based PCR technique, ligation-mediated PCR 46.
A more promising approach for developing a PCR-based typing system is to identify novel polymorphic loci, which are independent of existing techniques such as IS6110 typing. Mycobacterial interspersed repetitive units (MIRUs) are an example of such elements 50–52. They are a specific class of VNTR that have been identified at 41 different loci in the genome of M. tuberculosis. Each comprises strings of short repetitive sequences (<100 base pairs). The number of repeats at different loci varies between strains. PCR amplification across each MIRU, therefore, generates fragments of different sizes from different strains. If these fragments are accurately sized, the number of repeats at each loci can be determined. Analysing 12 of the most hypervariable loci resulted in a discriminative power close to that of IS6110 typing. This type of approach is particularly suited for use with global databases as each typed strain is assigned a 12-digit number corresponding to the number of repeats at each MIRU locus 53. This unambiguous coding system makes interlaboratory comparisons facile. The main limitation of MIRU typing is the technical difficulty associated with accurately sizing multiple small PCR fragments. This can be partly overcome by combining multiplex PCR with a fluorescence-based DNA analyser 52.
Molecular markers and phylogenetics
Most attempts at developing typing techniques for the study of tuberculosis have been aimed at distinguishing between individual strains of M. tuberculosis in order to define chains of transmission. The genetic mechanisms underlying such techniques need to be highly polymorphic, with rapid molecular clocks, to generate sufficient diversity amongst strains. However, these techniques are not particularly well suited to studying phylogenetic relationships between more distantly related strains. For example IS6110 copy number showed no relationship to a phylogeny of diverse M. tuberculosis strains based on single nucleotide polymorphisms in housekeeping genes 54. Interest in defining these phylogenetic relationships has developed recently with the identification of groups of closely related strains, such as the Beijing family (see below), that appear to have a specific phenotype.
Deletion analysis is a particularly attractive approach to studying the phylogeny of M. tuberculosis strains as it could simultaneously provide information about the biological basis of a unique strain phenotype 55. In essence, the genome of a strain is evaluated using a microarray to determine whether or not any deletions have occurred relative to the sequenced reference strain. Since these deletions rarely occur independently at exactly the same chromosomal locus, they can be seen as unique and irreversible genetic events. The number and distribution of these deletions, therefore, provides a genomic signature which can be used for constructing robust phylogenetic relationships. These deletions disrupt coding regions of the genome and it is tempting to speculate that the loss of specific genes could influence important characteristics of strains such as transmissibility or antigenicity 3, 56.
Lessons from molecular epidemiology
Dynamics of transmission within populations
Molecular epidemiological investigations have been very useful in providing an understanding of the transmission dynamics of tuberculosis within a community. These studies are based upon the premise that patients infected with strains showing identical fingerprints, termed “clustered cases”, are the result of recent transmission, whereas those infected with isolates with unique RFLP patterns are presumed to represent remote transmission and thus reactivation of strains acquired in the more distant past.
Before the availability of more reliable molecular epidemiological tools, the belief was that 10% of patients developed M. tuberculosis disease as a result of recent transmission. However, population-based molecular epidemiological studies in San Francisco, (CA, USA), New York (NY, USA) and Amsterdam (the Netherlands) have refined the understanding of this subject. These studies demonstrated that the rate of recent infection was much higher than the estimated 10% predicted by traditional epidemiological studies. For example, in San Francisco, almost one-third of new cases of tuberculosis were as a result of recent infection. In New York, clustering was estimated to be 40% and, in Amsterdam, 47% 57–59. In San Francisco, one patient was found to be the index case and accounted for the transmission of 6% of new cases of tuberculosis in the city during 1991–1992. In these studies, the risk factors associated with recent transmission were lower socioeconomic group, native ethnic minority and acquired immune deficiency syndrome. These studies have important ramifications for tuberculosis control. They demonstrate that ongoing transmission of infection contributed to the disease burden at much higher rates than previously thought, and highlighted the importance of control efforts in interrupting transmission, especially amongst groups at high risk.
Population-based studies in Norway and Switzerland showed percentages of clustering that were relatively low (16 and 17.5% respectively) compared to the other studies 60, 61. This low level of recent transmission suggested that tuberculosis control was more effective in these settings. RFLP fingerprinting studies, therefore, can be used as a tool to monitor the performance of a tuberculosis control programme. In addition, this type of study can be used to identify specific risk factors for tuberculosis transmission and assist in targeting interventions to the subpopulations that disproportionately contribute to transmission. For example, in San Francisco, intensification of tuberculosis control decreased the overall numbers of those with recent infection 62 but also demonstrated persistent transmission among difficult-to-target high-risk groups.
The general principle that clustering of IS6110 patterns equates with recent transmission might not always apply in all situations. For instance, studies have shown that the rate of clustering varies depending on the area under study. In resource-poor countries, clustering rates of 14–41% were found 9, 29, 63–66. These findings were unexpected since the rates of clustering in some of these studies are comparable to or lower than those seen in low incidence countries. In these higher-incidence areas, with poor tuberculosis control, more transmission is expected. However, most of these studies do not report the incidence of tuberculosis, and, in many cases, it is difficult to ascertain the percentage of samples analysed in the study of those available in the community. This is particularly important because the number of isolates sampled in the study may be small in comparison to the total number of circulating isolates contributing to transmission in the area. Since the degree of sampling affects clustering, the rates of transmission may have been grossly underestimated 67, 68. Consequently, it is imperative, in community studies, to include a high percentage of circulating isolates in order to accurately determine the rates of recent and reactivated disease.
Epidemiological impact of subpopulations on tuberculosis transmission
The impact of immigrant subpopulations on the epidemiology of tuberculosis in the population of a community has been perceived as an important public health issue in developed countries. For example, during 1986–1997, the number of tuberculosis cases diagnosed in foreign-born persons in the USA increased by 56%. These statistics suggest that immigrants could be transmitting tuberculosis to the native population. However, in one study, only one of 43 cases amongst immigrants resulted in two secondary cases of tuberculosis infection in US-born cases. Additionally one-fifth of Mexican-born patients acquired their tuberculosis infection in San Francisco 69. This study was particularly important because the native population were proved to have transmitted tuberculosis at a higher rate to immigrants than immigrants to natives. A study from the same city described two parallel epidemiological patterns of tuberculosis in foreign-born and US-born populations 70. Most foreign-born individuals develop tuberculosis from reactivation, whereas 20% of US-born cases developed tuberculosis from recent infection. In other settings, different patterns of transmission have been described within immigrant populations. Tuberculosis transmission has been documented from immigrants in Denmark 71, 72. In the Netherlands, almost half of the recent transmission occurs in immigrant groups, but most of the transmission occurs within the same nationality 73.
The transmission index, defined as the mean number of tuberculosis cases resulting from recent transmission of a potential source case, has been used to quantify transmission between different subpopulations 73. In San Francisco, the transmission index was found to be lower among foreign-born than US-born patients and was much higher among Black American patients of <35 yrs 74. These studies show that defining the nature of transmission between different population groups can be used to inform and strengthen tuberculosis control.
Epidemiological suspected and unsuspected transmission
In addition to the studies described above, molecular epidemiology has contributed to improving disease control in other ways. For example one study demonstrated the explosive potential for tuberculosis to progress to disease and spread amongst HIV-infected persons 75. In this study, conventional surveillance detected 12 cases of tuberculosis in a residential facility for persons with HIV disease. Analysis of isolates by IS6110 RFLP demonstrated that newly acquired tuberculous infection in HIV-infected patients spread readily and progressed within 3 months of exposure to disease, demonstrating the particular vulnerability of HIV-infected individuals to exogenous tuberculosis infection.
Molecular epidemiology has also documented the potential for spread of drug-resistant strains among hospitalised patients 14, 76–79. During one 43-month period, New York City accounted for almost one-quarter of all cases of MDR-TB in the USA. Most of these patients were infected with HIV and were found to have acquired their often-fatal MDR-TB whilst in hospital 80. The results of this and other similar studies have led to more rigorous adherence to infection control policies, particularly in settings in which there are many HIV-infected persons 81. Increased surveillance with prompt diagnosis and appropriate therapy in settings such as hospitals, prisons, schools and homeless shelters is now resulting in an overall decrease in tuberculosis transmission 81.
Population-based molecular typing studies have also shown the dramatic impact that an individual or a small group of individuals can have on tuberculosis transmission. In Minneapolis (MN, USA), during 1992, one individual was shown to have caused 35% of all new active cases of tuberculosis 80. One of the first large-scale molecular epidemiological studies uncovered extensive transmission of M. tuberculosis among a small group of substance abusers with significant “spillover” to the general population 82. Knowledge of the negative impact that poorly managed patients can have in a community emphasises the need for a sustained level of tuberculosis control for each and every disease case.
Active case finding, through the evaluation of individuals who have been in contact with infectious tuberculosis patients, is a traditional activity of tuberculosis control programmes in industrialised countries. A basic principle of this policy is that contacts are likely to have been infected by the infectious case and thus carry the same strain (with the same drug susceptibility pattern) as the index case. This hypothesis was tested in a study of index and contact cases 83. The authors compared the DNA fingerprints of pairs of indexes and contacts who were both ultimately diagnosed with tuberculosis. Thirty per cent (16 of 54) of pairs had different fingerprints, demonstrating that the contact had been infected from an unidentified third person. This illustrates that transmission links are often more complex than those assumed by conventional epidemiology. Similarly, a contact investigation among five large clusters in the Netherlands showed transmission occurred after only transient contact, contrary to the conventional view that tuberculosis is usually acquired following prolonged exposure to an infectious case 84. This complexity is also seen in population-based molecular studies. For example it is often difficult to establish an epidemiological link between individual cases in an RFLP cluster. Fingerprinting and contact investigation in different settings demonstrated epidemiological links in only 5–10% of cases 58–60, 84.
Identical fingerprints do not always correlate with an established epidemiological link. For example strains from a rural area in which patients were geographically dispersed and highly unlikely to have had previous contact were found to have identical fingerprints 33. Also, strains from cases in different states in the USA were found to be identical, although they had no history of previously having been in contact 26. Similarly, in South Africa, Tanzania, Zimbabwe, Kenya and Malawi, strains from different dispersed geographical regions were found to have identical patterns 85. These studies show that some strains are more prevalent and abundant than others and this could lead to the misattribution of epidemiological links. Therefore, caution should be exercised in reaching conclusions based solely on identical fingerprint results.
The interpretation of molecular epidemiological studies should largely depend on the study question, the area under study and the typing methods used. A combination of typing methods based on more rapid and slower molecular clocks should, in principle, be able to differentiate between the contributions of remote and recent transmission to clustering. Typing methods with a reliably slow molecular clock could be very useful in global tracking and evolutionary studies of tuberculosis. Thus, the future challenge is to identify techniques with different molecular clocks that can identify, with certainty, recent and remote evolutionary linkage between strains.
Quantification of the level of infectiousness among smear-negative patients
It is generally believed that patients with tuberculosis whose sputum microscopic examination fails to detect acid-fast bacilli (AFB) are significantly less infectious than those with positive smears. However, a molecular epidemiological study that compared transmission from AFB smear-positive and -negative patients suggested that AFB smear-negative patients were responsible for ≥21% of tuberculosis transmitted in the city of San Francisco 86. Thus intensifying tuberculosis control measures for smear-negative cases could significantly reduce the transmission of tuberculosis.
False-positive Mycobacterium tuberculosis cultures
There are now a plethora of studies describing the problem of false-positive M. tuberculosis cultures in the laboratory with the use of multiple markers 87–94. Laboratory cross-contamination represents a significant problem for the microbiologist and may result in unnecessary treatment and potential drug toxicity for a patient. For instance, a small but significant proportion (3%) of New York City patients had falsely positive cultures for M. tuberculosis as a result of contamination 79. Timely molecular analysis and appropriate changes to specimen processing have been identified as useful measures for avoiding false-positive cultures 95, 96.
Current topics of interest
Epidemiological prevalence of the Beijing genotype
A distinct family of M. tuberculosis strains (subsequently labelled the “W strain”) was associated with >350 cases in New York City and, at one point, accounted for 25% of all MDR-TB cases in the USA 97. These strains were later demonstrated to belong to a branch of a distinct family of strains named the “Beijing genotype” because of their predominance in the Beijing area of China. Although the IS6110 RFLP pattern of the Beijing strains is not unique and is often difficult to distinguish from other genotypes, the spoligotype is highly distinctive and easy to identify. With the use of spoligotyping, the Beijing strains have been associated with transmission in Azerbaijan, Thailand, Estonia, Iran, Vietnam, Malaysia, Estonia, China, Hong Kong, South Africa, Colombia and Russia 48, 98–103. Its presence has been detected throughout Southeast Asia and in Hong Kong, and, in one area of study, the Beijing genotype family accounted for 70% of all isolates 104.
On the island of Gran Canaria, this genotype was initially introduced by an immigrant from Africa. Over a period of 4 yrs after its introduction, this genotype became the most common isolate on the island 105. In some areas of the world, this strain has been associated with cases of tuberculosis in young individuals 48, 106. The association of this genotype with a younger age group is recognised as an indication of ongoing transmission. In addition, studies have found a significant correlation between the Beijing genotype and drug resistance 97, 103, 107. The global dissemination and apparent transmissibility of this strain has raised the tantalising possibility that these epidemiological characteristics are a reflection of an intrinsic biological property unique to this family. It has also been postulated that, if these strains can indeed replicate more efficiently in the host, this could favour the development of resistance 105. Intensive research efforts are now being directed towards elucidating the genetic basis of this apparent phenotype. However, it remains to be seen whether or not this family of strains is more virulent and what its role may be in the worldwide tuberculosis epidemic.
Epidemiological evidence of exogenous reinfection
It was not until molecular fingerprinting techniques became available that the exogenous reinfection phenomenon was conclusively demonstrated to occur. An evaluation of HIV-infected patients in a New York City hospital, who repeatedly yielded positive cultures for M. tuberculosis, identified 11 patients with sequential isolates that became resistant to antimicrobial agents. In four of these patients, the RFLP patterns of the isolates changed dramatically at the time that drug resistance was detected. In these patients, the clinical and microbiological evidence was consistent with the presence of active tuberculosis caused by a new strain of M. tuberculosis 107. Exogenous reinfection of persons whose only immunosuppressing condition was diabetes has also been demonstrated 108.
A small study in Africa demonstrated the importance of considering reinfection in patients with a past history of tuberculosis. Original and recurrent isolates of tuberculosis were analysed in five patients in this study who were known to be HIV-positive with recurrence of tuberculosis 109. Reinfection was demonstrated in one patient whose original and recurrent isolates had dramatically different fingerprints. This study demonstrates that, when tuberculosis recurs after standard treatment in HIV patients, reinfection with a new isolate and not just relapse should be considered.
In high-incidence settings, particularly those with high rates of HIV coinfection, it is anticipated that re-infection is important. This was the case in a study in South Africa in which up to 75% (12 of 16) of patients had different fingerprints in their initial and second episodes of disease 110. However, in another study carried out in South African miners, only 2% (1 of 48) of cases were found to be due to reinfection 109. Other studies in Hong Kong 8 and India 111 reported rates of reinfection ranging 12–31%. In these studies, different methodologies and possible contamination could have accounted for the discrepancies in the results. Thus, it is difficult to draw definite conclusions as to the rate of reinfection in high-incidence countries based on these studies.
In a study on the island of Gran Canaria, Spain, in a setting with good tuberculosis control measures and in conjunction with meticulous methodology, 44% (8 of 18) of cases were attributed to reinfection. Thus, even in communities with a low incidence of tuberculosis, reinfection appears to be important 112.
Accurately establishing the rates of reinfection in different settings is important in predicting the effects of control strategies, such as directly observed treatment, short course (DOTS), on the course of the current tuberculosis epidemic. For example, if DOTS rapidly shuts off transmission and reinfection is rare, the epidemic will die off slowly because cases will continue to appear through reactivation. Conversely, if reactivation is not common, the epidemic will decline relatively quickly as DOTS prevents new cases by stopping reinfection. Further, determining the rates of reinfection versus reactivation in areas with high levels of HIV coinfection is of critical importance for designing the most appropriate chemoprophylactic strategies. For example, in an area with high levels of exogenous reinfection, short-term chemoprophylaxis is unlikely to be of value.
The relative fitness of isoniazid-resistant strains
The development and transmission of MDR-TB, defined as resistance to at least isoniazid and rifampicin, is of particular concern because it requires prolonged costly therapy that often produces only low cure rates 113, 114. In spite of the implications of the spread of drug-resistant tuberculosis, little is known about the ability of drug-resistant strains to transmit, survive and reproduce in a population as compared to susceptible strains 115.
Experimental models suggest that antibiotic resistance imposes a biological cost on bacterial fitness 116, 117. It is thought that the mechanism of resistance, for example to isoniazid (mutations in the katG gene), causes an intrinsic cost to the fitness of the microorganism to be incurred 115. This is supported by in vitro studies that show that katG is required for optimal survival of M. tuberculosis in animal models 118.
Additional evidence that resistant strains are less fit comes from molecular epidemiological studies carried out in the Netherlands, Mexico and South Africa, in which less clustering was observed among drug-resistant strains 84, 109, 119. In a more in-depth analysis of clustering in San Francisco, the case reproduction number of circulating Mycobacterium tuberculosis strains resistant to isoniazid was found to be significantly reduced when compared to the case reproduction number of susceptible circulating phenotypes (data not shown). These molecular epidemiological studies suggest that development of isoniazid resistance incurs a significant biological cost for strains of M. tuberculosis. This decreased fitness could diminish the epidemiological impact of isoniazid-resistant tuberculosis and MDR-TB phenotypes in a population. However, it remains to be seen whether the magnitude of the fitness cost can overcome the many factors that collude to promote the transmission of drug-resistant tuberculosis in communities with a high burden of tuberculosis.
Conclusion
The development of molecular tools has added a new dimension to the classical epidemiology of tuberculosis and greatly enhanced understanding of the complex transmission dynamics within populations and between hosts. In the process, molecular epidemiology has demonstrated inadequacies in tuberculosis control programmes and helped accumulate motivation and resources for their improvement. Other technologies based on knowledge of the complete genome sequence of M. tuberculosis, which will provide newer tools for probing the epidemiology of tuberculosis, are now emerging. In spite of recent research advances, tuberculosis continues to remain a devastating infectious disease, disproportionately impacting on the world's poorest countries. The future challenge for molecular epidemiology is to provide better understanding of the transmisson dynamics of tuberculosis in those countries with the greatest burden of disease, and to stimulate an urgency to improving control measures on a more global scale.
- Received February 14, 2002.
- Accepted March 13, 2002.
- drug resistance
- fitness
- genomics
- molecular epidemiology
- Mycobacterium tuberculosis
- restriction fragment length polymorphism
- © ERS Journals Ltd