Chapter 9: Molecular Epidemiology: Breakthrough Achievements and Future Prospects
by Dick van Soolingen, Kristin Kremer, and Peter W.M. Hermans
Our understanding of the transmission of tuberculosis (TB) has been greatly enhanced since the introduction of desoxyribonucleic acid (DNA) fingerprinting techniques for Mycobacterium tuberculosis in the early ’90s. Historical enigmas have been solved in the last decade and classical dogmas are being evaluated. This review summarizes the most important and recent findings in the molecular epidemiology of TB and discusses essential knowledge still lacking. Furthermore, current developments in the introduction of typing techniques are described, as well as future challenges to improve the usefulness of molecular markers in the epidemiology of TB. Because the number of publications on the molecular epidemiology of TB has become too large to summarize in detail in a single review, only relatively new findings and subjects currently in the centre of attention are reviewed.
In the ’90s, a wide variety of genetic markers for M. tuberculosis were identified (Kremer 1999). However, only a minor number of these appeared to offer enough discrimination and reproducibility for wide scale implementation (Table 9-1) (Kremer 1999, Kremer 2005a). In 1993, IS6110 Restriction Fragment Length Polymorphism (RFLP) typing was adopted as the standard method for routine typing of M. tuberculosis (van Embden 1993). In this method, chromosomal DNA is digested with restriction enzyme PvuII. The digested DNA is separated on an agarose gel and, after Southern Blotting, hybridized with a DNA probe. This DNA probe is directed to the IS6110 insertion sequence and labelled with peroxidase, enabling enhanced chemiluminesence (ECL) detection of IS6110-containing restriction fragments (van Soolingen 1994). Another typing method, ‘spoligotyping’ has been used extensively as a secondary typing method (Bauer 1999, Kamerbeek 1997, Kwara 2003) and as a marker to study the phylogeny of the M. tuberculosis complex (Filliol 2002, Filliol 2003, Goyal 1997, Smith 2003). Spoligotyping exploits the polymorphism in the direct-repeat region of M. tuberculosis complex strains. This region consists of direct repeats interspersed with unique spacer sequences, and is amplified by Polymerase Chain Reaction (PCR) with primers directed to the repeats. The PCR-product is subsequently hybridized to known spacer sequences which are immobilized on a membrane through reversed-line blotting. Because one of the primers, and hence the PCR product, is labelled with biotin, ECL detection is achieved after incubation with peroxidase-labelled streptavidin (Kamerbeek 1997). Another DNA typing method frequently used for M. tuberculosis is Variable Numbers of Tandem Repeats (VNTR) typing. Typing results of this method are expressed as numerical codes. Each number of the code represents the number of tandem repeats at a particular repeat locus. The number of repeats varies by strain and is determined through PCR amplification of the repeat locus with primers directed to the regions flanking that repeat locus and determination of the PCR-product size. After an extended period of improvement and validation, VNTR typing is now ready to become the next gold standard for typing of M. tuberculosis complex isolates (Supply 2006).
Download of the entire textbook
(Tuberculosis 2007, 687 pages, PDF, 8.3 MB)
Table 9-1: Reproducibility and number of types obtained by using various DNA typing methods for differentiation of 90 M. tuberculosis complex strains and 10 non-M. tuberculosis complex mycobacterial strains (Kremer 1999, Kremer 2005a) DNA target Method used a Reference Repro-ducibility (%)b No. of types ob-tained IS6110 RFLP (PvuII) (van Soolingen 1994) 100 84 IS6110 Mixed-Linker PCR (Haas 1993) 100 81 IS6110 FLiP (Reisig 2005) 97 81 IS6110 IS6110 inverse PCR (Otal 1997) 6 nd c IS6110 LM-PCR (Prod’hom 1997) 81 73 IS6110/MPTR IS6110 ampliprinting (Plikaytis 1993) 39 nd IS6110/PGRS DRE-PCR (Friedman 1995) 58 63 15 loci VNTR typing (Supply 2006) nd 89 12 MIRUs VNTR typing (Supply 2001) 100 78 ETRs A-E VNTR typing (Frothingham 1998) 97 56 5 QUBsd VNTR typing (Roring 2004) 87 82 DR locus Spoligotyping (Kamerbeek 1997) 94 61 DR locus 2nd gen. spoligotyping (van der Zanden 2002) 90 61 DR locus RFLP (AluI) (van Soolingen 1993) 100 48 PGRS RFLP (AluI) (van Soolingen 1993) 100 70 (GTG)5 RFLP (HinfI) (Wiid 1994) 94 30 Total genome APPCR (Palittapongarnpim 1993) 71 71 4 conserved loci Amadio PCR (Amadio 2005) 74 13 EcoRI/MseI sites FAFLP typing (Ahmed 2003) 7 nd EcoRI/MseI sites FAFLP typing (Sims 2002) 0 nd BamHI/PstI sites FAFLP typing (Kremer 2005a) 0 nd a RFLP; Restriction Fragment Length Polymorphism, FLiP; Fast Ligation Mediated PCR, LM-PCR; Ligation-Mediated PCR, DRE-PCR; Double Repetitive Element PCR, VNTR; Variable Numbers of Tandem Repeats, APPCR; Arbitrarily Primed PCR, FAFLP; Fluorescent Amplified Fragment Length Polymorphism. b Fraction of duplicates showing identical types (31) c nd, not done d Results indicated exclude QUB locus 3232 The disclosure of suitable genetic markers to study the epidemiology of infectious diseases in the last decades has led to the widespread use of a new phrase; ‘mo-lecular epidemiology’. In fact, as pointed out by Foxman (Foxman 2001), this phrase is used in many articles on DNA fingerprinting (strain typing) of bacterial isolates, regardless of the inclusion of epidemiological data. Often, the availability of bacterial isolates dictates the design of the study, and not a fundamental, relevant epidemiological question in a given area. In many published studies, microbiolo-gists with an interest in molecular techniques were the main driving forces behind the described research. This was understandable in the initial stage of the imple-mentation of molecular typing techniques, when the main emphasis was on the evaluation of genetic markers. However, now that the value of genetic markers for M. tuberculosis has become clear, it is important to involve researchers of different disciplines in the design of any molecular epidemiological study, in order to ensure the validity of the research question, the sample size, the selection of cases and the interpretation of the results. 9.2. Historical context DNA fingerprinting of M. tuberculosis has been applied since the early ’90s to study transmission of TB at various scales. The first report on the use of IS986 RFLP to examine transmission of TB was published in September 1990 (Hermans 1990, McAdam 1990). Nine isolates with identical fingerprint patterns all origi-nated from an outbreak of TB among individuals who were all treated by the same physician, specialized in the treatment of arthritis patients. This finding led to the understanding that DNA polymorphism even in the genetically conserved M. tu-berculosis complex isolates could be applied as a strain-specific marker. In the years thereafter, the disclosure of many other genetic makers for M. tuberculosis complex would follow. Many investigators have tried to evaluate the reliability of strain typing by com-paring the clustering of M. tuberculosis isolates based on DNA fingerprints with the findings on the respective TB patients in contact tracing. However, this was highly cumbersome, as contact tracing by interviews in itself is not at all capable of finding even a quarter of the epidemiological links between sources and follow-up cases. Thus, contact tracing cannot serve as a gold standard to evaluate DNA fin-gerprint results. In contrast, DNA fingerprinting seems to be a much more sensitive tool to visualize epidemiological links between cases than conventional contact tracing. In the beginning, strain typing was mainly used to study outbreaks of TB and in-stitutional transmission. Soon thereafter, in multiple population-based studies, the rate of recent transmission and risk factors for transmission were determined (Diel 2002, Small 1994, van Soolingen 1999). Active transmission of TB in low-prevalence settings appeared to be associated for a large part to particular risk groups such as drug abusers, homeless people, and certain immigrant groups (Diel 2002, Small 1994, van Soolingen 1999). Transmission of drug resistant bacteria could be compared to that of drug-susceptible strains (van Doorn 2006, van Sool-ingen 2000). These findings are discussed in Section 9.5. With DNA fingerprinting, laboratory cross-contaminations were identified to occur at a considerable rate of 3-5 % of the positive cultures in low-prevalence settings, even though less than 10 % of the inoculated cultures were found positive in these areas (de Boer 2002, Small 1993). It is still not clear what the magnitude of this problem is in high-throughput laboratories in high-prevalence settings. Also, noso-comial infections by bacille Calmette-Guérin (BCG) have been disclosed by DNA fingerprinting and this contrasts the previous assumption that all M. bovis BCG infections are (late) complications of vaccination (Vos 2003b, Vos 2003a). Che-motherapeutics for the treatment of cancer patients were prepared in the same, non-disinfected biosafety cabinets that were used earlier to prepare BCG suspensions to treat bladder carcinoma patients. In this way, BCG bacteria were directly inocu-lated into cancer patients, in some cases with dramatic consequences. More recently, hypotheses on the infectiousness of individual patients have also been tested (see below). Another important finding in molecular epidemiology is that exogenous re-infections after curative treatment play a much larger role than previously anticipated (Das 1995, Sonnenberg 2001, van Rie 1999a). In the light of the description of exogenous re-infections it is interesting to read the recent obser-vations on the detection of mixed infections (see Section 9.7). Can a part of the exogenous re-infections be explained by the initial presence of more than one strain in diagnosed TB patients? Although M. tuberculosis may be one of the most widespread infectious agents in humans, not much is known about the evolution of this bacterium and whether there is an ongoing selection towards better adapted strains under the pressure of the measures introduced against TB in the last century. Because of the introduction of genetic markers for M. tuberculosis, the phylogeny of this bacterium can be studied in detail and the changes in the population structure can be disclosed. This has led to the recognition of a wide variety of genotype families worldwide (Bhanu 2002, Douglas 2003, Kremer 1999, Niobe-Eyangoh 2004, van Soolingen 1995, Victor 2004). In particular, the international database of spoligotyping patterns has been used most extensively for this purpose (Brudey 2006, Filliol 2002, Filliol 2003, Sola 2001). Although it has become clear that the phylogeny of M. tuberculosis differs signifi-cantly in several geographic areas, not much is known about the dynamics of the population structure and the reasons for the genetic conservation observed among M. tuberculosis isolates in high-prevalence areas. If particular genotypes of M. tuberculosis are selected, how fast does a shift towards more adapted variants oc-cur? Are we influencing the spread of particular genotypes of M. tuberculosis? Best studied in this respect is the Beijing genotype family of M. tuberculosis. There are indications that there is indeed a dramatic and relatively fast change in the compo-sition of the worldwide population of M. tuberculosis (see Section 9.6). If the cur-rent observations hold true, we may be facing a recurrent TB epidemic caused by bacteria with a higher level of evolutionary development. However, more research is needed to draw better conclusions. 9.3. Infectiousness of tuberculosis patients In most low-incidence settings, the majority of TB transmissions are limited to one or two persons. However, especially in high risk groups, such as the homeless and drug abusers in urbanized areas, ongoing transmission may take place for years and DNA fingerprint clusters sometimes grow over a hundred cases (unpublished ob-servations in the Netherlands). In these clusters, primary, secondary, and tertiary sources can usually not be distinguished. This makes it difficult to know how many cases are derived from individual sources. DNA fingerprinting, however, has dis-closed new information on the infectiousness of individual patients. For instance in San Francisco, 6 % of the TB cases in a two-year period seemed to have derived from a single source (Small 1994). In the Netherlands, a large outbreak in the small city of Harlingen was traced back to a single case diagnosed with a large doctor’s delay (Kiers 1996, Kiers 1997). It is only partly known what determines the transmissibility of TB. It is known that large patient- and/or doctor-originated delays play a significant role in the magni-tude of transmission. Furthermore, a more extensive pulmonary process and a bad coughing hygiene clearly contribute to disease transmission. However, the bacte-riological factor has not yet been established very well. It is, for instance, still not clear whether M. tuberculosis strains associated with large clusters on the basis of DNA fingerprinting are transmitted more easily than non-clustered strains. Is large-scale transmission only facilitated by risk factors, or do the bacterium’s character-istics also contribute to a more efficient transmission and breakdown to disease? Although there is a correlation between the smear status of a source case and the rate of transmission, smear-negative patients can also transmit TB. In San Fran-cisco smear-negative, but culture-positive cases were found to be responsible for 17 % of the cases (Behr 1999). This indicates that smear-negative pulmonary TB suspects should be considered infectious. 9.4. DNA fingerprinting, contact investigation and source case finding Case finding and treatment are the most important measures to inhibit the spread of TB in a community. In low-prevalence settings, where contact tracing has been routinely used for decades, a lot is known on how transmission of TB takes place. Prolonged exposure to an infectious source enhances the chance of transmission. Hence, direct and close contact with a TB patient is a main cause of infection in low-prevalence settings. However, in high-prevalence areas the transmission routes are less clear. What is the chance of acquiring an infection from an intimate contact in comparison to the chance of contracting TB from a casual contact in an envi-ronment with a high risk of infection? A recent study in South Africa (Verver 2004) pointed out that only 46 % of 313 TB patients had a matching fingerprint with an isolate of another member of the household they were living in. The pro-portion of transmission in the community that took place in the household was found to be only 19 %. This suggests that in this area, and presumably also in other high-incidence settings, TB transmission mainly occurs outside the household. In settings in Western countries where the incidence of TB has become very low, the role of contact investigation remains highly important. In each area, the risk factors for the transmission of TB may differ. Factors such as being homeless, a drug abuser, living in urban areas, and low age have commonly been found to in-crease the risk of transmission (Borgdorff 1999, Borgdorff 2001, Diel 2002, Small 1994, van Soolingen 1999). Usually, contact investigation is performed on the basis of the stone-in-the-pond principle and uses the Mantoux skin test (Veen 1990, Veen 1992) as an indicator of infection. Depending on the number of contacts found positive in the first ring of close contacts, the contact investigation is extended to the next ring of less intimate contacts. If again the ratio of positive contacts in that ring is high, the number is extended to the next circle of contacts. In many molecular epidemiological studies, it has been found that only a minority of the epidemiological links between TB cases disclosed by DNA fingerprinting, are also found by conventional contact tracing on the basis of interviews (Diel 2002, Lambregts-van-Weezenbeek 2003, Sebek 2000, Small 1994, van Deutekom 2004). This suggests that a large part of the TB transmission takes place through casual contacts in public places, such as bars, discothèques, public transportation, or other crowded settings. These contacts will generally not be found by interviews. Furthermore, in low incidence areas, where the skills of physicians to recognize TB adequately are waning, sources of transmission often spread the disease for extended periods and typing of isolated bacteria can help to find the source of an outbreak. In the Netherlands, nationwide DNA fingerprinting of M. tuberculosis has sup-ported contact investigations since 1993 (Lambregts-van-Weezenbeek 2003, Sebek 2000, van Soolingen 1999). All M. tuberculosis cultures are subjected to standard-ized IS6110 RFLP typing, and clustered cases are systematically reported to the regional TB services involved (cluster feedback). In an evaluation of six years of routine DNA fingerprint surveillance, it was found that among 2,206 clustered cases, 462 (21 %) of the epidemiological links between patients were expected on the basis of contact tracing information. After cluster feedback, an additional 540 (24 %) epidemiological links were established. Epidemiological links based on documented exposure increased by 35 % (Lambregts-van-Weezenbeek 2003) (Figure 9-1). Routine molecular typing also appears highly useful for evaluating the performance of TB control in a given area. In the Netherlands, each regional TB service quar-terly receives an overview of the growth of the active-transmission clusters of pa-tients to visualize in which populations ongoing transmission occurs and at what rate. In this way, municipal health services are able to deduce how much active transmission is ongoing in their region. Sometimes this leads to new measures, such as active screening of particular risk groups. Figure 9-1: Epidemiological linkage at diagnosis and after cluster-feedback, the Netherlands 1994-2004. The bars indicate the percentage of cases with a certain level of epidemiological linkage. Source: KNCV/RIVM DNA fingerprint surveillance project. One of the significant disadvantages of IS6110 RFLP typing is that it requires ex-tended culture incubation periods to obtain sufficient quantities of DNA. In the Netherlands, the typing results become available for contact tracing, on average, two months after the diagnosis of TB in a patient. At that time point, the contact investigation has usually already been finalized and not many TB services decide at that stage to re-open the contact investigations, even if the typing results provide new clues. However, the DNA fingerprint analysis clearly helps to evaluate the contact tracing process, and has therefore become an indispensable tool in TB con-trol in the Netherlands. It is expected that the yield of molecular typing in resolving epidemiological links between patients will sharply increase when faster finger-printing methods are implemented in the near future. In any case, nationwide mo-lecular epidemiological analysis contributes significantly to the evaluation of con-tact tracing and the performance of a TB control program. It clearly indicates the rate of recent transmission and to what extent, and in which populations and areas it occurs. Figure 9-2, available at http://www.tuberculosistextbook.com/pdf/Figure 9-2.pdf, summarizes the surveillance of active transmission of TB in the Nether-lands, 1997-2005. 9.5. Transmission of drug resistant tuberculosis In a recent paper by Zignol et al. (Zignol 2006), the global incidence of multidrug-resistant TB (MDR-TB) was described. The estimates of the World Health Organi-zation (WHO) on the global rate of MDR-TB have been updated from 272,906 MDR-TB cases in the year 2000 to 424,203 in the year 2004 because of the inclu-sion of countries that had previously not been surveyed. Zignol et al. underline the importance of expanding appropriate diagnostic and treatment services for MDR-TB patients, especially in countries with the highest burden of MDR-TB such as China, India, and the Russian Federation. Recently, the WHO also expressed its concern about the occurrence of extensively drug resistant (XDR) strains; M. tu-berculosis isolates resistant to at least isoniazid (INH), rifampicin (RIF), to one of the fluoroquinolones, and to one of the injectable anti-tuberculosis drugs (Anonymous 2006). These alarming observations trigger the question; are resistant strains as transmissible as susceptible ones? In as early as the ’50s, Mitchison observed that a large part of the INH resistant M. tuberculosis isolates revealed a lower degree of virulence in a guinea pig model (Mitchison 1954). For decades, it remained unclear whether resistant strains caused less transmission of TB than susceptible ones. This is important with respect to hygienic measures to prevent transmission from patients infected by MDR strains. Furthermore, for models predicting the development of the future TB epidemic, it is important to know if and how resistance interferes with transmission of TB. If resistant strains would be able to spread as efficiently as, or even better than sus-ceptible ones, the global rates of anti-tuberculosis drug resistance would rise stead-ily. Indeed, transmission of highly resistant strains has been reported in, for exam-ple, New York (Bifani 1996) and South Africa (van Rie 1999b, Gandhi 2006). However, observations of transmissibility of particular (multidrug) resistant strains should not be generalized to resistance in general. In a review by Cohen et al., describing the effect of drug resistance on the fitness of M. tuberculosis, it was concluded that the fitness estimates of drug-resistant M. tuberculosis strains are quite heterogeneous and that this confusion makes it difficult to predict the influ-ence of resistance on the trend of the TB epidemic (Cohen 2003). Indeed, various bacterial characteristics may influence the interference of resistance in transmissi-bility, including the drug susceptibility profile, the combination of mutations un-derlying drug resistance, presumably the genotype family the M. tuberculosis bac-teria represent, and possibly bacterial DNA repair mechanisms. In addition, non-bacterial factors may influence the interference of resistance and transmissibility, such as the immune status of the humans exposed, and the treatment regimen ap-plied. Because the above-mentioned factors have not been studied much, no meaningful conclusions can be drawn on the influence of the development of re-sistance on the worldwide TB epidemic. Yet, because of the contribution of DNA fingerprinting studies, some pieces of the puzzle have been unravelled in the last decade (van Doorn 2006, van Soolingen 2000). In a recent study in the Netherlands, in which 8,332 patients from the period 1993-2002 were included, the drug susceptibility profiles and transmissibility of the respective isolates were studied with the aid of DNA fingerprinting (van Doorn 2006). In total, 592 isolates were resistant to INH, of which 323 carried a mutation at amino acid position 315 (?315) of the catalase-peroxidase gene (katG). The remaining INH resistant strains had other mechanisms underlying INH resistance. As predicted by Mitchison (Mitchison 1954), in general INH resistant strains were less transmissible (i.e. less frequently present in DNA fingerprint clusters) than susceptible ones. However, strains with the ?315 were as frequently part of active transmission as susceptible ones. Moreover, the INH resistant strains with the ?315 had a higher level of INH resistance and were associated with multidrug resistance (van Doorn 2006, van Soolingen 2000). This suggests that the type of genetic mu-tation underlying INH resistance is an important factor in the fitness of the bacte-rium. Thus, particular strains may be the cause of MDR-TB transmission in both high and low-incidence settings, even though INH resistant strains in general are less fit than susceptible ones. In South Africa, most of the childhood contacts of adults with MDR-TB were more likely to be infected from these than other (drug susceptible) TB sources (Schaaf 2000). It would be highly interesting to know the mutations underlying resistance in these cases. In the Netherlands, transmission of MDR-TB is usually limited to incidental single person-to-person transmission. However, in the period 2003/2004 a single MDR-TB case infected nine other persons, of which two developed active disease. The respective MDR-TB strain had a mutation at amino acid position 315 of katG and exceptional mutations underlying RIF resistance (unpublished observations). It is not clear whether this type of resistant variant influences the epidemiology of TB in low and high-incidence areas. Therefore, further, more detailed and representative investigations into the basis of resistance in combination with the behaviour of the bacterium are needed. 9.6. Resistance and the Beijing genotype Another important factor that may determine the transmissibility of resistant strains is the genetic background of the bacterium. Based on several genetic markers, vari-ous M. tuberculosis genotype families have been identified, such as the Beijing family (van Soolingen 1995), the Haarlem family (Kremer 1999), Family 11 (Victor 2004), the Manila family (Douglas 2003), the Delhi family (Bhanu 2002), the Cameroon family (Niobe-Eyangoh 2004), the Latin American Mediterranean (LAM) family, the Central Asian clade, and the East African Indian clade (Brudey 2006, Filliol 2002, Filliol 2003, Sola 2001). It is important to study genotypic and phenotypic characteristics of the genotype families that fuel the worldwide TB epidemic. Up until now, the Beijing genotype has been studied most extensively. The Beijing genotype was first described in 1995 (van Soolingen 1995), and strains belonging to this genotype family appeared to be genetically highly conserved, which suggests that the spread of these strains started relatively recently. Moreover, in several areas, Beijing genotype strains are more frequently isolated from young patients than from older patients (Anh 2000, Borgdorff 2003, Glynn 2006). If, in high incidence areas, active transmission of TB is associated with lower age of the patients, as it is in low incidence settings (van Soolingen 1999), this suggests that Beijing genotype strains are emerging. The fact that Beijing strains have more often been found recently where population-based molecular epidemiological studies have been ongoing for several years points in that direction (Borgdorff 2003, Glynn 2006). Furthermore, the Beijing strains are associated with drug re-sistance in some areas (Glynn 2002, Glynn 2006). Thus, strains of the Beijing fam-ily may have a genetic background that favours their transmission, despite their drug resistance.In 2006, a large worldwide survey was published on the spread of the Beijing genotype of M. tuberculosis and its association with drug resistance (Glynn 2006). In this study, which included 29,259 patients from 35 countries, the overall prevalence of Beijing strains was 9.9 %, and the proportion of TB due to the Beijing genotype ranged from 0 % to over 72.5 % per area. The Beijing geno-type was endemic in East Asia and parts of the USA. In Cuba, the former Soviet Union, Vietnam, South Africa, and in parts of Western Europe this genotype was epidemic and associated with drug resistance (Glynn 2006). Previously, in New York outbreaks of MDR-TB were also caused by one of the evolutionary branches of the Beijing genotype family; the W strains (Bifani 1996, Kurepina 1998). The W strains, however, are a relatively minor branch on the evolutionary tree of the Beijing genotype family. It is to be determined to what extent the worldwide prevalence of MDR-TB is in-fluenced by the success of particular genotype families of M. tuberculosis in abso-lute terms, such as the Beijing strains. It is at least striking that in many areas with a high rate of MDR-TB, the Beijing strains are also highly prevalent (Glynn 2006, Kruuner 2001, Pfyffer 2001, World Health Organization 2004, Zignol 2006). It has yet to be determined whether there is a causal correlation between these observa-tions. It remains unclear whether transmission of highly resistant strains in high incidence settings are exceptions to the rule that resistance in general costs fitness of the bac-terium, or that particular genotypes of M. tuberculosis have developed efficient ways to become resistant to anti-tuberculosis drugs and maintain or even increase their ability to spread in a community. In the latter case, these genotypes will spread in the coming years and will influence the development of the worldwide TB epidemic. 9.7. Genetic heterogeneity of M. tuberculosis and multiple in-fections When talking about multiple M. tuberculosis sub-populations in sputum of TB patients, two phenomena are often confused, although they should be clearly dis-tinguished: · multiple strain populations derived from a single ancestral strain displaying genetic drift · multiple infections by more than one strain. In the case of multiple (or mixed) infections, the presence of more than one M. tuberculosis strain is demonstrated on one occasion of culturing from clinical mate-rial. This should not be confused with re-infection, usually after curative treatment, as this refers to a new episode of the disease caused by another strain. In South Africa, where the prevalence of TB is very high, the contribution of re-infection to new episodes of TB after curative treatment is considerable, and has been estimated at 75 % (van Rie 1999a, Verver 2005). Numerous observations in the molecular epidemiology of TB have pointed out that bacteria are subject to evolutionary change. Sometimes minor rearrangements of IS6110 RFLP profiles are noticed in epidemiologically related- and serial patient isolates. The rate of change of IS6110 RFLP patterns in such isolates has been studied by several investigators (de Boer 1999, Niemann 1999, Niemann 2000, Yeh 1998). However, also within clinical M. tuberculosis isolates, sub-populations of bacteria with minor genomic differences co-exist (de Boer 2000, Shamputa 2004, Shamputa 2006). For example, low-intensity bands in IS6110 RFLP profiles are a reliable indication of a sub-population of bacteria with, for example, a one-band difference in IS6110 RFLP. Preparation of single colony cultures and subsequent IS6110 RFLP typing of isolates with such low-intensity bands showed the co-existence of separate sub-populations of bacteria, either with or without a normal-intensity band at the position where the low-intensity band occurred in the original clinical isolate (de Boer 2000). Several recent papers describe the finding of multiple M. tuberculosis populations in sputum specimens of TB patients (Richardson 2002, Shamputa 2004, Shamputa 2006, van Rie 2005, Warren 2004). These findings point out that multiple infection of M. tuberculosis may be more prevalent than previously assumed. In the study by Warren et al., a PCR technique was used to specifically identify M. tuberculosis bacteria of the Beijing genotype family and other evolutionary lineages in sputum specimens of patients from South Africa (Warren 2004). These authors concluded that at least 19 % of the patients included were infected by both Beijing and non-Beijing strains. Multiple infections were more frequently observed in re-treatment cases than in new cases. The same group also explored IS6110 RFLP typing to detect multiple strain infections; a minor part of the IS6110 RFLP patterns exhib-ited background patterns suggestive of mixed infections (Richardson 2002). This was confirmed in three (2.3 %) of the cases. In addition, another interesting ap-proach was followed to study the occurrence of multiple infections in TB patients; by investigating M. tuberculosis strain diversity in autopsy material in South Africa (Plessis, 2001). In two out of 12 patients, pulmonary infection by two strains was demonstrated. The question remains about how this relates to the practical bacteri-ology: if this study had been performed at the time of diagnosis of TB in these patients, would one o