INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Tuberculosis (TB) is globally a major cause of morbidity and mortality, with most cases occurring in developing countries (20, 27). According to the World Health Organization, one-third of the world's population is infected with organisms of the Mycobacterium tuberculosis complex, with about 10 million cases of active TB disease reported each year, leading to 3 million deaths annually (6). An increased morbidity and mortality in TB patients is associated with the ongoing human immunodeficiency virus (HIV) pandemic. Guinea-Bissau has one of the highest incidences of TB in the world, with the annual incidence estimated to be 150/100,000 population. The current epidemic of HIV type 2 (HIV-2) infections (21) and the now emerging HIV-1 epidemic in Guinea-Bissau may contribute to a still higher incidence of TB.

It has long been observed that West African M. tuberculosis strains are biochemically more heterogeneous than European strains. Prat and colleagues (23) speculate that the biochemical heterogeneity observed among African strains represents a continuous spectrum linking the classical human and bovine variants of M. tuberculosis. David and coworkers (4), on the other hand, found that M. africanum strains tended to form two clusters on the basis of biochemical traits: one related to the human type and one related to the bovine type of M. tuberculosis complex. In a previous epidemiological study (16) we subtyped 56 M. tuberculosis complex strains isolated from patients with pulmonary TB in Guinea-Bissau into five biovars. On the basis of our biochemical methods, we identified five biovars representing a spectrum of strains ranging from classical human M. tuberculosis strains to classical M. bovis strains. Two biovars corresponded to the so-called African I variant, and one biovar corresponded to the African II variant described by Collins et al. (2).

To study the population structure of M. tuberculosis complex isolates in Guinea-Bissau, we have investigated 229 M. tuberculosis isolates by the same biochemical methods used in our previous study, as well as by genetic molecular fingerprinting, using the insertion sequence IS6110 and the spacer regions within the direct repeat (DR) locus of M. tuberculosis as targets. The insertion sequence IS6110 belongs to the larger IS3 family of insertion elements and is regarded to be specific for the M. tuberculosis complex. The DR locus of the M. tuberculosis complex genome contains multiple, highly conserved 36-bp DRs separated by 35- to 41-bp variable spacer sequences (9). On the basis of the results presented here, we hypothesize that the majority of M. tuberculosis complex isolates belong to a unique family of strains that originated and evolved in West Africa.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Patients. Approximately 900 consecutive patients with suspected TB (and without known previous treatment for TB) were examined from 1989 to 1994 as part of an epidemiological study of TB in Guinea-Bissau (5, 15, 16, 19, 21). The ages and sexes of the patients were recorded, and during 1992 and 1993, testing for HIV-1 and HIV-2 was performed as described previously (21). Patients with positive cultures for M. tuberculosis complex (n = 229) were included in this study.

M. tuberculosis complex isolates. Sputum samples were collected from all patients by a registered nurse and were transported at 4°C to the laboratory. Acid-fast microscopy was performed after staining by the Ziehl-Neelsen method. Before culture, the sputum samples were decontaminated of nonmycobacterial microorganisms by the sodium lauryl sulfate method (10). A 0.5-ml aliquot of the homogenized specimen was then inoculated into both conventional Löwenstein-Jensen egg medium (LJ) and LJ supplemented with 0.6% pyruvate. The samples were incubated at 37°C and were examined weekly for 7 weeks. Growth of mycobacteria was confirmed by microscopic observation of acid-fast bacilli. In total, 229 strains of M. tuberculosis complex were isolated from these patients. Fifty-six of these isolates were also part of a previous study (16).

Characterization of isolates. All isolates were further characterized by macroscopic and microscopic appearance and growth characteristics and were biochemically characterized as described earlier (15, 18). Resistance to thiophene-2-carboxylic acid hydrazide (TCH; 5 mg/liter) was determined by radiometric respirometry (BACTEC system; Becton Dickinson, Sparks, Md.). Detection of niacin was performed as described by Wayne (33), and the nitrate reductase test and pyrazinamidase test were performed as described by Kent and Kubica (18). Testing for susceptibility to the drugs ethambutol, isoniazid, rifampin, and streptomycin was performed by using the BACTEC system (15, 25). All isolates were also tested with a nucleic acid probe specific for M. tuberculosis complex (Accuprobe system; Gen-Probe, San Diego, Calif.).

IS6110 RFLP analyses. A standardized method for restriction fragment length polymorphism (RFLP) analysis with IS6110 (IS6110 RFLP analysis) (29) was used. Genomic M. tuberculosis DNA was extracted and digested with the restriction endonuclease PvuII, and Southern blotting was performed after separation of the restriction fragments by electrophoresis. Hybridization was performed with a 245-bp PCR fragment of the IS6110 sequence as a probe. The probe was nonradioactively labelled with peroxidase and was subsequently visualized by the use of an enhanced chemiluminescence kit (Amersham International plc, Little Chalfont, United Kingdom).

Spoligotyping. Spoligotyping relied on the amplification of the polymorphic DR region (17) to obtain hybridizations patterns of the amplified DNA by using multiple synthetic spacer oligonucleotides which are covalently bound to a membrane (Isogen Bioscience BV, Maarssen, The Netherlands).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

M. tuberculosis complex isolates from 229 consecutive patients with bacteriologically verified pulmonary TB were characterized by biochemical tests, drug susceptibility testing, IS6110 RFLP analysis, and spoligotyping.

Patient characteristics. All except eight of the patients were native to Guinea-Bissau. The eight patients from other countries came from the neighboring West African countries of Guinea Conakry (five patients), Senegal (two patients), and The Gambia (one patient). The mean age of the patients was 39 years, with 58% being male (mean age, 41 years) and 42% being female (mean age, 37 years). Of the 229 patients, 154 were tested for HIV; 118 were HIV negative, 3 (1.9%) were HIV-1 positive, 29 (18.8%) were HIV-2 positive, and 4 (2.6%) were infected with both HIV-1 and HIV-2.

Characterization of isolates. All isolates had typical macro- and microscopic appearances upon acid-fast staining and were positive by the M. tuberculosis complex-specific nucleic acid probe assay. Fifteen isolates were resistant to any of four major antituberculous drugs (six were resistant to streptomycin, eight were resistant to isoniazid, and one was resistant to streptomycin and isoniazid) (5). By using four biochemical tests (resistance to TCH, niacin production, nitrate reductase test, and pyrazinamidase test) the M. tuberculosis complex isolates could be assigned to five different biovars (Table 1). The characteristics of 4 strains conformed fully with the biochemical criteria for M. bovis (sensitivity to TCH and negativity by the other tests), while those of 85 isolates agreed with the biochemical criteria for M. tuberculosis (resistance to TCH and positivity by the other tests). The remaining 140 isolates could be allocated to one of three biovars representing a spectrum between the classical human and bovine tubercle bacilli (Table 1).

Characteristics of IS6110 RFLP patterns. The 229 M. tuberculosis complex isolates were analyzed by an RFLP assay with the insertion sequence IS6110 as the target. All isolates had RFLP patterns that contained one or more bands that hybridized with the IS6110 probe (Fig. 1), with the number of IS6110 insertions ranging between 1 and 17 copies. The Dice coefficient of similarity for all pairwise comparisons of the isolates in the whole collection ranged from 0 to 100%; 147 (64%) of the isolates belonged to clusters of strains with identical and/or closely related banding patterns (defined as a coefficient of similarity of >90%), while 36% had less similar patterns. Altogether, 119 distinct banding patterns were found; 33 patterns were shared by two or more isolates, constituting 37 different clusters (defined as at least two isolates with identical RFLP patterns). The different clusters each comprised 2 to 33 isolates (Fig. 1). The largest cluster comprised isolates with a four-band pattern (Fig. 1). Twelve isolates, which showed only a single band, could be separated into two groups of four and eight isolates, respectively, on the basis of slightly different band migration patterns.

View larger version (56K): [in this window] [in a new window]   FIG. 1.   IS6110 banding patterns and similarity matrices for 229 M. tuberculosis complex isolates from Guinea-Bissau. Banding patterns are ordered by similarity. The corresponding dendrograms are to the left of the panels. The positions of the bands in each lane are adjusted (normalized) so that band positions for all strains are comparable. Scale depicts similarity coefficients (which are defined elsewhere [29]). Groups A to C are as defined in the Results.

Spoligotyping. Spoligotyping was performed with all 229 isolates. We identified 73 different spoligotype patterns, and these were used to classify the isolates into three main groups, designated sptA, sptB, and sptC (Table 2). The sptA group, consisting of 88 isolates, lacked spacer 33 to 36, with 38 of these isolates also lacking spacer 21 to 24. The largest group (117 isolates), sptB, typically lacked spacers 7 to 9 and 39, with 30% of these strains also lacking between 1 and 16 additional spacers. The smallest group, sptC, consisting of 15 isolates, typically lacked spacer 34, with either spacers 29 to 33 or spacers 29 to 32 also absent. In addition to the three baseline patterns described above, many individual isolates further lacked one or more spacers, forming clusters of strains with identical spoligotype patterns.

Correlation between IS6110 RFLP pattern and spoligotype. Although the fingerprints obtained by spoligotyping were less polymorphic than those obtained by probing with the IS6110 probe, there was an overall correlation between spoligotype and IS6110 RFLP pattern, even for strains with small numbers of IS6110 copies, confirming the genetic relatedness of the isolates identified by RFLP analysis.

Correlation between IS6110 copy number and biovar. When the RFLP patterns of the individual isolates were compared to the results of the biochemical tests, a significant correlation was observed between IS6110 copy number and biovar classification (Fig. 2). Thus, the group of classical human M. tuberculosis isolates (biovar 5) had a significantly (P < 0.01 by the Wilcoxon rank sum test) larger number of IS6110 copies (mean copy number, 9; median copy number, 10), than the isolates that belonged to the intermediary biovars, biovars 2 to 4 (mean copy number, 5; median copy number, 5). The four M. bovis isolates carried three (one isolate), four (two isolates), and five (one isolate) copies of IS6110 (data not shown).

View larger version (6K): [in this window] [in a new window]   FIG. 2.   Number of IS6110 copies present in isolates of five different biovars of the M. tuberculosis complex. Differences in numbers of IS6110 copies between biovar 5 (classical human M. tuberculosis) and biovars 2, 3, and 4 were significant (P < 0.01 by the Wilcoxon rank sum test).

Correlation between IS6110 RFLP pattern and biovar. With a few exceptions, the biovar 5 isolates had the group A RFLP pattern and formed a distinct group. A total of 78 of the 85 biovar 5 isolates were in group A, while only 14 of the 51 biovar 4 isolates, 2 of the biovar 3 isolates, and none of biovar 1 and 2 isolates were in group A (Table 2). Additionally, many of the biovar 5 isolates in group A were highly related; 59 biovar 5 isolates belonged to two families of isolates with highly related banding patterns (similarity coefficient, >65%). In comparison to biovar 5, most of the isolates in biovars 1 to 4 had the main group B RFLP pattern.

View larger version (64K): [in this window] [in a new window]   FIG. 3.   Schematic representation of spoligotypes of 32 M. tuberculosis isolates belonging to cluster B:5. The isolates are sorted in the same order in which they were sorted by their IS6110 RFLP patterns in Fig. 1. X, positive hybridization signals; dots, lack of hybridization.

M. bovis isolates. The four strains that were biochemically classified as M. bovis all belonged to group B. They had between three and five IS6110 insertions, but intriguingly, none of the insertions was located on the 1.9-kb PvuII restriction fragment that is typical for most M. bovis and M. bovis BCG strains. Interestingly, one IS6110 insertion location was common to all four M. bovis isolates: the IS6110 sequence was located on a fragment identical in size to the single band of cluster C:1. Additionally, all M. bovis isolates possessed spacers 40 to 43, which, together with spacer 39, have been reported to be lacking in M. bovis strains (17).

Correlation between type of strain and patient characteristics. There was no correlation between age, sex, or positivity for HIV with any particular type of strain. Of the eight isolates from patients from neighboring countries, one biovar 2 isolate from a patient from The Gambia and one biovar 3 isolate from a patient from Guinea Conakry clustered with the group B isolates. The six other isolates were of biovar 5 and clustered with the group A isolates.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

As in a previous study (16), we have found a high degree of biochemical heterogeneity within strains of the M. tuberculosis complex isolated from Guinea-Bissau. In order to characterize this heterogeneity further at a genetic level, we subtyped isolates by DNA fingerprinting using the insertion element IS6110 and spoligotyping. Using these techniques, we were able to classify M. tuberculosis complex isolates into three groups. Group A isolates contained large numbers of IS6110 elements and a basic spoligotype pattern typical of classical human M. tuberculosis strains. Group B mainly comprised the intermediate biovars and the classical M. bovis strains, which are recognizable by the presence of few IS6110 copies per genome and a distinctive spoligotype pattern. In contrast, group C comprised a few isolates that mainly possessed a single IS6110 element. These DNA fingerprint results conform with recent observations from a study of M. africanum strains from Sierra Leone and Uganda (12). This study demonstrated that strains which had biochemical traits resembling those of M. bovis (Africanum I type) had IS6110 fingerprints that exhibited fewer bands, resulting in large clusters of strains with identical banding patterns, whereas strains that were biochemically more closely related to M. tuberculosis (Africanum II subtype) had a large number of fragments, with many individual banding patterns being observed.

European strains of M. tuberculosis are usually described as classical, are distinguishable by a large number of copies of IS6110 elements, and frequently lack the spoligotyping spacers 33 to 36 (8, 17). Interestingly, isolates of group A from Guinea-Bissau resembled European M. tuberculosis isolates in terms of biochemical traits, RFLP patterns, and spoligotypes when they were compared with isolates in an international database of the IS6110 patterns of M. tuberculosis complex isolates and also with strains from South and Central America (28a). Guinea-Bissau is a small country whose population has a rather high degree of mobility within the country but little contact with the populations of neighboring African countries. Historically, it is a former colony of Portugal, which is a country with a high incidence of TB caused by the classical human M. tuberculosis type. Even today, contacts between the populations of Guinea-Bissau and Portugal are probably still more numerous than contacts between the populations of Guinea-Bissau and other African countries. We therefore speculate that the group A M. tuberculosis strains, to which the majority of the biovar 5 isolates (i.e., the classical human variants) belonged, may have their origin within Europe, in particular, Portugal. Indeed, some of the isolates of group A had RFLP patterns that were identical to those of isolates from TB patients in Portugal (22).

Isolates of group B had few copies of IS6110 (mean copy number, 5) and highly conserved RFLP patterns as well as a homogeneous basic spoligotype pattern, which typically lacked spacers 7 to 9 and 39. This spoligotype pattern forms a baseline for all group B isolates. These common features indicate that the group B isolates are highly related to each other. Despite the genotypic similarities among these strains, the group B isolates showed a high degree of variability in terms of biochemical traits. Most of the larger clusters (on the basis of the IS6110 patterns) within group B contained isolates of different biovars. For example, the largest cluster (cluster B:5) of 32 isolates, presenting a typical four-band pattern, comprised isolates of all biovars, even including one isolate of biovar 5. Thus, despite the genetic stability in terms of the IS6110 fingerprints and the DR locus, the isolates varied in several phenotypic traits, which therefore could be used to further subdivide the different clusters determined by the DNA fingerprinting methods. One could speculate that the diversity in biochemical traits among these isolates reflects the ongoing evolution of M. tuberculosis complex in Guinea-Bissau or West Africa from classical M. bovis-like variants to more classical M. tuberculosis variants. This would mean, however, that the evolution of these properties is more rapid than the genetic evolution of IS6110 and the DR locus. Alternatively, the strains of biovars 1 to 4 may more readily switch on and off genes for, for example, niacin production during the adaptation to hosts of several species (and hence broaden their host tropism). Since only certain combinations of biochemical traits were found, such on-off switches for genes do not appear to occur randomly but, rather, occur as a stepwise turning on and off of genes in a defined sequence. For example, 69 isolates were positive for the production of niacin and negative for the production of nitrate reductase, but no isolate produced nitrate reductase but did not produce niacin. Recently, the pyrazinamidase gene (pncA) was cloned and a single point mutation in the gene was found to be unique to M. bovis (24). We sequenced the pncA genes of five cluster B:5 isolates, each representing one the five different biovars. Interestingly, none of them contained the M. bovis-specific mutation (data not shown).

Virtually all group B isolates contained spacers 33 to 36. These spacers were derived from the sequence of the BCG genome (17), supporting the relatedness of the group B isolates to M. bovis. It is generally supposed that M. bovis strains are more virulent for cattle, while classical M. tuberculosis strains are thought to be more virulent for humans. In Guinea-Bissau there is a high incidence of tuberculosis among cattle and goats. So far we have analyzed only one strain from a goat, with this isolate being of biovar 4 and showing an RFLP pattern and spoligotype identical to those of the four human biovar 4 isolates.

The four isolates in this study that were phenotypically characterized as M. bovis clustered with the group B isolates in terms of IS6110 RFLP and spoligotype patterns. Hence, they all had three to five copies of the IS6110 fragment, they all lacked the spacers 7 to 9 and 39, and they all contained spacers 33 to 36. M. bovis strains typically have small numbers of copies of the IS6110 insertion sequence (31, 32), while human and cattle M. bovis isolates in general harbor only a single copy of IS6110 (3, 11, 28), and it is usually located in the same position in the genome, on a characteristic 1.9-kb PvuII restriction fragment. In a report on the IS6110 RFLP patterns of 24 human M. bovis strains from The Netherlands, all strains had less than six copies, and one-third of the strains had only a single copy (31). By contrast, isolates from other animals appear to carry more IS6110 copies. In a study by van Soolingen et al. (30), the majority of isolates from cattle harbored a single IS6110 element, whereas the vast majority (29 of 34) of M. bovis isolates from a wide variety of other animals contained multiple IS6110 elements. During a study of Swedish M. bovis isolates from deer, we found that all isolates contained seven copies of the IS6110 element, yielding a highly specific RFLP pattern (28). Additionally, a study from Spain (11) demonstrated that M. bovis strains isolated from 23 goats carried six to eight copies of IS6110, while most isolates from cattle carried only a single copy of this element. Interestingly, the four M. bovis isolates in this study all had the sptB spoligotype pattern typical of the group B strains, which means that they all possessed spacers 40 to 43, which together with spacer 39 are reported to be lacking in all M. bovis and M. bovis BCG strains (17).

Cave and coworkers (1) suggested that the IS6110 insertion sequence originally appeared as a single copy in a progenitor of the M. tuberculosis complex and was retained as a single copy in M. bovis and then replicated to produce the multiple copies present in M. tuberculosis. They also suggested that these multiple insertional events may account for some of the phenotypic differences between M. tuberculosis and M. bovis. Our finding of a correlation between the phenotypic variants of strains of the M. tuberculosis complex and the number of IS6110 insertion elements supports such a hypothesis. Fomukong and coworkers (7) analyzed selected M. tuberculosis complex strains from Malaysia, Tanzania, and Oman as well as M. bovis isolates and M. bovis BCG strains, all of which carried a single copy of the insertion sequence, and suggested that in these organisms the IS6110 element is defective in transposition and that the loss of transposability may have occurred at an early stage in the evolution of the M. tuberculosis complex. The progenitor would therefore have had a single copy of the insertion sequence; the multicopy M. tuberculosis strains would therefore have evolved by replicative transposition of the element. It may be speculated that the small number of copies of the IS6110 fragment in the group B and C strains in this study reflects a similar defect in transposition of IS6110. Hermans and coworkers (14), however, found that the insertion sequence sequence element of M. bovis BCG was not defective in transposition but, rather, that it was located in a hot-spot integration region in the DR locus.

It has been reported earlier that M. tuberculosis strains isolated from patients in countries with a high prevalence of TB exhibit less DNA polymorphism than strains from patients in countries that have a low prevalence of TB, such as The Netherlands (31). In countries with a high prevalence of TB, such as Ethiopia and Tunisia, the majority of circulating M. tuberculosis strains belong to a limited number of strain families (13) and supposedly descend from a few clones that expanded in the recent past. In the present study there was a high proportion of clustering of the isolates by IS6110 RFLP patterns, resulting in a limited number of different RFLP banding patterns. This clustering was highest in group B, with clusters significantly larger than those in group A. On the basis of the large size of the major group B clusters, it appears that these strains have some selective advantage in this geographical area. In this study two of the eight isolates from patients from neighboring countries clustered with group B isolates; one was a biovar 2 isolate from a patient from The Gambia and one was a biovar 3 isolate from a patient from Guinea Conakry.

From an evolutionary point of view, we suggest that the group B and C isolates, including the four M. bovis isolates, belong to a unique branch of the M. tuberculosis tree, with their origin in West African countries. We suggest that these genetically similar strains have a recent common ancestral "M. bovis-like" origin and that they are better adapted for infection and survival in the range of hosts present in this geographical area. This genotype may have a short history of adaptation to humans, since TB in sub-Saharan Africa has been suggested to be a recently occurring disease, in contrast to TB in northern Europe and the Americas (26). We also propose that the M. bovis strains of the "European" type (which possess a single IS6110 fragment and which lack DR spacer sequences 39 to 43) branched off at an earlier stage than the M. bovis strains of the type found in this study.