INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The Mycobacterium tuberculosis complex includes four recognized species (M. tuberculosis, M. bovis, M. africanum, and M. microti), which together cause human and animal tuberculosis (37). M. tuberculosis and M. bovis are the most common isolates in clinical laboratories and can be distinguished by biochemical tests, growth phenotypes, and several genetic markers. However, M. africanum isolates have substantial phenotypic heterogeneity with some strains resembling M. bovis and others resembling M. tuberculosis (4, 10, 12, 15, 36). This has led several authors to question the validity of the species designation M. africanum (33, 37, 39).

We conducted a detailed genotypic analysis of a panel of 17 recent M. africanum isolates. All were isolated from humans in Sierra Leone between 1991 and 1993 and were identified as M. africanum on the basis of biochemical tests and growth patterns. These strains are a subset of a collection described previously (12, 14, 29). We sequenced portions of several genes and intergenic regions which are known to have polymorphisms in M. tuberculosis and M. bovis. We also analyzed six loci with variable numbers of tandem repeats (VNTRs). Our goal was to determine whether M. africanum strains have distinct genotypes which set them apart from M. tuberculosis and M. bovis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Mycobacterial strains. Mycobacterial strains are listed in Tables 1 and 2. M. africanum isolates were obtained from humans in Sierra Leone with pulmonary tuberculosis (12, 29). These strains were part of a large collection of 96 strains originally collected for the purpose of drug resistance surveillance organized by the National Tuberculosis Control Program in Sierra Leone and supported by the German Leprosy Relief Association. Isolates were characterized as M. africanum based on biochemical and biophysical properties as described previously (12). Susceptibility testing was performed by the proportion method using the critical concentrations of 40 µg/ml for rifampin, 0.2 µg/ml for isoniazid (INH), 4 µg/ml for streptomycin, and 2 µg/ml for ethambutol. Pyrazinamidase production was determined as described by Wayne (35). The 17 M. africanum isolates in this study were selected from the larger collection to include isolates with a maximal divergence in antibiotic susceptibility phenotypes. M. africanum ATCC 25420^T (the T indicates type strain), M. bovis ATCC 19210^T, M. microti ATCC 19422^T, M. tuberculosis ATCC 27294^T, and M. bovis ATCC 35734 (BCG Pasteur) were obtained from the American Type Culture Collection in Manassas, Va.

DNA preparation. Mycobacterial isolates were killed by immersion in 70% (vol/vol) ethanol (40). DNA was released from bacterial cells by snap freeze-thawing (41).

DNA sequence analysis. A portion of the rpoB gene was amplified by PCR and sequenced as described previously (41). Similarly, the first 500 codons of the katG gene, the upstream noncoding region of inhA, and the oxyR-ahpC intergenic region were amplified and sequenced as described previously (21, 26). Sequences were compared to wild-type sequences found in GenBank under accession no. L27989 (rpoB), Z97193 (katG), U02492 and Z79701 (inhA), and U16243 (oxyR-ahpC). We did not sequence the gyrA gene. Although polymorphisms at gyrA codon 95 are useful for separating isolates of M. tuberculosis, this codon is not variable in M. africanum (26).

VNTR analysis. Six loci containing variable numbers of tandem repeats (VNTRs) were amplified by PCR as previously described (9). The exact number of tandem repeats at each locus in each strain was determined by the length of the PCR products on agarose gels. One of these loci (MPTR-A) contains the major polymorphic tandem repeat (MPTR), and five loci (ETR-A through ETR-E) contained exact tandem repeats (ETRs). Primers, PCR conditions, map locations, and other details related to these VNTR loci were published previously (9).

Phylogenetic analysis. Phylogenetic trees were constructed based on the VNTR allele profiles by maximum-parsimony analysis using PAUP software (30). Maximum parsimony and other methods for phylogenetic analysis are described in recent reviews (22, 28). The data were resampled with 10,000 bootstrap replications (6), and a majority-rule consensus was computed from the most-parsimonious trees. The analysis was conducted twice, treating character states as either ordered or unordered. Treating the character states as ordered assumes single-step transitions in the number of tandem repeats at each locus. This assumption is consistent with the results of sequence analysis of the MPTR-A locus (7). Genetic distances between pairs of VNTR allele profiles were calculated based on the minimum number of single-step transitions.

Mixed-linker PCR. Genomic DNA was digested, ligated with mixed linkers, and amplified in a nested PCR as previously described (13). The PCR products were separated by agarose electrophoresis to generate a DNA fingerprint based on the mobile IS6110 insertion element. Mixed-linker fingerprints for this collection were previously published (12).

Other genetic analyses. RD1 is a 9.5-kb region deleted in substrains of M. bovis BCG. Two primers which flank RD1 and one primer within RD1 were combined in a multiplex-PCR amplification as previously described (31). Strains containing the RD1 region yield a 150-bp product, while strains with deletion of RD1 (all BCG substrains) yield a 200-bp product. Substrains of M. bovis BCG also have a sequence polymorphism at codon 252 of a 1,551-bp open reading frame (orf1) in GenBank under accession no. M15467 (gene homologous to Rv0422c). A portion of this gene was amplified by PCR, and the sequence at this codon was determined by digestion with the restriction enzyme BanI as previously described (7).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Tables 1 and 2 list phenotypic and genotypic results for 17 M. africanum isolates from Sierra Leone. These isolates were identified as M. africanum on the basis of biochemical and biophysical properties as listed in Table 1 (12, 36). M. africanum isolates grew as smooth, dysgonic, nonpigmented colonies. Growth was inhibited by thiophene-2-carboxylic hydrazide (TCH), with the exception of growth by strains resistant to INH. All strains had pyrazinamidase activity except two isolates (014 and 213) which were resistant to multiple drugs. Acquisition of pyrazinamide resistance is associated with the loss of pyrazinamidase activity. Most M. africanum isolates had catalase activity less than or equal to 10 mm (height of bubbles above agar) at 22°C, and all were negative for catalase activity at 68°C. All isolates except 370 grew at least 5 mm below the surface (niveau depth) on Lebek's semisolid medium, indicating microaerophilic growth. All isolates produced no color change when grown on bromcresol medium. The nitrate reduction and niacin accumulation phenotypes were variable, as described for other collections of M. africanum strains (4, 36).

katG sequence. katG codes for catalase peroxidase. Loss of catalase peroxidase activity can result from deletions or point mutations in katG and is associated with INH resistance. We sequenced the entire katG gene in the 17 M. africanum strains and the 5 reference strains and compared the sequences to the wild-type sequence (GenBank accession no. Z97193). Results are listed in Table 2. We found mutations at five katG codons (203, 253, 254, 315, and 463). Eleven strains were INH resistant; of these strains, seven had polymorphisms at katG codon 315. One INH-resistant strain had both a point mutation at codon 253 and a frameshift mutation at codon 254. Identical or similar katG mutations are associated with INH resistance in M. tuberculosis (21).

View larger version (12K): [in this window] [in a new window]   FIG. 1.   Broad evolutionary scenario for M. tuberculosis complex organisms. Sreevatsan et al. (26) defined groups 1, 2, and 3 based on the sequence of katG codon 463 and gyrA codon 95. We subdivided group 1 strains based on the sequence of katG codon 203. The distribution of the 22 strains described in this manuscript is shown.

rpoB sequence. rpoB codes for the  subunit of the DNA-dependent RNA polymerase which is the target of rifampin activity. Mutations in rpoB codons 509 to 533 are strongly associated with rifampin resistance in pathogenic mycobacteria, including M. tuberculosis, M. leprae, and M. avium (17, 41). Eight of the M. africanum isolates from Sierra Leone, seven of which had rpoB mutations, were resistant to rifampin. Nine of the M. africanum isolates were rifampin susceptible, and eight of these had the wild-type rpoB sequence in this segment of the gene. Interestingly, M. africanum 556 was rifampin susceptible but had a codon 533 Leu-to-Pro (LeuPro) mutation previously associated with rifampin resistance in clinical M. tuberculosis isolates (41). M. africanum 556 is no longer viable, so we were unable to repeat these analyses. Overall, we observed six distinct rpoB mutations in these M. africanum strains, all of which have been described in rifampin-resistant M. tuberculosis strains.

VNTR analysis. We amplified DNA at six loci with variable numbers of tandem repeats (9). The number of tandem DNA repeats at each locus in each strain was determined by the length of the PCR product on an agarose gel. The results of the six loci (MPTR-A, ETR-A, ETR-B, ETR-C, ETR-D, and ETR-E) were combined to yield a six-digit VNTR allele profile (9). Each digit represents the number of tandem repeats at a particular locus, with the exception of locus MPTR-A, where it represents the number of copies minus 10. For example, M. africanum^T has the VNTR allele profile 564544. This strain has 15 tandem repeat copies at locus MPTR-A, 6 copies at locus ETR-A, 4 copies at locus ETR-B, 5 copies at locus ETR-C, 4 copies at locus ETR-D, and 4 copies at locus ETR-E. VNTR allele profiles are listed in Table 2. Figure 2 displays phylogenetic relationships among the VNTR allele profiles, and Table 3 shows genetic distances between pairs of VNTR allele profiles.

View larger version (24K): [in this window] [in a new window]   FIG. 2.   Parsimony phylogenetic tree based on VNTR allele profiles. The tree includes 12 VNTR allele profiles generated from 22 strains. The number of clinical isolates is indicated in parentheses, as are the profiles associated with the type strains (T) and BCG Pasteur (BCG). VNTR allele profiles found in M. africanum group A are marked by an asterisk. Horizontal lengths represent genetic distances; vertical lengths are not meaningful. All clusters shown on this tree were found in 100% of the most-parsimonious trees. The data were resampled with 10,000 bootstrap replications (6, 30). The percentage of bootstrap replications which yielded each grouping is indicated. The tree was constructed with the assumption that the number of tandem repeats in each strain changes by a single copy at a time (ordered character states). When we conducted the same analysis with unordered character states, the M. africanum group A isolates remained clustered.

Mixed-linker PCR fingerprints. The mixed-linker PCR fingerprints for these strains were previously published (12). Eleven of the M. africanum isolates in group A were clustered by mixed-linker analysis, and three were unique. Two of the group B isolates were unique, and the third was not analyzed by mixed-linker PCR. Cluster 1 and cluster 2 have mixed-linker PCR fingerprints which differ by a single band and together accounted for 40% of all M. africanum isolates from Sierra Leone in a previous report (12).

Other analyses. All 17 M. africanum isolates had wild-type sequences in the upstream noncoding region of inhA and the oxyR-ahpC intergenic region (26). All 17 M. africanum strains had the wild-type sequence AGC Ser at orf1 codon 252 (7). All 17 M. africanum strains contained RD1, as previously reported (31).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The M. tuberculosis complex includes four species on the Approved List of Bacterial Names: M. tuberculosis, M. bovis, M. africanum, and M. microti (37). A fifth group, characterized by the Canetti strain, has distinct features, but has not been proposed as a separate species (34). Strains from all four species show >85% DNA-DNA relatedness and have identical 16S rDNA and 16S-23S rDNA internal transcribed spacer sequences (8, 16, 18). Comparative sequence analysis of multiple structural genes revealed minimal polymorphism among strains of all four species (26). Although several authors have suggested that the four species should be reclassified as subspecies of M. tuberculosis, no formal proposal has been made (34, 37).

M. tuberculosis is the major cause of human tuberculosis worldwide. M. bovis causes tuberculosis in multiple mammalian species, including humans. The distinction between M. tuberculosis and M. bovis has epidemiologic implications, since infected animals are often a source for human M. bovis infections, and therapeutic implications, since M. bovis strains are resistant to pyrazinamide. M. bovis may be less virulent than M. tuberculosis in humans, but it is clearly more virulent in some animal models. Strains of M. tuberculosis and M. bovis can be distinguished on the basis of epidemiology, biochemical tests, growth phenotypes, and several genetic markers. M. tuberculosis strains nearly always have a functional nicotinamidase, produce niacin, reduce nitrate, have enhanced growth in the presence of glycerol, and are resistant to TCH (36, 37). In contrast, M. bovis strains lack nicotinamidase function, do not produce niacin or reduce nitrate, are not stimulated by glycerol, and are sensitive to TCH (36, 37). M. tuberculosis and M. bovis strains can be distinguished by the sequence of oxyR nucleotide 285 (5, 25). Also, most, but not all, M. bovis strains have a guanine at pncA nucleotide 169 and lack the mtp40 gene (5, 20, 24, 27, 38).

M. africanum was described by Castets and colleagues in 1968 (2). The type strain (ATCC 25420) was isolated from the sputum of a tuberculosis patient in Senegal. Strains of M. africanum display substantial phenotypic heterogeneity, and there is no simple phenotypic definition of the species (10, 12, 15, 36, 37). Difficulties in precisely defining or identifying M. africanum have complicated recent studies of this species (15). Both genotypes and phenotypes of M. africanum vary by geographic area (4, 10, 12, 14), so reports from different regions are not directly comparable. Collins et al. (4) used the nitrate reduction assay to divide M. africanum strains into two groups, African I (negative for nitrate reduction) and African II (positive for nitrate reduction). African I isolates are generally associated with West Africa, and African II isolates are generally associated with East Africa (4, 11, 12). However, a recent survey identified similar numbers of African I and African II isolates in the West African nation of Guinea-Bissau (15).

Several researchers have questioned whether M. africanum is actually distinct from the other members of the M. tuberculosis complex (33, 37, 39). However, the existing M. africanum literature suggests that it differs from M. tuberculosis and M. bovis in host range, source of infection, geographic distribution, virulence, and pathogenesis. M. africanum has been reported in monkeys as well as humans (32). It is a common cause of human tuberculosis in both East and West Africa (12, 19, 23). Strains resembling M. africanum are isolated at low rates from humans on other continents (4, 11, 42). These strains cause a different pattern of disease than M. tuberculosis. For example, they are rarely isolated from humans with genitourinary tuberculosis (11). In experimental models, M. africanum has reduced virulence compared to M. tuberculosis (1, 3).

We identified isolates from Sierra Leone as M. africanum on the basis of biochemical and biophysical properties. In general, these strains differed from M. tuberculosis and M. microti in having negative or weak niacin production and differed from M. bovis in their pyrazinamidase production and their susceptibility to TCH. However, these phenotypes were variable (Table 1), as has been reported by other investigators. Phenotypic analysis is complicated by some changes associated with drug resistance. Acquisition of INH resistance is associated with TCH resistance (12), and acquisition of pyrazinamide resistance is associated with a loss of pyrazinamidase (24, 27).

These 17 M. africanum strains shared a number of genotypes with the rest of the M. tuberculosis complex, including the wild-type sequences of inhA, the oxyR-ahpC intergenic region, orf1 codon 252, and the presence of RD1. These M. africanum strains have mutations leading to INH and rifampin resistance similar to those seen in M. tuberculosis and M. bovis.

The 17 M. africanum strains fell into two distinct genotypic groups (Fig. 1). Fourteen group A strains had katG codon 203 ACT Thr and codon 463 CTG Leu and had closely related VNTR allele profiles. Their genotypes and phenotypes were similar to those of M. africanum^T from Senegal. The geographic distribution of group A M. africanum strains is unknown. In addition to the isolates described here, we have identified M. africanum strains from New York and The Netherlands with VNTR allele profiles that cluster with group A (data not shown). In contrast, the three group B strains had katG codon 203 ACC Thr and VNTR allele profile 642432. In these respects, they cluster with M. tuberculosis^T.

VNTR analysis and mixed-linker PCR fingerprinting are based on different genetic targets and were independently informative for strain differentiation. The 14 group A M. africanum strains formed 7 clusters by VNTR analysis, 7 clusters by mixed-linker analysis, and 10 clusters when the results of both analyses were combined. VNTR analysis and the sequence at katG codon 203 yielded identical separation of the M. africanum strains into genotypic groups A and B. Each cluster identified by mixed-linker analysis was also restricted to a single genotypic group. VNTR analysis may be useful for evolutionary study when combined with other techniques. VNTR analysis samples multiple independent genomic loci. VNTR results are digital and are amenable to phylogenetic and genetic distance analyses as shown in Fig. 2 and Table 3. However, the number of alleles at each VNTR locus is limited, so genetic convergence may occur frequently. Convergence would make strains appear more closely related than they actually are.

All 14 M. africanum group A strains had typical M. africanum phenotypes. One group B strain had two features more typical of M. tuberculosis, namely, it was positive for niacin production and showed 16 mm of catalase activity. Under the classification system proposed by Collins et al. (4), all 14 group A strains and one group B strain are African I (negative or weakly positive for nitrate reduction) and two group B strains are African II (positive for nitrate reduction). However, there was no phenotype which clearly separated the genotypic groups A and B.

Strains with M. africanum phenotypes have substantial genotypic variability. M. africanum group A isolates had M. africanum phenotypes and clustered with M. africanum^T in genotypic analyses. M. africanum group B strains also had M. africanum phenotypes but clustered with M. tuberculosis strains in genotypic analyses. Haas et al. (14) identified strains from Uganda with M. africanum phenotypes and katG codon 463 CGG Arg, a sequence previously found only in M. tuberculosis (26). These data suggest that multiple genetically distinct strains have converged toward an M. africanum phenotype. Our results support the retention of the designation M. africanum for the group A strains, since they cluster with M. africanum^T in multiple analyses and are distinct from the rest of the M. tuberculosis complex. Future studies of M. africanum should include both phenotypic and genotypic analyses.