Is Mycobacterium africanum Subtype II (Uganda I and Uganda II) a Genetically Well-Defined Subspecies of the Mycobacterium tuberculosis Complex?

In a retrospective study of the population structure of Mycobacterium tuberculosis complex strains from a single hospital in Kampala, Uganda, Niemann et al. (6) reported that M. africanum subtype II is a major cause of human tuberculosis in this area. The reclassification of strains previously isolated during the period during 1995 to 1997 as M. africanum subtype II relied on phenotypic characteristics such as dysgonic colony morphology, change of color of bromocresol medium, resistance of the strains to 1 μg of thiophen-2-carboxylic acid hydrazide (TCH) per ml, and growth on Lebek medium. Molecular markers such as gyrB single-nucleotide polymorphism, spoligotyping, and IS6110 restriction fragment length polymorphism (RFLP) were also used to further study these isolates. According to the authors, the subtype II isolates differed from the classical M. africanum subtype I isolates (including the type strain ATCC 25420) by the absence of spacers 33 to 36 and 40. Additional absence of spacer 43 further subdivided the so-called “subtype II” isolates into Uganda II (only spacer 40 missing) and Uganda I (both spacers 40 and 43 missing). This paper is an interesting description of the population structure of the M. tuberculosis complex isolates from Uganda. However, the classification of these isolates as M. africanum variants, based essentially on variable phenotypic characteristics, is in contradiction with the recent findings on the evolution and phylogenetic structure of the M. tuberculosis complex.

The M. tuberculosis complex encompasses five members or variants—M. tuberculosis, M. africanum, M. bovis, M. canettii, and M. microti. Contrary to the other variants of the complex, M. africanum isolates have substantial phenotypic heterogeneity that appears to fill the phenetic discontinuity between two major pathogens: M. bovis and M. tuberculosis (3). The members of the M. africanum species were initially subdivided by David et al. (3) into two major subgroups, which corresponded to their geographic origin and biochemical properties. Those from western Africa, or group G of David et al., also called “subtype I,” were further subdivided into G1, G2, and G3 (closer to M. bovis). The ones from eastern Africa, or group F of David et al., also called “subtype II,” were further subdivided into F1, F2, and F3 (closer to M. tuberculosis).

Biochemical characters can give discordant results among laboratories if the laboratories use different criteria for interpretation, such as for growth in the presence of TCH with critical concentrations ranging from 1 to 5 μg/ml according to the laboratories. Another example is the test used by the authors to determine oxygen preference, which included Lebek medium. The interpretation of this test was not described (6), but referred back to an article from the same laboratory (5): “Aerophilic growth is indicated by growth on the surface and above the surface on the glass wall of the tube, whereas microaerophilic growth is indicated by growth below the surface.” Based on this interpretation, the authors described 100% of the subtype II isolates from Uganda as microaerophilic. These results are in contradiction with previous results of David et al. (3) when using Lebek medium for F2 and F3 subclusters of M. africanum from Rwanda and Burundi (also eastern Africa). A more widely used medium today for such studies is the Middlebrook 7H9 broth with 0.1 to 0.2% agarose (11): “Aerobic strains grow at or near the surface, while microaerophilic strains grow as a band 10 to 20 mm below the surface, sometimes extending upwards.” An example of these results clearly differentiating aerophilic and microaerophilic growth can be found in a recent publication (7). Perhaps the authors should retest the putative M. africanum subtype II by using this currently recommended method.

Moreover, misidentification might also arise for strains with atypical features due to multidrug resistance, a frequent finding in African countries. Consequently differentiation of M. africanum should no longer be based on phenotypic characteristics exclusively (10). Many strains phenotypically classified as M. africanum probably include not only true M. africanum, but also other subgroups within the M. tuberculosis complex whose taxonomic position is yet uncertain, and only a finer genetic analysis may help classify them correctly. Although the formal genetic definition of M. africanum strains remains to be established, recent papers have shed light on their phylogenetic position within the M. tuberculosis complex. The first paper by Streevatsan et al. (9) defined three major genetic groups (I, II, and III) relying on the katG⁴⁶³-gyrA⁹⁵ single-nucleotide polymorphisms. According to this well-established and widely used scheme, the terms M. microti, M. africanum, and M. bovis should be reserved for M. tuberculosis complex group I organisms, whereas M. tuberculosis may be found in groups I, II, and III (9).

Based on the distribution of 20 variable regions in the genome of the M. tuberculosis complex, it was shown recently that M. africanum strains lack a specific region, RD9, and sometimes RD10 (2, 7), each of which is present in M. tuberculosis. M. africanum is also characterized by the presence of the region TbD1, a deletion specific for “modern” M. tuberculosis isolates (2). In a study by Brosch et al. (2), two strains from Uganda that were previously identified by phenotypic characters as M. africanum harbored this specific RD9 region, lacked TbD1, and presented a katG⁴⁶³ CGG (Arg) corresponding to group II or III of Sreevatsan and colleagues. Results from a second study (7) confirmed the presence of the RD9 region and also demonstrated a preference for aerophilic conditions in these two strains. These results suggest that these so-called M. africanum strains should be regarded as M. tuberculosis rather than M. africanum. After the publication of the paper by Niemann et al. (6), we reexamined three additional strains from East Africa (two from Burundi and one from Rwanda) identified as M. africanum phenotypically that also bore a spoligotype signature identical to the M. africanum subtype II strain (Uganda I) described by Niemann et al. (5). All of these three strains did carry the RD9 region, a feature suggesting that they were not a true M. africanum variant. These data are in agreement with the fact that Niemann et al. (6) could not distinguish the M. africanum subtype II from M. tuberculosis on the basis of the gyrB single-nucleotide polymorphism, which was used by the same authors to differentiate between the different members of the M. tuberculosis complex.

A last point concerns the use of the missing spacer 40 (and the absence of spacers 33 to 36) as a marker that may help to designate a tubercle bacillus originating from East Africa as M. africanum subtype II (6). The observation that spacer 40 is absent in many strains, whatever their geographical origin, was pointed out recently by Bifani et al. (1) for the Beijing group of M. tuberculosis. Besides, no strain showing simultaneous absence of spacers 33 to 36 and 40 and belonging to major genetic group I were reported in the Houston study (although many M. africanum subtype I strains with simultaneous absence of spacers 8 to 9 and 39 were reported in reference 8). Furthermore, we have independently observed in the Institut Pasteur de Guadeloupe spoligotype database a trend that the absence of this spacer 40 may be linked to an African origin of M. tuberculosis. (Reference 4 gives a current update with more than 21,000 isolates and 1,250 shared types from 100 countries.) However, spacer 40 was also missing from some other yet poorly defined clades, such as T2 and LAM4 (4; unpublished observations), which are all M. tuberculosis strains. Indeed, homoplasic events, due to a topological instability of this specific direct repeat (DR40), could be the cause for the frequent and independent loss of this spacer in different strain lineages. Consequently, putting too much emphasis on the absence of spacer 40 found in isolates from a single hospital in a single town of a single country and trying to generalize this observation to the dimension of an entire continent (Africa) in terms of prevalence may not be justified. Finally, it should be mentioned that in two independent studies using large sample sizes (2, 8), strains that lacked spacers 33 to 36, but had the flanking spacers present, exclusively carried the katG⁴⁶³ mutation CGG characteristic for M. tuberculosis strains of genetic group II or III. The absence of these spacers from the M. africanum subtype II strains described by Niemann et al. (6) represents an additional argument that these strains belong to M. tuberculosis group II or III strains rather than to M. africanum.

Based on several independent genetic markers, it appears that this interesting predominant group of strains in Uganda described as M. africanum subtype II phylogenetically represents a variant of M. tuberculosis that is distant from M. africanum. It seems plausible that the phenotypic similarities to M. africanum, such as the dysgonic colony morphology, may be due to yet-unidentified genetic changes that may have occurred independently in various members of the M. tuberculosis complex.

As a consequence, we therefore suggest that for future studies of M. africanum, the presence or absence of RD9 and the determination of the Sreevatsan's major genetic groups should be taken in consideration, before new isolates are designated as M. africanum. Based on these recommendations, a European network of experts is at present working to develop a consensus definition of M. africanum under the EU Concerted Action project QLK2-CT-2000-00630. Readers who wish to help contribute to this project may contact the undersigned investigators.