The variable efficacy of BCG vaccines under different conditions has been discussed almost ad nauseam (Bloom & Fine 1994, Fine 1995). In summary, we know that at least some BCG vaccines perform well, in terms of protection against pulmonary tuberculosis, under certain conditions. But the same vaccines can perform poorly in different contexts, and we remain ignorant of the determinants of this variance. Among the reasons often discussed are several that do not call immediately for genetical analyses: (1) that the differences reflect methodological differences between studies; (2) that they reflect physiological (e.g. nutritional) differences between the populations: or (3) that they reflect exposure to ultraviolet light, which either inactivates the vaccine or suppresses dermal Langerhans cells. But there are other explanations, which are explicitly genetic in flavour.
It has been proposed that differences in BCG vaccine efficacy might reflect genetical differences between Mycobacterium tuberculosis strains prevalent in different populations or parts of the world. This hypothesis first received attention after the 'failure' of both Paris and Danish strain BCG vaccines in the Chingleput trial in south India, when it was noted that many strains of M. tuberculosis which had been isolated in that region of the world were of relatively low virulence for guinea-pigs (Tuberculosis Prevention Trial 1980, Mitchison 1964). The argument ran something like this: that the M. tuberculosis strain current in the trial population had peculiar virulence properties, such that it led rarely to primary progressive disease (against which BCG is purported to be most effective), but that it led frequently to reinfection or reactivation of disease, against which BCG is less effective. This explanation was consistent with the observation of low incidence of disease among those considered tuberculin negative at the onset of the Chingleput trial, and with observed high risks of disease among those considered tuberculin positive at the onset of the trial (ten Dam & Pio 1982).
The hypothesis was tested in guinea-pigs by Hank et al (1981), who were unable to note any difference in the effect of either Danish or Prague strain BCGs between supposedly low virulence M. tuberculosis isolates from the trial area and recently isolated high virulence strains from North America. Although these animal studies related more to protection against primary disease than against reinfection or reactivation of disease, their publication eclipsed the argument about the so-called 'south India variant', which has since been labelled a 'red herring' by D. A. Mitchison, upon whose work the hypothesis was originally based (personal communication 1996).
That geographic differences in M. tuberculosis might have something to do with the variable efficacy of BCG has more recently arisen in the context of discussions of fingerprint analyses of isolates from different populations (Hermans et al 1995). Unfortunately, the studies appropriate to test this hypothesis have yet to be made (e.g. comparison of fingerprint genotypes between cases with and without history of prior BCG vaccination). Such comparisons will soon be made, using IS6110 fingerprints as markers. But we have a logical problem, in that even if the results are negative, this would of course not be sufficient to refute the hypothesis convincingly, given that the fingerprint-dependent strain designation is crude and may be totally irrelevant as far as virulence or antigenicity is concerned. Much more powerful and relevant markers will be revealed in the months to come, as we have a better understanding of the entire genome of M. tuberculosis, and these will need to be examined as covariates in the context of appropriate epidemiological studies of BCG. I look forward to such studies, and to the comparisons of field isolates from different cases and populations to the benchmark sequences being derived in the current sequencing of the H37Rv and Oshkosh (CSU isolate 93) 'strains'.
The story of the influence of host genetics on tuberculous infection and disease is discussed elsewhere in this volume. The classic evidence is based upon anecdotes implying differences in susceptibility between races, the recognition of family clustering, then twin studies, and, more recently, family segregation and case-control studies indicating that specific polymorphisms, e.g. the human leukocyte antigen DR2, influence susceptibility in certain populations (Fine 1981, Brahmajothi et al 1991, Comstock 1978). Inevitably, such studies have led to suggestions that genetical differences between populations might explain the observed differences in the behaviour of BCG. But the direct evidence is thin, at best.
Comstock & Palmer (1966) compared the efficacy of BCG between Caucasians and Blacks in the US Public Health Service trials in Georgia and Alabama. The observed protection was greater in Caucasians (46%; 95% confidence interval: -33% to 78%) than in Blacks (-6%; 95% confidence interval: -102% to 44%) but the numbers of cases were small (21 and 37, respectively), and the difference non-significant statistically. More recently, the finding of low efficacy in the south India Chingleput trial led to two case-control studies of BCG and tuberculosis among Asians in the UK, both of which showed evidence of appreciable protection in the UK (Rodrigues et al 1991, Packe & Innes 1988). Of course one could argue that the Asians included in these studies were not south Indian Dravidians, and hence were not comparable to the population of Chingleput District, and thereby declare the argument to be still open. Several studies are now employing genome scanning techniques to identify new genes associated with tuberculosis, and polymorphisms of candidate genes are being recognized, including those determining the various cytokines involved in cellular responses to mycobacteria. A next step in this evolution will be to look for differences in frequencies of particular polymorphisms between BCG-vaccinated and BCG-unvaccinated cases, and BCG-vaccinated and BCG-unvaccinated individuals similarly exposed but free of disease. The epidemiology implicit in such studies is not simple, and I hope we won't see too many overenthusiastic false positive (or negative) claims flooding our journals before sufficient substantial evidence accumulates allowing some firm conclusions.
BCG has never been cloned. It originated from an isolate of Mycobacterium bovis taken from a cow with mastitis, was attenuated over 13 years of passage in media containing ox bile, and has since been maintained in a variety of culture media in various laboratories around the world (Bloom & Fine 1994). As a consequence, there are many different strains of BCG, with different properties; and it has long been wondered whether these differences might not be responsible for some of the observed differences in efficacy. It would be most convenient if this were so, as it could point us towards particular antigens or epitopes that are important in protection.
There are several examples which appear (at least at first) to be consistent with this explanation. Comstock reanalysed case-control study data on the efficacy of BCG in Indonesia and in Colombia, and found evidence for a possible change (decline) in efficacy when the programmes shifted from one (Japanese or British) to another (French or Danish) vaccine (Comstock 1988).
More interestingly, a controlled trial comparing Paris and Glaxo vaccines was carried out in Hong Kong under the auspices of the World Health Organization (WHO) and the Hong Kong tuberculosis services. All infants born between 1978 and 1982 (more than 300 000) were randomized (by alternating vaccine supplies to vaccination centres) to receive one or the other vaccine, by either the intradermal or percutaneous route. They were followed up over the following six years, during which 129 cases were ascertained. Recipients of the Paris-type vaccine were approximately 40% less likely to contract tuberculosis than were those who received the Glaxo vaccine, by either route (^<0.05; ten Dam 1993). It is interesting that the differences implied by this trial go in the opposite direction to those suggested by the observational study analyses by Comstock (1988).
Molecular analyses have now identified particular genomic regions found in M. bovis, but which appear to be absent from some or all BCGs (Mahairas et al 1996, Philipp et al 1996). In particular, it has been found that the so-called RD-2 region, which contains the mpt-64 (also called MPB-64) gene, is present in the 'primitive' BCG strains (represented by current Brazilian [Moreau], Japanese and Russian substrains), but is absent from those substrains derived from the original BCG Pasteur strain after 1925, represented by today's Pasteur, Copenhagen and Glaxo-Evans substrains (Mahairas et al 1996). The full immunological implications of these deletions, if any, are as yet unknown.
It has recently been suggested that conscious selection by BCG vaccine manufacturers for decreased reactogenicity may have led to decreased immunogenicity of the vaccines (Behr & Small 1997). Despite the absence of human evidence that induction of tuberculin delayed-type hypersensitivity (DTH) was associated with protective immunity (Comstock 1988, Hart et al 1967, Fine 1993), the standard potency assay for BCG vaccines has long been their ability to induce tuberculin hypersensitivity in guinea-pigs. It is thus credible that manufacturers may have selected for strains which maintained (and perhaps maximized?) their ability to induce tuberculin hypersensitivity, but which minimized their induction of regional lymphadenopathy. Ironically, this selection might have favoured a decrease in protective immunizing potential. This is suggested by trends for decreased vaccine protection with increased passage number of several strains of BCG vaccines (Behr & Small 1997). The argument is interesting, perhaps even plausible, but not compelling, for two reasons.
First, there are several examples of identical vaccines providing different levels of protection in different populations. The most obvious example of this is the high efficacy of the Glaxo freeze-dried strain, which continues to provide good efficacy against pulmonary tuberculosis in the UK, but no evidence of any protection against pulmonary tuberculosis in Malawi (Karonga Prevention Trial Group 1996).
Second, it happens that the earlier trials, which used lower passage number vaccines, were carried out in northern Europe and the US, whereas the high passage number strains were evaluated at lower latitudes. A tendency for lower efficacy of BCG vaccines at lower latitudes has been noted by many authors (Palmer & Long 1966, Fine 1995), but has generally been attributed to the interference provided by exposure to environmental mycobacteria or to ultraviolet light. We may thus ask whether this latitude effect is likely to reflect genetical differences between vaccines or vice versa. Given the examples of fixed strains performing differently at different latitudes, and the abundant animal evidence for masking of BCG effects by prior exposure to environmental mycobacteria, it would seem inappropriate to attribute all the latitude effect in efficacy differences to variation between BCG strains.
The implications of environmental mycobacteria have been referred to above. This subject has attracted attention for more than three decades, and much evidence has accumulated indicating that regional differences in exposure to various mycobacteria are responsible for some of the observed differences in observed efficacy of BCG (Palmer & Long 1966, Fine 1995). Unfortunately, the arguments have yet to come down to species or molecular detail, largely because of the bewildering variety and complexity of the environmental mycobacterial flora in different regions of the world. Classic experiments on guinea-pigs suggested that exposure to Mycobacteriumfortuitum, Mycobacterium avium or Mycobacterium kansasii imparted 15%, 50% and 85% as much protection, respectively, as did BCG, as measured in terms of survival after challenge with H37Rv (Palmer & Long 1966). It is reasonable to presume that these differences somehow reflect antigenic relatedness to the tubercle bacilli.
On the other hand, the logic may be more perverse. Evidence to date suggests that BCG provides stronger protection against the more distantly related Mycobacterium leprae than against its closer relative, M. tuberculosis (Fine 1995). This has led some investigators (e.g. Rook & Hernandez-Pando 1998, this volume) to suggest that simple antigenic similarity is not the key to protection, and that it might even be detrimental by eliciting hypersensitivity, which is ultimately harmful. Not all mycobacterial immunologists would agree with this interpretation, but, given our ignorance of the natural history of mycobacterial infections, and of the nature of protective immune responses against them, we would be wise to keep such possibilities on the table.
Was this article helpful?