Main

An intriguing, but yet unanswered question is what defines resistance and susceptibility to Mycobacterium tuberculosis in man. Although TB is generally regarded as highly infectious, most individuals who have been exposed to M. tuberculosis fail to develop TB. Resistance to early growth of mycobacteria in some mouse strains is controlled by a single autosomal gene termed Bcg(1), which encodes for an intracellular protein Nramp1(2). Although this work focused attention on innate genetic resistance to infection, the extension of these findings from mice to man has yet to yield definitive results(3,4). No definitive association of resistance to TB with HLA haplotype has been found(4,5), although recently HLA-DQB1*0503 may be associated with susceptibility to TB in Cambodian patients(6). It is clear that cellular immunity plays a key role in limiting early disease(7). In this regard, infection with the human immunodeficiency virus, HIV, renders patients 170-fold more susceptible to TB(8–10). However, in most cases the factors that disturb the balance between host and pathogen remain undefined. Virtually no studies have examined factors that may play a role in the dissemination of M. tuberculosis. We therefore considered whether MBP might contribute to susceptibility to TB or to progression of disease in a well-defined South African population.

The human MBP is an acute-phase protein synthesized in the liver(11,12). MBP is a member of the collectin family(13) as it has a collagen region and a lectinlike domain(11,14). The overall organization of the hexamer of trimers resembles C1q, the first complement component. Overwhelming in vitro evidence indicates that MBP selectively recognizes the patterns of carbohydrate that decorate microorganisms, certain viruses, or virally infected cells. Microorganisms that are recognized by MBP include g-positive and g-negative bacteria, yeast, parasites, and mycobacteria(15–18). The MBP is an opsonin(19,20) and is considered as an "ante-antibody"(21). Human MBP can also activate complement via a novel so-called lectin pathway(22,23), substitute for C1q in the classical pathway(24), or activate the alternative complement pathway(25).

Low or absent MBP levels in children render them susceptible to recurrent infections(18,20,26,27). Three mutations in the first exon of the human MBP gene are associated with absent or very low levels of serum MBP(15,18,26,28–30). Two mutations, the B allele (G54D) and the C allele (G57E) encode for unstable MBP multimers as they disrupt the regular collagen helix(18). The high frequency of these mutations, which approaches 30% in certain populations, led to the idea that under certain circumstances MBP may be protective, whereas under other circumstances there may be a selective advantage for the heterozygous state(18,31). Accordingly, in diseases like TB in which intracellular tropism depends on phagocytosis of the bacillus, it has been suggested that high MBP concentrations facilitate this process(16,17,31). The fact that variant MBP alleles, associated with low serum MBP levels, occur with a relatively high frequency in certain populations (e.g. a frequency of 0.23 for allele C in sub-Saharan Africans(30)), has led to speculation that there may be selective pressure for the MBP variants to persist in populations with high exposure to specific intracellular pathogens(18,31). To explore this idea further, we investigated whether there was a correlation between MBP level or genotype and incidence of pulmonary or meningeal TB in a South African population with a TB incidence of approximately 700 per 100 000 population(32,33). It is important to note that the prevalence of HIV is low (1.66%) in this population(33), suggesting that inherited host factors may play a role in mediating resistance to the disease.

METHODS

Setting. This study was done in the Cape Town metropolitan area where the TB incidence was approximately 700 per 100 000 population(32–34) with an HIV prevalence of 1.66% (95% CI, 0.94-2.38%)(33). This study is a case-control study of retrospective design. The majority of patients and controls were from a single circumscribed suburb in which people were all of similar socioeconomic status.

The people studied were from the same ethnic [coloured(35,36)] background and unrelated. "Coloured" refers to a distinct ethnic group in this region of South Africa as distinct from black, white, or Asian. Family history of TB was not taken into account as this was an association study, but it is assumed that the vast majority of controls in this high-incidence community would have been exposed to TB.

Patients and controls. The sex and age were similar in both patient and control groups: the control group was 31% male, 69% female with a mean age of 27.2 y; the TB group was 45% male and 55% female with a mean age of 28.7 y. It is important to note that TBM is a disease of the very young so all TBM cases were younger than 10 y. Peripheral blood samples were obtained from the following groups of people as classified by the ATS. Ethical approval and informed consent was obtained in all cases.

Active TB: [TB clinically active (ATS class 3;Ref. (37))]. All these patients had clinical as well as radiographic or microbiological evidence of TB and had been on antituberculosis treatment for less than 2 wk at the time of blood sampling. Patients in this group had either pulmonary TB or TBM.

Previous TB: [TB not clinically active (ATS class 4;Ref. (36))] Patients in this group had all received antituberculosis therapy in the past for what was at that stage regarded as clinically active TB. At the time of blood sampling, there were no clinical symptoms or signs of TB and in those cases in which radiology or bacteriology was done, no evidence of TB was found.

Control: The controls were apparently healthy people from the same community, with no history of TB. The group consisted of individuals with no symptoms or signs of TB at the time of blood sampling and who had never been treated for TB in the past. Routine tuberculin skin testing was not performed of those individuals older than 20 y (BCG vaccination confounds interpretation of this test in our population), but they all lived in a community where one in every 3 houses has had at least one case of TB in the 10-y period 1985-1994(34). All adults were therefore assumed to have been exposed to TB and thus belonged to the ATS classification class 1 or 2(37).

Tuberculin skin testing was performed on 90% of the controls who were younger than 18 y. In 16% of these, the Mantoux skin test was negative (ATS class 0 or 1) and in 84%, the Mantoux skin test was positive (ATS class 2). Only an induration ≥15 mm was regarded as positive, because >98% of children in the Western Cape receive BCG in the neonatal period(38). The BCG strain used, Tokyo 172, is known to give a high percentage of Mantoux response >10 mm induration(39), and the hypersensitivity after BCG vaccination is known to persist for at least 2 y(39).

Assays. Blood was collected by venipuncture into EDTA-coated tubes. Whole blood was centrifuged at 1600 rpm for 15 min. The plasma was extracted and stored at -20°C. MBP concentration was measured in a double-antibody immune assay(40). CRP was measured by means of a latex agglutination slide test (the Humatex CRP test, Human, Tanusstein, Germany).

Genotyping by PCR restriction fragment length polymorphism. DNA was extracted by standard phenol-chloroform extraction. A 125-bp fragment of exon 1 including codons 52, 54, and 57 was amplified with primers(28), and digested with BamI, HhaI, MluI, and MboI, as described, to detect the B, C, and D alleles at codons 54, 57, and 52 respectively. A 349-bp fragment of exon 1 was also amplified(28) and digested with MboII to more accurately detect the C alleles.

Internal controls were developed for the C and D alleles, as their incidence is rare. PCR-amplified products from the KatG gene of M. tuberculosis(41) were added to the reaction mixtures before restriction enzyme digestion to provide a positive control for optimal digestion by the respective enzymes. A 223-bp product is digested by MboII to give 165-bp and 58-bp products, providing a control for the C allele, and a 221-bp product is digested by MluI to give 178-bp and 43-bp products, serving as a control for the allele D digestion. No other cutting sites for these enzymes were present in the amplified regions.

Statistical analysis. Median MBP concentrations among the different groups of individuals were compared by the Mann-Whitney U test(42). It is important to note that heterozygotes who are B, C, or D alleles have one tenth the level compared with wild-type as disruptions in the collagen region act as dominant mutations. For this reason, the levels of MBP in all populations studied are not normally distributed, hence the median rather than the mean and SD is how this heavily skewed serum range is routinely reported(28,29,31). The required sample size for genotyping was established using Epi-Info version 6. Association between genotype and disease (AB/BB and AC versus wild-type) was determined by the Mantel-Haenzel χ2 test(43). Fisher's exact test was used to compare genotype AD and wild-type. The OR was used to approximate the risk of disease with 95% CI calculated using the logit method(44). A two-tailed p < 0.05 was considered statistically significant. Data analysis was performed using version 6.11 of SAS for Windows (SAS Institute Inc., Cary, NC).

RESULTS

MBP plasma levels were determined in a total of 420 individuals by ELISA. The median MBP concentration in each group is shown in Table 1. During active disease, MBP levels are known to increase as part of the acute-phase response(12,15) and this was seen when comparing the median MBP levels of individuals who had a prior history of TB (previous TB) or prior history of TBM (previous TBM) with those sampled during active TB or active TBM. Control individuals had no history of TB despite living in the community where one in three houses had at least one case of TB in the period 1985-1994. This group showed a significantly lower (p = 0.004) median MBP level compared with the previous TB group and also significantly lower (p < 0.001) median level compared with the previous TBM group. The higher median MBP level in the previous TBM group compared with the control group is not a nonspecific consequence of the acute-phase reaction as all individuals studied had been sampled at least 15 mo (mean, 8.4 ± 4.5 y) after diagnosis of TBM and start of treatment. To confirm that the high MBP levels in the previous TBM group were not a result of an acute-phase response, plasma samples of individuals younger than 20 y were tested for the presence of CRP, a marker of the acute-phase response(45). The majority of samples had a CRP concentration of less than 6 mg/L (the lower detection limit of the assay). CRP was detectable in 1 of 7 samples in the control group, in 0 of 12 samples in the previous TB group, and in 3 of 30 samples in the previous TBM group, indicating a negligible contribution of the acute-phase reaction to the high MBP levels. These results suggested that low levels of MBP are protective against the development of TB and spread.

Table 1 MBP plasma levels
Full size table

We next determined the frequency in this population of the four MBP alleles, A (wild-type), B (G54D), C (G57E), and D (C52R) alleles, by PCR and restriction fragment length polymorphism (n = 234), shown in Table 2. Alleles B, C, and D were found only in the heterozygous form, except for a single BB homozygote. As expected, the presence of B, C, or D even in the heterozygous form had a dramatic effect on the MBP concentration(28,29) (results not shown). Almost all samples from heterozygous individuals have MBP concentrations below the median of that found in serum from wild-type AA individuals.

Table 2 Population frequency of MBP alleles
Full size table

To investigate the possibility of an association between disease (TB and TBM combined) and the presence of the wild-type genotype (AA) as opposed to any variant genotype (AB, BB, AC, and AD), χ2 analysis was done. This revealed a significant difference (p = 0.017, Mantel-Haenszel test) in the presence of disease between these genotype groupings. The different variant heterozygous genotypes were then compared individually with wild-type (AA) to determine which genotype was responsible for this difference. The single BB genotype (a TB individual) was included with the AB group. The analysis revealed that AB/BB is the genotype associated with protection against TB (Table 3). The OR for individuals of genotype AA having TB or TBM compared with those of genotype AB/BB was 3.21 (95% CI, 1.56-6.59; p = 0.001). No such association was found with the AC genotype, and the frequency of AD was too low for meaningful conclusions to be drawn. After stratifying the patients into TB and TBM, comparison of the AB/BB genotype with wild-type showed an association between the absence of the B allele and both pulmonary TB (OR, 2.59; 95% CI, 1.17-5.76; p = 0.017) and TBM (OR, 4.69; 95% CI, 1.64-13.41; p = 0.002). The stronger association between the absence of B allele and TBM suggests that MBP may facilitate the spread of TB via the blood stream.

Table 3 Association of MBP genotype with TB
Full size table

DISCUSSION

In this study, we found a highly significant association between MBP B allele and protection against the development of TBM (OR, 4.69; p = 0.002). This form of MBP has a mutation of glycine→aspartic acid at codon 54 within the collagen region, which disrupts the assembly of the protein. The net effect is reduction in circulating levels of MBP. Thus, if low levels of MBP are protective against the spread of TB beyond the lung, we would expect that TBM patients would have higher baseline MBP serum levels than controls who were exposed but did not develop disease. Our results are consistent with that premise. We have shown that the intrinsic baseline MBP levels in people who have fully recovered from TBM are significantly higher than those in people who have recovered from pulmonary TB, which are in turn higher than those seen in people who show some resistance to TB insofar as they have never had any form of TB disease (controls). Those with active disease show the acute-phase response and in each case have higher MBP levels than recovered patients, i.e. active TB is higher than previous TB, and active TBM is higher than previous TBM. Our hypothesis is that individuals with lower levels of MBP appear to be protected to some extent against developing TB. Conversely, those with higher baseline levels of MBP are more at risk of developing TB, and even more so of experiencing hematogenous spread and therefore TBM.

These findings support the idea that MBP may have a dual role in host defense(18,31). On the one hand, MBP protects against infection with encapsulated organisms especially within the first 18 months of life before the anticarbohydrate humoral response has matured(18), whereas on the other hand, MBP may facilitate the uptake and survival of intracellular pathogens such as TB within macrophages(31). This line of reasoning is supported by the fact that MBP serum levels were found to be significantly higher in Ethiopians infected with the intracellular Mycobacterium leprae bacillus than in a population of healthy blood donors. Similarly, recent studies on Leishmania chagasi, another obligate intracellular parasite, suggest that high levels of MBP increase susceptibility to development of symptomatic disease (visceral leishmaniasis) with this blood-borne pathogen (I.K.F. Santos, C. Costa, H. Krieger, M. Feitosa, D. Zurakowski, B. Fardin, R. Gomes, D. Weiner, R.A.B. Ezekowitz, D. Harn, and J. Epstein, manuscript in preparation).

Accordingly, MBP may facilitate intracellular tropism of the mycobacterium within macrophages. It may well be that complement-MBP-mycobacterial complexes provide a vehicle for dissemination by enhancing uptake by macrophages as they migrate in the blood. In this regard, MBP has been shown to bind and opsonize Mycobacteria, resulting in increased uptake by phagocytes(17). The MBP B allele (G54D) may bind M. tuberculosis inefficiently, resulting in ineffective opsonization, or there may be less MBP available in the serum. Furthermore, the B allelic form of the protein is known to abrogate complement activation as it fails to bind MBP-associated serine protease(46), which is a key factor for activation of MBP-mediated complement activation. The relevance of failure to engage the complement cascade to the pathogenesis of TB is congruent with the recent observation that complement cleavage products play an important role in enhancing mycobacterial recognition by macrophages(47). Together, all these findings support the idea that complement cleavage is an important mechanism used by mycobacteria to enhance intracellular tropism. What is not known is how MBP, alone or in conjunction with complement, influences the route of entry into the macrophage and whether the portal of entry influences the relative survival of the organism within the macrophage. It has recently been postulated that lack of the second complement component, C2, may confer some protection against TB(48). An interesting speculation in this regard is whether MBP B allele and C2 deficiency together may afford more protection than either alone.

Bellamy and colleagues(49) recently reported that the MBP C allele was only weakly associated with protection against pulmonary TB (p = 0.037) in Gambians compared with our findings in pulmonary TB (p = 0.017). However, this study did not examine the relationship between MBP mutations and TBM. It is important to note that the AC genotype did not appear to confer protection against disease in our study; however, the frequency of the C allele is much lower in the population we studied compared with that in Gambians (0.06 versus 0.27). The predominant mutant MBP allele in the population we examined was MBP B (frequency, 0.11). The findings of others that high serum MBP levels are associated with mycobacterial disease in Ethiopia(31) and that the C allele, which causes low serum MBP levels in Gambians, is associated with protection against TB(49), plus our own findings that high serum MBP is associated with TB and more strongly with TBM, would tend to support the hypothesis that it is MBP itself rather than a gene in linkage disequilibrium that is the cause of the association. The particular allele of a causative gene that is associated with a disease may be different in different populations(50). It is also possible that the products of the B and C allele are functionally distinct. We have expressed recombinant B allele protein and demonstrated that the protein is unstable(40) and unable to bind MBP-associated serine protease. Similar studies have not been performed with the C allele, and thus, although levels of the protein may be lower, the protein may be functional.

Our results suggest that the MBP gene may be regarded as one of the candidate genes that confers susceptibility to TB. Our hypothesis is that MBP plays a role particularly after infection in aiding the hematogenous spread of the disease to distant sites. It is likely that a number of gene products play a role in regulating the growth of the organisms within the macrophage, but the portal of entry may be critical in defining the ultimate fate of the organisms within the cell. The enhanced uptake to a privileged site may lead to a relative failure to contain the intracellular growth within macrophages. This may be the determining factor as these cells travel to the brain where they would have to negotiate the blood-brain barrier and seed the meninges. The availability of animals that have an MBP-null phenotype will greatly assist in providing further insights into this question.