Background
Tuberculosis (TB) remains a major health problem in most of sub-Saharan Africa. In March 2008, the incidence and mortality associated with tuberculosis in Uganda were 355 cases and 84 deaths per 100,000 population per year, respectively [
1]. In this setting, a further danger lies in the existence of TB/HIV co-infected individuals. Kampala, the capital city of Uganda, has a population of over two million people (National Housing and Census Survey, 2002) and carries 30% of the TB burden of the country (National TB and Leprosy Programme records). Records from the National referral hospital, Mulago, Kampala, show that 50% of TB patients are co-infected with HIV. Rubaga, one of the five administrative divisions surrounding the business capital of Kampala, has a resident catchment population of about 178,309 people and an average population density of 4,000/sq km (projection from the National Housing and Census Survey, 2002). About 250–300 TB cases are diagnosed from the TB clinics in the division quarterly (National TB and Leprosy Programme records). Poor housing conditions and overcrowding in most parishes of the division could facilitate the transmission of TB in the communities.
Molecular genotyping tools have enhanced our understanding of the epidemiology of TB by providing insight into the transmission dynamics, source and spread of
M. tuberculosis [
2,
3]. By helping to differentiate an ongoing exogenous infection from reactivation, these tools also help in evaluating TB control and treatment programmes in a given setting [
4]. One such genotyping tool, IS
6110 Restriction Fragment Length Polymorphism (IS
6110-RFLP) analysis, exploits variability in both the number and genomic position of the transposable element IS
6110 to generate strain-specific patterns after
PvuII restriction digestion of genomic DNA [
5]. Because it has been observed that the molecular clock for IS
6110-based RFLP runs at a faster rate than that for the PCR-based Spoligotyping and Mycobacterial Interspersed Repetitive Unit (MIRU) typing techniques [
6], IS
6110-RFLP is still referred to as the gold standard molecular epidemiological tool for TB in spite of its being labour intensive. In this study, we have used standard IS
6110-RFLP [
5] to determine the rate of recently transmitted TB in Rubaga division so as to understand the dynamics of transmission of the disease in this community. The findings are related to the HIV sero-status of the patients, socio-demographic characteristics of the patients, as well as drug susceptibility pattern of the isolates.
Discussion
The identification and differentiation of strains of
M. tuberculosis by IS
6110-RFLP has provided a better understanding of the epidemiology and tracing the transmission of TB in developed countries. However, not much is known in most developing countries where the disease burden is highest. Previous studies done in some sub-Saharan African countries have reported contrasting results in regard to risk factors for the transmission of TB [
10‐
13]. In this study, we have analyzed the transmission dynamics of TB in HIV seropositive and seronegative TB patients, and within age groups so as to study the association between these variables and ability to transmit TB in a high disease burden peri-urban setting of Uganda.
Analysis of clusters in our sample showed 118 patterns from the 183 strains fingerprinted. A large proportion (99/183 or 54.1%) of our sample belonged to clusters, suggesting a high prevalence of transmission. Even after removal of clusters containing strains with less than five copies of IS
6110, our result differs from that shown in a previous finger print analysis of 73 isolates from different villages of Kampala [
14] in which it was shown that only six (8.2%) of the isolates typed clustered, suggesting that multiple strains were transmitted in the community. It is generally known that high disease burden settings are characterized by a limited variety of
M. tuberculosis strains and a high proportion of isolates occurring in clusters [
15,
16]. The difference in the two results above, moreover from the same setting, could be due to the sampling strategy: while we sampled resident patients at TB clinics in a single division of Kampala, the previous study was based at the National referral hospital, with patients originating from villages of different divisions of the city. It is known that measured clustering depends on such factors as completeness of sampling, immigration and time period [
17]. While our previous study sampled 2,639 suspects from a resident population of a division of Kampala, only 344 (13%) turned out smear positive and were therefore recruited [
8]. Since smear microscopy has a detection limit below 50% in our setting (National TB Reference Laboratory records, Uganda), some TB cases could have been missed and therefore the true proportion of clustering in our study may have been underestimated. Additionally, in an attempt to compare an almost equal sample of HIV positive and HIV negative TB patients, only 183 of the 344 isolates (53.2%) were analyzed for clustering in the present study, a further limitation to estimating the actual clustering rate in the community studied.
Previous studies have shown varying clustering proportions dependent on length of the sampling period in settings with similar disease burdens. In South Africa for example, a one year study showed a proportion of clustering of 50% among gold miners [
18]. This result is close to that obtained in the present study in which 54.1% of the isolates obtained over a 9 months period clustered. These results, however, differ from those obtained in longitudinal studies over years as they allow a greater number of exposed persons to progress to active TB. Cases in point are clustering rates of 72% in a six year study in Cape Town, South Africa, [
13] and 73.6% in an area with a high HIV prevalence in Northern Malawi [
11]. In a more recent analysis, it has been shown that both epidemiologic and strain related factors may contribute to cluster size [
19]. It was further shown that spoligotype 59 in a Malawi population in all groups of patients, was clustered and unique, and was associated with a wide diversity of RFLP patterns. This is in agreement with our previous observation of predominance of RD 724 deleted – T2 spoligotypes of
M. tuberculosis in this locale [
8].
Our study shows that the average cluster size of 2.9 in our sample was twice that observed in another high burden setting, Tanzania, where Yang and colleagues [
20] showed an average cluster size of 1.33. This difference in average cluster size between ours and the Tanzania study may be due to the fact that they sampled the city of Dar es Salaam, with patients originating from other parts of Tanzania as well, while in our study only the resident population was considered and being a non-resident of Rubaga division was an exclusion criterion. The high average cluster size in our study however, is comparable with the 3.7 observed in isolates from Addis Ababa residents in Ethiopia [
10].
Only 10 of the 183 isolates were resistant to isoniazid, including five that were resistant to rifampicin as well, hence multidrug resistant. All the isolates from the sample analyzed were from newly presenting TB patients; hence the resistance of these isolates was considered primary drug resistance. There was no significant difference in the frequency of drug resistance between patients with clustered genotypes and those with unique genotypes (Table
2). Although the number of patients with strains resistant to any drug in our sample was small, this result contrasted with findings in Ethiopia in which 62.5% of 17 isolates resistant to any drug were found within clusters [
10].
The relationship between clustering and sex as well as between clustering and age groups was also analyzed. Generally, as was observed in a study in Botswana [
21], there was no statistical link between the risk of a patient belonging to a cluster and any of the demographic characteristics above (Table
2). These results further differ from observations in Ethiopia [
10] in which there was a trend toward increased clustering of isolates from tuberculous women residing in Addis Ababa. In northern Malawi, on the other hand, there was a weak association between clustering and sex, with clustering more common in women ≥ 45 years of age [
11]; while in South Africa, clustering was not associated with sex but clustered cases seemed to decrease with age, though not significantly [
13].
The IS
6110 copy number of the isolates ranged from 1 to 20, which is similar to that in the Tanzania study [
20] and close to the range of 1 to 17 copies in the earlier study in Kampala [
14]. These results are also comparable with that from Tunisia [
22] where 75% of the isolates were high copy number. However, there were significant differences in the distribution of the copy number between our strains and those in the Tanzania sample in that the proportion of strains with less than six copies (considered low copy number strains) in this study was 3.8% (Figure
2) and 4.1% (3/73) in the previous Kampala study [
14], while that in the Tanzania study was 26.1% (35/134). In fact, the isolates carrying only a single copy of the IS
6110 element in the Tanzania study was 8%, while only 2.7% (5/183) of our strain collection and 2.7% (2/73) of the previous Kampala study harbored a single copy of the element. More interestingly, the Tanzania study found a clear correlation between the occurrence of low copy number of IS
6110 and classification of the strains into the Asian subgroup, while in our previous analysis of the full set of 344 samples by spoligotyping, we observed that 70% (241/344) of our strain collection was of the T2 type (Euro-American) while only CAS1-Kili (3.5%) and CAS1-Delhi (2.6%) belonged to the Asian subgroup [
7]. These results further confirm observations that statistically significant association of a clone to the prevalence of disease in a community may reflect increased adaptation and fitness of the strain type [
23].
A majority (96.2%) of the strains in our sample and 95.9% of strains from the earlier Kampala study were high copy number, with histograms of the distribution of IS
6110 copies from both studies skewed to the right. In northern Malawi and Botswana, on the other hand, histograms showed a normal distribution for the IS
6110 copies. It has been observed elsewhere that high IS
6110 copy strains can be epidemiologically resolved using a smaller set of loci on MIRU-VNTR genotyping panel [
24]. Our data therefore suggests that it is possible to quickly establish epidemiological links for this group of strains using a less laborious and high throughput PCR-based technique.
A comparison of the RFLP patterns of HIV-related and non-HIV-related isolates showed no significant differences between the categories. The risk of belonging to a group of patients with identical RFLP patterns was not significantly different between the two categories (Table
2). A number of studies about clustering in TB/HIV co-infected patients have yielded contrasting results. In one of these, in a high disease incidence region of Spain, a study of 305 TB patients with known HIV serostatus showed that strains isolated from HIV-positive patients were not associated with clustering [
25] a result similar to observations by Yang
et al [
20] in Tanzania. More studies in sub-Saharan Africa have pointed to absence of association between HIV serostatus and clustering. A study in a high HIV prevalence district of northern Malawi did not find any association between clustering in TB patients and HIV serostatus [
19]. A similar result was obtained in a population-based prospective study of pulmonary TB patients in Botswana, in which HIV positivity was not associated with clustering even after stratification by site [
21]. In Ethiopia, on the other hand, HIV positive serostatus was significantly associated with clustering of isolates for patients of both sexes [
10], a result similar to that obtained in Lima, Peru, using fluorescent Amplification Fragment Length Polymorphism (fAFLP), in which the highest levels of relatedness were found among isolates from the same disease group, with the isolates grouping into two distinct clusters [
23]. It may argued that observation of clonal groupings among the AIDS-associated isolates in Peru may have been due to the effect of hospitalization of the patients since all the 25 isolates from HIV-1-negative individuals were found to be relatively heterogeneous and nonclonal.
Since a majority (144/183) of the patients in our sample was aged between 18 – 39 years inclusive, the absence of association between HIV serostatus and
M. tuberculosis clustering in this study may be due to a very high proportion of disease attributable to recent transmission in HIV negative patients in the younger category, as has been postulated elsewhere [
11].
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
BBA participated in the planning of the study, acquisition of samples and demographic data, culture and isolation of mycobacteria, RFLP assays, data analysis and drafting of manuscript; SG participated in running RFLP, data analysis and critical revision of manuscript; FAK, DPK participated in isolation of cultures and critical revision of the manuscript, TK participated in general supervision of the research in Sweden and critical revision of the manuscript; RP, AP paritcipated in running RFLP; GK & MLJ participated in the conception, design and general supervision of the study and critical revision of the manuscript. All authors read and approved the final manuscript.