Progression to active tuberculosis, but not transmission, varies by M. tuberculosis lineage in The Gambia

Abstract

Considerable variability exists in the outcome of M. tuberculosis infection. We hypothesized that M. africanum was less likely than M. tuberculosis to transmit and progress to tuberculosis disease.

In a cohort study of tuberculosis patients and their household contacts in the Gambia, we categorized 1,808 HIV negative tuberculosis contacts according to exposure to M. tuberculosis or to M. africanum. A positive skin test indicated transmission and development of tuberculosis during 2 years of follow-up indicated progression to disease. Transmission was similar, but progression to disease was significantly lower in contacts exposed to M. africanum than to M. tuberculosis (1.0% vs 2.9%; Hazard Ratio (HR) 3.1, 95% CI 1.1–8.7). Within M. tuberculosis sensu stricto, contacts exposed to a Beijing family strain were most likely to progress to disease (5.6%; HR 6.7 (2.0–22) relative to M. africanum).

M. africanum and M. tuberculosis transmit equally well to household contacts, but contacts exposed to M. africanum are less likely to progress to tuberculosis disease than those exposed to M. tuberculosis. The variable rate of progression by lineage suggests that TB variability matters in clinical settings and should be taken into account in studies evaluating tuberculosis vaccines and treatment regimens for latent tuberculosis infection.

Introduction

Tuberculosis remains a significant public health problem, particularly in resource limited settings [1]. The considerable variability in the outcome of M. tuberculosis infection has traditionally been attributed to host and environmental factors [2, 3] and indeed host immune suppression, such as caused by HIV, is the strongest known risk factor for development of active tuberculosis [4]. However, pathogen-related factors may also play a role [5]. While distinct genotypes have been identified in the M. tuberculosis complex, it is unclear whether these translate into phenotypic differences in humans [6].

Significant strain differences between M. bovis and M. tuberculosis were identified early in the history of mycobacteriology [7], and several experimental studies have shown that M. tuberculosis strains can differ in immunogenicity and virulence in animal models [8]. Recent advances in genotyping now allow more detailed analyses of the contribution of bacterial factors to the variability in transmission and progression to tuberculosis disease in its natural human host [9]. In population based studies these techniques have identified clustered isolates of M. tuberculosis, i.e. with identical genotypes [10, 11]. While clustering is suggestive of recent transmission, these studies were unable to distinguish between rates of transmission and of progression to tuberculosis disease. A different study described an outbreak strain, associated with high rates of tuberculin skin test (TST) conversion in contacts compared to historical controls [12]. In the context of a household study of tuberculosis in the Gambia, where M. africanum is endemic, we assessed the likelihood of transmission and of progression to tuberculosis disease according to mycobacterial lineage. We specifically tested two hypotheses: that M. africanum would be less likely than M. tuberculosis to be transmitted and less likely to cause disease.

Methods

In the Tuberculosis Case Contact (TBCC) study, we followed 317 adult sputum smear positive tuberculosis index cases and 2,381 of their household contacts. Participants were recruited between September 2002 and September 2004 and were followed for two years. Household members were eligible for inclusion in the study if they had been sleeping in the same compound (walled group of houses) as the index case during the index case’s period of illness with tuberculosis. Household contacts had a TST (PPD R23 2TU, Staten Serum Institute, Denmark) placed using the Mantoux technique. Those with a negative TST (induration < 10mm) had a repeat test after 3 months. There was no practice of treatment of asymptomatic TST positive persons for latent infection in The Gambia.

Follow-up

Five follow-up visits (at 3, 6, 12, 18 and 24 months after enrolment) were made to each of the 317 households. Any participant who reported tuberculosis symptoms at these visits was encouraged to present to the MRC tuberculosis clinic and had free access to treatment for any illness during this period. At each household visit we re-evaluated each individual for symptoms of tuberculosis. Any patients with symptoms of pulmonary disease received a chest radiograph and sputum analysis (three samples) for acid fast bacilli (AFB) smear analysis and culture. If tuberculosis disease was bacteriologically confirmed or clinically suspected in smear-negative or extra-pulmonary cases, patients were referred for the standard six month tuberculosis treatment course at the Gambian National Tuberculosis Treatment program. A diagnosis of tuberculosis disease among household contacts during the 2-year follow-up period was used as the main outcome variable for the analysis on progression to tuberculosis by mycobacterial lineage.

Case definition

All contacts with symptoms consistent with tuberculosis (fever, night sweats, persistent cough), or with a positive TST at enrolment or at the 3-month follow-up visit, were offered a chest radiograph and three sputum tests if they had a productive cough. Based on results of the chest radiograph, sputum smear and -culture results, and/or their response to a trial tuberculosis treatment course using the standard 4 drug regimen contacts were classified as non-diseased or diseased (“secondary cases”, see definitions in the section on statistical analysis). A positive TST was not an essential element of the case definition.

In addition to the identification of secondary cases presenting at MRC and during follow-up visits, the names and ages of all tuberculosis cases treated at the government health clinics during the course of the study were recorded. Those that matched with contacts participating in our study, using an age category matching within 5 years of the stated age on the government record, were re-visited to confirm whether or not they received tuberculosis treatment. Those that confirmed treatment were asked to have a chest radiograph to look for evidence of tuberculosis. The radiographs were reviewed by two physicians experienced in infectious diseases, and a pediatrician if the participant was a child. After review, a consensus opinion was formed on the presence or absence of tuberculosis. Contacts with symptoms compatible with tuberculosis but insufficient information to support a diagnosis of tuberculosis for this study were classified as “unconfirmed secondary cases”.

Laboratory procedures

Sputum from index cases and symptomatic contacts was examined for AFB using the auramine and Ziehl Nehlsen methods. Decontaminated sputum was cultured both in liquid media (Bactec 9000, Becton Dickinson) and on Lowenstein Jensen slopes, prepared as described previously [13]. Symptomatic contacts who presented directly to the Gambian Government TB clinic had sputum examined for AFB but not stored for culture and genotyping.

HIV testing was restricted to those who either wanted to know their result or were selected for full immunological testing by skin test and ELISPOT. Those individuals who tested positive for HIV-1 and/or HIV-2 were referred to the onsite MRC HIV clinic, where they were eligible for antiretroviral therapy.

Genotyping methods

Clinical isolates were molecularly characterized using spoligotype analysis [14], resulting in a binary pattern based on the presence or absence of 43 spacers. In addition, we performed PCRs for Large Sequence Polymorphisms [15] based on phylogenetically informative Regions of Difference (RD). This resulted in classification in one of 12 lineages, including the sub-species M. africanum (RD702) and 11 lineages within M. tuberculosis sensu stricto. The main lineages within M. tuberculosis sensu stricto included the Beijing family (RD105), lineage RD174 with a deletion in the DosR regulon, and lineages RD182 and RD219.

Genotypic definition

Recently, molecular techniques have identified two lineages within the true M. africanum type I: West African type 1, phylogenetically closer to M. tuberculosis and predominantly found around the gulf of Guinea, and West African type 2, phylogenetically closer to M. bovis and confined to the Western parts of West Africa [9, 16]. We refer to M. africanum type I, West- African type 2 when using “M. africanum” in this manuscript.

Identical spoligotype patterns (genotypes) in isolates of an index case and a diseased contact from the same household were referred to as ‘concordant genotypes’, different patterns as ‘discordant genotypes’. Isolates sharing the same spoligotype pattern except for one of the 43 spacers, were classified as ‘concordant’ because these patterns may reflect evidence of direct transmission between household members.

Ethical approval

The study was approved by the Gambian Government/MRC ethics committee and the Stanford University Institutional Review Board. All participants provided informed consent prior to enrollment.

Definitions

Contacts were classified as “prevalent TST positive” if they tested TST positive (≥10mm) at recruitment and as “incident TST positive” if they tested TST negative (<10mm) at recruitment and positive (≥10mm) at the 3-months follow-up visit, with induration ≥ 6mm increased since recruitment.

Contacts who developed TB disease (“secondary cases”) were classified as “co-prevalent tuberculosis” if they were diagnosed with tuberculosis at recruitment or within 3 months since recruitment, or “incident tuberculosis” if they were diagnosed 3 months or more since recruitment.

Statistical Analysis

Analysis was done stratifying by sub-species (M. africanum vs M. tuberculosis) and by lineage within the sub-species, where lineages with information on less than 70 contacts were grouped as ‘other’. In the absence of complete mycobacterial genotype data on the diseased contacts, we did not correct the analysis on the secondary cases by lineage for the likelihood of infection with a discordant genotype. Prevalent positive TST in household contacts and incident TST at 3 months, were used as outcome variables for transmissibility of the different mycobacterial lineages. After confirming that HIV was an independent predictor of lower TST indurations and of increased progression to disease, both analyses on transmission and on progression were limited to known HIV-uninfected contacts. We used the χ² - test to test for differences in TST positivity and progression to disease between contacts exposed to the different mycobacterial lineages. We calculated odds ratios (OR) and their corresponding 95 percent confidence intervals (95% CI) for having a positive TST using logistic regression, controlling for household clustering. Possible confounders such as age, presence of a BCG scar and sleeping proximity were added to the model. We used Cox regression models to estimate the hazards ratio (HR) and 95% CI of secondary cases among the contacts of M. tuberculosis index cases versus M. africanum index cases, controlling for household clustering, after confirming compliance with the proportional-hazards assumption.

We performed several different sensitivity analyses to assess potential biases. A degree of misclassification bias occurred, as not all secondary cases were infected by their index case. To determine whether misclassification bias was important, we removed the secondary cases known to be infected with a discordant genotype from the analysis. The potential effect of unconfirmed tuberculosis disease was assessed by repeating the analysis with unconfirmed secondary cases classified either as non-diseased contacts or as secondary cases. The potential effect of the inclusion of co-prevalent secondary cases in the analysis on progression was assessed by recalculating the hazard ratio limited to incident secondary cases.

All analyses were conducted using Stata (version 9, Stata Corp., College Station, Texas, USA).

Results

Study participants

In a tuberculosis case contact study, cultured isolates were successfully obtained from 301 (95%) index cases, of which 291 (97%) had interpretable genotyping patterns. These 291 cases had 2,174 household contacts, of whom 1,853 (86%) underwent HIV testing with 45 (2.4%) positive results; the 1,808 HIV uninfected contacts were eligible and included in this study. At recruitment, the TST was administered to 1,727 (96%) contacts, of whom 708 (41%) tested positive. Of the 1,808 HIV-uninfected contacts, 48% were male, 49% had a scar from Bacille Calmette Guerin (BCG) vaccination, and the median age was 16 (interquartile range 8–25).

Mycobacterial genotypes

The 291 isolates included 12 different lineages within the M. tuberculosis complex: 110 isolates of M. africanum (one lineage) and 181 isolates of M. tuberculosis sensu stricto (11 lineages). The genotypic findings are described in detail elsewhere (de Jong B, et al., submitted).

Follow-up

The HIV uninfected contacts had 3,301 person years of follow-up in total. This yielded 19 contacts with “co-prevalent tuberculosis” (presenting within three months of enrolment) and 20 contacts with “incident tuberculosis” (who developed symptoms consistent with tuberculosis more than three months after enrolment), for a total of 39 secondary cases. The incident tuberculosis cases have recently been described in detail elsewhere [17]. Their median age was 17 years (range 2 to 70) and 22 (56%) were males. Six contacts had extra-pulmonary tuberculosis and 33 had pulmonary tuberculosis. Four secondary cases were identified through the records of the government TB clinic, with x-ray findings suggestive of (treated) tuberculosis. Eight contacts had symptoms but insufficient information to support a diagnosis of tuberculosis, including three who had received TB treatment but had a normal chest x-ray afterwards; they were excluded from the analysis of the progression to disease.

Paired genotyped isolates of index case and secondary case were available for 18 of the 39 secondary cases. Eleven of these 18 paired isolates had identical genotypes, one secondary case isolate missed one additional spacer out of 43 compared with the index case isolate, and six had different genotypes.

Transmission by mycobacterial lineage

The prevalence of household contacts with a positive TST result at enrollment was not significantly different between M. africanum and M. tuberculosis (39% versus 42%, p=0.14), or between the lineages within M. tuberculosis (table 1). Similarly, the incidence of TST conversion did not differ between lineages or between M. africanum and M. tuberculosis (25% versus 26%, p=0.70). Adding known predictors of TST positivity, such as age and sleeping proximity to the index case, to the logistic regression model did not significantly change these results.

Table 1.

Tuberculin skin test results in HIV-uninfected contacts by lineage within the M. tuberculosis complex

Lineage (also known as)	Baseline*			Conversion at 3 months†
	Number of contacts	% TST positive n (%)	OR (95% CI)	Number of contacts	% TST positive n (%)	OR (95% CI)
RD702 (M. africanum)	653	253 (39)	1	280	69 (25)	1
RD105 (Beijing)	72	27 (38)	0.90(0.33–2.4)	24	6 (25)	1.3(0.31–5.7)
RD174 (DosR)	200	84 (42)	1.0(0.55–1.9)	91	26 (29)	1.2(0.49–2.7)
RD182	311	136 (44)	1.3(0.77–2.3)	127	32 (25)	(0.48–2.4)
RD219	84	39 (46)	1.4(0.58–3.3)	33	5 (15)	0.39 (0.09–1.7)
Other	407	169 (42)	1.2(0.76–2.0)	188	51 (27)	1.2(0.59–2.3)
Total	1,727	708 (41)		743	189 (25)

TST = tuberculin skin test

RD = region of difference; major phylogenetic groups are identified by major large sequence polymorphisms or deletions

^*= TST positivity at enrollment

^†= TST positivity at 3 months in those that tested TST negative at baseline, with conversion defined as ≥ 6 mm increase in induration.

OR= Odds Ratio

CI= Confidence Interval

Progression to tuberculosis disease by mycobacterial lineage

Comparing M. tuberculosis sensu stricto (all 11 lineages combined) with M. africanum, progression to tuberculosis was 2.9% in household contacts of index cases with M. tuberculosis sensu stricto compared with 1.0% in those with M. africanum (HR 3.1, 95% confidence interval (CI) 1.1–8.7). Figure 2 depicts the results using survival curves, showing significant divergence over time (p=0.036).

Figure 2.

Survival curves showing survival free of tuberculosis after exposure to M. africanum versus M. tuberculosis, based on a Cox regression model.

Table 2 shows the likelihood of progression to disease by lineage (p=0.033 across lineages). Household contacts whose index case had tuberculosis caused by a Beijing family isolate of M. tuberculosis were more likely to develop tuberculosis disease (5.6%) than contacts of an M. africanum case (1.0%, Hazard Ratio (HR) 6.7, 95% CI 2.0–22, table 2), but were not more likely to develop disease than other lineages within M. tuberculosis sensu stricto. Contacts of an index case with the lineage within M. tuberculosis sensu stricto defined by RD174, a deletion in part of the DosR regulon, were the second most likely to progress to disease (3.9% in two years, HR 4.4, 95% CI 1.4–14).

Table 2.

Incidence of tuberculosis in HIV-uninfected household contacts by lineage within the M. tuberculosis complex

Lineage (also known as)	Primary progression
	Number of contacts	Secondary cases n (%)	HR (95% CI) *	HR (95% CI) *†	HR (95% CI) *‡
RD702 (M. africanum)	681	7 (1.0)	1 (reference)	1 (reference)	1 (reference)
RD105 (Beijing)	72	4 (5.6)	6.7 (2.0–22)	7.7 (1.9–31)	16 (2.8–89)
RD174 (DosR)	204	8 (3.9)	4.4 (1.4–14)	5.8 (1.5–23)	10 (2.2–45)
RD182	317	9 (2.8)	2.8 (0.84–9.5)	3.6 (0.92–15)	4.2 (0.77–23)
RD219	86	1 (1.2)	1.4 (0.16–12)	§	4.2 (0.42–43)
Other	440	10 (2.3)	2.4 (0.74–7.6)	3.5 (0.95–13)	3.2 (0.60–17)
Total	1,800	39 (2.2)

TST = tuberculin skin test

HR = hazard ratio

CI = confidence interval

RD = region of difference; major phylogenetic groups are identified by major large sequence polymorphisms or deletions

^*= with Random Effects model, to account for household clustering

^†= Excludes secondary cases whose isolate differed by one or more spacers from the index case isolate

^‡= Excludes co-prevalent cases, i.e. contacts diagnosed with tuberculosis within three months of enrollment

^§= no observations

Progression to tuberculosis in contacts with prevalent or incident positive TSTing occurred in 4.9% of contacts of index cases with M. tuberculosis sensu stricto vs 2.2% of those with M. africanum (p=0.046). Repeating the analyses while excluding the secondary cases infected with isolates that differed by one or more spacers from the index case isolate (n=7), the association between genotype and the number of secondary cases was stronger (HR 4.0 in M. tuberculosis- relative to M. africanum households, 95% CI 1.2–13, p=0.026; Table 2). Similarly, when the analyses were repeated excluding the co-prevalent cases (n=19), incident tuberculosis was lower in M. africanum contacts than in M. tuberculosis contacts (HR 5.5, 95% CI 1.3–23, p=0.020; Table 2). Lastly, when we included contacts who had an unconfirmed TB diagnosis as cases (Figure 1, n=8), the incidence of tuberculosis was lower in M. africanum contacts than in M. tuberculosis contacts (HR 2.7, 95% CI 1.2–6.5, p=0.022).

Figure 1.

Flow diagram of analyses

Discussion

While transmission of the M. tuberculosis complex from tuberculosis patients to their household contacts did not vary by strain lineages in The Gambia, the percentage of contacts with incident tuberculosis during the two year follow-up varied 5- fold between contacts exposed to the different M. tuberculosis lineages. These percentages ranged from 1.0% of contacts in households exposed to M. africanum to 5.6% of those exposed to the Beijing family within M. tuberculosis sensu stricto (p=0.033). The different rate of progression to active tuberculosis between lineages provides strong evidence that pathogen-specific factors are important in the observed variability in outcomes of tuberculosis infection and indicates that the bacterial determinants of transmission and initial infection are distinct from the determinants of progression to disease.

The finding that fewer HIV- negative contacts of M. africanum index cases have incident tuberculosis is consistent with other characteristics of this lineage. We previously reported that M. africanum- infected cases and their contacts were less likely to mount an IFN- γ response against ESAT-6 than M. tuberculosis-infected cases and contacts[18]. Esat-6, encoded in the RD1 region [16], is a known virulence factor [19] and is thought to play a role in the cell-to-cell spread of M. tuberculosis [20]. The sequence of the RD1 region of M. africanum showed a frame shift mutation in gene Rv3879c compared with the same region in M. tuberculosis and M. bovis. As a result, Rv3879c is a pseudogene in M. africanum [18]. A transposon mutant of the Rv3879c homologue in M. marinum, a mycobacterium species that also contains the RD1 region, showed undetectable levels of ESAT-6 in the culture filtrate but normal levels of ESAT-6 in cell lysate [21]. When Rv3879c is non-functional, ESAT-6 secretion may be impaired and this would explain the attenuated ESAT-6 response induced by M. africanum. So while reduced secretion of ESAT-6 may impair the bacterium’s cell-to-cell spread and possibly explain its lower progression to disease, the reduction in ESAT-6 secretion could facilitate immune evasion, balancing selective pressures on the pathogen’s evolution. These findings suggest a role for RD1 in within-host but not between-host interactions.

While the above provides a potential molecular explanation (i.e., a proximal cause) for why M. africanum exhibits slower disease progression relative to M. tuberculosis, it does not address the ultimate evolutionary reason for this difference. The development of active disease is a sine qua non for the organism to spread to new susceptible hosts. Ecological theory predicts that under such conditions, a trade-off between transmission and virulence (or latency) can emerge, and that increased access to susceptible hosts could favor increases in virulence and/or reduced latency [22–24]. We hypothesize that M. africanum became endemic in West-Africa prior to the introduction of M. tuberculosis through European contact; a time when human populations were comparably small in West Africa. This scenario is supported by the higher diversity in genotype patterns within the M. africanum lineage relative to the different M. tuberculosis lineages in The Gambia, and the dominance of the Euro-American variants among the latter (de Jong B, et al., submitted). Longer latency in M. africanum might be an adaptation to low host densities, whereas a reduced latency period (i.e. increased “virulence”) in M. tuberculosis infections might be an adaptation reflective of the crowded conditions and high rates of tuberculosis in European cities at the time of European colonization. Interestingly, we showed that M. africanum was associated with members of the Fulani tribe who are nomadic pastoralists (de Jong B, et al., submitted). Nomadic populations tend to be significantly smaller than populations of sedentary agriculturalists such as the Mandinka, the other major ethnic group in The Gambia.

The finding that the M. tuberculosis Beijing lineage is more likely to lead to tuberculosis during the two year follow-up than M. africanum is consistent with experimental results. Phenoglycolipids produced by Beijing isolates prevent mice from mounting an effective immune response [8]. Our findings suggest that the global dissemination of the Beijing lineage [26], indicative of relatively increased pathogenicity, results from a propensity for progression and not from enhanced transmission. The RD174 lineage, the second most likely to lead to tuberculosis among household contacts, has a deletion in the DosR regulon, a region which is up-regulated under in vitro conditions thought to mimic latency [27]. Isolates with the RD174 deletion lead to chains of transmission with shorter periods between successive cases in San Francisco (K DeRiemer, unpublished).

We previously demonstrated an attenuated ESAT-6 response in cases and contacts infected with M. africanum [18], which precluded use of ELISPOT positivity to assess transmission. Yet using the skin test is not without limitations. Prevalent TST positivity may reflect prior exposure to the M. tuberculosis complex, to environmental non-tuberculous mycobacteria, or prior BCG vaccination. Moreover, the “boosting phenomenon”, whereby an initial negative TST result is followed by a positive result due to the activation of memory T cells, may have accounted for some of the incident TST positivity at the three month follow up visit. However, we have no grounds to suspect that these confounders were differentially distributed between individuals infected with the different genotypic lineages, and the TST has proven to be highly specific for recent M. tuberculosis complex infection in The Gambia [28].

We have shown that, by assessing concordance of genotypes between secondary cases and index cases in those with an available isolate, the majority of secondary cases were infected by their respective index case: the source of M. tuberculosis was the index case 67% of the time. We expect misclassification bias from the other 33% of situations to lead to an underestimate of the relatively reduced rate of M. africanum progression to disease. Indeed, when we excluded the six diseased contacts known to have a discordant genotype and re-analyzed the data, the significance of the association between mycobacterial sub-species and progression to disease increased despite the lower number of secondary cases analyzed. In the analysis of progression to disease, we did not make a distinction between co-prevalent- and incident secondary cases. However, when limiting the analysis to incident secondary cases, i.e. those who developed tuberculosis 3 months or more after enrolment, the analysis remained significant despite the halving in numbers. While HIV-infected contacts were excluded from this analysis of mycobacterial factors affecting rates of transmission and progression to disease, the rate of progression to active disease was significantly higher among HIV-infected contacts (8.7% vs 2.3%, p=0.005) [17].

In summary, our findings confirm the importance of mycobacterial factors in the variability in progression to tuberculosis disease in humans. The contribution of pathogen factors to the outcome of mycobacterial infections has implications for understanding tuberculosis epidemics and the development and assessment of new interventions, specifically vaccines and treatments for latent tuberculosis infection. Further studies in other settings can identify specific lineages with high and low predilection for progression to disease, while detailed molecular analyses may provide fresh insights for the development of new interventions.