Introduction

Tuberculosis (TB) is the leading cause of mortality due to an infectious disease [1]. In 2017, an estimated 10 million people developed TB globally, with 1.6 million dying of the disease. Tanzania is among the thirty high burden countries, with a national average TB notification rate of 129 cases per 100,000; however, some regions show higher notification rates [2]. Like in most sub-Saharan African countries, the HIV epidemic contributes substantially to the high TB incidence in Tanzania, where a-third of the TB patients are co-infected with HIV [2]. Contrarily, drug-resistant TB is still low in this setting [3]. Other risk factors such as poverty also influence the epidemiology of TB in Tanzania [4].

Transmission of TB occurs via infectious aerosols, where upon exposure individuals can either clear the infection, develop active disease or remain latently infected [5]. The complex dynamics of TB infection and disease are determined by the environment, the host and the pathogen [6]. Seven main phylogenetic lineages of the Mycobacterium tuberculosis complex (MTBC) (Lineage 1–7) cause TB in humans [7]. These lineages are phylogeographically distributed, partially reflecting human migration histories [8–10]. Genomic differences among the MTBC strains translate into relevant biological and epidemiological phenotypes [11]. Epidemiologically speaking, these phenotypes are demonstrated by indicators such as transmission potential, disease severity and progression rates from infection to disease [12–15]. In general, strains of the widely distributed lineages, Lineage 2 and 4 or “generalists”, appear to be more virulent than those of the geographically restricted lineages, Lineage 5 and 6 or “specialists” [7,11].

Studying genotype-phenotype relationships requires understanding the genetic diversity of MTBC clinical strains in a given clinical setting. In Tanzania, few studies have described the genetic diversity of the MTBC [16–19]. These previous works used conventional genotyping tools such as the spacer oligonucleotide typing (spoligotyping) technique and revealed the presence of mainly the East African Indian (EAI), Central Asian (CAS) and Latin American Mediterranean (LAM) spoligo families, and the Beijing family reported only at the lowest frequencies. Based on phylogenetically robust techniques, which include single nucleotide polymorphisms (SNPs), the spoligo families correspond to Lineage 1, 3, 4 and 2, respectively. These previous studies from Tanzania are limited as they only focused on few specific geographical locations on the country and only one study profiled MTBC on a countrywide scale albeit with low sampling coverage [18]. Moreover, despite the invaluable contribution of techniques like spoligotyping in the molecular epidemiology field, such techniques are suboptimal for phylogenetically robust strain classification due to high rates of convergent evolution [20,21].

In this study, we applied for the first time a robust SNP typing method to classify the largest so far nationwide representative collection of clinical isolates to gain insights into unknown patterns of MTBC diversity in different regions of Tanzania. Given that only few studies have assessed and identified lineage-specific differences in clinical settings, we then looked for potential associations between the MTBC lineages and the available clinical and epidemiological characteristics of the patients in the study population.

Material and methods

Results

Patients’ demographic and clinical characteristics

The patients’ demographic and clinical information in our study included age, sex, geographical location, HIV and disease category (new or retreatment case). Table 1 describes patients’ characteristics of the study population. The proportions of the observed and missing data for the study population are summarized in S2 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Clinical and demographic characteristics of the TB cases.

https://doi.org/10.1371/journal.pone.0206334.t001

Our study population consisted of TB patients ranging between the age of 2 and 89 years with a median age of 28 years (interquartile range [IQR] 27–44). To further probe the age distribution in the study population, we categorized the TB patients into five different age groups (Table 1). We detected approximately three-quarters of the TB cases to occur among the “young age” and “early adult” age groups. Further, our findings show that about 1.0% of the TB cases were pediatric cases (< 15 years).

Similar to other settings [1], we identified a higher proportion of male TB cases compared to female TB cases. However, the male-to-female ratio observed in our study population was higher than the national estimates for the two years of sampling (2.2:1 vs., 1.4:1). The striking gender imbalance among TB cases seems to peak at adolescence onwards and is less pronounced among pediatric TB cases (S1 Table). Additionally, a-third (32.2%, 517/1604) of the TB cases with available HIV status were HIV co-infected. In contrast, TB/HIV co-infected cases were more likely to be female (44.5%, CI 38.3–50.7% vs., 25.8%, 95% CI 20.6–31.0%) which is consistent with HIV being generally more prevalent in females than males in Tanzania [28]. We found that our study population comprised 16.1% (321/2000) of TB retreatment cases, which was four-fold higher than the overall countrywide notifications [29]. Finally, more than half (51.6%, 1029/1996) of the TB patients in our study population were diagnosed in the Coastal zone of Tanzania and about 40% were either diagnosed in the Lake and Northern zones. In addition to higher TB notification rates, the three former mentioned geographical zones contain the country’s zonal TB reference laboratories. The remaining 10% of the patients were diagnosed in any of the remaining four geographical zones of Tanzania.

Main MTBC lineages in Tanzania

Using SNP-typing, we detected four of the seven known MTBC lineages (Fig 1), albeit at varying proportions. In our study population, Lineage 4 and Lineage 3 were the most frequent (866, 42.5% and 681, 33.4%, respectively) followed by Lineage 1 (336, 16.5%). Lineage 2 was the least frequent (92, 4.5%). The remaining 64 clinical isolates (3.1%) could not be assigned to any of the MTBC lineages possibly because there was insufficient amount of DNA in the samples (below the detection limit). Of the seven geographical zones, four (Coastal, Northern, Lake and Southern Highlands) were highly represented with more than 100 clinical strains each (Table 2). The distribution of the MTBC lineages varied within the geographical zones (Fig 1 and S3 Fig). Our findings reveal that Lineage 1 strains were more frequent in the Lake zone compared to the overall average frequency (20.9% vs. 16.8%), whereas the frequency of Lineage 3 in this zone was lower (27.6% vs. 34.3%) compared to other geographical zones. By contrast, Lineage 4 was the most predominant in all geographical zones and showed relatively similar frequencies across the zones.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. MTBC lineages in Tanzania.

A. MTBC lineage classification of 2,039 nationwide clinical strains. B. MTBC lineage frequencies and geographical distribution in Tanzania.

https://doi.org/10.1371/journal.pone.0206334.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. MTBC lineage distribution across geographical regions in Tanzania.

https://doi.org/10.1371/journal.pone.0206334.t002

Association between lineage and patients’ characteristics

Having described the circulating main lineages of the MTBC we then assessed the relationship between these lineages and patients’ characteristics in the study population (Table 3). We detected a higher proportion of female sex among TB patients infected with Lineage 2 (52.1%) compared to those infected with the other three lineages (range from 31% to 34.5%, p = 0.009). Moreover, we observed that retreatment cases were frequently infected with Lineage 2 strains (26.8%), which was twofold higher compared to Lineage 1 and 4 strains (p < 0.001). We found no evidence for association between lineages and patients’ characteristics such as age and HIV status (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Frequency distribution of MTBC main lineages across patients’ characteristic groups.

https://doi.org/10.1371/journal.pone.0206334.t003

Lineage 2 has previously been associated with retreatment cases, drug resistance and lately also with female sex [15,25]. We therefore investigated if similar associations exist in our study population using a subset of TB cases with complete clinical and demographic information (n = 1,535). To assess these associations we performed logistic regression analyses comparing Lineage 2 to all other lineages pooled together (Table 4). Our analyses revealed Lineage 2 to be independently associated with female sex (adjusted odds ratio [aOR] 2.14; 95% confidence interval [95% CI] 1.31 – 3.50, p = 0.002) and retreatment cases (aOR 1.67; 95% CI 0.95 – 2.84, p = 0.065). We did not detect any association between the lineages and patients’ age and the HIV status.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Associations of patients’ clinical and demographic characteristics with MTBC Lineage 2 (n = 72) compared to all other lineages (n = 1,463).

https://doi.org/10.1371/journal.pone.0206334.t004

Mutations within rpoB gene in retreatment cases

To investigate whether drug resistance was linked to a particular lineage, we included in total 144 out of 321 retreatment cases for drug resistance genotyping of the rpoB gene that confers resistance to rifampicin. Out of these, 112 (77.8%) had no mutations compared to the H37Rv reference gene and 15 (10.4%) contained at least one mutation, either synonymous (3/15) or non-synonymous (12/15) (S4 Fig and S3 Table). We could not determine mutation status in the rpoB gene of 17 (11.8%) retreatment cases due to PCR and sequencing failure. Among the 12 strains detected with non-synonymous rpoB mutations, five belonged to Lineage 2, four to Lineage 4, and three to Lineage 3 (S4 Table). Table 5 summarizes the non-synonymous rpoB mutations detected.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Non-synonymous mutations detected on the rpoB gene among retreatment cases.

https://doi.org/10.1371/journal.pone.0206334.t005

Discussion

In this study, we classified the countrywide collection of 2,039 MTBC isolates representing 1.6% of all smear positive new and retreatment TB cases notified during 2012 and 2013 in Tanzania. Our findings show that the MTBC strains among the study population are diverse, comprising four main phylogenetic lineages (Lineage 1–4) which occur throughout the country. Specifically, we found that Lineage 4 was the most frequent, followed by Lineage 3 and 1. Despite Lineage 2’s recent global dissemination [13], it was the least frequent in our study population. Finally, our analysis on the relationship between MTBC lineages and patients’ characteristics revealed associations of Lineage 2 with female sex and retreatment TB cases included in the study population.

Among the 7 human–adapted MTBC lineages, Lineage 4 is the most broadly distributed and occurs at high frequencies in Europe, the Americas and Africa [24,34]. In our study, we observe that TB epidemics in Tanzania are also predominated by Lineage 4, which is regarded as the most successful of MTBC lineages [34]. In general, the wide geographical range of Lineage 4 is postulated to be driven by a combination of its enhanced virulence, high rates of human migration linked to its spread and ultimately its ability to infect different human population backgrounds [34,35]. In contrast, Lineage 1 and 3 are known to be mainly confined to the rim of the Indian Ocean [7], which is consistent with our observation that nearly 50% of the MTBC strains included in the study belong to these two lineages. This high frequency of Lineage 1 and 3 likely reflects the long-term migrations between Eastern Africa and the Indian subcontinent [36]. In addition, the distribution and frequency of Lineage 1 and 3 in the mainland subset did not vary from that of the coastal region, suggesting spread via internal migrations. Lineage 1 was proposed to have evolved in East Africa prior disseminating out of the continent [10]. Based on this, one might expect higher frequencies of Lineage 1 in the region. Instead, the so called “modern” (TbD1−) lineages (4 and 3 in this case) could be dominating in Tanzania despite presumably being introduced into the African continent only after the first European contact [34,37]. This perhaps illustrates the ability of “modern” lineages to thrive in co-existence with the pre-existing “ancient” (TbD1+) lineages such as Lineage 1 in our case, perhaps because of the comparably higher virulence [14,38]. The neighboring countries of Tanzania on the other hand show comparable MTBC lineage composition to our study [39,40], suggesting common demographic histories and ongoing exchanges that resulted into distinct MTBC populations. Our findings would suggest the frequency of Lineage 2–Beijing in Tanzania, like in most parts of the continent except for South Africa [39,40] to be relatively low, despite the long-standing African-Asian contacts [40]. Evidence from recent studies show that Lineage 2–Beijing was only recently introduced into Africa [13,41].

The burden of TB disease is generally higher in males [1,42], rendering male sex as a potential risk factor for TB. Furthermore, the male bias among TB patients is also observed in settings with no obvious sex-based differences in health-seeking behavior [43]. Whilst we show similar trends in this study, our findings reveal that the proportion of females was higher among TB patients infected with Lineage 2. This finding is consistent with several other previous studies conducted in different settings [15,25,44]. Social and physiological factors predisposing males to higher risk of TB have been indicated [45]. On the one hand, these include risk behaviors such as substance abuse (alcoholism, tobacco smoking) and gender specific roles such as risk occupations (e.g., mining) that are male dominated and known to increase the risk for TB. On the other hand, genetic makeup and sex hormones might contribute to the differences in TB susceptibility among females and males, as epidemiological and experimental studies have suggested female sex hormones to be protective [45]. These observations would propose that the sex imbalance in TB emerges after the onset of puberty. Of note, we observe less sex imbalance in “child” age group (<15 years) which also corroborates the national notification rates [29]. However, this observation can be confounded by BCG vaccination which might be most effective in this age group. Despite the high prevalence of HIV among young females in sub-Saharan Africa [28] and HIV being the strongest risk factor for TB, TB burden remains higher in males. While social and physiological aspects play an important role, findings from this study and others previously conducted in Nepal and Vietnam [15,25] suggest that bacterial factors could disrupt the trends towards male bias in TB, a finding which warrants further investigation. Our hypothesis is that because of higher virulence, Lineage 2 strains are able to overcome the resistance poised by female sex which could explain the less pronounced sex imbalanced.

In addition to its association with female sex, we found that retreatment TB cases were more likely infected with Lineage 2. A retreatment case in our study population represented recurrent TB case either due to relapse or reinfection. We hypothesized that this observation was possibly linked to drug resistance, given the previous reported association between Lineage 2 and drug resistance [46]. However, we detected only 8.3% (12/144) of strains among the retreatment subset tested to contain mutations conferring resistance to rifampicin, five of which belonged to Lineage 2. These findings would suggest that retreatment cases included in this study are mainly driven by reinfection as opposed to treatment failure or relapse.

Finally, based on the age distribution of TB cases in our study, recent or ongoing transmission in high burden countries is implicated as the main contributor to the TB burden rather than disease reactivation (following longer latency periods) [47]. Additionally, an association with young age has been used as an epidemiological proxy for highly transmissible strains and faster rates of disease progression [48,49]. In this study, we did not detect any differences in median age of TB patients infected with different lineages (S5 Fig), an observation that could speak for high ongoing transmission rates in general, irrespective of lineage.

Our study is limited by focusing on a convenient collection of MTBC clinical isolates that could be retrieved from the culture archives, representing 1.6% of all TB cases notified in 2012 and 2013. Given that our findings are based on a limited number of TB cases, the results particularly those related to associations between lineages and patients’ characteristics should be taken with caution as the strength or lack of such associations could likely be affected by the sampling. In addition, most of the geographical zones were underrepresented which could in turn underestimate the respective regional lineage composition and the overall countrywide distribution. Unfortunately, data on drug susceptibility based on other methods such as Xpert MTB/RIF, phenotypic DST and Line Probe Assay (LPA) were unavailable, which could have complemented the drug resistance genotyping performed on a limited subset of the retreatment cases. Systematic sampling would allow for better resolution on the distribution patterns, the frequencies and on epidemiological features of MTBC lineages, which might partially determine the regional specific epidemics.

In conclusion, this study addresses for the first time the countrywide MTBC population structure based on robust SNP-typing. We show that MTBC population in Tanzania is diverse with four of the seven known lineages detected. This study sets the stage for further in depth investigations on epidemiological impact of MTBC lineages in Tanzania.

Supporting information

Acknowledgments

We would like to thank the National Tuberculosis Leprosy Programme (NTLP) through the Central Reference Laboratory (CTRL) for permission to use the MTBC isolate collection for this study.