A first insight into genetic diversity of Mycobacterium bovis isolated from extrapulmonary tuberculosis patients in South Tunisia assessed by spoligotyping and MIRU VNTR

Introduction In Tunisia, almost 77% of clinically and bacteriologically diagnosed cases of extrapulmonary tuberculosis (EPTB) are zoonotic TB, caused by M. bovis. Although several studies have analyzed bovine TB in cattle in Tunisia, no study has evaluated the risk of transmission to humans in such an endemic country. We aimed to study the genetic diversity of M. bovis human isolates, to ascertain the causes of human EPTB infection by M. bovis and to investigate the distribution and population structure of this species in Tunisia. Materials and methods A total of 110 M. bovis isolates taken from patients with confirmed EPTB were characterized by spoligotyping and MIRU-VNTR typing methods. Results Among the 15 spoligotypes detected in our study, 6 (SB0120, SB0121, SB2025, SB1200, SB1003 and SB0134) were the most prevalent (83.5%) of which SB0120, SB0121 and SB2025 were the most prevailing. MIRU-VNTR typing method showed a high genotypic and genetic diversity. The genetic differentiation based on MIRU-VNTR was significant between populations from South East (Tataouine, Medenine) and Central West (Gafsa, Sidi Bouzid, Kasserine) regions. Of note, 13/15 (86.7%) spoligotypes detected in our study were previously identified in cattle in Tunisia with different frequencies suggesting a peculiar ability of some genotypes to infect humans. Using combined spoligotyping and MIRU-VNTR method, a high clustering rate of 43.9% was obtained. Our results underlined that human EPTB due to M. bovis was more commonly found in female gender and in young patients. Most of our patients, 66.4% (73/110) were raw milk or derivatives consumers, whereas 30.9% (34/110) patients would have contracted EPTB through contact with livestock. The findings suggest that the transmission of Zoonotic TB caused by M. bovis to humans mainly occurred by oral route through raw milk or derivatives. Conclusion Our study showed the urgent need of a better veterinary control with the implementation of effective and comprehensive strategies in order to reach a good protection of animals as well as human health.


Introduction
In developing countries, Mycobacterium bovis, the agent of bovine tuberculosis (bTB) represents a threat for livestock and human health [1,2]. Humans can acquire M. bovis either by aerogenous route due to a close contact with infected animals or by consuming unpasteurized dairy products [1]. According to the World Health Organization (WHO), among the estimated 10 million new cases of tuberculosis (TB) in 2017, almost 20% were extrapulmonary tuberculosis (EPTB) [3]. Furthermore, in the region with high bTB prevalence, M. bovis would frequently be associated with extrapulmonary disease in humans [1,3,4]. In developing countries, the exact percentage of M. bovis in human TB cases is generally underestimated, since the species diagnosis is rarely performed, especially in EPTB cases [1,3,4].
Human TB (30/100.000) is endemic in Tunisia with high frequency of EPTB cases (56.9%) despite the low rate of HIV infection [5]. Recently, Ghariani et al have reported high prevalence of M. bovis (76%) in lymphadenitis TB (LTB) cases in North Tunisia by conventional methods [6]. In South Tunisia, Siala et al have detected M. bovis by qPCR in almost 77% of extrapulmonary samples taken from patients with EPTB [7]. In addition, LTB was estimated to be 50% of all EPTB cases of which cervical localization was the most frequent (70% to 90%) [5,8]. Djemal [9,10]. Consequently, the risk of contracting EPTB is still high since the control measures for herd, livestock and unpasteurized dairy products as well as milk are consistently declining [1,7]. Nevertheless, no study evaluated the risk of transmission of zoonotic TB due to M. bovis to humans in such an endemic country. Furthermore, the genetic background of M. bovis causing human EPTB is still not known in Tunisia. The availability of such information is critically important in order to identify the source, the route of infection and thus to control the disease. The main goal of this work was to genetically characterize the M. bovis population involved in human EPTB in order to improve our understanding of the bTB epidemiology and M. bovis spread in Tunisia. Spoligotyping and mycobacterial interspersed repetitive units-variable number of tandem repeats (MIRU-VNTR) analysis were used, since the combination of these two genetic markers is known to be a powerful tool for the molecular epidemiology study of M. bovis [11,12]. We finally compared M. bovis data obtained from humans with data from Tunisian cattle [9,10].

Ethics statement
Approval for usage of M. bovis isolates, demographic, epidemiological and clinical data for our study was obtained by the Ethics Committee of Hedi Chaker Hospital-Sfax-Tunisia

Patients
The study was prospectively conducted over a 32-month period (January 2013 to September 2015) by the regional hygiene care mycobacteriology laboratory of the Department of Microbiology of the Hedi-Chaker University Hospital in Sfax (South Tunisia). During this period, all patients with a positive diagnosis of extrapulmonary tuberculosis (EPTB), microbiologically confirmed and attributed to M. bovis, were included in the study. All data were anonymized. The laboratory provided routine mycobacterial diagnostic tests for TB specimens obtained during consultation or hospitalization of suspected TB patients in Hedi-Chaker University Hospital.
In total, 110 patients were included in this study. One hundred and six patients were from 7 Tunisian governorates localized in East Central Tunisia (Sfax (n = 23), Gabes (n = 21)), West Central Tunisia (Gafsa (n = 13), Sidi bouzid (n = 7), Kasserine (n = 4)) and South East (Tataouine (n = 23), Medenine (n = 15)) of Southern Tunisia regions. Four patients were from Libya. The other demographic and clinical characteristics that were collected include sex, age, clinical site of TB, patient origin, past history of TB, raw milk consumption, contact with livestock and BCG vaccination status (Table 1 and S1 Table).

Isolation and identification of mycobacteria
Specimens from various extrapulmonary sites were collected: lymph nodes (n = 100), osteoarticular, intestinal, meningeal, peritoneal and pleural samples (n = 10). Solid specimens were homogenized with pestle and mortar. All specimens were directly inoculated on Lowenstein-Jensen (LJ) and Coletsos slants. After decontamination using the standard N-acetyl-L-Cysteine sodium hydroxide method, the samples were centrifuged at 3000 rpm for 15 min. The pellets were then inoculated on Lowenstein-Jensen (LJ), Coletsos media, and in liquid media Mycobacterial Growth Indicator Tube 960 (MGIT 960, Becton Dickinson Biosciences, Sparks, MD, USA). The tubes were incubated at 37˚C for 42 days and the cultures were incubated on solid media at 37˚C for up to 12 weeks.

Genotyping
DNA extraction from mycobacteria isolates and high-throughput spoligotyping on Luminex 200 (Luminex Corp., TX) were performed as previously described [15,16,17,18]. The obtained data were compared with those of the international database (www.mbovis.org). The  [19,20,21]. These eight loci were selected according to the recommendation of the European Union Reference Laboratory (EURL) [21]. The results were combined into 8-digit allelic profiles for each isolate [19]. The spoligotyping and MIRU-VNTR data obtained from M. bovis isolated from Tunisian cattle were used for comparison [9,10].

Genetic diversity and population structure analyses
Several diversity indices, including the genotypic diversity (Gd = the number of different genotypes divided by the total number of samples using the combination of MIRU-VNTR and Spoligotyping data), the allelic diversity per locus and the mean genetic diversity (Hs) were calculated. The population structure was explored by estimating the F st (index of genetic differentiation between samples) value (0 = no differentiation and 1 = fixation of alternative alleles). The allelic diversity, the Hs and the F st were calculated using F-STAT, version 2.9.3 using the 8 MIRU-VNTR loci data [22]. The discriminatory power of each locus and of the typing method was estimated using the Hunter-Gaston discriminatory Index (HGDI) according to the previously described formula [23].

Phenetic tree and statistical analyses
Neighbor-joining tree was constructed based on Cavalli-Sforza and Edwards distance methods, using Populations 1.2.30 (Olivier Langella, CNRS UPR9034, France) [24] and MEGA6 [25] software. The topology robustness was estimated by performing bootstrap analysis with 1000 replicates. Treedyn and Inkscape were used for tree visualization and annotation [26]. The clustering rate was calculated using the N-1 method formula as described previously [27] to estimate the proportion of TB potentially attributable to recent transmission. Univariate analysis was performed to test the association of risk factors with clustering (proportion of isolates in clusters versus non-clustered ones). Statistical analyses (chi-square, Fisher's exact 2-tailed and student test) were performed using SPSS 16.0 statistical software (SPSS Inc., Chicago, IL). Differences were considered significant at values of P � 0.05.

Sociodemographic and clinical data
The sociodemographic and clinical data linked to the EPTB human isolates are shown in Table 1 and S1 Table. For some patients, the epidemiological data could not be collected as mentioned in the tables.
The age of the patients ranged between 1 and 72 years, with a mean age of 30.36 ± 17.07 years ( Table 1). The majority of the patients ranged between 15-59 years (n = 84, (76.4%)) with 49% of cases between 20 and 40 years old. The male/female ratio in the study population was 0.43. Age distribution differed significantly between men and women (mean age 24.21 ±20.03 years versus 33 years ±15.02; P = 0.013, respectively).
Patients in contact with livestock (14/31 (45.2%)) and consuming raw milk (21/73 (28.8%)) were significantly more represented in the Central West region including Gafsa, Sidi Bouzid and Kasserine, than in the other regions (p = 0.004 and p = 0.039, respectively) ( Table 1). Of note, 100% of patients from the Central West region had LTB. Isolates from Libyan patients were excluded in this latter statistical analysis.
The other spoligotypes were detected differently between the regions and governorates as detailed in Fig 1 and S1 and S2 Tables. For the two last Libyan isolates, one belonged to SB1003 and the second to SB0134 (S1 Table).
Considering M. caprae, three out of the four isolates were grouped in one spoligotype (SB0866) and the remaining belonged to SB2024 (S1 Table). The M. caprae SB2024 pattern was found in Sidi Bouzid governorate and SB0866 was detected only in Gafsa governorate (Fig 1).
Concerning the four isolates taken from Libyan patients, three were not clustered (P52, P54 and P21) and one shared the P60.
The P3 MIRU-VNTR profile included two M. caprae. The two other M. caprae had specific patterns (P4 and P10).
The genetic analyses were performed on 102 M. bovis isolates using MIRU-VNTR data (excluding M. caprae isolates and isolates from Libyan patients). The genotypic diversity (Gd) and the mean genetic diversity (H S ) were 0.53 and 0.52, respectively.
Globally, the genetic differentiations estimated by Fst calculations were very low for the governorates and regions with a global genetic differentiation of 0.025 and 0.042, respectively. No genetic differentiation was found between M. bovis populations of the different governorates ( Table 3). The only significant differentiation was observed between the South East (Tataouine, Medenine) and Central West (Gafsa, Sidi Bouzid, and Kasserine) regions of Tunisia (Table 4).
The six most detected M. bovis spoligotypes in our human isolates were also reported in cattle in Tunisia [9,10] with different frequencies (S2 Table).

Recent transmission rate and risk factor analysis
The recent transmission rate among the 98 M. bovis isolates from Tunisia (excluding M. caprae isolates (n = 4), isolates from Libyan patients (n = 4) and isolates with missing genotyping data (n = 4)) was 43.9% based on the MIRU-VNTR patterns plus spoligotyping. The unique genotypes were detected in children (0-4 years), in old patients (� 60 years), in patients from Sidi bouzid, Kasserine and in patients with personal TB history and with no BCG vaccination (S3 Table).
The clustering was then analyzed in function of the demographic and clinical data (S3 Table). The number of isolates in clusters was high and statistically significant in female patients (75.8%), in patients between 15 to 59 years (79%), in patients with urban lifestyle (72.6%), in patients living in Gabes (24.2%), Sidi Bouzid (8.1%) and in Central Eastern Tunisia (43.5%) (S3/A Table). The univariate statistical analysis showed that the clustering was not affected by age, sex, area of residence, raw and/or unpasteurized milk consumption, contact with livestock, personal TB history and BCG vaccination status (P >0.05) (S3 Table).
Considering the three main M. bovis, SB0120, SB0121 and SB2025 (corresponding to 66 patients) detected in our samples, the Gd (0.58), the H S (0.12) and the CR (42.4%) values were equivalent to the data obtained in the whole Tunisian M. bovis samples based on the MIR-U-VNTR patterns plus spoligotyping (S4 Table). Among the 66 patients infected by the three main M. bovis SBs (SB0120, SB0121 and SB2025), the univariate analysis showed that the clustering was statistically associated with female patients (87.2% vs 63%, p = 0.021) but not affected by age, area of residence, raw and/or unpasteurized milk consumption, contact with livestock, personal TB history and BCG vaccination status (p >0.05) (S4 Table).

Discussion
In South Tunisia, almost 77% of clinically and bacteriologically diagnosed cases of EPTB are attributed to M. bovis [7]. To our knowledge, this is the first study that investigates molecular characterization of M. bovis causing human EPTB in the South of Tunisia.
The geographical distribution of human EPTB cases due to M. bovis in our study showed that the disease was widespread in the seven governorates in Southern Tunisia. SB0120, SB0121 and SB0134 have been described in humans and livestock across the globe [29; 30] indicating the zoonotic importance of these M. bovis spoligotypes. The most detected spoligotype in our study was SB0120 (26.9%) which is also the most common in humans in Italy (63.8%) [31], France (69%) [32], Germany (20%) [33] compared to 1% in the United Kingdom [34]. SB0120 is also detected in cattle in some European countries like France [35], Italy [30], Portugal [30,36], Spain [37], and Germany [33] as well as in neighboring countries (Algeria and Morocco) [38,39].
The frequency of SB0121 (19.3%) among our human isolates was in agreement with those mentioned in humans in previous studies (e.g. France (15%) [32] and Spain (14.1%) [40]). However, it was scarce in human M. bovis isolates from individuals in London and the southeast of England (6%) [29]. SB0134, detected in our study at 5.5%, was found in human M. bovis cases in France [32], in Spain (10.5% to 13.1%) [40; 41], in London, in the southeast of England (6%) [29], in UK (2%) [34] and in Italy (2%) [31]. SB0121 and SB0134 were previously detected in animal isolates all over the world [30] including neighboring countries such as Algeria [38] and Morocco [39]. Interestingly, this is the first study to identify SB2025 and SB1200 in human M. bovis isolates causing EPTB. It is worth noting that these two spoligotypes were isolated for the first time in milk and cattle samples in Tunisia, respectively [10; 28] and they were not detected in cattle in France [35]. Over the previous years, more emphasis was put on human infection due to M. caprae because of the increasing implication of this species in TB infection in cattle or other animals [42; 43]. Among our isolates, M. bovis subsp. caprae SB2024 was represented by one isolate and was previously found in Tunisian isolates from cattle [9,10]. This spoligotype was never reported in humans. In our study, M. bovis subsp. caprae SB0866 was localized in the same region (Gafsa in Central West Tunisia) but was never found in Tunisian cattle samples [9,10]. Only one EPTB case infected by this spoligotype was reported in humans in Germany [33].
The proportion of spoligotypes detected in our human isolates and shared with those previously found in cattle in Tunisia was of 86.7%. This was in agreement with the high values (61-90.9%) previously published [44,45]. Nevertheless, there are differences in frequency of certain spoligotypes among human and cattle isolates reported in Tunisia. For example, SB0134 was the second predominant spoligotype (11.4% to 20%) [9,10] in cattle while it was the fifth most common (5.5%) in our human strains. This finding suggests differential abilities of spoligotypes to infect humans and animals [30,44,45]. Regarding MIRU-VNTR data, human isolates shared some CVP genotypes based on the 6 common loci (ETR-A, ETR-B, QUB 11a, QUB 11b, QUB 3232 and MIR U4 (ETR-D) with cattle in Tunisia. However, humans and cattle showed different M. bovis populations suggesting a host-specific pool of genotypes as described in the Table 5. All these findings showed a specificity of M. bovis populations in terms of geography and hosts as observed for M. tuberculosis. It would be pertinent to go further in the evolutionary and biological mechanisms linked to host specificity in order to better understand the transmission mechanism between humans and cattle.
As previously reported in Tunisia [6,7], our results underlined that human EPTB cases due to M. bovis were more common in women [6,7,46,47]. Female gender represents a major risk factor for EPTB according to several published reports in Turkey, USA, Asia, Egypt and North Africa [48, 49; 50; 51]. The sociodemographic data in our current study showed that EPTB was also more frequent in "younger patients" between 15-59 years of whom 49% ranged between 20 and 40 years and 81% were female. This finding is in agreement with data reported in Tunisia as well as in several areas with moderate or high TB burden [6,7,46,47,49,51,52,53,54]. This can be explained by a higher consumption of raw milk or derivatives among "young people" than the elderly. Furthermore, due to the current lifestyle, females are generally more exposed to livestock or products from cattle. Indeed, our data showed that female patients are more than twice as likely as the male patients to have contact with livestock or other animals and 68% of "young patients" (ranging between 15-59 years) declared to consume raw milk or derivatives of whom almost 80% were females. Furthermore, the consumption of unpasteurized milk and dairy products has been indicated as the most likely source of transmission in clinical cases of LTB, which is the main form of zoonotic TB. Of note, most of our patients (73/110, (66.4%)) are consumers of raw milk or derivatives, whereas only 34/110 (30.9%) patients had contact with livestock, suggesting that, in our study, dairy products are one of the main sources of M. bovis human infection. This could explain the increased proportion of cases with positive LTB diagnosis (100/110 (91%)) caused by M. bovis among EPTB patients in Southern Tunisia. In these areas, people working in dairy farms commonly take raw milk home and sell it clandestinely to markets, increasing the risk of zoonotic TB transmission to humans. It has been shown that M. bovis was able to survive and to remain virulent for extended periods in a variety of cheeses made with raw milk [55; 56; 57].
MIRU-VNTR typing method showed a high genotypic and genetic diversity with 6/8 loci showing high allelic diversity and HGDI values. In our study, the global genetic differentiations based on MIRU-VNTR were very low between governorates and regions. Nevertheless, it was significant between populations from the South East (Tataouine, Medenine) and the Central West (Gafsa, Sidi Bouzid, Kasserine) regions of Tunisia. These findings suggest a slight structuring in the M. bovis population in humans. These findings are in concordance with those of Smith et al who mentioned that some M. bovis genotypes could be more localized compared to others [58]. The diversity observed in the South East (Tataouine, Medenine) and the Central West regions could be explained by the fact that these two regions are transit areas for livestock traffic through the Sahara coming from different neighboring areas with high bTB burden (e.g. Arab Maghreb (Libya and Algeria) and African countries).
Using combined spoligotyping and MIRU-VNTR method, a high clustering rate of 43.9% was obtained. These data do not corroborate previous reports showing much lower rates for LTB, 29.5% [59] and 35% [60] in Southwest Ethiopia and 33% for M. bovis TB infection in United Kingdom [61]. These results suggest a high level of recent transmission of zoonotic bTB in humans in Tunisia. SB0120, SB0121 and SB2025 represent the spoligotypes harboring the highest frequencies of isolates among clustered strains (42.4%). This indicates that the predominant spoligotypes, SB0120, SB0121 and SB2025, are mainly responsible for recent transmissions in Southern Tunisia.
According to the clustering analysis, female patients, patients aged between 15-59 years, from Central East Tunisia (Sfax, Gabes), South East regions, patients who not have a personal TB history and who have a positive BCG vaccination status were significantly clustered cases. Up to 60% and 58% of isolates from EPTB patients who consumed raw milk/milk products and who were in close contact to livestock, respectively, were also highly regrouped in clusters. Among human isolates that shared CVP genotypes with cattle isolates, 75.5% were in clusters. To explain these recent human transmission cases, oral transmission route seems evident since 72/73 (98.6%) and 33/34 (97%) patients who consumed raw milk/derivatives or who had animal contact, respectively, had extrapulmonary manifestations in Lymph nodes.
It is worth noting that, from the analysis, the elderly, children, young population, patients with personal TB history and with no BCG vaccination were significantly more associated with non-clustered isolates, which can be interpreted as ancient infection. Nevertheless, the recent transmission rate can be underestimated since only an exhaustive study including human and animal isolates could rigorously determine the bTB epidemiology.
Furthermore, our study was limited to the South of Tunisia, and the majority of patients were from urban areas. Thus, a larger study with sampling combining rural and urban areas from both North and South of the country, humans and animals would provide a more accurate view of M. bovis molecular epidemiology. This kind of study would give a better evaluation of bTB transmission risk to humans and a better identification of the sources and routes of transmission of M. bovis in humans in Tunisia. Besides, the complementary detailed genomic analysis would also allow study the genes/proteins involved in the biological processes responsible for host specificity and/or pathogenesis of the isolates.
In conclusion, our study described for the first time, in southern Tunisia, the transmission of zoonotic TB to humans due to M. bovis. Our findings underlined that recent transmission was the possible explanation of most M. bovis EPTB infections. The high genetic diversity reflects a large number of contamination sources. Nevertheless, the high clustering rates of certain genotypes such as the SB0120 and the strong association with milk or milk products consumption suggest that some genotypes have a higher ability to infect humans and that the source is mainly linked to oral ingestion.
Our study highlighted the urgent need for a better veterinary control with the implementation of effective and comprehensive strategies in order to reach an effective protection not only of humans but also of animals.
Supporting information S1  Table. Genetic data in function of the demographic and geographic characteristics (A) risk factors (B) of patients with extrapulmonary tuberculosis from Tunisia and with the 3 major detected SB (C). (DOC) S4 Table. Genetic data in function of the demographic and geographic characteristics (A) and risk factors of EPTB patients caused by Mycobacterium bovis bearing major SB (SB0120, SB0121 and SB2025). (DOC)