Molecular epidemiology of clinical Mycobacterium tuberculosis complex isolates in South Omo, Southern Ethiopia

As part of a study to investigate TB in South Omo using both molecular epidemiology and systems biology methods between 2014 and 2017, we established a human subject protocol and a consent form to recruit more than 2000 individuals from different ethnic groups and administrative districts. We discerned non-pastoral areas (Debub Ari, Semen Ari and part of Male) and pastoral areas (Bena Tsemay, Hamer, Dassenech, Selamago and part of Male). The minimum age for enrolment was 15 years. According to 2007 population census, 577,673 people (7.5% urban and 92.5% rural) lived in the area with equal proportions of men and women [14]. Basic demographic and clinical data were collected using a pre-structured questionnaire.

Local translators explained the study’s objectives, risks and benefits to those individuals who were not able to communicate in the Amharic language. Those individuals who consented to participate in the study were asked to provide fine needle aspirate (FNA) and sputum samples to test for EPTB and PTB, respectively. Samples were stored at + 4 °C and -20 °C following collection and transferred to the sample processing and microbial culture laboratory in order to screen for growth of MTBC isolates. Individuals who were diagnosed with TB were educated on and received DOTS treatment according to the national TB management guideline [18].

MTBC isolates were recovered from sputum and FNA samples for pulmonary TB (PTB) and extra-pulmonary TB (EPTB) cases, respectively. Mycobacterial cultures were performed at the Jinka Regional Laboratory (JRL). Samples were processed according to the modified Petroff’s method by decontaminating the specimen with an equal volume of 4% NaOH for 15 min; the remaining volume was filled with phosphate buffer saline (PBS) and centrifuged at speed of 3000 g for 15 min. A drop of phenol red was added to a pellet as pH indicator and neutralized using 10% HCl. The neutralized pellet was inoculated onto two LJ media. One supplemented with glycerol and the other with pyruvate. Inoculated media were incubated at 37 °C for up to 8 weeks. Mycobacterial growth was monitored every week. Culture was considered negative after 8 weeks if no growth was observed. Positive colonies were further confirmed by Ziehl-Neelsen (ZN) staining. Heat treatment of mycobacterial isolates in dH₂O at 80 °C for 50 min was used for genomic DNA extraction without extensive DNA purification. Extracts were stored at -20 °C until they were used for molecular characterization.

DNA of mycobacterial isolates were primarily subjected to region of difference (RD) 9 typing [11]. Secondly, a single tube amplification method [40] was applied for genus typing for isolates which were not detected by RD9 typing. Thirdly, spoligotyping was performed for MTBC isolates as described by Kamerbeek et al., [23]. Laboratory results of spoligotyping were interpreted in binary format and lineages were assigned using an updated version of the SITVITWEB [13] and major lineages were assigned using the “TB insight” database. Isolates which have similar patter to those in the SITVIT database were assigned a Spoligo International Type (SIT) number. Isolates not assigned to SIT numbers were referred as “Orphans” spoligotype. Finally, MTBC DNA samples were subjected to MIRU-VNTR typing following an established procedure [35]. Laboratory results of MIRU-VNTR typing were interpreted using the MIRU-VNTRplus database (http://​www.​miru-vntrplus.​org) to determine MTBC strain lineages and relatedness [2, 38]. A minimum spanning tree (MST) was constructed. Previously identified Ethiopian strains were assigned manually to Ethiopia_1 (Lineage 7), Ethiopia_2 and Ethiopia_3, which lack the spoligotype based spacer 4–24, 13 and 10–19, respectively [9, 17, 37]. The RD9 typing, genus typing and spoligotyping were performed at Aklilu Lemma Institute of Pathobiology (ALIPB), Addis Ababa University (AAU), Addis Ababa, Ethiopia whereas MIRU-VNTR typing was performed at J. Craig Venter Institute (JCVI), Maryland, USA.

Excel data were transformed into IBM SPSS version 20. The results were presented using descriptive statistics. The ability of genotyping methods to discriminate strains was calculated using Hunter-Gaston Discrimination Index (HGDI) [21]. The allelic diversity (h) based on MIRU-VNTR genotyping was calculated using HGDI and classified into relative discriminants based on previous [34] and newly proposed [27] index ranges. The recent transmission index (RTI) was calculated using a formula proposed by small et al. [33]. Isolates were said to be clustered if they have identical pattern based on spoligotyping and/or MIRU-VNTR. Clustering rate was determined using a formula (Nc-C)/N, where Nc = total number of isolates clustered, C = number of clusters, and N = total number of isolates [46]. The association between clustered strains and independent variables were computed using logistic regression models. The crude odds ratio (COR) and the adjusted odds ratio (AOR) were used to present the results. P-values of < 0.05 were considered statistically significant.

One thousand two hundred sputum and FNA samples of study participants were cultured for isolation of mycobacteria. Samples culture summary is presented in Table 1. In overall, 161 MTBC isolates were obtained upon RD9 and multiplex PCR based species and genus typing. Culture recovery rate vary from 2.5% (13/517) in community screened samples and 29.7% (33/111) in smear positive PTB samples from health facilities other than JGH to 67.3% (74/110) in smear positive PTB samples and 76.5% (26/34) in EPTB samples from JGH. Spoligotyping and MIRU-VNTR typing were carried out for a total of 156 isolates derived from 130 and 26 patients with clinical evidence of PTB and EPTB, respectively. Five samples were excluded due to insufficient quantities of DNA. Demographic and clinical characteristics of the corresponding 156 subjects are summarized in Table 2. The age range was 15–80 years with mean 32.9, median 30 and standard deviations 12.9. Female, PTB and new TB cases were 41.0% (64/156), 83.3% (130/156) and 95% (148/156), respectively.

All South Omo MTBC isolates were genotyped using spoligotyping as well as MIRU-VNTR at the strain level, as presented in Tables 3 and 4. The discriminatory power of spoligotying and 24-loci MIRU-VNTR were 0.9700 and 0.9995, respectively. The analysis of spoligotypes resulted in 66 different patterns of which 24 clustered and 42 were unique. MIRU-VNTR analysis resulted in 149 different patterns of which six clustered and 143 were unique. The clustering rates were 57.7% ((114–24)/156) for spoligotying and 3.9% ((12–6)/155) for the MIRU-VNTR analysis. The RTIs for South Omo were calculated to be 57.7% based on spoligotyping and 3.9% based on MIRU-VNTR data.

Allelic diversities (h) for each MIRU-VNTR locus are presented in Table 4. MIRU-VNTR loci are classified based on their allelic diversity and ability to discriminate among the isolates. Eleven MIRU-VNTR loci (424, 577, 802, 960, 1644, 1955, 2163b, 2347, 2996, 3192 and 4052) were highly discriminatory while 10 MIRU-VNTR loci (580, 2059, 2165, 2401, 2461, 2531, 3171, 3690, 4156 and 4348) were moderately discriminatory. The remaining three loci showed poor discriminatory values for the isolates under study. However, 11-loci are considered as moderately discriminatory according to newly proposed discriminatory values [27].

The results of spoligotyping and MIRU-VNTR are presented in Tables 5 and 6. According to spoliogotyping results 76.3% (119/156) of the isolates belonged to 36 shared types (SIT numbers). The remaining, about one-qaurter of the isolates (37/156) were orphan strains. Euro-American (EA) was the most prevalent lineage with 67.3% (105/156) followed by East-African-Indian (EAI) with 22.4% (35/156) and Indo-Oceanic (IO) with 6.4% (10/156) of the isolates. Interestingly, EA was more prevalent in non-pastoral areas while EAI was more prevalent in pastoral areas (Fig. 1). Two M. bovis isolates were identified in pastoral areas.

According to SITVIT analysis, the identified lineages were T 30.1% (47/156), CAS1_Delhi 10.9% (17/156), Haarlem 10.3% (16/156), Ethiopian 1.9% (3/156), Turkey 1.9 (3/156), MANU 1.3% (2/156), LAM 0.6% (1/156), EAI1_SOM 0.6% (1/156), BOV 0.6% (1/156) and X 0.6% (1/156). Almost one-third of isolates (56/156) were not registered in updated version of SITVITWEB database (SITVIT2). According to MIRU-VNTRplus analysis result (S1 Fig, Table 5 and Table 6), the identified lineages were Delhi/CAS 23.9% (37/155), Ethiopia_2 20.6% (32/155), Haarlem and URAL 14.2% (22/155) each, Ethiopia_3 8.4% (13/155), TUR 6.5% (10/155), Ethiopia_1 1.9% (3/155), Bovis and LAM 1.3% (2/155) each, and EAI, X and Ethiopia H₃₇Rv-like strain 0.6% (1/155) each. Around 6% of the isolates were not assigned into lineages using MIRU-VNTRplus database.

MST shows the relationship among strains and lineages based on MIRU-VNTR loci variant count. Five well-defined branches (Complex 1, 2, 3, 4 and 5) and four additional sub-branches (Complex 6, 7, 8 and 9) expanding from Complex 1 are featured in Fig. 2. Each complex largely consisted of distinct sub-lineages. Part of Complex 1 are the Ethiopia_2 sub-lineage and the additional sub-lineages H37Rv, Bovis, unidentified strains, and M. canetti (used as a reference). Complexes 2, 3, 4 and 5 are enriched in Ethiopian_3 and TUR, URAL, Delhi/CAS and Haarlem strains, respectively.

Clustering of M. tuberculosis strains association with different factors were analyzed using spoligotyping pattern. Clustering was significantly associated with the residential area, the major lineage type and the SIT status before adjusted to confounding factors (Table 7). However, only being part of the orphan group compared to shared type (AOR: 0.19, 95% CI: 0.04–0.25) was significantly associated with reduced odds of clustering in the area under study after adjustment with confounding factors.