The study was conducted in two cities, namely, Gondar and Mekelle, supposed to represent the emerging dairy areas of the northern and north-western parts of Ethiopia, aimed to explore the bTB situation in these regions. These urban centers held more than 700 dairy herds comprising of 7,400 dairy cattle, from which 59 and 61 herds comprising of 976 and 818 cattle from Gondar and Mekelle were considered for the study, respectively. In addition to the dairy herds in these areas, a dairy herd owned by the Agricultural, Technical and Vocational Education Training College (ATVET) at Alage, 215 km south of Addis Ababa, was also included in the study as it was one of the biggest herds (n = 284) that was distributing dairy animals to the southern and east central parts of the country, with potential risk of disease transmission. The investigated cattle herds were managed in intensive and semi-intensive systems, the majority of which were established within the last 10 years. The husbandry and farm settings differed somewhat from one site to the other depending on the level of awareness and experience on dairy farming.

bTB infected cattle for post-mortem examination were identified based on the 2016/17 survey aimed to estimate the prevalence and associated risk factors (14) using tuberculin antigens with standard interpretation—positive if the difference of skin thickness increase at the bovine site (PPD-B) was greater by 4 mm than the increase at avian site (PPD-A) of the SICCT test (20). Animals were selected based on the willingness of farmers to offer reactor cattle for a fair compensation or to consciously cull reactors without compensation to avoid further transmission within their herds. Accordingly, 27 out of 2,078 (22%) animals were selected. However, the authors would like to admit the possibility of bias introduction to some degree due to non-randomized selection and small sample size. Ante-mortem inspection of these animals was conducted before they were slaughtered for post-mortem inspection by an experienced pathologist, and relevant data were recorded. Age of the animals was determined based on dentition pattern; body condition score (BCS) was determined and categorized into three scales (poor, medium, and good) by modifying the five scales described by Kellogg (21) to better reflect the assessment in field conditions.

In post-mortem, each lobe of the lungs, liver, spleen, intestines, and kidneys were inspected externally and then sliced at a slice-thickness of 10–20 mm as appropriate to facilitate detection of any tubercle lesions. Lymph nodes of the head region—mandibular, parotid, retropharyngeal; thoracic region—pre-scapular, tracheobronchial, mediastinal; abdominal region—portal, mesenteric; and inguinal and thigh region—pre-femoral, supra-mammary were sliced into 2–5 mm sections and inspected for presence of typical tubercle lesions. Gross lesions were scored using a semi-quantitative approach as described by Ameni et al. (22) and Vordermeier et al. (23). Briefly, rated as “Score 3” when the lesions were multifocal and coalescing; “Score 2” when the lesions were multifocal but not coalescing; “Score 1” when the lesion was in one focus (just starting the tubercle lesion); and “Score 0” when there was no visible lesion. Lesion scoring was performed by the same operator for all animals to ensure scoring consistency. Summation of the pathological scores was considered to determine the total pathological score at the animal level. Furthermore, the type, scale, and size of the lesions were recorded.

Tissue samples from organs and tissues suspected of tubercle lesions identified during the post-mortem inspection were collected from all slaughtered cattle. Collected tissue samples were placed in 50 ml sterile universal tubes (with sterile PBS buffer) and kept on ice until and during transport to the laboratory where samples were stored at −20°C for 1 week and then at −80°C if processing was performed later than 1 week after collection.

Samples were processed according to a standard protocol described in Roberts et al. (24) but with modification on the neutralization step by adding PBS instead of HCl. Tissues were dissected, manually crashed and homogenized using a pestle and mortar, followed by decontamination in an equal volume of 4% NaOH for 15 min with frequent shaking. The sample mix was then filled to 50 ml with PBS buffer and centrifuged at 1,750 × g. The sediments were neutralized by refilling to 50 ml with PBS and concentrated by centrifugation at 1,750 × g for 15 min. The pellets were re-suspended with 1 ml PBS and then inoculated with 3–5 drops in triplicate of Löwenstein-Jensen (LJ) medium slants, i.e., two slants supplemented with pyruvate and the other with glycerol. The slants were incubated at 37°C for 8 weeks; slants were examined on a weekly basis to check for mycobacterial growth.

Cultures were considered negative if no visible growth was detected after 8 weeks of incubation. Microscopic examination of cultures using the Ziehl–Neelsen staining method was performed to identify acid-fast bacilli. Heat-killed cells of each isolate were prepared by mixing colonies in 500 μl distilled H₂O followed by incubation at 80°C for 1 h. Acid-fast positive cultures were preserved as stocks for future use in Dubos Tween-albumin broth and stored at −80°C.

Extraction of genomic DNA was made from heat-killed cultures using a Qiagen DNA extraction kit according to the manufacturer's instruction (25). RD4 deletion typing, a chromosomal deletion characteristic of M. bovis, was made using conventional PCR (26). PCR amplification was performed in a total volume of 27 μL consisting of 12.5 μL HotStarTaq Master Mix (Qiagen, United Kingdom), 3.5 μL Qiagen water, 2 μL of each oligonucleotide primer (10 μM), and 5 μL DNA template. The RD4 oligonucleotide primers used included: RD4-FlankFW 5′- CTC GTC GAA GGC CAC TAA AG - 3′, RD4-FlankRev 5′- AAG GCG AAC AGA TTC AGC AT - 3′, and RD4-InternalFW 5′- ACA CGC TGG CGA AGT ATA GC - 3′ (27). The PCR condition involved amplification in a FlexCycler² (Analytikjena, Germany), using an initial denaturation step at 96°C for 15 min, followed by 35 cycles of denaturation at 96°C for 30 s, annealing at 55°C for 1 min, and elongation at 72°C for 30 s. Cycling was completed by a final elongation step at 72°C for 10 min. The reaction product was resolved by electrophoresis on a 1.5% agarose gel (1.5 g agarose in 100 mL TAE buffer) containing SYBR safe stain (Invitrogen) for examination using a gel documentation system (BioRad Laboratories, USA). Identified strains of M. bovis and Mycobacterium tuberculosis were used as positive controls (PCR products of 446 bp (RD4 deleted) and 335 bp (RD4 present), respectively (27) and Qiagen water was employed as a negative control for deletion typing.

Mycobacterial isolates identified as M. bovis by RD4 deletion typing were further confirmed as M bovis and genotyped by spoligotyping as described by Kamerbeek et al. (28). PCR amplification of spacers were accomplished using two primers, namely, DRa (5′-GGT TTT GGG TCT GAC GAC-3′, biotinylated) and DRb (5′-CCG AGA GGG GAC GGA AAC-3′) in a thermal cycler (Applied Biosystem) using 35 cycles of denaturation at 96°C for 1 min, annealing at 55°C for 1 min and elongation at 72°C for 30 s. The primers were designed to anneal to all repeat sequences and thereby enabling amplification of all spacers that occur in the DR region of a specific strain. The amplified spacers were then hybridized with the known oligonucleotide spacers covalently bound to a non-commercial “biodyne C membrane,” produced by APHA (UK) according to Kamerbeek et al. (28). Hybridization was done by incubating the membrane for 1 hr at 60°C. Hybridized DNA was detected by incubating the membrane in enhanced chemiluminescence (ECL) detection fluid (GE Healthcare) for 2 min and visualized by exposing a light sensitive ECL film for 20 min. The membrane was then analyzed by recording the presence or absence of signals at the sites of the individual probes (29).

Genomic DNA of known M. bovis and H37Rv were used as positive controls for M. bovis and M. tuberculosis, respectively. Identified spoligotype patterns were compared at a global database of M. bovis genotypes (http://www.mbovis.org) (30). The discriminatory power of spoligotyping, i.e., the average probability that this assay assigns a different type for two unrelated strains randomly sampled among the available M. bovis isolates, was assessed by using Hunter-Gaston Diversity Index (HGDI) (31) using in silico website of the University of the Basque Country: http://insilico.ehu.es/mini_tools/discriminatory_power/index.php.

The total number of SICCT reactor cattle was 122 out of 2,078 tested cattle kept in 33 herds, from which 27 reactors (two male and 25 female) were recruited from 10 herds for post-mortem inspections. Seventeen animals were from four herds in Mekelle, and five animals were from five herds in Gondar. The remaining five reactor cattle were donated by Alage ATVET (Figure 1). The number of animals tested in this herd was 284, of which eight were reactors for the SICCT test. Animals sourced from Gondar and Mekelle were crosses of Holstein Frisian and Zebu breeds, while cattle from Alage were all Boran cattle (Zebu breed). In ante-mortem, three animals were highly emaciated and showed chronic cough, one animal had a swollen parotid lymph node on its right side, while the remaining animals showed no clinical signs of being diseased.

Post-mortem inspection showed visible tuberculous lesions in 21 out of 27 slaughtered cattle; the remaining six cattle showed no visible lesions in any of their organs or lymph nodes. All 21 animals identified with visible lesion(s) were suggestive of having tuberculosis in the lymph nodes. A total of 90 lesions were observed in the examined lymph nodes, with the most frequently affected being in mediastinal (MD), tracheobronchial (TRB), and retropharyngeal (RP) lymph nodes (Figure 2). Comparison of the total pathological scores for the examined lymph nodes also showed that higher scores were observed for these three lymph nodes, while the least was for medial iliac and prefemoral lymph nodes (Table 1).

Eleven of the slaughtered cattle showed tuberculous lesions in their visceral organs: in the lungs (7/11), intestine (4/11), liver (3/11), thoracic membrane (2/11), abdominal membrane (1/11), kidney (1/11) and mammary gland (1/11). In the lungs, the most affected lobes were the cardiac lobe followed by the diaphragmatic lobe, while the accessory lobe was the least affected. One of these 11 animals was diagnosed with generalized tuberculosis and all four of its lung lobes were affected; two of the seven animals having lesions in their lungs were detected with lesion(s) in only the cardiac lobe, while lesion(s) in the remaining four animals were detected in the apical or diaphragmatic lobes.

Animal level lesion severity, as determined based on the sum of pathological scores, showed no clear and strong correlation with either BCS or the skin test thickness, whether considering the bovine (PPD-B) only reaction or the comparative (PPD-B minus PPD-A) test results (correlations not shown but all Meta data are listed in Supplementary Table 1).

Mycobacterial culturing on LJ media yielded a total of 64 isolates, all confirmed deleted for RD4 (27), which is a characteristic feature for M. bovis. These isolates were obtained from 208 specimens collected from 27 slaughtered cattle with multiple isolates were recovered from several of the animals although no isolate was yielded from three of them. Of the 24 animals from which M. bovis was isolated, three animals were with non-visible lesions.

Spoligotype patterns were identified for 55 out of the 64 isolates and confirmed that all the isolates were M. bovis, while the results for the remaining were not interpretable due to either lower concentration or poor quality of genomic DNA in the samples. The identified spoligotypes lack spacers 3, 9, 16, and 39–43, a spoligotype signature unique to most M. bovis and is distinguished from other members of the Mycobacterium tuberculosis complex (MTC). The discriminatory power of the spoligotyping assay in the present study was moderately high (HGDI = 0.68). Considering identical spoligotypes from the same animal as one, a total of 28 unrelated strains were identified. The strains were grouped into seven spoligotypes, namely, SB0133, SB0134, SB1176, SB2233, SB2290, SB2467, and SB2520, defined after being compared with the global spoligotypes available in the M. bovis database (www.mbovis.org) (30). Spoligotype SB0134 was the most predominant spoligotype (47%, Figure 3). SB0133 was the second most frequently isolated spoligotype (25.5%), while SB2290 was the third frequently isolated spoligotype, accounting for 18% of the isolates.

Analysis of genetic similarity in the Direct Repeat (DR) region revealed that strains of SB0133, SB2290, SB2467, and SB1176 are likely related as marked by the lack of spacers 3–7, a characteristic of the Af2 clonal complex. On the other hand, strains of SB0134, SB2233, and SB2520 could not be categorized as Af2 strains as they had spacers 6 and 7 present (Figure 3).

Six of the identified spoligotypes, namely, SB0133, SB0134, SB1176, SB2290, SB2467, and SB2520, were from four herds that resided in Mekelle. Of these, one herd was recorded with only one spoligotype, two herds with two spoligotypes each, while the fourth herd had five spoligotypes recorded. In this fourth herd (herd size 100; 12% apparent prevalence), multiple strain types were identified in each of four of the 12 slaughtered reactor cattle, suggesting co-infection with at least two different strains of M. bovis in a single animal (Supplementary Table 1). This herd was established very recently and more than 90% of its cattle were sourced from known dairy herds or markets in Mekelle, Adigrat, and Humera. Trace backing the SICCT test history, some of the source herds had bTB infection history and can thereby be suspected as the main sources of infection for this herd from where we slaughtered the 12 reactors. In Gondar, spoligotypes SB0134 and SB2233 were identified from two herds, one from each, while in Alage spoligotypes SB2290 and SB0133 were identified from that single herd. The geographical locations of the dairy herds, where specific spoligotypes were found, are embedded in Figure 1.