Abattoir-based estimates of mycobacterial infections in Cameroon

Mycobacteria cause major diseases including human tuberculosis, bovine tuberculosis and Johne’s disease. In livestock, the dominant species is M. bovis causing bovine tuberculosis (bTB), a disease of global zoonotic importance. In this study, we estimated the prevalence of Mycobacteria in slaughter cattle in Cameroon. A total of 2,346 cattle were examined in a cross-sectional study at four abattoirs in Cameroon. Up to three lesions per animal were collected for further study and a retropharyngeal lymph node was collected from a random sample of non-lesioned animals. Samples were cultured on Lowenstein Jensen media and the BACTEC MGIT 960 system, and identified using the Hain® Genotype kits. A total of 207/2,346 cattle were identified with bTB-like lesions, representing 4.0% (45/1,129), 11.3% (106/935), 23.8% (38/160) and 14.8% (18/122) of the cattle in the Bamenda, Ngaoundere, Garoua and Maroua abattoirs respectively. The minimum estimated prevalence of M. bovis was 2.8% (1.9–3.9), 7.7% (6.1–9.6), 21.3% (15.2–28.4) and 13.1% (7.7–20.4) in the four abattoirs respectively. One M. tuberculosis and three M. bovis strains were recovered from non-lesioned animals. The high prevalence of M. bovis is of public health concern and limits the potential control options in this setting without a viable vaccine as an alternative.

The genus Mycobacterium belongs to the phylum Actinobacteria and is a large genus including over 153 recognized species of bacteria 1 . The genus includes a number of highly pathogenic species responsible for major human and animal epidemics including human and animal tuberculosis and leprosy. These are broadly grouped into those that form tuberculous lesions in their hosts and those that do not. Tuberculous mycobacteria are grouped in what is known as the Mycobacterium tuberculosis complex (MTBC) 2 . This group includes among other species Mycobacterium tuberculosis and M. africanum the cause of human tuberculosis and M. bovis the cause of both bovine (bTB) and zoonotic (zTB) tuberculosis 3 . There are many non-tuberculous mycobacterial species including M. leprae (causing human leprosy), M. avium subspecies paratuberculosis (causing Johne's disease) as well as other environmental mycobacteria that can be isolated from lymph nodes of apparently healthy carcasses such as M. avium, M. fortuitum and M. gordonae 4

.
M. bovis causes a chronic recrudescent disease which impacts production, and milk quality in cattle 5 and can be transmitted from cattle to humans through air borne spread or consumption of infected milk or meat causing pulmonary or extra pulmonary zTB 6 . The control and eradication of bTB has had varying success in high-income countries. Wildlife reservoirs, such as badgers the United Kingdom 7 and possums in New Zealand 8 have become major reservoirs complicating control. However, in most middle and low income countries there is limited or no control in cattle and a reliance on meat inspection to limit the public health risks 3,9,10 . African Republic [14][15][16][17] , with many also leaving to Nigeria to feed its large human population. Cattle movements within Cameroon are not well documented but reflect the extensive nature of cattle management including seasonal transhumance 18 . Study design. The four municipal abattoirs of Bamenda (Northwest Region), Ngaoundere (Adamawa Region), Garoua (North Region) and Maroua (Extreme North Region) were selected for sampling as they were the largest abattoirs in each Region. It was anticipated that these were most likely to be receiving cattle predominantly from the local administrative area and so reflect largely what was happening in that Region in terms of circulating mycobacterial strains and thus allow some spatial resolution of strain diversity.
Based on previous estimates from the Northwest Region we assumed a prevalence of lesions of ∼ 5% 12 and calculated a target sample size of ∼ 1,000 cattle per abattoir to ensure recovery of at least 25 isolates per abattoir assuming a 50% recovery from culture. This would allow the within abattoir prevalence of 5% to be estimated with a precision of ± 1.3% with a 95% level of confidence. Due to project constraints, we were unable to sample in Garoua and Maroua abattoirs for more than a week each, which severely limited the final sample sizes from these Regions.
The working practices at the 4 abattoirs varied considerably and at each abattoir we spent time observing operations and piloting abattoir specific sampling protocols in order to minimize disruption to the work. However, the basic principles were the same: animals were cast for slaughter and restrained with ropes by the slaughter men; the research team immediately tagged it, collected a blood sample (results not presented here) and recorded animal-level data on owner/butcher, sex, age, dentition, body condition 19 and origin. The slaughter men were given matching numbered tags for the offal so that at meat inspection the carcass, offal and head could be linked to the blood samples and animal data. The dentition score was based on the presence or absence of permanent incisor teeth with the scores ranging from 0 for no permanent incisor teeth present to 4 for all 4 pairs of incisors and an extra category, 5, for animals with broken teeth or excessively worn indicating older animals 20 .
Meat inspection was carried out following the 2002 MINEPIA meat inspection framework 21 . It should be noted that in Ngaoundere the research team did all the head inspections as these were not routinely being inspected for logistical reasons at this abattoir. Briefly, the lungs, liver and kidneys; lymph nodes of the thoracic and head regions; the mesenteric lymph nodes, other lymph nodes and tissues/organs of the body were visually examined, palpated and incised for the presence of granulomatous bTB-like lesions. Once identified by the veterinary inspectors, up to 3 macroscopic lesions from different anatomical sites per animal were collected into sterile 25 ml universal tubes using forceps and scalpel blades by our research team. Lesion type (purulent, caseous, or calcified), scale (single or multiple), size (< 10 mm, 10-50 mm, > 50 mm in diameter at widest axis) and pathological score (small lesion at 1 focus, small lesion at more than one focus, extensive necrosis) were also recorded following 22,23 . In addition to the tissue samples for culture from animals with lesions, a number of animals classed as non-lesioned by the meat inspectors at the Bamenda and Ngaoundere abattoirs were randomly sampled (using random number generator www.Random.org) and a single retropharyngeal lymph node per animal collected.
Tissue samples were stored on ice and transported back to the field lab or directly to the Tuberculosis Reference Laboratory (TBRL) in Bamenda. The samples from the Ngaoundere, Garoua and Maroua abattoirs were stored in liquid nitrogen and shipped to the TBRL Bamenda in dry shippers (Taylor-Wharton). Upon arrival at the TBRL, the samples were either thawed and processed immediately or stored at − 80 °C until processed. The universal tubes containing samples were only re-opened at the TBRL in a Class II biological safety cabinet (BSC). Samples were processed in batches in the hood for efficiency (maximum of 8 samples per batch) but only a single sample opened at a time to avoid cross contamination.
Mycobacterial culture. The tissue samples were prepared and cultured as described by the World Organisation of Animal Health (OIE) 24 with minor modifications. Briefly, thawed samples were removed from their tubes in the BSC and fatty tissues were trimmed off using sterile forceps and scalpel blade and then the infected area (or just a random section of the lymph node for non-lesioned lymph nodes) of lymph node cut into tiny pieces. Three to five grams of the sliced tissue was then ground into a paste using a sterile mortar and pestle with the addition of 1-2 g of fine sterile sand and 5 ml of sterile normal saline (0.85%). The paste was then transferred to a 50 ml falcon tube and decontaminated by adding an equivalent volume of 4% NaOH and left to stand for 15 min with intermittent vortexing. Sterile PBS (pH 6.8, 0.067 M) was then added up to the 50 ml mark and the tube was centrifuged at 3200 g for 20 minutes at 18 °C. The supernatant was then decanted, and the resulting pellet was re-suspended in 2 ml of sterile PBS and ∼ 0.5 ml was inoculated on a Mycobacterial Growth Indicator Tubes (MGIT) using the BACTEC MGIT 960 automated culture system following manufacturer's instructions, and ∼ 0.1 ml (2 drops) was also inoculated onto each of two slopes of Lowenstein Jensen (LJ) media (one supplemented with pyruvate and the other with glycerol). LJ cultures were observed for growth once a week until growth was observed, up to 12 weeks when it was classed as no growth observed.
The BACTEC MGIT 960 system includes a MGIT tube, MGIT 960 supplement kit and BACTEC MGIT 960 instrument. The MGIT tube contains 7 ml of modified Middlebrook 7H9 broth with casein peptone and an oxygen quenching fluorochrome (tris 4, 7-diphenyl-1, 10-phenonthroline ruthenium chloride pentahydrate) embedded in silicon at the bottom. The supplement kit contains antibiotics mixture -PANTA (Polymyxin B, Amphotericin B, Nalidixic acid, Trimethoprim, and Azlocillin) and an enrichment growth medium (OADC) containing oleic acid, bovine serum albumin with dextrose, catalase, and polyoxyethylene stearate. The MGIT instrument automatically monitors the fluorescence within each tube every hour to determine the culture status. A tube is indicated as positive when oxygen embedded in the fluorescent dye is used up by aerobic microorganism enabling the fluorochrome to fluoresce. The MGIT media was prepared according to the manufacturer's instruction with the addition of 0.8 ml of PANTA/OADC mixture to each tube. Tubes were incubated for 8 weeks as recommended by the manufacturer 25 .
Scientific RepoRts | 6:24320 | DOI: 10.1038/srep24320 Any observed growth on the LJ medium or MGIT indicated positive growth were scraped and used to prepare a suspension in 3% formal saline, heat-fixed on a hot plate at 70 °C for 30 minutes, stained by the Ziehl-Neelsen (ZN) method 26 and viewed under the 100× objective of a light microscope to determine the presence and morphology of acid-fast bacilli (AFB). MGIT Tubes that had no growth indicated by the BACTEC MGIT instrument after 56 days were visually examined for any growth before being discarded. If any growth was suspected a smear was also prepared, stained by ZN, and read to assess the presence of AFB.
Typing of acid fast bacilli (AFB). All AFB positive cultures were typed using the Hain GenoType MTBC assay and GenoType Mycobacterium CM/AS kits (Hain Lifescience, GmbH, Nehren, Germany). DNA was extracted from each positive culture using the GenoLyse kit (Hain Lifescience, GmbH, Nehren, Germany) strictly following the manufacturer instructions.
The Hain GenoType MTBC assay was performed following the manufacturer's instructions and as previously described 27 . Briefly, 45 μl of a PCR mix comprising of 10 μl of amplification mix ' A' (containing biotinylated primers) and 35 μl of amplification mix 'B' was prepared in a 0.2 ml PCR tube. Five microliters of extracted DNA was added in a separate room to give a final reaction volume of 50 μl. Amplification was carried out in a thermal cycler (Applied Biosystem, 2720 Thermal cycler, USA) using the following protocol: denaturation at 95 °C for 15 min; 10 cycles of denaturation at 95 °C for 30 seconds and elongation at 58 °C for 120 seconds; an additional 20 cycles of denaturation at 95 °C for 25 seconds, annealing at 53 °C for 40 seconds, and elongation at 70 °C for 40 seconds; and a final extension at 70 °C for 8 min. Twenty microliters of the resulting amplification products were hybridized on labelled membrane strips at 45 °C for 30 min. Hybridized products were developed by addition of a conjugate buffer (containing streptavidin conjugated with alkaline phosphatase) and a subsequent addition of the substrate buffer for colorimetric detection of the bands. Species were identified according to their characteristic band pattern using the interpretation chart provided by the manufacturer (http://www.hain-lifescience.de/en/products/ microbiology/mycobacteria/genotype-mtbc.html). Each MTBC strip contains 13 probes including amplification and hybridization controls to verify the test procedures. The kit with the exception of M. tuberculosis, M. africanum subtype II, and M. canettii, identifies unambiguously all of the MTBC species.
The GenoType Mycobacterium CM and AS assays detect 23 common NTM species frequently encountered in mycobacteriolosis plus an additional 14 species. This assay was done following the manufacturer's instructions and as previously described 28,29 . Briefly, 45 μl of PCR mix comprising of 35 μl primer-nucleotide mixture (provided with the kit), 5 μl of 10× PCR Buffer, 2 μl of MgCl 2 , 0.2 μl of HotStarTaq polymerase (QIAGEN, Hilden, Germany) and 3 μl of molecular grade water was prepared in a 0.2 ml PCR tube. Five microliters of isolated DNA was added in a separate room to give a final reaction volume of 50 μl. The amplification protocol, hybridization and detection were same as for the GenoType MTBC assay described above. Identification of the non-tubercular mycobacteria (NTM) species were according to the interpretation table provided by the manufacturer (http:// www.hain-lifescience.de/en/products/microbiology/mycobacteria/genotype-mycobacterium-cmas.html).
Data management and statistical analysis. Data in the field and in the laboratory were recorded by hand and then transcribed to an Access database and spreadsheet respectively (Microsoft Corp.) and uploaded to Dropbox (www.dropbox.com). The R statistical software tool (www.R-project.org) was used to manipulate the data and for statistical analyses. Using R scripts the data was read and merged into new tables in a PostgresSQL (http://www.postgresql.org) database. Using R scripts extracts of data were made as csv files and passed to the R 3.2.0 (https://cran.r-project.org/) statistical environment for analysis.
Summary statistics were performed using the epicalc package. Prevalence and binomial 95% confidence intervals (exact method) were estimated using the ci.binomial function. Summary graphics were drawn using the ggplot2 package. The chloropleth maps showing administrative Divisional/sub Divisional prevalences were produced using QGIS 2.2 ® and shape files obtained from the GADM database of Global Administrative Areas (www. gadm.org). Principal component analysis (PCA) was carried out using the princomp function to collapse the 4 ordinal lesion scores (lesion type, scale, size and pathology score) into the first 2 unrotated components and the scores visualised stratified by abattoir using the ggbiplot function to explore for evidence of differences in pathology between abattoirs. Comparisons between abattoirs of the continuous PCA scores for the first 2 components were examined using ANOVA and pairwise comparisons were made using the pairwise.t.test function with a Bonferroni correction in the R statistical environment.
Univariable odds ratios for categorical variables for the binary outcome variable bTB-like lesion or not and M. bovis culture positive or not, were estimated using the cci function in the epicalc package. A new "age" variable was made by collapsing the dentition variable; juvenile = animal with 2 or less permanent incisors (dentition score of ≤ 2) and adult = an animal with three or more pairs of permanent incisor teeth (dentition score > 2). A new condition score was also made: animals were categorised as 'poor' with a body condition score of less than 3, 'normal' for those with a condition score of 3 and 'good' for those with a score of 4 or 5.
An animal-level logistic regression model was developed to identify risk factors for the presence of bTB-like lesions and to test for differences in the prevalence between abattoirs after adjustment for other factors using the whole data set. The factors available were abattoir, breed, age, sex and condition score which were initially screened by univariable analysis. A second model was developed to identify risk factors for being M. bovis culture positive. Multivariable models were built by adding variables in a forward selection process in order starting with those with the smallest p-value from the univariable analysis. Variables with a p-value > 0.2 on the univariable analysis were not included in the multivariable analysis. Abattoir was forced into the model at the start to account for the stratification in the sampling design. The variables were allowed to be moved in and out to check if they were still significant (p < 0.05) after the addition of each new variable and their impact on the estimates checked (for potential confounding). The final model was compared to that based on the selection of the smallest Akaike Information Criterion (AIC) 30 and if variables that were not significant based on p-value and Wald statistic improved the fit of the model, based on the AIC, they were included in the final model. Post model diagnostics were done using the LogiDx package.
Ethical statement. This study did not involve the experimental use of any live vertebrate but reports the results of post mortem examinations of cattle for bovine tuberculosis carried out by the local veterinary inspectors in commercial abattoirs in Cameroon. Samples for culture were collected from carcases in accordance with best practice guidelines to minimise contamination. Local approval was given by the Head of Epidemiology at the Ministry of Livestock, Fisheries and Animal Industries responsible for supervision of activities at commercial slaughterhouses in each administrative Division. This project and the protocols had ethical approval from the University of Edinburgh Ethical Review Committee (Animal (Scientific Procedures) Act, 1986) (ERC No: OS02-13).

Demographics of samples from the 4 municipal abattoirs in cameroon.
A total of 2,346 animals were examined from the four municipal abattoirs. This was divided across the abattoirs with 1,129 cattle sampled from Bamenda in a 3-month period in 2012 (April-June); 935 cattle from Ngaoundere in a 4 week period in August 2013; and 122 from Garoua and 160 from Maroua each sampled over just a week each in late September and October 2013 respectively. The summary statistics for age, breed and sex of the animals sampled by abattoir are given in Table 1 and the distribution of animal ages are plotted in Fig. 2. This data suggests there is a very marked difference between the Bamenda and the northern abattoirs. Bamenda appears to slaughter a much higher proportion of younger males in comparison to the northern abattoirs. In addition, there is a strong breed difference between the abattoirs with more Red and White Fulani cattle being slaughtered in the Northwest and more Gudali and mixed breeds in the northern abattoirs (Fig. 3).

Prevalence of bTB-like lesions in slaughter cattle from the 4 municipal abattoirs in cameroon.
A total of 316 bTB-like lesions (mainly from lymph nodes) were collected from 207 of the 2,346 inspected animals. Due to logistical problems lesions from 6 animals in Ngaoundere did not get cultured so the denominator for cultures is 201. At least one tissue sample was collected from every animal that presented with one or more bTB-like lesions at different anatomic sites of the carcass. In addition, retropharyngeal lymph node samples were collected from a random sample of 179 apparently healthy animals with no visible lesions in the Bamenda and Ngaoundere abattoirs. Overall, bTB-like lesions were mostly observed in retropharyngeal, mediastinal and bronchial lymph nodes in all abattoirs (Fig. 4a). However, when the relative proportions are compared (Fig. 4b) there is a clear difference between abattoirs with a much higher proportion of lesions being in the retropharyngeal lymph nodes in Bamenda (0.376) and Ngaoundere (0.343) while mediastinal lesions predominate in Garoua (0.277) and Maroua (0.370). Also, interestingly, Ngaoundere and Garoua also had higher levels of liver lesions (although the numbers are small) and were also abattoirs with high fluke burdens. We also recorded the four measures of pathological changes and the comparison of the distribution of lesions, type and size is given in the PCA in Fig. 5. This suggests that the pattern and severity of lesions were different in the different abattoirs and both the first and second principal components were statistically significantly different (p < 0.01) between abattoirs based on an ANOVA. For PC1 pairwise comparisons showed Bamenda different from the northern abattoirs which appear to have larger and more extensive lesions and for PC2 all pairwise comparisons were significant except between Garoua and Maroua. The lesion data were collapsed to the animal-level and the summary results are presented in Table 2. Although the sample sizes are much smaller for Garoua and Maroua, the raw estimates suggest that the prevalence of animals with bTB-like lesions is significantly lower in Bamenda compared to the other 3 abattoirs in the northern Regions with Garoua having the highest prevalence, ∼ 6 times that of Bamenda.  Distribution and prevalence of mycobacterial species isolated from cattle from slaughter cattle from the 4 municipal abattoirs in cameroon. The majority of animals with bTB-like lesions had M. bovis isolated from at least one cultured lymph node (150/201) in comparison to only 3/179 randomly selected animals with no bTB-like lesions where a single retropharyngeal lymph node was cultured. Based on culture as a gold standard the sensitivity of lesions as a diagnostic test for M. bovis at the animal level was 86% (81-90%) and the specificity was 80% (74-85%). Further work is being done using no gold standard approaches to estimate these more reliably. This equates to a positive predictive value of 75% (61-86%) in Bamenda and 78 (69-85%) in Ngaoundere. In addition, a number of other mycobacteria were also isolated from these animals (   The spatial distribution of M. bovis positive cattle is plotted in Fig. 6 based on a subset of n = 2088 animals where the butcher or owner's view of the origin of the animal was reported. There were 52 missing origins in the Northwest Region and 206 for the 3 northern Regions. In the Northwest Region apart from the Mezam Division, which had the highest prevalence of M. bovis of 6.9%, all the other Divisions that supply Bamenda municipal abattoir had a prevalence of less than 5% (Fig. 6a). By comparison most of the cattle slaughtered in the Extreme North and North Regions of Cameroon came from Mayo-Kani and Benoue Divisions with extremely high apparent prevalences of M. bovis of 81.8% (9/11) and 48.6% (34/70) as shown in Fig. 6b. Conversely, cattle from the Vina Division in the Adamawa Region had the lowest burden 3.8% (19/498) of all the Divisions supplying the 3 northern abattoirs. The patterns are very similar if bTB-like lesions prevalence distribution is plotted instead of M. bovis prevalence (not shown). These estimates however need to be interpreted with some caution as the butchers often did not know the origins of the animals and so the maps are based on small and potentially biased subsets of the data.
The M. tuberculosis isolate came from a non-lesioned retropharyngeal lymph node from a bull, with a dentition score 2 and body condition score of 3 and it originated from the Momo Division. The unadjusted prevalence of M. tuberculosis in slaughter cattle in Bamenda was therefore an estimated 0.09% (0.00-0.49%).
The NTMs were broken down into the different species and whether they came from a bTB-like lesion or a "normal" lymph node in Table 3  three Gram-positive bacilli that did not belong to the genus Mycobacterium, these were recovered in samples in Bamenda and Ngaoundere municipal abattoirs but were not typed further. M. fortuitum was recovered from animals with visible lesions only at all abattoirs except Maroua. M. gordonae on the other hand was only recovered from animals that presented with bTB-like lesions at Bamenda municipal abattoir. M. phlei, which was the most dominant NTM species, was recovered from both animals that presented with and without bTB-like lesions at slaughter at the Bamenda and Ngaoundere municipal abattoirs (Table 3). If this pattern of NTMs were repeated across all non-lesioned animals, the estimated prevalence in Bamenda (17/91) would be 18.7% (12.4-30.8%) and in Ngaoundere (4/88) would be 4.5% (1.1-9.6%), from a single retropharyngeal LN from a random sample of non-lesioned animals. For animals with bTB-like lesions the prevalences of NTMs would be 6.7% (1.4-18.3%) for Bamenda and 6.6% (2.7-13.1%) for Ngaoundere. Although the odds were 1.67 (0.83-3.70) times higher for recovery of an NTM from a non-lesioned animal than form one with bTB-like lesions this was not significant at the 95% level.
The 4/2,346 animals that had mixed infections represents a possible prevalence of ∼ 0.17% (0.05-0.44) overall. All the four co-infections were with M. bovis and an NTM. Two of these were with the unknown mycobacteria species cultured from Garoua and Maroua municipal abattoirs. The other 2 were co-infections with M. phlei and M. fortuitum in cattle slaughtered at Ngaoundere and Garoua municipal abattoirs respectively (Table 2).

Risk factor analysis for bTB-like lesions in cattle at slaughter in cameroon.
The univariable analysis (Table 4) shows a large and significant increased odds of occurrence of bTB-like lesions in the northern abattoirs compared to Bamenda as already discussed but quantified here. One of the key questions was whether this difference was real or was due to the observed differences in structure of the populations in the different abattoirs. Interestingly there was also an increased odds of bTB-like lesions in thin animals, older animals and female cattle. A multivariable model was developed (Table 5) and all the variables from the univariable analysis were included in the final model. Interactions and confounding between age and sex were checked but there was no evidence of either although in the final model neither are statistically significant at the 5% level their inclusion improves the AIC so they were retained. Further the adjusted prevalence estimates based on a standardised population structure (in this case the structure of the Bamenda population of slaughter cattle) were 4.0% (2.8-5.1%) for Bamenda, 15.1% (10.3-21.9) for Ngaoundere, 17.5% (11.1-26.6%) for Garoua and 12.2% (6.8-20.1) for Maroua. This provides strong evidence that after adjusting for the various potential confounders and the population age structure the northern abattoirs had significantly higher bTB-like lesion prevalences and by implication much higher levels of M. bovis.

Risk factor analysis for M. bovis in cattle at slaughter in cameroon.
A univariable analysis for the isolation of M. bovis from a sampled tissue from an animal was carried out using the subset of animals from which at least one lymph node or tissue was collected and cultured (n = 323) ( Table 6). Animals with multiple tissues cultured were classified as confirmed positive for M. bovis if one or more cultures from that animal were typed as M. bovis using the Hain kit. The final dataset consisted of 144 animals with bTB-like lesions for which complete data was available for all variables of interest and 179 without bTB-like lesions. A multivariable logistic regression model was then developed and the final model included abattoir, sex and breed (Table 7). Ngaoundere

Discussion
The aim of this study was to get current estimates of the prevalence and identify risk factors for bTB-like lesions, describe the diversity of mycobacterial infections and estimate the minimum prevalence of M. bovis in slaughter cattle across Cameroon. This represents the most extensive study of mycobacterial infections in Cameroon and has provided important baseline data for understanding bovine tuberculosis and informing public health on the risks of zoonotic tuberculosis. Cameroon has a large cattle population ∼ 5-6 million with extensive movements of cattle within Cameroon and also from Chad and Central African Republic through Cameroon to Nigeria and these results also have useful information for public health authorities in the neighbouring countries.
We identified a number of mycobacterial species dominated by M. bovis from lesions found at slaughter with the exception of 3 isolates from grossly normal lymph nodes. In addition, we identified a bovid with M. tuberculosis as has been reported in other studies in Ethiopia, Burkina Faso, Nigeria, Zambia, India and   Table 5. Multivariable logistic regression model for the presence of a bTB-like lesion in an animal at four municipal abattoirs in Cameroon (n = 2,318).
Scientific RepoRts | 6:24320 | DOI: 10.1038/srep24320 Spain [31][32][33][34][35][36] . It is currently not clear whether this could be transmitted on to other cattle or spill back into the human population. Non-tuberculous mycobacteria (NTM) were predominantly recovered from cattle with and without lesions and from Bamenda and to a lesser extent Ngaoundere and as reported elsewhere in Africa and Europe [37][38][39] . M. phlei, a saprophyte and rare pathogenic Mycobacterium 40,41 was the most recovered NTM followed by M. fortuitum in Cameroon. This rare pathogen has been previously reported in both pulmonary and extra-pulmonary TB patients in Oregon in the USA and Asia although in lower proportions 42,43 . The diversity and species profile in Cameroon is slightly different from that reported in Chad 44 , but some commonly recovered species like M. gordonae were recovered. The former and M. fortuitum are so ubiquitous that they have previously been recovered from human, animals and the environment elsewhere [45][46][47][48] . Interestingly, a study in Nigeria found an association (OR = 2.54) with NTMs in human and seasonal Harmattan winds 49 . The vast majority of the NTMs in this study were recovered from the retropharyngeal lymph nodes consistent with oral exposure from the environment [50][51][52] .
The prevalence of bTB-like lesions after adjusting for the different population structure of slaughter cattle in each abattoir varied significantly between abattoirs with the lowest prevalence in Bamenda of ∼ 4% in the Northwest Region compared to 12-17% in the three northern Region abattoirs. Previous studies in the Northwest Region had shown a considerably lower prevalence of bTB-like lesions in the Bamenda abattoir 12 although a subsequent study suggested an increasing trend in prevalence from about 2003 to 2010 which if projected to 2012 seems consistent with our estimate 13 . It is not at all clear why there seems to have been such a dramatic increase in the last few years and this potential changing situation may have important implications particularly for the developing dairy industry in the region 53 . At the National level, the overall bTB-like lesion estimate is slightly higher than the 7.3%, 6.7% and 6.4% reported in Chad, Burkina Faso and Nigeria respectively [54][55][56] . It should be noted that cattle are known to be transported from Maroua for slaughter in N' djamena, Chad 15 , therefore the Extreme North Region of Cameroon is likely to be contributing to the lesion estimates documented in Chad. This study was a cross-sectional study and so temporal and seasonal variation, are not captured here.
The prevalence estimates for M. bovis were all slightly lower than those for bTB-like lesions but the overall pattern was not changed across the 4 abattoirs. Mycobacterium are notoriously difficult to culture so it is not surprizing that not all lesions produced a positive culture, however, given the very strong association between bTB-like lesions and a positive M. bovis culture it is likely that most of these were M. bovis.   Table 7. Multivariable logistic regression model for confirmed presence of M. bovis in an animal at two municipal abattoirs in Cameroon (n = 321).
Using data from the butchers on the market of origin of their animals, although incomplete particularly for the northern abattoirs, we were able to identify more precisely administrative division of the study area with particularly high prevalence such as Mezam in the Northwest Region and the Divisions of Mayo Rey and Benoue in the Northern Region and Mayo Kani and Diamare in the Extreme North Region. There is potentially much more mixing of cattle in the northern regions where animals are known to cross from the Central African Republic and Chad (from Sudan) and many continue on into Nigeria where there are even larger markets for meat than in Cameroon 57 . These areas are relatively isolated from the higher plateau areas of the Adamawa and mountainous areas of the Northwest and this information could help develop policies to limit the spread of bTB form these higher prevalence areas. As already mention, however, there may be biases both in the sampling as we were only able to sample for short periods in each abattoir and for Maroua and Garoua these were even shorter. We may have sampled on a particularly good or bad week and are unable to know what biases we have and so further studies in these areas is needed to build a more complete picture. Also the butchers were reporting what they believed to be the origins of the animals but may reflect the location of the last market the animal came from rather than it real origin.
The risk of having bTB-like lesions was associated with age and body condition as might be expected and reported in previous studies in Africa 58,59 . Interestingly there was a strong association with breed and risk of having a bTB-like lesion with Holstein-Frisians at high risk; ∼ 11 times higher odds compared to the white Fulani breed. Although the number of Holstein-Frisians in the study was small, this is consistent with other studies in Ethiopia 23 and in Cameroon (although the association was not statistically significant based on the authors design) 60 . Further the Gudali breed seems to be protected against bTB-like lesions when compared to White Fulani, which was supported by anecdotal reports from the butchers. Further the model also suggests there is some evidence of animals with lesions having increased odds of being in poor condition consistent with potential clinical impact of disease. Other factors which we were not able to record for these cattle but reported as risk factors are transhumance 61 and wildlife 62 . Importantly therefore even after adjusting for other variables and confounders there is a significantly higher prevalence of bTB-like lesions in slaughter cattle in the more northern abattoirs. This may reflect different stages of the ongoing epidemic of bTB across Africa and is important to understand particularly in the light of the growing dairy sector in Cameroon 53 and many other SSA countries.
The distribution of bTB-like lesions within animals showed that lesions were mostly affecting the retropharyngeal, mediastinal and bronchial lymph nodes reflecting most likely routes of infection through the respiratory route or ingestion although the dominance of site varied between abattoirs with more retropharyngeal involvement in animals slaughtered in Bamenda and Ngaoundere compared to more mediastinal involvement in those in Maroua in the Extreme North Region. We have no data to help unravel why this difference might exist but it may relate to route of exposure and factors such as feeding supplements, such as salt or cotton seed cake, which might be increasing close head to head contact and risk of ingestion. In general this finding agrees with previous reports from the Bamenda abattoir in Cameroon 12 and elswhere 22,63,64 . Interestingly there appear to be marked differences between the abattoirs in terms of lesion distribution (although the numbers are small for Garoua and Maroua) but of particular interest is that there seem to be possibly higher levels of liver involvement in Ngaoundere compared to Bamenda and this correlates with much high fluke infection levels observed at slaughter (unpublished). At the time of sampling the meat inspection procedure in Ngaoundere did not include the inspection of the carcase head (this was done by team members for the period of the study), therefore from a public health point of view more infected animals might be missed in light of the observed lesion distribution. This issue was discussed with the butchers and government meat inspectors at a stakeholder meeting after the end of the study. Previous reports from Ethiopia found high levels of infection in the respiratory related lymph nodes in housed cattle compared to digestive tract in extensively reared cattle 23 . Given that most if not all the cattle slaughtered in this study will have be extensively grazed it is unclear why there should be such a difference in lesion distribution but may be due to different strains in the different Regions.
Bovine tuberculosis remains an important disease problem both due to its impacts on livestock production and due to its zoonotic potential. The current study has highlighted the high prevalences of bTB in sections of the cattle population and this will be having significant impacts on production and animal welfare. Subsequently negatively impacting cattle rearing communities livelihoods and potential zoonotic transmission to farmers, herdsmen and slaughterhouse workers and the traders and consumers of milk 65 . The identification of M. tuberculosis in cattle further highlights the extremely complex nature of the problem with bi-directional transmission between humans and cattle in these settings. Although the current human disease burden is mostly due to M. tuberculosis, as control efforts from WHO and others as part of the "END-TB" programme by 2035 bring down this burden the importance of M. bovis and NTMs will increase 66 . This may be further exacerbated by the rapid rise in dairy production across Africa. Furthermore, these high rates of bTB in livestock populations, in this setting, make control options such as test and slaughter, as used in high-income countries, impractical. Hence in sub-Saharan Africa, possibly more than any other region of the world, is likely to benefit from development of an effective M. bovis vaccine. The levels of zTB in the associated human populations is not well documented for these areas but general estimates of ∼ 3% zTB in the human TB population represents around 100,000 cases in Cameroon 67 .
In the absence of any other controls these results emphasise the importance of maintaining high standards of meat inspection. Also given the relative low levels of knowledge of the risks of zoonotic tuberculosis in Cameroon but particularly in the Vina 68 , it also highlights the need for better public health engagement 68 and raising of awareness to reduce transmission via milk as well as direct contact.

Conclusion
Our study has shown that mycobacterial disease in cattle is widespread throughout Cameroon with M. bovis, M. tuberculosis and a number of NTMs present in cattle destined for human consumption. M. bovis infections dominated with much higher prevalences across the northern Regions (Ngaoundere, Garoua and Maroua) with higher cattle movements with neighbouring countries, compared to that in animals supplying Bamenda abattoir (Northwest Region). There also appears to be some breed differences in risk with the Fulani breeds at increased risk. This does however offer some potential for selection of more resistant breed lines. Finally the high prevalence of M. bovis and the presence of M. tuberculosis in cattle highlight the risks of (reverse) zoonotic transmission and the need for action if this important disease is to be controlled in humans and cattle.