An outbreak of tuberculosis due to Mycobacterium bovis infection in a pack of English Foxhounds (2016–2017)

Abstract Mycobacterium bovis can cause tuberculosis (TB) in social mammals including lions, cattle and man, but canine infections are considered rare. In 2016/17 we investigated a M. bovis TB outbreak in a pack of approximately 180 Foxhounds within the bovine TB Edge Area of England. We employed a combination of immunological tests including an interferon gamma release assay (IGRA) and a serological assay (DPP VetTB, Chembio). Test‐positive hounds were euthanased and subjected to post‐mortem examination (PME). Overall 164 hounds were tested; 97 (59%) responded positively to at least one test. Eighty‐five (52%) dogs responded to M. bovis antigens by IGRA while only 21 (12.9%) had detectable serological responses. At PME three hounds (3.1%) had visible lesions (VL) due to M. bovis infection, later confirmed by culture. Samples from 24 non‐VL hounds were cultured and M. bovis infection was confirmed in a further three hounds (11%). This study is the first investigation and report of an outbreak of M. bovis TB in a canine species. We establish that, in principle, diagnostic tests used for identifying infected individuals of other species can effectively be used in the dog. Further work is urgently needed to establish the sensitivity and specificity of the testing approach used in this study for future clinical application.

Discriminating between TB caused by different species of the MTBC is challenging as all member species share identical sequences across the 16s rRNA gene and 99.5% sequence homology across the remainder of the genome (Brosch et al., 2002). The most discriminating features between the species at the nucleotide level are genomic deletions, termed regions of difference (RD; Teo, Cheng, Jureen, & Lin, 2013). These have been shown to encode a variety of different virulence factors, for example, RD-1 is present on all MTBC mycobacteria other than M. bovis Bacillus-Calmette-Guérin (BCG) and M. microti and encodes for the immunodominant proteins and key virulence factors; early secreted antigenic target-6 KDa  and culture filtrate protein-10 KDa (CFP-10) (Gao et al., 2004;Guinn et al., 2004;Junqueira-Kipnis et al., 2006).
The human burden of disease is highest across Africa, South-East Asia and the Western Pacific. In South Africa, dogs have been shown to become infected with M. tuberculosis, as diagnosed using human interferon gamma release assays (IGRA), when exposed to high risk humans, that is, those with active TB disease (Parsons, Warren, Ottenhoff, Gey van Pittius, & van Helden, 2012). Similarly, African wild dogs (Lycaon pictus) have been found to be infected with M. bovis, presumptively from hunting exposure (Ayele, Neill, Zinsstag, Weiss, & Pavlik, 2004). In the UK, however, the significance of M. bovis in companion animals is largely limited to domestic cats, were frequent diagnoses are made (Broughan et al., 2013;Gunn-Moore, 2014;Pesciaroli et al., 2014;Rocha et al., 2017). Canine incidence of TB in the UK is currently considered to be rare, and almost all reported cases are limited to individual sporadic infection or small numbers of epidemiologically unrelated cases (Ellis et al., 2006;Gay et al., 2000;Liu, Weitzman, & Johnson, 1980;Park, Lim, Kwon, Bae, & Park, 2016;Parsons et al., 2012;Pesciaroli et al., 2014;Posthaus et al., 2011;Shrikrishna, de la Rua-Domenech, Smith, Colloff, & Coutts, 2009;Snider, 1971;Szaluś-Jordanow et al., 2016;Van Der Burgt, Crawshaw, Foster, Denny, & Schock, 2009  Areas and subject to additional surveillance and controls for this disease. This report describes the clinical features of the outbreak, the testing approach taken and the epidemiological aspects of the spread of disease as they are currently understood. Work to evaluate the testing approach further is currently ongoing and will be published separately. We go on to explain the new statutory controls which have been implemented in England as a result of this incident to minimize the low risk of spreading TB through the feeding of fallen stock to hounds in registered kennels.

| Index case
The hunt kennel where this outbreak occurred is situated in the south of England, UK, and housed up to 180 working Foxhounds ranging in age from juvenile puppies (<10 weeks old, n = 32), young adults (up to and including a year old, n = 21) and adults up to 8 years old (median age of 4 years).
The kennel is situated within the designated Edge Area of bovine TB incidence. The hounds work across six counties, four of which are also within the Edge Area, and two in the Low Risk Area, up to three times weekly during the peak of their hunting season (August-April).
In a similar fashion to many hunt kennels, the hounds were predominately fed raw meat, permitted offal and bone from fallen stock (so called "flesh feeding"), as permitted under Animal By-Products Clinical suspicion at this time was of leptospirosis infection as routine vaccination was, and remains, unused and clinical signs were compatible (Rissi & Brown, 2014). The hounds were not subjected to any confirmatory diagnostic testing at this time and carcasses were disposed of. When a third hound, a 4 year old entire male with no previous history of ill health, developed anorexia, plus marked polyuria and polydipsia, blood samples were taken for routine haematology, serum biochemistry and leptospirosis IgG titre testing.
The dog was euthanased and subjected to post-mortem examination (PME). Both kidneys were found to be grossly diffusely grey in colour and firmly textured. Across both the visceral and cut surfaces were multifocal to coalescing nodular, approximately round, areas containing hard white and soft black material (Figure 1). Representative samples were collected and fixed in formalin for histological evaluation by one of the authors (MD, FRCPath and experienced veterinary pathologist). Additional samples were frozen and later submitted for mycobacterial culture and PCR.

| Testing regime
Once M. bovis infection had been confirmed a prospective test and cull policy was implemented at the kennels in order to contain the spread of infection within the pack and remove any potentially infected animals. This was combined with immediate voluntary movement restrictions implemented by the kennel and increased biosecurity measures, for example, increased kennel disinfection protocols, cessation of feeding fallen stock and immediate repairs to limit wildlife access to the hounds.

| Interferon-gamma release assays
Interferon-gamma (IFN-γ) release assay tests were originally developed on the principle of quantitatively evaluating IFN-γ production by peripherally circulating antigen-specific effector memory T-cells upon in vitro stimulation, in order to aid the diagnosis of M. bovis TB (bTB) in cattle, where it has a reported sensitivity of 81.8%-100% and specificity of 88%-99% (Bezos et al., 2014;Schiller et al., 2009;Vordermeier et al., 2006;Wood & Jones, 2001). They have subsequently been adapted to identify active and latent TB in human patients with at least equivalent sensitivity and increased specificity compared to the tuberculin skin test, as well as being practically easier to perform (Eisenhut, 2014;Kim et al., 2011;Thillai, Pollock, Pareek, & Lalvani, 2014;Zhou et al., 2011). An IGRA test has been validated for use in domestic cats with up to 100% sensitivity for the detection of MTBC infections (Rhodes, Gruffydd-Jones, Gunn-Moore, & Jahans, 2008a, 2008bRhodes et al., 2011). Whereas intra-dermal testing has been shown to be of unreliable clinical utility in the dog, an IGRA has been used successfully for the detection of M. tuberculosis infections in dogs in a high risk setting (Parsons et al., 2012). The in vitro nature of the test protocol allowed us to test individual animal responses to multiple antigen combinations with technical replicates in this outbreak.
To interrogate the immune response of the hounds to mycobacteria, the antigens purified protein derivative (PPD) M. avium (PPDA), F I G U R E 2 A section of kidney from Case 0 (shown grossly in Figure 1), stained with haematoxylin and eosin. Image A shows the section at low power (×10 magnification) where multifocal areas of pathology (granulomas) are visible. Image B is a high power view of the same section (×40 magnification) and shows numerous epithelioid macrophages with neutrophils. The kidney is affected by severe, chronic active necrotizing granulomatous to pyogranulomatous nephritis F I G U R E 3 A section of kidney from Case 0 stained with Ziehl-Neelsen stain which reveals moderate numbers of intracellular and extracellular acid-fast organisms morphologically typical of Mycobacteria spp PPD M. bovis (PPDB) and a cocktail of the RD-1 specific immunodominant proteins ESAT6 and CFP-10 were selected. PPDA from M. avium is frequently used in both IGRA and tuberculin skin testing to assess for exposure/sensitization to, or infection with, environmental mycobacteria species (Pai, Dheda, Cunningham, Scano, & O'Brien, 2007;Rhodes et al., 2008a;Wood & Jones, 2001). Infection with an MTBC mycobacteria is confirmed if the IFN-γ response of any animal is greater to PPDB than to PPDA (Rhodes et al., 2008a;Wood & Jones, 2001). The presence of a concurrent response to the immunodominant antigens ESAT-6/CFP-10 indicates infection with an RD-1 positive MTBC mycobacteria (i.e., excludes infection with M. microti or previous vaccination with M. bovis-BCG) (Guinn et al., 2004;Teo et al., 2013).
The IGRA assays were conducted at Biobest Laboratories, Edinburgh in line with the assay protocol validated and previously published for use in the cat (Rhodes et al., 2008a. A 5 ml heparinized whole blood sample was taken from each hound and transported to the laboratory at ambient temperature within 18 hrs. Upon receipt, peripheral blood mononuclear cells (PBMC) were removed; blood was diluted 1:1 with Hanks Balanced Salt Solution (HBSS, Gibco, UK) and layered over Histopaque 1077 (Sigma, UK) before centrifugation at 800 g for 40 min at room temperature.
Cells were incubated for 4 days at 37°C/5% CO 2 , after which the supernatants were removed, and duplicates pooled, for quantifica- Prior to the instigation of testing, a prospective case definition was set to determine which hounds would be considered positive; it was decided that a statistically significant response to any of the three test antigens (PPDB, PPDA or ESAT6/CFP10) above the negative (culture medium control) condition was plausibly indicative of a biologically significant antigen specific T-cell response. Furthermore, such a response would be consistent with not just exposure to mycobacteria but rather would be suggestive of significant challenge and probable infection. To achieve maximum diagnostic sensitivity, and due to the fact that very little is known about the infection dynamics, clinical progression or probability of mycobacterial shedding by dogs once they are infected with mycobacteria, it was assumed that any hound with a significant IFN-γ response to at least one test antigen was infected and potentially infectious. Any such hound would therefore pose a risk to human and animal health as well as to the environment and so should be removed from the pack and euthanased.
As no cut off values or reference intervals exist for the dog, we defined the threshold for responsiveness to antigen stimulation as the mean OD value of the negative control replicates, plus two standard deviations ( x þ 2SD). Minimum antigen responses were defined as the mean OD value of the replicates, minus two standard deviations ( x À 2SD). If a calculated minimum antigen response value was greater than that calculated for the negative control, then the response was considered positive. For a test to be considered interpretable, the PMA/Ca (positive control) response had to be positive by the same criteria when compared to the negative control.
In total 164 hounds were tested by IGRA, only 11 tests failed and required repetition; the most frequent reason for this (n = 6) was that there was insufficient response to the PMA/Ca positive control, followed by clotted blood samples which precluded the isolation of PBMC (n = 4) and only one test required repetition due to significant disagreement between replicate values. All IGRA test result data are shown in Supporting Information Table S1.
Of the 164 hounds, 85 (52.0%) were found to be positive by the above case definition. Within those found to be positive, 77 (90.1%) displayed a significant response bias to PPDB which was greater than the PPDA response (which was also significantly elevated above the negative condition in 37, 48.1%, of these hounds), a pat- (1.8%) responded to PPDA and no other test antigens; whilst it was considered unlikely that these were infected with MTBC mycobacteria, they were removed from the pack, and euthanased, as a precaution.

| Serological assay
Concurrently to the IGRA testing, serological assessment of each hound was conducted. Blood was taken from each hound at the same time as the whole blood was collected for IGRA testing, 5mls of whole blood was placed into plain blood tubes and left upright at room temperature to allow clot formation. On arrival at the laboratory, blood was centrifuged at 800 g for 15 min at room temperature. Serum was removed and frozen at −20°C in 500 μl aliquots until needed. A total of 163 hounds were tested; of these 11 (6.7%) generated positive test bands visible by eye; of these, nine had antibodies directed against antigen MPB83 and two against ESAT-6/CFP-10.
No hounds were positive to both MPB83 and ESAT6/CFP10.
Hounds which produced a visibly positive test were considered to be infected, removed from the pack and euthanized.
In addition, a further 10 hounds (6.1%) showed RLU values greater than that of the majority of the remaining population and were assigned as "intermediate positive" test results ( Figure 4).
While interpreting these results is challenging, these hounds were considered to be at increased risk of infection in comparison to the background population and so were also removed from the group and euthanased. All test result data for the Chembio DPP VetTB test are given in Supporting Information Table S2.

| Screening of at risk, in-contact dogs
During the course of the outbreak, a number of dogs were identified out with the pack which had considerable potential for exposure to the hounds at the affected kennels. These dogs fell into three groups; (A) those dogs kept as pets by individuals who lived at or worked closely with the kennels and who spent a considerable amount of time with the hounds; (B) individual bitches from the affected kennel, that had been sent within the previous 18 months to other kennels for the purpose of breeding and remained at those kennels; and (C) a group of bitches that were sent from the affected kennel during the 18 months prior to the outbreak, to other kennels for purposes of breeding and were then returned, possibly bringing the infection with them.
After consultation with owners, 19 at-risk pet dogs (group A) were subjected to IGRA testing according to the same protocol as the hounds, outlined above (Section 2.2.1). Of these dogs only two were found to produce significant responses to PPDB and with a PPDB > PPDA bias, neither of these were responsive to ESAT-6/ CFP-10. Retrospective investigation revealed that one of these dogs was a hound which had been retired from the affected pack due to age-related poor performance only a few weeks prior to Case 0 becoming clinically sick. The second was a terrier dog, used as a "cutting room scrap dog" that is, it was regularly close to the preparation of carcasses before they were fed to the hounds.
In total, 13 bitches were sent from the affected kennel to one of two other kennels for breeding purposes, (groups B and C); both of these kennels are in the High Risk Area (HRA) of endemic bTB of Eng-

| Repeat interferon-gamma release assays testing protocol
Once the initial screening testing had been completed, the 66 hounds that remained were kept in isolation for 60 days. At this time, because the test accuracy of the two testing methods is unknown, the IGRA test was repeated. Nine hounds were found to be IGRA positive, none were visibly lesioned at PME.
Following these results it was decided that the remaining 57 hounds were unlikely to be infected/infectious and so voluntary restrictions on the kennels were lifted. The hounds continue to be closely monitored for any change in appetite, body weight and body condition score. All of the remaining hounds remain well at the time of writing.

HUMANS
Mycobacterium bovis is a known zoonotic bacteria. Once infection was confirmed by laboratory culture from samples taken from the hounds, Health Protection England (HPE) was informed to assess the risk to human health. An in-depth risk assessment was conducted, and contacts were stratified in to risk pool. A 'stone in the pond' approach was adopted for screening, with those at highest risk of exposure screened initially. This investigation and its findings are the subject of an additional study, where 11 people were screened (Phipps et al., submitted). One asymptomatic exposed person has tested positive for TB on initial screening by IGRA and has since been diagnosed with latent TB. Due to the nature of the contact between this individual and the infected hounds, it remains possible but unproven that the person was infected by contact with either the hounds and/or contaminated bovine material.

| Risk pathway identification
The APHA conducted an epidemiological investigation into the source of the outbreak to investigate possible transmission pathways. A qualitative risk assessment approach was used to address the likelihood of each of these routes of infection. The more probable contenders in order of likelihood were: However, not all animals could be followed up because of missing records or mortalities, only ten of the 16 traced animals were alive and available for testing at the time of the outbreak.

| Feeding of M. bovis-infected fallen stock (ingestion)
Details of the fallen stock collected from farms and fed to the hounds in the kennels over the 12-month period prior to the first hound death were examined to provide an evaluation of the likely risk posed. A total of 24 carcases were sourced from an Approved Finishing Unit (AFU) for negative-testing cattle from TB restricted farms. However, no tuberculous lesions had been confirmed at postmortem meat inspection in the abattoir in any cattle from the unit over the previous 3 years, reflecting the fact that most were sourced from low cattle TB incidence areas, that is, from across the Low Risk Area and/or low incidence parts of the Edge Area. Apart from the AFU, six carcases were collected from farms with a history of TB breakdowns caused by genotype 10:

| Exposure to infected livestock or wildlife during exercising (inhalation/ingestion/biting) either directly or via M. bovis-contaminated environment
About 85% of the geographical area to which the hounds were exposed had low cattle TB incidence and no indication of wildlife infection. The remaining 15% had some TB cattle breakdowns where APHA investigation indicated possible wildlife sources. The very infrequent reporting of M. bovis in farm dogs suggests that farmbased transmission pathways are not very effective (Wilkins et al., 2008). Kennel staff reported that direct contact between wildlife and working hounds during exercise was unlikely. However, the worst case scenario of an encounter with an infectious badger, or infected carcase could have been missed. The likelihood of infection whilst exercising was considered very low, but with high uncertainty associated with wildlife prevalence and general lack of quantification of the transmission pathways from the environment.

| Exposure to infected local wildlife at the kennels (inhalation/ingestion/biting)
The kennels are located in an area of the country where cattle TB incidence is low and there is no evidence of wildlife infection.
Access to the kennels by larger wildlife such as badgers was very unlikely. Likelihood was considered very low with low uncertainty.
Once in the kennels, infection appears to have spread horizontally between the hounds, suggesting that in certain conditions such As a proportionate risk mitigation strategy, DEFRA has introduced tighter restrictions on the collection and feeding of fallen stock to hounds in registered kennels. Since 10 October 2017 the feeding of offal from livestock species to dogs from recognized kennels or packs of hounds has been banned in England (Anonymous, 2017 M. bovis to spread readily between social animals, it is not surprising that a large group of kennelled hounds presented a pool of competent, susceptible hosts once the pathogen was introduced. Also remarkable in this outbreak was its fulminant nature; from the occurrence of the index case until the resolution of the outbreak (7 months), nearly one in ten animals on the premises became clinically sick. TB is usually considered a chronic disease with only a small percentage of infected individuals becoming clinically unwell (Eisenhut, 2014). Therefore, this rapid progression to disease in a relatively high proportion of animals can be considered unusual, although not unheard of, as sporadic fulminant human cases of renal TB leading to death by acute renal impairment have been reported (Adzic-Vukicevic et al., 2017;Dissanayake, 2004;Isao et al., 2006;Pathan et al., 2001;Punia & Kumar, 2008;da Silva Junior, Brito, Rabelo, & Saboia, 2016).
The reasons for the fulminant nature of this outbreak are likely to be multifactorial; a combination of pathogen virulence, host genetics and phenotype, as well as environmental conditions. There is a possibility that the pathogen became host-adapted. However, all animals from which viable organism was isolated were found to be infected with genotype 10:a. It may therefore also be the case that  nez-Miguel, 2016). This may also be the case for dogs; Foxhounds have previously been reported to be affected by renal amyloidosis (Mason & Day, 1996), and this kennel has had previously confirmed clinical cases, and one of the M. bovis infected hounds was found to be positive for renal amyloidosis (data not shown).
Other conditions which adversely affect the outcome of renal TB are the deposition of anti-mycobacterial antibodies within the glomeruli and/or hypercalcaemia which is a common complication of TB and is directly nephrotoxic (Bellendir, Kochanova, & Nakonechny 1978;Chan et al., 1994;Ko et al., 2004;Zhao, Sun, Xu, & Sun, 2017). It is possible that either of these conditions could have had an impact on the number of dogs affected and the severity of disease seen in this outbreak.
The relatively high level of renal immune-privilege may have allowed mycobacteria to replicate there whilst avoiding immune surveillance and clearance (Kurts et al., 2001). All of the hounds with renal lesions also had positive urine cultures (where these were available), which suggests that this could be a major source of infection for onward dog-to-dog transmission. However, if contaminated urine was the main source of infectious material, for example, via aerosolization, then lung pathology would be expected, but there were no pulmonary lesions in any of the dogs subjected to PME. In addition, the urine cultures were only found to be positive after 14 weeks of culture suggesting that the initial number of organisms present was low. Based on the distribution of lesions it would appear that the mycobacteria arrived in the kidney by haematogenous spread. If so this indicates a distant source of infection which has not been identified.
Infected urine was identified as a key risk factor for potential human exposure during the risk assessment of kennel staff, predominately due to the use of power washers to clean excreta from the hard standing, which posed a significant risk of aerosolization (Phipps et al., submitted). The outcome of TB screening for at-risk humans at the kennels was that a single individual was diagnosed with latent TB (Phipps et al., submitted). If urine from infected dogs was heavily contaminated with M. bovis it would have been likely that a higher proportion of individuals at risk would have been test positive.
Based on these observational data, it is possible to conclude that M. bovis infected dogs are likely to be infectious to each other, and their urine may contaminate their local environment, but that this may not be the only or even the major route of transmission, and that the risk to human health is low.
No ante mortem tests have been validated for the diagnosis of M. bovis infection in dogs; we therefore combined two established testing methods for other species to try and maximize our diagnostic sensitivity and bring this outbreak under control. The tests used here target the disparate arms of the immune response, with the IGRA evaluating the cell mediated Th1 response whilst the serological assay used to detect a humoral Th2 response. The percentage of tested animals deemed to be positive by each test was markedly different with 52% of dogs classed as infected by IGRA at initial screening but only 12.9% of dogs classed as infected or "at risk" by serology. This difference is likely due to the dominance of cell mediated immunity in an animals' response to mycobacterial infection (Khan, Hunter, & Jagannath, 2016) but further investigation of the canine immune response to M. bovis and mycobacterial antigens is needed before this can be shown conclusively, this work is currently ongoing and will be published separately.
Whilst uncertainty exists with regards to the source of infection, it seems likely that contaminated fallen stock carcasses were involved. All culture positive animals had the same genotype isolated, which has not been frequently detected in the area local to the kennels. This suggests a single introduction with a large dose of infectious material was likely and that infection then spread between cohabiting hounds (Reynolds, 2006). One kennel worker was also diagnosed with latent TB, potentially due to exposure to infected hounds and/or their contaminated feed, though this remains unproven. Though the risk posed by M. bovis infected hounds to other hounds, cattle herds, local wildlife, their environment and their human keepers is low, this outbreak demonstrates that the risk is plausible and real. Resultantly, to try to prevent such an outbreak from occurring again, the conditions for the collection and use of animal by-products for feeding to dogs in recognized kennels or packs of hounds have now been amended to ban the feeding of offal from fallen stock. Similarly, health surveillance requirements for hounds have been increased by the MFHA. To be successful, these policy changes will need to be aided, however, by the future evaluation of diagnostic test accuracy and test development for canine TB.

ACKNOWLEDG EMENTS
We thank all of the staff and management team of the kennels affected by this outbreak for facilitating the investigation and giving permission for its findings to be published. One of the authors,

Conor O'Halloran, is supported by a Biotechnology and Biological
Sciences Research Council (BBSRC) studentship. The remaining authors received no financial support for the research, authorship, and/or publication of this article.

CONFLI CT OF INTEREST
The authors declare no conflicts of interest.