Genomic variations associated with attenuation in Mycobacterium avium subsp. paratuberculosis vaccine strains

Background Mycobacterium avium subspecies paratuberculosis (MAP) whole cell vaccines have been widely used tools in the control of Johne’s disease in animals despite being unable to provide complete protection. Current vaccine strains derive from stocks created many decades ago; however their genotypes, underlying mechanisms and relative degree of their attenuation are largely unknown. Results Using mouse virulence studies we confirm that MAP vaccine strains 316 F, II and 2e have diverse but clearly attenuated survival and persistence characteristics compared with wild type strains. Using a pan genomic microarray we characterise the genomic variations in a panel of vaccine strains sourced from stocks spanning over 40 years of maintenance. We describe multiple genomic variations specific for individual vaccine stocks in both deletion (26–32 Kbp) and tandem duplicated (11–40 Kbp) large variable genomic islands and insertion sequence copy numbers. We show individual differences suitable for diagnostic differentiation between vaccine and wild type genotypes and provide evidence for functionality of some of the deleted MAP-specific genes and their possible relation to attenuation. Conclusions This study shows how culture environments have influenced MAP genome diversity resulting in large tandem genomic duplications, deletions and transposable element activity. In combination with classical selective systematic subculture this has led to fixation of specific MAP genomic alterations in some vaccine strain lineages which link the resulting attenuated phenotypes with deficiencies in high reactive oxygen species handling.


Background
Mycobacterium avium subspecies paratuberculosis (MAP) is a proven enteric pathogen with a wide host range that includes many domestic and wild animals [1]. It is the causal agent of Johne's disease (JD) in animals which is particularly common in countries with significant dairy industries leading to considerable economic losses [2]. MAP can infect, disseminate and persist in humans and has been suggested as a contributory factor in the development of Crohn's disease [3].
MAP vaccines are a major tool used in the control of JD in animals and can be highly profitable [4]. They have advantages over herd management [5] and culling strategies [6] in being more cost efficient, easier to implement on a wide scale and less reliant on diagnostic testing. It is clear however, that although able to prevent a majority of animals from reaching onset of clinical disease, their current formulations provide incomplete protection against infection and shedding [7][8][9], thus failing to eradicate the organism [10]. Most current whole cell vaccine preparations rely on subcultures of classic strains that were generated over 70 years ago [11] and some evidence suggests that, for killed preparations at least, more recently acquired local virulent strain types may be more effective [12]. Previous experience with BCG has shown that frequent in vitro passage of strains in different laboratories led to significant alterations in genomic profiles and diversities in attenuation and immunogenicity [13]. It is of importance therefore to derive accurate definitions of MAP vaccine genotypes to better standardize vaccine manufacture and understand the critical mechanisms determining vaccine attenuations and protective efficacies.
The distribution and worldwide use of MAP vaccines has continued since live 'attenuated' strains were selected in France (1924) and the UK (1940) using a method of sequential passage similar to that applied for the generation of BCG [14]. The degree and mechanism underlying their attenuation however is uncertain as virulence studies were not performed in any detail. Concerns in the 1980's regarding the use of live vaccine strains because of low shelf life and spread to the environment promoted the use of killed vaccine formulations. These were based on various combinations of three MAP strains comprising strain 2e from the UK, strain II from Canada and 316 F. A combination of these were used as live preparations for Norwegian goats in the 1960's and 1970's [15] and as a killed preparation delivered in mineral oil to Icelandic sheep in the 1950's [16]. Live vaccine formulations of 316 F alone were used in the 1960's and 70's in the UK [17] and Cyprus [18], 1980's in Hungary [19], 1990's in Germany [20] and Spain [21] and up until 2002 in New Zealand [22]. Killed preparations of 316 F alone have been used extensively worldwide [23] and are still available for commercial use. These strains, due to the difficulty in retaining mycobacteria in frozen seed stocks, have been maintained through regular subculture on a variety of laboratory in-house media. It is unsurprising therefore, that some reports suggest strain adaptation to growth in specialized media with loss of Mycobactin J dependence [24] and genome diversity [25] has occurred amongst some lineages.
In this work we demonstrate attenuation and differential virulence of vaccine strains 2e, II and 316 F in a mouse model and use a full MAP genome microarray, supported by PCR and sequencing to investigate the genomic shifts of vaccine strains from a variety of lineages, including one recently resuscitated 316 F strain, originally lyophilised in 1966. We describe large genomic regions with deletions and tandem duplications uniquely associated with each vaccine clade, demonstrate the functionality of some of these deleted genes and hypothesise as to their role in virulence attenuation.
Signal ratios from genes in multiple copies in the genome such as insertion elements are more difficult to interpret as changes in signal are related to copy number. Nevertheless IS900 signal ratios were significantly reduced in IIUK2000, 2eUK2000, 316FNOR1960, 316v, 316FUK2000 and elevated in 316FNLD1978 and 316FNLD2008. Increases in signals for IS1652-like elements were also seen in IIUK2000, 2eUK2000, 316FUK2000, 316FNLD1978 and 316FNLD2008 and decreases for this element in 316FNOR1960. IS256-like elements were more abundant in 2eUK2000 and also raised in 316v and 316FUK2000 but decreased in 316FNOR1960 relative to MAPK10. IS1311like elements were increased in 316v and decreased in 316FNOR1960. One of these elements (MAP0104) borders the previously described vGI-1b region which is bounded by short inverted repeat sequences and duplicated in MAP type III strains [26].
The tandem duplication vGI-21 was confirmed using primer set MAP2705.seq2 and MAP2733.R (Table 6) orientated to amplify only the end of one tandem repeat and the beginning of the other. This produced an amplicon of 651 bp only with vaccine strain IIUK2000 and IIUK2001 (Additional File 1). Sequencing of this amplicon showed the duplication event to occur relative to the reference strain K10 genome (GenBank accession AE016958) at position 3066285 within the first copy of MAP2733c (truncated at aa251) followed by a unique region of 16 bp (CCAGCTCGGCCACCCG) and then a second copy of MAP2704 (truncated at aa138) followed by the vGI-21 duplication thus comprising a tandem duplication of 24,971 bp spanning 29 ORFs (MAP2705c-MAP2733c). PCR using primer pairs specific for this tandem duplication MAP2705.seq2 and MAP2733.R was negative for all other strains included in this study.
vGI-22 was confirmed as a tandem duplication using primer set MAP1789.F and MAP1750.R (Table 6) which produced an amplicon of 2967 bp with vaccine strain 316 F-NLD1978 only. Sequencing of this region showed the duplication event to comprise 40,744 bp spanning 41 ORFs (MAP1749-MAP1789). The duplication site occurs at Genbank accession AE016958 position 1952589 within the first copy of MAP1790 (truncated at aa173) followed by an overlap region of 3 bp (GGG) and then the second copy of MAP1748c (truncated at aa143) followed by the vGI-22 duplication. PCR with primer pairs specific for this tandem duplication (MAP1789.F, MAP1750.R) was negative for all other strains included in this study.
Several attempts to localise the vGI-1b duplication using outward facing primers around the region in a similar manner to vGI-21 and vGI-22 were unsuccessful and this region may not be present in 316UK2000 as a tandem duplication. qPCR was performed using both MAP0101 and MAP0103 specific primer pairs located inside vGI-1b and the relative copy number of each compared against a single copy genome target control represented by primer pairs to MAP2114c. Both MAP0101 and MAP0103 pairs showed a doubling of signal strengths relative to MAP2114c in strain 316FUK2000 whilst all other strains included in this study showed signal consistent with single copies of both these genes (Table 7).

IS900 insertion site analysis
To determine which IS900 sites were absent relative to the K10 reference genome, PCR primers were designed to specifically amplify each of the known 17 IS900 loci (Table 6). These were used to confirm the insertion of IS900 into each locus in the reference strain K10 and were also all positive in all 316 F strains and a caprine isolate CAM87. Both vaccine strains IIUK2000 and 2eUK2000 were missing IS900(MAP1722) whereas IS900(MAP1033) was also missing from vaccine strain 2eUK2000 but present in all other strains including vaccine strain IIUK2000. Comparative qPCR of IS900 copy number relative to MAP2114c, demonstrated a range of IS900 copies in vaccine strains that corresponded to the trend in hybridisation signals observed in MAPAC scatterplots (Figure 1a & 1b). The ratio of copy number however was surprisingly higher than predicted, with vaccine strains IIUK2000 and 2eUK2000 having only 13 copies whilst MAPK10 and 316 F strains gave signals correspondent with 16-19 relative IS900 copy numbers (Table 7). a b Figure 1 Microarray scatterplots comparing genomes of test MAP vaccine strains against MAP K10 reference strain.

Functional analysis of tellurite resistance
One MAP specific gene predicted to be deleted in vGI-19 was MAP3730 (Table 1), a S-adenosylmethionine -dependent methyltransferase with homologues to tellurite resistance genes (tehB) involved in bacterial virulence and persistence [27,28]. The functionality of this gene in mycobacteria has not previously been investigated. Using a solid culture plate assay we compared tellurite resistance (MIC) of MAP strains with and without the vGI-19 deletion (Table 7). This demonstrated a wide MIC range (8 ->512 μg/ml) between strains, with significant reductions associated with vGI-19 (316FNOR1960) deletion over full genome complement strains. Of note however was the very low level of tellurite resistance (8 μg/ ml) found in strains containing the vGI-20 (IIUK2000 & 2eUK2000) deletion.

Assessment of virulence using a mouse model
The virulence of vaccine strains 316FUK2001, IIUK2001 and 2eUK2001 was compared with wild type strain JD87/107 in a mouse model. Ten mice from each of five groups (four inoculated with the different MAP strains and a negative control group inoculated with PBS) were killed at 4, 8 and 12 weeks post inoculation. Body, spleen and liver weights were recorded. Samples of the liver were taken for bacteriological culture and histopathology. Mean bodyweights increased with age, but no statistically significant difference was observed in mean body weight between any of the vaccine strains and the control wild type strain at any of the time points (p=0.11). There was strong evidence of a difference in pattern over time between the mean rank spleen weight expressed as a percentage of body weight of the mice inoculated with the different MAP strains relative to those receiving PBS, where those inoculated with MAP had increasingly higher mean rank percentage weights (P=0.006). Differences between the MAP strains were not formally statistically significant (p=0.06) although the control virulent strain JD87/107 showed an increase in mean rank spleen weight percentage between weeks 4 and 8, and 316FUK2001 had an increase between weeks 8 and 12. There was no statistical evidence for differences in the mean levels of liver weight expressed as a percentage of body weight either for different strains or over time for any of the MAP strains (p = 0.2). However, there was some evidence of a difference between the means for the MAP strains and the lower mean weights associated with PBS (p = 0.018). MAP was recovered from the liver tissue of mice four weeks post inoculation in all groups except the control group inoculated with PBS. By 12 weeks post infection, MAP was recovered from the tissues of only one mouse inoculated with vaccine strain 2eUK2001 (mean count 46 cfu/g), from 6 mice inoculated with IIUK2001 (mean counts between 46 and 315 cfu/g) and from all the mice inoculated with the virulent JD87/107 strain (mean counts 1.4-7 × 10 6 cfu/g) suggesting attenuation of each of the vaccine strains ( Figure 2a). Mean rank counts increase over time for the JD87/107 strain, while dropping for all the other MAP strains, this being most rapid for the 2eUK2001 strain but ultimately most notable for strain 316FUK2001. Statistical assessment of the effect of the strain by time interaction on the mean rank count indicate that differences exist in the abilities of the MAP vaccine strains to survive or persist in mice (p=0.02).
Infections of the liver result in multifocal hepatitis characterised by clusters of inflammatory cells. To measure the effect of the experimental infection with MAP, the densities of mononuclear leucocyte clusters were counted. Small numbers of leucocyte clusters (< 0.22 clusters per field) were observed in liver samples from mice inoculated with PBS, though no acid fast bacilli (AFB) were detected in any of these samples. Large differences in the mean ranked density of leucocyte clusters between strains were identified (p<0.001) with the wild type strain JD87/107 having the highest mean ranked densities of clusters ( Figure 2b). Strain 2eUK2001 showed evidence of higher mean rank densities than the 316FUK2001 and IIUK2001 strains (p = 0.03). The ranked density of leucocyte clusters with AFB showed highly statistically significant differences between the means of MAP strains (p<0.001), with the JD87/107 strain consistently showing higher mean densities, with this effect being more pronounced from 8 weeks post infection (Figure 2c). The vaccine strains all tended to exhibit increasingly lower mean ranked densities over the lifetime of the experiment (p=0.002), with consistent patterns of differences between strains (p=0.008): strain IIUK2001 showed the largest mean rank densities, strain 316FUK2001 the lowest, with 2eUK2001 intermediate. The histopathology results show that all strains elicited a similar inflammation at 4 weeks. Only thereafter some differences between the inflammatory responses to the strains became apparent. In addition, the analysis of mean bacterial counts and AFB positive clusters showed the reduced ability of the vaccine strains to survive and persist within mice. Overall, these results provide proof of attenuation of the vaccine strains with respect to a wild type MAP strain.

Discussion
In this study, we examined genomic and phenotypic characteristics of a panel of MAP vaccine strains obtained   (Table 6), located within the vGI-20, on 2eUK2001 and IIUK2001 were negative confirming that the vGI-20 region was not present in these strains (Additional File 1). The inclusion of MAP-specific genes in these deleted regions is an important observation as these genes could provide the basis for differentiating infected from vaccinated animals (DIVA). Indeed, the deleted region vGI-19 contains part of the 38 kb pathogenicity island described by Stratmann et al. (2004) [30] which contains genes encoding a number of antigens with diagnostic potential [31]. Both deleted regions vGI-19 and vGI-20 contain genes potentially involved in virulence and pathogenesis (Table 1) and their deletion could therefore have a profound effect on the virulence of these strains. In this study we demonstrated using a mouse model that both vaccine strains 2eUK2001 and IIUK2001 were attenuated with respect to a wild type MAP strain. In addition, vaccine strain IIUK2000 and IIUK2001 were found to contain a large 41 ORF tandem duplication (vGI-21) which includes copies of benzoate and lipid metabolic pathways. Vaccine strain 2eUK2000 comes from the same stock as 2eUK2001 and was maintained at the VLA, UK for over 50 years on a mineral deficient medium (Watson Reid ' A' block) whilst vaccine strain IIUK2000 was not. We suggest that the vGI-21 duplication in vaccine strain IIUK2000 was selected by these differences in media and fixed into the genome to compensate in vitro for the deletion of lipid biosynthesis and carbon usage repertoires, removed by the vGI-20 deletion.
The large deletion vGI-19 present in vaccine strain 316FNOR1960 was not present in any of the other 316 -F strains including an early low passage lineage (316FCYP1966) and a more recent isolate (316 F2001) shown to be significantly attenuated in our virulence studies. Notably, part of vGI-19 is also present in the same gene order within the related MAH104 genome (GenBank reference CP000479). Together these suggest that any ancestral precursor and therefore the original 316 F strain would be unlikely to be missing vGI-19. We hypothesise that the vGI-19 deletion appeared in the 316FNOR1960 strain some point after its acquisition and transfer to Norway in 1960 from the VLA, UK. This strain is recorded as having been maintained, uniquely on Dubos medium with added pyruvate [15] and we hypothesise that this medium was at some point selective. This is supported by the vGI-19 deletion in this strain including gene homologues of glyoxylate enzymes associated with pyruvate metabolism [32]. This strain previously has been used successfully as a live vaccine suggesting that it is attenuated. The knockout of the glyoxylate shunt could significantly affect the strain's ability to control anaerobic respiration [33] and intracellular persistence [34], which may indicate that attenuation in this strain may be related to this loss. A 2002 isolate of this strain was shown to be highly immunogenic in goats [35] and forms part of a 316 F/2e live commercial vaccine product (Paratuberkulose-vaksine, National Veterinaria Institute, Oslo, Norway) however there is no evidence that  vGI-19 was deleted when the strain was used, albeit very effectively, on goats in the 1960-70's [15]. We tested the potential impact provided by deletion of the putative tellurite resistance gene (tehB) included in vGI-19 on 316FNOR1960 phenotype. Tellurite is highly toxic to bacteria due to its action on DNA synthesis. It is an important mechanism by which animals combat intracellular microorganisms [27] and was used in early studies as a tuberculosis/leprosy therapeutic [36]. Bacterial resistance to tellurite is inducible, is associated with virulence [28] and is linked to catalases which are required to process the superoxide anions generated as a result of bacterial metabolic mechanisms used to inactivate tellurite. We show a significantly increased sensitivity to tellurite in 316FNOR1960 whilst other 316 F strains either matched or exceeded the resistance of the two wildtype strains tested (K10:bovine, CAM87:caprine). Interestingly the strains most sensitive to tellurite were IIUK2000 and 2eUK2000 which lack the tehB gene. The metabolism of tellurite generates high reactive oxygen species which subsequently need to be de-toxified by catalase [37]. Significantly the vGI-20 deletion in these strains includes loss of the catalase gene homologue MAP1725c. Both vaccine deletion regions thus involve alterations in metabolic pathways associated with deactivation of high reactive oxygen species toxicity, which suggests this may be an important mechanism underlying attenuation in these strains.
Several of the other vaccine strains tested are also reported to have been maintained on markedly different media which may have similarly promoted or selected for genomic and phenotypic diversities. 316FNLD1978, available as a heat killed vaccination for dairy cattle since 1985 [38], was found to contain a large tandem duplication (vGI-22) unique to this strain. It is notable that this isolate was selectively subcultured on potato starch medium to enhance its growth (P. Willemsen personal communication) and now grows with difficulty on other media. It is tempting to speculate that the acquisition of extra copies of 14 ORFs including cell wall, fatty acid biosynthesis genes and two extra copies of IS900 are a direct result of the selective process performed on this strain.
We demonstrated in this study that vaccine strain 316FUK2001 was clearly attenuated with respect to wild type MAP strain JD87/107. The vGI-19 deletion found in 316FNOR1960 and the vGI-20 deletion found in 2eUK2000 and IIUK2000 were not detected by PCR in this strain suggesting that attenuation in this strain is due to different genetic polymorphisms. A duplicated region (vGI-1b) was detected in vaccine strain 316FUK2000, which may possibly have arisen as an adaptation to growth on liquid Watson Reid media.
Insertion sequences bordering specific genomic regions have previously been associated with variably tandem duplicated [39] and possibly large genomic inversions [40] and contribute highly to the degree of plasticity in the MAP genome. Variations in copy number of insertion elements including IS900, IS1311, IS256 and IS1652-like elements were seen between vaccine strains and virulent isolates. An IS1311 was found immediately bordering the vGI-1b region duplicated in 316 F-UK2000 but not other 316 F strains. Similar genomic variations including vGI-1b  have been observed in virulent MAP strains [26]. IS900, a definitive element of MAP found in all clinical and vaccine strains, was also shown to be present in a variety of copy numbers. This work used comparative ratios of qPCR signals to estimate the average number of IS900 copies per cell per culture relative to two single copy MAP genes using an assumption determined from a MAP assembled genome sequence that MAPK10 would contain 17 copies. Our results confirm previous studies showing the vaccine strain 316v used in Australia for ELISA testing [41] contains one less genomic copy of IS900 [42] than most other 316 F strains [25]. Vaccine strain 316FNLD1978 exhibited higher gene signal ratios consistent with the two extra copies of IS900 copies inside the duplication of vGI-22. Vaccine strains IIUK2000 and 2eUK2000 contained lower signal ratios consistent with loss of an IS900 copy inside the deletion region vGI-20. Consistently however the calculated IS900 copy number in these strains was much lower than expected using the ratio method. Using site specific PCR we confirmed 16 IS900 filled insertion sites in the genomes of these strains whereas the ratio method, using MAPK10 as a standard, predicted only 13 copies. The reason could be technical, perhaps involving incomplete bacterial lysis of these unusual strains, however IS900 is known to replicate in episomal minicircles [43] and when all consensus insertion sites are filled they may exist as extra genomic components awaiting transposition. If this is indeed the case, virulent MAP strains would have the capacity to contain more than the predicted 17 IS900 copies per cell. This could be an important factor in studies relying on qPCR to determine accurate estimates of MAP load [44].
MIRU3 is a short tandem repeat sequence located within the sensX3-regX3 two component signalling system that controls carbon source usage and mechanisms reducing damaging reactive oxygen species generated by aerobic metabolism [45]. The attenuated BCG vaccine characteristically contains a low MIRU3 tandem repeat copy number which has been suggested to be involved in the control of sensX3-regX3 expression [46]. In this study 316 F strains (316FNLD1978, 316FUK2001, 316FNLD2008) had low MIRU3 copy numbers whilst others, mostly originating from older culture stocks, were larger. This may represent a potential for genomic drift within recent 316 F lineages and could be used as a marker for the most current 316 F VLA, UK strain but may also be suggestive that loss of intracellular persistence in full genome complement 316 F strains may be associated with reduced transcriptomic signalling capacity. It will be of interest therefore in future total genome sequencing studies to compare dysfunctional SNP variations within signalling features of 316 F strain genomes.

Conclusions
This study has shown that significant genomic diversity exists between MAP vaccine strains and within the 316 F lineage. These include large deletions, duplications and changes in insertion sequence copies. These mutations were probably derived in a classical manner by selective subculture on laboratory media and in some cases have led to significant alterations of phenotype and attenuation. There were 25 MAP specific gene deletions identified of which at least one could be linked to phenotypic change that would disadvantage its persistence in the host and thus associates it with virulence. Furthermore, these MAP-specific gene deletions could provide the basis for a DIVA diagnostic for use with these vaccines. Overall, this work illustrates that MAP genome plasticity can be influenced by in vitro culture over long periods and a robust definition of vaccine strain genome lineage will be necessary in the future to guarantee consistency between studies.

Strains and culture media
MAP strains used in this work, their origins, sources and media used for propagation are described in Table 8.

DNA extraction
DNA was extracted for typing and arrays using a previously described protocol [26]. Briefly, 1x10 9

MAPAC microarray hybridisation and analysis
DNA from the test strain and reference MAP K-10 strain were fluorescently labelled and hybridised to the microarray using protocols described previously [50]. Briefly, 1 μg of DNA was labelled by random priming with Klenow polymerase to incorporate either Cy3 or Cy5 dCTP (GE Healthcare) for the test strain and reference strain respectively. Equal amounts of the Cy3 and Cy5 labelled samples were co-purified through a Qiagen MinElute column (Qiagen), mixed with a formamidebased hybridisation solution (1×MES, 1 M NaCl, 20% formamide, 0.02 M EDTA, 1% Triton) and denatured at 95°C for 2 min. The labelled sample was loaded on to a prehybridised (3.5×SSC, 0.1% SDS, 10 mg/ml BSA) microarray under two 22×22 mm LifterSlips (Erie Scientific), sealed in a humidified hybridisation cassette (Corning) and hybridised overnight by immersion in a waterbath at 55°C for 16-20 h. Slides were washed once in 400 ml 1×SSC 0.06% SDS at 55°C for 2 min and twice in 400 ml 0.06×SSC for 2 min. Microarrays were scanned using an Affymetrix 428 scanner, and signal intensity data were extracted using BlueFuse for Microarrays v3.5 (BlueGnome). The intensity data was further postprocessed using BlueFuse to exclude both controls and low confidence data (p<0.1) prior to normalisation by 2D Lowess (window size=20) and median centring. Further analysis of the normalised data was undertaken using BlueFuse, GeneSpring 7.3.1 (Agilent Technologies) and Eisen Cluster [51]. Fully annotated microarray data has been deposited in BμG@Sbase (accession number: E-BUGS-264; http://bugs.sgul.ac.uk/E-BUGS-264) and also ArrayExpress (accession number: E-BUGS-264).

Tellurite MIC
Cultures of MAP strains were grown in conventional liquid media to exponential phase for 6 weeks then adjusted to 10 4 cfu/ml using OD550. Aliquots (10 μl) were inoculated onto solid RAF medium in petri dishes containing serial dilution of potassium tellurite to final concentrations of 512, 256, 128, 64, 32,16,8,4 or 0 μg/ml and incubated at 37°C. MIC's were taken as the least tellurite concentration able to inhibit >90% growth, seen as black colonies, after 6 weeks of growth and 12 weeks growth for strain IIUK2000 which was slower to grow in vitro.

Assessment of virulence using a mouse model
The virulence of vaccine strains 316FUK2001, IIUK2001 and 2eUK2001 was compared with wild type strain JD87/ 107 in a mouse model. 316FUK2001, IIUK2001 and 2eUK2001 were selected to represent the three different vaccine strains that have been used for control of JD over the years. JD87/107 was selected as the control strain as this was the virulent wild type strain that was used previously in our laboratory to optimise the mouse model and PBS was used as a negative control. C57BL/6 mice of approximately five weeks old and between 20 and 25 g in weight were purchased from Harlan UK, Shaws Farm, Blackthorn, Bicester, Oxon OX25 1TP. The mice were individually weighed and randomly assigned to five groups of 30. One negative control group was inoculated with 0.1 ml of sterile PBS. The remaining groups were inoculated intraperitoneally with 1.1 to 1.4 × 10 8 organisms in 0.1 ml PBS of one of the MAP strains 2eUK2001, IIUK2001, 316FUK2001 and JD87/107. The inocula were prepared as previously described [52] and enumerated by performing a microscopic count. Ten mice from each group were killed at 4, 8 and 12 weeks post inoculation by exposure to a mixture of carbon dioxide and halothane gas followed by cervical dislocation. Each mouse was weighed and the body weight recorded. The spleens and livers were removed aseptically and weighed. The respective weights were expressed as a percentage of body weight for each mouse. Approximately 0.1 g of liver was removed for