This study aimed at determining the strain diversity of MAP isolates responsible for paratuberculosis in Ugandan cattle and comparing them to strains reported from other countries. The results show that most of the MAP strains examined in the study (17 of 21) belonged to the cattle type of strains, while two isolates were of the bison type. A novel restriction pattern was revealed in two isolates (MapUg4 and MapUg2) indicating a type that has not been previously reported. This type was tentatively designated as X pending further characterisation. Cattle strains have been found in most areas where MAP has been detected in cattle. The bison type was first reported in the USA on cattle farms where bison infected with the same strain had been previously kept [10, 17, 18]. This strain has been reported to be widespread in India among buffaloes, cattle and goats [17, 19]. To our knowledge this is the first time the bison type is observed outside the USA and India. Since this is the first characterisation study of MAP in Uganda, it is not clear how widespread this type might be but it has the potential of spreading into more hosts than the cattle type [17, 20].
Based on the SSR locus 1, 15 isolates were divided into two allele groups having 6 and 7 G mononucleotide repeats respectively; while three allele profiles were revealed for SSR locus 2 (10 G, 11 G and 12 G) and four allele profiles for SSR locus 8: 4GGT, 2GGT and two polymorphic 3GGT repeats carrying mutations respectively, for all 21 isolates examined. Therefore, based on our results, the examination of the three loci is valuable in discriminating MAP isolates in this region, as has also been shown by El-Sayed et al. . Although it has been reported that loci 1 and 8 were the most discriminative , in our study the three SSR loci complement each other and the highest level of discrimination is reached when all the three are used.
Harris et al. , found that the majority of the 211 isolates from the USA had more than 14 G1 and 9 G2 repeats for SSR loci 1 and 2 respectively and 5 GGT for SSR locus 8. On the other hand, El-Sayed et al.  found 7–14 G1 repeats, the majority being 7 G1 repeats, for SSR locus 1; 9–13 G2 repeats for SSR locus 2, the most common being 11 G among 34 German isolates. Regarding SSR locus 8, El-Sayed et al.  found that their German isolates had 4GGT and 5GGT repeats. In another study where SSR has been used, the most common profiles of SSR loci 1, 2 and 8 were >14 G-9 G-5GGT, >14-10-5GGT . No study prior to this one has shown the occurrence of MAP strains with 6 G1 repeats in SSR locus 1, and 2GGT. Polymorphic 3GGT repeats has been observed in sheep strain  but the nature of the polymorphism was not defined.
According to Sevilla et al.  and Motiwala et al. , there is a relationship between the number of repeats in the three SSR loci of a particular strain and the host species. For instance, analysis of SSR locus 2 (G2 repeats) showed that 7 G were found in isolates from bison, impala, nyala, Thomson’s gazelles and goat, while 9 G were found in isolates from duikers, transcaspian urial and waterbucks. Elks were infected with isolates with 7 G and 13 G . Sevilla et al.  observed that cattle isolates in Spain had 8, 9 and 11 G repeats in the same SSR locus, and that sheep and goats were mainly infected with isolates having 3GGT repeats on SSR locus 8. In contrast to the findings of El-Sayed et al.  who observed that all isolates with seven or greater number of G1 repeats had 5GGT repeats on SSR locus 8, this was not the case in the present study.
The significance of these SSR genotypes with regards to the potential differences in pathogenesis is not yet known but it suffices to say that the 7 G-4GGT profile seen in one isolate, is associated with MAP isolated from Crohn’s disease patients  and has also been found to be the most common type among cattle herds in Ohio, USA .
According to the present study, a combination of different loci comprising IS1311, MIRU (2, 3), VNTR 32 and SSR (loci 1, 2 and 8) discriminated 21 MAP isolates into 11 distinct strains. This shows very high strain diversity in Uganda. It implies that either the Ugandan isolates have evolved over a very long period of time or that they have all come from different regions in the recent past. Further studies to map out the distribution of these isolates and the historical origins of the cattle might clarify this question. It would also be of interest to determine if the different strains found in this study have any association with any particular breed or genotype of cattle and if the strains themselves differ in their pathogenecity. The most common genotype was number 10 with 10 isolates, with a 9-5-9-6 G-10 G-4GGT profile according to MIRU 2 and 3, VNTR 32 and SSR loci 1, 2 and 8 respectively (Table 1). Nine of these isolates came from abattoir specimens. Although we could not get reliable information on the districts of origin of the cattle from which these isolates were derived, we conjecture that they might be from western Uganda, since the same genotype was seen in a cow from Masindi district which is located in Western Uganda and also due to the fact that most of the cattle slaughtered in these two abattoirs come from western Uganda. The remaining genotype profiles had 1–3 isolates each and were evenly distributed in the districts of Wakiso, Mpigi and Luwero, where there are also several breeds of cattle and different husbandry practices.
Castellanos et al.  observed that type II isolates of MAP from Spain had five repeats at MIRU 3 locus, while type III had three repeats in contrast to some German type III isolates which had five repeats at MIRU 3 locus. In our study, 18 of the 20 isolates amplified, had five repeats on MIRU 3 locus indicating that they may be related to type II isolates. However, five of our isolates which had five repeats at MIRU 3 locus also had 15 repeats at MIRU 2. The 15 repeats at MIRU 2 locus has been associated with type I [13, 24]. As more strains continue to be isolated, the classification of MAP could become more complex than the simple type I, II and III system that was predominant in the last decade or new criteria will have to be devised to classify them into those types for epidemiological reasons. The isolates in this study differ from isolates reported in other studies especially with regard to their SSR profiles and should therefore be considered as novel strains.
Characterisation using SSR locus 8 SNP profile alone was only slightly less discriminatory than the combination of all three systems: MIRU-VNTR/SSR/IS1311 PCR-REA. Therefore phylogenetic analysis of SSR locus 8 could offer an opportunity for strain discrimination and follow up of the transmission patterns in the country, but it also failed to distinguish between several diverse isolates including the cattle, sheep and bison types found in GenBank (see Figure 3).
In our case MIRU 2 was found to be more discriminative than MIRU 3 and VNTR 32. This is in agreement with the findings of El-Sayed et al.  and Castellanos et al. . It was also possible to amplify MIRU 2 from almost all the DNA templates that we had. Another positive aspect of the use of MIRU 2 is that it distinguishes MAP from M. intracellulare in which the locus is not amplified . Finally VNTR 32 was the least discriminative and most difficult to amplify.
As stated in the materials and methods, in-silico analysis indicated high melting temperatures (>100 °C) for the GC-rich MIRU, VNTR and SSR templates. This was the possible reason for the failure of amplification we encountered in many cases using published primers and protocols [9, 12, 15, 19] (data not shown). In an effort to improve the efficiency of amplification, new specific primers were designed and several additives such as DMSO, Betaine, 1, 2, Propan-diol and Ethylene glycol that facilitate DNA denaturation [12, 26], were tested. The new primers were more effective than published primers both with and without the additives. With the new primers, we were able to obtain products from samples which had been too difficult to amplify despite the use of the additives. A cyclic denaturation step of 97 °C for 30 seconds was also adopted. The best results were obtained using Ethylene glycol. Previous studies on MIRUs and SSR have used DMSO and Betaine [12–14] but in our case those two denaturants were not very effective. Unfortunately despite the improvement of the PCR protocols we were not able to obtain amplicons for some of the isolates especially for SSR locus 1 and VNTR 32. Motiwala et al.  were also unable to amplify SSR locus 1 from some isolates.