Results from the present study on 37 B. abortus isolates from milk and vaginal swabs taken from 340 cows (375 samples) during a 16 months period showed the in vivo genetic stability of the MLVA16 markers. Molecular typing methods are commonly used to investigate epidemiological relationships among isolates and sources of infection . However, before being used for those purposes, PCR methods for molecular typing require careful in-house validation of typeability, reproducibility, repeatability, stability, discriminatory power and epidemiologic concordance [5, 21]. The findings of this study, associated with previous data on the high discriminatory power and epidemiologic concordance of MLVA16, besides its good typeability and in vitro stability [6, 7, 9–11, 13, 15–17], corroborate the use of MLVA16 as suitable typing method for refining the understanding of the epidemiology of bovine brucellosis.
Our results showed a low genetic diversity and the existence of three different B. abortus strains within a focus of bovine brucellosis, the RB51 vaccine strain (genotype RB), recovered from animals that were twice vaccinated in the period, and two field strains (genotypes W and Z).
In the present study no RB51 was isolated from the milk of any animal during the post-partum period (30–60) but only from vaginal swabs, although some cows have been vaccinated in the last third of pregnancy and a highly sensitive diagnostic strategy have been employed . These results were corroborated by previous studies that also reported no recovery of B. abortus RB51 from any of the milk sample tested by conventional bacteriological methods of cows vaccinated during pregnancy or 30–60 days after delivery [23, 24]. Regarding the presence of viable RB51 in postpartum vaginal secretion, it is important to consider that all animals from which RB51 were recovered after delivery were over 5 months of gestation by the time of revaccination with this strain and that no abortion was related to RB51 isolation (Table 1). Thus, it is likely that the RB51 booster had contributed to the recovery of the vaccine strain in postpartum period, since it has been observed that during transition period a depression in cell-mediated response occurs, which leads to a decrease of the resistance to disease or increase of the residual virulence of vaccines .
Other important finding of this study was the demonstration of viable B. abortus in milk from infected animals associated with the outbreak, corroborating the public health risk of the consumption of raw milk and unpasteurized dairy products . The colonization of the mammary gland and associated lymph nodes by B. abortus with excretion of microorganisms in milk was already demonstrated . In addition, it is important to emphasize that the methodologies used for bacterial culture and molecular identification were able to differentiate RB51 vaccine strain from B. abortus field strains.
Interestingly, field strains isolated from this brucellosis outbreak showed distinct genotypes and biotypes, suggesting that the outbreak had two different sources of infection. The majority of B. abortus isolates from the outbreak (24/25) were classified as B. abortus biovar 3 genotype W, which remained the single cause of brucellosis in the herd for eleven months (from March 2009 to February 2010). This widely demonstrates that the introduction of this strain in the herd was responsible for the occurrence of the outbreak and consequently for the high abortion rate observed at the end of 2008. Furthermore, in February 2010, the introduction into the herd of another field strain, B. abortus biovar 1 genotype Z, was observed. This genotype Z was confirmed by a twice repetition of MLVA16 genotyping assay that showed the same results. There are two possible explanations for the presence of this infected animal, negative to the conventional serological tests adopted in the herd: first, this heifer was congenitally and persistently infected without seroconversion , and second, the negative serological result from this infected heifer was a false negative result inherent to any diagnostic test. However, the diagnostic strategy employed was very sensitive , which tends to minimized false negative results. Since this second field strain was represented by just one isolate, it is very likely that the control policy adopted in the herd, which included test-and-removal procedures, has prevented the dissemination of this strain to other animals and consequently a new increase in the brucellosis prevalence levels into the herd.
Besides the differences in the classification of biovar, the differences observed between the genotypes of the field strains were very large and not limited to differences in the more variable loci (Bruce04, Bruce16, Bruce18 and Bruce30), but were also observed in conserved loci, such as Bruce06 and Bruce11 (Table 2). Therefore, the identification of the new source of infection in the herd was only possible due the use of MLVA16, otherwise the genotype Z strain would have been considered just another B. abortus isolate within the outbreak.
Epidemiological data of the herd also confirmed that the B abortus biovar 1 genotype Z strain was a newly introduced strain in the farm, since the young heifer from which this strain was isolated was introduced in the herd shortly at the end of the outbreak (November/2009). In fact, it has been widely demonstrated that the purchase of infected animals is the main risk factor for the introduction of brucellosis in free herds . Unfortunately, it was not possible to obtain more epidemiological data about the heifer from which the second B. abortus field strain was isolated, such as the brucellosis status of the herd of origin, which would have allowed a better understanding about the epidemiology of the genotype Z strain.
The MLVA16 panel 1 profiles of both B. abortus isolates, genotypes 28 and 40 by the MLVAbank, have already been previously observed in Minas Gerais State . Moreover, the comparison between MLVA16 genotypes of B. abortus biovar 3 field strain (W) and those previously described by Minharro et al.  revealed differences restricted to hypervariable loci [panels 2A (Bruce19) and 2B (Bruce04)], suggesting a possible epidemiological link between these strains, since all were isolated in the state of Minas Gerais and were also identified as biovar 3. For the single field strain classified as genotype Z, the comparison with MLVA16 patterns previously described by Minharro et al.  showed a genetic distance of one locus (Bruce19) from a B. abortus biovar 2 also isolated from Minas Gerais State.
In a brucellosis control and eradication program the use of an accurate surveillance and highly discriminatory typing method is essential to characterize an outbreak and determine the source of infection and the transmission routes. The present results on typing multiple B. abortus isolates from an outbreak originated from an outbreak in the same herd depicted the in vivo stability of the MLVA16 markers. The set of loci that comprise the MLVA16 demonstrated to be very stable, even when assessed over the time span of one year and four months, since the field strain mainly responsible by the outbreak (W) and the RB51 vaccine strain recovered from vaccinated animals showed unchanged MLVA16 profiles. These findings are extremely important because they definitely confirm the ability of the MLVA16 to establish correct epidemiological correlations, since, besides having a high discriminatory power, MLVA16 markers were also stable under natural selection pressure exerted by the host, during the sixteen months assessed. Thus, these results increase the confidence in the traceback established from the results of MLVA16 and further support this technique as the choice one for typing B. abortus.
Futhermore, the examination of in vitro stability of the B. abortus RB51 vaccine strain, B. abortus strain 2308 and B. abortus field isolates by serial passages in culture medium showed no change in MLVA16 profile [9, 16]. Likewise, the analysis of B. abortus S19 vaccine strain from different batches of different manufacturers did not reveal significant differences in MLVA16 pattern . Concerning the evaluation of in vivo genetic stability of MLVA16 loci, it has also been demonstrated that passage of B. abortus RB51 in cattle and of B. abortus 2308 in mouse did not lead to changes in any marker of MLVA16 . Nevertheless, minor changes in VNTR pattern were observed in in vitro passages of B. abortus 544 and in genotyping of multiple B. abortus isolates from the same outbreak . Her et al.  found different allelic profiles in seven of twenty-three herds in which more than one B. abortus isolates were obtained. Those different genotypes from same outbreak showed mutations only in the loci Bruce 30 and 43, which did not seem to affect the identification of a possible common origin of the strains . However, our data showing the high in vivo stability of the MLVA16 loci have as main findings over previous data the large numbers of isolates (37) obtained from the same source, the long period of time assessed (16 months) and the typing of strains colonizing different sites (mammary gland or reproductive tract), and therefore under different selective environmental pressures.