Whole Genome Sequencing Reveals Local Transmission Patterns of Mycobacterium bovis in Sympatric Cattle and Badger Populations30 January 2013
UK - A new study has found evidence of Badger populations transmitting bovine tuberculosis to cattle following advanced genetic sequencing technology, pioneered by the Agri-Food and Biosciences Institute and Glasgow University.
Introduction to Study
Whole genome sequencing (WGS) offers the potential for unprecedented insight into infectious diseases spread at the individual-to-individual level. However, this potential can be compromised when a poorly sampled ‘reservoir’ population contributes to transmission, as strong biases in the obtained data are inevitable.
Therefore WGS data must be corroborated with epidemiological data in well-described systems, in order to enhance our confidence in their broader use. The epidemic of bovine tuberculosis (bTB) in British and Irish cattle has both economic and animal health importance; it also involves a management host (the cattle) whose demographic history is exceptionally well-documented, and with a reservoir host (the badgers) whose role in bTB spread has defied decades of study and observation.
Here, we show that the observed spatial patterns provide a good match to M. bovis WGS data, but cattle movement networks and within-herd transmission patterns generated by mathematical models do not.
Thus WGS offers considerable promise for revealing basic principles about bTB maintenance in British cattle and the role of badgers, as well as suggesting that similar approaches combining mathematical models and WGS data could be useful for the study of human TB and other infectious diseases where sampling biases are known to exist.
Herd locations indicate centroids of main holdings, and do not include isolated land parcels or rented land. doi:10.1371/journal.ppat.1003008.g001
Measuring genetic variation at the whole genome scale enabled us to genetically distinguish most isolates of M. bovis. This is particularly notable given the small spatial extent of the study cluster, with no two farms being more than 5 km apart. Compared to traditional typing methods, for which the same genotype may be distributed over many hundred square kilometres and only broad associations can be rigorously defined, WGS affords a level of resolution for epidemiological studies previously limited to rapidly evolving RNA viruses.
In addition to most isolates and outbreaks being genetically unique we found that subsequent outbreaks on the same farm tended to involve the same genetic lineage previously detected in that location. This indicates that lineages are commonly able to persist locally within the direct environment of a farm, although the mechanisms for this are not yet clear.
Results of our mathematical models based on individual cattle histories indicate that persistence on a farm is overall poorly explained by ongoing transmission within herds. In the cases of Herds 1 and 3 for example, the model identified independent introductions for subsequent outbreaks (Fig. 5), despite the fact that these serial outbreaks involved the same genetic lineage (Fig. 2).
Based on these findings, the detected infections (including unsequenced reactors) are insufficient in explaining local persistence on farms, instead suggesting a number of possible alternative mechanisms, such as re-introduction of the same lineage from neighbouring herds, environmental persistence, or alternative hosts.
In contrast, and despite the relative simplicity of the modelling approach used, the persistence of the outbreaks in Herds 3 and 5 from 2007 to 2010 are consistent with lineage persistence resulting from extensive within-herd transmission, despite the multiple clear whole herd tests that would have occurred in between dates for reactors.
While forward simulations were used to corroborate the robustness of our modelling approach, any extrapolation of our results for bTB epidemiology in general must be viewed with caution, both because of the small size of the dataset and because some of the modelling assumptions (in particular the assumption of explicitly time dependent generation times, see supplementary information) were not intended to be mechanistic descriptions of the underlying transmission process. Nevertheless, the fact that such a parsimonious model identifies a cattle-only contact structure largely consistent with the observed phylogeny generates confidence in our results.
Showing all individuals residing within the five study herds at some time from 1994 to 2010. Cattle residence times indicated by the length of the horizontal bars (each bar representing a single animal). In black, all cattle from which sequences are derived (herd indicated by surrounding type). Test dates on which one herd received a whole herd test are indicated by vertical dashed lines. Herd colours correspond to colours in Figure 1 (1 – pink, 2 – purple, 3 – blue, 4 – orange, 5 – red). doi:10.1371/journal.ppat.1003008.g004
In addition to local persistence, we also found evidence for the introduction of new genetic lineages onto farms and our analyses allow us to partially resolve how these introductions occur.
Though cattle movements are a known risk for between-herd spread of bTB in Britain, they do not appear to be important for the events observed here, as the probabilities of transmission amongst herds involved in the extensive network of all observed VNTR10 outbreaks only poorly predict the genetic divergence amongst isolates. In contrast, we found Cartesian distance to be a good predictor of genetic distance among isolates at a very fine scale.
Though the small sample size means that inferences regarding between-herd contacts should be viewed with caution, these results are consistent with the relatively low importance accorded to movements compared to local risk factors in bTB endemic areas that was previously observed at a national scale in GB, and suggests that a more extensive analysis of the balance between local spatial and network processes would be merited.
As it stands, the most parsimonious explanation for these outbreaks involve a local transmission process that could be due to a number of causes. A non-exhaustive list of these includes both cross-boundary contact or unrecorded local movements between herds and transmission from a common badger reservoir (where the interaction is spatially localised, consistent with the badger's stable social structure and strong territoriality, or a combination of these factors.
While our sample size for badgers is low, the badger-derived sequences are remarkably similar to those in cattle, demonstrating very recent cross-over events between the two populations, or alternatively recent infections from a common source, such as the environment. The demonstration of a high M. bovis diversity in a single badger suggests either a lengthy infection in that badger, or multiple exposures to different sources of infection.
The weighted, directed network shows the probability that a transmission path exists between cattle with sequenced isolates that does not pass through other sequenced isolates. Infection events poorly explained by transmission amongst reactor cattle are therefore more likely to be caused by a ‘reservoir’, which potentially encompasses infected badgers, between-herd interactions, latent infections, or environmental contamination. Sequences belonging to the same herd are surrounded by dashed outlines. doi:10.1371/journal.ppat.1003008.g005
Although our current estimate for the rate at which M. bovis genomes evolve must be considered preliminary, it is considerably slower than the rate observed in M. tuberculosis. Should it be confirmed, this has obvious implications for the level of temporal resolution that WGS can provide for unravelling epidemiological dynamics for bTB.
In our current data, we were unable to genetically resolve relationships among multiple isolates stemming from the same outbreak for example and saw serial outbreaks commonly involving identical genotypes.
This is corroborated by the poor correlation between the genetic distances and our estimates of the within-herd contact structure. However, apart from limiting opportunities for molecular epidemiological inference, these observations may also hold clues with respect to M. bovis biology and transmission.
A recent study conducting experimental infections with M. tuberculosis in primates, found mutation rates to be equivalent during latent and active infections and proposed oxidative damage as a potential mechanism.
If this is relevant to M. bovis, one could hypothesise that the slower rates of evolution seen here at the population level, could be caused by the bacterium spending extended periods outside the host, in the environment. Future studies and analyses are needed to obtain more accurate estimates for the genomic rate of evolution in M. bovis and to test for potential rate heterogeneity and its underlying factors.
While cattle movements and long-term, hidden persistence within herds have both been shown to contribute significantly to herd breakdowns, these previous analyses were aimed at the identification of statistical associations; here we have shown that WGS data are able to identify local interactions as the principle culprit in specific herds.
This makes WGS both a valuable tool for forensic epidemiology, and an aid in the construction of improved mathematical and statistical models of disease dynamics. In contrast, the poor correlation between network and genetic distance at the within-herd level suggests important limits to the resolution that WGS can provide for this system.
The local effects identified here may be due to the local infected badger population, but are also consistent with local herd- to-herd spread. Our simplified modelling approach was chosen to maximise the use of available epidemiological contact data, but at the expense of a more detailed exploration of the possible hypotheses regarding the sources of transmission. However, it is likely that WGS based on more extensive sampling will allow for more sophisticated approaches, that could be used to directly estimate the role of badgers in the maintenance of bTB in British and Irish cattle.
While insights into particular disease problems will depend on many factors we cannot consider here, our study supports the proposition that WGS data alone can provide insight into the impacts of unobserved populations on observed epidemics even in the absence of detailed transmission chain information, for M. bovis, other members of the M. tuberculosis complex, and other pathogens involving reservoir hosts.
Citation: Biek R, O'Hare A, Wright D, Mallon T, McCormick C, et al. (2012) Whole Genome Sequencing Reveals Local Transmission Patterns of Mycobacterium bovis in Sympatric Cattle and Badger Populations. PLoS Pathog 8(11): e1003008. doi:10.1371/journal.ppat.1003008
Editor: Oliver G. Pybus, University of Oxford, United Kingdom
Received: December 16, 2011; Accepted: September 17, 2012; Published: November 29, 2012
Further ReadingYou can view the full report by clicking here.