Appendix 1
BIOLOGICAL SCIENCES: Applied Biological Sciences or Population Biology or Genetics
Title: Genetic evidence that culling increases badger movement
Authors and Affiliations
Lisa C. Pope1, Roger K. Butlin1, Gavin J. Wilson2, Rosie Woodroffe3, Kristien Erven1, Chris M. Conyers2, Tanya Franklin2, Richard J. Delahay2, Chris L. Cheeseman2, Terry Burke1
1Department of Animal and Plant Sciences, University of Sheffield, Sheffield, S10 2TN, UK 2Central Science Laboratory, Sand Hutton, York, YO41 1LZ, UK. 3Department of Wildlife, Fish & Conservation Biology, University of California, 1 Shields Avenue, Davis CA 95616 USA
Corresponding author
Prof Terry A. Burke
Department of Animal and Plant Sciences, University of Sheffield, Sheffield, S10 2TN, UK
Telephone: +44 (0)114 222 0096
Fax: +44 (0) 114 2220002
Number of text pages: 18
Number of Figures: 4
Number of Tables: 2
Number of words in abstract: 247
Total number of characters in paper: 45011 exlcuding SI, including spaces
Abstract
The Eurasian badger (Meles meles) has been implicated in the transmission of bovine tuberculosis (TB, caused by Mycobacterium bovis) to cattle. However, evidence suggests that attempts to reduce the spread of TB among cattle in Britain by culling badgers have mixed effects. A large-scale field experiment (the Randomised Badger Culling Trial, RBCT) showed that widespread proactive badger culling reduced the incidence of TB in cattle within culled areas but that it increased in adjoining areas. Additionally, localised reactive badger culling increased the incidence of TB in cattle. It has been suggested that culling-induced perturbation of badger social structure may increase individual movements and elevate risks of disease transmission between badgers and cattle. Field studies support this hypothesis, by demonstrating increases in badger group ranges and the prevalence of TB infection in badgers following culling. However, more evidence on the effect of culling on badger movements is needed in order to predict the epidemiological consequences of this control strategy. Here, analysis of the genetic signatures of badger populations in the RBCT revealed increased dispersal following culling. While standard tests provided evidence for greater dispersal after culling, a novel method indicated that this was due to medium- and long-distance dispersal, in addition to increases in home-range size. Our results also indicated that badgers infected with M. bovis moved significantly further on average than did uninfected badgers. A disease control strategy that included culling would need to take account of the potentially negative epidemiological consequences of increased badger dispersal.
Introduction
Bovine tuberculosis (TB) has increased markedly in British cattle herds over the last 20 years and remains a significant economic and animal welfare problem (Krebs et al. 1997). Bovine TB has persisted in cattle in parts of the UK despite the implementation of control measures that have succeeded elsewhere. This persistence has been attributed to the existence of a significant reservoir of infection in wild badgers (Meles meles)(Krebs et al. 1998), although cattle-to-cattle transmission is also important (Gilbert et al. 2005) and other wildlife hosts may also play a role (Delahay et al. in press). Control of TB in the past has included badger culling, but the efficacy of this approach is variable. Following a review of TB control policy (Krebs et al. 1998), the impact of badger culling on TB incidence in cattle in Britain was experimentally assessed in the Randomised Badger Culling Trial (RBCT). This experiment aimed to measure the effect of reducing badger population density on rates of TB infection in cattle. The experimental design involved 30 100 km2 study areas throughout south and west England, grouped into 10 triplets. In one area of each triplet no culling was performed, in a second area localised badger culling took place in response to cattle herd breakdowns (average area, 5.3 km2) and in the third badgers were proactively culled, roughly annually, over an area of 100 km2 (Donnelly et al. 2003). Results from the RBCT have shown that reactive culling had at best little benefit, and at worst increased the incidence of TB in cattle (Donnelly et al. 2003), while proactive culling reduced the incidence of TB within culled areas but increased it in immediately-adjacent areas (Donnelly et al. 2006). In contrast, trials in Ireland suggested that badger culling over substantial areas reduced the incidence of TB (Eves 1999; Griffin et al. 2005). The lower badger density, potentially higher culling efficiency (Donnelly et al. 2006) and presence of geographical barriers to badger immigration surrounding trial areas (Woodroffe et al. 2006b) may have contributed to the different findings in the British and Irish studies.
Mathematical modelling supports the idea that the lethal control of badgers can reduce TB in cattle, if it is assumed that there is no impact on badger behaviour or social structure (Smith et al. 2001). However, phenomena such as culling-induced alterations to social organisation, dispersal and compensatory reproduction could increase contact rates between badgers in a culled population and result in decreased efficacy of control operations (Smith et al. 2001, Barlow 1996). In undisturbed medium to high-density populations badgers live in stable social groups, and the spatial distribution of TB can remain restricted (Delahay et al. 2000), supporting the idea that there is only limited transfer of TB under stable demographic conditions. However, studies of badger populations subjected to culling report increases in the overlap of social group ranges, which might be expected to increase the spread of infection between groups (Tuyttens et al. 2000a, Woodroffe et al. 2006a). Recent results also suggest that culling may increase the rate of transmission of M. bovis infection among badgers, particularly where there are few geographical barriers to impede badger immigration into culled areas (Woodroffe et al. 2006b). These results indicate that badger culling increases badger movement (Cheeseman et al. 1993), and are consistent with an increase in contact rates, and enhanced spread of disease between cattle and badgers.
Dispersal is of central importance in animal ecology, yet it remains difficult to measure (Greenwood 1980). Combining genetic and geographic information can be a powerful approach for studying dispersal, yet problems remain (Rousset 2001). Recent analytical advances in identifying sub-populations and landscape barriers to dispersal have been described (Corander et al. 2004; Francois et al. 2006; Guillot et al. 2005). These methods provide estimates of dispersal rates among sub-populations but little can be inferred about dispersal within sub-populations. Spatial autocorrelation analyses and Mantel tests of genetic versus geographic distance can be used to infer the average genetic ‘neighbourhood size’ (Peakall et al. 2003), often considered to be the basic unit of population structure (e.g. Wright 1943), but provide only limited information on individual movements (Rousset 2000). Assignment tests, most commonly used to identify long-distance migrants (Berry et al. 2004; Davies et al. 1999), can be adapted to compare dispersal patterns of different classes, such as sexes, within populations (Favre et al. 1997). However, these tests, along with the more traditional comparison of FST values, assume that there is no internal population structure, do not provide estimates of physical distances and have limited statistical power (Goudet et al. 2002). One method has been developed to infer the origin of individuals at a broad scale, based on genotypes from surrounding populations (Wasser et al. 2004), but this may not be applicable where the sample size at each location is limited.
Here, we examine the genetic profiles of adult badgers from eight RBCT proactive culling areas. We compare badgers from the first cull (representing a comparatively undisturbed, pre-cull population) with those taken in the second cull performed in each area (5 to 22 months after the initial cull; Fig. 1). In addition to using established methods for the detection of dispersal patterns, we developed a within population assignment method that utilises population structure to establish the most likely location of an individual within a sampled area, based on its genetic signature. From this we quantified changes in dispersal patterns after culling had taken place.
Results
In total, 3450 adult badgers from eight proactive culling areas were genotyped at 16 badger microsatellite loci (Carpenter et al. 2003). Badgers from the second culls showed significant evidence of greater dispersal than in the first culls for two out of three standard genetic marker tests for differential dispersal (Table 1). The three tests used were differences in mean corrected assignment index (AIC), variance AIC and FST. These three tests have differing sensitivities, with variance AIC being more suited to detecting rare long-distance dispersal, mean FST most sensitive to large frequency differences in dispersal on a smaller scale, and mean AIC best at detecting dispersal differences lying somewhere between these two measures (Goudet et al. 2002). By performing all three tests we increased the likelihood of detecting differences in dispersal patterns, though simulations have indicated that all three tests have limited power (Goudet et al. 2002). Comparisons of mean FST values were not significant (0.162 vs 0.149; p = 0.17), but the change in mean AIC (Favre et al. 1997; Paetkau et al. 1995) and the ratio of the variances of AIC were significantly different across the eight trial areas (p = 0.001 and 0.006, respectively) (Table 1).
We found that genetic similarity decreased with increasing distance in both the first and second cull samples, based on Mantel tests and spatial autocorrelation analyses (Fig. 2). The genetic similarity (r) of individuals in close proximity (< 500 m) was greater in the initial cull than in the second cull, suggesting that there was a higher proportion of relatives within social groups before culling took place. In survey only areas the median bait return distance was approximately 300 m, (Woodroffe et al. 2006a), and the average distance between main setts in proactively culled areas was approximately 600 m (R. Woodroffe, pers. com.), indicating that 500 m will on average represent the extent of a single social group. However, over larger distances (0.5 – 4 km), genetic similarity was greater among individuals in the second cull (Fig. 2). Genetic similarity fell to zero at a similar distance (between 4 and 4.5 km, Fig 2.) in both the first and second culls, indicating that there was no significant change in neighbourhood size (considered to be a basic population unit, defined as a product of population density and parent–offspring dispersal distance, (Wright 1943)).
To obtain a more detailed estimate of the change in movement patterns we predicted the most likely location of an individual based on a sub-sample from the first cull. We then calculated the distance from the location at which the individual was culled to the estimated location, using our software BadMove v1. We called this distance the displacement (D). We compared the difference in mean displacement (∆D) between successive culls. The ∆D was positive (i.e. greater after the second cull) in all study areas and was highly significant overall (based on 100 permutations, χ2(16 df) = 68.5; p10-8, Fig. 3). The majority of badgers sampled in the initial cull were predicted to originate less than 1 km from their culled location (Fig. 4a). Displacement distances greater than 1 km were inferred more often among badgers sampled from the second cull (Fig. 4a), indicating that movements were greater after culling. We also calculated the inferred displacement patterns in relation to TB infection status (Fig. 4b) and sex (Fig. 4c). Slight differences were observed between infected and uninfected badgers (Fig. 4b) and male and female badgers (Fig. 4c).
To further explore the effects of culling session, sex and TB infection status on displacement estimates we used a linear mixed-effects model (LME, see Methods). The minimal adequate model included all three predictors under consideration here – cull session, sex and TB infection status – but did not include any second-order interactions among these factors. The intraclass correlation (a measure of the variation accounted for by random factors, McCulloch & Searle 2001) was low (ICC = 0.07), indicating that differences among trial areas were not significant (p = 0.94). As found in our previous analyses, the difference between cull sessions was highly significant (642 m; 95% CI 581 to 809 m; p< 0.001, Table 2). Males were found to move significantly further than females, though the absolute difference in mean displacement was small (56 m, Table 2). Interestingly, badgers infected with bovine TB were found to move significantly further than uninfected animals, and while the difference in the estimated mean displacement was not large (99 m, Table 2), it was greater than that found between males and females. When we performed the same LME analysis using AIC data, the minimal adequate model retained only cull session and infection status as factors (not sex) and only cull session was found to be significant (p < 0.001).
To validate our new method we performed two tests. First, we estimated the mean displacement, and change in displacement over 24 months, in the intensively studied, undisturbed badger population at Woodchester Park, Gloucestershire. We found no difference between mean displacement calculated for badgers caught in Woodchester Park in 1999 compared with those captured in 2001 (with no replication of individuals in these two categories; mean ∆D = -0.015 km, sd 0.119). This demonstrates that displacement values within an undisturbed badger population did not increase measurably over the time scale of this study. Secondly, we performed simulations on the RBCT samples in reverse time sequence (see Methods). Under the null hypothesis that culling did not produce a change in movement, ΔD should be the same for both reversed and forward analyses. The ‘reverse tests’ on RBCT data supported the contention that culling had disrupted the genetic structure within trial areas. While estimates of ΔD between second and first culls were all positive, for six of the trial areas ΔD was not significantly different from zero, and in all cases was less than that of the ‘forward test’, the time simulations where the first cull preceded the second cull (Fig. 3). This result, in combination with the consistency in the estimated displacement over years at Woodchester Park, indicates that we can have confidence in our method and in the inference that culling was the cause of the increase in dispersal.
Discussion
In this paper we use genetic information to demonstrate an increase in badger movement following culling. We show, for the first time, that long-distance movement by badgers (> 1 km) increased following culling. Our work supports previous research from the RBCT which demonstrate that culling disrupts spatial organisation, resulting in an increase in badger movement at small scales (< 1km). Field studies have revealed a consistent increase in badger home-ranges following culling, demonstrating that culling alters badger social organization as well as population density (Woodroffe et al. 2006a). Our work supports this finding, but also suggests that the influence of culling on badger contact rates might on larger spatial scales than previously considered, though the detrimental effects on cattle TB appear much weaker over distances greater than 2 km (Donnelly et al. 2006).
Standard tests indicated that there was an overall significant increase in dispersal following culling. We have investigated this result in more detail using a novel analytical approach. The majority of badgers sampled in the initial cull were estimated to be displaced less than 1 km from their sampled location. This reflects the level of precision of our program and will also include temporary short-range movements such as foraging trips, and visits to neighbouring setts for mating, (Carpenter et al. 2005) as demonstrated by results from the well-studied Woodchester Park population (mean D = 0.75 km, Supplementary Information). Our estimates of long distance movement were conservative in that the maximum possible displacement was limited by the sampling boundaries plus approximately 500 m (< 11 km observed: maximum possible 17 km). Displacement distances greater than 1 km were inferred more often among badgers sampled from the second cull (Fig. 4a). Displacements across all the larger distance categories were also more common for the second-cull badgers, with the average increase in displacement estimated as 410 m. Increased overall displacement to this extent appears to have been caused by large numbers of animals changing their dispersal patterns, rather than by a few long-distance immigrants or a large number of individuals expanding their home ranges. The median nearest-neighbour distance for main setts in bait-marked no-culling areas was approximately 600 m on average (R. Woodroffe, pers. comm.). Therefore, an increase of the magnitude observed in the present study implies a substantial disruption of population structure.
While it could be argued that other factors may have been responsible for the increase in badger dispersal observed in all eight trial areas examined, several lines of evidence support our view that culling was responsible. In an un-culled badger population (Woodchester Park), badger dispersal was found to remain constant over a 24-month period, while the cull intervals considered in this study ranged from 5 to 22 months. It seems unlikely that time alone can account for the large increase in inferred badger movement. Furthermore, the culling events analysed here were spread over four years, from 1998 to 2002, making it unlikely that some other incident, such as the foot and mouth disease outbreak in 2001, was responsible for this increase.
Despite the significant increase in detected dispersal, strong genetic population structure persisted following the initial cull, with no detectable change in genetic neighbourhood size. Strong isolation-by-distance at a fine scale appears to be a general characteristic of badger populations (Pope et al. 2006), probably resulting primarily from mating among badgers from neighbouring social groups (Carpenter et al. 2005) and limited dispersal (da Silva et al. 1994; Rogers et al. 1998). Following culling, immigration of animals from outside culled areas would tend to reduce the correlation between genetic and geographic distance in badger populations. If culling resulted in the disruption of social groups we would also expect the strength of the correlation between genetic and geographic distance to have decreased. If, however, there was no effect of culling we would expect patterns of genetic similarity to remain constant. How can we account for this increase in genetic similarity among badgers at fine scales? Non-random sampling is one possible explanation. Population density greatly decreased following culling (Woodroffe et al. 2005). Social group ranges were demonstrated to expand (Woodroffe et al. 2005), and while overlapping social groups will decrease genetic similarity, in areas where social group ranges expanded without overlap due to low population density, genetic similarity will increase. It is also possible that, despite the exclusion of cubs from our analyses and the short time span involved (as little as 5 months), some change in the age structure of the population may have occurred, resulting in an increase in the proportion of relatives. A reduction in population density may have lead to greater survival among cubs and yearlings (Tuyttens et al. 2000b). Non-random sampling may then lead to a greater proportion of relatives in the population. Over greater time periods, reductions in populations have led to an increase in the number of females breeding within social groups (as observed in red grouse, Cattadori et al. 2005). Where such females are related this could also lead to an increase in the proportion of relatives at fine scales.