The Tasmanian devil is threatened with extinction by devil facial tumour disease (DFTD), a unique infectious cancer in which the tumour cells themselves, which derive from a single long-dead host devil, are the infective agent and the tumour is an infectious parasitic cell line.
Transmission is thought to occur via direct inoculation of tumour cells when susceptible and infected individuals bite each other or by fomitic transfer of tumour cells. The nature of transmission and the extent to which biting behaviour and devil ecology is associated with infection risk remains unclear. Until our recent study in north-west Tasmania showed reduced population and individual impacts, DFTD had caused massive population declines in all populations monitored.
In this paper, we investigate seasonal patterns of injuries resulting from bites between individuals, DFTD infection status and tumour location in two populations to determine whether the number of bites predicts the acquisition of DFTD and to explore the possibility that the reduced impacts of DFTD in north-west Tasmania are attributed to reduced bite rates.
Devils with fewer bites were more likely to develop DFTD and primary tumours occurred predominantly inside the oral cavity. These results are not consistent with transmission occurring from the biter to the bitten animal but suggest that dominant individuals delivering bites, possibly by biting the tumours of other devils, are at higher risk of acquiring infection than submissive individuals receiving bites. Bite rates, which were higher during autumn and winter, did not differ between sites, suggesting that the reduced population impacts in north-west Tasmania cannot be explained by lower bite rates.
Our study emphasizes the importance of longitudinal studies of individually marked animals for understanding the ecology and transmission dynamics of infectious diseases and parasites in wild populations.
Parasites and pathogens are ubiquitous in nature and are an important source of reduction in fitness in wild populations (Hudson et al. 2002). The key process in any host–parasite interaction is transmission, but understanding the dynamics of transmission in a wild population is often very difficult (McCallum, Barlow & Hone 2001; Tompkins et al. 2011). Inferring how parasite or pathogen transmission occurs in the field and determining the major routes or host behaviours that facilitate or limit disease transmission in wild animal populations is usually hampered by the difficulties of gathering empirical data on actual transmission events.
In a directly transmitted parasite or pathogen, transmission requires sufficient contact between an infected and a susceptible host to enable infective stages to pass from one to another. Host behaviour therefore plays an essential part in pathogen transmission and pathogens may in turn influence host behaviours responsible for transmission (Hart 1988; Loehle 1995; Wendland et al. 2010). For example, sick animals affected by pathogens may become lethargic, seek isolation, avoid contact with conspecifics (Hart 1990; Loehle 1995) or increase aggressive behaviour (Knobel & du Toit 2003). These pathogen-driven responses can alter the contact rates of host populations and alter the dynamics of infection (Artois et al. 1991). Theoretical and empirical studies suggest that behavioural changes because of parasites are usually adaptive and beneficial to the parasite rather than to the host (Dobson 1988; Hart 1990). Behavioural differences at the individual level are nonetheless, common in wild populations and evolutionary theory predicts that there should be fitness advantages for susceptible individuals that are capable of recognizing infectious conspecifics and avoiding parasite transmission (O'Donnell 1997; Boots et al. 2009).
Tasmanian devils (Sarcophilus harrisii) are threatened with extinction by devil facial tumour disease (DFTD) (Hawkins et al. 2006; McCallum et al. 2009), a directly transmissible cancer in which live tumour cells are transmitted directly between hosts through intimate contact. Transmissible cancers are a novel type of parasite in the spectrum described by Lafferty & Kuris (2002). They consist of cells that are clonally derived from an original tumourigenesis in a long-dead host and are genetically distinct from their current host individual. Thus, tumours can be regarded as a highly degenerate obligate parasitic mammal. Transmissible tumours share most of the characteristics of a typical microparasite, such as prolific replication within the host (McCallum & Jones 2012). Transmission of this cancerous cell line is possible due to low host genetic diversity, particularly in the Major Histocompatibility Complex (MHC) (Siddle et al. 2007), a complex of genes responsible for generating immune responses in vertebrates. Thus, the immune system of the devil fails to recognize foreign tumour cells as non-self, and the tumour is passed on as an allograft between susceptible and infected hosts (Pearse & Swift 2006). Until recently, DFTD has been proved to be consistently fatal, with infected individuals succumbing to the disease between 6 and 12 months after the detection of clinical signs (McCallum et al. 2009; Hamede et al. 2012).
Devil Facial Tumour Disease has spread from the point of its first detection in north-eastern Tasmania to now occupy most of the devil's distribution. The west and north-western areas of Tasmania now contain the only remaining DFTD-free populations (McCallum et al. 2007, 2009). The typical pattern following DFTD arrival is a rapid increase in disease prevalence to more than 50% and severe declines in population size and survival rates within 4 years (Lachish, Jones & McCallum 2007; McCallum et al. 2009). Our recent study in a north-western Tasmanian devil population at ‘West Pencil Pine’ demonstrated, for the first time, reduced impacts at the population and individual level (Hamede et al. 2012). Reduced impacts at West Pencil Pine were characterized by a much slower decline in population growth rates and population size, no change in age structure, increased longevity of infected individuals and a slower increase in prevalence of DFTD, compared to three eastern populations affected by DFTD (Hamede et al. 2012). Three different mechanisms could explain these patterns at West Pencil Pine compared with other eastern sites. First, differences in genetic structure of the host populations could manifest as either resistance or tolerance to DFTD (see Ebert & Bull 2003; Carval & Ferriere 2010), leading to reduced prevalence or slower disease progression, respectively. MHC genotypes, which are involved in tumour recognition, are different at West Pencil Pine from those of eastern populations and the tumour itself (Siddle et al. 2010). Second, the patterns could result from a less virulent strain of the tumour, a hypothesis supported by evidence of a tetraploid DFTD strain being present in the West Pencil Pine population (Pearse et al. 2012). Third and of relevance to this paper, reduced agonistic behaviour resulting in lower biting injury rates could explain the reduced infection risk at West Pencil Pine (Hamede et al. 2012).
Acquiring knowledge of transmission in wild animals is notoriously difficult. The clonal nature of the devil facial tumours is evidence for direct transmission of tumour cells from infected to susceptible host individuals (Pearse & Swift 2006; Siddle et al. 2007). One of the requirements for the transmission of an infectious cancer is intimate injurious contact that brings live tumour cells in direct contact with subepidermal tissue of the susceptible host in a suitable location where they can grow (McCallum & Jones 2012). Transmission must be swift as live cells do not survive long outside the body. This points towards biting as the major route of transmission, although transmission through fomites has not been excluded. Tasmanian devils are predatory carnivores with substantial canine teeth (Jones & Stoddart 1998; Jones 2003). When they bite each other, they inflict wounds that range from canine tooth puncture holes to sizeable avulsions (Hamede, McCallum & Jones 2008).
Several lines of evidence indicate that transmission of DFTD is strongly frequency-dependent. The force of infection remains high, at around 50%, even when populations are severely depleted by the disease to very low densities (McCallum et al. 2009). In addition, the disease is spreading into naturally very low-density populations (McCallum et al. 2007). This suggests frequency-dependent transmission, which is typical of sexually transmitted diseases, and which has no threshold population density below which the disease will die out (McCallum, Barlow & Hone 2001). Most of the biting injuries occur during the mating season (Hamede, McCallum & Jones 2008), because of male contests and females evading mate guarding (M. Jones, unpublished data), with just a low background level of bites the remainder of the year that would occur during feeding and other social interactions (Hamede, McCallum & Jones 2008). Tasmanian devils are specialized scavengers and social interactions around carcasses involving biting behaviour are common (Pemberton & Renouf 1993; Hamede, McCallum & Jones 2008). Whilst the rate of close proximity contacts (defined as two individuals making physical contact) during feeding interactions increases with population density (Hamede, McCallum & Jones 2008), few of these feeding encounters result in bites and even fewer in bite injuries that penetrate the dermal layer (Pemberton & Renouf 1993).
Our aim in this study is to investigate whether the frequency of biting injuries predicts future acquisition of DFTD in individuals. We relate the seasonal patterns of the number and location of bites in individuals to subsequent development of tumours in those individuals in two populations that subsequently differed in epidemic outcomes; disease prevalence and population effects of DFTD. We discuss our results in relation to devil behaviour and ecology that is relevant to DFTD transmission.
Materials and methods
Study Sites and Data Collection
We analysed data from two different DFTD affected populations in northern Tasmania (West Pencil Pine and Wisedale, Fig. 1) that were sampled using identical protocols. We set 40 custom built devil traps (constructed of 300 mm polypipe) baited with meat and checking commencing early each morning, over a 25 km2 area for 10 nights, four times a year at 3 month intervals. The timing was chosen to sample from four important seasons that represent key life-history events: summer (February, when juveniles are dispersing prior to the mating season), autumn (May, immediately after the mating season), winter (August, when females are carrying pouch young) and spring (November, when females are in late lactation). West Pencil Pine (41°31′S, 145°46′E) is a 25 km2 area situated on private production forestry land to the west of Cradle Mountain in north-west Tasmania. This population was monitored from August 2006 (in the early stages of disease arrival at the site) until May 2010. Wisedale (41°16′S, 146°39′E) is a 25 km2 area situated on a private farm in northern Tasmania, which was monitored from May 2006 (in the early stages of disease arrival at the site) until February 2008. To determine the pattern of injuries between individuals in both populations, we recorded the presence and location of all injuries that resulted in penetration of the dermal layer and therefore might have the potential to transmit DFTD. Bites resulting in this type of injury usually heal within 2–8 weeks depending on their severity (R. Hamede & M. Jones, unpublished data); thus, the 3 month intervals at which data were collected precluded double counting bites in subsequent trapping sessions. In addition, only unhealed bites, those bites that presented visible disruption of the epidermis, were included in the data analysis. Injuries resulting from agonistic interactions with other carnivores are extremely rare (M. Jones, unpublished data).
All devils were individually marked with microchip transponders (Allflex NZ Ltd, Palmerston North, New Zealand). Disease status was assessed by histopathological examination of biopsies from tumours, or when this was not possible, by visual inspection and identification of tumours (see Hawkins et al. 2006 for visual detection methods). In our data set, we have only categorized individuals as DFTD infected if confirmed by histopathology analyses or if visually scored as definitive DFTD (Hawkins et al. 2006). We recorded the location (inside the oral cavity or outside) of primary tumours (which are invariably on the head) in all individuals. When captured individuals had more than one tumour, the larger tumour was regarded as the primary tumour.
Devils were aged using a combination of molar eruption, molar tooth wear and canine over eruption (distance from dentine-enamel junction to the gum) (M.J., unpublished data). This method is considered precise for ageing devils up to 3 years of age. Because both sites were monitored regularly, most individuals in our data set were captured as juveniles or younger than 3 years old and were therefore of known age. In our analyses, we excluded subadult devils (1 year old) as they usually do not get bitten until they become sexually mature (2 years old).
All analyses were implemented in r version 2.12 (R Development Core Team 2011). At West Pencil Pine, to determine whether the number of bites on an individual influenced its probability of acquiring DFTD in the future, we used Generalized Mixed Models, implemented using the lme4 package in R (Bates, Maechler & Dai 2008) with individual devils as a random term (to allow for the fact that individual devils were repeatedly captured) and a binomial error distribution. All models were fitted using maximum likelihood, which is appropriate when examining fixed effects (Crawley 2007). For all cases in which devils had been captured and then recaptured either 6 or 9 months subsequently, we examined whether the DFTD status (healthy or diseased) was associated with the number of bites recorded 6 or 9 months previously. We chose 6 and 9 month intervals as these time periods represent the best estimates available for the latent period of DFTD based on field (R. Hamede, unpublished data) and experimental observations (Kreiss et al. 2010). Because almost all primary tumours occur on the head (R.H. and M.J., unpublished data), we ran separate analyses using bites to the head only, bites to the body tail and limbs and all bites pooled. We restricted this analysis to West Pencil Pine as this was the only site at which individual devils were consistently recaptured throughout the sampling period.
Seasonal patterns of bites in both sites were analysed using Generalized Linear Mixed Models (GLMM) with normal error distribution, using the package lme4 (Bates, Maechler & Dai 2008). There was substantial variation in the mean number of bites per season between years at each site. We therefore used GLMM to test the fixed effects of season and site, with the interaction of year, site and season as the random term. As we were again comparing fixed effects, the models were fitted using maximum likelihood rather than restricted maximum likelihood.
We obtained population size estimates from mark-recapture data using MARK (Cooch & White 2002). Closed population estimates were obtained including heterogeneity in capture probabilities with time and between individuals (Chao, Lee & Jeng 1992), implemented in the program CAPTURE (Rexstad & Burnham 1992). Population size estimates and 95% confidence intervals were converted to population density by dividing population size by the combined area of a minimum convex polygon constructed around the trapping grid (Kenward 1985) and an additional two kilometre boundary strip that represents half the home range diameter of an individual devil (Pemberton 1990).
At West Pencil Pine, adult Tasmanian devils with a large number of total bites or bites elsewhere than the head were significantly less likely to develop DFTD after either a 6 month delay or a 9 month delay (Table 1) than devils with fewer bites. Devils with <5 total bites were more than twice as likely to develop DFTD as those with between 5 and 9 bites (Fig. 2a,c). In contrast, we found no significant (P = 0·05) evidence that the number of bites to the head affected the likelihood of developing DFTD in either 6 months time or 9 months time (Table 1), despite the majority (55%) of all bites recorded being to the head. There was no significant evidence that the probability of developing a tumour was influenced either by sex or by age, and the effect of total bites on the likelihood of developing DFTD remained significant when either age or sex was included in the model (Table S1). Primary tumours were predominantly observed inside rather than outside of the oral cavity at both sites (Fig. 3). There were no subadults infected with DFTD at West Pencil Pine.
Table 1. Results of generalized models predicting the probability of developing devil facial tumour disease (DFTD) as a function of the number of bites recorded 6 or 9 months previously, for all bites, bites to the head only and ‘non-head bites’ (to the body, tail or limbs)
Chi-squared tests for likelihood ratios relative to a null model containing only the intercept are shown, together with the estimated coefficient and standard error for the effect of the number of bites on the logit transformed probability of acquiring DFTD. Also shown is the exponent of the coefficient of the logistic regression, which is the estimated odds ratio for becoming infected for each additional bite. In these analyses, the actual number of bites recorded was used as a predictor variable (i.e. the bites were not aggregated into categories as in Fig. 2). Analyses based on 600 observations from 107 Tasmanian Devils.
Bites to head
Bites to head
Bites in subadults were very rare at both sites regardless of season (Table S2). The number of bites in adult devils was greater at both sites in autumn and winter than in spring and summer, although there was substantial interannual variation, particularly at West Pencil Pine (Fig. 4). Generalized mixed modelling using the interaction between site, year and season as the random term showed that a model including a seasonal effect but no site effect was most strongly supported by the data, with the number of bites being lower in spring and summer than in autumn (Table 2). Estimated population density was consistently lower at Wisedale than at WPP, in all 3 years for which estimates were available at both sites (Fig. 5).
Table 2. Highest posterior density estimates (95% confidence) of parameters for a generalized mixed model with year/site/season as the random term, predicting the total number of bites as a function of season and site, with season parameters relative to autumn and the site parameter relative to West Pencil Pine
MCMC sample with n = 1000. Based on 748 individual capture events from West Pencil Pine and 116 from Wisedale.
Whilst there is very strong evidence that DFTD is transmitted through transfer of live tumour cells between hosts (Pearse & Swift 2006), how transmission occurs in the field has remained unclear. Transmission has been presumed to occur through biting (Pearse & Swift 2006; Siddle et al. 2007). Transfer through fomites has not been discounted but is unlikely given that tumour cells do not survive for more than a few minutes outside the body of a devil [Tasmanian Department of Primary Industries, Parks, Water and Environment (DPIPWE), unpublished data]. There are two ways in which transmission through biting could occur: from the biter to the bitten or from the bitten animal to the biter. If transmission occurred predominantly by the former route, one would expect that those animals with more bites, particularly to the head (where primary tumours occur), would be those that should subsequently have a higher probability of developing DFTD. Tumour cells and clusters of cells have been swabbed from canine teeth of infected devils with ulcerated tumours in close proximity to the tooth. Transmission potential is assumed to increase with tumour size, as ulceration (Loh et al. 2006) and thus friability increases with tumour size and age. No tumour cells were found in canine smears from devils that had non-ulcerated tumours or tumours outside of the oral cavity (Obendorf & McGlashan 2008).
Our results are more consistent with transmission occurring predominantly by a susceptible animal with injuries and exposed flesh inside or around the mouth biting into the tumour of an infected individual, a behaviour we have observed in wild devils (R.H. and M.J., unpublished data), than through transfer of cells from biter to bitten. We found that devils with fewer bites overall, or elsewhere than the head, were significantly more likely to have developed tumours after either 6 or 9 months (Fig. 2, Table 1) and that this was not related to either age or sex (Table S1). We did not find any evidence of a relationship between propensity to develop DFTD and the number of bites to the head only. Our unexpected result that devils with fewer bites overall or elsewhere than the head are more likely to acquire DFTD is explicable if more aggressive animals bite others more frequently than do subordinate individuals, but these aggressive animals are themselves less likely to be bitten in return, particularly on locations other than their head. This implies that subordinate diseased individuals or animals that subsequently become subordinate because of the debilitating effects of fast growing tumours around the head (i.e. physiological damage, removal of dentition, metastasis) are more likely to pass on infection than dominant and/or aggressive individuals, but that at least initially aggressive animals are more likely to acquire infection. At present, we have insufficient data to infer the effect of infection status on hierarchical dominance in individuals or on social structure. Furthermore, preliminary evidence suggest that different strains of DFTD (Pearse et al. 2012) could have higher or lower levels of host tolerance and affect tumour growth rates differently (R.H., unpublished data). We are currently investigating evolutionary changes in both the Tasmanian devil and the tumour in response to this epidemic. From the results reported here, we expect selection for less aggressive phenotypes to favour host survival, although this could be counterbalanced by sexual selection, given that biting is associated with the mating season (Hamede et al. 2009) and that more dominant (aggressive) males are likely to achieve higher rates of paternity (M. Jones, unpublished data). In addition, given the propensity for pathogens to manipulate their host (Dobson 1988), we would expect the tumour to select for more aggressive individual devils to increase transmission.
The location of the majority of primary tumours inside the oral cavity (Fig. 3) adds additional support to biting of infected devils as the behaviour most likely to lead to infection with DFTD. If tumours were largely transferred through an infected individual biting a susceptible individual, we would expect tumours to be predominantly external. Whilst we do not have unequivocal evidence that tumours occur close to the site of original inoculation, in the small number of infections that have been experimentally induced by injecting or implanting tumour cells, the primary tumour has occurred at the site of the original inoculation (S. Pyecroft, unpublished data, G. Woods, unpublished data). As specialized scavengers that eat the tough parts of large prey carcasses (Jones 1997), devils frequently have open wounds on their lips or inside their mouth, either from biting each other or from feeding on tough and sharp food items such as bones and echidna (Tachyglossus aculeatus) spines (R.H. and M.J, unpublished data). A limitation of this study is that, because wounds inside the mouth could result from feeding, we could not assess whether those injuries occurred through individuals biting each other inside the oral cavity or by feeding. As a devil bites into the tumour of another devil, tumour cells could easily be embedded in such open wounds, particularly if the tumour was friable or ulcerated, leading to the growth of a tumour at the site of injury inside the mouth.
Patterns of biting and of social contacts in wild Tasmanian devils show clear seasonal differences (this study, Hamede, McCallum & Jones 2008; Hamede et al. 2009), even with substantial interannual variation, although this does not translate into a seasonal distribution in new cases of DFTD (McCallum et al. 2009). Biting injury rates at West Pencil Pine, at Wisedale and at a third site, the Freycinet Peninsula, on the East Coast of Tasmania (Hamede, McCallum & Jones 2008) were higher in autumn and winter following the mating season when male–male contests and intersexual interactions (mate choice, mating and mate guarding) peak. These results are concordant with different patterns of social contact, as revealed by proximity sensing radiocollars, between mating and non-mating seasons (Hamede et al. 2009). A distributed delay in the latent period of the disease in the wild, because of the number and location of cells transferred, natural variability in host susceptibility (genotype) and immunological response, or overall individual fitness may explain the lack of a seasonal trend in prevalence of DFTD (McCallum et al. 2009) despite seasonality in contacts and biting injuries. The latent period from experimental inoculations of pieces of tumour and tumour cells from cultured cell lines range from 1 to 4 months (Kreiss et al. 2010). However, such inoculations probably transfer more cells directly into a potential site for tumour establishment than would be the case for field transmission, and thus are likely to have a shorter latent period than in the wild. Data on latent period from wild populations come from a single anecdotal case in which a wild devil brought into captivity developed DFTD 10 months after its removal from the wild (DPIPWE, unpublished data).
We found no evidence of a significant difference in the mean number of bites between sites, despite the higher population density at West Pencil Pine than at Wisedale (Fig. 5), suggesting that the reduced impact and low progression of DFTD reported at West Pencil Pine (Hamede et al. 2012) is unlikely to be due to reduced contact rates. Two separate observations support why this might be the case. First, whilst the frequency of contacts observed at carcasses placed at feeding stations increases with population density (Hamede, McCallum & Jones 2008), when devils have been subsequently trapped and assessed for injury following behavioural observations (not done in Hamede, McCallum & Jones 2008), agonistic interactions are found to rarely result in penetrating injuries (Pemberton & Renouf 1993). Second, population density probably does not play a major role in the transmission dynamics of DFTD. Epidemiological (McCallum et al. 2009) and social network studies (Hamede et al. 2009) have suggested that transmission of DFTD is more consistent with a frequency-dependent than a density-dependent mode. This is consistent with the majority of biting occurring during mating encounters rather than during foraging interactions (Hamede, McCallum & Jones 2008).
Options for managing the disease and mitigating its impact on wild populations are limited. In the absence of any treatment for DFTD or a vaccine, nor prospects for either in the foreseeable future, and with the difficulties of controlling DFTD by selective culling (Lachish et al. 2010), identifying and manipulating host or tumour traits that are under selective pressure are possible management actions that could reduce transmission. Individual level genotypic and phenotypic variation in aggression is common in animal populations (Sih, Bell & Johnson 2004; van Oers et al. 2005). Identifying, with the view to systematically removing, highly aggressive phenotypes or genotypes that are particularly important in disease transmission (i.e. super spreaders) could be a viable management for DFTD affected populations. Intriguingly, our results suggest that these highly aggressive individuals will be ‘super receivers’ rather than ‘super spreaders’. These individuals could be recognized through collection of behavioural data that are indicative of ‘bold’ and ‘shy’ phenotypes during handling of trapped animals. Fruitful research directions include acquiring high-resolution behavioural data related to disease transmission, longitudinal data on changes in behaviour with infection and disease progression, and long-term longitudinal studies across populations assessing the dynamics of aggressive and dominance interactions, disease prevalence, infectious period and/or tumour strains. This type of management could enhance existing natural disease-induced directional selection on behavioural traits and lead to the rapid evolution of a less aggressive devil species that is more resilient to DFTD.
Determining behavioural and ecological circumstances associated with high risk of disease transmission is a critical step for understanding the epidemiology and impact of pathogens and parasites in their hosts (Hart 1988; Courchamp et al. 1998). Our results highlight the importance of investigating parasite and pathogen transmission in the wild to understand the dynamics of infectious diseases of wildlife and emphasize the role that heterogeneity in individual behaviour may play in susceptibility to infectious disease.
We thank the Gooch family and Gunns Ltd. for facilitating land access and logistic support. Particular thanks to C. & M. Walsh from Discovery Holiday Parks and Narawntapu National Park staff for providing accommodation and logistic support during field work. Sincere appreciation is extended to S. Peck for field veterinary support and to B. & P. Hopcroft and a large number of volunteers for assistance in data collection. This research project was funded by an Australian Research Council Linkage grant to M.J. and H.M. (LP0561120) and by an Eric Guiler Tasmanian Devil Research grant through the University of Tasmania and the Save the Tasmanian devil appeal to M.J and R.H. The Tasmanian Department of Primary Industries & Water and the ‘Save the Tasmanian devil Program’ provided financial assistance and logistic support. The Tasmanian Department of Economic Development sponsored R.H. under the State and Territory Nominated Independent Scheme Ref. No. STNA-ITAS200501. This research was carried out with approval from the University of Tasmania's Animal Ethics Committee (A0010296).