Performed the modeling: JB. Helped write the paper: JB TJ MW. Helped conceive of the idea: JB TJ MW. Provided information on dragon life histories: TJ. Provided information on bacterial natural history and culturing: MW.
The authors have declared that no competing interests exist.
Komodo dragons, the world's largest lizard, dispatch their large ungulate prey by biting and tearing flesh. If a prey escapes, oral bacteria inoculated into the wound reputedly induce a sepsis that augments later prey capture by the same or other lizards. However, the ecological and evolutionary basis of sepsis in Komodo prey acquisition is controversial. Two models have been proposed. The “bacteria as venom” model postulates that the oral flora directly benefits the lizard in prey capture irrespective of any benefit to the bacteria. The “passive acquisition” model is that the oral flora of lizards reflects the bacteria found in carrion and sick prey, with no relevance to the ability to induce sepsis in subsequent prey. A third model is proposed and analyzed here, the “lizard-lizard epidemic” model. In this model, bacteria are spread indirectly from one lizard mouth to another. Prey escaping an initial attack act as vectors in infecting new lizards. This model requires specific life history characteristics and ways to refute the model based on these characteristics are proposed and tested. Dragon life histories (some details of which are reported here) prove remarkably consistent with the model, especially that multiple, unrelated lizards feed communally on large carcasses and that escaping, wounded prey are ultimately fed on by other lizards. The identities and evolutionary histories of bacteria in the oral flora may yield the most useful additional insights for further testing the epidemic model and can now be obtained with new technologies.
The Komodo dragon (
In some cases, the ultimate demise of a prey is purportedly due to more than just direct bite induced trauma, involving bacterial sepsis acquired from the lizard's bite
Here we focus on the possible origins and ecological bases of sepsis-inducing bacteria in the mouths of komodo dragons. Auffenberg
The purpose of this paper is to propose and analyze a third model, the ‘lizard-lizard epidemic’ model. Instead of interpreting the infectious oral flora of dragons as either beneficial to the lizards or as a byproduct of feeding on mammals and carrion, the model proposes that bacteria spread epidemically among lizard mouths via prey that escape an initial attack. Escaping, infected prey thus serve as vectors to spread the infection among lizard mouths. Predictions are developed for this model and compared to Komodo dragon life history details in attempting to refute the model and discriminate it from the alternatives. Original observations on wild dragons are reported here as part of this test.
The model describes the spread of a bacterium among lizard mouths. In humans and other highly social species, an oral infection might be spread directly by kissing or by other personal contact that led to contamination of the mouth. Like most lizards, Komodo dragons are largely solitary from birth and generally remain asocial
The oral bacterium causes sepsis. Sepsis has important effects on transmission both from lizard to prey and from prey back to lizard. First, a sepsis-causing, oral bacterium may be established in the mammalian prey easily by a lizard bite, because the bite initiates an infection. Auffenberg
We now develop formal predictions from this model. A useful concept of an infectious agent (‘disease’) is the basic reproductive number, R0, defined as the average number of new infections transmitted by the first infected individual in a population of naive hosts. R0 applies over the lifetime of the first infection, so it is, in essence, offspring number of a disease agent. R0 must exceed 1.0 for the disease to spread and be maintained; values progressively larger than 1.0 result in faster spread and higher equilibrium densities of infected hosts. (We use the term ‘disease’ without prejudice for whether it harms the lizard.) With this requirement, the first lizard to acquire the sepsis-inducing oral flora must on average spread it to more than one other lizard before the first lizard dies or loses the flora.
Under this model, the ecological characteristics fostering the spread of a sepsis-inducing oral flora include the following:
Prey escape lizard attacks after being bitten
Bacteria acquired from the lizard mouth cause a systemic infection in prey, achieving high densities in tissues that are normally consumed by lizards.
Infected prey are later eaten by lizards, enhanced by The bacteria weaken or kill large prey, facilitating subsequent capture by lizards Infected, large prey do not escape to habitats not frequented by dragons (e.g., savanna grassland, the majority of island habitat). Prey are sufficiently large to enable multiple (unrelated) lizards to consume a single carcass
Bacteria survive in a dead carcass, colonize and reproduce in a lizard mouth.
The main consequence of these properties/assumptions is that the same infectious bacteria come to exist in different lizard mouths. Given that large prey carcasses are typically eaten by the largest lizards, the epidemic spread of infectious bacteria is largely confined to large lizards.
These characteristics are presented as qualitative requirements; there are necessarily quantitative constraints on them as well. The
We discuss these four points in order of the evidence we have to address them. The third point will be discussed before the second.
Auffenberg
An injured Timor deer that has escaped an initial lizard has fled to the ocean and is being stalked by a different dragon. This second lizard succeeded in killing the deer. The Komodo dragon pictured is wearing a GPS collar used for tracking the animal. Occasional prey escape is essential to the operation of the lizard-lizard epidemic model. (Photo by Achmad Ariefiandy.)
It is perhaps no coincidence that prey escape rates are moderately high and that those prey are ungulate mammals as large as or larger than the lizards. Large prey size is likely key to the operation of this model: it (a) confers an increased probability of escape, (b) enables prey to be more tolerant of injury, enabling bacterial sepsis to develop and (c) imposes consumption limits on an individual lizard, so that multiple lizards feed concurrently or sequentially; an infected, large prey can potentially infect over a dozen lizards (see below).
Prey consumption by ‘other’ lizards (those not infecting the prey) is critical to the model. If injured prey are never encountered again or are only eaten by the lizard responsible for the initial attack, an oral infection would not spread. Perhaps surprisingly, dragon feeding behaviors are especially conducive to the spread of bacteria under this model. On a qualitative level, lizard density, movements and spatial structuring are sufficient and overlapping at least in some parts of the islands, that sick prey are likely to be consumed by one or more large lizards. The lizards scavenge carcasses (with a preference for fresh kills) and typically aggregate at kills of large prey (e.g., water buffalo, personal observations by TSJ and
In the course of field studies (by TSJ), twenty independent dragon feeding episodes on different large prey were observed: deer (8 observations), buffalo (7), pigs (4) and one sea turtle. Seventy per cent of these kills involved feeding by multiple lizards (
Incidence of single versus multiple Komodo dragons feeding on large prey (Timor deer, wild pigs, water buffalo, and a Hawksbill sea turtle). Based on 20 independent observations of Komodo dragon feedings noted during fieldwork activities between 2002–2009.
Three dragons are feeding on a wild pig. The large prey size of Komodo dragons with overlapping lizard home range generally precludes a single lizard from consuming its large prey alone. Multiple or shared feeding facilitates spread of infectious bacteria between lizard mouths. (Photo by Achmad Ariefiandy.)
There are miscellaneous reports of infections from dragon bites. Auffenberg
As discussed next, the data on dragon oral bacterial identities and characteristics are consistent with the lizard-lizard epidemic model but, on close inspection, are not detailed enough to provide much resolution. It should be realized that the ‘epidemic’ model, if correct, may apply to as little as a single species in dragon mouths but could also apply to a consortium of several species, no one of which is sufficient to cause an infection in prey. However, since the infectious flora is maintained via an epidemic among lizards, the same infectious bacteria should be found in different (adult) lizards, at least within each physically defined lizard population. Over time, the dominant infectious bacteria may turn over, and which bacterial species prevail(s) will depend on dynamical interactions in the lizard mouth as well as on the effects on prey.
The single published survey of bacteria cultured from wild Komodo dragon mouths observed a diverse bacterial population
The sampled oral flora from wild lizards included
Nonetheless,
(B) The most abundant bacterial species was found in only 14 of the 26 wild dragon mouths tested (an unidentified
The methodology used by Montgomery et al.
The data reported by the Montgomery study are thus inconclusive in supporting or refuting the epidemic model. They are useful in illustrating both the difficulty of using standard culture methods to assess bacterial communities and in revealing that dragon mouths harbor many bacterial species that could potentially spread as an oral epidemic under the right life history conditions.
The R0 of an oral bacterium is enhanced by its ability to colonize and persist in the lizard mouth and especially by its endurance in a carcass. As carrion is a common element in the lizard diet, a bacterium that causes prey death and also persists indefinitely after prey death has an excellent chance of introduction into lizard mouths. Even for septic prey that are killed by lizards, a large prey (e.g., water buffalo) may be consumed over several days, and survival in carrion affords the bacterium further opportunities to colonize lizards.
There are now three explanations for an infectious flora of lizard mouths, the ‘bacteria as venom’ model, the ‘passive acquisition’ model and the ‘lizard-lizard epidemic’ model. This paper has concentrated on developing and testing the latter. Here, we turn to consider all three.
The ‘bacteria as venom’ model was proposed casually, without specifying many of its properties
The two other models remain tenable at this point. There is, however, a strong asymmetry in the ease with which each is refuted. The ‘passive acquisition’ model, also proposed casually, potentially encompasses a wide variety of mechanisms by which the lizards acquire bacteria from their environment (
The arrows show the transmission of bacteria that ultimately colonize lizard mouths. In the lizard-lizard epidemic model, bacteria colonize new lizard mouths (L2, L3) when those lizards eat prey (P1) infected from another lizard (L1). The prey ultimately eaten by lizards 1 and 2 must be injured during but escape a lizard attack and survive long enough for the infection to develop. In contrast, lizards in the passive acquisition model acquire their oral bacteria either directly from the environment (E, lower arrow) or from prey that acquired bacteria environmentally. There are no chains of transmission between lizards in the latter model.
Possible Observation | Model refuted |
Infectious bacterial strain is found only in lizard mouths and bitten prey | PA |
Infectious bacterial strain found in young lizards | LLE |
Most adult lizards do not share the same infectious strain | LLE |
Bacterial strain tracked from lizard through prey to other lizards | PA |
The lizard-lizard epidemic model proposed here can operate in other predators, including mammals. Auffenberg
Since 2002, TSJ has been involved in field ecology studies of
To test whether our intuition about the process of the infectious spread of an oral flora is accurate, we used a quantitative model. This model is necessarily elementary, both because we cannot hope to capture the complexities of lizard ecology and population structure in any manageable set of equations and because there is no information with which to parameterize the model.
The model addresses whether an infectious oral flora can spread when rare in a population of uninfected lizards. This model does not address the full dynamics, but restricting the analysis to invasion dynamics simplifies the analysis, because it means that we can confine the dynamics to just two groups: infected lizards and infected prey. The much larger populations of uninfected lizards and uninfected prey can be considered constant as long as the infected types are rare, and interactions between infected prey and infected lizards can be ignored; the time scale considered is short enough to neglect lizard death as a source of change in the abundance of infected lizards.
In setting up the equations, we note that
Numbers of infected lizards change in two ways: a reduction when previously infected lizards lose their oral flora; a gain when uninfected lizards eat infected prey. This gain is higher with shared feeding (a function of prey size), captured with the parameter
Numbers of infected prey change in 3 ways: a gain when infected lizards bite uninfected prey that escape the initial attack; a loss when infected prey die without being eaten; a loss when infected prey are eaten by one or more lizards.
Equations for the changes in abundances of both types are thus
Note that some parameters encompass multiple steps. For example,
The characteristic equation for this system is
The infection spreads if
Several properties of this result agree with intuition. First, high rates of (i) communal feeding and colonization of lizard mouths (
The model enables us to present a verbal description about the type of life history needed to explain an infectious oral flora as an infectious ‘disease’ of lizards (as per
This analysis addresses only the spread of the oral flora when rare. A complete model would provide dynamics—how common the infection was among lizards at dynamical equilibrium, oscillations, and turnover rates. Furthermore, the model has omitted age structure of the lizards: an infectious oral flora should be more common in old animals than young ones, because (i) it must be acquired after birth, (ii) young dragons do not eat large prey, and (iii) even medium-sized dragons are less inclined to feed communally than large ones due to size-assortative competition for prey resources. There are thus obvious embellishments to include as data accumulate.
We thank Achmad Ariefiandy for the photos of dragons. We thank the reviewers for useful comments, and one reviewer specifically helped us appreciate the extent to which pathogenic bacteria are found in mouths of many vertebrates.