In general, the net effect of diversity upon disease dynamics in a focal host is likely to depend on the properties of that species relative to the entire community. For example, imagine a focal host species infected by a pathogen that can also infect other host species. If the focal host is a poor reservoir (i.e. does not transmit the pathogen effectively), adding other host species to the community might increase the prevalence of the pathogen in the focal host because the added hosts will be better reservoirs (‘spillover’sensuDaszak et al. 2000; Power & Mitchell 2004). To illustrate this, humans alone cannot sustain rabies because human-to-human transmission is negligible. The presence of additional species that can infect humans (e.g. raccoons); however, can sustain infection in humans because of animal-to-human transmission (encounter augmentation). In an experimental plant community, Power & Mitchell (2004) found that host communities containing grass species that were poor reservoirs had low rates of infection with barley yellow dwarf virus. More diverse systems had higher rates of infection because they contained a highly competent reservoir for the virus – the wild oat, Avena fatua (Fig. 4). Examples like these demonstrate how species diversity can sometimes amplify disease prevalence.
On the other hand, imagine a focal host species that is a highly competent reservoir. Adding host species that are less competent reservoirs might decrease disease risk if those added species decrease the probability of encounter between the pathogen and the focal host species (encounter reduction). This could occur, for example, for an environmentally transmitted disease, if the added species removed pathogens in a free-living, depletable pool of propagules. Intestinal parasites/pathogens that accumulate in latrines (e.g. raccoon roundworm, Baylisascaris procyonis; LoGiudice 2003) or decomposing tissues from infected carcasses (e.g. chronic wasting disease; T. Hobbs, personal communication) both serve as sites of infection, but both are at least potentially depletable by repeated visits from species that can harbour, but do not readily transmit, the propagules. These examples illustrate how host diversity could decrease disease prevalence.
Frequency-dependent vs. density-dependent transmission
Because of contradictory examples such as these, several recent theoretical studies have attempted to delineate under what general conditions host diversity should increase or decrease disease prevalence (Holt et al. 2003; Dobson 2004; Rudolf & Antonovics 2005). One key factor appears to be whether transmission of the pathogen is a function of the absolute density of infected hosts (density-dependent), or whether it is a function of the proportion of the total population that is infected with the pathogen (frequency-dependent). Density-dependent models of transmission are generally used to characterize diseases that are spread through environmental propagules or through random contact among individuals. Frequency-dependent transmission models are typically used to characterize sexually transmitted diseases (Getz & Pickering 1983; Thrall et al. 1993), because the number of sexual contacts is likely to be fixed, regardless of population density. Vector-borne diseases are also frequently considered to conform broadly to frequency-dependent models of transmission (e.g. Thrall et al. 1993), a situation that would apply if the number of contacts between vectors and hosts is fixed, e.g. because vectors actively search for their hosts and compensate for decreased density of hosts by increasing searching distances (Power 1987; Antonovics et al. 1995; Rudolf & Antonovics 2005).
Dobson (2004) and Rudolf & Antonovics (2005) argued that the effect on disease prevalence of adding host species will differ depending on whether the disease is characterized by density-dependent or frequency-dependent transmission. If pathogen transmission is density-dependent, adding hosts will typically decrease disease risk only if the added hosts reduce the abundance of the focal host (susceptible host regulation), assuming that transmission between species is lower than transmission within species. On the other hand, if the pathogen is transmitted in a frequency-dependent manner, adding hosts will decrease disease risk whether or not the added hosts reduce the abundance of the focal host. This is because adding a host species decreases the proportion of all infected individuals in the host community, resulting in a reduction in the number of contacts between susceptible and infected individuals (encounter reduction), again assuming that transmission between species is lower than transmission within species.
Between- vs. within-species transmission
The assumption that transmission is higher within species than between species is common to virtually all models of pathogen transmission among multiple host species and can even be required for host coexistence (e.g. Holt & Pickering 1985; Bowers & Begon 1991; Begon et al. 1992; Begon & Bowers 1994; Dobson 2004; Rudolf & Antonovics 2005). This assumption appears to be appropriate in many cases (Begon et al. 1999; Woolhouse et al. 2001). It also appears to be a necessary condition for host diversity to decrease disease risk. In cases with higher between- than within-species transmission, host diversity may increase disease prevalence. For example, Rhodes et al. (1998) found that side-striped jackal (Canis adustus) populations in Zimbabwe could not support rabies virus unless they were frequently reinoculated through contact with infected domestic dogs (encounter augmentation). Similarly, Caley & Hone (2004) used field and modelling efforts to establish that bovine tuberculosis (pathogen Mycobacterium bovis) in New Zealand was being maintained in low-density feral ferrets (Mustela furo) only through their contact with brushtail possums (Trichosurus vulpecula).
Holt et al. (2003) explored the consequences of relative rates of between- and within-species transmission for pathogen establishment in communities composed of two hosts. With only one host species and density-dependent transmission, there is a threshold density of that host above which the pathogen can become established; for pairs of hosts, there are various combined densities that permit establishment, depending on the amount of interspecific pathogen transmission. For example, at one extreme, if there is no interspecific transmission, at least one of the hosts must occur at or above its threshold density for the pathogen to become established. In contrast, if between-species transmission is greater than within-species transmission, the combination of host species more readily permits pathogen establishment than does either species alone – an example of disease amplification with increasing diversity. They also considered the possibility that one host cannot sustain the infection and, moreover, decreases the rate at which the other host becomes infected – an example of diversity diluting disease prevalence. In this case, as the density of the second host increases, the density of the first host required for pathogen establishment also increases. According to Holt et al. (2003), this latter situation is most plausible if transmission is via vectors or a depletable pool of environmental propagules.
Examples from multi-host systems
Most empirical investigations of the effects of diversity on disease risk have focused on vector-borne pathogens, despite the potential for host diversity to also influence the prevalence of pathogens that are directly or environmentally transmitted. Two recent studies of non-vector-borne diseases suggest that host diversity can reduce disease risk, though the mechanisms underlying these effects are not clear. In a study of the ecology of Laguna Negra virus (the aetiological agent for hantavirus pulmonary syndrome in Paraguay), Yahnke et al. (2001) found that host communities that had high proportions of the most competent reservoir – the vesper mouse, Calomys laucha– also had the highest antibody prevalence in this reservoir. Virus transmission appears to be primarily through direct contact (Yahnke et al. 2001). Thus, the probability of conspecific encounters between C. laucha individuals, and hence of potential transmission events, may have decreased as the relative abundance of this species declined with increasing diversity (encounter reduction). If there is a relatively fixed number of contacts per individual host, one expects frequency dependent transmission, and these results would then conform to the expectations of Dobson (2004) and Rudolf & Antonovics (2005). In a study of another hantavirus, Choclo virus, in Panamá, Ruedas et al. (2004) found that at sites where the virus was present (either in humans or wildlife), the host community was less diverse than at comparable sites where no virus was found, suggesting that diversity reduced disease prevalence. The mechanisms for this effect were not clear.
In a recent study of a vector-borne disease of wildlife, Telfer et al. (2005) found that the presence of bank voles (Clethrionomys glareolus) reduced the infection prevalence in wood mice (Apodemus sylvaticus) of species of Bartonella, bacteria vectored by fleas. Bank voles appear to be poor reservoirs for the pathogen, but good hosts for the flea vector. Flea prevalence did not increase with overall rodent density, suggesting that vector augmentation did not occur. Importantly, fleas were less abundant on wood mice when bank voles were present and wood mice were at high densities, strongly suggesting that encounter reduction between fleas and hosts may have taken place in this system.
Lyme disease, a vector-borne zoonosis in which a spirochete bacterium, Borrelia burgdorferi, is passed from host to host by the bite of an ixodid tick, provides one of the best-studied examples of the effects of host diversity on disease risk. The tick vectors in this system feed on a wide variety of vertebrate hosts, but the white-footed mouse (Peromyscus leucopus) is the most competent reservoir for the pathogen in eastern North America. Mice appear to be particularly abundant in small forest fragments because their predators and competitors are absent or scarce (Nupp & Swihart 1996; Krohne & Hoch 1999; Rosenblatt et al. 1999), providing examples of susceptible host regulation and infected host mortality. Encounter reduction also appears to operate in this system. When the density of chipmunks, an alternative host for ticks, is high, the number of ticks on mice is lower (Fig. 5; Schmidt et al. 1999; Ostfeld et al in press). This suggests that the presence of another species at high density (an increase in species evenness rather than species richness) reduces encounters between the vector and the most competent reservoir for the pathogen, though the evidence for this mechanism comes from correlative rather than experimental data. The presence of alternative (non-mouse) hosts in diverse host communities can lead to vector regulation, because ticks that feed on mice are more likely to survive to moult than are ticks that feed on some other hosts (Randolph 1979; Craig et al. 1996; LoGiudice et al. 2003). Thus, in a diverse community, ticks feed on a greater number of hosts, and these alternative hosts decrease their survival. Schauber & Ostfeld (2002) suggested that transmission reduction might also operate in the Lyme disease system.
Figure 5. The effect of eastern chipmunk (Tamias striatus) density, varying across years and sites, on the average number of larval blacklegged ticks (Ixodes scapularis) infesting white-footed mice (Peromyscus leucopus) in eastern New York state. In years of low chipmunk density, tick burdens on mice were variable. In years of high chipmunk density, however, tick burdens on mice were always low, suggesting that an abundance of an alternative host for the ticks reduced rates of encounter between ticks and white-footed mice, the most competent reservoir for the Lyme bacterium (Borrelia burgdorferi). These data provide an example of encounter reduction– a decline in encounters that could lead to infection as a result of increasing species diversity (see main text). Reprinted with permission from Ostfeld et al. (in press).
Download figure to PowerPoint
Allan et al. (2003) tested for a net effect of all of these mechanisms by evaluating disease risk in forest fragments in upstate New York. They predicted that the smallest fragments would have high densities of infected ticks, and thus high disease risk, because the small fragments had (a) high densities of white-footed mouse due to lower predation and/or competition; and (b) more tick meals being taken on mice because of loss of both encounter reduction and vector regulation. Consistent with these predictions, they found that densities of infected ticks were more than four times higher in small fragments than in larger fragments.
For louping ill, a similar tick-borne disease system, several studies (Norman et al. 1999; Gilbert et al. 2001; Laurenson et al. 2003) describe the results of modelling and empirical investigations in which the louping ill virus is transmitted among hosts by the bite of another ixodid tick (Ixodes ricinus). The roles of hosts in this system are complex: only sheep (Ovis aries) and red grouse (Lagopus lagopus) produce sufficient viraemia to pass the viral infection to ticks, but mountain hares (Lepus timidus) can both transmit the infection through co-feeding ticks and also sustain the vector population (Gilbert et al. 2001). Red deer (Cervus elephus) do not transmit the virus, but are the primary host for the tick vector and thus can sustain the tick population. Norman et al. (1999) and Gilbert et al. (2001) found that intermediate abundances of a non-viraemic host for the tick vector (e.g. red deer) permit viral persistence in a viraemic host (e.g. grouse), whereas high or low abundances lead to viral fadeout. At low deer abundance, there are too few ticks to sustain the pathogen (vector regulation); at high deer abundance, tick bites get ‘wasted’ on the non-viraemic deer (encounter reduction), and the pathogen cannot persist.
Whether the Bartonella, louping ill and Lyme disease systems conform to the predictions of Dobson (2004) and Rudolf & Antonovics (2005) is not clear because the relationship between host and vector abundances is not known. As Dobson (2004) pointed out, the net effect of host diversity for vector-borne diseases will be in part a consequence of whether vector abundance is a function of host abundance. In some cases, e.g. mosquitoes, vector abundance may be independent of host abundance (and limited instead by, for example, availability of breeding sites; Dobson 2004). But studies of the use of zooprophylaxis – the addition of non-human hosts to siphon vector meals away from humans – for malaria mitigation suggest that even for mosquitoes, this conclusion might not be straightforward, given that in some situations, adding hosts increases mosquito density (e.g. Saul 2003). Whether the abundance of tick vectors is a function of host abundance remains controversial (Van Buskirk & Ostfeld 1995; Norman et al. 1999; Gilbert et al. 2001; Schmidt & Ostfeld 2001, R.S. Ostfeld, personal communication). For example, white-tailed deer (Odocoileus virginianus) are the primary hosts for adult ticks in eastern North America, and in areas (e.g. islands) where they have been extirpated, tick abundance is essentially zero. However, an empirically based model developed by Van Buskirk & Ostfeld (1995) found that even very small numbers of deer sustained substantial tick populations, suggesting that tick abundance is not a linear function of deer abundance. A crucial area requiring attention is the determination of what factors limit and regulate vector populations.
In cases where vector abundance is independent of host density, frequency-dependent rather than density-dependent transmission may best describe transmission dynamics (Dobson 2004); these diseases would be predicted to show reduced disease prevalence with increasing diversity (Dobson 2004; Rudolf & Antonovics 2005). Cases in which vector abundance is dependent on host abundance are more complex, and outcomes are much less easy to predict (Van Buskirk & Ostfeld 1995; Norman et al. 1999; Gilbert et al. 2001; Schmidt & Ostfeld 2001; Dobson 2004), especially when there are nonlinearities in the relationship between host abundance and vector abundance (Van Buskirk & Ostfeld 1995; Norman et al. 1999). In these cases, a simple tally of species presence/absence may be insufficient to gauge the importance of diversity for disease dynamics in a focal host species, because different processes dominate at different population sizes. Going from zero to low densities, an alternative host that is critical to vector dynamics might boost vector numbers, leading to an increase in disease in the focal host species. However, if the alternative host is ineffective at sustaining the pathogen, further increases may lead to a reduction in disease prevalence (Van Buskirk & Ostfeld 1995; Norman et al. 1999; Gilbert et al. 2001). Similar non-monotonic effects arise broadly in trophic interactions due to the interplay of nonlinear functional and numerical responses (Holt 1997).