Special Viewpoints

Population Biology of Multihost Pathogens

See allHide authors and affiliations

Science  11 May 2001:
Vol. 292, Issue 5519, pp. 1109-1112
DOI: 10.1126/science.1059026


The majority of pathogens, including many of medical and veterinary importance, can infect more than one species of host. Population biology has yet to explain why perceived evolutionary advantages of pathogen specialization are, in practice, outweighed by those of generalization. Factors that predispose pathogens to generalism include high levels of genetic diversity and abundant opportunities for cross-species transmission, and the taxonomic distributions of generalists and specialists appear to reflect these factors. Generalism also has consequences for the evolution of virulence and for pathogen epidemiology, making both much less predictable. The evolutionary advantages and disadvantages of generalism are so finely balanced that even closely related pathogens can have very different host range sizes.

Most pathogens are capable of infecting more than one host species. This includes the 60% of human pathogen species that are zoonotic (1), causing diseases of major public health concern such as influenza, sleeping sickness, Lyme disease, food poisoning, and variant CJD. It also includes more than 80% of pathogens of domestic animals (2), notably those causing 57 of the 70 livestock diseases of greatest international importance (3), such as rinderpest, foot-and-mouth disease, and heartwater. Pathogens such as influenza A virus, rabies virus, and Blastocystis hominiscan infect hosts not only of different species but from different orders or classes (2). Yet, despite their ubiquity and importance, multihost pathogens have been largely neglected by population biologists in favor of the simpler paradigm of a single-host species.

Many, though not all, pathogens that can infect multiple hosts can also be transmitted by multiple hosts, and these can be regarded as ecological generalists rather than specialists. The advantages of generalism are poorly understood: it has been suggested that evolution should favor specialism, either because of the existence of functional trade-offs that limit the fitness of generalists in any one habitat or because evolution may proceed faster within narrower niches (4); these arguments apply especially to pathogens because they are under selection pressure to coevolve with their hosts (5). Yet paradoxically, only a minority of pathogens are specialists in the sense that they exploit a single host species.

So what processes lead to pathogens having multiple hosts, and why do multihost pathogens seem so pervasive? The evolution of generalism requires that pathogens have both the capability to exploit potential alternative host species and the opportunity to transmit to them. The subsequent maintenance of generalism depends on the consequences of an increased host range for pathogen population biology, especially such features as pathogenicity and epidemiology.

Capability to Infect Multiple Hosts

Pathogens are usually, though not always, less infectious to a different host species. This is referred to as the species barrier (6), and there are two main strategies for overcoming it. Some pathogens have an inherent ability to infect multiple host species; for example, Trypanosoma brucei rhodesiense has a number of variant surface glycoprotein genes that encode for receptors with different affinities to specific mammalian transferrins (7). More commonly, pathogens produce many different genetic variants, some of which become associated with different host species, e.g., rabies (8). Gene products involved in host specificity have been identified for some pathogens, such as human immunodeficiency virus (HIV), mouse hepatitis virus, and Citrobacter rodentium (9).

Genetic change associated with host switching constitutes host adaptation. This may involve a small number of nucleotide substitutions or more major genetic changes such as reassortment, e.g., influenza A (10), or the acquisition of genetic elements (sometimes associated with virulence as well as host specificity), e.g., Salmonella typhimurium (11). Host adaptation can be so rapid that pathogen lineages adapt to different host tissues (12) or to vector versus host cells (13).

Species barriers are routinely crossed by some pathogens (such as rabies virus, which is regarded as a true multihost pathogen), but much more rarely by others [such as simian immunodeficiency virus, which is thought to have been transmitted to humans from other primates only very rarely and to have diverged rapidly into new single-host pathogens, HIV-1 and HIV-2 (14)]. Another example of pathogen speciation associated with host shift is feline panleukopenia virus in cats evolving into canine parvovirus in dogs (15).

The extent of host adaptation is therefore linked to and limited by the genetic variability of the pathogen. In practice, pathogen populations are characterized by very high genetic diversity, which facilitates evasion of (i) within-host immune responses, (ii) herd immunity in the host population, and (iii) localized evolution of host resistance (16). A variety of mechanisms have evolved to generate this diversity, potentially affecting genes involved in all aspects of the pathogen-host interaction (17). In addition, most pathogens produce vast numbers of transmission stages (18). The implication is that a huge array of different pathogen genotypes will be available at any one time. Because a novel host may become infected even if exposed to a very small number of compatible pathogens (19), the frequent evolution of generalism is a likely outcome.

This interpretation of host adaptation and the species barrier leads to testable predictions. For example, pathogens with higher mutation rates should produce more genetic variants and are therefore more likely to be generalists. RNA viruses have a mutation rate per genome replication estimated to be 300 times higher than DNA viruses (20). Consistent with this, among pathogens infecting humans, RNA viruses are more likely to be zoonotic than DNA viruses (67% compared to 36%, respectively, for those transmitted by direct contact, which is the usual route for DNA viruses).

Opportunity to Infect Multiple Hosts

Pathogens may have differing opportunities to infect multiple hosts according to their route of transmission. Pathogens exit hosts, are transmitted between hosts, and enter new hosts by many different routes, which can be categorized as direct contact (physical contact or close proximity), indirect contact (including contamination of food, contact with environmental reservoirs, and contact with free-living infectious stages, including those emerging from intermediate hosts), and vector-borne (via biting arthropods, including mechanical transmission).

Some transmission routes (e.g., direct transmission by sexual contact or in utero) provide extremely limited opportunity for infecting other species, whereas others (e.g., indirect transmission involving widespread contamination of the environment) provide many such opportunities. For vector-borne pathogens, the vector itself determines whether there are opportunities for interspecific transmission. Many biting arthropods are generalist feeders, but some specialize on a single host species (21). As expected, most of the few vector-borne human pathogens that are not zoonotic [12 species (1)] are transmitted entirely or predominantly by anthropophilic vectors, e.g., Plasmodium falciparum(transmitted by anopheline mosquitoes), Wuchereria bancrofti(mainly by Culex mosquito species), and Borrelia recurrentis and Bartonella quintana (both by the lousePediculus humanus). However, even for generalist vectors, a combination of host preferences, host distribution, and vector dispersal pattern can limit opportunities for interspecific transmission (22).

Different transmission routes are also associated with different reproductive costs and benefits of generalism or specialism. This particularly applies to pathogens transmitted by generalist vectors. For these pathogens, the presence of an incompatible host species can have a high reproductive cost, because transmission opportunities for vector-borne pathogens are limited—a blood meal taken by a vector on one host means a blood meal not taken on another (Fig. 1A). In contrast, for pathogens transmitted by direct or indirect contact, the presence of a second host species need not have such an impact, because of the overproduction of infective stages, although there are exceptions: e.g., pathogens such as Gnathostoma spinigerum that are transmitted by predation.

Figure 1

Theoretical aspects of multihost pathogen biology. (A) Effects of a noncompatible host population on transmission potential. The density of a second, incompatible, host relative to the first, compatible, host (H 2/H 1 withH 1 constant, log scale) is related to the overall transmission potential (relative R 0). For pathogens transmitted by direct or indirect contact, according to the standard “SIR” model (28), R 0H 1 and is independent ofH 2 (top line). For vector-borne pathogens, according to the Ross-Macdonald model (28),R 0V[H 1/(H 1+H 2)]2, where V is the density of vectors. Here, V may be constant (bottom line) or proportional to overall host density,H 1 + H 2 (middle line). (B) Relationship between optimal virulence in the first host and transmissibility in the second host (where optimal refers to the value which maximizes overall R 0). The model is R 0 ∝ λ1/(α1 + μ1) + λ2/(α2 + μ2) where μi are the sums of natural mortality and recovery rates, αi are the additional mortality rates due to infection, i.e., virulence, and λi =c iαi qiwhere c i are indices of transmissibility andq i quantify the benefits of virulence for transmissibility (where 0 ≤ q i ≤ 1). Values of αi are related by α2 =k + mα1, where k,m, and q 2/q 1are constants. The model exhibits a range of behaviors, illustrated forq 1 = 0.5; μ1 = μ2 = 0.5; k = 0, m = 1 (bold lines); k = 1, m = −1 (narrow lines); and different values of q2 (as shown). (C) Relationship between expected final outbreak size (I , log scale) and R 0for different numbers of primary cases (I 0). The model is the recursive equation I =N − (NI 0)exp[−R 0 I /N] [adapted from (36)], where N is the total population size (assumed to be very large).

These arguments imply that transmission biology influences both opportunities for generalism and costs of specialism. Consistent with this, among human pathogens, the trend is that those transmitted by direct contact are less likely to be zoonotic than those transmitted by indirect contact, and those transmitted by vectors are most likely to be zoonotic (Fig. 2A). These patterns will at least partly reflect phylogeny (23), although different transmission routes are thought to have evolved independently many times (24).

Figure 2

Empirical aspects of multihost pathogen biology. (A) Relationship between the fraction of human pathogen species that are zoonotic and mode of transmission for different taxonomic groups [data from (1)]. Total numbers of species in each category are shown. Note that some pathogens have more than one possible mode of transmission and that for some species the mode of transmission is unknown. Very few species of fungi are transmitted by vectors, and no helminths are transmitted by direct contact. (B) Frequency distributions for reported sizes of outbreaks of disease caused by E. coli O157 in Scotland 1984–99, Ebola virus worldwide 1976–2000, and Staphylococcus aureus in Canada 1985–86 (37). All three distributions are highly aggregated with variance-to-mean ratios for numbers of cases 120.4, 183.1, and 34.6, respectively. (C) Relationship between log(variance) and log(mean) for outbreak sizes. The line shows a fitted linear regression with slope 2.25 (95% confidence limits ± 0.08, R 2 = 0.97). Data are published reports of numbers of clinical cases during outbreaks ofClostridium botulinum, S. aureus, Bacillus cereus, Campylobacter spp. (2 data points), human echovirus, Salmonella spp. (3 data points),Clostridium perfringens, E. coli O157 (2 data points), Shigella spp., small round structured virus,Listeria spp., Ebola virus, Cryptosporidium spp. (2 data points), Trichinella spiralis, measles virus, human polio virus, and Vibrio cholerae(38).

Consequences of Infecting Multiple Hosts

An important, but largely unconsidered, consequence of generalism is its effect on pathogenicity. Evidence from theoretical, laboratory, and field studies indicates that single-host pathogens evolve to an optimum level of virulence (25) determined by the trade off between virulence and transmissibility (26). For multihost pathogens, however, the situation is much more complicated. Depending on how transmission and virulence are related within and between host species, the pathogen can be more or less virulent in a second host than the first, and introduction of a second host can lead to an increase or a decrease in virulence in the first host (Fig. 1B). Moreover, if the second host contributes little to pathogen fitness, then there is no selective constraint on pathogen virulence in that host, which may explain why some zoonotic pathogens in which humans are “dead end” hosts, such as Echinococcus multilocularis or hantaviruses, are unusually virulent (27). Another complication is that some pathogens have only recently acquired the opportunity or ability to infect novel host species (e.g.,Escherichia coli O157 in humans) and so may not yet have evolved optimal levels of virulence in that host. Overall, theory suggests no simple rule as to whether multihost pathogens will be more or less virulent than single-host pathogens, and there are examples of human pathogens showing all possible combinations of high or low virulence, long-standing or recent associations with humans, and generalism or specialism.

Another consequence of generalism is its effect on pathogen epidemiology. Single-host pathogens must, by definition, be able to persist in their host species. This requires that the transmission potential (R 0) exceeds one and that the host population exceeds a critical size (28); thus, endangered species are rarely threatened by specialist pathogens. In contrast, multihost pathogens can affect hosts in which they do not persist independently, e.g., Escherichia coli O157 in humans or measles in nonhuman primates. Such pathogens must have reservoir host(s) in which they do persist (29), but occur outside that reservoir as localized outbreaks (30). Outbreak sizes are expected to be overdispersed, i.e., most outbreaks will be small but a few very large (Fig. 1C). This is because expected average outbreak size is nonlinearly related both to the number of primary cases (between-species transmission) and to the basic reproduction number (R 0) within the local population (within-species transmission), so that small variations in either can lead to large variations in outbreak size (31). This prediction is supported by data for a variety of infectious diseases (e.g., Fig. 2B). In general, the mean and variance of outbreak sizes obey Taylor's Power Law (32) with exponent greater than 2 (Fig. 2C), indicative of overdispersed distributions with longer tails. Outbreaks outside a major reservoir, characteristic of many multihost pathogens, can therefore be of variable and unpredictable timing and magnitude. The epidemiological and evolutionary processes that determine whether a pathogen occurs in an alternative host as sporadic infections (e.g., rabies in humans), minor outbreaks (e.g., Ebola), or major epidemics (e.g., influenza A) remain poorly understood, but are of crucial importance, notably in the context of emerging diseases.

These and other consequences of generalism have implications for macroevolutionary patterns of hosts use. Single-species pathogens are under strong selection pressure to, and appear predisposed to, cross the species barrier and exploit additional host populations, thereby becoming multihost pathogens. However, crossing the species barrier can itself lead to further pathogen specialization and subsequent speciation, and the ease with which they achieve this may explain why there are so many kinds of pathogens (33). Moreover, it is likely that generalist pathogens will be less prone to extinction, because their fate is not tied to that of a single host species (34). Thus, there are evolutionary advantages and disadvantages of generalism and specialism, and observed host range will be determined by a delicate balance between powerful selective pressures in both directions (5). The result is that pathogens may shift rapidly from one strategy to the other at different phases of their evolutionary history and that even closely related pathogens may have very different host range sizes (35).

Population biologists have been very successful in developing a formal understanding of the dynamics and evolution of single-host pathogens (27, 28). Understanding the more complex population biology of multihost pathogens will be one of the major challenges to biomedical science in the 21st century.

  • * To whom correspondence should be addressed. E-mail: mark.woolhouse{at}ed.ac.uk


View Abstract

Navigate This Article