Corrigendum to Streicker et al. (2013) Differential sources of host species heterogeneity influence the transmission and control of multi‐host parasites

Abstract

In a recent article, we described a conceptual and analytical model to identify the key host species for parasite transmission in multi‐host communities and used data from 11 gastro‐intestinal parasites infecting up to five small mammal host species as an illustrative example of how the framework could be applied. A limitation of these empirical data was uncertainty in the identification of parasite species using egg/oocyst morphology, which could overestimate parasite sharing between host species. Here, we show that the key results of the original analysis, namely that (1) parasites naturally infect multiple host species, but typically rely on a small subset of infected host species for long‐term maintenance, (2) that different mechanisms underlie how particular host species dominate transmission and (3) that these different mechanisms influence the predicted efficiency of disease control measures, are robust to analysis of a smaller subset of host–parasite combinations that we have greatest confidence in identifying. We further comment briefly on the need for accurate parasite identification, ideally using molecular techniques to quantify cross‐species transmission and differentiate covert host specificity from true host generalism.

In a recent paper (Streicker et al. 2013), we used parasite prevalence and egg oocyst shedding data collected from several parasite species from small mammal communities in the eastern United States to illustrate the application of a conceptual and analytical method of identifying different types of ‘key host’ species that contribute substantially to parasite transmission within ‘multihost’ communities. The data relied solely on egg or oocyst morphology to identify the different parasite species, which we acknowledged was an imperfect method of parasite identification and, for a few parasites, our analyses were limited to identification at the pseudospecies level. As such it is likely that in some cases different parasite species were grouped as the same species due to similarities in egg morphology, giving an inaccurate reflection of the true distribution of those parasite species across the host community. Here, we present a revised analysis of those data in which we focus only on the subset of host and parasite combinations for which we have the greatest confidence in identification (19 of the original 40 host‐parasite combinations). In Table 1 we provide an updated version of Table S2 from the original paper, showing the host–parasite combinations that we include here.

Table 1.

Geographic and host species distributions of parasites encountered in the study

	Siteb
Parasite species	MLBS	CF	GMF	IES	GSM‐L	GSM‐H
Coccidia
E. arizonensis A	PL, PMa	PL	PL, PM	PL	PL	PM
E. arizonensis B	PL, PM	PL	PL, PM	PL	PL	PM
E. delicata	PL, PM	PL	PL, PM	PL	PL	PM
Nematodes
A. americana	PL	PL	PL, PM, MG	PL
C. americana	PL, PM, TS	PL, TS, BB	MG	PL, TS, BB		PM, TS, MG
Pterogodermatites A	TS		PL, MG	PL	PL	PM, BB, MG
Pterogodermatites B	TS	BB	MG	PL, BB	PL	PM, BB
Strongyle A	PL	TS, BB		TS	PL	PM, MG
Cestodes
Cestode A		PL	PL, PM, MG	PL, BB
Hymenolepis A c	PL, PM	PL		TS, BB	TS	PM, BB, MG
Hymenolepis B d	PL	BB		BB	PL	PM, BB

^aSpecies abbreviations denote infection of at least 1 individual of that species; vacant cells indicate parasite absence that site. Abbreviations for infected host species names are as follows: PL, Peromyscus leucopus; PM, Peromyscus maniculatus; TS, Tamias striatus; BB, Blarina brevicauda; MG, Myodes gapperi. Data are pooled across multiple trapping grids in each site (see Fig. 2 from original paper).

^bAbbreviations for site names are as follows: MLBS, Mountain Lake Biological Station, Virginia; CF, Center Forest, Virginia; GMF, Great Mountain Forest, Connecticut; IES, Cary Institute of Ecosystem Studies; GSM‐L, Great Smoky Mountains National Park (≤ 638 m), Tennessee; GSM‐H, Great Smoky Mountains National Park (≥ 785 m), Tennessee.

^cTaxonomy uncertain to species level, formerly suggested to be Hymenolepis dimunata.

^dTaxonomy uncertain to species level, formerly suggested to be Hymenolepis citelli.

Underlined text indicates host‐parasite‐locality combinations that were removed for the more conservative analysis.

Even within our reduced data set, we still find an absence of single‐host parasites and a strong tendency for a small subset of the infected host community to dominate parasite transmission by contributing the majority of infective stages. Interspecific differences in infection prevalence and shedding of infectious stages (Table 1 in the original paper) are largely replicated in this more conservative analysis (Table 2). Exceptions are Cestode A and Hymenolepis B (formerly assigned as being H. citelli), where previously significant variation in prevalence among host species is less well supported (P = 0.14 and P = 0.08 respectively Table 2). Importantly though, as in the original analysis, different mechanisms underlie how certain host species dominate parasite transmission, with examples of ‘super‐abundant’ key hosts (e.g. Eimeria arizonensis A and E. arizonensis B) and ‘super‐infected’ key hosts (e.g. Capillaria americana), while others showed a combination of asymmetries, for example being both ‘super‐infected’ and ‘super‐shedding’ key hosts (e.g. Hymenolepis B) (Fig. 1). In two cases, previously identified key hosts were removed from the more conservative analysis. This resulted in identification of Peromyscus leucopus and P. maniculatus as key hosts for Cestode A and Hymenolepis B respectively. In both cases, the newly identified key hosts contributed over 90% of infectious stages. Thus, the more conservative analyses support our central conclusion that key host species are typical for multi‐host parasites and, importantly, these arise through distinct biological mechanisms.

Table 2.

Results of generalised linear mixed models of parasite prevalence and shedding

	Host species		Species density		Small mammal density
	d.f.	P a	d.f.	P	d.f.	P
Prevalence
E. arizonensis A	1	0.62	1	0.85	1	0.24
E. arizonensis B	1	0.33	1	0.62	1	0.54
E. delicata	1	0.04	1	0.08	1	0.01
A. americana	1	< 0.01	1	0.03	1	0.05
C. americana	3	< 0.001	1	0.76	1	0.68
Cestode A	1	0.14	1	0.07	1	0.05
Hymenolepis A	2	0.43	1	0.97	1	0.17
Hymenolepis B	1	0.08	1	0.32	1	0.18
Parasite sheddingb
E. arizonensis A	1	0.29	1	< 0.01	1	0.01
E. arizonensis B	1	0.72	1	0.08	1	0.29
E. delicate	1	< 0.001	1	< 0.001	1	0.14
A. americana	1	0.24	1	0.82	1	0.64
C. americana	3	0.10	1	0.24	1	< 0.01
Cestode A	1	< 0.01	1	<0.01	1	< 0.01
Hymenolepis A	2	0.19	1	0.47	1	0.08

^a P values were calculated from likelihood ratio tests following term removal from full models; all models contained random effects of sampling site and month of sampling.

^b Hymenolepis B could not be included in the revised analysis of parasite shedding due to lack of statistical power.Bold text indicates statistical significance, P < 0.05.

Figure 1.

Contributions of three sources of host heterogeneity for eight multi‐host parasites (Revised from Figure 3 of the original paper). Symbol sizes are proportional to the total contribution of infectious stages produced by each host species. Squares indicate the key host species (π _i > 0.5) for each parasite.

We also verified that our main conclusions regarding hypothetical control scenarios (figure 4 in the original paper) were robust to using the more conservative subset of host–parasite combinations. As before, targeted control of infected individuals of the key host species was always optimal, but the predicted benefits of this strategy over untargeted control of the key host species (without respect to infection status) or random control of any individual varied dramatically depending on the community context (Fig. 2).

Figure 2.

Efficacy of three control strategies for empirical multi‐host parasites (Revised from Figure 4 of the original paper). Each panel shows the expected reduction in the infectious pool size by random removal of individuals regardless of host species (green) and by targeted (blue) and untargeted (red) removal of the most influential host species (shown in the title of each panel). The dashed line shows the maximum reduction that can be achieved by removing all individuals of the key host species (i.e. the proportion of transmission due to non‐key host species). The J’ values relate to Pielou's evenness index, and quantify the degree of variability across the host community in contributions to parasite transmission; values of J’ lie between 0 (complete dominance by a single species) and 1 (equal contributions of all infected host species).

Two points are worth emphasising from these re‐analyses. First, as in the original paper, we use these data to illustrate the application of our theoretical framework and, given uncertainties in species identification from egg morphology alone, definitive statements regarding the true structure of these specific host–parasite communities are not appropriate. We hope though that this re‐analysis helps to clarify the possible extent of host sharing in these communities, based on currently available information from egg morphology. Second, we emphasise that these re‐analyses do not alter the aims or conclusions of the original paper, and indeed perhaps strengthen them – multi‐host parasites do not use the available host community equally, tending to rely on relatively few host species for maintenance, but those ‘key’ host species may arise in different ways, with potentially important implications for transmission and control. This leads to an important, more general point, that in order to truly quantify host sharing it is vital to be able to identify the parasites accurately, ideally using molecular techniques, to be able to differentiate ‘covert’ host specificity of parasites from those that are true host generalists.