A biological invasion modifies the dynamics of a host–parasite arms race

Abstract

By imposing novel selection pressures on both participants, biological invasions can modify evolutionary ‘arms races’ between hosts and parasites. A spatially replicated cross-infection experiment reveals strong spatial divergence in the ability of lungworms (Rhabdias pseudosphaerocephala) to infect invasive cane toads (Rhinella marina) in Australia. In areas colonized for longer than 20 years, toads are more resistant to infection by local strains of parasites than by allopatric strains. The situation reverses at the invasion front, where super-infective parasites have evolved. Invasion-induced shifts in genetic diversity and selective pressures may explain why hosts gain advantage over parasites in long-colonized areas, whereas parasites gain advantage at the invasion front.

1. Introduction

Interactions between parasites and their hosts are intricately shaped by antagonistic coevolutionary ‘arms races’ occurring over millennia [1]. In spatially structured populations, parasites are often more infective to locally occurring hosts than to allopatric hosts [2–5]. This local adaptation may arise because the parasites' shorter generation times and larger population sizes allow them to gain the upper hand in evolutionary conflicts [6,7]. In other cases, however, the outcome is reversed: hosts may gain advantage (possibly temporary) in the arms race by evolving defences that are highly effective against local strains of parasite [3,8,9]. Teasing apart the processes that generate these divergent outcomes is challenging in a stable system with coadaptations accumulating over a vast timescale. Biological invasions can provide an opportunity to better understand how parasites and their hosts adapt to each other, and the timescales over which those conflicts are won and lost, because invasions remove the spatial structuring of host and parasite populations and impose novel selective forces on one or both participants [7,10,11].

Parasites can play critical roles in biological invasions. An absence of native-range parasites (enemy release) in the newly colonized area can facilitate the success of an invasion by allowing individuals of the translocated species to redirect resources from parasite defence into dispersal, growth, reproduction and competitive ability [12,13]. By contrast, high rates of parasitism by native-range or newly encountered parasite taxa can slow an invasion if infection reduces dispersal rate or entails energy allocation for immune defence [11,14].

The invasion of cane toads (Rhinella marina) through tropical Australia has attracted intensive study. The descendants of 101 toads introduced to northeastern Queensland in 1935 now occupy more than one million square kilometres and are continuing to spread. Most of the cane toad's native-range parasites were left behind during the species' stepwise translocation to Australia [15,16], but a nematode lungworm Rhabdias pseudosphaerocephala was carried by the colonizing toads, and now occurs at high prevalence and intensity in most of the toad's invaded range [16]. However, the lungworm is absent from the toad invasion front, where low host densities reduce transmission opportunities [17,18] and thus are expected to increase selection on parasite infectivity [19].

We conducted a spatially replicated cross-infection experiment using lungworms and common-garden-reared toads from locations across their invaded range in tropical Australia. Our aim was to identify how the level of spatially stable sympatry between host and parasite populations affects infection success (i.e. the ability of a larval parasite to penetrate a toad and establish an infection in the lungs of the host).

Specifically, we predicted that:

• (i) Despite the brief timespan of the toad invasion, intense selection generates spatial variation in the ability of parasites to infect hosts. Thus, at the invasion front where host density is lowest, we expect parasites to exhibit increased infectivity.
• (ii) Evolutionary pressures at an invasion front alter the outcomes of host–parasite arms races compared to those in areas with a longer opportunity for coadaptation. Thus, we predicted that in long-established populations behind the invasion front, we would find evidence of local adaptation (parasites better able to infect local hosts) or local maladaptation (parasites better able to infect allopatric hosts).

2. Methods

(a) . Host–parasite system

Rhabdias sp. are lungworm parasites of amphibians and reptiles. They have a direct life cycle and do not require an intermediate host [20]. Toads become infected with Rhabdias by contacting soil that contains L3 larvae. These larvae penetrate the host's skin and migrate through tissues and along fascia to reach the lung [21]. In the lung, the parasite attaches to capillary beds and feeds on blood [22]. This parasitic stage is a protandrous hermaphrodite, initially forming a testis and producing and storing sperm, then developing ovaries, uterus and eggs that are self-fertilized with stored sperm [23]. Eggs are released into the lung lumen, carried up into the mouth via mucous and swallowed into the host's digestive system. Eggs hatch in the intestine and the first-stage larvae (L1) are shed into the environment in faeces. The free-living stage in the soil (L2) reproduces sexually. After mating, the males die and females develop one to four live offspring inside them [21,24–26]. These offspring consume the mother's organs and burst through her cuticle into the soil as nonfeeding infective larvae (L3). L3 larvae wait in the soil to infect a toad that comes into close contact [21,24].

(b) . Toad collection, breeding and rearing

Adult toads collected from 10 sites across tropical Australia in late 2020 (figure 1; electronic supplementary material, table S1) were used to generate common-garden offspring. Three sites (Cairns, Mareeba and Townsville) were in Queensland (Qld), the state to which toads were introduced in 1935 (all populations had been present for over 80 years). Three sites (Pine Creek, Middle Point, Jabiru) were in the Northern Territory (NT), where toads arrived in the mid-1980s (colonization times of 17, 15 and 17 years, respectively). Four sites (Wyndham, Drysdale, Mitchell Plateau, Fitzroy Crossing) were in Western Australia (WA), where toads arrived in 2009 (colonization times of 7, 4, 1 and 1 years, respectively).

Figure 1.

Map showing northern Australian collection sites for parental toads (white symbols) and lungworms (black symbols). The grey area represents approximate cane toad distribution in 2020. Dates after collection site names indicate the year of toad arrival.

Pairs of toads from each site were induced to breed with subcutaneous injections of 0.25 mg ml⁻¹ leuprorelin acetate (Lucrin, Abbot Australasia, Kurnell, Australia) in amphibian Ringers [27]. Females received 0.18 mg and males received 0.08 mg. Two clutches were obtained from Mareeba and from Townsville and single clutches from the other eight sites, resulting in a total of 12 clutches. A subset of 100 tadpoles randomly selected from each clutch were reared in 70 l tubs equipped with aerators (one clutch per tank) at our field station in Middle Point, Northern Territory. Tadpoles were fed daily on frozen lettuce and fish flakes.

Metamorphs began emerging from tadpole tubs after three weeks, and metamorphs from each clutch were reared in 700 l enclosures lined with dried sand and containing a shallow water dish. Toadlets were fed daily on termites mixed with crushed cat-food pellets. When they were approximately six-months old (51–89 mm snout–vent length, ‘SVL’), 6–24 toads from each clutch were experimentally exposed to infective Rhabdias larvae.

(c) . Lungworm collection

In early 2021, 10 adult toads infected with Rhabdias were collected from each of six sites, close to the places where parental toads had been collected eight months earlier (figure 1; electronic supplementary material, table S1). Two sites were in long-colonized areas in Queensland (over 80 years; Townsville and Cairns), two in intermediate areas in the Northern Territory (over 15 years; Jabiru, Marlow Lagoon) and two close to the invasion front in Western Australia (less than 11 years; Fitzroy Crossing, Kununurra; figure 1). The infected toads were brought to our field station at Middle Point in the Northern Territory. Toads were held separately by site in 700 l bins lined with wood chips and containing a shallow 400 mm diameter water dish. Toads were fed daily with mealworms mixed with cat-food pellets.

To obtain infective larvae from each site, we collected a pooled sample of fresh faeces from the bins housing the 10 toads from that site. A 2–3 g aliquot of faeces from the pooled sample was placed onto a small square of moist paper towel in a Petri dish. Water was added daily to each dish to maintain moisture. After 5 days the Petri dish was flooded with clean water and the mobile larvae that emerged from the paper towel were pipetted out and counted.

Experimental infections of toads were done in two batches: half in late June 2021 and the second half two weeks later in mid-July 2021. From each batch, we ethanol-preserved a subsample of the larvae used for infections from each of the six worm-collection sites. These larvae (N = 17–89 larvae per site per batch) were later examined under a dissecting microscope and measured for length and width.

(d) . Experimental infections

We randomly assigned 232 toads from 12 families to be exposed to lungworm larvae from one of the six Rhabdias collection sites. One to six toads from each of the 10 host sites were exposed to infective larvae from one of the six parasite sites (electronic supplementary material, table S2). Levels of sympatry between host and parasite pairings in cross-infections varied within and among states. Within each state the distances between host and parasite populations ranged from 0 to 400 km. Overall distances between host and parasite populations used in pairings ranged from 0 to 2280 km apart. Experimental infections were performed by placing a juvenile toad in a 250 ml plastic container lined with damp paper towel. We then added 50 infective larvae in 1 ml of water to the paper towel. After 12 h, we removed toads from the plastic cups, measured them for SVL and moved them to individual 15 l cages for long-term housing. These cages were lined with paper towel and contained a 200 ml water dish. Eighteen weeks later, we euthanized toads by an overdose of pentobarbitone sodium and dissected their lungs to count the number of adult Rhabdias.

(e) . Analysis

To quantify the size of larvae, we used principal component analysis to combine measures of the length and width of preserved larvae. The first component from this analysis (PC1) incorporated 68% of the variation in measures of length and width.

Our measure of infection outcomes was abundance—the number of worms found in the lungs of each toad at dissection, including 0s for toads that remained uninfected. Dissections and lungworm counts were done blinded to toad ID and treatment group. We used a generalized mixed model with a negative binomial distribution and a log link function to assess how host and parasite origins affected the outcome of experimental infections. We used the state (WA, NT or QLD) from which the toads and lungworms originated, and the interaction between host and parasite state, as independent variables. Population genetic analysis of cane toads indicates the presence of two genetically distinct groups in the north of Australia: those from the east of the Great Dividing Range (i.e. coastal Queensland) and those to the west (i.e. western QLD, NT and WA; [28]). Similarly, Rhabdias from coastal QLD are genetically distinct to those from WA and no within-region variation was identified, but genetic data from the NT are not available [29]. We also included the body size of the toad and the average size (PC1) of Rhabdias larvae used for infections as covariates in the model. As random effects, we included the clutch to which the toad belonged, and the host and parasite collection sites (within State). We ran the mixed model analysis using the Glimmix procedure in SAS 9.4 [30]. We examined residuals for violations of assumptions and used two-tailed tests to assess significance.

3. Results

The sizes of larvae used for experimental infections did not differ significantly among the three states (F_2,720 = 0.77, p = 0.4619; electronic supplementary material, figure S1). Of 232 toads that were each exposed to 50 lungworm larvae, 104 became infected. Among the infected toads, median infection intensity was five adult worms (interquartile range = 2–8 worms). Mixed model analysis indicated that infection abundance was affected by larval size (F_1,208 = 6.55, p = 0.0112; table 1, electronic supplementary material, table S3); larger larvae resulted in heavier infections (electronic supplementary material, figure 2). Infection abundance was not related to toad body size (F_1,208 = 0.96, p = 0.329).

Table 1.

Generalized linear mixed model analysis of factors affecting lungworm abundance in cane toads (Rhinella marina) following experimental infections. The model included toad clutch and collection site within state as random effects and used a negative binomial distribution with a log-link function. Italics indicate p-values < 0.05.

effect	d.f.	F-value	p-value
mean larvae size	1, 208	6.55	0.0112
toad SVL	1, 208	0.96	0.3290
toad state	2, 208	0.68	0.5053
worm state	2, 208	4.28	0.0151
toad state × worm state	4, 208	4.12	0.0031

Figure 2.

Least-square mean abundances of lungworm parasites in cane toads (correcting for host and parasite size) illustrating the significant interaction between host and parasite origins as a determinant of infection abundance. Experimental infections were carried out using toads and lungworms from sites in three states (Western Australia (WA), Northern Territory (NT) and Queensland (QLD)) that encompass the cane toad's invasion trajectory in tropical Australia. Bars represent standard errors around least-square means.

The most significant determinant of infection abundance was the interaction between state of collection of the host versus the parasite (F_1,218 = 4.11, p = 0.0031; table 1, figure 2). A likelihood ratio test comparing models with and without this interaction term indicated that inclusion of the interaction provided a significant improvement (likelihood ratio = 47.83, d.f. = 4, p < 0.0001).

Visual inspection of figure 2 suggested that parasites from populations close to the invasion front (WA) had the highest infection abundance in hosts from all locations. To explicitly test our prediction that frontal parasites were more infective, we ran a pre-planned contrast to compare average infectivity of WA parasites with average infectivity of parasites from elsewhere (NT and QLD). This contrast was significant (estimate = 1.33, t₂₀₈ = 2.92, p = 0.0039), supporting higher infectivity of frontal parasites.

Queensland parasites appear more infective to NT toads than to QLD toads, and vice versa (figure 2). To test our prediction that either local adaptation or local maladaptation would be present in older populations, we performed another pre-planned contrast. For crosses involving toads and parasites from NT and QLD, we compared mean infection abundances between sympatric host–parasite combinations (i.e. within-state) and allopatric (i.e. different-state) combinations. This contrast was highly significant (estimate = 1.58, t₂₀₈ = 4.02, p < 0.0001) and indicated that infection abundances were higher in allopatric pairings than in sympatric pairings, suggesting that parasites were maladapted to their local hosts.

4. Discussion

Evolutionary theory predicts that parasites will generally win in arms races with their host (local adaptation) because of shorter generation times, larger population sizes and greater genetic variation in pathogens than hosts [7,9,31]. This prediction of advantage to the parasite is supported by many empirical studies [2–5], but is not universal; in some systems, the hosts have advantage over the parasite, at least temporarily (e.g. [3–5,8,32–34]). Our data from a biological invasion reveal a more dynamic outcome. First, hosts from long-colonized areas were more capable of resisting infection by locally occurring parasite strains than by parasites from more distant sites (i.e. local maladaptation of the parasite, resulting in advantage to the host). Second, lungworms from the invasion front infected toads from all populations at far higher rates than did conspecifics from range-core populations (i.e. widespread advantage to the parasite rather than location-specific outcomes). Below, we address possible explanations for these effects.

In long-colonized areas within the cane toad's Australian range, we found that hosts have gained an advantage in encounters with their local parasite. Such outcomes have been attributed to higher rates of gene flow in host populations than in those of parasites, facilitating more effective anti-parasite adaptation [3–5,8,32,34]. Consistent with that explanation, adult lungworms are self-fertilizing hermaphrodites [23,24], whereas toads reproduce sexually. Although the lungworms also include a sexually reproducing phase (L2 larvae living in toad faeces and soil), most larvae within a faecal pile are likely to be siblings or at most, close relatives (especially if the host is infected by a small number of adult worms: [35]). Inbreeding among lungworms at a given site may decrease the parasite's ability to out-evolve the host [3,7].

Selection on the host to resist infection also may be greater than selection on the parasite to infect, because (i) the advantage to a toad of resisting infection may be high in all populations [36], whereas (ii) a parasite that fails to infect a host may soon have another opportunity with a different toad. Transmission requires a toad to come into prolonged contact with a patch of damp soil where an infected toad had defecated several days earlier. In long-colonized areas, toads frequently share moist diurnal retreat sites on a consistent basis [37,38], where their faeces create ideal opportunities for transmission of lungworms (either to the same individual or its siblings) from one host to another. Such a reduction in the fitness penalty to unsuccessful infection attempts would skew the arms race outcome in favour of the host.

Why, then, do parasites at the invasion front exhibit greater infectivity? The high infection abundance of invasion-front lungworms in invasion-front toads reflects traits of the parasite (superior performance against all hosts) rather than the host. Although selection and spatial sorting of dispersal-enhancing traits have altered immune investment of toads at the range edge [14,39,40], toads from the invasion front retained effective immune defences against lungworms from long-colonized areas (figure 2). The high infection success of invasion-front lungworms (consistent with earlier studies: [18,19,25]) cannot be explained by larger larvae, because the sizes of L3 used for experimental infections did not differ significantly among states. Instead, the higher infection abundances of invasion-front lungworms may result from circumstances that intensify selection on the ability of parasites to infect new hosts.

Low population densities of hosts at the invasion front reduce opportunities for infection [17,19], exacerbated by the fact that invasion front toads rarely share or reuse diurnal shelter sites [38]. At the invasion front, a lungworm larva that fails to infect the first host it encounters may never have another chance; and neither will its siblings. That shift in intensity of selection on the parasite, with no concurrent shift in selection on the host, may tip the scales in favour of advantage to the parasite. Additionally, arid conditions at the invasion front may have favoured changes to lungworm larval morphology or physiology (e.g. thicker cuticle, reduced permeability) that facilitate survival on dry substrates but also circumvent host defences and thereby achieve global not local fitness optima.

Mechanisms underlying the spatial difference in the arms race outcomes between lungworms and toads are unclear and identifying them would be a fruitful avenue for further study. The infectiveness of parasites can be enhanced by an ability to hide their antigenic surfaces from the host immune system, or to secrete products that interfere with host immune signals [41,42], thereby stimulating counter-adaptations in the host [1,4]. Parasite behaviour also may play a role. For example, invasion-front lungworms might be better able to locate areas of host skin that are easily penetrated [43] or offer a shorter route to the lungs. Structural or chemical changes to the larvae's cuticle or secretions might also allow it to evade host immune defences as the parasite migrates through tissue enroute to the lungs [42]. By bringing parasites and toads into direct contact, our experimental procedure excluded some of these mechanisms and revealed selection on infectivity, but not how it arises.

On a cautionary note, our study incorporated limited replication of families (N = 12) and sites (N = 10). Future research could incorporate expanded familial and geographical variation in toad and Rhabdias genetics and life history, to verify the shifts in infection outcomes revealed by our experiments and to clarify their basis [6,9,31]. For example, studies on the ultrastructure and differential gene expression of infective larvae across this invasive range could identify mechanisms underlying the high infectivity of invasion-front larvae. Investigating gene expression of both toads and lungworms when paired sympatrically and allopatrically could also clarify the pattern of local maladaptation that we found in long-colonized sites. Using such an experimental design, the relative contributions of hosts and their parasites to the outcomes of arms races can be investigated (e.g. Feis et al. [7]). Additionally, specific host–parasite genotype combinations and host genotype rarity can affect infection dynamics [44,45]. Both of these factors are reduced in populations with low genetic diversity, such as is often found at invasion fronts (e.g. in invasion-front cane toads [28]) and their lungworms [29]). In a study of Daphnia magna and the pathogen Pastueria ramosa, host population age was negatively correlated with pathogen establishment success of all genotypes [46], indicating that genotype–genotype interactions may be relaxed on invasion fronts. It is expected that parasite-induced selection on host life-history traits is relaxed when genotype–genotype interactions are strong [47], so it may be interesting to investigate whether differences in specificity between Rhabdias and cane toad genotypes differ across the range. Further, the effect of host genotype rarity may contribute to differences in sympatric versus allopatric host–parasite relationships. Testing the impact of these factors on host–parasite dynamics in an experimental framework will advance our understanding of these relationships.

Our study reinforces the value of biological invasions as model systems in which to explore rapid evolutionary changes not only in invasive species and the native biota affected by them [10], but also in co-adapted interactions among species [48]. Invasive host–parasite systems are ideal in this respect, because the novel evolutionary pressures engendered by range expansion can result in rapid shifts in adaptive optima for both participants in the arms race [11]. In the case of cane toads in Australia, invasion has produced striking disparities in the ability of co-evolved lungworms to infect their hosts, and in the ability of those hosts to resist infection. In such a system, experimental studies that bring allopatric hosts and parasites into contact can provide useful insights into the ecological foundations and mechanistic factors that underly arms races [7,31].

Ethics

The study was conducted under Macquarie University Animal Ethics approval 2021/001.

Data accessibility

Data available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.v41ns1s2z [49].

Supplementary material is available online [50].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

G.P.B.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft; R.S.: conceptualization, funding acquisition, project administration, supervision, writing—original draft; L.A.R.: conceptualization, funding acquisition, project administration, supervision, writing—original draft.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This research was funded by Australian Research Council grant # DP190100507 to R.S. and L.A.R.